Protein Kinase Cα (PKCα) Regulates p53 Localization and Melanoma Cell Survival Downstream of Integrin αv in Three-dimensional Collagen and in Vivo*

Background: Integrin αv promotes melanoma cell survival in in vivo and in vitro three-dimensional environments. Results: PKCα is up-regulated in an integrin αv- and three-dimensional collagen-dependent manner and promotes p53 relocalization and melanoma survival. Conclusion: PKCα functions downstream of integrin αv to promote melanoma cell survival. Significance: PKCα has been identified as a key component of the integrin αv-mediated melanoma survival pathway. Protein kinase C α (PKCα) is overexpressed in numerous types of cancer. Importantly, PKCα has been linked to metastasis of malignant melanoma in patients. However, it has been unclear how PKCα may be regulated and how it exerts its role in melanoma. Here, we identified a role for PKCα in melanoma cell survival in a three-dimensional collagen model mimicking the in vivo pathophysiology of the dermis. A pathway was identified that involved integrin αv-mediated up-regulation of PKCα and PKCα-dependent regulation of p53 localization, which was connected to melanoma cell survival. Melanoma survival and growth in three-dimensional microenvironments requires the expression of integrin αv, which acts to suppress p53 activity. Interestingly, microarray analysis revealed that PKCα was up-regulated by integrin αv in a three-dimensional microenvironment-dependent manner. Integrin αv was observed to promote a relocalization of endogenous p53 from the nucleus to the cytoplasm upon growth in three-dimensional collagen as well as in vivo, whereas stable knockdown of PKCα inhibited the integrin αv-mediated relocalization of p53. Importantly, knockdown of PKCα also promoted apoptosis in three-dimensional collagen and in vivo, resulting in reduced tumor growth. This indicates that PKCα constitutes a crucial component of the integrin αv-mediated pathway(s) that promote p53 relocalization and melanoma survival.

Protein kinase C ␣ (PKC␣) is a widely expressed, conventional member of the PKC family of serine/threonine kinases (1), which participates in numerous cellular functions such as proliferation, migration, survival, and apoptosis (2). Roles for PKC␣ have been identified in numerous cancer types, including breast (3), glioblastoma (4), and prostate (5). Strong links to melanoma have also been established (6), particularly to melanoma metastasis in patients (7)(8)(9)(10). Interestingly, PKC␣ was required for integrin ␣v␤3-mediated melanoma invasion (9) and in integrin ␣v␤3-mediated murine primary osteoclast bone resorption and migration (11). Additional possible roles in cancer come from the ability of PKC␣ to influence apoptosis in a number of cell types where it is regarded as an anti-apoptotic member of the PKC family (12).
PKC␣ may exercise its anti-apoptotic function through phosphorylation of Bcl-2 (13). However, the ability of PKC␣ to influence the tumor suppressor p53 may also be of importance. Upon activation, p53 can promote apoptosis via the transcriptional up-regulation of proapoptotic factors, including PUMA, Apaf 1, and Bax (14) or via direct effects at the mitochondria (15). PKC␣ can bind and phosphorylate the C terminus of p53 (16,17), although its effects upon p53 activity appear cell type specific (18,19). Meanwhile, a recent yeast model identified PKC␣ as an inhibitor of p53-induced growth arrest (20). Nevertheless, it has remained unclear how PKC␣ may be regulated and exert its role in malignant melanoma.
Melanoma progression toward a malignant phenotype is characterized by a shift from a radial to a vertical and invasive growth pattern, which includes the up-regulation of integrin ␣v␤3 expression (21). Integrin ␣v␤3 promotes melanoma cell invasion, growth, and survival in both three-dimensional dermal collagen and in vivo in mice and in human dermis (22)(23)(24)(25)(26). Integrin ␣v-mediated regulation of melanoma cell survival is dependent upon cell growth within a three-dimensional environment, emphasizing the importance of three-dimensional models that mimic physiological conditions (22,27). Indeed, integrin-mediated signal transduction and cell behavior in three-dimensional matrices are often different to that observed in two-dimensional tissue culture, and a closer approximation to physiological behavior (28 -31). Previous data using a threedimensional collagen model indicated that integrin ␣v expression promoted melanoma survival by inhibiting p53 activity and thereby apoptosis (22). Overcoming p53-mediated apoptosis is critical for cancer progression and although p53 is mutated in a significant number of cancers, the p53 gene is rarely mutated in melanoma (32). Despite this, malignant melanomas are typically extremely radio-and chemo-resistant and elevated p53 levels did not induce apoptosis (32,33). However, it has remained unclear how integrin ␣v may mediate p53 inhibition.
