KTH ROYAL INSTITUTE OF TECHNOLOGY SCHOOL OF ENGINEERING SCIENCES
IN
DEGREE PROJECT ENGINEERING PHYSICS, SECOND CYCLE, 30 CREDITS
STOCKHOLM SWEDEN 2021,
Characterization of Alcohol
Modulation of a Pentameric Ligand- gated Ion Channel with
Electrophysiology and Molecular Dynamics Simulations
SABINA VICTORIA GUTHEIM
Characterization of Alcohol Modulation of a Pentameric Ligand-gated Ion Channel with Electrophysiology and Molecular Dynamics
Simulations
SABINA VICTORIA GUTHEIM
Degree Programme in Engineering Physics Date: June 7, 2021
Supervisors: Erik Lindahl, Rebecca Howard Examiner: Hjalmar Brismar
School of Engineering Sciences
Swedish title: Karakterisering av alkoholmodulering av en pentamerisk ligandstyrd jonkanal med elektrofysiologi och molekylärdynamiksimuleringar
Abstract | i
Abstract
Pentameric ligand-gated ion channels (pLGICs) are membrane receptors that play a crucial role in every living organism. The pLGIC protein structure forms a pore through the membrane of a cell that can let specific ions pass through, upon activation by endogenous agonists. pLGICs are allosterically modulated by ligands binding at allosteric sites, that either stabilize a certain conformation or change the binding affinity of the endogenous agonist.
However, much remains unknown about the exact way in which these modulators bind to and affect pLGICs. An increased understanding could help in the search for novel and/or more effective target drugs. With this masters thesis, I hope to contribute by investigating the modulatory effect of ethanol on the bacterial Gloeobacter ligand-gated ion channel (GLIC).
This has been done by performing oocyte electrophysiology recordings and analysis of molecular dynamics simulations, both with and without ethanol, and of four separate variants of GLIC that are either potentiated or inhibited by ethanol. Two possible allosteric sites were discovered in a transmembrane intrasubunit pocket: a potentiating allosteric site close to the M2 helix and residue V242, as well as an inhibitory membrane- and M4 helix-close intrasubunit site. Finally, evidence was found that could support a previously suggested inhibitory allosteric site in the pore around the 9’ hydrophobic gate.
Keywords
Biophysics, Electrophysiology, MD simulations, Ligand-gated ion channels, GLIC, Ethanol modulation
Sammanfattning | iii
Sammanfattning
Pentameriska ligandstyrda jonkanaler (pLGICs) är membranreceptorer som utgör vitala delar av varje levande organism. pLGICs proteinstruktur formar en por genom cellmembranet, som kan släppa igenom specifika joner efter aktivering av endogena agonister. pLGICs är allostermodulerade av ligander som binder vid allostera säten och som därigenom antingen stabiliserar en viss form eller förändrar den endogena agonistens bindningsstyrka. Emellertid saknas fortfarande mycket kunskap på detaljnivå om hur dessa modulatorer binder sig till och påverkar kanalerna. En ökad förståelse skulle hjälpa forskningen efter nya och/eller mer effektiva mediciner. Mitt examensarbete hoppas bidra genom att studera hur etanol modulerar den bakteriella ligandstyrda jonkanalen GLIC från Gloeobacter. Det har gjorts genom elektrofysiologimätningar på oocyter och analys av molekulärdynamik- simuleringar, båda av fyra olika GLIC-varianter, som antingen potentieras eller hämmas av etanol, och med eller utan etanol. Två allostera säten upptäcktes i det transmembrana intrasubenhetområdet: ett säte för potentiering nära M2 helixen och aminosyran V242, och ett hämmande säte nära membranet och helix M4. Slutligen hittades tecken som kan styrka existensen av det tidigare föreslagna hämmande allostera sätet i poren kring den hydrophoba porten.
Nyckelord
Biofysik, Elektrofysiologi, MD simuleringar, Pentameriska ligandstyrda jonkanaler, GLIC, Etanolmodulering
Acknowledgments | v
Acknowledgments
I would like to thank my supervisor Erik Lindahl for, firstly, peeking my interest in the subject of Biophysics and teaching it to me in one of the greatest courses I have had during my years at KTH. Secondly, and most importantly, I am grateful to have gotten the opportunity to do my masters thesis in his group. The last year has been a great experience, the best that I have had so far in a research group, and to a large extent it is also due to Rebecca Howard. Many thanks to Rebecca for guiding me through the project, always with clear communication and structure, and for challenging me and answering my many questions along the way. I have learnt a great deal from her during this time. In addition, I want to thank Yuxuan Zhuang for his help and answers to the computational and mechanistic parts of the project. His excellent knowledge has been a great resource, especially every time I have gotten stuck. Lastly, I want to thank the group as a whole, for inviting me in with open arms and for the great atmosphere.
Stockholm, June 2021 Sabina Victoria Gutheim
CONTENTS | vii
Contents
1 Introduction 1
1.1 pLGIC Structure . . . 2
1.2 The bacterial GLIC and the Current Knowledge of its Allosteric Modulation . . . 5
1.3 Useful Methods to Study Allosteric Modulation with. . . 7
1.3.1 TEVC . . . 7
1.3.2 MD Simulations . . . 7
1.4 Thesis Hypothesis and Structure . . . 8
2 Methods 9 2.1 Two Electrode Voltage Clamp Electrophysiology . . . 9
2.1.1 Xenopus LaevisOocyte Injection. . . 12
2.1.2 TEVC . . . 13
2.1.3 TEVC Data analysis . . . 14
2.2 MD Simulations . . . 15
2.2.1 Construction of the MD Simulations of GLIC . . . . 15
2.2.2 Simulation Analysis and Definitions . . . 16
3 Results and Analysis 21 3.1 Electrophysiology . . . 21
3.1.1 Ethanol Modulation . . . 21
3.1.2 Functional Dynamics . . . 26
3.2 MD Simulation Analysis . . . 27
3.2.1 CHAP Average Pore Radius Profiles . . . 30
3.2.2 RMSF Curves. . . 31
3.2.3 RMSD . . . 33
3.2.4 β -Expansion . . . 33
3.2.5 VMD Ethanol Density . . . 37 3.2.6 Pore Hydration and Ethanol Occupancy . . . 42 3.2.7 Intersubunit Cavity Hydration, Ethanol and Lipid
Occupancy . . . 43 3.2.8 Intrasubunit Cavity Hydration, Ethanol and Lipid
Occupancy . . . 45
4 Discussion 51
5 Conclusions and Future work 57
5.1 Conclusion . . . 57 5.2 Future Work . . . 57 5.2.1 Ethical Reflections . . . 58
References 59
A Supplemental Results 67
A.1 RMSDs . . . 67 A.2 Pore Hydration and Ethanol Occupancy . . . 67
LIST OF FIGURES | ix
List of Figures
1.1 The GLIC WT protein structure . . . 3
1.2 GLIC WT TMD structure from above with M1-M4, the pore, intersubunit cavity and the intrasubunit region marked . . . 4
2.1 TEVC electrophysiology setup schematic . . . 10
2.2 M2 helices and pore lining residues . . . 11
2.3 Xenopus laevisoocytes and injection needle . . . 12
2.4 GLIC variant pH curves . . . 15
2.5 Ethanol binding in the M2-close and the M4-close intrasubunit cavities . . . 20
3.1 TEVC GLIC variant sample traces . . . 22
3.2 The mean % alcohol modulation and SEM of GLIC WT, F14’L, I9’T and F14’L/I9’T for 600 mM, 200 mM and 60 mM EtOH in Ringer’s buffer . . . 24
3.3 The mean % alcohol modulation and SEM of GLIC WT, F14’L, I9’T and F14’L/I9’T for 60 mM EtOH in Ringer’s buffer 25 3.4 The mean leak current and SEM of GLIC WT, F14’L, I9’T and F14’L/I9’T . . . 27
3.5 GLIC MD structure with the mutated residues 233 and 238 . . 28
3.6 Snap-shot of the ethanol present in GLIC F14’L at the end of the simulation . . . 29
3.7 Average pore radius profiles with of GLIC F14’L, I9’T and F14’L/I9’T to WT . . . 32
