Structural Insights How PIP2 Imposes Preferred Binding Orientations

Subscriber access provided by UNIV OF CALIFORNIA SAN DIEGO LIBRARIES

Article

Structural Insights How PIP2 Imposes Preferred Binding Orientations of FAK at Lipid Membranes Florian Adorin Herzog, Lukas Braun, Ingmar Schoen, and Viola Vogel J. Phys. Chem. B, Just Accepted Manuscript • DOI: 10.1021/acs.jpcb.6b09349 • Publication Date (Web): 26 Jan 2017 Downloaded from http://pubs.acs.org on January 28, 2017

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.

The Journal of Physical Chemistry B 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.

Page 1 of 34

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

The Journal of Physical Chemistry

Structural Insights How PIP2 Imposes Preferred Binding Orientations of FAK at Lipid Membranes Florian A. Herzog#, Lukas Braun#, Ingmar Schoen*, Viola Vogel* Laboratory of Applied Mechanobiology, Department of Health Sciences and Technology, ETH Zurich, Switzerland #

these authors contributed equally

Corresponding authors: * E-mail: [email protected], [email protected]. Phone: +41 44 632 0887 (V.V.), +41 44 633 7743 (I.S.).

ABSTRACT Focal adhesion kinase (FAK) is a multidomain protein (FERM-Kinase-FAT) with important signaling functions in the regulation of cell-substrate adhesions. Its inactive, autoinhibited form is recruited from the cytoplasm to the plasma membrane, where it becomes activated at PIP2 enriched regions. To elucidate the molecular basis of activation, we performed a systematic screening of binding orientations of FAK’s autoinhibited FERM-Kinase complex, as well as of the dissociated FERM and Kinase domains alone, on model plasma membranes using coarse-grained scFix MARTINI simulations, partially corroborated by atomistic MD simulations. The proteins adopted many more different orientations than previously thought. The presence of PIP2 tuned and narrowed the complex map of competing interfacial orientations. The dissociated FERM domain most frequently interacted with the membrane through its autoinhibitory interface rather than with the ‘basic patch’ residues. These findings suggest a PIP2-dependent activation mechanism in which membrane binding of the dissociated FERM domain competes with the rebinding of the Kinase domain. This competition could promote FAK autophosphorylation on Y397 and subsequent Src binding. The orientation of peripheral proteins at membranes is crucial to understand cell adhesion processes and is furthermore required to exploit steered molecular dynamics to predict how tensile forces might switch their active states. INTRODUCTION The Focal Adhesion Kinase (FAK) is a key signaling molecule in the formation and turnover of focal adhesions at the cell periphery.1-3 It has both scaffolding and enzymatic functions that are tightly regulated through post-translational modifications and interactions with other proteins or ligands. The kinase domain is flanked by the N-terminal FERM domain and the C-terminal focal adhesion targeting (FAT) domain (Fig. 1). Inactive FAK adopts an autoinhibited conformation where the FERM domain binds to the kinase domain

1 ACS Paragon Plus Environment


1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 2 of 34

and blocks its activity.4 A crucial step of activation is the autophosphorylation of Y397 by another FAK molecule in trans upon cell adhesion to the extracellular matrix,2 which requires local enrichment and dimerization of FAK.5 The phosphorylated tyrosine residue then forms a binding site for the Src kinase, which can in turn phosphorylate Y576 and Y577 in the activation loop of FAK. This abolishes the autoinhibition of FAK and increases the catalytic activity of the kinase.4,6 In the active conformation of the kinase, the activation loop is structured,4 and the phosphorylated Y577 interacts with the catalytic core through a network of charged amino acids.4,7 It has been suggested that FAK might be a mechano-regulated protein, whereby mechanical tension might regulate its activation.8,9 Steered molecular dynamics was introduced in 1998 to establish atomistic models how protein structures, and thus functions, are altered by tensile mechanical forces 10-12 . While the thermodynamic stability of a folded protein scales with the buried hydrophobic surface area of the hydrophobic side chains, the mechanical stability of proteins and of protein-protein contacts is determined by the locations at which the pulling forces are acting, and thus the geometry under which clusters of bonds are opened11,13-20. To run SMD simulations that have predictive value, it is thus crucial to grab the protein of interest in the right positions and to apply the force vector in a physiologically relevant orientation. For peripheral membrane proteins, prior knowledge about their orientation(s), and how these might depend on the lipid composition, is thus essential and was a major motivation for our current investigation.

Figure 1: The structure of the Focal Adhesion Kinase (FAK). The Kinase domain (yellow) is flanked by an N-terminal FERM domain (blue) and a C-terminal FAT domain (green). The ATP-binding pocket and important tyrosine residues that are phosphorylated during the activation process are highlighted.

FAK binding to negatively charged phosphatidylinositol-4,5-bisphosphate (PIP2) is required for its activation.21 PIP2 is localized to specific areas of the intracellular leaflet of the plasma membrane and is involved in many biological processes, such as exocytosis, endocytosis, membrane trafficking and the


Page 3 of 34

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


activation of enzymes.22,23 PIP2 binding leads to clustering and partial opening of the interface between the FERM and the Kinase domain, which allows access to the phosphorylation site Y397.24 Since mutation of the basic patch, a cluster of 4 positively charged amino acids, inhibits the binding of the isolated FERM domain as well as of the FERM-Kinase complex to PIP2 containing vesicles,21,24 the basic patch is thought to be responsible for PIP2 binding. In the crystal structure of the basic patch mutant (K216A, K218A, R221A and K222A), the loop Y180-K190 that interacts with the Kinase domain in the autoinhibited conformation could not be resolved, presumably because of its higher flexibility compared to the wildtype24. It was also observed that the basic patch mutant does not bind the Kinase domain as efficiently as the wildtype.25 Moreover, binding of the FERM domain to the receptor kinase c-Met,26 to the FAT domain5 or FAK translocation to the nucleus17 are blocked by this mutant, which indicates a multifunctional role of the basic patch residues. The interaction of autoinhibited FAK with lipid membranes was previously studied with atomistic and coarse-grained molecular dynamics simulations.8,24,27 For the atomistic simulations, the protein was docked to a single PIP2 molecule and this complex was then embedded into the membrane.8,24 However, the time scale of the atomistic simulations in the order of 150 ns did not allow for a sampling of wide ranges of orientations at the membrane. To then ask whether membrane-bound FAK might serve as a mechanosensor, membrane-bound FAK was stretched by SMD8,11, however, it is unclear whether the orientation chosen in these simulations24,8 is actually the most relevant. In contrast, coarse-grained MARTINI simulations which can more efficiently sample the configurational space were used in the past to study the binding process from the solvent to the membrane with the polarizable water model and observed additional binding orientations.27 While the non-bonded interactions in the MARTINI model have been validated for the interactions of lipids with short unstructured peptides,28 the MARTINI force fields have never been rigorously tested for membrane-bound proteins until recently.29 Since we repeatedly observed unrealistic side chain flips of beta-strand residues in several model proteins caused by the lack of physical constraints, we previously introduced and rigorously tested an improved version of the MARTINI protein model called side-chain fix (scFix). In scFix MARTINI, the side chain orientations of residues that are part of -strands are restrained to match the range of orientational fluctuations observed in atomistic simulations.29 This modification was necessary and sufficient to correctly predict the binding pose of the pleckstrin homology domain of PLC-δ1 at fluid POPC/POPE/PIP2 membranes, which we then verified in atomistic simulations.29 Since the elastic network of the MARTINI force field limits conformational changes within the protein, the transition from the closed to the open FERM



