The actin nucleation factors JMY and WHAMM enable a rapid Arp2/3 complex-mediated intrinsic pathway of apoptosis

The actin cytoskeleton is a well-known player in most vital cellular processes, but comparably little is understood about how the actin assembly machinery impacts programmed cell death pathways. In the current study, we explored roles for the human Wiskott-Aldrich Syndrome Protein (WASP) family of actin nucleation factors in DNA damage-induced apoptosis. Inactivation of each WASP-family gene revealed that two of them, JMY and WHAMM, are necessary for rapid apoptotic responses. JMY and WHAMM participate in a p53-dependent cell death pathway by enhancing mitochondrial permeabilization, initiator caspase cleavage, and executioner caspase activation. JMY-mediated apoptosis requires actin nucleation via the Arp2/3 complex, and actin filaments are assembled in cytoplasmic territories containing clusters of cytochrome c and active caspase-3. The loss of JMY additionally results in significant changes in gene expression, including upregulation of the WHAMM-interacting G-protein RhoD. Depletion or deletion of RHOD increases cell death, suggesting that RhoD normally contributes to cell survival. These results give rise to a model in which JMY and WHAMM promote intrinsic cell death responses that can be opposed by RhoD.


Author summary
The actin cytoskeleton is a collection of protein polymers that assemble and disassemble within cells at specific times and locations. Cytoskeletal regulators called nucleation factors ensure that actin polymerizes when and where it is needed, and many of these factors are members of the Wiskott-Aldrich Syndrome Protein (WASP) family. Humans express 8 WASP-family proteins, but whether the different factors function in programmed cell death pathways is not well understood. In this study, we explored roles for each WASPfamily member in apoptosis and found that a subfamily consisting of JMY and WHAMM is critical for a rapid pathway of cell death. JMY-mediated actin assembly in the cytoplasm is necessary for its pro-apoptotic function. Furthermore, the loss of JMY results in changes in gene expression, including a dramatic upregulation of the small G-protein RhoD, which appears to contribute to cell survival. Collectively, our results point to the a1111111111 a1111111111 a1111111111 a1111111111 a1111111111 Introduction acetyltransferase p300, and p53 [24,31]. Co-overexpression of JMY and p300 in p53-proficient epithelial cells increases the transcription of BAX [24,29], while introduction of a JMY siRNA in p53-overexpressing cells decreases the amount of BAX induction [29]. These data indicate that JMY can promote cell death by enhancing the transcription of at least one pro-apoptotic gene. Consistent with this conclusion, an siRNA targeting JMY reduces the amount of sub-G1 (presumably apoptotic) cells after UV treatment [27]. Although, under other conditions, RNAi of JMY can increase the number of sub-G1 cells [32].
Collectively, these studies on WAVE1 and JMY reveal an interesting dynamic between actin assembly proteins and apoptosis-associated processes, and suggest that the relationships between nucleation factors and cell death pathways merit further investigation. In the current study, using a systematic gene knockout approach, we tested roles for all WASP-family members in DNA damage-induced apoptosis. We discovered that JMY and WHAMM are key regulators of a rapid p53-dependent and Arp2/3-mediated cell death pathway, but that their proapoptotic functions are opposed by the small G-protein RhoD.

The WASP-family proteins JMY and WHAMM enable DNA damageinduced apoptosis
To evaluate the impact of WASP-family members on apoptosis, we employed a panel of knockout (KO) HAP1 or eHAP fibroblast-like cell lines lacking N-WASP, WAVE1, WAVE2, WAVE3, the WAVE Complex, the WASH Complex, WHAMM, JMY, or Cortactin (S1 Table and S1 Fig). The parental cells, derived from a chronic myeloid leukemia patient [33], are useful for studying apoptosis because they harbor a well-defined immortalizing BCR-ABL fusion [34], have easily manipulatable nearly-haploid (HAP1) or fully-haploid (eHAP) genomes [33,35], and are p53-proficient [36]. To examine intrinsic apoptotic responses, the parental and KO cells were treated with etoposide, a topoisomerase II inhibitor that induces dsDNA breaks. To monitor cell death, each sample was then imaged using fluorescent Annexin V (AnnV) to identify the apoptotic hallmark of phosphatidylserine externalization, propidium iodide (PI) to assess membrane permeability, and hoechst to visualize nuclear fragmentation (Figs 1A and S1).
After 6h of etoposide treatment, approximately 33% of parental HAP1 or eHAP cells were apoptotic, as determined by AnnV-positive staining, compared to only 3% of DMSO-treated controls ( Fig 1B). The proportion of cells displaying AnnV staining was not significantly different across the majority of the etoposide-or DMSO-treated WASP-family knockout lines compared to the parental lines (Figs 1B and S1). In one of three etoposide experiments, a statistically significant increase in AnnV-positive cells was observed for the WAVE1 KO line ( Fig  1B), a phenotype reminiscent of previous studies in which WAVE1 depletion in leukemia cells increased apoptosis [22,23]. In contrast, across all three etoposide experiments, two cell linesthe JMY KO and the WHAMM KO -exhibited significantly lower percentages of AnnV-positive cells (Fig 1A and 1B). The JMY KO and WHAMM KO cells also showed substantially less frequent PI staining and nuclear fragmentation compared to parental cells (Figs 1A and S2). These results indicate that among a set of cell lines lacking WASP-family members, only those missing JMY or WHAMM have significantly impaired apoptotic responses to DNA damage.
JMY and WHAMM are approximately 35% identical and 50% similar to one another [37], and comprise a subgroup within the WASP family. Since both appear to be important for apoptosis, we sought to more thoroughly compare the effects of JMY or WHAMM inactivation on cell death. For this, we characterized three independent JMY mutant cell lines derived from HAP1 cells and two independent WHAMM mutant cell lines derived from eHAP cells (Figs 1C and S3). Etoposide treatment caused similarly high levels of DNA damage across all of the parental and knockout samples, as evidenced by increased phosphorylated histone H2AX (γH2AX) protein levels and greater numbers of γH2AX foci at DNA breaks within nuclei (S4 Fig). Both the parental cell lines, as well as the JMY KO and the WHAMM KO lines, showed etoposide dose-dependent increases in AnnV-positive apoptotic cells, but the mutants displayed significantly lower percentages at virtually every concentration (Fig 1D). At the highest concentrations (4-5μM), nearly 40% of parental cells, but only 8-12% of JMY KO cells and 15-20% of WHAMM KO cells, were AnnV-positive (Fig 1D).
To assess early versus late apoptosis, the percentage of AnnV-positive cells without versus with PI staining was quantified (S2 Fig). Among the~40% of parental cells exhibiting AnnV staining, about 25% were AnnV-positive and PI-negative, indicating early apoptosis, while 15% were AnnV and PI double-positive, signifying late apoptosis (Fig 1E). In comparison, all of the JMY KO and WHAMM KO samples contained significantly fewer cells in both early and late apoptosis, with the JMY KO cells exhibiting 8% early and 2% late, and the WHAMM KO cells displaying 15% early and 5% late (Fig 1E). These results indicate that JMY and WHAMM each play important roles in apoptosis, but the more extreme phenotypes observed when JMY is mutated suggest that JMY is more prominent in enabling cell death.

