Nonmuscle myosin 2 without an assembly competence domain can incorporate into established filaments in cells

Abstract

Myosin 2 dynamically assembles into filaments that exert force on the actin cytoskeleton. To form filaments, myosin 2 monomers transition between folded and unfolded states. Monomer unfolding exposes an extended coiled-coil that interacts with other monomers in parallel and antiparallel fashions, enabling bipolar filament formation. A C-terminal domain of the coiled-coil, termed assembly competence domain (ACD), has been repeatedly identified as necessary for filament assembly. Here, we revisit ACD contribution when full-length filaments are present. Nonmuscle myosin 2A lacking the ACD (∆ACD) initially appears diffuse, but triton extraction of cytosolic fraction reveals cytoskeletal association. Disruption of the folded monomer enhances the cytoskeletal fraction, while inhibition of endogenous filament assembly appears to reduce it. Finally, high-resolution imaging of endogenous and exogenous myosin 2 reveals highly coincident filamentous structures, suggesting that ∆ACD constructs coassemble with endogenous myosin 2A filaments. Our data demonstrate that while the ACD is required for de novo filament assembly, it is not required for monomers to recognize and associate with established filaments in cells. More broadly, this highlights the existence of distinct pathways governing how monomers participate in nascent filament assembly and how monomers recognize and associate with established filaments to maintain steady-state contractile networks.

• For decades, researchers have demonstrated that a domain on the myosin 2 tail is necessary for filament assembly.

• We demonstrate that this domain is not required for myosin 2 monomers recognizing and associating with endogenous full-length myosin 2 filaments.

• This suggests that there are potentially unique mechanisms governing myosin 2 filament assembly and myosin 2 filament maintenance.

INTRODUCTION

Myosin 2 motors are the dominant contractile motor proteins in mammalian cells. To function, myosin 2 monomers assemble into bipolar filaments that engage filamentous actin to drive contraction (reviewed in (Quintanilla et al., 2023; Chinthalapudi and Heissler, 2024). Defining the process of myosin 2 filament formation informs how cells build and maintain contractile network dynamics to drive an array of cellular processes, including migration and division. In nonmuscle cells, roughly one to two-thirds of the myosin 2 exist in the filamentous state (Breckenridge et al., 2009; Smith et al., 2017; Schiffhauer et al., 2019). Although filaments are continually being assembled and disassembled throughout the cell, it is likely that many filaments exist for extended periods with a steady-state exchange of monomers moving into and out of the filament. Therefore, although identifying mechanisms by which myosin 2 monomers assemble nascent filaments is critical for understanding how contractile networks are built, identifying mechanisms by which myosin 2 monomers recognize and associate with established filaments is critical for understanding how contractile networks are maintained.

Myosin 2 monomers are hexameric ensembles of three components: two myosin heavy chains (MHC), two regulatory light chains (RLC), and two essential light chains (ELC). MHCs consist of a motor domain, a neck region with two light-chain binding IQ motifs, an extended alpha-helix that dimerizes into an extended coiled-coil “tail” and terminates in a short nonhelical tailpiece (Figure 1A) (Quintanilla et al., 2023). The coiled-coil is imperative for filament assembly and a focus of this work.

FIGURE 1:

EGFP-2A-∆ACD colocalizes with endogenous myosin 2A networks upon extraction of the cytosolic fraction. (A) Cartoon of full-length EGFP-2A (top) and EGFP-2A-∆ACD (bottom) with amino acids in parentheses. The ACD is displayed in orange. Minus and plus signs indicate the net charged regions along the coiled-coil tail. (B–D) Summed confocal zprojections. “Raw” panels and insets are scaled for intensity identical to preextraction images, while “Enhanced” images are scaled independently. (B) Fibroblasts transiently expressing diffuse EGFP imaged pre- and postextraction. (C and D) Halo-2A fibroblasts expressing EGFP-2A (C) or EGFP-2A-∆ACD (D) imaged pre- and postextraction. The rightmost column in D displays zoomed insets of the orange box in the EGFP-2A-∆ACD postextraction merged image. See also Supplemental Movie S1.

Movie S1.

Download video file ^{(10.6MB, mov)}

Time-lapse imaging EGFP constructs during triton extraction. Fibroblasts expressing EGFP alone or EGFP2A constructs as indicated were imaged on a Zeiss LSM 880 Airyscan confocal. A 3 µm stack with 7 × 500 nm steps at the ventral surface was collected every 15 seconds and sum projected. LUT applied is in bottom right corner (MQ_div-magma). Triton extraction buffer was added after the first two time points. Video frame rate = 5 fps.

