thesis.txt

INVESTIGATING THE RECOGNITION
AND INTERACTIONS OF NON-POLAR
α HELICES IN BIOLOGY

A thesis submitted to the University of Manchester
for the degree of Doctor of Philosophy
in the Faculty of Science and Engineering

2019

James A. Baker

School of Chemistry

Contents
Abstract

15

Lay Abstract

16

Declaration

17

Copyright Statement

18

Acknowledgements

19

1 Introduction

20

1.1

Transmembrane proteins . . . . . . . . . . . . . . . . . . . . . . . . . .
1.1.1

1.2

1.3

1.4

20

A brief history of the discovery and exploration of the transmembrane proteins . . . . . . . . . . . . . . . . . . . . . . . . .

20

1.1.2

Transmembrane proteins in disease . . . . . . . . . . . . . . . .

22

1.1.3

The transmembrane protein problem . . . . . . . . . . . . . . .

23

1.1.4

The transmembrane protein revolution . . . . . . . . . . . . . .

24

1.1.5

The role of bioinformatics in transmembrane biology . . . . . .

26

Biological membranes . . . . . . . . . . . . . . . . . . . . . . . . . . . .

28

1.2.1

Membrane lipids . . . . . . . . . . . . . . . . . . . . . . . . . .

28

1.2.2

Membrane potential . . . . . . . . . . . . . . . . . . . . . . . .

30

α helices in the membrane; structure and function . . . . . . . . . . . .

31

1.3.1

Transmembrane helix sequence composition . . . . . . . . . . .

31

1.3.2

The hydrophobicity of transmembrane segments . . . . . . . . .

35

1.3.3

Sequence complexity . . . . . . . . . . . . . . . . . . . . . . . .

38

Biogenesis of transmembrane proteins . . . . . . . . . . . . . . . . . . .

40

1.4.1

40

An overview of translocation . . . . . . . . . . . . . . . . . . . .
2

1.5

1.4.2

Signal peptides . . . . . . . . . . . . . . . . . . . . . . . . . . .

44

1.4.3

β sheets in the membrane . . . . . . . . . . . . . . . . . . . . .

45

The aims of this thesis . . . . . . . . . . . . . . . . . . . . . . . . . . .

48

2 The “negative-outside” rule
2.1

50

Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

50

2.1.1

Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

50

2.1.2

Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

50

2.1.3

Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

51

2.2

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

51

2.3

Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

56

2.3.1

Acidic residues within and nearby transmembrane helix segments are rare . . . . . . . . . . . . . . . . . . . . . . . . . . . .

2.3.2

56

Amino acid residue distribution analysis reveals a “negative-notinside/negative-outside” signal in single-pass transmembrane helix segments . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

2.3.3

60

Amino acid residue distribution analysis reveals a general
negative-charge bias signal in outside flank of multi-pass transmembrane helix segments — the negative outside enrichment
rule . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

2.3.4

64

Further significant sequence differences between single-pass and
multi-pass helices: distribution of tryptophan, tyrosine, proline
and cysteine . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

2.3.5

Hydrophobicity and leucine distribution in transmembrane helices in single- and multi-pass proteins . . . . . . . . . . . . . .

2.3.6

71

A negative-outside (or negative-non-inside) signal is present
across many membrane types . . . . . . . . . . . . . . . . . . .

2.3.7

69

73

Amino acid compositional skews in relation to transmembrane
helix complexity and anchorage function . . . . . . . . . . . . .

76

2.4

Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

82

2.5

Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

90

2.5.1

90

Datasets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3

2.5.2

On the determination of flanking regions for transmembrane helices and the transmembrane helix alignment . . . . . . . . . . .

97

2.5.3

Separating simple and complex single-pass helices . . . . . . . .

99

2.5.4

Distribution normalisation . . . . . . . . . . . . . . . . . . . . . 100

2.5.5

Hydrophobicity calculations . . . . . . . . . . . . . . . . . . . . 101

2.5.6

Normalised net charge calculations . . . . . . . . . . . . . . . . 101

2.5.7

Statistics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102

3 Collation and analysis of tail-anchored protein transmembrane helices reveals subcellular variation in flanking charged residue distribution and the transmembrane helix core hydrophobicity

104

3.1

Abstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104

3.2

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105

3.3

Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110

3.4

3.3.1

Building a list of tail-anchors . . . . . . . . . . . . . . . . . . . 110

3.3.2

Calculating hydrophobicity . . . . . . . . . . . . . . . . . . . . . 115

3.3.3

Calculating sequence information entropy . . . . . . . . . . . . . 116

3.3.4

Statistics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116

3.3.5

Modelling cytochrome b5 and PTP1b . . . . . . . . . . . . . . . 117

3.3.6

Availability of materials . . . . . . . . . . . . . . . . . . . . . . 118

Results and discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . 118
3.4.1

A comparison of up-to-date tail-anchored protein datasets . . . 118

3.4.2

It is difficult to observe any hydrophobic variation of tailanchored protein transmembrane helices from different species . 121

3.4.3

There are biochemical differences between tail-anchored transmembrane helices from different organelles . . . . . . . . . . . . 123

3.4.4

More annotation is required to identify chaperone interaction
factors of the transmembrane helix. . . . . . . . . . . . . . . . . 131

3.4.5

Spontaneous insertion may be achieved by polar strips in the
transmembrane helix of tail-anchored proteins . . . . . . . . . . 133

3.5

Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135
4

4 Sequence analysis of polarity in transmembrane helices suggests that
translocation of marginally hydrophobic helices could be facilitated
by neighbouring typically hydrophobic helices

139

4.1

Abstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 139

4.2

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 140
4.2.1

The ribosome-translocon complex in the biogenesis of membrane
proteins . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 141

4.2.2

Cooperative transmembrane helix insertion by the transloconribosome complex . . . . . . . . . . . . . . . . . . . . . . . . . . 143

4.2.3
4.3

4.4

Aims . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 145

Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 145
4.3.1

Datasets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 145

4.3.2

Gene ontology . . . . . . . . . . . . . . . . . . . . . . . . . . . . 147

4.3.3

Complexity and hydrophobic estimation . . . . . . . . . . . . . 147

4.3.4

Statistics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 148

4.3.5

Availability of materials . . . . . . . . . . . . . . . . . . . . . . 148

Results and discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . 149
4.4.1

Large contrasts in transmembrane helix hydrophobicity occur in
channels and receptors . . . . . . . . . . . . . . . . . . . . . . . 149

4.4.2

GPCRs contain conserved relatively polar TMH7, which follows
the typically hydrophobic TMH6 . . . . . . . . . . . . . . . . . 150

4.4.3

6TMH ion channels contain polar-hydrophobic transmembrane
helix pairs/groups indicative of conserved cooperative insertion . 155

4.4.4

The prevalence of the high hydrophobic discrepancy of transmembrane helices amongst other common ‘transporter’ transmembrane protein classes

4.5

. . . . . . . . . . . . . . . . . . . . . 159

Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 161

5 Conclusions and outlook

164

Word count 47,705
5

List of Tables
2.1

Acidic residues are rarer in transmembrane helices of single-pass proteins than in transmembrane helices of multi-pass proteins. . . . . . . .

2.2

Statistical significances for negative charge distribution skew on either
side of the membrane in single-pass transmembrane helices. . . . . . . .

2.3

67

Leucines at the inner and outer leaflets of the membrane in transmembrane helices. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

2.5

63

Statistical significances for negative charge distribution skew on either
side of the membrane in multi-pass transmembrane helices. . . . . . . .

2.4

57

74

Simple transmembrane helices are less similar than complex transmembrane helices to transmembrane helices from multi-pass proteins in UniHuman. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

2.6

79

Simple transmembrane helices are less similar than complex transmembrane helices to transmembrane helices from multi-pass proteins in ExpAll. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

80

2.7

The experimental evidences of TOPDB. . . . . . . . . . . . . . . . . .

92

2.8

Records with INTRAMEM and TRANSMEM flanking region overlap. .

99

3.1

Hydrophobicity statistical comparisons between mouse and human,
yeast, and plants in the SwissProt Filtered Dataset. . . . . . . . . . . . 121

3.2

Hydrophobicity statistical comparisons between mouse and human,
yeast, and plants in the UniProt Curated Dataset. . . . . . . . . . . . . 123

3.3

Statistical comparisons between TMH sequences from organelles in the
UniProt Curated Dataset. . . . . . . . . . . . . . . . . . . . . . . . . . 123

3.4

Statistical comparisons between transmembrane helix sequences from
organelles in the SwissProt Filtered Dataset. . . . . . . . . . . . . . . . 129
6

4.1

Dataset sizes of common transmembrane protein families of transporters
and channels. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 146

7

List of Figures
1.1

A selection of figures demonstrating important discoveries of membrane
proteins and their environment. . . . . . . . . . . . . . . . . . . . . . .

21

1.2

The structure of SecYE in a nanodisc at near atomic resolution. . . . .

25

1.3

A comparison of a single structure held by commonly used structural
transmembrane protein databases. . . . . . . . . . . . . . . . . . . . . .

1.4

A cartoon showing the general components of the membrane and a
typical transmembrane helix. . . . . . . . . . . . . . . . . . . . . . . . .

1.5

28

32

A cartoon of the topological clustering of positively-charged residues
near transmembrane helices in relation to subcellular location. . . . . .

33

1.6

The hydropathic index of rabbit cytochrome b5. . . . . . . . . . . . . .

37

1.7

The hydrophobic-complexity continuum distinguishes between transmembrane helix anchors and those with function beyond anchoring. . .

1.8

39

A simplified schematic of the co-translational Sec pathway and the posttranslational pathway. . . . . . . . . . . . . . . . . . . . . . . . . . . .

41

The pore, the plug, and the lateral gate of the translocon. . . . . . . .

42

1.10 The key components of a signal peptide. . . . . . . . . . . . . . . . . .

45

1.9

1.11 Cartoons showing the structural differences of the outer membrane β-barrel proteins, transmembrane α-helix channels, and transmembrane
α-helix signal transducers in the membrane. . . . . . . . . . . . . . . .

46

1.12 A cartoon of the biogenesis of β barrel membrane proteins in mitochondria and Gram-negative bacteria. . . . . . . . . . . . . . . . . . . . . .
2.1

47

Negatively-charged amino acids are amongst the rarest residues in transmembrane helices and ±5 flanking residues. . . . . . . . . . . . . . . .
8

58

2.2

Relative percentage normalisation reveals a negative-outside bias in
transmembrane helices from single-pass protein datasets. . . . . . . . .

2.3

Negative-outside bias is very subtle in transmembrane helices from
multi-pass proteins. . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

2.4

61

65

The net charge across multi-pass and single-pass transmembrane helices
shows a stronger positive inside charge in single-pass transmembrane
helices than multi-pass transmembrane helices. . . . . . . . . . . . . . .

2.5

66

Relative percentage heat-maps from predictive and experimental
datasets corroborate residue distribution differences between transmembrane helices from single-pass and multi-pass proteins.

2.6

. . . . . . . . .

There is a difference in the hydrophobic profiles of transmembrane helices from single-pass and multi-pass proteins. . . . . . . . . . . . . . .

2.7

72

There is a difference in the hydrophobic profiles of transmembrane helices from single-pass and multi-pass proteins. . . . . . . . . . . . . . .

2.8

70

73

Comparing charged amino acid distributions in transmembrane helices
of multi-pass and single-pass proteins across different species and organelles. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

2.9

75

Comparing the amino acid relative percentage distributions of simple
and complex transmembrane helices from single-pass proteins and transmembrane helices from multi-pass proteins. . . . . . . . . . . . . . . . .

77

2.10 Residue distributions of transmembrane anchors. A view showing additional residue distribution features that transmembrane helices with
an anchorage function display. . . . . . . . . . . . . . . . . . . . . . . .

89

2.11 The lengths of flanks and transmembrane helices in multi-pass and
single-pass proteins in the UniHuman and ExpAll dataset. . . . . . . .

98

2.12 Relative percentage heatmaps from the predictive datasets calculated
by fractions of the absolute maximum and by the relative percentage of
a given amino acid type. . . . . . . . . . . . . . . . . . . . . . . . . . . 102
3.1

An overview of the biogenesis of tail-anchored proteins. . . . . . . . . . 107

3.2

The sources, methods, and filters applied to the sequences in the tailanchored protein datasets. . . . . . . . . . . . . . . . . . . . . . . . . . 111
9

3.3

A Venn diagram showing tail-anchored protein UniProt ids present in
each of the datasets as well as those present in multiple datasets. . . . . 119

3.4

Average values of species datasets from UniProt manually curated set
and SwissProt automatically filtered dataset. . . . . . . . . . . . . . . . 122

3.5

Average sequence-based biochemical values of organelle datasets from
UniProt manually curated set and SwissProt automatically filtered
dataset. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124

3.6

The normalised skews of each amino acids from tail-anchored proteins
grouped by localisation from the SwissProt automatically filtered dataset.126

3.7

The normalised skews of each amino acids from tail-anchored proteins
grouped by localisation from the SwissProt automatically filtered dataset.128

3.8

The profile of transmembrane helix and flanks hydrophobicity from tailanchored protein groups stratified by chaperone interactors. . . . . . . 133

3.9

Structural biochemical analysis of a homology model of cytochrome b5. 135

3.10 Structural biochemical analysis of a homology model of PTP1b. . . . . 136
3.11 A cartoon of a potential method the cytochrome b5 and PTP1b transmembrane helix could integrate spontaneously into the membrane. . . . 137
4.1

A cartoon showing the generally accepted schematic of sequential multipass transmembrane helix insertion into the membranes. . . . . . . . 140

4.2

A cartoon of the ribosome in association with the translocon during
insertion. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 142

4.3

Pie charts of a non-redundant list of transmembrane proteins compared
to a list of transmembrane proteins containing the most hydrophobically
different transmembrane helix pairs. . . . . . . . . . . . . . . . . . . . . 149

4.4

The hydrophobicity and complexity of GPCR transmembrane helices. . 151

4.5

The hydrophobic difference observed between TMH6 and TMH7 in
GPCRs is not due to the choice of hydrophobic scale. . . . . . . . . . . 153

4.6

The hydrophobicity of transmembrane helices in GPCR subfamilies. . . 154

4.7

The hydrophobicity of transmembrane helices in ion channels. . . . . . 157

4.8

The hydrophobic difference observed between TMH4 and the neighbouring transmembrane helices in 6TMH ion channels is not due to the
choice of hydrophobic scale. . . . . . . . . . . . . . . . . . . . . . . . . 159
10

4.9

Sequence entropy is unsuitable for assessing function in TMH4 of ion
channels. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 160

4.10 High polarity discrepancy between sequentially adjacent transmembrane helices is not present in all transmembrane protein transporter
families. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 161
4.11 A cartoon of potential cooperative transmembrane helix insertion methods. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 162

11

Acronyms
AP Arrest Peptide. 141–144, 163, 165
EM Electron Microscopy. 21, 22, 25–27, 43, 163, 165
EMC ER Membrane protein Complex. 23, 43, 44, 107, 109, 162, 165
ER Endoplasmic Reticulum. 30–32, 37, 40, 43, 54, 86, 87, 106–110, 112–114, 123–126,
128–131, 140, 144, 155
GPCR G protein-coupled receptor. 28, 46, 145–147, 149–155, 162, 163, 165
K-S Kolmogorov-Smirnov. 78–81, 103, 116
K-W Kruskal-Wallis. 63, 67, 78–81, 102, 103, 116
MD Molecular Dynamics. 27, 29, 35, 165
MOM Mitochondrial Outer Membrane. 46–48, 109, 130, 131, 135, 136
PDB Protein Data Bank. 22, 28, 32, 55
PM Plasma Membrane. 40, 43, 86, 87, 112, 114, 123–125, 127, 129, 130
POPC Palmitoyloleoylphosphatidylcholine. 38, 73
RNA Ribonucleic Acid. 41, 140–142
SNARE Soluble N-ethylmaleimide-sensitive factor attachment protein receptor. 105,
109, 164
SP Signal Peptide. 24, 26, 43–45, 47, 48, 112, 113, 131
12

SR Signal Recognition Particle Receptor. 41, 43, 44, 106, 140, 141
SRP Signal Recognition Particle. 41, 43, 44, 106–108, 115, 140, 141
TA Tail Anchor. 27, 37, 40, 43, 49, 105–110, 113–115, 118–125, 127, 129–133, 135–
137, 164, 165
TM Transmembrane. 21, 26, 28, 34, 39, 41, 52–55, 69, 71, 78, 85, 98, 109
TMH Transmembrane Helix. 20, 23–46, 48, 49, 51–78, 81–91, 93, 94, 96–103, 105–
110, 112–114, 116–118, 120–138, 140–145, 147–158, 160–166
TMP Transmembrane Protein. 20–27, 29, 31, 35, 40, 42–45, 48, 49, 51, 53, 55, 58,
59, 68, 71, 73, 82–87, 108, 121, 131, 140–143, 149, 150, 155, 162–166
TMS Transmembrane Segment. 24, 32, 101, 117, 143
TOM Translocase of the Outer Membrane. 47, 48, 109

13

The University of Manchester
James A. Baker
Doctor of Philosophy
Investigating the Recognition and Interactions of Non-Polar α Helices in
Biology
February 1, 2019

14

Abstract
Non–polar helices feature prominently in structural biology. Transmembrane α helix
containing proteins make up around a third of all proteins, represent around 40% of
drug targets, and contain some of the most critical proteins required for life as we
know it. Yet they are fundamentally difficult to study experimentally. This is in part
due to the very features that make them so biologically influential; their non–polar
transmembrane helical regions.
By leveraging large data-sets of transmembrane proteins, this thesis is focused on
characterising features of transmembrane α helices en masse, particularly regarding
their topology, membrane–protein interactions, and intramembrane protein interactions.
In this study, we present statistical evidence demonstrating the ‘negative-outside’
rule in opposition to the ‘positive-inside’ rule. We also identify stabilising amino acid
distributions in anchoring transmembrane helices compared to transmembrane helices
with function beyond anchoring.
Tail-anchored proteins are a group of post-translationally inserting proteins. In this
thesis we show adaptations of hydrophobicity and residue distributions through the
transmembrane helices of tail-anchored proteins to different membrane environments
within the cell (the mitochondria, endoplasmic reticulum, the Golgi, and the plasma
membrane). However, we could not detect a hydrophobic difference between global
populations of tail-anchored proteins in different species (mammals, plants, and fungi).
A handful of these proteins are capable of integrating into the membrane without
the need for membrane integration proteins. Structural modelling of transmembrane
helices from PTP1b and cytochrome b5 reveals a 3D amphipathic arrangement of
residues. This structural feature may play a role in their spontaneous membrane
insertion.
Finally, we find a conserved pattern of typically hydrophobic transmembrane helices neighbouring marginally hydrophobic helices in some families of transmembrane
proteins. This feature corresponds to transmembrane helices that have the potential to cooperate in order to integrate the more polar, but functionally important,
transmembrane helix of the pair into the membrane.

15

Lay Abstract
The survival of each of our cells relies on a cellular barrier (called the membrane)
to separate themselves from the surrounding environment. The membrane works by
being chemically very different from both the outside environment and the inside of
the cell, which in both cases is mostly water. The membrane is fatty so repels water.
Proteins are the molecular machinery that forms much of the cell structure and
shape as well as carrying out many of the cell’s routine tasks. Around a third of
our genome codes for membrane-embedded proteins. But because these membraneproteins are adapted for a life in the water-repelling cell membrane, they are very hard
to study in laboratories which often rely on methods that hold proteins in water-based
solutions.
In this thesis, we focus particularly on the parts of the protein that are embedded in
the water repelling membrane. We computationally analysed the biochemical make-up
of thousands of proteins from openly available biological databases.
This thesis demonstrates three features of membrane proteins:
• the radically different evolutionary story that membrane-bound regions have
compared to other proteins; the sacrifices they make for their stability in order to
maintain their function, and their optimisation through evolutionary timescales
to adapt to the membrane as best they can.
• a distinct sub-group of membrane-proteins that have a radically different
membrane-insertion mechanism (tail-anchored proteins) have adaptations in
their membrane regions depending on where they are located in the cell.
• some types of membrane proteins may use several membrane elements to ensure
the least stable, but functionally important, elements are correctly inserted into
the membrane.
These results will go on to inform more specific studies about membrane proteins.
These findings will provide insight into the causes of some genetic diseases as well as
drug targets in the case of pathogenic infections and cancers.

16

Declaration
No portion of the work referred to in the thesis has been
submitted in support of an application for another degree
or qualification of this or any other university or other
institute of learning.

17

Copyright Statement
i. The author of this thesis (including any appendices and/or schedules to this thesis)
owns certain copyright or related rights in it (the “Copyright”) and s/he has given
The University of Manchester certain rights to use such Copyright, including for
administrative purposes.
ii. Copies of this thesis, either in full or in extracts and whether in hard or electronic
copy, may be made only in accordance with the Copyright, Designs and Patents
Act 1988 (as amended) and regulations issued under it or, where appropriate, in
accordance with licensing agreements which the University has from time to time.
This page must form part of any such copies made.
iii. The ownership of certain Copyright, patents, designs, trade marks and other intellectual property (the “Intellectual Property”) and any reproductions of copyright
works in the thesis, for example graphs and tables (“Reproductions”), which may
be described in this thesis, may not be owned by the author and may be owned by
third parties. Such Intellectual Property and Reproductions cannot and must not
be made available for use without the prior written permission of the owner(s) of
the relevant Intellectual Property and/or Reproductions.
iv. Further information on the conditions under which disclosure, publication and
commercialisation of this thesis, the Copyright and any Intellectual Property
and/or Reproductions described in it may take place is available in the University
IP Policy (see http://documents.manchester.ac.uk/DocuInfo.aspx?DocID=487), in
any relevant Thesis restriction declarations deposited in the University Library,
The University Library’s regulations (see http://www.manchester.ac.uk/library/aboutus/regulations) and in The University’s Policy on Presentation of Theses.

18

Acknowledgements
I would like to thank all members of both the Eisenhaber research group, as well as
the Curtis and Warwicker research group for their interesting and engaging discussions throughout my PhD. In particular, I would like to acknowledge the patience,
direction, and exceptional supervision from Dr Jim Warwicker of the University of
Manchester along with Dr Frank Eisenhaber and Dr Birgit Eisenhaber from the Singapore Bioinformatics Institute. I also express my gratitude towards Dr Wing Cheong
Wong and Dr Max Hebditch for working with me on many fiddly problems. I would
also like to thank my advisor, Professor Stephen High, for his stimulating discussions
and guidance throughout the project.
I also thank the University of Manchester and the ARAP programme at the
A*STAR for funding the project.
I thank my parents Janet and Martin, my brother Tim, and my partner Emily for
their unwavering support of me in undertaking a PhD based for two years on the other
side of the planet. Being so far from my family and loved ones was certainly the most
challenging part of this process.

19

Chapter 1
Introduction
Membrane biology is a huge and varied field that is ultimately the study of the interface between compartments of the cell. The Transmembrane Protein (TMP) group
include some of the most critical to life proteins as well as a large number of drug targets. However, the experimental inaccessibility of the Transmembrane Helix (TMH)
has hampered the progress of study compared to their globular structural analogues.
Despite progress over the last decade, the understanding of the relationship between
the sequence and function of a TMH is incomplete.
In this chapter we will place the TMH problem in context, discuss some of the tools
and methods that allow us to analyse the TMHs, describe the important biological
aspects of the TMH, and discuss other membrane spanning protein segments.

1.1
1.1.1

Transmembrane proteins
A brief history of the discovery and exploration of the
transmembrane proteins

Due to the ability to segregate biochemical environments, the cellular barrier and
resultant compartmentalisation has been described as one of the fundamental pillars
of life as we know it [1, 2].
The significance of the cellular barrier was first noted in 1665 with the dawn of
the microscope, when Robert Hooke described the cell wall of the cork plant (Figure
1.1A), with the clear distinction of the barrier giving rise to the term “cell” [3, 4].
20

1.1. TRANSMEMBRANE PROTEINS

A)

C)

21

B)

D)

Figure 1.1: A selection of figures demonstrating important discoveries of membrane
proteins and their environment. A) (Above/“Fig 1”) A diagram of the cork plant as viewed
through an optical microscope drawn by Hooke circa 1665. Note the observable cellular barrier
between what we now know to be the plant cells. (Below/“Fig 2”) The cork plant used in the
microscope observations. From Hooke, republished in 1961 [4]. B) The first near-atomic resolution
(7Å resolution map) structure of a TMP acquired by Electron Microscopy (EM). This is a model
of the 7Transmembrane (TM) bacteriorhodopsin single protein from the purple membrane viewed
roughly parallel to the plane of the membrane. From Henderson & Unwin, 1975 [5]. C) The electron
density map of the first crystal structure of a TMP [6]. The image shows 11 layers with contours
representing 1.2Å between layers at 3Å resolution. As in B, this structure is from phototrophic purple
bacteria. The chromophores are made up of cytochrome (can be seen at the top of the image) and the
H-subunit (which can be seen at the bottom). From Deisenhofer et al., 1984. D) The first drawing
in which the fluid mosaic model was presented in the bilayer. From Singer & Nicolson, 1972 [7]

Throughout the early 20th century several theories were explored regarding the

22

CHAPTER 1. INTRODUCTION

composition of the membrane. In 1925, by scrutinising the concentration of chromocytes in the blood, and the surface area of the cells in blood from a variety of
mammals Gorter and Grendel concluded that: “chromocytes are covered by a layer
of fatty substances that is two molecules thick” [8]. This established the awareness of
the lipid bilayer. Later, the popular Danielli and Davson model correctly identified a
phospholipid bilayer, however also incorrectly suggested a lipoid space existed within
the phospholipid bilayer [9]. Eventually, as these ideas were explored, the fluid mosaic
model was established in 1972 (Figure 1.1D) [7]; the membrane has a fluid behaviour,
due to the gel-like nature of its composition, which allows proteins to move throughout
the membrane bilayer.
In order to fully understand the relationship between TMPs and the membrane,
molecular details would be needed. By the 1970s, reasonably detailed structures of
membrane-spanning segments of TMPs were available by EM [5] (Figure 1.1B). However, the real goal was to have the atomic resolution available. The advent of x-ray
crystallography through the first half of the 20th century showed that it was possible to solve atomically resolved structures of small organic molecules [10, 11] and
larger proteins such as a steroid [12], penicillin [13], and vitamin B12 [14]. Dorothy
Hodgekin was awarded the Nobel prize for chemistry in 1964 for the elucidation of
these complex structures [15]. Seven years after the creation of the Protein Data Bank
(PDB) [16], the first TMP was successfully solved by x-ray crystallography in 1984
[6] (Figure 1.1C). Because of this discovery Johann Deisenhofer, Robert Huber, and
Hartmut Michel won the Nobel prize in chemistry in 1988 for solving with atomistic
resolution the 3D structure of a photosynthetic reaction centre [17]. In the following
decades, x-ray crystallography of TMPs, despite the challenges therein, remained the
predominant method of elucidating the structures of TMPs [18]. Currently, we are in
a revolution of TMP structural acquisition, which will be discussed further in section
1.1.4.

1.1.2

Transmembrane proteins in disease

Membrane-bound proteins underpin almost every biological process directly, or indirectly, from photosynthesis to respiration. Integral TMPs are encoded by between a
third to a half of the genes in the human genome [19–21] and account for 40% of drug

1.1. TRANSMEMBRANE PROTEINS

23

targets [22] which reflects their biological and physiological importance. TMPs allow
biochemical pathways that traverse the various biological membranes used in life, either by transporting molecules or transducing signals across the bilayer. Misfolding of
these proteins during the complex process of translocation is common in both normal
and diseased cells [23]. The mechanisms of folding are actively studied even recently
revealing new systems of translocation, such as the role of the ER Membrane protein
Complex (EMC) in both co-translational [24] and post-translational [25] TMP integration. Mutations tend to cluster in the TMH regions of the genes [26], which can
cause the misfolding. Glycine to arginine mutations occur statistically more frequently
in TMHs compared to their globular α-helix counterparts [27]. This can result in a
range of diseases including but not limited to: cystic fibrosis caused by mutations in
the cystic fibrosis conductance regulator [28], Charcot-Marie–Tooth disease caused by
mutations in the PMP22 gene [29] and the connexin 32 gene [30], diabetes insipidus
caused by mutations in the aquaporin 2 water channel [31], retinitis pigmentosa caused
by a point mutation in rhodopsin [32], and Niemann-Pick disease for which there are
over 200 disease-causing mutations identified [33–36].

1.1.3

The transmembrane protein problem

Studying TMPs in a laboratory typically presents several challenges not so often faced
when investigating their globular counterparts which are pointed out in a review by
Seddon et al., 2004 [37]. TMPs are typically difficult to obtain in a useful quantity.
They do not exist in high enough quantities naturally, and overexpression often results
in aggregation in the cytoplasm due to the hydrophobic nature of the TMHs. Also,
the complexity of their native environments limits structural studies. Even relatively
simple lipid bilayers cannot be probed by conventional structural techniques, so study
in the native state by traditional structural methods is difficult and involves suspending the protein in detergent or lipid environments, leading to difficulties in sample
preparations. Their insolubility in aqueous solutions also leads to the requirement of
specialised synthetic in vitro systems which have been difficult to introduce purified
proteins into [37].
In addition to the practical difficulties in studying TMPs, it has become clear that
they have a rich structural diversity. Throughout the 1990s the concept of a TMH

24

CHAPTER 1. INTRODUCTION

was simple and fairly assured: they were greasy peptides of around 30Å in length,
often bundled together and oriented perpendicularly to the membrane [38]. By 2007,
crystallography had elucidated 368 TMP structures from 148 unique proteins [18].
Whilst this accumulation of structures was growing exponentially, at the same time,
there were in excess of 50,000 protein structures in the PDB. What became clear when
examining the newly available large number of TMPs was that although the classic
TMH structures were broadly prevalent, these structures contained a plethora of unusual TMHs [38]. Transmembrane Segment (TMS)s are capable of partial spanning
of the membrane, spanning using oblique angles, and even lying flat on the membrane
surface [38, 39]. The classical model was incomplete. Even recently, there is a contingency in the TMP biology field that despite progress over the last decade there is still
a lack of information regarding the relationship between TMH sequences and function,
TMH structure, intra-membrane TMP assembly, and the behaviour of TMHs in the
lipid bilayer; the native biological environment of TMHs [1].
Furthermore, the insertion and formation of the unusually orientated TMHs and
of the more traditional TMHs have been shown to be underpinned by complex thermodynamic equilibrium and electrostatic interactions [40–42]. As well as being a
biophysically convoluted system, TMHs are biologically functional beyond anchoring
in many cases. TMSs have been identified as regulators of protein quality control and
trafficking mechanisms, shifting the idea away from TMHs exclusively functioning as
anchors [43], and crucially this function beyond anchoring can be revealed by sensitive
analysis of the sequence information alone [44, 45].
There is also an issue of silent homology extension from α-helices [46]. TMHs often
share sequence composition similarities due to membrane requirements and restraints,
not necessarily due to common ancestry which underpins homology-based domain
libraries like Pfam [47] and SMART [48]. It was found that between 2.1% and 13.6%
of Pfam hits for Signal Peptide (SP)s or TMSs are positive hits that were misreported
as homologous among distantly related proteins in Pfam [46].

1.1.4

The transmembrane protein revolution

Over the last decade or so, several significant steps have been made toward overcoming
the challenges in solving the structures of TMPs. Firstly, improvements in expression

1.1. TRANSMEMBRANE PROTEINS

25

and purification occurred thanks to cleavable green fluorescent protein (allowing protein tracking during purification) [49, 50] and antibody tagging for purification [51] at
the colony level [52] greatly increased the throughput of TMP expression and purification.

Membrane protein

Native-like lipid bilayer

Scaffold

Figure 1.2: The structure of SecYE in a nanodisc at near atomic resolution. A single
particle cryo-EM structure of SecYE in complex with the ribosome (not shown) with the TMHs held
in a reconstituted nanodisc (PDB code 4v6m [53]). The resolution is 7.1 Å. The protein structure
is depicted by a rainbow coloured cartoon in the centre of the assembly. The lipids are modelled
as spherical atoms coloured by atom (carbon and hydrogen are white, oxygen is red and nitrogen is
blue). The peptide scaffold is shown in pink.

The folding of the TMP often depends on the lipid environment. Finding conditions
that encourage protein stability was shown to improve crystallisation [18, 54]. The
technology behind nanodiscs was originally published in 2002 [55]. The method was
originally shown to be a viable method for obtaining native condition TMP crystals
for cytochrome P450 [56]. The structure of the nanodisc is formed from two scaffold
peptide helices wrapped around a group of self-associating phospholipids for which
the composition can now be tuned [57] (Figure 1.2A). Nanodiscs have been routinely
used to much more easily obtain crystal structures in native-like condition. Nanodiscs
overcome some of the major challenges caused by the hydrophobic helices and a more
faithful representation of the biological membranes than alternative model membranes

26

CHAPTER 1. INTRODUCTION

like liposomes which allows investigation into TMP-lipid interactions [58].
Still, the challenge remained of getting crystals of sufficient quality for x-ray diffraction. Although many TMPs readily formed low-resolution 2D crystals for cryo-EM [59,
60], and genome sequencing resulted in an exponential increase in structures derived
from x-ray crystallography [61], the dependency on crystallising TMPs has always
resulted in a hindrance.
In a full circle from the EM studies of the fluid mosaic model in the 1970s [7], once
more EM techniques were applied to TMPs albeit now in single molecule systems.
This method escapes the need for TMP crystals [59]. The resolution and dynamics of
structures have been captured in unprecedented detail in EM systems in recent years
thanks, in part, to the development of direct electron detectors which are radiation
hardened CMOS sensors that have deposed as the highest quality EM method plate
cameras and CCD detectors [62]. The elucidation of the sub 3Å structures utilising
this method [63, 64] have exceeded the near atomistic resolutions that were called a
revolution in TMP biology not a year earlier [65].

1.1.5

The role of bioinformatics in transmembrane biology

Due to the experimental difficulties of working with TMPs, often it is only possible
to experiment on the TMPs in silico, or at least common tasks that would involve
expensive time-consuming processes become relatively cheap and non-time-intensive.
Traditionally, the role of bioinformatics has been to predict TM regions of a protein.
There are countless methods that can perform this. Many rely on statistical patterns to
identify TMHs such as the neural network approach of MEMSAT3 [66] and the Hidden
Markov models used in TMHMM [67], HMMTOP [68], S-TMHMM [69], OCTOPUS
[70], Phobius [71]. Unsatisfied with a statistical approach, Scampi seq [72] uses a
first principles method following the Kyte & Doolittle suggestion 26 years previously
that a biophysical scale can predict TMHs [73] and applies a score based on the
von Heijne biological scale [74] (scales described in section 1.3.2). Predicting the
presence of a typical TMH based on a protein sequence is currently a fairly trivial
task and typically relies on a consensus of several prediction programs, such as the
automated TRANSMEM annotation used by UniProt [75]. Yet several challenges
remain in separating SPs from TMHs [76] and correctly predicting the topology of a

1.1. TRANSMEMBRANE PROTEINS

27

TMP. Generally, topology prediction algorithms score, using various scoring methods,
the likelihood of all possible combinations of TMH topologies within the TMP, and
pick the highest scoring viable conformation.

Due to the under-representation of TMPs among protein structures, understanding
the biochemical role TMHs play within the membrane from sequence alone is often
desirable. As more data is becoming available, bioinformatic methods that use multiple
biochemical traits are being developed. Sequence information entropy in conjunction
with hydrophobicity can predict the presence of a functional role of the TMH beyond
anchoring [44, 45]. Machine learning approaches have allowed prediction of which
residues could be involved in pore lining from sequence alone [77]. Protein trafficking
signals from a combination of hydrophobicity and charged residues may be involved
in the localisation of TMHs belonging to post-translational Tail Anchor (TA) proteins
[78].

Despite the advances in the availability of TMP structures over the last decade,
most are obtained from membrane free environments via x-ray diffraction of crystals,
single-molecule cryo-EM, and NMR techniques [59, 79]. The relationship between the
membrane and TMPs is underpinned by complex thermodynamic and electrostatic
equilibrium [40]. Although sometimes there are bound lipids and detergents included
in the crystal lattice, or the TMPs are integrated into a micelle or bicelle environment
this is often insufficient to account for the complex lipodynamics of the membraneTMH interactions [79, 80].

There have been several approaches used to attempt to find a more biologically relevant structure contextualised in the membrane (Figure 1.3). OPM [83] and PDBTM
[84] use simplified models of the membrane’s hydrophobic environment to orientate
the TMP into position. MemProtMD [79] uses explicit lipids, a phosphatidylcholine
bilayer, in a flexible Molecular Dynamics (MD) simulation along with TMPs accessed
from the PDB [82]. These approaches allow us to enhance our understanding of the
TMP structures in the context of the membrane environment.

28

CHAPTER 1. INTRODUCTION

A) MemProtMD

C) OPM

B) PDB / OPM

D) PDBTM

Figure 1.3: A comparison of the raw outputs from commonly used structural transmembrane protein databases. The example used here is a ligand-free G protein-coupled receptor
(GPCR) rhodopsin (PDB 3cap) [81]. A) The structure held by MemProtMD [79]. A cartoon of the
structure coloured according to hydrophobicity. Red indicates hydrophobic residues and the cream
colour represents hydrophilic residues. Note the flexing and distortion of the membrane. B) The
RCSB PDB [82] showing the hydrophobic patches on the surface (yellow is hydrophobic, cyan is
hydrophilic) of the membrane proteins using OPM for topological orientation [83]. The blue lines
indicate the cytoplasmic side whilst the red indicates the extracellular side. C) The cartoon structure
of 3cap from OPM [83] coloured by the protein chain. Again, the blue lines indicate the cytoplasmic
side whilst the red indicates the extracellular side. D) The PDBTM [84] record coloured by the
TM regions predicted by TMDET [85] in yellow, intramembrane regions in green. In this case, the
non-membrane components are coloured according to side 1 and side 2 rather than a topological
prediction.

1.2
1.2.1

Biological membranes
Membrane lipids

It is critical to understand that the lipid bilayer and the transmembrane α helices are
inextricably linked. Often TMHs reflect the membrane environments they exist in, for

1.2. BIOLOGICAL MEMBRANES

29

example, the hydrophobicity, asymmetry, and thickness of the membrane lipids are
reflected when looking at large numbers of TMPs [86].
The lipid membranes influence the local structure, dynamics, and activity of proteins in the membrane in non-trivial ways [87, 88], as well as protein folding via
repositioning of TMHs [89]. For example, cross-linking a β secretase, BACE, with
lipid rafts revealed that protein activity is modulated by lipids [90]. Furthermore,
bilayer deformations induced by rhomboid protease, a polypeptide cleavage protein,
allow substrate access to the active site [91].
The lipids that make up these membranes are very diverse. Not only do different
cells have different membrane compositions but so do different subcellular compartments which again is reflected in the TMH composition of TMPs [86, 92].
Individual lipid molecules consist of a polar head group and a hydrophobic region
and so are described as being amphiphilic molecules. This factor, along with geometric
constraints of the membrane lipids causes them to not only be insoluble molecules but
also that they readily self-associate into complex ordered structures that have been
demonstrated in silico with coarse grain MD simulations [93]. This self-association is
entropically driven by water molecules. There is a rich variety of lipid molecules that
make up the biological membranes. Furthermore, the bilayer maximises van der Waals
interactions between the closely-packed hydrocarbon chains, which contributes to the
stability of the bilayer. This can be seen even in relatively early MD simulations [94].
The majority of lipids in higher organism membranes are phospholipids, sphingolipids, and sterols. The hydrophobicity of a membrane lipid can be caused by
several features; (i) aliphatic chains such as in the most abundant membrane lipids,
glycerophospholipids and the sphingolipids, (ii) aromatic groups, or (iii) polycyclic
structures such as the sterols most abundantly of which in mammals is cholesterol
[95, 96]. Phospholipids are composed of a glycerol molecule. Bonded to the glycerol
molecule are two hydrophobic fatty acid tail groups and a negatively-charged polar
phosphate group. The polar phosphate group is modified with an alcohol group. Often
these are represented as a polar head group with two fatty chains extending into the
membrane, and a bilayer is formed from two sheets of these molecules with the polar
headgroups facing outwards towards either side of the membrane and the fatty side
chains filling the interior of the membrane space (Figure 1.1D).

30

CHAPTER 1. INTRODUCTION
It has been known for almost 4 decades that biological bi-layer membranes are

asymmetric [7, 97]. For example, in the outer membranes of Gram-negative bacteria,
the outer membrane leaflet contains lipopolysaccharide, whilst the inner is a mixture of
approximately 25 phospholipid types [92]. Adding to the membrane asymmetry composition story, a thorough analysis of residue composition in yeast and human TMH
regions revealed intra-membrane leaflet composition asymmetry in the Endoplasmic
Reticulum (ER), but not the Golgi [86]. Furthermore, protein-lipid interactions have
been shown to be determinants of membrane curvature [88], and undertake complex
orientations and conformations to allow for hydrophobic mismatch [98].

1.2.2

Membrane potential

Simply put, membrane potential is the voltage across a membrane. If the membrane
is permeable to a certain type of ion, then the ion will experience an electrical pulling
force during the diffusion process that pulls toward the “preferred” biological location.
This clearly depends on a chemical component involving both the charge and ion
concentration gradient. There are various ways of estimating the membrane potential.
The Nernst equation can be derived directly from the simplified thermodynamic
principles (i) the Boltzmann distribution, and (ii) a field charge interaction energy [99].
It is defined as:
RT
cout
× ln
(1.1)
zF
cin
is the membrane potential, z is the ion charge, c is the conEm =

Where charge Em

centration of an ion in that cell environment, R is the ideal gas constant, T is the
temperature (Kelvin), and F is the Faraday constant per molecule.
However, the Nernst equation is rife with caveats caused by the assumptions of
the simplified model. Such assumptions include ions having point charge and that the
potential is constant throughout the solution. This issue is compounded because it
assumes the constant potential is the same as the point of measurement which can be
heavily influenced by, for example, a specific adsorption of either part of the redox pair
or the competitive adsorption of a supporting ion in solution [99]. Considering that in
biology the compartments always involve multiple ion channels and constant flux of
biochemical environments, one should be cautious to understand the limitations and

1.3. α HELICES IN THE MEMBRANE; STRUCTURE AND FUNCTION

31

variability when extrapolating experimentally determined E0 particularly when using
such an idealised model in a biological context.
Several studies have attempted to quantify the various voltages across the intracellular membranes. Negativity was found in the ER, with a voltage between 75mV
to 95mV in the ER membrane [100, 101]. Negativity was found in the mitochondrial
matrix with a voltage across the mitochondrial membrane at 150mV [102]. No notable
membrane potential has been identified in the Golgi [103, 104].

1.3

α helices in the membrane; structure and function

1.3.1

Transmembrane helix sequence composition

TMHs are α helix structural elements of membrane proteins that span the membrane.
Measurements of the TMH regions have found that they are roughly 20 residues in
length; 17.3±3.1 from 160 TMHs [105], 27.1±5.4 residues based on 129 TMHs [106],
26.4 residues based on 45 TMHs [107], 25.3±6.0 residues based on 702 TMHs [108],
24.6±5.6 from 837 TMHs [109], and 28.6±1.6Å to 33.5 ±3.1Å from 191 proteins depending on membrane types [110]. There are a couple of reasons for this variation.
Primarily is that the boundaries of TMHs are extremely hard to precisely identify
since it is unclear exactly how far the TMH rises into the water interface region [38].
Secondly is that it is emerging that different membranes have different thicknesses [92]
and that this is directly reflected in the hydrophobic lengths of the TMH [86, 110].
A study by Baeza–Delgado et al. from 2013 [109] looked at TMHs with known
topology and structure in 170 integral membrane proteins from a manually maintained
database of experimentally confirmed TMPs; MPTopo [112]. The group examined the
distribution of residues along the TMHs. As expected, half of the natural amino acids
are equally distributed along transmembrane TMHs whereas aromatic, polar, and
charged amino acids along with proline are biasedly near the flanks of the TMHs [109].
It has been noted that transitions between the polar and non-polar groups at the ends
of the hydrophobic core occur in a more defined edge on the cytoplasmic side than at
the extracytoplasmic face when counting from the middle of the helix outwards [109].

32

CHAPTER 1. INTRODUCTION

Outer leaflet

Hydrophobic
groups

Inner leaflet

Membrane

Outside
(non-cytosolic)

Polar
groups
Positive
charge
Inside
(cytosolic)

Figure 1.4: A cartoon showing the general components of the membrane and a typical
transmembrane helix. The example used here for illustrative purposes is the transmembrane region
of therein (PDB 2LK9) [111]. Dark grey areas denote the area of lipid head groups. The residues
found in these areas are often described as flanking regions and are often in contact with the aqueous
interface of the membrane. The helix core is mostly composed of hydrophobic residues. Although the
regions labelled here generally hold true in terms of the statistical distribution of polar, non-polar,
and charged groups, it is by no means absolute laws and many proteins break these “rules” [86, 109,
110].

This is probably reflecting the different lipid composition of both leaflets of biological
membranes [109].
A previous study by Sharpe et al. from 2010 used 1192 human and 1119 yeast
predicted TMHs that were not structurally validated to further explore the difference
in TMH and leaflet structure by exploiting the evolutionarily conserved sequence differences between the TMH in the inner and outer leaflets [86]. The differences in
consensus TMH structure implies that there are general differences between the membranes of the Golgi and ER. The abundance of serines in the region following the
lumenal end of Golgi TMSs probably reflects the fact that this part of many Golgi
enzymes forms a flexible linker that tethers the catalytic domain to the membrane [86].

1.3. α HELICES IN THE MEMBRANE; STRUCTURE AND FUNCTION

33

The “positive-inside” rule
Two publications by von Heijne coined the “positive-inside” rule demonstrated the
practical value of positively-charged residue sequence clustering in topology prediction
of TMHs in bacteria [113, 114]. It was clearly defined and shown that positivelycharged residues more commonly were found on the “inside” of the cytoplasm rather
than the periplasm of E. coli. This holds true in the human, where the “inside” is the
cytoplasm. In the case of the mitochondria, net positive charge was found to localise
in the matrix [115] and so the intermembrane space is considered “outside” and the
matrix can be considered “inside” (Figure 1.5).

Mitochondria

Plasma membrane
Endoplasmic Reticulum
Outside
Golgi
Inside
“Outside”

“Outside”
“Inside”

“Outside”

Inside

Nucleus
Positively charged residues near the transmembrane
regions of proteins cluster on the inside
Figure 1.5: A cartoon of the topological clustering of positively-charged residues near
transmembrane helices in relation to subcellular location. First identified across the inner
bacterial membrane from the periplasm to the cytoplasm, the “positive-inside” rule generally is in
relation to inside the cytoplasm and outside the cell across the cellular membrane. The clustering
pattern also applies to other cellular locations denoted by “inside” and their opposing environment
across a membrane shown by the “outside” annotation. Throughout this thesis, “inside” and “outside” generally refer to inside and outside the cytoplasm unless explicitly stated otherwise.

More recently still, large-scale sequence analysis of TMHs from different organelle
membrane surfaces in eukaryotic proteomes, show the clustering of positive charge
being cytosolic [86, 109, 110].

34

CHAPTER 1. INTRODUCTION

The aromatic belt
Tyrosine and tryptophan residues commonly are found at the interface boundaries
of the TMH and this feature is called the “aromatic belt” [86, 109, 116–118]. Not
all aromatic residues are found in the aromatic belt; phenylalanine has no particular
preference for this region [117, 119]. However, it remains unclear if the aromatic belt
is to do with anchoring or translocon recognition [109].
A study of conserved tryptophan residues during folding of integrin αIIβ3 TM
complex demonstrated the anchoring effects of tryptophan (0.4 kcal/mol contribution
to membrane stability) in TMHs is greater than the other residues [120]. It was
suggested that the wide amphiphilic range (the stabilising energetic contribution in
either hydrophobic or polar sites) of tryptophan complements the heterogeneity and
asymmetry of mammalian membrane lipids in particular.
Phenylalanine, although aromatic, is hydrophobic, and unlike tyrosine and tryptophan is typically found in the transmembrane part rather than the interfacial parts of
TMHs.
The tyrosine side chain is a 6-carbon aromatic ring with an OH group attached.
Tryptophan has two aromatic rings that contribute to a large hydrophobic ringstructure. The aromatic rings of tryptophan and tyrosine are buried close to, or
within, the hydrophobic core, while the hydrophilic portion can interact with the polar lipid head-groups at the interface between the lipids and the water environment.
Other factors such as the aromaticity, size, rigidity and shape of tryptophan, rather
than its dipolar character, has also been suggested as the primary reasons for its interfacial preference and indeed interfacial localisation preference could be the result of
a combination of all of these factors [121].
Snorkelling
Broadly speaking, TMHs are non-polar. However, some contain polar and charged
residues in the helix itself. Whilst this might seem thermodynamically unstable at first
glance, a molecular dynamic feature called the “snorkel” effect explains in part how
this is possible [122, 123]. Simply put, the snorkelling effect involves the long flexible
side chain of lysine reaching out of the inner core of the membrane environment to the
water interface region to interact with the polar head-groups of the bilayer even when

1.3. α HELICES IN THE MEMBRANE; STRUCTURE AND FUNCTION

35

the α helix backbone is pulled into the hydrophobic layer [124]. This has also been
suggested to allow helices to adapt to varying thicknesses of the membrane [125]. More
recently it was found that although in simulations the energetic cost of arginine at the
centre of the TMH is large, in vivo experimentation with the Sec61 translocon reveals
a much smaller penalty [126]. That same study also found that in MD simulations,
snorkelling, bilayer deformation, and peptide tilting combined to be sufficient to lower
the thermodynamic stability penalty of arginine insertion so that hydrophobic TMHs
with a central arginine residue will readily insert into the membrane.

