High-resolution structures of malaria parasite actomyosin and actin filaments

We reconstituted the malaria parasite actomyosin filament in the rigor state in vitro using recombinantly expressed Act1 and MyoA:ELC:MTIP complex in the presence of a small cyclic peptide, jasplakinolide, which stabilizes actin filaments and has been used for cryo-EM studies on Act1 filaments previously [13,15,23]. After plunge-freezing and cryo-EM imaging, we observed three types of filaments: bare Act1 filaments and filaments either partially or fully decorated by the MyoA:ELC:MTIP complex. By picking well-decorated filaments from the micrographs (S1 Fig), we determined the malaria parasite actomyosin structure using helical reconstruction methods [31] to an overall resolution of 3.1 Å, as estimated by Fourier shell correlation with the 0.143 threshold criterion (Figs 1 and S1 and Table 1).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. High-resolution reconstructions of filamentous Act1 (left) and MyoA-decorated Act1 (right).

In the density maps, the central individual actin protomers and myosins are depicted in different colors. Jasplakinolide (JAS) and ADP are shown in red and yellow, respectively. The top middle part shows density maps around individual protein molecules and small molecule ligands as labeled. The bottom row shows density maps of selected structural motifs, as labeled, within the interfaces in the complexes. Additionally, density around phosphorylated Ser19 (S19-P) is shown.

https://doi.org/10.1371/journal.ppat.1010408.g001

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Data collection and refinement statistics.

https://doi.org/10.1371/journal.ppat.1010408.t001

For direct comparison, we also determined the structure of the non-decorated Act1 filament at an average resolution of 2.6 Å (Figs 1 and S2 and Table 1). Visual inspection of the density maps of both complexes agrees with the global resolution estimations (Figs 1 and S3–S5 and S1–S3 Movies). Based on the maps, we built atomic models for an actomyosin filament consisting of four Act1 molecules and two MyoA molecules and for the non-decorated Act1 filament consisting of five actin subunits (S5 Fig).

As typical for actomyosin cryo-EM structures [23,32–35], the resolution of the Act1:MyoA:ELC:MTIP structure decays radially away from the helical axis, having a well-resolved core, consisting of the actin filament and the myosin motor domain, whereas density for the MyoA neck region is weaker, but nevertheless allowed us to locate density for the light chains, ELC and MTIP (Fig 2). The actin core in both of our structures is particularly well-resolved compared to other actin filament structures (S3 and S4 Figs). Thus, the molecular details of F-actin, including the small molecules ADP and jasplakinolide as well as the active site metal and water molecules, can be visualized in greater detail (Fig 1) than in myosin-decorated or undecorated actin filament structures [13,32–41]. The majority of the MyoA motor domain is well-ordered, and side chains can be placed into density, especially within the actomyosin interface, allowing analysis of specific contacts between MyoA and Act1 in detail.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. A model of the light chains ELC and MTIP bound to MyoA.

(A) An unsharpened map (gray) filtered with LAFTER shows the location of the MyoA motor (blue) and the light chains ELC (magenta) and MTIP (cyan). (B) The Act1:MyoA:ELC:MTIP model in the rigor state shown in the context of the membrane-delimited sub-pellicular compartment. An outline of GAC (dotted lines) in two randomly chosen orientations with approximate dimensions derived from the T. gondii GAC small-angle X-ray scattering (SAXS) low-resolution envelope [16]. (C) Results of single molecule optical tweezer assays are presented. (Upper) A displacement histogram for MyoA:ELC:MTIP at 2.5 μM ATP. Power stroke was determined as the shift in the Gaussian peak from 0 nm (gray line), which has an average ∼4.5 nm. (middle) An attached lifetime histogram for muscle actin-MyoA:ELC:MTIP events measured at 2.5 μM ATP. The detachment rate (grey line) was determined by fitting a single-exponential to the histogram. The event lifetime distribution (middle panel) fitted well to a mono-exponential decay (least squares fit gives a rate constant 4.1 s^-1) consistent with detachment being controlled by ATP binding to the rigor complex (at 2.5 uM ATP was 1.4x10⁶ M^-1 s^-1). (Lower) The panel shows the ensemble average of 598 binding events. Each event was synchronized to its start [arrow] and end point [arrow] defined by the change in system stiffness, determined from the amplitude of a 200 Hz sinusoidal forcing function applied by the optical tweezers. The synchronized data were averaged. The average displacement at the start of the event (~4 nm) is smaller than that at the end of the event (~5.5 nm). This observation implies the power stroke is generated in two phases, ~4 nm followed by a further motion of ~1.5 nm.