Inactivation of JMY or WHAMM impairs multiple steps of intrinsic apoptotic signaling
Since the loss of JMY or WHAMM decreased the final apoptotic readouts of phosphatidylserine externalization and membrane permeability, we next wanted to determine if the inactivation of JMY or WHAMM affected earlier aspects of apoptotic signaling. Intrinsic pathways of apoptosis are characterized by the export of apoptogenic proteins from mitochondria into the cytosol upstream of the initiation of a caspase cleavage cascade. So we examined whether the loss of JMY or WHAMM affected the release of cytochrome c (cyto c), a protein that is maintained in the mitochondrial intermembrane space under normal conditions, but is exported to trigger the cytosolic assembly of macromolecular structures called apoptosomes, which serve as platforms for the activation of initiator caspases during intrinsic apoptosis [38,39]. In parental HAP1 and eHAP cells, as expected, mitochondrial-independent cyto c localization became apparent following etoposide treatment (Fig 2A and 2B). Categorization of cyto c localization as discrete mitochondrial versus diffuse cytosolic was confirmed by plotting the fluorescence intensity profiles of mitochondria and cyto c (Fig 2C and 2D). It was also notable that diffuse cytosolic cyto c staining was sometimes accompanied by non-mitochondrial clusters of cyto c puncta (described in a later section). Consistent with the previously-observed amounts of AnnV-positive cells, >30% of parental cells exhibited cytosolic cyto c localization patterns following 6h of etoposide treatment (Fig 2E). In contrast, the JMY KO and WHAMM KO cell lines had significant delays in cyto c release, and by 6h only 10% of JMY KO and around 15% of WHAMM KO cells showed cytosolic cyto c staining ( Fig 2E). These findings suggest that JMY and WHAMM are necessary for an efficient intrinsic apoptosis pathway, and that part of their contributions to apoptotic responses may occur prior to and/or during the accumulation of cyto c in the cytosol.
After mitochondrial cyto c release, apoptotic signaling is typified by the multimerization and proteolytic processing of initiator caspases in the apoptosome that cleave and activate executioner caspases, which in turn target multiple proteins in the cytosol and nucleus [40,41]. We next sought to measure the effects of JMY and WHAMM on the caspase cascade by examining both the expression and cleavage of representative initiator and executioner caspase proteins via immunoblotting. Control DMSO-treated parental, JMY KO , and WHAMM KO cells expressed indistinguishable levels of inactive initiator procaspase-9 and inactive executioner procaspase-3 ( Fig 3A and 3B), suggesting that the mutant cell lines did not have any inherent deficits in the steady-state abundance of these core components of the apoptotic machinery. However, upon etoposide treatment for 3h and 6h, the parental, JMY KO , and WHAMM KO cell lines displayed very different kinetics in the conversion of inactive procaspases to cleaved caspases. Whereas HAP1 and eHAP cells began to show processing of procaspases to cleaved caspases by 3h of etoposide treatment, the cleaved caspases were generally not detectable in the (D) Lines were drawn through the magnified images using ImageJ to measure the pixel intensity profiles for cyto c and mitochondria, highlighting the differences in scoring mitochondrial versus cytosolic cyto c localization. (E) The % of cells with cytosolic cyto c staining was calculated in ImageJ by counting cells that exhibited cytosolic rather than mitochondrial cyto c staining and dividing by the total number of DAPI-stained nuclei. Each point represents the mean % ±SD from 3-4 experiments (n = 533-856 cells per point). Significance stars refer to comparisons of parental to KO samples at the depicted timepoints. �� p<0.01; ��� p<0.001 (ANOVA, Tukey post-hoc tests).
https://doi.org/10.1371/journal.pgen.1009512.g002  KO , eHAP, and WHAMM KO cells were treated with DMSO for 6h or 5μM etoposide for 3h or 6h, and extracts were immunoblotted with antibodies to caspase-9 (Casp-9 Pro and Casp-9 Cleaved ), caspase-3 (Casp-3 Pro and Casp-3 Cleaved ), and tubulin. (C) Key for color-coded samples in panels (D-I). (D-E) For quantification, the caspase cleavage ratio was calculated by dividing the cleaved band intensity by the procaspase band intensity. Each point represents the mean ratio ±SD from 3 experiments. AU = Arbitrary Units. (F) Cells were treated with DMSO or etoposide, and stained with caspase-3/7 green detection reagent to label cleaved DEVD (green) and Hoechst to stain DNA (blue). Images are from the 6h timepoint. Scale bars, 100μm. (G) The % of cleaved DEVD-positive cells was calculated in ImageJ by counting cells that exhibited green nuclear fluorescence and dividing by the total number of Hoechst-stained nuclei. Each point represents the mean ±SD from 3 experiments (n = 2,741-6,646 cells per point). (H) Cells treated as above were fixed and stained with an antibody that JMY KO and WHAMM KO cells until 6h (Fig 3A and 3B). Even at the later timepoint, caspase-9 cleavage was 2-3-fold lower in both JMY KO and WHAMM KO cells compared to the parental cell lines (Fig 3C-3E). For caspase-3, cleavage was 10-fold less efficient in JMY KO cells and 3-fold less efficient in WHAMM KO cells relative to their parental lines (Fig 3C-3E). Together, these results show that following acute genotoxic damage, JMY and WHAMM are each required for the rapid activation of caspase cleavage cascades.
During the terminal execution stage of apoptosis, caspase-3 and caspase-7 cleave the motif Asp-Glu-Val-Asp (DEVD) [41], so we used a DEVD substrate peptide conjugated to a fluorescent reporter to quantify the amount of cells with active caspase-3/7 following treatment with etoposide. Caspase-3/7 activation was readily detectable in parental cells at 3h, but did not become clearly apparent in the knockout cells until the 6h timepoint, when 50% of parental cells versus only 15-25% of JMY KO or WHAMM KO cells were positive for cleaved DEVD ( Fig  3F and 3G). To further validate these results, we stained cells with antibodies that specifically recognize the active cleaved form of caspase-3. Immunofluorescence microscopy revealed that nearly 40% of parental cells contained bright clusters of active caspase-3 staining, whereas <10% of the JMY KO and <20% of the WHAMM KO cells stained positive for active caspase-3 ( Fig 3H and 3I). Therefore, the apoptotic deficiencies arising from JMY or WHAMM mutations include significantly delayed and less potent activation of executioner caspases, with JMY knockout cells exhibiting more extreme defects, particularly in caspase-3 activation.

Cells lacking JMY and/or WHAMM are defective at apoptosis in several cellular contexts
Because the loss of either JMY or WHAMM results in multiple apoptotic deficits, we next sought to determine the extent to which apoptosis would proceed in cells lacking both proteins. So we derived two independent WHAMM/JMY double knockout (DKO) cell lines from WHAMM KO-2 cells (S3 Fig). Similar to the other knockout cells, both WHAMM/JMY DKO cell lines had low levels of DNA damage when exposed to DMSO, but more numerous DNA breaks when incubated with etoposide (S4 Fig). After etoposide exposure and staining with fluorescent AnnV, significantly fewer WHAMM/JMY DKO cells were apoptotic compared to control cells, as only 10% of WHAMM/JMY DKO cells were AnnV-positive (Fig 4A and 4B). This frequency of apoptosis was noticeably lower than that of the single WHAMM knockouts and more closely resembled that of the single JMY knockouts (Fig 1E).
To directly compare the efficiency of apoptosis in the parental cells, single JMY or WHAMM mutants, and JMY/WHAMM double mutants, we stained each of these cell populations for active cleaved caspase-3. While nearly 40% of parental HAP1 or eHAP cells contained active caspase-3 after etoposide treatment, 20% of WHAMM KO cells and <10% of JMY KO or WHAMM/JMY DKO cells showed active caspase-3 staining (Fig 4C and 4D). Thus, inactivation of JMY in a WHAMM-deficient background does not result in any synthetic or synergistic death phenotypes, but simply reduces apoptosis levels down to that of a single JMY mutant. This frequency, typically 8-10% of the population, seems to be the lowest achievable amount of apoptotic death in HAP1 derivatives under the genotoxic conditions used in our experiments. recognizes caspase-3 cleaved at Asp175 (Cleaved Casp-3; green) and DAPI (DNA; blue). Images are from the 6h timepoint. (I) The % of cleaved caspase-3-positive cells was calculated in ImageJ by counting cells that contained bright clusters of green fluorescence and dividing by the total number of DAPI-stained nuclei. Each point represents the mean ±SD from 3 experiments (n = 1,136-7,844 cells per bar). Significance stars refer to comparisons of parental to KO samples. � p<0.05; �� p<0.01; ��� p<0.001 (ANOVA, Tukey post-hoc tests). https://doi.org/10.1371/journal.pgen.1009512.g003

PLOS GENETICS
To next determine the extent to which transient depletion, rather than permanent deletion, of JMY, WHAMM, or both, affects apoptosis, we treated HAP1 and eHAP cells with two independent siRNAs targeting either JMY (Fig 4E)   RNAi-mediated depletion of JMY in WHAMM KO cells resulted in significantly fewer AnnVpositive cells compared to control siRNA treatments (Fig 4E and 4F), a phenotype resembling that of the WHAMM/JMY DKO lines. Overall, the more severe apoptotic deficiencies that arise when JMY is permanently deleted or transiently depleted suggest that JMY is the nucleation factor with the most influential role in programmed cell death.
For further gauging the effects of JMY on apoptosis in different human cell lines, we targeted JMY using siRNAs in U2OS osteosarcoma cells and HeLa adenocarcinoma cells (S5