1.3.2

The hydrophobicity of transmembrane segments

Perhaps the most prevalent and important feature of the transmembrane regions is
the membrane-spanning region which is composed mostly of non-polar residues. The
importance of hydrophobicity on the effectiveness of membrane anchoring has been
known for some time [127]. More recently the hydrophobic group region has been
associated with cell localisation and a broad range of biochemical functions [45, 86,
128]. Once inserted the protein does not leave the membrane as a result of the TMH
being very hydrophobic. This hydrophobicity of TMHs and the hydrophobicity of the
lipid tails means that they self-associate and this association is entropically driven
by water. Another way of describing it is that they dissociate from the water. The
overall ∆G for an ideal TMH in the membrane is -12 kcal/mol [40]; the association of
the helix in the membrane is typically spontaneous but complex owing to interactions
with other TMHs in the membrane as well as TMP-lipid interactions.
Over the last 50 years or so, there have been many attempts to use hydrophobicity scales of residues to predict structural classifications of proteins. Due to the vast
amounts of scales, major efforts have been made to compare them to identify which
ones are better for which tasks of identifying structural elements [129, 130]. Simm
et al. 2016 [129] compared 98 scales and found that the accuracy of a scale for secondary structure prediction depends on the spacing of the hydrophobicity values of
certain amino acids but generally that the methods behind the scales do not affect the
separation capacity between β sheets or α helices.
Throughout this thesis, several scales are used to evaluate and estimate the hydrophobic values of peptide chains. All the scales aim for quantifying the hydrophobic

36

CHAPTER 1. INTRODUCTION

values of each residue. There are several key differences in their methodology, assumptions, and aims. Ultimately, all the scales are attempting to allow estimation of
∆Gwhf ; the free energy of a folded helix (f ) from the water (w) into the membrane
core (h). This free energy measurement is regarded as being currently experimentally
inaccessible [40].
Although as a trend most of the scales agree, because of the methodological differences, there are indeed variations of values even after normalisation. Due to these
discrepancies, it is preferable and typical amongst the literature to use several scales
to verify the observable trends resulting from interpretation from an individual scale.
Notably, one of the classic scales, Kyte & Doolittle Hydropathy scale shows a striking
similarity to the modern Hessa’s ∆Gaa
app scale, and that generally the “better” scales
for TMH prediction count proline as hydrophilic, and focus on helix recognition rather
than amino acid analogues [130]. Several of the scales used throughout this thesis are
overviewed below.

Kyte & Doolittle hydropathy scale
The Kyte & Doolittle scale [73] is based on the water–vapour transfer free energy
and the interior-exterior distribution of individual amino acids determined in previous
investigation [131]. Kyte & Doolittle gave a composite score to each amino acid based
on evidence from previous experiments in the literature, scaling them between -4.5
and +4.5. These experiments included the molal volumes of model compounds for the
side-chain R groups [132–134], the water vapour partition coefficients [132, 135], and
water-ethanol transfer free energies [136, 137]. However, alanine, tyrosine, leucine,
and proline scores were subjectively modified. The authors found it difficult to accept
that the single methyl group from alanine would have more hydrophobic force than
leucine, which contains 4 methyl groups so the score was arbitrarily lowered to halfway
between the originally determined score and the score of glycine. Tyrosine and leucine
had their hydropathy values raised to one closer to the water vapour transfer free
energy than experimental data suggested. In the case of proline, no suitable model
existed and it tends to become buried, indicating that it is fairly hydrophilic. However,
because it contains three methyl groups, the score considers it more hydrophobic than
the experiments suggested. The authors stated: “None of these last 3 adjustments,

1.3. α HELICES IN THE MEMBRANE; STRUCTURE AND FUNCTION

37

the result of personal bias and heated discussion between the authors, affects the
hydropathy profiles in any significant way”[73]. Arginine represents the bottom of the
scale, arbitrarily set at -4.5.
Hydrophobic, but not long enough
to indicate a TMH.

Hydrophobic, span indicative of a
TMH.

Figure 1.6: The hydropathic index of rabbit cytochrome b5. The hydropathic index according to the Kyte & Doolittle scale [73] is given on the vertical axis and the sequence number is on the
horizontal axis. Whilst there are several peaks above 0, only one is sustained for around 20 residues
near the c-terminal. Cytochrome b5 is a TA protein with an exceptionally polar TMH for an anchor
[138], demonstrating the sensitivity of this method. Figure adapted from Kyte & Doolittle, 1982 [73].

An algorithm called SOAP was used to scan a window of ±10 residues either side
of each residue (allowing for half windows) across a protein to take the mean average
of each position and plot them on a profile (Figure 1.6). Sustained areas of high
hydrophobicity were indicative of a TMH [73].

Hessa’s biological hydrophobicity scale
This is arguably the most biologically relevant scale [130] and is often called the ∆Gaa
app
scale. The scale is based on an experimental method where the free energy exchange
during recognition of designed poly-peptide TMH by the ER Sec61 translocon occurred [116]. These measurements were then used to calculate a “biological hydrophobicity scale.” The original study reported positional variance in some residues and is

38

CHAPTER 1. INTRODUCTION

strictly valid only for residues in the core of the TMH. A more refined study quantified
the positional dependencies of each amino acid type [74].

White and Wimley octanol – interface whole residue scale
This scale is calculated from two other scales; the octanol scale, and the interface
scale [139]. This scale is fundamentally based on the partitioning of host-guest pentapeptides (acetyl-WL-X-LL-OH) and another set of peptides (AcWLm) between water and octanol, as well as water to Palmitoyloleoylphosphatidylcholine (POPC).

The Eisenberg hydrophobic moment consensus scale
The Eisenberg scale is a consensus scale based on the earlier scales from Tanford [137],
Wolfenden [140], Chothia [131], Janin [141], Wolfenden [142], and the von Heijne
scale [143]. The scales are normalised according to serine [144]. The automatic
TRANSMEM annotation currently used in UniProt is according to TMHMM [67],
Memsat [66], Phobius [71] and the hydrophobic moment plot method of Eisenberg
and co-workers [144].

1.3.3

Sequence complexity

Whilst identifying TMH presence in a protein is somewhat trivial with modern tools,
the role a TMH plays within the membrane is more difficult to determine. Sequence
properties that can be analysed by bioinformatics, the sequence information complexity and hydrophobicity, of the TMH have been used to predict the role of the TMH as
either functional or structural, and as a discrete cluster from other SCOP-annotated
helices [45]. Those findings demonstrated that the sequence of the TMH holds valuable
information regarding biological roles, and forms the basis of our interest in the link
between the polarity of a helix and functional activity beyond structural anchoring.
TMSOC’s z-score is able to distinguish between functionally active TMHs and those
only associated with anchoring [44, 45]. This was determined with the annotation of
TMHs from the UniProt database (303 membrane anchors and 1741 functional TM
helices) from UniProt [75] and tested with 181132 TMH sequences. Two peaks can
be observed when stratifying the hydrophobicity and the sequence complexity of the

1.3. α HELICES IN THE MEMBRANE; STRUCTURE AND FUNCTION

39

TMHs (Figure 1.7). These peaks correspond with the “simple” and “complex” TMH
annotation.
A)

B)

Figure 1.7: The hydrophobic-complexity continuum distinguishes between transmembrane helix anchors and those with function beyond anchoring. A) Two distinct points
can be observed when plotting sequence complexity on the x-axis, frequency on the y-axis, and hydrophobicity on the z-axis of TMHs from UniProt. These peaks were shown to correspond to the
membrane anchor and functional TMH populations. From Wong et al., 2011 [44]. B) The α helix hydrophobicity complexity continuum stratified by hydrophobicity (vertical axis) and complexity
(horizontal axis). In blue are the membrane anchors, in red are the TMH with function, and in green
are non-TM α-helices from SCOP [145]. Clear distinctions can be made between the 3 groups and in
fact, a composite z-score of the distance of these two factors from the functional TMH group allows
prediction of which group a TMH belongs to [44, 45]. From Wong et al., 2012 [45].

The z-score is a product of both hydrophobicity and a Shannon like sequence
entropy [44, 45] of the character string in the TMH. This term is described below in
equation 1.2.
"
z(xΦ , xc ) = (−1)s

(xΦ − µΦ )2 (xc − µc )2
+
σφ2
σc2

#
(1.2)

Where xc and xΦ are moving window averages of c, the sequence entropy [146]. Φ
is the White and Wimley hydrophobicity [139] for a given segment and µ and σ are
the mean and standard deviation of the sequence entropy and hydrophobicity of the
functional TMH set, that is those TMHs containing active residues. s is set to equal
to one if equation 1.3 is true, and zero otherwise.

Where ρc,Φ

xΦ − µ Φ
1 xc − µ c
≥−
(1.3)
σΦ
ρc,Φ σc
is the correlation between sequence complexity and normalised hy-

drophobicity. Equation 1.3, above, is the normal to the regression line between that

40

CHAPTER 1. INTRODUCTION

can be read as in equation 1.4.
xΦ − µ Φ
x c − µc
= ρc,Φ
σΦ
σc

(1.4)

Sequence entropy is essentially an estimate of the linguistic entropy of a string. In
the context of biology can be thought of as an estimation of the non-randomness of
a sequence. Sequence complexity can be used to analyse DNA sequences [147–149],
however here we will focus on the analysis of the complexity of a sequence in protein
sequences.
The SEG algorithm [146, 150] finds stretches of non-random amino acids; subsegment regions of the protein which have the lowest probability of occuring. It uses a
Shannon entropy-like term [151] which identifies regions with high information entropy
density. It can distinguish between globular and non-globular regions of a protein
sequence [150]. Furthermore, by altering the window lengths and thresholds SEG
can be optimised to search for subtle compositional deviations, such as coil-coiled
regions [152]. However, sequence complexity alone fails to fully discriminate in the
low complexity high hydrophobicity search space of TMHs [46], hence the requirement
for the hydrophobic scoring contribution (Figure 1.7).

1.4
1.4.1

Biogenesis of transmembrane proteins
An overview of translocation

In this thesis we focus on cotranslational protein translocation and insertion via the Sec
pathway, and on the post-translational pathway of TA proteins in eukaryotic systems
(Figure 1.8).
The Sec pathway is conserved across all life [153] and is facilitated by the heterotrimeric SecYEG (prokaryotic) and Sec61αβγ (eukaryotic) translocons. SecYEG
targets proteins to the Plasma Membrane (PM) (inner membrane) of bacteria. In
bacteria, nearly all TMPs are inserted by either the Sec pathway via SecYEG machinery, or by the structurally unrelated the 5-fold more abundant YidC machinery [154,
155], and proteins can often use either pathway. Sec61 targets proteins to the ER
of eukaryotic cells. Broadly speaking, these proteins translocate hydrophilic peptides

1.4. BIOGENESIS OF TRANSMEMBRANE PROTEINS

41

Ribosome

Co-translational path

SRP
Post-translational
path

Chaperone
complex

SR

ER

Membrane insertion
machinery

Sec machinery

Figure 1.8: A simplified schematic of the co-translational Sec pathway and the posttranslational pathway. The ribosome translates Ribonucleic Acid (RNA) to a nascent polypeptide
chain. In the case of post-translational insertion (the left-hand-path), the nascent peptide is released,
albeit briefly, to the cytosol. Chaperones of various types then shield the protein TMH and bring
it into contact with various TM bound translocation machinery. In the case of co-translation (the
right-hand-path), the Signal Recognition Particle (SRP) binds to the emerging signal sequence of
the nascent peptide, which slows protein synthesis of the ribosome. SRP then targets the ribosome
protein complex to the Signal Recognition Particle Receptor (SR), which then moves into contact
with the translocon. As the nascent protein is inserted into the translocon, SR and SRP detach from
the complex and translocation begins.

across a membrane whilst also integrating sufficiently hydrophobic sequences to the
membrane [40, 128, 156, 157]. SecYEG and Sec61 also translocate soluble proteins.
In bacteria, the SecY protein is partly encircled by SecE (Sec61γ in eukaryotes) which
enhances its stability in the membrane [158]. SecG is made up of two TMHs and a
cytosolic loop. It has been associated with SecA dependent translocation of secretory
proteins [159, 160]. SecA is an ATPase and a membrane channel that can clamp
translocating substrates above the SecY pore [161].
Whilst ultimately the translocon machinery is built of multiple subunits, the TM
translocating protein itself is a 10 TMH protein with an hourglass-shaped interior
pore [163] (SecY in prokaryotes or Sec61α in eukaryotes). When translocation is not

42

CHAPTER 1. INTRODUCTION
B) Translating state

A) Idle state

Ribosome
Sec61/SecY

Nascent
peptide

Constriction ring
Cytosolic

Membrane
Lumenal/Extracellular Lumenal
plug

C)

Lateral gate opens for
hydrophobic segments

PTC
Small
ribosomal
subunit

Large
ribosomal
subunit

Translocon

Figure 1.9: The pore, the plug, and the lateral gate of the translocon. A) The idle
state of the translocon. The pore constriction ring is shown by a dashed transparent cylinder. The
plug is in red and the translocon (SecY or Sec61) is in grey. B) The translocating state of the
translocon coloured similarly to A, but the plug has moved to allow translocation through the channel,
the constriction ring has widened, and the lateral gate opens to allow hydrophobic TMHs into the
membrane environment. C) A sphere representation of the translocon-ribosome complex structure
during translocation. Highlighted are the Sec61 translocon (orange) the ribosome made up of the
large 60s subunit and the small 40S subunit and the peptidyl transferase centre (denoted by PTC),
the site of translation (PDB 3J7R [162]). Part C redrawn from Voorhees et al., 2014 [162].

occurring and the protein is idle, the hydrophobic core is constricted [128] and a
lumenal plug is in place over the pore resulting in a seal across the TMP pore [164,

1.4. BIOGENESIS OF TRANSMEMBRANE PROTEINS

43

165] that protects the membrane from ion flow [166] (Figure 1.9). SPs in elongationstalled nascent chains crosslinked to the lipid and to the translocon machinery [167],
showing that there may be a holding space which exposes the TMH to the membrane
interfacial environment before insertion is complete and this was verified by more
recent EM structural studies [156, 168].
During translation the 6-polypeptide single RNA ribonucleoprotein complex SRP
binds to hydrophobic signal sequences, typically at the N-terminal of the nascent
peptide, emerging from the ribosomal exit tunnel and translation elongation is slowed
[169]. The SRP targets the ribosome-SRP complex to the target membrane (either
the PM in bacteria or the ER in eukaryotes) [170, 171]. The SRP engages with
the membrane-bound SR, which mediates the association of the ribosome and the
translocon [172]. As the ribosome binds with the translocon and the signal sequence is
released [173], SRP54 moves away from the ribosomal protein L23a [174], the nascent
peptide is brought into proximity of the translocon [175] and translocation begins
(Figure 1.8).
During translocation the plug exits the pore, the constriction ring is opened, and a
lateral gate opens between TMH2 and TMH3, and TMH7 and TMH8 to integrate hydrophobic stretches of nascent peptide into the membrane [163, 168, 176–181] (Figure
1.9). Mutation experiments on these three features (the plug, constriction ring, and
lateral gate) destabilised the closed state of the translocon and resulted in a protein localisation phenotype that suppressed inactivating mutations in signal sequences [182–
185] and even caused transient channel openings in bacteria [186].
The EMC can act as a post-translational TA protein insertase [25]. Recently it was
shown that the EMC was also able to act cooperatively with the Sec translocon during
cotranslational insertion on a subset of multipass TMPs in both yeast and human cells
[24]. Proximity specific ribosome profiling in yeast confirmed that the EMC engages
the translocating multipass TMPs, especially those containing charged residues in
the TMH. This would explain why it also interacts with TA protein TMHs [24, 25].
Pulldown experiments revealed that the EMC stabilised the synthesised TMPs and
recruited chaperones (both substrate specific and general) that allowed unstable TMPs
to achieve viable conformations allowing function in the ER and transport to the Golgi
[24]. In the absence of the EMC, UPS (a proteasome) degrades the membrane proteins

44

CHAPTER 1. INTRODUCTION

[24]. This dual function of the EMC is analogous with YidC [24] in bacteria which has
the role of an insertase [154, 155, 187] as well as interacting with SecYEG to stabilise
translocated TMPs [188].
Additional proteins associated in cotranslational insertion during the TMP integration by Sec61 are the membrane-bound TRAP and TRAM protein complexes. TRAP,
otherwise known as the translocon-associated protein complex, is a multi-subunit complex. The TRAP complex facilitates the translocation of TMHs by moderating the
“positive-inside” rule, but does not seem to play a role in the integration of very hydrophobic TMHs [189]. TRAM, also called the translocating chain-associated membrane protein, is critical to the efficient translocation of TMPs [190], but has yet to be
fully structurally solved [191]. During translocation, the association between TRAM,
TRAP and the Sec61 translocon is strengthened [191].
Another group of proteins capable of Sec61 TMP integration are the membranebound Sec62 and Sec63 proteins. In mammals, these proteins associate with Sec61 and
facilitate membrane integration independently from the SRP-ribosome complex [192].
Post translational translocation in this case involves ATP hydrolysis driven Hsp70
chaperones delivering the unfolded TMP [193, 194] to the Sec61-Sec62-Sec63 machinery [195]. It was shown using crosslinking and co-translational and post-translational
translocation assays that SR displaces Sec62 which returns the Sec61 to the SRP dependent insertion pathway; Sec62 and ribosomal associations with Sec61 are mutually
exclusive [196].

1.4.2

Signal peptides

SPs are short peptide sequences present at the N-terminus of most secretory pathway destined proteins in both prokaryotes and eukaryotes. Mutations to the length
and composition of the SP revealed their composition is essential for effective localised transport, however, they have high compositional variability [197]. Like TMHs
they translocate into the membrane due to their hydrophobic α helix (albeit typically
shorter than a TMH) and follow the positive-inside rule.
Once the protein is fully translocated and at the correct location within the cell, a
three domain signal peptidase cleaves the SP from the rest of the protein. SPs can also
be experimentally interchanged to proteins from different organisms to target them to

1.4. BIOGENESIS OF TRANSMEMBRANE PROTEINS

45

Hydrophobic residues

Protein
sequence

N-terminus

Hydrophilic positively
charged residues

C-terminus

Cleavage site

Figure 1.10: The key components of a signal peptide. Signal peptides consist of between 1-5
hydrophilic charged residues (in blue) at the N terminal, a 7-15 residue hydrophobic region (in green),
and a c-region of 3-7 residues that invites cleavage by specific signal peptidase (in red). Redrawn
from von Heijne, 1990 [198].

cellular locations [199, 200].
SPs are rarely used by TMPs with the exception of most type I TMPs, which are
defined by the presence of an SP. Instead, TMPs use their first TMH and flanking
region as a targeting signal. Many bioinformatic methods over the last twenty years
have aimed to predict SPs and sort them from TMHs [201], however, these methods all
struggle to distinguish SPs from N-terminal TMHs [76]. This is due to the similarities
SPs have with TMHs, and that although TMHs are not cleaved, the cleavage site
pattern alone cannot effectively separate SPs from TMHs. This results in many false
positives for SP presence in genome-wide studies [76]. SignalP 4, a neural network
based approach, is less effective at predicting cleavage sites or SP prediction in the
absence of TMPs from the dataset, however currently is the most accurate method to
distinguish the currently vague differences between TMHs and SPs [76].

1.4.3

β sheets in the membrane

The membrane is critical in maintaining the separation of biochemically distinct compartments. However, there needs to be a flux of not only small molecules but also larger
complex structures across these membranes in order to facilitate cellular life. Whilst
the focus of this thesis is primarily on TMHs it is important to acknowledge that β
sheets are another secondary structural element capable of spanning the membrane.
β barrel proteins are present only in a few membrane types; the outer membrane of
Gram-negative bacteria for which they constitute between 2-3% of the genome [205],

46

CHAPTER 1. INTRODUCTION

B)

C)

Membrane

A)

90˚

90˚

90˚

Hydrophobic
residues

Hydrophilic
residues

β barrel pore

α helical channel

α helical signal transducer

Figure 1.11: Cartoons showing the structural differences of the outer membrane proteins, transmembrane helix channels, and transmembrane helix signal transducers in
the membrane. The top row shows the proteins traversing the membrane with the interfacial regions denoted by the black lines. A) A porous outer membrane protein using β-sheets to form a
pore-forming protein structure. PDB code 2JQY [202]. B) A highly selective ion channel made up of
membrane-spanning TMHs. PDB code 4HYO [203]. C) A 7-TMH GPCR that receives a signal on
one side of the membrane which causes a conformational change influencing the structural arrangement of the tertiary structure on the opposing side of the membrane. PDB code 1F88 [204]. Note
the pore in the β barrel protein not present in either the TMH channel or the TMH transducer.

and in the outer membranes of eukaryotic chloroplasts and the Mitochondrial Outer
Membrane (MOM). Their restriction to these membrane types is the result of ancestral
symbiogenesis [206–211].
These outer membrane proteins cause the membranes to be permeable, diminishing
membrane potential between the outer and the inner membrane. Unlike the TMHs,
β barrel proteins are permeable to water and many other small molecules often unspecifically due to their large pore (Figure 1.11). Integral β barrel proteins consist of
between 8 and 26 amphipathic antiparallel β strands forming a cylindrical pore. They
are capable of active and passive transport, molecule receptors, enzymatic catalysis,
structural formation, and translocon machinery [205].
The biogenesis of β barrel membrane proteins also differs from that of TMHs,
and there are differences between prokaryotic and eukaryotic cells. In Gram-negative

1.4. BIOGENESIS OF TRANSMEMBRANE PROTEINS

47

bacteria, the β barrel proteins are translated on cytoplasmic ribosomes (Figure 1.12).
An N-terminal SP sequence allows the protein to pass through the inner membrane to
the outer membrane [212, 213]. Instead of cotranslational insertion, SecB transports
the nascent polypeptide chain to the Sec translocon on the inner membrane [214]. βsheets do not trigger the opening of the lateral gate [211]. Once in the periplasm, the
SP is cleaved [215] and chaperones, such as the redundant SurA [216, 217], DegQ, and
Skp [217], bind to the protein to prevent premature folding. Once at the inner surface,
the BAM complex assembles and integrates the β barrel protein into the membrane
[218, 219] via a lateral opening of one of the BAM subunits, BamA [220].
A) Gram negative bacteria

B) Mitochondria
Cytosol

Chaperone
Extracellular
TOM

BAM
MOM

Outer membrane

Periplasm

TOB

Tim 8
Tim 13

Chaperone

Tim 9
Tim 10

Inner
membrane
space

MIM

Inner membrane
SecYEG

Cytoplasm

Matrix

SecB

SP

Ribosome

Figure 1.12: A cartoon of the biogenesis of β barrel membrane proteins in mitochondria and Gram-negative bacteria. This is an overview of the current model of β-barrel protein
membrane integration. A) In bacteria, SecB chaperones the nascent peptide to SecYEG where it
is transported across the inner membrane. Once the SP is cleaved, the protein is chaperoned to
the BAM machinery by SurA, DegQ, or Skp. The BAM machinery integrates the protein into the
membrane. B) After ribosomal translation, the nascent peptide is transported across the MOM by
the Translocase of the Outer Membrane (TOM) machinery. Several Tim recognition molecules then
attach to the protein and present it to the TOB insertion machinery, which integrates the protein
into the membrane. Redrawn from Ulrich & Rapoport, 2015 [211].

Unlike in the bacterial outer membrane, β-barrel proteins are rare in the mitochondria [211]. There is Tom40, Tob55, two isoforms of a porin, and Mdm10 [211]. And

48

CHAPTER 1. INTRODUCTION

yet, indeed it was shown that a β barrel protein fragments not found in the eukaryotic
proteome called trimeric autotransporter can be recognised and assembled by the mitochondria [221]. Since there is no need to cross the inner membrane, mitochondrial
proteins do not have the cleavable SP or any other known targeting signal [211]. After
cytosolic ribosomes complete the translation, the β barrel proteins are recognised on
the mitochondrial surface by the TOM complex import receptors, handed to the β
barrel protein Tom40 which is the central machinery for mitochondrial protein import
[222]. After being processed through the TOM machinery, hexameric chaperone complexes of Tim8 in complex with Tim13, and Tim9 in complex with Tim10 transport
the peptide to the TOB complex (also known as the SAM complex) also on the MOM
which integrates the β barrel protein into the membrane [223–225].
Whereas mitochondria originated from the ancestral incorporation of α-proteobacteria [207], chloroplasts were thought to originate from the incorporation
of cyanobacterium [206]. Similarly to mitochondrial β barrel proteins, there is no
cleavable targeting SP for the chloroplast since they integrate on the internal side of
the membrane, however much less is known about the precise insertion machinery of
β barrel proteins at a molecular level in chloroplasts [211].
Besides the differences in biogenesis between β barrel TMPs and TMPs with TMHs
as the membrane-spanning units, there are also biophysical differences. Membrane-spanning β sheets exhibit a “positive-outside” distribution of positively-charged
residues [110], as opposed to the much more typical positive-inside distribution [86,
109, 110, 113, 114]. Typically, β sheets are less hydrophobic than TMHs [226] as a
result of their amphiphilic nature.

1.5

The aims of this thesis

In this thesis, we explore three ideas surrounding the role of TMHs in the membrane.
Each chapter is presented in the general style of an article and contains a further
introduction to the specific topic of study.
In chapter 2 we explore the sequence composition of large datasets of TMHs stratifying them by species, organelle, single-pass and multi-pass. After considering the
rarity of certain amino acid types, we show a “negative-outside” tendency in concert

1.5. THE AIMS OF THIS THESIS

49

with the “positive-inside” rule. We go on to further divide the single-pass TMPs into
complex and simple TMHs according to TMSOC. This reveals that simple TMHs have
amino acid distribution features along the helix that would optimise the TMH for anchoring, however complex TMHs are more akin to TMHs from multi-pass TMPs. This
indicates that TMH function beyond anchoring outweighs the evolutionary pressure
to adhere to membrane constraints.
In chapter 3 we generate an up to date dataset of post-translationally inserted TA
proteins based on previous methods and compare the TMHs within to a manually curated TA protein dataset from UniProt. This cross-examination revealed adaptations
in the TMH sequences to mitochondrial membranes. This is likely to play a role in
trafficking of the TA TMPs as well as being an adaptation to the differences in the
biological composition of the mitochondrial membrane. We also generate homology
models of two spontaneously inserting TMHs from TA proteins and show that they
have an amphipathic surface which may be key to their spontaneous insertion into the
amphiphilic lipids of the bilayer.
In chapter 4 we look at families of TMPs that exhibit a high hydrophobic discrepancy between sequentially adjacent TMHs. The conservation of these pairs across the
families corroborates evidence of cooperative TMH insertion, showing that marginally
hydrophobic TMHs could use typical TMHs as part of their biogenesis on a larger
scale than previously thought.

Chapter 2
The “negative-outside” rule
The research presented in this chapter is published work presented in Baker et al., 2017
titled ‘Charged residues next to transmembrane regions revisited: “Positive-inside
rule” is complemented by the “negative inside depletion/outside enrichment rule”’ by
James Alexander Baker, Wing-Cheong Wong, Birgit Eisenhaber, Jim Warwicker, and
Frank Eisenhaber [227]. Here we include the supplementary tables and figures in the
text.

2.1
2.1.1

Summary
Background

Transmembrane helices frequently occur amongst protein architectures as means for
proteins to attach to or embed into biological membranes. Physical constraints such as
the membrane’s hydrophobicity and electrostatic potential apply uniform requirements
to transmembrane helices and their flanking regions; consequently, they are mirrored in
their sequence patterns (in addition to transmembrane helices being a span of generally
hydrophobic residues) on top of variations enforced by the specific protein’s biological
functions.

2.1.2

Results

With statistics derived from a large body of protein sequences, we demonstrate that,
in addition to the positive charge preference at the cytoplasmic inside (positive-inside
50

2.2. INTRODUCTION

51

rule), negatively-charged residues preferentially occur or are even enriched at the noncytoplasmic flank or, at least, they are suppressed at the cytoplasmic flank (negative-not-inside/negative-outside rule). As negative residues are generally rare within or
near transmembrane helices, the statistical significance is sensitive with regard to
details of transmembrane helix alignment and residue frequency normalisation and
also to dataset size; therefore, this trend was obscured in previous work. We observe
variations amongst taxa as well as for key organelles along the secretory pathway. The
effect is most pronounced for transmembrane helices from single-pass transmembrane
(bitopic) proteins compared to those with multiple transmembrane helices (polytopic
proteins) and especially for the class of simple transmembrane helices that evolved for
the sole role as membrane anchors.

2.1.3

Conclusions

The charged-residue flank bias is only one of the transmembrane helix sequence features with a role in the anchorage mechanisms, others apparently being the leucine
intra-helix propensity skew towards the cytoplasmic side, tryptophan flanking as well
as the cysteine and tyrosine inside preference. These observations will stimulate new
prediction methods for transmembrane helices and protein topology from a sequence
as well as new engineering designs for artificial membrane proteins.

2.2

Introduction

Two decades ago, the classic concept of a TMH was a rather simple story: typical TMPs were thought to be anchored in the membrane by membrane-spanning bundles of non-polar α–helices of roughly 20 residues length, with a consistent orientation
of being perpendicular to the membrane surface. Although this is broadly true, hundreds of high quality membrane structures have elucidated that membrane-embedded
helices can adopt a plethora of lengths and orientations within the membrane. They
are capable of just partial spanning of the membrane, spanning using oblique angles,
and even lying flat on the membrane surface [38, 39]. The insertion and formation
of the TMHs follow a complex thermodynamic equilibrium [40, 228, 229]. From the
biological function point of view, many TMHs have multiple roles besides being just

52

CHAPTER 2. THE “NEGATIVE-OUTSIDE” RULE

hydrophobic anchors; for example, certain TMHs have been identified as regulators of
protein quality control and trafficking mechanisms [43]. As these additional biological
functions are mirrored in the TMHs’ sequence patterns, TMHs can be classified as
simple (just hydrophobic anchors) and complex sequence segments [44–46].
The relationship between sequence patterns in and in the vicinity of TMHs and
their structural and functional properties, as well as their interaction with the lipid
bilayer membrane, has been a field of intensive research in the last three decades [1].
Besides the span of generally hydrophobic residues in the TMH, there are other trends
in the sequence such as with a saddle-like distribution of polar residues (depressed
incidence of charged residues in the TMH itself), an enriched occurrence of positively-charged residues in the cytosolic flanking regions as well as an increased likelihood
of tryptophan and tyrosine at either flank edge [86, 109, 113, 117, 230, 231]. Such
properties vary somewhat in length and intensity between various biological organelle
membranes, between prokaryotes and eukaryotes [232] and even among eukaryotic
species studied due to slightly different membrane constraints [86, 110]. These biological dispositions are exploitable in terms of TM region prediction in query protein sequences [233, 234] and tools such as the quite reliable TMHMM [67, 235], Phobius [71,
236] or DAS-TMfilter represent today’s prediction limit of TMHs’ hydrophobic cores
within the protein sequence [237–239]. The prediction accuracy for true positives and
negatives is reported to be close to 100% and the remaining main cause of false positive prediction are hydrophobic α–helices completely buried in the hydrophobic core
of proteins. To note, reliable prediction of TMHs and protein topology is a strong
restriction for protein function of even otherwise non-characterised proteins [240–242]
and thus, very valuable information.
The “positive-inside rule” reported by von Heijne [38, 113] postulates the preferential occurrence of positively-charged residues (lysine and arginine) at the cytoplasmic
edge of TMHs. The practical value of positively-charged residue sequence clustering
in topology prediction of TMH was first shown for the plasmalemma in bacteria [113,
243]. As a trend, the “positive-inside rule” has since been confirmed with statistical
observations for most membrane proteins and biological membrane types [109, 118,
244, 245]. However, more recent evidence suggests that, in thylakoid membranes, the
“positive-inside rule” is less applicable due to the co-occurrence of aspartic acid and

2.2. INTRODUCTION

53

glutamic acid residues together with positively-charged residues [110].
The positive-inside rule also received support from protein engineering experiments
that revealed conclusive evidence for positive charges as a topological determinant [113,
246–248]. Mutational experiments demonstrated that charged residues, when inserted
into the centre of the helix, had a large effect on insertion capabilities of the TMH
via the translocon. Insertion becomes more unfavourable when the charge was placed
closer to the TMH core [116].
It remains unclear exactly why and how exactly the positive charge determines
topology from a biophysical perspective. positively-charged residues are suggested to
be stronger determinants of topology than negatively-charged residues due to a dampening of the translocation potential of negatively-charged residues. This dampening
factor is the result of protein-lipid interactions with net zero charged phospholipid,
phosphatidylethanolamine and other neutral lipids. This effect favours cytoplasmic
retention of positively-charged residues [249].
The recent accumulation of TMP sequences and structures allowed revisiting the
problem of charged residue distribution in TMHs (see also http://blanco.biomol.
uci.edu/mpstruc/). For example, whilst β–sheets contain charged residues in the TM
region, α-helices generally do not [106]. Large-scale sequence analysis of TMH from
various organelle membrane surfaces in eukaryotic proteomes confirm the clustering
of positive charge having a statistical bias for the cytosolic side of the membrane. At
the same time, there are many TMH exception examples to the positive-inside rule;
however as a trend, topology can be determined by simply looking for the most positive
loop region between helices [86, 109].
When the observation of positively-charged residues preferentially localised at the
cytoplasmic edge of TMHs emerged, it was also asked whether negatively-charged
residues work in concert with TMH orientation. It was shown that a single additional lysine residue can reverse the topology of a model Escherichia coli protein,
whereas a much higher number of negatively-charged residues is needed to achieve the
same [248]; nevertheless, a sufficiently large negative charge can overturn the positiveinside rule [250, 251] and, thus indeed, negative residues are topologically active to a

54

CHAPTER 2. THE “NEGATIVE-OUTSIDE” RULE

point. negatively-charged residues were observed in the flanks of TMHs [109], especially of marginally hydrophobic TM regions [252]. It is known that the negativelycharged acidic residues in TM regions have a non-trivial role in the biological context.
In E. coli, negative residues experience electrical pulling forces when travelling through
the SecYEG translocon indicating that negative charges are biologically relevant during the electrostatic interactions of insertion [42, 253].
Unfortunately, there is a problem with statistical evidence for preferential negative charge occurrence next to TMH regions. Early investigations indicated overall
both positive and negative charge were influential topology factors, dubbed the charge
balance rule. If true, one would also expect to see a skew in the negative charge distribution if a cooperation between oppositely charged residues orientated a TMH [243,
254]. It might be expected that, if positive residues force the loop or tail to stay inside, negative residues would be drawn outside and topology would be determined not
unlike electrophoresis. Yet, there is plenty of individual protein examples but no conclusive statistical evidence in the current literature for a negatively-charged skew [86,
109, 110, 114, 117, 118].
There are many observations described in the literature that charged residues determine topology more predictably in single-pass proteins than in multi-pass TMH [251,
255]. It is thought that the charges only determine the initial orientation of the TMH
in the biological membrane; yet, the ultimate orientation must be determined together
with the totality of subsequent downstream regions [256].
With sequence-based hydrophobicity and volume analysis and consensus sequence
studies, Sharpe et al. [86] demonstrated that there is asymmetry in the intramembranous space of some membranes. Crucially, this asymmetry differs among the membrane
of various organelles. They conclude that there are general differences between the
lipid composition and organisation in membranes of the Golgi and ER. Functional
aspects are also important. For example, the abundance of serines in the region following the lumenal end of Golgi TMHs appears to reflect the fact that this part of
many Golgi enzymes forms a flexible linker that tethers the catalytic domain to the
membrane [86].
A study by Baeza-Delgado et al. [109] analysed the distribution of amino acid

2.2. INTRODUCTION

55

residue types in TMHs in 170 integral membrane proteins from a manually maintained database of experimentally confirmed TMPs (MPTopo [112]) as well as in 930
structures from the PDB. As expected, half of the natural amino acids are equally
distributed along TMH whereas aromatic, polar and charged amino acids along with
proline are biased near the flanks of the TM helices. Unsurprisingly, leucine and
other non-polar residues are far more abundant than the charged residues in the TM
region [86, 109].
In this work, we revisit the issue of statistical evidence for the preferential distribution of negatively-charged (and a few other) residues within and nearby TMHs.
We rely on the improved availability of comprehensive and large sequence and structure datasets for TM proteins. We also show that several methodical aspects have
hindered previous studies [86, 109, 110] to see the consistent non-trivial skew for
negatively-charged residues disfavouring the cytosolic interfacial region and/or preferring the outside flank. First, we show that acidic residues are especially rare within
and in the close sequence environment of TMHs, even when compared to positivelycharged lysine and arginine. Second, therefore, the manner of normalisation is critical:
Taken together with the difficulty to properly align TMHs relative to their boundaries,
column-wise frequency calculations relative to all amino acid types as in previous
studies will blur possible preferential localisations of negative charges in the sequence.
However, the outcome changes when we ask where a negative charge occurs in the
sequence relative to the total amount of negative charges in the respective sequence
region. Thus, by accounting for the rarity of acidic residues with sensitive normalisation, the “non-negative inside rule/negative-outside rule” is clearly supported by the
statistical data. We find that minor changes in the flank definitions such as taking
the TMH boundaries from the database or by generating flanks by centrally aligning TMHs and applying some standardised TMH length does not have a noticeable
influence on the charge bias detected.
Third, there are significant differences in the distribution of amino acid residues
between single-pass and multi-pass TM regions in both the intra-membrane helix and
the flanking regions with further variations introduced by taxa and by the organelles
along the secretory pathway. Importantly, we find that it is critical to weigh down the
effect of TMHs in multi-pass TMPs with no or super-short flanks to observe statistical

56

CHAPTER 2. THE “NEGATIVE-OUTSIDE” RULE

significance for the charge bias. To say it bluntly, if there are no flanks of sufficient
length, there is also no negative-charge bias to be observed.
The charge bias effect is even clearer when a classification of TMHs into so-called
simple (which, as a trend, are mostly single-pass and mere anchors) and so-called
complex (which typically have functions beyond anchorage) is considered [44–46]. We
also observe parallel skews with regard to leucine, tyrosine, tryptophan and cysteine
distributions. With these large-scale datasets and a sensitive normalisation approach,
new sequence features are revealed that provide spatial insight into TMH membrane
anchoring, recognition, helix-lipid, and helix-helix interactions.

2.3
2.3.1

Results
Acidic residues within and nearby transmembrane helix
segments are rare

In order to reliably compare the amino acid sequence properties of TMHs, we assembled datasets of TMH proteins from what are likely to be the best in terms of quality and comprehensiveness of annotation in eukaryotic and prokaryotic representative
genomes, as well as composite datasets to represent larger taxonomic groups and with
regard to sub-cellular locations (see Table 2.1). In total, 3292 single-pass TMH segments and 29898 multi-pass TMH segments were extracted from various UniProt [75]
text files according to TRANSMEM annotation (download dated 20–03–2016). The
UniProt datasets used only included manually curated records; however, it is still necessary to check for systematic bias due to the prediction methods used by UniProt
for TMH annotation in the majority of cases without direct experimental evidence.
Therefore, a fully experimentally verified dataset was also generated for comparison.
The representative 1544 single-pass and 15563 TMHs were extracted from the manually
curated experimentally verified TOPDB [257] database (download dated 21–03–2016)
referred to as ExpAll here (Table 2.1). TMH organelle residency is defined according
to UniProt annotation. To ensure reliability, organelles were only analysed from a
representative redundancy-reduced protein dataset of the most well-studied genome:
Homo sapiens (referred to as UniHuman herein). The several datasets from UniProt

2.3. RESULTS

57

are subdivided into different human organelles (UniPM, UniER, UniGolgi) and taxonomical groups (UniHuman, UniCress, UniBacilli, UniEcoli, UniArch, UniFungi) as
described in Table 2.1 (see also Methods section). As will be shown below, these various datasets allow us to validate our findings for a variety of conditions, namely with
regard (i) to experimental verification of TMHs, (ii) to origin from various species and
taxonomic groups, (iii) to the number of TMHs in the same protein as well as (iv) to
sub-cellular localisation. Data-sets and programs used in this work can be downloaded
from http://mendel.bii.a-star.edu.sg/SEQUENCES/NNI/.
Table 2.1: Acidic residues are rarer in transmembrane helices of single-pass proteins
than in transmembrane helices of multi-pass proteins.The statistical results when comparing
the number of acidic residues in single-pass or multi-pass TMHs within their database-defined limits
and excluding any flanks. The number of helices per dataset can be found in Table 2.2 for single-pass TMHs and Table 2.3 for multi-pass helices. µ SP is the average number of the respective
residues per helix in TMHs from single-pass proteins, while µ MP is the average number of the
respective residues per TMH from multi-pass proteins. The Kruskal-Wallis test scores (H statistics)
were calculated for the numbers of aspartic acid and glutamic acid residues in each helix from singlepass and the number of aspartic acid and glutamic acid residues in each helix from multi-pass TMHs
Acidic residues (D and E)
Data-set

µ SP

Aspartic acid (D only)

H statistic
µ SP
p-value

µ MP

Glutamic acid (E only)
H statistic µ SP
p-value

µ MP

µ MP

H statistic
p-value

ExpAll

0.086

0.309

148.1
4.50E-34

0.045

0.157

40.3
2.13E-10

0.042

0.161

46.6
8.64E-12

UniHuman

0.076

0.398

316.5
8.31E-71

0.034

0.191

91.6
1.05E-21

0.042

0.207

100.3
1.33E-23

UniER

0.106

0.43

34.4
4.39E-9

0.061

0.161

8.0
4.72E-3

0.045

0.268

26.8
2.24E-7

UniGolgi

0.097

0.381

39.8
2.88E-10

0.043

0.18

19.4
1.05E-5

0.053

0.201

20.2
7.01E-6

UniPM

0.039

0.4

121.0
3.86E-28

0.016

0.187

32.7
1.06E-8

0.022

0.213

36.9
1.26E-9

UniCress

0.062

0.434

163.5
1.99E-37

0.036

0.198

32.5
1.20E-8

0.025

0.241

66.0
4.59E-16

UniFungi

0.177

0.349

43.1
5.14E-11

0.044

0.166

24.5
7.60E-7

0.133

0.183

4.6
0.033

UniBacilli

0.089

0.352

24.1
9.16E-7

0.048

0.185

11.2
8.27E-4

0.04

0.176

12.3
4.54E-5

UniEcoli

0.148

0.315

2.7
0.100

0.111

0.15

0.1
0.729

0.037

0.163

2.2
0.140

UniArch

0.438

0.606

1.8
0.183

0.083

0.344

11.2
8.33E-4

0.354

0.247

3.5
0.0624

The hydrophobic nature of the lipid bilayer membrane implies that, generally,
charged residues should be rare within TMHs. For acidic residues, even the location
in the sequence vicinity of TMHs should be disfavoured because of the negatively-charged head groups of lipids directed towards the aqueous extracellular side or the
cytoplasm. In agreement with the biophysically justified expectations, the statistical
data confirms that acidic residues are especially rare in TMHs and their flanking

58

CHAPTER 2. THE “NEGATIVE-OUTSIDE” RULE

regions. In Figure 2.1 where we plot the total abundance of all amino acid types in
single-pass TMHs and multi-pass TMHs (including their ±5 flanking residues), acidic
residues were found to be amongst the rarest amino acids both in UniHuman and
ExpAll.
UniHuman
A)
10608
Absolute Count

Absolute Count

12000
Single-pass

ExpAll

B)

10000
8000
2299

6000

1597

666

4000

560

2000
0

10000
9000
8000
7000
6000
5000
4000
3000
2000
1000
0

59700

D)

60000
50000
40000
30000

11545

8972

7338

7251

80000
Absolute Count

Absolute Count

Multi-pass

70000

70000

1683

582

520

68201

60000
50000
40000
30000

20000

20000

10000

10000

0

2085

L V I A G F S T R K Y C M P WQ N E H D
Residue

L V A I G F S T R C K P Y MWQ N H E D
Residue

C)

8724

12907

11066
8156

7722

0
L V I F A S G T Y P M R N C K Q E DWH
Residue

L I A V F G S T Y M P R N K WQ C E D H
Residue

Figure 2.1: Negatively-charged amino acids are amongst the rarest residues in transmembrane helices and ±5 flanking residues. Bar charts of the abundance of each amino acid
type in the TMHs with flank lengths of the accompanying ±5 residues from the (a) UniHuman single-pass proteins, (b) ExpAll single-pass proteins, (c) UniHuman multi-pass proteins, and (d) ExpAll
multi-pass proteins. Amino acid types on the horizontal axis are listed in descending count. The bars
were coloured according to categorisations of hydrophobic, neutral and hydrophilic types according
to the free energy of insertion biological scale [116]. Grey represents hydrophilic amino acids that
were found to have a positive ∆G app, and blue represents hydrophobic residues with a negative ∆G
app, purple denotes negative residues and positive residues are coloured in orange. The abundances
of key residues are labelled.

The effect is most pronounced in single-pass TMPs (Figure 2.1). There are only
666 glutamates (just 1.24% of all residues) and 560 aspartates (1.05% respectively)
among the total set of 53238 residues comprised in 1705 TMHs and their flanks.
Within just the TMH regions, there are 71 glutamates (0.20% of all residues in TMHs
and flanks) and 58 aspartates (0.16% respectively). This cannot be an artefact of
UniProt TMH assignments since this feature is repeated in ExpAll. There are only
582 glutamates (1.22%) and 520 aspartates (1.09%) among the 47568 residues involved. Within the TMH itself, there are 64 glutamates (0.19%) and 69 aspartates
(0.21%). In both cases, the negatively-charged residues represent the ultimate end of

2.3. RESULTS

59

the distribution. To note, acidic residues are rare even compared to positively-charged
residues which are about 3–4 times more frequent. On a much smaller dataset of
single-spanning TMP, Nakashima et al. [258] made similar compositional studies. To
compare, they found 0.94% glutamate and 0.94% aspartate within just the TMH region
(values very similar to ours from TMHs with small flanks; apparently, they used more
outwardly defined TMH boundaries) but the content of each glutamate and aspartate
within the extracellular or cytoplasmic domains is larger by an order of magnitude,
between 5.26% and 9.34%. These latter values tend to be even higher than the average
glutamate and aspartate composition throughout the protein database (5–6% [258]).

In the case of multi-pass TMPs (Figure 2.1), glutamates and aspartates are still
very rare in TMHs and their ±5 residue flanks (1.94% and 1.92% from the total of
377207 in the case of UniHuman respectively, 1.79% and 1.70% from the total of
454700 in the case of ExpAll). Yet, their occurrence is similar to those of histidine and
tryptophan and, notably, acidic residues are only about ∼1.5 times less frequent than
positively-charged residues. The observation that acidic residues are more suppressed
in single-pass TMHs compared with the case of multi-pass TMHs is statistically significant. In Table 2.1, the acidic residues are counted in the helices (excluding flanking
regions) belonging to either multi-pass or single-pass helices. Indeed, single-pass helices appear to tolerate negative charge to a far lesser extent than multi-pass helices
as the data in the top two rows of Table 2.1 indicates (for datasets UniHuman and
ExpAll). The trend is strictly observed throughout sub-cellular localisations (rows 3–5
in Table 2.1) and taxa (rows 6–10). Statistical significance (P≤0.001) is found in all
but six cases. These are UniEcoli (D+E, D, E), UniArch (D+E, E) and UniFungi (E).
The problem is, most likely, that the respective datasets are quite small. Notably, the
difference between single- and multi-pass TMHs is greatest in UniPM; here, TMHs
from multi-pass proteins have on average 0.400 negative residues per helix, whereas
single-pass TMHs contained just 0.039 (P=3.86e-28).

60

CHAPTER 2. THE “NEGATIVE-OUTSIDE” RULE

2.3.2

Amino acid residue distribution analysis reveals a
“negative-not-inside/negative-outside” signal in singlepass transmembrane helix segments

The rarity of negatively-charged residues is a complicating issue when studying their
distribution along the sequence positions of TMHs and their flanks. For UniHuman
and ExpAll , we plotted absolute abundance of aspartic acid, glutamic acid, lysine,
arginine, and leucine at each position (i.e., it scales as the equivalent fraction in the total composition of the alignment column) (Figure 2.2). To note, the known preference
of positively-charged residues towards the cytoplasmic side is nevertheless evident. Yet,
it becomes apparent that any bias in the occurrence of the much rarer acidic residues
is overshadowed by fluctuations in the highly abundant residues such as leucine.
The trends become clearer if the occurrence of specific residues is normalised with
the total number of residues of the given amino acid type in the dataset observed in
the sequence region studied as shown for UniHuman and for ExpAll in Figure 2.2. For
comparison, we indicated background residue occurrences (dashed lines calculated as
averages for positions -25 to -30 and 25 to 30). The respective average occurrences in
the inside and outside flanks (calculated from an average of the values at positions -20
to -10 and 10 to 20 respectively) are shown with wide lines.
The “positive-inside rule” becomes even more evident in this normalisation:
Whereas the occurrence of positively-charged residues is about the background level
at the outside flank, it is about two to three times higher both for the UniHuman and
the ExpAll datasets at the inside flank. To note, the background level was found to
be 1.7% (lysine) and 1.6% (arginine) in UniHuman and 1.4% (lysine and arginine) in
ExpAll. The inside flank average is 4.3% (lysine) and 4.6% (arginine) in UniHuman
and 4.2% (lysine) and 4.6% (arginine) in ExpAll. The outside flank is similar to the
background noise levels: about 1.4% (lysine) and 1.5% (arginine) in UniHuman and
about 1.5% (lysine) and 1.4% (arginine) in ExpAll.
Most interestingly, a “negative-inside depletion” trend for the negatively-charged
residues is apparent from the distribution bias. The inside flank averages for glutamic
acid were 1.1% and 1.4% in UniHuman and ExpAll respectively; for aspartic acid,
1.2% and 1.4% in UniHuman and ExpAll respectively. Meanwhile, the outside flanks

2.3. RESULTS
L

61
UniHuman

R and K

ExpAll

D and E
A)

B)
Outside

Inside

Outside

Absolute count

Absolute count

Inside

C)
Outside

Inside

Inside

Outside

Outside

7

7

6

6

6

5

5

4

4

3

3

2

2

1

1

1

1

0

0

0

0

7

7

4
3
2

Relative percentage

5

Relative percentage

6

- 20
- 10
0
Distance from TMH

Outside

Inside

Outside

0
10
20
Distance from TMH

F)
Inside

- 20
- 10
0
Distance from TMH

Relative percentage

Inside

Inside

Relative percentage

E)

D)

Outside

5
4
3
2

0

10
20
Distance from TMH

Figure 2.2: Relative percentage normalisation reveals a negative-outside bias in transmembrane helices from single-pass protein datasets. All flank sizes were set at up to ±20
residues. We acknowledge that all values, besides the averaged values, are discrete, and connecting lines are illustrative only. On the horizontal axes (a–d) are the distances in residues from the
centre of the TMH, with the negative numbers extending towards the cytoplasmic space. For (e)
and (f), the horizontal axis represents the residue count from the membrane boundary with negative
counts into the cytoplasmic space. Leucine, the most abundant non-polar residue in TMHs, is in
blue. Arginine and lysine are shown in dark and light orange respectively. Aspartic and glutamic
acid are showing in dark and light purple respectively. In red are the uncharged polar amino acids
serine, asparagine, glutamine, threonine, tyrosine, and cysteine. (a) and (b) On the vertical axis is
the absolute abundance of residues in TMHs from single-pass proteins from (a) UniHuman and (b)
ExpAll. Note that no clear trend can be seen in the negative residue distribution compared to the
positive-inside signal and the leucine abundance throughout the TMH. c and d On the vertical axis is
the relative percentage at each position for TMHs from single-pass proteins from (c) UniHuman and
(d) ExpAll. The dashed lines show the estimation of the background level of residues with respect to
the colour; an average of the relative percentage values between positions 25 to 30 and –30 to –25.
The thick bars show the averages on the inner (positions –20 to –10) and outer (positions 10 to 20)
flanks coloured to the respective amino acid type. Note a visible suppression of acidic residues on the
inside flank when compared to the outside flank in single-pass proteins when normalising according
to the relative percentage. (e) and (f) The relative distribution of flanks defined by the databases
with the distance from the TMH boundary on the horizontal axis. The inside and outside flanks are
shown in separate subplots. The colouring is the same as in (a) and (b).

