Supervillin-mediated Suppression of p53 Protein Enhances Cell Survival*

Background: Supervillin is a membrane-associated actin- and myosin II-regulatory protein that promotes cell growth, adhesion, and invasion. Results: Supervillin inversely regulates p53 levels, binds the p53-stabilizing protein USP7, and regulates the USP7-p53 interaction. Conclusion: Supervillin regulation of p53 through USP7 contributes to cell survival. Significance: This study describes a new locus for cross-talk between cytoskeletal proteins and p53-regulated survival signaling. Integrin-based adhesions promote cell survival as well as cell motility and invasion. We show here that the adhesion regulatory protein supervillin increases cell survival by decreasing levels of the tumor suppressor protein p53 and downstream target genes. RNAi-mediated knockdown of a new splice form of supervillin (isoform 4) or both isoforms 1 and 4 increases the amount of p53 and cell death, whereas p53 levels decrease after overexpression of either supervillin isoform. Cellular responses to DNA damage induced by etoposide or doxorubicin include down-regulation of endogenous supervillin coincident with increases in p53. In DNA-damaged supervillin knockdown cells, p53 knockdown or inhibition partially rescues the loss of cell metabolic activity, a measure of cell proliferation. Knockdown of the p53 deubiquitinating enzyme USP7/HAUSP also reverses the supervillin phenotype, blocking the increase in p53 levels seen after supervillin knockdown and accentuating the decrease in p53 levels triggered by supervillin overexpression. Conversely, supervillin overexpression decreases the association of USP7 and p53 and attenuates USP7-mediated p53 deubiquitination. USP7 binds directly to the supervillin N terminus and can deubiquitinate and stabilize supervillin. Supervillin also is stabilized by derivatization with the ubiquitin-like protein SUMO1. These results show that supervillin regulates cell survival through control of p53 levels and suggest that supervillin and its interaction partners at sites of cell-substrate adhesion constitute a locus for cross-talk between survival signaling and cell motility pathways.

Cross-talk between signaling pathways involved in cell motility, invasion, and cell survival is of great interest in the search for novel cancer therapies (1). Cell migration and proliferation are controlled by integrin-based substrate interaction sites called focal adhesions (2,3). Focal adhesions also facilitate cell survival during stressful conditions, e.g. DNA damage, by decreasing levels of the p53 tumor suppressor protein (4 -6). Adhesion is proposed to mediate a feedback loop involving direct binding of p53 protein to the focal adhesion kinase (FAK) 3 protein and to the FAK promoter (7). In addition, the FAK-related protein Pyk2, which can be expressed at increased levels after FAK knockdown (8), increases cell proliferation by decreasing p53 levels (9).
Integrin signaling also is required for adhesion and matrix invasion by F-actin-enriched structures known as podosomes and invadopodia, or collectively, as invadosomes (10,11). Downstream signaling involving FAK and Src family tyrosine kinases, which include Lyn, promotes cell proliferation as well as invasion and correlates with poor prognosis in cancer patients (12). Depending on the cellular context (13), Lyn can promote cell survival by down-regulating p53 levels (14). Interestingly, wild-type p53 negatively regulates cell migration and invasion in vascular smooth muscle cells (15), and mutant p53 drives invasion of lung cancer cells by promoting integrin recycling (16). Taken together, these reports suggest cross-regulation of p53 and adhesion-based signaling pathways (17).
In previous studies, we found that the focal adhesion-regulatory, Lyn-associated protein supervillin inversely regulates tight cell-substrate adhesion and is required for normal cell division, cell motility, and matrix degradation (18 -24). Supervillin is tightly associated with cholesterol-rich lipid raft membranes and co-immunoprecipitates with Lyn and other signaling proteins (21). As is observed after FAK knockdown (25,26), supervillin knockdown increases the numbers of large, mature focal adhesions (23). Supervillin also increases podosome turn-over and function (18), regulates cell spreading (27), and promotes rapid recycling of integrins (28). Increased focal adhesion and podosome disassembly involve the myosin IIactivating and focal adhesion-targeting domains in the supervillin N terminus and its villin-like C terminus, which contains interaction sites for invadosome and cell cycle proteins (18,22,23,27). Supervillin targeting to focal adhesions and invadosomes requires myosin II activation (18,29), leading to a model in which supervillin increases contractility-induced turnover of these structures by scaffolding the long isoform of myosin light chain kinase onto preexisting myosin II filaments (18,27).