JMY functions in p53-dependent apoptosis after a p21-associated proliferation arrest
To begin to understand the underlying mechanisms that give rise to the apoptotic pathway defects in JMY-and WHAMM-deficient cells, we sought to determine the importance of p53 in the death of parental HAP1 and eHAP cells. First, to verify that apoptosis was dependent on p53, we treated these parental cell lines with three independent siRNAs targeting TP53 and measured the percentage of cells with AnnV staining. Compared to samples that were treated with a control siRNA, samples in which p53 levels were diminished contained significantly fewer AnnV-positive cells (Figs 5D, 5E, and S6). Moreover, greater degrees of p53 knockdown resulted in larger reductions in apoptosis, as the amount of p53 protein was positively correlated with the percentage of AnnV-positive cells (Figs 5F and S6). Extrapolation of the p53 trendline down to zero protein expression suggested that approximately 12% of HAP1 cells completely lacking p53 would undergo apoptosis (Figs 5F and S6), nearly matching the percentage observed for JMY-depleted samples ( Fig 5C). These results demonstrate that the majority of etoposide-induced apoptosis in HAP1 cells requires both p53 and JMY.
For evaluating the impact of p53 levels on apoptosis in the absence of JMY or WHAMM, we then targeted TP53 in WHAMM KO and JMY KO cells. Depletion of p53 in WHAMM KO cells resulted in significantly fewer AnnV-positive cells (S6 Fig), consistent with the idea that WHAMM inactivation causes only a partial inhibition of p53-dependent apoptosis which can be rendered complete by the subsequent removal of p53. Conversely, depletion of p53 in the JMY mutant background revealed that the already low apoptosis levels following etoposide treatment could not be reduced below 10% (Fig 5D-5F). Together, the JMY and p53 depletion studies support the conclusion that JMY and p53 function in the same intrinsic apoptotic death pathway.
The observation that HAP1 cells undergo apoptosis in a primarily p53-dependent manner, combined with the finding that JMY can associate with p53 and its acetyltransferase cofactor p300 [24], led us to next investigate the specific influence of JMY on p53 functions. Since apoptosis is typically accompanied by p53 protein stabilization, post-translational modification, and re-localization into the nucleus [25,26], we examined these properties. At steady-state, HAP1 and JMY KO  In response to genotoxic damage, nuclear p53 alters transcription and can trigger cell cycle arrest, DNA repair, apoptosis, senescence, and other stress responses [42,43]. Because a proliferation arrest is usually an early response to DNA damage, we compared the growth and death rates of HAP1 and JMY KO cultures after exposure to etoposide. For these experiments, we treated each cell line with DMSO or etoposide for 6h, removed the solvent or drug and replaced them with fresh media, and then quantified the numbers of total cells (live and dead) at regular intervals up to a 48h endpoint. While HAP1 and JMY KO cells multiplied at equivalent rates after treatment with DMSO, both cell types stopped proliferating after treatment with etoposide (S8 Fig). In contrast, when apoptotic cell quantities were measured using AnnV staining, the proportion of cells undergoing apoptosis was significantly higher in the HAP1 samples at every timepoint (S8 Fig). For parental samples, >35% were AnnV-positive by 6h, 50% were apoptotic by approximately 10h, and the frequency of apoptosis had reached about 90% of cells by 48h (S8 Fig). JMY KO samples never hit the 50% apoptosis mark, as only 35% were AnnV-positive at 48h (S8 Fig). Thus, without JMY, cells remain stuck in an arrested condition and fail to shift into a proper death signaling state.
One of the key p53 targets that triggers cell cycle arrest is CDKN1A, which encodes the cyclin-dependent kinase inhibitor, p21 [44]. To determine if CDKN1A expression is affected by JMY inactivation in the absence or presence of etoposide, we used RT-PCR for comparing CDKN1A transcript levels. Consistent with the proliferation assays described above, CDKN1A was expressed at low levels in DMSO-treated HAP1 and JMY KO cells, and significantly upregulated in both cell lines following a 6h exposure to etoposide (S8 Fig). Similarly, small amounts of p21 protein were found in HAP1 and JMY KO cells at steady-state, while treatment with etoposide resulted in a 3-fold increase in the abundance of nuclear p21 in both sets of cells (S8 Fig). Collectively, these experiments suggest that JMY specifically promotes a cell suicide program and is not required for several other aspects of nuclear p53 modification or function, including the transcriptional responses that lead to a proliferation arrest.

Expression of the small G-protein RhoD is turned on in JMY-knockout cells
While inactivation of JMY did not prevent arrest or CDKN1A upregulation, it could still impact other aspects of transcriptional programming. Indeed, given that JMY can enhance the expression of the pro-apoptotic p53 target BAX [24,29], one plausible explanation for the apoptotic defects in JMY KO cells could be that they possess different gene expression patterns than normal cells such that they are less 'equipped' to die. To characterize the transcriptomic changes that took place upon knocking out JMY, we performed differential gene expression analyses using RNA-sequencing (RNA-seq) on HAP1 cells and on one of the JMY knockout cell lines (JMY KO-1A ). When comparing the JMY KO-1A cells to parental HAP1 cells, <0.36% of protein-coding genes displayed expression differences of at least 2-fold and with a significance q-value of <0.05 (Figs 6A and S9). Expression of the genes encoding p53, caspases, Bax and other Bcl-2-family members, additional key modulators of apoptosis, or JMY-interacting proteins such as p300, Strap, and Mdm2 were not significantly different (S10 Fig). Moreover, genes for WASP-family members and other actin nucleation factors were not substantially changed (S10 Fig). Therefore, mutating JMY does not appear to adversely affect the basal expression of canonical apoptosis regulators or factors that are structurally or functionally related to JMY.
Since important differences in gene expression might not become evident until cells experience a pro-apoptotic stimulus, we next compared the transcriptomic changes that occurred in HAP1 and JMY KO cells after etoposide exposure. When comparing etoposide-treated to control samples, <0.37% of protein-coding genes in either cell line displayed expression differences of at least 2-fold and with a significance q-value of <0.05 (Figs 6B, 6C, and S9). Of the 67 total genes that met such criteria, 48 were shared between the HAP1 and JMY KO cells (S9 Fig).
The mutually-upregulated factors included the CDKN1A cell cycle inhibitor described above, NFκB signaling components, and AP-1 transcription factors (S9 Fig). Substantial changes in expression of genes that encode p53, caspases, Bcl-2 family members, or other prototypical apoptotic modulators were again not detected. Among the 19 genes that showed statistical differences between the parental and knockout cells, 16 displayed similar trends in either up-or down-regulation. The only sizable discrepancies between the two cell lines were in the cytokine genes IL8 and TNF, as IL8 was strongly upregulated in one of the HAP1 samples and TNF was greatly upregulated in the JMY KO samples (Figs 6B, 6C, and S9). These results imply that HAP1 and JMY KO cells are sufficiently equipped with enough potentially-apoptotic factors to be able to execute an intrinsic death program without a major increase in the abundance of pro-apoptotic transcripts.
A closer inspection of known p53 targets in the RNA-seq datasets indicated that BAX transcript levels were unchanged after etoposide treatment in both HAP1 and JMY KO cells (S10 Fig). Under the same genotoxic conditions, the BBC3 transcript, which encodes the p53

PLOS GENETICS
upregulated modulator of apoptosis PUMA, was increased 2.5-fold in HAP1 cells and 5-fold in the JMY KO , although neither of these values reached statistical significance (S10 Fig). To more specifically test whether changes in the amounts of particular transcripts could be detected by a targeted approach, we performed RT-PCRs for BAX and BBC3, as well as the anti-apoptotic factor BCL-2. Compared to untreated steady-state conditions, etoposide exposure resulted in slightly increased BAX and BBC3 transcript levels and decreased BCL-2 levels in both HAP1 and JMY KO-1A cells ( Fig 6D). Therefore, mutating JMY does not appear to drastically affect the expression patterns of canonical p53 target genes, further suggesting that transcriptional reprogramming is not the principal driver of death signaling under the conditions used in our experiments.
In light of these observations, we sought to better understand how unanticipated factors might be contributing to cell survival or death in the presence and absence of JMY. When examining the 51 genes whose expression differed in the JMY KO line at steady state, it was notable that the most upregulated gene was RHOD (Figs 6A and S9), which encodes a small Gprotein that was previously shown to interact with WHAMM [45,46]. Transcript levels for other small G-proteins were virtually the same between parental and KO cells (S10 Fig), implying that the change for RHOD might be functionally meaningful. Moreover, RT-PCR experiments confirmed that RHOD transcript levels were almost undetectable in HAP1 cells and significantly turned on in the JMY KO cells (Fig 6E and 6F). The expression of WHAMM and RAB1A/RAB1B, the latter of which encode small G-proteins known to bind WHAMM directly [47], were unchanged between the parental and JMY KO cells (Figs 6E and S10). RT-PCR further showed that one of the other two independently derived JMY knockout cell lines also had increased RHOD transcript levels ( Fig 6G), suggesting that RHOD upregulation may be a frequent, but not universal, response to the inactivation of JMY. These experiments reveal that although the loss of JMY may not cause any obvious compensatory changes in the transcription of actin nucleation factors, it can result in increased expression of a G-protein known to affect cytoskeletal dynamics.

