Claws, Disorder, and Conformational Dynamics of the C-terminal Region of Human Desmoplakin
CClaws, Disorder, and Conformational Dynamicsof the C-terminal Region of Human Desmoplakin
Charles E. McAnany and Cameron Mura ∗ Department of Chemistry, University of Virginia, Charlottesville, VA 22904, USA
E-mail: [email protected]
Phone: 1.434.924.7824. Fax: 1.434.924.3710 ∗ To whom correspondence should be addressed In press (2016) a r X i v : . [ q - b i o . B M ] J un bstract Multicellular organisms consist of cells that interact via elaborate adhesion com-plexes. Desmosomes are membrane-associated adhesion complexes that mechanicallytether the cytoskeletal intermediate filaments (IFs) between two adjacent cells, creat-ing a network of tough connections in tissues such as skin and heart. Desmoplakin(DP) is the key desmosomal protein that binds IFs, and the DP • IF association posesa quandary: desmoplakin must stably and tightly bind IFs to maintain the struc-tural integrity of the desmosome. Yet, newly synthesized DP must traffick along thecytoskeleton to the site of nascent desmosome assembly without ‘sticking’ to the IFnetwork, implying weak or transient DP ···
IF contacts. Recent work reveals that thesecontacts are modulated by post-translational modifications (PTMs) in DP’s C-terminaltail. Using molecular dynamics simulations, we have elucidated the structural basis ofthese PTM-induced effects. Our simulations, nearing 2 µs in aggregate, indicate thatphosphorylation of S2849 induces an ‘arginine claw’ in desmoplakin’s C-terminal tail(DP
CTT ). If a key arginine, R2834, is methylated, the DP
CTT preferentially samplesconformations that are geometrically well-suited as substrates for processive phospho-rylation by the cognate kinase GSK3. We suggest that DP
CTT is a molecular switchthat modulates, via its conformational dynamics, DP’s efficacy as a substrate for GSK3.Finally, we show that the fluctuating DP
CTT can contact other parts of DP, suggestinga competitive binding mechanism for the modulation of DP ···
IF interactions. ntroduction Desmosomes mediate cellular adhesion — Desmosomes are inter-cellular junctions foundin epithelial and cardiac tissue.
By connecting the intermediate filaments (IFs) of neigh-boring cells, desmosomes create a network of adhesive structural interactions that imparttensile strength and durability to these tissues. The general architecture of the desmosomeis shown in Figure 1. Desmosomes expose the extracellular regions of two transmembranecadherins, desmocollin and desmoglein, on the cell surface; these proteins bind the cadherinsof neighboring cells via Ca -dependent homo- or heterophilic interactions. The desmoso-mal cadherins traverse the plasma membrane and bind two other key proteins, plakoglobinand plakophilin, which in turn bind a large, essential protein known as desmoplakin (DP;Figure 1). DP binds to the cytoskeletal IFs and, because the cytoskeleton spans the cytosolof one cell and binds to other desmosomes (which in turn bind to other, neighboring cells),this extended network of adhesive molecular contacts links together cells into tissues.The IFs bind to DP’s three plakin repeat domains (PRDs), which correspond to residues1960-2208 and are denoted PRD A, PRD B, and PRD C (Figure 1). The C-terminal PRDsare connected to the plakin domain, in DP’s N-terminal region, via a fibrous rod (residues1057-1945, central coiled-coil in Figure 1). The coiled-coil region is responsible for DP dimer-ization and, ultimately, links an electron-dense region known as the outer dense plaque (nearthe cell membrane) to the inner dense plaque (proximal to the IF network), across a span of ≈ The plakin domain of spectrin repeats (residues 178-883, leftmoststructure in DP in Figure 1) provides a relatively rigid N -terminal connection that bindsto the plakophilin (PKP) and plakoglobin (PG) proteins, thereby helping target DP to thedesmosome. Plakoglobin binds to the intracellular regions of desmocollin and desmoglein,denoted as the cadherin cytoplasmic regions (CCR) in Figure 1. Crystallographic structuresof PRDs have revealed a basic groove that can sterically accommodate IFs, suggesting thatas a potential mode of DP ···
IF interactions.
Because desmosomes impart structural integrity and mechanical strength to cell ··· cell3unctions, aberrant desmosome function underlies several diseases of the skin and heart. Forexample, pemphigus is an autoimmune disease caused by antibodies to desmoglein, the DPmutation S2594P is linked to Carvajal syndrome, and several DP mutations are associatedwith the lethal heart disease arrhythmogenic right ventricular cardiomyopathy. Down-regulation of DP has been linked to metastasis of tumor cells, and desmosome functionin cancer remains an active area of research. Several point mutations in the desmoplakinC-terminal tail (DP
CTT ) have been examined previously, providing evidence that theDP
CTT region regulates DP • IF adhesion.
Include Figure 1 here.
Cellular adhesion by desmosomes is regulated by two principal mechanisms: (i) Ca -dependent adhesion of extracellular cadherin domains and (ii) phosphorylation-dependentadhesion of DP to IFs. During desmosome formation, DP must be translocated to thedesmosome along the cytoskeletal network, and therefore must bind only loosely to IFs.Once DP reaches the desmosome and is properly localized, it binds more tightly to the IFsin order to create stable and persistent intercellular connections in epithelial tissues (e.g.,skin) and cardiac muscle. The IF-binding site of DP is required for normal desmosomeassembly in vivo , suggesting that DP transport occurs along IFs; however, the S2849Gmutation, which is in the DP CTT , causes DP to associate abnormally strongly with IFs,thereby retarding desmosome assembly. Post-translational modifications alter the behavior of desmoplakin — Cell biologi-cal and proteomic work suggest that assembly of the DP • IF adhesion complex is regulated byspecific post-translational modifications (PTMs) in DP, including phoshporylation at S2849in the DP
CTT . At least two kinases are suspected to phosphorylate DP: protein kinaseC- α (PKC α ) and glycogen synthase kinase 3 (GSK3). PKC α binds to DP in the cytoplasmand phosphorylates DP CTT to initiate desmosome assembly. Once S2849 is phosphorylated,a second kinase, GSK3, further phosphorylates DP
CTT in a processive manner. GSK3 isa processive kinase that recognizes peptides with the sequence SXXXS PO and phosphory-4ates the serine. Suitable substrates for GSK3 are generally those peptides that have beenalready phosphorylated, and the processive phosphorylation cascade proceeds in a C → N direction. While its cellular activity is regulated by various factors, GSK3 does not displaystrong substrate specificity on its own.
Recent in vivo studies have shown that theS2849G mutation has the same effect on DP as does inhibiting PKC α or GSK3—namely,DP binds IFs tightly as soon as DP is synthesized, slowing its recruitment to the assem-bling desmosome. The site of these phosphorylation events (i.e., the DP CTT ) features aglycine/serine/arginine-rich region (GSRR) containing the sequence (GSRS) GSRRGS.In addition to the S2849G mutation, which exhibits deleteriously enhanced IF binding, anR2834H mutation causes cardiac dysfunction in mice, and various other mutations in DP
CTT are linked to various disease states. The residue R2834 in the DP
CTT is important becauseits dimethylation (giving R Me CTT is phosphorylated at multiple sites, and thisphosphorylation cascade is contingent on two PTMs: a phosphorylation (S PO Me Me CTT is phosphorylated only at S PO As mentioned above, GSK3 generally exhibits low substrate specificity. The R2834Hmutation provides an interesting counterpoint to this trend. Since GSK3 binds DP
CTT atS PO Therefore, the DP
CTT also provides a useful system to explore thestructural and dynamical basis of GSK3 substrate recognition.In addition to phosphorylation, mass spectrometry (MS) studies of DP
CTT have revealedmultiple methylated arginine residues, with up to six methyls concurrently in the DP
CTT . (In that case, three of the seven arginine residues in the DP CTT were dimethylated.) In caseswhere a single arginine is dimethylated, some evidence indicates that both methyls are on the5ame nitrogen, yielding an asymmetric dimethylarginine residue. In DP
CTT , methylationappears to be necessary before GSK3 can initiate processive phosphorylation. Since theirinitial discovery in the 1960s, methylated protein residues often have been found to occurin serine-rich region (SRR) regions; indeed, such sequences serve as a common substrate formethyltransferases. However, unlike phosphorylation, methylation is not known to bemetabolically reversible, at least not outside the context of histones. Apart from regulatingthe phosphorylation cascade of DP
CTT , any functional roles of these methylations remainunexplored.
Arginine claws can structurally rigidify disordered regions — The DP
CTT contains anSRR which is multiply phosphorylated, but any structural and dynamical effects of PTMsin the DP CTT remain unknown. The structural dynamics of heavily-phosphorylated SRRshave been studied in other systems, and phosphorylation of SRRs is a common regulatorymechanism in the Eukarya. A three-dimensional (3D) structure known as the arginine claw provides a rationale for some of these interactions and effects.The arginine claw (RC), a relatively recently-identified structural element of SRRs, wasfirst characterized in the C-terminal region of ASF/SF2, a protein involved in mRNAsplicing, spliceosome assembly, and mRNA nuclear trafficking. This protein is phosphory-lated in an SRR, and this modification serves as a nuclear import signal. Fundamentally,the compaction of a peptide region into an RC sequesters charged side-chains away fromthe protein surface (Figure 2). Implicit-solvent molecular dynamics (MD) simulations ofa fully-phosphorylated (RS PO ) peptide initially revealed a compact structure, with onephosphate group coordinated by the guanidinium moieties of several arginine residues. AnRC such as we find in the DP CTT (see below) is shown in Figure 2c, alongside an illustrationof the RC originally characterized by Hamelberg et al. in Figure 2d. Such structures asshown in Figure 2d were found to stably persist over the 200-ns, fully-atomistic, explicit-solvent MD simulations of the (RS) system. In multiply-phosphorylated SRRs, thosephosphate groups not involved in the RC are solvent-exposed, and this dynamically-varying6urface exposure has been proposed as the recognition mechanism for nuclear import of aserine / arginine–rich ASF/SF2 (this particular ‘SR protein’ is also known as SRSF1). NMRstudies of the ASF/SF2 system, as well as hPrp28 (another RNA-splicing–related system),have complemented the results of MD simulations, demonstrating that the phosphorylationof SRRs rigidifies the region. Further simulation-based studies of RCs showed that clawformation allows the SRR of the lamin B receptor to bind to histones, despite the largepositive charges of both interacting proteins. Crystallographic studies of the RNA splic-ing factor SF1 have also revealed a partial RC. As a final recent example, simulationshave detected a claw-like structure in the long-time dynamics of a small, apoptosis-relatedintrinsically disordered protein (IDP) known as Noxa. Include Figure 2 here.
