Abstract
The cyclical interaction of myosin with F-actin and nucleotides is the basis for contractility of the actin cytoskeleton. Despite a generic, highly conserved motor domain, ATP turnover kinetics and their activation by F-actin vary greatly between myosins-2 isoforms. Here, we present a 2.25 Å crystal structure of the human nonmuscle myosin-2C motor domain, one of the slowest myosins characterized. In combination with integrated mutagenesis, ensemble-solution kinetics, and molecular dynamics simulations approaches, this study reveals an allosteric communication pathway that connects the distal end of the motor domain with the active site. Genetic disruption of this pathways reduces nucleotide binding and release kinetics up to 85-fold and abolishes nonmuscle myosin-2 specific kinetic signatures. These results provide insights into structural changes in the myosin motor domain that are triggered upon F-actin binding and contribute critically to the mechanochemical behavior of stress fibers, actin arcs, and cortical actin-based structures.
Introduction
The extent to which filamentous actin (F-actin) can activate the enzymatic output of conventional myosins-2 varies by more than two orders of magnitude (1, 2). Despite a substantial sequence identity and a highly conserved actomyosin ATPase cycle, structural and allosteric adaptations causative for the tremendous enzymatic and hence physiological differences are largely unknown (Figure 1A) (1, 3).
Nonmuscle myosin-2C is one of the slowest myosins-2 as its steady-state ATPase activity lacks potent F-actin activation (4). Transient kinetic signatures of the nonmuscle myosin-2C enzymatic cycle include a high affinity for ADP, small kinetic (k−AD/k−D) and thermodynamic (KAD/KD) coupling ratios, and a higher duty ratio than it is commonly found in myosins-2 (Figure 1A - Figure supplement 1A). Together, these signatures qualify nonmuscle myosin-2C as a dynamic strain-sensing actin tether (3-5). The participation in the active regulation of cytoplasmic contractility in cellular processes including cytokinesis, neuronal dynamics, adhesion, and tension maintenance are in line with this interpretation (6-8). Nonmuscle myosin-2C has recently received great attention as the near atomic resolution cryo electron microscopic structure of its motor domain bound to F-actin has been determined (9). The structure of the actomyosin complex, together with a detailed kinetic characterization of the motor properties, qualifies nonmuscle myosin-2C as the ideal human myosin to decipher the structure-function relationships in the myosin motor domain that underlie its kinetic signatures (4, 9).
To understand how structural adaptations lead to characteristic kinetic features, we solved the 2.25 Å crystal structure of the human nonmuscle myosin-2C motor domain (NM2C) (Figure 1B, Table 1). The NM2C pre-powerstroke state structure suggests that structural fine-tuning at the active site and a reduced interdomain connectivity in the motor domain define its kinetic signatures. Comparative structural analysis, ensemble solution kinetic studies, and in silico molecular dynamics (MD) simulations collectively demonstrate a communication pathway that connects the active site and the distal end of the motor domain. Disruption of the pathway uncouples the myosin ATPase activity from actin-activation, alters the ATP/ADP sensitivity and results in the loss of nonmuscle myosin-2 kinetic signatures. The allosteric coupling pathway may be of significance in intermolecular gating and load-sensitivity of nonmuscle myosins-2 that is important for their physiological function as cytoskeletal strain-sensor in the context of stress fibers, actin arcs, and cortical actin-based structures.
Results
Overall topology of pre-powerstroke NM2C
We determined a 2.25 Å pre-powerstroke state structure of human NM2C, one of the slowest myosins-2 characterized. Crystallization was achieved after truncation of the flexible N-terminal 45 residues and the fusion of the NM2C C-terminus to Dictyostelium α-actinin tandem repeats 1 and 2 (Figure 1B, Figure 1 - Figure supplement 1B, Table 1). This approach has been used successfully in prior structural and kinetic studies on NM2C and allows the detailed analysis of structure-function relationships in the myosin motor domain (4, 9).
