Scholarly article on topic 'Dimerization of the EphA1 Receptor Tyrosine Kinase Transmembrane Domain: Insights into the Mechanism of Receptor Activation'

Dimerization of the EphA1 Receptor Tyrosine Kinase Transmembrane Domain: Insights into the Mechanism of Receptor Activation Academic research paper on "Biological sciences"

0
0
Share paper
Academic journal
Biochemistry
OECD Field of science
Keywords
{""}

Academic research paper on topic "Dimerization of the EphA1 Receptor Tyrosine Kinase Transmembrane Domain: Insights into the Mechanism of Receptor Activation"

BIOCHEMISTRY

including biophysical chemistry & molecular biology

Subscriber access provided by LAURENTIAN UNIV

Article

Dimerization of the EphA1 Receptor Tyrosine Kinase Transmembrane Domain: Insights into the Mechanism of Receptor Activation

Matthieu Chavent, Alan P. Chetwynd, Phillip James Stansfeld, and Mark S. P. Sansom

Biochemistry, Just Accepted Manuscript • DOI: 10.1021/bi500800x • Publication Date (Web): 06 Oct 2014 Downloaded from http://pubs.acs.org on October 7, 2014

Just Accepted

"Just Accepted" manuscripts have been peer-reviewed and accepted for publication. They are posted online prior to technical editing, formatting for publication and author proofing. The American Chemical Society provides "Just Accepted" as a free service to the research community to expedite the dissemination of scientific material as soon as possible after acceptance. "Just Accepted" manuscripts appear in full in PDF format accompanied by an HTML abstract. "Just Accepted" manuscripts have been fully peer reviewed, but should not be considered the official version of record. They are accessible to all readers and citable by the Digital Object Identifier (DOI®). "Just Accepted" is an optional service offered to authors. Therefore, the "Just Accepted" Web site may not include all articles that will be published in the journal. After a manuscript is technically edited and formatted, it will be removed from the "Just Accepted" Web site and published as an ASAP article. Note that technical editing may introduce minor changes to the manuscript text and/or graphics which could affect content, and all legal disclaimers and ethical guidelines that apply to the journal pertain. ACS cannot be held responsible for errors or consequences arising from the use of information contained in these "Just Accepted" manuscripts.

ACS Publications

High quality. High impact. Biochemistry is published by the American Chemical Society. 1155 Sixteenth Street

N.W., Washington, DC 20036

Published by American Chemical Society. Copyright © American Chemical Society. However, no copyright claim is made to original U.S. Government works, or works produced by employees of any Commonwealth realm Crown government in the course of their duties.

10 11 12

20 21 22

1699787_File000000_27025629.docx October 6, 2014

Dimerization of the EphA1 Receptor Tyrosine Kinase Transmembrane Domain: Insights into the Mechanism of Receptor Activation

Short title: EphA1 transmembrane domain dimer

Matthieu Chavent, Alan P. Chetwynd, Phillip J. Stansfeld, Mark S. P. Sansom*

Department of Biochemistry,

University of Oxford,

South Parks Road,

Oxford,

OX1 3QU,

*to whom correspondence should be addressed. Email mark.sansom@bioch.ox.ac.uk

Direct: +44 (0)1865-613306

PA: +44 (0)1865-613212

Biochemistry, ms. bi-2014-00800x

Abstract

EphA1 is a receptor tyrosine kinase (RTK) that plays a key role in developmental processes, including guidance of the migration of axons and cells in the nervous system. EphA1, in common with other RTKs, contains an N-terminal extracellular domain, a single transmembrane (TM) a-helix, and a C-terminal intracellular kinase domain. The TM helix forms a dimer, as seen in recent NMR studies. We have modelled the EphA1 TM dimer using a multiscale approach combining coarse grain (CG) and atomistic molecular dynamics (MD) simulations. The one dimensional potential of mean force (PMF) for this system, based on inter-helix separation, has been calculated using CG MD simulations. This provides a view of the free energy landscape for helix-helix interactions of the TM dimer in a lipid bilayer. The resulting PMF profiles suggest two states, consistent with a rotation-coupled activation mechanism. The more stable state corresponds to a right-handed helix dimer interacting via an N-terminal glycine zipper motif, consistent with a recent NMR structure (2K1K). A second metastable state corresponds to a structure in which the glycine zipper motif is not involved. Analysis of unrestrained CG MD simulations based on representative models from the PMF calculations or on the NMR structure reveals possible pathways of interconversion between these two states, involving helix rotations about their long axes. This suggests that the interaction of TM helices in EphA1 dimers may be intrinsically dynamic. This provides a potential mechanism for signalling whereby extracellular events drive a shift in the repopulation of the underlying TM helix dimer energy landscape.

10 11 12

20 21 22

Introduction

The Eph receptors play an important role in developmental processes including cell migration and axonal guidance (1). This group of receptors is the largest of the Receptor Tyrosine Kinase (RTK) family. The RTKs can be activated through the binding of extracellular ligands. In the case of Eph receptors, there are two classes of receptors, A and B, which mainly interact with the ligands ephrin A and B respectively, cross interactions between EphA receptors and ephrins B have also been reported (2, 3). The binding of ligand leads to the creation of receptor oligomers (dimers and/or higher order oligomers) (4). There is also growing evidence that RTKs also form (inactive) dimers in the absence of ligands. This preformed and inactive dimeric configuration has been discussed for several RTKs (5, 6) including the EGFR (7) and FGFR (8), and has sometimes been referred to as a "pre-dimerization" state (9). For the EphA2, data also support ligand-independent clustering of the receptor (10, 11). Furthermore, it has been postulated that Eph receptor dimers can switch between active and inactive configurations through a rotation coupling activation mechanism (12, 13) and indeed recent mutational experiments highlighted more or less activated forms of the dimer compared to the wild type EphA2 receptor in both ligand dependent and independent configurations (14). Taken together these studies emphasize the importance of a better understanding of how EphA2 and related receptors may adopt alternative dimeric configurations, and of how they may pass from one dimeric configuration to another.

The structure of Eph receptors is typical for RTKs. It is composed of an extracellular region which interacts with ephrin ligands (10, 15) and an intracellular region containing a juxtamembrane region, a tyrosine kinase domain, and a steril-alpha motif (SAM) domain, often followed by a PDZ binding motif. The two regions are linked by a transmembrane (TM) helix. This TM helix plays a role in the dimerization of the Eph receptor (12, 13, 16), a feature shared by numerous receptor tyrosine kinases (RTKs; (17)). Several studies have emphasized the role of the TM helix dimer and changes in its packing mode in signalling by RTKs including EGFR (7, 18) and FGFR3 (19, 20) in addition to other cellular receptors (21) including the insulin receptor (22). In particular recent mutational studies of the closely related EphA2 receptor have suggested that switching of the TM helix dimer between two packing modes is a possible mechanism underlying EphA2 signal transduction (14). Given that association of TM helix domains (23-25) is central to RTK signalling, understanding the nature of TM helix association is a necessary component of understanding mechanisms of signalling (9).

Structures of EphA1 and of EphA2 TM helix dimers in phospholipid bicelles have been determined using NMR (12, 13). The EphA1 TM domain forms a right-handed helix dimer interacting via an N-terminal glycine zipper motif (A550x3G554x3G558; (12) see also the sequence in Fig.1). The importance of the Small-x3-Small motif was highlighted more than 20 years ago in growth factor receptors (26). It is the main interaction motif of the first NMR structure determined for a TM helix dimer (in glycophorin A (27)) and has since been studied and identified in numerous TM domains (28). Although there is a high frequency of glycine and small residues in the 58 TM helix sequences of the human RTKs (29), direct conservation of a common interaction motif is not apparent. This diversity may reflect differences in local structure of the TM domains and/or different modes of activation for each RTK. Thus, the interaction through the Small-x3-Small motif is not the only mode of association available to TM domains (25). Indeed, the EphA2 TM domain has been shown to dimerize via a heptad repeat motif (13). On this basis it has been suggested that changes in helix packing mode may underlying signalling mechanisms by EphA and related receptors, with interactions via different motifs corresponding to active and inactive states of the receptor (14).

A number of experimental techniques exist for measuring association of TM helices (30), including TOXCAT and related assays (29, 31), biophysical approaches including FRET (32) and analytical ultracentrifugation (33, 34). Computational methods (including molecular dynamics (MD) or Monte Carlo (MC) simulations) provide a complementary approach for exploring the energy landscapes for association of TM helix dimers, and have been applied to model systems such as the TM domain of glycophorin A (35-37). Simulation approaches have more recently been applied to studying TM helix dimers from RTKs such as ErbB (35, 39). Polansky et al. have suggested a modelling framework to predict and analyse TM dimer interaction in which MD simulations and potential of mean force (PMF) calculations play a central role (40). A number of these studies have employed coarse-grained (CG) models (41, 42) and have shown that CG models yield similar free energies of dimerization to more detailed atomistic simulations of the well-studied glycophorin A TM helix dimer (compare e.g. (35) and (36)). Recent studies of e.g. the integrin TM helix dimer (43), glycophorin A (44), and the protein adaptor DAP12 (45) have shown that by combining CG and atomistic simulations a detailed view of the nature of helix/helix interactions in TM helix dimers may be obtained.