1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 4 of 34

Kinase complex cannot be simulated. The present study thus focused on the very endpoints of FAK activation at the membrane, namely on the initial membrane-contact of the closed (autoinhibited) FERM-Kinase complex and the fate of its individual domains in the fully dissociated (open) state. Using scFix MARTINI force fields, we systematically explored possible membrane orientations and analyzed their orientational movements on µs time scales in the absence or presence of PIP2. Different fluid membranes mimicking either the cell membrane in their composition or vesicles from in vitro experiments were tested. Although the obtained orientational maps do not directly reflect the underlying energy landscape, they provide important insights into transient docking poses and their dependence on membrane composition. We now used this approach here to ask how the population of possible binding orientations of the closed FERM-Kinase complex and the FERM domain alone are controlled by residues defining the basic patch, and finally to clarify which of the FAK binding sites are accessible in a membrane-bound state. METHODS The human FAK domains were modeled using Modeller30 based on the autoinhibited crystal structure (pdb 2J0J)4 and the basic patch mutant of the FERM domain (pdb 3zdt).24 The model was coarse-grained with the MARTINI force field v2.228,31 and an elastic network32 connecting the backbone beads was applied with a force constant of 400 kJ mol-1 nm-2 and a distance cutoff between 0.5 and 0.9 nm, except for the residues that are missing in the crystal structure. Additional dihedral potential controlled the side chain orientations of beta strands (scFix).29 The membrane was built with insane.33 The simplified membrane consisted of 6% PIP2 in the upper leaflet and POPC as a background lipid. The more complex membrane was based on the head group distribution of the plasma membrane34 and contained 78 POPC, 114 POPE, 42 DPSM, 48 POPS, 20 POPI, 48 PIP2 (POP2) and 129 cholesterol molecules in the upper leaflet, and 188 POPC, 29 POPS, 100 DPSM and 165 cholesterol molecules in the lower leaflet. The topology for phosphatidylinositol 4,5-bisphosphate (PIP2) is based on the topology for phosphatidylinositol 3,4-bisphosphate (PI34)35 without the dihedral potential for the head group. The membrane was neutralized with Na+ ions, minimized for 1000 steps and equilibrated during 1 µs with a 20 fs time step. To study the protein-lipid interactions, we placed the protein in 32 systematically spaced orientations in close proximity to the membrane (< 7 Å) using the Fibonacci spiral mapped on a sphere.29 The initial set-ups were minimized first for 10 steps in vacuum and then 5000 steps with coarsegrained water and 150 mM NaCl. Then, the system was equilibrated for 1 ns using a 20 fs time step and the velocity rescale temperature36 and Berendsen37 pressure coupling scheme with parameters of 1 and 3 ps-1. For production, the neighbor list was updated every 20th step with the Verlet neighbor search38 and a 20 fs time step. Temperature was controlled with the


Page 5 of 34

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


velocity rescale thermostat and the Parinello-Rahman barostat39 with coupling parameters of 1 and 12 ps-1 and reference temperature and pressure of 310°K and 1 bar. Lennard-Jones and Coulomb forces were cut off at 1.1 nm with the potentials shifted to zero at the cut off using the Potential modifiers. All simulations were run with Gromacs 5.0.6.40 The model of the coarse-grained ATP-Mg was built on the basis of adenine.41 Two Qa beads were added with charge of -2 and -1 to represent the additional phosphate groups and one Qd bead with a charge of +2 for the hydrated magnesium ion. The bond length and angles were based on their distribution in 100 ns long atomistic simulations of ATP bound to the kinase. Since the coarse-grained ATP-Mg moved within the binding pocket, the beads of ATP-Mg were connected to the backbone beads of surrounding residues with elastic bonds, the same way as the backbone beads are connected to each other to preserve the structure of the protein. This ‘ad hoc’ model of ATP-Mg was built to represent the bound ATP and should therefore not be used for other purposes. The phosphorylated amino acids Y576 and Y577 in the active structure of the kinase domain (pdb 2j0k) were modeled by adding a Qa bead with a charge of -2 to the coarse-grained side chain beads 2 and 3 (SC2, SC3) of the tyrosine residue. The bond length was set to 3.6 Å and the force constant to 5000 kJ mol-1 nm-2. The orientation of the protein was monitored over time using the geographical coordinate system with latitude and longitude. A rotation in ψ corresponds to a rotation around the longitudinal axis of the FERM-kinase complex. For all simulation frames where the protein was in contact with the membrane (< 6 Å), the orientation was used to compute the probability density map.29 For the PIP2 interaction, the same 6 Å cut off was used.42 All figures were rendered in VMD.43 To obtain atomistic models from snapshots of the coarse grained simulations we used the backwards.py script distributed by 44. As a target force field we chose CHARMM36 45 and all atoms including water and ions were back mapped. We used the provided mapping definitions for the protein residues. Since for POP2 and DPSM no definitions were available we created them from scratch. The files for the remaining lipids were adapted in order to fit our coarse-grained representation of the respective molecule, as well as the name conventions in the CHARMM topology files. We used NAMD 2.10 46 to minimize the resulting structure for 5000 steps with the conjugate-gradient algorithm and briefly (20 ps) simulated these to ensure that the system was stable. For these runs we used a 2 fs time step, pressure was set to 1 bar and the temperature to 310 K, controlled with a Langevin baro- and thermostat. NAMD structure files were then converted into GROMACS topologies using the VMD package topotools 47. Simulations were run in GROMACS 5.1.4 40 for 100 ns starting from the final coordinates of the



1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 6 of 34

NAMD runs. For these simulations we used a 2 fs time step, the pressure was set to 1 bar and the temperature to 310 K, controlled with a Berendsenbarostat and velocity rescaling. We used the CHARMM36 force field with the TIP3P water model for all atomistic simulations. The parameters for DPSM were taken from 48, those for POPI resp. POP2 were adapted from 49. To estimate the (relative) binding free energies of FAK-KD to the membrane in different orientations we performed Molecular Mechanics PoissonBoltzmann Surface Area (MM/PBSA) calculations with the GROMACS based g_mmpbsa application 50,51. We examined States I, IIb and III on the complex membrane containing PIP2. For each State we picked 3 frames from independent MARTINI simulations, fine grained and simulated them as described above. From each trajectory we extracted 10 snapshots from the last 45 ns. Since sodium ions play a crucial role in mediating the protein’s interactions with the inner leaflet of the membrane, we therefore decided to include the bound ions in the MM/PBSA calculations. The molecular mechanics component was calculated based on the CHARMM36 force field. For the non-polar contributions we performed SASA only calculations with water probe radius set to 1.4 Å and the constants g (gamma) and b (sasaconst) to 0.0226778 KJ/Å2 and 3.84982 KJ/mol, respectively. The polar contributions were calculated by solving the non-linear Poisson-Boltzmann equation, with a vacuum dielectric constant (vdie) of 1, a solvent (sdie) dielectric constant of 80, and a grid spacing of 0.5 Å. As this implementation of MM/PBSA 50 only allows for a constant value for both protein and membrane, we decided to perform a screening of the solute dielectric constant (pdie) over the range e=2-30 to evaluate its influence on the result (see Supplementary Text S1 for detailed discussion). RESULTS The autoinhibited FAK FERM-kinase complex at fluid POPC/PIP2 membranes To answer how FAK in its autoinhibited conformation binds lipid membranes, we first screened for possible membrane docking orientations of the FERMkinase complex on a fluid POPC membrane with 6% PIP2 using 32 coarsegrained simulations starting from evenly distributed initial orientations with 5 µs duration each (see Method section for details how the simulations were set up). This simulation time is sufficient for convergence of the membranebinding process (Fig. 2A). The membrane composition was chosen to match the experimental details in the vesicle pull-down assay performed by Goñi et al.24 First, we assessed the protein-membrane interactions of the FERM-Kinase complex throughout all simulations. For a total of 32 simulations, the number of incidences during which the FERM-Kinase complex was bound to the membrane during a 5 µs interval are shown in Fig. 2A. The average binding kinetics could be fitted with a single exponential function f(t) = a-b*exp(-ct), with an apparent binding rate of c = 1.6 µs-1 and a saturation at a = 31.5 (=