62

CHAPTER 2. THE “NEGATIVE-OUTSIDE” RULE

for aspartic acid and glutamic acid occurrences were measured at 2.9% and 2.4% respectively in UniHuman and, in ExpAll, these values for aspartic acid and glutamic
acid were found to be 2.5% and 2.1% respectively. Against the background level of
aspartic acid (2.8% and 2.9% in UniHuman) and glutamic acid (2.6% and 2.9% in
ExpAll), the inside flank averages were found to be about 2–3 times lower than the
background level while the outside flank averages were comparable to the background
level (Figure 2.2). Taken together, this indicates a clear suppression of negatively-charged residues at the inside flank of single-pass TMHs and a possible trend for
negatively-charged residues occurring preferentially at the outside flank. This is not
an effect of the flank definition selection since the trend remains the same when using
the database-defined flanks without the context of the TMH (Figure 2.2). For UniHuman, the negative charge expectancy on the inside flank doesn’t reach above 2%
until position -10 (D) and position -11 (E), whereas, on the outside flank, both D and
E start >2%. The same can be seen in ExpAll where negative residues reach above
2% only as far from the membrane boundary as at position -9 (D) and position -7 (E)
on the inside but exceed 2% beginning with position 1 (D) and 3 (E) on the outside
(Figure 2.2).
Residue presence is a zero-sum variable. If there is more likelihood of a positivelycharged residue being present at an inside position, then there must be less probability
of at least one type of amino acid at that position. To check if this probability was
spread throughout non-charged amino acids as well as negatively-charged amino acids,
we also examined non-charged polar residues for any inside versus outside preference
(Figure 2.2B and Figure 2.2C). As expected there was an increased prevalence at
the flanks (peaking at position +12 with 2.27% in Expall and 2.39 at position -10 in
UniHuman), however, there was no clear difference between the inside and outside flank
relative percentages. In ExpAll the inside flank (1.8% relative percentage average) to
outside (1.9% relative percentage average) was between 5 to 10 times less than the
negatively-charged residue inside-outside difference, and there was very little difference
in the UniProt inside (1.88% relative percentage average) to outside (1.94% relative
percentage average) relative abundance.
The observation of negative charge suppression at the inside flank, herein the
“negative-inside depletion” rule, is statistically significant throughout most datasets in

2.3. RESULTS

63

this study. The inside-outside bias was counted using the Kruskal-Wallis (K-W) test
comparing the occurrence of acidic residues within 10 residues of each TMH inside and
outside the TMH (Table 2.2). We studied both the database-reported flanks as well
as those obtained from central alignment of TMHs (see Methods). The null hypothesis (no difference between the two flanks) could be confidently rejected in all cases
(p-value<0.001 except for UniBacilli), the sign of the H-statistic (K-W) indicating
suppression at the inside and/or preference for the outside flank (except for UniArch).
Most importantly, acidic residues were found to be distributed with bias in ExpAll (pvalue<3.47e-58) and in UniHuman (p-value=1.13e-93). Whereas with UniBacilli, the
problem is most likely the dataset size, the exception of UniArch, for which we observe
a strong negative inside rule, is more puzzling and indicates biophysical differences of
their plasma-membrane.

Table 2.2: Statistical significances for negative charge distribution skew on either side
of the membrane in single-pass transmembrane helices. The “Helices” column refers to the
total TMHs contained in each dataset (ExpALL, TMHs from TOPDB [257]; UniHuman, human
representative proteome; UniER, human endoplasmic reticulum representative proteome; UniGolgi,
human Golgi representative proteome; UniPM, human plasma membrane representative proteome;
UniCress, Arabidopsis thaliana (mouse-ear cress) representative proteome; UniFungi, fungal representative proteome; UniBacilli, Bacilli class representative proteome; UniEcoli, Escherichia coli
representative proteome; UniArch, Archaea representative proteome; see Methods for details). In
the “Database-defined flanks” column, the “Negative residues” column refers to the total number of
negative residues found in the ±10 flanking residues on either side of the TMH and does not include
residues found in the helix itself. In the “Flanks after central alignment” column, the “Negative
residues” column refers to the total number of negative residues found in the –20 to –10 residues and
the +10 to +20 residues from the centrally aligned residues of the TMH. Unlike the other tables,
the global averages are derived from the ±20 datasets. The K-W scores were calculated for negative residues by comparing the number of negatively-charged residues that were within the 10 inside
residues and the 10 outside residues in either case
single-pass

Database-defined flanks

Flanks after central alignment

Negative residues
Data-set

Helices

ExpAll

1544

UniHuman
UniER

Negative residues
H statistic

p-value

258.59

3.47E-58 735

1541

1922

421.53

1.13E-93

652

156

23.76

1.09E-06

76

60

240

104.45

493

197

578

UniCress

632

314

UniFungi

729

449

UniBacilli

124

90

UniEcoli

54

UniArch

48

Inside

Outside

848

1648

1705

780

132

78

UniGolgi

206

UniPM

Inside

Outside

H statistic

p-value

262.29

5.44E-59

1865

501.86

3.74E-111

150

21.62

3.33E-06

1.61E-24 54

239

107.18

4.06E-25

177.68

1.56E-40

161

569

215.18

1.02E-48

450

18.23

1.96E-05

231

444

55.8

8.01E-14

631

28.15

1.12E-07 413

627

38.08

6.79E-10

113

3.73

5.35E-02

106

2.53

1.12E-01

32

77

17.24

3.30E-05 30

74

14.74

1.24E-04

113

8

49.66

1.83E-12 96

7

45.62

1.43E-11

86

64

CHAPTER 2. THE “NEGATIVE-OUTSIDE” RULE

2.3.3

Amino acid residue distribution analysis reveals a general negative-charge bias signal in outside flank of multi-pass transmembrane helix segments — the negative
outside enrichment rule

As a result of the rarity of negatively-charged residues, any distribution bias is difficult to be recognised in the plot showing the total abundance (or alignment column
composition) of residues in multi-pass TMHs and their flanks from UniHuman and
ExpAll (Figure 2.3). Yet, as with single-pass helices, the dominant general leucine
enrichment, as well as positive inside signal, can be identified with certainty. When
the residue occurrence is normalised by the total occurrence of this residue type in the
sequence regions studied (shown as a relative percentage of at each position for multi-pass helices from UniHuman and ExpAll in Figure 2.3), the bias in the distribution
of any type of charged residues becomes visible.
With regard to the positive-inside preference, positively-charged residues have a
background value of 2.0% for arginine and 2.2% for lysine in UniHuman, and 1.7%
for arginine and 1.9% for lysine in ExpAll. At the inside flank, this rises to 4.6% for
arginine and 4.1% for lysine in UniHuman and 4.6% for arginine and 4.2% for lysine
in ExpAll. The mean net charge at each position was calculated for multi-pass and
single-pass datasets from UniHuman and ExpAll (Figure 2.4). The positive inside rule
clearly becomes visible as the net charge has a positive skew approximately between
residues -10 and -25. What is noteworthy is that the peaks found for single-pass
helices were almost three times greater than those of multi-pass helices. For single-pass TMHs, the peak is +0.30 at position -15 in UniHuman and +0.31 at position
-14 in ExpAll, whereas TMHs from multi-pass proteins had lower peaks of +0.15
at position -13 in UniHuman and +0.10 at position -14 in ExpAll. Thus, there is a
positive charge bias towards the cytoplasmic side; yet, it is much weaker for multi-pass
than for single-pass TMHs.
Notably, a “negative outside enrichment” trend also can be seen from the distribution of the negatively-charged residues, though with some effort (Table 2.3) as the effect
is also weaker than in the case of single-pass TMHs. We studied the flanks under four
conditions: (i) database-defined flanks without overlap between neighbouring TMHs,

2.3. RESULTS

L
R and K

65

UniHuman

ExpAll

D and E
A)

B)
Outside

Inside
Inside

Outside
Outside

Inside

9

6

Outside

Inside

Outside
Outside

7

7

6

6

5

5

4

4

3

3

2

2

1

1

1

0

0

6
5
4
3
2

- 20
-10
0
Distance from TMH

Relative percentage

7

F)

7

8
Relative percentage

Inside

Inside
Inside

0
10
20
Distance from TMH

0
- 20
-10
0
Distance from TMH

Relative percentage

E)

D)

Relative percentage

Absolute count

C)

Outside

Absolute count

Inside

Outside

5
4
3
2
1
0
0
10
20
Distance from TMH

Figure 2.3: Negative-outside bias is very subtle in transmembrane helices from multipass proteins. The meaning for the horizontal axis is the same as in Figure 2.2, with the negative
sequence position numbers extending towards the cytoplasmic space. Leucine is in blue. Arginine
and lysine are shown in dark and light orange respectively. Aspartic and glutamic acid are shown
in dark and light purple respectively. In red are the uncharged polar amino acids serine, asparagine,
glutamine, threonine, tyrosine, and cysteine. All flank sizes were set at up to ±20 residues. (a) and
(b) On the vertical axes are the absolute abundances of residues from TMHs of multi-pass proteins
from (a) UniHuman and (b) ExpAll. c and d On the vertical axes are the relative percentages at each
position for TMHs from multi-pass proteins from (c) UniHuman and (d) ExpAll. As in Figure 2.2(c)
and (d), the dashed lines show the estimation of the background level of residues with respect to the
colour, and the thick bars show the averages on the inner and outer flanks coloured to the respective
amino acid type. e and f The relative distribution of flanks defined by the databases with the distance
from the TMH boundary on the horizontal axis for both the inside and outside flanks. The colouring
is the same as in (a) and (b).

CHAPTER 2. THE “NEGATIVE-OUTSIDE” RULE

Net charge

A) UniHuman
0.4
0.3
0.2
0.1
0
-0.1 -30

B)
Single-pass
Multi-pass

-20

-10

0

10

20

30

Distance from centre of helix

ExpAll

Net charge

66

0.4
0.3
0.2
0.1
0
-0.1 -30

-20

-10

0

10

20

30

Distance from centre of helix

Figure 2.4: The net charge across multi-pass and single-pass transmembrane helices
shows a stronger positive inside charge in single-pass transmembrane helices than multipass transmembrane helices. The net charge per TMH plotted at each position; the positive-inside
rule is stronger in TMHs from single-pass proteins than TMHs from multi-pass proteins. The net
charge was calculated at each position as described in the Methods section for the (A) UniHuman
and (B) ExpAll datasets. Net charge for TMHs from multi-pass proteins is shown in black, and the
profile of TMHs from single-pass proteins is drawn in blue.

(ii) flanks after central alignment of TMHs without flank overlap, (iii) database-defined
flanks but allowing overlap of flanks shared among neighbouring TMHs, (iv) same as
condition (ii) but only the subset of cases where there is at least half of the required
flank length at either side of the TMH. In UniHuman as calculated under condition
(i), aspartic acid is lower on the inside flank (2.3%) than on the outside flank (3.0%).
Glutamic acid is also lower at the inside flank (2.4%) than the 2.8% on the outside
flank (Figure 2.3C). Slight variations in defining the membrane boundary point do
not influence the trend (compare figures 2.3C and 2.3E). We find that, in all studied
conditions, the UniHuman dataset delivers statistical significances (p-values: (i) 6.10e34, (ii) 5.43e-41, (iii) 3.00e-57, (iv) 5.60e-41) strongly supporting negative-charge bias
(inside suppression/outside preference; see Table 2.3).
As with the single-pass proteins, we checked if this probability was spread throughout non-charged amino acids as well as negatively-charged amino acids by examining
non-charged polar residues for inside versus outside preference (Figure 2.3B and Figure 2.3C). There was no clear difference between the inside and outside flank relative
percentages for ExpAll since the inside flank was 1.9% (relative percentage average)
and the outside flank was 2.0% (relative percentage average). There was some small
difference in the UniHuman dataset with the inside average being 1.9% and the outside
average being 2.2%. This however is much less of a difference than the negativelycharged residue flank differences in the UniHuman dataset.
Surprisingly, the result could not straightforwardly be repeated with the considerably smaller ExpAll. Under condition (i), we find with ExpAll that aspartic acid has

2.3. RESULTS

67

Table 2.3: Statistical significances for negative charge distribution skew on either side
of the membrane in multi-pass transmembrane helices. The “Helices” column refers to the
total TMHs contained in each dataset (ExpALL, TMH from TOPDB [257]; UniHuman, human
representative proteome; UniER, human endoplasmic reticulum representative proteome; UniGolgi,
human Golgi representative proteome; UniPM, human plasma membrane representative proteome;
UniCress, Arabidopsis thaliana (mouse-ear cress) representative proteome, UniFungi, fungal representative proteome; UniBacilli, Bacilli class representative proteome; UniEcoli, Escherichia coli representative proteome; UniArch, Archaea representative proteome; see Methods for details). In (A)
the “Database-defined flanks” and in (B) the “Database-defined viable* flanks” and the “Overlapping flanks” columns, the “Negative residues” column refers to the total number of negative residues
found in the ±10 flanking residues on either side of the TMH and does not include residues found
in the TMH itself. (A) In the “Flanks after central alignment” column, the “Negative residues”
column refers to the total number of negative residues found in the –20 to –10 residues and the +10
to +20 residues from the centrally aligned residues with a maximum database defined flank length
of 20 residues. The total number of proteins is given in the IDs column. The “Helices” column
contains the total number of TMHs in the dataset (n), the average number of TMHs per protein in
that population (µ) and the standard deviation of that average (σ). The K-W scores were calculated
for negative residues by comparing the number of negatively-charged residues that were within 10
residues inside and 10 residues outside the TMH.
*Here, “viable” indicates that in each TMH used for both flanks either side of the TMH has a flank
length of at least half the maximum allowed flank length, in this case 10 (the viable length is 5)
multi-pass
Helices
Data-set

IDs

Database-defined flanks

Flanks after central alignment

Negative residues

Negative residues

n

µ

σ

Inside

Outside

H statistic

p-value

Inside

Outside

H statistic

p-value

ExpAll

2205

15,563

7.07

3.95

9709

9598

0.04

8.43E-01

9648

9659

0.35

5.56E-01

UniHuman

1789

12,353

6.93

3.2

7196

9164

147.5

6.10E-34

6740

8968

179.77

5.43E-41

UniER

155

898

5.85

3.2

630

584

0.44

5.08E-01

578

576

0.03

8.58E-01

61

383

6.28

2.97

274

261

0.02

8.75E-01

266

259

0.09

7.65E-01

UniPM

427

3079

7.22

3.3

1945

2499

47.98

4.30E-12

1791

2440

64.42

1.01E-15

UniCress

507

3823

7.55

3.32

2567

2426

0.73

3.93E-01

2398

2433

1.11

2.93E-01

UniFungi

1338

8685

6.5

3.75

5560

5266

5.83

1.57E-02

5140

5214

0

9.62E-01

(A) UniGolgi

UniBacilli

140

822

5.94

3.98

470

468

0.07

7.92E-01

450

471

0.92

3.38E-01

UniEcoli

529

3888

7.39

3.76

1990

1902

0.26

6.07E-01

1875

1887

0.18

6.71E-01

UniArch

59

327

5.97

2.73

245

175

7.98

4.72E-03

235

181

7.08

7.81E-03

multi-pass

Overlapping flanks

Database-defined viable* flanks

Negative residues
Data-set

Inside

Outside

Negative residues
H statistic

p-value

N

Inside

Outside

H statistic

p-value

ExpAll

11,969

12,615

22.54

2.05E-06

8808

6082

6916

59.93

9.81E-15

UniHuman

8645

11,181

254.3

3.00E-57

8183

5169

6915

179.71

5.60E-41

UniER

750

763

1.16

2.81E-01

516

398

441

3.16

7.55E-02

333

369

7.12

7.64E-03

195

162

186

3

8.30E-02

UniPM

2319

3107

99.68

1.79E-23

1977

1343

1960

98.63

3.05E-23

UniCress

3142

3298

9.21

2.41E-03

2110

1626

1741

6.4

1.14E-02

UniFungi

6724

6814

0.46

4.96E-01

4581

3340

3411

0.41

5.22E-01

UniBacilli

585

636

2.65

1.04E-01

382

230

306

12.73

3.61E-04

UniEcoli

2574

2800

17.88

2.35E-05

1596

951

1114

16.57

4.69E-05

UniArch

342

248

14.67

1.28E-04

132

120

104

0.28

5.97E-01

(B) UniGolgi

68

CHAPTER 2. THE “NEGATIVE-OUTSIDE” RULE

a background level of 1.0%, an average of 2.6% on the inside flank, and of 2.9% on the
outside flank but glutamic acid’s background is 1.2% but 2.8% on the inside flank and
2.5% on the outside flank. Statistical tests do not support finding a negative-charge
bias in conditions (i) and (ii). Apparently, the problem is TMHs having no or almost
no flanks at one of the sides. Statistical significance for the negative-charge bias is
detected as soon as this problem is dealt with – either by allowing extension of flanks
overlap among neighbouring TMHs as in condition (iii) or by removing examples without proper flank lengths from the dataset as in condition (iv). The respective p-values
are 2.05e-6 and 9.81e-15 respectively.

The issues we had with ExpAll raised the question that, maybe, sequence redundancy in the UniHuman set could have played a role. Therefore, we repeated all
calculations but with UniRef50 instead of UniRef90 for mapping into sequence clusters (see Methods section for detail). We were surprised to see that harsher sequence
redundancy requirements do not affect the outcome of the statistical tests in any major way. For the conditions (i)- (iv), we computed the following p-values: (i) 1.31e-28
(5940 negatively residues inside versus 7492 outside), (ii) 1.38e-36 (5516 versus 7320),
(iii) 5.60e-53 (7089 versus 9233) and (iv) 4.18e-41 (4232 versus 5730).

So, the amplifying effect of some subsets in the overall dataset on the statistical
test that might be caused by allowing overlapping flanks (condition (iii)) is not the
major factor leading to the negative charge skew. Similarly, the trend is also not
caused by sequence redundancy. Thus, we have learned that the negative-charge bias
does also exist in multi-pass TMPs but under the conditions that there are sufficiently
long loops between TMHs. Bluntly said: no loops equals to no charge bias. As
soon as the loops reach some critical length, there are differences between single-pass
and multi-pass TMHs with regard to occurrence and distribution of negative charges
and the inside-suppression/outside-enrichment negative-charge bias appears. Not only
are there more negative charges within the multi-pass TMH itself (in fact, negative
charges are almost not tolerated in single-pass TMHs; see Table 2.1), but also, there is
a much stronger negative outside skew in the TMHs of single-pass proteins than those
of multi-pass proteins.

2.3. RESULTS

2.3.4

69

Further significant sequence differences between singlepass and multi-pass helices: distribution of tryptophan,
tyrosine, proline and cysteine

Amino acid residue profiles along the TM segment and its flanks differ between singleand multi-pass TMHs also in other aspects. The relative percentages of all amino
acid types (normalisation by the total amount of that residue type in the sequence
segment) from single-pass helices of the UniHuman (Figure 2.5A; from 1705 TMHs
with flanks having 68571 residues) and ExpAll (Figure 2.5B; from 1544 TMHs with
flanks having 60200 residues) were plotted as a heat-map. The amino acid types were
listed on the Y axis according to Kyte & Doolittle hydrophobicity [73] in descending
order.
In accordance with expectations, enrichment for hydrophobic residues in the TMH,
for the positively-charged residues on the inside flank as well as a distribution the negative distribution bias was found in both datasets. Additionally, the inside interfacial
region showed consistent enrichment hotspots for tryptophan (e.g., 7.1% at position
-11 in ExpAll, 6.2% at position -10 in UniHuman with flanks after central TMH alignment) and tyrosine (6.4% at -11 in ExpAll, 7.1% at -11 in UniHuman), and some
preference can also be seen for the outer interfacial region (e.g., 5.2% at position 11
for tryptophan in ExpAll, and 5.8% at position 10 for tryptophan in UniHuman) albeit the “hot” cluster of the outer flank covers fewer positions than that of the inner
flank. Further, there is an apparent bias of cysteine on the inner flank and interfacial
region (e.g., 5.5% at position -10 in ExpAll, 5.9% at position -11 in UniHuman), and
a depression in the outer interfacial region and flank (up to a minimum of 0.3% in
both ExpAll and UniHuman). Proline appears to have a depression signal on the outer
flank. Note that, in a similar way to Figures 2.2 and 2.3, the distributions of the flanks
derived from centrally aligned TMHs are corroborated by the distributions from the
database defined TMH boundary flanks (see outside bands in Figures 2.5A-D).
A similar heatmap was generated for UniHuman multi-pass (Figure 2.5C; from
12353 TMHs with flanks having 452708 residues) TMHs and ExpAll multi-pass (Figure
2.5D; from 15563 TMHs with flanks having 535599 residues). Whereas Figures 2.5A-C
appear quite noisy, the plot for ExpAll multi-pass TMHs appears almost Gaussian-like

70

CHAPTER 2. THE “NEGATIVE-OUTSIDE” RULE

UniHuman
A)

Inner
flank

Central TMH residue
0

-20

-10

10

0

20

Outer
flank

Central TMH residue
0

-10

10

0

Inner
flank
-20

-10

0

10

20

10

0

Single-pass

-10

ExpAll
B)

Outer
flank

Relative
Percentage

Relative
Percentage
From TMH central residue 0
In flanks excluding TMH
0

C)

Inner
flank

From TMH central residue 0
In flanks excluding TMH
0

9.4
9.3

0

-20

-10

0

10

D)

Outer
flank

Central TMH residue
20

0

10

Inner
flank
-10

5.0
5.0

9.9
10.1

Outer
flank

Central TMH residue
0

-20

-10

0

10

20

0

10

Multi-pass

-10

4.7
4.7

Relative
Percentage

Relative
Percentage
From TMH central residue 0
In flanks excluding TMH
0

4.6
5.3

9.2
10.7

From TMH central residue 0
In flanks excluding TMH
0

4.1
5.5

8.2
11.0

Figure 2.5: Relative percentage heat-maps from predictive and experimental datasets
corroborate residue distribution differences between transmembrane helices from single-pass and multi-pass proteins. The residue position aligned to the centre of the TMH is on the
horizontal axis, and the residue type is on the vertical axis. Amino acid types are listed in order
of decreasing hydrophobicity according to the Kyte and Doolittle scale [52]. The flank lengths in
the TMH segments were restricted to up to ±10 residues. The scales for each heat-map are shown
beneath the respective subfigure. The darkest blue represents 0% distribution, whilst the darkest
red represents the maximum relative percentage distribution that is denoted by the keys in each
subfigure, with white being 50% between “cold” and “hot”. The central TMH subplots extend from
the central TMH residue, whereas the inner and outer flank subplots use the database-defined TMH
boundary and extend from that position. a TMHs from the single-pass UniHuman dataset. b
single-pass protein TMHs from the ExpAll dataset. c TMHs from the proteins of the multi-pass
UniHuman dataset. d TMHs from ExpAll multi-pass proteins. The general consistency in relative
distributions of every residue type between single-pass and multi-pass of either dataset including
flank/TMH boundary selection allows us to infer biological conclusions from these distributions that
are independent of methodological biases used to gather the sequences. The only residue that displays
drastically differently between the datasets is cysteine in multi-pass TMHs only. The most striking
differences in distributions between residues from TMHs of single-pass and multi-pass proteins include
a more defined Y and W clustering at the flanks, a suppression of E and D on the inside flank, a
suppression of P on the inside flank and a topological bias for C favouring the inside flank.

2.3. RESULTS

71

smoothed, thus, indicating the quality of this dataset. Tyrosine and tryptophan in the
multi-pass case do not appear as enriched in the interfacial regions of single-pass TMHs
from both UniHuman and ExpAll. Prolines are only suppressed in the TMH itself and
are not suppressed in the outer flank as in the single-pass case but, indeed, are tolerated
if not slightly enriched in the flanks.

2.3.5

Hydrophobicity and leucine distribution in transmembrane helices in single- and multi-pass proteins

Generally, we see in Figure 2.5 that compositional biases appear more extreme in
the single-pass case, particularly when it comes to polar and non-polar residues being
more heavily suppressed and enriched. To investigate this observation, we calculated
the hydrophobicity at each sequence-position averaged over all TMHs considered (after
having window-averaged over 3 residues for each TMH) using the Kyte & Doolittle hydrophobicity scale [73] (Figure 2.6A) and validated using White and Wimley octanolinterface whole residue scale [139], Hessa’s biological hydrophobicity scale [116], and
the Eisenberg hydrophobic moment consensus scale [144] (Figure 2.7). The total set
of TMHs was split into 15 sets of membrane-spanning proteins (1 set containing single-pass proteins, 13 sets each containing TMHs from 2-, 3-, 4-. . . 14-TMPs and another
of TMHs from proteins with 15 or more TMHs). In Figure 2.6B, we show the p-value at
each sequence position by comparing the respective values from multi-pass and singlepass TMHs using the 2-sample t-test (Figure 2.6B). Strikingly, the inside flank of the
single-pass TMHs is much more hydrophilic (e.g., see the Kyte & Doolittle score=-1.3
at position -18) than that of multi-pass TMHs (p-value=5.64e-103 at position -14).
Most likely, the positive inside rule, along with the interfacial clustering of tryptophan
and tyrosine, contribute to a strong polar inside flank in single-pass helices that is not
present in multi-pass helices en masse. Further, multi-pass TMHs cluster remarkably
closely within the TM core; the respective hydrophobicity is apparently not dependent
on the number of TMHs in a given multi-pass TMP. On average, single-pass TMHs
are more hydrophobic in the core than multi-pass TMHs (p-value<1.e-72 within positions -5. . . 5 and p-value=5.92e-190 at position 0). On the other hand, hydrophobicity
differences between TMHs from single- and multi-pass proteins fade somewhat at the

72

CHAPTER 2. THE “NEGATIVE-OUTSIDE” RULE

transition towards the flanks (p-value=1.85e-4 at position -10, and p-value=3.35e-31
at position 10).

UniHuman

Single-pass

Kyte & Doolittle
Hydrophobicity

3
2

Multi-pass

1
0

>15-pass

-1
-2
-25 -20 -15 -10 -5 0
Inside

B)

UniHuman
1.E+00
1.E-01
1.E-02
1.E-03
1.E-04
1.E-05
1.E-06
1.E-07
1.E-08
1.E-09
1.E-10
1.E-11
1.E-12
1.E-13

P-value

A)

5 10 15 20 25
Outside

-25 -20 -15 -10 -5

0

5

10 15 20 25

Position from centre of helix

Sequence position

Figure 2.6: There is a difference in the hydrophobic profiles of transmembrane helices
from single-pass and multi-pass proteins. a The hydrophobicity of single-pass TMHs compared
to multi-pass segments from the UniHuman dataset. The Kyte and Doolittle scale of hydrophobicity [73] was used with a window length of 3 to compare TMHs from proteins with different numbers
of TMHs. This scale is based on the water-vapour transfer of free energy and the interior-exterior distribution of individual amino acids. The same datasets also had different scales applied (Figure 2.7).
The vertical axis is the hydrophobicity score, whilst the horizontal axis is the position of the residue
relative to the centre of the TMH, with negative values extending into the cytoplasm. In black are
the average hydrophobicity values of TMHs belonging to single-pass TMHs, whilst in other colours
are the average hydrophobicity values of TMHs belonging to multi-pass proteins containing the same
numbers of TMHs per protein. In purple are the TMHs from proteins with more than 15 TMHs per
protein that do not share a typical multi-pass profile, perhaps due to their exceptional nature. b The
Kruskal-Wallis test (H statistic) was used to compare single-pass windowed hydrophobicity values
with the average windowed hydrophobicity value of every TMH from multi-pass proteins at the same
position. The vertical axis is the logarithmic scale of the resultant p-values. We can much more
readily reject the hypothesis that hydrophobicity is the same between TMHs from single-pass and
multi-pass proteins in the core of the helix and the flanks than the interfacial regions, particularly at
the inner leaflet due to leucine asymmetry ( Table 2.4)

Leucine is the most abundant residue in TMHs (Figure 2.1) and is considered one
of the most hydrophobic residues by all hydrophobicity scales. Therefore, it plays a
very influential role in TMH helix-helix and lipid-helix interactions in the membrane
and recognition by the insertion machinery. When looking at the difference in the
abundance of leucine between the inner and outer halves, we find that TMHs from
single-pass proteins have a trend to contain more leucine residues at the cytoplasmic side of TMHs, particularly in the case of TMHs from single-pass proteins (see
Figures 2.2 and 2.5).
This trend is statistically significant for TMHs in many biological membranes (Table 2.4, Figure 2.8). In the most extreme case of UniCress (single-pass), we see 49%
more leucine residues on the inside leaflet than the outside leaflet (p-value=5.41e-24).
This contrasts with UniCress (multi-pass), in which the skew is far weaker, albeit

73

0.5
0

C)

2.5
2
1.5
1
0.5
0
-0.5

-0.5
-25 -20 -15 -10 -5 0

5 10 15 20 25

Sequence position

1
Eisenberg
Hydrophobicity

B)

1

Hessa
Hydrophobicity

A)

White and Wimley
Hydrophobicity

2.3. RESULTS

0.5
0
-0.5
-1

-25 -20 -15 -10 -5 0

5 10 15 20 25

Sequence position

-25 -20 -15 -10 -5 0

5 10 15 20 25

Sequence position

Figure 2.7: There is a difference in the hydrophobic profiles of transmembrane helices
from single-pass and multi-pass proteins. The difference in hydrophobicity between the single-pass and multi-pass datasets stratified by number of TMHs is not due to the choice of scale. As
with Figure 2.6, UniHuman was stratified according to the number of TMHs in each protein. The
mean amino acid hydrophobicity values of TMHs with a sliding unweighted window of 3 residues from
UniHuman proteins at each position were plotted. To validate the findings presented in Figure 2.6A,
several scales of hydrophobicity were used. (A) The White and Wimley whole residue scale [139] is
based on the partitioning of peptides between water and octanol as well as water to POPC. A positive
score indicates a more polar score. (B) The Hessa biological scale [116]. The hydrophobicity values
represent the free energy exchange during recognition of designed peptide TMHs by the endoplasmic
reticulum Sec61 translocon and, therefore, negative values indicate an energetic preference for the
interior of a lipid bilayer. (C) The Eisenberg consensus scale [144] is a scale based on the earlier scales
from Nozaki and Tanford [137], Wolfenden et al. [142], Chothia [131], Janin [141] and the von Heijne
and Blomberg scale [143]. The scales are normalised according to serine. A positive score indicates
a generally more hydrophobic score.

yet statistically significant. There are 6% more leucine residues at the inside half (pvalue=2.08e-4). The trend of having more leucine residues at the cytoplasmic half of
the TMH is observed for all datasets (both single- and multi-pass) except for UniArch
(single-pass). The phenomenon is statistically significant with p-value<1.e-3 for ExpAll, UniHuman, UniPM and UniCress (both single- and multi-pass). As with negative
charge distribution, UniArch presents a reversed effect compared to other single-pass
protein datasets with a 57% reduction in leucine on the inside leaflet compared to the
outside leaflet (p-value=7.25e-6). However, leucine of TMHs from UniArch multi-pass
proteins have no discernible preference for the inside leaflets (4% more on the inside
leaflet, p-value=0.625).

2.3.6

A negative-outside (or negative-non-inside) signal is
present across many membrane types

We explored the presence of amino acid residue compositional skews described above
for human TMPs for those in other taxa and also specifically for human proteins with
regard to membranes at various subcellular localisations. Acidic residues for TMHs
from single-pass and multi-pass helices were plotted according to their relative percentage distributions (of the total amount of this residue type in the respective segment) for

74

CHAPTER 2. THE “NEGATIVE-OUTSIDE” RULE

Table 2.4: Leucines at the inner and outer leaflets of the membrane in transmembrane
helices. The statistical results when comparing the number of leucine residues from the inner and
outer leaflets in each protein in the dataset. The number of helices per dataset can be found in
Table 2.1. The Kruskal-Wallis test scores (H statistics) were calculated for leucine residues by comparing the number of leucine residues that were in the inner half of the leaflet with those in the outer
half of the leaflet of the database-defined TMH
single-pass
Dataset

multi-pass

Inside

Outside

Percentage H statis- p-value
tic

Inside

Outside

Percentage H statis- p-value
tic

ExpAll

4020

3403

118.13

40.07

2.44E-10

27,986

27,008

103.62

14.13

1.70E-04

UniHuman

4982

3697

134.76

193.02

6.99E-44

25,199

22,365

112.67

195.24

2.29E-44
4.61E-02

UniER

359

297

120.88

8.41

3.72E-03 1863

1764

105.61

3.98

UniGolgi

604

513

117.74

10.74

1.05E-03 753

677

111.23

5.61

1.79E-02

UniPM

1485

1006

147.61

98.9

2.65E-23

6221

5577

111.55

35.21

3.00E-09

UniCress

1495

1005

148.76

102.05

5.41E-24

6491

6099

106.43

13.76

2.08E-04

UniFungi

1389

1308

106.19

3.41

6.48E-02

14,505

14,099

102.88

6.74

9.41E-03
5.89E-03

UniBacilli

260

251

103.59

0.03

8.72E-01

1488

1335

111.46

7.59

UniEcoli

130

100

130

2.78

9.53E-02 7251

6975

103.96

5.92

1.50E-02

UniArch

51

118

43.22

20.13

7.25E-06 636

612

103.92

0.24

6.25E-01

five taxon-specific datasets UniCress (Figure 2.8A), UniFungi (Figure 2.8B), UniEcoli
(Figure 2.8C), UniBacilli (Figure 2.8D), UniArch (Figure 2.8E) and for three organellespecific datasets UniER (Figure 2.8F), UniGolgi (Figure 2.8G), UniPM (Figure 2.8H).
For single-pass proteins in all taxon-specific datasets (with the exception of
UniArch), there are more negative residues at the outside than at the inside. The
skew is statistically significant (see Table 2.2, P<0.001) except for UniBacilli. Despite
statistical significance found for UniFungi (p-value=1.12e-7 for database-defined and
p-value=6.79e-10 for flanks after central alignment; Table 2.2), however, the trend is
not very strong in this case (Figure 2.8B). Whereas the skew is just a suppression of
negatively-charged residues at the inside flank for ExpAll and UniHuman (as well as
in UniCress), the bias observed for UniEcoli involves also a negative charge enrichment at the outside flank. In the case of UniArch (Figure 2.8E), we see a negative
inside preference that is 6.0% in the case of aspartic acid, and 6.3% for glutamic acid
(not shown), with much lower values close to 0% on the outside. Whilst the difference is statistically significant for both TMHs (Table 2.2) from single-pass proteins
(p-value=1.83e-12 and p-value=1.43e-11 for two versions of flank determination) and
multi-pass proteins (p-values 4.72e-3, 7.81e-3, 1.28e-4 for three versions of flank determination, see Tables 3A and 3B), the distribution along the position axis is heavily
fluctuating, maybe as a result of the small size of the dataset. However, one can assuredly assign a “negative-inside” tendency to the flanking regions of Archaean TMHs.
In the human organelle datasets, we see trend shifts at different stages in the

2.3. RESULTS

75

Single-pass
UniCress

A)
Inside

B)

C)

D)

Multi-pass

Outside

UniFungi

UniEcoli

UniBacilli

UniArch

E)
Inside

Multi-pass

Single-pass

L
R and K
D and E

Inside

Outside

F)

G)

H)

Outside

Inside

Outside

UniER

UniGolgi

UniPM

Figure 2.8: Comparing charged amino acid distributions in transmembrane helices of
multi-pass and single-pass proteins across different species and organelles. The relative
percentage distribution of charged residues and leucine was calculated at each position in the TMH
with flank lengths of ±20 in different datasets. The distributions are normalised according to relative
percentage distribution. Aspartic acid and glutamic acid are shown in dark purple and light purple
respectively. Leucine, the most abundant non-polar residue in TMHs, is in blue. Arginine and lysine
are shown in orange. TMHs from single-pass proteins are on the left and TMHs from multi-pass
proteins are on the right for different taxonomic datasets: a UniCress, b UniFungi, c UniEcoli, d
UniBacilli, e UniArch, and different organelles: f UniER, g UniGolgi, h UniPM. As a trend, the
negative-outside skew is more present in TMHs from single-pass proteins than multi-pass proteins
(Tables 2 and 3). Another key observation is that in single-pass TMHs there is a propensity for
leucine on the inner over the outer leaflet (Table 2.4)

secretory pathway. In UniER, there is an enrichment of negative charge on the outside
flank of 1–1.5% that is comparable to the magnitude of the positive inside signal. In
UniGolgi, there is a suppression of negatively-charged residues on the inside flank as
well as an enrichment on the inside flank resulting in ∼2% distribution difference. For
UniPM, there is a negative-inside suppression (but no outside enrichment) as well as

76

CHAPTER 2. THE “NEGATIVE-OUTSIDE” RULE

a positive-inside signal. All observed trends are statistically significant (see Table 2.2,
P<1.e-5).
For multi-pass TMH proteins, we see either the same trends but in a weaker form
or no skews are observed at all as inspection of the graphs in Figure 2.8 shows. For
datasets UniER, UniGolgi, UniCress, UniFungi, and UniBacilli, the hypothesis of equal
distribution of negatively-charged residues cannot be rejected (p-value>0.001, see Table 2.3); thus, a skew is statistically non-significant. Although UniPM has a statistically significant bias (p-value<4.30e-12, Table 2.3), the trends are more subtle
and most present for aspartic acid of UniPM. We see many more negative and positive charges tolerated within the multi-pass TMHs themselves throughout all datasets
(Table 2.1). To note, there is a positive-inside rule for all multi-pass datasets studied
herein.
To conclude, we find that negative-charge bias distribution is a feature of singlepass protein TMHs that is present across many membrane types and it can have the
form of a negative charge suppression at the inside flank or an enrichment of those
charges at the outside flank.

2.3.7

Amino acid compositional skews in relation to transmembrane helix complexity and anchorage function

In previous work, we studied the relationship of TMH composition, sequence complexity and function [44–46] and concluded that simple TMHs are more probably
responsible for simple membrane anchorage, whereas complex TMHs have a biological
function beyond just anchorage. We wished to see how the skews observed in this
work relate to that classification. Therefore, the single-pass TMHs from UniHuman
and ExpAll were separated into subsets of simple, twilight, and complex TMHs using
TMSOC [44, 45]. The relative percentages of eight residue types (L, D, E, R, K, Y,
W, C; normalisation with the total amount of residues of that amino acid type in
all sequence segments considered) were plotted along the sequence position for simple
and complex helices (Figure 2.9). Of UniHuman single-pass proteins, there were 889
records with simple TMHs and 570 with complex TMHs (Figure 2.9B). In ExpAll,
769 TMHs from single-pass proteins were simple TMHs and 570 were complex TMHs.

D)

C)

2.3. RESULTS

77
E)

F)

G)

H)

Simple single-pass
Complex single-pass
Simple single-pass
Multi-pass
Complex single-pass

UniHuman

A)

Inside

ExpAll
Outside

B)

Inside

Outside

B)

Inside

C)

D)

C)

D)

E)

F)

E)

F)

G)

H)

G)

H)

I)

J)

ExpAll

Outside

Outside

UniHuman

ExpAll

I)

J)

K)

L)

M)

N)
S, N, Q, T, Y, C
Relative percentage

UniHuman

Inside

S, N, Q, T, Y, C
Relative percentage

Multi-pass

A)

Figure
2.9: Comparing the
amino acid relative percentage distributions of simple and
I)
J)
complex transmembrane helices from single-pass proteins and transmembrane helices
from multi-pass proteins. Comparing the amino acid relative percentage distributions of simple
and complex TMHs from single-pass proteins and TMHs from multi-pass proteins. TMSOC was
used to calculate which single-pass TMHs were complex and which were simple from ExpAll and
UniHuman datasets. Simple TMHs are typically anchors without necessarily having other functions
K)
L)
(Wong et al. [46]). The relative percentages from single-pass simple (shown in light blue), single-pass
K)
L)
complex (red), and multi-pass protein TMHs (black) were plotted for (a, c, e, g, i and k) UniHuman
and (b, d, f, h, j and l) ExpAll for (a and b) positive residues, (c and d) negative residues, (e and
f) tyrosine, (g and h) tryptophan, (i and j) leucine and (k and l) cysteine (m and n) uncharged
polar amino acids; serine, asparagine, glutamine, threonine, tyrosine, and cysteine. The slopes are
statistically compared in Tables 2.5 and 2.6, and as a trend, the profiles of complex TMHs are more
similar to multi-pass TMH profiles than simple TMHs are to multi-pass TMHs

78

CHAPTER 2. THE “NEGATIVE-OUTSIDE” RULE
It is visually apparent (Figure 2.9) that there are (i) stronger skews and more

inside-outside disparities in simple single-pass TMs than in complex single-pass TMs
and (ii) greater similarities between single-pass complex TM regions and those from
multi-pass proteins compared with simple single-pass TMs in comparison with either
of the other two distributions. To examine the statistical significance of these observations, we compared the amino acid distributions (K, R, K+R, D, E, D+E, Y, W, L,
C) across the range of TMHs with flank lengths ±10 residues using the KolmogorovSmirnov (K-S), K-W and the χ2 statistical tests. To note, the K-S test scrutinises for
significant maximal absolute differences between distribution curves; the glskw test is
after skews between distributions and the χ2 statistical test checks the average difference between distributions. Calculations were carried out over single-pass complex,
single-pass simple and multi-pass TMH datasets from both ExpAll and UniHuman (for
p-values and Bahadur slopes, Table 2.5 (dataset UniHuman) and Table 2.6 (dataset
ExpAll)).
There is also a visual difference between simple single-pass proteins, complex single-pass proteins, and multipass proteins with regard to uncharged polar amino acids
(serine, asparagine, glutamine, threonine, tyrosine, and cysteine), with complex single-pass TMHs being between multipass TMHs and simple single-pass TMHs in terms
of relative percentage profile across the membrane (Figure 2.9M and Figure 2.9N).
TMSOC uses hydrophobicity as part of the scrutinisation between simple and complex
TMHs, so it is not surprising that there are differences in polar residues between
simple and complex TMHs. However, it is interesting to note that the reduction in
polar residues is not reduced through the TMH and flanks of simple and complex
TMHs evenly; simple TMHs have less uncharged polar residues in the core of the
TMH than the complex TMHs relative to the flanking areas. Because there was no
observable inside-outside flank skews in the distributions, no further statistical analysis
was carried out on this set.
Many low p-values in Tables 2.5 and 2.6 indicate significant differences between
the three distributions studied. For the UniHuman dataset (Table 2.5), we find most
striking, significant differences between charged residue distributions (R, K, D, E)
of simple and complex single-pass TMH+flank regions (χ2 p-value<2.23e-3 for single

2.3. RESULTS

79

Table 2.5: Simple transmembrane helices are less similar than complex transmembrane
helices to transmembrane helices from multi-pass proteins in UniHuman. The statistical
results were gathered by comparing complex single-pass TMHs, simple TMHs from single-pass proteins and TMHs from multi-pass proteins in UniHuman. The abundance of different residues at each
position when using the centrally aligned TMH approach was compared with several statistical tests
(the K-S, K-W and the χ2 statistical tests) and the Bahadur slope values of those results
p-values for χ2
Residues

Bahadur slopes for χ2

Simple-vscomplex

Simple-vsmulti

Complexvs-multi

Simple-vscomplex

Simple-vsmulti

Complexvs-multi

R

3.20E-06

7.38E-02

1.24E-01

6.61E-03

2.20E-03

1.27E-04

K

2.23E-03

4.99E-02

2.14E-01

3.99E-03

3.70E-03

1.18E-04

D

1.67E-09

3.06E-01

3.02E-01

3.34E-02

3.24E-03

1.20E-04

E

3.80E-07

2.34E-01

2.31E-01

1.81E-02

3.05E-03

1.36E-04

Y

3.86E-01

3.97E-01

2.11E-01

1.06E-03

1.47E-03

8.25E-05

W

3.77E-03

2.97E-01

3.84E-01

8.52E-03

2.73E-03

1.13E-04

L

3.59E-01

2.88E-01

3.21E-01

1.52E-04

3.92E-04

1.69E-05

C

6.44E-01

3.97E-01

3.41E-01

4.29E-04

1.29E-03

8.57E-05

R+K

2.19E-02

2.83E-01

2.52E-01

1.11E-03

6.33E-04

4.68E-05

D+E

1.47E-03

2.86E-01

2.79E-01

4.59E-03

1.49E-03

6.15E-05

p-values for Kolmogorov-Smirnov

Bahadur slopes for KolmogorovSmirnov

Simple-vscomplex

Simple-vsmulti

Complexvs-multi

Simple-vscomplex

Simple-vsmulti

Complexvs-multi

R

2.31E-01

3.57E-04

1.08E-02

7.66E-04

6.71E-03

2.76E-04

K

4.31E-02

2.18E-03

8.93E-01

2.06E-03

7.56E-03

8.68E-06

D

1.39E-01

5.02E-06

1.08E-02

3.26E-03

3.34E-02

4.52E-04

E

7.96E-02

1.58E-05

1.08E-02

3.10E-03

2.32E-02

4.20E-04

Y

7.96E-02

2.22E-02

2.31E-01

2.81E-03

6.07E-03

7.78E-05

W

2.31E-01

9.06E-04

4.31E-02

2.24E-03

1.58E-02

3.70E-04

L

2.31E-01

2.31E-01

5.31E-01

2.17E-04

4.61E-04

9.42E-06

C

1.39E-01

3.61E-01

3.61E-01

1.93E-03

1.42E-03

8.10E-05

R+K

7.96E-02

1.33E-04

7.96E-02

7.35E-04

4.48E-03

8.60E-05

D+E

4.31E-02

1.58E-05

4.98E-03

2.21E-03

1.31E-02

2.55E-04

p-values for Kruskal-Wallis

Bahadur slopes for Kruskal-Wallis

Simple-vscomplex

Simple-vsmulti

Complexvs-multi

Simple-vscomplex

Simple-vsmulti

Complexvs-multi

R

2.19E-01

5.06E-02

2.37E-01

7.92E-04

2.52E-03

8.79E-05

K

2.90E-01

1.33E-01

7.00E-01

8.11E-04

2.49E-03

2.73E-05

D

3.50E-01

1.81E-02

2.81E-01

1.74E-03

1.10E-02

1.27E-04

E

2.59E-01

5.65E-02

1.78E-01

1.65E-03

6.04E-03

1.60E-04

Y

6.03E-01

4.53E-01

4.41E-01

5.62E-04

1.26E-03

4.34E-05

W

4.19E-01

1.84E-01

5.70E-01

1.33E-03

3.81E-03

6.62E-05

L

6.37E-01

4.88E-01

9.77E-01

6.68E-05

2.25E-04

3.47E-07

C

5.00E-01

2.22E-01

9.62E-01

6.76E-04

2.10E-03

3.11E-06

R+K

1.87E-01

8.67E-02

4.08E-01

4.86E-04

1.23E-03

3.05E-05

D+E

1.68E-01

4.52E-02

1.91E-01

1.25E-03

3.68E-03

7.97E-05

80

CHAPTER 2. THE “NEGATIVE-OUTSIDE” RULE

Table 2.6: Simple transmembrane helices are less similar than complex transmembrane
helices to transmembrane helices from multi-pass proteins in ExpAll. As in Table 2.5, the
statistical results were gathered by comparing complex single-pass TMHs, simple TMHs from single-pass proteins and TMHs from multi-pass proteins; however, in this case only ExpAll is used. The
abundance of different residues at each position when using the centrally aligned TMH approach was
compared with several statistical tests (the K-S, K-W and the χ2 statistical tests) and the Bahadur
slope values of those results
p-values for χ2
Residues

Bahadur slopes for χ2

Simple-vscomplex

Simple-vsmulti

Complexvs-multi

Simple-vscomplex

Simple-vsmulti

Complexvs-multi

R

5.10E-06

2.98E-01

5.10E-06

9.17E-03

1.61E-03

6.23E-05

K

2.35E-03

1.85E-01

2.35E-03

4.81E-03

3.88E-03

9.78E-05

D

2.61E-08

1.84E-01

2.61E-08

4.15E-02

7.90E-03

1.41E-04

E

2.38E-10

2.04E-01

2.38E-10

3.88E-02

7.08E-03

1.22E-04

Y

3.03E-01

3.11E-01

3.03E-01

2.01E-03

2.49E-03

5.51E-05

W

4.21E-03

4.29E-01

4.21E-03

1.11E-02

4.76E-03

6.46E-05

L

3.79E-01

3.04E-01

3.79E-01

2.28E-04

4.66E-04

1.50E-05

C

3.87E-01

2.52E-01

3.87E-01

1.75E-03

3.28E-03

1.48E-04

R+K

7.16E-04

2.52E-01

7.16E-04

2.80E-03

1.28E-03

3.76E-05

D+E

3.58E-05

2.94E-01

3.58E-05

1.03E-02

1.94E-03

4.90E-05

p-values for Kolmogorov-Smirnov

Bahadur slopes for KolmogorovSmirnov

Simple-vscomplex

Simple-vsmulti

Complexvs-multi

Simple-vscomplex

Simple-vsmulti

Complexvs-multi

R

3.61E-01

4.31E-02

3.61E-01

7.66E-04

7.79E-03

1.62E-04

K

4.31E-02

8.93E-01

4.31E-02

2.49E-03

1.05E-02

6.57E-06

D

1.39E-01

2.18E-03

1.39E-01

4.68E-03

3.61E-02

5.10E-04

E

5.31E-01

1.33E-04

5.31E-01

1.11E-03

2.81E-02

6.87E-04

Y

2.31E-01

9.06E-04

2.31E-01

2.47E-03

6.26E-03

3.30E-04

W

5.31E-01

4.98E-03

5.31E-01

1.29E-03

1.13E-02

4.04E-04

L

2.31E-01

2.31E-01

2.31E-01

3.45E-04

2.12E-03

1.85E-05

C

5.31E-01

3.61E-01

5.31E-01

1.16E-03

8.91E-04

1.09E-04

R+K

1.39E-01

2.31E-01

1.39E-01

7.61E-04

4.82E-03

4.00E-05

D+E

1.39E-01

9.06E-04

1.39E-01

1.99E-03

1.41E-02

2.80E-04

p-values for Kruskal-Wallis

Bahadur slopes for Kruskal-Wallis

Simple-vscomplex

Simple-vsmulti

Complexvs-multi

Simple-vscomplex

Simple-vsmulti

Complexvs-multi

R

4.37E-01

3.92E-01

4.37E-01

6.24E-04

2.52E-03

4.82E-05

K

3.83E-01

6.93E-01

3.83E-01

7.62E-04

2.88E-03

2.13E-05

D

4.49E-01

1.81E-01

4.49E-01

1.90E-03

1.06E-02

1.42E-04

E

7.64E-01

1.94E-01

7.64E-01

4.71E-04

9.05E-03

1.26E-04

Y

8.32E-01

3.36E-01

8.32E-01

3.09E-04

9.63E-04

5.15E-05

W

7.25E-01

1.36E-01

7.25E-01

6.53E-04

5.44E-03

1.52E-04

L

7.15E-01

7.95E-01

7.15E-01

7.90E-05

3.41E-04

2.90E-06

C

8.47E-01

9.54E-01

8.47E-01

3.05E-04

4.26E-05

5.06E-06

R + K

2.89E-01

5.13E-01

2.89E-01

4.79E-04

1.41E-03

1.82E-05

D+E

4.94E-01

2.07E-01

4.94E-01

7.11E-04

4.14E-03

6.29E-05

2.3. RESULTS

81

amino acid types). Similarly, simple single-pass TMH+flank segments differ significantly from multi-pass TMH+flank segments (K-W test p-values<3.e-2 for R, K, D,
E, Y, W amino acid types as well as for K+R and D+E). The trends are the same for
the ExpAll dataset (Table 2.6): simple and complex single-pass TMH+flank regions
differ in charged amino acid type distributions (χ2 p-value<4.21e-3 for all cases), as
well as simple single-pass and multi-pass ones, do (K-W test p-values<5.e-2 for R, D,
E, Y, W amino acid types and D+E).
Whereas p-value tests for significant differences between distributions depend
strongly on the amount of data, the more informative Bahadur slopes that measure the
distance from the zero hypothesis are independent of the amount of data [259–261]. As
we can see in Tables 2.5 and 2.6, the absolute Bahadur slopes for the simple single-pass
to multi-pass comparison are always larger (even by at least an order of magnitude):
(ii) for all three statistical tests applied (χ2 , K-S and K-W), (ii) for all amino acid
types, for K+R and E+D and (iii) for both datasets UniHuman and ExpAll. Thus,
complex single-pass TMH+flanks have compositional properties that are indeed very
similar to those of multi-pass ones (which are known to have a large fraction of complex TMHs [44, 45]). This strong evidence implies that the actual issue is not so much
about single- and multi-pass TMH segments but between simple and complex TMHs
where the first are exclusively guided by the anchor requirements whereas the latter
have more complex restraints to fulfil.
Several distribution features of simple TMHs from single-pass proteins when compared to complex TMHs from single-pass proteins and TMHs from multi-pass proteins
that contribute to the statistical differences (Figure 2.9) are especially notable. There
is a more pronounced trend for positively-charged residues and tyrosine to be preferentially located on the inside flanks and for negatively-charged residues to be on
the outside flanks. The symmetrical peaks in the percentage distribution of tyrosine
in complex single-pass TMHs are more akin to multi-pass TMHs, whereas in simple TMHs the distribution resembles a more typical single-pass helix (compare with
Figure 2.2). Furthermore, the depression of charged residues within the TMH itself is
strongest in simple single-pass TMHs.
To emphasise, tryptophan is essentially not tolerated within the simple TMHs and
there are higher peaks of tryptophan occurrence at either flank. We also see a strong

82

CHAPTER 2. THE “NEGATIVE-OUTSIDE” RULE

inside skew for leucine clustering within the core of simple TMHs which is not present
in the “flatter” distributions of complex single-pass TMHs and TMHs from multi-pass
proteins.
There is obviously a cysteine-inside preference for simple, single-pass TMHs but
less in complex, multi-pass TMHs (Figure 2.9). This conclusion is contrary to a
previous study [258] but that deduction was drawn from a much smaller dataset of 45
single-pass TMHs and 24 multi-pass TMPs.