https://doi.org/10.1371/journal.ppat.1010408.g002

ELC and MTIP bind to the MyoA neck in a compact manner, presumably stiffening the neck region so it acts as a relatively rigid lever arm (Fig 2), consistent with the requirement of both light chains for maximal actin gliding speed as measured in in vitro motility assays [10,11]. To confirm that our recombinant MyoA:ELC:MTIP complex was fully-functional we measured its ATP-driven power stroke, using single-molecule optical trapping nanometry [42] (see Materials and Methods). The average power-stroke, determined by histogramming the amplitude of many individual actomyosin binding events, was 4.5 +/- 0.4 nm s.e.m. (1535 events from 8 replicates, Fig 2C, upper). The distribution of event lifetimes was monotonic and controlled by the rate of ATP binding to the rigor complex (Fig 2C, middle). Ensemble-averaging of events [43] revealed the power stroke was biphasic, comprising a rapid initial movement of ~4 nm followed by a subsequent movement of ~1.5 nm (Fig 2C, lower). Previous studies, indicate the biphasic nature of the power stroke arises from structural changes associated first with phosphate and later with ADP release [44].

A consequence of the extended and rigid neck region is a space constraint for the actomyosin complex in the sub-pellicular compartment. The MyoA:ELC:MTIP complex would occupy approximately half the distance between the IMC and the plasma membrane (~30 nm; Fig 2B), which is similar to the spacing between thick and thin filaments in vertebrate muscle sarcomeres [45]. In the parasite, when the additional space taken up by the actin filament (diameter ~6.5 nm) and GAC (D_max in solution ~25 nm; [16]) is considered, it is clear that there must be significant geometric and steric constraints on how the actomyosin complex might be arranged (Fig 2A and 2B). It is possible that the attachment of MyoA via its MTIP light chain to the GAP45:GAP50 complex in the IMC might allow azimuthal movement so the motor domain might approach the actin helix from different angles. Also the structure of GAC may be different in the tightly packed pellicle than in solution. These structural considerations will need to be revisited when the 3D geometry of the full glideosome complex is understood.

The MyoA motor domain has a conserved overall structure with several known subdomains common to functional myosin motors (S6 Fig). The upper and lower 50 kDa subdomains are separated by the actin-binding cleft, which is in a closed conformation in our complex (Fig 3). The central transducer region consists of seven β strands (S6 Fig). The lower 50 kDa subdomain contains a helix-turn-helix (HTH) motif, involved in intimate contacts with Act1, while the rest of the actin-contacting residues are in the hypertrophic cardiomyopathy (HCM) loop and in loops 2, 3, and 4 (Fig 3). The more distal parts, the converter and the relay helix, are involved directly in the conformational change upon the power stroke (Fig 2, [46]). Additionally, class XIV myosins have an N-terminal SH3 domain, the role of which is unclear but could be related to the regulation of myosin function during different stages of the parasite life cycle [47]. Our actomyosin rigor structure shows that neither the SH3 domain nor the N-terminal helix is in direct contact with the actin filament and the closest segments are about 18 Å away (Fig 3).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Molecular interactions of MyoA with Act1.