In this study we explore the EphA1 TM domains using MD simulations in order to gain insights into their mechanism of dimerization and how it may effect the activation of the receptor. There have been a number of MD studies of the dimerization of EphA1 receptor TM domain (12, 40, 46, 47). Here we have used CG MD simulations to calculate a one dimensional free energy landscape

2 for the association of the EphAl TM domain in a phospholipid bilayer. This suggests two

5 the activation mechanism involving alternative packing modes of the TM helices. We have used

7 unrestrained MD simulations to compare representative models from the free energy calculations

10 the two packing modes via rotation of the TM helices, which in turn may provide a pathway to link

12 a potential inactivate configuration to the activated state of the receptor. Atomistic simulations of

metastable states for this system, consistent with recent results for other RTKs and with models of

with the NMR structure of the dimer. These simulations suggest a pathway for interconversion of

representative structures from the CG energy landscape allow refinement of these models and

15 provide a more detailed description of the underlying helix/helix interactions.

Methods

Titratable amino acids were in their default ionization states. The CG forcefield used was a local

20 Unrestrained CG simulations starting from the NMR structure

22 The NMR structure of the EphAl dimer (PDB id: 2K1K) (12) was converted into a CG model.

25 modification (48, 49) of the widely-employed MARTINI forcefield (41, 42). A harmonic restraint

27 was applied to backbone particles to mimic secondary structure stabilising hydrogen bonds, with an

28 equilibrium bond length of 0.6 nm and a force constant of 1000 kJ/mol/nm . We then added 269

30 DPPC (or DLPC) lipids, water molecule, and counterions using the self-assembly method (50, 51).

32 After an initial 5000 steps of steepest descent energy minimization, the simulation was run for 100

ns to allow the lipids to self-assemble around the protein. All simulations were performed using

35 Gromacs (http://www. gromacs.org/) (52). A Berendsen thermostat was used with a coupling

37 constant of 1.0 ps. The reference temperature was 323 K. Electrostatic interactions utilized a

40 between 0.9 and 1.2 nm and 0 and 1.2 nm, respectively. A Berendsen barostat with a coupling

41 _ _

42 constant of 1.0 ps, a compressibility value of 5.0 x 10_6 bar-1 and a reference pressure of 1 bar was

47 PMF calculations

relative dielectric constant of 20. Lennard-Jones and electrostatic interactions were shifted to zero

used. The integration time step was 10 fs.

50 This structure was subsequently used as a starting point for PMF calculations. Thus we initially ran

52 a 200 ns CG MD simulation during which 271 DPPC lipids were allowed to self-assemble to form a

As described above, we performed CG MD simulations to yield a structure of the EphAl dimer.

bilayer around a pair of restrained TM helices (aligned with their long axes parallel to the z axis and

55 separated by 7 nm). Titratable amino acids were in their default ionization states. Five 1.5 ^s

57 simulations were subsequently performed with these helices unrestrained, yielding dimers

59 interacting through their glycine zipper motifs. One such dimer was then used as a starting point for

PMF calculations, increasing the separation of the centres of mass of the glycine zipper motifs as the PMF reaction coordinate.

The PMF was calculated using the umbrella sampling technique (53). 24 independent simulations each of 7.5 ^s duration were run for the PMF calculation, i.e. the simulations corresponding to the 24 windows were run such that the initial configuration of one simulation did not depend on the outcome of the preceding simulation. Each independent simulation corresponds to a different separation distance of the centre of mass of the glycine zipper motifs on each helix. The timestep was 40 fs and the reference temperature was 323 K. The starting structure for each window in a given calculation was the same. The centre of mass of the chosen motif was restrained along both axes in the plane of the bilayer. The restraint was harmonic with a force constant of 1000 kJ/mol/nm2 In one axis (y) the restraint was centred around zero separation in each window. Along the other axis (x) the centre of the restraint varied from 0.4 nm to 2.7 nm, in 0.1 nm intervals. The restraints were applied to the chosen motifs of the starting structure at the beginning of the simulation; the helices moved rapidly (within 5 ns) to the vicinity of the centre point of restraint for all windows. It is important to note that the separation of the two helices i.e. the starting point to launch each simulation is set independently of the other 'adjacent' simulations. Thus, during each simulation, there is directional increment in the initial separation: there is only a constraint on the distance between the two helices. To estimate the convergence of the system we have used the approach of Yang et al. (54). Then, the PMF profile was calculated using the Weighted Histogram Analysis Method (WHAM) (55) as implemented by Grossfield (http://membrane.urmc.rochester.edu/content/wham/, version 2.0.9) and errors were estimated using the uncorrelated data (see Supporting Information, Text and Fig. S1). For unrestrained CG simulations, we used either systems extracted from the (restrained) PMF simulations or CG models derived from the NMR structure.

Structure selection for AT simulations

From NMR structures in DPPC bilayer: We performed clustering on all of the trajectories using the g_cluster tool with a cut-off of 0.2 nm. This analysis revealed 2 different configurations. The major cluster (representing 95% of the ensemble of trajectories) corresponded to the helices interacting via the glycine motif. The representative structure for this cluster was extracted from run 4 based on the NMR structure. The second cluster represents only 3% of the whole population and defined an intermediate state between States 1 and 2 (see Results below for definition of these states). This structure was characterised by a glycine zipper distance around 0.9 nm and a crossing angle around -10°.

10 11 12

20 21 22

From PMF structures: We have extracted the last frame from one run of the CG simulation to depict State 1 (see Supporting Information Fig. S3-A). This structure had a crossing angle around -20° and the glycine zipper distance was around 0.5 nm. For State 2, we selected the most representative structure from the unrestrained simulation with a crossing angle based on a structure centred around -5°. This structure has a glycine zipper distance around 1.2 nm.

AT simulation details

The conversion from CG models to AT representations was as described previously (56). This method uses a fragment-based protocol to convert CG protein and lipid molecules. Water and counterions molecules were added to equilibrate the systems. For each converted system, a 5000 steps steepest descent minimisation was performed followed by an equilibration phase. For this equilibration, we have gradually decreased the restraints on the protein backbone passing from a constant of 1000 to 250 kJ mol-1 nm-2 in 4 ns. We, then completely removed restraints to run production simulations for duration of 50 to 70 ns depending of the system. The simulations were performed using the GROMOS96 43a1 forcefield. Long-range electrostatic were modelled up to 0.1 nm using the Particle Mesh Ewald (PME) method (57). The same cut-off distance was used to model van der Waals interaction. The reference temperature was 323 K. All simulations were performed at constant temperature, pressure and particle number using semi-isotropic pressure coupling with the Parinello-Rahman barostat (58) and the V-rescale thermostat (59). The integration time step was 2 fs. We have also created a system with a single EphA1 TM helix in a DPPC bilayer using as a starting point a CG model of ideal a-helix embedded in the bilayer using the self-assembly methodology described before. This was then converted to an AT representation and used to launch a 1.2 ^s simulation.

Analysis

Representative structures of the dimer were obtained using the g_cluster tool. Contact analysis was performed on simulations which corresponded to low-energy states on the relevant PMF profiles. The fraction of the simulation for which each residue was in the closest 5 residues to its partner on the opposing helix was calculated. Crossing angles, inter-helix distances, and residue contact calculations were performed using locally written code (60). Structure visualization and some analysis were also performed using VMD (61).

2 Results

7 To avoid biasing our simulation by using the structure of an EphA1 helix extracted from the NMR

10 To evaluate the stability of this structure we performed a long (1.2 ^s) AT simulation of the single

12 helix inserted in a DPPC bilayer. Using the do dssp utility from GROMACS, we analyzed the

15 For the first 0.3 ^s the a-helix was stable (apart from the two extremities switching between helix

17 and coil). Subsequently we observed transient changes in helicity, but overall the secondary

Stability of a single EphA1 transmembrane helix

To avoid biasing our simulation by using the struc dimer structure (PDB id 2K1K), we instead used an ideal a-helix model to depict the TM domain. To evaluate the stability of this structure we performed a long (1.2 ^s) AT simulation of the single helix inserted in a DPPC bilayer. Using the do dssp utility from GROMACS, we analyzed the secondary structure of the TM domain as a function of time (see Supporting Information Fig. S2A).

structure of the TM domain stayed stable throughout the simulation. These transient changes may

EphAl TM domains to be used in our subsequent PMF calculations.