To prevent spurious filament assembly, mammalian myosin 2 monomers can be sequestered into a folded, inactive state termed the “10S” (Trybus et al., 1982). In the 10S, the motor domains autoinhibit by docking on one another and fold back to bind along the N-terminus of the coiled-coil, creating the interacting heads motif (IHM) (Trybus et al., 1997; Wendt et al., 1999). The coiled-coil tail also folds twice to wrap around the IHM, further stabilizing the sequestered 10S (Onishi and Wakabayashi, 1982; Jung et al., 2011; Yang et al., 2020). The presence of unphosphorylated RLCs is critical for 10S formation and stability (Trybus et al., 1982; Trybus and Lowey, 1988). Phosphorylation of the RLC at T18/S19, or deletion of the second IQ motif (∆IQ2) where the RLC binds, destabilizes the 10S and shifts the equilibrium to favor the assembly-competent “6S” monomer (Adelstein and Conti, 1975; Craig et al., 1983). This unfolding exposes the coiled-coil tail. Surface-exposed residues along the coiled-coil display alternating negative and neutral charge regions, with a positive charge region near the C-terminus (Figure 1A) (Parry, 1981; McLachlan and Karn, 1982; Ricketson et al., 2010; Billington et al., 2013). These alternating charge regions enable staggered electrostatic interactions that stabilize parallel and antiparallel monomer interactions to promote bipolar filament assembly (Ricketson et al., 2010; Kaufmann and Schwarz, 2020). Therefore, the standard assembly model is that the folded, inactive 10S diffuses throughout the cytoplasm until it is phosphorylated by RLC kinases to induce unfolding into the assembly-competent activated 6S that subsequently assembles into filaments. Alternative nonmutually exclusive assembly models have also been proposed, which suggest that the 10S monomer or 10S dimers first interact with established filaments, and then unfold into the 6S while already filament-associated (Liu et al., 2020).

Foundational studies to establish mechanisms of assembly have been performed using purified proteins. While examining both full-length and tail fragments of various myosin 2s, these studies defined an assembly competence domain (ACD) in the C-terminus of the coiled-coil, which universally includes the lone positive charge region (Nyitray et al., 1983; Cross and Vandekerckhove, 1986; Atkinson and Stewart, 1991; Sohn et al., 1997; Nakasawa et al., 2005). These studies using myosin 2 constructs either lacking the ACD (∆ACD) or consisting solely of the ACD established that the ACD is both necessary and sufficient for paracrystal formation or insolubility, suggesting that it is necessary and sufficient for filament assembly. However, these studies predominantly examined homogenous populations for their ability to assemble, and did not examine truncated myosin 2 in the presence of full-length myosin 2. Therefore, these studies were primarily examining the capacity for monomers to assemble new filaments, not the capacity for monomers to recognize and associate with established filaments in steady-state contractile networks.

Parallel to the in vitro work, several studies examined tail truncations in Dictyostelium and mammalian cells. Overexpression of EGFP-tagged myosin 2 with C-terminal truncations that removed all or part of the ACD was reported to abolish assembly and result in diffuse cytosolic myosin 2 (Fukui et al., 1990; Lee et al., 1994; Sohn et al., 1997; Liu et al., 1998; Ikebe et al., 2001; Breckenridge et al., 2009). This suggested that even in the presence of endogenous full-length myosin 2, the ACD is necessary for recognition and association with established filaments. Reciprocal to ∆ACD constructs preventing assembly, ∆IQ2 constructs that prevent RLC binding are thought to be constitutively filamentous (Uyeda and Spudich, 1993; Breckenridge et al., 2009). This is due to the inability to form the autoinhibited 10S monomer, thus locking the molecule into the unfolded 6S that readily assembles into filaments. A double deletion ∆IQ2∆ACD construct also appeared cytosolic when expressed in mammalian cells, but displayed slower diffusion kinetics than the ∆ACD alone, consistent with it being locked into the unfolded 6S monomer but unable to assemble (Breckenridge et al., 2009).

Here, we revisit myosin 2 truncation and deletion mutants using the mammalian nonmuscle myosin 2A isoform (hereafter myosin 2A) to test models of filament assembly and maintenance in cells. Contrary to previous studies, we find that ACD deletion does not prevent recognition of and association with established full-length myosin 2 filaments in cells. Moreover, simultaneous deletion of both the ACD and the IQ2 enhances filament association. Inhibition of endogenous filament assembly by either ROCK (rho-associated coiled-coil containing kinase) inhibition or genetic ablation of myosin 2A (Myh9) demonstrates that this filament association is dependent on the presence of endogenous filaments. Finally, high-resolution imaging reveals highly coincident exogenous and endogenous filamentous structures, consistent with these deletion constructs coassembling into full-length endogenous myosin 2 filaments.

RESULTS

Retention of myosin 2A filament association upon ACD deletion

To test the contribution of the ACD to the steady-state incorporation of myosin 2A monomers into established filaments, we expressed EGFP, EGFP-myosin 2A (“EGFP-2A”), or EGFP-myosin 2A-∆ACD (“EGFP-2A-∆ACD”) in an immortalized mouse fibroblast cell line (“JR20s”) (Rotty et al., 2015), where endogenous myosin 2A is labeled with a HaloTag (Halo-2A; Figure 1A; Supplemental Figure S1). Although the EGFP signal appeared diffuse (Figure 1B; preextract), EGFP-2A displayed cytoskeletal localization that was highly coincident with the endogenous Halo-myosin 2A signal (Figure 1C; preextract). Similar to EGFP, and in agreement with previous publications, EGFP-2A-∆ACD displayed a diffuse signal in fibroblasts, with no discernible cytoskeletal appearance (Figure 1, B and D; preextract). However, extraction of the cells with a physiological buffer containing Triton X-100, which releases the cytosolic EGFP (Figure 1B; postextract) and cytosolic myosin 2 fractions but retains actin-bound filamentous myosin 2, revealed a cytoskeletal myosin 2-like distribution of EGFP-2A-∆ACD (Figure 1D; postextract; Supplemental Movie S1). Although the postextracted raw EGFP-2A-∆ACD signal intensity was reduced relative to EGFP-2A (Figure 1; “Raw” insets), the enhanced signal was highly coincident with the endogenous Halo-2A signal (Figure 1, B and D; postextraction zoomed insets). These data suggest that while the majority of the EGFP-2A-∆ACD molecules are monomeric, a fraction might associate with and incorporate into endogenous myosin 2A filaments.