Mechanisms by which supervillin might contribute to cell proliferation and survival have previously focused on its regulation of cytokinesis and the prolongation and amplification of stimulus-mediated signaling through the lipid raft-based Raf/ MEK/ERK signaling cascade (22,28,30,31). The severe cell growth deficits observed after reducing supervillin levels with shRNAs or dsRNAs (22) caused us to hypothesize the presence of additional mechanisms. We report here that supervillin isoform 1 and, especially, a new isoform of supervillin (isoform 4) regulate cell survival, down-regulate the levels of p53, bind directly to the p53-deubiquitinating and stabilizing protein, USP7/HAUSP (32), and are themselves ubiquitinated under regulation by USP7.
Live Cell Imaging-U2OS cells on 25-mm 2 coverslips were transfected with dsRNA, incubated at 37°C with 5% CO 2 for 48 h, and sealed into live-cell imaging chambers. Time-lapse images were acquired every 3 min for 36 h using a 10ϫ objective lens (NA 0.30) on a DMIRE 2 inverted microscope with a mechanical stage (Leica Microsystems), a Hamamatsu ORCA-ER digital camera (Hamamatsu Photonics) and Simple PCI 6 software (Compix Inc., Sewickley, PA). Images were exported as AVI videos at 15 frames/s; selected stills were sized and contrast-enhanced in Adobe Photoshop.
Flow Cytometry-U2OS cells were trypsinized and resuspended with phosphate-buffered saline before overnight fixation at 4°C in 95% ethanol and staining with propidium iodide. DNA content was determined using FACSCalibur (BD Biosciences) and FlowJo software (Tree Star).
MTT Cell Proliferation and Survival Assay-U2OS cells transfected with dsRNA were incubated at 37°C with 5% CO 2 in 96-well flat-bottom microplates in a final volume of 100 l of DMEM/well. Cells were treated with etoposide or doxorubicin and assayed for tetrazolium reductase activity, which is proportional to cell proliferation and viability (39), according to the manufacturer's instructions (ATCC, kit 30-1010K). Cellular metabolic activity was determined as a percentage relative to untreated control cells.
Ubiquitination Assays-HEK293T cells were co-transfected with FLAG-tagged supervillin and HA-tagged ubiquitin plasmids for 24 h and lysed in denaturing lysis buffer containing 10 M MG132, 1 mM PMSF, and protease inhibitors. Competition assays were performed similarly with FLAG-p53 and GFPtagged USP7 and SV4. FLAG-tagged proteins were recovered, as described above.
RNAi-mediated reduction in the level of SV4 leads to Ͼ3-fold increases in the level of the cell guardian protein p53 and to increased cell death (Fig. 2, supplemental Fig. S2, and supplemental Movies S1-S3). Stealth dsRNA targeting coding exon 4, 5, or 16 reduced the amount of SV4 to ϳ10% of endogenous levels within 48 h ( Fig. 2A, lanes 2 and 3 versus lane 1) and to ϳ5% of control levels after 72 h (see below). Only the dsRNA targeting coding exon 16 effectively reduces the level of SV1 within 48 h (Fig. 2A, lane 4) although all three dsRNAs reduce both isoforms to ϳ5% of endogenous levels by 72 h (see below). All three supervillin-specific dsRNAs cause a ϳ3-fold increase in the level of p53 protein relative to actin and ERK1/2 loading controls within 48 h (Fig. 2, A and B). Consistent with supervillin knockdown experiments in HeLa cells (22), dsRNA targeting either coding exon 5 or coding exon 16 decreases the percentage of U2OS cells that undergo apparently normal cell division 24 -60 h later (Fig. 2, C and D, and supplemental Movies S1-S3). However, rather than failing primarily during early cytokinesis, as is seen in HeLa cells (22), supervillin knockdown in U2OS cells predominantly increases cell death (Fig. 2, C and D, and supplemental Movies S2 and S3). The supervillin knockdown U2OS cells mostly die after rounding without attempting cleavage or after a sudden, apparently apoptotic, fragmentation of cytoplasm and nuclei. The differences between cells treated with control and supervillin-specific dsRNAs are even more pronounced after 36 h of knockdown (Ͼ12 h of filming in supplemental Movies S1-S3), suggesting an increased effect as residual supervillin dwindles. As predicted by the numbers of phase-bright dead cells at the end of filming (Fig. 2C), cytometric analyses show increased amounts of fragmented, sub-G 1  DNA in cells deficient in SV4 only (Fig. 2E, Exon 5) or in both SV1 and SV4 (Fig. 2E, Exon 16). Taken together, these experiments show that p53 protein levels and the amount of cell death both increase in the absence of supervillin and suggest enhancement of p53-dependent cell cycle checkpoints (42).