Simulations of disordered structural ensembles are not straightforward — MDsimulations have been used to examine SRRs, IDPs, PTMs and, to a lesser extent, theinterplay between these.
The long timescales of conformational transitions and struc-ture formation in SRRs has often prompted the use of relatively inexpensive implicit-solventmodels. However, continuum solvent models likely overestimate the electrostatic effects ofsalt bridges in determining three-dimensional structure, and RC simulations performedwith implicit solvent models predict more compact structures than do analogous explicitsolvent simulations. Another important consideration is the force-field (FF) used to de-scribe the potential energy landscape of a system. Modern FFs have been used to predictprotein structures, albeit with limited success; any FF shortcomings are exacerbated insimulations of IDPs due to the small energy differences between conformations. Recentwork has shown that CHARMM36 and ff03* predict substantially different secondary struc-tures in glycosylated IDPs. Simulations of highly-charged systems are also affected by theinadequate representation of electronic polarizability in current FFs. The classical Coulombmodel of electrostatic interactions has been extended to include polarizability, though polar-izable FF parameters are not yet available for PTMs such as in the systems studied here. Disordered peptides are often underrepresented in these parameterizationprocesses, as validating a structural ensemble generated by simulations of an IDP may beexperimentally challenging (versus non-IDP systems). Not only are structural parametersdifficult to determine experimentally, trajectory analysis is seldom straightforward andmany complex techniques have been employed in analyzing IDP simulation results. Fi-nally, note that the RC is a somewhat unusual system insofar as it has a highly-chargedcore, while FFs are parameterized against the more common cases wherein charged residuesare solvent-exposed. For these reasons, we note that simulations of systems of this typeshould be considered more suggestive and predictive rather than conclusive.For slow processes and rare events, the computational cost of simulating a system suchas an IDP for a sufficient length of time may be untenable. Several enhanced samplingmethods have been developed.
However, the size of the DP
CTT , with its extended startingconformation (and requisite number of solvent molecules; Figure 2a), necessitates a largenumber of replicas for replica-exchange simulations, and correspondingly long trajectoriesare required for adequate mixing of the replicas (McAnany & Mura, data not shown). Our MD simulations of DP — We used classical, all-atom MD simulations to examinethe structural effects of PTMs in the DP
CTT , with a specific aim of elucidating the confor-mational dynamics of this 70-residue region (Figure 1) and the riddle of strong / weak DP ··· IFinteractions (might DP
CTT be a PTM-modulated molecular switch?). To mitigate the effectsof FF inaccuracies and limited sampling, each system was simulated under two independentFFs (from the Amber and CHARMM families), and each production trajectory is at least100-ns long. Simulations were extended to 200 ns for all phosphorylated systems; for consis-tency in scaling the figures, the 200-ns simulations were split into 100-ns chunks. When werefer to a simulation without explicitly mentioning a time, we refer exclusively to the first100 ns; when referring to the second 100 ns, we call this ‘cycle2’.8e begin by proposing a quantitative definition of an RC, and we show that simultaneousmethylation and phosphorylation cause DP
CTT to assume conformations that are compatiblewith GSK3–binding. We propose that DP
CTT , in its various claw and non-claw states, com-petes with IF molecules for binding sites on the neighboring PRD elements (Figure 1), thussuggesting a straightforward dynamical mechanism for the regulation of DP ···
IF interactions.
Methods of Procedure
Molecular dynamics simulations
Classical, all-atom MD simulations were performed using NAMD 2.9, with eitherthe Amber PARM99SB or CHARMM36 force-field. Parameters for modifiedresidues, such as diprotic phosphoserine (S PO ), were drawn from and as avail-able (see below). No crystallographic or NMR structure of the DP CTT is avail-able, so the peptide
LLEAASVSS KGLPSPYNMS SAPGSRSGSR SGSRSGSRSG SRSGSRRGSFDATGNSSYSY SYSFSSSSIG H was constructed using VMD’s
Molefacture plugin in pro-tein builder mode (VMD v1.9.1); note that the above sequence numbering matches humanDP (UniProt ID P15924), and the simulated DP CTT peptide ends at the very C -terminusof DP. The peptide was constructed in an extended conformation ( φ = 180 ◦ , ψ = 180 ◦ ),as shown in Figure 2a. PTMs were applied to specific residues (Figure 1) by using eitherleap (for PARM99SB, LEaP from AmberTools13 ) or patches in VMD’s psfgen tool (forCHARMM36). Each initially-extended peptide system was subjected to a brief conforma-tional relaxation simulation in implicit solvent. These relaxation simulations were performedwith rigid hydrogen atoms, a nonbonded cutoff distance of at least 11.0 ˚A, and a Langevinthermostat set to human physiological temperature (310 K). NAMD’s generalized Born im-plicit solvent model was used with an ion concentration of 0.15 M. A 2-fs integrationtimestep was used in all simulations. The relaxation simulation consisted of 10,000 steps of9onjugate gradient potential energy minimization, followed by 10 ns of unrestrained MD. Arepresentative relaxed structure is shown in Figure 2b.Periodic boundary conditions were set-up by solvating the final structures from the relax-ation simulations in a truncated octahedral cell of water molecules, of sufficient dimensionssuch that there would be at least 15 ˚A of water between the peptide and the envelope ofthe cell (this worst case scenario being reached if the peptide were to adopt the most ex-tended state found in the last 5 ns of the relaxation simulation). This heuristic was adoptedbecause of the periodic boundary conditions used in the explicit-solvent simulations: thepeptide will be flexible during the production runs, and any prolonged violation of a 30 ˚Adistance between periodic images of the DP CTT solute could introduce artifacts. To miti-gate computational costs, a “worst case” expanded size for the peptide was estimated basedon the last half of the relaxation run; the first half of the relaxation run was not used inour geometry calculations, as the peptide is still collapsing during that time from its initial(extended) state. Even with the relaxation simulation, most of our simulated systems stillcontained over 200,000 particles (mostly H Os). Waters were placed about the compactifiedpeptide using the SOLVATE program, with custom modifications introduced in-house toenhance its performance. Ions were placed by VMD’s Autoionize plugin (for CHARMM36)or LEaP (for PARM99SB) to reach 0.15 M NaCl. Because LEaP’s ion placement was ob-served to be non-random, a 10-ns water equilibration run was performed on those systemssimulated using PARM99SB; this run comprised 100 steps of energy minimization, followedby 10 ns of dynamics with the protein atoms harmonically restrained by a force constant of1 kcal / mol / ˚A . All other parameters were the same as in the equilibration runs.For consistency, all PARM99SB and CHARMM36 systems were equilibrated in the sameway, using the general approach of Mura & McCammon. Again, a 2-fs timestep, with atleast an 11.0 ˚A nonbonded cutoff and a 310 K Langevin thermostat, were used. Periodicboundary conditions were employed with particle mesh Ewald (PME) electrostatics and agrid spacing of better than 1/˚A per direction. NAMD’s langevinPiston feature was used to10aintain pressure at 1 atm. Protein atoms were initially harmonically restrained by a 50kcal / mol / ˚A spring. The systems were minimized for 1000 steps, then gradually heated in 10K increments, with 2 ps of dynamics at each new temperature. Once the system temperaturereached 310 K, the restraints were weakened to 0.01 kcal / mol / ˚A by repeatedly halving therestraint strength and simulating for 2 ps. Finally, the restraints were completely removedand the system was equilibrated for 10 ns in the NPT ensemble.Production trajectories were computed using the same simulation parameters as theequilibration runs described above, and were extended to at least 100 ns each (Table 1).Analyses were performed using VMD and custom scripts written in the Python and D languages. All simulation and analysis scripts are available upon request, as are dehydratedtrajectories. Methylarginine parameterization
Parameters for dimethylarginine, R Me , in the CHARMM family of FFs were generouslycontributed by the Dejaegere laboratory. These parameters lacked a term for the CK1–NH1–CK2 angle, subtended by the carbons of the added methyl groups and the nitrogento which they are bonded; therefore, the value of this term was estimated using ab initio quantum mechanical calculations on a single R Me residue. Specifically, the GAMESS program was used to perform geometry optimizations at the RHF/3-21G level in implicitwater. First, the optimal equilibrium geometry was determined, then the relevant bondangle was constrained 1 ◦ higher than the equilibrium angle and the equilibrium geometryre-calculated subject to this constraint. The derivative of energy with respect to angleprovides the necessary value for this new FF parameter. The angle constraint was found tobe 95 .
467 kcal / mol / rad , with an equilibrium angle value of 115 . ◦ .11 nalysis pipeline For the sake of data-processing consistency, comparability, and automation, software toolswere developed into a pipeline to analyze each simulation trajectory in a standardized man-ner. The detailed results of these analyses are shown in Figures S1 to S19. Figures wereprepared using matplotlib and Python 3.3, with some analysis steps performed in VMD andthe D programming language. Detailed descriptions of our analysis modules follow in theremaining subsections.
Arginine clawicity, Cy R ∗ (panel a) — Plots of the arginine clawicity, arginine clawicity(Cy R ∗ ), show, at each trajectory time-step, the Cy R ∗ of the simulated system. For each residuein the sequence, the number of hydrogen bonds made to arginine are calculated, and thelargest of these numbers (the Cy R ∗ , by definition, where the ‘*’ wildcard denotes any residue)is plotted as a blue point. A green trace, representing a 1-ns running average, smoothensthe noisy behavior of Cy R ∗ . On the right of the panel, a vertically-oriented histogram showsthe distribution of Cy R ∗ values over the entire simulation; an example is given in Figure 3a.These histograms (e.g., ) are also used within the text to succinctly convey the Cy R ∗ behavior of a given simulation. Residue-specific arginine clawicity, Cy R (panel b) — Plots of residue-specific arginineclawicity (Cy R ) show which residues are contained in an RC, as exemplified in Figure 3b.For each residue, at each time-step, the number of hydrogen bonds to arginine is calculated.These data are averaged with a 1-ns window before plotting, in order to avoid aliasing.White areas indicate that no hydrogen bonds were made to arginine by a particular residueat a particular time. For clarity, the DP CTT sequence is staggered (up/down) along thehorizontal axes of these plots: Residues on the top line align with inward-facing ticks andresidues on the bottom line align with the extended outward-facing ticks. The key residuesH2834 and S2849 are marked with asterisks.