Page 7 of 34

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


98.4%). While these values can be compared between different simulation conditions, they have to be interpreted with care since the simulation time did not allow for sufficient sampling of the off-rate (Fig. S1) and because of the accelerated dynamics in the coarse-grained MARTINI simulations. Next, the evolution of protein orientations in contact with the membrane was assessed over time. The FERM-Kinase fragment typically rotated after it bound the membrane and then adopted relatively well-defined orientations (see Supplementary Movies S1 and S2). The resulting trajectories were then used to compute two-dimensional maps in a geographic coordinate system with longitude ψ and latitude λ that visualize the occupancy of different binding orientations during the (limited) simulation time, as described before.29 Three main orientations (called States) were observed in the resulting map (Fig. 2B), albeit not all States were equally occupied. While State I was consistently sampled in 8 simulations on average, the number of simulations sampling State II decreased over time, while the number of simulations in State III increased (Fig. 2C). In some simulations, the FERM-Kinase complex rotated from one state to another (see Movie S2). A relatively frequent exchange between States II and III with a net flux towards State III was observed, whereas few transitions from State II into State I, and none in the reverse direction, happened during the simulated timeframe of 5 µs (Fig. 2D). This analysis thus indicates that State II was less stable than State I or State III.



1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 8 of 34

Figure 2: Binding dynamics and orientations of the autoinhibited FERM-Kinase complex in contact with the inner leaflet of a fluid membrane containing only PC lipids and 6% PIP2 in the inner leaflet. See also Supplementary Movies S1 and S2. A) Time evolution of the number of simulations in which the FERMKinase complex made contact with the membrane (See also Fig.S1). The data was fitted to an -1 exponential model f(t) = a-b*exp(-c*t) that yielded a = 31.5, b = 16.5, c = 1.6 µs . The total number of 5 µs-simulations was 32. B) Three main peaks were observed in the orientational map derived from all 32 5 µs long simulations showing three distinct orientations of the FERM-Kinase complex at the membrane. C) Number of simulations in State I, II and III over time. The solid lines are moving averages of the actual data (transparent) with a window size of 0.5 µs. D) Number of transitions between States I, II and III. E) The orientation of the FERM-Kinase complex facing the lipid phosphate headgroups in the membrane (gray spheres) is shown for States I, II and III (same color code as in Fig. 1). F) Views of the domains that face the membranes for the three different States. Color-coded is the fraction of time that each amino acid interacted with PIP2. Note that the residues which showed the highest PIP2 interaction probabilities for the three different States are distinctly different, and that many of them are not part of the basic patch.

In State I, the FERM-Kinase complex was lying on one of its flat sides (Fig. 2E, left) and PIP2 frequently interacted with amino acids in the vicinity of the ATP-binding pocket at the putative substrate-binding site (Fig. 2F, left). In the transiently populated State II, FAK was in a position where only the F2 lobe of the FERM domain and the C-terminal lobe of the Kinase domain were in contact with the membrane (Figs. 2E, middle). PIP2 frequently interacted with the basic patch on the FERM domain, but also with positively charged residues in the C-terminal lobe of the Kinase domain (Fig. 2F, middle). In State III, the FERM-Kinase complex was flipped by 180° with respect to State I (Fig. 2E, right), and basic amino acids that are spread over the FERM and the Kinase domains interacted frequently with PIP2 (Fig. 2F, right). Next, we asked if these orientations depend on ATP binding or the charges at the N- and C-termini of the backbone beads. ATP binding has been shown to induce only minor ( III >> IIb was robust for ε>10 (see Supplementary Table S1). This free energy hierarchy directly reflected the amount of clustering of PIP2 molecules underneath the FERM-kinase complex and their mutual interactions (Figs. 4C and S6). Moreover, a flexible loop (aa362-394) interacted tightly with the membrane in State I (Fig. 4D), as also seen in our MARTINI simulations, and thereby contributed substantially to the higher binding energy. The atomistic MD simulations and energy calculations thus consistently supported the conclusions from our coarse-grained orientational screening approach and gave additional insights into how PIP2 affects the binding orientation.

Figure 4: Interactions of the FERM-Kinase complex with the PIP2 cell mimetic membrane in State I revealed by reverse coarse graining and atomistic MD simulations. See also Movie S5. A) A reversecoarse grained and energy-minimized representation of a MARTINI snapshot from State I (color coded as in Fig. 1). As a comparison, the crystal structure aligned by the Kinase domain (yellow) is shown in gray. B) Final frame of a 100 ns atomistic simulation (color coded) based on the MARTINI snapshot (black, same as in A). C) Clustering of PIP2 (purple) and ions (yellow) underneath the FERM-kinase complex. D) Interactions of the flexible loop (licorice) with the membrane.


Page 13 of 34

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


In summary, 10% of PIP2 had by far the biggest effect of all tested lipids on the interaction of FAK’s FERM-kinase complex with lipid membranes, but also the other lipids changed the orientational sampling. In order to more closely mimic the natural situation and without loss of generality, we used the complex membrane mimetic composition for further simulations of the isolated FERM or kinase domain. The FERM domain of FAK at a PIP2-containg cell mimetic membrane Since the basic patch of the FERM domain is thought to predominantly drive the binding of FAK to PIP2,21 we probed the orientation of the FERM domain (amino acids 35-362) alone. To facilitate the comparison between the FERMKinase complex and the individual FAK domain, we chose similar coordinate systems and used the same label for similar states such that the orientation in State “i” of the FERM domain is comparable to State “i” of the FERM-Kinase complex. The individual FERM domain bound to the PIP2-containg cell mimetic membrane much faster (Fig. 5B) than to the PIP2-free membrane (Fig. 5A). On the PIP2-free membrane, the individual FERM domain adopted around four preferential orientations (Fig. 5C) which were approximately equally occupied (Fig. 5E) and showed a similar evolution of occupancy over the time course of the simulation (Fig. 5G). In contrast, in the presence of 10% PIP2, State IX was by far the most frequently adopted orientation (Fig. 5D,F,H). PIP2 thus effectively modulated the binding energy landscape and drove the system towards State IX, where the F1 and F2 modules of the isolated FERM domain are in contact with the cell membrane (Fig. 5I, see also Movie S6). Most importantly, State IX is not seen for the FERM-Kinase complex. Frequent interactions with PIP2 were observed for several amino acids of F1 and F2, including those which bind the Kinase domain and the short β-strand that includes the phosphorylation site Y397 in the autoinhibited conformation (Fig. 5J,K,L and Movie S6). Since these residues are not accessible in the inactive form of FAK,4 these interactions of the FERM domain with the membrane can only occur if the kinase domain and the linker are dissociated.



1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 14 of 34

Figure 5: Membrane binding dynamics of the individual FERM domain to the inner leaflet of a cell mimetic membrane (see Table 1). See also Movie S6. Binding kinetics to the membrane A) without -1 PIP2 (fitted model parameters a=32, b=22.8 and c=0.7 µs ) and B) with 10% PIP2 (parameters a=30.9, b=9.4, c=6.8 µs 1). C,E,G) In the absence of PIP2, four main orientations (States V, VII, VIII, IX) were equally occupied. D,F,H) In the presence of 10% PIP2, similar states were sampled, but State IX was occupied much more frequently than the other states. I) In State IX, residues of F1 and F2 modules interacted with PIP2 molecules in the membrane J) Percentage of time each residue was in contact with PIP2 in state IX. K) The FERM residues involved in PIP2 binding, depicted as a side view and from below (solid surface, color scale), correspond to the autoinhibitory interface, as shown by the overlay of the crystal structure (pdb 2j0j, gray ribbon structure). L) Residues with frequent contact with PIP2 are highlighted and regions of high PIP2 density are shown in red as iso-surfaces.