(A) MyoA is shown as cartoon and Act1 as surface representation. The two adjacent actin protomers are colored with distinct colors. The MyoA motifs contributing to the interface as well as MyoA subdomains, and the N-terminal helix and SH3 domain are highlighted and labeled. (B) The HCM-loop lies on one actin protomer, contacting actin SD1. The basic loop 2 (Lys632 and Lys634) interacts with the acidic N-terminus of Act1. (C) The HTH motif in the lower 50K domain contributes most specific hydrogen bonds in the interface, interacting with the D-loop (SD2) and the C-terminal patch of Act1 between SD1 and SD3 in close vicinity of the D-loop in the adjacent actin subunit. (D) Loop 3 and the HTH motif visualized approximately 190° rotated with respect to panel A. Loop 3 (576–577) interacts with the adjacent actin monomer on SD1. The activation loop in the lower 50K domain protrudes towards the next SD1, with Lys534 interacting with the actin N-terminus.

https://doi.org/10.1371/journal.ppat.1010408.g003

Overall, the MyoA structure in complex with actin is very similar (with an overall root mean square deviation (r.m.s.d.) for the C_α positions ~1.3 Å) to the rigor-like conformer found as one of the molecules in the asymmetric unit in the recently determined MyoA crystal structure [20]. However, there are several small but important changes in the complex. Notable differences allow interactions and avoid clashes with actin, including movement of the HCM loop (Ile409-Arg422) by nearly 4 Å, slightly smaller shifts in loop 4 (Glu369-Ala380), loop 3 (Pro572-Phe581), and the HTH motif (Val538-Asn563) as well as the ordering of loop 2 (Lys632-Ser639), which is not visible in the truncated MyoA crystal structure [20], but can be seen in the full-length complex with light chains [21]. In addition, there are several smaller rearrangements in more distal areas, in particular in the converter domain. The N-terminal helix, which is specific to apicomplexan MyoAs and accommodates the phosphorylated Ser19, is stacked between the SH3 domain and the motor domain, suggesting an important structural or functional role (Figs 3 and S6; [20]). The phosphorylation of Ser19 in the MyoA:MTIP:ELC complex is visible in the density map (Fig 1) and was confirmed using mass spectrometry, showing ~50% phosphorylation of the relevant peptide. In the crystal structure of T. gondii MyoA, the entire N-terminal helical extension is not visible in the electron density map [19] but is seen in most MyoA crystal structures [20,21] and MyoA:Act1 complex [23]. The MyoA:Act1 complex described by Robert-Paganin et al. (2021) has slightly different helical parameters, half a degree in twist and 1 Å smaller rise. The largest differences are in the HCM loop and at the tip of HLH with overall r.m.s.d. ~1.4 Å.

Naked Act1 and myosin decorated Act1 filaments are nearly identical

We reconstructed the Act1 filament alone and obtained helical parameters that were similar to our Act1:MyoA complex (Table 1) with a similar rise and a 0.7° smaller azimuthal rotation for the complex. As previously noted for other actomyosin complexes [23,32,33,35], neither MyoA nor Act1 undergo large conformational changes when compared to the structures in isolation, and actin subunits superpose well over 372 residues (RMDS 0.34 Å). However, our highly-resolved density of the Act1 filament allows us to see most of the side chain orientations and details of ligand binding for ADP•Mg²⁺ and jasplakinolide, as well as putative water molecules, including several in the vicinity of the active site (Figs 1 and 5 and S4 Movie). Active sites of filamentous Act1 or Act1:MyoA filament models are nearly identical, showing presence of ADP and its coordinated magnesium ion.