Supervillin-p53 Cross-talk during DNA Damage-Supervillin levels decrease as p53 protein levels increase following DNA damage by the topoisomerase II inhibitors etoposide (Fig. 4A) and doxorubicin (Fig. 4B). These reagents cause DNA damage that triggers p53-mediated cell cycle arrest (44). Down-regulation of SV4 alone by a dsRNA that targets coding exon 5 or the down-regulation of both SV4 and SV1 with dsRNA against coding exon 16 or 18 exacerbates cell death following DNA damage (Fig. 4, C and D), without itself causing increased nuclear staining for the DNA damage marker ␥H2AX (data not shown). This increase in cell death is partially reversed by simultaneous RNAi-mediated depletion of p53 (Fig. 4, E and F) or by inhibition of p53 mitochondrial or transcriptional functions with pifithrin-(PFT) or pifithrin-␣ (PFT␣), respectively (Fig. 4, G and H) (45,46). These results suggest that part of the effect of supervillin on cell survival is through control of p53 protein levels. However, cell metabolic activity after double knockdown of p53 and supervillin is lower than after knockdown of p53 alone (Fig. 4, E and F), consistent with p53-independent effects of supervillin.
Supervillin Itself Can Be Regulated by Ubiquitination-The stability of the supervillin protein itself can be regulated by USP7 and by covalent modification with ubiquitin and SUMO1 (Fig. 7), a small ubiquitin-related modifier peptide that functionally cross-talks with ubiquitin (50). The UbPred program (51) predicts 19 high confidence ubiquitination sites in SV1 and 34 high confidence sites in SV4, nearly all in the N terminus. In agreement with this prediction, supervillin levels increase in cells that overexpress GFP-USP7 (Fig. 7A). Immunoprecipitated FLAG-tagged SV1 (Fig. 7B) and SV4 (Fig. 7C) are both modified by HA-tagged ubiquitin (Fig. 7, B and C, lanes 3). The amount of HA-ubiquitin covalently bound to immunoprecipitated SV1 (Fig. 7B) and, especially, to SV4 (Fig. 7C) is reduced in the presence of GFP-USP7 (Fig. 7, B and C, lanes 4), consistent with USP7-mediated deubiquitination. As predicted by the SUMOplot TM Analysis Program, FLAG-tagged SV1 also is covalently modified by GFP-tagged SUMO1 (Fig. 7D, lane 2,  SUMO-SV1). Most SUMO1 binding is eliminated by mutagen-  Overexpression of SV1 or SV4 reduces p53 levels. A, immunoblots of HeLa CLL-2 cells that transiently overexpress FLAG-tagged SV4 or SV1 stained for supervillin, p53 protein, and actin as a loading control. B, quantification of p53 protein levels after overexpression of SV4 or SV1. Each supervillin isoform reduced p53 levels. n ϭ 3; **, p Ͻ 0.01; ***, p Ͻ 0.001. Error bars, S.D. The difference between SV4 and SV1 was not statistically significant. esis of 2 of the 13 potential SUMOylation sites (Fig. 7D, lane 5). Cellular levels of SV1 and SV4 (Fig. 7E, top panel), but not USP7 (Fig. 7E, second panel), increase in the presence of GFP-tagged SUMO1 (Fig. 7E), suggesting that SUMOylation of supervillin may stabilize the protein. Consistent with this possibility, levels of SUMOylation-deficient SV4 decrease when co-expressed with HA-ubiquitin (Fig. 7F, lane 5). In the absence of HA-ubiquitin, all the SUMOylation-deficient SV4 mutants are expressed at similar levels, and all lead to decreased levels of p53 protein (Fig. 7G). Taken together, these results suggest that supervillin levels are regulated by ubiquitination, USP7-mediated deubiquitination, and SUMOylation.