The overall topology of NM2C shares the structural four-domain architecture of myosin-2 motor domains comprising the N-terminal subdomain (Nter), the upper 50 kDa subdomain (U50 kDa), the lower 50 kDa subdomain (L50 kDa), and the converter that terminates in the lever (Figure 1C) (10). The nucleotide-binding active site is formed by structural elements of the U50 kDa and allosterically communicates with the actin binding interface that is formed by distant structural elements of the U50 kDa and the L50 kDa. The NM2C Cα atoms superimpose with a root mean square deviation (r.m.s.d.) of 0.57 Å, 0.78 Å, and 0.73 Å to the closely related motor domain structures from chicken smooth muscle myosin-2 (PDB entry 1BR2), scallop striated muscle myosin-2 (PDB entry 1QVI), and Dictyostelium nonmuscle myosin-2 (PDB entry 2XEL) (Figure 1 - Figure supplement 1C) in the pre-powerstroke state.
NM2C pre-powerstroke state characteristics include a closed conformation of the nucleotide binding motifs switch-1 and -2 in the active site (Figure 1B). In this conformation, the characteristic salt bridge between switch-1 R261 and switch-2 E483 that is pivotal for the hydrolysis of ATP is formed (10-12). The relay helix is in a kinked conformation, the central seven-stranded β-sheet is untwisted, and the converter domain and the adjacent lever arm in the up-position (Figure 1B). NM2C switch-1 and lever dihedral φ, ψ angles are substantially changed when compared to the pre-powerstroke state structures of chicken smooth muscle myosin-2 (PDB entry 1BR2), indicating that small arrangements in the active site are coupled to conformational changes at the distal end of the motor domain (Figure 1 - Figure supplement 1C-D, Supplementary table 1). Concomitantly, the relative orientation of converter and lever arm deviates by ~8-10° when compared to the high resolution pre-powerstroke state crystal structure of chicken smooth muscle myosin-2 (PDB entry 1BR2) and resembled the orientation of the respective regions of scallop striated muscle myosin-2 (PDB entries 1QVI, 1DFL) motor domain structures (Figure 1 - Figure supplement 1D).
Unique structural rearrangements in the NM2C active site in the pre-powerstroke state
Switch-1, switch-2, P-loop, and the purine-binding A-loop are the prototypic nucleotide binding motifs in the myosin active site that undergo conformational changes in response to nucleotide binding and release. The active site is flanked by loop-3 and a loop that connects helices J and K in the U50 kDa, hereafter referred to as JK-loop (Figure 2A). The JK-loop constitutes a major connection between switch-1 and the nucleotide is of great functional significance as a hotspot for human cardiomyopathy-causing mutations in cardiac myosin-2 and the site of exon 7 in Drosophila muscle myosin-2 (Figure 2 - Figure supplement 1A) (13-15).
A slight reduction in the JK-loop length in NM2C causes an 8.8 Å shift between switch-1 and the U50 kDa and abolishes the formation of a tight interaction network with residues of the active site found in other class-2 myosins (Figure 2B,C, Figure 2 - Figure supplement 1A,B,C). Comparative analysis of interactions between the JK-loop and the switch-1 region in NM2C and scallop striated muscle myosin-2 (PDB entry 1QVI) shows that in the latter, JK-loop N321 is in hydrogen bond interaction with switch-1 N238 and located at a distance of 4.6 Å to the hydroxyl group of the C2’ of the ADP ribose. The connectivity between switch-1 and the nucleotide is further strengthened by a hydrogen bond between N237 and the ADP ribose. Moreover, A-loop residue R128 forms hydrogen bonds with the ADP adenosine (3.2 Å) and E184 (2.8 Å) of the P-loop in the active site of striated muscle myosin-2. Strikingly, NM2C lacks all interactions described for scallop striated muscle myosin-2 due to the replacement of R128 with Q150 and the above mentioned JK-loop shortening. Both structural alterations increase the distance to the adenosine in the active site to 5.8 Å and disrupt constrains between swich-1 and the JK-loop (Figure 2B,C,E). The shift also increases the volume and hence the accessibility of the NM2C active site when compared to the transition state structures of chicken smooth muscle myosin-2 (PDB entry 1BR2), Dictyostelium nonmuscle myosin-2 (PDB entry 2XEL), and scallop striated muscle myosin-2 (PDB entry 1DFL,1QVI) (Figure 2A, Figure 2 - Figure supplement 1B,D,E). A 4.3-fold volume increase from 697 Å3 to 3054 Å3 in the active site is observed for example between scallop striated muscle myosin-2 (PDB entry 1QVI) and NM2C (Figure 2 - Figure supplement 1D,E). Strikingly, the enlarged accessibility of the active site region observed in the NM2C pre-powerstroke state structure resembles the conformation of the active site found in the actin-bound NM2C rigor state (PDB entry 5J1H), indicating that F-actin binding does not induce major structural rearrangements in the vicinity of the active site (Figure 2C).