20 reflect limitations of the force field as revealed in long simulations (62). Due to the relative

22 simplicity of the CG model, we considered that an ideal a-helix was a good starting model for the

26 A Free Energy Landscape for Helix-Helix Interactions

28 To understand the free energy landscape for helix-helix interactions within a EphA1 TM helix

30 dimer we undertook potential of mean force (PMF) calculations based on a reaction coordinate

31 corresponding to inter-helix separation. These were aimed to provide a (one dimensional) profile of

33 the free energy of interaction of a pair of parallel EphA1 TM helices in a lipid bilayer as a function

35 of the inter-helix separation. As a prelude to these calculations, we performed five simulations

36 (each of 1.5 ^s duration) starting from two parallel EphA1 TM helices inserted initially 7 nm apart

38 in a CG bilayer containing 271 DPPC molecules. A dimer was formed in each of these five self-

40 assembly simulation within 20 to 800 ns. In four of the five simulations the helices interacted via

41 their N-terminal glycine zipper motifs (i.e. A550x3G554x3G558) as is the case in the NMR structure.

43 (One self-assembly simulation did not yield a stable dimer).

46 On the basis of the self-assembly simulations, we used a TM dimer in which the helices interacted

48 via the N-terminal glycine zipper motif as the starting point for the PMF calculation. In these

50 calculations we progressively moved the two helices apart, using the separation of the centres of

51 mass of their glycine zipper motifs as the PMF reaction coordinate. These configurations were used

53 as starting points to launch 24 independent simulations, each of 7.5 ^s duration, each corresponding

55 to a glycine zipper separation window of width 0.1 nm with the reaction coordinate ranging from

56 0.4 nm to 2.7 nm (Fig. 1A). The profile shows a globally stable state at a glycine zipper separation

58 of 0.5 nm, which we will refer as State 1. The profile also shows a metastable state at a glycine

2 zipper separation of 1.2 nm, which we will call State 2. Cluster analysis was used to produce a

representative structure of each state. In the State 1 structure the glycine zippers pack against one

5 another (Fig. 1B). In contrast, in State 2 the glycine zipper motifs face away from each other. Thus,

7 State 1 is similar to the NMR structure of the EphA1 TM dimer (12). In State 2 (Fig. 1C) the

8 helix/helix interactions are predominantly via residues T544, A559, L563 and V567. Thus the one

10 dimensional free energy landscape suggests at least two possible modes of interaction of the TM

12 helices within an EphA1 TM helix dimer.

15 To help understand the differences between the two states we analysed helix crossing angles and the

17 structural dynamics of the two restrained dimer simulations (Fig. 2), i.e. of the two 7.5 ^s

20 State 1 the main crossing angle is -21° (Fig. 2AB), corresponding to right-handed helix packing.

22 Significantly, two minor populations are seen with crossing angles of -3° and +3°, and the time

simulation trajectories for the glycine zipper separation at 0.5 nm (State 1) and 1.2 nm (State 2). For

course of the crossing angle shows frequent switching between all three crossing angles. For the

25 State 2 simulation, two overlapping major populations are seen, with crossing angles of -5° and +3°

27 (Fig. 2AB). Thus it can be seen that even in the restrained simulations corresponding to individual

'states' along the free energy profile, there is flexibility in the helix/helix interactions as indicated

30 by the multi-modal helix crossing angle distributions.

33 Unrestrained CG simulations of State 1 and State 2 Structures

35 To further characterize the dynamic behaviour of the two states (and also to check that artefacts

38 several representative structures extracted from each state (three for State 1 and two for State 2; see

40 Fig. 2A and Supporting Information Fig. S2 B) as starting points for unrestrained CG simulations,

43 peaks in the crossing angle distribution in Fig. 2B; see above), the three unrestrained simulations

45 yielded comparable crossing angle distributions, with a major population corresponding to a

48 crossing angle distributions are the same as that derived from the State 1 window of the restrained

50 PMF simulations (see also Supporting Information. S3A). For the two representative structures

53 crossing angle of -5° i.e. the same as that from the State 2 window of the restrained PMF

55 simulations. Nevertheless, due to the relatively short time simulation, we cannot exclude that, after

58 State 1.

were not introduced by the reaction coordinate restraint in the PMF calculations (63)), we used

each of 1 p,s duration. Starting from three State 1 structures (selected to correspond to the three

crossing angle of -21° and two minor populations peaks at -3° and +3° (see Fig. 3A). Thus these

extracted from State 2, one of the unrestrained simulations yielded a population with a main

a longer simulation time, the TM domain may pass from the metastable State 2 to the more stable

Indeed, the second unrestrained simulation starting from a State 2 structure (with a crossing angle of +3°) lead to a clear switch to a distribution of crossing angles similar to those seen when starting from State 1, with a major population centred around -20° (Fig. 3A). Examining the crossing angle as a function of the time revealed a shift (at around 200 ns) from a values centred around 0 to ones centred around -20° i.e. from State 2-like behaviour to State 1-like behaviour (see Supporting Information Fig. S3A). We have repeated this simulation three more times, changing initial random velocity seeds, and in each case observed the same behaviour (see Supporting Information Fig. S4).

We further analysed the four unrestrained simulations which showed a transition from crossing angles centred around 0° to crossing angles centred around -20° in terms of the distance between centres of mass of helices and of the residues in contact along these trajectories (Fig. 4; also Supporting Information Fig. S5). This revealed that the EphA1 TM helix dimer can indeed switch from a structure in which the glycine zipper is pointing away from the helix/helix interface (i.e. State 2) to a structure in which this motif forms the interface (i.e. State 1). This transition from State 2 to State 1 passes through an intermediate conformation in which residues involved in the State 2 interface (A559, L563 and V567) are not completely released while the residues composing the glycine motif (i.e. A550, G554 and G558) interact together (Supporting Information Fig. S5A between 530 and 715 and Fig. S5BC after 153.5 ns and 117 ns).

The transition between the two main states is also seen if one tracks the distance between the centres of mass of each helix (Fig. 4B). Thus, for State 1 this distance fluctuates between ca. 0.8 and 0.9 nm (see Fig. 4B after 620 ns and Supporting Information Fig. S5A after 715 ns). In contrast, for State 2 the interhelix distance averages ca. 1.0 nm and exhibits only small fluctuations (see the first 200 ns in Fig. 4B).

Unrestrained CG Simulations Based on the NMR Structure

It has been suggested that the EphA1 dimer may adopt multiple conformations (64). In the context of this and our observations of switching between the two states defined by the PMF calculations we performed CG MD simulations starting from the NMR structure (2K1K) of the EphA1 TM dimer embedded in a lipid bilayer. The results show that over 1 p,s of CG MD simulation the distribution of crossing angles exhibits a major population centred around -20° (Fig. 3B), i.e. the same major population as for the State 1 simulations. This is not surprising as both the State 1 and the NMR structures exhibit right-handed crossing angles and in both structures the helices interact via the glycine zipper motif. The crossing angle for the initial NMR dimer is somewhat larger (ca. -30°) than the modal value of -20° seen for the simulations, but the experimental value is clearly

The structures within the NMR-based simulations remained close to the initial NMR structure. Thus

Significantly, the representative structure of State 1 (see above) had a Ca RMSD of 0.46 nm from

2 within the range observed in either the unrestrained State 1 or the NMR-structure based simulations.

5 the average Ca RMSD from the NMR structure is 0.43 nm, although of course there were

7 considerable dynamic fluctuations during the simulations, with RMSDs ranging from 0.2 to 0.7 nm.

10 the NMR structure, suggesting that the PMF based and NMR based simulations were sampling the

12 same region of conformational space.

15 Interestingly, in all of the simulations initiated from the NMR structures (especially in run 3 in Fig.

17 3B), a minor population was observed with a positive crossing angle, i.e. left-handed packing of the

18 helices. On closer inspection it can be seen that this minor population corresponds to an asymmetric

20 dimer. In this asymmetric dimer one helix interacts through its glycine zipper motif whilst the other

22 helix interacts mainly via residues of the 'State 2' interface, i.e. A559, L563 and V567. (Indeed, this

23 asymmetric dimer conformation constituted the main population for one simulation in NMR

25 structure initiated simulations in a thinner, DLPC, bilayer - see Supporting Information Figs. S3C

27 and S6). Thus, CG MD simulations based either on the PMF-derived stable and metastable states or

28 on the NMR structure reveal complex and dynamic behaviour of the EphAl TM dimer, involving

30 both the canonical glycine zipper interface, but also a more C-terminal interface and an asymmetric

32 conformation. Interestingly, two recent computational studies of the conformational dynamics of

33 the EphAl TM dimer reveal comparable behaviour (46, 64). Both these studies also found the most

35 stable state to correspond to the NMR structure (i.e. a right-handed, glycine-zipper mediated

37 packing of the helices). Interestingly, the simulations of Li et al. (using an implicit bilayer model)

38 also revealed a left-handed dimer conformation with a leucine zipper interaction motif (64).