3.8 apo vs. ethanol RMSF curves of the Cα residues. . . 34
3.9 Mutant vs. WT RMSF curves of the TMD region . . . 35
3.10 GLIC variant loop RMSDs . . . 36
3.11 GLIC variant β -expansion . . . 38
3.12 Visualization of the ethanol density at prime level 9 . . . 39
3.13 Visualization of the ethanol density at prime level 13 . . . 40
3.14 Visualization of the ethanol density at prime level 16 . . . 41 3.15 GLIC variant water and ethanol occupancy around the 9’ M2
level . . . 44 3.16 GLIC variant water, ethanol and lipid occupancy of the
intersubunit cavity. . . 46 3.17 GLIC variant water, ethanol and lipid occupancy of the M2-
close intrasubunit cavity . . . 48 3.18 GLIC variant water, ethanol and lipid occupancy of the M4-
close intrasubunit cavity . . . 49 4.1 The proposed mechanistic model for the allosteric modulation
of GLIC by alcohol . . . 53 A.1 The RMSD of the Cαs of a) the entire subunits and b) M2. c)
the Pro-loop RMSD . . . 68 A.2 Average number of a) water molecules and b) ethanol molecules
in the pore at the 13’ M2 level . . . 69 A.3 Average number of a) water molecules and b) ethanol molecules
in the pore at the 16’ M2 level . . . 70
LIST OF TABLES | xi
List of Tables
2.1 CDNA concentrations injected into GLIC . . . 13 3.1 The mean % alcohol modulation, SEM and number of oocytes
recorded on (N) of GLIC WT, F14’L, I9’T and F14’L/I9’T for 600 mM, 200 mM and 60 mM EtOH in Ringer’s buffer . . . . 23 3.2 Average and SEM oocyte leak currents . . . 26
List of acronyms and abbreviations | xiii
List of acronyms and abbreviations
EC50 half maximal effective concentration 4HFI pH 4.6 open state
4NPQ pH 7.0 closed state apo without ethanol
cDNA complementary DNA
cryo-EM cryo-electron microscopy ECD extracellular domain
F14’L F238L
F14’L/I9’T F238L/I233T
GLIC Gloeobacter ligand-gated ion channel GluCl glutamate-gated chloride channel GlyR Glycine receptor
I9’T I233T
ICD intracellular domain MD molecular dynamics
nAChR nicotinic acetylcholine receptor pLGIC pentameric ligand-gated ion channel POPC 1-palmitoyl-2-oleoyl phosphatidylcholine RMSD root mean square deviation
RMSF root mean square fluctuation SEM standard error of mean
StD standard deviation
TEVC two electrode voltage clamp TMD transmembrane domain WT wild type
Introduction | 1
Chapter 1 Introduction
The membrane receptor is an essential structure in any living organism and it acts as receiver and transducer of information to and from cells. All membrane receptors are formed by proteins and sit in the membrane lipid bilayer. However, the way in which they relay information can differ [1, pp. 9–10, 529–530]. One membrane receptor superfamily is thepentameric ligand-gated ion channel (pLGIC) [2], whose protein structure forms a pore through the membrane of the cell that can let specific ions pass through.
pLGICsare activated by endogenous agonists (i.e. ligands) that bind to the, so called, orthosteric binding sites. What this means is that, upon binding with the ligand, the protein structure can undergo a conformational change (i.e. opening the pore to ions), thereby, changing the membrane potential.
pLGICscan also be allosterically modulated by chemical compounds (called modulators) that bind at non-orthosteric sites, called allosteric sites. Binding there can stabilize one of the conformations, or change the sensitivity of the channel to ligands binding at the orthosteric site [3], by changing the energy landscape of the channel’s activation [4].
pLGICsare abundant in the nervous system and their malfunctioning can be connected to numerous neurological and psychological conditions. In addition, they are popular targets of drugs that either potentiate or inhibit the sensitivity of the channel’s activation [5]. The binding sites of these drugs, and the way in which they affect the channels, depend on the structure and dynamics of the drug and the specific channel. By increasing understanding of these pLGICstructures, binding sites and gating mechanisms, the nervous system can be better understood and novel and/or more effective target drugs can be developed.
With this masters thesis, I hope to contribute to this field. Specifically, this
work has focused on the interaction between alcohol and pLGIC. Alcohol is one of the most dangerous substances in terms of overall harm to users and others [6], [7]; contributions to the field could help in the search for specialized drugs to treat alcohol overdose. More importantly, alcohol can serve as a simple model molecule for the allosteric modulation of lipophilically modulated agents. One such example is general anaesthetics that anaesthetize patients by targeting pLGICs in the nervous system and brain. Thus, this work has the potential to increase knowledge of the effect of lipophilically modulated agents onpLGICsin general.
1.1 pLGIC Structure
AllpLGICstructures have a universal organization, despite the, in some cases, major amino acid sequence variability between the different channels [8]. The pLGICstructure consists of five identical, or homologous, protein monomers.
These are arranged in anextracellular domain (ECD)with the ligand binding domain (LBD), a transmembrane domain (TMD)that forms the membrane pore, and anintracellular domain (ICD)that differs in size and can be absent for some pLGICs. In both prokaryotic and eukaryotic pLGICs, each ECD monomer is folded into a β -sandwich. Moreover, each TMD monomer is organized into a bundle made out of four consecutive α-helices, named M1- M4. The M2 helices face the pore, surrounded by the M1s and M3s in a ring formation, while the M4s all sit towards the lipid bilayer [3]. The helices are visualized in Figure 1.2. In addition, all pLGICs have an activation period, during which the conformation changes into an open state. This is followed by a refractory period, where the channel is in a desensitized, non-conducting state but the ligand affinity is still high. Thereafter, the circle is closed by a closed or resting state that can be activated once more [9]. An example of a pLGIC structure, with its different parts highlighted, can be found in Figures1.1and2.2.
Advances in the field of X-ray crystallography andcryo-electron microscopy (cryo-EM), have lead to the discovery of new high-resolution, atomistic structures in recent years and have furthered the understanding of pLGICs’
gating mechanisms and overall structures [11]. Such structures, realized in all the different states and conditions, together with drugs like, in this case, alcohol, are needed for a more extensive understanding of a specific channel and its modulation, together with complementary experimental and computational methods. Naturally, the eukaryotic human channels are of highest interest to researchers and drug developers alike. However, the
Introduction | 3
TMD ECD
α-helix β-sheet
loop
loop C
loop 2 loop F
Figure 1.1 – The GLIC WT protein structure. Each subunit is colored differently and the ECD and TMD is marked, as well as the general appearance of α-helices, β -sheets and loops. In addition, three of the channel’s specific loops are pointed to. Made with VMD [10].
B A C M4
M1
M2
M3
Figure 1.2 – GLIC WT TMD from above with M1-M4, the pore, the intersubunit cavity and the intrasubunit region marked. The subunits are in different colors. The M1-M4 α-helices in a subunit are marked, as well as the A) the pore, B) the intersubunit cavity and C) the intrasubunit region.
