a study of electrogenic transient and steady-state ... a study of electrogenic transient and...

174
A Study of Electrogenic Transient and Steady-State Cotransporter Kinetics: Investigations with the Na + /Glucose Transporter SGLT1 by Daniel Krofchick A thesis submitted in conformity with the requirements for the degree of Doctor of Philosophy Institute of Medical Science University of Toronto © Copyright by Daniel Krofchick 2012

Upload: doananh

Post on 20-May-2018

235 views

Category:

Documents


1 download

TRANSCRIPT

A Study of Electrogenic Transient and Steady-State Cotransporter Kinetics: Investigations with the

Na+/Glucose Transporter SGLT1

by

Daniel Krofchick

A thesis submitted in conformity with the requirements for the degree of Doctor of Philosophy

Institute of Medical Science University of Toronto

© Copyright by Daniel Krofchick 2012

ii

A Study of Electrogenic Transient and Steady-State

Cotransporter Kinetics: Investigations with the Na+/Glucose

Transporter SGLT1

Daniel Krofchick

Doctor of Philosophy

Institute of Medical Science

University of Toronto

2012

Abstract

Significant advancements in the field of membrane protein crystallography have provided in

recent years invaluable images of transporter structures. These structures, however, are static

and require complementary kinetic insight to understand how their mechanisms work.

Electrophysiological studies of transporters permit the high quality kinetic measurements

desired, but there are significant difficulties involved in analyzing and interpreting the data.

Current methods allow a variety of kinetic parameters to be measured but there is a disconnect

between these parameters and a fundamental understanding of the carrier. The intent of this

research was to contribute new tools for studying the electrogenic kinetics of membrane

transport proteins, to understand the link between these kinetics and the carrier, and to

ultimately understand the mechanisms involved in transport. In this vein, two projects are

explored covering two important kinetic time domains, transient and steady-state. The transient

project studies the conformational changes of the unloaded carrier of SGLT1 through a multi-

exponential analysis of the transient currents. Crystal structures have potentially identified a

gated rocker-switch mechanism and the transient kinetics are used to support and study this

kinetically. A protocol taking advantage of multiple holding potentials is used to measure the

iii

decay time constants and charge movements for voltage jumps from both hyperpolarizing and

depolarizing directions. These directional measurements provide insight into the arrangement of

the observed transitions through directional inequalities in charge movement, by considering the

potential for a slow transition to hide a faster one. Ultimately, four carrier decays are observed

that align with the gated rocker-switch mechanism and can be associated one-to-one with the

movement of a gate and pore on each side of the membrane. The steady-state project considers a

general theoretical model of transporter cycling. Recursive patterns are identified in the steady-

state velocity equation that lead to a broad understanding of its geometric properties as a

function of voltage and substrate concentration. This results in a simple phenomenological

method for characterizing the I–V curves and for measuring the kinetics of rate limiting patterns

in the loop, which we find are the basic structures revealed by the steady-state velocity.

iv

Acknowledgments

To my grandparent, who began this journey with me and are no longer here. My parents, who’s

support and love kept me going. And most of all Kate, who made this all possible.

v

Table of Contents

Acknowledgments .......................................................................................................................... iv

Table of Contents ............................................................................................................................. v

List of Tables ................................................................................................................................... x

List of Figures ................................................................................................................................. xi

List of Abbreviations .................................................................................................................... xiv

List of Variables ............................................................................................................................ xv

Introduction ................................................................................................................................. 1 1

1.1 Types of Transport ............................................................................................................... 4

1.1.1 Facilitated Diffusion ................................................................................................ 4

1.1.2 Primary Active Transport ........................................................................................ 5

1.1.3 Secondary Active Transport .................................................................................... 5

1.2 Familial Relations of SGLT1............................................................................................... 6

1.2.1 SLC5 Family ............................................................................................................ 7

1.2.2 The Solute:Sodium Symporter Family .................................................................. 12

1.2.3 Phylogenetic Homology ........................................................................................ 12

1.2.4 Structural Homology ............................................................................................. 13

1.2.5 Figures ................................................................................................................... 15

1.3 Crystal Structures of the LeuT Fold .................................................................................. 19

1.3.1 Timeline ................................................................................................................. 19

1.3.2 Architecture ........................................................................................................... 20

1.3.3 Solved Conformations ........................................................................................... 20

1.3.4 Pores ...................................................................................................................... 21

1.3.5 Thin Gates .............................................................................................................. 22

vi

1.3.6 Substrate Site ......................................................................................................... 23

1.3.7 Cation Sites ............................................................................................................ 24

1.3.8 Inhibitor Sites......................................................................................................... 25

1.3.9 Thick Gates ............................................................................................................ 26

1.3.10 Conformational Changes ....................................................................................... 27

1.3.11 Transport Model .................................................................................................... 28

1.3.12 Figures ................................................................................................................... 29

1.4 Rationale ............................................................................................................................ 40

Dissecting the Transient Current of SGLT1 ............................................................................. 43 2

2.1 Introduction........................................................................................................................ 43

2.1.1 The T156C Mutant ................................................................................................ 45

2.1.2 Historical Perspective ............................................................................................ 46

2.1.3 This Study .............................................................................................................. 49

2.2 Materials an Methods ........................................................................................................ 50

2.2.1 Molecular Biology ................................................................................................. 50

2.2.2 Oocyte Collection, Injection, and Maintenance .................................................... 50

2.2.3 Two-Electrode Voltage-Clamp .............................................................................. 51

2.2.4 Voltage-Clamp Protocol ........................................................................................ 51

2.2.5 Exponential Curve Fitting...................................................................................... 52

2.3 Voltage Jump Experiment Theory ..................................................................................... 53

2.3.1 Anatomy of a Voltage Jump .................................................................................. 53

2.3.2 The Transient ......................................................................................................... 53

2.3.3 How the Voltage Jump Affects the Transient........................................................ 54

2.3.4 How the System Affects the Transient .................................................................. 54

2.3.5 Voltage Jump Protocols ......................................................................................... 55

2.4 Data Analysis ..................................................................................................................... 56

vii

2.4.1 Form of the Transient Currents ............................................................................. 56

2.4.2 Defining the Data Set............................................................................................. 56

2.4.3 Fitting ..................................................................................................................... 57

2.4.4 Fit Quality: Residuals and χ2 ................................................................................. 58

2.4.5 Nonsense Fits ......................................................................................................... 58

2.4.6 Seeding .................................................................................................................. 59

2.4.7 Stopping ................................................................................................................. 61

2.4.8 Parameter Variation with the Number of Terms ................................................... 61

2.4.9 Looking at the Dataset as a Whole ........................................................................ 62

2.5 Results ............................................................................................................................... 63

2.5.1 Transient Kinetics of wt and the T156C mutant .................................................... 63

2.5.2 Capacitive Decay ................................................................................................... 63

2.5.3 Carrier Decays: Charge Movements ...................................................................... 64

2.5.4 Carrier Decays: Time Constants ............................................................................ 64

2.5.5 Phloridzin and Non-Injected Controls ................................................................... 65

2.5.6 Limiting Terms and its Effect on the Measured Kinetics ...................................... 65

2.5.7 Using Expected Behavior to Predict the Number of Terms .................................. 66

2.5.8 Dependence of the Decay Charges on the Holding and Test Potentials................ 67

2.5.9 Supplementary Data............................................................................................... 68

2.6 Discussion .......................................................................................................................... 68

2.6.1 Ordering the Transitions ........................................................................................ 68

2.6.2 Carrier Conformations ........................................................................................... 69

2.6.3 Functional Insights................................................................................................. 70

2.6.4 Building a Kinetic Model ...................................................................................... 71

2.7 Summary and Conclusion .................................................................................................. 71

2.8 Future Work ....................................................................................................................... 73

viii

2.8.1 Simulations ............................................................................................................ 73

2.8.2 Strategic Mutants ................................................................................................... 74

2.8.3 Substrate Transients ............................................................................................... 76

2.8.4 Other Carriers and Mechanisms ............................................................................ 76

2.8.5 Miscellaneous ........................................................................................................ 78

2.9 Figures ............................................................................................................................... 79

A Practical Method for Characterizing the Voltage and Substrate Dependence of 3

Membrane Transporter Steady-State Currents........................................................................ 101

3.1 Introduction...................................................................................................................... 101

3.1.1 Historical Perspective .......................................................................................... 102

3.1.2 This Study ............................................................................................................ 104

3.2 Steady-State Velocity of a Cyclical Model ..................................................................... 105

3.3 Voltage Dependence ........................................................................................................ 106

3.3.1 Introducing Voltage Dependence ........................................................................ 106

3.3.2 The General Voltage Dependent Equation .......................................................... 108

3.3.3 Geometric Properties of the I–V Curves ............................................................. 110

3.4 Substrate Dependence ...................................................................................................... 112

3.4.1 Introducing Substrate Dependence ...................................................................... 112

3.4.2 Characteristics of Substrate Dependence............................................................. 113

3.5 Results ............................................................................................................................. 115

3.5.1 Characterizing Experimental Data ....................................................................... 115

3.5.2 Modeling the Steady-State Velocity .................................................................... 116

3.6 Summary and Conclusion ................................................................................................ 118

3.7 Future Work ..................................................................................................................... 119

3.8 Appendix.......................................................................................................................... 120

3.8.1 Deriving and Arranging the Steady-State Equation ............................................ 120

3.8.2 Simplifying the Voltage Dependent Expressions ................................................ 121

ix

3.8.3 Simplifying the Substrate Dependent Expression ............................................... 123

3.8.4 Two Substrate Binding Events ............................................................................ 124

3.9 Figures ............................................................................................................................. 126

Conclusion and Future Work .................................................................................................. 138 4

References.................................................................................................................................... 140

x

List of Tables

Table 1: Properties of SLC5 and select TC 2.A.21 members ........................................................ 15

Table 2: Properties of the LeuT fold transporters .......................................................................... 29

Table 3: Gate, substrate and cation interacting residues for the LeuT fold transporters ............... 30

Table 4: Nonsense fit examples ..................................................................................................... 88

Table 5: Seeding examples for multi-exponential fitting .............................................................. 89

Table 6: Example voltage dependent map of a transient kinetics exponential fit analysis. .......... 91

xi

List of Figures

Fig. 1: Classical SGLT1 transport model ...................................................................................... 16

Fig. 2: Homology of the SLC5 and SSS families .......................................................................... 17

Fig. 3: Phylogenic tree of secondary active transport families with solved crystal structures ...... 18

Fig. 4: Timeline of discovery for the LeuT fold structures ........................................................... 31

Fig. 5: Organization of the TM segments for the LeuT fold structures ......................................... 32

Fig. 6: Extracellular and intracellular pores demonstrated by LeuT, BetP and vSGLT ................ 33

Fig. 7: Gating mechanisms ............................................................................................................ 34

Fig. 8: Substrate binding site ......................................................................................................... 35

Fig. 9: Cation binding sites ............................................................................................................ 36

Fig. 10: LeuT with multiple bound substrates ............................................................................... 37

Fig. 11: Competitive and non-competitive inhibitor binding sites in LeuT .................................. 38

Fig. 12: Transport model predicted by the various conformations of the LeuT architecture

captured in crystal structures ......................................................................................................... 39

Fig. 13: Characteristics of the T156C mutant ................................................................................ 79

Fig. 14: Position of the T156 and K157 residues of SGLT1 in the vSGLT structure ................... 80

Fig. 15: Anatomy of a voltage jump .............................................................................................. 81

Fig. 16: Relationship between voltage jumps and transient kinetics ............................................. 82

Fig. 17: Hypothetical three-state system illustrating the masking effect....................................... 83

Fig. 18: Single and multi-holding voltage clamp protocols .......................................................... 84

Fig. 19: Example multi-holding data set........................................................................................ 85

xii

Fig. 20: Multi-exponential fit of a transient data set ..................................................................... 86

Fig. 21: Fit residuals ...................................................................................................................... 87

Fig. 22: Transient charge movements by component .................................................................... 90

Fig. 23: Transient kinetics of SGLT1 ............................................................................................ 92

Fig. 24: An expanded overview of the transient kinetic data ........................................................ 93

Fig. 25: Close up of the transient kinetic data ............................................................................... 94

Fig. 26: Changes in measured kinetics when fitting with limited exponential terms .................... 95

Fig. 27: Component charge dependence on the holding and test potential ................................... 96

Fig. 28: Additional wt and T156C transient kinetic data sets........................................................ 97

Fig. 29: Arranging decay charge profiles using the masking effect .............................................. 98

Fig. 30: Assigning transient decays to conformational changes of the carrier predicted by the

crystal model .................................................................................................................................. 99

Fig. 31: Revised SGLT1 transport model ...................................................................................... 99

Fig. 32: Rough state-model of SGLT1 transport ......................................................................... 100

Fig. 33: Types of time constant voltage dependent behavior ...................................................... 100

Fig. 34: Example I–V data ........................................................................................................... 126

Fig. 35: General -state cyclical model ....................................................................................... 127

Fig. 36: Example showing the form of the steady-state equation ............................................... 128

Fig. 37: Example 1. Solution of the voltage dependent general velocity equation ..................... 129

Fig. 38: Geometric properties of a sigmoid function ................................................................... 129

Fig. 39: Effect of a dominant denominator term ......................................................................... 130

xiii

Fig. 40: Geometric properties of a sigmoid function with two terms .......................................... 131

Fig. 41: Example 2. Voltage dependent properties of the general velocity equations ................ 132

Fig. 42: Characteristics of the logarithmic exponential shifts ..................................................... 133

Fig. 43: Example 3. Substrate dependence of the I–V curves ..................................................... 134

Fig. 44: Analyzing experimental I–V data .................................................................................. 135

Fig. 45: Steady-State velocity models ......................................................................................... 136

Fig. 46: Voltage dependence of the Na+ and αMG apparent affinities for the Q170C and Q170E

mutants of rSGLT1 ...................................................................................................................... 137

xiv

List of Abbreviations

AdiC arginine/agmatine antiporter

AcrB multidrug efflux transporter

APC amino acid-polyamine-organocation family

ApcT H+/amino-acid symporter

BCCT betaine/carnitine/choline transporter family

BetP Na+/betaine transporter

CHT Na+/choline transporter

EmrD multidrug efflux pump

EmrE multidrug efflux pump

GlpT glycerol-3-phosphate/Pi exchanger

K157C lysine to cysteine mutant of SGLT1

LacY lactose permease

LeuT Na+/leucine transporter

MFS major facilitator superfamily

Mhp1 Na+/benzyl-hydantoin transporter

NCS1 nucleobase cation symporter-1 family

N Na+

NIS Na+/iodide symporter

NSS neurotransmitter:sodium symporter family

OxlT oxalate/formate exchanger

PutP Na+/proline transporter

Pz phloridzin

RND resistance nodulation-cell division superfamily

S sugar, glucose, phloridzin

SGLT(1–5) Na+/glucose transporter

SMCT(1–2) Na+/monocarboxylate transporter

SMIT(1–2) Na+/myo-inositol transporter

SMVT Na+/multivitamin transporter

SSS(F) solute:sodium symporter family

T156C threonine to cysteine mutant of SGLT1

TC transport classification system

TCA tricyclic antidepressant

TM transmembrane segment

vSGLT Vibrio parahaemolyticus Na+/galactose transporter

xv

List of Variables

amplitude

( )⁄

capacitance

elementary charge

Faraday constant

current

, sum of voltage and substrate independent snake terms

dissociation constant for transition

, coefficient of substrate dependent snake term

,

coefficient of voltage dependent snake term

,

coefficient of voltage and substrate dependent snake term

Michaelis-Menten apparent affinity

, voltage independent factor of a voltage dependent rate constant

, rate constants for transition

number of expressed carrier proteins

Hill coefficient

charge,

net charge translocated per cycle

gas constant

net steady-state velocity,

, steady-state velocities

, maximum steady-state velocity

substrate, substrate concentration

absolute temperature

time constant

reduced voltage, ( )⁄

, exponential shift of voltage dependent snake term

,

exponential shift of voltage and substrate dependent snake term

, voltage

, voltage at half maximum amplitude, ( ( ))⁄

holding potential

Michaelis-Menten maximum velocity

resting potential

at saturating substrate concentrations for I–V curves

test potential

valence

, combined valence of voltage dependent snake term

,

combined valence of voltage and substrate dependent snake term

1

Introduction 1

Secondary active transport was first proposed by Robert Crane in 1960 to explain the uphill

transport of glucose and other molecules in the intestinal brush border. He hypothesized that the

translocation of substrate was coupled to Na+, which would energize this process through the

Na+ gradient maintained by the Na

+/K

+ pump

1. This mechanism was clarified a few years later

in 1966 when Jardetzky outlined an hypothesis whereby transport could occur through

alternating access to a central substrate binding site2. Throughout the 60’s and 70’s research was

focused on determining substrate and ion specificity of sugar and amino acid absorption using in

vitro preparations of various intact tissues and radioactive tracers3. Although much work at this

time focused on kinetic studies, and theoretical derivations, the available techniques posed

significant difficulties (unstirred layers, cell metabolism of the substrate, multiple endogenous

transport pathways, and an uncontrolled membrane potential) that limited the precision of the

measurements and depth of analysis.

A significant advancement was the use of isolated membrane vesicles, initially prepared from E.

coli (1966)4, and later adapted to the brush border (1973)

5. This new technique solved a number

of concerns and allowed for much more control over the experimental conditions, including

limited manipulation of the membrane potential6,7

. For the most part, these studies involved

steady-state and equilibrium measurements with radiotracers that led to more precise

characterization of affinities and coupling ratios. However, the presence of heterologous

transport pathways remained a major drawback that complicated kinetic analysis8.

Expression cloning, originally demonstrated with rabbit SGLT1 in 19879, marked a new era in

the study of membrane transport10,11. For the first time, carrier proteins could be cloned and

studied in a heterologous system. Not only were the majority of earlier complications resolved,

but a new powerful and precise tool became available for controlling and monitoring these

proteins. The two-electrode voltage-clamp and the oocyte system brought electrophysiological

techniques to secondary active transporters, whose expression levels and currents are too small

to observe with the patch clamp in native tissues (turnover of ~10 s−1

for carriers12

versus ~200

s−1

for pumps and ~106–10

8 s

−1 for channels

13,14). Soon apparent affinities, turnover rates, and

coupling ratios (and their voltage dependencies) were being measured in greater detail than ever

2

before. In addition, rapid transient studies became possible, allowing for measurement of

expression level, valences, ’s, decay time constants and amplitudes.

Once cloning was available, mutation studies became a popular means of studying the

relationship between structure and function, and often these mutants were characterized in the

oocyte system with electrophysiological techniques. Our group used cysteine scanning

mutagenesis to identify a region of SGLT1 critical for substrate binding (K157, T156)15-17

,

while others found a key residue in translocation (Q457)18

, and a structurally significant

disulfide bridge (C255–C511)19

.

Progress in elucidating mechanism, however, was tempered by the sheer difficulty of inferring

potentially complex interactions without a tertiary structure. Much of these kinetic studies

supported the alternating access hypothesis proposed by Jardetzky, but none were able to verify

it. However, as discussed in §1.3 Crystal Structures of the LeuT Fold, this began to change in

2008 when the crystal structure of a bacterial Na+/galactose transporter (vSGLT) was solved, a

homolog of SGLT1. Quite surprisingly, the vSGLT architecture was found to be the same as the

leucine transporter (LeuT), which shares no sequence similarity and has two fewer

transmembrane domains. In the years that followed, this structural superfamily has grown to

contain members from five genetically distinct transporter families. Furthermore, as these

structures have appeared they have been found in various conformations that sketch out a series

of potential conformational changes involved in transport. These conformational changes have

finally demonstrated the existence of the alternating access mechanism, and in the case of the

LeuT architecture it has become known more specifically as a gated rocker-switch. However,

even with this invaluable structural detail a real limitation is the static nature of these crystals,

and it seems that kinetic studies may be the crucial piece needed to understand the moments in

between, and confirm the gated rocker-switch kinetically.

The goal of this project has been to find ways to extract as much kinetic information as possible

from electrophysiological kinetic studies of cotransporters, to help interpret the crystallographic

data and understand the transport mechanism. Although significant work has been done already

with transient and steady-state studies, the results are often kinetic parameters that help

characterize transporters ( , , , , ), yet fall short of revealing a definitive model. The

models that have been built are often over parameterized, because of insufficient kinetic features

3

and the inherent difficulty of the problem. For decades kinetic studies have progressed in the

dark, but now with the availability of crystal structures it is important to reinvest in this basic

research to take advantage of new synergies that have become available.

The experiments performed in this study are based around the two-electrode voltage-clamp

technique and the heterologous Xenopus laevis (African clawed frog) oocyte expression system.

Protein is overexpressed in the oocytes, and then manipulated and monitored via the membrane

potential by the two electrodes. This research studies two time domains of cotransporter

membrane currents, transient and steady-state, with SGLT1 as a model system; SGLT1’s

properties are discussed in more detail in §1.2.1.1 Sugar: SGLT1–4. Each domain provides

information on different aspects of transport, and both are needed for a complete picture. The

transient currents contain detailed information on individual conformational changes of the

empty carrier, while the steady-state currents measure lumped parameters from the transport

loop. As we will show in §2 Dissecting the Transient Current of SGLT1, the transient currents

of SGLT1 can be decomposed into four carrier decays that reveal the gated-rocker switch

mechanism. Measuring and fitting transient currents is not an easy task, and so we present new

methods for analyzing them in as much detail as possible. What we find is that when all the

decay components are identified it is much easier to build a kinetic model. The steady-state

project presented in §3 A Practical Method for Characterizing the Voltage and Substrate

Dependence of Membrane Transporter Steady-State Currents, uses the concept of a general

cyclical system to model the voltage and substrate dependence of the steady-state velocity.

Recursive patterns in the velocity equation allow for a general understanding of its geometric

features. This reveals that the I–V curves can be modeled phenomenological with Boltzmann

functions whose parameters report on the kinetics of rate limiting segments of the transport

loop. These rate limiting segments hide other parts of the loop and are the most we can hope to

extract from steady-state studies. Only by combining the strengths of transient and steady-state

studies with crystallographic research can we complete the picture of how transporters work.

Although most of this research is centered around SGLT1, this carrier is an excellent model

system for ion-coupled cotransporters and these results can be extended to many others.

4

1.1 Types of Transport

The flow of molecules across biological membranes is mediated by three major classes of

intrinsic membrane proteins known as channels, pumps, and cotransporters, which comprise

~3% of all protein encoding genes. These classes of carrier proteins are differentiated based on

the size of the molecules they transport, the energy source that drives them, and their basic

architecture and mechanism. On the surface of each cell a diverse ecosystem of hundreds of

these proteins work together to maintain homeostasis, although the size and composition of this

ecosystem can vary considerably between organisms to suit their different needs as shown

below20

:

transport genes channels pumps cotransporters

Human 754 43% 11% 44%

E coli 354 4% 20% 66%

Asian rice 1200 15% 21% 63%

1.1.1 Facilitated Diffusion

Facilitated diffusion involves the transport of ions and small molecules down an electrochemical

gradient. These proteins are not directly coupled to an energy source, and instead rely on other

membrane proteins to maintain the electrochemical gradient that allows them to function.

Because of this they are unable to concentrate their substrate.

Ion channels mediate the rapid and selective movement of ions (~106–10

8 s

−1) such as K

+, Na

+,

Ca2+

, and Cl−, and are fundamental to biological processes that require speed, such as neuronal

signaling and muscle contraction. The channels contain a pore, which the ions flow through, and

a gating mechanism, which controls access to the pore and is regulated by the membrane

potential or another substrate. The first channel crystal structure was solved in 1998 for the K+

channel KcsA (homotetramer, 108 amino acids, 2 transmembrane segments—TMs), which won

McKinnon a shared 2003 Nobel Prize in Chemistry14

with Agre for his discovery of the water

channel aquaporin (AQP1) in 199221

, which was crystalized in 200022

(monomer, 269 amino

acids, 6 TMs).

5

Larger molecules like many dietary sugars are rapidly equilibrated (~1200 s−1

23

) by the GLUT

family of uniporters (monomer, ~500 amino acids, 12 TMs). Although the GLUTs have not

been crystalized, several of their distant relatives including LacY have24

, and belong to the same

major facilitator superfamily (MFS). They are thought to share a similar rocker-switch

mechanism, whereby the substrate binding site, located at the center of the protein, is alternately

exposed to either side of the membrane as two symmetrical halves of the protein rock back and

forth23

.

1.1.2 Primary Active Transport

These carriers, also known as pumps, use an energy source, usually ATP or light, to fuel uphill

transport, and are responsible for maintaining the electrochemical ion gradients that drive

channels and cotransporters, including repolarization after an action potential. The Na+/K

+ pump

is the most well-known member because it energizes most animal cells, while the H+ pump fills

a similar role in plants and fungi. The Ca2+

pump SERCA1a from skeletal muscle was the first

high resolution pump crystal, published in 200025

, with the Na+/K

+ 26

and H+ pumps

27 following

in 2007. They are among the most complex transport proteins (monomer, ~1000 amino acids).

The Na+/K

+ pump, for example, has a two part transmembrane module (10 TMs), three

cytoplasmic domains (A, N, and P), and interacts with two extracellular subunits (β and γ). It

functions by switching between two primary conformational states, which have different

affinities for Na+ and K

+, and alternate exposure of the binding sites on either side of the

membrane28

.

The microbial rhodopsin family of transporters, which includes bacteriorhodopsin, use light to

pump H+, and are much smaller than their ATP counterparts (monomer, 250–350 amino acids,

~7 TMs). Bacteriorhodopsin was first crystalized in 199629

.

1.1.3 Secondary Active Transport

Secondary active transporters, also known as cotransporters, are able to concentrate substrate

like pumps, but use the electrochemical ion gradients (often H+ or Na

+) maintained by pumps as

fuel instead of ATP. They transport a wide variety of substrates and are the predominant mode

of entry for most medium and large sized molecules.

6

There are many different families of secondary active transporters, but some prominent

members include the major facilitator superfamily (MFS), the neurotransmitter:sodium

symporter (NSS) family, and the solute:sodium symporter (SSS) family. The first secondary

active transporter to be crystalized, and the last of the three major classes of transporters, was

the bacterial MFS oxalate:formate exchanger OxlT in 2002 (monomer, 418 amino acids, 12

TMs)30

. Since then three other MFS members have been crystalized, all transporting very

different substrates but with the same rocker-switch architecture; the H+/lactose symporter

LacY24

, the glycerol-3-phosphate/Pi exchanger GlpT31

, and the H+/multidrug exchanger EmrD

32.

This has turned out to be a common theme for secondary active transporters, where the bacterial

NSS 2Na+/Cl

−/leucine symporter LeuT (monomer, 513 amino acids, 12 TMs) architecture is

also shared by a diverse group of carriers belonging to five phylogenetically distinct families

that includes the SSS family. The architecture of this structural superfamily is similar to the

MFS transporters, but has additional gates on either side of the membrane to control access to

the intra and extracellular pores, and is referred to as a gated rocker-switch. SGLT1 belongs to

this structural superfamily which will be discussed in more detail in the next section (§1.2).

1.2 Familial Relations of SGLT1

SGLT1 can be considered a member of three families. It is the most well-known of the 12

member solute carrier 5 (SLC5) family of membrane transport proteins, and was the first to be

cloned (SLC5A1)33

. All together there are 51 SLC families formed from the pool of human

genesa. These groupings are based on a minimum of 20–25% sequence similarity with other

members34

. The Transport Classification (TC)b is a more general system, with transporters of all

species grouped according to function and phylogeny35

, and is modeled after the classification

system for enzymes adopted by the Enzyme Commission (EC). Within the TC system SGLT1

belongs to the solute:sodium symporter (SSS) family (TC 2.A.21); sometimes referred to as the

sodium/substrate symporter family (SSF36

or SSSF37

). In more recent years a structural

superfamily has emerged from the crystal structure literature, where SGLT1 and several other

a Curated by the HUGO Gene Nomenclature Committee (HGNC), an extension of the Human Genome

Organization (HUGO).

b Curated by the Saier lab and adopted by the International Union of Biochemistry and Molecular Biology

(IUBMB).

7

phylogenetically distinct members have been found to share a common architecture based on the

LeuT fold (named after its first member, the Na+/leucine transporter). With SGLT1 belonging to

such a diverse group of genetically or structurally similar transport proteins, it is likely that the

methods and conclusions presented in this thesis can be extended to these other members as

well.

1.2.1 SLC5 Family

1.2.1.1 Sugar: SGLT1–4

Properties of the SLC5 proteins are outlined in Table 1. Four members, SGLT1–4, are primarily

aldohexose transporters. This group includes two of the more prominent members of the SLC5

family, SGLT1 and SGLT2. SGLT1 (SLC5A1) is found mainly in the small intestine, but also

in the heart and kidney and to a lesser extent in some other tissues, and is responsible for the

vast majority of dietary glucose and galactose absorbed across the brush border membrane, from

the lumen into enterocytes and ultimately the bloodstream38,39

. SGLT2 (SLC5A2) is found in

many tissues but is highest in concentration in the kidney proximal tubule, where it reabsorbs

glucose from the glomerular filtrate preventing its loss in the urine38,40,41

. Of these two

transporters, SGLT1 has a high affinity for glucose, 0.5 mM, and couples two Na+ ions to the

transport process42. While SGLT2 is a low affinity transporter, 1.6 mM, that couples a single

Na+ ion.

Defective mutations in SGL1 lead to glucose-galactose malabsorption, a rare autosomal

recessive disease. The disease appears in newborn infants as diarrhea and dehydration, which is

ultimately fatal unless all glucose, galactose, and lactose are removed from the diet. As of 2003

there were ~300 cases worldwide with 56 identified mutations from 82 patients43,44

.

Malfunctioning SGLT2 results in familial renal glucosuria (FRG), an inability to reabsorb

glucose from the glomerular filtrate. The glucose is instead released in the urine, at a rate of a

few to more than 100 g per day45

. FRG is asymptomatic, with patients exhibiting no major

negative outcomes. As of 2008 132 patients with FRG had been studied, finding 44 unique

mutations in 52 families45-52

. Because of the mildness of the disease, the majority of cases

probably go unreported. In a study of all South Korean children, who are given mandatory

annual urinalysis, 0.07% test positive for glucosuria53

. The favorable outcome of FRG has made

8

SGLT2 an attractive target for selective inhibition, with potential medical applications

controlling plasma glucose levels in diabetics and weight loss. Research in this area has

increased rapidly in the last decade with a number of promising drugs identified54,55

, and clinical

trials underway56-59

.

When pig SGLT3 (pSGLT3) was originally cloned from LLC-PK1 cells and expressed in COS7

cells it was thought to be a sodium dependent amino acid transporter (SAAT1)60

. However, this

changed when pSGLT3 was shown to be a sodium glucose cotransporter in the oocyte

expression system61-65

. Almost a decade later the human isoform was characterized and quite

surprisingly found unable to transport glucose, yet glucose would enhance Na+ and H

+

dependent membrane currents, classifying it as glucose gated Na+ transporter

66. hSGLT3,

therefore, appears to be a glucose sensor, a theory supported by the protein's localization in the

cholinergic neurons of the small intestine and skeletal muscle66

. More recently it has been

shown that a single residue is responsible for the majority of differences between hSGLT1 and

hSGLT3. Q457E-hSGLT1 is a glucose gated Na+ transporter, and E457Q-hSGLT3 a sodium

glucose cotransporter67

. The role 457 plays in sugar transport and coupling is understandable

considering its involvement in the sugar-binding site of vSGLT68

.

Very few studies have been done on SGLT4. It has a preference for mannose, yet still transports

glucose, and galactose, and is thought to account for dietary absorption in the intestine and

reclamation from the glomerular filtrate in the kidney69

.

1.2.1.1.1 Classical SGLT1 Functional Models

The SGLT1 transport cycle is often represented by the six-state model in Fig. 1A70,71. There are

two main conformations of the carrier, one where the binding sites for Na+ and glucose face the

extracellular solution (1–3, outside facing), and another where they face intracellularly (4–6,

inside facing). Traversing the cycle counterclockwise beginning with the outside facing empty

carrier (1), two sodium ions bind first with a high cooperativity (2) followed by glucose (3), the

transporter reorients to the inside (4) where glucose (5) and then Na+ (6) are released before the

empty carrier (6) returns to an outside facing conformation (1).

Although it is has been established that Na+ must be present to allow glucose to bind, based on

phloridzin binding studies,72 there has been no definitive proof on whether one or both Na+ ions

9

bind before glucose. There have been a few studies which have attempted to distinguish

between the Na+/glucose/Na

+ and Na

+/Na

+/glucose binding orders but they have been

inconclusive, with some suggesting the former73-75 and others the latter76. Regardless, for some

time now the Na+/Na

+/glucose binding order has been the prevalent scheme12,71,77.