Inhibition of 10S monomer increases assembly of EGFP-2A-∆ACD

We hypothesized that EGFP-2A-∆ACD follows a canonical assembly process, whereby RLC phosphorylation inhibits the folded 10S monomer, generating unfolded 6S monomers that transiently assemble into established endogenous myosin 2 filaments. To explore this, we used a previously published construct that inhibits 10S formation by deleting the RLC-binding IQ2 motif, termed EGFP-2A-∆IQ2∆ACD (Figure 2A) (Breckenridge et al., 2009). Previous studies theorized that this construct would reside in the 6S intermediate, unable to form the folded 10S and unable to assemble into filaments. When expressed in our fibroblasts, EGFP-2A-∆IQ2∆ACD displayed a diffuse cytosolic signal preextraction in a subpopulation of the cells (Figure 2B; preextract). However, many cells also displayed clear cytoskeletal structures before any extraction (Figure 3C; Supplemental Movies S1 and S2). Upon triton-extraction, we readily observed a robust cytoskeletal, myosin 2-like network in all cells (Figure 2B; postextract). Similar residual cytoskeletal signal was observed following triton extraction for both EGFP-2A-∆ACD and -∆IQ2∆ACD in other cell types previously used to investigate these deletion/truncation constructs (Hela and U2OS cells; Supplemental Figure S2), demonstrating that these observations are not cell-type dependent.

FIGURE 2:

Deletion of both the IQ2 motif and ACD enhances association with endogenous myosin filaments. (A) Cartoon of EGFP-2A-∆IQ2∆ACD with amino acids in parentheses. Note, the loss of RLC binding via IQ2 deletion. (B) Summed confocal z-projection of Halo-2A fibroblasts transiently expressing EGFP-2A-∆IQ2∆ACD. “Raw” inset is scaled for intensity identical to preextraction image, while “Enhanced” image is scaled independently. The rightmost column displays zoomed insets of the orange box in EGFP-2A-∆IQ2∆ACD postextraction merged image. (C) Live cell imaging of the EGFP-2A-∆IQ2∆ACD construct in fibroblasts. The zoomed inset (right) of the magenta box on the left shows the cytoskeletal localization of EGFP2A-∆IQ2∆ACD. See also Supplemental Movies S1 and S2.

FIGURE 3:

Quantification of myosin 2A insoluble fractions in triton fractionation assays. (A) Cartoon of the analysis pipeline (for details, see Supplemental Figure S3). Images of Halo-2A fibroblasts transiently expressing EGFP-2A constructs pre- and postextraction were used for analysis. The consistent endogenous Halo-2A signal was used to generate a cell mask, which was applied to the pre- and postextraction EGFP images. The insoluble fraction, as a correlate for filament assembly, was calculated as the ratio of postextraction intensity to preextraction intensity. (B) Comparison of the insoluble fractions for EGFP or EGFP-2A constructs without (left) or with (right) ROCK inhibition using 10 µM Y27632 for 30 min. Data points indicate individual cells with error bars, indicating mean ± SD. Statistical comparisons were made using nested one-way ANOVA and Tukey's multiple comparison tests. The p values are indicated for “not significantly different” comparisons between untreated and ROCK-inhibited datasets. See Supplemental Table S1 for replicates and full statistical comparisons. Data points outside the scale were removed from the graph but included in statistical analyses.

Movie S2.

Download video file ^{(7.9MB, mov)}

Time-lapse imaging of ∆IQ2∆ACD. A fibroblast expressing EGFP-2A-∆IQ2∆ACD was imaged on a Zeiss LSM 880 Airyscan confocal. A 1.5 µm stack with 4 × 500 nm steps at the ventral surface was collected every 30 seconds and sum projected. Magenta box on left indicates zoomed area on right. Video frame rate = 25 fps.

To quantify these observations, we developed an analysis pipeline to measure the intracellular EGFP intensity pre- and postextraction based on a cell mask generated from the consistent endogenous Halo-2A signal (Figure 3A; Supplemental Figure S3). The postextraction to preextraction ratio provides an “insoluble fraction,” which should predominantly represent filamentous actin-bound myosin 2 filaments (see Materials and Methods). As expected, the mean insoluble fraction for the EGFP control was only 0.004 (0.4%), consistent with its cytosolic localization (Figure 3B). In agreement with previous reports (Breckenridge et al., 2009; Smith et al., 2017; Schiffhauer et al., 2019), approximately half (48.4%) of the EGFP-2A molecules were in the insoluble fraction (Figure 3B). Although about one-tenth (9.5%) of EGFP-2A-∆ACD was present in the insoluble fraction, nearly half (47.9%) of EGFP-2A∆IQ2∆ACD was insoluble, comparable with that of full-length EGFP-2A. Plotting the preextraction EGFP intensities versus the insoluble fraction for each construct revealed that their insoluble fractions were independent of their expression levels (Supplemental Figure S1B).