To clarify how integrin ␣v may inhibit p53-mediated apoptosis, a three-dimensional collagen model was utilized. As the dermal extracellular matrix is comprised of ϳ90% collagen type I, this model provides an environment comparable to physiological conditions (25). Through microarray-based expression profiling, we identified PKC␣ to be up-regulated downstream of integrin ␣v in a three-dimensional collagen-dependent manner. We also found that p53 translocation to the cytoplasm in melanoma cells was three-dimensional, integrin ␣v-, and PKC␣-dependent. Furthermore, PKC␣ knockdown inhibited integrin ␣v-mediated melanoma survival in three-dimensional collagen and in vivo tumor growth. Our results indicate that in melanoma, PKC␣ up-regulation is critical for integrin ␣v␤3mediated p53 regulation and suppression of apoptosis.
Analysis of p53 Localization by Immunofluorescence-For imaging of p53, 20 l of collagen mixed with cells was pipetted into each well of an eight-well chamber slide (Nunc, Rochester, NY). Slides were inverted to produce a hanging drop of collagen and incubated at 37°C, 5% CO 2 for 30 min. Once set, the slide was placed the correct way up, and three-dimensional medium was added. After the period of time stated, cells were fixed in 4% paraformaldehyde for 10 min at room temperature and then permeabilized with 0.1% Triton X-100/PBS for 1 min at room temperature. Cells were blocked in 5% BSA/PBS for 20 min at room temperature and then probed with anti-p53 mAb (DO-1, Santa Cruz Biotechnology, Santa Cruz, CA; 1/200) in 5% BSA/ PBS for 1 h at room temperature. After washing, Alexa Fluor 568 goat anti-mouse (Invitrogen, 1/200) and FITC-phalloidin (Invitrogen, 1/1000) were added for 1 h at room temperature in the dark. After further washing, cells were incubated with DRAQ5 DNA stain (Biostatus Ltd., Shepshed, UK, 1/1000) at 37°C for 3 min in the dark. DRAQ5 was aspirated off, and the collagen was mounted in fluorescent mounting medium (DAKO, Glostrup, Denmark) and covered with a glass coverslip. For two-dimensional analysis of p53 localization, nutridoma-treated cells were cytospun onto slides, fixed, and prepared for imaging as in three dimensions.
Confocal images were taken using a Zeiss LSM510 microscope with a C-Apochromat 63ϫ/1.2 numerical aperture water objective and LSM510 software. p53 fluorescent intensities in the cell cytoplasm and nucleus were analyzed using NIH ImageJ software and converted into a cytoplasmic/nuclear ratio for comparison between cells and experiments.
In Vivo p53 Imaging-p53 localization was established in vivo using the chick chorioallantoic membrane (CAM) 3 model as described previously (34). Briefly, a small window was made in the shell above the CAM, and 5 ϫ 10 6 M21 (␣v ϩ ) or M21L (␣v Ϫ ) cells in 30 l of sterile PBS were pipetted onto the membrane. The window was covered using tape and incubated for 7 days to allow tumor growth. Tumors were removed from the CAM, fixed with 4% paraformaldehyde overnight, rinsed in 70% ethanol, and then stored in fresh 70% ethanol overnight at 4°C. Tumors were dehydrated, embedded in paraffin, and sectioned. Samples were mounted on glass slides at 60°C for 30 min and then rehydrated in xylene followed by 100, 95, and 70% ethanol steps. Antigen retrieval was performed in citrate buffer, pH 6.0, heated for 4 min. After cooling and washing, the sample was permeabilized in 0.1% Triton X-100/PBS and blocked in 20% FBS/PBS for 1 h at room temperature. Samples were incubated with anti-p53 mAb (DO-1, Santa Cruz Biotechnology, 1/100) and anti-␤-catenin pAb (Invitrogen, 1/250) in 10% FBS/ PBS for 1 h at room temperature. After washing, Cy-5 goat anti-mouse (Jackson ImmunoResearch Laboratories, West Grove, PA, 1/200) and Rhodamine Red-X goat anti-rabbit antibodies (Jackson ImmunoResearch Laboratories, 1/250) in 10% FBS/PBS were added for 1 h at room temperature. Hoechst DNA stain (Invitrogen, 1/4000) was added in PBS for 3 min at room temperature. Samples were mounted in fluorescent mounting medium (DAKO) and imaged using an Andor spinning disc confocal with an Olympus UPlanSApo 60ϫ/1.35 numerical aperture oil objective. NIH ImageJ software was used to determine the p53 localization ratio by identifying cytoplasmic (␤-catenin) and nuclear (Hoechst) areas of cells before the mean p53 intensity of each area was measured. Mean p53 intensities were converted into a cytoplasmic/nuclear ratio to allow cellular and population comparisons.