2.4

Discussion

The “negative-outside/non-negative inside” skew in TMHs and their flanks is statistically significant We have seen that, consistently throughout the datasets, there is
a trend for generally rare negatively-charged residues to prefer the outside flank of
a TMH rather than the inside (and to almost completely avoid the TMH itself); be
it by suppression on the inside and/or enrichment on the outside. The trend is much
stronger in single-pass protein datasets than in multi-pass protein datasets. However
as we elaborated on further, the real crux of the bias appears to be associated with
the TMH being simple or complex [44, 45], thus, whether or not the TMH has a role
beyond anchorage. The existence of this bias has implications for topology prediction
of proteins with TMHs, engineering membrane proteins as well as for models of protein
transport via membranes and protein-membrane stability considerations.
It should be noted that the controversy in the scientific community about the existence of a negative-charge bias at TMHs was mainly with regard to multi-pass TMPs.
Despite having access to much larger, better annotated sequence datasets and many
more 3D structures than our predecessors, we also had our share of difficulties here
(see Results section 2.3.3 and Table 2.3). The straightforward approach results in
inconclusive statistical tests if datasets become small (for example, if selections are
restricted to subcellular localisations, 3D structures or if very harsh sequence redundancy criteria are applied) and, especially, if TMHs with very short or no flanks are
included. Therefore in the case of multi-pass proteins, we studied flanks as taken from
the TM boundaries in the databases under several conditions: (i) without allowing
flank overlap between neighbouring TMHs, (ii) as subset of (i) but with requiring

2.4. DISCUSSION

83

some minimal flank length at either side, (iii) with overlapping flanks. We also studied flanks after central alignment of TMHs and assuming standardised TMH length.
multi-pass TMHs (without overlapping flanks) do not show statistically significant
negative-charge bias under condition (i) but, apparently, due to many TMHs without
any or super-short flanks at least at one side. Significance appears as soon as subsets
of TMHs with flanks at both sides are studied. Not surprisingly, there is no charge bias
if there are no flanks in the first place. It is perhaps worth noting that the results from
multi-pass TMHs with overlapping flanks may involve amplification of skews since it
involves multiple counting of the same residues. Given the redundancy threshold of
UniRef90, we cannot rule out that these statistical skews are the result of a trend from
only a small sub-group of TMPs which is being amplified. Hence, we also needed to
observe if these same observed biases were true in condition (ii), which is indeed the
case.
As the “negative-outside/negative-not-inside” skew is widely observed among varying taxa and subcellular localisations with statistical significance, it appears to, at
least to a certain extent, be caused by physical reasons and be associated with
the background membrane potential.

Several earlier considerations and observa-

tion support this thought: (i) Firstly, a concert between the negative and positive
charge on the TMH flanks drives anchorage and the direction of insertion of engineered TMHs [243, 254]. (ii) The inner leaflet of the plasmalemma tends to be more
negatively-charged [262]. Specifically, phosphatidylserine was found to distribute in
the cytosolic leaflets of the plasma membrane and it was found to electrostatically interact with moderately positive-charged proteins enough to redirect the proteins into
the endocytic pathway [263]. The negative charge of proteins at the inside of the
plasma-membrane would decrease the anchoring potency of the TMH via electrostatic
repulsion. (iii) Thirdly in membranes that maintain a membrane potential, there are
inevitably electrical forces acting on charged residues during chain translocation as
this influences the translocon machinery when orienting the TMH. Therefore, it is no
surprise that we see an inside-outside bias for negatively-charged residues that is opposite to the one for positively-charged residues. The negative charges in TMH residues
have been shown to experience an electrical pulling force as they pass through the
bacterial SecYEG translocon import [42, 253]. Also, they are known to be involved in

84

CHAPTER 2. THE “NEGATIVE-OUTSIDE” RULE

intra-membrane helix-helix interactions [264]. For example, aspartic acid and glutamic
acid can drive efficient di- or trimerisation of TMHs in lipid bilayers and, furthermore,
that aspartic acid interactions with neighbouring TMHs can directly increase insertion
efficiency of marginally hydrophobic TMHs via the Sec61 translocon [264]. In support
of this, less acidic residues are found in single-pass TMHs, among which only some
will undergo intra-membrane helix-helix interactions. As the mutation studies have
shown negative charge as a topological determinant [248], therefore, it is perhaps no
surprise that we observe a skew in negatively-charged residues in a similar manner to
the skew in positively-charged residues.
Whereas the “negative-outside/negative-not-inside” skew is observed for distantly
related eukaryotic species and it is also present in Gram-negative bacteria such as
E. coli, this sequence pattern was not observed for the Gram-positive bacteria in
which there is no observable bias. In contrast, Archaea have a statistically significant
“negative-inside” propensity both for single- and multi-pass TMPs. It is known that
Archaea have remarkably different membranes compared to other kingdoms of life due
to their extremophile adaptations to stress [265]. Whilst it is unclear why negative
charge is distributed so differently in UniArch to the other taxonomic datasets, one
must appreciate that a much more nuanced approach would be needed to draw formal conclusions about Archaea, which current databases cannot provide due to the
relatively limited information and annotation of Archaean proteomes.
Methodological issues made previous studies struggle to identify negatively-charged
skews with statistical significance
Whereas the influence of a negative-charge bias in engineered proteins with TM
regions on the direction of insertion into the membrane was solidly established [114,
248, 250, 251, 266], the search for the negative charge distribution pattern in the
statistics of sequences of TM proteins from databases failed to find significance for the
expected negative charge skew [86, 109, 110, 114, 117, 118].
Generally speaking, the datasets from previous studies have been considerably
smaller compared with those in our work (only Sharpe et al. had a similar order of
magnitude [86]), especially those with experimental information about 3D structure
and membrane topology that we used for validation. And they might not have had
the luxury of using UniProt’s improved TRANSMEM consensus annotation based on

2.4. DISCUSSION

85

a multitude of TM prediction methods and experimental data, but this is also not
the major issue. We found that there are other factors that are critical for observing
sequence bias such as negative charge skew in the case of TMHs.
i Acidic residues are rare near and within TMH and biases in their distribution
are easily blurred by minor fluctuations of much more frequent amino acid types,
most notably leucine. Therefore, the method of normalisation is critical. We have
shown that normalising by the total amount of residues of the amino acid type
studied within the sequence region under consideration is appropriate to answer
the question where to find a negatively-charged residue if there is any at all (called
“relative percentage” in this work).
ii The alignment of the TMHs is critical. It was common practice to align TMH
according to the most cytosolic residue [86] although it is known that the membrane/cytosol boundary of the TMH is not well defined (and the exact boundary is
even less well understood at the non-cytosolic side). Aligning the TM regions and
their flanks from the center of the TMH was first proposed by Baeza-Delgado et
al. [109]. Since we know now that acidic residues are often suppressed in the cytosolic flank and within the TMH, this implies that the few acidic residues found in the
cytosolic interface would appear more comparable to those in the poorly defined
non-cytosolic interface as the respective residues are spread over more potential
positions, diminishing any observable bias.
iii We find that separation into single- and multi-pass TM datasets (or, even better,
simple and complex TMHs [44, 45]) is critical to study the inside/outside bias. As
many TMHs in multi-pass TMPs have essentially no flanks or very short flanks if
the condition of non-overlap is applied to flanks of neighbouring TMHs, this might
also obscure the observation of the negative-charge bias. If there are no flanks,
then there will be no residue distribution bias in these flanks. The problem can be
alleviated by either studying only subsets with minimal flank lengths on both sides
(although datasets might become too small for statistical analysis) or by allowing
flank overlaps between neighbouring TMHs.
iv This classification is even more justified in the light of previous reports about the
“missing hydrophobicity” in multi-pass TMHs [74, 248, 267, 268]. Otherwise, the

86

CHAPTER 2. THE “NEGATIVE-OUTSIDE” RULE
distribution bias well observed among the exclusive anchors could be lost to noise.
This addresses the more biologically contextualised issue that there are different
evolutionary pressures on different types of TMHs. The negative charge skew is
most pronounced for dedicated anchors frequently found with simple TMHs typically observed in single-pass TM proteins. These TMHs are pressured to exhibit
residue biases that may aid anchorage in a topologically correct manner. Complex TMHs, typically within multi-pass membrane proteins that have a function
beyond anchorage, comply with a multitude of restraints structural and functional
constraints and the negative charge skew is just one of them.
The most representative precedent papers are those of Sharpe et al. [86] from

2010 (with 1192 human and 1119 yeast single-pass TMHs), Baeza-Delgado et al. [109]
(with 792 TMHs mixed from single- and multi-pass TMPs) and Pogozheva et al. [110]
(TMHs from 191 mixed from single- and multi-pass TMPs with structural information)
both from 2013. Whereas the first analysis would have benefited from the central
alignment approach and the first two studies from another normalisation as described
above, the third study did come close to our findings. To note, their dataset mixed
with single- and multi-pass proteins was too small for revealing the negative-charge
bias with significance; yet, they observed total charge differences at either sides of the
membrane varying for both single- and multi-pass proteins. Membrane asymmetry due
to positively-charged residues occurring more frequently on the cytosolic side causes
net charge unevenness at both sides of the membrane. This observation has been
known to correlate with orientation for decades [109, 113, 264]. Our data shows that
the negative charge skew contributes to this asymmetry.
There are differences in charged amino acid residue biases in TMH flanks through
each stage of the secretory pathway
Here, we observe differences throughout sub-cellular locations along the secretory
pathway. We found that negative charges are enriched at the outside flank (in the ER),
both enriched outside and suppressed inside for the Golgi membrane, and suppressed
on the inside flank in the PM. It has been suggested that the leaflets of different membranes have different lipid compositions throughout the secretory pathway [92] and this
has led to general biochemical conservation in terms of TMH length and amino acid
composition in different membranes [86, 110]. However, herein only the organelles with

2.4. DISCUSSION

87

the most protein record annotation were used. Further investigation into the TMPs
of lysosomes, endosomes, and other ER-Golgi transition vesicles would yield more information on this. Furthermore, there could be study into TMPs not associated with
the signal peptide TMPs destined for the secretory pathway such as those TMPs embedded in the membranes of the mitochondrion, apicoplast, chromoplast, chloroplast,
cyanelle, thylakoid, amyloplast, peroxisome, glyoxysome, and hydrogenosome.
Lipid asymmetry in the Golgi and PM (in contrast to the ER) has been known
about for over a decade [269, 270]. To note, the Golgi and PM have lipid asymmetry with sphingomyelin and glycosphingolipids on the non-cytosolic leaflet, and
phosphatidylserine and phosphatidylethanolamine enriched in the cytosolic leaflet. Although the ER is the main site for cholesterol synthesis, it has markedly low concentrations of sphingolipids [271]. Golgi synthesises sphingomyelin, a lipid not present in
the ER, but present in both the Golgi [272] and in the PM [184, 273]. The PM is also
enriched with densely packed sphingolipids and sterols [274]. Another factor influencing the sequence patterns of TMHs and their along the secretory pathway appears to
be the variation in membrane potentials [100, 101, 103].
Several sequence features can be assigned to anchor TMHs: Charged-residue flank
biases, leucine intra-helix asymmetry, and the “aromatic belt”.
We investigated the difference between TMHs from single-pass and multi-pass proteins and found significant differences in sequence composition that are reflective of the
biologically different roles the TMHs play. To emphasise and validate these findings,
we separated TMHs from single-pass proteins into simple and complex TMHs [44,
45]; ones that likely contains mostly TMHs that act as exclusive anchors, and another
that have roles beyond anchorage. This leaves us with “anchors” (simple TMHs from
single-pass proteins) and “non-anchors” (complex TMHs from single-pass proteins,
and TMHs from multi-pass proteins). If there are strong sequence feature differences
between anchors and non-anchors, it is likely that the sequence feature has a role in
satisfying membrane constraints to act as an energetically optimally stable anchor.
Future studies in the area would desirably directly include a comprehensive analyses
of datasets oligomerised TMHs from single-pass proteins and ascertain if they appear
to be more similar to simple anchors, multi-pass, or generally neither. Currently, no
sufficiently complete set of intra-membrane oligomerised single-pass proteins exists

88

CHAPTER 2. THE “NEGATIVE-OUTSIDE” RULE

that can be compared to a large set of known non-oligomerising proteins. The current
work sidesteps this issue by comparing single-pass proteins with simple TMHs, which
tend to be simple anchors (as shown in previous work [44, 45]), against datasets that
contain TMHs that will form intra-membrane bundles. Bluntly, the simple/complex
status of a TMH can be easily computed from its sequence with TMSOC whereas
the oligomerisation state of most membrane proteins still needs to be experimentally
determined.
Unsurprisingly, both positively and negatively-charged residues can be seen to be
more strongly distributed with bias in anchors than non-anchors. Both the “positiveinside” rule as well as the “negative-outside/non-negative-inside” bias are mostly observable in simple single-pass TMHs (although they are statistically significant elsewhere). It is perhaps true that where a bias is clearly present in both non-anchors and
anchors alike, it is a strong topological determinant, whereas if the residue is only distributed with topological bias in exclusively anchoring TMHs, we can attribute these
features more specifically to biophysical anchorage. This being said, we should not
rule out that the same features aid topological determination since negative charge has
been shown to be a weaker topological determinant than positively-charged residues
(35).
Tyrosine and tryptophan residues commonly are found at the interfacial boundaries of the TMH and this feature is called the “aromatic belt” [86, 109, 116–118]
and this was thought to be caused by their affinity to the carbonyl groups in the
lipid bilayer [275]. Not all types of aromatic residues are found in the aromatic belt;
phenylalanine has no particular preference for this region [117, 119]. It is still unclear
if the aromatic belt has to do with anchorage or with translocon recognition [109].
Here, TMHs with exclusively anchorage functions showed stronger preferences for the
W and Y in the aromatic belt region, otherwise known as the water-lipid interface
region than TMHs with function beyond anchorage. This is strong evidence that the
aromatic belt indeed assists with anchorage, and is less conserved where the TMH
must conform to other restraints beyond membrane anchorage. Furthermore, we see
that the tyrosine’s preference for the inside interface region also appears to be to do
with anchorage and this trend is somewhat true for tryptophan, too.
Finally, our findings corroborate earlier reports that many multi-pass TMHs are

2.4. DISCUSSION

89

much less hydrophobic than typical single-pass TMH and about 30% of them fail the
hydrophobicity requirements of ∆G TMH insertion prediction (“missing hydrophobicity”) [74, 116, 267, 268]. We also find that the leucine skew and the hydrophobic
asymmetry towards the cytosolic leaflet of the membrane is more pronounced in simple,
single-pass TMHs than in complex or multi-pass ones; thus, it appears to be another
anchoring feature. It was found previously that the hydrophobic profiles of TMHs of
multi-pass proteins share similar hydrophobicity profiles on average irrespective of the
number of TMHs and TMHs from single-pass proteins have been found to be typically more hydrophobic than TMHs from multi-pass proteins [44]. Sharpe et al. [86]
report an asymmetric hydrophobic length for single-pass TMHs. Our study reiterates
the hydrophobic asymmetry and attributes it mainly to the leucine distribution. The
leucine asymmetry might be linked to the different lipid composition of either leaflet
of biological membranes.
B) Additional features of
anchoring TM αhelices

A) TM αhelix features

Inner leaflet Outer leaflet

Membrane

Outside
(non-cytosolic)

Inside
(cytosolic)

Negative-outside
enrichment

Polar group
enrichment

Hydrophobic
stretch

Aromatic Belt
(Y,W)

Additional
charged residue
suppression
(positive and
negative)

Leucine inside
preference
(additional
hydrophobicity)

Aromatic Belt
inside bias (Y,
and W)
“Positive-Inside”
rule

Cysteine inside
preference

Negative-inside
suppression

Greater net
positive charge

Figure 2.10: Residue distributions of transmembrane anchors. A view showing additional residue distribution features that transmembrane helices with an anchorage
function display. a The more classic model of a TMH showing the “positive-inside” rule [113], the
hydrophobic core [73], the polar enrichment that flanks the hydrophobic stretch [109] and the aromatic
belt [117]. b Simple anchors may display additional features that conform to the membrane biophysical constraints: further suppression of charge in the hydrophobic core (Table 2.1), intra-membrane
leucine asymmetry that likely causes hydrophobic skew [86] (Table 2.4, Figure 2.6), a higher preference for cysteine on the inside flanking region (Figure 2.9K and L), a higher net “positive-inside”
charge (Figure 2.4), asymmetric skew of the hydrophobic belt favouring the inner leaflet interface
(Figure 2.9E, F, G, and H) and a negative-outside bias via suppression on the inside flanking region
or enrichment on the outside flanking region (Figure 2.9C and D, Tables 2.2 and 2.3)

In summary, three key features can be assigned to aiding TMH stability in the
membrane (Figure 2.10): (i) charge, (ii) the aromatic belt, and (iii) leucine leaflet
preference. What is most novel here is that each of these features are furthermore

90

CHAPTER 2. THE “NEGATIVE-OUTSIDE” RULE

distributed with preference for a particular side of the bilayer in the case of anchoring TMHs. These differences in inside-outside topology that are most present in anchoring TMHs further supports the notion that there are broad lipid compositional
differences between the inner and outer leaflets of the bilayers [86]. Furthermore, while
some TMHs conform and complement to the properties of the bilayer, other TMHs
with function beyond anchorage are less constrained to biophysically complement the
bilayer. For these TMHs, any advantage gained by adhering to the membrane restrictions is outweighed by more complicated protein dynamics, topological frustration and
protein functional requirements.
To conclude, the large fraction of functionally uncharacterised genomic sequences is
the great bottleneck in life sciences at this moment that hinders many biomedical and
biotechnological applications, some with tremendous societal need [241, 276]. Among
these uncharacterised genomic regions, there is ∼10000 protein-coding genes, especially
many membrane-embedded proteins. It is hoped that the NNI/NO-rule as well as the
other sequence properties of membrane anchoring TMHs described in this article will
add new insights for membrane protein function discovery, design and engineering.

2.5

Methods

2.5.1

Datasets

Databases.
All datasets used for analysis are listed in Table 2.1. Transmembrane protein sequences
and annotations were taken from TOPDB [257] and UniProt [75]. UniProt derived
datasets are the most comprehensive datasets built with (i) robust transmembrane
prediction methods providing the limit of today’s achievable accuracy with regard to
hydrophobic core localisation and (ii) subcellular location annotation that can be used
for orientation determination. However, they mostly rely on predicted transmembrane
regions. TOPDB has experimental verifications of the orientation from the literature
that are independent of prediction algorithms [257]. Unfortunately, this dataset is
much smaller with too few entries to have it divided with regard to taxonomy or
subcellular locations.

2.5. METHODS

91

UniProt database files were downloaded by querying the server for different taxonomic groups as well as different subcellular membrane locations; UniHuman (human
representative proteome), UniCress (Arabidopsis thaliana, otherwise known as mouse
eared cress, representative proteome), UniER (human endoplasmic reticulum representative proteome), UniPM (human plasma membrane representative proteome),
UniGolgi (human Golgi representative proteome). To enforce a level of quality control,
the queries were restricted to manually reviewed records and transmembrane proteins
with manually asserted TRANSMEM annotation [75]. Proteins were then sorted into
multi-pass and single-pass groups according to having more than one or exactly one
TRANSMEM region respectively. TRANSMEM regions are validated by either experimental evidence [75], or according to a robust transmembrane consensus of the
predictors TMHMM [67], Memsat [66], Phobius [71, 236] and the hydrophobic moment plot method of Eisenberg and co-workers [144]. TMHs and flanking regions were
oriented according to UniProt TOPO DOM annotation according to the keyword “cytoplasmic”. If a “cytoplasmic” TOPO DOM was found in the previous TOPO DOM
relative to the TRANSMEM region then the sequence remained the same. If “cytoplasmic” was found in the next TOPO DOM, relative to the TRANSMEM section
then the sequence was reversed. Proteins without the “cytoplasmic” keyword in their
TOPO DOM annotation were omitted from further analysis.
The TOPDB database [257] is a manually curated database composed of experimental records from the literature that allow determination of the protein topology.
Experiments include fusion proteins, posttranslational modifications, protease experiments, immunolocalization, chemical modifications as well as revertants, sequence
motifs with known mandatory membrane-embedded topologies, and tailoring mutants
(Table 2.7).
Length cut-offs for the TMH were set at 16 as the shortest length and 38 as the
longest.
To note, we are aware that proteome datasets are a moving target that have dramatically changed over the years and, probably, will continue to do so to some extent
in the future[83]. Yet, we think that currently available protein sequence sets are sufficiently good for the purpose as we search for statistical properties in the TMH context
only.

92

CHAPTER 2. THE “NEGATIVE-OUTSIDE” RULE

Table 2.7: The experimental evidences of TOPDB. The total number of experimental
evidences that contribute to ExpAll according to the TOPDB database (More information at
http://topdb.enzim.hu/?m=exptype&mid=14). “*” refers to the total number of a subsection being
larger than the total of the subcategories, likely due to lack of annotation where ambiguous literature
evidence is counted toward the total, but cannot be categorised further.
Experiment
Fusion

PostTransMod

Protease

Immunolocalisation

Chemical modification

Structure

Other

Bitopic (single-pass)

Polytopic (multi-pass)

PhoA

97

2332

PhoAS

0

90

LacZ

20

433

PhoALacZ

0

224

BlaM

162

570

BAD

0

2

PL

0

47

GFP

18

591

HIS

4

2

SplitUbiquitin

0

11

Suc2

0

96

Other

1

137

Total Fusion

316*

4600*

NGlyc

4634

1130

Cman

0

6

Phosphorylation

4

1

Ubiquitination

47

102

Total PostTransMod

4685

1239

Partial Proteolysis

51

264

Signal Peptidase

1

0

TID

13

15

Total Protease

64

279

Epitope Insertion

33

313

Endogen Epitope

8

41

Total Immunolocalisation

53*

451*

Cys

0

361

Lys

0

3

Quenching

0

2

Total Chemical Modification

0

368*

PDBTM TMDET

5968

41977

Revertants

0

14

SeqMotif

2

32

Tailoring

1

67

Total other

3

115*

2.5. METHODS

93

The following datasets are used throughout this work:
ExpAll
TOPDB contained 4190 manually annotated transmembrane proteins at the time of
download [257]. CD-HIT [277] identified 3857 representative sequences using sequence
clusters of >90% sequence identity. This choice of similarity threshold was chosen since
CD-HIT ultimately underlies the clustering behind UniRef. Unlike the other datasets,
which by definition contain reasonably typical TMHs, many of the transmembrane
segments annotated in TOPDB are extremely short or long and this would cause
severe unrealistic hydrophobic mismatches. Especially, the short segments could be
the result of miss-annotation, TMHs broken into pieces due to kinks or segments that
peripherally insert only into the interface of the membrane bilayer. To remove the
atypical lengths, cut-offs were set at 16 as the lower cut-off and 38 as the upper cut-off
after inspecting the length histogram. We found that, for the single-pass TMHs in
TOPDB, 1215 out of 1544 are within the length limits (78.7%). Among the 17141
multi-pass TMHs, we find 15563 within our global length limits (from 2205 TOPDB
records corresponding to 2281 UniProt entries). This removed 1578 very short TMHs
and none of the long TMHs. Our cut-off selection is very similar to the one by BaezaDelgado et al. [109].
To get an idea of the taxonomical breakdown in the ExpAll dataset, the UniProt
ID tags were extracted and mapped to UniProtKB. The combined dataset of multi-pass (single-pass) proteins was mapped to 1288 (1343) eukaryotic records, 404 (776)
of which were human records, 926 (191) bacterial records, 46 (5) archaea records, and
14 (22) viral records.
UniHuman
This is a set of mostly human TMH-containing proteins or their close mammalian
homologues.

UniProtKB contains 5187 human protein records that are manu-

ally annotated with TRANSMEM regions (query = “annotation:(type:transmem)
AND reviewed:yes AND organism:“Homo sapiens (Human) [9606]” AND proteome:up000005640”. To reduce sequence redundancy, these sequences were submitted
to UniRef90 [278]. To note, UniRef90 was chosen over UniRef50 to maintain a viable

94

CHAPTER 2. THE “NEGATIVE-OUTSIDE” RULE

size of datasets for statistical analysis of occurrence of negatively-charged residue,
which are very rare in the vicinity of TMHs. 5015 UniRef90 clusters represented the
5187 sequences. A list of sequences representing those clusters was submitted back
to UniProtKB resulting and 5014 representative entries were recovered. There is a
small issue in that the list of representatives from UniRef includes non-canonical isoforms, while the batch retrieve query of UniProtKB only supports complete entries, i.e.
canonical isoforms. This resulted in the loss of one record at this point is due to two
splice isoforms acting as representative identifiers. Of those 5014 records, 4714 were
records from human entries, 197 were from mice, 94 from rats, 5 from bovine, 2 from
chimps, 1 from Chinese hamsters, and 1 from pigs. Although the TMH length variations within the UniHuman dataset are much smaller than for ExpAll, we applied
the same length cut-offs for the sake of comparability. Out of the 1709 single-pass
cases, 1705 entered the final dataset. Of those, 1596 were from human records, 87
were from mouse, 19 were from rat, and 2 were from chimpanzee. Among the 12390
multi-pass TMHs, 12353 were included into UniHuman. The other, multi-pass record
identifiers were mapped to 1789 UniProtKB entries. 1660 of these were human entries,
63 from rat, 61 from mouse, 4 from bovine, and 1 from Chinese hamster. This clustered human dataset was then queried for subcellular locations to make the UniER,
UniGolgi, and UniPM datasets (detailed below).

UniER

The clustered UniHuman dataset was queried using UniProtKB for endoplasmic reticulum subcellular location (locations:(location:“Endoplasmic reticulum [SL-0095]” evidence:manual)). This returned 487 protein entries, 457 of which belonged to human,
24 to mouse and 6 to rat. 287 of these records contained sufficient annotation for
orientation determination. 132 were single-pass entries of which 120 records were from
humans, 11 from mouse, and 1 from rat. 155 were multi-pass entries containing 898
transmembrane helices. 144 were records from human, 8 were from mouse and 3 were
from rat.

2.5. METHODS

95

UniGolgi
The clustered human dataset was queried using UniProtKB for Golgi subcellular location (locations:(location:“Golgi apparatus [SL-0132]” evidence:manual)). This returned 323 protein entries, 301 of which belonged to human, 19 to mice, 2 to rat and
1 to pig. 269 of these records contained sufficient annotation for orientation determination. 206 were single-pass entries of which 195 records were from human, 9 from
mouse, and 1 from rat. 61 were multi-pass entries containing 383 transmembrane
regions. 54 were records from human, 6 were from mouse and 1 was from rat.

UniPM
The clustered human dataset was queried using UniProtKB for the cell membrane subcellular location (locations:(location:“Cell membrane [SL-0039]” evidence:manual)).
This returned 1036 protein entries, 948 of which belonged to humans, 62 to mice, and
26 to rats. 920 of these records contained sufficient annotation for orientation determination. 493 were single-pass entries of which 451 records were from human, 37 from
mouse, and 5 from rat. 427 were multi-pass entries containing 3079 transmembrane
regions. 394 were records from human, 17 were from mouse and 16 were from rat.

UniCress
For the mouse ear cress, a representative proteome dataset was acquired with the query
annotation:proteomes:(reference:yes) AND reviewed:yes AND organism:“Arabidopsis
thaliana (Mouse-ear cress) [3702]” AND proteome:up000006548. This returned 3174
records in UniProtKB. UniRef90 identified 3111 clusters. 3110 of the representative
sequences were mapped back to UniProtKB. Of those, 3090 were from Arabidopsis
thaliana, 2 from Hornwort, 1 from cucumber, 1 from tall dodder, 1 from soybean
(Glycine max), 2 from Indian wild rice, 2 from rice, 2 from garden pea, 1 from potato,
4 from spinach, 1 from Thermosynechococcus elongatus (thermophilic cyanobacteria),
1 from wheat, and 2 from maize. Of those there were 1146 with suitable TOPO DOM
annotation for topological orientation determination. 632 of those records were identified as single-pass, all of which were from Arabidopsis thaliana. 507 protein records
were from multi-pass records, which contained 3823 transmembrane helices. 506 of

96

CHAPTER 2. THE “NEGATIVE-OUTSIDE” RULE

those records were from Arabidopsis thaliana, whilst 1 was from Thermosynechococcus elongatus.
UniFungi
For the Fungi dataset, the query “annotation:(type:transmem) taxonomy:“Fungi
[4751]” AND reviewed:yes” was used. This returned 5628 records that were submitted to UniRef90. UniRef90 identified 4934 representative records, all of which were
successfully mapped back to UniProtKB. Of those, 2070 had suitable annotation for
orientation. 1990 records belonged to Ascomycota including 1243 Saccharomycetales.
73 were Basidomycota, and 6 were Apansporoblastina. 729 records contained a single TMH region, 702 of which belonged to Ascomycota, 26 to Basidomycota and one
to Encephalitozoon cuniculi, a Microsporidium parasite. 8698 helices were contained
in 1338 records of multi-pass proteins. Of these records 1285 were Ascomycota, 47
were Basidomycota, and 5 were Apansporoblastina. One TMH from UniFungi was
discounted from P32897 due to an unknown position.
UniEcoli
This dataset was generated by querying UniProt with “reviewed:yes AND organism:”Escherichia coli (strain K12)[83333]”” which returned 941 hits. The hits were
submitted to UniRef90, which returned 935 clusters. The representative IDs were
then resubmitted to UniProtKB, all of which returned successfully. 934 were from
Bacteria, whilst one were from lambdalike viruses. Of the bacterial records, 862 were
from various Escherichia species of which 565 were from E. coli strain K12, 28 were
from Salmonella choleraesuis, 25 were from Shigella and the rest all also fell under
Gammaproteobacteria class. This dataset contains 54 single-pass proteins and 3888
helices from 529 multi-pass proteins with sufficient annotation for topological determination.
UniBacilli
The Bacilli dataset was constructed by querying UniProt for “reviewed:yes AND taxonomy:”Bacilli””. This returned 5044 records, which were submitted to UniRef90.
2,591 clusters were found in UniRef from these records. The representative IDs were

2.5. METHODS

97

successfully resubmitted to UniProtKB. 2031 of these were of the genus Bacillales
whilst 560 were also of the genus Lactobacillales. This dataset contains 124 singlepass proteins and 822 helices from 140 multi-pass proteins.

UniArch
The Archaea dataset was constructed by querying UniProt for “reviewed:yes AND
taxonomy:”Archaea [2157]””. This returned 1,152 records, which were submitted to
UniRef90. 1,054 clusters were found in UniRef from these records. The representative IDs were successfully resubmitted to UniProtKB. 946 records belonged to the
Euyarchaeota, 101 to Thermoprotei, 4 to Thaumarchaeota, and 3 to Korarchaeum
cryptofilum. This dataset contains 48 single-pass proteins and 59 multi-pass proteins
containing 327 helices from 59 proteins.

2.5.2

On the determination of flanking regions for transmembrane helices and the transmembrane helix alignment

The determination of the boundary point at the sequence between the TMH in a
membrane and the sequence immersed in the cytoplasm, extracellular space, vesicular
lumen, etc. is not that trivial as it initially appears. There is a lot of dynamics in
the TMH positioning and the actual boundary point will be represented by various
residues at different time points. Whilst the TMH core region detection from a sequence is trivial with modern software, the exact determination of TMH boundaries
remains difficult since it is unclear exactly how far in or out of the membrane a given
helix extends [232]. Previous studies have dealt with this issue in various ways [86,
109, 110, 279].
Here in this work, we explore two boundary definitions. First, we assign TMH
boundary locations as described in the respective databases. These flanks are the ones
that are reported in our TMH data files that are available at the WWW-site associated
with this paper. We studied flank lengths of ±5, ±10, and ±20 residues preceding
and following the inside and outside TMH boundaries. In these cases, the flanks are
aligned relative to the residue closest to the TMH.
In cases where the loops before and after the TMH are shorter than the predefined

98

CHAPTER 2. THE “NEGATIVE-OUTSIDE” RULE

flank lengths, further precautions are necessary. In the multi-pass datasets particularly
(Figure 2.11 & Figure 2.4), the flanks overlap with other membrane region flanks. We
explore several variants. On the one hand, we work with data files where the flank
residue stretches are equally truncated so that no overlap occurs. If the loop length was
uneven, the central odd residue was not included into any flank. We find surprisingly,
that a large number of TMH has no or just a super-short flank, a circumstance that
should disturb any statistical analysis due to the absence of objects. Therefore, we also
work with alternative datasets (i) with flanks overlapping between consecutive TMH
(e.g., in Table 2.3B; yet, it leads to some residues being counted more than one time)
as well as (ii) with subsets of the data where the flanks at both sides have a defined
minimal length (50% or 100% of the required flanks; unfortunately, some of them
become too small for analysis).
Outside

Inside

ExpAll
TMH

Outside

Flank overlaps
not allowed

UniHuman
TMH

Multipass

Flank overlaps
allowed

Flank overlaps
not allowed

Flank overlaps
allowed

Singlepass

Inside

Figure 2.11: The lengths of flanks and transmembrane helices in multi-pass and single-pass proteins in the UniHuman and ExpAll dataset.On the horizontal axis are the lengths
of the TM segment regions in residues. On the vertical axis are the percentages of the population.
There are three regions: the inside flank, the TMH and the outside flank. These regions are acquired
according to the TMH boundary of the respective database. Where no overlap is permitted, if the
flank encroaches the flank of another TMH, the flank length becomes half the number of residues in
the loop region between the two features. Where they are allowed to overlap, flanking residues may
include other flanks, or indeed other TMHs.

The problem of flanks overlapping does affect also some single-pass and multi-pass TMH proteins with INTRAMEM regions as described in some UniProt entries.
We do not include INTRAMEM regions in the datasets as TMHs but, sometimes,

2.5. METHODS

99

the flanking regions of TMHs were truncated to avoid overlap with INTRAMEM
flanking regions (Table 2.8). The identifiers affected for single-pass TMH proteins
are Q01628, P13164, Q01629, Q5JRA8, A2ANU3 (UniHuman), P13164, Q01629,
A2ANU3 (UniPM) and Q5JRA8 (UniER).
Table 2.8: Records with INTRAMEM and TRANSMEM flanking region overlap. The
total number of TMHs from UniProt datasets with flanking region overlap between INTRAMEM
and TRANSMEM regions. The number of multi-pass records that the TMHs belong to are shown in
brackets.

Flank length
5

Dataset
singlepass

10
multipass

singlepass

20

multipass

singlepass

multipass

UniHuman

0

96 (80)

1

151 (90)

5

204 (96)

UniER

0

6 (6)

1

13 (8)

1

16 (8)

UniGolgi

0

1 (1)

0

2 (2)

0

4 (2)

UniPM

0

57 (46)

0

93 (51)

3

113 (52)

UniCress

0

17 (17)

0

24 (18)

0

46 (18)

UniFungi

0

0

0

0

0

0

UniBacilli

0

11 (3)

0

12 (3)

0

13 (3)

UniEcoli

0

22 (8)

0

25 (9)

0

31 (9)

UniArch

0

0

0

8 (8)

0

17 (9)

The second form of boundary point definition for flank determination was achieved
with gaplessly aligning all TMHs relative to their central residue at the position equal
to half the length of the TMHs at either side. Though there is some length variation
among TMHs, most of them are centred around a length of 20-22 residues. In this case,
flanks are the sequence extensions beyond the standardised-length 21-residues TMHs.
We define the inside flanking segments as the positions -20 to -10 and the outside
flanking regions to be +10 to +20 from the central TMH residue (with the label
“0”). Instead of emphasising some artificially selected boundary residue, this definition
allows the average TMH boundary transition to become apparent.

2.5.3

Separating simple and complex single-pass helices

Single-pass helices from ExpAll and UniHuman datasets helices were split into two
groups: simple and complex following a previously described classification [44, 45]

100

CHAPTER 2. THE “NEGATIVE-OUTSIDE” RULE

to roughly distinguish simple hydrophobic anchors and TMHs with additional structural/functional roles.

Simple and complex helices were determined using TM-

SOC [45]. The complexity class is determined by calculating the hydrophobicity
and sequence entropy. The resulting coordinates cluster with anchors being more
hydrophobic and less complex whilst more complex and more polar TMHs are associated with non-anchorage functions. In UniHuman there were 889 simple helices and
570 complex TMHs. In ExpAll there were 769 simple helices and 570 complex helices.

2.5.4

Distribution normalisation

In this work, we have used normalisation techniques described in previous investigations as well as new approaches designed to more sensitively identify biases of
rare residues. Baeza-Delgado and co-workers used LogOdds normalisation columnwise in TMH alignments. Critically, this is based on their definition of probability,
which takes into account the total number of amino acids in the dataset as a denominator [109]. Since aliphatic residues such as leucine and other highly abundant
slightly polar residues dominate the denominator, the distribution of the rare acidic
residues will be easily lost in the “background noise” of those highly abundant residues.
Pogozheva and co-workers used two approaches, (i) the total accessible surface area
(ASAtotal) and (ii) total number of charged residues (Ntotal ) as a denominator in their
distribution normalisation [110].
In this work, two methods for measuring residue occurrence in the TMH and its
flanks were used. Similarly to previous work, we compute the occurrence of an amino
acid type at a certain sequence position in a set of aligned sequences TMHs and their
flanks. Following [86], the absolute relative occurrence of this amino acid type at the
sequence position is then given by Equation 2.1 as:

pi,r =

ai,r
max(ar )

(2.1)

r

Here, the denominator is the maximal number of all residues in any alignment
column (i.e., the number of sequences in the alignment) and, to emphasise, this will
make mostly dependent on the most abundant residue types. This type of normalisation reveals the most preferred residue types at given sequence positions.

2.5. METHODS

101

Our second normalisation method is independent of the abundance of any amino
acid types other than the studied one; it answers the question: “If there is a residue of
type in the TMH-containing segment, where would it most likely be?” This relative
occurrence calculated in Equation 2.2 as:

qi,r =

100 · ai,r
ai

(2.2)

The value ai is the total abundance of residues of just amino acid type i in a
given alignment of TMH-containing segments (i.e., in the TMH together with its two
adjoining flanks summed over all cases of TMHs in the given dataset). Peaks in qi,r
as function of r reveal the preferred positions of residues of type i. The difference in
qi,r and pi,r normalisation is visualised in Figure 2.12.

2.5.5

Hydrophobicity calculations

Hydrophobicity profiles were calculated using the Kyte & Doolittle hydrophobicity
scale [73] and validated with the Eisenberg scale [144], the Hessa biological scale [116],
and the White and Wimley whole residue scale [139](Figure 2.7). The hydrophobicity
profile uses un-weighted windowing of the residue hydrophobicity scores from end to
end of the TMS slice. Three residues were used as full window lengths and partial
windows were permitted.

2.5.6

Normalised net charge calculations

Charge was calculated at each position by scanning through each position of the transmembrane helices and flanking regions and subtracting one from the position if an
acidic residue (D or E) was present, or adding one if a positively-charged residue (K or
R) was present. The accumulative net-charge was then divided by the total number
of transmembrane helices that were used in calculating the accumulative net-charge.
Thus, the charge distribution is calculated by:

cr =

(aK,r + aR,r ) − (aD,r + aE,r )
N

(2.3)

102

CHAPTER 2. THE “NEGATIVE-OUTSIDE” RULE
A)

B)
Inner Inner
flank leaflet
-15

Fraction of
Maximum

-10

Outer Outer
leaflet flank
0

0

0.5

pi,r =

ai,r
max ( ar )
r

10

1

15

Inner Inner
flank leaflet
-15

-10

Outer Outer
leaflet flank
0

Relative
Percentage 0

9.0

qi,r =

10

15

17.9

100 ⋅ ai,r
ai

Figure 2.12: Relative percentage heatmaps from the predictive datasets calculated by
fractions of the absolute maximum and by the relative percentage of a given amino acid
type.The residue position aligned to the centre of the TMH is on the horizontal axis, and the residue
type is on the vertical axis. Amino acid types are listed in order of decreasing hydrophobicity according
to the Kyte and Doolittle scale [73]. The flank lengths in the TMH segments were restricted to up
to ±5 residues. The scales for each heatmap are shown beneath the respective subfigure. All TMHs
and flank lengths are from the UniHuman dataset. (A) The heatmap has been coloured according
to a scale that uses column-wise normalisations used in previous studies [86]. See Equation 2.1. As
an illustrative example, we show how the value for E at position ±12 is obtained. There are in total
91/22 Es at these positions in 1705 sequences; thus, the represented value is 0.013 at –12 and 0.053 at
12. Note that L is clearly a hotspot as well as trends for other hydrophobic residues, I and V, as is to
be expected. A positive inside effect can also be seen. (B) The heatmap has been coloured according
to the relative percentage of each amino acid type (Equation 2.2). Here, 91/22 Es at position ±12 are
compared with 615 Es seen within the flanks and the TMH section itself amongst all sequences in the
alignment. So, the expectation of an E at position ±12 if there is any E in the TMH + flanks region
at all is 0.036 at –12 and 0.148 at position 12. With this type of normalisation, not surprisingly, we
see the positive-inside rule is hotter than in subfigure A. There are also hotspots in the flanks for the
negatively-charged residues on the outside flank. The leucine hotspot is no longer very pronounced,
as the leucines are quite evenly spread over many positions.

2.5.7

Statistics

The inside/outside bias of negative residues was quantified by computing the independent K-W and the 2-sample t-test statistical method from the Python scipy stat
package v0.15 python package [280]. This test answers the question whether two means

2.5. METHODS

103

are actually different in the statistical sense. The p-values in Table 2.2 and Table 2.3
are calculated by comparing two lists of the number of the given residues in either the
inside and outside flank with each entry belonging to an individual TMH. The p-values
in Table 2.1 compare a list of the total number of negatively-charged residues in TMHs
from either single-pass or multi-pass TMHs. For the leucine residues, each TMH region was divided into two sections, representing the inner and outer leaflets (Table 2.4)
and the lecine residues in the inner and outer leaflets were compared similarly to the
negatively-charged residues in the inside and outside flanks in Table 2.2 and Table 2.3.
For the hydrophobicity plot, 3 window values of hydrophobicity were taken for
each TMH at each position. The statistical analyses were separately performed for
single-pass and multi-pass transmembrane proteins. At each position, the two groups
were compared using the K-W test.
The zero hypothesis of homogeneity of two distributions was examined with the KS, the K-W and the χ2 statistical tests. To note, the K-S test scrutinises for significant
maximal absolute differences between distribution curves; the K-W test is after skews
between distributions and the χ2 statistical test checks the average difference between
distributions. As the statistical significance value (“p-value”) is a strong function
of N, the total amount of data used in the statistical test (for example, Table 2.5
and Table 2.6 N would be the total number of TMHs from both datasets in the
comparisons), we rely on the (absolute) Bahadur slope (B) as a measure of distance
between two distributions [259, 260]:

B=

ln(p value)
N

(2.4)

The larger the absolute Bahadur slope, the greater the difference between the two
distributions.

Chapter 3
Collation and analysis of
tail-anchored protein
transmembrane helices reveals
subcellular variation in flanking
charged residue distribution and
the transmembrane helix core
hydrophobicity
3.1

Abstract

Tail-anchored proteins are a functionally diverse group of post-translationally inserted
membrane proteins that have a cytosolic N-terminal, a transmembrane α helix, and
a short C-terminal tail. The transmembrane helix and flanking regions of the tail-anchored protein have been shown to contain sufficient information for subcellular
targeting. Here, we built datasets based on sequence definitions of plausible tailanchored proteins using the UniProt database. We show that any statistical differences
between the hydrophobicity of the transmembrane helix between tail-anchored proteins
104

3.2. INTRODUCTION

105

belonging to mammalian, plant, or yeast organisms are, if at all present, small. Yet,
the hydrophobicity of transmembrane helices from different subcellular locations is, on
average, different between the mitochondria and other membranes along the secretory
pathway. Notably, in the case of mitochondria, this appears to be a difference in the
hydrophobic residue preference of alanine instead of leucine or isoleucine and does
not seem to be an increase of intramembrane polar residues. In the endoplasmic
reticulum, Golgi, and plasma membrane there are positively charged residues inside
the cytoplasm and negatively charged residues outside. Yet we identify a charge skew
reversal of the “positive-inside” and “negative-outside” rules in the mitochondrial tailanchored proteins to negative-inside and positive-outside. It is unclear to what extent
these differences are the result of membrane environment adaptations or biological
features useful for the biogenesis and accurate localisation of the protein. Furthermore,
structural homology modelling of the spontaneously inserting transmembrane helices of
PTP1b and cytochrome b5 revealed that tail-anchored proteins may gain their ability
to integrate into the membrane unaided due to a strip of conserved relatively polar
residues and strong flanking charge that would allow effective membrane coupling and
anchoring.

3.2

Introduction

TA proteins are defined by their single carboxy-terminal TMH with a cytosolic facing
amino-terminus and are a topologically distinct class of intracellular proteins. The
integration of TA proteins into the membrane is post-translational rather than cotranslational; the ribosome is not in complex with the membrane bound insertion
machinery. TA proteins are involved in a range of key cellular functions such as
translocation (Sec61β and Sec61γ) [281] and apoptosis (Bcl-2 apoptotic protein family) [282]. Additionally, within the TA class of proteins are a set of vesicle fusion
proteins called Soluble N-ethylmaleimide-sensitive factor attachment protein receptor
(SNARE) proteins [283], which contain typically hydrophobic TMHs [284]. The idea
that SNARE proteins are modular and capable of spontaneous insertion has significant
implications for biomedical application in liposome-based drug delivery. Furthermore,
this could aid future research for testing complex biological molecular networks [285,

106

CHAPTER 3. COLLATION AND ANALYSIS OF TAIL-ANCHORED...