The temperature factors obtained from the refined crystal structure show that the NM2C JK-loop is highly flexible despite its reduced length. We attribute the flexibility to the lack of constraints with switch-1 residues and the bound nucleotide (Figure 2A,B,C,E). Specifically, the connectivity of the JK-loop to switch-1 is reduced by the replacement of a conserved arginine (R280 in PDB entry 1QVI) with cysteine (C299) in the loop preceding helix I. This arginine further establishes an interaction with the highly conserved switch-1 K233 in scallop striated muscle myosin-2. The arginine to cysteine substitution in NM2C also abolishes the formation of a salt bridge with JK-loop residue E341 and hence disrupts the connection between the JK-loop and switch-1 (Figure 2C,E, Figure 2 - Figure supplement 1C). Consequently, the JK-loop loses its ability to sense conformational changes in the nucleotide binding motif switch-1 in response to ATP binding, hydrolysis, and product release.
The interaction between the JK-loop and the nucleotide is weakened by the replacement of an asparagine with G339, which abolishes hydrogen bond formation with switch-1 D257. Comparison with data from previous crystallographic studies shows that D257 replaces an asparagine (N238 in PDB entry 1QVI) in the switch-1 in striated muscle myosins-2. The asparagine interacts weakly with the ADP moiety in the pre-powerstroke state (Figure 2B,D,E, Figure 2 figure supplement 1B,C) and strongly in the near-rigor state (16, 17). The nucleotide coordination in NM2C is further weakened by residue Q150 that replaces an arginine (R128 in PDB entry 1QVI) in the A-loop of scallop striated muscle myosin-2. The substitution disrupts coordinating interactions between the A-loop and the ADP adenosine (Figure 2D, Figure 2D, figure 2 supplement 1B,C). The replacement additionally impairs the formation of a salt bridge with the invariant E206 (E184 in PDB entry 1QVI) at the distal end of the P-loop that is predicted to be involved in nucleotide recruitment to the active site (Figure 2D, Supplementary table 2) (16).
Previous studies established that the number of hydrogen bond interactions between P-loop, switch-1, and the Nter positively correlate with the thermodynamic coupling (KAD/KD), the efficiency of F-actin to displace ADP during the catalytic cycle (16). Substitution of an arginine with K255 and a glutamate with H700 in NM2C is expected to weaken the tight and highly conserved interaction network and is expected to contribute to the low thermodynamic (KAD/KD) and kinetic (k−AD/k−D) coupling efficiency, and the slow actin-activated ADP release in nonmuscle and smooth muscle myosins-2 compared to cardiac and striated muscle myosins-2 (Supplementary table 2) (16).
Taken together, comparative structural analysis of the active site of myosins-2 suggests that the interconnectivity of the highly conserved nucleotide switches and myosin subdomains is weaker in NM2C and other nonmuscle myosins-2 compared to fast sarcomeric myosins-2.
Interdomain connectivity between converter, Nter, and lever
The interdomain connectivity between the converter and the Nter in the myosin-2 motor domain is established by the relay and the SH1-SH2 helix. The Nter controls the movement of the converter, which undergoes a large-scale rotation that drives the powerstroke during force generation (18-22).
Residue R788 at the interface of converter and lever is highly conserved in nonmuscle and smooth muscle myosins-2 and the only connecting hub between structural elements of the L50 kDa and the Nter (Figure 3A,B). Notably, replacement of R788 with a lysine in muscle myosins-2 weakens the interaction between the aforementioned structural elements in all states of the myosin and actomyosin kinetic cycle (Figure 3 - Figure supplement 1A,B). In NM2C, R788 forms 3 main chain and 5 side chain interactions with residues of the SH1-SH2 helix, the converter, and the lever in the pre-powerstroke state. Specifically, the guanidinium group of the R788 side chain of NM2C forms hydrogen bonds with the main chain carbonyls of residues Q730 and G731 of the SH1 helix. The hydrophobic methylene groups of R788 are stabilized by SH1 helix F732 and the main chain oxygen and hydroxyl groups of N776 of the converter. Y518 from the relay helix interacts with residues G731 and F732 of the converter, thereby interlinking R788 with both structural elements. The side chain of R788 is in van der Waals distance (4.8 Å) from W525 of the relay loop. The main chain carbonyl of R788 further interacts with V791 at the converter/lever junction (Figure 3A,B). This interaction is important for the stabilization of the converter fold and the interface with the lever in the absence of F-actin. Notably, the R788 side chain-main chain interaction is also evident in crystal structures of nonmuscle myosins-2 in the nucleotide-free (PDB entry 4PD3) and the phosphate-release state (PDB entry 4PJK) (not shown). We therefore hypothesize that (i) the tight interaction network formed by R788 is required for the precise positioning and coupling of the converter, the SH1-SH2 helix, the relay helix, and the lever arm throughout all steps of the myosin ATPase cycle (Figure 1 - Figure supplement 1A), and (ii) different communication pathways between the active site and the distal end of the motor domain are employed by NM2C in the presence and absence of F-actin.