SASA of residues 2849 and 2834 (panels c and d) — Solvent-accessible surface areas12ere calculated using VMD’s SASA tool, with a solvent probe radius of 1.4 ˚A. As with Cy R ∗ ,the solvent-accessible surface area (SASA) for each frame is shown as a blue point and agreen trace shows a 1-ns running average. SASA values were calculated for the entirety ofa residue, so comparison between systems with different residue modifications or mutationsrequires caution, as the residues are of different size. The histogram adjoined to the rightaxis (200 bins) shows the distribution of SASA over the entire simulation. The S2849 ·····
S2845 distance (panel e) — For each frame in the simulation, the distancebetween the hydroxyl oxygens of S2849 and S2845 was calculated and plotted as a blue point.The green trace shows a 1-ns running average, and the histogram on the right (200 bins)shows the distribution of distances for the entire simulation.
GSK3 clash scores (panel f) — The GSK3 steric clash scores were evaluated via whateffectively became a one-dimensional docking procedure (Figure 5). We began with the 3Dstructure of GSK3, taken as chain A from PDB entry 1I09. The (side-chain) oxygen ofresidue 338 of chain B (the recognition site landmark), and the solvent-facing oxygen of thephosphate docked to chain A (the active site landmark), were used as reference points foralignment. These two reference points correspond to the recognition site and active site ofGSK3. Note that only those chain A protein atoms built into the crystal structure wereconsidered in the evaluation of clash scores. The corresponding pair of atoms from DPare the side-chain oxygens of S2849 (phosphorylated prior to GSK3 interaction) and S2845(destined for phosphorylation by GSK3). In phosphorylated systems, the oxygen attachedto the carbon was used. For each frame of each trajectory, DP and GSK3 were aligned basedon the two pairs of atoms described above. GSK3 was then rotated, in 1 ◦ increments, aboutthe axis defined from these four reference points. For each configuration, the number ofclashes was taken as the number of contacts between atoms in GSK3 and atoms in DP (sanshydrogens for computational efficiency), with a 2 ˚A sweep radius. The minimum numberof contacts, considered among all rotated positions for the trajectory frame in question, is13efined as the clash score for that frame; it is this quantity which is plotted in the panels f. Contact maps (panel g)—
The contact map shows the pairwise contacts within a protein3D structure, measured as a symmetric matrix of interatomic distances, d i,j , for all pairs ofresidues i and j . The distance is defined so as to account for side-chain interactions: for agiven residue pair, all pairs of atoms within each of the two residues ( i x , j y ) are considered,where atom x ( i x ) is from residue i and atom y ( j y ) is from residue j . The contact mapdistance for ( i, j ) is then taken as the distance between the closest pair of atoms for allof those pairs within the residue pair. In our illustrations, the lower-left triangle of thecontact map shows the average inter-residue distance for the duration of the simulation,while the upper triangle gives the minimum distance considered over the entire trajectory.The horizontal axis is identical to that used for Cy R , and the vertical axis is marked everyten residues and at the residues that were PTM sites in this study (asterisks). Ramachandran plots (panel h)—
Ramachandran plots show the distribution of peptidebackbone torsion angles, ( φ, ψ ), for each system, along the entire trajectory. Colors aregraded by the logarithm of the probability density of a given ( φ, ψ ) configuration. Regionscorresponding to canonical secondary structures are demarcated by guidelines, with theboundaries drawn from the MolProbity source code. The percent of observations in eachregion is given at the top of the panel, and these regions roughly correspond to secondarystructures: ‘L α ’ = left-handed α -helix; ‘L α +’ = generously-allowed left-handed α -helix; ‘e’= (cid:15) -turn regions, often found ahead of a helix or strand; ‘ α ’ = standard (right-handed) α -helix; ‘ β ’ = β -strand; ‘g+’ = generously-allowed helix or strand; ‘o’ = other structures.14 esults Arginine claws occur in the DP
CTT
A claw can be quantitatively defined, and occurs in the DP
CTT — Past efforts havequalitatively detected RCs based on visual analysis of trajectories, such as the one shownin Figure 2d.
These past claws (i) were characterized as multiple arginines interactingwith a phosphate, (ii) were found to be stable on the timescale of a 100-ns simulation, and(iii) had estimated free energies of formation of ≈ –5 kcal/mol. While those attributesdescribe the behavior of a claw, they are not suitable metrics for determining the claw-forming propensity across a number of trajectories, which is a goal in our current study.First, the above set of descriptors does not, in and of itself, provide an algorithmic solution tothe decision problem of whether a particular structure is or is not an RC. Second, the abovedescription involves kinetic and thermodynamic information, both of which require morecomputationally expensive calculations than would a straightforward geometric definition ofan RC. Finally, the above description of a claw does not work well for a trajectory thattransiently adopts a claw or claw-like conformation. Therefore, we propose a definition of aclaw that is akin to that of a protein secondary structural element.Our definition is purely geometric, based only on a definition of the hydrogen bond, and our parameter is easily evaluated for an arbitrary 3D structure. We define the clawicity ,Cy AB , as the maximum number of hydrogen bonds made by any residue in B to all residues inA. For example, Cy RS51 refers to the number of hydrogen bonds made by S51 to all arginine(R) residues. Cy RS refers to the number of hydrogen bonds made to an arginine by the serine( any serine) with the greatest number of hydrogen bonds to arginine. We define the arginineclawicity (Cy R ∗ ) as the number of hydrogen bonds made to arginine residues by the residuewith the most hydrogen bonds to arginine (here, the ‘*’ wildcard means any residue ). The residue-specific arginine clawicity (Cy R i ) is defined as the number of hydrogen bonds made toarginine by each residue, i . Thus, for a peptide containing n residues, Cy R would contain n R1 , Cy R2 , Cy R3 , ..., Cy Rn − , Cy Rn . In the current work, we consider a hydrogen bondto have a donor ····· acceptor distance below 3 ˚A and a donor–hydrogen–acceptor angle lessthan 20 ◦ . This definition is trivially extended to other residues and may be made smootherby incorporating a definition of hydrogen bonds with non-integer order. For example, theorder of a hydrogen bond might smoothly decrease from 1 to 0 as the donor ····· acceptordistance varies from 3 to 4 ˚A.As an initial observation, note that representative plots of the arginine clawicity Cy R ∗ (Figure 3a) and the site-specific Cy R (Figure 3b) reveal a rather strong RC when the R2834HS PO Include Figure 3 here.
To facilitate communication in this text, we represent Cy R ∗ values using histograms as in-linestrip charts, e.g. . Each bar denotes the frequency of a particular Cy R ∗ value acrossa trajectory, with the leftmost bar representing an Cy R ∗ of zero. For example, tendsto adopt structures of Cy R ∗ equal to 1, 2 or 3. Conversely, shows a system with aparticularly strong RC. Distributions of Cy R ∗ values for the last 100 ns of each simulationsystem are shown in Table 1. Table 1 NEAR HERE.
We discovered an RC in the conformational states sampled by the DP
CTT , as shown in Fig-ure 2. Several arginine residues in DP
CTT surround S PO CTT are long-lived structures,such as were those identified by Hamelberg et al. To our knowledge, DP
CTT is the largestunstructured peptide wherein an RC has been found.
Non-phosphorylated DP
CTT systems do not form strong claws — The unmodified(non-phosphorylated) wild-type DP
CTT peptide does not adopt a strong RC, as shown inFigures S1a and S2a for the Amber and CHARMM force-fields, respectively. Simulationsunder PARM99SB show little RC formation ( ), and Figure S1b shows that noresidue consistently hydrogen-bonds with any arginine with an Cy R exceeding unity. The16imulations using CHARMM36 predict slightly higher average Cy R ∗ , , than do thoseusing PARM99SB. Amino acids D2851 and H2871 (the final C -terminal residue) accountfor most of the Cy R ∗ , as shown in Figure S2b.As mentioned above, a newly-discovered PTM in the DP CTT is asymmetric dimethylationof R2834, yielding R Me For this modified peptide system, we find a slight increasein the average Cy R ∗ when PARM99SB is used, . D2851 is the primary contributorto this weak RC (Figure S14b). CHARMM36 predicts a slightly higher Cy R ∗ than thatseen in the unmodified peptide: . Consistent with the PARM99SB simulation ofthis system, D2851 is the primary residue creating the RC in the CHARMM36 trajectories(Figure S15b). This particular RC structure does not appear to be dynamically stable: itbriefly dissociates 40 ns into the trajectory, and then re-forms at ≈
60 ns. This observationsuggests that, although a claw can form in this system, the DP
CTT would be unlikely toadopt a collapsed RC conformation as a stable, long-lived structure.The R2834H mutant exhibits low Cy R ∗ values under PARM99SB ( ), with Fig-ure S8b showing E2804 forming the center of a weak RC. Under CHARMM36, D2851 formsno RC and the overall Cy R ∗ is low: . For R2834H simulations under both FFs, theclawicity behavior is similar to that in the unmodified system. The behavior of DP
CTT is sensitive to force-field — The backbone dihedral angledistributions for PARM99SB and CHARMM36 are shown in Figure 4. A recent method-ological study of an arginine/serine (RS)-rich peptide (unrelated to DP), using several FFs,found that CHARMM36 tends to favor the formation of left-handed helices. We found thatDP
CTT , which also contains an SRR, does not show this trend, at least not on the timescalesof our present simulations. Instead, CHARMM36 frequently predicts more β -strand char-acter (54.5%) than does PARM99SB (40.9%), as indicated in Figure 4. The total helicalcontent (including left-handed helices) is somewhat higher under PARM99SB (23.0%) thanit is under CHARMM36 (18.4%). Include Figure 4 here. Me (Figures S16d and S17d) than that predicted by CHARMM36 (Figures S18dand S19d). Phosphorylation of S2849 leads to claw formation in the wild-type system —Trajectories computed under both CHARMM36 and PARM99SB are consistent, inasmuchas the S PO R ∗ profile, . The S PO R ∗ , , in the first 100 ns of the production run, followed byin the next 100 ns (cycle2). Figures S4a and S5a indicate that this system’s DP CTT ’s claw isless stable than that reported for the (RS) peptide, and Figure S4a also shows a dramaticre-structuring at ≈
70 ns in the production run. CHARMM36 shows a similar trend, movingfrom the unmodified Cy R ∗ ( ) to in the first 100 ns, and then inthe second 100 ns (cycle2). Figures S6a and S7a show a more stable RC, akin to that seenpreviously. For both FFs, the RC that forms is centered around position S PO HPO PO ), using the Amber FF. Thissystem, under PARM99SB, exhibited essentially no Cy R ∗ ( ). Figure S3b shows thatthe RC does not form around S HPO RD2851 values of 2, and this tendency18iminishes after ≈
40 ns.