286].
The TA protein’s TMH is both the anchor and the targeting factor for the ER [287].
Furthermore, the hydrophobicity of the TMH appears to be a determining factor in
the delivery pathway that TA proteins use for insertion [288, 289], for which there
is evidence demonstrating that there are several mechanisms of protein-membrane
integration [289, 290](Figure 3.1).
TA proteins have several pathways for biogenesis in the ER membrane. TA proteins were originally thought to be inserted into the membrane via different machinery
than the co-translational machinery, but unexpectedly SRP was found to be a factor
for post-translational targeting confirmed by both cross-linking studies [291] and an in
vitro pull-down experiment [292]. It was shown that SRP could deliver the TA protein
to the membrane-bound SR in association with a highly conserved Sec translocon.
Further cross-linking experiments confirmed Sec61 is also involved during TA protein
membrane insertion [293]. Previous studies had shown the Sec61 translocon is not necessary for TA protein membrane integration by biochemical reconstitution experiments
[287] and conditional mutants in yeast [294, 295]. So whilst it is hard to determine if
Sec61 can be part of the post-translational pathway, we can conclude that it is certainly not essential, indeed almost no observable impact on biogenesis is had when it
is removed [287, 294, 295]. Nevertheless, this suggests the possibility of at least one
insertion mechanism that is related to the co-translational method of insertion. Most
likely is that SRP binds to the TA protein after it is released from the ribosome and
chaperones it until it is close enough to the established post translational machinery
[296] (Figure 3.1iii).
A second redundant system is also known to be involved in TA protein biogenesis
and is referred to as the TRC40 (also known as Asna1) pathway in mammals (Figure
3.1i). A conserved homologue was found in Saccharomyces cerevisiae, Get3 [297],
named after the action it has in the GET pathway. Unlike co-translational insertion,
the post-translational chaperone proteins do not couple with the ribosome, so the TA
protein must be exposed to the cytosolic environment for at least some time [78].
At some point after the TA protein emerges from the ribosomal exit tunnel, the TA
protein TMH associates with Sgt2 (SGTA in mammals). An in vitro assay revealed
that Sgt2 associates with Get5 [298] (UBL4A) as part of a dimerised Get4 (TRC35)

ADP

(WRB+CAML)
Get1+Get2

ATP

(TRC40)
Get3

(TRC35+Ubl4A)
Get4+Get5

(Sgta)
Sgt2

Snd2+Snd3

Snd1

ii
iii

?

SRP

iv

N terminal

TA Protein

TMH

C terminal

v

Calmodulin

C-terminal Tail
charge

EMC

vi

Mitochondria

ATP

?
Unassisted

Hsp40
(Hsc70)

Peroxisome

Pex3

Pex 19

Figure 3.1: An overview of the biogenesis of tail-anchored proteins.
(i) The intensively studied Get (yeast) or TRC40 (mammalian) pathway can target either for membrane integration or degradation of the TA protein. Text
in brackets indicates mammalian proteins. (ii) In yeast, a novel mechanism was identified in which Snd1 binds to the folded TA protein and delivers it to the
membrane-bound Snd2 and Snd3 complex. (iii) SRP of the co-translational insertion mechanism has been shown to be able to integrate TA proteins. Sec61s role
in this is disputed, and it is likely other proteins are involved in membrane partitioning instead. (iv) A recently identified insertase, the EMC, can integrate TA
proteins with relatively polar TMHs into the ER. (v) A handful of TA proteins with relatively polar TMH regions have been observed spontaneously integrating
into the membrane using Hsp40 and Hsc70 as chaperones. This system may be employed for mitochondrial localisation. (vi) Peroxisomal proteins with an
abundance of charge in the tail region are chaperoned by Pex19 into association with the membrane-bound Pex3.

ER

(Bag6)
degradation

i

TMH
hydrophobicity

Ribosome

3.2. INTRODUCTION
107

108

CHAPTER 3. COLLATION AND ANALYSIS OF TAIL-ANCHORED...

and Get5 (Ubl4A) complex (two copies of each)[299–302].
In mammals, at this point SGTA either associates with preferential TRC40 which
targets the TA protein for ER membrane biogenesis or if there are excess TA proteins
SGTA also associates with Bag6 which targets the TA protein for degradation [303].
This “race” between Bag6 and TRC40 ensures a level of quality control within the
system.
Get3 associates first with the Sgt2-Get4-Get5 complex via an interaction with the
N-terminal of Get4 [298]. A dimerised ADP-bound Get3 [304–308] associates with
and shields the C-terminal region of the TA protein [297, 309, 310]. This shielding
may be especially important since Get3 is involved in reducing the misfolding of any
nascent TA proteins, the TMH of which would be unviable in the cytosol due to their
hydrophobicity [311]. Fluorescence studies revealed that tagged Get3 appears at both
the cytosol and the ER membrane so apparently shuttles the TA protein between the
transmembrane complex of Get1 and Get 2 (WRB and CAML in mammalian cells),
that contains cytosolic domains that receive the Get3, Get4, Get5, Sgt2 complex [312,
313]. Yet it is an interesting note that a single molecule fluorescence study revealed
that the minimum machinery required for TA protein insertion from this system is a
Get1 and Get2 heterodimer [313]. The Get pathway exclusively delivers TA proteins
to the ER membrane, and indeed has been recently shown to be responsible for some
of the mislocalisation of mitochondrial TA proteins during overexpression or signal
masking to the ER [314]. The significance of this is that the Get machinery can
recognise and tolerate integration of non-ER proteins. Yet there is also evidence that
the deep groove of Get3 [315, 316] predisposes it to only effectively integrating the
more hydrophobic TMHs of TA proteins [298, 317]. As an example, increasing the
hydrophobicity of the TMH in squalene synthase (an ER TA TMP involved in sterol
synthesis) in a TRC40 inhibited system reduced the biogenesis of the protein, where
the wild-type was unaffected by TRC40 inhibition [25]. Around a half of TA proteins
are estimated to not use the TRC40 pathway [25].
Redundancy of the GET/TRC40 pathway and SRP pathway may be explained
in part by a novel SRP and Get independent pathway. This pathway utilises the
Snd pathway and was discovered in yeast [318] (Figure 3.1ii). Snd1 binds to the TA

3.2. INTRODUCTION

109

protein after it exits the ribosome and delivers it to the Snd2 and Snd3 membrane-bound complex which integrates the TA protein into the membrane. So far only the
human homologue of Snd2 has been identified (hSnd2) with relatively low sequence
identity [319]. However, it is suspected that functional mammalian homologues exist
for Snd1 and Snd3 also, albeit with low sequence similarity.
Even more recently identified as a TA protein insertase was the EMC [25] (Figure
3.1iv). In the interaction study, squalene synthase did not effectively crosslink with any
TRC pathway machinery. Calmodulin (a eukaryotic highly conserved calcium-sensing
globular protein with a flexible symmetrical structure that performs a variety of functions) sufficiently prevented aggregation and acted as a chaperone in ER microsomes
in a chaperone free E. coli translation system with purified translation factors. In
this system, calmodulin acted similarly to the mammalian chaperone protein SGTA,
and although in native cytosol calmodulin was preferred, SGTA could also be used by
the protein. By contrast VAMP2 (a TA protein involved in vesicle binding), which is
known to interact with TRC40 as a chaperone [303], was unable to insert in this system. Abolition by knock-outs of the EMC components greatly reduced the insertion
potential of squalene synthase, along with five mutants thereof, and six other TRC40
independent proteins in ER membranes from semi-permeabilised cultured cells, but
not that of VAMP2. Insertion of Sec61β (a post-translationally inserted protein ) was
partially dependent on both systems, perhaps indicating the midway point between
the two [25] (Figure 3.1).
Some TA proteins, for example BCL-2 involved in apoptosis, are targeted to the
MOM instead of the ER, however the mitochondrial mechanism of membrane integration works independently from the TOM [320, 321]. In the absence of the mitochondrial inner and outer translocons and their associated machinery, insertion machinery
and GET/TRC machinery, heat shock proteins Hsp40 and Hsc70 chaperones along
with ATP are also sufficient for enough biogenesis of TA proteins for viable cell growth
[288, 289, 320–324] (Figure 3.1v). Synaptobrevin, one of the first identified SNARE
proteins, is capable of spontaneous insertion if the naturally occurring TA domain is
replaced by the TM that is known to be able to spontaneously insert into the ER.
One can, therefore, conclude that spontaneous insertion is a property of the TMH, not
of the rest of the protein [286]. Molecular dynamics simulations showed that direct

110

CHAPTER 3. COLLATION AND ANALYSIS OF TAIL-ANCHORED...

insertion TMHs thermodynamically mimics the energies of TMHs integrated by the
translocon [325] so in theory, no integration machinery is strictly necessary if the TMH
can “correctly” interact with the membrane interface. Furthermore, it was revealed
that scrambling the TMH sequence, but maintaining hydrophobicity, reduced the insertion potential of spontaneously inserting TMHs [138]. This phenomenon cannot,
therefore, be explained entirely by the marginal hydrophobicity of the TMH.
The few peroxisomal TA proteins first associate with Pex19 which forms a complex
with the membrane-anchored Pex3 protein from which the TA protein is integrated
into the membrane [326–328](Figure 3.1vi).
Given a “choice”, it is speculated that hydrophobicity determines the integration
pathway [288, 289, 329]. Altering the hydrophobicity, at least in the case of the
spontaneously inserting PTP1b, also determines the localisation of the TA protein to
either the mitochondrial membrane or the ER membrane, or rather a more hydrophobic
TA protein TMH is less likely to localise to the mitochondrial membrane [330]. Broader
analysis has shown that hydrophobicity [139] stratified by TM tendency score [234] can
distinguish between the ER and mitochondrial localised TA proteins [78]. However,
the diversity and known biosynthetic redundancy of these proteins may mean that no
single factor applied en masse is able to distinguish the TMH recognition factors and
investigation into this area is becoming increasingly complex [78].
By regenerating a list of likely TMHs [284] and using a manually curated list of TA
proteins [75], this investigation aims to find relationships between biochemical factors
and a disposition to a certain insertion mechanism and terminal localisations. Here,
we also present evidence for a conserved polar strip along the spontaneously inserting
TA protein TMHs, which may be the key to the initial interaction of these TMHs with
the membrane interface.

3.3
3.3.1

Methods
Building a list of tail-anchors

Steps carried out by Kalbfleisch et al. (Traffic 8: 16871694) to generate a list of all
TA proteins in the human proteome [284] were recreated using up to date tools and
applied to other model representative species. Whilst their study focused on the human

3.3. METHODS

111

proteome, here we take into account the entire TrEMBL and SwissProt database and
then stratify the datasets by the organism at the end of the pipeline. An overview of
the generation of these datasets is shown in Figure 3.2.

SwissProt and
TrEMBL

NCBI

SwissProt

1 TMH

Single
transmembrane
annotation

25 residues or less
between TMH and C
terminal residue

Non-splice isoform

Single transmembrane
annotation

No signal peptide

No signal peptide

Signal peptide and
mitochondrial signals
filtered out

N-terminal is
cytoplasmic

Biogrid chaperone
interactors
Type IV membrane
protein query

UniProt
remapping

25 residues or less
between TMH and
C terminal residue

25 residues or less
between TMH and
C terminal residue
Kalbfleisch et al., 2007

UniProt
remapping
“Redundant SwissProt
automatically filtered list”

CD-Hit

“S1”

Tail-anchored
chaperone interactors

CD-Hit

“Non-redundant
SwissProt automatically
filtered list”

Mouse and human
(mammal), plants,
yeast

“Redundant Manually
Curated UniProt list”

ER, Golgi, PM,
Miochondria

“Non-redundant
UniProt manually
curated list”

Mouse and human
(mammal), plants,
yeast

ER, Golgi, PM,
Miochondria

Figure 3.2: The sources, methods, and filters applied to the sequences in the tail-anchored protein datasets. From top to bottom are the sources of the sequences and the filters
and methods applied to each of the datasets of sequences. The database symbol is used to denote
when the dataset was used to capture results and is available as supplementary material. For the
dataset size and more information, see the methods section text.

112

CHAPTER 3. COLLATION AND ANALYSIS OF TAIL-ANCHORED...

SwissProt tail anchored dataset according to filters
There were 557012 protein records downloaded from SwissProt via UniProt [75] (downloaded 24–04–2018). 106149 TMHs (TRANSMEM annotation) were found between 76953
records (annotation:(type:transmem)ANDreviewed:no). This keyword is contained
in a record according to either experimental evidence [75] or a conservative metaanalysis of TMH prediction using TMHMM [67], Memsat [66], Phobius [71, 236] and
the hydrophobic moment plot method of Eisenberg and co-workers [144]. 11141 of
those records had only a single TMH. 11110 of those TMHs were within the length
thresholds of 16 to 30 residues (None of those had the annotation for splice isoforms
according to NON_TER annotation). 5548 of those had had no SP annotation (SIGNAL).
4332 of those had annotation (based on TOPO_DOM annotation) that the N terminal was
cytoplasmic. 615 of those had the TMH within 25 residues of the C-terminal, the same
threshold used by Kalbfleisch et al., [284]. Running CD-Hit 4.5.3 on the WebMGA
web-server [277, 331] at 90% identical sequence at 90% coverage thresholds resulted
in 443 representative proteins. This threshold was chosen as a compromise between
avoiding over-representation of a certain protein and maintaining a viable sample size.
From this representative list, 46 were Archaeal, 66 were bacterial, and 320 were
eukaryotic and 11 came from dsDNA viruses. When counting proteomes with greater
than 20 records, 49 belonged to the A. thaliana proteome, 48 to Mus musculus
(Mouse), 46 to the Homo sapiens proteome, 24 to S.cerevisiae.
65 were annotated under the mitochondrion location (query locations:
(location:"Mitochondrion[SL-0173]")), 157 in the PM (query locations:
(location:"Cellmembrane[SL-0039]"),

82 in the Golgi (query locations:

(location:"Golgiapparatus[SL-0132]")), and 98 in the ER (query locations:
(location:"Endoplasmicreticulum[SL-0095]"). Only 16 records were found for
the peroxisome (query locations:(location:"Endoplasmicreticulum[SL-0204]")
which is not high enough a sample size for accurate statistical analysis.
TrEMBL tail anchored dataset according to filters
111425234 records were stored in the TrEMBL database at the time of download (downloaded 25–04–2018). 22107826 of those contained TRANSMEM annotation
(annotation:(type:transmem)ANDreviewed:no). 18053 of these were single-pass

3.3. METHODS

113

proteins. All of these were within the length restrictions of between 16 and 35 residues
for the TMH region. 17973 of those did not contain a signal sequence when looking
for SIGNAL annotation. 5157 of those contained a cytoplasmically located N terminal according to TOPO_DOM annotation. 155 records had a TMH within 25 residues
of the C-terminal residue. When considering which species these records come from,
no more than 1 record belonged to any given species. To avoid representing a well
annotated SwissProt record that includes species annotation by a poorly annotated
TrEMBL record without species annotation, these TrEMBL records were omitted from
the sequence redundancy protocol and further analysis.

UniProt curated list
A query for locations:(location:"Single\--passtypeIVmembraneprotein[SL-9908]
") was used in UniProt which returned 2633 UniProtKB IDs; 463 SwissProt results
and 2170 TrEMBL results. Type IV anchors are sometimes split into two topological
groups; A (A cytosolic facing N terminal domain) and B (The N terminal is targeted
to the lumen), however, the UniProt nomenclature is strictly N terminal being
cytosolic.

This manually created list contained some TA proteins that did not

exactly fit the generally accepted definition of a TA protein and were excluded from
further analysis.

These could be examples of misannotation in the databases or

exceptional TA proteins that behave as post-translationally inserted TA proteins
despite not matching the exact criteria. 101 exceeded the TA length restrictions
of 25 residues between the TMH and the terminal residue. 8 contained annotation
for SIGNAL, indicating an SP, inconsistent with the TA protein definition. 20 were
multipass proteins. A full list of which records exceeded these limits and by how
much is included in the supplementary files. Running these records through CD-HIT
at 90% redundancy yielded 956 clusters; 269 SwissProt records and 687 TrEMBL
records [277, 331]. No further filters were applied to this list. Proteomes represented
by more than 20 records include A. thaliana (53 records), Human (30), Mus musculus
(30), and S. cerevisiae (27).
426 were annotated under the mitochondrion location (query locations:
(location:"Mitochondrion[SL-0173]")) 47 from SwissProt and 379 automatically assigned in TrEMBL. 397 in the ER (query locations:(location:

114

CHAPTER 3. COLLATION AND ANALYSIS OF TAIL-ANCHORED...

"Endoplasmicreticulum[SL-0095]")), 88 from SwissProt and 308 automatically annotated in TrEMBL. 1 TrEMBL record (UniProt ID A0A1E5RT24) in the ER set
contained an “X” residue in the C-terminal flank and was omitted from the analyses. Two subcellular location datasets had no automatically ascribed records and only
contained manually annotated SwissProt records; 31 in the PM (query locations:
(location:"Cellmembrane[SL-0039]"), and 83 in the Golgi (query locations:
(location:"Golgiapparatus[SL-0132]")). There were only 8 TA proteins located
in the peroxisome (locations:(location:"Golgiapparatus[SL-0204]")), making
them an unsuitable dataset for statistical analysis.
Remapping the previous dataset
189 of the 411 proteins from the previous Kalbfleisch et al., 2007 study [284] were
successfully mapped to 222 UniProtKB IDs using the UniProt mapping tools with the
RefSeq Protein to UniProtKB option [75].
Tail anchor protein chaperone interactors
As discussed in the introduction text, there is evidence surrounding the TMH biochemical factors that determine which chaperone will interact with a given TA protein. To
gain a more quantitative understanding of the relationship between the TA protein
TMH and the potential chaperones, it would be ideal to have a large TA dataset
stratified by known chaperone interactions.
Known interactor lists from BioGrid from the chaperones were checked against the
SwissProt automatically filtered and the UniProt curated TA protein datasets. There
were 91 interaction pairs for Hsp40 (Biogrid ID 119699) which mapped to 206 UniProt
records. Hsc70 (Biogrid ID 109544) had 534 interaction pairs 61 of which were mapped
to 91 UniProt records. Hsp40 and Hsc70 both returned 0 record hits after filtering the
UniProt records IDs through the UniProt manually curated list and the automatically
generated SwissProt list both before redundancy removal.
Snd1 (Biogrid ID 32240) had 237 interaction pairs which mapped to 239 records.
Snd1 returned 15 hits when filtering it through the TA datasets.
Sgt2 (BioGrid ID 34410) had 260 BioGrid interactor ids which mapped to 264
UniProt records. SGTA (BioGrid ID 112347) had 155 interactor ids from BioGrid,

3.3. METHODS

115

153 of which were mapped to 274 records. Sgt2 and SGTA returned 14 and 5 hits
respectively.
50 BioGrid ids from TRC40 (Biogrid ID 106931) interaction pairs were mapped to
90 UniProt records. Get3 (BioGrid ID 31962) had 456 Biogrid interactor ids, which
mapped to 465 UniProt records. After filtering those records through the TA anchor
datasets TRC40 and Get3 returned 7 and 22 hits respectively.
In yeast, Pex19 (BioGrid ID 31994) contained 466 interactor pairs which mapped
to 384 UniProt IDs. The Homo sapiens Pex19 (BioGrid ID 111782) contained 230
interaction pairs which successfully mapped to 218 UniProt records. When the TA
list filters were applied, 2 Homo sapiens and 7 Saccharomyces cerevisiae records were
found.
For SRP54, ideally the plant version of the protein was required, for which there
is more precedent for post-translational protein interaction and biogenesis into the
chloroplasts [291], however, of the 3 plant SRPs available on Biogrid, between them
only 10 interactor pairs were available, none of which were common with our TA lists.
It is also worth noting that plants have two forms of SRP; one for the chloroplastic
and one cytoplasmic which would have to be compared separately. The human SRP54
(Biogrid ID 112607) had 37 interactors which mapped to 85 UniProt IDs, but again,
none of which were in our TA lists. On the other hand, the yeast SRP54 (BioGrid
ID 36258) had 270 interactors which mapped to 273 UniProt records. 4 of those were
found in our TA lists.

3.3.2

Calculating hydrophobicity

Windowed hydrophobicity was calculated using a window length of 5 residues, and half
windows were permitted. Average hydrophobicity takes the total of the raw amino
acid hydrophobicity values and divides them by the number of amino acids in the
slice. Unless explicitly stated, values reported in the results are based on the Kyte &
Doolittle scale [73] which is based on the water–vapour transfer free energy and the
interior-exterior distribution of individual amino acids.

116

3.3.3

CHAPTER 3. COLLATION AND ANALYSIS OF TAIL-ANCHORED...

Calculating sequence information entropy

Information entropy is essentially an estimate of the entropy of a string of characters.
In the context of biology, it can be thought of as an estimation of the non-randomness
of a sequence. Sequence complexity can be used to analyse DNA sequences [147–
149] and is a component of the TMSOC z-score which can predict function beyond
anchoring of a TMH; an increase in complexity is associated with increased likelihood
of function [44, 45, 227]. Here we focus on the analysis of the complexity of a string
of characters in protein sequences.
Broadly speaking, the entropy of a string of characters in information theory can
be defined as in equation 3.1, and we treat the protein sequence TMH as a string with
or without its flanking regions.

H (S) = −

n
X

P (si ) log2 P (si )

(3.1)

i=1

Where H is the entropy of a sequence S with values {s1 , . . . , sn }, and P (si ) is the
probability (P ) of a character i through each position in S. This allows us to quantify
the average relative information density held within a string of information [151].

3.3.4

Statistics

The null hypothesis of homogeneity of two distributions was examined with the Kolmogorov Smirnov, the Kruskal-Wallis, and the 2-sampled Student’s T-test statistical
tests. These tests were all ran through the Python SciPy stat v0.17 package [280].
To note, the K-S test scrutinises for significant maximal absolute differences between
distribution curves; the K-W test is after skews between distributions and the student
T-test statistical test checks the average difference between distributions.
Since the p-value is a product of a fraction of test statistics obtained from a permutated set of the samples, it exponentially increases as N increases; the p-value is
a strong function of N . We rely on the Bahadur slope (B) as a measure of distance
between two distributions [227, 259–261]. A larger Bahadur slope shows a greater
difference between the two distributions.

B=

| ln(p value)|
N

(3.2)

3.3. METHODS

117

In the heatmaps (Figure 3.6, Figure 3.7), the relative percentage normalisation was
used rather than a fraction of the absolute value. This aims to answer the question of
“if we have a certain amino acid, which position is it likely to be in?” and are able to
sensitively identify clusters of skewed preference [227].
100 · ai,r
(3.3)
ai
ai is the total abundance of residues of a specific amino acid type (i) of an aligned
qi,r =

set of TMH-containing segments. Peaks in qi,r as a function of r (the position index)
reveal the preferred positions of residues of type i. TMHs are oriented according
to “inside” to “outside” from left to right respectively where “inside” refers to the
cytoplasm.

3.3.5

Modelling cytochrome b5 and PTP1b

In order to assess the 3D arrangement of residues, the HHpred web server was used
to query homologues of and model templates for Cytochrome b5 (UniProt accession
code P00167) and PTP1b (UniProt accession code P18031) [332]. Homologues were
queried using three iterations of HHblitscd against the sequence database version
uniprot20 2016 02 to generate the query Hidden Markov Model. For cytochrome b5,
a multiple alignment was generated from PDB accession codes and residues positions
were approximated for the TMH regions based on those of PDB accession codes 5NAO,
5DOQ, 5NAM, and 2MMU covering the TMH. Modeller was run from within the HHPRED server to generate the homology model [333, 334]. Similarly for PTP1b, the
TMS was modelled using a homology model derived from a single sequence alignment
of 5NAO based on the TMH±6 (the length of the C-terminal tail) residues of PTP1b.
Both the TMH regions for that of PTP1b and cytochrome 1b were verified by a
consensus of sequence TMH predictions (Scampi seq [72], Phobius [71], TMHMM [67],
MEMSAT3 [66], TMpred [335], HMMTOP [68], DAS–TMfilter [238], MINNOU [336],
OCTOPUS [70], PRODIV [69], PRO–S [69], S–TMHMM [69], and proteus [337]).
However, it should be noted that not all these predictions unanimously agreed. For
PTP1b, several methods identified more than 1 TMH (HMMTOP, TMPred) whilst
Memsat identified a short TMH in a completely different region for the TMH (35A
to 45R). Besides that, only S–TMHMM and PRO agreed on the exact start and stop

118

CHAPTER 3. COLLATION AND ANALYSIS OF TAIL-ANCHORED...

positions, and it so happens that these are also the majority consensus positions [338]
(409F to 429F).
APBS as a PyMol plugin was used to map the electrostatic surface of the model
[339]. Consurf [340] was used to map the conservation scores based on 5 iterations
of PSI-BLAST [341] with an E-value cut-off of 0.0001. Hydrophobicity was mapped
according to the Eisenberg aggregated hydrophicity scale [144] using a script accessed
at https://pymolwiki.org/index.php/Color_h.

3.3.6

Availability of materials

The scripts and datasets associated with this study can be accessed at https://
github.com/JamesABaker/TA-protein-seq.

3.4
3.4.1

Results and discussion
A comparison of up-to-date tail-anchored protein
datasets

Here, we use two sources for TA protein datasets. One dataset is based on a previous
method [284] to obtain TA datasets and consists of 9296 TMH residues (13279 including up to ±5 flanking residues) from 443 SwissProt entries with 90% redundancy
removal. Another dataset contains the UniProt curated set of Type IV membrane proteins again with 90% redundancy removal. This dataset contains 20528 TMH residues
(27950 including up to ±5 flanking residues) from 956 UniProt protein records.
In order to get an understanding of the consistency of the datasets, before removing
redundant proteins, we compared these two datasets to a remapped set of proteins from
a previous 2007 method by Kalbfleisch et al., which aimed to predict TA proteins from
the NCBI database using predictive tools [284] (see Figure 3.2 for an overview of these
methods). The S1 dataset was built with an aim to gather TA proteins in the human
genome from the NCBI. The greatest source of uncertainty here is that the original S1
list includes 411 records, however, only 222 of these were successfully mapped to the
UniProt dataset. This figure is closer to the 202 proteins from the original S1 list that
excluded proteins that were either hypothetical or splice isoforms. That being said,

3.4. RESULTS AND DISCUSSION

Redundant UniProt Curated

119

Redundant SwissProt Filtered

8
2077

18
16

31

5

170

Kalbfleisch et al., 2007 “S1”
dataset
Figure 3.3: A Venn diagram showing tail-anchored protein UniProt ids present in each
of the datasets as well as those present in multiple datasets. The number of ids present
in redundant versions of i) the supplementary materials table of a previous study by Kalbfleisch et
al., predicting the complete set of human tail-anchored proteins was mapped to UniProtKB and is
denoted by “S1” [284] is in red, ii) and in green is the SwissProt dataset filtered according to typical TA
features limited to the Homo sapiens proteome [75], and iii) in blue is the UniProt curated list of TA
proteins [75]. Note that to avoid losing IDs to redundancy reduction this diagram was generated
without the use of CD-HIT [277, 331], which is applied in the later statistical analysis.

120

CHAPTER 3. COLLATION AND ANALYSIS OF TAIL-ANCHORED...

this mapping step prevents us from directly comparing the entire original S1 dataset.
We compared the up-to-date datasets to S1 to see how many records are shared, how
many are now obsolete, and how many are unique.
Figure 3.3 shows that S1 has 175 record ids of 222 records (78.8%) which do not
share overlap the up-to-date manually curated UniProt dataset [75]. Of the 170 unique
records of that S1 dataset, 4 were manually annotated as not belonging to the Homo
sapiens proteome, 20 have the C-terminal as annotated being cytoplasmic, only 125
had TRANSMEM annotation indicating a bona fide TMH. If we apply equivalent filters,
only 42 have annotation verifying that they are TA proteins.
Equivalent criteria to the original Kalbfleisch et al., 2007 [284] study was applied
to the entire SwissProt database and then restricted to the human proteome dataset.
24 of these 47 records (51.1%) are in the curated UniProt TA dataset. 21 of the 49
(44.7%) records from SwissProt filtered Homo sapiens dataset can be found in the
original S1 list.
The same method applied to an up-to-date dataset overlaps more with a manually
curated dataset. There is also a large degree of what we now believe to be mistakes
that occurred in the older prediction tools and datasets, even when using similar
methods. As a trend, this shows that up-to-date datasets improve the reliability of
this automated predicted method. These automated criteria still do not fully align
with the manually curated list. Of 2633 records in the manually curated list, only
2241 have the TRANSMEM annotation. Furthermore, it is not only the transmembrane
annotation itself but also the type of transmembrane protein. Small integral membrane
protein 1 is a blood group antigen (UniProt ID B2RUZ4) that is just one example of a
protein we know to be a post-translationally inserted TA protein, and yet in UniProt it
is annotated as a type II, not a type IV, transmembrane protein. As a result of which
it appears in the SwissProt automatically filtered list and not the manually curated
list. In an ideal database, where there are instances of discrepancy, a note on post-translational or co-translational biogenesis would address this issue. Ultimately, this
points to the idea that datasets are a moving target as they are constantly updated
with more accurate information using evermore reliable tools and methods.

3.4. RESULTS AND DISCUSSION

3.4.2

121

It is difficult to observe any hydrophobic variation of
tail-anchored protein transmembrane helices from different species

In single-pass proteins of eukaryotic species, there are typically various adaptations
of the TMH to adhere to the membrane constraints of the specific membrane. For
single-pass proteins, previous studies have observed differences in terms of TMH hydrophobicity between Saccharomyces cerevisiae and Homo sapiens TMPs [86], or in
cress, yeast, bacteria, and Homo sapien datasets [227]. We would expect to see a similar trend between the TMHs of TA proteins from different species. However, when
assuming a zero-difference hypothesis, in these TMH TA protein datasets we cannot
observe any species-level differences between the datasets at this sample size for TMH
hydrophobicity.
When comparing the average Kyte & Doolittle [73] hydrophobicity values for
the TMHs from Homo sapiens and Mus musculus, A. thaliana, and S. cerevisiae,
we can see little difference between the mean values. All of the mean values lie between 2.3-2.6 when we only consider the TMH and at 1.3-1.6 when also considering
flanking residues in close proximity to the TMH (±5 residues) (Figure 3.4). Indeed, we
see no strong observable statistical differences in hydrophobicity (P > 3.35E −1 in the
SwissProt automatically filtered list Table 3.1, and P > 2.40E − 1 in the UniProt curated list Table 3.2). There are also no consistent trends among the absolute Bahadur
slopes; no datasets are greatly different from any other.
Table 3.1: Hydrophobicity statistical comparisons between mouse and human, yeast,
and plants in the SwissProt Filtered Dataset. Here, we compare a mammalian set of TA
proteins (Homo sapiens N=46 and Mus musculus N=48) to A. thaliana (N=49) representing plants
and S. cerevisiae (N=24) representing yeasts. The hydrophobicity was predicted as the mean average
of the values of the sequences of the TMH, as well another group including up to ±5 flanking residues,
since predicting the boundary of TMHs is difficult, according to the Kyte & Doolittle hydrophobicity
scale [73]. The Test column refers to the statistical score obtained from the test; H statistic for the
Kruskal Wallis, the KS statistic for the Kolmogorov Smirnov test, and the t-statistic for the T-test. p
is the p-value of that statistical score. B refers to the Bahadur slope, an interpretation of the p-value
that accounts for the sample size powering the test [259, 260].

TMH

TMH and flanks

KW
KS
T-test
KW
KS
T-test

Mammal and Plant
Test
p
B
0.93
3.35E-1
7.64E-3
0.13
6.36E-1
3.17E-3
-0.86
3.90E-1
6.58E-3
0.04
8.52E-1
1.12E-3
0.11
7.72E-1
1.81E-3
-0.22
8.23E-1
1.37E-3

Mammal and Yeast
Test
p
B
0.10
7.56E-1
2.37E-3
0.12
9.24E-1
6.69E-4
0.21
8.31E-1
1.57E-3
0.12
7.28E-1
2.69E-3
0.13
8.79E-1
1.09E-3
-0.38
7.04E-1
2.97E-3

Test
0.84
0.19
0.79
0.04
0.11
-0.19

Plant and Yeast
p
B
3.60E-1
1.40E-2
5.28E-1
8.76E-3
4.33E-1
1.15E-2
8.33E-1
2.51E-3
9.80E-1
2.81E-4
8.50E-1
2.22E-3

Here, we are dealing with datasets at least an order of magnitude smaller than

122

CHAPTER 3. COLLATION AND ANALYSIS OF TAIL-ANCHORED...

3.50
TMH
3.00

Kyte & Doolittle
Hydrophobicity

2.50
2.00
1.50
1.00
0.50

TMH
and flanks

0.00
Mammal

Plant

Yeast Mammal

UniProt Curated Set

Plant

Yeast

SwissProt Filtered Set

Figure 3.4: Average values of species datasets from UniProt manually curated set and
SwissProt automatically filtered dataset.
The average hydrophobicity values from the Kyte & Doolittle scale [73].for both the TMH and the
TMH±5 residues. Values are shown for both the UniProt manually curated set and the SwissProt
filtered set. In the UniProt manually curated set we compare the mammalian set of TA proteins
(Homo sapiens N=30 and Mus musculus N=30) to A. thaliana (N=57) representing plants and S.
cerevisiae (N=27) representing yeasts. For the SwissProt filtered set we compare the mammalian set
of TA proteins (Homo sapiens N=46 and Mus musculus N=48) to A. thaliana (N=49) representing
plants and S. cerevisiae (N=24) representing yeasts. Error bars are shown at ±1σ from the mean of
the respective dataset.

those broad studies [86, 227] which could explain the absence of the effect. However,
this only goes to show that if there is a biochemically distinct effect in TA proteins in
terms of hydrophobicity between species, it is indeed weak.

3.4. RESULTS AND DISCUSSION

123

Table 3.2: Hydrophobicity statistical comparisons between mouse and human, yeast, and
plants in the UniProt Curated Dataset. Here, we compare a mammalian set of TA proteins
(Homo sapiens N=30 and Mus musculus N=30) to A. thaliana (N=53) representing plants and S.
cerevisiae (N=27) representing yeasts. The hydrophobicity was predicted as the mean average of
the values of the sequences of the TMH, as well another group including up to ±5 flanking residues,
since predicting the boundary of TMHs is difficult, according to the Kyte & Doolittle hydrophobicity
scale [73]. The Test column refers to the statistical score obtained from the test; H statistic for the
Kruskal Wallis, the KS statistic for the Kolmogorov Smirnov test, and the t-statistic for the T-test. p
is the p-value of that statistical score. B refers to the Bahadur slope, an interpretation of the p-value
that accounts for the sample size powering the test [259, 260].

TMH

TMH and flanks

3.4.3

KW
KS
T-test
KW
KS
T-test

There

Test
0.71
0.13
-0.93
1.37
0.19
-1.17

are

Mammal and Plant
P
B
4.01E-01
8.09E-03
6.93E-01
3.24E-03
3.55E-01
9.15E-03
2.42E-01
1.26E-02
2.40E-01
1.26E-02
2.45E-01
1.24E-02

Test
0.03
0.13
-0.11
0.38
0.14
-0.79

biochemical

Mammal and Yeast
P
B
8.72E-01
1.57E-03
9.11E-01
1.08E-03
9.13E-01
1.04E-03
5.36E-01
7.17E-03
8.13E-01
2.38E-03
4.35E-01
9.58E-03

differences

Test
0.57
0.20
0.64
0.08
0.09
0.20

Plant and Yeast
P
B
4.48E-01
1.00E-02
4.16E-01
1.10E-02
5.22E-01
8.12E-03
7.80E-01
3.11E-03
9.97E-01
3.21E-05
8.43E-01
2.14E-03

between

tail-

-anchored transmembrane helices from different organelles
Although the species datasets appeared to have no significant differences between them
in terms of hydrophobicity, we also investigated the subcellular membranes. We see
clear differences in the biochemistry of the TMH (Figure 3.5).
In the UniProt manually curated dataset, the Kyte & Doolittle hydrophobicity
scores range from 1.7 in mitochondria to 2.7 in the PM (Figure 3.5A).
Table 3.3: Statistical comparisons between TMH sequences from organelles in the
UniProt Curated Dataset. Here, we compare an organelle subset from the UniProt curated
dataset of TA proteins. We compare ER (N=397) to Golgi (N=83), PM (N=31), and the mitochondria (N=426). The hydrophobicity was predicted as the mean average of the values of the sequences
of the TMH, as well another group including up to ±5 flanking residues, since predicting the boundary of TMHs is difficult, according to the Kyte & Doolittle hydrophobicity scale [73]. The linguistic
information entropy was calculated according to the methods section [151]. The Test column refers to
the statistical score obtained from the test; H statistic for the Kruskal Wallis (KW), the KS statistic
for the Kolmogorov Smirnov test (KS), and the t-statistic for the student’s T-test (T-test). p is the
p-value of that statistical score. B refers to the Bahadur slope, an interpretation of the p-value that
accounts for the sample size powering the test [259, 260].

Hydrophobicity of TMH

... and flanks

Sequence Entropy of TMH

... and flanks

KW
KS
T-test
KW
KS
T-test
KW
KS
T-test
KW
KS
T-test

Test
21.83
0.34
-6.45
0.21
0.19
0.32
4.66
0.24
3.22
0.52
0.13
1.08

ER and Golgi
P
B
2.98E-06
2.66E-02
1.61E-07
3.27E-02
2.72E-10
4.61E-02
6.48E-01
9.07E-04
1.10E-02
9.44E-03
7.48E-01
6.07E-04
3.09E-02
7.28E-03
4.78E-04
1.60E-02
1.37E-03
1.38E-02
4.70E-01
1.58E-03
2.06E-01
3.31E-03
2.82E-01
2.65E-03

Test
28.53
0.57
-8.86
17.53
0.50
-4.85
27.54
0.46
6.42
19.50
0.41
4.47

ER and PM
P
B
9.21E-08
3.80E-02
5.32E-09
4.47E-02
2.30E-17
8.99E-02
2.83E-05
2.46E-02
4.69E-07
3.42E-02
1.75E-06
3.11E-02
1.54E-07
3.68E-02
4.20E-06
2.91E-02
3.71E-10
5.10E-02
1.01E-05
2.70E-02
7.97E-05
2.22E-02
1.00E-05
2.70E-02

Test
377.02
0.67
23.53
490.46
0.82
34.60
24.03
0.18
-4.55
40.11
0.23
-5.84

ER and mito
P
5.54E-84
4.22E-82
6.58E-94
1.13E-108
5.58E-123
2.19E-162
9.48E-07
2.10E-06
6.28E-06
2.40E-10
5.53E-10
7.51E-09

B
2.34E-01
2.28E-01
2.61E-01
3.03E-01
3.43E-01
4.53E-01
1.69E-02
1.59E-02
1.46E-02
2.70E-02
2.60E-02
2.28E-02

In the UniProt curated list, there are clear hydrophobic differences between all the
organelle TMH datasets excluding flanks (P < 2.98E−6) which as a trend becomes less

124

CHAPTER 3. COLLATION AND ANALYSIS OF TAIL-ANCHORED...

A)

B)

Figure 3.5: Average sequence-based biochemical values of organelle datasets from
UniProt manually curated set and SwissProt automatically filtered dataset.
A) The average hydrophobicity values from the Kyte & Doolittle scale [73], B) the average information
entropy [151] (see methods) for both the TMH and the TMH±5 residues. Values are shown for both
the UniProt manually curated set and the SwissProt filtered set. In the UniProt manually curated
set we compare TA proteins from the ER (N=400) to the Golgi (N=82), the PM (N=37), and the
mitochondria (N=401). For the SwissProt filtered set we compare TA proteins from the ER (N=98)
to the Golgi (N=82), the PM (N=157), and the mitochondria (N=65). Error bars are shown at ±1σ
from the mean of the respective dataset.

3.4. RESULTS AND DISCUSSION

125

clear when considering the TMH±5 flanking residues except for mitochondria which
increases in significance when considering the flanks also (Table 3.3). The ER and
mitochondrial tests are very significant (P < 4.22E − 82). Consistently the Bahadur
slope is at least an order of magnitude greater in the ER and mitochondrially located
protein comparison than for the other considerations, so these differences cannot be
accounted for by the larger sample size. This gap in hydrophobicity appears to be
due to a trend of the ER, PM, and Golgi using isoleucine, valine, and leucine as
their most common TMH residues, whereas in the case mitochondrially located TA
proteins, the most common residue type is alanine in the UniProt manually curated
dataset (16.3% of total residues) followed by valine (12% total residues)(Figure 3.6).
Similarly, alanine is the second most common residue in mitochondrial located TA
proteins from the SwissProt automatically generated dataset at 11.9% of the total
residues after leucine which is 13.4% of the total residues(Figure 3.7).
Analysis from 16 TA proteins with known subcellular locations showed that both
the C-terminal tail charge and hydrophobicity are determinants of the terminal destination to the ER, mitochondria, and the peroxisome intracellular subcellular locations
[328]. They found that less hydrophobicity and more charge in the “tail” determined
the TA protein for the mitochondria rather than the ER. This corroborates what we
see in terms of hydrophobicity (Figure 3.5A). When we consider charge difference between organelles on larger datasets, we see trends that reinforce this idea, however,
rather than net charge, we see differences in charge distribution along the TMH and
the neighbouring flanks. In the SwissProt automatically filtered dataset, in the ER
9.4% of the residues are positively charged, and 2.5% are negatively charged. Most
of the positively charged residues cluster following the “positive-inside” rule between
positions -15 and -10 for R and K, but so do the negatively charged residues D and
E, effectively reducing this local charge by 2.5%. In mitochondria, we find that the
proportion of charge is similar (10.4% R and K, 3.3% D and E) however the negatively
charged residues cluster on the N flank (-15 to -8) and the positively charged residues
cluster more strongly on the outside flank (positions 9 to 15) (Figure 3.7).
In the UniProt manually curated ER set, 6.6% of residues are positively charged
and 3.3% of residues are negatively charged. K clusters strongly on the inside flank
as expected, yet R clusters strongly between positions 7 to 15 and rather weakly at

126

CHAPTER 3. COLLATION AND ANALYSIS OF TAIL-ANCHORED...

A)

B)

ER

0

3

Golgi

27

C)

0

2

41

D)

PM

0

3

Mitochondria

66

0

2

37

Figure 3.6: The normalised skews of each amino acids from tail-anchored proteins
grouped by localisation from the UniProt manually curated dataset.
The residue position aligned with the centre of the TMH is on the horizontal axis, and the residue type
is on the vertical axis. Amino acid types are listed in order of decreasing hydrophobicity according
to the Kyte and Doolittle scale [73]. Flank lengths were restricted to ±5 residues. The edge residues
from proteins with flank lengths and TMH lengths that exceeded the plotted 31 residues were still
included in the normalisation calculations despite not being plotted. The colour scale represents the
relative percentage of a particular amino acid and is shown with dark blue as 0, white as the 50th
percentile value of the entire heatmap, and dark red as the highest percentage on the heat map.
The panels are constructed from TA proteins derived from the SwissProt automatic method with
redundancy removal applied detailed in the methods section. The datasets were further separated by
subcellular locations: (a) the ER, (b) the Golgi, (c) the cell membrane, (d) the mitochondria. These
datasets are more thoroughly outlined in the methods section.

3.4. RESULTS AND DISCUSSION

127

the inside flank (positions -15 to -10) (Figure 3.6). Similarly to the SwissProt sets,
D prefers the inside flank but is tolerated in the outside flank. The more abundant
E residues behave very unusually and cluster at positions 5-10. Generally, charged
residues are suppressed in the TMH core [86, 109], especially in anchoring TMHs [227].
It is unclear why this is observed, yet, altogether the 313 glutamic acid residues and
397 arginine residues that appear unusually deep in the TMH core may be to an extent
neutralising one another in the compacted TMH arrangement, but are ultimately not
that abundant compared to the total number of residues in this organelle dataset
(11351 total residues).
In mitochondria, 1413 positively charged residues (11% of the total residues in the
mitochondrial dataset) were preferentially located at the outside flank and somewhat
into the core (positions 6 to 15) than the expected “inside” flank (positions -15 to
-5). The 221 negatively charged residues (1.7%) unusually cluster at the inside. This
results in a strong net positive-outside charge signal since there are more positively
charged residues on the outer flank un-countered by the negatively charged residues,
which are skewed with a preference for the inside flank.
Information entropy has been known to identify cryptic function in TMHs when
considered along with hydrophobicity [44, 45]. In terms of information entropy, there
is a marked decrease in entropy in the PM subset (mean entropy = 3.15 in the TMH,
2.67 including ±5 flanking residues) from the UniProt curated dataset compared to the
other organelle datasets (entropy > 3.29 and > 2.85 including the flanks). However,
this stark difference in TMHs from PM-bound TA proteins and the other organelle
datasets cannot be observed in the SwissProt set (Figure 3.5).
No clear significant differences can be observed for the information entropy (P >
6.33E − 2). This is unsurprising given that the hydrophobic nature of the TMHs demands that certain residues must be over-represented, which lowers the information
entropy. In this case, we have a highly hydrophobic set, the PM UniProt set, which
likely contains a higher proportion of the most hydrophobic residues. As a trend,
the information entropy mirrors the hydrophobicity albeit with less range between
dataset means (2.67-3.15 in the TMH for information entropy, 1.72-2.74 for hydrophobicity)(Figure 3.5).
Similarly, in the SwissProt filtered dataset, the mean TMH hydrophobicity for

128

CHAPTER 3. COLLATION AND ANALYSIS OF TAIL-ANCHORED...

A)

B)

ER

0

2

Golgi

39

C)

0

2

42

D)

PM

0

2

Mitochondria

33

0

2

30

Figure 3.7: The normalised skews of each amino acids from tail-anchored proteins
grouped by localisation from the SwissProt automatically filtered dataset
Similarly to figure 3.6, the residue position aligned with the centre of the TMH is on the horizontal
axis, and the residue type is on the vertical axis. Amino acid types are listed in order of decreasing
hydrophobicity according to the Kyte and Doolittle scale [73]. Flank lengths were restricted to ±5
residues. The edge residues from proteins with flank lengths and TMH lengths that exceeded the
plotted 31 residues were still included in the normalisation calculations despite not being plotted.
The colour scale represents the relative percentage of a particular amino acid and is shown with
dark blue as 0, white as the 50th percentile value of the entire heatmap, and dark red as the highest
percentage on the heat map. The panels are constructed from TA proteins derived from the SwissProt
automatic method with redundancy removal applied detailed in the methods section. The datasets
were further separated by subcellular locations: (a) the ER, (b) the Golgi, (c) the cell membrane, (d)
the mitochondria. These datasets are more thoroughly outlined in the methods section.