RhoD promotes cell survival in the absence or presence of JMY
RhoD is a Rho-family GTPase that regulates many cellular functions, including actin assembly during filopodia formation, cell migration, endosome dynamics, and Golgi trafficking [45][46][47][48][49][50][51][52][53][54][55][56]. Additionally, RhoD appears to participate in other processes that affect cell proliferation, such as cell cycle regulation and cytokinesis [49,57]. Since JMY is involved in proteostasis and genotoxic stress responses through its activities in autophagy and apoptosis, we reasoned that the increase in RHOD expression in JMY KO cells might be part of a compensatory pro-survival mechanism that allows the cells to better tolerate the loss-of-function mutation in JMY. To explore this possibility, we targeted RhoD for depletion using two independent siRNAs in JMY KO cells. RT-PCRs verified that each RhoD siRNA reduced RHOD transcript levels (S11 Fig). DMSO-or etoposide-treated JMY KO cells were then subjected to AnnV, PI, and Hoechst staining for measuring apoptosis and for evaluating necrosis (S11 Fig). Compared to control siRNA-treated samples, RHOD-depleted samples that were exposed to just DMSO showed modest increases in the fraction of cells with AnnV staining (S11 Fig), consistent with the possibility that cells are slightly more prone to apoptosis when RHOD levels are decreased. For the samples that were treated with etoposide, the low frequency of AnnV staining of JMY KO cells was equivalent whether or not RHOD was silenced (S11 Fig). Since diminishing RHOD expression in these experiments did not restore apoptosis to normal levels, the intrinsic apoptotic defects in JMY KO cells must not be due simply to the upregulation of RHOD. Perhaps more importantly, the percentage of AnnV-negative cells that were PI-positive, or considered to be necrotic, was significantly higher when RHOD was depleted (S11 Fig). This latter result supports the idea that RhoD normally promotes cell survival.
Because of the complexities in studying RhoD activities during apoptosis in JMY KO cells, which appear to be fundamentally defective in their intrinsic apoptotic responses, we next wanted to characterize the effects of transiently depleting or permanently inactivating RHOD in otherwise healthy JMY-proficient cells. While RHOD transcript is below the limit of detection in the HAP1 cell line, it is expressed at measurable levels in eHAP cells (Fig 7A). So eHAP samples were treated with siRNAs for depleting RHOD (Fig 7B), exposed to DMSO or etoposide, and stained with AnnV and PI to assess the proportion of apoptotic and necrotic cells (Fig 7C). For the DMSO-treated populations, apoptotic death rose from 1% in cells receiving the control siRNA to 5% in cells receiving either of the RhoD siRNAs (Fig 7D), suggesting that RhoD depletion in healthy cells leads to a higher incidence of death under normal growth conditions. For the etoposide-treated samples, RHOD silencing caused apoptosis to occur in approximately 80% of eHAP cells, frequencies which were approximately double what was observed in cells treated with a negative control siRNA (Fig 7C and 7D). Moreover, the amount of RhoD transcript was inversely correlated with the percentage of AnnV-positive cells ( Fig 7E). Necrotic death frequencies also rose, in this case from 2% in the presence of RHOD to 5% when it was silenced ( Fig 7D). Thus, decreasing RHOD expression can increase stress-induced apoptosis and necrosis when normal JMY and p53 responses are present, further strengthening the conclusion that RhoD plays a basic pro-survival role in cells.
To assess the impact of a permanent loss of RHOD on cell survival and death, we next derived two independent RhoD KO cell lines from parental eHAP cells (S3 Fig). RT-PCRs verified that wild type RhoD mRNA was absent in the mutant cells ( Fig 7F). Similar to other knockout cells, each of the RhoD KO cell lines had low levels of DNA damage when exposed to DMSO, but more numerous DNA breaks when incubated with etoposide (S4 Fig). Consistent with the transient depletion findings in etoposide-treated samples, apoptotic and necrotic death were more common in RhoD KO cells than in eHAP cells (Figs 7G and S11). Additionally, in dose-response experiments, at each etoposide concentration the RhoD KO cell lines showed a death frequency that was approximately twice that of parental cells (Fig 7H). These studies reveal that the permanent loss of RhoD makes cells more susceptible to dying during normal culture conditions and also enhances their apoptotic responses following acute DNA damage.
Since JMY and WHAMM are necessary for an efficient pathway of cyto c release, initiator caspase cleavage, and executioner caspase activation, we tested if RHOD inactivation affected these aspects of intrinsic apoptosis. Following 3h and 6h of etoposide treatment, compared to the eHAP parental cell line, the RhoD KO samples had significantly higher percentages of cells with diffuse cyto c localization, greater levels of initiator caspase-9 processing, more cells undergoing cleavage of DEVD-containing substrates, and more cells staining positive for active caspase-3 (Figs 7I-7K and S11). Together, the above results show that knocking out RHOD has the opposite apoptotic phenotypes of knocking out JMY or WHAMM. Thus, while JMY and WHAMM are WASP-family proteins that have important pro-apoptotic roles in intrinsic pathways of cell death, such functions may be constrained by pro-survival or antiapoptotic activities of the small G-protein RhoD.

The Arp2/3 complex is required for efficient intrinsic apoptosis
The actions of Rho-family G-proteins and WASP-family nucleation factors converge on the activation of the Arp2/3 complex, which nucleates actin into filaments [58,59]. To better understand the contribution of the actin cytoskeleton in intrinsic apoptosis, we first tested the importance of the Arp2/3 complex by treating HAP1 and eHAP cells with siRNAs targeting two Arp2/3 subunits, ACTR3 and ARPC4, whose depletion results in loss of the entire complex [37]. Similar to the knockdowns of p53, JMY, or WHAMM described earlier, Arp2/3 depletions resulted in significantly fewer AnnV-positive cells (Fig 8A and 8B) and significantly fewer cells exhibiting active cleaved caspase-3 staining (Fig 8C). Further, in HAP1 and eHAP cells, greater degrees of Arp2/3 complex knockdown resulted in larger reductions in apoptosis, and the amount of Arp3 protein was positively correlated with the percentage of AnnV-positive cells (Fig 8D). To validate the apparent requirement for the Arp2/3 complex in apoptosis, we also performed etoposide treatments in the presence of the small molecule Arp2/3 inhibitor, CK666. Consistent with the RNAi experiments, CK666-treated samples contained fewer active caspase-3-positive cells than did DMSO-treated control samples (Fig 8C).
To determine the effect of the Arp2/3 complex on apoptosis in the absence of JMY or WHAMM, we similarly performed Arp2/3 depletions in JMY KO and WHAMM KO cells ( Fig  8A). Analogous to the observation that knockdown of p53 in JMY KO cells did not further diminish apoptosis levels from their already low amount of~10% (Fig 5D-5F), the depletion of Arp2/3 in this knockout background did not affect AnnV or active caspase-3 staining frequencies (Fig 8B and 8C). In contrast, Arp2/3 silencing in WHAMM KO samples did result in significantly fewer apoptotic cells (Fig 8A-8C). One interpretation of these observations is that among the 20% of WHAMM-deficient cells that successfully execute apoptosis, approximately half rely on a death mechanism involving JMY and Arp2/3. Overall, it appears that the major intrinsic apoptotic pathway requires p53, JMY, and the Arp2/3 complex, and can be enhanced by WHAMM.