Methylation of R2834 weakens the RC — Simulations of the phosphorylated peptidewith PARM99SB show that methylation of R2834, in conjunction with the phosphoryla-tion at S2849, significantly weakens the RC, with in the first 100 ns followed byin the second 100 ns. The second cycle even has slightly lower Cy R ∗ than thenon-phosphorylated R2834H system. Figures S16b and S17b show that the principal residueinvolved in the RC is still S PO R ∗ values remain similar to the non-methylated system in terms of their distribution, butFigures S18a and S19a show that the RC is more labile in this system. The sliding-windowaverage (green trace) shows an increased variability compared to the nearly-constant behav-ior seen in Figures S6a and S7a (compare also the panels (c) [SASA of S2849] in Figures S6,S7, S18 and S19). Again, the CHARMM36 RC is centered on S PO Mutation R2834H may disrupt the RC structure — For simulation systems containingthe R2834H point-mutant as well as phosphorylation at S2849 (i.e., S PO R ∗ , with for the first 100 ns andfor the next 100 ns; Figures S10b and S11b show that the RC stably settles at S PO Cs typically exclude solvent
Analysis of the solvent-accessible surface area (SASA) of S PO PO residue will be electrostatically attracted to arginineside-chains and, as expected, this is borne out in our observations of Cy R ∗ values. In thosesimulation systems containing S PO PO R ∗ was relativelylow in the first (Figure S10a) and second (Figure S11a) 100-ns bins, and this agrees with thehigher SASA observed for S PO HPO R ∗ and high SASA values for S HPO R ∗ and SASA values. Under PARM99SB, S PO PO R (Figure S17b). S PO Methylation and phosphorylation prime DP
CTT for GSK3 activity
The DP
CTT sequence (Figure 1) contains several potential phosphorylation sites, includingconsensus sites for the GSK3 kinase. Recent experiments have revealed that DP is phospho-rylated in its CTR by GSK3. Thus, we used two simple metrics to assess the ability (notnecessarily the propensity) of the DP
CTT to interact with GSK3 throughout the entire MDtrajectory: (i) the S PO ····· S2845 distance, and (ii) the extent of steric clash between theDP
CTT and GSK3 molecules. First, the simple geometric distance between S PO PO this distanceis ≈
12 ˚A (some variability in this value is expected, as the active site was not occupied bya substrate in this GSK3 structure). As a rudimentary gauge of DP
CTT ’s ability to bind toGSK3, we suggest that DP
CTT conformations wherein the S PO ····· S2845 distance is ≈
12 ˚A will be more favored to bind to GSK3 as a result of simple geometric matching, withoutrequiring substantial structural rearrangement of the DP
CTT .While DP
CTT systems that are not phosphorylated at S2849 would not be expected (bio-logically) to interact with GSK3, it is nevertheless informative to consider, as a backgrounddistribution, how these distances compare for the non-phosphorylated and phosphorylatedsystems. We find that the distances in the non-phosphorylated systems show a strong depen-dence on FF. PARM99SB yields distances that are substantially less than 12 ˚A for the com-pletely unmodified wild-type system (Figure S1e) and the methylated, non-phosphorylatedwild-type system (Figure S14e). The non-phosphorylated R2834H mutant system startswith GSK3-compatible distances, but collapses at ≈
70 ns to incompatible distances (Fig-ure S8e). In general, the CHARMM36 simulations predict longer distances than PARM99SB,and tend to predict distances that are more compatible with GSK3 binding (see Figures S2e,219e and S15e).Simulations of the phosphorylated DP
CTT systems exhibit good agreement between thedistance distributions for PARM99SB and CHARMM36. Our distance parameter consis-tently lies between ≈ PO PO Me CTT . Specifically, we align (i) the active site of GSK3 with S2845 of DP
CTT (as thisis where the next phoshporylation event will occur), and (ii) the recognition site of GSK3to S PO CTT that is recognized). These spatial trans-formations and geometric constraints effectively reduce the problem to a one-dimensionalprotein • protein docking exercise, the one degree-of-freedom being rotation about the linedefined by constraints (i) and (ii); this construction is schematized in Figure 5. If thereexists a rotation wherein GSK3 and DP CTT can be brought together without substantialsteric clash (literally, overlap of atomic van der Waals envelopes), then this suggests thatGSK3 can readily bind to that conformation of DP
CTT (or at least that there is no en-thalpic barrier to doing so). By this measure, we find that the only phosphorylated DP
CTT systems which exhibit steric compatibility along the trajectory frames are the methylatedsystems (Figures S17f and S19f). This accommodation is seen with both the PARM99SBand CHARMM36 FFs in the last 50 ns of the production run. Therefore, based on thesedata we suggest that methylation at R2834, yielding R Me CTT for pro-cessive phosphorylation by biasing its structural ensemble towards conformations that areamenable to GSK3 phosphorylation. 22 nclude Figure 5 here.
The serine-rich region of DP
CTT is not entirely free in solution
Potential interactions between the SRR of the DP
CTT and the rest of the large DP protein(Figure 1) were explored by analyzing pairwise inter-residue distances. As detailed in theMethods section, the full suite of contact maps, shown in panels (g) of Figures S1 to S19,give the mean inter-residue distances (lower triangle), averaged over entire trajectories, whilethe upper-right triangle gives the minimum inter-residue distance across an entire trajectory.One may be tempted to view the DP
CTT as a disordered string that thermally fluctuatesin solution, but this is not entirely accurate: the dynamical DP
CTT may in fact double backon the plakin repeat domains (as a reminder, see the PRDs in Figure 1). While simulations oflarger DP systems, including an entire PRD in addition to the DP
CTT , are beyond the scopeof this work, the first few residues of our DP
CTT simulation system are from a PRD (thethird PRD in Figure 1). Therefore, if the phosphorylated S PO CTT , then that suggests that regions within theDP
CTT may interact directly with the PRDs to regulate IF binding (and also that simulationslimited to only the SRR might not account for all the factors that govern the structure anddynamics of this region). In all of our simulations, the SRR comes into close spatial proximityto other regions of DP
CTT , including the more N -terminal residues that are part of PRD-C. A possible mechanism by which the DP CTT can attenuate the overall strength of DP • IFbinding may involve a simple binding competition between IFs and DP
CTT for the IF-bindingsite of the plakin repeat domain; in this model, the precise pattern of PTMs, and thereforethe clawicity and dynamics of the DP
CTT , would modulate the competitive binding events.When DP
CTT is fully phosphorylated, its strongly negative charge could compete with the(negatively-charged) IFs for the binding groove on the PRD, as suggested by crystallographicstudies.
Include Figure 6 here. iscussion Arginine claws can form in partially phosphorylated systems — Past work on thearginine claw considered only fully-phosphorylated (RS) n repeats. To our knowledge, ourpresent study provides the first evidence that RCs can form in other systems too. Experimen-tal and computational studies of a 36-residue peptide from myelin basic protein suggestedthat phosphorylated threonines can confer structure to disordered regions, via electrostaticinteractions with basic residues. However, unlike an RC, the interactions in that system didnot result in burial of the phosphate group within the protein. Our present simulations,focused on the DP
CTT , predict that a strong RC can form in protein segments with onlyhalf the arginine density of RS-repeat peptides, and even when only a single serine is phos-phorylated. Therefore, the RC may be a common, or at least underappreciated, structuralelement in phosphorylation-based regulation of protein function via molecular switches, evenfor protein sequences that lack canonical (RS) n repeat regions. Claws are predicted by several force-fields — The original RC was described as “verystable and, once formed, persist[ing] for the rest of the simulation”; that initial studyemployed only the Amber FF03 parameter set. A subsequent study of another RS-richpeptide found that RCs form under the Amber PARM99SB-ILDN FF. Our simulations ofDP
CTT show that RCs can form under both PARM99SB and CHARMM36. Nevertheless,the fine details of RC dynamics are sensitive to the FF; for instance, for many of our systemsCHARMM36 frequently predicts higher Cy R ∗ values than does PARM99SB.The FF-dependence of our Cy R ∗ parameter is substantial, and this may reflect the some-what unusual chemical nature of RC sequences, versus most protein sequences. In additionto charged moieties buried in a proteinaceous core, arginine ··· phosphate interactions arecharacterized by a “covalent-like” stability that may be only inadequately described aspoint-charges interacting via simple Coulombic electrostatics. An RC was not detected inrecent NMR experiments with another RNA splicing-related, serine/arginine-rich system; and the trajectories in thatwork sampled shorter ( ≈ In that system, the RCacts as a secondary structural element in an otherwise disordered region; notably, electrondensity could be detected for residues immediately upstream of the phosphoserine, but onlyin the phosphorylated, not the non-phosphorylated, system. Methylation in the DP
CTT may promote GSK3 binding — From the simulationspresented here, we suggest that the R Me PO ··· GSK3 interactions. This claim is based upon three lines of evidence. First, ourmodified DP
CTT systems were found to present the phosphate group on the surface, ratherthan buried within the protein. This surface exposure did not occur in phosphorylated sys-tems with unmodified R2834, suggesting that methylation is coupled to the dynamics ofS PO PO PO ····· S2845 distance closely matches the distance betweenthe active site and substrate recognition site of GSK3. Upon GSK3 binding to S PO Me • GSK3 complex. Inour mechanistic model for GSK3 regulation, DP
CTT essentially self-regulates its processive25hosphorylation by GSK3; DP
CTT achieves this by sampling conformational states that varyin their suitability as substrates for GSK3.
The serine-rich region of DP
CTT contacts other parts of DP — Past studies ofRCs have examined short, (RS) n –containing peptides in isolation. The serine-rich regionof DP CTT is not well-described by these past models, as we have shown that the SRR caninteract with other regions of DP. In particular, the SRR can contact residues that havebeen resolved in a crystal structure of a plakin repeat domain. The charge-complementaritybetween a fully-phosphorylated SRR in DP
CTT and the positively-charged IF-binding grooveon a PRD, combined with the tendency for DP CTT to explore the surface of DP, suggeststhat a simple competition for PRD binding sites may account for the cellular effects of DP
CTT phosphorylation. That the DP
CTT is covalently linked to the upstream PRDs (Figure 1)implies a high local density of negative charge, and this could compete with the negatively-charged IFs to cause DP to detach from the IF network; examination of the ionic strength-dependence of this process would be telling. Finally, note that in our mechanistic modelany structural role for arginine claw conformational dynamics (apart from its role in GSK3processive phosphorylation) would require a further series of simulations, ideally includingas many structured PRD regions as possible.