Kinetic consequences after the disruption of the converter/Nter/lever interface
To experimentally probe for a possible effect of R788 on myosin motor function, we performed comparative ensemble solution kinetic studies with NM2C and R788E, a mutant in which a glutamate replaces the converter R788. Rationale for the design of the R788E charge reversal mutant was to disrupt side chain:side chain and side chain:main chain interactions, an approach that has been successfully used in in vitro and in vivo studies to probe for the influence of the converter/U50 kDa interface on myosin-2 performance (19, 23-25).
R788E shows prominent changes in virtually all parameters of the myosin and actomyosin ATPase cycle (Supplementary table 3). Confirming our hypothesis, transient kinetic changes mainly affect nucleotide binding and release kinetics and are more pronounced in the absence of F-actin, as seen in single-turnover measurements (Figure 4A,B). Most importantly, nonmuscle myosin-2 specific transient kinetic signatures such as a high k+AD/K1k+2 ratio (k+AD/K1k+2 ~ 2-20) are absent in R788E (k+AD/K1k+2 ~ 0.26) due to a 20-fold acceleration of the second-order rate constant for ATP binding (K1k+2) by F-actin (Table 2, Figure 4 - Figure supplement 1A-D) (4, 26, 27). This feature is also described for conventional myosins-2 from cardiac and striated muscle that bind ATP and ADP with similar rates (Supplementary table 2) (28). Actin-activation of the ADP release (k−AD = 0.68±0.01 s-1 for NM2C and k−AD = 0.19±0.01 s-1 for R788E) results in a kinetic coupling constant k−AD/k−D of ~ 2.5 for R788E, whereas neutral or negative kinetic coupling is a feature of NM2C (k−AD/k−D =0.7) and other human nonmuscle myosins-2 (Table 2) (4, 26, 27). The changes in ADP binding and release kinetics of R788E result in a 10-fold increase in the ADP dissociation equilibrium constant KAD (KAD ~ 0.29 μM for NM2C and ~ 2.68 μM for R788E). The thermodynamic coupling (KAD/KD=5) for R788E is 42-times higher than for NM2C (Supplementary table 3). R788E displays an extraordinary slow ADP binding rate constant (k+AD=0.03±0.001 μM-1s-1) (Table 2, Figure 4 - Figure supplement 1D) in the presence of F-actin whereas the kinetic constants for the interaction between actomyosin and ATP, K1K+2 and 1/K1 are marginally affected when compared to NM2C (Supplementary table 3). The F-actin affinity in the absence and presence of ADP (KA, KDA) of NM2C and R788E is similar (Supplementary table 3) and F-actin can activate the steady-state ATPase activity of R788E to approximately half the kcat of NM2C (kcat = 0.2±0.01 s-1 for R788E and kcat = 0.37±0.02 s-1 of NM2C) (Supplementary table 3) under steady-state conditions. The duty ratio at an F-actin concentration of 190 μM increases from ~ 0.3 for NM2C to ~ 1 for R788E due to the decreased kcat and the decreased actin-activated ADP release rate k−AD. This feature makes k−AD likely to rate-limit the kinetic cycle of R788E, whereas the actin-activated Pi release is expected to rate-limit the kinetic cycle of NM2C and other myosins-2 (1, 4, 26, 27, 29, 30).
