Cellular Control of Cortical Actin Nucleation

1Here, we took an unbiased approach to study cortical actin nucleation. We used natural and induced cellular blebs as tools; expanding blebs are initially devoid of F-actin and progressively reassemble a contractile cortex prior to retraction, making them an ideal model to study de novo cortex assembly. Thus, we reasoned that the proteins necessary for the regrowth of cortical actin should be present in blebs. Cortex assembly could occur via elongation of F-actin seeds or mediated by nucleators. We first examined several seed elongation cortical growth mechanisms and concluded that these were not supported by experimental evidence (see Figure S1 available online). Therefore, we investigated the role of actin nucleators in cortex assembly using two independent unbiased approaches.

2First, we used proteomics on isolated cortices to identify the actin nucleators present in the cortex. To this aim, we separated blebs from constitutively blebbing M2 melanoma cells by mechanical shearing as previously described, a procedure that allows isolation of dynamic actin cortices (Figure 1A). We investigated the presence of actin nucleators in the actin-rich detergent-insoluble fraction of isolated blebs using mass spectrometry analysis. We detected the presence of only two actin nucleators: the formin mDia1 and the Arp2/3 complex (Figure 1B), consistent with some reports but in contradiction with others. All seven subunits of the Arp2/3 complex were detected along with the Arp2/3 nucleation-promoting factors cortactin and two subunits of the WAVE complex (SRA1 and NAP1).

3To verify these results using an independent approach, we undertook a screen based on mRNA expression profiles, protein localization, and small hairpin RNA (shRNA) depletion. First, we determined which nucleators were expressed in both M2 and HeLa cells and localized to their cortex, reasoning that basic mechanisms of cortex nucleation should be conserved across cell types. Quantitative PCR revealed that ten nucleators were expressed in both cell lines: the Arp2/3 complex subunits ACTR2 and ACTR3; the formins DAAM1, DIAPH1, DIAPH3, FHOD1, FMNL1, INF1, and INF2; and SPIRE1 and SPIRE2 (Figure S2A). To distinguish potential cortical actin nucleators, we examined the localization of the expressed nucleators in M2 cells using imaging assays, reasoning that proteins clearly localized to the cell periphery were good candidates for initiators of cortex growth, whereas nucleators strongly enriched in other locations were likely fulfilling other functions. Of the expressed nucleators, only Daam1, Fhod1, mDia1, and the Arp2/3 complex were found to localize to the cell periphery (Figures 1C–1F and S2B–S2I). Daam1 was present in the membrane of both expanding and retracting blebs (Figure S2B). Fhod1 accumulated at the actin cortex late in bleb retraction (Figure S2C). Immunostaining for mDia1 and the ARPC2 subunit of the Arp2/3 complex revealed clear localization to the cortex of M2 cell blebs (Figures 1C and 1E) and mitotic HeLa cells (Figures 1D and 1F).