Page 15 of 34

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


Even though positively charged residues of the basic patch (K216, K218, R221, K222) made contact with PIP2 molecules for 57-79% of the time, other residues (R35, R57, R184, K190, K191, R236) did interact even more frequently (more than 95% of the time, Fig. 5J). Among these, lysines K190 and K191 are spatially in close proximity to the basic patch. To test whether these basic patch amino acids were required for the membrane binding or significantly contributed to the stabilization of a specific binding orientation of the isolated FERM domain, we performed a membrane orientation screening with the basic patch mutant (K216A, K218A, R221A, K222A) based on the respective crystal structure (pdb 3zdt).24 Surprisingly, the mutated FERM domain was in contact with the membrane 97% of the time in our simulations, comparable to the wild type FERM (95%), and the main peaks in the orientational map were conserved, including the highest occupancy of State IX (Fig. S7). We conclude from our simulations that the basic patch of the FAK FERM domain does not change the preferred orientation of the isolated FERM domain at our PIP2-containing lipid model membrane in major ways. The individual Kinase domain at a PIP2-containing cell mimetic membrane Even though the Kinase domain is not known to contribute to the membrane binding of FAK, our simulations indicate that some residues of the Kinase domain engaged in lipid binding (Fig. 2F). Therefore, we investigated the orientation of the isolated Kinase domain with ATP-Mg bound to its pocket in an inactive and active conformation at the PIP2-membrane, which both bound the membrane (Fig. 6A,B). The Kinase domain in the autoinhibited conformation is not phosphorylated and has an unstructured activation loop. For this inactive Kinase domain, the same two main orientations (States I and III) as for the autoinhibited FERM-Kinase complex were observed, and both states were approximately equally occupied (Fig. 6C,E,G). In State I, the active site was facing towards the membrane (Fig. 6I, left, and Movie S7), while it was solvent-accessible in State III (Fig. 6I, right). Amino acids that frequently interacted with PIP2 in State I included R426, K587 and K621 (Fig. 6K), while in State III, amino acids at the other side of the protein (R413, R541, R665) interacted frequently with PIP2. Upon Src binding, the two tyrosine residues 576 and 577 in the activation loop become phosphorylated and a conformational change leads to its adoption of secondary structure, thus referred to as structured activation loop.4 To test for the effect of phosphorylation on membrane binding, we also simulated this ‘active’ phosphorylated Kinase domain and found surprisingly that State III was much less frequently sampled (Fig. 6D,F,H). The most prominent peak in the probability density map was State Ib, that was shifted by about 30° compared to the non-phosphorylated kinase. In this orientation, access to the active site of the Kinase was limited (Fig. 6J and Movie S8). PIP2 frequently interacted with basic residues (R508, R581, R583, R597, K621) in the vicinity of the ATP-binding pocket at the predicted substrate binding site (Fig. 6L).



1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 16 of 34

Small and membrane-proximal substrates could potentially compete with PIP2 for binding these residues.

Figure 6: Binding dynamics of the isolated Kinase domain to the inner leaflet of a cell mimetic membrane containing 10% PIP2. See also Movies S7 and S8. Binding kinetics of A) the inactive (fitted -1 model parameters a=30.4, b=16.4, c=2.3 µs ) and B) the active (parameters a = 29.9, b = 18, c = 7.3 -1 µs ) C) The inactive kinase with the unstructured activation loop mainly sampled four states, with a strong preference towards State I and III (E, G). D) The active conformation with the structured activation loop different orientations. State Ib is shifted and preferred over other states (F, H). I) In the inactive FAK, the ATP-binding pocket is facing the membrane in State I, while it is solvent accessible in State III. J) In the active conformation, the ATP binding site faces the membrane in State Ib, but in a


Page 17 of 34

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


more tilted orientation than in State I of the inactive Kinase. K) Interaction of the inactive Kinase with PIP2 is shown for each amino acid in State I and III. L) PIP2 interaction of the active Kinase in State Ib.

DISCUSSION Since little is known about the orientational distribution of FAK at the cell membrane, we studied the effect of lipid composition, binding of ligands, conformational changes and posttranslational modifications on the membrane binding pose of FAK using scFix MARTINI simulations. We found that PIP2 had a major impact on the membrane binding velocity and orientation of FAK and its domains (Figs. 2-6), suggesting an important role of PIP2 during early activation events. In particular, the orientational map of the FERM domain is highly sensitive to the presence of PIP2 (Fig. 5, Movie S6), as are those of the FERM-Kinase complex (Figs. 2, 3, 4 and Movies S1-S4) and the isolated kinase domain (Fig. 6, Movies S7 and S8). Upon integrin-mediated cell adhesion, FAK accumulates at nascent adhesions and becomes activated to orchestrate the formation and turnover of focal adhesions.3 Activation of FAK at the membrane is accompanied by conformational changes as revealed by experiments with FRET-labeled FAK constructs.55,21,24 It is thus commonly assumed that FAK makes the first intimate contact with the membrane in its autoinhibited conformation.24 Our simulations show that the autoinhibited FERM-Kinase complex had a rather low affinity for and a relatively undefined orientation at the inner leaflet of cell membranes without PIP2 (Fig. 3A,C). In contrast, it consistently bound PIP2enriched plasma membranes in two main orientations which we termed here States I and III (Figs. 2B-C, 3B+D, Movies S1-S5). The PIP2-induced orientations were robust with respect to background lipids, the charge of the N- and C-terminus, or whether ATP-Mg was bound or not (Figs. 2, 3, S1). Importantly though and not predicted before, the complex lay flat on the PIP2membrane, with large areas of the FERM and the Kinase domain in contact with the membrane. The ATP and substrate binding sites were only accessible in State III, whereas the binding pockets were hidden by the membrane in State I (Figs. 2E, 3J) which has major functional implications. Most significantly, neither of these States matched the expected docked orientation as assumed in previous computational studies24,52. The docked orientation corresponds roughly to State II in our simulations which was only transiently occupied and showed a net flux towards States I and III (Fig. 2C-D and Movie S2). MM/PBSA energetic calculations confirmed this result (Table S1), providing an independent validation of our screening approach beyond previous evidence29.



1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 18 of 34

Figure 7: Schematic model of the first activation steps of FAK. 1) PIP2 favors the membrane bound state of the autoinhibited FERM-Kinase complex. 2) Dissociation of the FERM-Kinase domain is followed by 3) a reorientation of the FERM domain and Y397 unbinding. 4) Subsequent autophosphorylation of Y397 (marked by an asterisk) blocks the rebinding of pY397 to the FERM domain. 5) Further phosphorylation of Y576 and Y577 by Src.