41 A 2D Conformational Landscape for TM Helix Dimers

43 To characterize the pathway(s) between the two states, we generated a map of the conformational

44 landscape as a function of crossing angle and the glycine zipper interhelix separation from the

46 multiple restrained simulation trajectories used for the PMF calculations. It is also useful to map the

48 unrestrained CG simulations onto this crossing angle/separation landscape (Fig. 5). For the

49 unrestrained simulations starting from State 2 models (Fig. 5A), after fluctuations between -10° and

51 +10° at a separation of ca. 1.2 nm (i.e. State 2 behaviour), the structures switch to a separation of

53 ca. 0.5 nm with a crossing angle fluctuating around ca. -20° (i.e. State 1 behaviour). We also note

54 that the trajectory of the switch between States 1 and 2 passed through an area centred on a glycine

56 motif separation of 0.8 nm and corresponding to the asymmetric dimer. This path broadly follows

58 the more populated areas of the underlying 2D map. For the NMR structure initiated simulations

(Fig. 5B), in three of the simulations the helix dimer stayed mainly in the State 1 region of the map (i.e. at a separation ca. 0.5 nm with fluctuations in the crossing angle between -40° and +10°) while for the fourth (run3 in Fig. 3B) simulation the dimer evolved to a separation of ca. 0.9 nm, again following the underlying 2D landscape, corresponding to an asymmetric intermediate between States 1 and 2. Thus, the 2D map of the conformational landscape of the unrestrained simulations is globally in agreement with possible pathway(s) for interconversion involving an asymmetric dimer (separation ca. 0.8 nm) intermediate between the stable between the stable (State 1) and metastable (State 2) states identified from the one dimensional PMF.

A (Simplified) Model for Interconversion Between States 1 and 2

On the basis of the simulations and analysis presented above we can propose a (simplified) model for interconversion between the two major states of the EphA1 TM helix dimer, as summarised in Fig. 6, in which States 1 and 2 may interconvert via an ensemble of intermediate states which have a glycine zipper separation of ca. 0.8 nm, and which includes asymmetric helix dimers. This suggests that the repacking of the dimer may occur via mechanisms of either concerted or decoupled rotations. In the case of the concerted rotation pathway, the upper parts of the dimer (i.e. residues from T544 to A559) undergo a combined sliding and rotation movement leading to the creation of the glycine zipper (Supporting Information Fig. S5 and Movie S1). This yields an intermediate structure in which the glycine zipper interacts at the dimer interface in addition to residues L555, A559, L563 or A567. The interconversion of States 1 and 2 can also occur via decoupled rotations. In this case, one helix rotates (e.g. Fig. 4A and Supporting Information Movie S2) to create an asymmetric dimer (as also seen in our simulations based on the NMR structure) and subsequently the second helix switches to form an alternative symmetric dimer. Thus, in both cases rotational motions about the helix axes provide a potential mechanism for interconversion of States 1 and 2.

Model Refinement with Atomistic Simulations

To increase the resolution of our description of the two TM dimer states and their dynamic properties, we selected representative structures from the CG simulations and converted them to atomistic resolution (see Methods). Thus, two structures from the unrestrained CG simulations (one structure representing State 1 and the other State 2; see Supporting Information for details of the structure selection method) and also two structures from the NMR-based CG simulations were selected. These four systems (protein and lipids) were converted to atomistic models (56). For each resultant system, we performed a short (50 ns) AT-MD simulation (which was extended to 70 ns for the model based on State 2 to obtain a stable structure). For each of the four simulations the dimer

2 structure evolved during the first few nanoseconds (the Ch RMSD increasing between 0.25 and

0.45 nm). The State 1 model and the two NMR initiated simulations all retained a negative crossing

5 angle (i.e. right handed packing) whereas the State 2 model retained a crossing angle of ca. 0°

7 throughout their respective atomistic simulations.

10 The atomistic simulations reveal exploration of the crossing angle versus glycine zipper distance

12 landscape even though the timescale was small in comparison of with the CG simulations.

Nevertheless, the AT simulations agree quite well with the (CG) map of the landscape (see

simulations. Thus the simulation started from a structure from the NMR run4 CG simulation

15 Supporting Information Fig. S7). Furthermore, the AT simulations explored a substantial range in

17 term of glycine zipper separation, covering the majority of the distance range seen in the CG

20 described the State 1 configuration with a glycine zipper distance mostly around 0.5 nm and a

22 crossing angle centred around -20°. The residues at the interface are mainly from the motif

24 G546X3A550X3G554X3G558, along with some interactions of residues F553, L557, I559, and I565 (Fig.

25 7A). The AT simulations starting with structures from PMF State 1 and NMR run 3 described an

27 intermediate state between States 1 and 2 in which residues involved in the glycine motif as well as

28 some residues involved in State 2 packing were present at the interface (Fig. 7B; also Supporting

30 Information Fig. S8). In particular, these two simulations reveal that residues T544 and of E547 can

32 form inter and intra-helix hydrogen bonds (see Supporting Information Table S1 and Fig. S9 and

37 Significantly, the atomistic simulation based on State 2 revealed a (partial) TM helix dissociation

38 544

39 event. During this event, disruption of the T544 interaction initiated a structural transition in which

40 the glycine zipper distance increased to 1.7 nm. This disruption was driven by the creation of an

42 intra-helix interaction between the carboxyl group of the side-chain of E and backbone amide of

44 T544 (see Supporting Information S9 and Table S1). This interaction was also identified by

45 Bocharov et al. (13). At that point only the C-terminal residues continued to interact (Fig. 7C).

47 Extending this simulation to 200 ns did not result in a complete dissociation of the helices, as a

49 number of residues e.g. R572, R571, S570, F568 or V567 continued to interact (data not shown).

52 Due to the short time scale of AT simulations (<100 ns), we have not explored beyond a small,

55 our CG models. Nevertheless, the atomistic simulations suggest that by starting from different

57 structures yielded by the CG simulations, one may explore the conformation and local dynamics of

58 the EphA1 TM helix dimer, highlighting a stable right handed (State 1) conformation, intermediate

local fraction of the energy landscape and so these AT simulations were more important to refine

states, and a (incomplete) helix dissociation event starting from the (metastable) State 2 configuration.

Discussion

Mechanistic and Structural Implications

Our simulations have revealed two different states of the EphA1 TM helix dimer, as was recently also suggested by Li et al. (64). In the current study it was also possible to characterize potential paths from one state to the other. This allows for an improved understanding of the transition between the two states, emphasising a potential asymmetric dimer as also suggested recently for other RTKs (e.g. (19, 65)) but not previously for EphA1 or its close homologues. This has possible mechanistic implications which are consistent with a number of experimental and computational studies.

The State 1 configuration of the EphA1 dimer is in agreement both with the NMR structure (12) and with other theoretical studies (46, 64, 66). In particular, our model reveals a dimer with a principal interaction around the glycine zipper motif A550X3G554X3G558 as seen in the NMR structure. This interaction is extended to residues in N and C terminal parts e.g. G546 and L561L562 or I565L566 (see Fig. 4 and Fig. S5). We note that Bocharov et al. proposed that this dimer may adopt different conformations and may be quite flexible as suggested by high chemical shift changes (12). Furthermore, other models for this dimer also imply configurations with a large range of crossing angles e.g. (66). On average, the crossing angle for State 1 is around -20°, in comparison with a value of -35° in the NMR structure. From our unrestrained CG simulations, this difference in crossing angles does not seem to be due to changes in membrane thickness. It is conceivable that it may reflect differences between the bilayer and bicelle environments, especially e.g. interactions of the termini with lipid headgroups, and also the possible effects of membrane curvature, as it was suggested by a recent computational study of an unrelated system (67), but further studies will be required to explore this more systematically.

In contrast the State 2 configuration exhibits an interface spread along the whole helix (interacting via residues T544x2E547I548x2V551x3L555x3A559x3L563x2L566V567). Bocharov et al. postulated that the EphA1 dimer could adopt a second conformation (12), proposing a second interaction site for the EphA1 dimer structure at two different potential positions: around the GG4-like motif A560x3G564 or via a heptad like motif IV549x5LL556x5LL563. Our model is broadly in agreement with their second

2 proposed motif. We note that this is also in partial agreement with a left-handed model of EphA1

5 through the A560X3G564 motif, although this motif was seen in transient helix interactions (e.g. Fig.