The Act1 filament has an internal cavity, which is extended towards the filament exterior (Fig 4A and S4 Movie). We modelled density in the cavity as a chain of water molecules (Fig 5), but we cannot exclude presence of other solute molecules. The end of the cavity is formed by the side chains of Asn116 and Trp80 (Fig 4 and S4 Movie). In monomeric Act1, the cavity is much shorter compared to the F-form (Fig 4C). In muscle F-actin, the cavity is limited by the bulkier Ile76 instead of Val77 in Act1 (Fig 4). Notably, residue 76/77 is an Ile in α- and γ-actins, but is also Val in β-actin as well as plant and yeast actins, which release phosphate faster than muscle actin [48,49]. We have previously noted the importance of Gly115 substitution to alanine in Act1; alanine at this position slows down phosphate release compared to glycine, in particular in the Mg-bound form [15,28]. In skeletal or smooth muscle α-actin, Asn115 and Arg116 are sites of disease-associated mutations (N115S, N115T, R116Q, R116H) [50,51]. In yeast, mutations N115T and R116Q affect both nucleotide exchange as well as polymerization kinetics, in particular in the nucleation phase [52]. Therefore, these features of Act1 active site may be relevant to understand actins generally.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. Internal cavities in filamentous and monomeric actin structures.

The internal cavities were calculated using the CASTp server [74] and visualized as transparent surfaces using Chimera [66]. (left) The internal cavity in the Act1 filament and (middle) the corresponding channel in monomeric Act1 (4cbu; [15]). (right) The corresponding cavity in muscle actin (6djm; [38]) is shorter due to change from Ile76 to Val77 in Act1.

https://doi.org/10.1371/journal.ppat.1010408.g004

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 5. Details of the nucleotide binding site in the Act1 filament.

(left) Density map around the nucleotide-binding site, including ADP (sticks), Mg²⁺ (green sphere) with coordinating water molecules and putative water molecules (red spheres) in the internal cavity. (right) Schematic presentation of ADP interactions in the filamentous Act1 active site.

https://doi.org/10.1371/journal.ppat.1010408.g005

Based on crystal structures of monomeric actin, the side chain of Ser15 (14 in canonical actins) has been established as a nucleotide sensor, which is turned away from bound ATP but hydrogen bonded to the β-phosphate of ADP [53]. In our structure of the ADP-bound filamentous Act1, the side chain of Ser15 points away from the nucleotide, not contacting the ADP phosphates, thus more closely resembling the conformation in crystal structures of the ATP-bound state (Fig 5). Instead, the Ser15 main chain NH group interacts with the β-phosphate.

Actomyosin interface has conserved features

The MyoA domain architecture is conserved, as in other myosins, regardless of the myosin class (Figs 2 and 3), but residues in the actin-binding interface are generally not conserved (S6 and S7 Figs). The Act1 surface is relatively flat, and thus MyoA lacks large protrusions, which would penetrate deep into the filament pockets (Fig 3A). The interface is mainly formed by contacts between the HTH motif, the HCM loop, and loops 2, 3, and 4 (Fig 3) similarly to previous structures [23,34,35,46]. We analyzed several actomyosin filament structures in the rigor conformation (S7 Fig) by structure-based sequence alignment (DALI, [54]) of the motor domains and compared the fractional buried surface area (PDBePISA, [55]) of MyoA residues in the actin contacting segments. Although, most interacting amino acids are not conserved among different myosin classes, especially in loops 2,3 and 4 (S6 Fig), they have a strikingly conserved pattern of buried surface area within the actin interface (horizontal blue bars in S7 Fig). Interestingly, when we set a threshold for interface residues that are >25% buried in all six myosins, a common interface consists of nine residues (long horizontal red boxes; Thr412, Ala414, Ile419, Gly421, Asp544, Leu547, Ala548, Pro550, Ile635) mostly contacts from HCM-loop and HLH substructures.

Recently, Robert-Paganin et al. (2021) defined conserved (15 residues) and ancillary (9 residues) interface in MyoA (green and cyan boxes in S7 Fig, respectively). Their conserved interface partially overlaps with our analysis, but our core interface consists of 9 residues. Additionally, we noted that each respective myosin motor makes extensive contacts with actin that are not present in the other myosins (gray boxes and yellow-highlighting in S7 Fig), which we summarize here. Within loop 3, Myo 1b has extensive contacts with actin that are not present in other myosins. In loop 4, MyoA uses it to interface actin more (yellow-shaded box) similarly to Myo 6. At the end of loop 2, Myo 7 has an additional patch of non-charged residues in addition to a charged-patch within loop 4. Within the activation loop of the HLH motif, Myo 2c has more interfacing residues; we noted double proline, and three additional residues at loop 2. The HCM-loop is 5 residues longer in Myo 6 than others, and the tip of loop makes extensive contacts with actin.