PIP2 binding is thought to partially open up the FERM-Kinase complex, which allows for phosphorylation of Y397.24 Indeed, salt bridges that stabilize the FERM-Kinase interface (Glu636 with Arg184/Lys190) were opened up in 100 ns long atomistic simulations (Figs. 4, S3-S5, and Movie S5). What would happen if the FERM-Kinase complex completely dissociated at the membrane? Interestingly, FAK’s FERM domain alone preferentially adopted a different orientation than the FERM-Kinase complex in the presence of PIP2 (compare Fig. 5 with Fig. 3). While State III was common to the FERM-Kinase complex and to the individual FERM domain, a new State IX appeared when the inhibitory Kinase domain was removed from the complex. A strong tendency towards State IX was observed in the presence of PIP2 (Fig. 5D,F,H and Movie S6). Amino acids that interact with the Kinase domain in the autoinhibited conformation are found here to be in contact with the cell membrane (Fig. 5K,L). Our simulations thus reveal for the first time that these protein-lipid interactions compete with the rebinding of the Kinase domain to the FERM domain. Based on these simulations, we would thus like to propose the following sequential steps by which rapid rebinding of the FERM to the Kinase domain is prevented such that sufficient time is allowed for the residues Y576/Y577 to


Page 19 of 34

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


be phosphorylated (Fig. 7): Once the FERM domain is released from the Kinase domain, it can rotate at the membrane from State III to IX, thereby burying the autoinhibitory interface and the binding site of Y397 on its F1 module. Since burying these residues in the lipid membrane is expected to significantly lower the rebinding rate of the FERM with the Kinase domain, we now propose that the rotational movement of the FERM domain at the PIP2 containing inner membrane leaflet will significantly increase the lifetime of the Kinase domain in a FERM-dissociated state. In fact, two different FRET sensor were developed to investigate the activation process.21,55 Despite the similar design of the probes, an opposite response in FRET was observed upon activation, which does support a change in orientation rather than a change in the FERM-Kinase distance.55 Our data now suggest that there might be a dynamic equilibrium between the FERM-Kinase and the FERMmembrane interactions. Since the previous FRET measurements correspond to ensemble averages,24 it is possible that the ‘partial opening’ corresponds to the average of two populations: one where the FERM domain is bound to the Kinase, and another, where the lipids bind the FERM domain. Please note though that we did not assess the binding affinity of these states from the orientational maps because of the small number of sampled transitions. Moreover, the predicted binding orientation of the FERM domain could explain the lower affinity of the FERM mutant 180/183A for PIP2 containing lipid vesicles compared to wild-type,24 since these amino acid are part of the protein-lipid interface in State IX. PIP2 directly interacts with amino acids on the F1 lobe of the FERM domain (see Movie S6) that bind the short linker segment that includes the phosphorylation site Y397. Our simulations now predict that this binding site might not be accessible when the FERM domain is bound to a PIP2 containing membrane (Fig. 5L). It is thus tempting to speculate that PIP2 binding would prevent not only the rebinding of the Kinase domain, but also the rebinding of the short linker segment that includes Y397. In the dissociated case, Y397 would be accessible for phosphorylation. Since the posttranslational modification of Y397 prevents the rebinding to the FERM domain and enables the further phosphorylation of Y576 and Y577 that block the FERM-kinase interaction, the phosphorylation of Y397 would effectively stabilize the fully dissociated state (Fig. 7). How might membrane binding affect the activity of the Kinase domain? The two main orientations of the FERM-Kinase complex at the membrane are also preferentially sampled by the Kinase domain alone (State I and III, Fig. 6C and Movie S7). In fact, it seems that the Kinase domain is mainly responsible for the orientation of the autoinhibited FERM-Kinase complex at the membrane. However, the orientation of the activated Kinase domain at the membrane is changed upon phosphorylation of Y576 and Y577 and the associated conformational change in the activation loop (Fig. 6D). State III is shifted upon phosphorylation of Y576 and Y577 and much less sampled (Fig. 6F, H). In the main orientation, the substrate-binding pocket is facing the lipid membrane (Movie S8). There is only limited access to the binding sites, but



1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 20 of 34

the Kinase could potentially phosphorylate unstructured regions of peripheral or transmembrane proteins that are able to compete with the lipid head groups that bind to the Kinase domain. It is also possible that the Kinase is not catalytically active when bound to the cell membrane. Only by binding to different ligands or by detaching the kinase from the membrane, for example by mechanical force, the substrate-binding site would become fully accessible. Would the PIP2-directed membrane binding orientation of the FERM-Kinase complex be compatible with FAK dimerization as it has been shown that its dimerization is required for autophosphorylation of Y397?5 Indeed, our data suggest that this might occur. The FERM F3 lobe forms the dimerization interface for an elongated, arch-shaped dimer.5 If we consider a certain flexibility of the membrane bound states along the longitudinal axis of the FERM-Kinase complex, the two main States I and III are thus in a predisposed orientation for dimerization. Moreover, PIP2 mediates FAK clustering at the cell membrane via the basic patch in the F2 lobe of the FERM.24 Importantly, the basic-patch mutant of the FERM domain (K216A, K218A, R221A, K22A) bound the membrane in our simulations with similar dynamics and orientations as the wild-type (Fig. S2). In the main orientation (State IX), the basic patch residues are not at the direct protein-membrane interface, but the relatively large head group of PIP2 or other ligands can interact with these residues (Fig. 5K) which is in contrast to assumed models where the basic patch residues are completely buried in the membrane.24,52 In our simulations, the basic patch did not directly determine the binding affinity of the FERM domain for PIP2 containing membranes. As an alternative explanation, we hypothesize that the basic patch could mediate clustering of FAK at PIP2 containing membranes and thereby enhance association with membranes by an avidity effect. CONCLUSIONS The landscape of possible orientations of FAK at the cell membrane is far more complex than previously thought and has major implications regarding the recruitment of FAK to the inner lipid leaflet, as well as for the PIP2 induced activation process. Since phosphorylation of Y576/Y577 by Src and Y397 by FAK requires the dissociation of the FERM and the Kinase domains, a multistep activation process might be required to prevent rapid rebinding. Rotation of the dissociated FERM domain at the PIP2 membrane might be essential to extend the lifetime of the dissociated state, such as to enhance the probability that the Kinase is given sufficient time to diffuse towards another FAK molecule and bind Y397 as required for activation. The preferential orientation of the FERM domain on the membrane can be experimentally tested for example by electron paramagnetic resonance (EPR) to probe for the solvent accessibility of amino acids.56 The combination of EPR with coarse-grained simulation is in general a powerful method, since it can avoid the screening of many mutations and guide the experimentalist


Page 21 of 34

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


towards more specific and fewer target mutations. The back-transformation of promising coarse-grained states into all-atom representations can further help to mechanistically address the effect of specific residues. Potential PIP2directed FAK dimerization could also be further studied by coarse-grained57 or with large scale atomistic molecular dynamics simulations58 to refine the structural models. Given the complexity of interactions involved, only a combined theoretical and experimental effort will provide the necessary information to decipher the detailed activation mechanism of FAK at cellular membranes. The here presented systematic screening by coarse-grained simulations and its extension to back-transformed atomistic simulations and energetic calculations substantially facilitates such an integrated computational-experimental research strategy. ASSOCIATED CONTENT This article is accompanied by Supporting Information on the website containing Supplementary Figures S1-S7, Supplementary Movies S1-S8, Supplementary Table S1, and Supplementary Text S1 which discusses the MM/PBSA calculations.

ACKNOWLEDGEMENTS We deeply thank and remember Klaus Schulten for a longlasting friendship and collaboration that started over 20 years ago. With his enormous energy and talent, he pushed the frontiers of atomistic simulations, from single proteins solvated by explicit water molecules to finally resolving the capsule of a virus. Rather than going an entrepreneurial route, he was passionately training the next generation of computational and experimental scientists and was making his tools available to a broad community worldwide. His educational training workshops profoundly changed the careers of five generations of my graduate students alone (André Krammer, David Craig, Eileen Faucher, Samuel Hertig, Garif Yalak, Florian Herzog and Lukas Braun), and our views how mechanical forces can switch the structurefunction relationships of proteins. His giant footprints will stay, yet we greatly miss him. This work was supported by an ERC Advanced Grant (VV), by ETH Zurich and by a grant from the Swiss National Supercomputing Centre (CSCS) under project ID s514 and s723.