7 4B). It is useful to note that the presence of a sequence motif does not seems sufficient to predict

10 al. indicated that this latter motif in EphA1 seems not to be conserved across different receptor

12 tyrosine kinase families (68). We superimposed our model of State 2 obtained from AT MD

presented in recent studies by Li et al. (64). We have not observed long-lasting dimerization

and explain TM dimerization (25). Furthermore a sequence alignment presented by Muhle-Goll et

simulations with recent left-handed dimer structures obtained by NMR for EphA2 and for PDGFR

15 (Fig. 8). The Ca RMSD is 0.3 nm between our State 2 model and the EphA2 structure, and 0.33 nm

17 between our State 2 model and the PDGFR structure. Furthermore, this structural alignment

20 interact in our State 2 model. The interface of this model seems also to be in partial agreement with

22 recent MD simulations of the Neu TM dimer in a DPPC bilayer (69). The crossing angle obtained

confirms that those residues at the interface in the two experimental left-handed structures also

in our model for State 2 is distributed around 0° (between -5° and +3°). This value is smaller than

25 that seen in NMR structures for a left-handed dimer (between +15° and + 23° (28)). Whilst we

27 cannot exclude the possibility that the CG forcefield may have an influence on the crossing-angle

30 range: for example models presented by Volynsky et al. have crossing angle values around 5° (see

32 Supplementary Table 1 in (19)), Prakash et al. have angles are distributed around 0° (figure 4 in

values, models constructed with other methodologies present crossing angle values in the same

(39)). More recently, Li et al. proposed left-handed models for the EphA1 dimer with crossing

35 angle ranging from +5° to +11° (figure 4 in (64)). This suggests a convergence of computational

37 approaches.

40 Overall, our results suggest a rotational mechanism for the transition from State 1 to State 2 (Fig. 6

42 and Supporting Information Movies 1 & 2), with the TM helices rotating relative to one another

along their respective axis. Such a rotational movement has been proposed for other related systems

45 e.g. (65, 70). For example, Beevers et al. have also noticed small rotation phenomenon during their

47 100 ns simulations of the Neu TM helix dimer (69) and Prakash et al. also postulated this type of

50 lead to an asymmetric dimer. Recently, Volinsky et al. highlighted asymmetric conformations for

52 the FGFR3 dimer (19), and Reddy et al. also found an asymmetric dimer for FGFR3 mutants using

rotation for the ErbB2 receptor TM domains (39). This rotational mechanism may sometimes also

MD simulations (65). Bocharov et al. also mentioned that the TM domain is involved in a micro to

55 millisecond conformational exchange (12) and found that the Glu547 can play an important role in

56 544

57 the dimer interaction, interacting with Thr544 and thus presenting a substantive chemical shift

59 difference as a function of pH (12). In our proposed pathway, the evolution from the State 2 to State

1 involves the interactions of Thr544 and Glu547. These two residues can be involved in both inter-and intra-helical interactions (see Supporting Information Table S1), so it may be that the creation of intra-helical interactions weaken the inter-helical interactions (Supporting Information Fig. S9 and S10) facilitating the passage from State 2 to State 1, perhaps leading to a partial dissociation of the dimer as seen for one AT simulation (Fig. 7C). This implies a reorganization of the area around residues 547-550, in accordance with the variations of this area noticed by Bocharov et al.

These observations are consistent with a rotation-coupled activation mechanism for the EphAl receptor (12). Although one must exercise caution in extrapolating from the behaviour of a TM helix dimer to the whole EphAl receptor, we suggest that the rotations of TM helices along their axes may result in a rotation of the kinase domains. State 2, through an interaction stabilized along the whole helix, may hold the kinases in a "locked" position while the C-terminal part of the helices in State 1 seem to be more flexible and may provide greater freedom for rearrangements of the juxtamembrane and kinase domains (see Fig. 8-C). In the case of State2, the locked conformation may hold the kinase domain close to the membrane and so obscure phosphorylation and activation sites, as suggested by studies on EphA3 (71) and EGFR (7), while State 1 would provide more flexibility to expose the different sites to be phosphorylated (see Fig. 8C). The substantial ectodomain is also flexible and so that it is more difficult to infer an effect of the TM dimerization mode on it. Nevertheless, it has recently been postulated that a large rearrangement of the ectodomain occurs between the activated and inactivated states for the EphA2 receptor is linked to changes in packing of the TM domains (14). Using this model as a template for EphAl, we suggest that the State2 conformation will lock the ectodomain in a preformed inactive state where the two ectodomains are staggered, as seen in a structure of unliganded EphA2 ectodomain dimer (10). State1 will allow greater displacements of the FN2 domains near the membrane to adopt a wider spacing between the two FN2 domains as seen for the ligand bound form of the EphA2 dimer (10). Thus, State 1 may be related to an active state while State 2 may be linked to an inactive state (14), as has also been suggested for FGFR3 (19) and for EGFR (18). A comparable mechanism was

recently postulated by Bocharov et al. in their "string-puppet" model for the FGFR3 receptor (8). A

question remains: does the free energy landscape of the dimer directly drive the adoption of activate vs. inactivate states? Due to the size of the cytosolic and extracellular domains, we doubt that this is the case but the dimer energetics may determine the passage from one state to the other. Thus, receptor activation may be the result of fine dynamic balances at each level, extracellular, membrane, and cytosolic, and it is the reinforcement or the antagonism of these different

10 11 12

20 21 22

components that would drive a global movement. It will therefore be important to model the whole receptor to better understand these combinatorial and multifaceted dynamic balances.

Free Energies of Dimerization

Our PMF calculations, based on the glycine motif separation, reveal two states of the EphA1 TM helix dimer, and these are supported by unrestrained simulations. However, there are difficulties inherent in comparing absolute dimerization free energies both between simulations and experiments and between different simulations. In particular, experimental as well as theoretical analyses of TM dimerization may be quite sensitive to parameter changes making quantitative comparisons difficult at present.

The free-energy difference between the fully-separated EphA1 TM helices (at a separation of > 2.5 nm) and the dimerized State 1 is ca. -60 ± 2 kJ mol-1. This is significantly larger than an experimental estimate of the association free energy of -15 kJ mol-1 as measured using FRET in DMPC liposomes (16). It is important to remain aware that estimates of the association free energy of TM helices may be sensitive to both the exact extent of the TM helix construct (66) and to the membrane (or membrane-like) environment employed (72, 73). Furthermore, FRET is to some extent an indirect probe of dimerization free energy, corresponding to a finite local concentration in contrast to the limit of infinite dilution when the helices are separated. Thus, it is informative to deconstruct the disagreement between experiment and simulation.

As a point of reference, one may consider the free energy of dimerization of the TM domain of glycophorin A (GpA). Experimentally this can range from ca. -30 kJ mol-1 in e.g. C8E5 detergent micelles (74) to ca. -15 kJ mol-1 in plasma membrane derived vesicles (75). Computationally, estimates include ca. -45 kJ mol-1 (from atomistic simulations in a membrane-mimetic dodecane slab) (35) and ca. -30 (37) to -40 (36) kJ mol-1 via CG simulations in lipid bilayers. Thus both experimental and computational estimates range quite widely.

CG simulations have been used to estimate PMFs for a number of TM helix dimers, with free-energy differences between the fully-separated TM helices and the dimerized state as follows: -30 kJ/mol (37) and -45 kJ/mol for GpA (36); -21 kJ/mol for ErbB1 to -33 kJ/mol for ErbB4 homodimers (38); and -58 kJ/mol for the neuropilin 1/plexin A1 TM helix heterodimer (76). Thus our estimate for the EphA1 State 1 TM dimer is comparable to estimates for related systems using the MARTINI forcefield.

From a purely simulation perspective sampling in our free energy calculations was checked by calculating the statistical inefficiency following Yang et al. (54). This allows one to define the converged part of the simulation. Using this criterion we considered the last 10% of the simulation as fully converged (see Supporting Information Text and Fig. Sl). Furthermore, the depth and shape of the PMF profiles generated are similar to those for GpA, calculated using an either atomistic model (35) or the MARTINI CG model (36).

Although our multiscale approach reveals key aspects of the dynamics and energetics of the EphAl TM dimer, our understanding of this system remains incomplete. The PMF profile is based on a one dimensional helix separation and so depends on the reaction coordinate chosen (here the glycine-motif separation distance). Other reaction coordinates such as crossing angle, helix tilt, etc. may have an influence on the energy profile. Therefore, we would like in the future to extend these studies results to 2D PMF calculations (64), possibly combined with advanced sampling methods (77), in order to provide a fuller characterization of the free energy landscape of helix-helix interactions in this system. Another point on which to reflect is the stability of the State 2 model in our simulations. Even if our CG results compare reasonably well with other computational studies, many of these studies are based on the same forcefield. Indeed, during the short AT MD simulations starting from the State 2 we have seen the early stages of helix-dimer dissociation. It would be interesting to perform longer time scale simulations on TM dimers using different forcefields (in a study comparable to recent work on small water soluble proteins (62)) to enable a more critical comparison of TM helix modeling at different granularities.