Helix-Turn-Helix (HTH) interactions

In MyoA, the majority of specific contacts are formed by the HTH motif in the lower 50K domain, which contacts the actin-actin interface (buried surface area 554 Ų), where the D-loop of one actin inserts into the cleft between subdomains 1 and 3 of the adjacent actin (Fig 3B and 3C). The first half of the HTH motif (Ser540-Gly550) forms extensive contacts with both actin subunits, whereas the latter part (Gly551- Lys565) interacts mostly with the D-loop. The Asp544 side chain of MyoA forms hydrogen bonds with the backbone amine of Thr352 in Act1 and the side-chain or backbone amine of Ser351. Asp544 also interacts via its backbone carbonyl with the backbone amine of Ala548 in MyoA. This interaction network between Act1 and MyoA occurs just before the kink in the relay helix, possibly stabilizing it. The tip of the HTH motif (Ala548-Pro549-Gly550) inserts into a hydrophobic cleft in Act1, lined by Ile347-Leu352 and Tyr145-Thr150 as well as the D-loop residues Met46-Val47 from the adjacent actin subunit.

Hypertrophic Cardio-Myopathy and other surface loop interactions

The HCM loop (Fig 3B) forms a second important interface to Act1 (buried surface area 284 Ų) that is complementary in shape, and we detect no hydrogen bonds at the interface. Additional contact points are created by loop 2, which is in close proximity to the HCM loop. The backbone atoms of Ile635 and Gly633, hydrogen bond the actin backbone atoms Gly24 and Asp26 respectively and the Ile635 side-chain inserts into a small patch on actin lined by Gly25, Asp26, Ser345, Ile346, and Ser349. The basic loop2 residues; Arg606 (in the strut segment), Lys634, and Lys637 coordinate actin Asp26, Asp25, and Ser349/Ser351, respectively, with distances ranging between 3.8 and 4.5 Å. The TEDS phosphorylation site [56] Thr417, located at the tip of the HCM-loop, is solvent accessible, and the density map suggests that it is not phosphorylated. Interestingly, our structural study of the actomosin complex did not require phosphomimetic mutation Thr417Asp in contrast to studies of the truncated MyoA [23], which results in a shift of the position of the HCM loop with respect to MyoA, and where the aspartic acid contributes buried surface area to the interaction.

Loop 3 has a small interaction point with Act1 sub-domain I, mediated by Val575/Ala576 in MyoA and Tyr92/Asn93 in actin. Finally, Loop 4 residues, Ala373-Asp377, have extensive interactions with Act1 throughout Lys329 to Glu335.

Concluding remarks

The malaria parasite actomyosin filament structure reported here will serve as a foundation for understanding gliding motility in Apicomplexa as well as force generation in a variety of actomyosin motor systems. We visualize the actin-myosin interface at high-resolution and also present the structure of the entire core motor complex including the light chains, which are essential for fast motility. In addition, we show that the actomyosin light chain complex produces a power stroke similar to other myosins with two light chains [57]. The extremely well-resolved bare Act1 filament assigns accurate side-chain conformations and reveals new active site features. The structures describe several protein-protein interfaces and will be important tools for evaluating the actomyosin motor as a drug target.