JMY-driven actin assembly is associated with cyto c clustering and caspase-3 activation
JMY activates the Arp2/3 complex using a C-terminal domain consisting of 3 actin-binding WASP-homology-2 (W) motifs and an Arp2/3-binding Connector-Acidic (CA) motif, and can also use its WWW segment to nucleate actin directly [60]. To investigate whether the actin polymerization capacity of JMY is important for its participation in apoptosis, we generated 3 GFP-tagged JMY derivatives for performing rescue experiments in JMY KO cells: a wild type JMY construct (JMY WT ); a JMY mutant missing all 3 W motifs (JMY ΔWWW ) that can neither nucleate actin directly nor cooperate with the Arp2/3 complex; and a JMY mutant lacking the significantly non-zero. (F) RNA from eHAP and RhoD KO cells was subjected to RT-PCR using primers to RHOD (forward primer: nt 124-147; reverse primer: nt 523-545), RHOD (forward primer: nt 225-247; reverse primer: nt 523-545), and GAPDH. Nucleotide 147 is predicted to be absent in mRNA from both knockouts, whereas nucleotide 247 should be present in both, albeit reflecting transcript levels that are lower than the parental control due to nonsense-mediated decay. Bands from this gel are also shown in panel (A). (G) eHAP and RhoD KO  CA segment (JMY ΔCA ) that cannot activate Arp2/3 but retains its actin nucleating WWW region. The GFP-JMY plasmids, or a vector control, were introduced into the JMY KO-1A line, which was then cultured in selective media, exposed to etoposide, and subjected to immunoblotting ( Fig 8E) or immunostaining (Fig 8F) for active cleaved caspase-3. Compared to the vector-transfected knockout cells, which contained low levels of cleaved caspase-3, cells transfected with the wild type JMY construct showed a 4.5-fold increase in the amount of cleaved caspase-3 and a 3.5-fold increase in the cellular frequency of cleaved caspase-3 staining ( Fig  8G), indicating that full-length JMY can restore intrinsic apoptosis functions to JMY KO cells. In contrast, the JMY mutant with the WWW segment deleted failed to rescue apoptosis in the knockout cells, as cleaved caspase-3 levels were indistinguishable from those of the vectortransfected controls (Fig 8E and 8G). The JMY mutant lacking the CA portion was also ineffective at increasing the frequency of cleaved caspase-3 staining (Fig 8G), and actually decreased the total levels of active caspase-3 in JMY KO cells (Fig 8E and 8G). These results show that JMY requires its actin assembly activity, particularly its Arp2/3 complex binding segment, in order to drive intrinsic apoptosis.
JMY is primarily cytoplasmic in healthy cells [27,28], and is typically redistributed into both the cytoplasm and nucleus upon exposure to apoptosis-inducing stressors [29,30]. Consistent with this expectation, GFP-JMY exhibited a fairly uniform localization in the cytosol and nucleus (Fig 8F). Interestingly, however, GFP-JMY additionally appeared in intense juxtanuclear clusters (Fig 8F), and those cytoplasmic structures overlapped with clusters of cleaved caspase-3 ( Fig 8F). Further, staining with fluorescent phalloidin demonstrated that filamentous-(F-) actin also assembled throughout the caspase-3 region and was most intense adjacent to JMY (Fig 8F). JMY ΔCA and JMY ΔWWW could also be found in bright cytoplasmic spots, but cleaved caspase-3 staining was diffuse within these areas and F-actin was not enriched ( Fig  8F). These findings highlight important relationships between JMY-mediated actin polymerization and the formation of cytoplasmic territories containing active caspase-3.
Because caspase-3 activation is driven by apoptosomes, macromolecular platforms consisting of cyto c, procaspase-9, and other scaffolding factors [38,61,62], we revisited the potential connections among JMY, cyto c, and actin in the cytoplasm. Reminiscent of the GFP-JMY and cleaved caspase-3 clusters, punctate juxtanuclear structures of endogenous JMY were visible in a subset of etoposide-treated HAP1 cells (Fig 9A and 9B). The same type of staining pattern was also observed in 15-20% of U2OS cells, which are larger and easier to image (Fig 9A and  9B). Interestingly, in both HAP1 and U2OS cells, many of these JMY puncta colocalized with cyto c puncta that were distinct from mitochondria (Figs 9C, 9D, and S12). They were also associated with bright clouds of the cleaved DEVD reporter (prior to its nuclear import), indicative of active caspase-3 being present in proximity to these JMY-and cyto c-containing Magnifications represent examples of a JMY WT cluster that overlaps with a cleaved caspase-3 cluster and is associated with an F-actin-rich territory (i), and a JMY ΔCA structure lacking both cleaved caspase-3 and F-actin enrichment (ii). Scale bars, 25μm, 10μm. (G) For quantification of (E), the cleaved caspase-3 band intensities were normalized to the loading control intensities and the vector sample was set to 1. Each bar represents the mean intensity ±SD from 4 experiments. AU = Arbitrary Units. For quantification of (F), the % of cleaved caspase-3-positive cells was calculated and each bar represents the mean ±SD from 3 experiments (n = 3,447-6,084 cells per bar). �� p<0.01, ��� p<0.001 (t-tests).
https://doi.org/10.1371/journal.pgen.1009512.g008  Fig 9E). To test whether JMY was required for cells to form the distinct cytoplasmic cyto c structures, we quantified the proportion of HAP1 and JMY KO cells containing mitochondria-independent cyto c puncta. While the fraction of parental cells harboring cyto c puncta increased during 3h and 6h exposures to etoposide, the JMY KO lines did not form cytosolic cyto c puncta under any of these conditions (Fig 9F). In addition, inspection of U2OS cells revealed that the overlap of JMY and cyto c puncta took place within F-actin-rich territories (Figs 9G and S12). F-actin appeared to be specifically reorganized, rather than globally polymerized or upregulated, as the inactivation of JMY did not significantly affect the overall levels of actin filaments, stained with phalloidin, or the total amounts of actin in the cytoplasm or nucleus, stained with an actin antibody (S12 Fig). These observations are consistent with a role for cytoplasmic JMY and F-actin during the period of apoptosis encompassing apoptosome assembly and executioner caspase-3 activation.

PLOS GENETICS
Finally, we investigated whether the formation of the cytosolic JMY puncta was impacted by WHAMM or RhoD. After etoposide treatment, 20% of HAP1 and eHAP cells contained clusters of JMY puncta while only 10% of WHAMM KO cells but more than 35% of RhoD KO cells displayed such cytosolic JMY puncta (Fig 10A and 10B). These phenotypic differences are reminiscent of the reduced versus increased apoptotic characteristics previously observed in the WHAMM and RhoD knockout cells, respectively. To address how overexpression of RhoD might influence the formation and organization of the JMY-containing structures, we transiently transfected U2OS cells with constructs encoding GFP-tagged wild type RhoD (RhoD WT ) or a GDP-locked dominant negative mutant RhoD (RhoD DN ) (Fig 10C). JMY puncta were formed at indistinguishable frequencies in cells expressing either GFP-RhoD WT or the GFP vector control (Fig 10D), suggesting that elevating RhoD protein levels does not inherently disturb this pattern of JMY localization during apoptotic signaling. In contrast, only half as many cells expressing the RhoD DN mutant contained JMY puncta ( Fig 10D). Additionally, this dominant interfering RhoD protein increased JMY levels in the nucleus (Fig 10C-10E), signifying that the nucleotide-bound state of RhoD affects both the formation of the punctate cytosolic structures and the nuclear accumulation of JMY. Further examination of GFP-RhoD localization and Factin staining revealed that RhoD WT surrounded the clusters of JMY puncta and allowed actin polymerization (Fig 10C and 10E). Conversely, RhoD DN did not localize in analogous juxtanuclear regions or permit actin filaments to form there (Fig 10C and 10E), suggesting that RhoD-GDP can act as an 'off switch' for JMY-mediated cytoskeletal rearrangements. Collectively, these results indicate that although JMY and RhoD exhibit opposing pro-apoptotic versus pro-survival activities, both proteins apparently function in the assembly and organization of cytoplasmic territories containing cyto c, active caspases, and F-actin.

Discussion
The polymerization, organization, and turnover of actin filaments have been thoroughly characterized during many cellular functions that maintain viability [9,10]. However, the mitochondria, highlighting the differences in cyto c maintained within mitochondria (M) and cytosolic cyto c that colocalizes with JMY puncta (P). (E) HAP1 cells were treated with 5μM etoposide for 6h, stained with caspase-3/7 green detection reagent to label cleaved DEVD, fixed, and stained with antibodies to JMY (green), cyto c (magenta), and DAPI. Scale bars, 25μm, 10μm. Magnifications represent examples of JMY and cyto c puncta that overlap with an area containing cleaved DEVD. (F) HAP1 and JMY KO cells were treated with DMSO or etoposide before being fixed and stained to visualize AIF and cyto c. The % of cells with non-mitochondrial cyto c puncta was calculated in ImageJ. Each point represents the mean % ±SD from 3 experiments (n = 194-332 cells per point). Significance stars refer to comparisons of parental to KO samples at the depicted timepoints. (G) U2OS cells were treated with etoposide, fixed, and stained with antibodies to JMY (green) and cyto c (magenta), and with phalloidin (F-actin; yellow). Magnifications (i) depict clusters of JMY, cyto c, and F-actin. Scale bars, 25μm, 10μm. ��� p<0.001 (ANOVA, Tukey post-hoc tests).