The slow nucleotide binding and release kinetics of R788E suggest that the R788-mediated interaction at the converter/Nter/lever interface is allosterically communicated to the active site. This is in line with the observation that R788E does not show a nucleotide-induced change in the intrinsic tryptophan fluorescence signal during steady-state and transient-state kinetic assays (Figure 5D,F, Figure 4 - Figure supplement 1E, Table 3). The fluorescence change observed with NM2C is attributed to a conformation-induced change in the microenvironment of the conserved relay loop W525, which is in van der Waals distance to R788 and a direct indicator for the switch-2 induced converter rotation (31).
Allosteric communication pathway between the active site and the distal end of the motor domain
R788 is a hub amino acid and forms the center of a cluster of interactions that connect the converter, the SH1-SH2 helix, the relay helix, and the lever (Figure 3A,B). To understand its dynamic interactions that are allosterically communicated to the active site, we performed comparative molecular dynamics simulations of NM2C and R788E in explicit water with Mg2+. ATP bound to the active site over a time of 100 ns. As expected, the characteristic salt bridge between switch-1 R261 and switch-2 E483 of the active site that is critical for the hydrolysis of ATP is stable throughout the time course of the simulation for NM2C (Figure 5 - Figure supplement 1A). The relay helix that connects the active site and the converter straightens and likewise the converter undergoes a 27° rotation that directs lever arm motion (Figure 5A,C). This straightening is caused by the exchange of the co-crystallized ADP-VO4 to ATP, since the replacement with ADP did not lead to either relay straightening nor converter rotation in a control simulation. Hence, ATP appears to trigger the structural transition from the pre-powerstroke state (up position) towards the post-rigor state (down position). After 100 ns, the position of the converter and the adjacent lever show an orientation between the post-rigor and pre-powerstroke state, as compared to the crystal structures of Dictyostelium nonmuscle myosin-2 motor domain (PDB entries 1FMW, 2XEL) (Figure 5C). With the movement of the relay helix, the indole ring of relay loop W525 changes its conformation by 70-80°, which agrees with the experimentally observed nucleotide-induced change in the intrinsic tryptophan fluorescence signal in NM2C (Figure 5B,C, Figure 4 - Figure supplement 1E). The hub amino acid R788 is in transient interactions with main chain and side chain atoms of 10 amino acids of the SH1 helix and the converter during the 100 ns time course of the simulation (Figure 5B).
MD simulations for R788E indicate that all interdomain interactions between R788E of the converter and the SH1 helix are lost (Figure 5B). The straightening angle of the relay helix remains constant and abolishes a detectable converter rotation (Figure 5A,E). This allosteric decoupling at the distal end of the myosin motor domain is translated further upstream via the relay helix and results in a pronounced 6 Å conformational change of a loop that connects the γ-phosphate sensor switch-2 of the active site and the relay helix. As a consequence, the salt bridge between switch-2 E483 and switch-1 R261 appears less stable in the MD simulations of R788E as compared to NM2C (Figure 5 - Figure supplement 1A,B). The importance of the salt bridge between both switches for the ATP hydrolysis is established and expected to directly contribute to the experimentally observed impaired nucleotide binding, hydrolysis, and release kinetics of R788E in transient-state kinetic assays (Supplementary table 3, Figure 4A, Figure 4 - Figure supplement 1A,D) (11, 32, 33). Moreover, the side chain of switch-1 N256 changes its orientation, thereby impairing hydrogen bond interactions with the α- and β-phosphate moieties of the nucleotide and constrains of switch-1. The lack of coordinating interactions is reflected in the very slow nucleotide binding and release rate constants (Figure 4A, Figure 4 - Figure supplement 1A, Supplementary table 3).
The side chain of relay loop W525 does not undergo a conformational change throughout the time course of the simulation for R788E, supporting our experimental observation that R788E does not exhibit a nucleotide-induced change in its intrinsic tryptophan fluorescence signal of R788E (Figure 5D,F, Figure 4 - Figure supplement 1E). The direct comparison of the dynamic interaction and conformational signatures of W525 in NM2C and R788E reveals that R788 is in van der Waals distance (5.4 Å) from relay loop W525 at the start of the simulations. W525 transiently interacts with I523 (18%) and Q519 (14%), changes its conformation by 70-80° and increases the distance to NM2C R788 to 7.3 Å after 100 ns simulation time. In comparison, the distance between W525 and E788 increases to 9.9 Å during the simulation for R788E. The lack of a conformational change in W525 along the simulation trajectory results in transient interactions with Q515 (55%) and Q730 (14%) instead of the interactions with I532 and Q519 as seen in NM2C.