DISCUSSION
We show here that the lipid raft-associated protein supervillin promotes cell survival by suppressing levels of p53. The knockdown of the newly described supervillin isoform 4 by itself or with supervillin isoform 1 increases cell death and p53 protein levels. Conversely, overexpression of SV1 or SV4 decreases the level of p53. In DNA-damaged cells, endogenous supervillin proteins decrease as p53 levels rise; further reduction of supervillin by RNAi accentuates p53-dependent effects on metabolic activity. RNAi-mediated knockdown of p53 or addition of a p53 inhibitor significantly reverses the supervillinknockdown effect, strongly suggesting that increased cellular p53 is an important intermediate in the control of cell proliferation by supervillin.
One mechanism of action is the association of supervillin with the p53 deubiquitinating enzyme USP7/HAUSP (Fig. 8). USP7 binds to the supervillin N terminus and is required for the increase in p53 levels observed after supervillin knockdown. Moreover, supervillin overexpression reduces the USP7-p53 interaction in co-transfected cells and inhibits USP7-mediated p53 deubiquitination. In the absence of supervillin, USP7 is better able to deubiquitinate p53, which protects p53 from deg-radation and decreases the rate of cell survival, especially after DNA damage. Upon supervillin overexpression, USP7 is less able to deubiquitinate p53, resulting in increased destruction of p53 and enhanced rates of cell survival. USP7 knockdown inhibits the effects of supervillin knockdown and synergizes with supervillin overexpression, suggesting that USP7 acts downstream of supervillin. We thus propose that supervillin negatively regulates the USP7 interaction with ubiquitinated p53 (Fig. 8). Supervillin SUMOylation and USP7-mediated deubiquitination of supervillin could reinforce this regulatory loop by stabilizing supervillin protein levels.
The effects of supervillin knockdown on cell division are probably related to p53 status. In a previous study with HeLa cells, the most prominent cell cycle defect after supervillin knockdown was the failure of the cleavage furrow to close during early cytokinesis (22). Cell division also is aberrant in U2OS cells deficient in supervillin, but increased cell death, appar- ently due to apoptosis, is the predominant phenotype after 48 h of supervillin knockdown in U2OS cells (Fig. 2). The difference between these cell lines likely arises from the presence in HeLa cells, but not in U2OS cells, of human papillomavirus oncoproteins, which target p53 for degradation by a pathway that is unaffected by USP7 (48,52). The supervillin contributions to other survival signaling pathways (28,31) may contribute to the incomplete rescue by p53 knockdown of the supervillin knockdown effect in U2OS cells.
The intracellular location(s) for the supervillin/USP7/p53 functional interactions remain an open question. At physiological levels, supervillin associates primarily with membranebound cortical myosin II and F-actin and overlaps with focal adhesion proteins, including the p53-binding protein FAK (18,19,21,24,27,41), whereas most USP7 and p53 are concentrated within the nucleus (53,54). On the other hand, USP7 and p53 can be cytoplasmic (55,56), and the central region of supervillin contains a strong nuclear localization signal that can target supervillin fragments to the nucleus (24). This central supervillin region also promotes androgen receptor signaling at an undetermined intracellular location (57). We have been unable to document localization of USP7 at focal adhesions or physiological levels of supervillin in the nucleus. However, we cannot exclude the possibility that a minor fraction of supervillin or a proteolytic fragment translocates to the nucleus, where it could interact with nuclear populations of USP7, FAK, and/or Lyn (13,53,58). Alternatively, supervillin may affect the deubiquitination and/or nuclear translocation of cytoplasmic p53, either alone or in complexes with other proteins, such as FAK or Lyn (13,14,58).
