High-resolution helix orientation in actin-bound myosin determined with a bifunctional spin label

Significance

The interaction between actin and myosin is responsible for driving a vast array of essential biological processes, including the production of force in contracting muscle. However, the structural behavior of the two proteins in complex is not well understood, because high-resolution atomic models are not yet available by traditional methods. We use a bifunctional spin label and site-directed electron paramagnetic resonance spectroscopy to determine orientations of individual α-helices within the complex. We thus quantify for the first time, to our knowledge, structural changes within the motor domain of actin-bound myosin on nucleotide binding and dissociation. Our results provide valuable insight into the mechanism of muscle contraction while showcasing a method with wide applicability to other oriented biological systems.

Abstract

Using electron paramagnetic resonance (EPR) of a bifunctional spin label (BSL) bound stereospecifically to Dictyostelium myosin II, we determined with high resolution the orientation of individual structural elements in the catalytic domain while myosin is in complex with actin. BSL was attached to a pair of engineered cysteine side chains four residues apart on known α-helical segments, within a construct of the myosin catalytic domain that lacks other reactive cysteines. EPR spectra of BSL-myosin bound to actin in oriented muscle fibers showed sharp three-line spectra, indicating a well-defined orientation relative to the actin filament axis. Spectral analysis indicated that orientation of the spin label can be determined within <2.1° accuracy, and comparison with existing structural data in the absence of nucleotide indicates that helix orientation can also be determined with <4.2° accuracy. We used this approach to examine the crucial ADP release step in myosin’s catalytic cycle and detected reversible rotations of two helices in actin-bound myosin in response to ADP binding and dissociation. One of these rotations has not been observed in myosin-only crystal structures.

The myosin family of molecular motors is responsible for numerous vital functions in eukaryotes, including the contraction of striated muscle. Bundled within an intricate and highly regulated myofibril lattice, muscle myosin II converts the chemical energy released by ATP binding and hydrolysis into mechanical work, executing a series of structural transitions that generate force on actin and shorten each muscle cell (1, 2). Coupling of actin binding, nucleotide hydrolysis, and lever arm movement within myosin’s catalytic domain (CD) is essential for proper function of the contractile apparatus (3, 4).

Myosin function requires actin, and thus an understanding of its mechanism requires analysis of both proteins in complex. However, no crystals of actin–myosin complexes have been reported, so the resolution of actin-bound myosin structures is currently limited to that of electron microscopy. Furthermore, X-ray crystallography and electron microscopy produce only static structures in frozen or crystalline environments, which cannot accurately render the dynamics, disorder, and structural transitions that are essential to understanding function and pathology (4, 5).

In contrast, site-directed spectroscopy can be used to examine the actin–myosin complex under more physiological conditions. Both fluorescence and electron paramagnetic resonance (EPR) have been used in complement to examine the structural dynamics of myosin bound to actin (6, 7). EPR offers superior orientational resolution, due to the high sensitivity of the EPR spectrum to alignment of a spin label in the applied magnetic field. A well-placed spin label can provide direct information about orientation and dynamics in the vicinity of the labeling site, a strategy that has proven powerful in the study of myosin in oriented muscle fibers (7–9). However, conventional methods for site-directed spin labeling impose significant limits on the effective resolving power of EPR. Spin labels are typically incorporated into proteins through covalent attachment to Cys, resulting in a flexible linker that permits the label to undergo ns rotational motion independent of the peptide backbone. Such motion obscures the orientation dependence of the spectrum (10, 11).

Our solution is to eliminate local probe motions by using a spin label that becomes strongly and stereospecifically immobilized with respect to the target protein on attachment. In site-directed fluorescence, this has been achieved by using probes that react with di-Cys (12–14) or tetra-Cys (15) labeling sites, but these fluorescent probes are typically at least twice the size of spin labels, and fluorescence lacks the high orientational resolution of EPR (6, 7). The spin-labeled amino acid TOAC provides stereospecific attachment to the peptide backbone, but this probe is currently only practical for peptides on the order of 50 amino acids or less (16). For larger proteins, spin labels have been synthesized with bulky substituents to reduce mobility (17), or substitution with additional reactive moieties to confer bifunctionality (18–20). The smallest and simplest of these derivatives shares its basic structure with the widely used methanethiosulfonate spin label [1-oxyl-2,2,5,5-tetramethyl-∆3-pyrroline-3-methyl methanethiosulfonate spin label (MTSSL)], with a second MTS group that allows bifunctional targeting of two Cys residues (20) (Fig. 1A). This bifunctional spin label (BSL) is rigidly immobilized when reacted with a pair of Cys residues, so that it undergoes negligible ns rotational motion relative to the protein and can thus be used reliably to measure μs protein rotational motions by saturation transfer EPR (21–23).