Next, we considered an alternative possibility for our observations: that the ACD deletion constructs are triton insoluble due to formation of heteromeric monomers with endogenous myosin 2. In this model, a single exogenous MHC without an ACD would form an incomplete coiled-coil with an endogenous full-length MHC, potentially generating peculiar monomer or dimer species with an ACD that could associate with filaments (diagram in Supplemental Figure S4A). To test this, we immunoprecipitated (IP) EGFP2A, EGFP-2A-∆ACD, or EGFP-2A-∆IQ2∆ACD monomers from wild-type fibroblasts. Monomers were enriched by treating the cells for 1 h with ROCK inhibitor (Y27632) and using high salt concentrations throughout lysis and IP (Supplemental Figure S4B). For EGFP-2A, we observed some endogenous myosin 2 in the IP (Supplemental Figure S4C, magenta arrows). However, for ACD deletion constructs, we observed little to no myosin 2 in the IP. Although these traced levels of heteromeric monomers could contribute to some of the triton insolubility we observe (Figure 3B), they certainly cannot account for the ∼50% EGFP-2A-∆IQ2∆ACD in the insoluble fraction, and support a model of coassembly.

These data in wild-type fibroblasts also enable ballpark quantification of overexpression, as the endogenous Halo-2A is no longer obscuring the exogenous EGFP-2A Western blot band. We observed ∼30% transfection efficiency and ∼3×-fold greater intensity for endogenous myosin 2A relative to EGFP-2A (Supplemental Figure S4). Therefore, the mean overexpression level is approximately equal to endogenous myosin 2A.

Endogenous myosin 2A filaments promote filament association of EGFP-2A-∆ACD and -∆IQ2∆ACD

To test our hypothesis that the EGFP-2A-∆ACD and EGFP-2A-∆IQ2∆ACD truncation constructs coassemble into endogenous filaments, we used two parallel assays to reduce endogenous myosin filament levels. First, we inhibited the dominant RLC kinase in fibroblasts, ROCK, using the small-molecule Y27632. As expected, mild ROCK inhibition reduced EGFP-2A assembly by 2.7-fold (48.3% vs. 17.9%), reduced EGFP-2A-∆ACD 2.1-fold (9.5% vs. 4.4%), and EGFP-2A-∆IQ2∆ACD assembly by 1.4-fold (47.9% vs. 33.5%; Figure 2B). Notably, EGFP-2A-∆IQ2∆ACD may be slightly less impacted by ROCK inhibition compared with EGFP-2A, as it does not have any RLC-binding sites and therefore is not a direct target of ROCK. These data are consistent with the majority of the insoluble fraction for these constructs coassembling with endogenous myosin 2A.

Our second approach to reducing endogenous filament levels was to genetically ablate endogenous myosin 2A (Myh9 gene) in fibroblasts using CRISPR/Cas9. These fibroblasts predominantly express myosin 2A, with little expression of myosin 2B and no detectable myosin 2C (Supplemental Figure S5). Therefore, despite the ability of these NM2 isoforms to coassemble (Beach et al., 2014; Shutova et al., 2014), deletion of myosin 2A should remove the majority of endogenous myosin 2, significantly reducing the filamentous appearance of EGFP-2A-∆ACD and -∆IQ2∆ACD. Transient expression and triton extraction of full-length EGFP-2A in myosin 2A-knockout (KO) cells revealed that about 50% was filamentous (Figure 4, A and B), similar to what we observed in wild-type fibroblasts (Figure 3B). For EGFP-2A-∆ACD, we again observed a diffuse signal preextraction and little residual signal postextraction (Figure 4, A and B). For EGFP-2A-∆IQ2∆ACD, filament incorporation was more apparent than for EGFP-2A∆ACD, with a higher insoluble fraction postextraction (Figure 4, A and B). However, this was about 2-fold lower than the insoluble fraction of EGFP-2A-∆IQ2∆ACD in wild-type fibroblasts (Figure 3B). Our data show that while EGFP-2A-∆ACD and EGFP2A-∆IQ2∆ACD still assemble into filaments, their ability to do so is reduced when endogenous myosin filament levels are reduced.

FIGURE 4:

Myosin 2A KO reduces the assembly of EGFP-2A-∆IQ2∆ACD. (A) Summed confocal z-projection of myosin 2A-KO fibroblasts transiently expressing EGFP-2A (top), EGFP-2A-∆ACD (middle), and EGFP-2A-∆IQ2∆ACD (bottom) pre- and postextraction. The EGFP signal was used to generate a cell mask, which was applied to the pre- and postextraction images. “Raw” inset is scaled for intensity identical to preextraction image, while “Enhanced” image is scaled independently. (B) Comparison of the insoluble fractions for EGFP-2A constructs. Data points indicate individual cells with error bars, indicating mean ± SD. Statistical comparisons were made using nested one-way ANOVA and Tukey's multiple comparison tests. See Supplemental Table S1 for replicates and full statistical comparisons.

High-resolution imaging of endogenous and exogenous myosin 2A is consistent with a coassembly model.

Finally, to test whether the triton insolubility and cytoskeletal association of the exogenous myosin 2 constructs used in this study were due to actin binding instead of coassembly with endogenous myosin2 filaments, we used Structured Illumination Microscopy (SIM). SIM provides a theoretical lateral resolution under 100 nm, sufficient to resolve individual myosin 2 filaments or small filament stacks, with fluorophore-tagged N-termini creating puncta ∼300 nm apart (Figure 5A) (Beach et al., 2014). We imaged lamellar regions of extracted Halo-2A fibroblasts expressing either EGFP-2A, EGFP-2A∆ACD, or EGFP-2A-∆IQ2∆ACD and stained for filamentous actin with phalloidin. We observed highly coincident structures consisting of endogenous Halo-2A and each of the exogenous EGFP-2A constructs (Figure 5B; columns 2, 3, and 5). In contrast, although these EGFP-2A constructs overlapped with some actin fibers, as expected, the majority of the actin filaments showed no associated EGFP-2A signal. This observation is particularly relevant for the EGFP-2A-∆IQ2∆ACD construct, which is theoretically constitutively active and capable of binding actin independently of filament assembly. We further quantified potential colocalization by performing cross-correlation analysis (McCall, 2024) between each EGFP-2A construct and Halo2A, and between each EGFP-2A construct and actin, in myosin-rich lamellar regions. For every cell analyzed, each EGFP-2A construct showed a higher correlation with Halo-2A than with actin. Therefore, although actin binding could contribute to some of the cytoskeletal association and triton-insolubility observed, these data strongly suggest that filament coassembly is the primary contributing mechanism.