REFERENCES (1)

(2)

Mitra, S. K.; Hanson, D. A.; Schlaepfer, D. D. Focal Adhesion Kinase: in Command and Control of Cell Motility. Nat. Rev. Mol. Cell Biol. 2005, 6 (1), 56–68. Schaller, M. D. Cellular Functions of FAK Kinases: Insight Into Molecular Mechanisms and Novel Functions. J Cell Sci 2010, 123 (Pt 21 ACS Paragon Plus Environment


1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

(3)

(4)

(5)

(6)

(7)

(8)

(9)

(10)

(11)

(12) (13)

(14)

(15)

(16)

(17)

Page 22 of 34

7), 1007–1013. Arold, S. T. How Focal Adhesion Kinase Achieves Regulation by Linking Ligand Binding, Localization and Action. Curr. Opin. Struc. Biol. 2011, 21 (6), 808–813. Lietha, D.; Cai, X.; Ceccarelli, D. F. J.; Li, Y.; Schaller, M. D.; Eck, M. J. Structural Basis for the Autoinhibition of Focal Adhesion Kinase. Cell 2007, 129 (6), 1177–1187. Brami-Cherrier, K.; Gervasi, N.; Arsenieva, D.; Walkiewicz, K.; Boutterin, M.-C.; Ortega, A.; Leonard, P. G.; Seantier, B.; Gasmi, L.; Bouceba, T.; et al. FAK Dimerization Controls Its Kinase-Dependent Functions at Focal Adhesions. The EMBO Journal 2014, 33 (4), 356– 370. Calalb, M. B.; Polte, T. R.; Hanks, S. K. Tyrosine Phosphorylation of Focal Adhesion Kinase at Sites in the Catalytic Domain Regulates Kinase Activity: a Role for Src Family Kinases. Molecular and Cellular Biology 1995, 15 (2), 954–963. Herzog, F. A.; Vogel, V. Multiple Steps to Activate FAK's Kinase Domain: Adaptation to Confined Environments? Biophys. J. 2013, 104 (11), 2521–2529. Zhou, J.; Aponte-Santamaría, C.; Sturm, S.; Bullerjahn, J. T.; Bronowska, A.; Gräter, F. Mechanism of Focal Adhesion Kinase Mechanosensing. PLoS Comput Biol 2015, 11 (11), e1004593. Mofrad, M. R.; Golji, J.; Rahim, N. A. Force-Induced Unfolding of the Focal Adhesion Targeting Domain and the Influence of Paxillin Binding. Mech. Chem. Biosyst. 2004. Lu, H.; Isralewitz, B.; Krammer, A.; Vogel, V.; Schulten, K. Unfolding of Titin Immunoglobulin Domains by Steered Molecular Dynamics Simulation. Biophys. J. 1998, 75 (2), 662–671. Isralewitz, B.; Gao, M.; Schulten, K. Steered Molecular Dynamics and Mechanical Functions of Proteins. Curr. Opin. Struc. Biol. 2001, 11 (2), 224–230. Sotomayor, M.; Schulten, K. Single-Molecule Experiments in Vitro and in Silico. Science 2007, 316 (5828), 1144–1148. Marszalek, P. E.; Lu, H.; Li, H.; Carrion-Vazquez, M.; Oberhauser, A. F.; Schulten, K.; Fernandez, J. M. Mechanical Unfolding Intermediates in Titin Modules. Nature 1999, 402 (6757), 100–103. Craig, D.; Krammer, A.; Schulten, K.; Vogel, V. Comparison of the Early Stages of Forced Unfolding for Fibronectin Type III Modules. PNAS 2001, 98 (10), 5590–5595. Gao, M.; Craig, D.; Lequin, O.; Campbell, I. D.; Vogel, V.; Schulten, K. Structure and Functional Significance of Mechanically Unfolded Fibronectin Type III1 Intermediates. PNAS 2003, 100 (25), 14784– 14789. Bayas, M. V.; Schulten, K.; Leckband, D. Forced Dissociation of the Strand Dimer Interface Between C-Cadherin Ectodomains. Mech. Chem. Biosyst. 2004. Lim, S.-T.; Chen, X. L.; Lim, Y.; Hanson, D. A.; Vo, T.-T.; Howerton, K.; Larocque, N.; Fisher, S. J.; Schlaepfer, D. D.; Ilic, D. Nuclear FAK 22 ACS Paragon Plus Environment

Page 23 of 34

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


(18)

(19)

(20)

(21)

(22)

(23)

(24)

(25)

(26)

(27)

(28)

(29)

(30)

Promotes Cell Proliferation and Survival Through FERM-Enhanced P53 Degradation. Molecular Cell 2008, 29 (1), 9–22. Schoeler, C.; Malinowska, K. H.; Bernardi, R. C.; Milles, L. F.; Jobst, M. A.; Durner, E.; Ott, W.; Fried, D. B.; Bayer, E. A.; Schulten, K.; et al. Ultrastable Cellulosome-Adhesion Complex Tightens Under Load. Nature Communications 2014, 5, 5635. Schoeler, C.; Bernardi, R. C.; Malinowska, K. H.; Durner, E.; Ott, W.; Bayer, E. A.; Schulten, K.; Nash, M. A.; Gaub, H. E. Mapping Mechanical Force Propagation Through Biomolecular Complexes. Nano Lett. 2015, 15 (11), 7370–7376. Kufer, S. K.; Puchner, E. M.; Gumpp, H.; Liedl, T.; Gaub, H. E. SingleMolecule Cut-and-Paste Surface Assembly. Science 2008, 319 (5863), 594–596. Cai, X.; Lietha, D.; Ceccarelli, D. F.; Karginov, A. V.; Rajfur, Z.; Jacobson, K.; Hahn, K. M.; Eck, M. J.; Schaller, M. D. Spatial and Temporal Regulation of Focal Adhesion Kinase Activity in Living Cells. Molecular and Cellular Biology 2007, 28 (1), 201–214. McLaughlin, S.; Wang, J.; Gambhir, A.; Murray, D. PIP(2) and Proteins: Interactions, Organization, and Information Flow. Annu Rev Biophys Biomol Struct 2002, 31 (1), 151–175. McLaughlin, S.; Murray, D. Plasma Membrane Phosphoinositide Organization by Protein Electrostatics. Nature 2005, 438 (7068), 605– 611. Goñi, G. M.; Epifano, C.; Boskovic, J.; Camacho-Artacho, M.; Zhou, J.; Bronowska, A.; Martín, M. T.; Eck, M. J.; Kremer, L.; Gräter, F.; et al. Phosphatidylinositol 4,5-Bisphosphate Triggers Activation of Focal Adhesion Kinase by Inducing Clustering and Conformational Changes. Proc. Natl. Acad. Sci. U.S.A. 2014, 111 (31), E3177–E3186. Dunty, J. M.; Gabarra-Niecko, V.; King, M. L.; Ceccarelli, D. F. J.; Eck, M. J.; Schaller, M. D. FERM Domain Interaction Promotes FAK Signaling. Molecular and Cellular Biology 2004, 24 (12), 5353–5368. Chen, S. Y.; Chen, H.-C. Direct Interaction of Focal Adhesion Kinase (FAK) with Met Is Required for FAK to Promote Hepatocyte Growth Factor-Induced Cell Invasion. Molecular and Cellular Biology 2006, 26 (13), 5155–5167. Feng, J.; Mertz, B. Novel Phosphotidylinositol 4,5-Bisphosphate Binding Sites on Focal Adhesion Kinase. PLOS ONE 2015, 10 (7), e0132833. de Jong, D. H.; Singh, G.; Bennett, W. F. D.; Arnarez, C.; Wassenaar, T. A.; Schäfer, L. V.; Periole, X.; Tieleman, D. P.; Marrink, S. J. Improved Parameters for the Martini Coarse-Grained Protein Force Field. J. Chem. Theory Comput. 2013, 9 (1), 687–697. Herzog, F. A.; Braun, L.; Schoen, I.; Vogel, V. Improved Side Chain Dynamics in MARTINI Simulations of Protein–Lipid Interfaces. J. Chem. Theory Comput. 2016, 12 (5), 2446–2458. Webb, B.; Sali, A. Comparative Protein Structure Modeling Using MODELLER; John Wiley & Sons, Inc.: Hoboken, NJ, USA, 2014; pp 5.6.1–5.6.32. 23 ACS Paragon Plus Environment