Wider Implications

Overall, this study suggests that the interaction of TM helices in EphAl RTK dimers is intrinsically dynamic, reflecting a complex energy landscape for interaction with multiple local minima. Our study provides a novel interpretation of previous experimental results (12, 66) showing two states for the EphAl dimerization. We suggest there is a rotational pathway to pass from one state to the other one. These observations are consistent with a rotation-coupled activation mechanism which has been postulated to govern the EphAl system. This provides a potential mechanism for signalling whereby both extracellular and intracellular events can be linked to a 'repopulation' of the underlying TM helix dimer landscape, rather than a simple process of switching between two states. Related models have been suggested for EphA2 (14), FGFR3 (19) and for EGFR (7, 18). This suggests a degree of convergence on a model in which the TM dimers of RTKs transmembrane dimers readily exchange between conformations. Thus, these studies stress the need to

2 fully understand the dynamic behaviour of the TM dimers of RTKs both from a mechanistic

perspective, and also in order to exploit our understanding for the design of e.g. peptides which

5 target receptor TM domains (78).

10 11 12

20 21 22

http ://pubs .acs.org.

2 Supporting Information

4 Details of additional analysis of the simulations and movies of the State 2 to State 1 transition are

6 provided in the Supporting Information. This material is available free of charge via the Internet at

10 11 12

20 21 22

10 11 12

20 21 22

References

1. Kullander, K., and Klein, R. (2002) Mechanisms and functions of Eph and ephrin signalling, Nat. Rev. Mol. Cell Biol. 3, 475-486.

2. Qin, H., Noberini, R., Huan, X., Shi, J., Pasquale, E., and Song, J. (2010) Structural characterization of the EphA4-Ephrin-B2 complex reveals new features enabling Eph-ephrin binding promiscuity, J. Biol. Chem. 285, 644-654.

3. Bowden, T. A., Aricescu, A. R., Nettleship, J. E., Siebold, C., Rahman-Huq, N., Owens, R. J., Stuart, D. I., and Jones, E. Y. (2009) Structural plasticity of Eph receptor A4 facilitates cross-class ephrin signaling, Structure 17, 1386-1397.

4. Nikolov, D. B., Xu, K., and Himanen, J. P. (2013) Eph/ephrin recognition and the role of Eph/ephrin clusters in signaling initiation, Biochim. Biophys. Acta 1834, 2160-2165.

5. Maruyama, I. N. (2014) Mechanisms of activation of receptor tyrosine kinases: monomers or dimers, Cells 3, 304-330.

6. Schlessinger, J. (2014) Receptor tyrosine kinases: legacy of the first two decades, Cold Spring Harbor Perspectives in Biology 6.

7. Arkhipov, A., Shan, Y., Das, R., Endres, N. F., Eastwood, M. P., Wemmer, D. E., Kuriyan, J., and Shaw, D. E. (2013) Architecture and membrane interactions of the EGF receptor, Cell 152, 557-569.

8. Bocharov, E. V., Lesovoy, D. M., Goncharuk, S. A., Goncharuk, M. V., Hristova, K., and Arseniev, A. S. (2013) Structure of FGFR3 transmembrane domain dimer: implications for signaling and human pathologies, Structure 21, 2087-2093.

9. He, L. J., and Hristova, K. (2012) Physical-chemical principles underlying RTK activation, and their implications for human disease, Biochim. Biophys. Acta 1818, 995-1005.

10. Seiradake, E., Harlos, K., Sutton, G., Aricescu, A., and Jones, E. (2010) An extracellular steric seeding mechanism for Eph-ephrin signaling platform assembly, Nat. Struct. Mol. Biol. 17, 398-402.

11. Himanen, J. P., Yermekbayeva, L., Janes, P. W., Walker, J. R., Xu, K., Atapattu, L., Rajashankar, K. R., Mensinga, A., Lackmann, M., Nikolov, D. B., and Dhe-Paganon, S. (2010) Architecture of Eph receptor clusters, Proc Natl Acad Sci U S A 107, 10860-10865.

12. Bocharov, E. V., Mayzel, M. L., Volynsky, P. E., Goncharuk, M. V., Ermolyuk, Y. S., Schulga, A. A., Artemenko, E. O., Efremov, R. G., and Arseniev, A. S. (2008) Spatial structure and pH-dependent conformational diversity of dimeric transmembrane domain of the receptor tyrosine kinase EphA1, J. Biol. Chem. 283, 29385-29395.

13. Bocharov, E. V., Mayzel, M. L., Volynsky, P. E., Mineev, K. S., Tkach, E. N., Ermolyuk, Y. S., Schulga, A. A., Efremov, R. G., and Arseniev, A. S. (2010) Left-handed dimer of EphA2 transmembrane domain: helix packing diversity among receptor tyrosine kinases, Biophys. J. 98, 881-889.

14. Sharonov, G., Bocharov, E., Kolosov, P., Astapova, M., Arseniev, A., and Feofanov, A. (2014) Point mutations in dimerization motifs of transmembrane domain stabilize active or inactive state of the EphA2 receptor tyrosine kinase, J. Biol. Chem. 289, 14955-14964.

15. Himanen, J., Yermekbayeva, L., Janes, P., Walker, J., Xu, K., Atapattu, L., Rajashankar, K., Mensinga, A., Lackmann, M., Nikolov, D., and Dhe-Paganon, S. (2010) Architecture of Eph

receptor clusters, Proc. Natl. Acad. Sci. USA 107, 10860-10865.

Artemenko, E., Egorova, N., Arseniev, A., and Feofanov, A. (2008) Transmembrane domain of EphA1 receptor forms dimers in membrane-like environment, Biochim. Biophys. Acta 1778, 2361-2367.

Hubert, P., Sawma, P., Duneau, J.-P., Khao, J., Henin, J., Bagnard, D., and Sturgis, J. (2010) Single-spanning transmembrane domains in cell growth and cell-cell interactions: More than meets the eye?, Cell Adh. Migr. 4, 313-324.

Endres, N. F., Das, R., Smith, A. W., Arkhipov, A., Kovacs, E., Huang, Y. J., Pelton, J. G., Shan, Y. B., Shaw, D. E., Wemmer, D. E., Groves, J. T., and Kuriyan, J. (2013) Conformational coupling across the plasma membrane in activation of the EGF receptor, Cell 152, 543-556.

Volynsky, P. E., Polyansky, A. A., Fakhrutdinova, G. N., Bocharov, E., and Efremov, R. G. (2013) Role of dimerization efficiency of transmembrane domains in activation of fibroblast growth factor receptor 3, J. Amer. Chem. Soc. 135, 8105-8108.

Sarabipour, S., and Hristova, K. (2013) FGFR3 transmembrane domain interactions persist in the presence of its extracellular domain, Biophys. J. 105, 165-171. Brooks, A. J., Dai, W., O'Mara, M. L., Abankwa, D., Chhabra, Y., Pelekanos, R. A., Gardon, O., Tunny, K. A., Blucher, K. M., Morton, C. J., Parker, M. W., Sierecki, E., Gambin, Y., Gomez, G. A., Alexandrov, K., Wilson, I. A., Doxastakis, M., Mark, A. E., and Waters, M. J. (2014) Mechanism of activation of protein kinase JAK2 by the growth hormone receptor, Science 344, 710-+.

Lee, J., Miyazaki, M., Romeo, G. R., and Shoelson, S. E. (2014) Insulin receptor activation with transmembrane domain ligands, J. Biol. Chem. 289, 19769-19777. Langosch, D., and Arkin, I. T. (2009) Interaction and conformational dynamics of membrane-spanning protein helices, Prot. Sci. 18, 1343-1358.

Matthews, E., Zoonens, M., and Engelman, D. (2006) Dynamic helix interactions in transmembrane signaling, Cell 127, 447-450.

Li, E., Wimley, W., and Hristova, K. (2012) Transmembrane helix dimerization: beyond the search for sequence motifs, Biochim. Biophys. Acta 1818, 183-193.

Sternberg, M. J. E., and Gullick, W. J. (1990) A sequence motif in the transmembrane region of growth factor receptors with tyrosine kinase activity mediates dimerization, Prot. Engng. 3, 245-248.

MacKenzie, K. R., Prestegard, J. H., and Engelman, D. M. (1997) A transmembrane helix dimer: structure and implications., Science 276, 131-133.