FIGURE 5:

High-resolution imaging reveals likely co-assembled endogenous and exogenous myosin 2A filaments. (A) Cartoon of co-assembled endogenous and exogenous myosin 2A filaments imaged at high-spatial resolution. Halo-2A (magenta) monomers co-assembled into bipolar filaments with EGFP-2A monomers (green) should appear as two puncta ∼300 nm apart, with significant magenta and green overlap, when imaged with SIM (see Materials and Methods). (B) Postextraction images of Halo-2A fibroblasts expressing the indicated EGFP-2A construct and stained for filamentous actin. First column provides overview of cell morphology with all three channels merged. Remaining columns are zoomed insets of rear lamellar regions with inverted gray scale for the individual channels (columns 2–4) or merged images as indicated (columns 5–6). (C) Quantification of colocalization by cross correlation comparing the indicated EGFP-2A construct with either Halo-2A or actin in the same lamellar region. Values below 0.1, between 0.1 and 0.2, and above 0.2 indicate low, medium, and high confidence levels, respectively, in molecule colocalization (McCall, 2024).

Collectively, these data argue that removal of the ACD does not preclude monomers from recognizing and associating with established myosin 2 filaments and that disruption of the 10S increases filament assembly even in the absence of an ACD.

DISCUSSION

Here, we provide novel insight into how myosin 2 monomers recognize and incorporate into established myosin 2 filaments. We posit that there are potentially unique mechanistic differences between monomers assembling into a nascent filament (Figure 6, top) and monomers recognizing and incorporating into established filaments (Figure 6, bottom). This necessitates unique, but not mutually exclusive, molecular models to test. To assemble nascent filaments, we maintain that the ACD is absolutely required. Definitive experiments with purified protein and modeling studies inform us that the positively charged regions within the ACD are needed to stabilize electrostatic interactions between myosin 2 monomers in both a parallel and antiparallel manner (Nyitray et al., 1983; Cross and Vandekerckhove, 1986; Atkinson and Stewart, 1991; Sohn et al., 1997; Nakasawa et al., 2005). These interactions could occur between 10S dimers that subsequently unfold into nascent filaments (Liu et al., 2020), or between unfolded 6S monomers that encounter each other in the cytosol. Regardless, the positively charged regions within the ACD, and therefore the ACD itself, are required for de novo filament assembly.

FIGURE 6:

Model of myosin 2 filament assembly and association, see Discussion.

In contrast, we argue that the ability of myosin monomers to recognize and incorporate into an established filament is a distinct concept, which has often either been overlooked or incorrectly grouped into the former concept of nascent filament assembly. Although there is likely significant overlap in the molecular interactions involved in filament assembly and filament association, there could also be exclusive interactions that enable association but are not sufficient for assembly. By removing the ACD in our study, we were able to distinguish between filament assembly and filament recognition. We demonstrate that even in the absence of an ACD, myosin 2 molecules can associate with and incorporate into established filaments composed of full-length myosin 2. Furthermore, through additional removal of the IQ2 motif and inhibition of the 10S, we demonstrate that the extended 6S monomer, even without an ACD, can readily incorporate into established filaments composed of full-length myosin 2. We note that our results do not invalidate nonmutually exclusive models, suggesting that 10S monomers or 10S dimers can associate with established filaments and then unfold and incorporate into filaments (Liu et al., 2018).

Our work elicits consideration of relative contribution of de novo filament assembly versus the maintenance of established filaments. Certainly, de novo filament assembly is critical throughout cell biology. However, the contribution of de novo assembly may be skewed toward processes involving major morphological changes, where many new filaments need to be generated (e.g., cytokinesis, rapid cell migration, etc.). In contrast, steady-state maintenance of established filaments is likely the dominant pathway in many “static” cells throughout the body. We do not know what the lifetime of an established filament is, but fluorescence recovery after photobleaching (FRAP) studies show that the half-times of recovery for nonmuscle myosin 2 are ∼0.5 to 1.5 min, indicating that monomers are readily exchanging into and out of established filaments (Sandquist and Means, 2008; Breckenridge et al., 2009). Surprisingly, relatively rapid exchange kinetics have also recently been reported for cardiac myosin 2 (Kelly et al., 2024). Therefore, in steady-state contractile networks (e.g., stress fibers or sarcomeres), it is possible that the original monomers in the filament are entirely replaced multiple times during the filament's lifetime. One could even ponder, a la the Ship of Theseus, “if all monomers of a filament are replaced is it still the same filament?” Regardless of the philosophical answer, the collective evidence argues that replacing monomers within filaments happens and is critical to maintaining physiological levels of contractility.