Taken together, our in silico and experimental in vitro data support a model in which R788 is a distant allosteric modulator of switch-2 dynamics at the active site that impacts nucleotide binding and release kinetics in the actin-detached states. Moreover, our data support a model of R788 as a quencher of W525 fluorescence.
Discussion
The experiments presented here collectively demonstrate an allosteric communication pathway from the distal end of the myosin motor domain that, together with substitutions of several key residues in or in vicinity to the active site, account for NM2C-specific kinetic properties. Disruption of the pathway by mutation of R788 to a glutamate causes the loss of its enzymatic signatures and results in a high duty ratio motor.
It is of note that several key residues involved in the communication pathway are near residues that are mutated in patients with autosomal dominant hearing loss (34). Missense mutation G376C is in proximity to residues C324 of helix J and R328 of the JK-loop (Figure 4 - Figure supplement 1F). Based on its location, we suggest that this substitution may interfere with the nucleotide binding and release kinetics from the NM2C active site. Missense mutation R726S is in the SH1-SH2 helix and the guanidinium group of the wild type arginine interacts (3.3 Å) with the NM2C Nter. The serine residue is expected to disrupt this interaction because of the shorter side chain and different charge when compared to arginine. It is therefore likely that these mutations impact the interaction of the motor domain with nucleotides, thereby contributing to impaired tension-sensing and maintenance of nonmuscle myosin-2C in the human cochlea (34-36).
R788 is part of a conserved pathway that connects the active site and the converter
R788 is part of an allosteric communication pathway that connects the converter at the distal end of the myosin motor domain via the relay helix with switch-2 of the active site. Uncoupling of the converter from the motor domain in R788E slows down all kinetic steps of the myosin ATPase cycle and decreases its actin-activation, due to altered switch-2 dynamics. The observed kinetic phenotype of the R788E myosin ATPase cycle is similar to the Dictyostelium nonmuscle myosin-2 non-hydrolyzer mutants in which the salt bridge between switch-1 and switch-2 is destroyed by mutagenesis (11, 32). Non-hydrolyzer mutants are characterized by long-lived ATP states and reduced nucleotide release and binding kinetics, including a drastic decrease in k+AD (11, 32). Like R788E, Dictyostelium myosin-2 non-hydrolyzer mutants do not exhibit a nucleotide-induced change in the intrinsic fluorescence signal (11). It is of note that the lack of an intrinsic fluorescence signal in R788E is caused by the blockage of the communication pathway on the relay helix before or at Y518 (Figure 6A,B) whereas the impairment of the nucleotide switches to form a salt bridge and the resulting lack of the switch-2 induced conformational change of the relay helix is the expected cause in the non-hydrolyzer mutants.
F-Actin affinities are largely unaffected by the R788E mutation, which is in line with the observation that switch-1 dynamics are only marginally affected in comparative MD simulations. The presence of F-actin however establishes ATP/ADP selectivity in the actomyosin ATPase cycle: ActoNM2C preferentially binds ADP over ATP, whereas actoR788E preferentially binds ATP over ADP (Supplementary table 3). The ATP selectivity is caused by a 85-fold decreased second-order ADP binding rate constant k+AD, a key signature of R788E actomyosin ATPase cycle. ATP/ADP selectivity is established by actin-induced conformational changes in the myosin motor domain and underlines that the coupling mechanism from the active site to the converter and vice versa is different in the presence and absence of F-actin. This observation is in line with recent reports on distinct pathways for the myosin and actin-activated ATPase cycle (21).
Implications for NM2C function in vivo
Nonmuscle myosin-2C assembles into small bipolar filaments that dynamically tether actin filaments in cells (8, 37). The actomyosin interaction and hence the tension exerted by a sarcomeric array of nonmuscle myosin-2C is of importance for the function and organization of the apical junctional complex in the organ of Corti and actin-rich structures including stress fibers and actin arcs (8).
Based on our kinetic data, the calculated duty ratio of NM2C suggests that ~ 14 motor domains would be needed to be geometrically capable of interacting with F-actin at any given time. This number is identical to the number of motor domains per nonmuscle myosin-2C half filament, suggesting that the filament may be at the threshold of being processive in the absence of external loads (37). It is likely that this threshold is crossed in the presence of other actin binding proteins, intermolecular loads, and gating between the F-actin bound motor domains (38).