In the absence of glucose there is a Na+-dependent phloridzin-sensitive steady-state current

observable in oocytes, which was originally presumed to be caused by a Na+-leak pathway

between states 2 and 5, but has more recently been shown to be the result of a cationic leak

pathway (Na+, Li

+, Cs

+, K

+) that is independent of the carrier translocation steps

78. This leak

pathway is much weaker than Na+ coupled glucose transport, generating a current that is only

2.5% of the αMG inducible current42. Transient currents are only observed in the absence of

glucose (2↔1↔6, yellow shading), as glucose promotes an inside facing conformation where

the carrier is electrically silent. Internal binding of Na+ (6→5) is unfavorable under normal

conditions because of the low intracellular Na+ concentration in oocytes (8.5±0.5 mM79) and a

low intracellular Na+ affinity measured with the reverse transport mode (54.3±7.8 mM 80, 12±4

mM81, 6–50 mM82). A weak pathway between 6↔5 is often (although not always19) used to

rationalize a low occupancy of state 5 in the absence of glucose, limiting these transient models

to sates 1, 2 and 670,83,84. In the presence of glucose substrate is translocated across the

membrane and released intracellularly through the pathway 3↔4↔5 (blue shading).

The demonstration by Chen et al. of two transient decays in the absence of Na+, and our

supporting finding of three in the presence of Na+, necessitates the existence of an intermediate

state of the empty carrier (Fig. 1B, state 7)85,86; there must be at least one transition per decay.

This same reasoning was used to extend the model again when Loo et al. observed three decays

in the absence of Na+ (Fig. 1B, state 8)84.

1.2.1.2 Myo-inositol: SMIT1 and SMIT2

The SLC5 family contains two myo-inositol transporters, SMIT1 (SLC5A3)87

and SMIT2

(SLC5A11)88,89

. Inositol is an important compound in a number of physiological processes, as

an osmolyte90

and a precursor of the inositol phosphate group of secondary messengers91

.

Supplements of inositol have been shown to be effective in treating depression92

, panic

disorder93

, obsessive-compulsive disorder94

, and insulin resistance due to polycystic ovary

syndrome95

.

10

Both transporters have a similar expression profile that includes brain, neuron, heart, kidney,

skeletal muscle, placenta, pancreas and lung—while SMIT2 is additionally found in liver and

intestine 89,96-99

. SMIT1 knockout mice die soon after birth unless given myo-inositol

supplements, from what appears to be undeveloped peripheral nerves leading to the loss of

brainstem control and respiration100,101

. Furthermore, SMIT1 expression is up regulated in cases

of Down syndrome102

and bipolar disorder103

, and down regulated by lithium and other bipolar

drugs104

.

The Na+:substrate stoichiometry for SMIT1

105,106 and SMIT2

105,106 is 2:1. Both carriers share

similar substrates, including myo-inositol, glucose and xylose, but differ in their affinities for

particular isomers. SMIT2 transports D-chiro-inositol, SMIT1 does not; SMIT1 transports both

the L and D isomers of glucose and xylose, SMIT2 only transports the D isomers. SMIT1 is a

high affinity myo-inositol transporter, 50 μM107

, while SMIT2 is low affinity, 120 μM105

.

1.2.1.3 Monocarboxylates: SMCT1 and SMCT2

Two members of the SLC5 family, SMCT1 (SLC5A8) and SMCT2 (SLC5A12), transport short

chain monocarboxylates such as butyrate and lactate108

. Much like the relationship between

SGLT1 and SGLT2, one is a high affinity transporter of lactate (0.16 mM) with a Na+:substrate

stoichiometry of 2:1 (SMCT1)109-111

, while the other is a low affinity transporter (~40 mM) with

a 1:1 stoichiometry (SMCT2)112-114

. The substrates of these transporters carry a −1 charge,

making SMCT2 the only electroneutral member of the SLC5 family, and the only one that

cannot be studied with the two-electrode voltage-clamp (studies of SGLT2, SGLT5 and CHT

are also difficult because of low expression levels in cultured cells and oocytes115

). The rate of

SMCT1 transport was originally thought to be stimulated by a factor of two in the presence of

saturating Cl−, without Cl

− itself being transported

110. However, a more recent study has

discovered that Cl− does not in fact stimulate transporter current, and that the earlier

interpretation was caused by inhibition of the carrier by the Cl− replacement anion cyclamate

116.

The SLC5A8 gene was originally found in the apical membrane of thyroid cells and thought to

passively transport iodine, leading it to be called the apical iodine transporter (AIT)117

.

However, subsequent groups were unable to observe I− transport, and instead found

monocarboxylates to be better substrates and changed the name to SMCT1109,110

. Both

transporters are expressed in the kidney, intestine, brain, and retina, while SMCT1 is

11

additionally found in the thyroid, colon, and salivary glands, and SMCT2 in skeletal muscle108

.

In the kidney SMCT1 is thought to be the high affinity lactate transporter118

, with silencing of

this gene causing lactaturia119

.

Interest in SMCT1 increased recently when it was identified as a silenced gene in most colon

cancers120,121

, and two of its substrates, butyrate and pyruvate, were found to induce apoptosis in

cancerous cells108,120,122,123

. Butyrate is produced at high concentrations in the colon by the

bacterial fermentation of dietary fiber, and it seems silencing of SMCT1 by these cancer cells is

necessary for their survival.

1.2.1.4 Choline: CHT

One of the more unique members of the SLC5 family is the sodium choline transporter, CHT

(SLC5A7)124,125

. CHT is the only SLC5 member to transport a cationic substrate and is located

exclusively in the presynaptic terminals of cholinergic neurons126

, where it mediates the uptake

of choline for the intracellular production of acetylcholine by choline acetyltransferase

(ChAT)127

. The Na+:substrate stoichiometry increases from 2:1 to 9:1 as the membrane potential

is hyperpolarized. Cl− is required to initiate transport but is itself not transported and can be

partially replaced by Br− 124,128

.

CHT has the largest Na+-leak of the SLC5 family at ~40% of the transport current, which may

explain the large Na+ stoichiometry at hyperpolarizing potentials. A remarkably low level of

surface expression complicates the study of CHT in heterologous expression systems. Choline

uptake in CHT expressing oocytes is only 3–4 times greater than background (compared with

1000 times for SGLT19) with maximal choline induced currents of ~13 nA. However, resealed

membrane vesicles made from transiently transfected COS7 cells show significantly more

activity, implying a large intracellular pool129

. A L530A/V531A double mutant appears to

increase trafficking to the membrane by ~2.5 times and permits some electrophysiological

measurements in oocytes124

.

1.2.1.5 Iodide: NIS

The sodium iodide symporter, NIS (SLC5A5)130

, is the primary transporter of iodine into

thyroid cells, but is also found in several other tissues including salivary gland, stomach and

lactating mammary gland131

. Because of this, it plays an important role in delivering radioactive

12

iodine for the treatment of thyroid cancer, and has been proposed as a treatment for some breast

cancers that express NIS132

. NIS has been well characterized in the Xenopus oocyte expression

system where it has been shown to transport Na+:I

− with a 2:1 ratio

133. A single amino acid

mutation (T354P) has been identified as the cause of hypothyroidism in several patients134

.

1.2.1.6 Multivitamin: SMVT

The sodium multivitamin transporter, SMVT (SLC5A6)135

, is expressed most in placenta136

, is

the primary multivitamin transporter in the intestine and liver137

, and can also be found in the

pancreas, kidney and heart. It is known to transport pantothenate, lipoate and biotin with a

stoichiometry of two Na+ per substrate, but all together there are very few kinetic studies of the

transporter135,136

.

1.2.1.7 Orphan: SGLT5

Only a handful of studies have been done on SGLT5 (SLC5A10), which is an orphan transporter

found mostly in rabbit138

and bovine kidney139

, and chicken intestine140

.

1.2.2 The Solute:Sodium Symporter Family

Several non-human transporters that share homology with the SLC5 family are members of the

SSS family, TC 2.A.21. A relatively small number have been characterized, and they are listed

in Table 1. vSGLT from Vibrio parahaemolyticus is the most significant member, because of its

recently solved crystal structure68

, and is a close cousin to SGLT1 transporting Na+ and

galactose with a 1:1 stoichiometry141,142

. The Na+ proline transporter PutP, also with a 1:1:

stoichiometry, is essential for the survivability of the infectious bacteria Staphylococcus

aureus36,37,143

, making it a promising drug target144,145

. As well, a homology mapping of the

PutP sequence with that of the crystalized vSGLT structure has confirmed that they share

similar structures146

. There are only a few studies on the sodium pantothenate transporter

PanF147,148

.

1.2.3 Phylogenetic Homology

An unrooted phylogenic tree of the 12 SLC5 members, 10 SGLT1 species homologs, and three

prokaryotic members of the SSS family is shown in Fig. 2A. All of the SLC5 members that bind

neutral substrates (glucose, inositol, and mannose) are clustered on one branch of the tree

13

(bottom), while anionic binding carriers (monocarboxylates, multivitamin, and iodide) form

another (top left). The only cationic transporter (choline) and the three prokaryotic transporters

are relatively distinct from each other and the other carriers (top). Therefore, with the human

genes the transporters are clearly grouped according to the charge carried by the substrate.

The sequence similarities and identities are shown in Fig. 2B. Amongst the sugar transporters,

SGLT3, a glucose sensor, is most similar to SGLT1, followed by the low-affinity glucose

transporter SGLT2, the mannose transporter SGLT4 and the orphan transporter SGLT5, and the

two myo-inositol transporters SMIT1 and SMIT2. All of the SGLT1 species homologs but

Atlantic salmon are more related to human SGLT1 than any of the other SLC5 members.

1.2.4 Structural Homology

Crystal structures of bacterial secondary active membrane transporters began appearing in 2002

with the oxalate formate exchanger OxlT30

of the major facilitator superfamily (MFS), and the

proton driven multidrug efflux transporter AcrB149

of the resistance nodulation-cell division

(RND) superfamily. Since then, 16 different proteins have been solved and six architectures

have been identified as shown in Fig. 3. A common feature of most of these architectures is a

two-fold axis of symmetry between N and C-terminal 4–6 helix bundles. Evolutionary clues

about the origin of this architectural symmetry have come from the multidrug transporter EmrE

of E. coli, a four-transmembrane homodimer that has also been crystalize150

. It suggests an

evolutionary path where a single gene encoding a dual topology homodimer like EmrE may

undergo a gene duplication event and become an inverse-topology heterodimer, these two genes

can later fuse to encode a monomer with inverted repeats151,152

.

The LeuT fold (the first protein to be crystalized from this family) is the most genetically

diverse, having been found in 7 transporters (LeuT153

, vSGLT68

, Mhp1154

, BetP155

, AdiC156,157

,

ApcT158

, CaiT159

) belonging to 5 phylogenetically distinct families (neurotransmitter:sodium

symporter NSS; solute:sodium symporter SSS; nucleobase:cation symporter-1 NCS1;

betaine/carnitine/choline transporter BCCT; amino acid-polyamine-organocation APC). Each of

these transporter families forms a separate branch in Fig. 3, highlighting their genetic diversity.

Their basic structure is a 5 helix repeat, which is discussed in more detail in §1.3 Crystal

Structures of the LeuT Fold.

14

Other secondary active transporter architectures include the 6 helix repeat MFS transporters

(OxlT30

, eLacY24

, GlpT31

, EmrD32

), the 6 helix repeat RND transporters (AcrB149

, MexB160

), the

unique and intricate 6 helix repeat of the NhaA transporter161

, the unique 8 helix structure with

minor symmetry of the homotrimer dicarboxylate/amino acid:cation symporter (DAACS)

GltPh162

, and the 4 helix homodimer small multidrug resistance (SMR) transporter EmrE150

.

The existence of shared architectures that transcend substantial differences in primary structure

reveals familial relationships at the tertiary level. This is an important finding, because it

suggests the likelihood of shared mechanisms adopted by a wide range of proteins, and as the

field becomes more interconnected the applications of each new discovery are amplified.

15

1.2.5 Figures

Table 1: Properties of SLC5 and select TC 2.A.21 members115

.

16

Fig. 1: Classical SGLT1 transport model. (A) The transport of Na+ and glucose follows an

ordered process; 1 outside facing empty carrier, 1→2 two Na+ bind simultaneously with a high

cooperativity, 2→3 glucose binds, 3→4 fully loaded carrier orients to the inside, 4→5 glucose

unbinds, 5→6 both Na+ unbind simultaneously, 6→1 inside facing empty carrier orients back to

the outside. In the absence of glucose the transient model is restricted to 2↔1↔6 (yellow

shading). The full transport pathway involves 3↔4↔5 (blue). The transporter cycles in an

anticlockwise direction at hyperpolarizing potentials and in a clockwise direction at

depolarizing. (B) Extensions to the classical model, spurred by the discovery of additional

transient decays. An intermediate state of the empty carrier (7) was first discovered by Chen et

al.85 and further supported by Krofchick and Silverman86, while a second intermediate state (8)

was established later by Loo et al.84.

17

Fig. 2: Homology of the SLC5 and SSS families. (A) Phylogenetic tree including SGLT1

species homologues (grey). (B) Sequence similarity and identity shared between SLC5 and SSS

members. Species abbreviation: human, pig, rabbit, rat, Atlantic salmon, mouse, bovine, dog,

sheep, Eurasian common shrew, horse, Escherichia coli, Vibrio parahaemolyticus.

18

Fig. 3: Phylogenic tree of secondary active transport families with solved crystal structures

(stars). See Fig. 2 for species abbreviations.

hSGLT

1hS

GLT3hSG

LT2

hSGLT4hSGLT5

hSMIT2

vSGLT

hSMIT1

hNIS

hSMCT1

hSMCT2

hSMVT

ePutP

ePanFhC

HT

AdiC

Cad

B

PotE

Cad

C

ApcT

SteT

xC

T

CAT

1

NK

CC

2N

CC

GltP

h

LeuT

Aa

hGlyT

1b

hGAT1

hDAT

hSERT

Mhp1HyuP

PUCI

DAL4

FUI1

BetPO

puDButA

BetTEctPCaiTN

haA

MexB

AcrB

eLac

Y

hG

LU

T1

HM

IT

Glp

TEmrD

Oxl

T

0.027

SSS/

SLC

5

APCN

SS/SLC6

NC

S1

BCCT

MFS/SLC2

DAACS

NhaARND

★★★

EmrE

SMR ★

★ crystal

19

1.3 Crystal Structures of the LeuT Fold

The LeuT fold is of particular interest to this thesis because it is shared by a close cousin of

SGLT1, the bacterial Na+/galactose symporter vSGLT of Vibrio parahaemolyticus (Fig. 2B),

and mutation studies have confirmed that vSGLT and SGLT1 have similar structures17,68

.

Understanding the features and conformations of this architecture is essential for developing a

working model that, with the help of kinetic data, may one day explain the mechanisms

involved.

1.3.1 Timeline

The structure of the 2Na+/Cl

−/leucine symporter LeuT of the NSS family was solved in 2005

153.

The core was found to be made out of two symmetrically intertwined bundles, TMs 1–5 and 6–

10, with 11 and 12 trailing on the periphery (see Table 2 for transporter properties and Fig. 4 for

timeline). Over the next few years it was crystalized with a range of amino acid substrates, a

competitive inhibitor163

, and several noncompetitive inhibitors164,165

.

In 2008 another transporter crystal appeared with the same structure, vSGLT68

, but from a

different family (SSS). The 5+5 core was the same, but this time there was one additional TM at

the front and three at the end for a total of 14. Two months later the Na+/benzyl-hydantoin

transporter Mhp1, of the nucleobase:cation symporter (NCS1) family, was solved with the same

TM arrangement as LeuT154

, followed in 2009 by the 2Na+/betaine transporter BetP, of the

betaine/carnitine/choline transporter (BCCT) family, with two extra TMs at the front and one at

the end155

.

At this point it was becoming obvious that transporters with no relation in amino acid sequence

(these were phylogenetically distinct families) or number of TMs (12–14 so far) could share the

same basic architecture, and that other methods of classification might be helpful. With this in

mind, a hydropathy profile alignment of LeuT and vSGLT was used to identify a similar fold for

the amino-acid-polyamine-organocation (APC) superfamily166

, and within a year this hypothesis

was confirmed for two members; the arginine:agmatine antiporter AdiC156,157

and the H+/amino-

acid symporter ApcT158,167

.

20

1.3.2 Architecture

A distinctive feature of the LeuT fold is an internal symmetry relating the first five and last five

TMs of a ten-helix bundle. These 5 TM halves are interwoven to form the carrier, and,

depending on the variant, can be flanked by an additional 2–4 TMs of questionable significance

(Table 2). To keep the notation simple we will adopt the numbering of Abramson and Wright

for the rest of this discussion168

: extra N-terminal helices ( −2, −1), core (1–10), extra C-

terminal helices (11, 12, 13).

As Fig. 5 shows, TMs 1–5 and 6–10 are twisted together to form structural and functional pairs

(1/6, 2/7, 3/8, 4/9 and 5/10) that are related by a two-fold axis of symmetry. Two central pairs,

1/6 and 3/8, make up the core of the transporter by defining the majority of the pore surface, and

substrate and cation binding sites. A unique and important feature are unwound regions at the

midpoint of TMs 1 and 6 that participate in substrate and ion binding, and conformational

changes. Typically one end of the substrate binds at the unwound regions, and the other extends

towards TMs 3/8 (Fig. 5, orange triangle). In some cases (BetP and vSGLT) TMs 1/6 are

wedged apart by the substrate and 2/7 become involved in binding, and in others (AdiC and

vSGLT) 10 bends over the pore to interact as well. Ion binding occurs mostly between TMs 1

and 8, but there is another site between 1, 6 and 7 in LeuT.

The outer shell is made up of TMs 2/7 on one side and two opposing V’s formed by 4-5 and 9-

10 on the other. These V’s pinch the long diagonal 3/8 pair on both ends, while 5/10 support 1/6

from the side, and 2/7 buttress them from behind.

1.3.3 Solved Conformations

The structures captured so far have been found in a variety of conformations, and with various

substrates (Fig. 4). Comparing them provides clues about the mobility of TMs, possible

conformations, and conformational changes, but it is important to keep in mind that there may

be variations in the mechanisms between carriers. For this reason, different conformations of the

same carrier are more valuable, with the ultimate prize remaining one carrier visualized in all

conformations of the transport cycle.

All of the LeuT structures have been found in an outside facing conformation, characterized by

an extracellular pore leading to the binding site, at the unwound regions of TMs 1 and 6, and

21

covered by a thin gate. The first structure was determined with bound leucine153

, with later

structures showing alanine, glycine, leucine, methionine, 4-F-phenylalanine and tryptophan at

the same site163

. Based on these structures, and kinetic data, tryptophan’s mechanism of

competitive inhibition was explained. Other structures with bound tricyclic antidepressants

(TCA) desipramine, imipramine, and clomipramine uncovered the mechanism of

noncompetitive inhibition164,165

.

Mhp1154

and AdiC156,157,167

were also found in an outside facing conformation similar to LeuT,

and were able to be crystalized with and without substrate. These structures have provided

important clues about the conformational changes that take place after substrate binding, and the

gating mechanism.

The existence of an intracellular pore was confirmed with the vSGLT1 structure, which was

found in an inside facing conformation with bound substrate, blocked from release by a thin

intracellular gate68

. ApcT was also crystallized in an inside facing conformation, but without

substrate and the intracellular pore partly occluded158

. The BetP structure is unique, existing in

an intermediate conformation with bound substrate and partially closed pores on either side155

.

1.3.4 Pores

The substrate binding site is located midway across the membrane, at the unwound regions of

TMs 1 and 6, and, depending on the conformation, is accessed by an intra or extracellular

solvent accessible pore. When one pore is open the other is closed by a thick layer of collapsed

TMs, and, in the process of switching, the carrier appears to pass through an intermediate state

where both pores are closed (BetP, ApcT).

As shown in Fig. 6A with LeuT, but also observed for Mhp1 and AdiC, the extracellular pore is

visible between TMs 1, 3, 6, 8 and 10. TMs 1 and 6 bend in unison at their breaks away from

the opening, and along with 3 and 10 line the exit with 8 filling the rear. The extracellular pore

is closed by the inward bending of the tip of TM 10, and the extracellular halves of 1and 6 (see

B and C). This change may occur in two steps, first with the bending of TM 10 seen with Mhp1,

followed by the tilting of 1 and 6 observed with BetP and vSGLT.

vSGLT shows the intracellular pore (C), which is formed by TMs 1, 2, 3, 6, 8 and 10. Again

TMs 1 and 6 bend away, highlighting the flexibility of their unwound centers to allow rotations

22

on both sides. TMs 1, 6 and 8 form the exit, while 2, 3 and 10 line the rear. Tracing the transport

path through these pores sees the substrate enter between TMs 6 and 10 and exit in an S-shaped

motion through 1 and 5.

An intermediate conformation, with both pores closed, is seen in BetP and ApcT (B). In both

cases the cytoplasmic pore is partly open, although not enough to allow the release of substrate.

The presence of three distinct conformations, open-to-out, occluded, and open-to-in suggests

two large scale conformational changes during transport.

1.3.5 Thin Gates

Some of the structures with bound substrate have a fully open pore (LeuT, vSGLT, and AdiC)

yet the release of substrate is prevented by one or more gates, often consisting of the large

aromatic residues phenylalanine, tryptophan and tyrosine. In all three cases the gate residues lie

directly on top of the substrate, locking it in the binding site (Fig. 7A–C).

In the LeuT structure (A) a tyrosine on TM 3 (Y108) and a phenylalanine on TM 6 (F253) pin

the substrate down, with the gates held in place by a salt bridge (R30/ D404) linking TMs 2 and

10 at the mouth of the pore. A single tyrosine on TM 6 (Y263) stacks with galactose in vSGLT

(B), preventing its release to the cytoplasm, a feature commonly shared by sugar binding

proteins24,169

. Although the extracellular pore is closed for vSGLT, three gates lay directly on

top of the galactose (M73, Y87, and F424) which bonds with the OH group of Y87. Three layers

of gates have been proposed for the AdiC structure (C). A middle gate (W293) separates distinct

intracellular and extracellular binding sites that are enclosed by their own intracellular (Y93,

E208, and Y365) and extracellular (W202 and S26) gates. This unique gate structure is thought

to be necessary to accommodate the antiporter nature of AdiC. Arginine would be transported

intracellularly by first passing through the extracellular and then intracellular binding sites, and

once in an inward facing conformation agmatine would be transported out by passing through

the two binding sites in reverse order. This mechanism would allow the extracellular binding

site to have a higher affinity for arginine, and the intracellular binding site a higher affinity for

agmatine. How far might a gate travel between open and closed conformations? Comparing

AdiC structures with and without arginine shows a tryptophan gate on TM 6 (W202) travelling

10 Å to reach the closed conformation, caused by a 40° rotation of TM 6 around the unwound

region167

.

23

1.3.6 Substrate Site

The substrate binding site is located at the breaks in TMs 1 and 6 at the center of the protein,

about 6 Å from either Na+ site

153, and in a pocket devoid of water (see Fig. S5

153 and Fig.

S15158

). Structures captured in an outside facing conformation (LeuT, Mhp1 and AdiC) show

the substrate extending away from TMs 1 and 6 towards 3 and 8 which assist in binding (Fig.

8A–C). In contrast, the two inward facing structures (vSGLT and BetP) show the binding site

shifted towards TMs 2 and 7 and in-between 1 and 6 (Fig. 8D and E). This correlation between

the location of the binding site and the conformation of the transporter may reflect a movement

of the binding site as the transporter switches between outside and inside facing, or conversely,

the location of the binding site may affect which conformation is lower in energy and therefore

crystalized.

Two of the structures transport amino acids, LeuT (leucine) and AdiC (arginine), and this carries

over to similarities in their binding sites153,167

(Fig. 8A and C). In both cases the substrate

carboxyl is oriented towards the unwound region of TM 1 and is coordinated by mostly

backbone nitrogen, while the amide nitrogen points towards the unwound region of TM 6 and is

coordinated by multiple oxygens on TM 6 and one on TM 1 (see Table 3). Helical dipole

moments caused by the unwound regions of TMs 1 and 6 also contribute to binding, with the

positive dipole of TM 1 interacting with the carboxyl group and the negative dipoles of TMs 1

and 6 with the amino group153

. The aliphatic portion of the substrate extends towards TM 3 and

is surrounded by multiple hydrophobic side chains from TMs 3, 6 and 8, which include the

gates. The guanidinium group of arginine spans the binding site with the tip grazing TM 3.

These nitrogen are coordinated by three oxygen on TM 3, one on TM 10 and cation-π

interactions with a tryptophan on TM 8. LeuT has been crystallized with amino acid substrates

of various size, including alanine, glycine, leucine, methionine, 4-F-phenylalanine and

tryptophan and these are all shown in Fig. 10163

. Comparing these structures demonstrates that

there is excess space within the binding pocket towards TMs 3 and 8, and that as the side chain

increases in size it extends in this direction.

Although betaine has a similar structure to the carboxyl and nitrogen group of amino acids, its

binding site is significantly different from the site in LeuT and AdiC. Betaine is enclosed in a

tryptophan box built from four residues on TMs 2 and 6, a common motif of betaine-specific

24

binding proteins required to prevent the repulsion of this highly hydrophilic osmolyte by the

protein backbone155

(Fig. 8D). This tryptophan box is one section of a hydrophobic pathway

spanning the membrane (Fig. 7E). It has been proposed that betaine travels through several of

these binding sites during transport.

Galactose is bound in vSGLT by a perimeter of two charged and four polar residues that

hydrogen bond with the six substrate oxygen through two backbone oxygen, two side chain

oxygen, and four side chain nitrogen situated on TMs 1, 2, 6, 7 and 10 (Fig. 8E and Table 3).

The galactose is then sandwiched on both sides by one intracellular and three extracellular

hydrophobic gates (Fig. 7B).

In the Mhp1 structure benzyl-hydantoin is folded into a V-shape with the tip pointing

intracellularly (Fig. 8B). The hydantoin and benzene rings form pi-stacking interactions with a

tryptophan on TM 3 and 6, and the remaining oxygen and nitrogen atoms bond with four polar

residues on TMs 1, 3 and 8.

1.3.7 Cation Sites

Given the similar architectures of the LeuT fold structures, there are a surprising variety of

cations that can be accommodated. Two of the six structures are Na+ symporters, transporting

one Na+ ion (Mhp1 and vSGLT), two transport two Na

+ ions (LeuT and BetP), one is a virtual

proton antiporter (AdiC) and another a proton symporter (ApcT) (Table 2). This diversity

appears to be accomplished through a shared cation binding site between all of the structures (no

site was found for AdiC). This site, referred to as Na2 based on its naming in the LeuT structure,

is located between TM 8 and the unwound region of TM 1 half way across the membrane (Fig.

9). Coordination of Na+ at the Na2 site is remarkably similar for all of the transporters (Table 3).

It involves five oxygen, 60% backbone from hydrophobic residues (Ala, Ile, Met, Val and Gly)

of which 50% are alanine, and 40% side chain oxygen of polar residues (Ser and Thr). The bend

in TM 1 contributes two backbone oxygen, with one backbone and two side chain oxygen

coming from TM 8 (TM 5 is also involved in BetP). This corresponds to a trigonal bipyramid

arrangement in LeuT153

and a square pyramid in Mhp1154

. Cation coordination at the Na2 site of

ApcT is unique because it involves the binding of a proton, instead of Na+, to a lysine residue

protruding from TM 5 into the Na2 site. This protonated lysine is then coordinated by one

backbone oxygen on TM 1 and one side chain oxygen on TM 8.

25

Two of the crystallized transporters, LeuT and BetP, operate with a 2:1 cation:substrate

stoichiometry. As expected, a second Na+ site, Na1, adjacent to Na2 and the ligand carboxyl

group has been identified for both. Na1 is located between TMs 1 and 7 in LeuT and between

TMs 1, 3, 6 and 8 in BetP (Fig. 9D and E, and Fig. 5). In both cases, multiple backbone and side

chain oxygen are involved as well as the substrate carboxyl group (Table 3). In the LeuT

structure these coordinating oxygen are arranged in an octahedral. Participation of the substrate

carboxyl in Na+ binding demonstrates the potential for direct coupling between the substrate and

ligand.

LeuT is the only protein with the LeuT fold that interacts with a chloride ion, and is also the

only structure containing an extracellular helix between TMs 3 and 4, adjacent to the chloride

(Fig. 5). This extracellular helix, along with TMs 4 and 8, form the Cl− binding site. However, it

is not immediately clear how Cl− affects the transporter from this position at the periphery.

1.3.8 Inhibitor Sites

LeuT is currently the only transporter with the LeuT fold that has been crystallized in the

presence of inhibitors. However, because of the simplicity of these mechanisms there is a good

chance that the other structures are inhibited in a similar way.

A number of amino acids of various sizes are transported by LeuT including glycine, alanine,

leucine, methionine and tyrosine. However, tryptophan is not, and instead acts as a competitive

inhibitor, becoming wedged in the binding cavity because of its large indole ring (Fig. 10F). The

extracellular gates, Y108 and F253, are forced apart 3 Å by tryptophan (Fig. 11A, and compare

Fig. 10F and E), and the R30/D404 salt bridge is separated by 8 Å, preventing its formation and

allowing solvent access to the binding site163

. In this state the extracellular pore cannot close,

nor can the transporter progress to an inside facing conformation, thus preventing intracellular

release. Three additional low-affinity binding sites were also identified (Fig. 11B). One is

located in the extracellular pore between R30 and D404, and blocks the extracellular release of

tryptophan at the high affinity site. There are also two other sites on the periphery, but they are

not thought to be in functionally significant areas163

.

LeuT belongs to the neurotransmitter sodium symporter (NSS) family, which is inhibited non-

competitively by tricyclic antidepressants (TCA). TCAs were commonly used to treat

26

depression, presumably through an interaction with NSS members, but have been replaced in

recent years by selective serotonin reuptake inhibitors (SSRI), which are considered to have

more favorable side effects. The mechanism of TCA inhibition is elucidated by the LeuT

structures bound with leucine and the TCAs clomipramine, imipramine and desipramine. In all

cases the TCAs bind directly above the extracellular gates, Y108 and F235, and occupy the

extracellular pore164,165

(Fig. 11C). The TCAs reinforce the salt bridge between R30 and D404

by bringing the residues closer together, and the bulky TCA shields the salt pair from solvent. In

addition, R30 is held firmly in place by a cation-π interaction with the extracellular gate F253

directly below, the TCA directly above, and a hydrogen bond network between its guanidine

group, the sodium at the Na1 site, and the substrate’s carboxyl164

. Lastly, a salt bridge between

the TCA’s N2 atom and D404 anchors the inhibitor (Fig. 11C). These interactions work together

to stabilize the TCA within the extracellular pore, and outline the mechanism of non-

competitive inhibition. An additional low affinity TCA binding site is located at the intracellular

tip of TMs 4–5, but it is not clear if this site is functionally significant (Fig. 11D).

Competitive inhibition is possible when the substrate has the key carboxyl and amide motif

common to amino acids, but a side chain that is too bulky to fit within the binding pocket.

However, as illustrated by tryptophan binding to LeuT, a secondary binding site within the

extracellular pore (alternate site 1) may be required to prevent the inhibitor’s release (Fig. 11B).

It is possible that the other cotransporter structures are inhibited in similar ways, as may be the

case for phloridzin inhibition of SGLT1, considering that phloridzin is the natural substrate

glucose bonded at the C1 position to phloretin. However, there is some evidence that phloridzin

is transported, in which case the inhibition mechanism may be a very slow translocation

step170,171

. In comparison, non-competitive inhibition involves binding within the extracellular

pore, and stabilization of the protein conformation, possibly through a salt bridge interaction.

1.3.9 Thick Gates

Not all of the structures retain the substrate with a thin gate layer. Mhp1 holds the substrate in

the extracellular binding pocket by bending TM 10 over the pore, essentially using the helix as a

gate (Fig. 7D). BetP and ApcT were crystallized in an intermediate conformation with both

pores partially closed to various degrees by the collapsing of the surrounding TMs (Fig. 7E and

F). In the BetP structure there is a rather long pathway spanning the membrane between TMs 1,

27

2, 6 and 7, lined by 23 highly conserved aromatic residues in BetP and the

betaine/choline/carnitine transporter (BCCT) family155

. It has been proposed that this

hydrophobic pathway allows the hydrophilic osmolyte betaine to travel through the protein

without being repelled by the protein backbone. In the crystal structure betaine is trapped within

a four residue tryptophan box, but is thought to travel through multiple binding sites of this type

that line the pathway and are potentially separated by a series of gates155

. There is no substrate

bound in the closed ApcT structure but both pores are blocked by inward bending TMs.