Currently, we can only speculate on the precise molecular interactions enabling our truncation/deletion constructs to associate with established contractile networks. We hypothesize that the negatively charged regions in the proximal tail of the EGFP2A-∆ACD and -∆IQ2∆ACD constructs are able to recognize and interact with the positively charged regions in the ACD of full-length myosin 2 already incorporated into established filaments. This suggests that these positively charged regions are dynamically and transiently available for monomer recruitment into established filaments rather than being permanently engaged in electrostatic interactions within the filament. This model is reminiscent of previous studies demonstrating that myosin 18A can coassemble with myosin 2 (Billington et al., 2015). Myosin 18A possesses an extended coiled-coil domain with alternating negatively charged regions but no positively charged region, as is found in the myosin 2 ACD. Therefore, myosin 18A has domain similarity to a myosin 2-∆ACD construct. Alternatively, or in addition to this electrostatic tail recruitment model, many low-affinity intramolecular interactions between the tail and the ELC, RLC, or motor domain that stabilize the folded 10S (Yang et al., 2020) could facilitate intermolecular interactions between monomeric and filamentous myosin 2, thereby recruiting the monomer.

Although our imaging and cross-correlation data argue that most of the EGFP-2A-∆IQ2∆ACD signal is interacting with endogenous myosin filaments, there remains the possibility that this is not the exclusive mechanism driving retention after triton extraction. If there are additional interactions, the most likely alternative interaction is with actin, as the absence of RLCs should disrupt the IHM, enabling actin binding. Our data might even shed light on this possibility: First, we occasionally observe some low-level background intensity postextraction for the EGFP-2A-∆IQ2∆ACD (see lamella of Figure 2B, for example), although the colocalization by cross-correlation (CCC) quantification did not reveal a notable increase to support this. This could indicate some residual actin-binding capacity by monomers. Second, if EGFP-2A-∆IQ2∆ACD insolubility was entirely dependent on endogenous filament assembly, ROCK inhibition would have impacted both EGFP-2A and EGFP-2A-∆IQ2∆ACD similarly. However, EGFP-2A insolubility decreased 3-fold while EGFP-2A-∆IQ2∆ACD decreased only ∼1.5-fold. Some of this resistance to ROCK inhibition could be because EGFP-2A-∆IQ2∆ACD is not a direct target of ROCK due to the absence of an RLC to phosphorylate, leaving unfolded molecules that can bind actin even when endogenous filament levels are reduced.

It is worth restating that several previous studies, which similarly overexpressed EGFP-tagged myosin 2 truncations in mammalian cells, did not observe filamentous structures (Breckenridge et al., 2009; Weißenbruch et al., 2021). Minor sequence differences in the creation of the truncated ∆ACD construct might account for some discrepancy, but we think this is unlikely. We performed our triton extraction assay in the same cell types used in those earlier studies (e.g., HeLa and U2OS) and observed similar cytoskeletal localization of both EGFP-2A-∆ACD and -∆IQ2∆ACD, ruling out cell-type specificity. Considering our quantitative analysis suggests that over 90% of EGFP-2A-∆ACD is cytosolic, we believe that this cytosolic component masked the remaining filament-incorporated component in earlier studies, rendering it undetectable without removal of the cytosolic fraction. This argues that any fluorophore-tagged protein that initially appears diffuse should be interpreted carefully, and researchers should consider experiments to remove cytosolic fractions to determine whether underlying stable associations with the cytoskeleton are present.

Our observation suggesting that myosin 2 can form heteromeric monomers (Supplemental Figure S4) is worth discussing. We previously assumed that MHC alpha helical tails form coiled-coils with neighboring proteins in polysomes during protein translation. We further assumed that once the monomer forms, the stability of such an extended coiled-coil would render it a life-long monogamous interaction. Our current data suggest that one or both of these assumptions is false. This is important when experimentally introducing exogenous myosin 2 constructs into cells, but also suggests that disease-causing mutations in heterozygotic patients may not be an independent species of homomeric monomer, but are potentially a combination of homomeric and heteromeric monomer.

From a technical perspective, we highlight that myosin 2-∆ACD should not be used as a negative control for assembly assays where full-length myosin 2 is present. Furthermore, although there may be differences in the assembly mechanisms between nonmuscle myosin 2 and smooth/striated myosin 2 paralogues, it is likely that our observations apply to all myosin 2 paralogues. Therefore, experimentation with any ∆ACD construct should be performed with caution.

In conclusion, our key observation that the ACD is not a requisite domain for recognition and incorporation into established filaments highlights that the terms “filament assembly” and “filament association” are not synonymous. We need to consider potentially exclusive mechanisms by which monomers recognize and incorporate into established filaments that are unique from mechanisms by which monomers are able to assemble de novo filaments. Isolating these concepts is experimentally challenging. However, at minimum, we argue that the field should be careful not to conflate these concepts when interpreting and discussing data.