Fig. 1.

(A) Chemical structure of BSL. (B) BSL bound stereospecifically to an α-helix at positions i and i+4, as in ref. 24. (C) Angles θ_NB and ϕ_NB that define the orientation of the nitroxide spin label (defined by axes x_N, y_N, z_N) relative to the applied magnetic field B, which directly determine the orientation dependence of the EPR spectrum. (D) Orienting the helically ordered muscle fiber (and thus the actin filament axis) with B permits direct measurement of the nitroxide orientation relative to actin.

Crystallography has shown that BSL exhibits a rigid and stereospecific linkage when reacted with Cys four residues apart on successive turns of an α-helix, with great potential for accurate spin-spin distance measurements (Fig. 1B) (24). However, the potential advantages of BSL for enhanced orientational resolution have not been explored. We hypothesize that if BSL is used in the context of an intrinsically oriented system (e.g., myosin in the myofilament lattice), the resolution of EPR will be sufficient to detect the orientation of individual protein structural elements with unprecedented accuracy (Fig. 1 C and D).

Results

Comparison of Spectra from Myosin Labeled with BSL and MTSSL.

A solvent-exposed location on the C-terminal end of myosin’s relay helix was chosen for initial study, because crystal structures and spectroscopic studies of isolated myosin have shown that the orientation of this helical segment is sensitive to nucleotide binding (25–28). MTSSL was reacted monofunctionally at position 492 (Fig. 2A, Upper), and BSL was reacted bifunctionally at positions 492 and 496 (Fig. 2 A, Lower, and B).

Fig. 2.

(A) MTSSL and BSL on the relay helix. (B) BSL labeling sites chosen on three stable helices throughout the myosin CD. (C–F) (Left) EPR spectra of the spin-labeled constructs on oriented fiber bundles, in the absence of nucleotide (rigor), with the fiber axis aligned parallel (red) and perpendicular (blue) to the magnetic field. (Right) EPR spectra of randomly-oriented (minced fiber) preparations of all spin-labeled constructs. (C) S1dC labeled at residue 492 with MTSSL (relay helix). (D) S1dC labeled at residues 492 and 496 with BSL (relay helix, bifunctional analog to C). (E) S1dC labeled at residues 325 and 329 with BSL (helix K). (F) S1dC labeled at residues 639 and 643 with BSL (helix W).

Skinned muscle fiber bundles, decorated with spin-labeled myosin S1dC in the absence of nucleotide (“rigor”), were oriented either parallel or perpendicular to the spectrometer’s applied magnetic field. Spectra from monofunctionally labeled (MTSSL) protein show virtually no dependence on fiber orientation (Fig. 2C, Left), indicating that the spin label ensemble is not well ordered in the myofilament lattice. In contrast, BSL spectra are highly sensitive to fiber orientation (Fig. 2D, Left), implying orientational order of the spin label relative to the myosin CD and of the myosin CD relative to actin.

Spectra of minced (randomly oriented) fibers show no dependence on sample orientation, as expected (red and blue spectra identical in Fig. 2 C and D, Right), but the spectra of monofunctionally (C) and bifunctionally (D) attached probes are quite different. Because these samples lack orientation, such differences can only arise from ns rotational motion. The BSL spectrum has a “powder” lineshape, with a wide splitting of 71.2 ± 0.2 G between the outer extrema, indicating no significant ns rotational motion (Fig. 2D, Right). Thus, BSL is strongly immobilized and highly ordered at this site, enabling it to directly report structural states in an oriented system. In contrast, the spectrum of monofunctional MTSSL bound to the same site shows substantial narrowing (Fig. 2C, Right), indicating large-amplitude dynamic disorder on the ns timescale.

BSL Is Strongly Immobilized and Ordered on Myosin at Sites Across the Myosin CD.