There appear to be two complementary mechanisms for occluding the binding sites of these

carriers. Thin gates that interact directly with the substrate, often sandwiching it in place and

leaving the pore intact (LeuT, vSGLT, and AdiC), and thick gates that involve occlusion of the

pore by multiple collapsed TMs (Mhp1, BetP, and ApcT). There is also variation in the extent of

thick gate collapse between an initial lightly closed state (Fig. 7D–F), and a tightly packed one

formed when the carrier orients to the other side (Fig. 6A and C).

1.3.10 Conformational Changes

The Mhp1 and AdiC structures were solved with and without substrate. In both cases

conformational changes take place after binding that result in partial closing of the extracellular

pore. In the case of Mhp1 the extracellular tip of TM 10 bends substantially at its midpoint

towards the pore and traps the substrate154

. This motion is enabled by an adjacent glycine and

proline at positions 371 and 372. Mhp1 is the only transporter seen in an outside facing

conformation with this bend in TM 10. However vSGLT and BetP whose structures have been

solved in inside facing states both show this bend, which occurs at G432 in vSGLT and P528 in

BetP (see Fig. 6B and C).

After substrate binding to AdiC there is a large movement of the extracellular half of TM 6, and

smaller movements in TMs 2 and 10, towards the pore (see Fig. S8167

). TM 6 rotates 40° about

its unwound region, moving W202 10 Å to interact with the aliphatic portion of the substrate.

Considering that a residue on TM 6 moves to make contact with the substrate in both Mhp1 and

AdiC, it is possible that the substrate pulls on TM 6 causing the motion. All of the LeuT fold

structures make contact on TM 6 with the substrate and therefore this mechanism may be

common to all of the transporters.

28

Other conformational changes can be deduced by comparing different transporters in different

conformations. TMs 1 and 6 seem to play a major role in allowing access to both the intra- and

extra-cellular pores. Each half of these split TMs seem to move independently to block or

expose each pore with the unwound region acting as a hinge (see Fig. S14172

). Outside facing

structures (LeuT, Mhp1 and AdiC) have the extracellular half of TMs 1 and 6 bending away

from the extracellular pore allowing access, while the intracellular halves lean in blocking the

inside pore. Intracellular facing structures (vSGLT and ApcT) show an opposite trend. BetP,

which is in an intermediate conformation, has TMs 1 and 6 bending away from both pores,

confirming the potential for each half of TMs 1 and 6 to move independently.

1.3.11 Transport Model

Using the variety of crystal structures as a guide we can put together a sequence of

conformational changes that are likely to take place during the translocation of substrate168,173

.

The proposed model is shown in Fig. 12 and the corresponding crystal states are arranged

below. Mhp1 and AdiC were found in an outside facing state with a clear path to the binding

site (state 1). Thin gates can close and block the binding site (2), but a solvent accessible pore

remains (AdiC and LeuT). The pore is then covered by one or more TMs acting as thick gates

(3), with no clear path on either side of the membrane (Mhp1, ApcT, BetP). The thick

intracellular gates open to reveal the intracellular pore (4), but a thin gate blocks release of the

substrate (vSGLT). The thin gate must then open to release substrate (5), although no structures

have been captured in this state yet. This model suggests that four transitions occur during

transport. Two large scale conformational changes, where whole TMs act as thick gates to

alternately expose pores on either side of the membrane. Two smaller conformational changes,

where thin gates made of several hydrophobic residues move to expose the binding site to either

pore.

This mechanism has been referred to as a gated rocker-switch173

, where the rocker-switch

motion involves the pore switching between inside and outside facing conformations. This is

similar to the rocker-switch mechanism proposed for the major facilitator superfamily, except

without the gates. However, the MFS rocker-switch motion may occur in one step because they

lack the unwound helix structures of TMs 1 and 6 likely required for independent pore

movement.

29

1.3.12 Figures

Table 2: Properties of the LeuT fold transporters.

30

Table 3: Gate, substrate and cation interacting residues for the LeuT fold transporters.

Associated TM segments are colored and labeled, and the interacting O, N or Se is indicated at

the end. Residues with hydrophobic interactions are marked with a ∆, and cation-π with π. Gate

residues are labeled with E (extracellular), M (middle) and I (intracellular) and salt bridge pairs

are labeled ES (extracellular) IS (intracellular).

31

Fig. 4: Timeline of discovery for the LeuT fold structures. Structures are shown in the various

conformations they were obtained, and with transported substrate (green) and/or inhibitor (red). Outside

facing (up), inside facing (down), ions (blue circles), closed gate (black bar), and in some cases a

partially closed pore (small opening). The same structures are sorted by conformation on the bottom.

32

Fig. 5: Organization of the TM segments for the LeuT fold structures. Crystal view is from the extracellular side perpendicular

to the membrane, with a schematic below. The substrates are drawn as space filling spheres C (white), H (white), N (blue) and O

(red) in the crystals and as an orange triangles in the schematics; Na+ (purple sphere), and Cl− (green sphere) in both. In the

schematics, grey lines represent intracellular (dotted) and extracellular (solid) loops. Substrate and ion interactions are indicated

as coloured dotted lines, and gates as solid black lines. The entrance to the outside pore is between TM’s 10 and 6 and the inside

pore between 1 and 5. Leading and trailing TMs not part of the LeuT fold have been omitted. All of the crystal structures

visualized here and in the following figures were generated with MacPyMol from the original Protein Data Bank structures.

33

Fig. 6: Extracellular and intracellular pores demonstrated by LeuT, BetP and vSGLT. (A) LeuT,

an extracellular solvent accessible pore is found between TMs 1, 3, 6, 8 and 10, while at the

inside the substrate is occluded by a cluster of TMs. (B) BetP, both pores are partially closed.

(C) vSGLT, the intracellular pore is lined by TMs 1, 2, 3, 6, 8 and 10, while the extracellular

pore is collapsed.

34

Fig. 7: Gating mechanisms. (A) Two gates (Y108 and F253) and a salt bridge (R30/D404) block

the extracellular pore of LeuT. (B) One gate blocks the intracellular pore (Y263), and three line

the extracellular side (M73, Y87, and F424) of vSGLT. (C) Three layers of gates in AdiC, two

at the extracellular side (S26 and W202), one in the middle (W293), and three at the cytoplasm

(Y93, E208, and Y365). (D) TM 10 folds over the binding site locking in the Mhp1 substrate.

(E) A partially closed pathway lined with aromatic residues follows the length of the protein,

suggesting multiple gated binding sites. (F) Inside and outside pores of ApcT partially closed by

the surrounding TMs, in the absence of substrate.

35

Fig. 8: Substrate binding site. The substrate binds between TMs 1, 3, 6 and 8 in LeuT (A), Mhp1 (B) and

AdiC (C), structures captured in an outside facing conformation. In the two inside facing conformations

BetP (D) and vSGLT (E), the substrate is moved away from TMs 3 and 8, towards TMs 2 and 7 behind.

The AdiC and vSGLT structures also show involvement of TM 10. Interacting residues are shown while

TMs not involved in substrate binding are transparent; for residues see Table 3.

36

Fig. 9: Cation binding sites. All of the structures have a common cation binding site at the Na2

position between TM 8 and the bend in TM 1. For vSGLT (A) and Mhp1 (B) this is the only

binding site for Na+, while in ApcT (C) this site is occupied by a proton carried by a lysine on

TM 5. LeuT (D) has an additional Na+ site between TM 7, the bends in TM 1 and 6, and the

substrate carboxyl. For BetP a second Na site has been predicted between TM 8, the bend in

TM1, and the substrate carboxyl; while the Na2 site is assisted by two residues on TM 5.

Substrates are shown as white (carbon), red (oxygen), blue (nitrogen), and purple (Na+) spheres.

Interacting residues are shown (see Table 3).

37

Fig. 10: LeuT with multiple bound substrates. In all of the structures the substrate carboxyl group is

oriented towards TM 1 and the amino group towards TM 6. As the substrate increases in size the new

mass extends down and towards TMs 3 and 8. Tryptophan is too large to be transported and is wedged in

the binding site, inhibiting the protein. 4-F-phenylalanine was used as a replacement for tyrosine,

because of its low solubility163

.

38

Fig. 11: Competitive and non-competitive inhibitor binding sites in LeuT. Tryptophan competes for the leucine

binding site in LeuT (A), where it wedges the extracellular gates, Y108 and F253, and salt bridge, R30/D404, open.

There are three alternate tryptophan binding sites (B), located in the extracellular pore (1), the extracellular loop

between TMs 7 and 8 (2) and intracellularly at the base of TMs 1 and 7 (3). Clomipramine is a non-competitive

inhibitor that binds in the extracellular pore between the R30/D404 salt bridge. An alternate clomipramine binding

site is located intracellularly between TMs 4 and 5 (D). Desipramine and imipramine bind to the same high and low

affinity sites as clomipramine. Most intra and extracellular loops have been removed for clarity.

39

Fig. 12: Transport model predicted by the various conformations of the LeuT architecture

captured in crystal structures. The core of the carrier is represented here by the two unwound

helices 1 and 6, which bind substrate at their center and facilitate the rocker-switch motion. Thin

gates (black) composed of several hydrophobic amino acids have been found at the entrances to

the intra and extracellular pores. Thick gates comprised of one or more bent or collapsed TMs

fill the pore on either side, and transition the carrier between inside and outside facing

conformations. Transport begins in state 1 with an open extracellular gate and pore, substrate

binds and the gate closes (2), the outside pore closes leading to an intermediate state (3), the

inside pore opens resulting in an inside facing conformation (4), the inside gate opens and

releases substrate (5). Captured transporter structures associated with each of the model states

are shown below.

40

1.4 Rationale

Electrophysiology has become an important tool in the study of membrane transporter proteins

because it provides superior control over them via the membrane potential, and rapid high-

resolution measurement of transporter activity via the membrane current. It is often the preferred

method of characterization by providing a wide range of kinetic measurements through the

study of transient and steady-state transporter currents. Despite being an excellent experimental

system with access to a variety of kinetic parameters ( , , , , ), there remains a

disconnect between these parameters and an understanding in terms of a model or mechanism.

The goal of this thesis is to find ways to extract as much kinetic information as possible from the

transient and steady-state currents, to achieve a more intimate view of carriers in action, and

come closer to understanding how they work.

All of the experimental work presented in this thesis—the transient kinetic studies of SGLT1 in

§2 Dissecting the Transient Current of SGLT1 and the sampling of steady-state data analyzed in

§3 A Practical Method for Characterizing the Voltage and Substrate Dependence of Membrane

Transporter Steady-State Currents—were collecting with the Xenopus laevis oocyte expression

system using the two-electrode voltage-clamp technique. SGLT1 DNA or RNA is injected into

the oocytes and over a period of 4–6 days transporter protein is grown in abundance and inserted

into the membrane. The large size of these oocytes allows them to accommodate two electrodes,

of which one is used to measure the membrane potential, while the other controls it by injecting

current through a feedback loop to maintain the desired voltage; this injected current is the

current measured during experiments. This system is one of the few ways available for studying

the electrogenic properties of cotransporters because of their small currents and normally small

expression levels in native tissues. The majority of the data presented here is from experiments

with SGLT1 and several SGLT1 mutants, as this carrier has been the primary focus of our lab

for some time. SGLT1, however, is an excellent model system for ion-coupled cotransport in

general, and, therefore, the ideas presented here can be applied to many other ion-coupled

cotransporters and, perhaps, some pumps and cannels. SGLT1 was used to initially demonstrate

the viability of expression cloning in oocytes, and this head start has put it at the forefront of

cotransporter kinetic studies, making it, arguably, the most kinetically well characterized

cotransporter.

41

The current that arises in response to a step-change in membrane potential has two phases. An

initial transient phase, marked by an exponentially decaying current as the carriers settle,

followed by a steady-state, during which the carriers cycle as they transport substrate. Thinking

about the transport model in terms of a sequence of states connected in a loop (e.g. Fig. 1A), the

transient and steady-state currents provide uniquely tinted windows of different regions of this

loop. As will be shown in the two research chapters that follow, the transient current is a family

of exponential decays that directly monitor conformational changes of the protein, while the

steady-state current measures the cycling rate, which is affected by rate limiting segments of the

loop. Transient experiments are typically more difficult to implement and analyze than steady-

state, but the payout in kinetic detail is worth it, as this is the only means to observe individual

transitions of the protein. In contrast, steady-state currents measure lumped parameters, but the

data is easy to collect and analyze, and reveals parts of the loop that are normally inaccessible to

transient studies. Since each type of study exposes different segments of the transport loop in

different ways, both are needed to build a complete a model of the transporter.

Within the past decade significant advancements in the field of membrane protein

crystallography have resulted in the first high resolution structures of channels, pumps, and

cotransporters. While more recently the LeuT architecture has been found throughout a broad

superfamily of Na+ coupled cotransporters, of which SGLT1 is a member. The variety of

conformations that this LeuT architecture has been captured in seems to identify a gated rocker-

switch mechanism that confirms the alternating access hypothesis proposed 45 years ago by

Jardetzky2. Although we are now at the point where we can visualize these structures in great

detail, their dynamics remain a mystery. It is the view of this thesis that by combining these

structures with comparably detailed kinetic data, we can fill in the missing pieces and complete

the picture of how transporters work.

In §2 Dissecting the Transient Current of SGLT1, the transient currents of SGLT1 are

decomposed into as many decay components as possible, to study the conformational changes

they report on and learn about the gated rocker-switch mechanism. This is normally a hard

problem to solve, yet we present a methodology that is successful and can be repeated with

other ion coupled cotransporters. In §3 A Practical Method for Characterizing the Voltage and

Substrate Dependence of Membrane Transporter Steady-State Currents, a general theoretical

model of membrane transport is developed and used to understand the voltage and substrate

42

dependence of the steady-state currents. This ultimately leads to a phenomenological method for

characterizing the classic I–V curves, and an understanding of the parameters that are derived as

representations of rate limit segments of the transport loop. These parameters can then be used

to reconstruct a steady-state model of the carrier. For both the transient and steady-state phases

we have attempted to extract as much kinetic information as possible, and, in each case, this has

resulted in more direct and simple ways of understanding the data in terms of a kinetic model.

43

Dissecting the Transient Current of SGLT1 2

2.1 Introduction

Electrophysiological studies of cotransporters, and in particular their kinetic characterization,

have been common practice ever since expression cloning was demonstrated 25 years ago in the

heterologous Xenopus laevis oocyte system using SGLT1 as a model protein9. This system

provides an opportunity to stimulate overexpressed carrier proteins residing within the cell

membrane by controlling the membrane potential with electrodes. The stimulated carriers

undergo conformational changes that can be studied by monitoring at the membrane current

generated by charged amino acids on the protein or ions as they move, or are isolated, within the

membrane electric field.

This electrophysiological system provides a unique opportunity to observe the transient kinetics

of membrane proteins, generated by step-changes in membrane potential, because of the high

speed with which the membrane potential can be controlled and the carrier currents measured.

These transient currents reflect the kinetics of the overexpressed carriers as they equilibrate

between their initial distribution prior to the voltage jump and the steady-state distribution after

the voltage jump. In 1992, several years after expression cloning was introduced, the first

transient experiments with SGLT1, and one of the first for any cotransporter, were performed174

.

They found that SGLT1 produces transient currents in the absence of glucose, and the addition

of glucose inhibits them. This behavior has been explained recently by showing that glucose

promotes an inside facing conformation where the carrier is electrically silent, hence the absence

of transient current—this corresponds to movement between states 4, 5, and 6 in the classical

model of SGLT1 shown in Fig. 1A77,175

. In the absence of glucose the carrier is restricted to

movement between states 1, 2, and 6, and it is the conformational changes of the empty carrier

1↔6 that are believed to generate current, as Na+ binding is typically much too fast to observe

(0.004176–0.06177 ms). Although state 5 is technically active in the absence of glucose, its

occupancy is highly unfavorable under normal conditions (low intracellular Na+), and is

typically ignored.

44

Recent crystallographic data of a superfamily of cotransporters that share a common architecture

with SGLT1 has provided deeper insights into the conformational changes of the carrier, as

discussed in §1.3 Crystal Structures of the LeuT Fold. As shown in Fig. 12, the various states

that these cotransporters have been crystalized in sketch out a transport mechanism, termed a

gated rocker-switch, that involves four conformational changes of the carrier, situated between

states 1↔6 (and 3↔4) in the classical model of Fig. 1. This mechanism works by mediating

movement in and out of a central substrate binding pocket by way of a mobile pore and gate on

each side of the membrane. One of the few ways of observing these conformational changes in

action is through the transient kinetics. The transient current consists of a number of exponential

decays equal to one less than the number of carrier states, and for a non-cyclical system, like

1↔6, this results in one decay per transition. As we will show, by measuring the decay time

constants (τ) and amplitudes (A) we can learn about the rates and ordering of these transitions.

There are, however, significant challenges involved in the measurement and analysis of these

transient currents. In particular, with respect to the problem of multi-exponential fitting, and the

stimulation of sufficiently large transient currents for accurate fitting.

Multi-exponential fitting is a notoriously difficult task, because of the interconnectedness of the

decays. As will be demonstrated, fitting with an insufficient number of decays skews the results

into a type of weighted average of the true carrier kinetics, and these results can have the

unintended consequence of obscuring the underlying mechanism. However, it can be difficult to

find the correct number of decays, since, at the onset, there is no way of knowing how many to

look for. In addition, it becomes geometrically harder with each exponential term added to the

fit equation to perform the fits and find a successful solution. This project is focused on the

general problem of transient analyses using SGLT1 as a model system; SGLT1 was one of the

first cotransporters to be studied using these techniques, and has become one of the most well

characterized kinetically. A unique protocol is presented for collecting transient data that uses

multiple holding potentials to produce large decays. Also, a methodology for multi-exponential

fitting is demonstrated that provides measures for evaluating the quality of a fit and indicators

for helping to decide when too few or too many exponential terms are being used.

As discussed in the next section initial transient studies of SGLT1 characterized the transient

currents with a single exponential decay, but over time this number has gradually increased to

three. However, the current model based on the crystallographic data suggests four

45

conformational changes of the carrier, and therefore four decays. In this study the transient

kinetics of SGLT1 and a threonine to cysteine mutant at position 156 (T156C) were

characterized using these new methods. Four carrier decays were found for each, in agreement

with the crystallographic model, with rates of 2, 5, 25, and 50–200 ms. In addition, charge

movements for each decay (Q=Aτ) were found to differ depending on the direction of the

voltage jump. This phenomena suggested that some transitions might be masked when moving

in one direction, and this was interpreted as an effect seen when a slow transition is in front of a

faster one and hides it. Using this directional charge information, four conformational changes

were aligned in series in a way that could explain the data. When this kinetic model was then

aligned with the crystallographic, the fast decays (2 and 5 ms) were found to correspond with

movement of the intra and extracellular gates, and the slow decays (25 and 50–200 ms) with the

intra and extracellular pores. Although these experiments were performed with SGLT1, the

techniques presented here are general enough to be used on a wide variety of electrogenic

cotransporters that produce transient currents. Furthermore, now that the gated rocker-switch

mechanism has been described kinetically, other carriers can be tested for a similar or different

mechanism.

2.1.1 The T156C Mutant

We decided to study the T156C mutant because it is located in a critical region for substrate

binding. This is exemplified by the neighboring lysine to cysteine mutation K157C, which

completely abolishes glucose and phloridzin binding16. It is our hypothesis that access to the

binding site is blocked, and that the lysine at this position is possibly involved in a salt bridge

that mediates a conformational change that provides access to the binding site. The T156C

mutant retains full function, making it more suitable than K157C for kinetic studies, but has a

significantly reduced apparent affinity for the inhibitor phloridzin (30–170 μM versus 1.6±0.6

µM for wt, Fig. 13A). This reduced affinity alters the interaction of phloridzin with the carrier

into a weak inhibitor that unbinds when the membrane potential is depolarized. This behavior

can be observed as a new slow decay that appears in the presence of phloridzin (Fig. 13C), and

stands out from phloridzin`s normal role as a tight-binding inhibitor178. Lastly, the Q–V

distribution, measured by integrating the transient charge movements, is right-shifted by a large

amount (38 mV, Fig. 13B), indicating that the carrier has difficulty changing into an inside-

facing conformation (i.e. with the extracellular pore closed and the intracellular open). These

46

characteristics appeared to indicate an affected conformational change, making this mutant an

interesting candidate for transient studies. As shown in Fig. 14, by the placement of their aligned

residues in the vSGLT structure T156 and K157 are near, and equidistant to, the Na+ and

glucose binding sites. This places them in an active region of the carrier that is consistent with

effect of their mutations.

2.1.2 Historical Perspective

The first electrogenic study of SGLT1 transient kinetics was performed in 1992 on the rabbit

isoform, by observing conformational changes of the empty carrier in the absence of

glucose70,174. The transient decays were fit with a double exponential and steady-state term

( ⁄ ⁄ ), finding a fast (0.95–2.3 ms) and slow (4–18 ms) component.

Despite some voltage dependence the fast was concluded to be voltage independent and was

assigned to the membrane capacitance (typically ~0.5 ms), while the slow peaked at negative

potentials, was sensitive to αMG and phloridzin, and was attributed to the carrier. These

observations were repeated the following year for the human carrier (fast 0.54–0.82 ms, slow 2–

8 ms), with the fast again being assigned to the membrane, and the slow peaking, instead, at

positive potentials83. Attempting to find the amino acids responsible for the difference between

the slow decay of rabbit and human, the D176A mutant of rabbit was characterized in 1994179.

In this case, the fast decay was too slow (2.5–4.6 ms) to attribute to the membrane capacitance,

yet this was overlooked despite hinting at the presence of a second carrier decay. The slow time

constant peak did shift to positive potentials (4.6–28 ms) to resemble human, but there were

other factors that led to the conclusion that other amino acids were responsible for the

differences between species.

With these early studies two protocols were commonly used to analyze the transient currents of

SGLT1 and other carriers, and in particular in how they dealt with the membrane capacitance.

The subtraction method involved subtracting readings with and without an inhibitor to remove

the membrane and endogenous currents, and isolate the carrier current which was then fit with a

single exponential (SGLT183, GAT1180, SERT181, EAAT2182). Alternatively, the fitted method

would fit raw data with a double exponential and assign one component to the capacitive and the

other to the carrier (SGLT183,174,179, STP1183, SMIT184, PEPT1184).

47

The first attempt at a more detailed analysis of SGLT1 transient currents was made by Chen et

al. in 1996 using the human carrier85. They used the subtraction protocol and the recently

introduced cut-open oocyte technique (1992)185

, because of its higher initial time resolution

(0.08–0.35 ms) and ability to control the intracellular solution. The inhibitor-subtracted currents

were fit better with a double exponential (fast 0.4–0.8 ms, slow 2–10 ms), which indicated the

presence of two carrier decays and identified the faster decay hinted at earlier by the D176A

mutant. Also, by taking advantage of the cut-open oocyte they were able to show that both

decays remained in the absence of intra and extracellular Na+. This finding allowed them to

conclude that both decays were generated by the empty carrier, and not by binding of

extracellular Na+, leading to a proposed extension of the standard model with two

conformational changes of the empty carrier (see Fig. 1B). Although unaware of it at the time,

this was the initial step in decomposing the gated rocker-switch mechanism. However, despite

the conclusiveness of this finding, quite a few studies continued for some time to use the fitted

method, with the faster decay being assigned to the membrane capacitance (SGLT1175,186,

hCNT3187, GAT1178).

Up to 2003 transient studies of SGLT1 commonly evaluated the quality of the exponential fits

by eye, and limited the analysis to two decays. Encouraged by the findings of Chen et al., we

wanted to evaluate the number of carrier decays in a more rigorous way, and did so by

introducing a number of advancements86. We extended the transient analysis to 150 ms, in

comparison with the 40–100 ms used by others15,18,85

, to allow the detection of slower decays.

Residuals were used to evaluate the quality of the fits and help decide if additional decays were

present. Lastly, the protocol used to stimulate the transient currents incorporated multiple

holding potentials to obtain the largest decays possible for improved exponential fits. At the

time most protocols used one holding potential of −50 mV, and small voltage jumps, like −50 to

−30 mV, would result in small decays that were difficult to fit accurately. With these methods,

and using the subtraction protocol, three carrier decays were found by fitting with a three

exponential function (fast 0.5–1 ms, medium 0.5–4 ms, and slow 8–50 ms). The fast and slow

decays were analogous to those identified by Chen et al., while the medium was new. We were

unable to test for all three decays with zero trans Na+ because we used the two-electrode

voltage-clamp and not the cut-open oocyte. To be conservative, we refrained from

hypothesizing an additional empty carrier conformational change, and instead left open the

48

possibility that the medium decay was related to extracellular Na+ binding. However, in

hindsight Na+ binding is much too fast to account for the medium decay (0.004176 and

0.06177ms), which we now know is related to the empty carrier. Subsequent studies by our group

have applied this technique to other mutants, including Q170C188, Q170E189

, Q457C and

Q457R190.

Another detailed study of SGLT1 transient kinetics was published in 2005 by Loo et al84. They

used a curve peeling strategy to fit the decays over different time domains to a simple

exponential function with one or two terms. Recordings of 500 ms were fit with a single

exponential to measure the slow decay (25–150 ms). This decay was then subtracted from 100

ms recordings, which were then fit with a double exponential to obtain the medium (2–30 ms)

and capacitive (~0.5 ms) decays. Finally, 5 ms recordings made with the cut-open oocyte were

fit with a single exponential to characterize a novel fast rising component (0.17–0.55 ms). All

three decays remained with zero trans Na+, requiring another empty carrier conformation in the

model, for a total of three (Fig. 1C).

Overall there remains minimal interest in qualifying transient analyses, and this appears to have

limited advancements in the field. In a review of the literature that included transporters other

than SGLT1 the most exponential decays that were found to be used was four (61, 767, 3278

and 5424 ms) in a study of voltage-activated outward K+ currents in ventricular myocytes191.

This study used a statistical method to determine the correct number decays by taking advantage

of a large number of repeated measurements. The above study, an investigation of the Na+/K

+

pump with the cut-open oocyte192, and the neuronal excitatory amino acid carrier (EAAC1)

studied in human embryonic kidney cells with the patch clamp193, used residuals to qualify the

fits. Three exponential decays were used in several other studies of ventricular myocyte outward

K+ currents (79, 310, and 1802 ms194; 40, 350, and 1600–2000 ms195; 8–12, 30–40, and 500–

600 ms196), and the EAAC1 study mentioned above (0.4–0.7, 1.2–1.7, and 8.1–12 ms)193. Using

curve peeling, three decays (0.06, 0.22 and 4 ms) were found with the Na+/K

+ pump in squid

giant axon177. Other studies we came across fit the transient current with one or two

exponentials.

49

2.1.3 This Study

This project is an evolution of our earlier SGLT1 transient studies86,188-190

. It began out of a

desire to characterize the T156C mutant, and, in particular, the unique phloridzin transient

currents that it produces (Fig. 13C). Without an effective inhibitor for this mutant, we switched

from the subtraction to the fitted method for handling the capacitive decay. This had the added

benefit of removing the potential for subtraction artifacts, and simplified interpretation of the

results. We had also come across data showing an unexpected inequality of charge movement

between the on and off Q–V distributions collected to 150 ms—an on distribution represents

charge mobility for a range of voltage jumps away from a fixed holding potential, while an off

distribution represents returned charge when jumping back to the same holding potential—, and

we hypothesized that this might be caused by a very slow decay who’s charge was inadequately

collected in one direction. This was confirmed when Loo et al. found a 150 ms decay with

hSGLT184. To accommodate this slower decay, we began collecting data out to 300 ms. We

refrained from using a 500 ms pulse like Loo et al. because we found this to be excessively

stressful on the oocyte, often resulting in current drift.

Because of these changes we were expecting to potentially fit five decays, three original,

hypothesized slow, and the membrane capacitance. Knowing that this would be challenging, we

focused on indicators that would help decide how many terms to use. These, ultimately, came

down to the fit residual as an indicator of the need for an additional term, and the standard error

of the fit parameters as an indicator of potentially too many terms. These indicators worked

together in a complementary way to add and subtract terms from the fit equation while

searching for valid solutions.

We decided to first characterize the transient kinetics of the T156C mutant, and for comparison

wt, in a standard 100 mM Na+ solution, with and without phloridzin to test for inhibition. As

expected, this resulted in the observation of five decays (0.5, 2, 5, 25, and 50–200 ms). What

was noticed this time around was the importance of the charge information associated with each

decay ( ). This provided to some degree information about the level of activity of the

conformational changes associated with the decays. As will be discussed in more detail later,

each decay’s charge movement was found to be significantly different depending on the

direction of the voltage jump (i.e. x→y versus y→x), and we hypothesized that this was caused

50

by a masking effect that occurred when a slow transition was situated in front of a faster one.

Using concepts from this masking effect, we were able to build a model relatively easily that

could account for the unequal directional charge movements. This kinetic model turned out to

correspond remarkably well with the gated rocker-switch mechanism hypothesized by the

crystallographic studies, and we were able to make a one-to-one assignment between each decay

and the two-gate and two-pore conformational changes.

This report is split into four main sections. Some basic concepts that were used in the design of

the voltage-clamp protocol and interpretation of the data are explained in §2.3 Voltage Jump

Experiment Theory. A detailed walkthrough of the fitting procedure is illustrated for a single

data set in §2.4 Data Analysis. The complete results for wt and the T156C mutant are presented

in §2.5 Results, and conclusions are given in §2.6 Discussion.

2.2 Materials an Methods

2.2.1 Molecular Biology

The cDNA of rSGLT1 was subcloned into the EcoRI site of the eukaryotic expression vector

pMT3 (Genetics Institute, Boston, MA) after removal of the multicloning site by digestion with

PstI and KpnI. The megaprimer protocol of polymerase chain reaction mutagenesis was used to

generate the T156C mutation, which was then confirmed by sequencing15

.

2.2.2 Oocyte Collection, Injection, and Maintenance

Oocytes were extracted from Xenopus laevis frogs in conformity with protocols approved by the

University of Toronto Animal Care Committee. The frogs were anesthetized with a 0.2%

aqueous solution of 3-aminobenzoic acid ethyl ester for 30–40 min. The oocytes were then

removed via an incision in the abdomen and the ovarian sacs were separated in a solution of

modified Barth’s saline supplemented with MgCl2 (MBS: 0.88 mM NaCl, 1 mM KCl, 2.4 mM

NaHCO3, 15 mM HEPES-NaOH, 1 mM MgCl2, pH 7.4). The vitelline membrane surrounding

the oocytes was removed by digestion with 2 mg/mL of type IV collagenase (Sigma, Oakville,

Canada) dissolved in MBS for 25–60 min. When digestion was complete the oocytes were

washed several times with MBS and then maintained in Leibovitz L-15 solution (Sigma)

supplemented with 0.08 mg/mL gentamicin, 0.736 g/L L-glutamine, and 10 mM HEPES-NaOH

at pH 7.4.

51

After collagenase treatment the oocytes were left to rest overnight, and to allow time for

potential damage to the membrane to show. Healthy oocytes were then selected for injection. A

Drummond Scientific Nanoject II (Broomall, PA, USA) was used to inject 9.2 nL of 150 ng/μL

of rabbit SGLT1 wt or T156C mutant cDNA into the nucleus of the oocyte via the animal pole.

Over the following 4–6 days the protein was given time to accumulate in the membrane and

once per day the oocyte bath was changed and dead or dying eggs were discarded.

2.2.3 Two-Electrode Voltage-Clamp

The two-electrode voltage-clamp technique was used to control the oocyte membrane potential

while simultaneously measuring the membrane current174. A GeneClamp 500 amplifier and

Digidata 1440A analog-to-digital converter were used along with pClamp 9.0 data acquisition

software (Molecular Devices, CA, USA). Electrode tips were made from 150 μm borosilicate

glass capillary tubes pulled with a model p-97 Flaming/Brown micropipette puller (Sutter

Instrument Company, CA, USA), and filled with 3M KCl. Electrodes typically had a resistance

of 0.1 MΩ when first inserted into the oocyte, and in some cases this resistance would rise and

fall as membrane was lodged and dislodged from the tip. The resistances of both tips were

checked regularly, and if they were equal to or greater than 1 MΩ the tip was replaced. The

amplifier gain and stability were adjusted with each new egg, and typically set to values of 1000

and 50 μs, respectively. Data was recorded at a sampling interval of 10 μs, with the built in low-

pass filter on the GeneClamp amplifier set to its highest setting of 50 kHz to minimize any

preprocessing of the data. Only oocytes with a resting potential more hyperpolarizing than −30

mV were used.

During experiments the oocytes were perfused with a voltage clamping solution (VC: 100 mM

NaCl, 2mM KCl, 1 mM MgCl2, 1 mM CaCl2, 10 mM HEPES-NaOH, pH 7.4 with Tris Base).

This VC solution was then supplemented further with phloridzin or glucose as called for by the

experiment. Because of phloridzin’s low solubility in water it was first dissolved in 100%

ethanol at a concentration of 50 mM before being added to the VC solution.