Optical trap assay

All chemical reagents were obtained from Sigma unless stated otherwise. Experiments were performed using a custom-built optical trapping instrument based around an inverted, Zeiss Axiovert 135 microscope, that has been described previously [44]. Briefly, microscope flow-cells were prepared from a microscope slide and coverslip fixed together with double-sided adhesive tape leaving a flow-channel of ~30 μl volume. Coverslips were previously coated with 2.4 μm diameter glass microspheres diluted into a 0.1% nitrocellulose solution. MyoA:ELC:MTIP was infused into the flow-cell at 1 μg ml^-1 in an assay buffer “AB-” containing 25 mM KCl, 25 mM imidazole, 4 mM MgCl₂, and 1 mM EGTA, pH 7.4, at ∼23°C. Flow-cell surfaces were then passivated using four chamber volumes of AB− with 1 mg ml^-1 BSA. The flow-cell was then infused with “AB+” buffer comprising AB- supplemented 0.3 μM ATP, 2 mM creatine phosphate, 0.1 mg ml^-1 creatine kinase, 50 mM DTT, 3 mM glucose, 100 μM glucose oxidase, and 20 μM catalase, 0.2 nM rhodamine-phalloidin labeled, 10% biotinylated, filamentous porcine actin and neutravidin-coated 1μm diameter polystyrene beads. Actin filaments were visualised using 532 nm laser excitation. The incoming 532 nm laser was reflected by a dual-dichroic mirror (532 nm and 1064 nm, Semrock custom dichroic) which also reflected the near infra-red (1064nm) optical trapping laser onto the specimen. Rhodamine fluorescence from the labelled actin was imaged via a 580-640nm bandpass filter using a custom-built, image-intensified CMOS camera. The polystyrene beads were visualised using bright-field optics which allowed a single actin filament (5–7 μm long) to be captured at either end between two beads held in two, independently controlled, optical traps. The so-called “dumbbell” of bead-actin-bead was then held taut and positioned in close proximity to one of the surface-immobilized MyoA:ELC:MTIP coated beads. Under conditions used here only about 1-in-3 of the surface beads produced actomyosin interactions due to the low surface-density of MyoA molecules. Bead positions were measured by projecting their bright-field images onto two, 4-quadrant, photodiode detectors (bandwidth >10kHz, resolution <1nm). Digital data was collected at 10kHz and saved to computer. Optical trap stiffness was 0.02 pN nm^-1 and thermal noise, governed by the equipartition principle, was ~15 nm r.m.s. When actin bound the surface-immobilized MyoA positional noise fell dramatically because system stiffness increased to >0.25 pN nm^-1 as the actomyosin bond stiffness acts in parallel to the trap stiffness. To increase the time-resolution of our measurements the optical tweezers were moved sinusoidally (50 nm peak-to-peak, 200 Hz) producing a ~12 nm p-p bead motion. Motion was quantified by computing a running discrete Fourier transform (software phase-locked-loop) of the bead motion. This enabled the start- and end-time of individual binding events to be identified with <10 ms jitter so that events could be synchronised in software and averaged to give an ensemble representation of the powerstroke time-course (see [43]).

Preparation of proteins and cryo-EM grids

MyoA:ELC:MTIP and Act1 were expressed and purified as previously described [10,24]. For Act1 samples, ammonium acetate was removed by a spin column, and Act1 (13.1 μM) was polymerized in the presence of 13 μM jasplakinolide by adding 10X KMEI, in final 1x concentration of 10 mM Hepes, pH 7.5, 50 mM KCl, 4 mM MgCl2, 1 mM EGTA. Polymerized filamentous Act1 was diluted down to 0.13 μM, and treated with apyrase (77 μg/ml) for 20 min before addition of the MyoA:ELC:MTIP complex in a 1:1 ratio.

For Act1-only samples, no apyrase was used. After 30 min incubation, negatively stained or cryo-EM grids were prepared. The quality of the decorated filaments was first checked with negative staining, and after confirming the presence of decorated Act1 filaments, frozen-hydrated samples were prepared on Quantifoil 2/2, 200 mesh in-air glow-discharged grids using an Mk III Vitrobot (FEI). 3 μl of sample was applied on a grid and blotted for 3 s before plunge-freezing in liquid ethane.