We generated two additional myosin constructs with bifunctional labeling sites on helix K (325C and 329C) and helix W (639C and 643C), located, respectively, in the upper and lower 50-kDa domains that form myosin’s actin-binding cleft (Fig. 2B). In all three cases, spectra of oriented fibers show very sharp lines that are quite sensitive to fiber orientation (Fig. 2 D–F, Left), consistent with highly ordered probes. Spectra of minced fibers all exhibit identical lineshapes between sample cell orientations (Fig. 2 D–F, Right) and show a very wide outer splitting, indicating strong immobilization on the ns timescale. Spectra of fibers oriented parallel to the field (Fig. 2 D–F, Left, red) are clearly different across all sites, indicating that probes at these sites have distinct orientations relative to the fiber (actin) axis.

Spin-Labeling with BSL Does Not Significantly Alter Myosin Function.

Actin-activated ATPase assays were performed on all BSL-labeled and unlabeled di-Cys myosin samples to assess the effect of labeling on protein function (Fig. S1 and Table S1). Extent of labeling on BSL-labeled samples was determined by spin-counting; spin-to-protein ratios approaching 1 were obtained for all three constructs (Table S1), demonstrating virtually complete bifunctional labeling. In addition, spectra of single- and di-Cys constructs share no significant features in oriented fibers, indicating the negligible presence of monofunctionally-attached BSL (Fig. 2 C and D and Fig. S2). For all three constructs, substantial actin activation of myosin ATPase was observed for both labeled and unlabeled samples (Fig. S1 and Table S1). There was significant variation among the three constructs, but the observed kinetics fall within the range previously observed for Dicty myosin S1dC (28, 29), and labeling affected neither V_max nor K_ATPase by more than a factor of 2. Thus, labeling does not alter the fundamental mechanism of myosin’s enzymatic activity.

Fig. S1.

Actin-activated ATPase data for bifunctional constructs. (Upper) Di-Cys myosin S1dC constructs with labeling sites highlighted and spin labels modeled in red. (Lower) Actin-activated ATPase data for each mutant construct, before labeling (red) and after labeling with BSL (blue). Errors are ±SEM (n = 3).

Table S1.

Actin-activated ATPase activity of Dicty S1dC constructs

Labeling site	Unlabeled		BSL labeled
	Vmax (s−1)	KATPase (µM)	Vmax (s−1)	KATPase (μM)	Spins/protein
492.496-BSL (relay helix)	2.32 ± 0.10	7.77 ± 0.75	4.05 ± 0.30	15.40 ± 1.60	0.97
325.329-BSL (helix HK)	5.33 ± 0.30	15.06 ± 1.19	3.85 ± 0.99	22.00 ± 7.06	0.87
639.643-BSL (helix HW)	3.75 ± 0.18	13.43 ± 1.51	2.05 ± 0.06	11.80 ± 0.78	0.96

V_max is the activity at saturating actin, and K_ATPase is the actin concentration needed for half-maximal activation (V = V_max/2). Errors are ±SEM (n = 3). Label-to-protein ratios were determined by spin counting.

Fig. S2.

Comparison of BSL spectra from single-Cys and di-Cys constructs. Spectra were acquired on parallel-oriented fiber bundles in the absence of nucleotide. Black, spectrum of BSL attached to a di-Cys S1dC construct at positions 639 and 643 (helix W); blue, spectrum of MTSSL attached to a single-Cys S1dC construct at position 639 (helix W); orange, spectrum of BSL attached to same construct as in blue. Absence of blue or orange spectral components from the black spectrum indicates complete bifunctional labeling on the di-Cys construct.

BSL Orientation Relative to Actin Is Determined by Direct Spectral Simulation and Fitting.

The results in Fig. 2 show that BSL undergoes negligible ns rotational motion, so the shape of the spectrum is dictated entirely by the orientation of the nitroxide spin label (N) relative to the instrument’s static magnetic field (B), defined by θ_NB and ϕ_NB (Fig. 1C). Thus, when spectra are acquired with the muscle fiber (actin filament) axis parallel to the magnetic field (Fig. 1D), spectra directly report the orientational distribution of the nitroxide spin label relative to actin (8, 9, 30), defined by θ_NA and ϕ_NA, which are identical to θ_NB and ϕ_NB (Fig. 1C)_. Therefore, we focus on spectra acquired using the parallel fiber alignment.