2.2.4 Voltage-Clamp Protocol

The two-electrode voltage-clamp experiments performed in this study were designed to

investigate the transient kinetics of SGLT1. These experiments used a multi-holding voltage-

52

clamp protocol (see §2.3.5 Voltage Jump Protocols). The basic premise is to use a range of

holding potentials to study the transient kinetics at each test potential.

There are three phases to the waveform, accompanied by three voltage step-jumps (see Fig. 18C

and D): 1, at t=0 the membrane potential is jumped to one of several possible holding potentials

which is maintained for 300 ms to allow the system to equilibrate; 2) at t=300 ms the membrane

potential is jumped to the test potential and held for 300 ms to measure the transient current; 3)

at t=600 ms the membrane potential is returned to the resting potential and held for 300 ms to

allow the system to return to resting steady-state.

With the T156C mutant, 23 holding potentials were used ranging between −150 and 70 mV in

10 mV steps. Therefore, for each test potential studied, 23 measurements with different holding

potentials were made in series during one experimental run. All together 12 test potentials were

used ranging between −150 and 70 mV in 20 mV steps for a total of 276 measurements per

experiment.

The wt transporter and non-injected control used a shorter voltage range to limit the stress

placed on the oocyte. Holding potentials spanned from −130 to 50 mV in 10 mV steps (19

traces), and test potentials covered the same range in 20 mV steps (10 runs). The T156C

transporter was the first to be studied with this technique, and so the wider range was used to

avoid missing any important charge movements. From these initial studies it became apparent

that the charge V0.5 of the T156C mutant was relatively depolarizing (32 mV, Fig. 13B), and

that the 70 mV test potential was required. The wt carrier has a charge V0.5 which is more

negative (−6 mV), and therefore the 70 and −150 mV test potentials could be dropped to lower

the oocyte stress.

2.2.5 Exponential Curve Fitting

Data fitting was performed with the software package Origin 8.0 (OriginLab, MA, USA), which

uses the Levenberg-Marquardt algorithm. This software has a Nonlinear Curve Fit tool

containing built in exponential functions with 1–3 terms, and allows for custom user defined

functions. We created custom functions with 4–6 exponential terms. The built in functions are

reasonably fast, while the user defined functions are orders of magnitude slower. In both cases,

as the number of exponential terms are increased the fits take longer to compute. For example,

53

using a 3 GHz Core2 Duo E8400 with 2GB of RAM and Windows Vista, the calculation time

per iteration of the Levenberg-Marquardt algorithm was 0.46, 0.64, 0.93, 51, and 73 s for one

through five exponential terms, respectively. These times correspond with fitting,

simultaneously, five 30000 data-point traces. Calculations would typically take 10 to 200

iterations to complete, depending on the distance between the solution and parameter seeds.

This would result in fitting times of 5 seconds to 4 hours. Multiple traces were fit

simultaneously using the Global Fit option, with a tolerance of 10−15

. We experimented with

tolerances between 10−5

and 10−15

and found that in some cases, especially with 4 or more

terms, the latter produced slightly better results.

Sometimes when the carrier signal was small (i.e. from short jumps, phloridzin, non-injected

oocyte) a small slow decay (~250 ms) was detected in response to depolarizing jumps that

moved in the opposite direction of the faster decays—the decay had a negative amplitude

compared with the others which were positive. This decay appeared to be caused by oocyte drift

and was, therefore, omitted from the transient analysis.

2.3 Voltage Jump Experiment Theory

2.3.1 Anatomy of a Voltage Jump

Transporter transient kinetics are often studied by measuring the membrane current generated by

these proteins in response to a fast step-change in membrane potential. This voltage jump

involves a rapid transition (1–1.6 ms for the two-electrode voltage-clamp83,197

) from a holding

potential to a test potential (Fig. 15, jump 2). The test potential is then maintained until steady-

state is reached, to collect as much signal as possible from the slowest decays. Between

measurements the membrane potential is returned to the resting potential, typically −50 mV,

which is close to the unclamped potential of the oocyte. When the resting and holding potentials

are different, a preliminary jump is used to transition between them (jump 1), and the holding

potential is maintained until steady-state is reached—if not, the amplitude data is harder to

interpret.

2.3.2 The Transient

The transient current takes the form of a multi-exponential function,

54

.

Eq. 1

Each exponential corresponds with a decay in the signal that to a first approximation reflects a

redistribution of the protein, characterized by a rate ( ) and amplitude ( ). If a system contains

states decays are expected. The steady-state current is represented by .

Since in practice the amplitudes vary inversely with their paired time constant, a normalized and

better measure of the decay magnitudes is the charge contained within them, calculated as

. For example, the amplitude/time-constant pairs (10000 nA, 1 ms) and (100 nA, 100

ms) both account for a charge displacement of 10 nC.

2.3.3 How the Voltage Jump Affects the Transient

The holding and test potentials each have different effects on the transient kinetics, as

demonstrated by the examples in Fig. 16. The time constants ( ) are only affected by the test

potential, because they depend on the kinetics of the conformational changes which are voltage

dependent. This can be seen in A, where different holding potentials (−150 and 70 mV) have no

effect when the test potential is the same (30 mV), and in B, where different test potentials

(−150 and 70 mV) change the kinetics.

The amplitudes ( ) are proportional to the quantities of charge passing at various rates through

the conformational changes. These depend on the differences between the starting and ending

carrier distributions, and are therefore a function of the holding and test potentials. C shows that

a large jump (−150→−10 mV, red) does not guarantee a large decay, but a smaller jump

(70→−10 mV, blue) can generate one if it crosses the active region of the carrier (30 mV). D

shows the range of charge movements possible in this example, compare −150 and 70 mV

holding potentials. Note how small the charge movements are from the standard −50 mV

holding potential (orange) in C and D, compared with 70 mV (blue).

2.3.4 How the System Affects the Transient

The rates of the state transitions can have a large impact on the composition of the transient

signal. Consider the example in Fig. 17, where a slow (2↔3) and fast (1↔2) transition are

55

adjacent, and the system is loaded into state 1 at large hyperpolarizing potentials and state 3 at

large depolarizing. If we start at a large depolarizing holding potential (state 3) and jump to

hyperpolarizing test potentials, the fast transition will be hidden by the slow. In the transient

signal the fast decay will have a negligible amplitude and the slow a large one, as depicted by

the sigmoidal charge curves above. Moving in the other direction (beginning in state 1) both

transitions will be visible (red curves), because the fast transition is unable to hide the slow. An

interesting corollary is that this same reasoning can work backwards. If we were to see these

charge profiles for transitions with a large difference in rate, we could deduce that the slow

transition was on the right. For this method to work we would require measurements from both

sides to compare.

2.3.5 Voltage Jump Protocols

Transporter transient kinetics at different test potentials are typically measured from a single

holding potential, often with a protocol like the one shown in Fig. 18A and B84,174. The resting

and holding potentials are usually the same to keep the protocol simple (often around −50 mV),

and this results in a series of simple jumps away from and back to the shared resting/holding

potential. An advantage of this protocol is the minimal amount of stress exerted on the oocyte,

but this also comes with a number of limitations. As demonstrated earlier in Fig. 16, small

jumps such as −50→−10 mV in C (orange) generate small transients, and even larger jumps like

the −50→−130 mV in D (orange) will produce a small signal if the active region of the carrier is

not crossed. Furthermore, because the charge movements are only measured from one direction,

the masking effect discussed above (§2.3.4 How the System Affects the Transient) cannot be

taken advantage of to order the transitions.

An enhanced approach that addresses these limitations is to use multiple holding potentials86

,

like the example in Fig. 18C and D. An initial jump equilibrates at one of several holding

potential (0–300 ms) followed by a second jump to the test potential (300–600 ms). For each

test potential a series of readings are taken over a range of holding potentials, like the twelve

holding potentials shown in C for the −10 mV test potential (compared with the one orange

trace in A). The largest signals are analyzed and charge movements from either direction can be

compared, and because the traces decay at the same rate, several can be fit together to increase

the sample size. A disadvantage of this technique is the extra stress placed on the oocyte,

56

because of the doubling in clamp time, and the quadratic increase in measurements; each grey

line in C represents another series of twelve measurements at the other test potentials. In our

experience stress was a factor because of the large variability in oocyte health amongst animal

providers, individual frogs, and the seasons, and robust eggs needed to be screened for.

2.4 Data Analysis

2.4.1 Form of the Transient Currents

An example of a typical transient data set is shown in Fig. 19. The data was produced by jumps

to a −50 mV test potential from a range of holding potentials, and corresponds with the 300–600

ms time window in Fig. 18. We used a high density of holding potentials (Δ10 mV) for a total of

23 traces between −150 and 70 mV. Early time points are dominated by large-amplitude fast

decays (2–20 ms) that vanish rapidly to expose low-amplitude slow decays, which continue on

for several hundred milliseconds (2–300 ms). To show more detail these two regions, which are

shaded, are expanded in B and C. In this data set hyperpolarizing jumps (blue) produce

significantly more charge than depolarizing (red). This occurs because the active region of the

carrier ( mV, Fig. 13B) is to the right of the −50 mV test potential.

2.4.2 Defining the Data Set

When a voltage clamp is applied there is a short period of time at onset (1–2 ms for the two-

electrode voltage-clamp) where the membrane potential is transitioning to the new value and is

unstable. During this period large capacitive currents are also produced that saturate the

equipment198,199. Both phenomena interfere with multi-exponential fitting and, therefore, need

to be avoided. This was done by omitting data between 0–2 ms from the analysis. The early time

points contain information on the fastest decays, however even without data for the first 2 ms

we were able to reliably measure decays with time constants as fast as ~0.4 ms. Faster decays

(0.2–0.3 ms) that appeared in some solutions had excessively large parameter fit errors. The

remaining signal from 2–300 ms was fit as one segment. Others have used a curve peeling

strategy84, but it can introduce artifacts and was not necessary.

An advantage of a multi-holding protocol is a large sample of recordings with the same test

potential. Fitting several traces together is possible because traces with the same test potential

share the same set of decay time constants, and this can be used to increase the signal to noise

57

ratio. Initially we fit all 23 traces simultaneously, but the computational load was too large and

caused the calculations to take prohibitively long. The low amplitude traces were found to

contribute disproportionately to this load, because of the larger uncertainty associated with their

fit parameters, and we decided, therefore, to move forward with a reduced data set that only

included the five most extreme holding potentials on either end (red and blue traces in Fig. 19).

Although all ten of these traces could be fit simultaneously, they were instead fit in two separate

groups based on the voltage jump direction (red depolarizing, and blue hyperpolarizing) to test

for any directional dependence to the time constants.

2.4.3 Fitting

The red and blue traces in Fig. 19 were fit with a series of multi-exponential functions

containing 1–5 exponential terms, as shown Fig. 20. The data are magnified over the short (2-20

ms) and long (2-300 ms) time domains to give a clearer view of these different kinetic domains,

with the fits overlaid in black. The red and blue groups were fit separately, with each group

sharing a set of time constants amongst all five traces. The blue decays, which were larger than

the red, were successfully fit with five exponential terms, while the red could only be fit with

four. In each case there was only one best solution, and its time constants are given below for

each direction.

With 1–2 terms differences between the fits and traces are noticeable by eye, especially on the

longer time scale (2–300 ms). There is a large capacitive decay (~0.5 ms) that carries the

majority of charge, and the fit is weighted towards it when only one exponential term is used (1

and 2 ms). These single exponential fits are slightly slower than the capacitive decay alone

because of the presence of slower, and smaller, decays produced by the transporter. As more

terms are added to the fit equation these slower decays are detected, but it becomes increasingly

more difficult to ascertain the quality of the fit by eye. In the supplemental document Fits and

Residuals these fits are expanded to provide a clearer view. To best appreciate the progression in

fit quality it helps to flip between these images on a display. This highlights how difficult it is to

evaluate the quality of a fit by eye, and the need for other measures, several of which are

discussed below.

58

2.4.4 Fit Quality: Residuals and χ2

A more direct representation of the quality of a fit is the residual, which shows the difference

between the data and fitting curve. Examples of these residuals are presented in Fig. 21 for the

end holding potential traces (−150 and 70 mV). These plots show deviations in the fit, as an

optimal fit will resemble white noise. With only a few terms the deviations are clearly non-

random, but as more terms are added the residuals shrink and the deviations become smaller. By

four terms with the red trace and five with the blue the residuals become flat. As the blue

residuals flatten between four and five terms, there is an obvious smoothing in the first 75 ms

accompanied by a more subtle straightening out of a shallow U-shape lasting to 300 ms. It is

more difficult to see differences between the three and four term red residuals, but there are

small changes in the first 60 ms. These residuals are also presented in an enlarged format in the

accompanying Fits and Residuals document, and we recommend flipping between them to see

the differences more clearly. The red trace failed with five terms and the blue with six, for

reasons that will be discussed in the next section §2.4.5 Nonsense Fits. These residuals show no

significant improvements over the successful solutions with one fewer term. Although obvious

distortions in the residuals are a strong indicator that more terms are needed in the fit equation,

as the residuals straighten out it becomes harder to make this evaluation. In this sense the

residuals can identify a need for more terms, but they cannot determine if too many have been

used. The fit χ2 values were used to rank the overall fit quality (see Fig. 21). The drops in χ

2

eventually became quite small with a large number of terms (for example 78.41 to 78.36),

making it difficult to use the χ2 as an indicator of fit completeness as there was no optimal value

to target.

2.4.5 Nonsense Fits

An important part of the fitting process is a method for distinguishing between valid and invalid

solutions. Some statistical approaches have been used with large data sets191, but they are less

suited to the types of data collected with SGLT1. In this study the standard errors of the fit

parameters were used to identify invalid fits, which often occurred because of over

parameterization. The errors were initially small with only a few terms, but as more terms were

added the errors always increased until eventually one or more became too large for the solution

to make any sense, and the solution had to be rejected. In the more obvious cases the standard

59

error was several times the parameter value, and we called them nonsense fits. However,

sometimes the situation was more ambiguous, such as a standard error of 50–100% the

parameter value. To be conservative these cases were avoided by using 50% error as a hard

cutoff. A fit was rejected, and earned the nonsense label, if any parameter error was greater or

equal to 50%.

Along with large parameter errors there were other illogical markers that helped identify and

explain failed solutions, and these are categorized in Table 4. In row A are conventional

nonsense fits with large parameter errors (beige shading). In rare cases the χ2 would increase

when a term was added, as shown in B (green); for example 67.49 to 67.58 in B2. Sometimes

when the system was over parameterized: one component would have a very small charge

(grey); the fastest component would become too fast (yellow) with a large charge (red); the

slowest component would approximate a straight line with a very slow time constant (orange)

and a large charge (red); the time constants of two terms would converge on the same value

(purple). Bad parameter seed values, or over parameterization, would sometimes result in terms

with alternating amplitude signs (blue).

2.4.6 Seeding

Finding a valid multi-exponential solution is most dependent on the choice of parameter seed

values, since different seeds can lead to different solutions. With just one term the fitting

process is straightforward because of a well-defined global minimum. However, as exponential

terms are added to the fit equation the topology of the solution space roughens, and the potential

to get stuck in a local minimum increases. Often, these local minima are failed fits with illogical

parameter values that are reached because the parameter seeds are too far from the global

solution. Finding the global minimum is typically an iterative process of trying different seed

values while searching for a valid solution. Our strategy for fitting involved multiple seeding

steps. First a guess was made for a set of time constant seeds. These time constant seeds were

then used to find appropriate seed values for the amplitudes and , by performing a preliminary

fit where the amplitudes and were optimized from initial values of zero while holding the

time constants fixed. In a subsequent fit the time constants were released and all the parameters

were optimized together.

60

To give some perspective on the solution space, Table 5 shows seed values that resulted in valid

and invalid solutions when analyzing the blue data set in Fig. 19. With a one-term fit, any time

constant between 0.01–100,000 ms returned the same solution of 2.0 ms. Only when the time

constant extended beyond this range did the fits fail. C1 failed because the Origin Levenberg-

Marquardt algorithm “cannot find a direction to change parameters to reduce χ2”, while C2

failed because of a dependency between , and a slow that approximates a straight line.

Since the settling time of the clamp (2 ms) and the clamp duration (300 ms) allow decays from

~0.4–300 ms to be detected, any reasonable time constant seed could be used with a one-term fit

to arrive at the same result. The same was also true of the two and three-term fits, where a wide

range of time constant seeds produced the same solution.

When fitting with four or more terms the choice of time constant seeds became increasingly

more important. Two simple and effective strategies that we often used for choosing these seeds

revolved around modifying the time constants from the valid solution one order lower. One

strategy would insert a new time constant midway between the time constants of the lower order

valid solution. For example, the three-term solution time constants (1.0, 2.7, and 46 ms) would

lead to the following four-term seeds: (0.5, 1.0, 2.7, and 46 ms), (1.0, 1.9, 2.7, and 46 ms), (1.0,

2.7, 24, and 46 ms), and (1.0, 2.7, 46, and 92 ms). Another strategy would pick seeds that

straddled the time constants from the valid lower order solution. For example, the three-term

solution would lead to the following four-term seeds (0.5, 1.9, 24, and 92 ms). All of these seed

sets, using the insertion or straddling methods, lead to the optimal four-term solution. Other

seeds like (1, 2, 3, and 4 ms) and (0.1, 0.2, 100, and 200 ms) return failed fits with inter-

dependencies caused by several time constants converging on the same value. The five-term

solution was found in a similar was as the four-term, with all of the seed sets generated by the

insertion and straddling methods leading to the optimal solution.

When beginning to fit a dataset this seeding process is very important, as it may not be obvious

what time constants to use. However, after several iterations it is possible to narrow in on

solutions. These solutions can then be reused as seeds for neighboring test potentials or other

datasets to speed up the fitting process.

61

2.4.7 Stopping

The fitting process begins with one exponential term, and then more are added until the fits fail

and valid solutions can no longer be found. This is demonstrated in Fig. 22B which shows

complete solutions with charge movements for the −150 and 70 mV holding potentials. With

each exponential term added to the fit equation the parameter errors increase until only nonsense

fits can be found and the residuals go flat (Fig. 21); the red fails with five terms (±462%) and

the blue with six (±77%). Although there is no direct test for the correct number of terms,

nonrandomness in the residuals suggests more terms are needed while nonsense fits suggest too

many are being used. The one caveat is that nonsense fits are not an absolute indication that the

limit has been reached, since bad seeds may have been chosen. However, the more thorough the

parameter space is searched without success, the more confident one can be that the limit has

been reached.

2.4.8 Parameter Variation with the Number of Terms

The Q–τ pairs in Fig. 22B are plotted in A to illustrate how they change with the number of

exponential terms used to fit the data. Charge magnitudes are drawn as horizontal bars on a

vertical and logarithmic time constant scale. These pairs spread out in a horizontal pyramid

pattern, and like oranges stacked in a market, the time constants are staggered between

neighboring solutions with more or fewer terms. Unsurprisingly, this staggering is weighted by

the charge movements. For example, the blue one-term time constant (2 ms) is significantly

closer to 1.7 than 19 ms from the two-term solution, because of their respective charges (−14.9

and −3.1 nC). This staggered pattern was always observed, and suggests that when fewer terms

than the intrinsic number of decays are used, the solution is a form of weighted average of the

intrinsic kinetics.

Ultimately, the −150 and 70 mV holding potentials should measure the same set of time

constants. The solutions are initially divergent when a small number of exponential terms are

used, because of the large difference in carrier contribution (blue −19.6 nC, red 1.6 nC), but as

more terms are added they become closer (four-term: 0.48–0.46, 1.6–1.5, 3.5–6.1 and 24–39

ms). Although the slowest decay (blue 141 ms) in not observed directly in the red trace, the

62

slower red terms are weighted upwards (3.5–6.1 and 24–39 ms) suggesting that it may be

present but too small to separatec.

2.4.9 Looking at the Dataset as a Whole

Some amount of time is spent searching at one test potential for a valid solution with as many

terms as possible. Once it is found, the time constants often make good seed values for

neighboring test potentials, and the solution is propagated in both directions across all of them.

Sometimes new solutions will be found as the data set is being built and these are propagated as

well. In this way all the test potentials are connected and form part of a larger whole.

As solutions are obtained they are collected in a table like the one shown in Table 6, which

contains the time constants from the blue and red groups, and the charge components from the

−150 and 70 mV holding potentials. The decays are then sorted by row so that the time

constants and charges form continuous functions along each row. This process is mostly

straightforward when the signal is large and all the decays are present, but as the jumps become

smaller and terms drop off some amount of reasoning is necessary.

For example consider the blue 10 and 30 mV holding potential data, where one of the −10 mV

decays has disappeared and only four remain. Looking only at the time constants it is not

immediately clear which decay is gone. Decay 1 must remain (0.50 and 0.51 ms), but one can

imagine a scenario where any of the others is removed and the remaining three time constants

are distributed between components 2–5. Help comes from the charge movements, where

decreasing trends are expected, and are only possible if decay 4 is removed. Decay 4 has the

smallest charge movement at −10 mV (−1.3 nC), which already makes it the most likely

candidate to drop to zero, but it is also doubtful that its charge magnitude would increase

sharply to −2.1 or −3.2 nC.

c The failed red five-term fit detects a slow decay (217±17%) but the charge movement is very small and the error

is too large (0.04±0.17 nC, 462% error) to accept this fit.

63

2.5 Results

2.5.1 Transient Kinetics of wt and the T156C mutant

In §2.4 Data Analysis we demonstrated how multi-exponential fitting of SGLT1 transient

current data was performed at one test potential (−50 mV). A complete data set repeats this

process over a range of test potentials to measure the voltage dependence of each decay’s rate

( ) and charge contribution ( ). Here we will show the results of this full analysis for wt SGLT1

and the T156C mutant. These kinetics were measured in the presence and absence of saturating

concentrations of the inhibitor phloridzin to test its ability to inhibit SGLT1 decays, and an

experiment was also done on a non-injected oocyte to measure the background signal. These

results are presented in Fig. 23, which shows and data for each decay as a function of the

test-potential, measured from two direction—hyperpolarizing (red) and depolarizing (blue)

holding potentials. Five decays were found at most, and were sorted and numbered by rate (1,

0.4–0.85 ms; 2, 1–5 ms; 3, 2–12 ms; 4, 10–60 ms; 5, 60–380 ms), with their data plotted in

separate rows. The data indicates the strength of the decays, which were found in many cases

to be different depending on the direction of the voltage jump. Total charge (black = red + blue)

represents the overall charge mobility of each decay. Certainty of the measurements was

directly proportional to the corresponding of the decay, since it was harder to estimate the

decay rate when the signal was small. Because of this, it is important to consider the

corresponding data when interpreting the . The top row of Fig. 23 gives an overview of the

data sets by plotting multiple components together. Here the directional charge data is overlaid

for the carrier decays (2–5) with the total carrier charge shown in black, while the time constants

are all plotted on a logarithmic scale that shows their spread. This data is also replotted on its

own in Fig. 24 to give an expanded view of the relationships between the decays.

2.5.2 Capacitive Decay

The fastest decay (1) had a number of features that identified it as the membrane capacitance: a

voltage and direction independent time constant of ~0.5 ms; linear, and equally sloped, charge-

voltage relationships in both directions (red and blue), with a voltage independent total charge

(black); was not inhibited by phloridzin; contributed the largest proportion of charge (28±4 nC);

was the only sizeable decay in the non-injected. The time constant was the same for wt and the

T156C mutant with and without saturating phloridzin (0.47±0.07 ms), but increased slightly in

64

the non-injected (0.74±0.03 ms), possibly because of the absence of overexpressed carriers. At a

few depolarizing potentials a second capacitive decay appeared (0.59±0.03 ms), and could be

identified as such because adding the charge contribution of both capacitive decays (grey and

white filled) produced the expected linear dependence (red).

2.5.3 Carrier Decays: Charge Movements

The four remaining decays (2–5) exhibited properties that identified them as belonging to

SGLT1. They were inhibited by phloridzin (wt, 18±2 to 2.9±1.4 nC; T156C, 25±4 to 5.6±1.0

nC), absent from the non-injected (1.4±1.0 nC), and the sum of the individual charges was

voltage independent (black, 2–5 top row). Charge profiles for all four carrier decays are overlaid

in the top row for comparison, and this data is also reproduced in Fig. 24A on a larger scale. wt

was a lower expressor than the T156C mutant and, therefore, its charge movements were

displayed on a smaller scale to make comparisons easier (15 versus 30 nC).

Often, the directional charge movements (red and blue) of the carrier decays were unequal. To a

first approximation, decays 2, 3 and 5 shared similar features between wt and the T156C

mutant: 2 had a small red curve and larger blue; 3 had a moderate blue and larger red; 5 had a

large blue and smaller red. There were, however, some differences in their relative magnitudes,

with the T156C mutant having a larger 2-blue and 3-red, and smaller 5-red. By far, the key

difference caused by the T156C mutant was a significant inhibition of decay 4 charge (1.2±0.6

nC).

2.5.4 Carrier Decays: Time Constants

The four carrier decays were split into two fast (2 and 5 ms) and two slow (25 and 50–200 ms).

Their time constant curves displayed minimal voltage dependence, however, decay 5 was

sigmoidal for both (wt, 150–50 ms; T156C, 200–100 ms), and decays 2 (1.5–4 ms) and 3 (3.5–

7) increased to right for the T156C mutant. The time constant curves were noisier because of

their dependence on . When the corresponding charge was small, the time constant estimates

had a larger error. This can be seen in 3-red, where fluctuates at hyperpolarizing potentials

(−110, −90 and −70 mV) because the charge is small (0.05, 0.10 and 0.17 nC, respectively). The

time constant curves can be compared with each other more easily by looking at the log plots in

the tope row. These figures are also reproduced, for convenience, on a larger scale in Fig. 24.

65

The log plots are helpful when sorting the Q–τ pairs, as discussed in §2.4.9 Looking at the

Dataset as a Whole, by considering options for continuous segments. In theory the directional

time constant measurements should give the same result, and within experimental error they do.

2.5.5 Phloridzin and Non-Injected Controls

The smaller charge movements produced in the presence of phloridzin and by the non-injected

can be seen in greater detail in Fig. 25. This figure shows the same data as Fig. 23, but with each

sub-plot expanded to use its own scale, as indicated on the left by the scale’s maximum. The

background signal is typified by the non-injected data. The capacitive decay (1) was strong, but

the remaining four (2–5) were not always observed, the charges were very small (decay 3 and 4,

<0.3 nC), there were gaps in the data (2, 4, and 5), and sharp discontinuities (2–5). We attributed

this behavior to small background signal and/or noise. The lack of any regularities, besides the

capacitive decay, indicates that there are no major contributions from endogenous membrane

proteins.

As discussed earlier in §2.1.3 This Study, the T156C mutation significantly reduces the carrier’s

affinity for phloridzin (Fig. 13A). Even at the relatively high concentration of 2000 µM,

phloridzin inhibition is incomplete. All four T156C carrier decays exhibit some charge for

hyperpolarizing jumps (blue). This is because hyperpolarizing jumps are first preconditioned at

a 70 mV, which causes phloridzin to unbind and free a small pool of carriers. In comparison, at

200 µM phloridzin decays 3–5 of wt are similar to the non-injected background. Surprisingly,

however, decay 2 retains the same charge profile and 50% activity. This suggests that the carrier

may retain some mobility with phloridzin bound of a ~2 ms conformational change.

2.5.6 Limiting Terms and its Effect on the Measured Kinetics

Quite often transient studies of SGLT1 and other carriers fit the transient currents with a few

exponential terms (1–3), and without testing to see if a maximum was reached. In many cases,

much effort is then put into building a kinetic model by interpreting this data. The question

presents itself, to what extent does a partial analysis represent a carrier’s real kinetics? To

explore this question, we will examine how the Q–τ pairs morph as the number of exponential

terms used to fit the data is artificially reduced from five down to one. This behavior was

66

demonstrated earlier at one test potential (Fig. 22), but we will now consider the entire T156C

data set, in the absence of phloridzin (Fig. 26).

On the left of Fig. 26, the full analysis with five exponential terms is shown. When terms are

removed from the fit equation some decays are no longer detected, and, instead, are absorbed by

another one. To make sense of these changes, it helps to look at the charge data, which signifies

the relative strengths of the decays. When the first term is removed (five to four), decay 4,

which contributes the smallest charge, disappears. The remaining charge profiles adjust slightly,

while the time constants shift in the direction of the missing component (i.e. 1–3 move up, while

5 moves down). With three exponential terms, 2-red is mostly absorbed by 3-red, because their

time constants are similar (see log scale in top row), while 3-blue disappears and pulls the

remaining time constants in (1 and 2 move up, while 5 moves down). At two terms, 2-blue is

absorbed by 5, and 5-red is absorbed by 3 (see small humps added to charge profiles). Lastly,

with only one term, the red and blue curves become a changing average of the most dominant

decays. At hyperpolarizing potentials blue is an average of decays 1 and 2 (~1 ms), while red

only follows 1 (~0.6 ms), and at depolarizing potentials they both track decay 3 (~ 6.5 ms).

These results show that the measured kinetics can change considerably, depending on the

number of exponential terms used. When an insufficient number are used, the results are often

some form of a weighted average of the real decays. Because the problem of multi-exponential

fitting involves separating multiple distinct decays, their measurement is unavoidably coupled.

Consequently, an incomplete analysis may give results that do not represent the underlying

mechanism at all, and using this as the basis for building a model may have the unintended

consequence of obscuring the mechanism further.

2.5.7 Using Expected Behavior to Predict the Number of Terms

As alluded to earlier, certain behavior is expected of the membrane capacitance and carrier

kinetics. The capacitance should have linear directional charge-voltage curves, of equal slope

( ⁄ ); a voltage independent total charge (conservation of charge); voltage independent

and equal directional time-constant curves (fixed time constant). While the carrier’s total charge

(∑ – ) should be voltage independent (conservation of charge). In Fig. 26 we can see that

using less than five terms results in a capacitive decay with differently sloped directional

67

charge-voltage curves (red and blue)d, and a sloped total charge-voltage curve (black). As well,

the carrier total charge-voltage curve is sloped (black). In all cases this sloping gets

progressively stronger with fewer terms. Lastly, the blue capacitive time-constant-voltage curve

pulls away from the red. Based on these criteria it is easy to see that only five terms reproduces

all the expected behavior. By combining these measures with the residuals and the standard

errors of the fit parameters, it is possible test the completeness of the transient analysis.

2.5.8 Dependence of the Decay Charges on the Holding and Test Potentials

The initial transient analysis was limited to a handful of the largest decays produced by holding

potentials at each end of the voltage range (i.e. the red and blue traces in Fig. 19), to speed up

the exponential fitting. This provided a measure of the decay time constants, and maximum

charge mobility moving in both directions. It would, however, also be useful to see how the

charge mobility is affected by the holding potential, which can be found by fitting the middle

decays (grey traces in Fig. 19). Since the time constants had already been solved for the larger

decays on either end, it was easy and fast enough to solve for the decay amplitudes of the

middle holding potentials by performing a follow up fit on the entire data set (red, blue, and

grey traces), with the time constants fixed to these solved values.

This amplitude data, plotted as charge, is shown Fig. 27 for the T156C mutant. Each decay

contains a family of curves (various colours) that show the charge mobility as a function of the

holding potential for each test potential. The carrier decays (2–5) had a sigmoidal dependence,

while the capacitive (1) was linear, as would be expected in both cases. The sigmoidal curves

were fit with a Boltzmann function to obtain the average valences ( ) and ’s shown. The

sum of the individual carrier decays (2–5) had Boltzmann parameters that were almost identical

to a conventional Q–V distribution calculated by integrating the transients ( =28±8 versus

32; =0.93±0.14 versus 0.94; see Fig. 13B), while decays 2, 3 and 5 shared similar properties

( =27–36 mV, =0.84–1.2). The one exception was decay 4 which split into two pools

d Irregularity in the 1-blue charge curve at hyperpolarizing potentials is likely caused by crossover between decays

1 and 2. Their time constants converge here (0.56±0.01 and 1.1±0.2 ms n=5), and there is a strong correlation

between the upward hump in the 2-blue curve and downward hump in the 1-blue.

68

of 5±1 and 61±14 mV. This aberration may provide an explanation for the inhibition of decay 4

in the T156C mutant. It shows that when jumping to negative potentials decay 4 requires a

larger hyperpolarization (5±1 mV) than the others (27–36 mV) to fully activate, and this also

occurs when jumping to depolarizing potentials (61±14 mV versus 27–36 mV). It appears that

the conformational change associated with decay 4 is restricted from moving in both directionse.

The charge curves for the capacitive decay (1) were fit with a straight line, and a membrane

capacitance of 132±19 nF was calculated from the slopes, which is similar to other reported

measurements for oocytes (185 nF200, and 175–250 nF83).

2.5.9 Supplementary Data

Additional transient kinetic data sets are included in Fig. 28 for wt and the T156C mutant. As

can be seen, the characteristics of the charge and time constant curves are identical between

oocytes.