1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

(31)

(32)

(33)

(34)

(35)

(36) (37)

(38)

(39)

(40)

(41)

(42)

(43) (44)

(45)

Page 24 of 34

Monticelli, L.; Kandasamy, S. K.; Periole, X.; Larson, R. G.; Tieleman, D. P.; Marrink, S.-J. The MARTINI Coarse-Grained Force Field: Extension to Proteins. J. Chem. Theory Comput. 2008, 4 (5), 819– 834. Periole, X.; Cavalli, M.; Marrink, S.-J.; Ceruso, M. A. Combining an Elastic Network with a Coarse-Grained Molecular Force Field: Structure, Dynamics, and Intermolecular Recognition. J. Chem. Theory Comput. 2009, 5 (9), 2531–2543. Wassenaar, T. A.; Ingólfsson, H. I.; Böckmann, R. A.; Tieleman, D. P.; Marrink, S. J. Computational Lipidomics with Insane: a Versatile Tool for Generating Custom Membranes for Molecular Simulations. J. Chem. Theory Comput. 2015, 11 (5), 2144–2155. Ingólfsson, H. I.; Melo, M. N.; van Eerden, F. J.; Arnarez, C.; López, C. A.; Wassenaar, T. A.; Periole, X.; de Vries, A. H.; Tieleman, D. P.; Marrink, S. J. Lipid Organization of the Plasma Membrane. J. Am. Chem. Soc. 2014, 136 (41), 14554–14559. López, C. A.; Sovova, Z.; van Eerden, F. J.; de Vries, A. H.; Marrink, S. J. Martini Force Field Parameters for Glycolipids. J. Chem. Theory Comput. 2013, 9 (3), 1694–1708. Bussi, G.; Donadio, D.; Parrinello, M. Canonical Sampling Through Velocity Rescaling. J. Chem. Phys. 2007, 126 (1), 014101. Berendsen, H. J. C.; Postma, J. P. M.; van Gunsteren, W. F.; DiNola, A.; Haak, J. R. Molecular Dynamics with Coupling to an External Bath. J. Chem. Phys. 1984, 81 (8), 3684–3690. Verlet, L. Computer “Experiments” on Classical Fluids. I. Thermodynamical Properties of Lennard-Jones Molecules. Phys. Rev. 1967, 159 (1), 98. Parrinello, M.; Rahman, A. Polymorphic Transitions in Single Crystals: a New Molecular Dynamics Method. Journal of Applied Physics 1981, 52 (12), 7182–7190. Pronk, S.; Páll, S.; Schulz, R.; Larsson, P.; Bjelkmar, P.; Apostolov, R.; Shirts, M. R.; Smith, J. C.; Kasson, P. M.; van der Spoel, D.; et al. GROMACS 4.5: a High-Throughput and Highly Parallel Open Source Molecular Simulation Toolkit. 2013, 29 (7), 845–854. Uusitalo, J. J.; Ingólfsson, H. I.; Akhshi, P.; Tieleman, D. P.; Marrink, S. J. Martini Coarse-Grained Force Field: Extension to DNA. J. Chem. Theory Comput. 2015, 11 (8), 3932–3945. Psachoulia, E.; Sansom, M. S. P. Interactions of the Pleckstrin Homology Domain with Phosphatidylinositol Phosphate and Membranes: Characterization via Molecular Dynamics Simulations. Biochemistry 2008, 47 (14), 4211–4220. Humphrey, W.; Dalke, A.; Schulten, K. VMD: Visual Molecular Dynamics. Journal of Molecular Graphics 1996, 14 (1), 33–38. Wassenaar, T. A.; Pluhackova, K.; Böckmann, R. A.; Marrink, S. J.; Tieleman, D. P. Going Backward: a Flexible Geometric Approach to Reverse Transformation From Coarse Grained to Atomistic Models. J. Chem. Theory Comput. 2014, 10 (2), 676–690. Best, R. B.; Zhu, X.; Shim, J.; Lopes, P. E. M.; Mittal, J.; Feig, M.; 24 ACS Paragon Plus Environment

Page 25 of 34

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


(46)

(47)

(48)

(49)

(50)

(51)

(52)

(53)

(54)

(55)

(56)

(57)

Alexander D MacKerell, J. Optimization of the Additive CHARMM AllAtom Protein Force Field Targeting Improved Sampling of the Backbone Φ, Ψ and Side-Chain Χ1 and Χ2 Dihedral Angles. J. Chem. Theory Comput. 2012, 8 (9), 3257–3273. Phillips, J. C.; Braun, R.; Wang, W.; Gumbart, J.; Tajkhorshid, E.; Villa, E.; Chipot, C.; Skeel, R. D.; Kalé, L.; Schulten, K. Scalable Molecular Dynamics with NAMD. Journal of Computational Chemistry 2005, 26 (16), 1781–1802. Vermaas, J. V.; Hardy, D. J.; Stone, J. E.; Tajkhorshid, E.; Kohlmeyer, A. TopoGromacs: Automated Topology Conversion From CHARMM to GROMACS Within VMD. J. Chem. Inf. Model. 2016, 56 (6), 1112– 1116. Venable, R. M.; Sodt, A. J.; Rogaski, B.; Rui, H.; Hatcher, E.; MacKerell, A. D.; Pastor, R. W.; Klauda, J. B. CHARMM All-Atom Additive Force Field for Sphingomyelin: Elucidation of Hydrogen Bonding and of Positive Curvature. Biophysical Journal 2014, 107 (1), 134–145. Lupyan, D.; Mezei, M.; Logothetis, D. E.; Osman, R. A Molecular Dynamics Investigation of Lipid Bilayer Perturbation by PIP2. Biophysical Journal 2010, 98 (2), 240–247. Kumari, R.; Kumar, R.; Open Source Drug Discovery Consortium; Lynn, A. G_Mmpbsa--a GROMACS Tool for High-Throughput MMPBSA Calculations. J. Chem. Inf. Model. 2014, 54 (7), 1951–1962. Baker, N. A.; Sept, D.; Joseph, S.; Holst, M. J.; McCammon, J. A. Electrostatics of Nanosystems: Application to Microtubules and the Ribosome. PNAS 2001, 98 (18), 10037–10041. Zhou, J.; Bronowska, A.; Le Coq, J.; Lietha, D.; Gräter, F. Allosteric Regulation of Focal Adhesion Kinase by PIP2 and ATP. Biophys. J. 2015, 108 (3), 698–705. Lomize, A. L.; Pogozheva, I. D.; Mosberg, H. I. Anisotropic Solvent Model of the Lipid Bilayer. 2. Energetics of Insertion of Small Molecules, Peptides, and Proteins in Membranes. J. Chem. Inf. Model. 2011, 51 (4), 930–946. Hou, T.; Wang, J.; Li, Y.; Wang, W. Assessing the Performance of the MM/PBSA and MM/GBSA Methods. 1. the Accuracy of Binding Free Energy Calculations Based on Molecular Dynamics Simulations. J. Chem. Inf. Model. 2011, 51 (1), 69–82. Papusheva, E.; de Queiroz, F. M.; Dalous, J.; Han, Y.; Esposito, A.; Jares-Erijmanxa, E. A.; Jovin, T. M.; Bunt, G. Dynamic Conformational Changes in the FERM Domain of FAK Are Involved in Focal-Adhesion Behavior During Cell Spreading and Motility. J Cell Sci 2009, 122 (5), 656–666. Chen, H.-C.; Ziemba, B. P.; Landgraf, K. E.; Corbin, J. A.; Falke, J. J. Membrane Docking Geometry of GRP1 PH Domain Bound to a Target Lipid Bilayer: an EPR Site-Directed Spin-Labeling and Relaxation Study. PLOS ONE 2012, 7 (3), e33640. Parton, D. L.; Klingelhoefer, J. W.; Sansom, M. S. P. Aggregation of Model Membrane Proteins, Modulated by Hydrophobic Mismatch, 25 ACS Paragon Plus Environment