Residues I233 (I9’) are visible facing the pore while the F238s (F14’s) face the area of the intersubunit cavity. Made with VMD [10].
Introduction | 5
complexity of their structures, and the difficulty to express them in larger quantities, means that not all of those states and conditions have yet been discovered, especially at a high resolution. This is where the study of prokaryotic homologs comes in.
1.2 The bacterial GLIC and the Current Knowledge of its Allosteric Modulation
The origin of eukaryotic pLGICs has been established as prokaryotic, and more than 20 bacterial and archea homologs have so far been found [12].
One such example is the bacterial Gloeobacter ligand-gated ion channel (GLIC), which can easily be expressed in Escherichia coli (E. coli) and has been co-crystallized with several ligands in various conditions, at a high resolution. GLIChas been argued to represent the minimal structure required for signal transduction. This is due to its compact, homopentameric structure that lacks the amino-terminal helix, canonical disulphide bridge and large ICD, usually found in eukaryotes [13]. GLIC is proton-activated, unlike its closest eukaryotic pLGIC homolog: the nicotinic acetylcholine receptor (nAChR), which is activated by the neurotransmitter Acetylcholine, amongst others. All the same, compared to nAChR, GLIC presents similar allosteric modulation by compounds such as general anaesthetics [13], and the same weakly potentiating effect by methanol and ethanol, and inhibition by larger n-alcohols [14]. Furthermore, GLIC has been concluded as a valid model system for allosteric modulation of other pLGICs [3], such as GABAA [15]
and α1 Glycine receptor (GlyR)[16], despite that these channels are anionic and usually highly potentiated by ethanol and larger n-alcohols [17].
Previous studies ofpLGICandGLICwith regard to alcohol modulation, have revealed several potential binding sites [14], all of which are marked A- C in Figure1.2. First and foremost, general anaesthetics binding in the center and middle of pore by the M2 pore-lining helices have been connected to inhibition of GLIC in closed state structures [18], [19]. Alcohol molecules are much smaller than general anaesthetics, but since both are lipophilic, it is possible that they modulate with the same underlying mechanism.
The alcohol 2-bromoethanol that inhibits the bacterial Erwinia chrysanthemi ligand-gated ion channel (ELIC), was found in the pore right below the center of the M2 helices in a closed state crystal structure [20]. This area is believed to be important for the gating mechanism ofGLICthat ensures the closure of the channel [21]. Hence, it is reasonable to believe that an inhibitory allosteric
site is present, either by blocking ions to pass through the channel in the open conformation or stabilizing the closed one.
Another region of interest is aTMDpocket in-between subunits, referred to as the intersubunit cavity. Ethanol has been shown to bind spontaneously there in computational models of GlyR based on GLIC [17], [22]. The same intersubunit site has been occupied by another modulator in crystal structures of glutamate-gated chloride channel (GluCl) [23]. Worth noting, is the polarity of one or more of theGluClresidues that line their intersubunit pocket, and that are non-polar in cationic channels like nAChR and GLIC.
Notwithstanding, in molecular simulations of bothGLIC wild type (WT)and mutated versions, alcohol was observed to occupy the intersubunit cavity, as well as an intrasubunit cavity (i.e. a TMD pocket inside each subunit).
Since the ethanol occupancy was higher in the intersubunit cavity for a highly ethanol-sensitive mutant of GLIC, the intersubunit pocket was proposed as the potentiating allosteric site, while the intrasubunit cavity was suggested to be inhibitory [24]. A similar conclusion was reached after analyzing the crystal structure of the same highly ethanol-sensitive mutant, co-crystallized with ethanol, and finding ethanol in the intersubunit cavity. Some ethanol was also found in the intrasubunit cavity, but at a too low resolution to model it with certainty [25].
The intrasubunit cavity has been connected to the potentiating allosteric modulation of GLIC by general anaesthetics, both in structures [26], electrophysiology [27] and with computational methods [26], [27]. Other data, instead support a theory of an inhibitory allosteric site in the intrasubunit region; bromoform, which inhibits GLIC at submillimolar concentrations, was found to occupy several intrasubunit sites inGLIC[19], in addition to the pore; octanol was found in the intrasubunit cavity ofnAChR, in-between the M1, M2 and M3 helices, through photo-labeling studies [28]. Since octanol is a longer n-alcohol, it inhibits the channel. The same site innAChRwas found occupied by propofol in a desensitized state in another study [29]. Whether the same allosteric site can be potentiating in one instance and inhibitory in the next, has currently not been ruled out in the field. For instance, this was suggested to be the case for an intersubunit allosteric site of a type of GABAA receptor [30]. Another possibility is that there are several allosteric sites within the intrasubunit region responsible for separate effects.
Further investigation of the roles of the intersubunit and intrasubunit pockets, as well as the center of the TMD pore, with regard to either potentiation or inhibition, is needed for a more definitive understanding of the modulatory effects of alcohol onGLICand, in extension,pLGICsin general.
Introduction | 7
1.3 Useful Methods to Study Allosteric Modulation with
Atomic structures of pLGICs, although informative, only serve as snapshots of the channels’ gating mechanisms. These snapshots are crucial for the structural understanding of the superfamily, with conclusive data of the key conformations the channels go through and ligand binding sites. However, in living organisms, pLGICs are not static, nor are the ligands that bind to them. Structures from X-ray crystallography and cryo-EMcan therefore be complemented by other techniques of study that give more information about the dynamics of the systems. Two such methods are two electrode voltage clamp (TEVC)electrophysiology andmolecular dynamics (MD)simulations.
1.3.1 TEVC
TEVCis a valuable experimental method in the study ofpLGICs, in addition to those for the extraction of the protein structures, mentioned above. DNA or RNA can be injected into cells like Xenopus laevis oocytes, to make them express a specific pLGIC. In TEVC, two microelectrodes are probed into the oocyte. One is to control the cell’s membrane potential, with the help of a voltage follower and a clamping amplifier. The other is to deliver current to keep the membrane potential intact. A schematic of the setup is visualized in Figure 2.1. As more of the channels open, more ions will flux across the membrane. The deflection of the whole cell current from baseline is then recorded, showing whole cell statistics of the response of a certain pLGICto a ligand [31].
1.3.2 MD Simulations
As computational capacity has advanced, so has the expansion of MD simulations applied to pLGICs and other biomolecular complexes. In short,MDsimulations can convey possible physical movements of molecules over time and indicate its interactions with its surroundings. These MD simulations are governed by the laws of physics and are generated through the use of extensive molecular mechanics force fields and energy calculations.
Although each trajectory is characterized by randomness, the overall trend and average behavior between replicates can still be significant and give valuable insights into the channel’s behavior. In addition, further restraining potentials to a specificpLGICcan be added, by translating the data from their
experimental crystal structures. Therefore, MD simulations have become a common, complementary tool to experimental techniques in the field of pLGICs, amongst others [32]. Specifically, simulation analysis can help explain the mechanism or behavior behind an experimental result of for instanceTEVCelectrophysiology.
1.4 Thesis Hypothesis and Structure
The main hypothesis of this thesis, is that alcohol’s potentiating allosteric effect on GLIC is mainly caused by a high occupancy of the intersubunit cavity, given most structures implicate its importance. Additionally, binding of alcohol by the hydrophobic gate was hypothesized as allosterically inhibitory. The varied evidence of the intrasubunit region, that indicates it as either inhibitory or potentiating, meant that its significance toGLICallostery would in addition be investigated.