In vivo localization of p53 was also established in mouse xenograft tumors. FM88sicon, shGFP; and FM88sicon, shPKC␣ tumors were stored at Ϫ80°C upon excision and later embedded in Cryomount medium (Histolab, Gothenburg, Sweden), frozen, and sliced into 10-m sections. Sections were placed onto coverslips and fixed (4% paraformaldehyde, 10 min at room temperature), washed, and blocked in 10% donkey serum/PBS containing 0.3% Triton X-100 for 1 h at room temperature. Sections were probed with FITC-conjugated anti-p53 mAb (DO-1, 1/100) and anti-␤-catenin pAb (Invitrogen, 1/250) in 10% donkey serum/PBS overnight at 4°C. After washing, Rhodamine Red-X goat anti-rabbit IgG (Jackson ImmunoResearch Laboratories, 1/500) in 10% donkey serum/PBS was added for 1 h at room temperature. DRAQ5 (Biostatus Ltd., 1/1000) was added for 3 min at 37°C before the sections were mounted in fluorescent mounting medium (DAKO). Images were acquired using a Nikon A1R microscope with a Plan-Apochromat VC 60ϫ/1.4 numerical aperture oil objective and Nikon A1R software. Quantitative analysis was performed as for the chick CAM images.
Determination of Apoptosis by Flow Cytometry-Cells were cultured in three-dimensional collagen for 3 or 5 days before being extracted as described above. The cells were resuspended in 0.05% BSA/PBS and fixed with ice cold 70% ethanol on ice for 30 min. Cells were spun down, resuspended in PBS with 1 g/ml RNase A and 10 g/ml propidium iodide (both from Sigma Aldrich), and incubated at 37°C for 1 h. Cells were analyzed using the FACSCalibur flow cytometer and CellQuest software (both from Becton Dickinson, Franklin Lakes, NJ). The sub-G 1 population of cells was regarded as apoptotic.
Determination of Apoptosis by Cleaved Caspase-3 Immunofluorescence-Cells were cultured in three-dimensional collagen on chamber slides and prepared as for p53 localization imaging. Apoptotic cells were visualized using anti-cleaved caspase-3 Asp-175 pAb (Cell Signaling Technology, 1/2000) and subsequent incubation with Alexa Fluor 568 goat anti-rabbit (Invitrogen, 1/200), FITC-phalloidin (Invitrogen, 1/1000), and DRAQ5 (Biostatus, Ltd., 1/1000). Once mounted, slides were imaged using a Nikon A1R confocal microscope with a Plan Apochromat 20ϫ/0.75 numerical aperture air objective. 3 ϫ 3 montage images were acquired for each planar view with two planes imaged per collagen drop. Analysis was performed using Nikon NIS-Elements software, nuclear outlines were identified using DRAQ5 staining, and the mean cleaved caspase-3 intensity within the nucleus was ascertained.
Cleaved caspase-3 determination of apoptosis in vivo utilized mouse xenograft tumors mounted as for p53 localization studies. Permeabilized sections were probed with anti-cleaved caspase-3 Asp-175 (Cell Signaling Technology, 1/500) and subsequent incubation with Rhodamine Red-X goat anti-rabbit IgG (Jackson ImmunoResearch Laboratories, 1/500) and DRAQ5 (Biostatus Ltd., 1/1000). Images were acquired using a Nikon A1R confocal microscope with a Plan Apochromat 20ϫ/ 0.75 numerical aperture air objective and analyzed using ImageJ software. Nuclear regions were identified from the DRAQ5 staining, and the mean intensity of cleaved caspase-3 in each region was measured.
Determination of Apoptosis by Scoring for Blebbing Phenotype-Dimethyl sulfoxide (DMSO) or 100 nM of the PKC␣ inhibitor Gö6983 (Merck KGaA, Darmstadt, Germany) was mixed with collagen prior to addition of cells. Collagen containing cells was added to a 24-well plate. Cells were cultured for 5 days after which apoptotic cells, using membrane blebbing as a read out, were counted on an inverted light microscope.
Microarray-Total RNA from 2 ϫ 10 6 M21 (␣v ϩ ) and M21L (␣v Ϫ ) cells grown for 72 h in three-dimensional collagen was isolated using the RNeasy kit (Qiagen, Solna Sweden). Three independent biological replicates were used for each condition. Raw data from Affyetrix hgu133a chips (Affymetrix Inc., Santa Clara, CA, USA) were normalized using the robust multiarray average with GC correction (GCRMA) procedure, implemented in the R package gcrma (quantile-based normalization with background correction based on gene chip content of probes (35)). Differential expression between conditions was estimated using significance analysis of microarrays procedure (36) implemented in the R package siggenes. Each gene found to be differentially expressed was assigned to the most similar pattern of expression. First, the gene profile was normalized to the 0 -1 range, and then the Euclidean distance to each of the possible expression patterns (also range normalized, a schematic is shown in Fig. 1A) was computed. The gene was assigned to the pattern with the smallest distance.