PLOS GENETICS
contributions of actin assembly factors to apoptotic cell death are not well understood. Among the best-known cytoskeletal features of apoptosis are the actin filament rearrangements and disassembly that accompany changes in cell adherence and morphology [7,8]. Actin itself can be cleaved by caspases [63,64], resulting in non-polymerizable fragments that promote apoptotic cellular phenotypes [65][66][67]. The actin turnover machinery has also been implicated in controlling apoptosis at multiple steps. Cofilin, which depolymerizes actin filaments [68], influences the early stages of apoptosis, as it translocates to mitochondria and may increase the release of cytochrome c [69][70][71] and the localization of p53 [72]. Actin is additionally recruited to mitochondria at around the time of mitochondrial permeabilization [73][74][75]. During the later stages of apoptosis, the F-actin-severing protein gelsolin [76] is cleaved by caspases, resulting in a fragment with unregulated severing activity that drives depolymerization [77] and allows chromatin fragmentation [78]. In contrast to these described roles for actin disassembly proteins in apoptosis, less is known about the actions of the actin assembly machinery during cell death. Our results reveal intrinsic apoptotic functions for JMY, WHAMM, the Arp2/3 complex, and RhoD, and create a framework for understanding how actin nucleation factors and small G-proteins control pro-apoptotic and pro-survival processes.
WASP-family proteins have many well-recognized activities, but the extent to which they participate in cell death pathways is relatively underexplored. WAVE1 has been one of the most studied members, as it can affect the localization or modification of Bcl-2-family proteins. Several transient overexpression and depletion experiments suggest that WAVE1 limits mitochondrial permeabilization [20][21][22][23]. But in our initial knockout screen using BCR-ABLimmortalized human cells, the permanent loss of WAVE1 did not cause a consistent increase in the frequency of apoptosis, so the degree to which WAVE1 acts as an anti-apoptotic factor in a cell type specific manner requires further investigation.
In addition to WAVE1, JMY has been studied in the context of apoptosis. JMY was discovered as a cofactor that enhanced the transcriptional activity of p53 [24], and subsequent RNAi studies indicated that JMY has pro-survival [32] or pro-death [27,29,79] functions under different experimental circumstances. Our work advances the understanding of its pro-death role by demonstrating that cells with permanent deletions or transient depletions of JMY exhibit significant defects in apoptotic processes following DNA damage. In fact, inactivation of each WASP-family gene revealed that only cells lacking JMY, or its closest homolog WHAMM, had substantial deficiencies in their intrinsic apoptotic responses. WHAMM did not have any previously described roles in apoptosis, so the latter observation adds a third WASP-family member to the repertoire of actin assembly factors that can function in programmed cell death.
WHAMM and JMY comprise a subgroup within the WASP family, are approximately 50% similar to one another, and both promote anterograde membrane transport [37,47,80] and autophagy [32,81,82]. Although the two proteins participate in similar cellular activities, they exhibit some key molecular differences. WHAMM has a conventional C-terminal WWCA domain that activates Arp2/3 complex-mediated actin assembly [37], whereas JMY has an extended WWWCA domain wherein the WCA produces Arp2/3-branched networks, and the WWW directly nucleates linear filaments [60]. WHAMM localizes to the ER and cis-Golgi and binds microtubules to mediate membrane tubulation and transport [37,47,83], while JMY acts later in the secretory pathway to promote trafficking away from the trans-Golgi [80]. WHAMM and JMY also both function in multiple aspects of autophagy. WHAMM binds to the phospholipid PI(3)P and stimulates the Arp2/3 complex to accelerate lipidation of LC3-family proteins during autophagosome biogenesis [82]; WHAMM additionally drives Arp2/3-dependent autophagosome movement [81] and autolysosome remodeling [84]. JMY binds to LC3 and polymerizes actin at autophagosomes during their maturation [32] and actin-based motility [28]. While it is reasonable to assume that the functions of WHAMM and JMY in secretion and autophagy help sustain cells during normal growth and allow them to adapt to conditions of nutritional stress, the fact that both factors also enable cell death indicates that they function as pivotal players in the cellular responses to multiple other stressors including genotoxic insults.
Following etoposide-induced DNA damage, parental HAP1 cells executed an immediate p53-dependent cell death program. In contrast, JMY or WHAMM knockout cells showed significant reductions in the terminal apoptosis phenotypes of phosphatidylserine externalization and membrane permeabilization. Further comparisons of JMY, WHAMM, and WHAMM/ JMY knockout cells revealed that an individual WHAMM deficiency results in partial apoptotic deficits, whereas JMY ablation, either alone or in conjunction with WHAMM, causes a more severe defect in apoptosis-one equivalent to a loss of p53.
Importantly, cells lacking JMY or WHAMM displayed inefficiencies and/or kinetic delays in multiple apoptotic processes, including the mitochondrial export of cyto c, caspase activation, and effector caspase cleavage of target proteins. These results indicate that both cytoskeletal regulators are key participants in a rapid, intrinsic, p53-dependent cell death pathway, with JMY acting as the more prominent contributor. They further suggest that JMY and WHAMM function at one or more steps in the apoptotic pathway that take place after the recognition of DNA damage and arrest of cell proliferation, including during transcriptional reprogramming, mitochondrial permeabilization, apoptosome assembly, and caspase cleavage/activation.
Under normal cell culture conditions, JMY is found predominantly in the cytosol [27,28], but in response to DNA damage, it accumulates in the nucleus while still maintaining a cytosolic presence [29,30]. At steady-state, WHAMM is mostly associated with cytoplasmic membrane-bound organelles, although it displays a nuclear localization in a small subpopulation of cells [37]. Such observations are consistent with the existence of both cytoplasmic and nuclear roles for these proteins in apoptosis. JMY was previously shown to interact with the stressresponse protein Strap, the acetyltransferase p300, and p53 [24,31]. p300 is one of many cofactors known to modify p53 to promote the transcription of genes necessary for initiating cell cycle arrest and stimulating pro-death pathways in response to cell damage [26,[85][86][87]. While the precise function of JMY in a complex with Strap, p300, and p53 is unknown, co-overexpression of JMY with p53, or with p53 plus p300, augments BAX transcription in several epithelial cell contexts [24,29,79], and introduction of a JMY siRNA into p53-overexpressing cells limits BAX induction [29]. However, our current work suggests that the activity of JMY in transcriptional regulation via p53 may not be the primary apoptotic driver following DNA damage.
p53 shuttles between the cytoplasm and nucleus, and although it has pro-apoptotic functions in the cytosol [88,89], its most extensively characterized activities are as a nuclear transcription factor [43,90,91]. In our experiments, etoposide-induced increases in the overall abundance, nuclear accumulation, phosphorylation, and acetylation of p53 occurred in the presence or absence of JMY. Moreover, JMY inactivation had little effect on transcriptional upregulation of the p53 target CDKN1A, which encodes the key cell cycle inhibitor p21. In fact, JMY-deficient cells exhibited a prolonged arrest in response to DNA damage, indicating that JMY is not necessary for the gene expression changes that stop proliferation. These results collectively support the idea that JMY participates in only a subset of the p53-mediated transcriptional responses to genotoxic stress.
Given the speed and efficiency with which p53-and JMY-proficient parental cells underwent apoptosis, we hypothesized that they constitutively express a pool of potentially apoptotic factors that enable the cells to respond rapidly to DNA damage. Indeed, global and targeted gene expression profiling experiments implied that inactivation of JMY did not adversely affect the expression of common apoptosis regulators either before or after etoposide exposure.
However, it was striking that the JMY knockout cells experienced a significant upregulation of RHOD, which encodes a small G-protein previously shown to interact with WHAMM [45,46]. RhoD has wide-ranging functions in regulating actin dynamics at the plasma membrane, endosomes, and Golgi, and in influencing cell proliferation [45][46][47][48][49][50][51][52][53][54][55][56][57]. While the selective pressures that led to greater RHOD expression in two out of the three JMY knockout cell lines are unknown, we nevertheless reasoned that increased RhoD levels may be part of a compensatory pro-survival mechanism that allows cells to better tolerate a permanent JMY mutation. Consistent with this possibility, RhoD depletion or deletion experiments resulted in more necrosis and enhanced apoptotic responses-the opposite phenotype of JMY-or WHAMM-depleted cells. These findings reveal complex relationships among two actin nucleation factors and one small G-protein during cell death and survival (S12 Fig). Given that the best understood molecular behaviors of JMY and WHAMM reside in their abilities to promote actin assembly in the cell cytoplasm, we tested whether the Arp2/3 complex was also important for the execution of apoptosis. Indeed, RNAi-mediated depletion or pharmacological inhibition of the Arp2/3 complex also impaired apoptosis, and JMY truncations that cannot activate Arp2/3 failed to increase the apoptotic capacity of JMY knockout cells. Most notably, wild type and actin polymerization-deficient mutants of JMY were both able to congregate into discrete cytosolic structures, but only the polymerization-proficient protein generated distinct juxtanuclear zones containing actin filaments, active caspase-3, and cytochrome c. In light of these observations, we propose that JMY-mediated actin assembly is required for creating cytoplasmic territories that direct the activation of executioner caspase-3 by cytochrome c-containing apoptosomes (S12 Fig). The structural and biophysical properties of apoptosomes have begun to come into focus [38,61,92], yet their cell biological characteristics remain less clear. Since JMY clusters form less frequently in WHAMM KO cells and with different frequencies depending on RhoD and its nucleotide-bound state, it will be important to gain a better understanding of the physical connections among RhoD, WHAMM, JMY, Arp2/3, and actin, and how they impact the organization and function of apoptosomes and caspases.
Continuing to define the roles of different elements of the actin assembly and regulatory machinery in cellular adaptations to endogenous or exogenous stress could provide new avenues for understanding tumorigenesis. The Arp2/3 complex and WASP-family members influence cell motility [13,14], and elevated levels of nucleation factors are associated with increased metastasis [93,94], indicating that such actin cytoskeletal proteins have proto-oncogenic features. Similarly, the functions of WHAMM and JMY in autophagy [32,81,82] could enable cancer cells to better survive in diverse physiological environments and in response to chemotherapy [95,96]. However, our current results describing apoptotic requirements for JMY and WHAMM support the idea that these factors also possess key tumor-suppressive features. TP53 is well known as the most commonly mutated gene in human cancers, and mutations that inactivate p53 or other p53-associated proteins are linked to poor prognoses [97][98][99]. Interestingly, JMY expression also appears to be lost in several B-cell lymphomas and invasive carcinomas [100]. Given the new positions of JMY and WHAMM as important players in p53-dependent apoptotic pathways, a greater understanding of how their actin nucleation, proto-oncogenic, and tumor-suppressive activities are coordinated will likely shed further light on how programmed cell death mechanisms are impacted by the cytoskeleton.