To test the hypothesis,TEVCandMDsimulations were applied toGLIC and three of its mutants. One mutant targeted a residue lining the hydrophobic gate, which should then affect GLIC inhibition, assuming there is in fact an inhibitory allosteric site present. Another mutant’s mutation targeted a residue that faces the intersubunit pocket, similar to what Murail, Howard, Broemstrup, et al. [24] did but with a residue less different in size. The final mutant contained both of these mutations.
This masters thesis report is divided into a methods chapter, where the methodsTEVCandMDsimulations used, together with relevant settings and definitions, is presented. It is followed by a results and analysis section, where the outcome and analysis of both the experimental and computational data is given. Lastly, the results and their implications is discussed in the chapter called Discussions, followed by conclusions and future work. Supplemental results can be found in the Appendix for the interested reader.
Methods | 9
Chapter 2 Methods
2.1 Two Electrode Voltage Clamp Electrophysiology
TEVCis a powerful experimental tool to measure whole cell currents through pLGICs. The TEVC experimental setup, explained in the introduction, can be found in Figure2.1.
In the following experiments, the non-modifiedGLIC WTand threeGLIC mutants were injected into Xenopus laevis oocytes and the whole-cell currents recorded. The three mutants were GLIC F238L (F14’L), where the amino acid Phenylalanine (F) at residue number 238 in the M2 α-helix was replaced with a Leucine (L),GLIC I233T (I9’T)with a Threonine (T) residue at place 233 instead of a Isoleucine (I), and the double mutant GLIC F238L/I233T (F14’L/I9’T) with both alterations. Residues in the M2 α-helix are often described with prime notation instead of referencing the residue numbers.
Prime notation begins at ∼1’ at the N-terminal of the intracellular end of the M2 α-helix and ends at ∼20’ at the C-terminal extracellular end. The pore lining TMD M2 residues are marked with the prime notation in Figure2.2.
The reason for using this notation, is the ability to more easily compare between different channels that have similar structure but, for instance due to anICD, have shifted residue numbers in the M2 α-helix compared to one another. This is where the names of the mutations in parentheses come in; the GLICmutations areF14’L,I9’TandF14’L/I9’Twith prime notation. These will henceforth be used to describe both the specific mutation and theGLIC mutants with those mutations.
Worth noting are the different attributes of the amino acids at the 9’ and 14’ position. Leucine is a non-polar amino acid like Phenylalanine, but
Figure 2.1 – TEVC electrophysiology setup. The schematic shows an oocyte in the bath, the electrodes that clamp it, amplifiers, KCL bridges etc.
Illustration made by Urška Rovšnik, reproduced with permission.
Methods | 11
16'
13'
9'
6'
2'
-2'
Figure 2.2 – The M2 helix and pore lining TMD residues. The residues are numbered with the prime notation and visualized with two M2 helices on opposite sides of the pore. Made with VMD [10].
smaller. Threonine and Isoleucine are the opposite: they are similar in size and are both uncharged, but Threonine is polar. This means that it will effect compounds like water and ions differently. However, both mutations are of interest.
In the case of the I9’T mutation, the 9’ residues in GLIC have been suggested as key residues of a hydrophobic gate, that ensure the closure of the channel [21]. As already mentioned, binding of anaesthetics in that region has also been connected to allosteric inhibition [18], [19], same as for an alcohol binding in ELIC [20]. Mutations in that area can, therefore, either reinforce or weaken that notion. If the mutation causes a higher ethanol occupancy in the pore due to the increased polarity of Threonine, it is interesting to see if the channel is inhibited.
F14’L, on the other hand, is fascinating to study for two main reasons.
Firstly, the mutation targets one of the residues that faces and takes part in the intersubunit pocket. The intrasubunit cavity has been hypothesized to be a potentiating allosteric site of alcohol for GLIC [24], [25], so a mutation at the 14’ position can investigate the role of the intrasubunit cavity further;
secondly, the F14’L mutation is larger than the Alanine mutation in F14’A
Figure 2.3 – Xenopus laevis oocytes and injection needle. The brown region is the so called animal pole and the white is the vegetable pole. For cDNA injections, the needle needs to be injected into the center of the animal pole, to ensure the tip and the cDNA reaches the nucleus of the oocyte. Photo by the author.
used in [24], [25], which could mean that the mutation causes less disruption to the overallGLICstructure. When implementing a mutation to investigate the importance of a certain residue or region, it is always important that the mutation in turn does not cause a major change to the system or structure that is responsible for the changes in behavior. In addition, Leucine is the same residue as one of the eukaryotic GABAA receptors has at the 14’ position, amongst others. This could mean that F14’L is a better model system for them.
2.1.1 Xenopus Laevis Oocyte Injection
Xenopus laevis oocytes were ordered from EcoCyte Bioscience (Castrop- Rauxel, Germany).Complementary DNA (cDNA)ofGLIC WTand the three mutants were injected into the nucleus of oocytes with the help of a Nanoject II microinjector (Drummond Scientific, PA, USA). The needles were made with Drummond #3-000-203-G/XL glass capillaries (Drummond Scientific, PA, USA) and pulled with a PC-10 Narishige Puller (Narishige Group, Tokyo, Japan) into a ∼20 µm tip. After filling the needles with mineral oil, thecDNA was taken up into the needle, while making sure to avoid air bubbles. See
Methods | 13
Table 2.1 –CDNAconcentrations injected intoGLIC mutant concentration [ng/32.2 nL]
WT 0.25, 0.5, 1.0
F14’L 3.0
I9’T 0.5
F14’L/I9’T 3.0
Figure2.3for an example of oocytes and injection needle.
The different injected concentrations for the GLIC WT and its mutants can be found in Table2.1. The expression level forGLIC WTvaried a great deal at times, depending on the condition of the oocytes. In those cases, the three different concentrations in Table2.1were used, and the oocytes with the most optional level of expression was selected. The needles were made sure to enter the center of the brown, animal pole in Figure2.3.
After injection, the oocytes were stored in individual wells on a 96- well or 48-well cell culture plate (Corning incorporated, ME, USA) and incubated at 13C◦ in pH 7.5 incubation medium (88 mM NaCl, 10 mM HEPES, 2.4 mM NaHCO3, 1 mM KCl, 0.91 mM CaCl2, 0.82 mM MgSO4, 0.33 mM Ca(NO3)2, 2 mM sodium pyruvate, 0.5 mM theophylline, 0.1 mM gentamicin, 17 mM streptomycin, 10,000 U/I penicillin, filtered through a 0.22 µm filter) for GLIC WT and pH 8.5 incubation medium (with 10 mM Trizma instead 10 mM Hepes) for the mutants. The oocytes were incubated for 2-4 days.
2.1.2 TEVC
For the TEVC recordings, the oocytes were probed with borosilicate glass electrodes (Harvard Apparatus, MA, USA), that were pulled with a PC-10 Narishige Puller (Narishige Group, Tokyo, Japan). These were filled with 3M KCl, to ensure a resistance of 5-50 MΩ. After the probing, the oocytes were clamped at a membrane potential of -70 mV with an OC-725C voltage clamp (Warner Instruments, CT, USA). The system was perfused with Ringer’s buffer, which consists of 123 mM NaCl, 2 mM KCl, 2 mM MgSO4, 2 mM CaCl2 and 10 mM HEPES (for pH 7.5) or 10 mM Trizma (for pH 8.5), and filtered through a 0.22 µm filter. The flow rate of the perfusion system was
∼1 mL/min and the digitizer Axon CNS 1440A Digidata and pCLAMP 10 software (Molecular Devices, Sunnyvale, CA) converted the currents to the corresponding digital signals.