Conclusion
Recent experiments have revealed that desmoplakin’s activity is regulated by PTMs in itspresumably-disordered C-terminal tail. Using MD simulations, we have elucidated the struc-tural effects of three modifications in the 70-residue DP
CTT region: phosphorylation of S2849,methylation of R2834, and mutagenesis of R2834 to histidine. Our simulations indicate thatan RC can form in some of the phosphorylated systems, sequestering the phosphate withinthe protein. To our knowledge, DP
CTT is the largest system that has been shown to form anRC by MD simulation. Our findings build on past studies of RC formation in SR repeats,26nd are corroborated by recent crystallographic results for other SR systems. Upon methy-lation of R2834, the S PO CTT ’s SRR is not isolated from the rest of DP, suggesting thatstudies of short peptides excised from larger systems may miss some of the interactions thatdefine the conformational ensemble of such regions. This point is illustrated by the effectsof R2834 methylation: The position of the RC, and the overall conformation of the DP
CTT ,are affected by this seemingly minor chemical modification, many residues away from thesite of phosphorylation. The common self-contact in DP
CTT , seen in contact maps for allour simulated systems, suggests that a regulatory mechanism of DP ···
IF adhesion may bea simple binding competition between DP
CTT and the IFs for the positively-charged grooveon plakin repeat domains.By elucidating the roles and linkages between protein conformational dynamics, PTMs,and claw-like structural elements, our simulations of the C-terminal region of human desmo-plakin synthesize several strands of evidence and shed light on the underlying molecularmechanism of DP ···
IF interactions, including the riddle of strong/weak interactions withthe IF network. We predict that RCs can form when S2849 is phosphorylated, and thatmethylation of the disease-associated site R2834 promotes processive phosphorylation byGSK3. Our data also are consistent with DP
CTT binding to a PRD, thus providing a simple,atomically-detailed competition mechanism for the regulation of DP • IF adhesion.
Associated Content
Supporting Information — Detailed results of the analysis suite described in the Methodssection of the main text, as applied to each of our simulation systems. Each trajectory has27een analyzed in terms of (a) Cy R ∗ , (b) Cy R , (c) SASA of S2849, (d) SASA of R2834, (e) theS2849 ····· R2834 distance, (f) GSK3 steric clash scores, (g) inter-residue contact maps, and(h) the distribution of peptide backbone torsion angles ( φ, ψ ). Acknowledgments
We thank D. Hunt & L. Zhang (UVA), as well as K. Green & L. Albrecht (Northwestern), forhelpful discussions. We thank A. Dejaegere (IGBMC) for providing force-field parametersfor R Me , and D. Hamelberg (Georgia State) for providing the coordinates used to generateFigure 2d. K. Holcomb & A. Munro (UVA) are thanked for exceptional computer support;UVA’s Advanced Research Computing Services and Information Technology Services pro-vided computational resources and technical support that contributed to the results reportedherein. Portions of this work were supported by UVA, the Jeffress Memorial Trust (J-971),and NSF Career award MCB-1350957. 28 eferences (1) Stokes, D. L. Desmosomes from a Structural Perspective.
Current Opinion in CellBiology , , 565 – 571.(2) Holthofer, B.; Windoffer, R.; Troyanovsky, S.; Leube, R. E. Structure and Function ofDesmosomes. Int. Rev. Cytol. , , 65–163.(3) Al-Amoudi, A.; Frangakis, A. S. Structural Studies on Desmosomes. Biochemical Soci-ety Transactions , , 181–187.(4) Garrod, D.; Chidgey, M. Desmosome Structure, Composition and Function. Biochimicaet Biophysica Acta (BBA)–Biomembranes , , 572 – 587.(5) Delva, E.; Tucker, D. K.; Kowalczyk, A. P. The Desmosome. Cold Spring Harbor Per-spectives in Biology , .(6) Desai, B. V.; Harmon, R. M.; Green, K. J. Desmosomes at a Glance. J. Cell. Sci. , , 4401–4407.(7) Kowalczyk, A. P.; Green, K. J. Structure, Function, and Regulation of Desmosomes. Prog Mol Biol Transl Sci , , 95–118.(8) Choi, H. J.; Park-Snyder, S.; Pascoe, L. T.; Green, K. J.; Weis, W. I. Structures ofTwo Intermediate Filament-binding Fragments of Desmoplakin Reveal a Unique RepeatMotif Structure. Nat. Struct. Biol. , , 612–620.(9) Green, K. J.; Parry, D. A.; Steinert, P. M.; Virata, M. L.; Wagner, R. M.; Angst, B. D.;Nilles, L. A. Structure of the Human Desmoplakins. Implications for Function in theDesmosomal Plaque. J. Biol. Chem. , , 2603–2612.(10) Choi, H. J.; Weis, W. I. Crystal Structure of a Rigid Four-Spectrin-Repeat Fragmentof the Human Desmoplakin Plakin Domain. J. Mol. Biol. , , 800–812.2911) Smith, E. A.; Fuchs, E. Defining the Interactions Between Intermediate Filaments andDesmosomes. J. Cell Biol. , , 1229–1241.(12) Kang, H.; Weiss, T. M.; Bang, I.; Weis, W. I.; Choi, H. J. Structure of the IntermediateFilament-Binding Region of Desmoplakin. PLoS ONE , , e0147641.(13) Fogl, C.; Mohammed, F.; Al-Jassar, C.; Jeeves, M.; Knowles, T.; Rodriguez-Zamora, P.;White, S.; Odintsova, E.; Overduin, M.; Chidgey, M. Mechanism of Intermediate Fil-ament Recognition by Plakin Repeat Domains Revealed by Envoplakin Targeting ofVimentin. Nature Communications , , 10827.(14) Shimizu, A.; Ishiko, A.; Ota, T.; Tsunoda, K.; Amagai, M.; Nishikawa, T. IgG Bindsto Desmoglein 3 in Desmosomes and Causes a Desmosomal Split Without KeratinRetraction in a Pemphigus Mouse Model. Journal of Investigative Dermatology , , 1145 – 1153.(15) Rasmussen, T. B.; Hansen, J.; Nissen, P. H.; Palmfeldt, J.; Dalager, S.; Jensen, U. B.;Kim, W. Y.; Heickendorff, L.; Mølgaard, H.; Jensen, H. K. et al. Protein ExpressionStudies of Desmoplakin Mutations in Cardiomyopathy Patients Reveal Different Molec-ular Disease Mechanisms. Clin. Genet. , , 20–30.(16) Al-Jassar, C.; Bikker, H.; Overduin, M.; Chidgey, M. Mechanistic Basis of Desmosome-targeted Diseases. J. Mol. Biol. , , 4006–4022.(17) Al-Jassar, C.; Knowles, T.; Jeeves, M.; Kami, K.; Behr, E.; Bikker, H.; Overduin, M.;Chidgey, M. The Nonlinear Structure of the Desmoplakin Plakin Domain and the Ef-fects of Cardiomyopathy-linked Mutations. J. Mol. Biol. , , 1049–1061.(18) Papagerakis, S.; Shabana, A. H.; Pollock, B. H.; Papagerakis, P.; Depondt, J.;Berdal, A. Altered Desmoplakin Expression at Transcriptional and Protein Levels Pro-vides Prognostic Information in Human Oropharyngeal Cancer. Hum. Pathol. , , 1320–1329. 3019) Stappenbeck, T. S.; Lamb, J. A.; Corcoran, C. M.; Green, K. J. Phosphorylation ofthe Desmoplakin COOH Terminus Negatively Regulates its Interaction with KeratinIntermediate Filament Networks. J. Biol. Chem. , , 29351–29354.(20) Zhang, L. Identification of Post-translational Modifications of Human Desmoplakinsin Regulating Interactions with Intermediate Filaments. Ph.D. thesis, University ofVirginia, 2014.(21) Hobbs, R. P.; Green, K. J. Desmoplakin Regulates Desmosome Hyperadhesion. J InvestDermatol , , 482–485.(22) Godsel, L. M.; Hsieh, S. N.; Amargo, E. V.; Bass, A. E.; Pascoe-McGillicuddy, L. T.;Huen, A. C.; Thorne, M. E.; Gaudry, C. A.; Park, J. K.; Myung, K. et al. DesmoplakinAssembly Dynamics in Four Dimensions: Multiple Phases Differentially Regulated byIntermediate Filaments and Actin. J. Cell Biol. , , 1045–1059.(23) Bass-Zubek, A. E.; Hobbs, R. P.; Amargo, E. V.; Garcia, N. J.; Hsieh, S. N.; Chen, X.;Wahl, J. K.; Denning, M. F.; Green, K. J. Plakophilin 2: a Critical Scaffold for PKCAlpha that Regulates Intercellular Junction Assembly. J. Cell Biol. , , 605–613.(24) ter Haar, E.; Coll, J. T.; Austen, D. A.; Hsiao, H. M.; Swenson, L.; Jain, J. Structureof GSK3beta Reveals a Primed Phosphorylation Mechanism. Nat. Struct. Biol. , , 593–596.(25) Beurel, E.; Grieco, S. F.; Jope, R. S. Glycogen Synthase Kinase-3 (GSK3): Regulation,Actions, and Diseases. Pharmacol. Ther. , , 114–131.(26) Medina, M.; Wandosell, F. Deconstructing GSK-3: The Fine Regulation of Its Activity. Int J Alzheimers Dis , , 479249.(27) Albrecht, L. V.; Zhang, L.; Shabanowitz, J.; Purevjav, E.; Towbin, J. A.; Hunt, D. F.;31reen, K. J. GSK3- and PRMT-1-dependent Modifications of Desmoplakin ControlDesmoplakin-cytoskeleton Dynamics. J. Cell Biol. , , 597–612.(28) Bedford, M.; Clarke, S. Protein Arginine Methylation in Mammals: Who, What, andWhy. Mol. Cell , , 1–13.(29) Paik, W. K.; Paik, D. C.; Kim, S. Historical Review: the Field of Protein Methylation. Trends Biochem. Sci. , , 146–152.(30) Chen, C.; Nott, T. J.; Jin, J.; Pawson, T. Deciphering Arginine Methylation: TudorTells the Tale. Nat. Rev. Mol. Cell Biol. , , 629–642.(31) Bedford, M. T. Arginine Methylation at a Glance. J. Cell. Sci. , , 4243–4246.(32) Clarke, S. G. Protein Methylation at the Surface and Buried Deep: Thinking Outsidethe Histone Box. Trends Biochem. Sci. , , 243–252.(33) Chang, B.; Chen, Y.; Zhao, Y.; Bruick, R. K. JMJD6 is a Histone Arginine Demethylase. Science , , 444–447.(34) Patwardhan, P.; Miller, W. T. Processive Phosphorylation: Mechanism and BiologicalImportance. Cell. Signal. , , 2218–2226.(35) Hamelberg, D.; Shen, T.; McCammon, J. A. A Proposed Signaling Motif for NuclearImport in mRNA Processing via the Formation of Arginine Claw. Proc. Natl. Acad.Sci. U.S.A. , , 14947–14951.(36) Zhong, X.-Y.; Wang, P.; Han, J.; Rosenfeld, M. G.; Fu, X.-D. { SR } Proteins in VerticalIntegration of Gene Expression from Transcription to { RNA } Processing to Translation.