MATERIALS AND METHODS

Request a protocol through Bio-protocol

Cell culture and transfection

Halo-2A (clone 2; “H2A2”), myosin 2A-KO, U2OS, and Hela cells were cultured in DMEM (10-013-CV, Corning) supplemented with 10% FB Essence (10803-034, Avantor Seradigm) and 1% antibiotic–antimycotic (30004-Cl, Corning) at 37°C and 5% CO₂. To generate CRISPR knockin lines, 1 million cells were transfected with 5 µg each of the target and donor plasmid using the LipoD293 (SL100668, SignaGen) system. Cells were transfected 24 h before experiments. For H2A2, U2OS or Hela, 400,000 cells were transfected with 3 µg (EGFP) or 4 µg (EGFP-2A constructs) plasmid DNA using a Neon electroporation system (Thermo Fisher Scientific) with a single 20 ms pulse of 1600V, and plated on glass coverslips precoated with 20 µg/ml human plasma fibronectin (FC010, EMD Millipore) at room temperature (RT) for 1 h. Alternatively, 200,000 myosin 2A-KO cells were transfected with 5 µg DNA, or for GFP IP experiments, 800,000 cells were transfected with 5 µg of plasmid DNA, using the same electroporation settings. Four hours posttransfection, cells were incubated overnight with 20 nM JFx650-HaloTag ligand (Lavis Laboratory, Lot: Sep-1-152).

Generation of CRISPR knockin and KO cell lines

Halo-2A knockin cells were described previously (Quintanilla et al., 2024). Briefly, immortalized parental fibroblasts (JR20) (Rotty et al., 2015) were transfected with pSpCas9 (BB)-2APuro (PX459) V2.0 (Plasmid 62988, Addgene) with target sequence 5′-AAACTTCATCAATAACCCGC-3′ generated using established protocols (Ran et al., 2013), and a donor plasmid (pUC57 with 794 bp 5′ HDR of genomic sequence immediate upstream of the endogenous start codon, a HaloTag with an 18 amino acid GS-rich linker, an 802 bp 3′ HDR of genomic sequence immediately downstream of the endogenous start codon with silent PAM site mutation). Single-cell sorting was performed 5 to 10 d after transfection. Individual clones were evaluated for knockin via Western blot analysis and microscopy. Clones used in this study include Halo-2A clone 2 (H2A2).

To generate myosin 2A-KO lines, two lentiviral plasmids targeting Myh9 were constructed: pLenti-CRISPRv2-puro (Plasmid 98290, Addgene; target, ACCCTGCATCATGCTCCGGT-AGG) and pLenti-CRISPRv2-puro (Plasmid 98291, Addgene; target, GCAGCACCGAAGCTTCGTTG-AGG). Lentivirus was generated in HEK293 cells and placed directly on fibroblasts. Cells were selected in both puromycin and hygromycin and single-cell sorting was performed to isolate KO clones. KOs were validated using Western blot analysis. JR20-myosin 2A-KO clone 3 was used for these experiments.

Western blotting

Cells were pelleted, resuspended in 2x Tris-Glycine Laemelli sample buffer (1610737, Bio-Rad) supplemented with 10% β-mercaptoethanol and 1x Halt protease and phosphatase inhibitor cocktail (78442, Thermo Fisher Scientific), and boiled for 5 min at 95°C. Cell lysates were analyzed with SDS–PAGE using 4 to 15% Mini-PROTEAN

TGX Stain-free gels (4568083, Bio-Rad). Proteins were transferred to a nitrocellulose membrane (1704271, Bio-Rad) using a Trans-Blot Turbo system (Bio-Rad). Membranes were blocked for 1 h at RT in 3% BSA/PBS supplemented with 0.1% Tween 20 (BP337-100, Thermo Fisher Scientific), incubated at 4°C overnight with primary antibody in 3% BSA/PBS + 0.1% Tween 20, washed three times for 5 min in blocking buffer, incubated for 1 h at RT with secondary antibody in blocking buffer, washed three times for 5 min in PBS + 0.1% Tween 20, rinsed in Milli-Q-H2O, and incubated with clarity Western ECL Substrate (1705061, Bio-Rad) for 5 min and imaged on a ChemiDoc MP Imaging System (Bio-Rad) for signal detection. The following primary antibodies were used: rabbit anti–β-actin (GTX109639, GeneTex, 1/5000 dilution), mouse anti-GFP (sc-9996, Santa Cruz Biotechnology, 1/2000 dilution), and rabbit anti-myosin IIA HC (MP3791, ECM Biosciences, 1/5000 dilution). The following secondary antibodies were used at a 1/5000 dilution: HRP goat anti-rabbit IgG (H+L) antibody (K1223, Apex bio), horseradish peroxidatse  goat anti-mouse IgG (H+L) antibody (K1221, Apex bio).

Triton-insolubility assays

Triton buffer consisted of 0.6% Triton X-100 (Millipore, 9002-93-1), 4% PEG 8000 (Promega, V3011), 5 mM NaCl, 140 mM potassium acetate, 100 mM PIPES, 1 mM EGTA, 1 mM MgCl2. For ROCK inhibition, cells were treated with 10 µM Y27632 (688001, EMD Millipore) for 30 min prior. Cells were either imaged before and 2 min after triton extraction, or fixed 2 min after extraction followed by fixation for high-resolution imaging (see below). We note that this extraction buffer does not contain ATP. Therefore, a small fraction of the postextraction signal could come from monomeric myosin 2 entering a rigor state on actin in the absence of ATP.

Imaging

Observational imaging

Confocal images were collected on a Zeiss LSM 880 Airsycan confocal in SR mode with a 63 × 1.4 NA objective (Figures 1, 2, and 4). Three-slice z-stacks were collected at the ventral surface with optimal step size (0.17 µm). Images were processed in Zen Blue with “joint Deconvolution” processing using 10 iterations for EGFP channel and 20 iterations for JFx650-HaloTag channel. Maximum intensity orthogonal projections were generated using FIJI/ImageJ.