TheGLICvariants were activated during 2 min by changing the perfusion buffer to a Ringer’s buffer 2 pH levels lower than the running buffer.
Accordingly, Ringer’s buffer pH 5.5 was used for the pH peaks of GLIC WT and pH 6.5 for the others. For the alcohol responses, the buffer was exchanged for a solution of the running buffer with ethanol (EtOH) during 1 min, to isolate variables. The pH 7.5 or pH 8.5 Ringer’s buffer with EtOH was, thereafter, exchanged again for the same alcohol concentration but with Ringer’s pH 5.5 or 6.5 respectively, for a duration of 2 min. After a total of 3 min, the perfusion system was returned to the respective running buffers.
Three different concentrations of EtOH in Ringer’s buffer were used for the alcohol peaks: 600 mM, 200 mM and 60 mM. In humans, a concentration of 17 mM ethanol in the blood corresponds to 0.08 % volume/volume [33], so the concentrations used in these experiments are notably high. This is to more easily be able to see the effect alcohol has. In between each ethanol response curve, a control pH peak was performed, so that the ethanol peaks could be compared between oocytes, mutants and different levels of expression. All peaks were recorded with approximately 5 min washout in between with the oocytes in the running buffer, apart fromGLIC F14’L/I9’Tthat needed 15-20 min in order to recover to a level close to base line.
The reason for using higher pH buffers for the GLIC mutants than WT, stems from previous work performed in the group, the result of which can be found in Figure 2.4. The pH curves of the GLIC mutants are left shifted, compared to WT, which means that a higher pH buffer is needed for the majority of the channels to be non-conducting. The pH values of the activation buffers were chosen so that the response was between 10 − 20% of thehalf maximal effective concentration (EC50)value.
2.1.3 TEVC Data analysis
The ethanol response curves and pH peaks for the different GLICs from TEVC were transferred into a spreadsheet and analysed with the PRISM software (Graphpad, San Diego, CA). The percentage ethanol modulation was calculated by taking the
peak EtOH current value − base line current
average pH peaks (peak − base line) on either side of the EtOH peak, for each ethanol concentration peak of each oocyte. Thereafter, these were averaged for each concentration and the standard error of mean (SEM) calculated. A similar procedure was followed for the leak currents, taking
Methods | 15
Figure 2.4 – GLIC variant pH curves of GLIC WT, F14’L, I9’T and F14’L/I9’T. The mutants show a shift in the response to the left. This means that the mutants are activated at a higher pH value than WT is. Made in previous work by the Lindahl group.
the initial leak current absolute value and calculating the mean and SEM values for eachGLICvariant. Example trace signals were extracted and noise reduced, one for each variant.
2.2 MD Simulations
TheMD simulations analyzed in this work, were ofGLIC WT, F14’L, I9’T andF14’L/I9’Tin an approximately open state, and a closed state, both with and without ethanol.
2.2.1 Construction of the MD Simulations of GLIC
TheMDsimulations ofGLIC WT,F14’L, I9’TandF14’L/I9’Twith ethanol andwithout ethanol (apo)analysed in this work, were made by Luise Zeckey and Yuxuan Zhuang respectively.
The protein structures used as the basis of the simulations, were the GLIC WT crystal structures of the open state from [34] and the closed state from [35]. These were point-mutated to their corresponding mutants with the Rotamers tool of Chimera [36]. The GLIC variant protein structures were protonated, such that the open structures would correspond to a pH 4.6 open state (4HFI) and the closed structures to a pH 7.0 closed state (4NPQ). All protein structures were embedded into a 1-palmitoyl-2-oleoyl phosphatidylcholine (POPC) lipid bilayer. The system box was solvated
with water molecules and filled with ions to an ionic strength of 0.1M and electric neutrality. The end product at this stage was a system with 1 GLIC variant protein, ∼ 300 POPC lipids, ∼ 40000 water molecules and ∼ 150 ions (Na+ and Cl-). In addition, 1 % of the water molecules were replaced with ethanol for the simulations of the states with alcohol. 1 % ethanol corresponds to a concentration of ∼ 600 mM, which is the same as the highest ethanol concentration tested experimentally. The reason for using the highest concentration, is to more easily be able to find where ethanol binds and possibly see bigger differences between theGLICvariants.
The simulations were run with Gromacs 2018 [37]. The force-fields used, were the Amber99sb-ILDN force-field, the Amber99sb-ILDN_berger for the lipid parameters and TIP3P for the water molecules [38]. The systems were energy minimized for 10000 steps with a force tolerance of 1000 kJ/mol/nm.
Thereafter, they were equilibrated using V-rescale [39], and equilibrated with positional restraints on heavy atoms, backbone atoms, and alpha carbons with berendsen semiisotropic coupling [40]. At this stage, seven ethanol molecules were placed in the respective channels of the states with ethanol, assuming they were not already in position. One ethanol molecule was placed at each intersubunit cavity, like they did in [24]. Two ethanol molecules were placed in the pore. The approximate position of the intersubunit cavity and the pore can be seen in Figure1.2. Finally, the MDsimulations were all run without restraints for up to 1000 ns. The end product was three replicates for theapo states of theGLICvaraints (4HFIand4NPQ), and four replicates for the ones with ethanol.
2.2.2 Simulation Analysis and Definitions
The simulations of the different GLIC channels, 4HFI and 4NPQ, apo and with ethanol, were analyzed with the libraries CHAP [41], MDTraj [42] and MDAnalysis [43], [44] and visualized with the visualization application VMD [10]. A full explanation of how these tools are used is beyond the scope of this thesis, but can be found in the referenced literature and on their respective websites. The methods used, with needed definitions and specifications is explained below. The water, ethanol and lipid occupancy calculations are particularly high-lighted because of their more complex nature. In all the analysis, the first 500 ns of the simulations were disregarded.
Methods | 17
Pore Radius Profiles
The pore radius profiles of the Cα, both over time and time-averaged, were made with CHAP [41] to visualize the pores and make sure they had not collapsed. The average radius profiles of the pore can also give an indication of how open the channel is, and if ether ethanol or the mutations have a major effect on the overall structures.
RMSF and RMSD
Root mean square fluctuations (RMSFs) and root mean square deviations (RMSDs) were calculated with the built-in MDTraj [42] methods. The RMSFs can show the flexibility of the structure for the Cα around a certain residue, which can show a change in mobility and, therefore, flexibility in- between mutants and apo or ethanol states. RMSDs, in general, show how stable the initial model is, by showing the deviation from said initial model.
By contrasting the ethanol states to theapostates, theRMSDcan indicate the effect the ligand is having. In some cases, theRMSDcan also be connected to a conformation of the protein structure. Key areas, such as loop C, are known to have certain conformations in the open vs. the closed state. Therefore, the RMSDof these key regions can be used as a metric for how open or closed a state is.
β -expansion
β -expansion has been determined as one of the key movements in gating [45].
It has been quantified with MDTraj [42], as the distance between the ECD residues with numbers 32 and 192 in each subunit. β -expansion shows how constricted this extracellular region is for one type of channel compared to another, which can indicate a possible opening of the channel. The 4HFI and 4NPQ states should correspond to pH 4.6 and pH 7.0 respectively, so the behavior of the β -expansion can be contrasted with both the pH curves in Figure 2.4 and experimental TEVC alcohol results. For instance, if a β - expansion of a4NPQstate with ethanol has a smaller value than for theapo state (making it more similar to the 4HFI state), it can help to explain why that specific channel is potentiated by ethanol. Furthermore, a shift of anapo state towards a lower β -expansion value, can be connected to a possible left shift of the corresponding pH curve in Figure2.4.