3.4. RESULTS AND DISCUSSION

129

Table 3.4: Statistical comparisons between transmembrane helix sequences from organelles in the SwissProt Filtered Dataset. Here, we compare organelle subsets from the
SwissProt automatically filtered dataset of TA proteins. We compare ER (N=98) to Golgi (N=82),
PM (N=157), and the mitochondria referred to as “mito” (N=65). The hydrophobicity was predicted
as the mean average of the values of the sequences of the TMH, as well another group including up
to ±5 flanking residues, since predicting the boundary of TMHs is difficult, according to the Kyte &
Doolittle hydrophobicity scale [73]. The linguistic information entropy was calculated according to
the methods section [151]. The Test column refers to the statistical score obtained from the test; H
statistic for the Kruskal Wallis (KW), the KS statistic for the Kolmogorov Smirnov test (KS), and
the t-statistic for the student’s T-test (T-test). p is the p-value of that statistical score. B refers to
the Bahadur slope, an interpretation of the p-value that accounts for the sample size powering the
test [259, 260].
Test

ER and Golgi
p
B

Test

ER and PM
p
B

ER and mito
Test

p

B

TMH Hydrophobicity

KW
KS
T-test

11.96
0.27
-3.47

5.43E-4
1.98E-3
6.50E-4

4.18E-2
3.46E-2
4.08E-2

0.02
0.08
-0.17

8.77E-1
8.48E-1
8.67E-1

5.14E-4
6.44E-4
5.60E-4

8.46
0.27
3.45

3.64E-3
4.62E-3
7.24E-4

3.45E-2
3.30E-2
4.44E-2

... including flanks

KW
KS
T-test

5.92
0.21
-2.52

1.50E-2
2.85E-2
1.25E-2

2.33E-2
1.98E-2
2.43E-2

9.14
0.26
-3.09

2.50E-3
4.88E-4
2.23E-3

2.35E-2
2.99E-2
2.40E-2

26.42
0.43
4.95

2.75E-7
4.93E-7
1.87E-6

9.27E-2
8.91E-2
8.09E-2

TMH entropy

KW
KS
T-test

2.96
0.13
1.58

8.56E-2
4.32E-1
1.15E-1

1.37E-2
4.66E-3
1.20E-2

0.66
0.10
0.79

4.17E-1
5.27E-1
4.32E-1

3.43E-3
2.51E-3
3.29E-3

0.69
0.18
1.03

4.05E-1
1.40E-1
3.06E-1

5.54E-3
1.20E-2
7.26E-3

... including flanks

KW
KS
T-test

2.62
0.15
1.84

1.06E-1
2.48E-1
6.75E-2

1.25E-2
7.75E-3
1.50E-2

2.87
0.17
1.66

9.04E-2
6.56E-2
9.84E-2

9.42E-3
1.07E-2
9.09E-3

0.05
0.21
0.42

8.31E-1
6.33E-2
6.72E-1

1.14E-3
1.69E-2
2.44E-3

mitochondria is the lowest at 1.9. It appears to be the Golgi apparatus that is the
peak at 2.4. In the SwissProt dataset, when we compare each subset of only the TMH
to the ER subset, we find significance between the ER and the Golgi (p < 1.98E − 3),
and the ER and the mitochondria (p < 4.62E − 3), however, the ER and PM are
more similar considering the Bahadur values are < 6.44E − 4, two orders of magnitude
smaller than the other sets (Bahadur values > 3.3E − 2) (Table 3.4). When we take
into account the flanks, the ER and PM dataset can be distinguished (P < 2.50E − 3),
however, as a trend the other two comparisons, ER and Golgi become less significant,
and ER and mitochondria become more significant.
The information entropy of the TMH string was also examined. No significance
was observed in any consideration of the information entropy, but similarly to the
UniProt subset, as a trend, the entropy mirrors the hydrophobicity (Figure 3.5).
Lipid asymmetry is caused by sphingomyelin and glycosphingolipids on the noncytosolic leaflet and phosphatidylserine and phosphatidylethanolamine in the cytosolic
leaflet in the Golgi and PM, however, this asymmetry is not present in the ER [269,
270]. Another consideration is that sphingomyelin is not present in the ER but is
present in the Golgi [272] and PM [184, 273]. Furthermore, the PM contains densely
packed sphingolipids and sterols [274]. Mitochondria have bacterial lipids in their

130

CHAPTER 3. COLLATION AND ANALYSIS OF TAIL-ANCHORED...

membrane and uniquely contain cardiolipin [342], which is also confirmed to be present
in the MOM [343]. The compositional differences of the membrane environments
inevitably places different constraints on the TMHs, which may drive the observed
variations in their biochemistry.
The hydrophobicity of TMHs in TA proteins is lower in proteins targeted to the
mitochondria compared to those targeted to the ER [344]. Here we observe that average biochemical features are evidently of significance. Furthermore, we see that the
typically positive-inside negative outside tandem in positively and negatively charged
residues is reversed in the mitochondria to positive-outside negative-inside. This variation in TMH hydrophobicity and the charged residue skew reversal may be yet another
nuance of the system which goes some way to explaining how signals are maintained
in local environments even when the average values are ambiguous. We also identify
that alanine is a key reason behind the hydrophobic difference between subcellular
organelles, with alanine being highly selected, if not favoured over other hydrophobic
residues like leucine, in mitochondrial TA proteins compared to the ER, the Golgi, and
the PM. This could be an adaptation to the mitochondrial membrane, which contains
a higher level of cardiolipins than other membranes [92, 343]. Regarding the charged
residue distribution, there should also be a consideration of membrane potential. Although membrane potentials are in flux, typically, the PM has a potential of 70mV
with the negativity being on the cytoplasmic side. It has been shown that the ER is
again between 75-95mV with negativity on the lumenal side [100, 101]. There is no
detectable potential across the Golgi [103], and the mitochondrial inner membrane has
a potential of 150-180 with negativity on the matrix side [102]. However, whilst those
numbers go some way to satisfy the flanking charge reversal we see, they do not apply
to the MOM in which the TA proteins are localised; β barrel porins on the MOM allow
passive ion and water transfer across the membrane effectively diminishing the membrane potential. Furthermore, these observed membrane potential values are by their
nature not fixed values. Not only would this be most striking in the nerve and muscle
cells of course, but also would contrast between resting and stimulated secretory cells.
It has been known for some time that both the charge and the hydrophobic length
of the TA protein TMH region are indicative of subcellular targeting of the protein
[344]. ER localised TA proteins typically have a more hydrophobic TMH than those

3.4. RESULTS AND DISCUSSION

131

targeted to the MOM [298]. The C-terminal tail charge is particularly important for
determining subcellular localisation, and can even override the hydrophobic signal if
strong enough [328]. Here we see that while hydrophobicity is statistically different
between the subcellular membranes, there is overlap. Whilst there are differences in
total average charge at the C-terminal flank, this is not an absolute rule. Perhaps these
TA protein anchor TMHs are mechanistically similar to those of the signal anchor
proteins, a group of singlepass TMPs containing an uncleaved SP that remains in
the membrane. The single hydrophobic segment in mitochondrial signal-anchored
proteins serves as both a mitochondrial targeting signal and a membrane anchor.
Signal-anchored proteins, along with some TA proteins, have been shown to be able to
spontaneously insert into the membrane independently from the translocon [41, 323,
345].
Nevertheless, it is tempting to conclude that these biochemical differences between
differently localised TA proteins are an adaptation to the membrane composition and
environment. But this must be tempered by noting that the spontaneously inserting
cytochrome b5 localises to the mitochondrial membrane in the absence of cytosol, and
to the ER in the presence of cytosol [329]; there are also biological factors determining
localisation.
In summary, the mitochondria located TA protein TMHs typically have a preference for alanine over leucine unlike their secretory counterparts and have a negative-inside positive-outside tendency counter to the overwhelming majority of TMPs. It
is unclear if these features are a biophysical adaptation or part of a biological sorting
process.

3.4.4

More annotation is required to identify chaperone interaction factors of the transmembrane helix.

TA proteins known to interact with certain chaperones were acquired by filtering the
interactor partner IDs for chaperones from BioGrid through the redundant versions of
these UniProt manually curated lists and SwissProt automatically generated lists.
Hsp40, Hsc70, SRP54 (both plant and human) returned 0 hits, indicating a lack

132

CHAPTER 3. COLLATION AND ANALYSIS OF TAIL-ANCHORED...

of annotation regarding TA proteins with these chaperons probably due to the relatively polar, and non-trivially predictable, TMHs of TA proteins that these chaperones
interact with.
Snd1 has 15 records that were in our TA lists. The average Kyte & Doolittle
hydrophobicity of these records was 2.60 in the TMH itself and 1.58 including ± 5
flanking residues. Sgt2, with 14 records, had a TMH hydrophobicity of 2.47 and 1.51
including the flanks. 5 records were captured for SGTA with a TMH hydrophobicity of
2.27 and 1.19 including the flanks. TRC40 had the highest TMH hydrophicity of 2.77
and 1.82 including flanks. However, TRC40 also only had 7 records. Get3 had 22 TA
interactor records with an average TMH hydrophobicity of 2.36 and 1.48 including the
flanks. The 2 records for human Pex19 had an average hydrophobicity of 1.33 for the
TMH and 0.70 including the flanking residues. The yeast Pex19 had a TMH average
hydrophobicity of 2.48 and 1.41 including the flanking residues. The 4 yeast SRP54
interactors had an average TMH of 2.43 and 1.98 including the flanking residues.
At the time of the investigation, these sample sizes are not statistically viable
for analysis. Whilst it appears TRC40 interactors have notably hydrophobic TMHs,
TRC40s yeast homologue Get3 has interactors with much more polar TMHs, yet SGTA
was lower than Sgt2 on average. So although these average values differ and overlap
between various chaperone systems, we tried to identify clearer patterns from the TMH
hydrophobic profiles (Figure 3.8). Similarly, whilst a clear dip in hydrophobicity at
position +5 in Get3 from between 2-3 across the rest of the TMH core to 0.32, there is
no such spike for TRC40 meaning this is probably not of any functional importance,
but rather an artefact of overrepresented proteins in the Get3 dataset. Snd1 also lies
among these values, reinforcing Snd1 as a biological redundancy system [289, 290,
297].
As expected, the human Pex19 interactors are as a trend among the most polar
throughout the TMH core, however, when we consider the yeast Pex19, this trend is
less clear (Figure 3.8). In figure 3.8, at least at a handful of locations (-3, 3, 4 and 5)
the SRP54 interactors have the most hydrophobic TMH cores.
TRC40 has the highest TMH mean average hydrophobicities at a given point at
3.9 at position -1. However, the yeast homologue Get3 does not appear to have any
preference for especially hydrophobic TMHs (Figure 3.8).

3.4. RESULTS AND DISCUSSION

133

5
Kyte and Doolittle
Hydrophobicity

4

Get3

3

TRC40

2

SGT2

1

SGTA

0

snd1

-1

Pex19 (human)

-2

Pex19 (yeast)

-3
-15

-10
-5
0
5
10
Sequence position from central TMH residue

15

SRP54 (yeast)

Figure 3.8: The profile of transmembrane helix and flanks hydrophobicity from tail-anchored protein groups stratified by chaperone interactors. On the horizontal axis is the
position relative to the central TMH residue defined by UniProt. On the vertical axis is the Kyte
& Doolittle hydrophobicity windowed across 5 residues allowing for half windows. The chaperone
interactors are colour coded according to the key.

In order to remove redundant proteins and investigate this further, more records
with greater levels of accurate annotation need to be available to both BioGrid and
UniProt. We also observe a great deal of overlap between the profiles, indicating that
as a trend this is more complex than hydrophobicity alone and that net/average TMH
polarity is not the absolute determinant of chaperone association.
However, this method demonstrates a potential way that this chaperone-interaction
problem can be investigated to verify that indeed hydrophicity plays a deterministic
role in chaperone selection.

3.4.5

Spontaneous insertion may be achieved by polar strips
in the transmembrane helix of tail-anchored proteins

The TMHs of cytochrome b5 and PTP1b are among the least hydrophobic of the TA
proteins [288, 289]. Indeed the TMH is so polar that it is not trivial to predict and is
not found in either dataset prepared herein. Structural modelling and analysis thereof
reveal features that may explain the “missing hydrophobicity” [74, 116, 267, 268] of
these particular TMHs.
A positive patch can be seen on the cytoplasmic side of either protein TMH, and

134

CHAPTER 3. COLLATION AND ANALYSIS OF TAIL-ANCHORED...

a negative patch on the lumenal tip at the C-terminus. There are large “positive-inside” patches [86, 109, 110, 113, 114, 227] and a strong “negative-outside” charge
[227] (Figure 3.9C) at the respective flanks of the TMH. Such a distribution of charge
is typical of a TMH anchor. Once in the membrane, this may allow the TMH to
be an effective anchor despite such poor hydrophobicity since it satisfies electrostatic
coupling to the membrane potential.

Furthermore, there is the question of overcoming the unfavourable interaction most
TMHs would hypothetically face when coming into contact with the highly polar
membrane interface without the presence of membrane integration machinery. We
observe a highly conserved strip of relatively polar / non-hydrophobic residues on one
side of the TMH core (in cytochrome b5 these are N112, P116, A120, A124 Y127, and
R128). Similarly, a polar face exists for the PTP1b TMH (R430, N434, Y426, T422,
and T419). These polar faces would not be as repulsed by the interfacial environment
as either a more hydrophobic TMH or an equally hydrophobic TMH with a different
sequence and structure order (Figure 3.9 and Figure 3.10). Scrambling the cytochrome
b5 TMH sequence whilst maintaining the same hydrophobicity (DSNSS W W T N W
V I P A I S A L I V A L M YR to DSNSS W W A S A I I A T M I P L L V N V
W YR) reduces the insertion potential [138]. Therefore we can conclude that there is
more to this phenomenon than hydrophobicity alone. It becomes apparent that the
3D arrangement of these relatively polar TMH residues is conserved and is probably
the key to spontaneous insertion of TMHs.

We speculate that this polar face allows the unassisted approach to the membrane’s
polar phospholipid head groups, and once in sufficiently close proximity, the hydrophobic side of the helix is entropically driven by the water environment into the membrane
core (Figure 3.11). This close proximity is less likely to be achieved if there is no side
of the TMH to favourably interact with the lipid head groups, even though average
hydrophobicity could be similar. Once integrated, the charged TMH flanking regions
help sustain integration in lieu of a more hydrophobic TMH core.

3.5. SUMMARY

135
C)

B)

A)

D)

TMH

180°

103

134

-2.530

1.060

+5KbT/ec

-5KbT/ec

1

9

Figure 3.9: Structural biochemical analysis of a homology model of cytochrome b5. (A)
The secondary structure of the protein coloured from the N terminus in blue to the C terminus in
red coloured through the rainbow according to the residue number. (B) The hydrophobicity of the
TMH from white representing relatively polar residues to red showing relatively hydrophobic residues
[144]. (C) The electrostatic surface with a threshold of ±5 KT/e calculated by APBS in PyMol [339].
Red patches are negatively charged whilst blue is positively charged. (D) The consurf scores on a
scale of 1-9 (all residues had sufficient data) [340]. Purple represents the most conserved whilst blue
is the least. Note the correlation between the highly and modestly conserved TMH residues and the
relatively polar residues. Another observable feature is the very strong “positive inside” [86, 109,
110, 113, 114] and “negative outside” features which are associated with anchoring [227].

3.5

Summary

Here, we have observed a clear biochemical distinction between TA proteins with
different terminal subcellular destinations. Previously it was known that both hydrophobicity and charge are involved in targeting to distinct subcellular locations
[328]. In this study, we find that the location of the charge along and around the
TMH of TA proteins differs among subcellular compartments. Crucially, there is a
shift in the charged residue inside-outside tandem in the TMH flanking residues in different organelles. In the secretory pathway, the “positive-inside” “negative-outside”
skew adheres to the membrane-potential for electrostatic coupling, but in the MOM
where there is no membrane potential, positively charged residues are skewed to be

136

CHAPTER 3. COLLATION AND ANALYSIS OF TAIL-ANCHORED...
C)

B)

A)

TMH

180°

404

435

-2.530

1.060

+5KbT/ec

-5KbT/ec

Figure 3.10: Structural biochemical analysis of a homology model of PTP1b. (A) The
secondary structure of the protein coloured from the N terminus in blue to the C terminus in red
coloured through the rainbow according to the residue number. (B) The hydrophobicity of the TMH
from white representing relatively polar residues to red showing relatively hydrophobic residues [144].
(C) The electrostatic surface with a threshold of ±5 KT/e calculated by APBS in PyMol [339]. Red
patches are negatively charged whilst blue is positively charged. Note the one hydrophobic face of
the TMH and the opposing relatively polar face. Another observable feature is the “positive inside”
[86, 109, 110, 113, 114] and “negative outside” features which are associated with anchoring [227].

located outside the cytoplasm similarly to β barrels [110] which are also located on the
MOM, and the negatively charged residues are preferentially found inside the cytoplasm. Furthermore, the missing hydrophobicity of mitochondrial TA proteins can be
in part attributed to the high abundance of alanine rather than leucine or isoleucine in
the TMH; it is not solely the tolerance of polar residues in the TMH. We expected to
see evolutionary adaptations of TMH hydrophobicity to species-specific membranes,
even within eukaryotes [86, 227]. In this study using both a manually curated dataset

3.5. SUMMARY

A)

137

B)
Even
hydrophobicity

Relatively
amphiphilic

Figure 3.11: A cartoon of a potential method the cytochrome b5 and PTP1b transmembrane helix could integrate spontaneously into the membrane. A) The marginally
hydrophobic TMH in teal cannot approach the membrane interface. B) Although the average hydrophobicity is the same as the teal TMH, by having a more hydrophobic side (blue), and a more
polar side (green), the TA protein now has a TMH surface that may interact more favourably with
the interfacial region. Once interacting with the membrane sufficiently close to the interfacial region, the hydrophobic face is still being entropically driven by water molecules from the cytosol,
which would lead to partitioning into the membrane. Once integrated, the strong “positive-inside”
“negative-outside” charges on the TMH flanks compensate for the lack of hydrophobicity in the core
of the TMH.

from UniProt and an automatically filtered list using SwissProt annotation, we do
not observe any strong differences. Since we could not scrutinise a difference in the
species, the strong hydrophobic differences between organelle TMHs are indicative of
a stronger adaptation pressure than between species as a whole. These differences are
likely to be partially adaptations to the organelle location membrane type and also
possible cryptic biological factors that play a role in their targeting via chaperonebinding affinity. We could not find any clear trends or perform statistical work on the
chaperone interactor datasets due to the small sample sizes, however, as the databases
are enriched, this same method will be able to answer the questions about chaperone
affinity with more accuracy in the future.
Furthermore, the spontaneously inserting TA proteins PTP1b and cytochrome b5
appear to share a polar face that emerges in structural models and a strong “positive-inside” “negative-outside” electrostatic surface. The polar face may be responsible

138

CHAPTER 3. COLLATION AND ANALYSIS OF TAIL-ANCHORED...

for the promotion insertion potential in the absence of insertion proteins since when
the sequence is scrambled, the insertion potential is reduced [138]. The positively and
negatively charged residues are distributed like an ideal anchoring TMH [227], which
could allow the marginally hydrophobic TMH to perform as a suitable membrane
anchoring feature.

Chapter 4
Sequence analysis of polarity in
transmembrane helices suggests
that translocation of marginally
hydrophobic helices could be
facilitated by neighbouring
typically hydrophobic helices
4.1

Abstract

Marginally hydrophobic transmembrane helices perform a variety of essential for life
intra-membrane molecular biochemistry. However, due to their relatively polar composition, they often lack sufficient hydrophobicity by themselves to incorporate into
the membrane efficiently via the translocon. In some cases other elements of the sequence context, such as other transmembrane helices or flexible loop regions, allow
the marginally hydrophobic transmembrane regions to integrate into the membrane
environment. Here we analyse large numbers of multipass transmembrane protein sequences stratified by protein families and molecular functions to show that there are
conserved stark differences between sequentially adjacent transmembrane helices in
139

140

CHAPTER 4. SEQUENCE ANALYSIS OF POLARITY...

terms of hydrophobicity and TMSOC complexity. Not all classes of transmembrane
protein have these highly polar-hydrophobic pairs. In the instances of GPCRs and ion
channels, there is evidence in the literature to support cooperative insertion between
these highly polar-hydrophobic transmembrane helix pairs highlighted in our results.

4.2

Introduction

Translocation is the process of incorporating a TMP into the membrane. A ribosome translates the RNA to a nascent peptide chain which is handed directly (cotranslational insertion) or indirectly (post-translational insertion) to the translocon
insertion machinery. This translocon assembly then threads the chain through the
membrane and releases hydrophobic TMHs into the membrane environment (Figure
4.1).
From Ribosome
C terminal

Cytoplasm

Membrane
Lumen
TMH
Translocon

Plug
N terminal

Lateral
exit gate

Figure 4.1: A cartoon showing the generally accepted schematic of sequential multipass
transmembrane helix insertion into the membranes. The two key concepts are that one at a
time, the TMHs emerge from the ribosome into the translocon. This appearance of hydrophobicity
triggers the lateral gate to open. As the nascent TMH is exposed to the membrane, it begins to
partition. The downstream protein from the TMH is then threaded through the translocon until the
next TMH is recognised. This implies that the TMHs ultimately have no meaningful interactions
with one another until the protein has been threaded into the membrane and the multiple TMHs
form a bundle.

The overwhelming majority of TMPs use the co-translational method of translocation. It has long been understood that this method is essentially the SRP recognising
and attaching to the nascent peptide chain whilst it is still associated with the ribosome, which targets the peptide and ribosome to an SR in association with the
membrane insertion machinery on the ER membrane [116, 171].
Crystal structures showed the SRP targets the nascent peptide chain for membrane insertion via a GTPase in both the SRP and the membrane bound translocon

4.2. INTRODUCTION

141

associated SR [175]. The SRP-ribosome-nascent peptide complex associates with
the translocon-SR complex thus bringing the nascent peptide chain in proximity to
the translocon [175]. Mutant studies of SRP revealed key discrete conformational
stages [175]. These are the specific recognition of signal sequences on cargo proteins,
the targeting of the package to the membrane, the handing over of the cargo to the
translocation machinery all the while maintaining precise spatial and temporal coordination of each molecular event [346].
The prevailing view of membrane insertion by the translocon is that the TMHs
partition in the membrane one at a time as the translocon lateral gate opens, exposing
the TMH to the membrane (Figure 4.1) [40].

4.2.1

The ribosome-translocon complex in the biogenesis of
membrane proteins

A ribosome translates the mRNA into amino acid chains and are present in all living
cells. Indeed the ribosomal complexes presence and activity is considered an important distinction in biology between ribosome-encoding organisms and capsid encoding
organisms [347]. They are a highly conserved RNA-protein complex with a multitude
of accessory proteins and targetting factors.
During translation of a TMP protein, the SRP binds to the ribosome after recognising the nascent protein as a TMP [169]. This complex then binds to the SR in
association with the membrane-bound translocon [172]. The nascent peptide is then
fed into the translocon as it is being translated; hence “co-translational insertion”. The
journey of the TMH through this machinery has been studied using both cross-linking
experiments and the relatively new technique of Arrest Peptide (AP)s [40].
APs are typically 10-15 residues long that bind to the upper end of the ribosomal
exit tunnel. Once a specific mRNA codon is recognised, ribosomal stalling is induced
[352] and the translation is halted unless a strong enough pulling force from the downstream insertion is acting on the nascent chain at that time [353]. Several “strengths”
of AP have been identified. For example, SecM from E. coli is 17 residues long and relatively weak, whereas a mutated SecM from Mannheimia succiniciproducens (Ms-Sup1)
is much stronger and 8 residues long ending in a proline which will halt translation

142

CHAPTER 4. SEQUENCE ANALYSIS OF POLARITY...

PTC
40S
subunit

60S subunit

Ribosome

Exit
tunnel
Nascent
peptide
Cytosol
Membrane

Translocon
Lumen

Lateral
exit gate

Plug

Figure 4.2: A cartoon of the ribosome in association with the translocon during insertion.
The structure used is of a translating mammalian ribosome-Sec61 translocon complex (PDB code
3j7r [162]). The Sec61 translocon is the membrane-embedded element of the structure that feeds
the protein into the membrane. The ribosome is made of two elements; 60s and 40s. These vary in
composition and size between eukaryotes and prokaryotes. PTC refers to the peptidase transferase
centre, a high conserved catalytic site across all the kingdoms of life that performs the translation of
RNA to amino acid polypeptides. After translation, the nascent peptide moves down the exit tunnel
located in the larger ribosomal subunit (a simplified version of this channel is highlighted in red).
There is evidence that the secondary structural elements of TMHs in ion channels are pre-folded in
the ribosomal exit tunnel [348–351]. The ribosome hands the nascent chain to the translocon. Once
the plug moves to expose the translocon pore, the translocon has a hydrophobic core which regulates
the integration of the TMH via the lateral gate [128].

[253]. There are several other SecM proteins of other strengths from various bacterial
species [354]. Therefore APs are a technique that can be used to measure precise
forces acting on a specific part of the nascent chain during co-translational membrane
protein integration allowing the study of TMP kinetics during insertion and folding.
The force profile of a single residue can now be obtained in vivo [253]. In an idealised
TMH segment composed of alanine and leucine being inserted into E. coli membrane

4.2. INTRODUCTION

143

through SecM AP with SDS-PAGE, hydrophobicity is more able to overcome the arrest peptide when it is near the N-terminal of a TMH with N-terminal-inside topology
[253]. This could be either the TMH finally coming into contact with the cytoplasmic
face of the lipid bilayer, or an interaction between the N-terminal and the tip of the
lateral gate as previously shown in Sec61; part of a pre-integration TMH interrogation
[355].
The ribosome passes the nascent chain to the Sec translocon machinery. Like the
ribosome, the Sec pathway is widely conserved across life [153]. Sec61 (in eukaryotes)
and SecYEG (in prokaryotes) translocate hydrophilic peptides across a membrane
whilst also integrating sufficiently hydrophobic sequences to the membrane [40, 128,
156, 157]. SecY in prokaryotes or Sec61α in eukaryotes are TMPs with 10 TMHs
that perform the translocation. These proteins have a lumenal plug [164, 165] and
a constricted core to prevent permeability whilst the protein is idle [128]. During
translocation the plug moves out of the pore [356], the channel is in an open unconstricted state [128], and a lateral gate of TMH2 and TMH7 [181] periodically opens
to release a TMH to the membrane environment if sufficient hydrophobicity is present
in the nascent peptide chain [128, 179, 180](Figure 4.1).

4.2.2

Cooperative transmembrane helix insertion by the
translocon-ribosome complex

Early evidence of secondary structure folding of the Kv ion channel showed that the S6
TMH (referred to later in this study as TMH6) is indeed compacted in the ribosomal
tunnel by using pegylation and calmodulation of the tagged cysteine-scanned S6 TMS
[348]. Furthermore, accessibility assays and an improved intramolecular cross-linking
assay showed that the helical transmembrane S3-S4 hairpin (the “paddle”) of a voltagegated potassium (Kv ) forms in the ribosome tunnel [350]. Ribosomal folding of the
TMHs in Kv 1.3, a potassium channel, is maintained in the translocon [349]. Therefore,
some of the final structural folding of the voltage sensor domain occur within the
ribosomal exit tunnel.
Furthermore, it has recently been suggested that larger structures fold as the ribosomal exit tunnel widens [351]. This size-dependent folding was observed by using the

144

CHAPTER 4. SEQUENCE ANALYSIS OF POLARITY...

SecM translational AP. Two ribosome mutants of deleted uL23 and a uL24 variant
that does not contain the hairpin loop were compared to the wild-type (uL23 is a
globular domain buried within the ribosome that is close to the exit tunnel and uL24
contains a hairpin loop that obstructs the tunnel exit). These mutants resulted in
reduced space for folding in the ribosomal exit tunnel at separate points. ADR1a, a
29- residue zinc finger motif, was found to fold deeper in uL23 mutant than the wildtype and the uL24. Two domain folds of 109 and 89 residues in length folded deeper
in the uL24 mutant than in the wild-type and the uL23 mutant. This shows that
cotranslational folding can occur in the ribosomal exit tunnel once sufficient space is
available [351].
The ribosomal tunnel also slows down elongation of positively charged peptide
segments. This is attributed to the sporadic negative patches within the ribosomal
exit tunnel [357].
The ribosome clearly has the potential to pre-fold secondary structures and some
motifs before translocon insertion into the membrane.
Multiple TMHs in a nascent protein can be associated with the eukaryotic translocon simultaneously. It was shown that TMHs can stay in association with the translocon in order to mediate integration of downstream TMHs demonstrated by crosslinking analysis in Aquaporin-4 [358] and a ligand-gated ion channel, the P2X2 receptor channel [359]. Not only this, but it was shown that there is a direct interaction
between the TMHs; more recently APs were used to show pulling forces between a
TMH and more C-terminally located TMH during the C-terminal TMH membrane
partitioning from the translocon in vivo [360]. This could be facilitated during the
probing of a TMH from the translocon as the lateral gate “cracks” open in an intermediate stage before the TMH satisfies the full hydrophobic requirements to open the
gate fully, an intermediate stage observed in a SecY crystal structure [179].
Sixteen marginally hydrophobic TMHs were screened and revealed that independently they did not efficiently insert via the translocon into the ER [267]. The study
showed that in the presence of the neighbouring loops and TMHs, insertion often became sufficient with the exception of two of the marginally hydrophobic TMHs. The
orientational preference of neighbouring TMHs was shown to increase insertion efficiency into the ER of a marginally hydrophobic TMH [268]. A glycosylation study

4.3. METHODS

145

of three relatively polar TMHs revealed that proteins have even evolved to exhibit
stronger “positive-inside” tendencies when following another TMH [361]. Subsequent
TMHs with a strong topological preference (increased arginine presence) increase the
insertion potential of marginally hydrophobic TMHs, however, no specific interaction
between the TMHs was responsible. The nascent polypeptide in the cavity between
the ribosome and the translocon can alter the hydrophobic threshold of the translocon which therefore is dynamic and influenced by downstream flexible sequences [362].
Sequence context is clearly an important factor for the efficient insertion of marginally
hydrophobic TMHs. These TMHs rely on other components of their protein in order
to incorporate into the membrane.

4.2.3

Aims

Marginally hydrophobic TMHs often miss the energetic requirements of insertion via
the translocon [74]. In the literature, there are examples of cooperative TMH interactions during insertion of marginally hydrophobic TMHs in 6TMH ion channels and
GPCRs [40, 268, 350, 358–361, 363–367].
In this chapter, we investigate the relationship between biochemical sequence
derived properties such as TMH hydrophobicity and TMSOC z-score [44, 45] of
marginally hydrophobic TMHs and their sequential neighbouring TMHs to identify
family wide patterns that may use a cooperative TMH insertion folding mechanism.

4.3

Methods

4.3.1

Datasets

Membrane protein families
Datasets were attained by querying the UniProt database for controlled vocabulary
keywords (Table 4.1) [75]. After filtering out redundant proteins using UniRef50, the
datasets were stratified according to the total number of TMHs per protein. Only
families with a total N >20 were included in the analysis, for example, there are ion
channels with 12 TMHs, but only 16 examples fitted the criteria and were omitted
from the study. Analysis can only be done if the number of TMHs is conserved across

146

CHAPTER 4. SEQUENCE ANALYSIS OF POLARITY...

the family.
Table 4.1: Dataset sizes of common transmembrane protein families of transporters and
channels. The type column refers to the family name. The keyword identifier column refers to the
specific query term that UniProt uses as the controlled vocabulary identifier. UniProt hits refers to
the number of SwissProt hits, which are manually curated, however, the SwissProt and TrEMBL
hits are also given in the brackets. The UniRef column denotes how many representative sequences
resulted from the UniProt hits, again with the number in the brackets including the TrEMBL hits.
Because UniRef often returns representatives of splice isoforms as a separate hit, the lists were reuploaded to UniProtKB to get the final hit. The lists are then stratified by the total number of
TMHs.
Type
Ion channel

KW
0407

UniProt Hits
2452 (176417)

UniRef
912 (21024)

Records
882

Ligand-gated ion channel
Voltage-gated channel
Calcium channel
Potassium channel
Sugar transport
Chloride channel
Ion transport

1071
0851
0107
0631
0762
0869
0406

452 (14051)
683 (27,662)
272 (5056)
337 (11852)
1217 (107721)
303 (3697)
10264 (443600)

189 (3055)
285 (3876)
128 (1003)
160 (1893)
467 (16748)
117 (1039)
3070 (42259)

185
267
120
152
464
117
3023

TMHs
4
6
24
6
6
6
6
12
4
4
6
8
10
12
24

Final hits
234
188
34
33
97
38
88
134
50
338
390
162
189
292
35

GPCR subfamilies
The 7TMR list is a list distributed by UniProt containing GPCRs available at
http://www.uniprot.org/docs/7tmrlist.txt [75]. The entire list contains 3115
UniProt IDs which mapped to 3092 records on the date of download (12/9/2017).
After removing redundant records using UniRef50 to identify cluster representatives,
1142 records made the final dataset. The original list is also sub categorised by function. Here, we also looked at opsins, the T2R taste receptors, frizzled/smooth GPCRs,
metabotropic glutamate receptors (Family C), and fungal mating proteins.
218 opsin records were mapped through UniRef50 to 40 records. 211 T2R taste
receptors were mapped to 45 records after UniRef50. There were 82 records for frizzled/smooth, of which 41 were cluster representatives after querying with UniRef50.
114 were metabotropic glutamate receptors which became 44 after redundancy removal. In serotonin, 91 original records were represented by 28 records. There were 13
fungal mating proteins, represented by 9 after UniRef50. 560 olfactory records were
mapped to 274 records after Uniref50.

4.3. METHODS

147

After limiting all these sets to records containing 7 TRANSMEM features, there
were 39 opsins, 38 frizzled smooth GPCRs, 9 fungal mating GPCRs, 41 metabolic
glutamate receptors, 45 t2r taste receptor proteins, and 263 olfactory receptor GPCR
proteins. TRANSMEM is an annotation by UniProt denoting a TMH evidenced by
experimental documentation or a robust TMH prediction method [75].

4.3.2
77153

Gene ontology
protein

records

matched

ANDreviewed:yes in UniProt [75].

the

query

annotation:(type:transmem)

30045 records of these represented those

proteins after submitting the list to UniRef50. To gain an approximate idea of what
proteins may have cooperative TMHs, we submitted that non-redundant SwissProt
transmembrane dataset to PANTHER and used their pie chart visualisation [368].
We also submitted a constricted list of the non-redundant SwissProt transmembrane
dataset to PANTHER by sorting the list according to proteins with sequentially
adjacent TMHs that had the greatest difference in Kyte & Doolittle hydrophobicity
[73]. This consisted of the top 102 pairs from a list of 101604 (the top 0.1%) TMH
pairs from 87 unique UniProt records. In PANTHER, these hit 59 genes with 58
functions.

4.3.3

Complexity and hydrophobic estimation

Primarily we used the Kyte & Doolittle scale in this work [73]. We verified this scale
with the Eisenberg scale [144], the Hessa biological scale [116], and the White and
Wimley scale [139]. The Eisenberg scale is potentially double counted for many of these
proteins since it is used by UniProt during their automatic TMH annotation where
experimental evidence is not available. The Hessa biological scale originally reported
positional dependencies that affected the hydrophobicity values which were refined in
a later version of the scale [74], however, this later version was not programmatically
available to us.
When stratifying TMHs by hydrophobicity and sequence information entropy, there
is a clear distinction between anchoring TMHs and those with a function beyond anchoring. The TMSOC z-score quantifies this relationship and is calculated by equation

148

CHAPTER 4. SEQUENCE ANALYSIS OF POLARITY...

4.1.

"
z(xΦ , xc ) = (−1)s

2

2

(xΦ − µΦ )
(xc − µc )
+
2
σφ
σc2

#
(4.1)

xc and xΦ are moving window averages of c, the sequence entropy determined by
SEG [146, 150]. Φ is the White and Wimley hydrophobicity [139] for a given segment and µ and σ are the mean and standard deviation of the sequence entropy and
hydrophobicity of the functional TMH sets. s is set to either zero or one depending on the the normal to the regression line between normalised hydrophobicity and
complexity [44] (Equation 1.3, equation 1.4)
Broadly speaking, the information theory entropy of a string of characters can be
defined as in equation 4.2, which we use to calculate the sequence information entropy
of TMH sequences.

H (S) = −

n
X

P (si ) log2 P (si )

(4.2)

i=1

Where H is the entropy of a sequence (S) with values {s1 , . . . , sn }, and P (si ) is
the probability of a character i through each position in S. This allows us to quantify
the average relative information density held within a string of information [151].

4.3.4

Statistics

We used the Welch’s t-test to scrutinise the means of the datasets without assuming
equal variance. To examine any differences in the skew of the datasets, the Kruskal
Wallis, and the Kolmogorov Smirnov tests were used. All these tests were performed
using the Python scipy package [369].

4.3.5

Availability of materials

The scripts and datasets associated with this study can be accessed at https://
github.com/JamesABaker/multipass_tmh_analysis.

4.4. RESULTS AND DISCUSSION

4.4
4.4.1

149

Results and discussion
Large contrasts in transmembrane helix hydrophobicity
occur in channels and receptors

Large differences in sequentially adjacent TMHs could indicate cooperative insertion
between a marginally hydrophobic TMH and a more typically hydrophobic TMH. In
order to identify which type of TMP may contain cooperative TMHs, the hydrophobicities for TMHs from 30045 records in a non-redundant version of SwissProt were
calculated according to the Kyte & Dolittle hydropathy scale [73]. The absolute difference between each sequentially adjacent TMH was calculated. The full list, alongside
a list of the 0.1% highest hydrophobic discrepancy TMH pairs, were submitted to
PANTHER [368]. We found that transporters were by far the most abundant in the
0.1% sub-list relative to the global TMP list, however, it should be considered that
“transporters” in this sense also includes both ion transporters and ion channels (Figure 4.3). Among the other prevalent functions are receptors, binding proteins, and
signal transducers. All these lists were predominantly made up of GPCRs.
A) Global TMP list
antioxidant activity (GO:0016209)
transporter activity
(GO:0005215)

B) List of top 0.1%
hydrophobic pair difference

binding
(GO:0005488)

structural molecule
activity (GO:0005198)
signal transducer
activity (GO:0004871)

catalytic activity
(GO:0003824)

receptor activity
(GO:0004872)
channel regulator
activity (GO:0016247)

Figure 4.3: Pie charts of a non-redundant list of transmembrane proteins compared to
a list of transmembrane proteins containing the most hydrophobically different transmembrane helix pairs. The output of PANTHER Gene ontology server of A) a non-redundant list
of all TMPs in SwissProt and B) a list of the TMPs with the highest (0.1%) hydrophobic discrepancy between adjacent TMHs. Note that the proportions of binding TMPs stay roughly the same,
receptors and signal transducers reduce slightly, catalytic proteins are reduced by two thirds and
transporter activity proteins more than double in proportion.

150

CHAPTER 4. SEQUENCE ANALYSIS OF POLARITY...

The relative proportion of GO terms for catalytic function decreased when considering the top 0.1% of hydrophobic difference between sequential TMH pairs. This
trend suggests that a catalytic function of the TMP may not be associated with a
high hydrophobic discrepancy between TMHs within the protein. Figure 4.3 gives us
an idea of search space where highly polar TMHs are proximally adjacent to typically
hydrophobic TMHs; GPCRs and membrane transporters.

4.4.2

GPCRs contain conserved relatively polar TMH7,
which follows the typically hydrophobic TMH6

GPCRs are a diverse family of membrane surface receptors with 7 TMH segments.
GPCRs have long been known to be overrepresented among genomes [370]. They have
adapted to respond to a wide range of specific signals ranging from macromolecules
to photons. The specific signal triggers a conformational change of the GPCR that
is translated across the membrane. GPCRs have been associated with tumorigenesis
[371], metastasis [372] and in cancers [373] and are a potential target for therapies
[374]. Their ubiquitous presence in cellular life and medical relevance makes them an
important topic of study.
Here, we structurally aligned 7 structures of monomeric GPCRs using PyMol [375];
PDB codes 1u19 (rhodopsin) [376], 2z73 (rhodopsin) [377], 2vt4 (β1-adrenergic receptor) [378], 2lnl (CXCR1, rhodopsin-like) [156], 4mbs (CCR5, chemokine receptor)
[379], 4xt1 (viral chemokine) [380], and 4ea3 (opioid receptor) [381] (Figure 4.4A).
Even in a relatively small dataset of functionally varied GPCRs, we can see structural conservation of the TMH arrangement (Figure 4.4A Figure 4.4C). Hydrophobic
patterns are hard to identify structurally beyond the clear membrane boundary where
we would expect to see hydrophobic residues in the TMH (Figure 4.4B). TMSOC [44,
45] is an algorithm that takes into account the White and Wimley hydrophobicity
[139] and the information entropy of the TMH sequence. Note that the scale is the
inverse of the Kyte & Doolittle scale, i.e a more negative TMSOC z-score indicates
a higher hydrophobic contribution to the TMSOC z-score. The resulting z-score has
been shown to be able to scrutinise between TMHs that serve as anchors and those
that have a function beyond TMP anchoring. The lower the z-score (low sequence

4.4. RESULTS AND DISCUSSION
A)

151
B)

180°

180°

D)

6
C)
5

7

1
2

3

4
1

E)

2

3

4

5

6

7

F)

Figure 4.4: The hydrophobicity and complexity of GPCR transmembrane helices. A)
A cartoon of 7 structurally aligned GPCRs of various molecular functions coloured according to the
residue sequence position. Blue is the N terminal residue, working through the rainbow until red at
the C terminal position. B) The same 7 GPCRs aligned structurally but instead coloured by the
Eisenberg hydrophobic scale [144] at 50% transparency with red being hydrophobic and white being
polar. C) A top-down (birds-eye) version of the alignment coloured similarly to A taken at a slice
halfway through the TMHs. D) On the vertical axis is the TMSOC z-score of each of the helices and
on the horizontal axis is the TMH number. The points are coloured by the corresponding structure.
E) The axes are similar to D, however, the datasets are from non-redundant UniProt GPCR sequence
datasets from 1016 GPCR records that contained the annotation for 7 TMHs. The data is represented
by a violin plot, the thickness indicating the distribution of the data. The mean average is the dash
in black, and the median is the dash in orange. F) As in E, however, the vertical axis is Kyte &
Doolittle hydrophobicity [73].

information entropy and high hydrophobicity), the more likelihood of the TMH being
solely a membrane anchor. When considering the functional/anchoring potential of
these GPCR TMHs using TMSOC, we see trends among the structures with an average of 1.96 for TMH1, 2.30 for TMH2, 2.43 for TMH3, 2.34 for TMH4, 2.05 for TMH5,

152

CHAPTER 4. SEQUENCE ANALYSIS OF POLARITY...

2.24 for TMH 6 and 2.83 for TMH7 (Figure 4.4D). TMH3 and TMH7 are therefore
the most likely to contain function beyond anchorage, whilst the low z-scores of TMH1
and TMH5 indicate they are more optimal anchors [227].
When we consider much larger sequence datasets containing 1016 TMHs at each
TMH number from non-redundant GPCRs obtained from UniProt, the trend remains
the same; that TMH3 (mean z-score of 2.52) and TMH7 (mean z-score of 2.55) have the
highest z-score, whilst TMH1 (mean z-score of 2.09) and TMH5 (mean z-score of 2.20)
have the lowest (Figure 4.4E). To investigate the statistical differences between the zscore of the TMHs, we applied the Kruskal Wallis and 2-sample Kolmogorov Smirnov
tests from each TMH number to each other TMH number. TMH5 and TMH7 are
statistically significantly distinct in terms of TMSOC z-score (Welch’s t-test p-value
= 1.57E-146, Kruskal Wallis p-value = 1.66E-132, Kolmogorov Smirnov p-value =
1.38E-105) as are TMH6 and TMH7 (p-value = 1.48E-51, Kruskal Wallis p-value =
4.96E-52, Kolmogorov Smirnov p-value = 1.74E-37). This is also mirrored in the
relationship between the preceding TMHs of TMH3. TMH1 and TMH3 are distinct
(Welch’s t-test p-value = 3.12E-229, Kruskal Wallis p-value = 1.06E-192, Kolmogorov
Smirnov p-value = 1.06E-174) and so are TMH2 and TMH3 (Welch’s t-test p-value =
3.66E-39, Kruskal Wallis p-value = 2.70E-36, Kolmogorov Smirnov p-value = 1.06E-31). We also find it difficult to observe any significant differences between TMH2
and TMH6 (Welch’s t-test p-value >0.99, Kruskal Wallis p-value = 0.99, Kolmogorov
Smirnov p-value = 0.07).
As expected, these trends are mirrored in the Kyte & Doolittle [73] hydrophobicity
plots. Where there is a high TMSOC z-score, there is a low hydrophobicity score.
However, whereas TMH6 (mean hydrophobicity = 1.69) to TMH7 (mean hydrophobicity = 1.13) are very different (Welch’s t-test p-value = 3.95E-163), the difference
between TMH2 (mean hydrophobicity = 1.42) and TMH3 (mean hydrophobicity =
1.39) is less stark (Welch’s t-test p-value = 0.12). Even though there is a discrepancy between the scales, the significant change in hydrophobicity between TMH6 and
TMH7 can be seen consistently across the Eisenberg consensus scale [144] (Welch’s
t-test p-value = 5.61E-130), the White and Wimley biophysical scale [139] (Welch’s
t-test p-value = 5.21E-62), and the von Heijne/Hessa biological scale [116] (Welch’s
t-test p-value = 3.56E-210) emphasising the significance of this particular TMH pair

4.4. RESULTS AND DISCUSSION

153

(Figure 4.5).
A)

B)

C)

Figure 4.5: The hydrophobic difference observed between TMH6 and TMH7 in GPCRs
is not due to the choice of hydrophobic scale. Three different scales were applied to the same
dataset used in figure 4.4F to verify the observed differences were not caused by the choice of scale.
A) The White and Wimley biophysical scale [139] B) The Eisenberg consensus scale [144]. C) The
von Heijne/Hessa biological scale [116].

In order to determine if this is conserved through the GPCR subfamilies, a variety of GPCRs with different functions were examined in terms of Kyte & Doolittle
hydrophobicity [73]. When considering different GPCR subfamilies, we see a consistent difference between the TMH hydrophobicities of TMH6 and TMH7 besides
metabolic glutamate GPCRs (Welch’s t-test p-value = 0.90) and Frizzled-smooth
GPCRs (Welch’s t-test p-value = 0.07) (Figure 4.6). For opsins, Welch’s t-test p-value is 5.62E-15 between TMH6 and TMH7. After restricting the dataset to only
those records with 7 TMHs, fungal mating GPCRs were reduced to a set of 9 proteins, however as a trend, they appear to be very typical GPCRs in terms of TMH
hydrophobicity. They follow a very similar pattern to the much more popular olfactory
GPCRs (N=263) for which the TMH6 to TMH7 Welch’s t-test p-value was 4.73E-35.

154

CHAPTER 4. SEQUENCE ANALYSIS OF POLARITY...

For taste receptors, there are much bigger step changes between sequentially adjacent
TMHs than between TMH6 to TMH7. But there is still a real statistical difference
between TMH6 and TMH7 (Welch’s t-test p-value = 2.89E-12).
A)

Opsins

C) Fungal mating

E) Metabolic Glutamate

B) Frizzled smooth

D) Olfactory

F) Taste receptors

Figure 4.6: The hydrophobicity of transmembrane helices in GPCR subfamilies. Violin
plots of the Kyte & Doolittle hydrophobicity [73] of TMHs of GPCRs after selecting a non-exhaustive
sample of GPCR subfamilies with a wide variety of molecular functions. A) Opsins (N=39). B)
Frizzled-smooth GPCRs (N=38). C) Fungal mating GPCR proteins (N=9). D) Olfactory receptor
proteins (N=263). E) Metabolic glutamate receptor (N=41). F) T2R taste receptors (N=45)

Opsins are a group of light-sensitive GPCRs that exhibit the TMH6-TMH7 hydrophobic discrepancy (Figure 4.6A). It was shown by cross-linking studies that opsin

4.4. RESULTS AND DISCUSSION

155

TMHs 5-7 are retained in the ER translocon and only partition into the membrane
once biosynthesis is complete [363]. The timing of this partitioning is controlled by the
hydrophobicity of the TMH, not protein length or the relative position of the TMH
within the protein. Although artificially extending the C-terminal did not result in
the release of the TMHs, by replacing native TMH7 with a more hydrophobic TMH,
the speed of insertion was decreased. TMHs 1-4 are inserted independently, and the
5-7 TMHs partition into the membrane at the same time [363].
Lowering the difference in hydrophobicity between TMH6 and TMH7 impedes
efficient insertion of the GPCR [363], however, one would expect the terminal TMH to
be more hydrophobic rather than relatively polar [361]. This hydrophobic difference is
highly conserved throughout GPCRs generally (Figure 4.4F), and especially in many
of the subfamilies (Figure 4.6A, C, D, F). We believe that the hydrophobic difference
of TMH6 and TMH7 is highly important to the biogenesis of GPCRs. This could be
evidence for widespread biologically employed TMH cooperation during translocation.
A typically hydrophobic TMH remains in association with the translocon until the
relatively polar TMH is processed by the insertion machinery.

4.4.3

6TMH ion channels contain polar-hydrophobic transmembrane helix pairs/groups indicative of conserved
cooperative insertion

Another group of proteins that featured heavily on the list of top TMH pairs with
a high hydrophobic discrepancy were ion channels (Figure 4.3B). Ion channels are a
much more structurally varied class of TMP than GPCRs, varying greatly in their
number of TMHs, so a summary structural alignment equivalent to Figure 4.4A, B,
and C was not possible.
Voltage-gated ion channels are another example of cooperative TMH insertion.
The 3rd and 4th TMH of the potassium channel (shaker family) has been shown
to insert either sequentially or cooperatively [40, 366]. This is especially notable
in the case of KAT1, that is a plant Kv channel that is thought to mediate longterm potassium influx into guard cells causing the stomata to open. In the case
of KAT1, N-glycosylation of various mutant fusion KAT1 constructs revealed that

156

CHAPTER 4. SEQUENCE ANALYSIS OF POLARITY...

there is no choice of sequential insertion since TMH3 and TMH4 have no insertion
potential and no topogenic functions themselves [364, 365]. In TMH4 this is due to
the charged residues making it relatively polar. However, previous experimentation
in Kv 1.3 had found that while TMH4 did not initiate insertion, it did have insertion
potential and that when constructs contained multiple TMHs, membrane insertion
efficiency increased [367]. Without the ability to stop the translation through the
translocon and form a TMH, it was suggested that a different mechanism was needed
than classic sequential insertion, and even that TMH3 and TMH4 are integrated by
the translocon at the same time post-translationally, i.e the TMHs are folded prior
to insertion [365]. They achieve this in part because the previous TMHs, TMH1 and
TMH2, form a firm “base” within the membrane environment [365].
In the 6TMH class of ion channels, we observe a conserved difference in the relatively polar TMH4 known to carry charged residues (Figure 4.7A). When using the
Kyte & Doolittle hydrophobic scale, the hydrophobic step change between TMH3
(mean hydrophobicity of 1.34) and TMH4 (mean hydrophobicity of 1.01) is statistically significantly different (Welch’s t-test p-value = 1.83E-8). This is the most significant step change besides TMH4 to TMH5 (mean hydrophobicity of 1.77, Welch’s
t-test p-value = 1.88E-31). TMH5, the pore-forming region, and TMH6 can be integrated independently into the membrane, but TMH3 and TMH4, at least in the case
of KAT1, cannot [364]. However, there is strong evidence that TMH3 and TMH4
cooperate with one another during membrane integration and are inserted at the same
time [40, 364–366]. With this in mind, the hydrophobic gap here indicates a conserved
cooperative role between TMH3 and TMH4 rather than TMH4 and TMH5.
This statistical difference between the hydrophobicity of TMH3 and TMH4 holds
true for voltage-gated ion channels (Welch’s t-test p-value = 1.57E-17) (Figure 4.7C).
However, the pattern is more obscure in ligand-gated ion channels (Welch’s t-test P–
value = 5.32E-2). The difference does appear to match the general pattern, however,
albeit as a trend, in the 33 records; there is a decrease in the mean hydrophobicity
from 1.31 to 1.03 (Figure 4.7D).
The TMH3 TMH4 hydrophobic step also appears to be independent of the ion
being transported. For calcium ion channels the drop from TMH3 (mean hydrophobicity 1.52) to TMH4 (mean hydrophobicity 1.21) is significant (Welch’s t-test p-value

4.4. RESULTS AND DISCUSSION
A)

157
B)

Ion channels

Ion channels

C)

D)

Ligand gated ion
channels

Voltage gated ion
channels

E)

F)

Calcium channels

G)

NH2

Potassium channels

+ve Charge

COOH

P

Figure 4.7: The hydrophobicity of transmembrane helices in ion channels. A) The Kyte
& Doolittle TMH hydrophobicity of TMHs according to their TMH number of the 6TMH class of ion
channels. N=188 (The total number of TMHs analysed is 1128). The data is presented as a violin
plot. The thickness of the bar indicates a higher density of points at that hydrophobic value. B) A
similar violin plot as in (A), however with the 24 TMH class of ion channel. N=35 (A total of 840
TMHs analysed). C) A subset of (A) with the keyword of ion channels restricted to voltage-gated
ion channels. N=97. D) A subset of (A) with the keyword of ion channels restricted to ligand-gated
channels. N=33. E) A subset of (A) with the keyword of ion channels restricted to calcium channels.
N=38. F) A subset of (A) with the keyword of ion channels restricted to potassium channels. N=88.
G) The general schematic of an ion channel. P indicates the pore-forming area of the protein, which
is integrated into the membrane but is not a membrane-spanning unit. In the case of the 24 class,
this pattern is repeated 4 times.