2.6 Discussion

2.6.1 Ordering the Transitions

The presence of four carrier decays indicates four transitions, likely related to conformational

changes of the carrier. The question is how are these transitions arranged? As discussed earlier

in §2.3.4 How the System Affects the Transient, the unequal directional charges seen with the

carrier decays (i.e. the blue and red charge data in Fig. 23A) can be caused by what we call a

masking effect that occurs when a slow transition is in front of a faster one. By comparing decay

charge movements and rates for jumps from different directions, we can figure out the order of

the transitions.

The process of arranging the transitions to arrive at a model of SGLT1 transient kinetics is

illustrated in Fig. 29, using the wt data. Different arrangements are considered in turn, beginning

e Note that a conformational change can only be associated with a particular decay when it is much slower or faster

than the conformational changes that are connected with it in sequence. A good example that satisfies this criterion

is the rapid equilibrium condition. When rapid equilibrium occurs, isolated fast conformational changes can be

considered to equilibrate first followed by isolated slow conformational changes. The decay time constants of

isolated slow conformational changes are mostly determined by the rate constants of each individual slow

conformational change, with some modulation by faster adjacent conformational changes.

69

with the slowest decays. In A, the decay 4 (25 ms) and 5 (50–150 ms) charge profiles are placed

beside each other, with 5 on the left. With this arrangement when moving from left to right

decay 4 should be masked by 5, which is much slower. Instead, 4-red is large and invalidates

this arrangement. Reversing the transitions in B does make sense, because 4-blue is now masked

by transition 5 when moving from right to left, and 5-red is not masked by 4. If we try and place

the two faster decays 2 (2 ms) and 3 (5 ms) in the middle as they are in C, they would be

masked in both directions by the slower transitions 4 and 5 on either end. The outside

arrangement in D is inconsistent because 2-blue and 5-red are large, but should be masked by 4

and 5 in the middle. Swapping sides makes the most sense, resulting in the final arrangement in

E.

2.6.2 Carrier Conformations

Because the transient kinetics were measured in the absence of glucose they correspond with a

reduced model that includes conformational changes of the unloaded carrier and extracellular

Na+ binding (see Fig. 1). However, because Na

+ binding/debinding is too fast to see with the

two-electrode voltage-clamp (0.004176 to 0.06177ms), all four decays are likely related to carrier

movements as it reorients between inside and outside facing conformations. As discussed

earlier, the classical model has been extended over time by inserting additional empty carrier

conformations as new decays have been revealed in the transient currents. The latest model

accommodates three conformational changes, but our transient analysis reveals four.

Recent success in crystalizing membrane cotransporters, and in particular the LeuT architecture

shared by SGLT1, has led to the proposition of a gated rocker-switch transport model, as

discussed in §1.3.11 Transport Model and illustrated in Fig. 12, that involves four

conformational changes of the carrier. When this crystal model is aligned with our transient

model we arrive at the assignments shown in Fig. 30, where each decay is associated one-to-one

with the movement of a gate or pore. What is striking about this mapping is that the fast decays

(2 and 5 ms) are generated by opening/closing of the intra and extracellular gates, while the

slower decays (25 and 50–150 ms) are produced by opening/closing of the intra and

extracellular pores. It makes sense that the gate movements, which only involve the rotation of a

few amino acids, would be faster than the pores, where several transmembrane segments must

70

collapse. We now have a physical mapping for each of the decays that allows us to discuss the

relative rates of these conformational changes.

If we now take the gated rocker-switch model for representing the carrier conformational

changes and merge it with the classical transport model, we end up with the model in Fig. 31.

This model includes all the steps known to date involved in SGLT1 transport.

2.6.3 Functional Insights

If we look at how the T156C mutant fits within the context of the model presented in Fig. 30,

we see that decay 4, which is inhibited by this mutation, is associated with opening/closing of

the extracellular pore. We also know that this transition’s charge mobility is restricted, with

respect to the other transitions, by the presence of the two pools (5 and 61 mV, see Fig. 27).

Together, these observations suggest that movement of the extracellular pore is restricted by the

T156C mutation. How does this understanding align with other properties of the mutant, such as

difficulty orienting towards an inside-facing conformation, decreased phloridzin affinity and

phloridzin dissociation (see Fig. 13)? Difficulty orienting towards an inside-facing conformation

can be explained directly by the restricted extracellular pore. If we consider that phloridzin’s

high affinity for the carrier and tight-binding might be the result of the extracellular pore closing

and trapping it in the binding site, then the reduced affinity and dissociation seen with the

T156C mutant are easily explained by an extracellular pore that cannot close properly. This

theory is supported by reports that phloridzin is transported by SGLT1170, and a chimera of

SGLT1 and SGLT2171, and suggests a mechanism of inhibition that involves slow transport,

much like indican175

, as opposed to a static locked conformation as commonly beleived201

. With

the K157C mutant, glucose and phloridzin are unable to bind to the carrier, which could be

caused by an extracellular pore that is unable to open. It is our hypothesis that the lysine at

position 157 is involved in a salt bridge interaction that facilitates movement of the extracellular

pore.

The wt carrier exhibits 50% activity of decay 2 in the presence of saturating phloridzin (see Fig.

25). This seems to indicate that the intracellular gate retains some mobility when phloridzin is

bound. Perhaps the intracellular gate is able to move independently of the other conformational

changes. However, it is difficult to imagine what sort of mobility is available, since there is little

free space when the intracellular pore is collapsed (see Fig. 6). An alternative explanation is that

71

if phloridzin is transported, it is possible for the carrier to progress towards an inside facing

conformation with the intracellular pore open, at which point the intracellular gate would be

mobile.

2.6.4 Building a Kinetic Model

The next logical step would be to build a kinetic model that could be used to simulate the

transient currents and zero in on values for the transition rate constants. Conceptually this

should be rather straightforward, by working with a five-state model and assigning rate

constants to the transitions based on the decay time constants, as shown in Fig. 32. There is a

problem, however, with assigning voltage dependence and charge to the transitions. If any of the

transitions that are observed through the transient decays are voltage dependent, in theory, their

decay time constant should asymptotically approach zero as the absolute membrane potential

increases, as illustrated in Fig. 33A. However, all of the carrier time constants are relatively

voltage independent, with none varying by more than a factor of 3, let alone approaching zero

(see Fig. 24). The transient time constants, instead, resemble the situation in Fig. 33B that

occurs when a voltage independent transition is bordered by much faster voltage dependent

transitions. In this case the time constant varies between two finite rates. It is my hypothesis that

the most appropriate model of SGLT1 will look something like Fig. 32, but with one or more

rapid voltage dependent transitions inserted somewhere. These voltage dependent transitions

would likely be faster than the time resolution of the two-electrode voltage-clamp (<0.4 ms), to

not be detected, and their role would be to push/pull the system from side-to-side by harnessing

the electrochemical gradient. The charge movements could then be generated by either these

rapid transitions, or by the slower voltage independent transitions as the gates/pores close and

isolate charge within the membrane electric field. The problem of building this kinetic model

will involve considering various combinations for the placement of these rapid voltage-

dependent transitions, and the assignment of charge production. Although this would be an

excellent problem to tackle, it is unfortunately beyond the scope of this thesis.

2.7 Summary and Conclusion

The goal of this project was to figure out how to extract as much kinetic inform as possible from

SGLT1 transient currents in order to better understand the transport mechanism. We found that

this involved learning how to fit transient currents with a large number of exponential decays,

72

and evaluate the quality of these fits using a number of different measures, to find the right

number of terms. This analysis is laborious and time consuming, but the end result is kinetic

parameters that map one-to-one with conformational changes of the protein, insight that is not

available by any other means. In the process we discovered that a transient analysis with a

limited number of terms does not measure a subset of the decays, but instead provides a form of

weighted average that may not be indicative of the underlying mechanism—highlighting the

importance of using measures to test for the correct number of decays.

Although an emphasis is often placed on measuring the decay time constants, we found that they

contained minimal features and were, for the most part, voltage independent, while the

accompanying charges contained important information about the arrangement of the

transitions. The magnitude of each decay’s charge contribution is highly dependent on the

direction of the voltage jump, an observation that reinforces the need to use multiple holding

potentials to detect all the decays. This multi-holding strategy also provides a direct and easy

way to use these unequal directional charges to build a model of the transporter transitions, by

taking advantage of the masking effect that occurs when a slow transition hides a faster one.

Ultimately we end up with a linear five-state model with two slow transitions in the middle (25

and 50–150 ms) flanked by two faster ones (2 and 5 ms) on either end.

This kinetic data becomes significantly more relevant when combined with structural insight

into the gated rocker-switch mechanism of SGLT1 revealed by crystallographic studies of the

LeuT architecture superfamily. The gated rocker-switch mechanism predicts four

conformational changes, and the kinetic studies have confirmed this in an elegant way by

matching each gate and pore conformational change to a carrier decay. This reveals that the

gates move on the order of 2–5 ms while the pores are an order of magnitude slower at 25–150

ms. By looking within the crystal structure at the gate and pore structures in more detail, we can

form theories about how these mechanisms work on the atomic level by matching them with

their known rates. Going one step further, we can use strategic mutants to test these theories by

mapping a mutations effect back on to the structure by monitoring changes to the four carrier

decays. There is remarkable synergy between transient and crystallographic studies. Crystal

structures provide unparalleled structural information, while the transient kinetics provide

unparalleled dynamic information. Now, more than ever, it is important to reinvest in kinetic

studies to take advantage of the structures we now have and have waited so long for. This

73

analysis is general enough that it can be applied to any electrogenic cotransporters, and,

therefore, presents an opportunity to expose other mechanisms and possibly classify

cotransporter mechanisms kinetically.

2.8 Future Work

There are a quite a few different paths worth exploring with the transient currents kinetics that

are discussed in more detail below. These include simulations; verifying the connection between

the four predicted conformational changes and the four carrier decay components using strategic

mutants; expanding the analysis to involve substrate binding steps and explore the question of

binding order; study the kinetics of other carriers with the gated rocker-switch mechanism, as

well as other mechanisms.

2.8.1 Simulations

As discussed earlier in §2.6.4 Building a Kinetic Model, the path to building a state-model

simulation of SGLT1 has been laid out with a five–state model provided by the transient

kinetics. What remains is figuring out where to place the voltage dependent steps, and the

assignment of charge displacement, which should be possible by trying a number of different

configurations. This would be an important test of the transient model (fast-slow-slow-fast), as it

would demonstrate the capability of this model to produce the observed transient kinetics. It

would also provide a way to tune the rate constants through fitting and find more precise values.

It would be helpful to demonstrate the masking effect directly with a simple state model. A good

place to start would be a three-state model with one slow and one fast transition. However, there

are complications with the placement of voltage dependence. Adding voltage dependence

directly to either or both transitions makes their rate and time constants variable, and, therefore,

difficult to enforce separate transition rates. A better setup might be voltage independent fast

and slow transitions connected in the middle by a rapid voltage dependent transition that can

move the system from side to side. Once the model is built, large differences in rates could be

used to establish the masking effect and then these rates can be brought closer together to find a

threshold where the effect dissipates. It would also be interesting to compare the threshold for

the masking effect with the one for the rapid equilibrium assumption.

74

2.8.2 Strategic Mutants

Prior to the crystal structure of vSGLT quite a few SGLT1 mutations were characterized in an

effort to elucidate structure and function. Amongst these are several that cause uniquely

debilitating effects, which seem to indicate a malfunction of one of the carrier’s conformational

changes. Characterizing the transient kinetics of these mutants will help identify the affected

transition, much like the T156C mutant, and reveal the amino acid’s functional significance

within the tertiary structure. Carrying out this strategy over a series of mutants has the potential

to map each step of the gated rocker-switch mechanism to key residues of the architecture. Once

these associations are confirmed, the functional significance of new mutations can be

understood in a comprehensive way by looking at their effect on the four carrier conformational

changes.

Despite not being able to bind glucose or phloridzin the K157C mutant does generate transient

currents. Studying their kinetics should corroborate the findings of T156C and reveal the greater

extent of this mutation’s malfunction. While T156C retains minimal mobility of the transition

associated with component 4 (30 ms), it is our hunch that the K157C mutation will abolish it.

Component 4 is putatively related to movements of the extracellular pore, and an important

question is what other conformational changes can and cannot take place when this is

completely inhibited. This leads to ideas of coupling between the conformational changes and

how dependent they are on each other.

The Q457R mutation has been found to be the cause of glucose galactose malabsorption in some

patients. This mutant can bind glucose but is unable to transport it and transiently unbinds it,

suggesting an inability to orient to an inside facing conformation18

. It is likely that the

intracellular pore (decay 5, 150–50 ms) is unable to open, although it is also possible that the

extracellular pore cannot close to initiate transport. The vSGLT crystal shows that this residue

forms part of the galactose binding site with the amide group interacting directly with two

substrate oxygens68

. Furthermore this residue is responsible for turning hSGLT3 into a glucose

sensor67

. An earlier study of ours with three decays and 150 ms pulses found that this mutation

decreases the slowest component from 45 to 15 ms, seeming to indicate indirectly that decay 5

(150 ms), which we were not monitoring then, had disappeared190

. A full transient analysis

should confirm this and implicate the intracellular pore.

75

No mutations have been linked to the gating mechanisms of SGLT1 (except for an identified

F453L mutation in several GGM patients45

), yet it would be valuable to confirm that

components 2 (2 ms) and 3 (5 ms) monitor these processes. This should be possible by studying

conservative mutations of the intracellular (Y290) and extracellular gates (L87, F101, and

F453), such as swapping phenylalanine and tyrosine or leucine and isoleucine. We suggest

conservative mutations because we are only looking for a positive indicator, such as a change in

rate, that the associated component is affected. A more drastic mutation may cripple the carrier

in a way that makes it difficult to study.

A disulfide bridge has been identified between C255 and C51119

, which when disrupted by

mutating either or both residues (C255A, C511A, C255A/C511A), leads to transients in the

presence of glucose77

. This suggests difficulty orienting to an inside facing conformation similar

to Q457R. Characterizing any of these mutants and identifying the affected conformational

change can help with understanding the structural role of this disulfide bridge, and the roles

sulfur bridges play in conformational changes.

Mutations of residue D204 have a large impact on ion coupling and affinity, but most

significantly D204N turns the carrier into a glucose gated H+ channel with a H

+:glucose

stoichiometry as large as 145:1202

. It would be interesting to see what conformational changes

this residue affects, since the D204E mutation has already been shown to alter the transient

kinetics. Understanding how SGLT1 might be altered to take on channel-like properties can give

insight into possible conformations that would provide a channel-like pathway. D204N and

D204C are expressed poorly making D204E the best mutant to study.

A stretch of TM4 and its extracellular tip (143–180) has been studied by our group with cysteine

scanning mutagenesis17

, and a key series of residues (F163, A166, Q170, and K173) were found

to line one side of an alpha helix forming a surface of an extracellular pore that is thought to be

involved in Na+ interaction

15,188,189,203. Each of these mutations shifts the Q–V distribution (10–

20 mV) and double and triple mutants have an additive shifting effect (45–90 mV). To

understand the functional significance of this region we would investigate the kinetics of these

key mutants beginning with A166C because of its ~4 times higher αMG affinity, followed by

Q170C, K173C, and F163C.

76

2.8.3 Substrate Transients

A natural extension of the kinetic studies presented here would be to study how altering the

extracellular Na+ concentration affects each decay component. My hypothesis is that lowering

the extracellular Na+ concentration will decrease the rate of the extracellular gate (decay 3),

without significantly altering the other decays. This might give insight into how Na+ binds, and

if it is involved in a gating mechanism that exposes the binding site to glucose, since Na+ is

required to allow glucose to bind. Studying transient currents with the cut-open oocyte and zero

trans Na+ would also verify that there are four conformational changes in the complete absence

of Na+. The cut-open oocyte might also provide an opportunity to measure Na

+ binding directly

because of the faster settling time of the clamp.

Because substrates like glucose substantially diminish the transient signal, by promoting an

inside facing conformation where the carrier is electrically silent77,175

, most transient studies of

cotransporters are done in the absence of the non-ionic substrate. Some mutants, however, resist

an inside facing conformation, and this provides an opportunity to study substrate transients and

the steps immediately before and after binding. The Q457R mutant is perhaps the best example,

since transport is abolished but substrate binding is intact. However, the T156C mutant is

similar in its ability to transiently unbind phloridzin with depolarizing pulses. A strategy

designed around one or both of these mutants could identify new carrier conformations related

to substrate binding, and would also provide a great opportunity to study the elusive binding

order question, which has been debated since the dawn of SGLT1 kinetic studies. The

dependence of the transient parameters on glucose, phloridzin, and Na+ should provide enough

kinetic detail, if not the most available so far, for discriminating between the three competing

models (N/N/S, N/S/N, and N/random; N=Na+, S=glucose/phloridzin). The disulfide bridge

mutants (C255A, C511A, and C255A/C511A) are also potential candidates that generate

glucose transients, but in their case retain transport function. Lastly, the wt carrier generates

transients with some poorly transported substrates like indican175

.

2.8.4 Other Carriers and Mechanisms

Given that SGLT1 belongs to a larger structural superfamily that shares the LeuT architecture, it

would be valuable to confirm that other members participate in the same conformational

changes of the proposed gated rocker-switch mechanism. This would validate the findings from

77

SGLT1 and provide contrasting variations to help understand this class of transporters. A

number of other solute:sodium symporter members that have been vetted in the oocyte system

and are suited for this include SMIT1107

SMIT2105

, SMCT1116

, and NIS133

.

Perhaps more interesting though are members of the neurotransmitter:sodium symporter family,

such as GAT1178,204

and SERT205

, which are expressed well in oocytes and generate a slow ~150

ms decay similar to component 5 of SGT1, while also appearing to have faster decays that have

not yet been measured. These carriers are close relatives of the crystalized LeuT and have a high

biological interest because of their role in synaptic signaling and psychological disorders and

treatment. If the gated rocker-switch conformations can be identified in these carriers as well,

this will demonstrate a strategy for characterizing transport mechanisms kinetically.

This approach could also be extended to other carrier families to see if they share a similar

mechanism or adopt a unique one, with a number of candidates listed in 3.1.1 Historical

Perspective.

Of most value may be a member of the major facilitator superfamily. These carriers are

expected to operate with a rocker-switch mechanism, and so it is our prediction that they will

lack the faster gate transitions (2 and 5 ms). Unfortunately, because of difficulties being targeted

to the plasma membrane, these carriers have mostly evaded electrogenic study. However, a

recent study of purified LacY reconstituted in proteoliposomes with a solid-supported

membrane has been successful206

. Using concentration jumps this carrier generates decays on

the order of ~500 ms. Exponential analysis was not performed but from our observation there

appear to be two decays of ~50 and ~200 ms. There has also been success with inducing plasma

membrane expression of HMIT in oocytes with strategic mutations106,207

, and this perhaps

represents the best opportunity for a MFS carrier. This superfamily is of interest not only

because it is one of the largest, but because it contains the GLUT family of hexose facilitators,

and has a large number of crystalized members.

Lastly, there are the excitatory amino acid transporters (EAAT), which have been characterized

in oocytes182,208,209

and studied transiently193

. These carriers belong to the DAACS family,

which has a unique architecture identified by the Gltph crystal.

78

2.8.5 Miscellaneous

The T156C mutant appears to have reduced mobility of the extracellular pore, which makes it a

good candidate for crystal studies, similar to how LacY was originally crystalized with an

immobilizing mutation24

. However, it would be difficult to crystalize T156C (or K157C)

directly as mostly bacterial transporters have been crystalized so far. One option would be to try

and find the analogous amino acids in vSGLT, but there are no obvious candidates.

A compliment to the electrophysiology studies would be an analysis of fluorescent transients.

The big question would be to see if the same four transitions were present or if different

conformational changes were exposed. Also, by placing fluorescent tags on key areas associated

with each conformational change, each mechanism could be explored in more detail.

The multi-holding protocol is rather stressful on the oocytes, with many unable to endure it. To

decrease this stress it would be helpful to validate a reduced protocol. This could be achieved by

reducing the number of holding potentials, potentially to only two with one jump from each

side. Depending on how many holding potentials were used, one might lose the ability to

measure Q–V distributions as a function of the holding and test potential (Fig. 27), but this cost

might be acceptable if viable oocytes were easier to come by.

79

2.9 Figures

Fig. 13: Characteristics of the T156C mutant. (A) Phloridzin apparent affinity calculated from inhibition

of the Q–V distribution; titration to 200 μM for wt and 2000 μM for the T156C mutant. The wt apparent

affinity is voltage independent (1.6±0.6 μM, n=10), compared with a highly voltage dependent and

significantly reduced affinity for the T156C mutant (30±3 μM, n=9 plateau). The y-axis is linear from 0–

5 and 5–170 μM. (B) Normalized Q–V distributions (wt n=17, T156 n=16), and fits with a Boltzmann

function ( (

( ))⁄ ; wt , ; T156C , ). (C) Effect

of phloridzin on the transient produced by a −50 to 70 mV jump. Phloridzin progressively reduces the wt

transient without affecting the kinetics, characteristic of a tight binding inhibitor. In contrast, phloridzin

unbinds from the T156C mutant and produces a slow transient decay. Intermediate phloridzin

concentrations (5–100 μM) enhance this unbinding decay, but eventually larger concentrations (>500

μM) are able to overcome this effect and inhibit the carrier.

80

Fig. 14: Position of the T156 and K157 residues of SGLT1 in the vSGLT structure, where they

are represented by Val-141 and Asn-142. (A) looking at the collapsed extracellular pore. (B)

zoomed view from the same angle with extraneous TM’s removed. T156 and K157 appear to be

at the midpoint of TM 3, near the binding site of Na+ and glucose and forming an equidistant

triangle with both substrates.

81

Fig. 15: Anatomy of a voltage jump. Three potentials known as the resting ( or ),

holding ( or ), and test potentials ( or ) define a voltage jump. (jump 1) If the

resting and holding potentials are different, an initial jump is needed to allow the system to

equilibrate at the holding potential before jumping to the test potential. (jump 2) The properties

of the transient signal are determined by the jump from the holding potential to the test

potential. (jump 3) After a recording the oocyte is returned to the resting potential.

82

Fig. 16: Relationship between voltage jumps and transient kinetics. (A) Two jumps with the

same test potential (30 mV) but different holding potentials (−150 and 70 mV) decay at the

same rate. (B) Reversing the jumps in A results in different kinetics because of the different test

potentials (red and blue charge movements are approximately equal). (C and D) Different

holding potentials can result in large variations in charge movement. Charge movements are

dependent on the voltage jump crossing the active region of the transporter (30 mV in this

example). Traces have been baseline adjusted, and absolute current values are plotted. Data is

from the T156C mutant.

83

Fig. 17: Hypothetical three-state system illustrating the masking effect. When a voltage jump

occurs and a slow transition (2↔3) precedes a faster one (1↔2), the fast transition is hidden. In

this example, when a jump starts on the right and ends on the left (depolarizing holding potential

to hyperpolarizing test potential), only the slow transition is observed (blue Q– curve

above). In the other direction, both are seen because the fast transition is unable to mask the

slow. This logic works backwards as well, since the transitions can be ordered using information

about the decay rates and their unequal directional charge movements.

84

Fig. 18: Single and multi-holding voltage clamp protocols. (A) The voltage waveform of a single-

holding protocol with a −50 mV holding/resting potential shared by all twelve test potentials. (B)

Transient currents produced by A. (C) Multi-holding protocol for a −10 mV test potential with twelve

holding potentials. A phase has been inserted (0–300 ms) for transitioning between the resting and

holding potentials. Protocols for alternate test potentials are indicated by the grey waveforms. (D)

Transient currents produced by C. Orange traces show the data that would be analyzed with each type of

protocol to measure the kinetics at −10 mV. Data is from the T156C mutant.

85

Fig. 19: Example multi-holding data set. (A) Transient decays shown at full scale for jumps

from 23 holding potentials (−150…Δ10…70 mV) to a −50 mV test potential. The two shaded

regions are expanded in B and C to give a clearer view of the fast large-amplitude (B), and slow

low-amplitude (C) decays. The five furthest holding potentials on either end are colour coded

and labeled in B. Traces in C are smoothed using 100 nearest neighbor averaging (1 ms

window). Data is from the T156C mutant.

86

Fig

. 20:

Mult

i-ex

ponen

tial

fit

of

a tr

ansi

ent

dat

a se

t. R

ed (

−15

0,

−14

0,

−13

0,

−12

0 a

nd −

110

mV

) an

d b

lue

(70,

60,

50,

40 a

nd

30

mV

) tr

aces

wer

e fi

t in

sep

arat

e g

roups

wit

h a

ser

ies

of

mult

i-ex

po

nen

tial

fun

ctio

ns

conta

inin

g 1

–5 t

erm

s, a

nd

a s

har

ed s

et o

f ti

me

const

ants

wit

hin

eac

h g

roup

; cu

rve

fits

are

over

laid

in b

lack

. (A

) V

iew

of

the

2–20 m

s, a

nd (

B)

2–

300 m

s ti

me

win

dow

s. (

C)

Solv

ed

tim

e co

nst

ants

%er

ror,

gre

y)

are

colo

r co

ded

. B

asel

ine

is i

ndic

ated

by t

he

gre

y t

race

(−

50 m

V h

old

ing p

ote

nti

al).

87

Fig

. 21:

Fit

res

idual

s. R

esid

ual

s ar

e sh

ow

n f

or

the

−150 a

nd 7

0 m

V h

old

ing p

ote

nti

als

wit

h 1

–6 t

erm

s. (

A a

nd B

) R

esid

ual

s dec

reas

e

as t

erm

s ar

e ad

ded

(χ2

indic

ated

), u

nti

l o

nly

nonse

nse

fit

s ar

e re

turn

ed (

fad

ed r

ed f

ive-

term

and b

lue

six

-ter

m).

Plo

ts a

re s

had

ed t

o

emphas

ize

dev

iati

ons

fro

m z

ero. D

ata

has

bee

n s

mooth

ed w

ith 1

00 n

eare

st n

eighbor

aver

agin

g (

1 m

s w

indow

).

88

Table 4: Nonsense fit examples. A variety of fit results are shown for the different nonsense cases that came up.

The fit parameters ± %error are shown along with the best χ2 that was achieved for each number of exponential

terms. The voltage jump is shown above each solution. (A) %error > 50% (beige shading). (B) χ2 greater than a fit

with fewer terms (green). (C) Low Q (grey) (<0.05 nC). (D) Low τ (yellow), and high Q (red). (E) High τ (orange).

(F) Some Q with opposite sign (blue). (G) Two τ converging (pink).

89

Table 5: Seeding examples for multi-exponential fitting of the hyperpolarizing jump dataset (Fig. 19,

blue). Final solutions are shown on the left for different numbers of exponential terms used in the fit

equation. To the right are parameter seeds that led successfully to the final solution, as well as seeds that

failed to arrive at a valid solution. Some failed fits are shown in two steps, with the initial seed followed

by the failed solution (grey shading). #I, number of iterations to final solution. Percent standard errors of

the fit parameters are shown in grey. Charge data is from the 70 mV holding potential trace.

90

Fig. 22: Transient charge movements by component. Charge/time-constant pairs from the −150

and 70 mV holding potential fits in Fig. 20 are displayed. (A) Charge magnitudes are shown as

horizontal bars positioned along a vertical and logarithmic time-constant scale. (B) The

corresponding data points are displayed in the table below. The failed 5 and 6-term fits are faded

and the invalidating parameter errors are shaded beige.

91

Table 6: Example voltage dependent map of a transient kinetics exponential fit analysis.

92

Fig. 23: Transient kinetics of SGLT1. Multi-exponential analysis of wt ±phloridzin (200 μM), T156C ±phloridzin (2000 μM), and non-injected oocyte transient currents.

Five decay at most were found (1–5). The charge (A) and time constant data (B) are shown for each decay as a function of the test potential and direction of the voltage jump

(depolarizing, red; hyperpolarizing, blue; total charge, black = red + blue). Decays are numbered by the rate of their decay (1, 0.4–0.85 ms; 2, 1–5 ms; 3, 2–12 ms; 4, 10–60 ms;

5, 60–380 ms).Decay 1 is produced by the membrane capacitance and decays 2–5 by the carrier. Top row: carrier charges are overlaid (B, 2–5), with the total in black indicating

the size of the free carrier pool; time constants are plotted on a logarithmic scale (A, 1–5). wt and non-injected experiments used a shorter voltage range (−130 to 50 mV) than

T156C (−150 to 70 mV) to reduce oocyte stress (see §2.2.4); also, their decay 2–5 charges are plotted on a smaller scale (15 nC) than T156C (30 nC) because of lower protein

expression. Two capacitive decays (grey and white filled) were observed at some depolarizing potentials for non-injected. Representative data sets are shown. The T156C data is

an average of three measurements in the same oocyte without phloridzin and two with; one measurement for wt ±phloridzin and control.

93

Fig. 24: An expanded overview of the transient kinetic data. Charge and time constant data from

the top row of Fig. 23 is enlarged here to give a clearer picture of the relationship between the

different components.

94

Fig. 25: Close up of the transient kinetic data given in Fig. 23. Each plot is expanded with its own ordinate scale

ranging between zero and the maximum indicated on the left.

95

Fig. 26: Changes in measured kinetics when fitting with limited exponential terms. The T156C mutant data set in the

absence of phloridzin, shown in Fig. 23, was analyzed with limited numbers of exponential terms (5–1, separate

columns decreasing to the right). As exponential terms are removed from the fit equation some decays are no longer

distinctly observed, and are instead absorbed by the others.

96

Fig. 27: Component charge dependence on the holding and test potential for the T156C mutant. In a second round

of fits the charge contributions of the intermediate holding potentials were found, for each test potential, as

described in the text. These dependencies were sigmoidal for the carrier (2–5) and linear for the capacitive (1), as

expected. The sigmoidal curves were fit with a two-state Boltzmann equation, fits shown, to obtain valence ( ) and

data for each decay as a function of the test potential, averages indicated on the figure; curves with insufficient

detail for fitting are faded. The component 4 curves separated into two pools with distinct ’s. The capacitive

data was fit with a line, fits shown, to find the membrane capacitance from the slope, average indicated.

97

Fig. 28: Additional wt and T156C transient kinetic data sets, in the absence of phloridzin. In Fig.

23–Fig. 25 data from a single oocyte is shown and reproduced here in column 1, while

additional data sets from the same or different oocytes are shown in columns 2 and 3. Charge

scales are indicated in each subplot.

98

Fig. 29: Arranging decay charge profiles using the masking effect. (A) The two slowest decays (4 and 5) are

placed side by side but there is an inconsistency with the large decay 4-red charge, which is expected to be

masked by 5 when moving from left to right. (B) The alternate arrangement is consistent and accepted. (C)

The faster decays (2 and 3) cannot be placed in the middle, in any arrangement, because the slow decays (4

and 5) on either end will mask them when moving in both directions. (D) 2-blue and 3-red generate more

charge in the same direction that they would be masked by the slow transitions in the middle (4 and 5). (E)

This arrangement makes the most sense as it is consistent with all the charge profiles.

99

Fig. 30: Assigning transient decays to conformational changes of the carrier predicted by the crystal model.

Conformational changes predicted by crystal structures sharing the SGLT1 architecture, as shown in Fig. 12, are

aligned with the decay charge profiles arranged in Fig. 29. Most of these states were captured in the presence of

substrate, but are expected to also occur when the empty carrier reorients between sides of the membrane. The two

fast decays on either end (3 and 2) align with the outside and inside gates, and the inner slow decays (4 and 5) with

movements of the two pores. The T156C data shows that decay 4, which is inhibited by the mutation, is associated

with opening/closing of the extracellular pore.

Fig. 31: Revised SGLT1 transport model. This model incorporates the four carrier conformational changes

predicted by the crystal and transient studies (Fig. 30), for reorientation of the carrier across the membrane, with the

classical model of ordered substrate binding (Fig. 1). Closed gates and pores are represented by black and maroon

bars, respectively.

100

Fig. 32: Rough state-model of SGLT1 transport. The rate constants were assigned based on the

wt time constants for each transition. The model, however, needs to consider voltage

dependence and charge contributions before it can be used to simulate the transient currents.

Fig. 33: Types of time constant voltage dependent behavior. (A) when an isolated transition is

voltage dependent (red), its time constant will asymptotically accelerate to zero at large absolute

potentials. (B) if a voltage independent transition is straddled by two rapid voltage dependent

transitions (where the rapid equilibrium assumption holds), its time constant will vary between two

finite rates.