We determined the orientational distribution in each sample through least-squares fitting of EPR data to spectral simulations (Fig. S3), as described previously (8, 9, 30). First, magnetic and hyperfine tensors were obtained by fitting spectra acquired from minced fibers (Fig. 2 D–F, Right), assuming immobile and randomly oriented spin labels. After fixing these values, analysis of oriented samples depends exclusively on θ_NA and ϕ_NA (Fig. 1). Each spectrum was assumed to arise from a distribution of static probe orientations relative to the applied magnetic field B (here equivalent to the actin filament axis A). Spectra were fitted allowing for the presence of up to two independent populations, each characterized by central angles θ_NA and ϕ_NA, and a Gaussian full width at half maximum (Δθ,ϕ) describing static disorder of the ensemble. For each of the three BSL samples, one principal oriented Gaussian component was observed (Fig. S3), and that component is displayed in Fig. 3B, with parameters given in Table 1.

Fig. S3.

EPR spectra of fibers oriented parallel to the applied magnetic field. (Left) Data (black), best fit simulated spectrum (red) and residual (purple). (Center and Right) Angular distributions associated with each simulated spectrum, showing angular component 1 (cyan), angular component 2 (violet), and envelope (black). EPR is subject to some intrinsic ambiguity in the determination of angular centers, because the orientation dependence in EPR depends on squared cosines of angles. For simplicity, a single distribution on the interval [0,90] is reported here for each parameter. In all cases in this study, only one plausible helix orientation was found. For each construct, we observed two well-resolved spectral components, corresponding to a major, highly ordered spin label population, and a minor, more weakly ordered population. We hypothesize that the second component arises predominantly either from unbound S1dC trapped in the myofilament lattice during fiber decoration, or (in the case of the site on helix W, where the minor component is relatively well ordered) from a small population of alternate BSL conformers stabilized by the structural environment of the specific labeling site. Thus, in the main text, only the primary component is considered.

Fig. 3.

(A) BSL-labeled myosin bound to actin in skinned muscle fibers in the absence of nucleotide (black) and the presence of 5 mM MgADP (green). (B) θ_NA distributions for nucleotide-free (black) and ADP-bound (green) biochemical states, obtained by spectral simulation and least-squares fitting of the data in A.

Table 1.

Orientation of the myosin-bound nitroxide spin label (N) relative to the actin (fiber) axis (A), derived from spectra of muscle fibers oriented parallel to the magnetic field (Fig. 3)

Labeling site	Actin-BSL axial angle, θNA (°)		Actin-BSL azimuthal angle, ϕNA (°)		Angular width FWHM, Δθ,ϕ (°)
	Apo (rigor)	5 mM MgADP	Apo (rigor)	5 mM MgADP	Apo (rigor)	5 mM MgADP
492.496-BSL (relay helix)	70.5 ± 0.1	60.9 ± 0.2	27.0 ± 1.8	13.6 ± 2.4	11.9 ± 1.5	18.1 ± 0.7
325.329-BSL (helix K, U50)	88.2 ± 0.5	89.2 ± 0.6	33.3 ± 1.8	35.3 ± 1.0	10.0 ± 1.7	9.2 ± 0.2
639.643-BSL (helix W, L50)	28.0 ± 0.1	25.0 ± 0.3	32.6 ± 2.1	22.6 ± 1.8	10.4 ± 0.2	9.9 ± 0.7

Mean ± SEM (n = 3).

BSL Resolves Significant Structural Changes in the Force-Generating Domain of Myosin During ADP Binding and Release.

Next, we asked whether this approach is capable of resolving transitions between distinct nucleotide-dependent structural states. We focused on identifying differences between nucleotide-free and MgADP-bound complexes, because of the physiological relevance of the ADP-release transition (Discussion).

We acquired spectra from oriented fibers in the absence (black) and presence (green) of 5 mM MgADP (Fig. 3A) by switching the source of peristaltic flow through the fiber capillary. After acquisition in the presence of MgADP, flow was switched back to a nucleotide-free solution containing 10 mM EDTA to dissociate the residual bound ADP. In all cases, this spectrum was identical to that of the original nucleotide-free sample.

In the relay helix construct, a marked increase in the outer splitting was observed with MgADP (Fig. 3A, Top). Subsequent fitting revealed that this corresponds to a change in θ_NA of −9.6 ± 0.3°, a change in ϕ_NA of −13.4 ± 4.2°, and a change in disorder (Δθ,ϕ) of +6.2 ± 2.2° (full width half maximum) (Fig. 3B and Table 1). In the helix K construct (upper 50-kDa domain), no significant changes were detected (Fig. 3, Middle, and Table 1). In the helix W construct (lower 50-kDa domain), a significant increase in outer splitting was observed, corresponding to a change in θ_NA of −3.0 ± 0.4° and a change in ϕ_NA of −10.0 ± 3.9°; there was no significant change in disorder (Fig. 3, Bottom, and Table 1). The different effects observed in these three helices indicate ADP-induced structural changes internal to the CD rather than a concerted rotation of the entire domain relative to actin.