That said, the results reported here are consistent with the associations and functions of supervillin at cell-substrate adhesion sites. Like FAK and Src family kinases (25,59,60), super-   1 and 3) or GFP-USP7 (lanes 2 and 4), without (lanes 1 and 2) and with (lanes 3 and 4) HA-ubiquitin (HA-Ub). The positions of presumably monoubiquitinated SV1 and SV4 are indicated (Ub-SV1 and Ub-SV4). Loading controls were antibody against ␤-actin (A) or anti-FLAG staining (B and C). D, immunoblots of anti-FLAG immunoprecipitates and WCE from HeLa cells that co-express FLAG-tagged SV1 or FLAG-SV1 with the indicated point mutations in one or both of two predicted SUMOylation sites (K968R, K1479R) and either GFP alone (lane 1) or GFP-SUMO1 (lanes 2-5). Blots were stained for FLAG or GFP, as shown. E, immunoblots of WCE from cells that co-express FLAG-SV1 (lanes 1 and 2) or FLAG-SV4 (lanes 3 and 4) and either GFP (lanes 1 and 3) or GFP-SUMO1 (lanes 2 and 4) after staining with antibodies against FLAG, USP7, actin, and GFP. F, immunoblots of IP or WCE from HeLa cells co-expressing either empty vector (lane 1) or HA-tagged ubiquitin (HA-Ub, lanes 2-5) and FLAG-SV4 without (lanes 1 and 2) or with mutations in SUMOylation sites that correspond to the SV1 mutations in D (lanes 3-5). G, immunoblots of WCE from HeLa cells expressing either empty vector (lane 1) or wild-type (lane 2) or SUMOylation-deficient (lanes 3-5) FLAG-SV4 mutants stained for supervillin (H340 antibody), p53 and actin. Molecular mass markers, in kDa, are on the left. villin increases cell motility and promotes the turnover of cellsubstrate adhesions (18,23,28). Although other processes also occur (19, 60 -62), adhesion turnover mediated by FAK, Src, and supervillin involves activation of myosin II contractility by myosin light chain kinase (18,23,26). We now show that supervillin also regulates p53 levels, as do FAK, PYK2, and Src family kinases. Taken together, these observations raise questions about commonality in signaling mechanisms that link decreases in p53 levels with increased rates of adhesion turnover and cell motility.
Signaling cross-talk between cell proliferation pathways and sites of cell-substrate adhesion goes in both directions. Overexpression of wild-type p53 or activation of p53 with doxorubicin can down-regulate expression of the ␣5 integrin subunit and inhibit Src-induced podosome formation, cell motility, and cell invasion (15,63). Conversely, mutated p53 proteins can promote cell invasion and prolong survival signaling by increased recycling of integrins and growth factor receptors (16). In addition, the stability of many focal adhesion proteins, including FAK and Src, is regulated by ubiquitination (64), and many deubiquitinases, although not USP7, are required for epithelial cell scattering (65). A role for supervillin in the regulation of cell survival, as well as adhesion and matrix degradation (18 -20, 23), sheds new light on the observation that missense somatic mutations in supervillin occur in ϳ22% of sequenced tumors (26/120 unique samples) (66). We thus suggest that the machinery involved in adhesion protein turnover may be integrally involved in the poorly understood survival signaling pathways promoted by integrin-based cell-substrate adhesion.
In summary, we report the cloning of a new supervillin isoform and show for the first time a role for supervillin in regulation of cell survival through control of p53 levels. This study provides new insights into cross-talk between the cytoskeletal motile machinery and cell survival mechanisms. FIGURE 8. Summary of USP7-supervillin interactions described here. We propose that the inverse relationship between levels of supervillin and p53 is caused by direct or indirect inhibition by supervillin of the ability of USP7 to deubiquitinate ubiquitinated p53 (Ub-p53). Decreases in USP7 enzymatic activity or accessibility to Ub-p53 will increase p53 degradation and cell survival, especially after stressors like DNA damage. Supervillin SUMOylation and supervillin deubiquitination by bound USP7 could accentuate this effect by stabilizing supervillin itself against ubiquitin-mediated degradation.