101

A Practical Method for Characterizing the Voltage 3

and Substrate Dependence of Membrane Transporter Steady-State Currents

3.1 Introduction

After the transient current subsides a transporter will continue to cycle in the presence of

transported substrate as it is acted upon by electrochemical forces. This cycling velocity ( ; is

used so that can be reserved for voltage) can be measured electrophysiologically by observing

the steady-state current generated as a net quantity of charged substrate is translocated across the

membrane. This cycling velocity is analogous to the turnover frequency (both with units of

cycles/second), and this rate is related to the membrane current by a scalar involving the number

of expressed carriers ( ), the net charge translocated per cycle ( ), and the elementary charge

( ),

Eq. 2

As stated above, the cycling velocity is driven by the electrochemical potential which is affected

by the concentration of transported substrates and the strength of the membrane potential.

Measuring the steady-state current while varying the membrane potential produces the familiar

I–V curves (Fig. 34A), which often have a sigmoidal voltage dependence that varies between a

maximum velocity at saturating voltage and zero. Titrating a transported substrate scales and

alters the shape of the I–V curve. Typical I–V studies are concerned with measuring the

standard Michaelis-Menten parameters , , and the Hill coefficient by analyzing, at

each voltage, the saturation curve as a function of substrate concentration (Fig. 34B). These

parameters are then often plotted as a function voltage to observe their voltage dependence.

The sigmoidal voltage dependence of the steady-state velocity curves is seldom analyzed,

because no methods of a general nature are available to do so, and the problem is generally

believed to be complex. There is, however, important kinetic information contained in the shape

of these curves that is wasted, especially considering that this data is commonly and easily

collected for substrate apparent affinity measurements. The goal of this project was to extract as

102

much kinetic information as possible from the steady-state velocity by modeling its voltage and

substrate dependence, and in so doing increase our understanding of how cotransporters work. A

general model will be presented that turns out to be rather simple, yet has broad applications to

the field of membrane transport. Using the concept of a general -state cyclical system, a

mathematical representation of the steady-state velocity will be derived, and its voltage and

substrate dependence described. It turns out that the steady-state velocity consists of a series of

terms that can be written down by following a simple recursive pattern. Furthermore, these

terms have units of time with the slowest ones of each type of dependence (voltage, substrate,

voltage and substrate, and none) dominate the expression under different conditions. This allows

the steady-state velocity equation to be reduced to a few terms that extend the Michaelis-Menten

equation to include voltage dependence. This reduced equation shows how the I–V curves can

be fit with the familiar Boltzmann equation, and this allows for a simple measurement of

valences ( ) and ’s, which are typically only obtainable from the transient currents.

3.1.1 Historical Perspective

A number of attempts at describing the properties of the I–V curves have been made, but they

have either been too specific, too complicated, or both, to be widely adopted. Some of the

earliest treatments were in the 70’s. Geck and Heinz derived a general n-state model210

, while

Stein used standard 3 and 4-state representations211

. In both derivations, they considered several

cases for assignment of the voltage dependent transitions, and described trends for some of the

parameters (linear, nonlinear, increasing, decreasing, etc.), with Stein noting that the expressions

were too complex to go into more detail. Other more specific derivations include a six-state

model of H+/substrate symport by Sanders et al.

212, a six-state model of ion coupled transport by

Lauger and Jauch213

, and a four-state model of ion pumps by Lauger focused on the shape of the

energy barrier214

.

Hansen et al. were among the first to describe the geometric properties of the curves

themselves215

. A reduced two-state model ignoring substrate was used, where all the voltage

independent steps were combined together and one voltage dependent step was used. They

considered the net velocity (forward and reverse velocities subtracted), and then used a change

of coordinate system and hyperbolic trigonometric functions to simplify the expression.

103

Although they did have success identifying symmetries, the expression remained too complex,

and the model too simple, to be of general use.

At the time of these earlier investigations quality I–V measurements were difficult to obtain, and

there was an emphasis on the saturation behavior at extreme potentials for pumps and

ionophores, such as currents measured from a Neurospora crassa H+ pump

216,217, a Na

+ pump in

sheep cardiac Purkinje fibres218

, active Na+ currents across frog skin

219, and ion transport by

valinomycin in lipid bilayers220

. Glucose activated currents were measured several years later in

intact epithelial cells with ionic control of the membrane potential221

, and in intact Necturus

small intestine222

.

The patch clamp technique was developed by Neher and colleagues in 1978223

, and during the

80’s grew in popularity for studying channels and pumps and measuring I–V curves, such as the

first I–V measurement of the Na+/K

+ pump in myocardial cells (1985)

224. Expression cloning

soon followed in 1987, and attention shifted towards characterizing various cloned transporters

and designed mutants, and interest in a mathematical solution of the I–V curves waned. The first

I–V measurements of SGLT1 expressed in oocytes were reported in 1990225

, and were soon

followed by a wide range of carrier proteins (in chronological order): GABA transporter

GAT1180,226

; serotonin transporter SERT181

; H+/hexose transporter STP1

183; pig SGLT3

61;

H+/peptide transporter PepT1

184,227; Na

+/myo-inositol transporter SMIT1

107; excitatory amino

acid transporter EAAT2182

; EAAC1208

; amino acid permease AAP1228

; H+/myo-inositol

transporter MIT229

; H+/sucrose transporter StSUT1

230; SGLT2

62; Na

+/I

− symporter NIS

133; type

II Na+/Pi cotransporter NaPi-5

231; NaPi-2

197,232; Xenopus SGLT1-like protein xSGLT1L

233;

Na+/multivitamin transporter SMVT

136; reverse SGLT1 transport

80,82; flounder renal high-

affinity-type Na+/dicarboxylate cotransporter fNADC-3

234; amino acid permease AAP2–6

235;

SMIT2105

; Na+/nucleoside transporter CNT1

236; Na

+/monocarboxylate transporter SMCT1

109,110;

NaPi-IIb237

; choline transporter CHT124

; CNT2238

; zebrafish Na+/monocarboxylate cotransporter

zSMCTe114

; CNT3187

.

With a plethora of kinetic data, and advancements in computing power, there was a shift in

emphasis towards model simulation and away from mathematical representations: SGLT1

rat239,240

, rabbit70,86,240-242

(D176A179

, Q170E189

, and BBMV76

), and human42,85,240

with

substrate77,175,243

and/or fluorescence175,243,244

; SGLT1 reverse transport82

; SGLT262

; SMIT1107

;

104

EAAC1193

; GAT1245

; PEPT1184

; StSUT1230

; NaPi-IIa197,232

; theoretical work246,247

. However, a

problem with these simulations is an inability to deduce fundamental properties of a system.

3.1.2 This Study

As mentioned above, earlier theoretical models of the steady-state velocity’s voltage and

substrate dependence have tended to rely on models tailored for a particular transporter. A

complex mathematical expression would typically be derived, and its characteristic studied for

different configurations of the voltage and substrate dependent steps. This task, however, is

laborious, requires a significant level of expertise, and can ultimately be futile if the wrong

model is chosen. In this study we have instead considered a general -state cycle without

limiting the placement of substrate and voltage dependent steps, as discussed in §3.2 Steady-

State Velocity of a Cyclical Model. This ends up making the derivation easier because we can

focus on overarching patterns as opposed to complex rate constant products. The trick is in

realizing that term in the general velocity equation can be conceptualized as a recursive family

of patterns. The effect voltage has on this equation is considered in §3.3.1 Introducing Voltage

Dependence and §3.3.2 The General Voltage Dependent Equation, and the geometric properties

of this curve are explored through examples in §3.3.3 Geometric Properties of the I–V Curves.

The ultimate conclusion from this section is that voltage dependence transforms the

denominator of the steady-state velocity equation into a series of voltage dependent exponential

terms, yet, in most cases, a single exponential term dominates and allows the others to be

ignored, resulting in the familiar Boltzmann function. This provides a phenomenological

method for characterizing the I–V curves by fitting them with Boltzmann functions, as is shown

later. Substrate dependence is considered in §3.4 Substrate Dependence, where we see how

changing the substrate concentration shifts the exponential denominator terms, and in so doing

can change the dominant exponential term. Lastly in §3.5 Results, steady-state data from our lab

for SGLT1 and other carriers taken from the literature are fit with Boltzmann functions and a

mathematical model is constructed. We see how this model can be related back to the original

cycle and few rate limiting patterns. To provide simple examples mostly finite models are

considered in the main text, while general derivations are given in §3.7 Appendix.

105

3.2 Steady-State Velocity of a Cyclical Model

Often transporters can be represented by a simple cycle, if substrate interactions are ordered and

leak pathways are small12,175,215,248. The general case is accounted for by the -state cycle

shown in Fig. 35. The steady-state velocity of this system is well defined (Eq. 29) and is derived

in the Appendix (§3.8.1 Deriving and Arranging the Steady-State Equation). However, for our

purposes all that is important is the form of this equation, which is easier to demonstrate with

the small three-state cycle shown in Fig. 36. Understanding these patterns will clarify the

general solution and help later on when we introduce voltage dependence.

The steady-state velocity ( ) of any cycle can be separated into counterclockwise ( ) and

clockwise ( ) componentsf,

Eq. 3

which have the following solution for the three-state system in Fig. 36,

( )

( )

( )

(

)

(

)

(

)

Eq. 4

with,

Eq. 5

f (rate) is used for the steady-state velocity so that and can be reserved for voltage and sigma notation,

respectively.

106

Each component describes the velocity of the reaction in one direction and the net velocity is

arrived at by subtraction.

Because the model is symmetric there are natural symmetries between the counterclockwise and

clockwise expressions as well as the denominator terms, which can be organized into recursive

patterns. The fundamental repeating pattern is a block of terms (grey, orange, and yellow in Fig.

36) containing a common “head” ( ⁄ ) and a series of expanding “tails” ( ). These

head and tail pairs, which we will refer to their product as “snakes”, form expanding patterns

that start at the head transition and grow around the loop. Consider for example terms in the

grey block of : the head ( ⁄ ) starts at transition 1, and tails are grown in the clockwise

direction through transition 3 ( ) and then 2 ( ), stopping when the circuit is complete.

This pattern is then repeated for each possible head (orange and yellow blocks) to fill in the

remaining denominator terms.

The counterclockwise velocity heads are counterclockwise rate constants ( ⁄ ) and the tails are

grown in the clockwise direction ( ), while the clockwise velocity has clockwise

heads ( ⁄ ) and tails that grow in the counterclockwise direction (

). It is not hard

to see how these patterns can be extended to any number of states by adding blocks for

additional heads and traversing the tails further around the loop. It is important to note that these

snake patterns have units of time, as we will show later how they are related to the observed

behavior of the carrier.

3.3 Voltage Dependence

3.3.1 Introducing Voltage Dependence

Voltage dependence is accounted for with Eyring rate theory by adding an exponential term to

each voltage dependent rate constant213,215

,

Eq. 6

( )⁄ is the reduced membrane potential, where is the membrane potential, the

Faraday constant, the gas constant, and the absolute temperature. The voltage dependence

of a step is described by a valence , which is proportional to the quantity of charge moved and

107

the fraction of the field crossed, and when the voltage is zero the rate constant reduces to ;

this notation is also used for the and factors.

To see what effect Eq. 6 has on the velocities we can substitute it into one of the blocks in Eq. 4

(grey),

( ( )

( ) ( ) )

( ( )

( ) ( ) )

Eq. 7

By combining coefficients and exponents each term can be reduced to a simple exponential

( ), and because all blocks behave the same, and larger cycles will just have more terms, the

general form of these velocities is an inverted series,

Eq. 8

Since all , the greatest effect on these curves’ behavior is the signs of the exponents

determined by . If the signs are all the same, the curves will be monotonically increasing

( ) or decreasing ( ) with increasing . However, with a mixture they will switch

between increasing and decreasing, and be much harder to describe in a general way.

For the remainder of this discussion we will only consider the behavior of curves similar to Eq.

8 without mixed signs. In practice this includes any carriers that have monotonic I–V curves,

and excludes those that demonstrate biphasic behavior. Fortunately, most carriers that transport

like-charged substrates, such as SGLT1, have monotonic curves, while some pumps have

biphasic curves. This condition is achieved by restricting the valences of rate constants in each

108

direction around the loop to opposite signsg. This implies that all the rate constants around each

loop accelerate with the same polarity, and that the two loops have opposite polarity.

The convention is for models to rotate counterclockwise at negative potentials, which occurs

when,

Eq. 9

and this will be the orientation that we adopt for the rest of this discussion.

3.3.2 The General Voltage Dependent Equation

Eq. 8 can be written in a more general and simplified form by considering voltage independent

terms ( and ), and moving their coefficients into the exponents to be expressed as horizontal

shifts ( and ) as outlined in §3.8.2 Simplifying the Voltage Dependent Expressions,

∑ ( )

∑ ( )

Eq. 10

where,

(

)

(

)

Eq. 11

g This becomes clear when we examine the form of the exponents in Eq. 7 for ( or ) and ( or

).

109

Reduced voltage dependent terms have combined valences ( and ) and coefficients ( and

),while the reduced voltage independent terms have been grouped together ( ∑ and

∑ ), and limit the maximum velocity in both directions ( ⁄ and ⁄ ).

3.3.2.1 Example 1

The steady-state velocities of the three-state model in Fig. 37 will be solved for and arranged

into the form of Eq. 10. We will begin by writing down the head and tail pairs using the patterns

illustrated in Fig. 36,

(

)

(

)

(

)

(

)

(

)

(

)

Eq. 12

and then collect and reduce the terms,

Eq. 13

Finally the voltage independent terms are rearranged and the amplitudes are brought into the

exponent,

( ) ( )

( ) ( )

Eq. 14

110

3.3.3 Geometric Properties of the I–V Curves

The steady-state expressions in Eq. 10 are sigmoid-like functions, ( ( ))⁄ , but with

the potential for multiple exponential terms in the denominator instead of just one. Since the

counterclockwise and clockwise expressions have the same form, and are simply reflections of

one another in the y-axis, we will only consider the counterclockwise velocity. As we will show,

the shape of most I–V curves can be accounted for by one or two terms, and so it helps to study

their geometric properties.

With one term the curve is a simple sigmoid, as shown in Fig. 38. It varies asymptotically

between 1 and 0, reaching its midpoint and steepest part at . For the rest of this

discussion we will use the notation and instead of and , because it is simpler to

write, but also because in the presence of multiple denominator terms it is not necessarily the

value of or that results in half maximal velocity. The majority of change (90%) occurs

within a band ( ⁄ ) on either side of , with the band becoming smaller, and the

slope steeper, as increases. In this way, and describe the curve’s voltage dependence.

3.3.3.1 Last Man Standing

If we think about the steady-state velocity equation in terms of the denominator terms (Eq. 4),

and recall that they represent snake patterns with units of time, we can see that the voltage

dependent patterns are infinitely slow at depolarizing potentials and increase in speed

exponentially to zero at hyperpolarizing potentials. In turn, the steady-state velocity starts off

negligible at depolarizing potentials and speeds up as the patterns become faster with

hyperpolarization. This behavior is illustrated in Fig. 39, which shows that when there are

multiple voltage dependent terms in the denominator the one with the most negative is the

last to speed up and therefore has the largest impact on the shape of the curve. If we examine the

curve at depolarizing potentials and move negative, we see that at all the terms are to the

right of their and much larger than 1, forcing the expression to 0. The first term reaches its

at , and drops to 1, but the other two terms are still large and the expression remains 0; this

is repeated again for the second term at . Only when the last term reaches its at

does this exponential series approach 1 and the curve begins to rise. This shows how even

though there may be many exponential terms in the velocity equation, often the one with the

most negative is dominant.

111

3.3.3.2 Overlapping Terms

Sometimes two terms can contribute to the shape of the velocity curve, if they have the most-

negative ’s and their bands overlap. An example of this behavior is shown in Fig. 40

for a low valence ( ) and high valence ( ) term. This large difference in valences is

needed to see the properties of both curves. In real systems the range of valences is usually

much smaller (0–2), and therefore it may be difficult to see the properties of more than one. In

this example the low valence (red) is held fixed while the high valence (blue) is shifted to

observe the effect this has on the compound curve (black). Because the compound curve is

limited by both terms, it can be drawn heuristically by tracing a path along the lower of the two.

The shallow slope and wider band of the low valence term ( ) forces the high valence to fit

underneath it. This results in a combined curve that has properties of the high valence term on

the right and the low valence on the left. This combined curve usually has one inflection point,

but if the high valence term is far enough to the right there will be three; one from each sigmoid

and another connecting them.

3.3.3.3 Example 2

Solving the model in Fig. 41 results in the following velocity equations,

( )

( )

Eq. 15

which are plotted along with the net velocity. The counterclockwise velocity has three inflection

points (green dots), because the higher valence term is far enough to the right ( ). In

contrast, the clockwise velocity only has one (red dot), because the two terms have similar

valences (−5.5 and −5) and are close together ( ). Notice that at extreme potentials

the net velocity approaches the directional velocities, as the opposing velocity becomes

negligible.

112

3.4 Substrate Dependence

3.4.1 Introducing Substrate Dependence

To see how substrate affects the denominator terms it will help to recall some examples from

Eq. 4,

Eq. 16

An important feature of these, and all the other terms, is the separation of counterclockwise and

clockwise rate constants across the numerator and denominator. If one substrate binding event is

introduced in each direction, we would need to consider three dependencies ( , independent,

), and for two binding events five ( , , independent, , ).

This many cases become cumbersome, and so a practical way to reduce them is to only consider

substrate binding in one direction. This simplification is used in the Michaelis-Menten

derivation and is natural since most I–V measurements are made under conditions that attempt

to minimize reverse transport, such as low intracellular substrate concentrations. We will

therefore limit this discussion to the counterclockwise velocity (i.e. forward transport) with

binding events in the counterclockwise direction.

Substrate dependence for one binding event can be added to the counterclockwise velocity in

Eq. 10 by adding voltage dependent (

) and independent (

) terms with a ⁄

dependence, as described in §3.8.3 Simplifying the Substrate Dependent Expression.

∑ ( ) ∑ (

)

Eq. 17

113

where,

and,

(

)

(

)

Eq. 18

This results in a maximum velocity that scales with substrate, , and two types of

exponential terms with different exponential shifts, and .

3.4.2 Characteristics of Substrate Dependence

Unsurprisingly, from Eq. 18 follows Michaelis-Menten kinetics,

where,

Eq. 19

as it represents the substrate dependence of the steady-state velocity at saturating voltage. This

shows that the maximum velocity at saturating voltage and substrate, , is governed by the

sum of the non-dependent (voltage/substrate independent) snake patterns ( ), while depends

on the ratio of the non-dependent and substrate dependent ( ) patterns (see Eq. 36).

114

The two exponential shifts in Eq. 18 have logarithmic substrate dependencies that can be

expressed in a simplified from as,

(

)

( )

Eq. 20

Terms representing voltage dependent snake patterns shift negative ( ), while those for

voltage/substrate dependent patterns shift positive ( ) as increases, as illustrated in Fig. 42.

The ’s of the voltage dependent terms decrease from infinity to a plateau of ( ) (red),

while the ’s of the voltage/substrate dependent terms increase from ( ) to infinity (blue).

This tells us that at low a voltage/substrate dependent exponential term will have the most

negative and be dominant, while at high a voltage dependent term will take over the

dominant role.

A special case occurs when the substrate binding step is voltage dependent, because none of the

substrate dependent snake terms will be voltage independent, and . This alters the

expressions,

(

) ( )

(

) ( )

Eq. 21

The maximum velocity loses its substrate dependence, since voltage can compensate for low

substrate concentrations. The ’s of the voltage dependent snake terms become substrate

independent, while the ’s of the voltage/substrate dependent terms now increase from

negative infinity to positive infinity.

115

3.4.2.1 Example 3

Some substrate dependent properties of the I–V curves are demonstrated with the examples in

Fig. 43. The model in Fig. 41, originally used to demonstrate voltage dependence, is modified

by adding a substrate binding event to either the voltage dependent transition (2→3) or the one

following it (3→1). The substrate dependence of the – curves is shown in A, and trajectories

of the – curves are in B; the paths followed by the are also overlaid in A.

In case 1, the substrate binding event is voltage dependent, and, therefore, is constant and

both increase (shift to the right) with increasing (see Eq. 21). At low concentrations (0.01–

0.1) is more negative and dominates, and the curve has a shallow slope ( ), but as

increases (1–10) the bands overlap (see B) and ’s steeper valence ( ) begins to

appear. Eventually is passed by (≥100) and becomes the most negative and dominant

term.

In case 2, substrate binding is voltage independent, and we see an that increases with

and a decreasing in (see Eq. 19 and Eq. 20). Over most substrate concentrations all three

bands overlap (0.01–1000, B) and both valences (0.5 and 5.5) appear in the – curves (A).

Only at very large concentrations (1000–1000×105) does drift above the band and slowly

disappear.

3.5 Results

3.5.1 Characterizing Experimental Data

To test the theory derived above, we analyzed substrate titration I–V data for a number of

cotransporters that are shown in Fig. 44A. The glutamine to cysteine mutant at position 170 of

rSGLT1 (Q170C) studied by our lab, and data digitized from the literature for the

2Na+/nucleoside human concentrative nucleoside transporter 3 (hCNT3)

187, hSGLT1

249, and the

3Na+/Pi rat renal type II cotransporter (NaPi-2)

197. These sigmoidal I–V curves were fit with a

Boltzmann function, ( ( ))⁄ , to measure a substrate dependent series of ’s

and ’s.

All of the carriers besides hCNT3 responded similarly, in that their ’s were relatively substrate

independent and their ’s initially followed an increasing trajectory (blue) and then plateaued

116

(red line). This increasing trajectory can be attributed to a positive shifting of the dominant

exponential term (see Fig. 42). Eventually, however, this positive shifting term passes another

exponential term that is following a negative, or flat, trend that has plateaued, and which

becomes the new dominant term. As shown, the increasing trajectory was fit with a logarithmic

function, ( ) ⁄ , to measure its parameters. However, only the Q170C with Na+

trajectory had enough detail to solve for , while the others, instead, used the average values

from B.

In the case of hCNT3 the dropped rapidly (3.5–0.5) as the substrate concentration increased.

This indicates that the dominant exponential term is replaced multiple times with lower valence

terms as the substrate concentration increases. Because the dominant term is changing, the

decreasing trajectory cannot be fit, and instead shows the ’s of several different terms as

they shift negative and take over the dominant position.

In the case of Q170C and hCNT3, where the charged substrate Na+ was used, we can tell that

Na+ binding to Q170C is voltage dependent because of its substrate independent . On the

other hand, Na+ binding to hCNT3 is clearly voltage independent because is substrate

dependent and the shifts negative. These findings provide clues about the pathway to the Na+

binding site, in that SGLT1’s is likely narrower, causing Na+ to sense the membrane electric

field, while hCNT3’s is wider and open to the extracellular environment.

3.5.2 Modeling the Steady-State Velocity

If we look at the general form of the steady-state velocity from Eq. 35, which is reproduced

below,

Eq. 22

we see that the series of terms in the denominator have different combinations of dependence on

substrate and voltage (none, , , and ). These terms come about from the snake patterns

which are then grouped based on these dependencies. As discussed in §3.4.2 Characteristics of

117

Substrate Dependence, ∑ and ∑

are related to the standard Michaelis-Menten parameters

, and (see Eq. 19 and Eq. 36), and can be calculated from them,

{

Eq. 23

When substrate binding is voltage dependent there is no ∑

term. We also know that, often,

one of the voltage dependent terms will be dominant, and so we can simplify Eq. 22 by

considering just one of each type of voltage dependent term,

Eq. 24

At low substrate concentrations,

is the dominant voltage dependent term, because

substrate is rate limiting and it represents the slowest substrate/voltage dependent snake pattern.

As substrate increases this term shifts positive following the blue trajectories in Fig. 44 (Eq. 20

or Eq. 21). Eventually, though, substrate stops being rate limiting and , which either shifts

negative or is substrate independent, becomes dominant, as it becomes the slowest voltage

dependent snake pattern (red dotted lines in Fig. 44). Looking at these different regions, we can

use the and information from the I–V fits, along with Eq. 39 and Eq. 20, to calculate the

remaining snake patterns,

Eq. 25

118

where is the saturating indicated by the red dotted line, and and are from the

logarithmic fit of the upward trajectory indicated by the blue line. The results of these

calculations are shown for several of the data sets in Fig. 45, and then used to construct a model

of the steady-state velocity equation using published and data. The calculation of

using always gave better simulation results and were therefore used, likely because of

inaccuracy in measuring the logarithmic slope ( ) with such few data points. The simulated

curves are in good agreement with the original data. This confirms that the steady-state velocity

can be modeled simply with one term of each type of dependence in the denominator (none, ,

, and ), and that the two voltage dependent terms ( and ) switch dominance as substrate

increases and ceases to be rate limiting. The Q170C rSGLT1 simulation deviated slightly at low

Na+ concentrations, indicating that a ⁄ term might be needed since two Na

+ ions bind to

SGLT1.

What is remarkable about these findings is that steady-state velocity gives information about

rate limiting snake patterns in the cycle, and that under different combinations of saturating

substrate and voltage conditions, different snake patterns are revealed. It appears that these rate

limiting snake patterns are the most that can be observed with the steady-state velocity, because

the rest of the patterns are faster and remain hidden. The voltage dependent terms and

are natural extensions of the Michaelis-Menten equation that take into account voltage

dependence.

3.6 Summary and Conclusion

The substrate dependence of the steady-state velocity has been of interest in enzyme kinetics for

over a century, marked most significantly by the publication of the Michaelis-Menten equation

in 1913250

. These studies are fundamental for characterizing substrate apparent affinities,

cooperativity, and turnover. With membrane transporters, their location within the membrane

provides a unique opportunity to study the ability of voltage to accelerate the cycling velocity,

and indirectly the voltage dependent transitions, using electrophysiological techniques.

Expression cloning has facilitated these types of studies and the measurement of the familiar

sigmoidal I–V curves. However, the approach to studying their kinetics is largely unchanged

from the classical Michaelis-Menten (or Hill) methods. The Michaelis-Menten parameters are

measured separately at each membrane potential to gauge their voltage dependence, yet the

119

sigmoidal shape of the I–V curves have remained mostly unexplored. Most attempts at

describing the I–V curves have failed because of the inherent complexity of the steady-state

velocity expression, which grows geometrically with the size of the model, and the added

complexity of working with exponential terms introduced through voltage dependence.

In order to get at the kinetic information contained within the I–V curves, we studied the

patterns formed by the denominator terms of a general -state cycle. The trick was in realizing

that these terms can be written down by following a simple recursive pattern, using the concept

of head and tail pairs. Conceptually, these snake terms have units of time and so the slowest

term, which corresponds with a pattern in the model, dominates the others and allows the

velocity expression to be greatly reduced. The reduced velocity expression is a simple

Boltzmann function, with one exponential term in the denominator, and provides a

straightforward phenomenological method for characterizing the I–V curves, by measuring their

’s and ’s. Not only is this analysis easy to do, but it provides a way to obtain and data—

which normally have required transient studies—from steady-state data that is often readily

available from apparent affinity measurements. We can go one step further and use this and

data, in combination with the Michaelis-Menten parameters, to calculate the values of these rate

limiting terms and build a reduced/equivalent model of the carriers voltage and substrate

dependence. This provides additional measures for characterizing and comparing transporters

and their mutants. These voltage dependent terms and are a natural extension of

the Michaelis-Menten parameters (∑ ⁄ and ∑

⁄ ) into the voltage

dependent space. The ultimate finding is that the steady-state velocity provides access to these

rate limiting snake patterns, which are also the most we can extract from the I–V curves. The

next step would be to try and build a model of the transport loop by guessing at the rate limiting

patterns and using them as building blocks.

3.7 Future Work

This project began because several SGLT1 mutants we were studying had highly voltage

dependent apparent affinities for Na+ and αMG and we wanted to model this behavior, thinking

that it might give insight into substrate binding order (see Fig. 46). At first the models designed

were specific to SGLT1, much like earlier theoretical studies, but this approach turned out to be

difficult and cumbersome. Eventually, on the suggestion of Dr. Backx, we began modeling the

120

steady-state current, thinking that it could be used to understand the apparent affinities. This

project is a completion of that steady-state work, and it would make sense now to return to the

apparent affinities, as they may contain additional information. It should be relatively straight

forward to model the apparent affinity as the expression is directly related to the steady-state

velocity (when , ⁄ ).

Ideally, it would be great to do a comprehensive experimental study of a cotransporter’s

kinetics, to validate the theory further and gain a deeper understanding of the carrier. With a

carrier like SGLT1, titrating with both substrates (Na+ and glucose) would provide information

on six–eight snake patterns (none, , , , , , , ; Na= , glucose= , voltage= ).

Using the cut-open oocyte, these measurements could be made in the forward and reverse

directions with complete control over the intra and extracellular substrate concentrations. This

would allow the destination solution to be clamped at zero substrate to eliminate reverse

transport and give more accurate readings. Working with patterns in both directions, it might be

possible to piece them together into a model of the carrier’s cycle by taking into account

substrate binding order and the relationships between the substrate dependent terms. The end

result would be simplified rate equations in both directions, and a full model built from the

patterns. Experiments with substrate on both sides of the membrane could then be used to test

these models in a mixed mode that they were not designed for.

3.8 Appendix

3.8.1 Deriving and Arranging the Steady-State Equation

Working with the three-state system in Fig. 36, we will first show how to transform the standard

form of the steady-state equation into one that is more useful for a voltage dependent analysis,

and then extrapolate to the general case.

The steady-state velocity equation in its general form,

Eq. 26

121

can be written down directly using the schematic methods of Wong–Hanes for the numerator

terms (∑ ), and King–Altman for the denominator (∑ )251,

Eq. 27

The net velocity, , is then split into forward and reverse components,

Eq. 28

which are each divided by their numerator and organized to arrive at the form in Eq. 4.

The patterns displayed in Eq. 4, and described in §3.2 Steady-State Velocity of a Cyclical Model,

can then be extended to the general case with states (see Fig. 35).

∑ [

( ∑ ∏

)]

∑ [

( ∑ ∏

)]

Eq. 29

Because the model is cyclical, indices in Eq. 29 that fall outside the range , need to be

mapped using modular arithmetic.

3.8.2 Simplifying the Voltage Dependent Expressions

We can express Eq. 8 in a more formal way by considering voltage independent terms and

writings separate expressions for the counterclockwise and clockwise velocities,

122

∑ ∑

∑ ∑

Eq. 30

Voltage dependent terms have combined valences ( and ) and combined coefficients with a

notation ( and

). The voltage independent terms ( and ) come about if there is at

least one voltage independent step in either directionh. Eq. 30 can then be simplified by

grouping the voltage independent terms,

Eq. 31

moving them into the numerator,

Eq. 32

and then bringing the new exponential coefficients into the exponents to be expressed as

horizontal shiftsi,

h For example, if the rate constants and are voltage independent, there will be a ⁄ term in and a ⁄

term in .

i ( ),

(

).

123

∑ ( )

∑ ( )

Eq. 33

where,

(

)

(

)

Eq. 34

3.8.3 Simplifying the Substrate Dependent Expression

To add substrate dependence for one binding event to the counterclockwise velocity in Eq. 30,

we need to include voltage dependent (

) and independent (

) snake terms

with a ⁄ dependence,

Eq. 35

The voltage independent terms can then be grouped,

Eq. 36

moved,

124

(

)⁄

Eq. 37

and the new coefficients brought into the exponents as shifts,

(

)⁄

∑ ( ) ∑ (

)

Eq. 38

where,

(

)

(

)

Eq. 39

3.8.4 Two Substrate Binding Events

Extending the derivation in Eq. 35–Eq. 39 to two binding events adds a quadratic term,

(

)⁄

∑ ( ) ∑ (

) ∑

(

)

Eq. 40

where,

125

(

)

(

)

(

)

Eq. 41

126

3.9 Figures

Fig. 34: (A) Example I–V curves measured over a range of substrate concentrations. (B) At each

potential the saturating I–s curves are fit to the Michaelis-Menten or Hill equation to determine

, and . Data is from the Q170C mutant of rSGLT1.

127

Fig. 35: General -state cyclical model.

128

Fig. 36: Example showing the form of the steady-state equation. The three-state model at the top is used to

illustrate these patterns for the steady-state velocity ( ), which has been split into the counterclockwise ( ) and

clockwise ( ) expressions written in full below. The denominator terms are organized into blocks of repeating

patterns that are colour coded to match the diagrams. The diagrams show how the terms within each block can be

generated iteratively by starting with a head term ( ⁄ ) and then growing the tail (∏ ) in a cyclical pattern until

the model is traversed. Note the symmetry between the and patterns, and how there is one block for each

possible head term. This pattern is easily applied to larger models by adding more blocks and extending the tails

around a larger loop.

129

Fig. 37: Example 1. Three-state model used to demonstrate a solution of the voltage dependent

general velocity equation.

Fig. 38: Geometric properties of a sigmoid function. The curve varies asymptotically between 1

and 0, reaching half-height at where it is also the steepest (slope, ⁄ ). A majority of

the change in magnitude (90%) occurs within the band .