Ethics statement
Research with biological materials in the Campellone Lab was approved by the UConn Institutional Biosafety Committee. This study did not include research with human subjects or live animals. Human cell lines were acquired from Horizon Genomics or the UC Berkeley cell culture facility as described below.

Biological materials
Cell lines are listed in S1 Table, nucleic acids are listed in S2 Table, and key reagents are listed in S3 Table. HAP1 cells are nearly-haploid fibroblast-like cells that contain an immortalizing BCR-ABL fusion and a single copy of all chromosomes except for a heterozygous 30Mb fragment of chromosome 15, which is integrated within the long arm of chromosome 19 [33,101]. This diploid portion encompasses 330 genes, including WHAMM. CRISPR/Cas9-engineered eHAP cells were derived from the HAP1 line and are fully haploid [35]. In the current study, CRISPR/Cas9-mediated recombination using guide RNAs against target genes resulted in frameshift and/or splicing mutations (S1 Table and  , while eHAP cells were used for mutagenesis of WHAMM and RHOD. The latter two cell lines were made in an eHAP background due to the WHAMM diploidy in HAP1 cells, and because under normal culture conditions RHOD transcript was detectable in eHAP but not HAP1 cells (Horizon Genomics). One of the WHAMM mutant cell lines (WHAMM KO-2 ) was studied previously [82], and was used to make the WHAMM/JMY DKO cell lines. Attempts to generate WASH1 KO cells were unsuccessful, as all viable clones contained in-frame mutations (Horizon Genomics). HAP1 cell derivatives were cultured in Iscove's Modified Dulbecco's Medium (IMDM) supplemented with 10% fetal bovine serum (FBS) and penicillin-streptomycin. U2OS and HeLa cells (UC Berkeley cell culture facility) were cultured in Dulbecco's Modified Eagle's Medium (DMEM) supplemented with 10% FBS and antibiotic-antimycotic. All cell lines were grown at 37˚C in 5% CO 2 . All assays were performed using cells that had been in active culture for 2-10 trypsinized passages.

Transgene expression
For JMY cloning (S2 Table), DNA fragments were amplified from cDNA templates by PCR, digested, and ligated into the BamHI and NotI restriction sites of pKC425 [102]. JMY WT contains amino acids 1-983 of murine JMY, JMY ΔWWW contains amino acids 1-856/934-983, and JMY ΔCA contains amino acids 1-941. For RhoD cloning (S2 Table), a BamHI-EcoRI restriction fragment from GST-RhoD [45] was inserted into the BglII and EcoRI sites of pKC-LAP-C1 [37]. RhoD WT contains amino acids 1-210 of murine RhoD, and RhoD DN harbors a T31N mutation that was introduced by PCR-based site-directed mutagenesis. Plasmids were maintained in E.coli XL-1 Blue (Stratagene). For rescue studies, JMY KO-1A cells cultured in 12-well plates were transfected with 2μg of linearized plasmid encoding GFP or GFP-JMY derivatives using LipofectamineLTX (Invitrogen). 24h later, the cells were transferred to media containing 1.5mg/mL G418 for 12-16 days. Surviving colonies were collected, expanded in 24-well plates, 6-well plates, and ultimately T-25 flasks prior to cryopreservation. Upon reanimation, G418 concentrations were reduced stepwise to 1.0mg/mL, 0.5mg/mL, and 350μg/mL in 6-well plates. The cell populations were subjected to experimental manipulations within 5 passages in media containing 350μg/mL G418. For transient expression, U2OS cells were transfected in 6-well plates with 500ng of DNA prior to reseeding onto 12mm glass coverslips in 24-well plates and fixing as described below. Transfection increased cell death in

Chemical treatments and RNA transfections
Cells were treated with different concentrations of etoposide (Sigma Aldrich) diluted from an initial stock of 10mM in DMSO or with 100μM CK666 (Millipore) from a 100mM stock. Equivalent volumes of DMSO were used as controls. For RNA-seq experiments, the control condition consisted of regular media, while the treated condition consisted of media in which the etoposide powder was dissolved directly to 5μM. For JMY rescue experiments, to be able to detect cleaved caspase-3 in the JMY KO cell line, etoposide was dissolved to 10μM. For RNAi experiments, cells were grown in 6-well plates for 24h, transfected with 40nM siRNAs (S2 Table) using RNAiMAX (Invitrogen), incubated in growth media for 24h, reseeded into 6-well plates and 24-well glass-bottom plates (MatTek), and incubated for an additional 24h. Cells cultured in 6-well plates were collected and processed for immunoblotting or RT-PCR, and cells cultured in 24-well plates were used in live fluorescence microscopy assays.

RT-PCR and RNA-sequencing
For RT-PCRs, approximately 10 6 cells were seeded in 6-well plates, and for RNA-seq 2.5x10 6 cells were seeded in 6cm dishes. After 24h of growth, cells were subjected to control-, DMSO-, or etoposide-containing media changes and rinsed with phosphate buffered saline (PBS). RNA was isolated using TRIzol reagent (Ambion), followed by a chloroform extraction, isopropanol precipitation, and 75% ethanol wash before resuspending RNA in water. For RT-PCRs, RNA was reverse transcribed into cDNA using Superscript III RT (Invitrogen) and subsequently amplified using Taq polymerase (New England Biolabs) and gene-specific primers (S2 Table). Primers were designed to amplify~230-460bp products from each cDNA template. After agarose gel electrophoresis, band intensities for CDKN1A, WHAMM, and RHOD were quantified using ImageJ software [103] and normalized to β-ACTIN and/or GAPDH bands for each sample. For RNA-seq, Illumina cDNA library preparation was based on the Illumina TruSeq Stranded mRNA sample preparation guide. Total reads from RNA-seq were aligned with TopHat, and differential expression was calculated with the CuffLinks/CuffDiff program. Gene expression values were given as Fragments Per Kilobase of transcript per Million mapped reads (FPKM), differential expression was calculated as the Log2(FPKM ratio JMY KO-1A :HAP1), and data from 3 independent experiments were merged. When comparing the transcriptomic changes that occurred in HAP1 or JMY KO-1A lines upon etoposide treatment, differential expression was calculated as the Log2(FPKM Etoposide:Control) for each cell line and data from 2 independent etoposide experiments were merged. Volcano plots were generated by plotting the -Log10(q-value) against Log2(FPKM ratio) using Microsoft Excel software. The data summarized in this publication can be found in Supporting Information described below.

Immunoblotting and quantification
Detached cells were collected from 6-well plates and combined with adherent cells in PBS containing EDTA, centrifuged, washed with PBS, and centrifuged again to ensure collection of all live and dead material. Cell pellets were resuspended in lysis buffer (20mM HEPES pH 7.4, 100mM NaCl, 1% Triton X-100, 1mM Na 3 VO 4 , 1mM NaF, plus 1mM PMSF, and 10μg/ml each of aprotinin, leupeptin, pepstatin, and chymostatin), diluted in SDS-PAGE sample buffer, boiled, centrifuged, and subjected to SDS-PAGE before transfer to nitrocellulose membranes (GE Healthcare). Membranes were blocked in PBS + 5% milk (PBS-M) before being probed with primary antibodies (S3 Table) diluted in PBS-M overnight at 4˚C plus an additional 2-3h at room temperature. Membranes were rinsed twice with PBS and washed thrice with PBS + 0.5% Tween-20 (PBS-T). Membranes were then probed with secondary antibodies conjugated to IRDye-800, IRDye-680, or horseradish peroxidase (S3 Table) and diluted in PBS-M. Membranes were again rinsed with PBS and washed with PBS-T. Blots were imaged using a LI-COR Odyssey Fc imaging system. Band intensities were determined using the Analysis tool in LI-COR Image Studio software, and quantities of proteins-of-interest were normalized to tubulin, actin, and/or GAPDH loading controls.

Apoptosis and caspase activation assays
For live fluorescence-based assays, approximately 2.5x10 5 cells were seeded into 24-well glassbottom plates and allowed to grow for 24h prior to DMSO or etoposide treatments. For endpoint apoptosis assays, Alexa488-AnnexinV (Invitrogen), Propidium Iodide (Invitrogen), and Hoechst (Thermo Scientific) were added directly to the media (S3 Table) and incubated for 15min at 37˚C in 5% CO 2 . For caspase activation assays, CellEvent Caspase-3/7 Detection Reagent (Invitrogen) and Hoechst were added to the media and incubated for 30min. All live imaging was performed at 37˚C as described below. For mitochondrial visualization in Fig 9, MitoTracker Red (Invitrogen) was added for 30min prior to fixation.