Visualization
The channel visualizations and ethanol density plots were made with VMD and the Volmap tool [10]. The volume slices orthogonal to the pore axis were specified at a level corresponding to a specific M2 residue prime level. The scale is ethanol atoms per ų with a range between 0 and 0.6 /ų. The density plots were made, one for each channel and state, on the systems where the replicates were concatenated. The result is, therefore, an overview of the general ethanol densities of the channels.
Pore Hydration and Ethanol Occupancy
The hydration and ethanol occupancy of the pore was calculated with the help of MDAnalysis [43], [44]. The oxygen atoms of the water molecules and either of the atoms in the ethanol molecules respectively, were selected and appended to the output value, if they were within a cylindrical volume of a certain radius and height 4 Å, centered at the center of a specific level of the pore. The levels were the M2 prime levels, where 9’, 13’ and 16’ were of highest interest. The radii were chosen based on the average radius profiles made with CHAP, described above. The radii used were 3 Å for 9’ and 4 Å for the other two. The average time each ethanol stayed at the certain prime level was measured with the MDAnalysis function Survival Probability, allowing for an intermittency of 2 ns. However, no difference between the mutants was observed, so they were omitted from this report.
Intersubunit Cavity Hydration, Ethanol and Lipid Occupancy
The hydration, ethanol and lipid occupancy of the intersubunit cavity was computed with MDAnalysis [43], [44]. Firstly, the region of the intersubunit cavity was defined, by closeness to certain residues. The residues and cutoff distances were decided by following the ethanol molecules in VMD and trying out different values using selected atoms in the graphical representation tool, around the known region close to residue 238 and 263. The approximate region can be seen as circle B in Figure 1.2. The most narrow definition was chosen that captured all the ethanol that stayed more or less stable in the region of interest.
The final selection became within 5 Å of residue 238 and 7 Å of residue 200 of subunit i, and within 10 Å of residue 239 and 7Å of residue 263 of subunit i+1 for the open state. For the closed state it was changed to within 10 Å of residue 238 and 9 Å of residue 200 of subunit i, and within 8 Å or
Methods | 19
residue 239 and 5 Å of residue 263 of subunit i+1, to account for the increased volume of the region. Most of these residues are highly conserved in-between pLGICs, supporting their importance for the structure and function of the channel.
Like for the calculations of the pore hydration and ethanol occupancy, all oxygen atoms of the water molecules and either of the atoms of the ethanol molecules within that region were selected and added to the output. For the lipid content of the intersubunit cavity, all atoms of thePOPClipid within that region were selected, such that more atoms of a lipid within the region would give a higher value.
Intrasubunit Cavity Hydration, Ethanol and Lipid Occupancy
After studying the ethanol occupancy of the intrasubunit cavity region in VMD, it was clear that ethanol occupied different regions within the intrasubunit pocket seen as circle C in Figure 1.2. One of the regions was close to the M2 helix and residue V242, and another faces out towards the membrane and M4. They are visualized with bound ethanol molecules in Figure2.5. As mentioned in the introduction, an intrasubunit cavity close to V242 in GLICwas suggested to be important for the potentiating allosteric effect of the anaesthetic Propofol [27]. The whole intrasubunit pocket has also been correlated to allosteric inhibition. Therefore, to understand what causes this discrepancy and if there are two regions within each subunit that possibly cause the two very separate allosteric effects of alcohol, the occupancy of the two regions will be quantified. Henceforth, the two regions will be referred to as two separate cavities in this thesis: the M2-close intrasubunit cavity and the M4-close, or membrane-close, intrasubunit cavity.
The procedure for calculating the two different intrasubunit cavity hydrations, ethanol and lipid occupancy, was the same as for the intersubunit cavity. The only difference was the specified residues and cutoff distances to those residues. For the intrasubunit cavity close to M2, the definition was within 5 Å of residue 205, 8 Å of residue 255 and 5 Å of residue 242 within one subunit. For the membrane-close intrasubunit cavity, the specification was instead within 5 Å of residue 205, 10 Å of residue 255 and within 7 Å of residue 307, which is a part of the M4 helix instead of the M2 pore lining helix, just as 242 is.
The survival probability of the ethanol was also measured for the inter- and intrasubunit cavities in the same way as for the pore hydration. However, these were excluded due to too large error margins and small differences in-
EtOH in the M2-close intrasubunit cavity
EtOH in the M4-close intrasubunit cavity
Pore Pore
Figure 2.5 – Ethanol binding in the M2-close (left) and the M4-close (right) intrasubunit cavities. The left figure is of the GLIC F14’L 4HFI state and the right figure is of theI9’T 4HFIstate. Only one subunit is shown.
Visualized with VMD [10].
betweenGLICs.
Results and Analysis | 21
Chapter 3
Results and Analysis
In this chapter, the results from theTEVC electrophysiology recordings and analysis of the MD simulations is presented. All are of the four different GLICvariantsWT,F14’L,I9’TandF14’L/I9’T, as well as with and without ethanol. The three concentrations of ethanol used for the recordings were 600 mM, 200 mM and 60 mM, and theMDsimulations were with the equivalent of 600 mM. Starting with the TEVC experimental results, the chapter then continues with a section with the computational results.
3.1 Electrophysiology
TEVC electrophysiology is a powerful tool to that can give whole-cell statistics in real time of a channel’s response to a change in environment and exposure to a ligand. By comparing the response of different GLICvariants to different concentrations of ethanol, it allows for a deeper understanding of the effect a certain mutation has. The characteristics of each variant can, thereafter, be compared and put in the context of the results and analysis of the simulations.
3.1.1 Ethanol Modulation
Examples ofTEVCsample traces ofGLIC WT,F14’L,I9’TandF14’L/I9’T can be found in Figure 3.1. The resulting average and SEM percentage modulations of the different GLIC variants is shown in Figure 3.2 and Table3.1.
The appearance ofWTin Figures3.1and3.2will serve as a reference for the other variants. WTis potentiated to almost 40 % by 600 mM EtOH, not
WT
F238L
I233T
F238L/I233T
Figure 3.1 –TEVC GLICvariant sample traces. The sample traces are of WT, F14’L, I9’T and F14’L/I9’T for 600 mM, 200 mM and 60 mM EtOH in Ringer’s buffer. The time and current scales are the same for all traces.
WT andF14’L/I9’T are slightly potentiated by especially 600 mM ethanol, while F14’L is substantially potentiated and I9’T is inhibited. F14’L/I9’T and partiallyI9’Thave a greater base line offset and the former also has slow kinetics.
Results and Analysis | 23
Table 3.1 – The mean % alcohol modulation,SEMand number of oocytes recorded on (N) ofGLIC WT, F14’L, I9’TandF14’L/I9’Tfor 600 mM, 200 mM and 60 mM EtOH in Ringer’s buffer. F14’L is substantially potentiated by ethanol, while WT and F14’L/I9’T are less so and I9’T is inhibited.
mutant Ethanol Concentration [mM] Mean % SEM% N
WT 600 38,5 3,68 11
200 1,7 5,66 12
60 -7,9 2,47 11
F14’L 600 224,1 8,45 12
200 43,5 6,06 11
60 -1,7 3,31 11
I9’T 600 -8,4 3,47 13
200 -7,6 1,74 11
60 -8,7 2,67 11
F14’L/I9’T 600 27,8 11,34 13
200 -3,2 2,87 12
60 -11,7 5,92 11
Figure 3.2 – The mean % alcohol modulation and SEM of GLIC WT, F14’L, I9’T and F14’L/I9’T for 600 mM, 200 mM and 60 mM EtOH in Ringer’s buffer. The ’*’, or absence of them, indicate how significant the difference of the % alcohol modulation is compared to GLIC WT with Welch’s t test. F14’Lis substantially potentiated by ethanol, whileWT and F14’L/I9’Tare less so andI9’Tis inhibited. The inhibition of 600 mM and 200 mM onGLIC I9’T has significance ’*’ and ’**’ respectively, compared to zero, with the one sample t test.