4Having narrowed down potential candidates to mDia1, Fhod1, Daam1, and the Arp2/3 complex, we depleted these proteins with shRNA and used the size of blebs in M2 cells as reporters of perturbations to cortical actin. We reasoned that silencing of a cortical nucleator might lead to either larger blebs due to weakening of the cortex as a result of lower F-actin polymerization (as observed upon cytochalasin D treatment, Figure S2J) or smaller blebs due to disruption of contractility because of perturbation of cortical F-actin organization. Stable knockdown of Fhod1 did not lead to significant changes in bleb size in M2 cells (p = 0.5 compared to nonsilencing shRNA [NS]; Figures S2K, S2M, S2O, and S2Q; Table S1). Similar results were observed with transient transfection with two other shRNAs (∼50% mRNA depletion; data not shown). Stable depletion of Daam1 led to significant changes in bleb size distribution in one out of two knockdown lines (two different shRNAs, p = 0.03 and p < 0.001, respectively, compared to NS; Figures S2K, S2L, S2P, and S2Q; Table S1). Stable knockdown of mDia1 led to cells with significantly larger blebs (Figures 1G, 1J, S2K, and S2Q; Table S1; Movie S1). A similar phenotype was observed with transient transfection for three other shRNAs (Figure S2N). Stable knockdown of the Arp3 or Arp2 subunits of the Arp2/3 complex did not inhibit blebbing (consistent with) but led to cells with significantly smaller blebs (Figures 1H, 1K, 1L, S2K, and S2Q; Movie S2; Table S1). When we acutely perturbed Arp2/3 activity using the selective inhibitor CK666, we observed a phenotype with small blebs in ∼60% of cells (Figure 1I), consistent with depletion experiments, thus suggesting that CK666 is a good substitute for gene depletion. Finally, combined perturbation of mDia1 and Arp2/3 by gene depletion and CK666 treatment led to two distinct phenotypes: 34% of cells became amorphous and retained only a few discernible foci of cortical actin, while 30% of cells formed very large blebs (Figure 1M; t > 0 s; Movie S3). Together, these data suggest that in M2 cells, mDia1, Arp2/3, and Daam1 play an important role in the control of bleb size and may thus be involved in actin cortex nucleation.

5We then investigated whether the formins Daam1 and mDia1 were bound to cortical F-actin. In apparent contradiction with our immunostaining data (Figures 1E and 1F), we have previously reported cytoplasmic localization for GFP-tagged mDia1. Therefore, we tested whether this was due to a high mDia1 unbound/bound ratio by imaging mDia1 dynamics in live cells using single-molecule imaging, a technique that enables visualization of the localization of proteins with high cytoplasmic background. mDia1-GFP speckles clearly localized to the cortex of blebs (Figure 2A, arrowheads) and mitotic HeLa cells (Figure 2B), remaining visible for several seconds before disappearing. To determine whether cortical speckles represented active mDia1 proteins, we imaged the localization of constitutively active mDia1 (CA-mDia1; Figure 2C). In contrast to full-length mDia1 that displays cytoplasmic localization, CA-mDia1 exclusively localized to the plasma membrane during all stages of the bleb life cycle in M2 cells (Figures 2D and 2E) and at the cortex of prometaphase HeLa cells (Figure 2F). Actin depolymerization did not alter CA-mDia1 localization (Figure S3A), indicating F-actin-independent recruitment. We concluded that mDia1 speckles primarily represented active proteins localized to the cortex or the cell membrane. To determine whether mDia1 and Daam1 bind to the actin cortex, we compared the half-time of fluorescence recovery after photobleaching of CA-mDia1 and Daam1 using actin as a positive control. The GTPase-formin homology 3 (FH3) domain of mDia1 (GBD+FH3), which lacks an actin-binding domain but localizes to the cell membrane (similar to the GTPase-binding domain of Daam1; Figures 2C, S3B, and S3C), was used as a negative control. We reasoned that binding to cortical F-actin should stabilize the cortical localization of formins and therefore slow down their fluorescence recovery compared to GBD+FH3. CA-mDia1 fluorescence recovered at a rate similar to F-actin, but ∼4-fold slower than GBD+FH3 and Daam1 (Figures 2G and S3D). Together, these results indicate that CA-mDia1 is strongly associated with the actin cortex but that Daam1 is not, suggesting that Daam1’s effect on bleb size in one cell line was indirect, perhaps by modulating cytoplasmic pressure or downregulating adhesion.

6Overall, our independent screening and proteomic approaches identified the Arp2/3 complex and mDia1 as major contributors to the actin cortex, though contributions from Fhod1 and Daam1 cannot be entirely excluded, notably due to the incomplete mRNA depletion achieved in our experiments. We then focused on characterizing the specific functions of mDia1 and Arp2/3 in cortex generation, organization, and dynamics.