The structural prerequisites underlying gating and load-sensitivity in nonmuscle myosins-2 have not been investigated but include a distortion at the converter/lever or the converter/motor domain interface in the nanometer range that is allosterically communicated through the lever arm via the converter to the active site (39). Resisting load applied to the lever slows down the actin-activated ADP release from the lead motor of nonmuscle myosin-2A around 5-fold thereby increasing the duty ratio, but does not alter the rate of ATP binding (38, 39). The kinetic signatures of the strained nonmuscle myosin-2A lead motor are very similar to the observed kinetic features of R788E. The reduction of the steady-state ATPase of R788E and the concomitant increase in duty ratio and a rate-limiting ADP release rate goes in line with this finding. We propose that internal strain in the nonmuscle myosin-2 dimer distorts the lead motor at the converter/lever interface and leads to its axial translation as seen in electron microscopic studies (40). This translation abolishes the interaction of the converter R788 with residues of the SH1-SH2 helix and the relay helix, thereby uncoupling the myosin subdomains and disrupting the communication pathway from the converter to the active site, which is expected to result in a similar kinetic effect as in R788E. Consequently, a motor in a nonmuscle myosin-2 filament would decrease its enzymatic activity and stay strongly bound to F-actin. This feature is required to generate processivity and cytoskeletal tension and of physiological significance in the maintenance of cell shape and tensional homeostasis in the actin cytoskeleton.
Material and Methods
Protein production
For structural studies, a His8-tagged human NM2C construct comprising amino acids 45-799 directly fused to spectrin repeats 1 and 2 from α-actinin was generated based on the vector pFastBac1-NMHC-2C0-2R-His8 (4). Mutagenesis was accomplished by sequence-specific deletion using a whole-plasmid amplification approach. For kinetic studies, an equivalent motor domain construct comprising amino acids 1-799 of NM2C was cloned into a modified pFastBac1 vector containing a cDNA sequence encoding a C-terminal Flag-tag (Figure 1 - Figure supplement 1B). This construct was used as a template to introduce the R788E mutation by InFusion cloning (Clontech, Mountain View, CA 94943, USA). All proteins were recombinantly overproduced in the Sf9/baculovirus system, purified to electrophoretic homogeneity, and concentrated to ~10 mg/ml using Vivaspin ultrafiltration units (Sartorius, Göttingen, Germany) as previously described (4, 41). Throughout this manuscript, numbering refers to the amino acid sequence of full-length nonmuscle myosin-2C (GenBank accession number NP_079005).
Crystallization of NM2C
NM2C at a concentration of ~10 mg/ml was complexed with the ATP analogue ADP·VO4 and crystallized using the hanging drop vapor diffusion method by mixing 2 μl of protein solution with 2 μl of reservoir solution containing 50 mM Tris pH 8.2, 10 % (w/v) PEG-5K MME, 1 % (v/v) MPD, and 0.2 M NaCl at 8°C. Rectangular plate shaped crystals grew up to 400 × 300 × 400 μm3 within 4 weeks. Crystals were soaked in the corresponding mother liquor supplemented with 100 mM NaCl and 20 % (w/v) ethylene glycol. Protein crystals were transferred in serial steps of increasing concentrations of the cryo-solution. Crystals were transferred into liquid nitrogen using MiTeGen loops and stored at 100 K until data collection.
Data collection, processing and refinement
X-ray diffraction data of NM2C crystals were collected at the beam line BL14.1 at Bessy II (Helmholtz-Zentrum, Berlin, Germany) to 2.25 Å resolution (Table 1). Data processing was performed using XDS and SADABS (42). The structure of NM2C in the pre-powerstroke state was solved by molecular replacement using Phaser (43). The crystal structure of the chicken smooth muscle myosin-2 motor domain (PDB entry 1BR4) was used as a search model and the two α-actinin repeats were manually traced and rebuilt. The electron density map was sharpened using Coot (44) to ensure the directionality and identity of the α-helices for the two α-actinin repeats. Maximum likelihood crystallographic refinement was performed using iterative refinement cycles in autoBUSTER (45). Iterative cycles of model building were performed using Coot, and model bias was minimized by building into composite omit maps. The model was initially validated in Coot and final validation was performed using MolProbity (46).