Helix Orientation with Respect to Actin.

Angles in Fig. 3 and Table 1 correspond to θ_NA and ϕ_NA, defined by the relative orientation of BSL and the actin filament (Fig. 4 A and B). To develop structural constraints that are independent of the spin label, we considered the spatial relationship between the label and its associated α-helix. BSL is highly ordered at all sites tested, indicating that the label is stereospecific in its attachment, so our data support the use of a single model to define the label–helix relationship. We started with a crystal structure of BSL attached to T4 lysozyme at positions i and i + 4 on an α-helix [Protein Data Bank (PDB) ID code 3L2X] (24). We determined the axis of the labeled helix using a previously described technique (31) and defined a nitroxide reference frame using the geometry of BSL (Fig. 4C). Vector algebra yields two new angles, θ_NH and ϕ_NH, which describe the orientation of the helix vector in the nitroxide frame (Fig. 4D). With both helix and actin vectors defined in the nitroxide reference frame, the angle between them (θ_AH) describes the tilt of the myosin helix with respect to the actin filament (Fig. 4 E and F).

Fig. 4.

Visualization of coordinate transformations. (A) Actin (yellow) in complex with myosin S1 (blue) labeled with BSL (red). The actin long axis (gold, same as the magnetic field axis B in Fig. 1C, because fibers are aligned parallel to B) has a well-defined orientation within BSL’s nitroxide coordinate frame (white), defined by angles θ_NA and ϕ_NA (B). (C) BSL bound to a helix, with the nitroxide frame (gray) and helix axis (red) highlighted, defining angles θ_NH and ϕ_NH that describe the helix orientation relative to BSL (D). (E) Actin (yellow) and myosin (blue) in complex, as in A, showing the two vectors of interest, actin (gold) and a representative helix axis (red), defining angle θ_AH between them (F).

To verify our derivation, we compared the helix orientations calculated from our nucleotide-free EPR results to a recent model of the actomyosin complex, assembled from cryo-EM and crystallographic data on nucleotide-free myosin II in complex with rabbit skeletal actin (32). In this model, θ_AH values were measured for the three labeled helices relative to the actin axis, and these values were compared with the centers of our experimental θ_AH distributions. Table 2 shows that each EPR experiment produces a θ_AH value that gives good agreement with the model calculation, with a difference of 12.7 ± 1.9° in the most extreme case.

Table 2.

Axial tilt angles of labeled helices with respect to actin in the absence of nucleotide

Labeling site	Model-based θAH (°)	Initial (θNH = 74.2°, ϕNH = −73.8°)		Optimized (θNH = 74.9°, ϕNH = −61.2°)
		Experimental θAH (°)	Difference (°)	Experimental θAH (°)	Difference (°)
Relay helix (apo)	84.4	74.8 ± 1.8	−9.5 ± 1.8	83.4 ± 1.8	−1.0 ± 1.8
Helix K (U50)	29.6	42.3 ± 1.9	12.7 ± 1.9	30.6 ± 1.9	0.9 ± 1.9
Helix W (L50)	77.5	83.5 ± 2.1	6.0 ± 2.1	78.4 ± 2.1	1.0 ± 2.1

Values are derived using probe orientations obtained either from EPR (Experimental columns) or directly from a cryo-EM model of actomyosin (Model-based column) (32).

Although all labeling sites on our myosin constructs were chosen within straight helices, the T4 lysozyme helix used in our derivation is kinked at the labeling site. It is likely that BSL’s conformation on straight helices is different from that observed in this crystal, and we hypothesized that this discrepancy could account for some of the observed disagreement in θ_AH values. We varied the angles θ_NH and ϕ_NH, assuming the same values at all three sites, to minimize the difference between experimental and model-based results, yielding the optimized values given in Table 2. It is striking to observe that a small change in θ_NH and ϕ_NH is sufficient to bring all three of the experimental θ_AH angles to within 1° of their model-based predictions. This result demonstrates that we can obtain structural constraints for the actomyosin complex using our EPR-based method that are in precise agreement with previous literature while avoiding the inherent difficulties and caveats associated with frozen and crystalline samples. The agreement across all three labeling sites, when a single conformation of BSL is assumed, provides further evidence that BSL is indeed highly stereospecific in its attachment to α-helices.