130

Fig. 39: Effect of a dominant denominator term. With multiple exponential terms in the

denominator the one with the most negative dominates. The bottom expression is plotted and

evaluated at different values of . As terms pass their ’s they drop from to 1 to 0. When the

last term passes its the others are already small and the curve begins to rise.

131

Fig. 40: Geometric properties of a sigmoid function with two terms. A composite sigmoid

(black) contains two exponential terms, a low valence ( ) and high valence ( ).

Separate curves are also drawn for single term functions with the low valence (red) and high

valence (blue). The low valence term is held fixed while the high valence is shifted (magenta,

) to show the effect this has on the composite curve. The composite curve is

roughly drawn by following the lower of the two single term curves at any point. Inflection

points are shown as coloured dots. When the high valence term is far enough to the right of the

low valence the composite curve has three inflection points ( ), but otherwise there is only

one.

132

Fig. 41: Example 2. The model shown is used to illustrate voltage dependent properties of the

general velocity equations. The counterclockwise ( , green), clockwise ( , red), and net velocity

( , black) are shown. Inflection points are marked (dots), and the ’s are indicated by the

dotted lines ( is shared by both curves, black).

133

Fig. 42: Characteristics of the logarithmic exponential shifts. For a single binding event the ’s

can shift positive (blue) or negative (red). The red curve starts at infinity and asymptotically

approaches ( ) at large , while the blue starts at ( ) and increases to infinity. The

substrate concentrations where are indicated at the bottom.

134

Fig. 43: Example 3. Substrate dependence of the I–V curves. A substrate binding event is added

to two different steps of the model in Fig. 41, shown in the diagrams on the left; voltage

dependent transition in red, and substrate binding in blue. (A) Substrate dependence of the –

curves with overlaid paths showing ( ) (direction of the trajectories indicated by the

arrow). (B) – curves plotted on a log- scale with the bands shown in grey. curves

with a dependence are dashed, and are solid; those with increasing log

dependence are dark blue and a decreasing light blue. The equations are shown on the right.

135

Fig. 44: Analyzing experimental I–V data. (A) I–V substrate titration data were analyzed for the

Q170C mutant of rSGLT1 (1), hCNT3 (2)187, hSGLT1 (3)249 and rat NaPi-2 (4)197. These

curves were fit (black lines) with a one-term Boltzmann function, ( ( ))⁄ ;

in 1, is shared across Na+ concentrations, as expected for voltage dependent binding. The

fits returned ’s and ’s at each substrate concentration and they are plotted in B and C (axis tic

colours match data in A); ’s are also overlaid in A (open circle). (B) The valences were

mostly substrate independent and their average is indicated (dotted line). (C) The rising phase of

the – data at low was fit with the logarithmic function shown (blue), with results given on

the right (

). The intercept of this rising phase at zero can be calculated from and is

shown ( ( ) ( )⁄ , blue dotted line). A plateaued , from another term following a

downward or flat trajectory, would become dominant at high and was not fit as part of the

rising phase (red dotted line). The hCNT3 – data could not be fit, because the changing

valence indicated a changing dominant term.

136

Fig. 45: Steady-State velocity models. (B) The steady-state velocity equation was constructed

using the published turnover ( ) and apparent affinity ( ) data shown in A, and values for

and were calculated from the fit results in Fig. 44 using Eq. 25. It was found

that values calculated with gave better results than , and, therefore, these were used. (C)

The steady-state velocity equations in B are simulated (coloured lines). Turnover data for

Q170C rSGLT1188

, turnover12

and apparent affinity249

data for hSGLT1, and turnover and

affinity data for NaPi-2197

.

137

Fig. 46: Voltage dependence of the Na+ and αMG apparent affinities for the Q170C and Q170E

mutants of rSGLT1. The affinities have been fit with an exponential function plateauing at zero

(red) or above zero (blue).

138

Conclusion and Future Work 4

For a long time studying membrane transporters has been a block box analysis. Little was

known about their structure, let alone their inner workings, and the primary way to learn about

them was to stimulate and observe the output. Electrophysiological experiments have become

the pinnacle of these types of studies because of their unmatched ability to both stimulate and

observe. For many years, electrophysiological studies have uncovered curiosities and measured

parameters in greater detail, but without a physical representation of the carrier to work with it

was extremely difficult to unravel mechanisms with kinetic data alone. We are fortunate now to

have the structural data that we do from the crystal structures, as it provides a context for

interpreting kinetic data by essentially opening up the black box so we can gaze at its constituent

parts. Now that we can see how these biological machines are put together, kinetic studies are

more important than ever, as they provide a way to reverse engineer these proteins by observing

them in action. Kinetic and structural studies are complementary because they each fill in a

missing part of the other. It would be a mistake to assume that transporters can be understood by

looking only at their structures; a mistake on par with assuming that kinetics could do the same.

With that said, kinetic studies can only be helpful in elucidating mechanism, as opposed to a

tool for characterization, if we understand what the kinetic parameters measure. To do this we

need good models that can explain where the parameters come from, and accurate

measurements that provide authentic observations of the carrier. In this regard there are two

prominent time domains for kinetic studies, the transient and steady-state; equilibrium studies

could provide additional information, but they are less suited to electrophysiology and more so

for radioactive uptake. In this thesis we approached both time domains with a goal of

understanding the composition of their kinetics, and to then use this understanding to retrieve as

much kinetic information as possible. The hope was that by elucidating the nature of these

kinetics they could be related more easily back to an understanding of the carrier.

The transient project demonstrated that it was necessary to find all the decay components in

order to measure the kinetics of real conformational changes and not some form of weighted

average. When this was done it became much easier to interpret the kinetic data in terms of a

model, leading to a kinetic representation of the gated rocker-switch mechanism with its four

139

conformational changes. The steady-state project relied on theory to understand what factors

affect the cycling velocity. In this case we learned that the I–V curves are governed by rate

limiting patterns in the transport loop, and this led to new ways of measuring them, and a

method for building descriptive models from them. Most importantly, this allows us to think

about the transport loop in terms of these rate limiting patterns.

The transient and steady-state kinetics provide different windows into the carrier transport cycle.

Transient currents are mostly available only in the absence of the non-ionic substrate, and

therefore monitor conformational changes of the unloaded carrier (top of Fig. 31), while the

steady-state currents incorporate the entire transport loop, and are one of the few ways of

studying states in the presence of substrate. Although kinetics from transient studies are

theoretically superior because they provide information on individual conformational changes,

steady-state kinetics are necessary to study substrate binding. An ideal project would be to

combine transient and steady-state kinetic studies for a single carrier, such as SGLT1. The first

step would be to build a state model of the empty carrier conformations using transient kinetics

(top of Fig. 31). This model could then be used as a template for the carrier conformations in the

presence substrate (bottom of Fig. 31). Steady-state studies, in the forward and reverse

directions, could measure the kinetics of the various rate limiting patterns in each direction, and

the form and overlap of these patterns could be guessed at and incorporated into the model.

These techniques have introduced a number of kinetic parameters that can be used to

characterize and understand ion-coupled cotransporters. These provide new opportunities to

classify and compare cotransporters kinetically and study the effects of mutations. Using the

crystal structures as a guide, strategic mutations can be used to single out key parts of the

transport mechanism, while monitoring the effect kinetically with parameters that map onto

specific functions. As our understanding of the connection between these parameters and the

structure grows, these techniques will only become more effective. By understanding these

mechanisms we can better understand mutations that lead to disease, we can build better

inhibitors, and perhaps the ultimate goal is to one day build biological machines of our own.

140

References

1. Crane RK. The gradient hypothesis and other models of carrier-mediated active

transport. Rev Physiol Biochem Pharmacol. 1977;78:99–159.

2. Jardetzky O. Simple allosteric model for membrane pumps. Nature. 1966 Aug.

27;211(5052):969–970.

3. Schultz SG, Curran PF. Coupled transport of sodium and organic solutes. Physiol Rev.

1970 Oct.;50(4):637–718.

4. Kaback HR, Stadtman ER. Proline uptake by an isolated cytoplasmic membrane

preparation of Escherichia coli. Proc Natl Acad Sci USA. 1966 Apr.;55(4):920–927.

5. Hopfer U, Nelson K, Perrotto J, Isselbacher KJ. Glucose transport in isolated brush

border membrane from rat small intestine. J Biol Chem. 1973 Jan. 10;248(1):25–32.

6. Murer H, Kinne R. The use of isolated membrane vesicles to study epithelial transport

processes. J Membr Biol. 1980 Jul. 15;55(2):81–95.

7. Stevens BR, Kaunitz JD, Wright EM. Intestinal Transport of Amino Acids and Sugars:

Advances Using Membrane Vesicles. Annu Rev Physiol. Annual Reviews 4139 El

Camino Way, P.O. Box 10139, Palo Alto, CA 94303-0139, USA; 1984 Oct.;46(1):417–

433.

8. Silverman M. Structure and function of hexose transporters. Annu. Rev. Biochem.

1991;60:757–794.

9. Hediger MA, Coady MJ, Ikeda TS, Wright EM. Expression cloning and cDNA

sequencing of the Na+/glucose co-transporter. Nature. 1987;330(6146):379–381.

10. Wright EM, Hager KM, Turk E. Sodium cotransport proteins. Curr Opin Cell Biol.

1992 Aug. 1;4(4):696–702.

11. Hediger MA, Kanai Y, You G, Nussberger S. Mammalian ion-coupled solute

transporters. J Physiol (Lond). 1995 Jan.;482:7S–17S.

12. Longpré J-P, Lapointe J-Y. Determination of the Na+/glucose cotransporter (SGLT1)

turnover rate using the ion-trap technique. Biophys J. 2011 Jan. 5;100(1):52–59.

13. Bamberg E, Clarke RJ, Fendler K. Electrogenic properties of the Na+,K

+-ATPase

probed by presteady state and relaxation studies. J Bioenerg Biomembr. 2001 Oct.

1;33(5):401–405.

14. Doyle DA, Morais Cabral J, Pfuetzner RA, Kuo A, Gulbis JM, Cohen SL, et al. The

structure of the potassium channel: molecular basis of K+ conduction and selectivity.

Science. 1998 Apr. 3;280(5360):69–77.

141

15. Lo B, Silverman M. Cysteine scanning mutagenesis of the segment between putative

transmembrane helices IV and V of the high affinity Na+/Glucose cotransporter SGLT1.

Evidence that this region participates in the Na+ and voltage dependence of the

transporter. J Biol Chem. 1998 Nov. 6;273(45):29341–29351.

16. Liu T, Lo B, Speight P, Silverman M. Transmembrane IV of the high-affinity sodium-

glucose cotransporter participates in sugar binding. Am J Physiol, Cell Physiol. 2008

Jul.;295(1):C64–72.

17. Liu T, Speight P, Silverman M. Reanalysis of structure/function correlations in the

region of transmembrane segments 4 and 5 of the rabbit sodium/glucose cotransporter.

Biochem Biophys Res Commun. 2009 Jan. 2;378(1):133–138.

18. Loo DD, Hirayama BA, Gallardo EM, Lam JT, Turk E, Wright EM. Conformational

changes couple Na+ and glucose transport. Proc Natl Acad Sci USA. 1998 Jun.

23;95(13):7789–7794.

19. Gagnon DG, Bissonnette P, Lapointe J-Y. Identification of a disulfide bridge linking the

fourth and the seventh extracellular loops of the Na+/glucose cotransporter. J Gen

Physiol. 2006 Feb.;127(2):145–158.

20. Nagata T, Iizumi S, Satoh K, Kikuchi S. Comparative molecular biological analysis of

membrane transport genes in organisms. Plant Mol. Biol. 2008 Apr.;66(6):565–585.

21. Preston GM, Carroll TP, Guggino WB, Agre P. Appearance of water channels in

Xenopus oocytes expressing red cell CHIP28 protein. Science. 1992 Apr.

17;256(5055):385–387.

22. Murata K, Mitsuoka K, Hirai T, Walz T, Agre P, Heymann JB, et al. Structural

determinants of water permeation through aquaporin-1. Nature. 2000 Oct.

5;407(6804):599–605.

23. Carruthers A, DeZutter J, Ganguly A, Devaskar SU. Will the original glucose

transporter isoform please stand up! Am J Physiol Endocrinol Metab. 2009

Oct.;297(4):E836–48.

24. Abramson J, Smirnova I, Kasho V, Verner G, Kaback HR, Iwata S. Structure and

mechanism of the lactose permease of Escherichia coli. Science. 2003 Aug.

1;301(5633):610–615.

25. Toyoshima C, Nakasako M, Nomura H, Ogawa H. Crystal structure of the calcium

pump of sarcoplasmic reticulum at 2.6 A resolution. Nature. 2000 Jun.

8;405(6787):647–655.

26. Morth JP, Pedersen BP, Toustrup-Jensen MS, Sørensen TL-M, Petersen J, Andersen JP,

et al. Crystal structure of the sodium-potassium pump. Nature. 2007 Dec.

13;450(7172):1043–1049.

27. Pedersen BP, Buch-Pedersen MJ, Morth JP, Palmgren MG, Nissen P. Crystal structure

142

of the plasma membrane proton pump. Nature. 2007 Dec. 13;450(7172):1111–1114.

28. Morth JP, Pedersen BP, Buch-Pedersen MJ, Andersen JP, Vilsen B, Palmgren MG, et

al. A structural overview of the plasma membrane Na+,K

+-ATPase and H

+-ATPase ion

pumps. Nat Rev Mol Cell Biol. 2011 Jan.;12(1):60–70.

29. Grigorieff N, Ceska TA, Downing KH, Baldwin JM, Henderson R. Electron-

crystallographic refinement of the structure of bacteriorhodopsin. J Mol Biol. 1996 Jun.

14;259(3):393–421.

30. Hirai T, Heymann JAW, Shi D, Sarker R, Maloney PC, Subramaniam S. Three-

dimensional structure of a bacterial oxalate transporter. Nat Struct Biol. 2002 Jul.

15;9(8):597–600.

31. Huang Y, Lemieux MJ, Song J, Auer M, Wang D-N. Structure and mechanism of the

glycerol-3-phosphate transporter from Escherichia coli. Science. 2003 Aug.

1;301(5633):616–620.

32. Yin Y, He X, Szewczyk P, Nguyen T, Chang G. Structure of the multidrug transporter

EmrD from Escherichia coli. Science. 2006 May 5;312(5774):741–744.

33. Wright EM, Loo DDF, Hirayama BA, Turk E. Surprising versatility of Na+-glucose

cotransporters: SLC5. Physiology (Bethesda). 2004 Dec.;19:370–376.

34. Hediger MA, Romero MF, Peng J-B, Rolfs A, Takanaga H, Bruford EA. The ABCs of

solute carriers: physiological, pathological and therapeutic implications of human

membrane transport proteins. Pflugers Arch. 2004 Feb.;447(5):465–468.

35. Busch W, Saier MH. The transporter classification (TC) system, 2002. Crit Rev

Biochem Mol Biol. 2002;37(5):287–337.

36. Jung H. Towards the molecular mechanism of Na+/solute symport in prokaryotes.

Biochim Biophys Acta. 2001 May 1;1505(1):131–143.

37. Jung H. The sodium/substrate symporter family: structural and functional features.

FEBS Lett. 2002 Oct. 2;529(1):73–77.

38. Zhou L, Cryan EV, D'Andrea MR, Belkowski S, Conway BR, Demarest KT. Human

cardiomyocytes express high level of Na+/glucose cotransporter 1 (SGLT1). J Cell

Biochem. 2003 Oct. 1;90(2):339–346.

39. Banerjee SK, McGaffin KR, Pastor-Soler NM, Ahmad F. SGLT1 is a novel cardiac

glucose transporter that is perturbed in disease states. Cardiovasc Res. 2009 Oct.

1;84(1):111–118.

40. Wells RG, Pajor AM, Kanai Y, Turk E, Wright EM, Hediger MA. Cloning of a human

kidney cDNA with similarity to the sodium-glucose cotransporter. Am J Physiol. 1992

Sep.;263(3 Pt 2):F459–65.

143

41. Kanai Y, Lee WS, You G, Brown D, Hediger MA. The human kidney low affinity

Na+/glucose cotransporter SGLT2. Delineation of the major renal reabsorptive

mechanism for D-glucose. J Clin Invest. 1994;93(1):397–404.

42. Chen XZ, Coady MJ, Jackson F, Berteloot A, Lapointe JY. Thermodynamic

determination of the Na+: glucose coupling ratio for the human SGLT1 cotransporter.

Biophys J. 1995 Dec. 1;69(6):2405–2414.

43. Wright EM, Turk E, Martín MG. Molecular basis for glucose-galactose malabsorption.

Cell Biochem. Biophys. 2002;36(2-3):115–121.

44. Wright EM, Martín MG, Turk E. Intestinal absorption in health and disease--sugars.

Best Pract Res Clin Gastroenterol. 2003 Dec.;17(6):943–956.

45. Santer R, Kinner M, Lassen CL, Schneppenheim R, Eggert P, Bald M, et al. Molecular

analysis of the SGLT2 gene in patients with renal glucosuria. J Am Soc Nephrol. 2003

Nov. 1;14(11):2873–2882.

46. Calado J, Sznajer Y, Metzger D, Rita A, Hogan MC, Kattamis A, et al. Twenty-one

additional cases of familial renal glucosuria: absence of genetic heterogeneity, high

prevalence of private mutations and further evidence of volume depletion. Nephrol Dial

Transplant. 2008 Dec. 1;23(12):3874–3879.

47. Calado J, Loeffler J, Sakallioglu O, Gok F, Lhotta K, Barata J, et al. Familial renal

glucosuria: SLC5A2 mutation analysis and evidence of salt-wasting. Kidney Int. 2006

Mar.;69(5):852–855.

48. Magen D, Sprecher E, Zelikovic I, Skorecki K. A novel missense mutation in SLC5A2

encoding SGLT2 underlies autosomal-recessive renal glucosuria and aminoaciduria.

Kidney Int. 2005;67(1):34–41.

49. Francis J, Zhang J, Farhi A, Carey H, Geller DS. A novel SGLT2 mutation in a patient

with autosomal recessive renal glucosuria. Nephrol. Dial. Transplant. 2004

Nov.;19(11):2893–2895.

50. Kleta R, Stuart C, Gill FA, Gahl WA. Renal glucosuria due to SGLT2 mutations. Mol

Genet Metab. 2004 May;82(1):56–58.

51. Calado J, Soto K, Clemente C, Correia P, Rueff J. Novel compound heterozygous

mutations in SLC5A2 are responsible for autosomal recessive renal glucosuria. Hum.

Genet. 2004 Feb.;114(3):314–316.

52. den Heuvel van LP, Assink K, Willemsen M, Monnens L. Autosomal recessive renal

glucosuria attributable to a mutation in the sodium glucose cotransporter (SGLT2).

Hum Genet. 2002 Dec. 1;111(6):544–547.

53. Cho B-S, Kim S-D. School urinalysis screening in Korea. Nephrology (Carlton). 2007

Dec. 1;12 Suppl 3:S3–7.

144

54. Katsuno K, Fujimori Y, Takemura Y, Hiratochi M, Itoh F, Komatsu Y, et al.

Sergliflozin, a novel selective inhibitor of low-affinity sodium glucose cotransporter

(SGLT2), validates the critical role of SGLT2 in renal glucose reabsorption and

modulates plasma glucose level. J Pharmacol Exp Ther. 2007;320(1):323–330.

55. Meng W, Ellsworth BA, Nirschl AA, McCann PJ, Patel M, Girotra RN, et al. Discovery

of dapagliflozin: a potent, selective renal sodium-dependent glucose cotransporter 2

(SGLT2) inhibitor for the treatment of type 2 diabetes. J Med Chem. 2008 Mar.

13;51(5):1145–1149.

56. Hussey EK, Clark RV, Amin DM, Kipnes MS, O'Connor-Semmes RL, O'Driscoll EC,

et al. Single-dose pharmacokinetics and pharmacodynamics of sergliflozin etabonate, a

novel inhibitor of glucose reabsorption, in healthy volunteers and patients with type 2

diabetes mellitus. J Clin Pharmacol. 2010 Jun. 1;50(6):623–635.

57. Hussey EK, Dobbins RL, Stoltz RR, Stockman NL, O'Connor-Semmes RL, Kapur A, et

al. Multiple-dose pharmacokinetics and pharmacodynamics of sergliflozin etabonate, a

novel inhibitor of glucose reabsorption, in healthy overweight and obese subjects: a

randomized double-blind study. J Clin Pharmacol. 2010 Jun. 1;50(6):636–646.

58. Marsenic O. Glucose control by the kidney: an emerging target in diabetes. Am J

Kidney Dis. 2009 May 1;53(5):875–883.

59. Trial watch: SGLT2 inhibitor shows promise in type 2 diabetes. Nat Rev Drug Discov.

2010 Mar.;:182.

60. Kong CT, Yet SF, Lever JE. Cloning and expression of a mammalian Na+/amino acid

cotransporter with sequence similarity to Na+/glucose cotransporters. J Biol Chem. 1993

Jan. 25;268(3):1509–1512.

61. Mackenzie B, Panayotova-Heiermann M, Loo DD, Lever JE, Wright EM. SAATl Is a

Low Affinity Na+/ Glucose Cotransporter and Not an Amino Acid Transporter. J Biol

Chem. 1994 Sep. p. 22488–22491.

62. Mackenzie B, Loo DD, Panayotova-Heiermann M, Wright EM. Biophysical

characteristics of the pig kidney Na+/glucose cotransporter SGLT2 reveal a common

mechanism for SGLT1 and SGLT2. J Biol Chem. 1996 Dec. 20;271(51):32678–32683.

63. Díez-Sampedro A, Lostao MP, Wright EM, Hirayama BA. Glycoside binding and

translocation in Na+-dependent glucose cotransporters: comparison of SGLT1 and

SGLT3. J Membr Biol. 2000 Jul. 15;176(2):111–117.

64. Díez-Sampedro A, Eskandari S, Wright EM, Hirayama BA. Na+-to-sugar stoichiometry

of SGLT3. Am J Physiol Renal Physiol. 2001 Feb. 1;280(2):F278–82.

65. Voss AA, Díez-Sampedro A, Hirayama BA, Loo DDF, Wright EM. Imino sugars are

potent agonists of the human glucose sensor SGLT3. Mol Pharmacol. 2007 Feb.

1;71(2):628–634.

145

66. Diez-Sampedro A, Hirayama BA, Osswald C, Gorboulev V, Baumgarten K, Volk C, et

al. A glucose sensor hiding in a family of transporters. Proc Natl Acad Sci USA. 2003

Sep. 30;100(20):11753–11758.

67. Bianchi L, Díez-Sampedro A. A single amino acid change converts the sugar sensor

SGLT3 into a sugar transporter. PLoS ONE. 2010;5(4):e10241.

68. Faham S, Watanabe A, Besserer GM, Cascio D, Specht A, Hirayama BA, et al. The

crystal structure of a sodium galactose transporter reveals mechanistic insights into

Na+/sugar symport. Science. 2008 Aug. 8;321(5890):810–814.

69. Tazawa S, Yamato T, Fujikura H, Hiratochi M, Itoh F, Tomae M, et al.

SLC5A9/SGLT4, a new Na+-dependent glucose transporter, is an essential transporter

for mannose, 1,5-anhydro-D-glucitol, and fructose. Life Sci. 2005 Jan. 14;76(9):1039–

1050.

70. Parent L, Supplisson S, Loo DD, Wright EM. Electrogenic properties of the cloned

Na+/glucose cotransporter: II. A transport model under nonrapid equilibrium conditions.

J Membr Biol. 1992;125(1):63–79.

71. Wright EM, Hirayama BA, Loo DF. Active sugar transport in health and disease. J.

Intern. Med. 2007 Jan.;261(1):32–43.

72. Moran A, Davis LJ, Turner RJ. High affinity phlorizin binding to the LLC-PK1 cells

exhibits a sodium:phlorizin stoichiometry of 2:1. J Biol Chem. 1988 Jan. 5;263(1):187–

192.

73. Chen XZ, Coady MJ, Jalal F, Wallendorff B, Lapointe JY. Sodium leak pathway and

substrate binding order in the Na+-glucose cotransporter. Biophys J. 1997 Nov.

1;73(5):2503–2510.

74. Oulianova N, Falk S, Berteloot A. Two-step mechanism of phlorizin binding to the

SGLT1 protein in the kidney. J Membr Biol. 2001 Feb. 1;179(3):223–242.

75. Kimmich GA. Membrane potentials and the mechanism of intestinal Na+-dependent

sugar transport. J Membr Biol. 1990 Mar.;114(1):1–27.

76. Berteloot A. Kinetic Mechanism of Na+ -Glucose Cotransport through the Rabbit

Intestinal SGLT1 Protein. Journal of Membrane Biology. 2003 Apr. 1;192(2):89–100.

77. Gagnon DG, Frindel C, Lapointe J-Y. Effect of substrate on the pre-steady-state

kinetics of the Na+/glucose cotransporter. Biophys J. 2007 Jan. 15;92(2):461–472.

78. Longpré J-P, Gagnon DG, Coady MJ, Lapointe J-Y. The actual ionic nature of the leak

current through the Na+/glucose cotransporter SGLT1. Biophys J. 2010 Jan.

20;98(2):231–239.

79. Charron FM, Blanchard MG, Lapointe J-Y. Intracellular hypertonicity is responsible for

water flux associated with Na+/glucose cotransport. Biophys J. 2006 May

146

15;90(10):3546–3554.

80. Sauer GA, Nagel G, Koepsell H, Bamberg E, Hartung K. Voltage and substrate

dependence of the inverse transport mode of the rabbit Na+/glucose cotransporter

(SGLT1). FEBS Lett. 2000 Mar. 3;469(1):98–100.

81. Quick M, Tomasevic J, Wright EM. Functional asymmetry of the human Na+/glucose

transporter (hSGLT1) in bacterial membrane vesicles. Biochemistry. 2003 Aug.

5;42(30):9147–9152.

82. Eskandari S, Wright EM, Loo DDF. Kinetics of the reverse mode of the Na+/glucose

cotransporter. J Membr Biol. 2005 Mar.;204(1):23–32.

83. Loo DD, Hazama A, Supplisson S, Turk E, Wright EM. Relaxation kinetics of the

Na+/glucose cotransporter. Proc Natl Acad Sci USA. 1993 Jun. 15;90(12):5767–5771.

84. Loo DDF, Hirayama BA, Cha A, Bezanilla F, Wright EM. Perturbation analysis of the

voltage-sensitive conformational changes of the Na+/glucose cotransporter. J Gen

Physiol. 2005 Jan.;125(1):13–36.

85. Chen XZ, Coady MJ, Lapointe JY. Fast voltage clamp discloses a new component of

presteady-state currents from the Na+-glucose cotransporter. Biophys J. 1996 Nov.

1;71(5):2544–2552.

86. Krofchick D, Silverman M. Investigating the conformational states of the rabbit

Na+/glucose cotransporter. Biophys J. 2003 Jun.;84(6):3690–3702.

87. Kwon HM, Yamauchi A, Uchida S, Preston AS, Garcia-Perez A, Burg MB, et al.

Cloning of the cDNa for a Na+/myo-inositol cotransporter, a hypertonicity stress

protein. J Biol Chem. 1992 Mar. 25;267(9):6297–6301.

88. Hitomi K, Tsukagoshi N. cDNA sequence for rkST1, a novel member of the sodium

ion-dependent glucose cotransporter family. Biochim Biophys Acta. 1994 Mar.

23;1190(2):469–472.

89. Roll P, Massacrier A, Pereira S, Robaglia-Schlupp A, Cau P, Szepetowski P. New

human sodium/glucose cotransporter gene (KST1): identification, characterization, and

mutation analysis in ICCA (infantile convulsions and choreoathetosis) and BFIC

(benign familial infantile convulsions) families. Gene. 2002 Feb. 20;285(1-2):141–148.

90. Beck FX, Schmolke M, Guder WG. Osmolytes. Curr Opin Nephrol Hypertens. 1992

Oct. 1;1(1):43–52.

91. York JD, Guo S, Odom AR, Spiegelberg BD, Stolz LE. An expanded view of inositol

signaling. Adv Enzyme Regul. 2001;41:57–71.

92. Levine J, Barak Y, Gonzalves M, Szor H, Elizur A, Kofman O, et al. Double-blind,

controlled trial of inositol treatment of depression. AJP. American Psychiatric

Association; 1995 May 1;152(5):792–794.

147

93. Benjamin J, Levine J, Fux M, Aviv A, Levy D, Belmaker R. Double-blind, placebo-

controlled, crossover trial of inositol treatment for panic disorder. AJP. American

Psychiatric Association; 1995 Jul. 1;152(7):1084–1086.

94. Fux M, Levine J, Aviv A, Belmaker RH. Inositol treatment of obsessive-compulsive

disorder. Am J Psychiatry. 1996 Sep.;153(9):1219–1221.

95. Nestler JE, Jakubowicz DJ, Reamer P, Gunn RD, Allan G. Ovulatory and metabolic

effects of D-chiro-inositol in the polycystic ovary syndrome. N. Engl. J. Med. 1999

Apr. 29;340(17):1314–1320.

96. Berry GT, Mallee JJ, Kwon HM, Rim JS, Mulla WR, Muenke M, et al. The human

osmoregulatory Na+/myo-inositol cotransporter gene (SLC5A3): molecular cloning and

localization to chromosome 21. Genomics. 1995 Jan. 20;25(2):507–513.

97. Poppe R, Karbach U, Gambaryan S, Wiesinger H, Lutzenburg M, Kraemer M, et al.

Expression of the Na+ D Glucose Cotransporter SGLT1 in Neurons. J Neurochem.

1997;69:84–94.

98. Lahjouji K, Aouameur R, Bissonnette P, Coady MJ, Bichet DG, Lapointe J-Y.

Expression and functionality of the Na+/myo-inositol cotransporter SMIT2 in rabbit

kidney. Biochim Biophys Acta. 2007 May 1;1768(5):1154–1159.

99. Aouameur R, Da Cal S, Bissonnette P, Coady MJ, Lapointe J-Y. SMIT2 mediates all

myo-inositol uptake in apical membranes of rat small intestine. Am J Physiol

Gastrointest Liver Physiol. 2007 Dec.;293(6):G1300–7.

100. Berry GT, Wu S, Buccafusca R, Ren J, Gonzales LW, Ballard PL, et al. Loss of murine

Na+/myo-inositol cotransporter leads to brain myo-inositol depletion and central apnea.

J Biol Chem. 2003 May 16;278(20):18297–18302.

101. Chau JFL, Lee MK, Law JWS, Chung SK, Chung SSM. Sodium/myo-inositol

cotransporter-1 is essential for the development and function of the peripheral nerves.

FASEB J. 2005 Nov. 1;19(13):1887–1889.

102. Berry GT, Wang ZJ, Dreha SF, Finucane BM, Zimmerman RA. In vivo brain myo-

inositol levels in children with Down syndrome. J Pediatr. Elsevier; 1999 Jul.

1;135(1):94–97.

103. Willmroth F, Drieling T, Lamla U, Marcushen M, Wark H-J, van Calker D. Sodium-

myo-inositol co-transporter (SMIT-1) mRNA is increased in neutrophils of patients with

bipolar 1 disorder and down-regulated under treatment with mood stabilizers. Int J

Neuropsychopharmacol. 2007 Feb. 1;10(1):63–71.

104. van Calker D, Belmaker RH. The high affinity inositol transport system--implications

for the pathophysiology and treatment of bipolar disorder. Bipolar Disord. 2000 Jun.

1;2(2):102–107.

105. Coady MJ, Wallendorff B, Gagnon DG, Lapointe J-Y. Identification of a novel

148

Na+/myo-inositol cotransporter. J Biol Chem. 2002 Sep. 20;277(38):35219–35224.

106. Bourgeois F, Coady MJ, Lapointe J-Y. Determination of transport stoichiometry for two

cation-coupled myo-inositol cotransporters: SMIT2 and HMIT. J Physiol (Lond). 2005

Mar. 1;563(Pt 2):333–343.

107. Hager K, Hazama A, Kwon HM, Loo DD, Handler JS, Wright EM. Kinetics and

specificity of the renal Na+/myo-inositol cotransporter expressed in Xenopus oocytes. J

Membr Biol. 1995;143(2):103–113.

108. Ganapathy V, Thangaraju M, Gopal E, Martin PM, Itagaki S, Miyauchi S, et al.

Sodium-coupled monocarboxylate transporters in normal tissues and in cancer. AAPS J.

2008;10(1):193–199.

109. Miyauchi S, Gopal E, Fei Y-J, Ganapathy V. Functional identification of SLC5A8, a

tumor suppressor down-regulated in colon cancer, as a Na+-coupled transporter for

short-chain fatty acids. J Biol Chem. 2004 Apr. 2;279(14):13293–13296.

110. Coady MJ, Chang M-H, Charron FM, Plata C, Wallendorff B, Sah JF, et al. The human

tumour suppressor gene SLC5A8 expresses a Na+-monocarboxylate cotransporter. J

Physiol (Lond). 2004 Jun. 15;557(Pt 3):719–731.