Molecular Cell , , 1 – 10.(37) Xiang, S.; Gapsys, V.; Kim, H.-Y.; Bessonov, S.; Hsiao, H.-H.; M¨ohlmann, S.;Klaukien, V.; Ficner, R.; Becker, S.; Urlaub, H. et al. Phosphorylation Drives a Dy-namic Switch in Serine/Arginine-Rich Proteins. Structure , , 2162 – 2174.3238) Sellis, D.; Drosou, V.; Vlachakis, D.; Voukkalis, N.; Giannakouros, T.; Vlassi, M. Phos-phorylation of the Arginine/serine Repeats of Lamin B Receptor by SRPK1-insightsfrom Molecular Dynamics Simulations. Biochim. Biophys. Acta , , 44–55.(39) Wang, W.; Maucuer, A.; Gupta, A.; Manceau, V.; Thickman, K. R.; Bauer, W. J.;Kennedy, S. D.; Wedekind, J. E.; Green, M. R.; Kielkopf, C. L. Structure of Phos-phorylated SF1 Bound to U2AF65 in an Essential Splicing Factor Complex. Structure , , 197–208.(40) Karim, C. B.; Espinoza-Fonseca, L. M.; James, Z. M.; Hanse, E. A.; Gaynes, J. S.;Thomas, D. D.; Kelekar, A. Structural Mechanism for Regulation of Bcl-2 proteinNoxa by Phosphorylation. Sci Rep , , 14557.(41) McCammon, J. A.; Harvey, S. C. Dynamics of Proteins and Nucleic Acids ; CambridgeUniversity Press, 1988.(42) Mura, C.; McAnany, C. E. An Introduction to Biomolecular Simulations and Docking.
Molecular Simulation , , 732–764.(43) Lu, S. Y.; Jiang, Y. J.; Zou, J. W.; Wu, T. X. Molecular Modeling and MolecularDynamics Simulation Studies of the GSK3/ATP/substrate Complex: Understandingthe Unique P+4 Primed Phosphorylation Specificity for GSK3 Substrates. J Chem InfModel , , 1025–1036.(44) Terakawa, T.; Takada, S. Multiscale Ensemble Modeling of Intrinsically DisorderedProteins: p53 N-terminal Domain. Biophys. J. , , 1450–1458.(45) Vacic, V.; Markwick, P. R.; Oldfield, C. J.; Zhao, X.; Haynes, C.; Uversky, V. N.; Iak-oucheva, L. M. Disease-associated Mutations Disrupt Functionally Important Regionsof Intrinsic Protein Disorder. PLoS Comput. Biol. , , e1002709.3346) Petrov, D.; Margreitter, C.; Grandits, M.; Oostenbrink, C.; Zagrovic, B. A SystematicFramework for Molecular Dynamics Simulations of Protein Post-Translational Modifi-cations. PLoS Comput Biol , , 1–9.(47) Vassall, K. A.; Bessonov, K.; De Avila, M.; Polverini, E.; Harauz, G. The Effects ofThreonine Phosphorylation on the Stability and Dynamics of the Central MolecularSwitch Region of 18.5-kDa Myelin Basic Protein. PLoS ONE , , e68175.(48) Stanley, N.; Esteban-Mart´ın, S.; De Fabritiis, G. Kinetic Modulation of a DisorderedProtein Domain by Phosphorylation. Nature Communications , .(49) Das, R. K.; Ruff, K. M.; Pappu, R. V. Relating Sequence Encoded Information toform and Function of Intrinsically Disordered Proteins. Current Opinion in StructuralBiology , , 102–112.(50) Zerze, G. H.; Mittal, J. Effect of O-Linked Glycosylation on the Equilibrium StructuralEnsemble of Intrinsically Disordered Polypeptides. J Phys Chem B , , 15583–15592.(51) Mittal, J.; Yoo, T. H.; Georgiou, G.; Truskett, T. M. Structural Ensemble of an Intrin-sically Disordered Polypeptide. J Phys Chem B , , 118–124.(52) Zhou, R.; Berne, B. J. Can a Continuum Solvent Model Reproduce the Free EnergyLandscape of a Beta -hairpin Folding in Water? Proc. Natl. Acad. Sci. U.S.A. , , 12777–12782.(53) Freddolino, P. L.; Harrison, C. B.; Liu, Y.; Schulten, K. Challenges in Protein FoldingSimulations: Timescale, Representation, and Analysis. Nat Phys , , 751–758.(54) Rauscher, S.; Gapsys, V.; Gajda, M. J.; Zweckstetter, M.; de Groot, B. L.; Grub-muller, H. Structural Ensembles of Intrinsically Disordered Proteins Depend Strongly34n Force Field: A Comparison to Experiment. J Chem Theory Comput , ,5513–5524.(55) Lopes, P. E.; Huang, J.; Shim, J.; Luo, Y.; Li, H.; Roux, B.; Mackerell, A. D. ForceField for Peptides and Proteins based on the Classical Drude Oscillator. J Chem TheoryComput , , 5430–5449.(56) Mackerell, A. D. Empirical Force Fields for Biological Macromolecules: Overview andIssues. J Comput Chem , , 1584–1604.(57) Vanommeslaeghe, K.; Hatcher, E.; Acharya, C.; Kundu, S.; Zhong, S.; Shim, J.; Dar-ian, E.; Guvench, O.; Lopes, P.; Vorobyov, I. et al. CHARMM General Force Field: AForce Field for Drug-like Molecules Compatible with the CHARMM All-atom AdditiveBiological Force Fields. J Comput Chem , , 671–690.(58) Mittag, T.; Forman-Kay, J. D. Atomic-level characterization of disordered protein en-sembles. Curr. Opin. Struct. Biol. , , 3–14.(59) Sgourakis, N. G.; Merced-Serrano, M.; Boutsidis, C.; Drineas, P.; Du, Z.; Wang, C.;Garcia, A. E. Atomic-level Characterization of the Ensemble of the Abeta(1-42)Monomer in Water Using Unbiased Molecular Dynamics Simulations and Spectral Al-gorithms. J. Mol. Biol. , , 570–583.(60) Sugita, Y.; Okamoto, Y. Replica-exchange Molecular Dynamics Method for ProteinFolding. Chemical Physics Letters , , 141 – 151.(61) Wang, L.; Friesner, R. A.; Berne, B. J. Replica Exchange with Solute Scaling: a MoreEfficient Version of Replica Exchange with Solute Tempering (REST2). J Phys ChemB , , 9431–9438.(62) Ostermeir, K.; Zacharias, M. Advanced Replica-exchange Sampling to Study the Flex-35bility and Plasticity of Peptides and Proteins. Biochim. Biophys. Acta , ,847–853.(63) Abraham, M. J.; Gready, J. E. Ensuring Mixing Efficiency of Replica-Exchange Molec-ular Dynamics Simulations. Journal of Chemical Theory and Computation , ,1119–1128.(64) Phillips, J. C.; Braun, R.; Wei, W.; Gumbart, J.; Tajkhorshid, E.; Villa, E.; Chipot, C.;Skeel, R. D.; Kale, L.; Schulten, K. Scalable Molecular Dynamics with NAMD. Journalof Computational Chemistry , , 1781 – 1802.(65) Case, D. A.; Darden, T. A.; Cheatham, T. E.; Simmerling, C. L.; Wang, J.; Duke, R. E.;Luo, R.; Walker, R. C.; Zhang, W.; Merz, K. M. et al. AMBER 12. 2012; http://ambermd.org/ , Accessed 2016-05-10.(66) Best, R. B.; Zhu, X.; Shim, J.; Lopes, P. E.; Mittal, J.; Feig, M.; Mackerell, A. D.Optimization of the Additive CHARMM All-atom Protein Force Field Targeting Im-proved Sampling of the Backbone φ , ψ and Side-chain χ (1) and χ (2) Dihedral Angles. J Chem Theory Comput , , 3257–3273.(67) MacKerell, A. D.; Bashford, D.; Bellott, M.; Dunbrack, R. L.; Evanseck, J. D.;Field, M. J.; Fischer, S.; Gao, J.; Guo, H.; Ha, S. et al. All-atom Empirical Poten-tial for Molecular Modeling and Dynamics Studies of Proteins. J Phys Chem B , , 3586–3616.(68) Homeyer, N.; Horn, A. H.; Lanig, H.; Sticht, H. AMBER Force-field Parameters forPhosphorylated Amino Acids in Different Protonation States: Phosphoserine, Phospho-threonine, Phosphotyrosine, and Phosphohistidine. J Mol Model , , 281–289.(69) Papamokos, G. V.; Tziatzos, G.; Papageorgiou, D. G.; Georgatos, S. D.; Politou, A. S.;Kaxiras, E. Structural Role of RKS Motifs in Chromatin Interactions: a Molecular36ynamics Study of HP1 Bound to a Variably Modified Histone Tail. Biophys. J. , , 1926–1933.(70) Humphrey, W.; Dalke, A.; Schulten, K. VMD: Visual Molecular Dynamics. J Mol Graph , , 33–38.(71) Tanner, D. E.; Chan, K. Y.; Phillips, J. C.; Schulten, K. Parallel Generalized BornImplicit Solvent Calculations with NAMD. J Chem Theory Comput , , 3635–3642.(72) Grubmuller, H.; Heymann, B.; Tavan, P. Ligand Binding: Molecular Mechanics Calcu-lation of the Streptavidin-biotin Rupture Force. Science , , 997–999.(73) Mura, C.; McCammon, J. A. Molecular Dynamics of a κ B DNA Element: Base Flip-ping via Cross-strand Intercalative Stacking in a Microsecond-scale Simulation.