Effects of ADP on Helix Orientation.

For the relay helix site, our results are consistent with crystal structures and previous spectroscopic studies on actin-free myosin, which show evidence of a nucleotide-dependent deformation in the C-terminal end of the relay helix (25, 33–35) (Fig. 5A). In each of these structures, a kink in the helix is introduced at residue M486. To model the changes in myosin reflected by EPR data, we assume that the observed reorientation of BSL corresponds to movement at the end of the relay helix, comprising residues 486–496. Analysis of MgADP spectra using our optimized θ_NH and ϕ_NH values (Table 2) yields a set of four potential relay helix orientations (θ_AH) in the presence of MgADP. Comparison with an alignment of myosin-only crystal structures highlights the most likely solution: a θ_AH value of 71.0 ± 2.4° (Fig. 5B, green) places the helix in good alignment with crystal structures of myosin with nucleotide and reflects a conformation highly similar to the MgADP-bound structures observed in crystallo. The other solutions bend the relay helix in a direction directly opposite from the crystal structures or deform the helix so severely as to disrupt its secondary structure. Thus, we conclude that the change observed by EPR correlates with a reorientation of the relay helix from 83.4 ± 1.8° to 71.0 ± 2.4° relative to the actin filament (Δθ_AH = −12.4 ± 3.9°; Fig. 5 B and C). The helix also becomes more disordered with the addition of MgADP (ΔΔθ_AH = +6.2 ± 2.2°; Fig. 5C).

Fig. 5.

(A) Structural alignment of three Dicty myosin relay helix crystal structures, showing the nucleotide-induced C-terminal bend. (B) Relay helix from our actomyosin model (32), showing the EPR-derived change induced by ADP (black to green), with the ADP.P_i structure from A shown for reference. (C) Actin (yellow) in complex with myosin (blue), with EPR-derived mean relay helix orientations highlighted. (D and E) EPR-derived amplitudes of angular disorder.

In contrast to the relay helix, the orientation of helix W is virtually invariant in myosin-only crystal structures. Thus, it is not clear whether our observed change corresponds to an internal bending of the helix, or to a reorientation of the entire helix. However, repeating the analysis discussed above and assuming that the helix behaves as a rigid rod, the best solution gives a change in θ_AH of −4.7° ± 1.1°, pulling helix W away from the bent relay helix.

Discussion

BSL Greatly Enhances the Resolution of EPR Within Oriented Systems.

Bifunctional derivatives of nitroxide spin labels simplify EPR spectra by reducing ns probe motion and disorder. Here we showed the power of this technology in intrinsically oriented systems, granting EPR the capacity to measure absolute orientations of individual protein structural elements.

Previously, it was possible to detect changes in protein structure by observing shifts in orientation and mobility of spin labels at well-chosen sites. Although informative, these results were typically unsuitable for accurate structural interpretation because of the weak coupling between probe and backbone (36). Measuring orientations precisely required the use of synthetically incorporated groups such as the amino acid TOAC, which are difficult to use in proteins too large to synthesize (7). The alternative presented here allows for the introduction of rigidly attached probes after expression, enabling high orientational resolution for structural elements within much larger targets.

We demonstrated that BSL consistently becomes strongly immobilized and well oriented on α-helices in myosin across multiple sites throughout the CD (Fig. 2). The result is a single, dominant, narrow angular distribution in oriented samples, implying a high degree of stereospecificity at each site (Fig. 3 and Table 1). Our derivations of helix-to-actin angles reinforce this observation and provide strong evidence that BSL binds all three target helices with a well-defined preferred conformation (Fig. 4 and Table 2).

ADP Induces Internal Structural Changes Within the CD in Actin-Bound Myosin.

The physiological significance of the myosin ADP release step has become increasingly apparent. Actin decreases myosin’s ADP affinity by a factor of roughly 100 (37, 38), implying crucial allosteric coupling between the nucleotide-binding pocket (ATPase) and the actin-binding interface (cross-bridge dynamics). Strain on the lever arm affects the rate of ADP release in several myosins (39–42), and single-molecule studies have correlated this rate with lever arm movement (39, 43). Thus, ADP release is directly related to a mechanical event that either represents the completion of the powerstroke, or functions as a load-dependent tuning mechanism (42). Electron microscopy of smooth muscle myosin bound to actin shows a significant effect of ADP on the orientation of the lever arm relative to the catalytic domain (44). Thus, a large body of evidence suggests an important structural change within the CD of myosin that affects both actin binding and lever arm position upon ADP release, but a direct measurement of this internal structural change has not previously been provided for actin-bound myosin.