Quantitative triton assays

Cells were imaged with a 20 × 0.8 NA Plan-Apochromat objective (Zeiss) on a spinning disk confocal microscope from 3i (Intelligent Imaging Innovations) consisting of an Axio Observer 7 inverted microscope (Zeiss) attached to a W1 confocal spinning-disk (Yokogawa) with Mesa field flattening (Intelligent Imaging Innovations), and Prime 95B sCMOS (Photometrics) camera with the 488 (EGFP constructs) and 640 (HaloTag) lasers at 50% laser power and 500 ms exposure time, using a 20 × 0.8 NA Plan-Apochromat objective (Zeiss) (Figures 3 and 4). Preextraction images were acquired, media were removed, triton buffer was added and allowed to permeabilize cells for 2 min, and postextraction images were acquired. Following a series of postacquisition processing steps, as diagrammed and described in Supplemental Figure S3, post–triton-extraction intensity/pre–triton-extraction intensity ratios were calculated.

High-resolution imaging

After 2 min of triton extraction, Halo-2A fibroblasts expressing EGFP-2A constructs and stained with JFx650-HaloTag ligand were fixed (1X cystoskeletal buffer, 4% PFA, 0.5% Triton, 1.5% BSA) for 1 h at RT (Figure 5). Cells were stained for actin using Alexa Fluor 647 Phalloidin (A22287, Invitrogen) for 1 h at RT, washed in 1 × PBS, mounted in Prolong Gold (P36930, Invitrogen), and allowed to cure at RT for 24 h before imaging. An additional alignment sample was prepared with near-confluent untransfected fibroblasts without Halo dye that were fixed and stained with three phalloidins (A22287, A12379, and A34055, Invitrogen). Imaging was performed on a Zeiss SIM5 with a 63 × 1.4 NA objective and two Hamamatsu Fusion BT cameras. Z-stacks of ∼1 µm acquired at the ventral surface with optimal step size. Processing and alignment were performed in Zen Blue (version 3.10). Each raw channel image was processed with SIM² with three-dimensional, sampling of 2, auto sharpness, Wiener Filter, no regularization, three iterations, fast fitting, and processed images scaled to raw. Multiphalloidin images were collected with similar acquisition settings with optimized laser and exposure times, and identical processing. Alignment was performed with the Channel Alignment (Advanced) method with Roberts preprocessing, highest quality, translation plus rotation plus iso scaling, and linear interpolation.

Colocalization by cross-correlation

We followed published protocols (McCall, 2024) and available online support (https://imagej.net/plugins/colocalization-by-cross-correlation) to perform CCC using ImageJ/FIJI plugin. Notably, improved spatial resolution imaging has minimized the overlap of fluorescent signals. CCC tolerates imperfect overlaps to reveal correlations in signal, necessary as which we found more useful than traditional pixelwise colocalization tools (e.g., Pearson's). SIM images were used for all analyses. A mask of myosin-rich lamellar region to be analyzed and a background mask of cytoplasmic signal largely devoid of filamentous actin or myosin was manually identified. The mean intensity in the background region was subtracted from the respective channel in 32-bit images (to enable negative values in the analysis). CCC was performed with ImageJ/FIJI plugin to compare EGFP signal with endogenous Halo-2A signal and to compare EGFP signal with phalloidin-stained actin in the lamellar mask region. Confidence values are reported here, with additional analysis output available upon request. We followed the author's guidance to consider confidence values below 0.1 as not colocalized, values between 0.1 and 0.2 as likely colocalized, and values above 0.2 as a high confidence of colocalization.

GFP immunoprecipitation

Forty eight hours after transfection, cells were treated with 20 µM Y27632 (10005583, Cayman Chemical Company) for 1 h. Cells were collected and washed three times with 1 × PBS and lysed in high salt lysis buffer (50 mM Tris–HCl pH 8, 300 mM NaCl, 0.5 mM EDTA and, 0.5% Nonidet P40 Substitute) supplemented with 1× protease inhibitor cocktail (PI78442, Thermo Fisher Scientific). Lysates were kept on ice for 30 min and cleared by centrifugation (17,000 × g, 30 min at 4°C). Cleared lysates were incubated with 40 µl of equilibrated GFP-Trap agarose beads (gta, Chromotek) for 2 h at 4°C on a rotating wheel. Beads were washed three times with high salt-wash buffer (10 mM Tris/Cl pH 7.5, 300 mM NaCl, 0.05% Nonidet P40 Substitute, 0.5 mM EDTA). The samples were eluted in 100 µl of 2 × Laemmli buffer by boiling for 5 min at 95°C. Eluates were analyzed by Western blotting on 3 to 8% Tris acetate gel with indicated antibodies.

Statistical analysis

Statistical analyses were performed using Prism (GraphPad). Data are presented as mean ± SD, and comparisons were made using values from each independent experiment. We performed nested one-way ANOVA and Tukey's multiple comparison tests. The p values <0.05 were considered significant. *, p ≤ 0.05; **, p ≤ 0.01; ***, p ≤ 0.001; and ****, p ≤ 0.0001.

Abbreviations used:

ACD

assembly competence domain

EGFP

enhanced green fluorescent protein

ELC

essential light chain

FRAP

fluorescence recovery after photobleaching

IHM

interacting heads motif

MHC

myosin heavy chain

RLC

regulatory light chain

ROCK

rho-associated coiled-coil containing kinase

SIM

structured-illumination microscopy.

DATA AVAILABILITY

The data that support the findings of this study are available upon reasonable request from the corresponding author (J.R.B.).