7We first investigated what portion of the cortical F-actin was nucleated by each nucleator. CK666 treatment resulted in an ∼60% decrease in cortical GFP-actin fluorescence intensity compared to DMSO (Figures 1I and 2H). A similar effect was observed when mDia1-depleted cells were treated with CK666 compared to DMSO (Figure 2H), indicating that Arp2/3 still contributes to cortex formation when mDia1 is depleted. Consistent with this result, immunostaining revealed the presence of Arp2/3 at the cortex of mDia1 knockdown cells (Figure S3E). Conversely, mDia1 was present at the cortex of Arp2/3-depleted and CK666-treated cells (Figures S3F–S3I), suggesting that the localization of both nucleators was independent from one another. Flow cytometry measurements of the total F-actin fluorescence intensity in cell populations indicated that perturbation to mDia1 and Arp2/3 activity led to significant decreases in total F-actin fluorescence (Figure 2I). When both were perturbed simultaneously, total cellular F-actin decreased additively. Because both nucleators had strong effects on cortex stability and were localized to the cortex, we assumed that the decrease in total F-actin intensity observed using flow cytometry was entirely imputable to changes in cortical actin. Because cortical F-actin contributes roughly 50% of total cellular F-actin fluorescence measured in flow cytometry (n = 33 cells examined; Figure S3J), the ∼27.5% decrease observed in mDia1-depleted cells treated with CK666 suggests that, together, mDia1- and Arp2/3-mediated actin polymerization account for at least 55% of cortical F-actin in M2 cells (Figure 2I). The remainder may be due to residual activity of mDia1 or Arp2/3 as a result of incomplete depletion/inhibition or contributions from other less abundant nucleators.

8We next used scanning electron microscopy to examine the impact of nucleator depletion on cortex organization in blebs, because we reasoned that the effect would be most apparent in structures necessitating de novo cortical actin polymerization. Depletion of mDia1 led to extensive changes in cortical actin organization, with areas of high filament density (Figure 3C; white arrow in Figure 3D) alternating with large gaps (100–200 nm) devoid of actin filaments (arrowhead in Figure 3D), very different from the homogeneous mesh with a gap size of ∼30 nm observed in control conditions (Figures 3A and 3B). Arp2/3 complex inhibition and depletion both led to apparently longer filaments, easily distinguishable compared to control cortices (Figure 3E; arrows in Figure 3F; Figure S3K). When both nucleators were perturbed simultaneously, the actin cortex consisted of a loose array of poorly interconnected filaments with a mesh size > 100 nm (Figure 3G; arrow in Figure 3H).

9Along with ultrastructural network organization, the dynamics of cortex assembly are a key determinant of the mechanical functions of the cortex in cell morphogenesis. Therefore, we examined how mDia1 and the Arp2/3 complex participated in setting the rate of de novo cortical actin accumulation in mitotic HeLa cells, where bleb formation can be induced in a controlled manner by laser ablation of the cortex (Figures 3I and 3J). De novo actin assembly in mDia1-depleted cells was 2-fold slower than in controls (Figures 3K and 3L). In contrast, perturbation of Arp2/3 activity by CK666 or ARPC2 small interfering RNA (siRNA) both led to an approximately 2-fold increase in actin assembly speed (Figures 3M and 3N). In control conditions, the actin assembly speed is the weighted average of the speeds of all nucleators. Thus, if a slower-than-average nucleator is depleted, assembly speed will increase, and conversely, if a faster-than-average nucleator is depleted, assembly speed will decrease. Hence, our results suggest that, consistent with rates of elongation measured for each nucleator in vitro, mDia1-mediated assembly of cortical actin is substantially faster than average (Figure 3K), whereas that mediated by Arp2/3 is substantially slower (Figure 3M). Taken together, our results indicate that perturbing Arp2/3 and mDia1 leads to profound changes in actin cortex organization and dynamics (Figure 3).