We used BSL to probe specific sites on myosin while the protein is in complex with actin. We observe a reversible change in label orientation at the relay helix site on addition and removal of MgADP (Fig. 3 and Table 1). Deriving the orientation of the labeled helix from EPR data (Fig. 4), we calculate a change of −12.4 ± 3.9° in the orientation of the relay helix relative to actin (Fig. 5). These results align well with existing crystal structures of myosin alone, but have not previously been observed within the actomyosin complex. We also detect a change of −4.7 ± 1.1° in the orientation of helix W relative to actin. This change is not observed in myosin-only crystal structures and thus serves as a rare example of actin directly modulating myosin’s structural dynamics. Further characterization of this behavior with additional labeling sites may help to refine our understanding of the mechanism behind actin activation of myosin ATPase.

Additional Labeling Sites Would Provide More Detailed Structural Information.

These results demonstrate BSL’s potential for dramatically improving the accuracy of structure determination in an oriented assembly. Although BSL’s high EPR resolution permits the precise determination of helix axial tilt angles in myosin with respect to actin (θ_AH), the measurements taken here are not sensitive to the orientation of the helix in the plane normal to the actin filament. This missing information means that a single labeling site on a helix cannot completely define how that helix is oriented within each myosin head, in a way that would allow direct modeling de novo. Such ambiguity should be alleviated by combining this approach with distance measurements between pairs of sites, detected by dipolar spin–spin interaction (24, 45). Such an approach would be essentially equivalent to the method used in structure determination by NMR, where both orientation and distance constraints are combined (46). The superior sensitivity of EPR makes it applicable to rapid measurements in dynamic samples, such as contracting muscle fibers; such measurements are needed to resolve crucial questions about the structural transitions that generate contractile force. The present work introduces a powerful application of EPR to intrinsically oriented biological systems in general, allowing for the derivation of accurate structural constraints not only for sarcomeric proteins, but also for proteins associated with nucleic acid chains, microtubules, and lipid bilayers.

Materials and Methods

Protein and Muscle Fiber Preparations.

Mutant myosin constructs were prepared in a Cys-lite S1dC Dictyostelium myosin II background truncated at residue 758 and containing only one native (nonreactive) Cys at position 655 (47). Constructs were expressed and purified from Dictyostelium orf+ cells. Skinned rabbit psoas muscle fiber bundles were dissected and permeabilized as described previously (48).

EPR Spectroscopy.

EPR spectra were recorded at X-band (9.6 GHz) using an E500 EleXsys spectrometer (Bruker Instruments). Acquisition of spectra for oriented samples was performed as described previously (21, 22). BSL spectra were analyzed to determine the orientational distribution of the nitroxide coordinate frame with respect to the applied magnetic field for parallel-oriented samples, using computational simulation and least-squares minimization as described previously (8, 30).

A detailed description of methods and reagents can be found in SI Materials and Methods.

SI Materials and Methods

Protein and Muscle Fiber Preparations.

Mutant myosin constructs were prepared in a Cys-lite S1dC Dictyostelium myosin II background truncated at residue 758 and containing only one native (nonreactive) Cys at position 655 (47), using a QuikChange II XL site-directed mutagenesis kit (Invitrogen). One construct was engineered for labeling with MTSSL, containing a single Cys substitution at position 492 on the C-terminal end of the relay helix. One additional construct presented here was engineered with a single Cys substitution at position 639. Additionally, three constructs were engineered for labeling with BSL, each containing a di-Cys motif wherein residues on successive turns of an α-helix (positions i and i + 4) were mutated to Cys. Helices targeted for mutagenesis in the three constructs were helix K in the upper 50-kDa domain (positions 325 and 329), helix W in the lower 50-kDa domain (positions 639 and 643), and the C-terminal end of the relay helix (positions 492 and 496). Following mutagenesis, constructs were expressed and purified from Dictyostelium orf+ cells. F-actin was prepared from rabbit skeletal muscle as described previously (49, 50). Actin-activated ATPase activity of myosin constructs, labeled and unlabeled, was measured using an NADH-coupled assay (51).