Results and Analysis | 25
Figure 3.3 – The mean % alcohol modulation and SEM of GLIC WT, F14’L,I9’TandF14’L/I9’Tfor 60 mM EtOH in Ringer’s buffer. The ’*’, or absence of them, indicate how significant the difference of the % alcohol modulation is compared to zero with one sample t test. 60 mM EtOH inhibits WTandI9’Tsignificantly.
especially affected by 200 mM, and verifiable inhibited by 60 mM in fig.3.3 with a P-value of 0,0095.
GLIC F14’Linstead has a mean of more than 200 % ethanol modulation for 600 mM, and more than 40 % even for 200 mM. The 60 mM concentration of ethanol does not have a significant effect on F14’L. Thus, the variant distinguishes itself by being substantially potentiated by ethanol, in comparison toWTwith a P-value of less than 10−4.
GLIC I9’T, on the other hand, is inhibited by all three ethanol concentrations. The significance of this difference to zero with the one sample t testis with a P-value of 0,0323 for 600 mM, 0,0014 for 200 mM and 0,0086 for 60 mM. Consequently, the I9’T mutation is inhibited by ethanol. This is interesting, considering the mentioned importance of the 9’ residue in the suggestedGLIChydrophobic gate [21] and the possible 9’ pore allosteric site ofGLIC.
GLIC F14’L/I9’Tshows no significant difference in modulation toWT. It is potentiated by 600 mM EtOH, 200 mM has no significant effect, nor does 60 mM due to largeSEMs. It appears that the potentiating effect of theF14’L mutation counteractsI9’T’s inhibitory effect.
Table 3.2 – Average andSEMoocyte leak currents.
mutant Average SEM N
WT -0.07400 0.03328 13 F14’L -0.07283 0.03854 12
I9’T -0.4274 0.2805 13
F14’L/I9’T -0.4917 0.1833 14
3.1.2 Functional Dynamics
The main focus of this work is the allosteric effect, whose experimental results are found in the subsection above. Notwithstanding, other features of the differentGLICvariants are worth mentioning:
In the sample traces in Figure 3.1, the time axis and current axis is the same for all variants. GLIC WT and F14’L both show quick responses and recoveries to base line. I9’T mostly showed the same quickness but was slightly more variable. TheF14’L/I9’Tmutant stands out with its slow kinetics, both in the activation period and in the recovery to base line. The others could be activated after a recovery of ∼ 5 min, where they had quickly returned to base line. F14’L/I9’T required 15-20 min, especially in the beginning, and even then it never reached a steady base line. Waiting much more than this time was not an option, since the longer the oocyte is clamped, the higher the risk for it dying. Therefore, theF14’L/I9’Tmutant was allowed to recover long enough for its "recovery gradient" to be small, and until it reached the approximate previous base line. The reason for the slow kinetics of F14’L/I9’T is something that would be interesting to investigate in the future. The total simulation times of the MDsimulations are sadly too low for this work to be able to shed any light on that question.
In general, the double mutant was also much more leaky compared to the otherGLICs, which can partly be seen as the base line offset in the sample traces. In Figure3.4 and Table 3.2, the full statistics of the leak current are presented, by taking the average and SEMs of all the leak currents right after clamping the oocytes. The statistical significance of the F14’L/I9’T mutation’s leak compared to that ofWTwas with a P-value of 0.0419.
The leak currents are for oocytes with expressed channels that have roughly similar levels of activity (e.g. within 1–5 µA at pH 5.5 or 6.5).
However, variations in expression level and/or agonist efficacy may account for variability between oocytes and variants. Still, the difference is quite suggestive. I9’T showed a great deal of variability, with a P-value higher
Results and Analysis | 27
WT F238L I233T F238L/I233T
0.0 0.2 0.4 0.6 0.8
Leak Current
LeakCurrent(µA)
*
Figure 3.4 – The mean leak current and SEM of GLIC WT, F14’L, I9’Tand F14’L/I9’T. The ’*’, or absence of them, indicate how significant the difference of the leak is in comparison to GLIC WT. F14’L/I9’T has a significantly higher leak whileI9’Tis more variable.
than 0.05, as it did with its kinetics, while F14’L/I9’Tpresents a significant difference of 0,0419 compared toWT.
3.2 MD Simulation Analysis
The analysis of the simulations of the four GLIC variants, in 4HFI and 4NPQ states and with and without ethanol, in this section can potentially provide an explanation for the mechanisms underlying ethanol potentiation and inhibition in the experimental TEVC results. The pore radius, RMSFs, RMSDs, pore occupancy, inter- and intrasubunit occupancy were analyzed.
To start off, theTMDstructure ofWTis shown from above in Figure3.5, with its original I233 and F238 residues as well as the mutated residues T233 and L238. Visible is the variable direction of F238, that either faces the intersubunit cavity or is flipped towards M1 and the intrasubunit pocket.
The latter position increases the size of the intersubunit cavity. Only the intersubunit-facing position is seen for L238. Secondly, to visualize regions where ethanol can be found, a snap-shot of GLIC F14’L with ethanol is presented in Figure3.6.
Figure 3.5 –GLIC MD structure with the mutated residues 233 and 238.
The originalWTI233 and F238 residues are in grey. In cyan is T233, present in I9’T and F14’L/I9’T. F14’L and F14’L/I9’T’s L238 is instead found in magenta, where the rigid position of L238 is visible, compared to the variable direction of F238 inWT. Made with VMD [10].
Results and Analysis | 29
Figure 3.6 – Snap-shot of the ethanol present in GLIC F14’Lat the end of the simulation. It exemplifies the places that the modulator can occupy.
Made with VMD [10].
3.2.1 CHAP Average Pore Radius Profiles
Time-averaged pore radius profiles of theGLICvariants, show cross sections of the general morphology of the pore. By comparing the different variants, they can indicate constrictions or openings of the pore structure responsible for the characteristic behavior of a certain variant. After verifying the stability over time by looking at the time series, that is omitted from this report, the time-averaged pore radius profiles were analysed.
The time-averaged pore radius profiles in Figure 3.7a with standard deviations (StDs)show small differences between theapoand ethanol states for each variant. Around 9’ the highest difference of theGLICvariants is an increased average radius of ∼0.1 Å forI9’T 4HFIethanol state compared to apo, and ∼0.2 for the 4NPQstate equivalent, followed by a slightly smaller difference in radius betweenWTwith ethanol compared to the case without.
This can serve as an indication of the reason for the ethanol inhibition of I9’T and slight potentiation of WT in Figure 3.2. Around 16’ of the 4HFI states there is a visible trend; ethanol increases the average radii ofI9’Tand F14’L/I9’T with ∼0.5 Å and decreases the radius of F14’L with the same amount, which is more difficult to connect to an experimental result. Around -2’ 4HFI states, ethanol expands the average radius of F14’L with ∼0.5 Å and decreases I9’T’s radius with approximately the same, again connected to the potentiation ofF14’Land inhibition ofI9’T. Around -2’4NPQstates, ethanol expands the radii ofF14’LandI9’Twith approximately 0.4 Å and 0.8 Å respectively, and reduces the radii ofWTwith ∼ 0.5 Å andF14’L/I9’Twith even less than that. Worth noting, however, is the size of theStDs. All radii are within the others’StDs. The fact that these differences also are small, shows that ethanol does not on its own destabilize or expand the pore substantially.