Act1:MyoA filament

Data consisting of 2815 movies were collected with a Titan Krios electron microscope equipped with a Falcon 3 camera, operated at 300 kV in counting mode. The magnification was 75 000x, corresponding to 1.09 Å pixel^-1, using a dose rate of 0.539 e^-/Ųs^-1 with exposure time of 92 s for a stack of 46 frames. The dose per frame was 1.07 e^-/Ų, and we used defoci from 0.8 to 2.6 μm. The movies were aligned with MotionCor2 [58] within the Scipion framework [59]. Contrast transfer function parameters were estimated by CTFFind 4.1.5 [60], and the resulting 2786 good images were used for further data processing. To confirm data processability, we picked decorated filaments by hand and performed low-resolution reconstructions from 2x binned images of a few hundred particles using symmetry parameters defined for Act1 [13] or refined helical symmetry parameters in Relion 2.1 [31,61] using a featureless cylinder as a starting reference. The reconstruction converged to a filamentous structure with protruding densities typical for actomyosin structures. Further inspection revealed that homologous myosin motor domains [34] as well as a filamentous Act1 model (PDB code: 5ogw [13]) fit the density well. Decorated filaments were iteratively picked with Relion autopick using reference-free Relion 2D classes as templates with an inter-box distance of two subunits with a rise of 28 Å. The resulting 381 256 particles were subjected to reference-free 2D classification in Relion, resulting in 239 225 fully decorated actomyosin particles. High-resolution volumes were generated in Relion 2.1 auto-refine using low-resolution models from 2x binned data as a reference. A soft mask was applied to the reference model, and reconstruction was continued from the last iteration. When the reconstruction converged, we continued refinement of the contrast transfer function and subsequent reconstructions with Relion 3 beta for two additional rounds. Refined half-maps were sharpened automatically [62], and global resolution was corrected for the effects of a mask [63] using the Relion postprocessing tool. We estimated local resolution with Blocres in Bsoft package version 1.8.6 with a Fourier shell correlation threshold of 0.3 [64,65].

Act1 filament

We collected 3953 movies with a Titan Krios electron microscope fitted with a Falcon 3 camera in electron counting mode, operated at 300 kV. We used a nominal magnification of 75 000x, corresponding to 1.09 Å pixel^-1. Each movie was recorded as 46 frames with 1.17 e^-/Ų dose/frame using defoci from 0.8 to 2.6 μm. The fractionated movies were aligned globally in Relion 3.0.7. Contrast transfer function was estimated by CtfFind 4.1.10 from dose-weighted micrographs. Particles were picked automatically using four times the rise (28 Å). The resulting 336 552 particles were extracted and binned two times at 2.18 Å/pixel in a 328 pixel box before they were subjected to a reference free 2D classification, and resulted in 305 480 particles in the final data set. These particles were re-extracted as unbinned, and processed with Relion auto-refine with a reference mask and solvent flattening from a previous reconstruction (nominal resolution 4.4 Å) low pass filtered to 20 Å. After polishing and contrast transfer function refinement, the resulting map reached an average resolution of 2.6 Å, based on the 0.143 Fourier shell criterion. Visual inspection of the density map features agreed with the resolution estimate. After reconstruction, the map was masked and sharpened by the Relion postprocessing tool.

Model building

Act1 filament

Our previous model of the Act1 filament (PDB code: 5ogw; [13]) was placed in the density map in Chimera [66], and was manually built using model building tools in ISOLDE 1.0b3.dev6 [67], Coot [68], and real-space refinement in Phenix 1.17 [72]. We built five copies of Act1, re-assigning side chain rotamers as required, and waters around Act1 monomers into density. Additionally, we built the active site with the Mg²⁺-coordinated waters. Density for the Mg²⁺-water complex was anisotropic, and the Mg²⁺-coordinated waters were merged into Mg²⁺-ion density. In order to keep ideal coordination of the water molecules, we created custom restraints for waters to restrain ideal distance and angles to the Mg²⁺ ion.

Low-resolution model of light-chain-decorated myosin A

Unfiltered 2x binned half-maps were filtered with LAFTER [73], and the MyoA model (6tu7) and a trimeric light chain model (6zn3; [22]) were placed into the filtered LAFTER map.