10Finally, we asked how interfering with the activity of each nucleator affected cell morphogenesis, focusing on cell division, where cortex reorganization is known to drive dramatic shape changes. Although previous studies indicate a role for mDia1 and Arp2/3 in mitosis, their role in assembling the mitotic cortex remains unclear. We found that both mDia1 and Arp2/3 localized to the cell cortex at all stages of mitosis in HeLa cells (Figures S4A and S3B). CK666 treatment or ARPC2 depletion did not significantly affect progression through cytokinesis (Figures 3N and S4C). However, ARPC2 depletion gave rise to increased cell blebbing in interphase, consistent with previous studies. In contrast, mDia1 depletion significantly affected progression past mitosis, and the cortex appeared patchy (Figures 4B, S4C, and S4D). Simultaneous depletion of mDia1 and inhibition of Arp2/3 significantly accentuated these effects (Figure S4C). When examined by fluorescence microscopy, the cortex of cells depleted in ARPC2 or treated with CK666 remained homogeneous (Figures 4C and 4D), whereas cells in which we perturbed both mDia1 and Arp2/3 possessed significantly less cortical actin, their cortex appeared fragmented, and cells blebbed vigorously at all stages of mitosis (Figure 4E). In cytokinesis, the polar cortex disappeared nearly entirely (Figure 4E) and cell shape was very unstable, a phenotype reminiscent of that observed in M2 blebbing cells (Figure 1M). These experiments suggest that mDia1 and Arp2/3 play different roles in mitosis progression.

11Our results demonstrate a crucial role for the formin mDia1 and the Arp2/3 complex in nucleating the submembranous F-actin cortex in M2 blebbing cells and HeLa cells. Although our experiments identify mDia1 and the Arp2/3 complex as major contributors to nucleation of the submembranous actin cortex, other studies suggest a possible role for different formins. This, together with the greater tissue specificity of formins compared to the ubiquitous expression of the Arp2/3 complex, suggests that the importance of mDia1 may vary between cell types. Future studies will have to examine the generality of the role of mDia1 and explore potential complementary contributions from other nucleators, such as Fhod1, Daam1, and mDia2, to cortical actin generation. Whether mDia1 and Arp2/3 act independently from one another in the cortex remains unclear. Reports of synergistic action of pointed-end nucleators with barbed-end nucleators suggest that mDia1 could elongate new branches nucleated by Arp2/3, or that Arp2/3 could catalyze formation of branches from the side of filaments nucleated by mDia1. Some of our data support independent action. When we reduced the activity of each nucleator separately, M2 cells retained a well-defined cortex, but when we inhibited both simultaneously, most of the cortex disappeared (Figures 1G–1I and 1M). Furthermore, loss of F-actin due to simultaneous perturbation of mDia1 and Arp2/3 was comparable to the sum of the losses due to perturbation of each separately (Figure 2I). Finally, depletion of one nucleator did not affect localization of the other (Figures S3E–S3I). Together, these data suggest that the cortex is composed of two actin networks that grow concurrently but largely independently from one another. However, other observations point to synergistic action. During mitosis, mDia1 depletion resulted in an increase in cell division failures, and whereas Arp2/3 inhibition alone had no effect on cell division, it potentiated the effect of mDia1 depletion (Figures 4 and S4C). This suggests that mDia1 may act upstream of Arp2/3 in nucleating cortical actin. Future experiments will need to investigate the level of interdependency between mDia1- and Arp2/3-mediated actin assembly in the cortex, and whether this dependency is cell-cycle dependent.

12Why generation of the cortex and other cellular F-actin structures, such as the leading edge of migrating cells, requires the combined action of two F-actin nucleators represents an intriguing question. Our ultrastructural examination suggests that the actin networks assembled by each nucleator (Figures 3A–3H) are not spatially separated as in the leading edge but rather intertwined. Furthermore, perturbation of each nucleator had distinct effects on organization and kinetics (Figure 3), suggesting that each contributes differently to assemble a cortical F-actin network with the requisite mechanical properties. Nucleators likely affect cortical F-actin density, thickness, and organization, all of which influence the emerging physical properties of the cortical network. Understanding how each nucleator contributes to cortical mechanics from the molecular level up will represent a challenge for the future.