Figure3.7bshows the difference in the time-averaged pore radius profiles betweenWTand the mutants. For the 4HFIstate ofF14’L, the difference in radius around the 9’ center of theTMDpore is minor, but larger towards the top and bottom. Towards the lower part of the pore,F14’Lhas a bigger radius, which could be a contributing factor to its increased ethanol potentiation; a larger pore usually means that more ions can go through it, increasing the conductance. F14’Lwith ethanol has a reduced radius around 16’, but since the pore is already wider in that region it should not affect the conductance.
The difference is not especially big for the4NPQstate.
I9’Tvs. WT, shows a ∼ 0.7 Å increase of the radius ofI9’Tat 9’ for the 4HFIstate compared toWT, which is 2/3 as large for the4NPQstate. For the 4NPQstate, it can help explain the left shift of the pH curves in Figure 2.4.
Results and Analysis | 31
For the 4HFI states, it will be interesting to see if the 9’ mutation with its increased polarity, means that more ethanol or water is there that expands the pore compared to WT. The radius of I9’T at the lower pore region, follows that pattern for the4HFIstate. Interestingly for the4NPQstate,I9’Tapo and WTethanol and vise versa have around the same average radii.
F14’L/I9’Tmostly shows the same behavior as I9’T does, compared to WT, even though the variability between with and without ethanol is slightly smaller. The average radii is larger around 9’, smaller around 16’ and smaller than bothWT apoand ethanol in the4NPQstate.
With all this said, the standard deviations that can be seen in Figure3.7a (and that were excluded in Figure 3.7b to increase the visibility of the average tendencies), show that the deviations are mostly within the StDs of the others. The averages are still interesting, since they can give some evidence connected to the potentiation and inhibition of ethanol seen in the experimental results in Figure 3.2and Table 3.1. In addition, they show that the mutations do not collapse or destabilize the pore structure substantially.
3.2.2 RMSF Curves
RMSFs of the Cα around a certain residue can show the change in mobility or flexibility in-between ethanol and apostates and in-between the different GLICvariants. This can in turn indicate the stability of the simulations, to a certain extent. If the simulation structure shows a great deal of flexibility for a certain state, that might imply disruption of its structure that might make the other results less reliable.
The curves of the RMSFs in Figure 3.8, show no major differences between apo and ethanol states. The only region that has some alterations, is around the residues with number 60-70 and 90-100 for the 4NPQ F14’L state. This is in theECD, in its β -sheets and loops, but not any of the ones in either of the loops important for allostery and the orthosteric site. In addition, the difference in flexibility is very small and solely for the 4NPQstate, that we know experimentally ethanol has no effect on.
In Figure 3.9, the RMSF between the mutants and WT in the TMD is presented. The 4HFI GLIC mutants with ethanol all show an increased RMSFin the region between approximate residue numbers 220 and 228. This increased mobility is in the -2’ to 2’ region, so it is not where the previously proposed hydrophobic gates lay, that constrict ions from passing through the channel [21]. Moreover, the difference is no more than 0.2 Å, so it is minor.
Another small difference is seen in Figure3.9bfor theF14’L/I9’Tmutant,
3 2 1 0 1
2 4HFI WT
apo ethanol
4HFI F238L
apo ethanol
4HFI I233T
apo ethanol
4HFI F238LI233T
apo ethanol
0.0 0.2 0.4 0.6 0.8
3 2 1 0 1
2 4NPQ WT
apo ethanol
0.0 0.2 0.4 0.6 0.8
4NPQ F238L
apo ethanol
0.0 0.2 0.4 0.6 0.8
4NPQ I233T
apo ethanol
0.0 0.2 0.4 0.6 0.8
4NPQ F238LI233T
apo ethanol
Radius [nm]
Distance along pore [nm]
Average Radius Profile
9'
-2' -2' 9' 16'
16'
(a)apovs. ethanol for allGLICs
3 2 1 0 1
2 4HFI F238L vs WT
F238L ethanol WT ethanol F238L apo WT apo
4HFI I233T vs WT
I233T ethanol WT ethanol I233T apo WT apo
4HFI F238LI233T vs WT
F238LI233T ethanol WT ethanol F238LI233T apo WT apo
0.0 0.2 0.4 0.6 0.8
3 2 1 0 1
2 4NPQ F238L vs WT
F238L ethanol WT ethanol F238L apo WT apo
0.0 0.2 0.4 0.6 0.8
4NPQ I233T vs WT
I233T ethanol WT ethanol I233T apo WT apo
0.0 0.2 0.4 0.6 0.8
4NPQ F238LI233T vs WT
F238LI233T ethanol WT ethanol F238LI233T apo WT apo
0.0 0.2 0.4 0.6 0.8 1.0
Radius [nm]
0.0 0.2 0.4 0.6 0.8 1.0
Distance along pore [nm]
Average Radius Profile
(b)GLICmutant vs.WT.
Figure 3.7 – Average pore radius profiles with of GLIC F14’L, I9’T and F14’L/I9’T to WT. Ethanol does not alter the appearance of the pore substantially, but the mutations increase the radius of I9’T around 9’ and around -2’ forF14’L. Made with CHAP [41].
Results and Analysis | 33
in the region between residue numbers 230-240. This region covers the 6’
to 16’ level, where two hydrophobic gates of the GLIC channels have been proposed to lay [21]. An increased flexibility, could be a reason for the increased leak current of theF14’L/I9’Tmutant seen in Figure3.1. However, the difference is only ∼ 0.1 Å.
3.2.3 RMSD
RMSDscan, in general, show how stable the simulations are, by showing the deviation from the initial model of the simulation.
The majority of the RMSD results showed small differences between mutants and states of no more than 0.1 Å. Thus, they cannot give any insight into the gating mechanism and ethanol modulation ofGLIC, and were placed in the Appendix. The RMSDs of the loops however, were of some interest.
The approximate positions of the loops can be found in Figure1.1.
Loop C, and therefore itsRMSD, was shown in [35] to have a different conformation when it was in a closed state than in an open state. This is reasonable, considering loop C is located by the orthosteric binding site and thus connected to allostery. As Figure 3.10ashows, theRMSDof the4NPQ states is higher than for the 4HFI states by almost 1 Å. This is especially the case for WTthat, unlike the mutants, we know from Figure 2.4 is non- conducting at pH 7.0. In [35], loop F was also observed to have different RMSD, depending on whether it was open or closed. The same can not be seen in Figure 3.10c. It can in Figure 3.10b of loop 2, that interacts with loop F. The difference is maximally ∼ 0.9 Å, forWT, and less than 0.5 Å for the others. The proton activation of GLIC does not occur by binding to the orthosteric site, so perhaps this is a reason for the partly inconsistent result, compared to the one in [35]. However, it is reassuring that the difference between closed and open state, where WTshould have the largest variation, is reflected in theRMSDsof the loops.
3.2.4 β -Expansion
As one of the key movements of gating, β -expansion shows how constricted the lower region of the ECD pore is and can indicate how open or closed a certain state is.
The β -expansion in Figure3.11, shows a similar behavior of the GLICs in-between 4HFI states, which is consistent with all the states being open.
Their major peaks are all around 1.05 Å, even though the error bars are quite