111. Coady MJ, Wallendorff B, Bourgeois F, Charron F, Lapointe J-Y. Establishing a

definitive stoichiometry for the Na+/monocarboxylate cotransporter SMCT1. Biophys J.

2007 Oct. 1;93(7):2325–2331.

112. Srinivas SR, Gopal E, Zhuang L, Itagaki S, Martin PM, Fei Y-J, et al. Cloning and

functional identification of slc5a12 as a sodium-coupled low-affinity transporter for

monocarboxylates (SMCT2). Biochem J. 2005 Dec. 15;392(Pt 3):655–664.

113. Gopal E, Umapathy NS, Martin PM, Ananth S, Gnana-Prakasam JP, Becker H, et al.

Cloning and functional characterization of human SMCT2 (SLC5A12) and expression

pattern of the transporter in kidney. Biochim Biophys Acta. 2007 Nov.;1768(11):2690–

2697.

114. Plata C, Sussman CR, Sindic A, Liang JO, Mount DB, Josephs ZM, et al. Zebrafish

Slc5a12 encodes an electroneutral sodium monocarboxylate transporter (SMCTn). A

comparison with the electrogenic SMCT (SMCTe/Slc5a8). J Biol Chem. 2007 Apr.

20;282(16):11996–12009.

115. Wright EM, Turk E. The sodium/glucose cotransport family SLC5. Pflugers Arch. 2004

Feb.;447(5):510–518.

116. Coady MJ, Wallendorff B, Bourgeois F, Lapointe J-Y. Anionic leak currents through

the Na+/monocarboxylate cotransporter SMCT1. Am J Physiol, Cell Physiol. 2010

Jan.;298(1):C124–31.

117. Rodriguez A-M, Perron B, Lacroix L, Caillou B, Leblanc G, Schlumberger M, et al.

Identification and characterization of a putative human iodide transporter located at the

149

apical membrane of thyrocytes. J Clin Endocrinol Metab. 2002 Jul. 1;87(7):3500–3503.

118. Gopal E, Fei Y-J, Sugawara M, Miyauchi S, Zhuang L, Martin P, et al. Expression of

slc5a8 in kidney and its role in Na+-coupled transport of lactate. J Biol Chem. 2004 Oct.

22;279(43):44522–44532.

119. Frank H, Gröger N, Diener M, Becker C, Braun T, Boettger T. Lactaturia and loss of

sodium-dependent lactate uptake in the colon of SLC5A8-deficient mice. J Biol Chem.

2008 Sep. 5;283(36):24729–24737.

120. Li H, Myeroff L, Smiraglia D, Romero MF, Pretlow TP, Kasturi L, et al. SLC5A8, a

sodium transporter, is a tumor suppressor gene silenced by methylation in human colon

aberrant crypt foci and cancers. Proc Natl Acad Sci USA. 2003 Jul. 8;100(14):8412–

8417.

121. Ganapathy V, Gopal E, Miyauchi S, Prasad PD. Biological functions of SLC5A8, a

candidate tumour suppressor. Biochem Soc Trans. 2005 Feb. 1;33(Pt 1):237–240.

122. Thangaraju M, Gopal E, Martin PM, Ananth S, Smith SB, Prasad PD, et al. SLC5A8

triggers tumor cell apoptosis through pyruvate-dependent inhibition of histone

deacetylases. Cancer Res. 2006 Dec. 15;66(24):11560–11564.

123. Gupta N, Martin PM, Prasad PD, Ganapathy V. SLC5A8 (SMCT1)-mediated transport

of butyrate forms the basis for the tumor suppressive function of the transporter. Life

Sci. 2006 Apr. 18;78(21):2419–2425.

124. Iwamoto H, Blakely RD, De Felice LJ. Na+, Cl

−, and pH dependence of the human

choline transporter (hCHT) in Xenopus oocytes: the proton inactivation hypothesis of

hCHT in synaptic vesicles. J Neurosci. 2006 Sep. 27;26(39):9851–9859.

125. Okuda T, Haga T, Kanai Y, Endou H, Ishihara T, Katsura I. Identification and

characterization of the high-affinity choline transporter. Nat Neurosci. 2000 Feb.

1;3(2):120–125.

126. Ferguson SM, Savchenko V, Apparsundaram S, Zwick M, Wright J, Heilman CJ, et al.

Vesicular localization and activity-dependent trafficking of presynaptic choline

transporters. J Neurosci. 2003 Oct. 29;23(30):9697–9709.

127. Ferguson SM, Blakely RD. The choline transporter resurfaces: new roles for synaptic

vesicles? Mol Interv. 2004 Feb. 1;4(1):22–37.

128. Okuda T, Haga T. Functional characterization of the human high-affinity choline

transporter. FEBS Lett. 2000 Nov. 3;484(2):92–97.

129. Apparsundaram S, Ferguson SM, George AL, Blakely RD. Molecular cloning of a

human, hemicholinium-3-sensitive choline transporter. Biochem Biophys Res

Commun. 2000 Oct. 5;276(3):862–867.

130. Smanik PA, Liu Q, Furminger TL, Ryu K, Xing S, Mazzaferri EL, et al. Cloning of the

150

human sodium lodide symporter. Biochem Biophys Res Commun. 1996 Sep.

13;226(2):339–345.

131. Riedel C, Dohán O, la Vieja De A, Ginter CS, Carrasco N. Journey of the iodide

transporter NIS: from its molecular identification to its clinical role in cancer. Trends

Biochem Sci. 2001 Aug. 1;26(8):490–496.

132. Dohán O, la Vieja De A, Paroder V, Riedel C, Artani M, Reed M, et al. The

sodium/iodide Symporter (NIS): characterization, regulation, and medical significance.

Endocr Rev. 2003 Feb. 1;24(1):48–77.

133. Eskandari S, Loo DD, Dai G, Levy O, Wright EM, Carrasco N. Thyroid Na+/I

-

symporter. Mechanism, stoichiometry, and specificity. J Biol Chem. 1997 Oct.

24;272(43):27230–27238.

134. Levy O, Ginter CS, la Vieja De A, Levy D, Carrasco N. Identification of a structural

requirement for thyroid Na+/I

− symporter (NIS) function from analysis of a mutation

that causes human congenital hypothyroidism. FEBS Lett. 1998 Jun. 5;429(1):36–40.

135. Prasad PD, Wang H, Kekuda R, Fujita T, Fei YJ, Devoe LD, et al. Cloning and

functional expression of a cDNA encoding a mammalian sodium-dependent vitamin

transporter mediating the uptake of pantothenate, biotin, and lipoate. J Biol Chem. 1998

Mar. 27;273(13):7501–7506.

136. Wang H, Huang W, Fei YJ, Xia H, Yang-Feng TL, Leibach FH, et al. Human placental

Na+-dependent multivitamin transporter. Cloning, functional expression, gene structure,

and chromosomal localization. J Biol Chem. 1999 May 21;274(21):14875–14883.

137. Balamurugan K, Ortiz A, Said HM. Biotin uptake by human intestinal and liver

epithelial cells: role of the SMVT system. Am J Physiol Gastrointest Liver Physiol.

2003 Jul. 1;285(1):G73–7.

138. Pajor AM. Sequence of a putative transporter from rabbit kidney related to the

Na+/glucose cotransporter gene family. Biochim Biophys Acta. 1994 Sep.

14;1194(2):349–351.

139. Zhao F-Q, Zheng Y-C, Wall EH, McFadden TB. Cloning and expression of bovine

sodium/glucose cotransporters. J. Dairy Sci. 2005 Jan.;88(1):182–194.

140. Gilbert ER, Li H, Emmerson DA, Webb KE, Wong EA. Developmental regulation of

nutrient transporter and enzyme mRNA abundance in the small intestine of broilers.

Poult Sci. 2007 Aug. 1;86(8):1739–1753.

141. Xie Z, Turk E, Wright EM. Characterization of the Vibrio parahaemolyticus

Na+/Glucose cotransporter. A bacterial member of the sodium/glucose transporter

(SGLT) family. J Biol Chem. 2000 Aug. 25;275(34):25959–25964.

142. Leung DW, Turk E, Kim O, Wright EM. Functional expression of the Vibrio

parahaemolyticus Na+/galactose (vSGLT) cotransporter in Xenopus laevis oocytes. J

151

Membr Biol. 2002 May 1;187(1):65–70.

143. Hediger MA, Turk E, Wright EM. Homology of the human intestinal Na+/glucose and

Escherichia coli Na+/proline cotransporters. Proc Natl Acad Sci USA. 1989

Aug.;86(15):5748–5752.

144. Schwan WR, Coulter SN, Ng EY, Langhorne MH, Ritchie HD, Brody LL, et al.

Identification and characterization of the PutP proline permease that contributes to in

vivo survival of Staphylococcus aureus in animal models. Infect Immun. 1998 Feb.

1;66(2):567–572.

145. Bayer AS, Coulter SN, Stover CK, Schwan WR. Impact of the high-affinity proline

permease gene (putP) on the virulence of Staphylococcus aureus in experimental

endocarditis. Infect Immun. 1999 Feb. 1;67(2):740–744.

146. Olkhova E, Raba M, Bracher S, Hilger D, Jung H. Homology model of the Na+/proline

transporter PutP of Escherichia coli and its functional implications. J. Mol. Biol. 2011

Feb. 11;406(1):59–74.

147. Jackowski S, Alix JH. Cloning, sequence, and expression of the pantothenate permease

(panF) gene of Escherichia coli. J. Bacteriol. 1990 Jul.;172(7):3842–3848.

148. Reizer J, Reizer A, Saier MH. The Na+/pantothenate symporter (PanF) of Escherichia

coli is homologous to the Na+/proline symporter (PutP) of E. coli and the Na

+/glucose

symporters of mammals. Res Microbiol. 1990 Oct.;141(9):1069–1072.

149. Murakami S, Nakashima R, Yamashita E, Yamaguchi A. Crystal structure of bacterial

multidrug efflux transporter AcrB. Nature. 2002 Oct. 10;419(6907):587–593.

150. Chen Y-J, Pornillos O, Lieu S, Ma C, Chen AP, Chang G. X-ray structure of EmrE

supports dual topology model. Proc Natl Acad Sci USA. 2007 Nov. 27;104(48):18999–

19004.

151. Bowie JU. Flip-flopping membrane proteins. Nat Struct Mol Biol. 2006 Feb. 1;:94–96.

152. Rapp M, Seppälä S, Granseth E, Heijne von G. Emulating membrane protein evolution

by rational design. Science. 2007 Mar. 2;315(5816):1282–1284.

153. Yamashita A, Singh SK, Kawate T, Jin Y, Gouaux E. Crystal structure of a bacterial

homologue of Na+/Cl

−-dependent neurotransmitter transporters. Nature. 2005 Sep.

8;437(7056):215–223.

154. Weyand S, Shimamura T, Yajima S, Suzuki S, Mirza O, Krusong K, et al. Structure and

molecular mechanism of a nucleobase-cation-symport-1 family transporter. Science.

2008 Oct. 31;322(5902):709–713.

155. Ressl S, Terwisscha van Scheltinga AC, Vonrhein C, Ott V, Ziegler C. Molecular basis

of transport and regulation in the Na+/betaine symporter BetP. Nature. 2009 Mar.

5;458(7234):47–52.

152

156. Gao X, Lu F, Zhou L, Dang S, Sun L, Li X, et al. Structure and mechanism of an amino

acid antiporter. Science. 2009 Jun. 19;324(5934):1565–1568.

157. Fang Y, Jayaram H, Shane T, Kolmakova-Partensky L, Wu F, Williams C, et al.

Structure of a prokaryotic virtual proton pump at 3.2 Å resolution. Nature. 2009 Aug.

20;460(7258):1040–1043.

158. Shaffer PL, Goehring A, Shankaranarayanan A, Gouaux E. Structure and mechanism of

a Na+-independent amino acid transporter. Science. 2009 Aug. 21;325(5943):1010–

1014.

159. Tang L, Bai L, Wang W-H, Jiang T. Crystal structure of the carnitine transporter and

insights into the antiport mechanism. Nat Struct Mol Biol. 2010 Apr. 1;17(4):492–496.

160. Sennhauser G, Bukowska MA, Briand C, Grütter MG. Crystal structure of the

multidrug exporter MexB from Pseudomonas aeruginosa. J Mol Biol. 2009 May

29;389(1):134–145.

161. Hunte C, Screpanti E, Venturi M, Rimon A, Padan E, Michel H. Structure of a Na+/H

+

antiporter and insights into mechanism of action and regulation by pH. Nature. Nature

Publishing Group; 2005 Jun. 30;435(7046):1197–1202.

162. Yernool D, Boudker O, Jin Y, Gouaux E. Structure of a glutamate transporter

homologue from Pyrococcus horikoshii. Nature. 2004 Oct. 14;431(7010):811–818.

163. Singh SK, Piscitelli CL, Yamashita A, Gouaux E. A competitive inhibitor traps LeuT in

an open-to-out conformation. Science. 2008 Dec. 12;322(5908):1655–1661.

164. Singh SK, Yamashita A, Gouaux E. Antidepressant binding site in a bacterial

homologue of neurotransmitter transporters. Nature. 2007 Aug. 23;448(7156):952–956.

165. Zhou Z, Zhen J, Karpowich NK, Goetz RM, Law CJ, Reith MEA, et al. LeuT-

desipramine structure reveals how antidepressants block neurotransmitter reuptake.

Science. 2007 Sep. 7;317(5843):1390–1393.

166. Lolkema JS, Slotboom D-J. The major amino acid transporter superfamily has a similar

core structure as Na+-galactose and Na

+-leucine transporters. Mol Membr Biol. 2008

Sep.;25(6-7):567–570.

167. Gao X, Zhou L, Jiao X, Lu F, Yan C, Zeng X, et al. Mechanism of substrate recognition

and transport by an amino acid antiporter. Nature. 2010 Feb. 11;463(7282):828–832.

168. Abramson J, Wright EM. Structure and function of Na+-symporters with inverted

repeats. Curr Opin Struct Biol. 2009 Aug. 1;19(4):425–432.

169. Sujatha MS, Balaji PV. Identification of common structural features of binding sites in

galactose-specific proteins. Proteins. 2004 Apr. 1;55(1):44–65.

170. Walle T, Walle UK. The beta-D-glucoside and sodium-dependent glucose transporter 1

153

(SGLT1)-inhibitor phloridzin is transported by both SGLT1 and multidrug resistance-

associated proteins 1/2. Drug Metab Dispos. 2003 Nov. 1;31(11):1288–1291.

171. Panayotova-Heiermann M, Loo DD, Kong CT, Lever JE, Wright EM. Sugar binding to

Na+/glucose cotransporters is determined by the carboxyl-terminal half of the protein. J

Biol Chem. 1996 Apr. 26;271(17):10029–10034.

172. Shaffer PL, Goehring A, Shankaranarayanan A, Gouaux E. SOM: Structure and

mechanism of a Na+-independent amino acid transporter. Science. 2009 Aug.

21;325(5943):1010–1014.

173. Diallinas G. Biochemistry. An almost-complete movie. Science. 2008 Dec.

12;322(5908):1644–1645.

174. Parent L, Supplisson S, Loo DD, Wright EM. Electrogenic properties of the cloned

Na+/glucose cotransporter: I. Voltage-clamp studies. J Membr Biol. 1992;125(1):49–62.

175. Loo DDF, Hirayama BA, Sala-Rabanal M, Wright EM. How drugs interact with

transporters: SGLT1 as a model. J Membr Biol. 2008 May 1;223(2):87–106.

176. Hilgemann DW. Channel-like function of the Na,K pump probed at microsecond

resolution in giant membrane patches. Science. 1994 Mar. 11;263(5152):1429–1432.

177. Holmgren M, Wagg J, Bezanilla F, Rakowski RF, De Weer P, Gadsby DC. Three

distinct and sequential steps in the release of sodium ions by the Na+/K

+-ATPase.

Nature. 2000 Feb. 24;403(6772):898–901.

178. Hirayama BA, Díez-Sampedro A, Wright EM. Common mechanisms of inhibition for

the Na+/glucose (hSGLT1) and Na

+/Cl

−/GABA (hGAT1) cotransporters. Br J

Pharmacol. 2001 Oct. 1;134(3):484–495.

179. Panayotova-Heiermann M, Loo DD, Lostao MP, Wright EM. Sodium/D-glucose

cotransporter charge movements involve polar residues. J Biol Chem. 1994 Aug.

19;269(33):21016–21020.

180. Mager S, Naeve J, Quick M, Labarca C, Davidson N, Lester HA. Steady states, charge

movements, and rates for a cloned GABA transporter expressed in Xenopus oocytes.

Neuron. 1993 Feb.;10(2):177–188.

181. Mager S, Min C, Henry DJ, Chavkin C, Hoffman BJ, Davidson N, et al. Conducting

states of a mammalian serotonin transporter. Neuron. 1994 Apr.;12(4):845–859.

182. Wadiche JI, Arriza JL, Amara SG, Kavanaugh MP. Kinetics of a human glutamate

transporter. Neuron. 1995 May;14(5):1019–1027.

183. Boorer KJ, Loo DD, Wright EM. Steady-state and presteady-state kinetics of the

H+/hexose cotransporter (STP1) from Arabidopsis thaliana expressed in Xenopus

oocytes. J Biol Chem. 1994 Aug. 12;269(32):20417–20424.

154

184. Mackenzie B, Loo DD, Fei Y, Liu WJ, Ganapathy V, Leibach FH, et al. Mechanisms of

the human intestinal H+-coupled oligopeptide transporter hPEPT1. J Biol Chem. 1996

Mar. 8;271(10):5430–5437.

185. Taglialatela M, Toro L, Stefani E. Novel voltage clamp to record small, fast currents

from ion channels expressed in Xenopus oocytes. Biophys J. 1992 Jan.;61(1):78–82.

186. Meinild A-K, Hirayama BA, Wright EM, Loo DDF. Fluorescence studies of ligand-

induced conformational changes of the Na+/glucose cotransporter. Biochemistry. 2002

Jan. 29;41(4):1250–1258.

187. Gorraitz E, Pastor-Anglada M, Lostao MP. Effects of Na+ and H

+ on steady-state and

presteady-state currents of the human concentrative nucleoside transporter 3 (hCNT3).

Pflugers Arch. 2010 May 22.

188. Huntley SA, Krofchick D, Silverman M. Position 170 of Rabbit Na+/glucose

cotransporter (rSGLT1) lies in the Na+ pathway; modulation of polarity/charge at this

site regulates charge transfer and carrier turnover. Biophys J. 2004 Jul.;87(1):295–310.

189. Huntley SA, Krofchick D, Silverman M. A glutamine to glutamate mutation at position

170 (Q170E) in the rabbit Na+/glucose cotransporter, rSGLT1, enhances binding

affinity for Na+. Biochemistry. 2006 Apr. 11;45(14):4653–4663.

190. Liu T, Krofchick D, Silverman M. Effects on conformational states of the rabbit

sodium/glucose cotransporter through modulation of polarity and charge at glutamine

457. Biophys J. 2009 Jan.;96(2):748–760.

191. Liu J, Kim K-H, London B, Morales MJ, Backx PH. Dissection of the voltage-activated

potassium outward currents in adult mouse ventricular myocytes: Ito,f, Ito,s, IK,slow1,

IK,slow2, and Iss. Basic Res. Cardiol. 2011 Mar.;106(2):189–204.

192. Vasilyev A, Khater K, Rakowski RF. Effect of extracellular pH on presteady-state and

steady-state current mediated by the Na+/K

+ pump. J Membr Biol. 2004 Mar.

15;198(2):65–76.

193. Watzke N, Bamberg E, Grewer C. Early intermediates in the transport cycle of the

neuronal excitatory amino acid carrier EAAC1. J Gen Physiol. 2001 Jun. 1;117(6):547–

562.

194. Sun H, Oudit GY, Ramirez RJ, Costantini D, Backx PH. The phosphoinositide 3-kinase

inhibitor LY294002 enhances cardiac myocyte contractility via a direct inhibition of

Ik,slow currents. Cardiovasc Res. 2004 Jun. 1;62(3):509–520.

195. Zhou J, Kodirov S, Murata M, Buckett PD, Nerbonne JM, Koren G. Regional

upregulation of Kv2.1-encoded current, IK,slow2, in Kv1DN mice is abolished by

crossbreeding with Kv2DN mice. Am. J. Physiol. Heart Circ. Physiol. 2003

Feb.;284(2):H491–500.

196. Zhou J, Jeron A, London B, Han X, Koren G. Characterization of a slowly inactivating

155

outward current in adult mouse ventricular myocytes. Circ. Res. 1998 Oct.

19;83(8):806–814.

197. Forster I, Hernando N, Biber J, Murer H. The voltage dependence of a cloned

mammalian renal type II Na+/Pi cotransporter (NaPi-2). J Gen Physiol. 1998 Jul.

1;112(1):1–18.

198. Rakowski RF. Charge movement by the Na/K pump in Xenopus oocytes. J Gen Physiol.

1993 Jan.;101(1):117–144.

199. Krofchick D, Huntley SA, Silverman M. Transition states of the high-affinity rabbit

Na+/glucose cotransporter SGLT1 as determined from measurement and analysis of

voltage-dependent charge movements. Am J Physiol, Cell Physiol. 2004

Jul.;287(1):C46–54.

200. Schmitt BM, Koepsell H. An improved method for real-time monitoring of membrane

capacitance in Xenopus laevis oocytes. Biophys J. 2002 Mar.;82(3):1345–1357.

201. Hirayama BA, Loo DDF, Díez-Sampedro A, Leung DW, Meinild A-K, Lai-Bing M, et

al. Sodium-dependent reorganization of the sugar-binding site of SGLT1. Biochemistry.

2007 Nov. 20;46(46):13391–13406.

202. Quick M, Loo DD, Wright EM. Neutralization of a conserved amino acid residue in the

human Na+/glucose transporter (hSGLT1) generates a glucose-gated H

+ channel. J Biol

Chem. 2001 Jan. 19;276(3):1728–1734.

203. Lo B, Silverman M. Replacement of Ala-166 with cysteine in the high affinity rabbit

sodium/glucose transporter alters transport kinetics and allows methanethiosulfonate

ethylamine to inhibit transporter function. J Biol Chem. 1998 Jan. 9;273(2):903–909.

204. Gonzales AL, Lee W, Spencer SR, Oropeza RA, Chapman JV, Ku JY, et al. Turnover

rate of the gamma-aminobutyric acid transporter GAT1. J Membr Biol. 2007

Dec.;220(1-3):33–51.

205. Li M, Lester HA. Early fluorescence signals detect transitions at mammalian serotonin

transporters. Biophys J. 2002 Jul.;83(1):206–218.

206. Garcia-Celma JJ, Smirnova IN, Kaback HR, Fendler K. Electrophysiological

characterization of LacY. Proc Natl Acad Sci USA. 2009 May 5;106(18):7373–7378.

207. Uldry M, Ibberson M, Horisberger JD, Chatton JY, Riederer BM, Thorens B.

Identification of a mammalian H+-myo-inositol symporter expressed predominantly in

the brain. EMBO J. 2001 Aug. 15;20(16):4467–4477.

208. Kanai Y, Nussberger S, Romero MF, Boron WF, Hebert SC, Hediger MA. Electrogenic

properties of the epithelial and neuronal high affinity glutamate transporter. J Biol

Chem. 1995 Jul. 14;270(28):16561–16568.

209. Larsson HP, Tzingounis AV, Koch HP, Kavanaugh MP. Fluorometric measurements of

156

conformational changes in glutamate transporters. Proc Natl Acad Sci USA. 2004 Mar.

16;101(11):3951–3956.

210. Geck P, Heinz E. Coupling in secondary transport. Effect of electrical potentials on the

kinetics of ion linked co-transport. Biochim Biophys Acta. 1976 Aug. 4;443(1):49–63.

211. Stein WD. How the kinetic parameters of the simple carrier are affected by an applied

voltage. Biochim Biophys Acta. 1977 Jun. 16;467(3):376–385.

212. Sanders D, Hansen UP, Gradmann D, Slayman CL. Generalized kinetic analysis of ion-

driven cotransport systems: a unified interpretation of selective ionic effects on

Michaelis parameters. J Membr Biol. 1984;77(2):123–152.

213. Läuger P, Jauch P. Microscopic description of voltage effects on ion-driven cotransport

systems. J Membr Biol. 1986;91(3):275–284.

214. Läuger P. Thermodynamic and kinetic properties of electrogenic ion pumps. Biochim

Biophys Acta. 1984 Sep. 3;779(3):307–341.

215. Hansen UP, Gradmann D, Sanders D, Slayman CL. Interpretation of current-voltage

relationships for “active” ion transport systems: I. Steady-state reaction-kinetic analysis

of class-I mechanisms. J Membr Biol. 1981;63(3):165–190.

216. Gradmann D, Hansen UP, Long WS, Slayman CL, Warncke J. Current-voltage

relationships for the plasma membrane and its principal electrogenic pump in

Neurospora crassa: I. Steady-state conditions. J Membr Biol. 1978 Mar. 20;39(4):333–

367.

217. Warncke J, Slayman CL. Metabolic modulation of stoichiometry in a proton pump.

Biochim Biophys Acta. 1980 Jul. 8;591(2):224–233.

218. Eisner DA, Lederer WJ. Characterization of the electrogenic sodium pump in cardiac

Purkinje fibres. J Physiol (Lond). 1980 Jun.;303:441–474.

219. Mandel LJ, Curran PF. Response of the frog skin to steady-state voltage clamping. II.

The active pathway. J Gen Physiol. 1973 Jul.;62(1):1–24.

220. Stark G. Rectification phenomena in carrier-mediated ion transport. Biochim Biophys

Acta. 1973 Mar. 16;298(2):323–332.

221. Kimmich GA, Randles J, Restrepo D, Montrose M. The potential dependence of the

intestinal Na+-dependent sugar transporter. Ann N Y Acad Sci. 1985;456:63–76.

222. Lapointe JY, Hudson RL, Schultz SG. Current-voltage relations of sodium-coupled

sugar transport across the apical membrane of Necturus small intestine. J Membr Biol.

1986;93(3):205–219.

223. Neher E, Sakmann B, Steinbach JH. The extracellular patch clamp: a method for

resolving currents through individual open channels in biological membranes. Pflugers

157

Arch. 1978 Jul. 18;375(2):219–228.

224. Gadsby DC, Kimura J, Noma A. Voltage dependence of Na/K pump current in isolated

heart cells. Nature. 1985 Apr.;315(6014):63–65.

225. Umbach JA, Coady MJ, Wright EM. Intestinal Na+/glucose cotransporter expressed in

Xenopus oocytes is electrogenic. Biophys J. 1990 Jun. 1;57(6):1217–1224.

226. Lu CC, Hilgemann DW. GAT1 (GABA:Na+:Cl

−) cotransport function. Steady state

studies in giant Xenopus oocyte membrane patches. J Gen Physiol. 1999

Sep.;114(3):429–444.

227. Fei YJ, Kanai Y, Nussberger S, Ganapathy V, Leibach FH, Romero MF, et al.

Expression cloning of a mammalian proton-coupled oligopeptide transporter. Nature.

1994 Apr. 7;368(6471):563–566.

228. Boorer KJ, Frommer WB, Bush DR, Kreman M, Loo DD, Wright EM. Kinetics and

specificity of a H+/amino acid transporter from Arabidopsis thaliana. J Biol Chem.

1996 Jan. 26;271(4):2213–2220.

229. Klamo EM, Drew ME, Landfear SM, Kavanaugh MP. Kinetics and stoichiometry of a

proton/myo-inositol cotransporter. J Biol Chem. 1996 Jun. 21;271(25):14937–14943.

230. Boorer KJ, Loo DD, Frommer WB, Wright EM. Transport mechanism of the cloned

potato H+/sucrose cotransporter StSUT1. J Biol Chem. 1996 Oct. 11;271(41):25139–

25144.

231. Forster IC, Wagner CA, Busch AE, Lang F, Biber J, Hernando N, et al.

Electrophysiological characterization of the flounder type II Na+/Pi cotransporter (NaPi-

5) expressed in Xenopus laevis oocytes. J Membr Biol. 1997 Nov. 1;160(1):9–25.

232. Virkki LV, Forster IC, Biber J, Murer H. Substrate interactions in the human type IIa

sodium-phosphate cotransporter (NaPi-IIa). Am J Physiol Renal Physiol. 2005

May;288(5):F969–81.

233. Nagata K, Hori N, Sato K, Ohta K, Tanaka H, Hiji Y. Cloning and functional

expression of an SGLT-1-like protein from the Xenopus laevis intestine. Am J Physiol.

1999 May 1;276(5 Pt 1):G1251–9.

234. Burckhardt BC, Steffgen J, Langheit D, Müller GA, Burckhardt G. Potential-dependent

steady-state kinetics of a dicarboxylate transporter cloned from winter flounder kidney.

Pflugers Arch. 2000 Dec.;441(2-3):323–330.

235. Fischer W-N, Loo DDF, Koch W, Ludewig U, Boorer KJ, Tegeder M, et al. Low and

high affinity amino acid H+-cotransporters for cellular import of neutral and charged

amino acids. Plant J. 2002 Mar. 1;29(6):717–731.

236. Larráyoz IM, Casado FJ, Pastor-Anglada M, Lostao MP. Electrophysiological

characterization of the human Na+/nucleoside cotransporter 1 (hCNT1) and role of

158

adenosine on hCNT1 function. J Biol Chem. 2004 Mar. 5;279(10):8999–9007.

237. Forster IC, Virkki L, Bossi E, Murer H, Biber J. Electrogenic kinetics of a mammalian

intestinal type IIb Na+/Pi cotransporter. J Membr Biol. 2006;212(3):177–190.

238. Larráyoz IM, Fernández-Nistal A, Garcés A, Gorraitz E, Lostao MP. Characterization

of the rat Na+/nucleoside cotransporter 2 and transport of nucleoside-derived drugs

using electrophysiological methods. Am J Physiol, Cell Physiol. 2006 Dec.

1;291(6):C1395–404.

239. Panayotova-Heiermann M, Loo DD, Wright EM. Kinetics of steady-state currents and

charge movements associated with the rat Na+/glucose cotransporter. J Biol Chem. 1995

Nov. 10;270(45):27099–27105.

240. Hirayama BA, Lostao MP, Panayotova-Heiermann M, Loo DD, Turk E, Wright EM.

Kinetic and specificity differences between rat, human, and rabbit Na+-glucose

cotransporters (SGLT-1). Am J Physiol. 1996 Jun.;270(6 Pt 1):G919–26.

241. Hazama A, Loo DD, Wright EM. Presteady-state currents of the rabbit Na+/glucose

cotransporter (SGLT1). J Membr Biol. 1997 Jan. 15;155(2):175–186.

242. Hirayama BA, Loo DD, Wright EM. Cation effects on protein conformation and

transport in the Na+/glucose cotransporter. J Biol Chem. 1997 Jan. 24;272(4):2110–

2115.

243. Loo DDF, Hirayama BA, Karakossian MH, Meinild A-K, Wright EM. Conformational

dynamics of hSGLT1 during Na+/glucose cotransport. J Gen Physiol. 2006 Dec.

1;128(6):701–720.

244. Gagnon DG, Frindel C, Lapointe J-Y. Voltage-clamp fluorometry in the local

environment of the C255-C511 disulfide bridge of the Na+/glucose cotransporter.

Biophys J. 2007 Apr. 1;92(7):2403–2411.

245. Hilgemann DW, Lu CC. GAT1 (GABA:Na+:Cl

−) cotransport function. Database

reconstruction with an alternating access model. J Gen Physiol. 1999 Sep.;114(3):459–

475.

246. Falk S, Guay A, Chenu C, Patil SD, Berteloot A. Reduction of an eight-state

mechanism of cotransport to a six-state model using a new computer program. Biophys

J. 1998 Feb.;74(2 Pt 1):816–830.

247. Falk S, Oulianova N, Berteloot A. Kinetic mechanisms of inhibitor binding: relevance

to the fast-acting slow-binding paradigm. Biophys J. 1999 Jul.;77(1):173–188.

248. Lieb WR, Stein WD. Testing and characterizing the simple carrier. Biochim Biophys

Acta. 1974 Dec. 10;373(2):178–196.

249. Bissonnette P, Noël J, Coady MJ, Lapointe JY. Functional expression of tagged human

Na+-glucose cotransporter in Xenopus laevis oocytes. J Physiol (Lond). 1999 Oct.

159

15;520 Pt 2:359–371.

250. Michaelis L, Menten ML. Die Kinetik der InwertinWirkung. Biochem., No. 49. (1913),

pp. 333-369. 1913;(49):333–369.

251. Cornish-Bowden A. Fundamentals of enzyme kinetics. Third edition. Portland Press;

2004. p. 422.