NucleicAcids Res. , , 4941–4955.(74) van Rossum, G. The Python Language Reference. 2014; http://docs.python.org/py3k/reference/index.html , Accessed 2016-05-10.(75) Alexandrescu, A. The D Programming Language ; Pearson Education, 2010.(76) Grauffel, C.; Stote, R. H.; Dejaegere, A. Force Field Parameters for the Simulation ofModified Histone Tails.
J Comput Chem , , 2434–2451.(77) Schmidt, M. W.; Baldridge, K. K.; Boatz, J. A.; Elbert, S. T.; Gordon, M. S.;Jensen, J. H.; Koseki, S.; Matsunaga, N.; Nguyen, K. A.; Su, S. et al. General Atomicand Molecular Electronic Structure System. Journal of Computational Chemistry , , 1347–1363.(78) Gordon, M. S.; Schmidt, M. W. In Theory and Applications of Computational Chem-istry: the first forty years ; Dykstra, C. E., Frenking, G., Kim, K. S., Scuseria, G. E.,Eds.; Elsevier: Amsterdam, 2005; pp 1167–1189.3779) Tomasi, J.; Mennucci, B.; Cammi, R. Quantum Mechanical Continuum Solvation Mod-els.
Chem. Rev. , , 2999–3093.(80) Chen, V. B.; Arendall, W. B.; Headd, J. J.; Keedy, D. A.; Immormino, R. M.;Kapral, G. J.; Murray, L. W.; Richardson, J. S.; Richardson, D. C. MolProbity: All-atom Structure Validation for Macromolecular Crystallography. Acta Crystallogr. DBiol. Crystallogr. , , 12–21.(81) Latimer, W. M.; Rodebush, W. H. Polarity and Ionization from the Standpoint of theLewis Theory of Valence. Journal of the American Chemical Society , , 1419–1433.(82) Arunan, E.; Desiraju, G. R.; Klein, R. A.; Sadlej, J.; Scheiner, S.; Alkorta, I.;Clary, D. C.; Crabtree, R. H.; Dannenberg, J. J.; Hobza, P. et al. Definition of theHydrogen Bond (IUPAC Recommendations 2011). Pure and Applied Chemistry , , 1637–1641.(83) Woods, A. S.; Ferre, S. Amazing Stability of the Arginine-phosphate Electrostatic In-teraction. J. Proteome Res. , , 1397–1402.(84) Choi, H. J.; Weis, W. I. Structure of the Armadillo Repeat Domain of Plakophilin 1. J. Mol. Biol. , , 367–376.(85) Choi, H. J.; Gross, J. C.; Pokutta, S.; Weis, W. I. Interactions of Plakoglobin andBeta-catenin with Desmosomal Cadherins: Basis of Selective Exclusion of Alpha- andBeta-catenin from Desmosomes. J. Biol. Chem. , , 31776–31788.38 igure Captions , McAnany & Mura, 2016 Figure 1: Desmoplakin in context: Architecture of the desmosome.
Key components of thedesmosome are diagrammed here, with the length of each rectangular protein schematic corresponding tothe number of amino acids (scale bar, lower-right). Dsg and Dsc are the transmembrane cadherin proteinsdesmoglein-1 and desmocollin, respectively; plakoglobin (PG) and plakophilin (PKP) are adapter proteinsthat bind the N -terminal region of desmoplakin to the cadherin cytoplasmic regions (CCR) of Dsg andDsc, as indicated. The PG crystal structure is inset, as are the structures of PKP and two PRDs.DP is shown in the middle, and regions of known structure are inset. Crystal structures are drawnas ribbon diagrams, with the color graded from the N (blue) to C (red) terminus. The locations of theR2834 and S2849 modifications in the DP CTT sequence are marked. Our various DP simulation systemsincluded: (i) the unmodified, wild-type sequence of DP
CTT , (ii) the wild-type sequence phosphorylated atS2849, (iii) the wild-type sequence methylated at R2834, (iv) the wild-type sequence phosphorylated atS2849 and methylated at R2834, (v) the R2834H mutant, and (vi) the R2834H mutant phosphorylated atS2849. All phosphate PTMs were diprotic, with the exception of one test system containing a monoproticphosphoserine. (See also Table 1.)
Figure 2: A recognizable RC in the DP
CTT . All simulation systems started in an extended backboneconformation (with bends at each proline residue), as exemplified in (a). After 10 ns of implicit-solventsimulation under CHARMM36, the R2834H S2849S PO system can be seen to have collapsed and formedan RC (b). A sample RC, at 95 ns for the R2834H S2849S PO system under CHARMM36, is shown in(c). The gray surface surrounds all residues other than arginines (shown as bonds), and the S PO is shown in (d), rendered similarly asin (c); the coordination geometry of arginines and a phosphoserine is similar to that seen for the DP CTT in (c). The structural stability of an RC is demonstrated in (e) by overlaying multiple frames from thesimulation trajectory of (c). The regions near the RC remain stable for the duration of the simulation, withthe chelating arginines (shown as bonds) moving very little relative to S PO CTT backbone (thin ribbons) does not adopt a single, stable structure.
Figure 3: Representative results from the analysis pipeline, showing a strong RC.
The Cy R ∗ forthe first 100 ns of the R2834H S2849S PO simulation under CHARMM36 is shown in (a), demonstrating theappearance of a strong claw (large, persistent clawicity value). Blue points show the arginine clawicity, Cy R ∗ ,defined as the number of hydrogen bonds made to arginines by the residue with the most hydrogen bondsto arginine; the green line is a 1-ns running average. The marginal distribution on the right is a histogramof piled-up Cy R ∗ values, ranging from 0 to 8: . The residue-specific Cy R (b) shows that Cy RS2849 frequently exceeds 6, and that only D2851 (immediately to the right of the dark strip) makes any othersubstantial contribution to this system’s Cy R ∗ . Figure 4: Backbone conformations across all simulations.
To gauge the frequency of any unusual(non-canonical) secondary structures, Ramachandran plots are shown for all of our (a) CHARMM36 and(b) PARM99SB data, compiled across all trajectories for all simulation systems. Regions corresponding tocanonical secondary structural elements are indicated as countour lines (see also the Methods section), andthe color-scale is graded by the log-likelihood of a particular ( φ , ψ ) conformation. In contrast to a recentreport, we find that DP CTT does not show a preference for left-handed helices under CHARMM36; thoserecent simulations of a non-phosphorylated SRR, unrelated to our DP systems, found that CHARMM36predicts that over 40% of the residues in the simulated peptide are in left-handed helices, even though only ≈
6% of the residues in a reference set of known protein structures exhibit such a structure. Our CHARMM36trajectory data do not indicate that left-handed helices are a problem, at least in our simulation systems.Combining all frames of every simulation, CHARMM36 predicts 9% left-handed helix, while PARM99SBpredicts 7%. 1 igure 5: A method to evaluate potential DP ···
GSK3 geometric complementarity via one-degree-of-freedom docking.
We begin with two molecules to be docked, as schematized in (a). Say weknow that the red atom on the yellow molecule aligns with the red atom on the blue molecule when themolecules interact. (In this two-dimensional case, only one atom is needed per molecule; in three dimensions,two atoms per molecule are needed to define an axis of rotation.) In step (b), the molecules have been aligned,in arbitrary angular orientation, based only on the positions of their red atoms. Next, the yellow molecule isrotated about the axis defined by the red atoms (c). At each step in the full rotational sweep (1 ◦ incrementsin our implementation), the clash score is computed as the number of pairs of atoms (one from DP, onefrom GSK3) with an interatomic distance less than 2 ˚A. The best pose (d) is taken as the one with theminimal clash score. This procedure is repeated for each frame along the trajectory. In a realistic example(e), two atoms (red, orange) from GSK3 are used to perform the alignment. The outward-facing oxygen ofthe phosphate (orange sphere) defines the recognition site, while the active site is defined by the side-chainoxygen of S261 (red sphere) of chain B (gray ribbon). The protein atoms from chain A (dark surface) wereused to calculate the clash. The atoms in DP that were used to perform the alignment are highlighted in (f).This frame, from 173.48-ns in the production trajectory of R Me PO Figure 6: S2849 in close proximity to the PRD.
This frame, from 71-ns in the R2834H, S2849S PO simulation under PARM99SB, exemplifies the contacts made between S2849 and residues that are part ofthe last plakin repeat domain (PRD C) in DP. Residues 2802–2805 are shown as van der Waals sphereson the left, and S2849S PO is shown as vdW spheres in the center. These close contacts suggest that theDP CTT can directly interact with the PRDs.
Table 1: Simulation systems and their Cy R ∗ histograms. Cy R ∗ values from the last 100 ns of eachsimulation are presented as histograms, where the intensity in a particular bin represents the frequencythat the system had the corresponding Cy R ∗ value. As an example, the bin numbers are explicitly shown in, which represents a simulation that frequently displayed Cy R ∗ values of 2, 3, and 4 (highest peaksin the histogram). CHARMM36 was consistently found to predict higher Cy R ∗ values than PARM99SB; interms of clawicity, CHARMM36 also predicts a stronger response to phosphorylation.2 cAnany and Mura (2016), Figure 1 cAnany and Mura (2016),
Figure 2 cAnany and Mura (2016),
Figure 3 cAnany and Mura (2016),
Figure 4 cAnany and Mura (2016),
Figure 5 cAnany and Mura (2016),
Figure 6 cAnany and Mura (2016),
Table 1
Force-fieldSimulation system Duration (per FF) PARM99SB CHARMM36Wild-type, unmodified 100 nsS2849S PO
200 nsR2834H 100 nsR2834H and S2849S PO
200 nsR2834R Me
100 nsR2834R Me and S2849S PO
200 nsS2849S
HPO (PARM99SB only) 100 ns – upporting Information for:Claws, Disorder, and Conformational Dynamicsof the C-terminal Region of Human Desmoplakin Charles E. McAnany and Cameron Mura ∗ Department of Chemistry, University of Virginia, Charlottesville, VA 22904, USA
E-mail: [email protected] ∗ To whom correspondence should be addressed S1 verview This document provides the detailed results of the analysis suite described in the Methods section of themain text, as applied to each of our simulation systems. Each trajectory has been analyzed in terms of (a)Cy R ∗ , (b) Cy R , (c) SASA of S2849, (d) SASA of R2834, (e) S2849 ····· R2834 distance, (f) GSK3 steric clash,(g) interresidue contact, (h) and backbone dihedral angles.