Cymer, F., Veerappan, A., and Schneider, D. (2012) Transmembrane helix-helix interactions are modulated by the sequence context and by lipid bilayer properties, Biochim. Biophys. Acta 1818, 963-973.

Finger, C., Escher, C., and Schneider, D. (2009) The single transmembrane domain of human tyrosine kinases encode self interactions, Science Signal. 2, ra56, 51-58. Fink, A., Sal-Man, N., Gerber, D., and Shai, Y. (2012) Transmembrane domains interactions within the membrane milieu: principles, advances and challenges, Biochim. Biophys. Acta 1818, 974-983.

Russ, W. P., and Engelman, D. M. (1999) TOXCAT: a measure of transmembrane helix association in a biological membrane, Proc. Natl. Acad. Sci. USA 96, 863-868.

You, M., Li, E., Wimley, W., and Hristova, K. (2005) Forster resonance energy transfer in liposomes: measurements of transmembrane helix dimerization in the native bilayer environment, Anal. Biochem. 340, 154-164.

Fleming, K. G., Ackerman, A. L., and Engelman, D. M. (1997) The effect of point mutations on the free energy of transmembrane alpha-helix dimerization, J. Mol. Biol. 272, 266-275.

Fleming, K. G., and Engelman, D. M. (2001) Specificity in transmembrane helix-helix interactions can define a hierarchy of stability for sequence variants, Proc. Natl. Acad. Sci. USA 98, 14340-14344.

Henin, J., Pohorille, A., and Chipot, C. (2005) Insights into the recognition and association of transmembrane a-helices. The free energy of a-helix dimerization in glycophorin A, J. Amer. Chem. Soc. 127, 8478-8484.

Sengupta, D., and Marrink, S. J. (2010) Lipid-mediated interactions tune the association of glycophorin A helix and its disruptive mutants in membranes, Phys. Chem. Chem. Phys. 12, 12987-12996.

Janosi, L., Prakash, A., and Doxastakis, M. (2010) Lipid-modulated sequence-specific association of glycophorin A in membranes, Biophys. J. 99, 284-292. Prakash, A., Janosi, L., and Doxastakis, M. (2010) Self-association of models of transmembrane domains of ErbB receptors in a lipid bilayer, Biophys. J. 99, 3657-3665. Prakash, A., Janosi, L., and Doxastakis, M. (2011) GxxxG motifs, phenylalanine, and cholesterol guide the self-association of transmembrane domains of ErbB2 receptors, Biophys. J. 101, 1949-1958.

Polyansky, A. A., Volynsky, P. E., and Efremov, R. G. (2012) Multistate organization of transmembrane helical protein dimers governed by the host membrane, J. Amer. Chem. Soc. 134, 14390-14400.

Marrink, S. J., Risselada, J., Yefimov, S., Tieleman, D. P., and de Vries, A. H. (2007) The MARTINI forcefield: coarse grained model for biomolecular simulations, J. Phys. Chem. B. 111, 7812-7824.

Monticelli, L., Kandasamy, S. K., Periole, X., Larson, R. G., Tieleman, D. P., and Marrink, S. J. (2008) The MARTINI coarse grained force field: extension to proteins, J. Chem. Theor. Comp. 4, 819-834.

Kalli, A., B.A., H., Campbell, I. D., and Sansom, M. S. P. (2011) A helix heterodimer in a lipid bilayer: structure prediction of the structure of an integrin transmembrane domain via multiscale simulations, Structure 19, 1477-1484.

Psachoulia, E., Bond, P. J., Fowler, P. W., and Sansom, M. S. P. (2008) Helix-helix interactions in membrane proteins: coarse grained simulations of glycophorin helix dimerization, Biochem. 47, 10503-105012.

Wei, P., Zheng, B.-K., Guo, P.-R., Kawakami, T., and Luo, S.-Z. (2013) The association of polar residues in the DAP12 homodimer: TOXCAT and molecular dynamics simulation studies, Biophys. J. 104, 1435-1444.

Zhang, L. Q., Sodt, A. J., Venable, R. M., Pastor, R. W., and Buck, M. (2013) Prediction,

refinement, and persistency of transmembrane helix dimers in lipid bilayers using implicit

and explicit solvent/lipid representations: Microsecond molecular dynamics simulations of

ErbB1/B2 and EphA1, Proteins: Struct. Func. Bioinf. 81, 365-376.

Volynsky, P. E., Mineeva, E. A., Goncharuk, M. V., Ermolyuk, Y. S., Arseniev, A. S., and Efremov, R. G. (2010) Computer simulations and modeling-assisted ToxR screening in deciphering 3D structures of transmembrane alpha-helical dimers: ephrin receptor A1,

Physical Biol. 7, 15.

Bond, P. J., and Sansom, M. S. P. (2006) Insertion and assembly of membrane proteins via simulation, J. Amer. Chem. Soc. 128, 2697-2704.

Bond, P. J., Wee, C. L., and Sansom, M. S. P. (2008) Coarse-grained molecular dynamics simulations of the energetics of helix insertion into a lipid bilayer, Biochem. 47, 1132111331.

Bond, P. J., Faraldo-Gomez, J. D., Deol, S. S., and Sansom, M. S. P. (2006) Membrane protein dynamics and detergent interactions within a crystal: a simulation study of OmpA, Proc. Natl. Acad. Sci. USA 103, 9518-9523.

Bond, P. J., Holyoake, J., Ivetac, A., Khalid, S., and Sansom, M. S. P. (2007) Coarsegrained molecular dynamics simulations of membrane proteins and peptides, J. Struct. Biol. 157, 593-605.

Hess, B., Kutzner, C., van der Spoel, D., and Lindahl, E. (2008) GROMACS 4: algorithms for highly efficient, load-balanced, and scalable molecular simulation, J. Chem. Theor. Comp. 4, 435-447.

Roux, B. (1995) The calculation of the potential of mean force using computer simulations, Comp. Physics Comm. 91, 275-282.

Yang, W., Bitetti-Putzer, R., and Karplus, M. (2004) Free energy simulations: use of reverse cumulative averaging to determine the equilibrated region and the time required for convergence, J. Chem. Phys. 120, 2618-2628.

Kumar, S., Rosenberg, J. M., Bouzida, D., Swendsen, R., and Kollman, P. (1992) The weighted histogram analysis method for free-energy calculations on biomolecules. I. The method, J. Comput. Chem. 13, 1011-1021.

Stansfeld, P. J., and Sansom, M. S. P. (2011) From coarse-grained to atomistic: a serial multi-scale approach to membrane protein simulations, J. Chem. Theor. Comp. 7, 11571166.

Darden, T., York, D., and Pedersen, L. (1993) Particle mesh Ewald - an N.log(N) method for Ewald sums in large systems, J. Chem. Phys. 98, 10089-10092.

Parrinello, M., and Rahman, A. (1981) Polymorphic transitions in single-crystals - a new molecular-dynamics method, J. Appl. Phys. 52, 7182-7190.

Bussi, G., Donadio, D., and Parrinello, M. (2007) Canonical sampling through velocity rescaling, J. Chem. Phys. 126, 014101.

Hall, B. A., Halim, K. A., Buyan, A., Emmanouil, B., and Sansom, M. S. P. (2014) Sidekick for membrane simulations: automated ensemble molecular dynamics simulations of transmembrane helices, J. Chem. Theor. Comp. 10, 2165-2175.

Humphrey, W., Dalke, A., and Schulten, K. (1996) VMD - Visual Molecular Dynamics, J. Molec. Graph. 14, 33-38.

Lindorff-Larsen, K., Maragakis, P., Piana, S., Eastwood, M. P., Dror, R. O., and Shaw, D. E. (2012) Systematic validation of protein force fields against experimental data, PLoS One 7.

10 11 12

20 21 22

63. Neale, C., Bennett, W., Tieleman, D., and R, P. (2011) Statistical convergence of equilibrium properties in simulations of molecular solutes embedded in lipid bilayers, J. Chem. Theory Comput. 7, 4175-4188.

64. Li, P. C., Miyashita, N., Im, W., Ishido, S., and Sugita, Y. (2013) Multidimensional umbrella sampling and replica-exchange molecular dynamics simulations for structure prediction of transmembrane helix dimers, J. Comp. Chem. 35, 300-308.

65. Reddy, T., Manrique, S., Buyan, A., Hall, B. A., Chetwynd, A., and Sansom, M. S. P. (2014) Primary and secondary dimer interfaces of the FGFR3 transmembrane domain: characterization via multiscale molecular dynamics simulations, Biochem. 53, 323-332.