Kinetic experiments
The actin-activated ATPase assays under steady-state conditions was performed as described earlier in buffer containing 10 mM MOPS pH 7.0, 50 mM NaCl, 2 mM MgCl2, 2 mM ATP, 0.15 mM EGTA, 40 U/ml l-lactic dehydrogenase, 200 U/ml pyruvate kinase, 200 μM NADH, and 1 mM phosphoenolpyruvate at 25°C with a Cary 60 Bio spectrophotometer (Agilent Technologies, Wilmington, DE 19808, USA) (41). Transient state kinetic assays were carried out as described previously with a TgK Hi-tech Scientific SF-61 DX stopped-flow system (TgK Hi-tech Scientific Ltd., Bradford-on-Avon, UK) in SF-buffer (25 mM MOPS pH 7.0, 100 mM KCl, 5 mM MgCfe and 0.1 mM EGTA) unless stated otherwise (4). Initial data fitting was performed with Kinetic Studio Version 2.28 (TgK Hi-tech Scientific Ltd., Bradford-on-Avon, UK). Plots were generated with OriginPro 8.5 (OriginLab Corp., Northampton, MA 01060, USA). Data interpretation is according to the kinetic scheme of the myosin and actomyosin ATPase cycle as presented in Figure 1 - Figure supplement 1A.
Fluorescence measurements
Tryptophan fluorescence emission spectra of 4 μM NM2C or R788E in the presence and absence of 0.5 mM ADP and 0.5 mM ATP, respectively were measured after excitation at 297 nm at a temperature of 20 °C in a QuantaMaster fluorescence spectrophotometer (Photon Technology International, Birmingham, NJ 08011, USA) as described previously (47). Prior to the assay, proteins were transferred to SF-buffer with zebra spin desalting columns (Thermo Fisher Scientific GmbH, Dreieich, Germany). The fluorescence was normalized to the maximum fluorescence of NM2C or R788E in the absence of nucleotide.
Molecular dynamics simulations
Molecular dynamics based in silico site directed mutagenesis and simulations were performed using NAMD 2.9 and the CHARMM27 force field (48, 49). The X-ray crystal structure of the NM2C motor domain in the pre-powerstroke state, encompassing residues 49-807, served as the starting structure for NM2C and R788E simulations. Mutations were introduced in silico and the proteins were prepared and optimized prior to MD simulations using the Protein Preparation Wizard of the Schrödinger software suite (Schrödinger Suite 2012 Protein Preparation Wizard; Epik version 2.3; Impact version 5.8; Prime version 3.1; Maestro version 9.3. Schrödinger LLC, New York, NY, USA). The proteins were fully hydrated with explicit solvent using the TIP3P water model and charge neutralization was accomplished by adding Na+ counter ions (50). Short-range cutoffs of 12 Å were used for the treatment of non-bonded interactions; while longrange electrostatics was treated with the particle-mesh Ewald method (51). All simulations were carried out in an NpT ensemble at 310 K and 1 atm using Langevin dynamics and the Langevin piston method. A 1 fs time step was applied. Prior to production runs the solvated systems were subjected to an initial energy minimization and subsequent equilibration of the entire system for 5 to 10 ns. All MD simulations were carried out at the Computer Cluster of the Norddeutscher Verbund für Hoch- und Höchstleistungsrechnen.
Author Contributions
D.J.M. conceived the study. K.C. performed the crystallization, collected the diffraction data, and solved the structure. S.M.H. cloned, expressed and purified the proteins, and performed the biochemical and kinetic assays. M.P. performed and analyzed the molecular dynamics simulations. K.C., S.M.H., M.P., J.R.S., and D.J.M. designed experimental approaches and wrote the manuscript. All authors read and approved the final manuscript.
Competing Interests
The authors declare no conflicts of interest with the contents of this research article.
Acknowledgements
We thank the staff at the beamline BL14-1 at BESSY for technical support. We thank the Norddeutscher Verbund für Hoch- und Höchstleistungsrechnen (HLRN) for providing computational resources and the Biophysics Core of the National Heart, Lung, and Blood Institute (NHLBI) for advice, support and the use of the facility. Data deposition: The atomic coordinates and structure factors have been deposited in the Protein Data Bank, www.pdb.org (PDB entry 5I4E).