158

CHAPTER 4. SEQUENCE ANALYSIS OF POLARITY...

= 3.57E-3) given the relatively low sample size (38 protein records) (Figure 4.7E). In
potassium channels, the difference is even greater between TMH3 (mean hydrophobicity of 1.37) and TMH4 (mean hydrophobicity of 0.61) (Welch’s t-test p-value =
2.52E-19) in potassium ion channels (Figure 4.7F).
The observed difference between TMH3 and TMH4 is somewhat independent of
the scale chosen (Figure 4.8), however, the Kyte & Doolittle scale is the most sensitive
to the hydrophobic differences between TMH3 and TMH4. Statistical significance was
found in the difference between TMH3 and TMH4 in the Eisenberg scale (Welch’s t-test
p-value = 4.74E-23) and the White and Wimley scale (Welch’s t-test p-value = 1.26E3). The Hessa/von Heijne biological scale was particularly insensitive to the mean
difference (Welch’s t-test p-value = 7.56E-2), but can scrutinise some difference in the
skews of the data (Kolmogorov Smirnov test p-value = 1.56E-4, Kruskal Wallis p-value
= 4.00E-2) indicating that whilst biophysically the TMHs are distinct, the translocon
may not readily scrutinise the hydrophobicity between the two. All the scales however
corroborate the large hydrophobic difference between TMH4 and TMH5 (Welch’s t-test p-values for Hessa = 1.23E-35, White and Wimley = 4.82E-23, Eisenberg =
5.72E-41)
The 24TMH class is essentially a modular variant of the 6TMH class and has
even more extreme discrepancies between the equivalent 4 TMHs (Figure 4.7B). In
the first cluster, TMH3 (mean Kyte & Doolittle hydrophobicity = 0.96) to TMH4
(mean hydrophobicity = 0.47) is different (Welch’s t-test p-value = 5.16E-6), but even
more so is the step up from TMH4 to TMH5 (hydrophobicity = 2.26, Welch’s t-test
p-value = 5.16E-32). The preceding step is weaker, and the proceeding step is greater
throughout the 6TMH “modules”. A similar step change is found between TMH9 and
TMH10 (p-value = 3.02E-13) and TMH10 to TMH11 (p-value = 5.51E-44), TMH15
to TMH16 (p-value = 1.79E-23), TMH16 to TMH17 (p-value = 1.67E-27), TMH21 to
TMH22 (p-value = 1.64E-9), and TMH22 to TMH23 (p-value = 3.42E-31).
As a trend, the TMSOC z-score was not sensitive to detecting function in the
extremely hydrophobic TMH4 (Figure 4.9). This is due to the other component of the
z-score, the sequence information entropy from SEG [44, 45, 146, 150]. The relative
probability of a polar/charged residue is so high in TMH4 that it apparently has
the same expectancy as the hydrophobic residues, and the “simple” entropy mutes

4.4. RESULTS AND DISCUSSION
A)

159
B)

C)

Figure 4.8: The hydrophobic difference observed between TMH4 and the neighbouring
transmembrane helices in 6TMH ion channels is not due to the choice of hydrophobic
scale. Three different scales were applied to the same dataset used in figure 4.7A to verify the
observed differences were not caused by the choice of scale. A) The White and Wimley biophysical
scale [139] B) The Eisenberg consensus scale [144]. C) The von Heijne/Hessa biological scale [116].

somewhat the effect of the extreme hydrophobicity. This is compounded by the fact
that the White and Wimley scale (used to generate the z-score) is already less sensitive
to the relative polarity difference between TMH3 and TMH4 than the Kyte & Doolittle
scale (Figure 4.7A, Figure 4.8A).

4.4.4

The prevalence of the high hydrophobic discrepancy of
transmembrane helices amongst other common ‘transporter’ transmembrane protein classes

Whilst the gene ontology recognises “transporters” as having disproportionately more
highly different hydrophobic sequential pairs, the list is predominantly ion channels,
which are molecularly and functionally different from ion transporters. It also appears

160
A)

CHAPTER 4. SEQUENCE ANALYSIS OF POLARITY...
B)

Figure 4.9: Sequence entropy is unsuitable for assessing function in TMH4 of ion channels. A) The TMSOC z-score [44, 45] of TMHs from the ion channel dataset used in Figure 4.7A.
B) The sequence information entropy of those same TMHs.

that ion transporters have no conserved hydrophobicity discrepancy between any of
their TMHs for the 4TMH, 6TMH, 8TMH, or 10TMH classes (Figure 4.10 C, D, E,
F, G).
Another large family of transporters are the sugar transporters. In terms of the
hydrophobic discrepancy between sequentially adjacent TMH pairs, sugar transporters
also have a few peaks and dips at TMH6 and TMH11 (Figure 4.10A). TMH5 (mean
hydrophobicity = 1.32) to TMH6 (mean hydrophobicity = 1.81) (Welch’s t-test P–
value = 3.09E-19) was less significant than TMH6 to TMH7 (mean hydrophobicity =
1.23) (Welch’s t-test p-value = 8.75E-29). TMH10 (mean hydrophobicity = 1.66) to
TMH11 (mean hydrophobicity = 1.23) (Welch’s t-test p-value = 1.03E-14) was also
less significant than TMH11 to TMH12 (mean hydrophobicity = 1.84) (Welch’s t-test
p-value = 3.22E-22).
The ion channel 4TMH class also has a discrepancy between TMH2 (Kyte & Doolittle hydrophobicity = 1.49) and TMH3 (hydrophobicity = 1.81) (Welch’s t-test P–value
= 4.42E-10, from 234 records) (Figure 4.10C). The largest family within the 4TMH
class is the chloride channel 4TMH family. A jump of hydrophobicity is observed
in TMH2 (hydrophobicity = 1.05) to TMH3 (hydrophobicity = 1.50) (Welch’s t-test
P–value = 5.34E-5, from 50 records) (Figure 4.10B).

4.5. SUMMARY

161
E)

A)

Sugar Transporters

Ion transport

F)

B)

Ion transport

Chloride channel

G)

C)

Ion channel

D)

Ion transport

H)

Ion transport

Ion transport

Figure 4.10: High polarity discrepancy between sequentially adjacent transmembrane
helices is not present in all transmembrane protein transporter families. The mean average
Kyte & Doolittle hydrophobic scores of a TMHs from various families and classes of membrane
transporter proteins. Datasets are represented as a violin plot where thickness denotes the relative
density of the population distribution. A) A dataset of 134 sugar transporters containing 12 TMHs.
B) A dataset of 50 chloride channels containing 4 TMHs. C) 234 ion channels with 4 TMHs. D-H)
Ion transporters restricted to the number of TMHs on the horizontal axis. Sample sizes can be found
in table 4.1

4.5

Summary

Marginally hydrophobic TMHs are often essential physiological components for carrying out complex intra-membrane tasks beyond membrane anchorage, and widely

162

CHAPTER 4. SEQUENCE ANALYSIS OF POLARITY...

prevalent among TMPs [74]. However, this presents a significant biophysical and biological challenge; how to incorporate and maintain such necessarily relatively polar
TMHs into the membrane. We show that among certain classes of TMPs that this is
a conserved issue. It was recently discovered that the EMC protein assists the translocon with the marginally hydrophobic TMHs, particularly those containing charged
residues [24]. There is also evidence to suggest that multipass TMH arrangements
work cooperatively during translocation and insertion to achieve correct folding of
marginally hydrophobic TMHs [40, 268, 350, 358–361, 363–367] (Figure 4.11).
Less hydrophobic

More hydrophobic

Translocon
A)

Lateral Gate B)

TMHs during
sequential
insertion

C)

TMH hydrophobic-polar pair during
cooperative insertion.
Lateral Gate

Figure 4.11: A cartoon of potential cooperative transmembrane helix insertion methods. A) Typically hydrophobic TMHs remain in association with the translocon machinery until
the marginally hydrophobic TMH is released into the membrane. This was observed in an opsin
protein [363]. B) TMH-TMH interactions allow the marginally hydrophobic TMH to shield the polar
residues. The marginally hydrophobic TMH can associate with the typical hydrophobic TMH in the
ribosomal exit tunnel [350] or in the translocon tunnel [366]. This was observed in a potassium ion
channel [40, 350, 366].

Here, we have shown that high discrepancy of TMSOC z-score (simple-complex
score) [44, 45], and several hydrophobicity scales [73, 116, 139, 144], between sequentially adjacent TMHs corroborates TMH groups known to insert cooperatively [40,
363, 365, 366]. The hydrophobic relationships between these cooperatively inserted
TMHs may be conserved throughout their larger families indicating that cooperative
insertion may be a routine mechanism for the insertion of some families of TMPs.
So far, this method has shown that the TMH7 of GPCRs is a conserved marginally
hydrophobic TMH with the typically hydrophobic TMH6 preceding it. In opsins,

4.5. SUMMARY

163

TMH5-TMH7 are incorporated into the membrane together simultaneously by staying in association with the translocon until TMH7 emerges [363]. We see this pattern
across several GPCR sub-families and the GPCR family as a whole.
Furthermore, we found that the charged TMH4 of 6TMH ion channels has a typically hydrophobic TMH3 preceding it and a more hydrophobic TMH5 following it.
Indeed there is evidence that TMH3 and TMH4 cannot incorporate alone [365], that
TMH3 and TMH4 can insert as a pair from the translocon simultaneously [40, 366].
Advanced cross-linking experiments, single molecule cryoEM, and APs can probe
energetics and TMH-TMH interactions and associations dynamically during ribosomal
translation and membrane insertion. These methods could explore the TMH-TMH relationships such as cooperation during insertion in more detail with temporal contextualisation. Understanding the biogenesis of TMPs is essential to solving misfolding in
disease phenotypes, and for developing viable TMP chassis in synthetic biology [382].

Chapter 5
Conclusions and outlook
As the evidence of positive residues preferentially localising to the inside the cytoplasm
emerged during the late 1980s [113], so did the idea of negative residues working in
concert with the positive residues for TMH orientation. It was shown that removing
a single lysine residue reversed the topology of a model E. coli protein, whereas much
higher numbers of negatively charged residues are needed to reverse topology [248].
However, there was no conclusive evidence in the literature for an opposing negatively
charged skew [86, 109, 110, 117, 118]. Yet these previous studies in various ways
did not account for the rarity of the negatively charged residues relative to the more
abundant residues. In E. coli negative residues do experience electrical pulling forces
when travelling through the SecYEG translocon indicating that negative charges are
biologically relevant [42], and as mentioned, they did to an extent have the ability
to reverse topology [248]. In this thesis, we have demonstrated that the “negative-outside” skew exists in anchoring TMHs. Additionally, we have made observations
regarding how anchoring TMHs are optimised to their membranes. This new insight
into the behaviour of TMHs will not only help us understand specific cases in biology,
but can also inform synthetic biology regarding TMPs [382–384].
Post-translationally inserted TA proteins integrate into the membrane independently of the co-translational ribosome-translocon complex insertion machinery. They
are also of interest as a liposome drug delivery system due to the role of TA SNARE
proteins in membrane fusion [283, 285, 286]. We found compositional differences between the mitochondrial and secretory localised TMHs of TA proteins. Alanine is used
more commonly in the mitochondrial TMHs, perhaps even more so than isoleucine and
164

165
leucine. We also observed a reversal of the positive-inside and negative-outside distributions of charged TMH flanking residues in mitochondrially localised TA proteins.
This re-emphasises the notion that there is more to TA protein trafficking than the
hydrophobicity of the TMH alone [78], and could inform more specific studies of individual proteins. Still, several important questions remain unanswered that limit our
understanding of TA protein biosynthesis. How truly redundant are the different pathways? What are the precise biophysical and biological signatures of TA proteins that
determine their path? Why do some post-translationally inserted TA proteins have
C-terminal tail lengths exceeding the traditional 25 residue length, whereas other proteins with far shorter tails than this cut-off are co-translationally inserted? Ultimately,
as we have demonstrated, annotation of the protein databases must be completed for
these questions to be more thoroughly answered.
We have identified amphipathic TMHs that may contribute to the spontaneous
insertion of some TA proteins. However, MD simulations of the interactions between
the membrane and the anchoring TMH of spontaneously inserting TA proteins would
be needed to reveal the molecular energetic contributions involved in the spontaneous
insertion.
The biological scale of hydrophobicity revealed that many TMHs do not make the
energetic requirements of insertion by the translocon [74]. It is now understood that
these so-called marginally hydrophobic TMHs rely on sequence context for correct
topology [268] and efficient insertion [267, 361, 362]. Here we show that marginally
hydrophobic TMHs located sequentially adjacent to typically hydrophobic TMHs is a
conserved feature throughout some families of TMP such as ion channels and GPCRs.
In these two cases, there is evidence demonstrating that these hydrophobic discrepancy
pairs are capable of inserting cooperatively [40, 363–366]. It remains unclear as to what
extent the newly discovered role of the EMC integration of marginally hydrophobic
TMHs [24] works in respect to cooperatively inserted TMHs.
Understanding cooperative TMH insertion and the integration of marginally hydrophobic TMHs will require a much more thorough understanding of the temporal
and energetic aspects of the process. It appears that the relatively new technologies of
single molecule EM and APs may be able to grant insight into the energetics and dynamics of the structures throughout the process of TMP integration. Understanding

166

CHAPTER 5. CONCLUSIONS AND OUTLOOK

the relationship TMHs have with one another, particularly the temporal and spacial
contextualisation, is critical for developing complex synthetic TMPs.
More generally, there are several challenges still facing our understanding of TMPs.
The description of a TMH remains incomplete. Therefore, it remains difficult to interpret the role of a TMH in the membrane beyond anchoring. Furthermore, the
prediction of TMP topology is often erroneous and requires experimental validation.
Despite a wealth of structures the general model of helix-helix and helix-lipid interactions remains speculative and requires intensive analysis to generate a working model
of a particular TMP. Nevertheless, progress is being made in our general understanding
of TMPs.

Bibliography
1. Ladokhin, A. S. Membrane Protein Folding & Lipid Interactions: Theory &
Experiment. Journal of Membrane Biology 248, 369–370 (2015).
2. Koshland Jr., D. E. Special essay: The seven pillars of life. en. Science 295,
2215–2216 (2002).
3. Donaldson, I. Robert Hooke’s Micrographia of 1665 and 1667. The Journal of
the Royal College of Physicians of Edinburgh 40, 374–376 (2010).
4. Hooke, R. Micrographia: Or Some Physiological Descriptions of Minute Bodies
Made by Magnifying Glasses. With Observations and Inquiries Thereupon. By
R. Hooke, Fellow of the Royal Society (Jo. Martyn, and Ja. Allestry, printers to
the Royal Society, 1961).
5. Henderson, R. & Unwin, P. N. Three-dimensional model of purple membrane
obtained by electron microscopy. Nature 257, 28–32 (1975).
6. Deisenhofer, J., Epp, O., Miki, K., Huber, R. & Michel, H. X-ray structure
analysis of a membrane protein complex: Electron density map at 3 Å resolution
and a model of the chromophores of the photosynthetic reaction center from
Rhodopseudomonas viridis. Journal of Molecular Biology 180, 385–398 (1984).
7. Singer, S. J. & Nicolson, G. L. The fluid mosaic model of the structure of cell
membranes. Science (New York, N.Y.) 175, 720–31 (1972).
8. Gorter, E & Grendel, F. On bimolecular layers of lipoids on the chromocytes of
the blood. The Journal of experimental medicine 41, 439–443 (1925).
9. Danielli, J. F. & Davson, H. A contribution to the theory of permeability of thin
films. Journal of Cellular and Comparative Physiology 5, 495–508 (1935).
10. Lonsdale, K. The structure of the benzene ring. Nature 122, 810 (1928).
167

168

BIBLIOGRAPHY

11. Dickinson, R. G. & Raymond, A. L. The crystal structure of hexamethylenetetramine. Journal of the American Chemical Society 45, 22–29 (1923).
12. Carlisle, C. H. & Crowfoot, D. The crystal structure of cholesteryl iodide. Proceedings of the Royal Society A: Mathematical, Physical and Engineering Sciences 184, 64–83 (1945).
13. Hodgekin, D. C. The X-ray analysis of the structure of penicillin. Advancement
of science 6, 85–9 (1949).
14. Hodgkin, D. G., Pickworth, J, Robertson, J. H., Trueblood, K. N., Prosen, R. J.
& White, J. G. The crystal structure of the hexacarboxylic acid derived from
B12 and the molecular structure of the vitamin. Nature 176, 325–328 (1955).
15. Nobel Media. Dorothy Crowfoot Hodgkin - Nobel Lecture: The X-ray Analysis
of Complicated Molecules 2018.
16. Bernstein, F. C., Koetzle, T. F., Williams, G. J., Meyer, E. F., Brice, M. D.,
Rodgers, J. R., Kennard, O., Shimanouchi, T. & Tasumi, M. The protein data
bank: A computer-based archival file for macromolecular structures. Archives of
Biochemistry and Biophysics 185, 584–591 (1978).
17. Nobel Media. Hartmut Michel - Biographical 2018.
18. Carpenter, E. P., Beis, K., Cameron, A. D. & Iwata, S. Overcoming the challenges of membrane protein crystallography. Current Opinion in Structural Biology 18, 581–586 (2008).
19. Hopkins, A. L. & Groom, C. R. The druggable genome. Nature Reviews Drug
Discovery 1, 727–730 (2002).
20. Almén, M. S., Nordström, K. J., Fredriksson, R. & Schiöth, H. B. Mapping the
human membrane proteome: A majority of the human membrane proteins can
be classified according to function and evolutionary origin. BMC Biology 7, 50
(2009).
21. Wang, X., Wang, R., Zhang, Y. & Zhang, H. Evolutionary survey of druggable
protein targets with respect to their subcellular localizations. Genome Biology
and Evolution 5, 1291–1297 (2013).

BIBLIOGRAPHY

169

22. Overington, J. P., Al-Lazikani, B. & Hopkins, A. L. How many drug targets are
there? Nature Reviews Drug Discovery 5, 993–996 (2006).
23. Ruggiano, A., Foresti, O. & Carvalho, P. Quality control: ER-associated degradation: protein quality control and beyond. The Journal of cell biology 204,
869–79 (2014).
24. Shurtleff, M. J., Itzhak, D. N., Hussmann, J. A., Schirle Oakdale, N. T., Costa,
E. A., Jonikas, M., Weibezahn, J., Popova, K. D., Jan, C. H., Sinitcyn, P., Vembar, S. S., Hernandez, H., Cox, J., Burlingame, A. L., Brodsky, J. L., Frost, A.,
Borner, G. H. & Weissman, J. S. The ER membrane protein complex interacts
cotranslationally to enable biogenesis of multipass membrane proteins. eLife 7
(2018).
25. Guna, A., Volkmar, N., Christianson, J. C. & Hegde, R. S. The ER membrane
protein complex is a transmembrane domain insertase. Science 359, 470–473
(2018).
26. Sanders, C. R. & Myers, J. K. Disease-related misassembly of membrane proteins. Annual Review of Biophysics and Biomolecular Structure 33, 25–51
(2004).
27. Partridge, A. W., Therien, A. G. & Deber, C. M. Polar mutations in membrane
proteins as a biophysical basis for disease. Biopolymers 66, 350–358 (2002).
28. Riordan, J. R., Rommens, J. M., Kerem, B, Alon, N, Rozmahel, R, Grzelczak,
Z, Zielenski, J, Lok, S, Plavsic, N, Chou, J. L. & et Al. Identification of the cystic fibrosis gene: cloning and characterization of complementary DNA. Science
(New York, N.Y.) 245, 1066–73 (1989).
29. Roa, B. B., Garcia, C. A., Suter, U., Kulpa, D. A., Wise, C. A., Mueller, J.,
Welcher, A. A., Snipes, G. J., Shooter, E. M., Patel, P. I. & Lupski, J. R.
Charcot-Marie-Tooth disease type 1A – Association with a spontaneous point
mutation in the PMP22 gene. New England Journal of Medicine 329, 96–101
(1993).

170

BIBLIOGRAPHY

30. Fairweather, N., Bell, C., Cochrane, S., Chelly, J., Wang, S., L.Mostacciuolo,
M., P.Monaco, A. & E.Haites, N. Mutations in the connexin 32 gene in X-linked
dominant Charcot- Marie - Tooth disease (CMTX1). Human Molecular Genetics
3, 29–34 (1994).
31. Van Lieburg, A. F., Verdijk, M. A., Knoers, V. V., van Essen, A. J., Proesmans,
W, Mallmann, R, Monnens, L. A., van Oost, B. A., van Os, C. H. & Deen,
P. M. Patients with autosomal nephrogenic diabetes insipidus homozygous for
mutations in the aquaporin 2 water-channel gene. American journal of human
genetics 55, 648–52 (1994).
32. Dryja, T. P., McGee, T. L., Reichel, E., Hahn, L. B., Cowley, G. S., Yandell,
D. W., Sandberg, M. A. & Berson, E. L. A point mutation of the rhodopsin
gene in one form of retinitis pigmentosa. Nature 343, 364–366 (1990).
33. Gelsthorpe, M. E., Baumann, N., Millard, E., Gale, S. E., Langmade, S. J.,
Schaffer, J. E. & Ory, D. S. Niemann-Pick type C1 I1061T mutant encodes a
functional protein that is selected for endoplasmic reticulum-associated degradation due to protein misfolding. The Journal of biological chemistry 283, 8229–
36 (2008).
34. Park, W. D., O’Brien, J. F., Lundquist, P. A., Kraft, D. L., Vockley, C. W.,
Karnes, P. S., Patterson, M. C. & Snow, K. Identification of 58 novel mutations in Niemann-Pick disease type C: Correlation with biochemical phenotype
and importance of PTC1-like domains in NPC1. Human Mutation 22, 313–325
(2003).
35. Scott, C. & Ioannou, Y. The NPC1 protein: structure implies function. Biochimica et Biophysica Acta (BBA) - Molecular and Cell Biology of Lipids 1685, 8–
13 (2004).
36. Fernandez-Valero, E., Ballart, A, Iturriaga, C, Lluch, M, Macias, J, Vanier, M.,
Pineda, M & Coll, M. Identification of 25 new mutations in 40 unrelated Spanish Niemann-Pick type C patients: genotype-phenotype correlations. Clinical
Genetics 68, 245–254 (2005).

BIBLIOGRAPHY

171

37. Seddon, A. M., Curnow, P. & Booth, P. J. Membrane proteins, lipids and detergents: not just a soap opera. Biochimica et Biophysica Acta (BBA) - Biomembranes 1666, 105–117 (2004).
38. Von Heijne, G. Membrane-protein topology. Nature Reviews Molecular Cell Biology 7, 909–918 (2006).
39. Elofsson, A. & von Heijne, G. Membrane protein structure: Prediction versus
reality. Annual Review of Biochemistry 76, 125–140 (2007).
40. Cymer, F., von Heijne, G. & White, S. H. Mechanisms of integral membrane protein insertion and folding. Journal of Molecular Biology 427, 999–1022 (2015).
41. Merklinger, E., Gofman, Y., Kedrov, A., Driessen, A., Ben-Tal, N., Shai, Y. &
Rapaport, D. Membrane integration of a mitochondrial signal-anchored protein
does not require additional proteinaceous factors. Biochemical Journal 442,
381–389 (2012).
42. Ismail, N., Hedman, R., Lindén, M. & von Heijne, G. Charge-driven dynamics
of nascent-chain movement through the SecYEG translocon. Nature Structural
and Molecular Biology 22, 145–149 (2015).
43. Hessa, T., Sharma, A., Mariappan, M., Eshleman, H. D., Gutierrez, E. & Hegde,
R. S. Protein targeting and degradation are coupled for elimination of mislocalized proteins. Nature 475, 394–399 (2011).
44. Wong, W. C., Maurer-Stroh, S. & Eisenhaber, F. Not all transmembrane helices
are born equal: Towards the extension of the sequence homology concept to
membrane proteins. Biology Direct 6, 57 (2011).
45. Wong, W.-C., Maurer-Stroh, S., Schneider, G. & Eisenhaber, F. Transmembrane
helix: simple or complex. Nucleic Acids Research 40, W370–W375 (2012).
46. Wong, W. C., Maurer-Stroh, S. & Eisenhaber, F. More than 1,001 problems
with protein domain databases: Transmembrane regions, signal peptides and
the issue of sequence homology. PLoS Computational Biology 6 (ed Bourne,
P. E.) 6 (2010).

172

BIBLIOGRAPHY

47. Finn, R. D., Coggill, P., Eberhardt, R. Y., Eddy, S. R., Mistry, J., Mitchell,
A. L., Potter, S. C., Punta, M., Qureshi, M., Sangrador-Vegas, A., Salazar,
G. A., Tate, J. & Bateman, A. The Pfam protein families database: towards a
more sustainable future. Nucleic Acids Research 44, D279–D285 (2016).
48. Letunic, I., Doerks, T. & Bork, P. SMART 7: recent updates to the protein
domain annotation resource. Nucleic Acids Research 40, D302–D305 (2012).
49. Drew, D., Slotboom, D.-J., Friso, G., Reda, T., Genevaux, P., Rapp, M., MeindlBeinker, N. M., Lambert, W., Lerch, M., Daley, D. O., Van Wijk, K.-J., Hirst,
J., Kunji, E. & De Gier, J.-W. A scalable, GFP-based pipeline for membrane
protein overexpression screening and purification. Protein Science 14, 2011–
2017 (2005).
50. Kawate, T. & Gouaux, E. Fluorescence-detection size-exclusion chromatography
for precrystallization screening of integral membrane proteins. Structure 14,
673–681 (2006).
51. Eshaghi, S., Hedrén, M., Nasser, M. I. A., Hammarberg, T., Thornell, A. &
Nordlund, P. An efficient strategy for high-throughput expression screening of
recombinant integral membrane proteins. Protein Science 14, 676–683 (2005).
52. Cornvik, T., Dahlroth, S.-L., Magnusdottir, A., Herman, M. D., Knaust, R.,
Ekberg, M. & Nordlund, P. Colony filtration blot: a new screening method for
soluble protein expression in Escherichia coli. Nature Methods 2, 507–509 (2005).
53. Frauenfeld, J., Gumbart, J., van der Sluis, E. O., Funes, S., Gartmann, M.,
Beatrix, B., Mielke, T., Berninghausen, O., Becker, T., Schulten, K. & Beckmann, R. Cryo-EM structure of the ribosome–SecYE complex in the membrane
environment. Nature Structural & Molecular Biology 18, 614–621 (2011).
54. Rosenbusch, J., Lustig, A., Grabo, M., Zulauf, M. & Regenass, M. Approaches
to determining membrane protein structures to high resolution: do selections of
subpopulations occur? Micron 32, 75–90 (2001).
55. Bayburt, T. H., Grinkova, Y. V. & Sligar, S. G. Self-assembly of discoidal phospholipid bilayer nanoparticles with membrane scaffold proteins. Nano Letters
2, 853–856 (2002).

BIBLIOGRAPHY

173

56. Duan, H., Civjan, N. R., Sligar, S. G. & Schuler, M. A. Co-incorporation of
heterologously expressed Arabidopsis cytochrome P450 and P450 reductase into
soluble nanoscale lipid bilayers. Archives of Biochemistry and Biophysics 424,
141–153 (2004).
57. Schuler, M. A., Denisov, I. G. & Sligar, S. G. in Methods in molecular biology
(Clifton, N.J.) 415–433 (2013).
58. Borch, J. & Hamann, T. The nanodisc: A novel tool for membrane protein
studies. Biological Chemistry 390, 805–814 (2009).
59. Vinothkumar, K. R. Membrane protein structures without crystals, by single
particle electron cryomicroscopy. Current Opinion in Structural Biology 33,
103–114 (2015).
60. Raunser, S. & Walz, T. Electron crystallography as a technique to study the
structure on membrane proteins in a lipidic environment. Annual Review of
Biophysics 38, 89–105 (2009).
61. Vinothkumar, K. R. & Henderson, R. Structures of membrane proteins. Quarterly Reviews of Biophysics 43, 65–158 (2010).
62. Murata, K. & Wolf, M. Cryo-electron microscopy for structural analysis of dynamic biological macromolecules. Biochimica et Biophysica Acta (BBA) - General Subjects 1862, 324–334 (2018).
63. Grant, T. & Grigorieff, N. Measuring the optimal exposure for single particle
cryo-EM using a 2.6 Å reconstruction of rotavirus VP6. eLife 4 (2015).
64. Bartesaghi, A., Merk, A., Banerjee, S., Matthies, D., Wu, X., Milne, J. L. S.
& Subramaniam, S. 2.2 Å resolution cryo-EM structure of β-galactosidase in
complex with a cell-permeant inhibitor. Science (New York, N.Y.) 348, 1147–
51 (2015).
65. Kühlbrandt, W. The resolution revolution. Science 343, 1443 LP –1444 (2014).
66. Jones, D. T. Improving the accuracy of transmembrane protein topology prediction using evolutionary information. Bioinformatics 23, 538–544 (2007).

174

BIBLIOGRAPHY

67. Krogh, A., Larsson, B., von Heijne, G. & Sonnhammer, E. L. Predicting transmembrane protein topology with a hidden Markov model: Application to complete genomes. Journal of Molecular Biology 305, 567–580 (2001).
68. Tusnády, G. E. & Simon, I. The HMMTOP transmembrane topology prediction
server. Bioinformatics 17, 849–850 (2001).
69. Viklund, H. & Elofsson, A. Best α-helical transmembrane protein topology predictions are achieved using hidden Markov models and evolutionary information.
Protein Science 13, 1908–1917 (2004).
70. Viklund, H. & Elofsson, A. OCTOPUS: Improving topology prediction by
two-track ANN-based preference scores and an extended topological grammar.
Bioinformatics 24, 1662–1668 (2008).
71. Käll, L., Krogh, A. & Sonnhammer, E. L. L. A combined transmembrane topology and signal peptide prediction method. Journal of Molecular Biology 338,
1027–1036 (2004).
72. Bernsel, A., Viklund, H., Falk, J., Lindahl, E., von Heijne, G. & Elofsson, A.
Prediction of membrane-protein topology from first principles. Proceedings of
the National Academy of Sciences 105, 7177–7181 (2008).
73. Kyte, J. & Doolittle, R. F. A simple method for displaying the hydropathic
character of a protein. Journal of Molecular Biology 157, 105–132 (1982).
74. Hessa, T., Meindl-Beinker, N. M., Bernsel, A., Kim, H., Sato, Y., LerchBader, M., Nilsson, I., White, S. H. & von Heijne, G. Molecular code for
transmembrane-helix recognition by the Sec61 translocon. Nature 450, 1026–
1030 (2007).
75. Bateman, A. et al. UniProt: A hub for protein information. Nucleic Acids Research 43, D204–D212 (2015).
76. Petersen, T. N., Brunak, S., von Heijne, G. & Nielsen, H. SignalP 4.0: Discriminating signal peptides from transmembrane regions. Nature Methods 8, 785–786
(2011).
77. Nugent, T. & Jones, D. T. Detecting pore-lining regions in transmembrane protein sequences. BMC Bioinformatics 13, 1 (2012).

BIBLIOGRAPHY

175

78. Guna, A. & Hegde, R. S. Transmembrane comain recognition during membrane
protein biogenesis and quality control. Current Biology 28, R498–R511 (2018).
79. Stansfeld, P. J., Goose, J. E., Caffrey, M., Carpenter, E. P., Parker, J. L., Newstead, S. & Sansom, M. S. MemProtMD: Automated insertion of membrane
protein dtructures into rxplicit lipid membranes. English. Structure 23, 1350–
1361 (2015).
80. Coskun, Ü. & Simons, K. Cell membranes: The lipid perspective. Structure 19,
1543–1548 (2011).
81. Park, J. H., Scheerer, P., Hofmann, K. P., Choe, H.-W. & Ernst, O. P. Crystal
structure of the ligand-free G-protein-coupled receptor opsin. Nature 454, 183–
187 (2008).
82. Berman, H. M., Westbrook, J., Feng, Z., Gilliland, G., Bhat, T. N., Weissig,
H., Shindyalov, I. N. & Bourne, P. E. The Protein Data Bank. Nucleic Acids
Research 28, 235–242 (2000).
83. Lomize, M. A., Pogozheva, I. D., Joo, H., Mosberg, H. I. & Lomize, A. L.
OPM database and PPM web server: resources for positioning of proteins in
membranes. Nucleic Acids Research 40, D370–D376 (2012).
84. Kozma, D., Simon, I. & Tusnády, G. E. PDBTM: Protein Data Bank of transmembrane proteins after 8 years. Nucleic Acids Research 41, D524–D529 (2012).
85. Tusnady, G. E., Dosztanyi, Z. & Simon, I. TMDET: web server for detecting
transmembrane regions of proteins by using their 3D coordinates. Bioinformatics
21, 1276–1277 (2005).
86. Sharpe, H. J., Stevens, T. J. & Munro, S. A comprehensive comparison of
transmembrane domains reveals organelle-specific properties. Cell 142, 158–
169 (2010).
87. White, S. H., Ladokhin, A. S., Jayasinghe, S. & Hristova, K. How membranes
shape protein structure. Journal of Biological Chemistry 276, 32395–32398
(2001).

176

BIBLIOGRAPHY

88. Jensen, M. & Mouritsen, O. G. Lipids do influence protein function - The
hydrophobic matching hypothesis revisited. Biochimica et Biophysica Acta Biomembranes 1666, 205–226 (2004).
89. Kauko, A., Hedin, L. E., Thebaud, E., Cristobal, S., Elofsson, A. & von Heijne,
G. Repositioning of transmembrane α-helices during membrane protein folding.
Journal of Molecular Biology 397, 190–201 (2010).
90. Kalvodova, L., Kahya, N., Schwille, P., Ehehalt, R., Verkade, P., Drechsel, D. &
Simons, K. Lipids as modulators of proteolytic activity of BACE: Involvement
of cholesterol, glycosphingolipids, and anionic phospholipids in vitro. Journal of
Biological Chemistry 280, 36815–36823 (2005).
91. Bondar, A. N., del Val, C. & White, S. H. Rhomboid protease dynamics and
lipid interactions. Structure 17, 395–405 (2009).
92. Van Meer, G., Voelker, D. R. & Feigenson, G. W. Membrane lipids: Where they
are and how they behave. Nature Reviews Molecular Cell Biology 9, 112–124
(2008).
93. Scott, K. A., Bond, P. J., Ivetac, A., Chetwynd, A. P., Khalid, S. & Sansom, M. S. Coarse-grained MD simulations of membrane protein-bilayer selfassembly. Structure 16, 621–630 (2008).
94. Goetz, R. & Lipowsky, R. Computer simulations of bilayer membranes: Selfassembly and interfacial tension. Journal of Chemical Physics 108, 7397–7409
(1998).
95. Helenius, A. & Simons, K. Solubilization of membranes by detergents. Biochimica et Biophysica Acta (BBA) - Reviews on Biomembranes 415, 29–79 (1975).
96. Lichtenberg, D., Robson, R. J. & Dennis, E. A. Solubilization of phospholipids
by detergents structural and kinetic aspects. Biochimica et Biophysica Acta
(BBA) - Reviews on Biomembranes 737, 285–304 (1983).
97. Op den Kamp, J. A. F. Lipid asymmetry in membranes. Annual Review of
Biochemistry 48, 47–71 (1979).

BIBLIOGRAPHY

177

98. De Planque, M. R. R. & Killian, J. A. Protein-lipid interactions studied with
designed transmembrane peptides: Role of hydrophobic matching and interfacial
anchoring (Review). en. Molecular Membrane Biology 20, 271–284 (2003).
99. Feiner, A.-S. & McEvoy, A. J. The Nernst equation. Journal of Chemical Education 71, 493 (1994).
100. Qin, Y., Dittmer, P. J., Park, J. G., Jansen, K. B. & Palmer, A. E. Measuring
steady-state and dynamic endoplasmic reticulum and Golgi Zn2+ with genetically encoded sensors. Proceedings of the National Academy of Sciences 108,
7351–7356 (2011).
101. Worley, J. F., McIntyre, M. S., Spencer, B., Mertz, R. J., Roe, M. W. & Dukes,
I. D. Endoplasmic reticulum calcium store regulates membrane potential in
mouse islet β-cells. Journal of Biological Chemistry 269, 14359–14362 (1994).
102. Perry, S. W., Norman, J. P., Barbieri, J., Brown, E. B. & Gelbard, H. A. Mitochondrial membrane potential probes and the proton gradient: A practical
usage guide. BioTechniques 50, 98–115 (2011).
103. Schapiro, F. B. & Grinstein, S. Determinants of the pH of the Golgi complex.
Journal of Biological Chemistry 275, 21025–21032 (2000).
104. Llopis, J., McCaffery, J. M., Miyawaki, A., Farquhar, M. G. & Tsien, R. Y.
Measurement of cytosolic, mitochondrial, and Golgi pH in single living cells with
green fluorescent proteins. Proceedings of the National Academy of Sciences 95,
6803–6808 (1998).
105. Hildebrand, P. W., Preissner, R. & Frömmel, C. Structural features of transmembrane helices. FEBS Letters 559, 145–151 (2004).
106. Ulmschneider, M. B. & Sansom, M. S. Amino acid distributions in integral
membrane protein structures. Biochimica et Biophysica Acta - Biomembranes
1512, 1–14 (2001).
107. Bowie, J. U. Helix packing in membrane proteins. Journal of Molecular Biology
272, 780–789 (1997).

178

BIBLIOGRAPHY

108. Cuthbertson, J. M., Doyle, D. A. & Sansom, M. S. P. Transmembrane helix
prediction: A comparative evaluation and analysis. Protein Engineering, Design
and Selection 18, 295–308 (2005).
109. Baeza-Delgado, C., Marti-Renom, M. A. & Mingarro, I. Structure-based statistical analysis of transmembrane helices. European Biophysics Journal 42, 199–
207 (2013).
110. Pogozheva, I. D., Tristram-Nagle, S., Mosberg, H. I. & Lomize, A. L. Structural adaptations of proteins to different biological membranes. Biochimica et
Biophysica Acta - Biomembranes 1828, 2592–2608 (2013).
111. Skasko, M., Wang, Y., Tian, Y., Tokarev, A., Munguia, J., Ruiz, A., Stephens,
E. B., Opella, S. J. & Guatelli, O. HIV-1 Vpu protein antagonizes innate restriction factor BST-2 via lipid-embedded helix-helix interactions. Journal of
Biological Chemistry 287, 58–67 (2012).
112. Jayasinghe, S. MPtopo: A database of membrane protein topology. Protein Science 10, 455–458 (2001).
113. Von Heijne, G. Control of topology and mode of assembly of a polytopic membrane protein by positively charged residues. en. Nature 341, 456–458 (1989).
114. Andersson, H., Bakker, E. & von Heijne, G. Different positively charged amino
acids have similar effects on the topology of a polytopic transmembrane protein
in Escherichia coli. Journal of Biological Chemistry 267, 1491–1495 (1992).
115. Rojo, E. E., Guiard, B, Neupert, W & Stuart, R. A. N-terminal tail export from
the mitochondrial matrix. Adherence to the prokaryotic ”positive-inside” rule of
membrane protein topology. The Journal of biological chemistry 274, 19617–22
(1999).
116. Hessa, T., Kim, H., Bihlmaier, K., Lundin, C., Boekel, J., Andersson, H., Nilsson, I. M., White, S. H. & von Heijne, G. Recognition of transmembrane helices
by the endoplasmic reticulum translocon. Nature 433, 377–381 (2005).
117. Granseth, E., von Heijne, G. & Elofsson, A. A study of the membrane-water
interface region of membrane proteins. Journal of Molecular Biology 346, 377–
385 (2005).

BIBLIOGRAPHY

179

118. Nilsson, J., Persson, B. & von Heijne, G. Comparative analysis of amino acid
distributions in integral membrane proteins from 107 genomes. Proteins: Structure, Function, and Bioinformatics 60, 606–616 (2005).
119. Braun, P. & von Heijne, G. The aromatic residues Trp and phe have different effects on the positioning of a transmembrane helix in the microsomal membrane.
Biochemistry 38, 9778–9782 (1999).
120. Situ, A. J., Kang, S. M., Frey, B. B., An, W., Kim, C. & Ulmer, T. S. Membrane anchoring of α-helical proteins: Role of tryptophan. Journal of Physical
Chemistry B 122, 1185–1194 (2018).
121. Yau, W. M., Wimley, W. C., Gawrisch, K. & White, S. H. The preference of
tryptophan for membrane interfaces. Biochemistry 37, 14713–14718 (1998).
122. Chamberlain, A. K., Lee, Y., Kim, S. & Bowie, J. U. Snorkeling preferences
foster an amino acid composition bias in transmembrane helices. Journal of
Molecular Biology 339, 471–479 (2004).
123. Strandberg, E. & Killian, J. A. Snorkeling of lysine side chains in transmembrane
helices: How easy can it get? FEBS Letters 544, 69–73 (2003).
124. Krishnakumar, S. S. & London, E. The control of transmembrane helix transverse position in membranes by hydrophilic residues. Journal of Molecular Biology 374, 1251–1269 (2007).
125. Kandasamy, S. K. & Larson, R. G. Molecular dynamics simulations of model
trans-membrane peptides in lipid bilayers: A systematic investigation of hydrophobic mismatch. Biophysical Journal 90, 2326–2343 (2006).
126. Ulmschneider, M. B., Ulmschneider, J. P., Freites, J. A., von Heijne, G., Tobias,
D. J. & White, S. H. Transmembrane helices containing a charged arginine are
thermodynamically stable. European Biophysics Journal 46, 627–637 (2017).
127. Davis, N. G. & Model, P. An artificial anchor domain: hydrophobicity suffices
to stop transfer. Cell 41, 607–614 (1985).
128. Junne, T., Kocik, L. & Spiess, M. The hydrophobic core of the Sec61 translocon defines the hydrophobicity threshold for membrane integration. Molecular
biology of the cell 21, 1662–70 (2010).

180

BIBLIOGRAPHY

129. Simm, S., Einloft, J., Mirus, O. & Schleiff, E. 50 years of amino acid hydrophobicity scales: Revisiting the capacity for Peptide classification. Biological research 49, 31 (2016).
130. Peters, C. & Elofsson, A. Why is the biological hydrophobicity scale more accurate than earlier experimental hydrophobicity scales? Proteins: Structure, Function and Bioinformatics 82, 2190–2198 (2014).
131. Chothia, C. The nature of the accessible and buried surfaces in proteins. Journal
of Molecular Biology 105, 1–12 (1976).
132. Wolfenden, R. V., Cullis, P. M. & Southgate, C. C. Water, protein folding, and
the genetic code. Science (New York, N.Y.) 206, 575–7 (1979).
133. Cohn, E. J., McMeekin, T. L., Edsall, J. T. & Blanchard, M. H. Studies in
the physical chemistry of amino acids, peptides and related substances. I. The
apparent molal volume and the electrostriction of the solvent. Journal of the
American Chemical Society 56, 784–794 (1934).
134. Traube, J. Samml. them. u. them.-techn. Vortr 4, 19–322 (1899).
135. Hine, J. & Mookerjee, P. K. Structural effects on rates and equilibriums. XIX.
Intrinsic hydrophilic character of organic compounds. Correlations in terms of
structural contributions. The Journal of Organic Chemistry 40, 292–298 (1975).
136. Cohn, E. J., Edsall, J. T., Kirkwood, J. G., Mueller, H., Oncley, J. L. &
Scatchard, G. Proteins, amino acids and peptides as ions and dipolar ions 6,
884 (1943).
137. Nozaki, Y. & Tanford, C. The Solubility of Amino in Aqueous Ethanol Acids and
Two Glycine Dioxane Solutions Peptides. The Journal of Biological Chemistry
246, 2211–2217 (1971).
138. Brambillasca, S., Yabal, M., Makarow, M. & Borgese, N. Unassisted translocation of large polypeptide domains across phospholipid bilayers. Journal of Cell
Biology 175, 767–777 (2006).
139. White, S. H. & Wimley, W. C. Membrane protein folding and stability: Physical
principles. Annual Review of Biophysics and Biomolecular Structure 28, 319–
365 (1999).

BIBLIOGRAPHY

181

140. Rose, G. D. & Wolfenden, R. Hydrogen bonding, hydrophobicity, packing, and
protein folding. Annu. Rev. Biophys. Biomol. Struct. 22, 381–415 (1993).
141. Janin, J. Surface and inside volumes in globular proteins [20]. Nature 277, 491–
492 (1979).
142. Wolfenden, R., Andersson, L., Cullis, P. M. & Southgate, C. C. Affinities of
amino acid side chains for solvent water. Biochemistry 20, 849–855 (1981).
143. Von Heijne, G. & Blomberg, C. Trans-membrane translocation of proteins: The
direct transfer model. European Journal of Biochemistry 97, 175–181 (1979).
144. Eisenberg, D., Schwarz, E., Komaromy, M. & Wall, R. Analysis of membrane
and surface protein sequences with the hydrophobic moment plot. Journal of
Molecular Biology 179, 125–142 (1984).
145. Murzin, A. G., Brenner, S. E., Hubbard, T. & Chothia, C. SCOP: A structural classification of proteins database for the investigation of sequences and
structures. Journal of Molecular Biology 247, 536–540 (1995).
146. Wootton, J. C. & Federhen, S. [33] Analysis of compositionally biased regions
in sequence databases. Elsevier, 554–571 (1996).
147. Pinho, A. J., Garcia, S. P., Pratas, D. & Ferreira, P. J. DNA sequences at a
glance. PLoS ONE 8 (ed Gibas, C.) e79922 (2013).
148. Oliver, J. L., Bernaola-Galván, P., Guerrero-Garcia, J. & Román-Roldán, R.
Entropic profiles of DNA sequences through chaos-game-derived images. Journal
of Theoretical Biology 160, 457–470 (1993).
149. Troyanskaya, O. G., Arbell, O., Koren, Y., Landau, G. M. & Bolshoy, A. Sequence complexity profiles of prokaryotic genomic sequences: A fast algorithm
for calculating linguistic complexity. Bioinformatics 18, 679–688 (2002).
150. Wootton, J. C. Non-globular domains in protein sequences: Automated segmentation using complexity measures. Computers and Chemistry 18, 269–285
(1994).
151. Shannon, C. E. A mathematical theory of communication. Bell System Technical
Journal 27, 379–423 (1948).

182

BIBLIOGRAPHY

152. Lupas, A, Dyke, M. V. & Stock, J. Predicting coiled coils from protein sequences.
Science 252, 1162–1164 (1991).
153. Cao, T. B. & Saier, M. H. The general protein secretory pathway: phylogenetic analyses leading to evolutionary conclusions. Biochimica et Biophysica
Acta (BBA) - Biomembranes 1609, 115–125 (2003).
154. Drew, D., Fröderberg, L., Baars, L. & de Gier, J.-W. L. Assembly and overexpression of membrane proteins in Escherichia coli. Biochimica et Biophysica
Acta (BBA) - Biomembranes 1610, 3–10 (2003).
155. Dalbey, R. E., Kuhn, A., Zhu, L. & Kiefer, D. The membrane insertase YidC.
Biochimica et Biophysica Acta (BBA) - Molecular Cell Research 1843, 1489–
1496 (2014).
156. Park, S. H., Das, B. B., Casagrande, F., Tian, Y., Nothnagel, H. J., Chu, M.,
Kiefer, H., Maier, K., De Angelis, A. A., Marassi, F. M. & Opella, S. J. Structure
of the chemokine receptor CXCR1 in phospholipid bilayers. Nature 491, 779–
783 (2012).
157. Shao, S. & Hegde, R. S. Membrane protein insertion at the endoplasmic reticulum. Annual Review of Cell and Developmental Biology 27, 25–56 (2011).
158. Kihara, A, Akiyama, Y & Ito, K. FtsH is required for proteolytic elimination of
uncomplexed forms of SecY, an essential protein translocase subunit. Proceedings of the National Academy of Sciences of the United States of America 92,
4532–6 (1995).
159. Duong, F & Wickner, W. Distinct catalytic roles of the SecYE, SecG and SecDFyajC subunits of preprotein translocase holoenzyme. The EMBO journal 16,
2756–68 (1997).
160. Koch, H. G. & Müller, M. Dissecting the translocase and integrase functions of
the Escherichia coli SecYEG translocon. The Journal of cell biology 150, 689–94
(2000).
161. Zimmer, J., Nam, Y. & Rapoport, T. A. Structure of a complex of the ATPase
SecA and the protein-translocation channel. Nature 455, 936–943 (2008).

BIBLIOGRAPHY

183

162. Voorhees, R. M., Fernández, I. S., Scheres, S. H. W. & Hegde, R. S. Structure
of the mammalian ribosome-Sec61 complex to 3.4 Å resolution. English. Cell
157, 1632–1643 (2014).
163. Van den Berg, B., Clemons, W. M., Collinson, I., Modis, Y., Hartmann, E.,
Harrison, S. C. & Rapoport, T. A. X-ray structure of a protein-conducting
channel. Nature 427, 36–44 (2004).
164. Tam, P. C. K., Maillard, A. P., Chan, K. K. Y. & Duong, F. Investigating the
SecY plug movement at the SecYEG translocation channel. The EMBO journal
24, 3380–8 (2005).
165. Junne, T., Schwede, T., Goder, V. & Spiess, M. The plug domain of yeast
Sec61p is important for efficient protein translocation, but is not essential for cell
viability. Molecular Biology of the Cell 17 (ed Gilmore, R.) 4063–4068 (2006).
166. Park, E. & Rapoport, T. A. Preserving the membrane barrier for small molecules
during bacterial protein translocation. Nature 473, 239–242 (2011).
167. Martoglio, B., Hofmann, M. W., Brunner, J. & Dobberstein, B. The proteinconducting channel in the membrane of the endoplasmic reticulum is open laterally toward the lipid bilayer. Cell 81, 207–214 (1995).
168. Gogala, M., Becker, T., Beatrix, B., Armache, J.-P., Barrio-Garcia, C., Berninghausen, O. & Beckmann, R. Structures of the Sec61 complex engaged in nascent
peptide translocation or membrane insertion. Nature 506, 107–110 (2014).
169. Walter, P & Blobel, G. Translocation of proteins across the endoplasmic reticulum III. Signal recognition protein (SRP) causes signal sequence-dependent
and site-specific arrest of chain elongation that is released by microsomal membranes. The Journal of cell biology 91, 557–61 (1981).
170. Rapoport, T. A. Protein translocation across the eukaryotic endoplasmic reticulum and bacterial plasma membranes. Nature 450, 663–669 (2007).
171. Pool, M. R. Signal recognition particles in chloroplasts, bacteria, yeast and mammals. Molecular Membrane Biology 22, 3–15 (2005).