66. Volynsky, P., Mineeva, E., Goncharuk, M., Ermolyuk, Y., Arseniev, A., and Efremov, R. (2010) Computer simulations and modeling-assisted ToxR screening in deciphering 3D structures of transmembrane alpha-helical dimers: ephrin receptor A1, Phys. Biol. 7, 16014.

67. Dominguez, L., Meredith, S. C., Straub, J. E., and Thirumalai, D. (2014) Transmembrane fragment structures of amyloid precursor protein depend on membrane surface curvature, J. Amer. Chem. Soc. 136, 854-857.

68. Muhle-Goll, C., Hoffmann, S., Afonin, S., Grage, S., Polyansky, A., Windisch, D., Zeitler, M., Burck, J., and Ulrich, A. (2012) Hydrophobic matching controls the tilt and stability of the dimeric platelet-derived growth factor receptor (PDGFR) p transmembrane segment, J. Biol. Chem. 287, 26178-26186.

69. Beevers, A., Nash, A., Salazar-Cancino, M., Scott, D., Notman, R., and Dixon, A. (2012) Effects of the oncogenic V(664)E mutation on membrane insertion, structure, and sequence-dependent interactions of the Neu transmembrane domain in micelles and model membranes: an integrated biophysical and simulation study, Biochem. 51, 2558-2668.

70. Fleishman, S. J., Schlessinger, J., and Ben-Tal, N. (2002) A putative molecular-activation switch in the transmembrane domain of erbB2, Proc. Natl. Acad. Sci. USA 99, 1593715940.

71. Janes, P. W., Wimmer-Kleikamp, S. H., Frangakis, A. S., Treble, K., Griesshaber, B., Sabet, O., Grabenbauer, M., Ting, A. Y., Saftig, P., Bastiaens, P. I., and Lackmann, M. (2009) Cytoplasmic relaxation of active Eph controls Ephrin shedding by ADAM10, PLoS Biol. 7.

72. Fisher, L., Engelman, D., and Sturgis, J. (2003) Effect of detergents on the association of the glycophorin a transmembrane helix, Biophys J85, 3097-3105.

73. Zhang, J., and Lazaridis, T. (2006) Calculating the free energy of association of transmembrane helices, Biophys. J. 91, 1710-1723.

74. Fleming, K. G. (2002) Standardizing the free energy change of transmembrane helix-helix interactions, J. Mol. Biol. 323, 563-571.

75. Sarabipour, S., and Hristova, K. (2013) Glycophorin A transmembrane domain dimerization in plasma membrane vesicles derived from CHO, HEK 293T, and A431 cells, Biochim. Biophys. Acta 1828, 1829-1833.

76. Aci-Seche, S., Sawma, P., Hubert, P., Sturgis, J. N., Bagnard, D., Jacob, L., Genest, M., and Garnier, N. (2014) Transmembrane recognition of the semaphorin co-receptors neuropilin 1 and plexin A1: coarse-grained simulations, PLoS ONE 9, e97779.

77. Barducci, A., Bonomi, M., Prakash, M., and Parrinello, M. (2013) Free-energy landscape of protein oligomerization from atomistic simulations, Proc. Natl. Acad. Sci. USA 110, E4708-E4713.

2 78. Li, R. H., Mitra, N., Gratkowski, H., Vilaire, G., Litvinov, R., Nagasami, C., Weisel, J. W.,

3 Lear, J. D., DeGrado, W. F., and Bennett, J. S. (2003) Activation of integrin alpha IIb beta 3

5 by modulation of transmembrane helix associations, Science 300, 795-798.

10 11 12

20 21 22

10 11 12

20 21 22

Figure Legends

Figure 1:

(A) PMF profile for the EphA1 TM helix dimer. The reaction coordinate is the separation of the centres of mass of the glycine zipper motifs. A stable minimum is seen at a glycine zipper separation of 0.5 nm (State 1) corresponding to the NMR configuration. A metastable state is seen at a glycine zipper separation of 1.2 nm (State 2). The error estimation of the PMF profile (red curve) is superimposed in orange. Above the PMF profile, the sequence of the EphA1 TM domain, from G542 to R572' is shown, with residues forming the glycine zipper in red, and in orange additional residues involved in the dimer interaction presented in (B) and (C). (B,C) The EphA1 TM helix, in CG representation, with particles coloured according to the fraction of the simulation for which the particle was one of the closest five particles to its corresponding particle on the other helix for the simulation windows corresponding to (B) State 1 and (C) State 2. Red corresponds to a particle which was one of the closest five particles to its corresponding particle on the other helix for the whole simulation; white to a particle which was never one of the closest five particles to its equivalent particle on the other helix.

Figure 2:

(A) The most representative structures for State 1 and State 2 with helix crossing angles of -21° and -5° respectively. The glycine motif residues are shown in red.

(B) Helix crossing angle distribution based on 7.5 ^s restrained simulations starting from State 1 (green) and State 2 (red). The insets on the right show the corresponding helix crossing angles as a function of time. Note that we have calculated the crossing angle from residue Val549 to Val567 on each helix to limit the calculation to the helix core.

Figure 3:

(A) Helix crossing angle distributions based on the 1 ^s unrestrained dynamics for the most representative structures for State 1 and State 2 in a DPPC bilayer.

(B) Helix crossing angle distribution for the four simulations (each of 1 ^s duration) using the NMR structure of the EphA1 dimer in a DPPC bilayer.

The inset figures show the dimer structure, with red particles depicting residues forming the glycine zipper and orange particles showing additional residues involved in the dimer interaction for State 2. The transparent envelope depicts the surface of one CG helix.

Figure 4:

Detailed view of one of the unrestrained simulations (duration 1 ^s) starting from a State 2 structure, with the vertical arrow at ca. 220 ns indicating the start of a switch to State 1.

(A) Snapshots of the dimer illustrating the progress of the simulation. Red particles depict residues forming the glycine zipper, and orange particles represent residues involved in the C-terminal dimer interaction seen in State 2. Tan empty spheres represent the phosphate groups of the lipid molecules.

(B) Diagrams depicting the number of contact at the dimer interface as a function of residue number (vertical axis) and time (horizontal axis). In between the 'helix 1' and 'helix 2' diagrams, the red curve shows the inter-helix distance as function of time.

Figure 5:

(A) The trajectories of the unrestrained CG simulations starting from the PMF-generated structure of State 2 are shown on top of the conformational map from the (restrained) PMF simulations. Thus the map of the conformational landscape of the helix dimer is shown as a function of crossing angle and the glycine zipper interhelix separation. This map was obtained from the different trajectories used for the PMF calculations. For each 0.1 nm window from the set of PMF trajectories, the crossing angle was calculated and the relative frequencies are shown on a blue-green-red colourscale.

(B) The trajectories of the unrestrained CG simulations starting from the NMR structure are shown on top of the same conformation map as in (A).

The stars depict the final configuration of the dimer for each trajectory. In plotting these trajectories we sample averaged positions (based on 10 steps) every 0.5 ns.

Figure 6:

Schematic representing the putative pathway(s) of interconversion between State 1 and State 2 derived from the unrestrained dynamics simulations. See main text for further details.

Figure 7:

Atomistic simulations starting from two structures were selected from the NMR-based CG simulations and two structures from the unrestrained CG simulations based on PMF models (one structure representing State 1 and the other State 2). In each case a contact map and a representative structure of the helix dimer derived from the last 10 ns of the corresponding AT simulations are

10 11 12

20 21 22

shown. For one helix, the surface is represented as a transparent envelope. Residues shown in red and orange are as previously described.

Figure 8:

(A) Sequence alignment of the EphA1 TM helix with those of two recently determined TM dimer structures, namely for EphA2 and PDGF.

(B) Structural alignment of the EphA1 TM helix dimer (structure in tan; our model of State 2 was extracted from AT MD simulation starting with a State 2 structure having a crossing angle ca. +5°) with recently determined NMR structures for left-handed TM dimers of EphA2 (structure in green; PDB id 2K9Y) and of PDGFR (structure in yellow; PDB id 2L6W).

(C) Schematic diagram depicting potential movements of the EphA1 receptor domains relative to the membrane. On left, State 2 may lock the kinase domains in a configuration where some phosphorylation sites are not accessible and so the receptor is in an inactive state. On right, due to the greater flexibility of State 1, the kinases sites are accessible and so may phosphorylated thus allowing the receptor to adopt an active state.

10 11 12

20 21 22

10 11 12

20 21 22

10 11 12

20 21 22

10 11 12

20 21 22

Chavent et al. Fig. 4

0 ns 220 ns 361 ns 580 ns 618 ns

0 200 400 600 800 1000 Time (ns)

10 11 12

20 21 22

10 11 12

20 21 22

10 11 12

20 21 22

10 11 12

20 21 22

Chavent et al. Fig. 8

10 11 12

20 21 22

ACS Paragon Plus Environment