Immunostaining assays
For immunofluorescence microscopy, approximately 2.5x10 5 cells were seeded onto 12mm glass coverslips in 24-well plates and allowed to grow for 24h. After DMSO or etoposide treatments, cells were washed with PBS, fixed in 2.5% or 3.7% paraformaldehyde (PFA) in PBS for 30min, washed, permeabilized with 0.1% TritonX-100 in PBS, washed, and incubated in blocking buffer (1% FBS + 1% bovine serum albumin (BSA) + 0.02% NaN 3 in PBS) for a minimum of 15min. Cells were probed with primary antibodies (S3 Table) diluted in blocking buffer for 45min. Cells were washed and treated with AlexaFluor-conjugated secondary antibodies, DAPI, and/or AlexaFluor-conjugated phalloidin (S3 Table) for 45min, followed by washes and mounting in Prolong Gold anti-fade reagent (Invitrogen).

Fluorescence microscopy
All live and fixed images were captured using a Nikon Eclipse Ti inverted microscope with Plan Apo 100X/1.45, Plan Apo 60X/1.40, or Plan Fluor 20x/0.5 numerical aperture objectives using an Andor Clara-E camera and a computer running NIS Elements software. Live cell imaging was performed in a 37˚C chamber (Okolab). All cells were viewed in multiple focal planes, and Z-series were captured at 0.2-0.4μm steps. Images presented in the figures represent one slice or two-slice projections. All images were processed and/or analyzed using Ima-geJ [103].

Image processing and quantification
The ImageJ Cell Counter plugin was used for live apoptosis or caspase-3/7 activation assays, and for fixed cytochrome c, cleaved caspase-3, or JMY staining assays by manually counting the total number of cells in the Hoechst or DAPI channel and the number of cells that were positive for AnnV fluorescence, AnnV fluorescence plus nuclear PI fluorescence, nuclear PI fluorescence, nucleic acid-associated DEVD reporter fluorescence, cytosolic non-mitochondrial cytochrome c staining, clustered cleaved caspase-3 staining, and cytosolic punctate cytochrome c or JMY staining. For analyses of nuclear γH2AX, p21, JMY, or actin levels, the Threshold, Watershed, and Analyze tools were used in the DAPI channel to separate individual nuclei, and the ROI Manager Measure tool was used in the γH2AX, p21, JMY, or actin channels to measure the mean fluorescence intensity per nucleus. To count the number of γH2AX foci per nucleus, the Find Maxima and ROI Manager Measure tools were used in the γH2AX channel. For analyses of total cellular p53, JMY, or actin levels, the Selection tool was used in the phalloidin channel to select individual cells, and the Measure tool was used in the p53, JMY, or actin channel to measure the mean fluorescence intensity per cell. For analyses of cytoplasmic JMY or actin levels, the nuclear intensity was subtracted from the total intensity for individual cells. Pixel intensity plots were generated using the Plot Profile tool. Lines were drawn through mitochondria-or JMY-rich regions after identification in the AIF/MitoTracker or JMY channels.

Reproducibility and statistics
All conclusions were based on observations made from at least 3 separate experiments and quantifications were based on data from 3-5 representative experiments, except where noted in the Figure Legends. To capture a breadth of apoptotic phenotypes, multiple fields-of-view containing cells at different densities were imaged. The sample size used for statistical tests was the number of times an experiment was performed, except where noted in the Legends. Statistical analyses were performed using GraphPad Prism software. Statistics on data sets with 3 or more conditions were performed using ANOVAs followed by Tukey's post-hoc test unless otherwise indicated. P-values for data sets comparing 2 conditions were determined using unpaired t-tests unless otherwise noted. P-values <0.05 were considered statistically significant. Numerical data underlying graphs or summary statistics are shown in S1 and S2 Datasets.
Supporting information S1   (HAP1, eHAP) and WASP-family knockout (JMY KO-1A , WHAMM KO-2 , WAVE1 KO ) cells were treated with DMSO or 5μM etoposide for 6h and stained with Alexa488-AnnV (green), PI (magenta), and Hoechst (DNA; blue). Scale bar, 100μm. These panels display the individual grayscale channels that comprise the merged images in Fig 1. (B-C) Representative examples of AnnV-negative/PI-negative (AnnV -PI -) non-apoptotic, AnnV-positive/PI-negative (AnnV + PI -) early apoptotic, AnnV/PI double-positive (AnnV + PI + ) late apoptotic, or AnnV-negative/PI-positive (AnnV -PI + ) necrotic cells are shown. Arrowheads in C highlight examples of hoechst-stained DNA condensation and nuclear fragmentation in HAP1 cells. Scale bars, 25μm.  KO cells were treated with DMSO or 5μM etoposide for 6h before washout (triangles) and replacement with regular media. Samples were stained with Alexa488-AnnV, PI, and Hoechst at the indicated time points. In (A-B), the total # of cells (live and dead) was counted in ImageJ and normalized to the # at 0h for each sample. The total # of cells is displayed as the fraction of live AnnV-negative (AnnV -) or apoptotic AnnV-positive (AnnV + ) cells (n = 1,461-9,108 cells per bar for DMSO; n = 1,415-2,992 cells per bar for etoposide). Panels (A) and (B) depict the same data in two different formats. In (C), the % of AnnV-positive cells was calculated and each point represents the mean ±SD from 3 experiments. Significance stars refer to comparisons among the etoposide-treated samples. RU = Relative Units. (D) HAP1 and JMY KO cells were treated with DMSO or etoposide for 6h before collecting RNA and performing RT-PCR with primers for CDKN1A and GAPDH. Agarose gel band intensities were quantified in ImageJ, and values for CDKN1A were normalized to GAPDH and plotted in the adjacent bar graph. AU = Arbitrary Units. (E) Cells were treated with DMSO or etoposide for 6h before being fixed and stained with a p21 antibody (green) and DAPI (DNA; blue). Scale bar, 30μm. (F) Nuclear p21 fluorescence intensity was measured in ImageJ, and each point represents the mean ±SD from 3 fields-of-view in a representative experiment (n = 216-418 cells per point). �� p<0.01; ��� p<0.001 (ANOVA, Tukey post-hoc tests). were treated with control siRNAs or independent siRNAs for the RHOD gene before performing RT-PCRs with primers to RHOD, GAPDH, and β-ACTIN. (B) Cells were treated with DMSO or 5μM etoposide for 6h and stained with Alexa488-AnnV, PI, and Hoechst. The % of AnnV-positive cells was calculated and displayed as the fraction of AnnV-positive/PI-negative (AnnV + PI -) or AnnV/PI double-positive (AnnV + PI + ) cells. Significance stars refer to comparisons of total AnnV + counts for siControl vs siRhoD samples. The % of AnnV-negative/PI-positive (AnnV -PI + ) necrotic cells was also quantified. Each bar represents the mean ±SD from 3 experiments (n = 3,166-4,998 cells per sample). (C) RHOD band intensities were normalized to β-ACTIN and plotted versus the % of AnnV-positive cells. Each point represents the mean ±SD from 3 fields-of-view in a given experiment (n = 670-1,919 cells per point). (D) eHAP and RhoD KO cells were treated with DMSO or etoposide and stained with Alexa488-AnnV (green), PI (magenta), and Hoechst (DNA; blue). Scale bar, 100μm. (E) Cells were treated with DMSO or etoposide, fixed, and stained with antibodies to detect cytochrome c (Cyto c; green), AIF (Mito; magenta), and DAPI (DNA; blue). Scale bar, 25μm. (F-G) Cells were treated with DMSO for 6h or etoposide for 3 or 6h, and extracts were immunoblotted with antibodies to caspase-9 (Casp-9 Pro and Casp-9 Cleaved ), caspase-3 (Casp-3 Cleaved ), and tubulin. (H) eHAP and RhoD KO cells were treated with DMSO or etoposide for 6h and stained with caspase-3/7 green detection reagent (Cleaved DEVD; green) and Hoechst (DNA; blue), or fixed and stained with an antibody that recognizes active caspase-3 cleaved at Asp175 (Cleaved Casp-3; green) and DAPI (DNA; blue). Scale bars, 100μm. � p<0.05, �� p<0.01 (ANOVA, Tukey post-hoc tests). Significances?. (TIF) S12 Fig. F-actin is reorganized in the cytosol of HAP1 cells in response to DNA damage. (A) HAP1 cells were treated with 5μM etoposide for 6h before being fixed and stained with a JMY antibody (green), a cyto c antibody (magenta), and DAPI (DNA; blue). Arrowheads and magnifications represent examples of juxtanuclear JMY and cyto c puncta. Scale bars, 25μm, 10μm. (B) U2OS cells were treated with 10μM etoposide for 6h before being incubated with MitoTracker (Mito; cyan), fixed, and stained with a JMY antibody (green), a cyto c antibody (magenta), and phalloidin (F-actin; yellow). Arrowheads and magnifications depict clusters of JMY and cyto c puncta in an F-actin territory. Scale bars, 25μm, 10μm. (C) HAP1 and JMY KO cells were treated with 5μM etoposide for 6h before being fixed and stained with a JMY antibody, DAPI, and either phalloidin to visualize F-actin or an actin antibody to visualize total actin. Nuclear and cytoplasmic F-actin, total actin, and JMY fluorescence intensities were measured in ImageJ and the total intensity was normalized to the HAP1 DMSO sample. Each bar represents the mean ±SD