184

BIBLIOGRAPHY

172. Song, W, Raden, D, Mandon, E. C. & Gilmore, R. Role of Sec61alpha in the
regulated transfer of the ribosome-nascent chain complex from the signal recognition particle to the translocation channel. Cell 100, 333–43 (2000).
173. Fulga, T. A., Sinning, I, Dobberstein, B & Pool, M. R. SRbeta coordinates
signal sequence release from SRP with ribosome binding to the translocon. The
EMBO journal 20, 2338–47 (2001).
174. Pool, M. R., Stumm, J., Fulga, T. A., Sinning, I. & Dobberstein, B. Distinct
modes of signal recognition particle interaction with the ribosome. Science (New
York, N.Y.) 297, 1345–8 (2002).
175. Shan, S. O. & Walter, P. Co-translational protein targeting by the signal recognition particle. FEBS Letters 579, 921–926 (2005).
176. Petriman, N.-A., Jauß, B., Hufnagel, A., Franz, L., Sachelaru, I., Drepper, F.,
Warscheid, B. & Koch, H.-G. The interaction network of the YidC insertase with
the SecYEG translocon, SRP and the SRP receptor FtsY. Scientific Reports 8,
578 (2018).
177. Hizlan, D., Robson, A., Whitehouse, S., Gold, V. A., Vonck, J., Mills, D.,
Kühlbrandt, W. & Collinson, I. Structure of the SecY complex unlocked by
a preprotein mimic. Cell reports 1, 21–8 (2012).
178. Bischoff, L., Wickles, S., Berninghausen, O., van der Sluis, E. O. & Beckmann, R.
Visualization of a polytopic membrane protein during SecY-mediated membrane
insertion. Nature Communications 5, 4103 (2014).
179. Egea, P. F. & Stroud, R. M. Lateral opening of a translocon upon entry of
protein suggests the mechanism of insertion into membranes. Proceedings of the
National Academy of Sciences 107, 17182–17187 (2010).
180. Niesen, M., Muller-Lucks, A., Hedman, R., von Heijne, G. & Miller, T. F. Forces
on nascent polypeptides during membrane insertion and translocation via the
Sec translocon. bioRxiv, 310698 (2018).
181. Van Den Berg, B., Clemons, W. M., Collinson, I., Modis, Y., Hartmann, E.,
Harrison, S. C. & Rapoport, T. A. X-ray structure of a protein-conducting
channel. Nature 427, 36–44 (2004).

BIBLIOGRAPHY

185

182. Emr, S. D., Hanley-Way, S. & Silhavy, T. J. Suppressor mutations that restore
export of a protein with a defective signal sequence. Cell 23, 79–88 (1981).
183. Veenendaal, A. K., van der Does, C. & Driessen, A. J. The protein-conducting
channel SecYEG. Biochimica et Biophysica Acta (BBA) - Molecular Cell Research 1694, 81–95 (2004).
184. Li, W., Schulman, S., Boyd, D., Erlandson, K., Beckwith, J. & Rapoport, T. A.
The plug domain of the SecY protein stabilizes the closed state of the translocation channel and maintains a membrane seal. Molecular Cell 26, 511–521
(2007).
185. Junne, T., Schwede, T., Goder, V. & Spiess, M. Mutations in the Sec61p channel
affecting signal sequence recognition and membrane protein topology. Journal
of Biological Chemistry 282, 33201–33209 (2007).
186. Saparov, S. M., Erlandson, K., Cannon, K., Schaletzky, J., Schulman, S.,
Rapoport, T. A. & Pohl, P. Determining the conductance of the SecY protein
translocation channel for small molecules. Molecular Cell 26, 501–509 (2007).
187. Samuelson, J. C., Chen, M., Jiang, F., Möller, I., Wiedmann, M., Kuhn, A.,
Phillips, G. J. & Dalbey, R. E. YidC mediates membrane protein insertion in
bacteria. Nature 406, 637–641 (2000).
188. Nagamori, S., Smirnova, I. N. & Kaback, H. R. Role of YidC in folding of
polytopic membrane proteins. The Journal of cell biology 165, 53–62 (2004).
189. Sommer, N., Junne, T., Kalies, K.-U., Spiess, M. & Hartmann, E. TRAP assists
membrane protein topogenesis at the mammalian ER membrane. Biochimica et
Biophysica Acta (BBA) - Molecular Cell Research 1833, 3104–3111 (2013).
190. Voigt, S, Jungnickel, B, Hartmann, E & Rapoport, T. A. Signal sequencedependent function of the TRAM protein during early phases of protein transport across the endoplasmic reticulum membrane. The Journal of cell biology
134, 25–35 (1996).
191. Conti, B. J., Devaraneni, P. K., Yang, Z., David, L. L. & Skach, W. R. Cotranslational stabilization of Sec62/63 within the ER Sec61 translocon is controlled by
distinct substrate-driven translocation events. Molecular cell 58, 269–83 (2015).

186

BIBLIOGRAPHY

192. Meyer, H. A., Grau, H, Kraft, R, Kostka, S, Prehn, S, Kalies, K. U. & Hartmann, E. Mammalian Sec61 is associated with Sec62 and Sec63. The Journal of
biological chemistry 275, 14550–7 (2000).
193. Chirico, W. J., Waters, M. G. & Blobel, G. 70K heat shock related proteins
stimulate protein translocation into microsomes. Nature 332, 805–810 (1988).
194. Deshaies, R. J., Koch, B. D., Werner-Washburne, M., Craig, E. A. & Schekman, R. A subfamily of stress proteins facilitates translocation of secretory and
mitochondrial precursor polypeptides. Nature 332, 800–805 (1988).
195. Steel, G. J., Fullerton, D. M., Tyson, J. R. & Stirling, C. J. Coordinated activation of Hsp70 chaperones. Science (New York, N.Y.) 303, 98–101 (2004).
196. Jadhav, B., McKenna, M., Johnson, N., High, S., Sinning, I. & Pool, M. R.
Mammalian SRP receptor switches the Sec61 translocase from Sec62 to SRPdependent translocation. Nature Communications 6, 10133 (2015).
197. Von Heijne, G. Signal sequences: The limits of variation. Journal of Molecular
Biology 184, 99–105 (1985).
198. Von Heijne, G. The signal peptide. J Membr Biol 115, 195–201 (1990).
199. Izard, J. W. & Kendall, D. A. Signal peptides: exquisitely designed transport
promoters. Molecular Microbiology 13, 765–773 (1994).
200. Gierasch, L. M. Signal sequences. Biochemistry 28, 923–930 (1989).
201. Choo, K., Tan, T. & Ranganathan, S. A comprehensive assessment of N-terminal
signal peptides prediction methods. BMC Bioinformatics 10, S2 (2009).
202. Liang, B. & Tamm, L. K. Structure of outer membrane protein G by solution NMR spectroscopy. Proceedings of the National Academy of Sciences of the
United States of America 104, 16140–5 (2007).
203. Posson, D. J., McCoy, J. G. & Nimigean, C. M. The voltage-dependent gate in
MthK potassium channels is located at the selectivity filter. Nature Structural
& Molecular Biology 20, 159–166 (2013).

BIBLIOGRAPHY

187

204. Palczewski, K, Kumasaka, T, Hori, T, Behnke, C. A., Motoshima, H, Fox, B. A.,
Le Trong, I, Teller, D. C., Okada, T, Stenkamp, R. E., Yamamoto, M & Miyano,
M. Crystal structure of rhodopsin: A G protein-coupled receptor. Science (New
York, N.Y.) 289, 739–45 (2000).
205. Wimley, W. C. The versatile β-barrel membrane protein. Current Opinion in
Structural Biology 13, 404–411 (2003).
206. McFadden, G. I. Chloroplast origin and integration. Plant physiology 125, 50–3
(2001).
207. Gray, M. W., Burger, G & Lang, B. F. Mitochondrial evolution. Science (New
York, N.Y.) 283, 1476–81 (1999).
208. Fischer, K, Weber, A, Brink, S, Arbinger, B, Schünemann, D, Borchert, S, Heldt,
H. W., Popp, B, Benz, R & Link, T. A. Porins from plants. Molecular cloning
and functional characterization of two new members of the porin family. The
Journal of biological chemistry 269, 25754–60 (1994).
209. Zeth, K. & Thein, M. Porins in prokaryotes and eukaryotes: common themes
and variations. Biochemical Journal 431, 13–22 (2010).
210. Fairman, J. W., Noinaj, N. & Buchanan, S. K. The structural biology of β-barrel
membrane proteins: a summary of recent reports. Current Opinion in Structural
Biology 21, 523–531 (2011).
211. Ulrich, T. & Rapaport, D. Biogenesis of beta-barrel proteins in evolutionary
context. International Journal of Medical Microbiology 305, 259–264 (2015).
212. Driessen, A. J. & Nouwen, N. Protein translocation across the bacterial cytoplasmic membrane. Annual Review of Biochemistry 77, 643–667 (2008).
213. Papanikou, E., Karamanou, S. & Economou, A. Bacterial protein secretion
through the translocase nanomachine. Nature Reviews Microbiology 5, 839–851
(2007).
214. Bechtluft, P., Nouwen, N., Tans, S. J. & Driessen, A. J. M. SecB—A chaperone
dedicated to protein translocation. Mol. BioSyst. 6, 620–627 (2010).

188

BIBLIOGRAPHY

215. Paetzel, M. Structure and mechanism of Escherichia coli type I signal peptidase.
Biochimica et Biophysica Acta (BBA) - Molecular Cell Research 1843, 1497–
1508 (2014).
216. Lazar, S. W. & Kolter, R. SurA assists the folding of Escherichia coli outer
membrane proteins. Journal of bacteriology 178, 1770–3 (1996).
217. Volokhina, E. B., Grijpstra, J., Stork, M., Schilders, I., Tommassen, J. & Bos,
M. P. Role of the periplasmic chaperones Skp, SurA, and DegQ in outer membrane protein biogenesis in Neisseria meningitidis. Journal of bacteriology 193,
1612–21 (2011).
218. Wu, T., Malinverni, J., Ruiz, N., Kim, S., Silhavy, T. J. & Kahne, D. Identification of a multicomponent complex required for outer membrane biogenesis in
escherichia coli. Cell 121, 235–245 (2005).
219. Hagan, C. L., Kim, S. & Kahne, D. Reconstitution of outer membrane protein
assembly from purified components. Science 328, 890–892 (2010).
220. Noinaj, N., Kuszak, A. J., Balusek, C., Gumbart, J. C. & Buchanan, S. K. Lateral opening and exit pore formation are required for BamA function. Structure
(London, England : 1993) 22, 1055–62 (2014).
221. Müller, J. E., Papic, D., Ulrich, T., Grin, I., Schütz, M., Oberhettinger, P.,
Tommassen, J., Linke, D., Dimmer, K. S., Autenrieth, I. B. & Rapaport, D.
Mitochondria can recognize and assemble fragments of a β-barrel structure.
Molecular Biology of the Cell 22 (ed Glick, B.) 1638–1647 (2011).
222. Chacinska, A., Koehler, C. M., Milenkovic, D., Lithgow, T. & Pfanner, N. Importing mitochondrial proteins: machineries and mechanisms. Cell 138, 628–44
(2009).
223. Wiedemann, N., Kozjak, V., Chacinska, A., Schönfisch, B., Rospert, S., Ryan,
M. T., Pfanner, N. & Meisinger, C. Machinery for protein sorting and assembly
in the mitochondrial outer membrane. Nature 424, 565–571 (2003).
224. Paschen, S. A., Waizenegger, T., Stan, T., Preuss, M., Cyrklaff, M., Hell, K.,
Rapaport, D. & Neupert, W. Evolutionary conservation of biogenesis of β-barrel
membrane proteins. Nature 426, 862–866 (2003).

BIBLIOGRAPHY

189

225. Gentle, I., Gabriel, K., Beech, P., Waller, R. & Lithgow, T. The Omp85 family of
proteins is essential for outer membrane biogenesis in mitochondria and bacteria.
The Journal of cell biology 164, 19–24 (2004).
226. Tamm, L. K., Hong, H. & Liang, B. Folding and assembly of β-barrel membrane
proteins. Biochimica et Biophysica Acta (BBA) - Biomembranes 1666, 250–263
(2004).
227. Baker, J. A., Wong, W.-C., Eisenhaber, B., Warwicker, J. & Eisenhaber, F.
Charged residues next to transmembrane regions revisited: “Positive-inside rule”
is complemented by the “negative inside depletion/outside enrichment rule”.
BMC Biology 15, 66 (2017).
228. Moon, C. P., Zaccai, N. R., Fleming, P. J., Gessmann, D. & Fleming, K. G.
Membrane protein thermodynamic stability may serve as the energy sink for
sorting in the periplasm. Proceedings of the National Academy of Sciences 110,
4285–4290 (2013).
229. MacCallum, J. L. & Tieleman, D. P. Hydrophobicity scales: A thermodynamic
looking glass into lipid-protein interactions. Trends in Biochemical Sciences 36,
653–662 (2011).
230. Von Heijne, G. The distribution of positively charged residues in bacterial inner
membrane proteins correlates with the trans-membrane topology. The EMBO
Journal 5, 3021–3027 (1986).
231. Von Heijne, G. & Gavel, Y. Topogenic signals in integral membrane proteins.
European Journal of Biochemistry 174, 671–678 (1988).
232. Öjemalm, K., Botelho, S. C., Stüdle, C. & von Heijne, G. Quantitative analysis
of SecYEG-mediated insertion of transmembrane α-helices into the bacterial
inner membrane. Journal of Molecular Biology 425, 2813–2822 (2013).
233. Beuming, T. & Weinstein, H. A knowledge-based scale for the analysis and prediction of buried and exposed faces of transmembrane domain proteins. Bioinformatics 20, 1822–1835 (2004).

190

BIBLIOGRAPHY

234. Zhao, G. & London, E. An amino acid “transmembrane tendency” scale that
approaches the theoretical limit to accuracy for prediction of transmembrane
helices: Relationship to biological hydrophobicity. Protein Science 15, 1987–
2001 (2006).
235. Sonnhammer, E. L., von Heijne, G & Krogh, A. A hidden Markov model for
predicting transmembrane helices in protein sequences. Proceedings 6, 175–182
(1998).
236. Käll, L., Krogh, A. & Sonnhammer, E. L. L. Advantages of combined transmembrane topology and signal peptide prediction-the Phobius web server. Nucleic
Acids Research 35 (2007).
237. Cserzö, M., Eisenhaber, F., Eisenhaber, B. & Simon, I. On filtering false positive
transmembrane protein predictions. Protein Engineering, Design and Selection
15, 745–752 (2002).
238. Cserzo, M., Eisenhaber, F., Eisenhaber, B. & Simon, I. TM or not TM: Transmembrane protein prediction with low false positive rate using DAS-TMfilter.
Bioinformatics 20, 136–137 (2004).
239. Käll, L. & Sonnhammer, E. L. Reliability of transmembrane predictions in
whole-genome data. FEBS Letters 532, 415–418 (2002).
240. Eisenhaber, B., Kuchibhatla, D., Sherman, W., Sirota, F. L., Berezovsky, I. N.,
Wong, W. C. & Eisenhaber, F. The recipe for protein sequence-based function
prediction and its implementation in the ANNOTATOR software environment.
Methods in Molecular Biology 1415, 477–506 (2016).
241. Eisenhaber, F. A decade after the first full human genome sequencing: when will
we understand our own genome? Journal of Bioinformatics and Computational
Biology 10, 1271001 (2012).
242. Sherman, W. A., Kuchibhatla, D. B., Limviphuvadh, V., Maurer-Stroh, S.,
Eisenhaber, B. & Eisenhaber, F. HPMV: Human protein mutation viewer
— relating sequence mutations to protein sequence architecture and function
changes. Journal of Bioinformatics and Computational Biology 13, 1550028
(2015).

BIBLIOGRAPHY

191

243. Sipos, L. & von Heijne, G. Predicting the topology of eukaryotic membrane
proteins. European Journal of Biochemistry 213, 1333–1340 (1993).
244. Gavel, Y., Steppuhn, J., Herrmann, R. & von Heijne, G. The ’positive-inside
rule’ applies to thylakoid membrane proteins. FEBS Letters 282, 41–46 (1991).
245. Wallin, E. & von Heijne, G. Genome-wide analysis of integral membrane proteins
from eubacterial, archaean, and eukaryotic organisms. Protein Science 7, 1029–
1038 (2008).
246. Beltzer, J. P., Fiedler, K., Fuhrer, C., Geffen, I., Handschin, C., Wessels, H. P.
& Spiess, M. Charged residues are major determinants of the transmembrane
orientation of a signal-anchor sequence. Journal of Biological Chemistry 266,
973–978 (1991).
247. Kida, Y., Morimoto, F., Mihara, K. & Sakaguchi, M. Function of positive charges
following signal-anchor sequences during translocation of the N-terminal domain. Journal of Biological Chemistry 281, 1152–1158 (2006).
248. Nilsson, I. M. & von Heijne, G. Fine-tuning the topology of a polytopic membrane protein: Role of positively and negatively charged amino acids. Cell 62,
1135–1141 (1990).
249. Bogdanov, M., Dowhan, W. & Vitrac, H. Lipids and topological rules governing
membrane protein assembly. Biochimica et Biophysica Acta - Molecular Cell
Research 1843, 1475–1488 (2014).
250. Andersson, H. & von Heijne, G. Position-specific Asp-Lys pairing can affect
signal sequence function and membrane protein topology. Journal of Biological
Chemistry 268, 21389–21393 (1993).
251. Kim, H., Paul, S., Jennity, J. & Inouye, M. Reversible topology of a bifunctional
transmembrane protein depends upon the charge balance around its transmembrane domain. Molecular Microbiology 11, 819–831 (1994).
252. Delgado-Partin, V. M. & Dalbey, R. E. The proton motive force, acting on
acidic residues, promotes translocation of amino-terminal domains of membrane
proteins when the hydrophobicity of the translocation signal is low. Journal of
Biological Chemistry 273, 9927–9934 (1998).

192

BIBLIOGRAPHY

253. Ismail, N., Hedman, R., Schiller, N. & von Heijne, G. A biphasic pulling force
acts on transmembrane helices during translocon-mediated membrane integration. Nature Structural and Molecular Biology 19, 1018–1023 (2012).
254. Hartmann, E., Rapoport, T. A. & Lodish, H. F. Predicting the orientation of
eukaryotic membrane-spanning proteins. Proceedings of the National Academy
of Sciences 86, 5786–5790 (1989).
255. Harley, C. A., Holt, J. A., Turner, R. & Tipper, D. J. Transmembrane protein
insertion orientation in yeast depends on the charge difference across transmembrane segments, their total hydrophobicity, and its distribution. Journal of
Biological Chemistry 273, 24963–24971 (1998).
256. Sato, M., Hresko, R. & Mueckler, M. Testing the charge difference hypothesis
for the assembly of a eucaryotic multispanning membrane protein. Journal of
Biological Chemistry 273, 25203–25208 (1998).
257. Dobson, L., Langó, T., Reményi, I. & Tusnády, G. E. Expediting topology data
gathering for the TOPDB database. Nucleic Acids Research 43, D283–D289
(2015).
258. Nakashima, H. & Nishikawa, K. The amino acid composition is different between
the cytoplasmic and extracellular sides in membrane proteins. English. FEBS
Letters 303, 141–146 (1992).
259. Bahadur, R. R. Rates of convergence of estimates and test statistics. The Annals
of Mathematical Statistics 38, 303–324 (1967).
260. Bahadur, R. R. Some limit theorems in statistics (Society for Industrial and
Applied Mathematics, 1971).
261. Sunyaev, S. R., Eisenhaber, F., Argos, P., Kuznetsov, E. N. & Tumanyan, V. G.
Are knowledge-based potentials derived from protein structure sets discriminative with respect to amino acid types? Proteins: Structure, Function and
Genetics 31, 225–246 (1998).
262. Zachowski, A. Phospholipids in animal eukaryotic membranes: transverse asymmetry and movement. Biochemical Journal 294, 1–14 (1993).

BIBLIOGRAPHY

193

263. Yeung, T., Gilbert, G. E., Shi, J., Silvius, J., Kapus, A. & Grinstein, S. Membrane phosphatidylserine regulates surface charge and protein localization. Science 319, 210–213 (2008).
264. Meindl-Beinker, N. M., Lundin, C., Nilsson, I., White, S. H. & von Heijne,
G. Asn- and Asp-mediated interactions between transmembrane helices during
translocon-mediated membrane protein assembly. EMBO Reports 7, 1111–1116
(2006).
265. Oger, P. M. & Cario, A. Adaptation of the membrane in Archaea. Biophysical
Chemistry 183, 42–56 (2013).
266. Rutz, C., Rosenthal, W. & Schülein, R. A single negatively charged residue affects the orientation of a membrane protein in the inner membrane of Escherichia
coli only when it is located adjacent to a transmembrane domain. Journal of
Biological Chemistry 274, 33757–33763 (1999).
267. Hedin, L. E., Öjemalm, K., Bernsel, A., Hennerdal, A., Illergård, K., Enquist,
K., Kauko, A., Cristobal, S., von Heijne, G., Lerch-Bader, M., Nilsson, I. &
Elofsson, A. Membrane insertion of marginally hydrophobic transmembrane helices depends on sequence context. Journal of Molecular Biology 396, 221–229
(2010).
268. Öjemalm, K., Halling, K. K., Nilsson, I. & von Heijne, G. Orientational preferences of neighboring helices can drive ER insertion of a marginally hydrophobic
transmembrane helix. Molecular Cell 45, 529–540 (2012).
269. Devaux, P. F. Phospholipid flippases. FEBS Letters 234, 8–12 (1988).
270. Devaux, P. F. & Morris, R. Transmembrane asymmetry and lateral domains in
biological membranes. Traffic 5, 241–246 (2004).
271. Bell, R. M., Ballas, L. M. & Coleman, R. A. Lipid topogenesis. Journal of Lipid
Research 22, 391–403 (1981).
272. Futerman, A. H. & Riezman, H. The ins and outs of sphingolipid synthesis.
Trends in Cell Biology 15, 312–318 (2005).

194

BIBLIOGRAPHY

273. Tafesse, F. G., Huitema, K., Hermansson, M., Van Der Poel, S., Van Den Dikkenberg, J., Uphoff, A., Somerharju, P. & Holthuis, J. C. M. Both sphingomyelin
synthases SMS1 and SMS2 are required for sphingomyelin homeostasis and
growth in human HeLa cells. Journal of Biological Chemistry 282, 17537–17547
(2007).
274. Di Paolo, G. & De Camilli, P. Phosphoinositides in cell regulation and membrane
dynamics. Nature 443, 651–657 (2006).
275. Killian, J. A. & von Heijne, G. How proteins adapt to a membrane-water interface. English. Trends in Biochemical Sciences 25, 429–434 (2000).
276. Kuznetsov, V., Lee, H. K., Maurer-Stroh, S., Molnár, M. J., Pongor, S., Eisenhaber, B. & Eisenhaber, F. How bioinformatics influences health informatics:
usage of biomolecular sequences, expression profiles and automated microscopic
image analyses for clinical needs and public health. Health Information Science
and Systems 1, 2 (2013).
277. Huang, Y., Niu, B., Gao, Y., Fu, L. & Li, W. CD-HIT Suite: A web server
for clustering and comparing biological sequences. Bioinformatics 26, 680–682
(2010).
278. Suzek, B. E., Wang, Y., Huang, H., McGarvey, P. B. & Wu, C. H. UniRef clusters: A comprehensive and scalable alternative for improving sequence similarity
searches. Bioinformatics 31, 926–932 (2015).
279. White, S. H. & von Heijne, G. How translocons select transmembrane helices.
en. Annual Review of Biophysics 37, 23–42 (2008).
280. Van Der Walt, S., Colbert, S. C. & Varoquaux, G. The NumPy array: A structure for efficient numerical computation. Computing in Science and Engineering
13, 22–30 (2011).
281. Osborne, A. R., Rapoport, T. A. & van den Berg, B. Protein translocation by
the Sec61/Secy channel. Annual Review of Cell and Developmental Biology 21,
529–550 (2005).

BIBLIOGRAPHY

195

282. Hockenbery, D., Nuñez, G., Milliman, C., Schreiber, R. D. & Korsmeyer, S. J.
Bcl-2 is an inner mitochondrial membrane protein that blocks programmed cell
death. Nature 348, 334–336 (1990).
283. Ungar, D. & Hughson, F. M. SNARE protein structure and function. Annual
Review of Cell and Developmental Biology 19, 493–517 (2003).
284. Kalbfleisch, T., Cambon, A. & Wattenberg, B. W. A bioinformatics approach
to identifying tail-anchored proteins in the human genome. Traffic 8, 1687–1694
(2007).
285. Allen, T. M. & Cullis, P. R. Liposomal drug delivery systems: From concept to
clinical applications. Advanced Drug Delivery Reviews 65, 36–48 (2013).
286. Nordlund, G., Brzezinski, P. & Von Ballmoos, C. SNARE-fusion mediated insertion of membrane proteins into native and artificial membranes. Nature Communications 5, 4303 (2014).
287. Kutay, U., Hartmann, E. & Rapoport, T. A. A class of membrane proteins with
a C-terminal anchor. Trends in Cell Biology 3, 72–75 (1993).
288. Rabu, C., Wipf, P., Brodsky, J. L. & High, S. A precursor-specific role for
Hsp40/Hsc70 during tail-anchored protein integration at the endoplasmic reticulum. Journal of Biological Chemistry 283, 27504–27513 (2008).
289. Rabu, C., Schmid, V., Schwappach, B. & High, S. Biogenesis of tail-anchored
proteins: the beginning for the end? Journal of Cell Science 122, 3605–3612
(2009).
290. Johnson, N., Powis, K. & High, S. Post-translational translocation into the
endoplasmic reticulum. Biochimica et Biophysica Acta - Molecular Cell Research
1833, 2403–2409 (2013).
291. Abell, B. M., Pool, M. R., Schlenker, O., Sinning, I. & High, S. Signal recognition
particle mediates post-translational targeting in eukaryotes. EMBO Journal 23,
2755–2764 (2004).
292. Leznicki, P., Clancy, A., Schwappach, B. & High, S. Bat3 promotes the membrane integration of tail-anchored proteins. Journal of Cell Science 123, 2170–
2178 (2010).

196

BIBLIOGRAPHY

293. Abell, B. M., Jung, M., Oliver, J. D., Knight, B. C., Tyedmers, J., Zimmermann, R. & High, S. Tail-anchored and signal-anchored proteins utilize overlapping pathways during membrane insertion. Journal of Biological Chemistry
278, 5669–5678 (2003).
294. Steel, G. J., Brownsword, J. & Stirling, C. J. Tail-anchored protein insertion into
yeast ER requires a novel posttranslational mechanism which is independent of
the SEC machinery. Biochemistry 41, 11914–11920 (2002).
295. Yabal, M., Brambillasca, S., Soffientini, P., Pedrazzini, E., Borgese, N. &
Makarow, M. Translocation of the C terminus of a tail-anchored protein across
the endoplasmic reticulum membrane in yeast mutants defective in signal
peptide-driven translocation. Journal of Biological Chemistry 278, 3489–3496
(2003).
296. Casson, J., McKenna, M., Haßdenteufel, S., Aviram, N., Zimmerman, R. &
High, S. Multiple pathways facilitate the biogenesis of mammalian tail-anchored
proteins. Journal of Cell Science 130, 3851–3861 (2017).
297. Schuldiner, M., Metz, J., Schmid, V., Denic, V., Rakwalska, M., Schmitt, H. D.,
Schwappach, B. & Weissman, J. S. The GET complex mediates insertion of
tail-anchored proteins into the ER membrane. Cell 134, 634–645 (2008).
298. Wang, F., Brown, E. C., Mak, G., Zhuang, J. & Denic, V. A chaperone cascade
sorts proteins for posttranslational membrane insertion into the endoplasmic
reticulum. Molecular Cell 40, 159–171 (2010).
299. Chang, Y. W., Chuang, Y. C., Ho, Y. C., Cheng, M. Y., Sun, Y. J., Hsiao, C. D.
& Wang, C. Crystal structure of Get4-Get5 complex and its interactions with
Sgt2, Get3, and Ydj1. Journal of Biological Chemistry 285, 9962–9970 (2010).
300. Chang, Y. W., Lin, T. W., Li, Y. C., Huang, Y. S., Sun, Y. J. & Hsiao, C. D.
Interaction surface and topology of Get3-Get4-Get5 protein complex, involved in
targeting tail-anchored proteins to endoplasmic reticulum. Journal of Biological
Chemistry 287, 4783–4789 (2012).
301. Chartron, J. W., Suloway, C. J. M., Zaslaver, M. & Clemons, W. M. Structural characterization of the Get4/Get5 complex and its interaction with Get3.
Proceedings of the National Academy of Sciences 107, 12127–12132 (2010).

BIBLIOGRAPHY

197

302. Chartron, J. W., VanderVelde, D. G., Rao, M. & Clemons, W. M. Get5
carboxyl-terminal domain is a novel dimerization motif that tethers an extended
Get4/Get5 complex. Journal of Biological Chemistry 287, 8310–8317 (2012).
303. Shao, S., Rodrigo-Brenni, M. C., Kivlen, M. H. & Hegde, R. S. Mechanistic
basis for a molecular triage reaction. Science 355, 298–302 (2017).
304. Mateja, A., Szlachcic, A., Downing, M. E., Dobosz, M., Mariappan, M., Hegde,
R. S. & Keenan, R. J. The structural basis of tail-anchored membrane protein
recognition by Get3. Nature 461, 361–366 (2009).
305. Hu, J., Li, J., Qian, X., Denic, V. & Sha, B. The crystal structures of yeast Get3
suggest a mechanism for tail-anchored protein membrane insertion. PLoS ONE
4 (ed Mayer, C.) e8061 (2009).
306. Bozkurt, G., Stjepanovic, G., Vilardi, F., Amlacher, S., Wild, K., Bange, G.,
Favaloro, V., Rippe, K., Hurt, E., Dobberstein, B. & Sinning, I. Structural
insights into tail-anchored protein binding and membrane insertion by Get3.
Proceedings of the National Academy of Sciences of the United States of America
106, 21131–6 (2009).
307. Suloway, C. J. M., Chartron, J. W., Zaslaver, M. & Clemons, W. M. Model
for eukaryotic tail-anchored protein binding based on the structure of Get3.
Proceedings of the National Academy of Sciences of the United States of America
106, 14849–14854 (2009).
308. Yamagata, A., Mimura, H., Sato, Y., Yamashita, M., Yoshikawa, A. & Fukai,
S. Structural insight into the membrane insertion of tail-anchored proteins by
Get3. Genes to Cells 15, 29–41 (2010).
309. Stefanovic, S. & Hegde, R. S. Identification of a targeting factor for posttranslational membrane protein insertion into the ER. Cell 128, 1147–1159 (2007).
310. Favaloro, V., Spasic, M., Schwappach, B. & Dobberstein, B. Distinct targeting
pathways for the membrane insertion of tail-anchored (TA) proteins. Journal of
Cell Science 121, 1832–1840 (2008).

198

BIBLIOGRAPHY

311. Jonikas, M. C., Collins, S. R., Denic, V., Oh, E., Quan, E. M., Schmid, V.,
Weibezahn, J., Schwappach, B., Walter, P., Weissman, J. S. & Schuldiner, M.
Comprehensive characterization of genes required for protein folding in the endoplasmic reticulum. Science 323, 1693–1697 (2009).
312. Huh, W. K., Falvo, J. V., Gerke, L. C., Carroll, A. S., Howson, R. W., Weissman,
J. S. & O’Shea, E. K. Global analysis of protein localization in budding yeast.
Nature 425, 686–691 (2003).
313. Zalisko, B. E., Chan, C., Denic, V., Rock, R. S. & Keenan, R. J. Tail-anchored
protein insertion by a single Get1/2 heterodimer. Cell Reports 20, 2287–2293
(2017).
314. Vitali, D. G., Sinzel, M., Bulthuis, E. P., Kolb, A., Zabel, S., Mehlhorn, D. G.,
Costa, B. F., Farkas, Á., Clancy, A., Schuldiner, M., Grefen, C., Schwappach,
B., Borgese, N. & Rapaport, D. The GET pathway can increase the risk of
mitochondrial outer membrane proteins to be mistargeted to the ER. Journal
of cell science 131, 734–764 (2018).
315. Mariappan, M., Mateja, A., Dobosz, M., Bove, E., Hegde, R. S. & Keenan, R. J.
The mechanism of membrane-associated steps in tail-anchored protein insertion.
Nature 477, 61–69 (2011).
316. Stefer, S., Reitz, S., Wang, F., Wild, K., Pang, Y. Y., Schwarz, D., Bomke, J.,
Hein, C., Löhr, F., Bernhard, F., Denic, V., Dötsch, V. & Sinning, I. Structural basis for tail-anchored membrane protein biogenesis by the Get3-receptor
complex. Science 333, 758–762 (2011).
317. Rao, M., Okreglak, V., Chio, U. S., Cho, H., Walter, P. & Shan, S. O. Multiple
selection filters ensure accurate tail-anchored membrane protein targeting. eLife
5 (2016).
318. Aviram, N., Ast, T., Costa, E. A., Arakel, E. C., Chuartzman, S. G., Jan,
C. H., Haßdenteufel, S., Dudek, J., Jung, M., Schorr, S., Zimmermann, R.,
Schwappach, B., Weissman, J. S. & Schuldiner, M. The SND proteins constitute
an alternative targeting route to the endoplasmic reticulum. Nature 540, 134–
138 (2016).

BIBLIOGRAPHY

199

319. Haßdenteufel, S., Sicking, M., Schorr, S., Aviram, N., Fecher-Trost, C.,
Schuldiner, M., Jung, M., Zimmermann, R. & Lang, S. hSnd2 protein represents an alternative targeting factor to the endoplasmic reticulum in human
cells. FEBS Letters 591, 3211–3224 (2017).
320. Setoguchi, K., Otera, H. & Mihara, K. Cytosolic factor- and TOM-independent
import of C-tail-anchored mitochondrial outer membrane proteins. EMBO Journal 25, 5635–5647 (2006).
321. Kemper, C., Habib, S. J., Engl, G., Heckmeyer, P., Dimmer, K. S. & Rapaport,
D. Integration of tail-anchored proteins into the mitochondrial outer membrane
does not require any known import components. Journal of Cell Science 121,
1990–1998 (2008).
322. Ngosuwan, J., Wang, N. M., Fung, K. L. & Chirico, W. J. Roles of cytosolic Hsp70 and Hsp40 molecular chaperones in post-translational translocation
of presecretory proteins into the endoplasmic reticulum. Journal of Biological
Chemistry 278, 7034–7042 (2003).
323. Colombo, S. F., Longhi, R. & Borgese, N. The role of cytosolic proteins in the
insertion of tail-anchored proteins into phospholipid bilayers. Journal of Cell
Science 122, 2383–2392 (2009).
324. Meineke, B., Engl, G., Kemper, C., Vasiljev-Neumeyer, A., Paulitschke, H. &
Rapaport, D. The outer membrane form of the mitochondrial protein Mcr1
follows a TOM-independent membrane insertion pathway. FEBS Letters 582,
855–860 (2008).
325. Ulmschneider, M. B., Ulmschneider, J. P., Schiller, N., Wallace, B. A., von Heijne, G. & White, S. H. Spontaneous transmembrane helix insertion thermodynamically mimics translocon-guided insertion. en. Nature Communications 5,
4863 (2014).
326. Chen, Y., Pieuchot, L., Loh, R. A., Yang, J., Kari, T. M. A., Wong, J. Y. &
Jedd, G. Hydrophobic handoff for direct delivery of peroxisome tail-anchored
proteins. Nature Communications 5, 5790 (2014).

200

BIBLIOGRAPHY

327. Yagita, Y., Hiromasa, T. & Fujiki, Y. Tail-anchored PEX26 targets peroxisomes
via a PEX19-dependent and TRC40-independent class I pathway. Journal of
Cell Biology 200, 651–666 (2013).
328. Costello, J. L., Castro, I. G., Camões, F., Schrader, T. A., McNeall, D., Yang, J.,
Giannopoulou, E.-A., Gomes, S., Pogenberg, V., Bonekamp, N. A., Ribeiro, D.,
Wilmanns, M., Jedd, G., Islinger, M. & Schrader, M. Predicting the targeting of
tail-anchored proteins to subcellular compartments in mammalian cells. Journal
of Cell Science 130, 1675–1687 (2017).
329. Costa, E. A., Subramanian, K., Nunnari, J. & Weissman, J. S. Defining the
physiological role of SRP in protein-targeting efficiency and specificity. Science
(New York, N.Y.) 359, 689–692 (2018).
330. Fueller, J., Egorov, M. V., Walther, K. A., Sabet, O., Mallah, J., Grabenbauer,
M. & Kinkhabwala, A. Subcellular partitioning of protein tyrosine Phosphatase
1b to the endoplasmic reticulum and mitochondria Dependssensitively on the
composition of its tail anchor. PLoS ONE 10 (ed Chen, Y.) e0139429 (2015).
331. Wu, S., Zhu, Z., Fu, L., Niu, B. & Li, W. WebMGA: A customizable web server
for fast metagenomic sequence analysis. BMC Genomics 12, 444 (2011).
332. Söding, J., Biegert, A. & Lupas, A. N. The HHpred interactive server for protein
homology detection and structure prediction. Nucleic Acids Research 33, W244–
W248 (2005).
333. Eswar, N., Webb, B., Marti-Renom, M. A., Madhusudhan, M., Eramian, D.,
Shen, M.-y., Pieper, U. & Sali, A. Comparative protein structure modeling using
MODELLER. Current Protocols in Protein Science 50, 2.9.1–2.9.31 (2007).
334. Webb, B. & Sali, A. in Current Protocols in Protein Science 1, 2.9.1–2.9.37
(John Wiley & Sons, Inc., Hoboken, NJ, USA, 2016). arXiv: NIHMS150003.
335. Hofmann, K & Stoffel, W. 103. Konferenz der Gesellschaft für Biologische Chemie Studiengruppe Membranstruktur und -Transport Membranforum
Frankfurt. Biological Chemistry Hoppe-Seyler 374, 143–170 (1993).

BIBLIOGRAPHY

201

336. Cao, B., Porollo, A., Adamczak, R., Jarrell, M. & Meller, J. Enhanced recognition of protein transmembrane domains with prediction-based structural profiles. Bioinformatics 22, 303–309 (2006).
337. Montgomerie, S., Cruz, J. A., Shrivastava, S., Arndt, D., Berjanskii, M. &
Wishart, D. S. PROTEUS2: a web server for comprehensive protein structure
prediction and structure-based annotation. Nucleic acids research 36, W202–
W209 (2008).
338. Kurowski, M. A. & Bujnicki, J. M. GeneSilico protein structure prediction metaserver. Nucleic Acids Research 31, 3305–3307 (2003).
339. Baker, N. A., Sept, D., Joseph, S., Holst, M. J. & McCammon, J. A. Electrostatics of nanosystems: Application to microtubules and the ribosome. Proceedings
of the National Academy of Sciences 98, 10037–10041 (2001).
340. Ashkenazy, H., Erez, E., Martz, E., Pupko, T. & Ben-Tal, N. ConSurf 2010:
Calculating evolutionary conservation in sequence and structure of proteins and
nucleic acids. Nucleic Acids Research 38, W529–W533 (2010).
341. Ramsay, L., Macaulay, M., Degli Ivanissevich, S., MacLean, K., Cardle, L.,
Fuller, J., Edwards, K. J., Tuvesson, S., Morgante, M., Massari, A., Maestri,
E., Marmiroli, N., Sjakste, T., Ganal, M., Powell, W. & Waugh, R. A simple
sequence repeat-based linkage map of Barley. Genetics 156, 1997–2005 (2000).
342. Choi, J. Y., Wu, W. I. & Voelker, D. R. Phosphatidylserine decarboxylases
as genetic and biochemical tools for studying phospholipid traffic. Analytical
Biochemistry 347, 165–175 (2005).
343. Gebert, N., Joshi, A. S., Kutik, S., Becker, T., McKenzie, M., Guan, X. L.,
Mooga, V. P., Stroud, D. A., Kulkarni, G., Wenk, M. R., Rehling, P., Meisinger,
C., Ryan, M. T., Wiedemann, N., Greenberg, M. L. & Pfanner, N. Mitochondrial cardiolipin involved in outer-membrane protein biogenesis: Implications for
Barth syndrome. Current Biology 19, 2133–2139 (2009).
344. Borgese, N., Brambillasca, S. & Colombo, S. How tails guide tail-anchored proteins to their destinations. Current Opinion in Cell Biology 19, 368–375 (2007).

202

BIBLIOGRAPHY

345. Lan, L, Isenmann, S & Wattenberg, B. W. Targeting and insertion of Cterminally anchored proteins to the mitochondrial outer membrane is specific
and saturable but does not strictly require ATP or molecular chaperones. The
Biochemical journal 349, 611–621 (2000).
346. Saraogi, I. & Shan, S.-o. Molecular mechanism of co-translational protein targeting by the signal recognition particle. Traffic 12, 535–542 (2011).
347. Raoult, D. & Forterre, P. Redefining viruses: lessons from Mimivirus. Nature
Reviews Microbiology 6, 315–319 (2008).
348. Lu, J. & Deutsch, C. Secondary structure formation of a transmembrane segment in Kv channels. Biochemistry 44, 8230–8243 (2005).
349. Tu, L. W. & Deutsch, C. A folding zone in the ribosomal exit tunnel for Kv1.3
helix formation. Journal of Molecular Biology 396, 1346–1360 (2010).
350. Tu, L., Khanna, P. & Deutsch, C. Transmembrane segments form tertiary hairpins in the folding vestibule of the ribosome. Journal of Molecular Biology 426,
185–198 (2014).
351. Kudva, R., Pardo-Avila, F., Sandhu, H., Carroni, M., Bernstein, H. D. & von
Heijne, G. The shape of the ribosome exit tunnel affects cotranslational protein
folding. bioRxiv, 274191 (2018).
352. Ito, K., Chiba, S. & Pogliano, K. Divergent stalling sequences sense and control cellular physiology. Biochemical and Biophysical Research Communications
393, 1–5 (2010).
353. Butkus, M. E., Prundeanu, L. B. & Oliver, D. B. Translocon ”pulling” of nascent
SecM controls the duration of its translational pause and secretion-responsive
secA regulation. Journal of Bacteriology 185, 6719–6722 (2003).
354. Yap, M. N. & Bernstein, H. D. The plasticity of a translation arrest motif
yields insights into nascent polypeptide recognition inside the ribosome tunnel.
Molecular Cell 34, 201–211 (2009).
355. MacKinnon, A. L., Paavilainen, V. O., Sharma, A., Hegde, R. S. & Taunton, J.
An allosteric Sec61 inhibitor traps nascent transmembrane helices at the lateral
gate. eLife 2014 (2014).

BIBLIOGRAPHY

203

356. Li, L., Park, E., Ling, J. J., Ingram, J., Ploegh, H. & Rapoport, T. A. Crystal structure of a substrate-engaged SecY protein-translocation channel. Nature
531, 395–399 (2016).
357. Lu, J. & Deutsch, C. Electrostatics in the ribosomal tunnel modulate chain
elongation rates. Journal of Molecular Biology 384, 73–86 (2008).
358. Sadlish, H., Pitonzo, D., Johnson, A. E. & Skach, W. R. Sequential triage of
transmembrane segments by Sec61α during biogenesis of a native multispanning
membrane protein. Nature Structural and Molecular Biology 12, 870–878 (2005).
359. Cross, B. C. S. & High, S. Dissecting the physiological role of selective
transmembrane-segment retention at the ER translocon. Journal of Cell Science 122, 1768–1777 (2009).
360. Cymer, F. & von Heijne, G. Cotranslational folding of membrane proteins
probed by arrest-peptide–mediated force measurements. Proceedings of the National Academy of Sciences 110, 14640–14645 (2013).
361. Virkki, M. T., Peters, C., Nilsson, D., Sörensen, T., Cristobal, S., Wallner, B. &
Elofsson, A. The positive inside rule is stronger when followed by a transmembrane helix. Journal of Molecular Biology 426, 2982–2991 (2014).
362. Junne, T. & Spiess, M. Integration of transmembrane domains is regulated by
their downstream sequences. Journal of Cell Science 130, 372–381 (2017).
363. Ismail, N., Crawshaw, S., Cross, B., Haagsma, A. & High, S. Specific transmembrane segments are selectively delayed at the ER translocon during opsin
biogenesis. Biochemical Journal 411, 495–506 (2008).
364. Sato, Y., Sakaguchi, M., Goshima, S., Nakamura, T. & Uozumi, N. Integration
of Shaker-type K+ channel, KAT1, into the endoplasmic reticulum membrane:
synergistic insertion of voltage-sensing segments, S3-S4, and independent insertion of pore-forming segments, S5-P-S6. Proc Natl Acad Sci U S A 99, 60–5
(2002).
365. Sato, Y., Sakaguchi, M., Goshima, S., Nakamura, T. & Uozumi, N. Molecular
dissection of the contribution of negatively and positively charged residues in

204

BIBLIOGRAPHY
S2, S3, and S4 to the final membrane topology of the voltage sensor in the K+
channel, KAT1. Journal of Biological Chemistry 278, 13227–13234 (2003).

366. Zhang, L., Sato, Y., Hessa, T., von Heijne, G., Lee, J.-K., Kodama, I., Sakaguchi,
M. & Uozumi, N. Contribution of hydrophobic and electrostatic interactions to
the membrane integration of the Shaker K+ channel voltage sensor domain.
Proceedings of the National Academy of Sciences 104, 8263–8268 (2007).
367. Tu, L., Wang, J., Helm, A., Skach, W. & Deutsch, C. Transmembrane biogenesis
of Kv1. 3. Biochemistry 39, 824–836 (2000).
368. Mi, H., Huang, X., Muruganujan, A., Tang, H., Mills, C., Kang, D. & Thomas,
P. D. PANTHER version 11: expanded annotation data from Gene Ontology
and Reactome pathways, and data analysis tool enhancements. Nucleic Acids
Research 45, D183–D189 (2017).
369. Oliphant, T. E. SciPy: Open source scientific tools for Python 2007.
370. Remm, M & Sonnhammer, E. Classification of transmembrane protein families
in the Caenorhabditis elegans genome and indentification of human orthologs.
Genome Res. 10, 1679–1689 (2000).
371. O’Hayre, M., Vázquez-Prado, J., Kufareva, I., Stawiski, E. W., Handel, T. M.,
Seshagiri, S. & Gutkind, J. S. The emerging mutational landscape of G proteins
and G-protein-coupled receptors in cancer. Nature Reviews Cancer 13, 412–424
(2013).
372. Singh, A., Nunes, J. J. & Ateeq, B. Role and therapeutic potential of G-protein
coupled receptors in breast cancer progression and metastases. European Journal
of Pharmacology 763, 178–183 (2015).
373. Bar-Shavit, R., Maoz, M., Kancharla, A., Nag, J. K., Agranovich, D., GrisaruGranovsky, S. & Uziely, B. G protein-coupled receptors in cancer. International
Journal of Molecular Sciences 17, 1320 (2016).
374. Arakaki, A., Pan, W.-a. & Trejo, J. GPCRs in cancer: protease-activated receptors, endocytic adaptors and signaling. International Journal of Molecular
Sciences 19, 1886 (2018).

BIBLIOGRAPHY

205

375. DeLano, W. L. The PyMOL molecular graphics system. PyMOL. DeLano Scientific, San Carlos, CA, USA, Version 1.2r3pre (2002).
376. Okada, T., Sugihara, M., Bondar, A.-N., Elstner, M., Entel, P. & Buss, V. The
retinal conformation and its environment in rhodopsin in light of a new 2.2 Å
crystal structure. Journal of Molecular Biology 342, 571–583 (2004).
377. Murakami, M. & Kouyama, T. Crystal structure of squid rhodopsin. Nature
453, 363–367 (2008).
378. Warne, T., Serrano-Vega, M. J., Baker, J. G., Moukhametzianov, R., Edwards,
P. C., Henderson, R., Leslie, A. G. W., Tate, C. G. & Schertler, G. F. X. Structure of a β1-adrenergic G-protein-coupled receptor. Nature 454, 486–491 (2008).
379. Tan, Q., Zhu, Y., Li, J., Chen, Z., Han, G. W., Kufareva, I., Li, T., Ma, L.,
Fenalti, G., Li, J., Zhang, W., Xie, X., Yang, H., Jiang, H., Cherezov, V., Liu, H.,
Stevens, R. C., Zhao, Q. & Wu, B. Structure of the CCR5 chemokine receptorHIV entry inhibitor maraviroc complex. Science (New York, N.Y.) 341, 1387–
90 (2013).
380. Burg, J. S., Ingram, J. R., Venkatakrishnan, A. J., Jude, K. M., Dukkipati,
A., Feinberg, E. N., Angelini, A., Waghray, D., Dror, R. O., Ploegh, H. L. &
Garcia, K. C. Structural biology. Structural basis for chemokine recognition and
activation of a viral G protein-coupled receptor. Science (New York, N.Y.) 347,
1113–7 (2015).
381. Thompson, A. A., Liu, W., Chun, E., Katritch, V., Wu, H., Vardy, E., Huang,
X.-P., Trapella, C., Guerrini, R., Calo, G., Roth, B. L., Cherezov, V. & Stevens,
R. C. Structure of the nociceptin/orphanin FQ receptor in complex with a peptide mimetic. Nature 485, 395–399 (2012).
382. Kang, M. K. & Tullman-Ercek, D. Engineering expression and function of membrane proteins. Methods (2018).
383. Findlay, H. E. & Booth, P. J. The folding, stability and function of lactose permease differ in their dependence on bilayer lipid composition. Scientific Reports
7, 13056 (2017).

206

BIBLIOGRAPHY

384. Lu, P., Min, D., DiMaio, F., Wei, K. Y., Vahey, M. D., Boyken, S. E., Chen, Z.,
Fallas, J. A., Ueda, G., Sheffler, W., Mulligan, V. K., Xu, W., Bowie, J. U. &
Baker, D. Accurate computational design of multipass transmembrane proteins.
Science 359, 1042–1046 (2018).