List of Figures
Figure Simulation system Page S1 WT_PARM99SB . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . S3S2
WT_CHARMM36 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . S4S3
S2849S1P_PARM99SB . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . S5S4
S2849S2P_PARM99SB . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . S6S5
S2849S2P_PARM99SB_cycle2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . S7S6
S2849S2P_CHARMM36 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . S8S7
S2849S2P_CHARMM36_cycle2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . S9S8
R2834H_PARM99SB . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . S10S9
R2834H_CHARMM36 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . S11S10
R2834H_S2849S2P_PARM99SB . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . S12S11
R2834H_S2849S2P_PARM99SB_cycle2 . . . . . . . . . . . . . . . . . . . . . . . . . . S13S12
R2834H_S2849S2P_CHARMM36 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . S14S13
R2834H_S2849S2P_CHARMM36_cycle2 . . . . . . . . . . . . . . . . . . . . . . . . . . S15S14
R2834MeMe_PARM99SB . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . S16S15
R2834MeMe_CHARMM36 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . S17S16
R2834MeMe_S2849S2P_PARM99SB . . . . . . . . . . . . . . . . . . . . . . . . . . . . S18S17
R2834MeMe_S2849S2P_PARM99SB_cycle2 . . . . . . . . . . . . . . . . . . . . . . . S19S18
R2834MeMe_S2849S2P_CHARMM36 . . . . . . . . . . . . . . . . . . . . . . . . . . . . S20S19
R2834MeMe_S2849S2P_CHARMM36_cycle2 . . . . . . . . . . . . . . . . . . . . . . . S21S2 a) Cy R ∗ (b) Cy R (c) SASA of S2849(d) SASA of R2834 (e) S2849 ····· S2845 distance (f) GSK3 steric clash(g) Contact map (h) Ramachandran plot
Figure S1: Behavior of
WT_PARM99SB : Time-series plots mark each observation as a blue point and contain a 1-ns running averageas a green trace, and the marginal distributions are shown on the right axis, in each of panels a, c, d, e, and f. S a) Cy R ∗ (b) Cy R (c) SASA of S2849(d) SASA of R2834 (e) S2849 ····· S2845 distance (f) GSK3 steric clash(g) Contact map (h) Ramachandran plot
Figure S2: Behavior of
WT_CHARMM36 : Time-series plots mark each observation as a blue point and contain a 1-ns running averageas a green trace, and the marginal distributions are shown on the right axis, in each of panels a, c, d, e, and f. S a) Cy R ∗ (b) Cy R (c) SASA of S2849(d) SASA of R2834 (e) S2849 ····· S2845 distance (f) GSK3 steric clash(g) Contact map (h) Ramachandran plot
Figure S3: Behavior of
S2849S1P_PARM99SB : Time-series plots mark each observation as a blue point and contain a 1-ns runningaverage as a green trace, and the marginal distributions are shown on the right axis, in each of panels a, c, d, e, and f. S a) Cy R ∗ (b) Cy R (c) SASA of S2849(d) SASA of R2834 (e) S2849 ····· S2845 distance (f) GSK3 steric clash(g) Contact map (h) Ramachandran plot
Figure S4: Behavior of
S2849S2P_PARM99SB : Time-series plots mark each observation as a blue point and contain a 1-ns runningaverage as a green trace, and the marginal distributions are shown on the right axis, in each of panels a, c, d, e, and f. S a) Cy R ∗ (b) Cy R (c) SASA of S2849(d) SASA of R2834 (e) S2849 ····· S2845 distance (f) GSK3 steric clash(g) Contact map (h) Ramachandran plot
Figure S5: Behavior of
S2849S2P_PARM99SB_cycle2 : Time-series plots mark each observation as a blue point and contain a 1-nsrunning average as a green trace, and the marginal distributions are shown on the right axis, in each of panels a, c, d, e, and f. S a) Cy R ∗ (b) Cy R (c) SASA of S2849(d) SASA of R2834 (e) S2849 ····· S2845 distance (f) GSK3 steric clash(g) Contact map (h) Ramachandran plot
Figure S6: Behavior of
S2849S2P_CHARMM36 : Time-series plots mark each observation as a blue point and contain a 1-ns runningaverage as a green trace, and the marginal distributions are shown on the right axis, in each of panels a, c, d, e, and f. S a) Cy R ∗ (b) Cy R (c) SASA of S2849(d) SASA of R2834 (e) S2849 ····· S2845 distance (f) GSK3 steric clash(g) Contact map (h) Ramachandran plot
Figure S7: Behavior of
S2849S2P_CHARMM36_cycle2 : Time-series plots mark each observation as a blue point and contain a 1-nsrunning average as a green trace, and the marginal distributions are shown on the right axis, in each of panels a, c, d, e, and f. S a) Cy R ∗ (b) Cy R (c) SASA of S2849(d) SASA of R2834 (e) S2849 ····· S2845 distance (f) GSK3 steric clash(g) Contact map (h) Ramachandran plot
Figure S8: Behavior of
R2834H_PARM99SB : Time-series plots mark each observation as a blue point and contain a 1-ns runningaverage as a green trace, and the marginal distributions are shown on the right axis, in each of panels a, c, d, e, and f. S a) Cy R ∗ (b) Cy R (c) SASA of S2849(d) SASA of R2834 (e) S2849 ····· S2845 distance (f) GSK3 steric clash(g) Contact map (h) Ramachandran plot
Figure S9: Behavior of
R2834H_CHARMM36 : Time-series plots mark each observation as a blue point and contain a 1-ns runningaverage as a green trace, and the marginal distributions are shown on the right axis, in each of panels a, c, d, e, and f. S a) Cy R ∗ (b) Cy R (c) SASA of S2849(d) SASA of R2834 (e) S2849 ····· S2845 distance (f) GSK3 steric clash(g) Contact map (h) Ramachandran plot
Figure S10: Behavior of
R2834H_S2849S2P_PARM99SB : Time-series plots mark each observation as a blue point and contain a 1-nsrunning average as a green trace, and the marginal distributions are shown on the right axis, in each of panels a, c, d, e, and f. S a) Cy R ∗ (b) Cy R (c) SASA of S2849(d) SASA of R2834 (e) S2849 ····· S2845 distance (f) GSK3 steric clash(g) Contact map (h) Ramachandran plot
Figure S11: Behavior of
R2834H_S2849S2P_PARM99SB_cycle2 : Time-series plots mark each observation as a blue point andcontain a 1-ns running average as a green trace, and the marginal distributions are shown on the right axis, in each of panels a, c, d, e, andf. S a) Cy R ∗ (b) Cy R (c) SASA of S2849(d) SASA of R2834 (e) S2849 ····· S2845 distance (f) GSK3 steric clash(g) Contact map (h) Ramachandran plot
Figure S12: Behavior of
R2834H_S2849S2P_CHARMM36 : Time-series plots mark each observation as a blue point and contain a1-ns running average as a green trace, and the marginal distributions are shown on the right axis, in each of panels a, c, d, e, and f. S a) Cy R ∗ (b) Cy R (c) SASA of S2849(d) SASA of R2834 (e) S2849 ····· S2845 distance (f) GSK3 steric clash(g) Contact map (h) Ramachandran plot
Figure S13: Behavior of
R2834H_S2849S2P_CHARMM36_cycle2 : Time-series plots mark each observation as a blue point andcontain a 1-ns running average as a green trace, and the marginal distributions are shown on the right axis, in each of panels a, c, d, e, andf. S a) Cy R ∗ (b) Cy R (c) SASA of S2849(d) SASA of R2834 (e) S2849 ····· S2845 distance (f) GSK3 steric clash(g) Contact map (h) Ramachandran plot
Figure S14: Behavior of
R2834MeMe_PARM99SB : Time-series plots mark each observation as a blue point and contain a 1-ns runningaverage as a green trace, and the marginal distributions are shown on the right axis, in each of panels a, c, d, e, and f. S a) Cy R ∗ (b) Cy R (c) SASA of S2849(d) SASA of R2834 (e) S2849 ····· S2845 distance (f) GSK3 steric clash(g) Contact map (h) Ramachandran plot
Figure S15: Behavior of
R2834MeMe_CHARMM36 : Time-series plots mark each observation as a blue point and contain a 1-nsrunning average as a green trace, and the marginal distributions are shown on the right axis, in each of panels a, c, d, e, and f. S a) Cy R ∗ (b) Cy R (c) SASA of S2849(d) SASA of R2834 (e) S2849 ····· S2845 distance (f) GSK3 steric clash(g) Contact map (h) Ramachandran plot
Figure S16: Behavior of
R2834MeMe_S2849S2P_PARM99SB : Time-series plots mark each observation as a blue point and containa 1-ns running average as a green trace, and the marginal distributions are shown on the right axis, in each of panels a, c, d, e, and f. S a) Cy R ∗ (b) Cy R (c) SASA of S2849(d) SASA of R2834 (e) S2849 ····· S2845 distance (f) GSK3 steric clash(g) Contact map (h) Ramachandran plot
Figure S17: Behavior of
R2834MeMe_S2849S2P_PARM99SB_cycle2 : Time-series plots mark each observation as a blue point andcontain a 1-ns running average as a green trace, and the marginal distributions are shown on the right axis, in each of panels a, c, d, e, andf. S a) Cy R ∗ (b) Cy R (c) SASA of S2849(d) SASA of R2834 (e) S2849 ····· S2845 distance (f) GSK3 steric clash(g) Contact map (h) Ramachandran plot
Figure S18: Behavior of
R2834MeMe_S2849S2P_CHARMM36 : Time-series plots mark each observation as a blue point and containa 1-ns running average as a green trace, and the marginal distributions are shown on the right axis, in each of panels a, c, d, e, and f. S a) Cy R ∗ (b) Cy R (c) SASA of S2849(d) SASA of R2834 (e) S2849 ····· S2845 distance (f) GSK3 steric clash(g) Contact map (h) Ramachandran plot
Figure S19: Behavior of