Skinned rabbit psoas muscle fiber bundles, dissected and permeabilized as described previously (48), were further dissected down to a diameter of 0.5 mm and a length of ∼3 cm, drawn inside a 25-μL glass capillary (Drummond), and secured at both ends with silk thread.

Site-Directed Spin Labeling of Myosin.

Single-Cys or di-Cys myosin mutants in labeling buffer (30 mM Tris, 50 mM KCl, and 3 mM MgCl₂, pH 7.5) were incubated at 4 °C with 5 mM DTT for 1 h to ensure reduction of engineered Cys residues before labeling. DTT was removed using Zeba Spin desalting columns (Thermo Scientific), and myosin was incubated at 4 °C for 1 h in twofold molar excess of either the bifunctional BSL, or the monofunctional MTSSL. Following incubation, excess spin label was removed, and the protein was exchanged into EPR (rigor) buffer (40 mM Hepes, 1 mM EGTA, 2 mM MgCl₂, and 15 mM KPr, pH 7.0) using Zeba Spin desalting columns.

EPR Spectroscopy.

Spectra were recorded at X-band (9.6 GHz) using an E500 EleXsys spectrometer (Bruker Instruments). Acquisition of spectra for oriented samples was performed essentially as described previously (21, 22). Briefly, skinned rabbit psoas muscle fiber bundles were incubated at 4 °C with 80 μM spin-labeled myosin for 1 h inside glass capillaries (described above) to decorate free actin filaments in the fiber bundle with spin-labeled S1dC. Spectra were acquired with the long axis of the fiber bundle capillary aligned parallel (using a modified TM₁₁₀ cavity) or perpendicular (using a TE₁₀₂ cavity) to the instrument’s magnetic field. Samples were kept at 4 °C throughout acquisition by flowing temperature-controlled nitrogen gas through a quartz dewar fixed to the bottom of the cavity (TM₁₁₀) or through the cavity’s anterior optical port (TE₁₀₂). The fiber bundle was perfused with buffer under a constant flow of 340 μL/min, to eliminate unbound myosin and unbound spin label and to allow for changing the biochemical environment of the bundle during the experiment. Three spectra were acquired in both cavities for each sample, in the presence of EPR (rigor) buffer, ADP buffer (40 mM Hepes, 1 mM EGTA, 2 mM MgCl₂, and 5 mM MgADP, pH 7.0), and EDTA buffer (40 mM Hepes, 1 mM EGTA, 2 mM MgCl₂, and 10 mM EDTA, pH 7.0). Following acquisition of spectra from an oriented fiber sample, the spectrum from randomly oriented fibers was recorded after removing the fiber bundle from its capillary, mincing it, and placing it in a TE₁₀₂-compatible flat cell.

BSL spectra from oriented fibers were analyzed to determine the orientational distribution of the applied magnetic field in BSL’s nitroxide coordinate frame, using computational simulation and least-squares minimization as described previously (8, 30). Briefly, minced fiber spectra were analyzed to determine the orientation-independent parameters (g and T tensors). These parameters were then fixed and applied to a simulation model that assumes a sum of Gaussian distributions of static (no nanosecond rotational diffusion) spin label orientations, each defined by a center (θ₀, ϕ₀) and a width (Δθ,ϕ, full width at half maximum). This static model is justified by previous observations of BSL attached to myosin (21–23) and other proteins (24) and by the spectra of randomly oriented samples of actin-bound myosin (Fig. 2).

EPR on S1dC in solution (for spin counts) was carried out using 80 μM spin-labeled protein in EPR buffer. Quartz capillaries containing S1dC (0.6 mm ID, 20-μL sample volume) were sealed with critoseal and placed in an SHQ cavity (ER4122 ST). Samples were kept at 4 °C throughout acquisition using a quartz dewar as described above. Quantification of the extent of spin labeling (spin labels bound per S1) was carried out as previously described (22).

Molecular Modeling.

Atomic coordinates of Dicty myosin bound to rabbit skeletal actin were derived from previously published cryo-EM data by Kenneth Holmes, Max Planck Institute for Medical Research, Heidelberg (32). The atomic coordinates for BSL bound to an α-helix were obtained from PDB ID code 3L2X (24). Angle calculations from atomic structures were calculated using custom extensions written for Visual Molecular Dynamics 1.9 (VMD) (52). Structure images were rendered using the PyMOL Molecular Graphics System, Version 1.3 (Schrödinger) and Blender, Version 2.72b (Blender Foundation).