1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

(58)

Page 26 of 34

Membrane Curvature, and Protein Class. Biophys. J. 2011, 101 (3), 691–699. Sener, M. K.; Olsen, J. D.; Hunter, C. N.; Schulten, K. Atomic-Level Structural and Functional Model of a Bacterial Photosynthetic Membrane Vesicle. Proc. Natl. Acad. Sci. U.S.A. 2007, 104 (40), 15723–15728.


Page 27 of 34

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


TOC



1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

The structure of the Focal Adhesion Kinase (FAK). The Kinase domain (yellow) is flanked by an N-terminal FERM domain (blue) and a C-terminal FAT domain (green). The ATP-binding pocket and important tyrosine residues that are phosphorylated during the activation process are highlighted. Fig. 1 82x77mm (300 x 300 DPI)

ACS Paragon Plus Environment

Page 28 of 34

Page 29 of 34

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


Binding dynamics and orientations of the autoinhibited FERM-Kinase complex in contact with the inner leaflet of a fluid membrane containing only PC lipids and 6% PIP2 in the inner leaflet. A) Time evolution of the number of simulations in which the FERM-Kinase complex made contact with the membrane (See also Fig.S1). The data was fitted to an exponential model f(t) = a-b*exp(-c*t) that yielded a = 31.5, b = 16.5, c = 1.6 µs-1. The total number of 5 µs-simulations was 32. B) Three main peaks were observed in the orientational map derived from all 32 5-µs long simulations showing three distinct orientations of the FERMKinase complex at the membrane. C) Number of simulations in State I, II and III over time. The solid lines are moving averages of the actual data (transparent) with a window size of 0.5 µs. D) Number of transitions between States I, II and III. E) The orientation of the FERM-Kinase complex facing the membrane (gray spheres) is shown for States I, II and III (same color code as in Fig. 1). F) Views of the domains that face the membranes for the three different States. Color-coded is the fraction of time that each amino acid interacted with PIP2. Note that the residues which showed the highest PIP2 interaction probabilities for the three different States are distinctly different, and that many of them are not part of the basic patch. Fig. 2 159x133mm (300 x 300 DPI)



1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Binding of the FERM-Kinase complex to the cell mimetic membrane (see Table 1) without and with 10% PIP2 in the inner leaflet. A) In the absence of PIP2, the binding kinetics of the complex to the inner membrane leaflet were slower in the absence versus presence of 10% PIP2 (B). The fitted model parameters are a = 27.7, b = 18.7, c = 0.4 µs-1 without PIP2, and a = 31.8, b = 17.8 c = 2.9 µs-1 with PIP2. C) A multitude of orientations were sampled when compiling 32 5 µs-long simulations indicating a slight preference towards State Ia and III as shown in panel E. D) In the presence of 10% PIP2, some peaks were shifted (State I), while others disappeared (State IV). F) Mainly two orientations were sampled (States I and III). Evolution of States I and III in the absence (G) and presence (H) of PIP2. The orientation of the protein in these states is visualized without (I) and with PIP2 (J) of PIP2 (same color code as in Fig. 1). K) The fraction of time each amino acid interacted with PIP2 is color-coded for States I and III. Notice the many residues that are firmly anchored in the membrane and that only K218 is part of the so-called basic patch. Fig. 3 171x212mm (300 x 300 DPI)


Page 30 of 34

Page 31 of 34


1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 ACS Paragon Plus Environment


1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Interactions of the FERM-Kinase complex with the PIP2 cell mimetic membrane in State I revealed by reverse coarse graining and atomistic MD simulations. A) A reverse-coarse grained representation of a MARTINI snapshot from State I (color coded as in Fig. 1). As a comparison, the crystal structure aligned by the kinase domain (yellow) is shown in gray. B) Final frame of a 100 ns atomistic simulation (color coded) based on the MARTINI snapshot (black, same as in A). C) Clustering of PIP2 (purple) and ions (yellow) underneath the FERM-kinase complex. D) Interactions of the flexible loop (licorice) with the membrane. Fig. 4 160x166mm (300 x 300 DPI)


Page 32 of 34

Page 33 of 34

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


Membrane binding dynamics of the individual FERM domain to the inner leaflet of a cell mimetic membrane (see Table 1). Binding kinetics to the membrane A) without PIP2 (fitted model parameters a=32, b=22.8 and c=0.7 µs-1) and B) with 10% PIP2 (parameters a=30.9, b=9.4, c=6.8 µs-1). C,E,G) In the absence of PIP2, four main orientations (States V, VII, VIII, IX) were equally occupied. D,F,H) In the presence of 10% PIP2, similar states were sampled, but State IX was occupied much more frequently than the other states. I) In State IX, residues of F1 and F2 modules interacted with PIP2 molecules in the membrane J) Percentage of time each residue was in contact with PIP2 in state IX. K) The FERM residues involved in PIP2 binding (solid surface, color scale) correspond to the autoinhibitory interface, as shown by the overlay of the crystal structure (pdb 2j0j, gray ribbon structure). L) Residues with frequent contact with PIP2 are highlighted and regions of high PIP2 density are shown in red as iso-surfaces. Fig. 5 147x203mm (300 x 300 DPI)




Page 34 of 34

Page 35 of 34

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


Binding dynamics of the isolated Kinase domain to the inner leaflet of a cell mimetic membrane containing 10% PIP2. Binding kinetics of A) the inactive (fitted model parameters a=30.4, b=16.4, c=2.3 µs-1) and B) the active (parameters a = 29.9, b = 18, c = 7.3 µs-1) C) The inactive kinase with the unstructured activation loop mainly sampled four states, with a strong preference towards State I and III (E, G). D) The active conformation with the structured activation loop different orientations. State Ib is shifted and preferred over other states (F, H). I) In the inactive FAK, the ATP-binding pocket is facing the membrane in State I, while it is solvent accessible in State III. J) In the active conformation, the ATP binding site faces the membrane in State Ib, but in a more tilted orientation than in State I of the inactive Kinase. K) Interaction of the inactive Kinase with PIP2 is shown for each amino acid in State I and III. L) PIP2 interaction of the active Kinase in State Ib. Fig. 6 150x198mm (300 x 300 DPI)




Page 36 of 34

Page 37 of 34

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


Schematic model of the first activation steps of FAK. 1) PIP2 favors the membrane bound state of the autoinhibited FERM-Kinase complex. 2) Dissociation of the FERM-Kinase domain is followed by 3) a reorientation of the FERM domain and Y397 unbinding. 4) Subsequent autophosphorylation of Y397 (marked by an asterisk) blocks the rebinding of pY397 to the FERM domain. 5) Further phosphorylation of Y576 and Y577 by Src.. Fig. 7 196x142mm (300 x 300 DPI)


Structural Insights How PIP2 Imposes Preferred Binding Orientations

Recommend Documents