In Vivo Tumor Growth-Animal experiments were performed according to a permit from the Stockholm South ethical committee in Sweden. Balb/c nude female mice (6 -8 weeks) were obtained from Taconic. 5 ϫ 10 5 FM88sicon, shGFP, FM88sicon; and shPKC␣ melanoma cells were mixed with Matrigel (Becton, Dickinson) in a 1:1 ratio and injected subcutaneously at the back of the mice. The wet weight of dissected tumors was determined at the experiment's conclusion, and tumors were utilized in downstream applications as described.

Identification of Integrin ␣vand Three-dimensional Collagen-dependent Transcriptional Changes-Integrin-mediated
outside-in signaling can activate a number of downstream pathways that regulate gene transcription (37). To establish integrin ␣v-mediated changes to gene transcription, the metastatic melanoma cell line M21 (␣v ϩ ) and the integrin ␣v-negative M21-derived M21L cells (23) were utilized. M21 (␣v ϩ ) and M21L (␣v Ϫ ) cells were cultured in two-and three-dimensional collagen for 72 h to assess the integrin ␣vand three-dimensional collagen dependence of gene transcriptional changes (Fig. 1, A and B, and supplemental Table S1). A number of genes showed significant change between two-and three-dimensional culture conditions, which were independent of the integrin ␣v status (orange group, Fig. 1, A and B), whereas the transcription of other genes was identified to be integrin ␣v-dependent but two-and three-dimensional collagen independent (blue group, Fig. 1, A and B). The transcription of other genes was specifically up-regulated in three-dimensional collagen upon lack of integrin ␣v expression (pale yellow group, Fig. 1, A-C), whereas some genes were down-regulated in an integrin ␣v and three-dimensional collagen-dependent manner (purple group, Fig. 1, A-C). Genes undergoing significant integrin ␣v and three-dimensional collagen-dependent transcriptional up-regulation (pink group, Fig. 1, A, B, and D) were also identified. A number of targets linked to p53 function that displayed differential regulation on the microarray were validated by quantitative real-time PCR (qPCR). ASPP1, CK1␦, NOX4, and PIAS1 qPCR showed the same regulation as observed with the microarray (Fig. 1E). The microarray data also identified a number of integrin ␣v-dependent transcriptional changes that specifically occur upon growth on a twodimensional substrate or within three-dimensional-collagen (supplemental Table S1, full microarray data available at the GEO database (www.ncbi.nlm.nih.gov/geo/, accession no. GSE34219). This information aids our understanding of how integrin ␣v promotes melanoma cell survival and highlights specific hits that can clarify the molecular pathways implicated in this process.
Integrin ␣v-mediated Up-regulation of PKC␣ in Three-dimensional Collagen Matrices-Interestingly, the PKC␣ gene was among the genes found to undergo transcriptional regulation in an integrin ␣vand three-dimensional collagen-dependent manner (Figs. 1D and 2A). Importantly, PKC␣ has previously been identified to promote melanoma progression (6,9). The transcriptional regulation of PKC␣ was verified by qPCR; M21 cells (␣v ϩ ) showed a substantial increase in PKC␣ mRNA, whereas FM88sicon cells (␣v ϩ ) also showed an increase in PKC␣ mRNA levels compared with FM88si␣v cells (␣v Ϫ ), although the fold-change observed was smaller than that in M21 cells (Fig. 2B). PKC␣ protein levels were also significantly higher in M21 cells (␣v ϩ ), compared with M21L cells (␣v Ϫ ) (Fig. 2, C and D). These results show that in three-dimensional collagen, PKC␣ is elevated in an integrin ␣v-dependent manner, suggesting a potential role for PKC␣ downstream of integrin ␣v in melanoma.
PKC␣ Is Critical for Integrin ␣v-mediated Relocalization of p53-Previous data showed that integrin ␣v promoted melanoma cell survival in three-dimensional collagen through inhibition of p53-mediated apoptosis (22). Nuclear localization of p53 has been linked to its oligomerization state (38) and hence transcriptional activity. However, although cytoplasmic p53 is linked to mitochondrial-mediated apoptosis (15), promotion of a cytoplasmic p53 localization has also been observed to inhibit p53-mediated apoptosis in a number of survival mechanisms (39 -41). Here, we identified a three-dimensional-dependent relocalization of p53. Under two-dimensional conditions, p53 was predominantly nuclear in both M21 (␣v ϩ ) and M21L (␣v Ϫ ) cells (Fig. 3A) but in three-dimensional collagen, M21 (␣v ϩ ) cells had a higher ratio of cytoplasmic to nuclear p53 as compared with two-dimensional collagen (Fig. 3, A and B). Importantly, M21 (␣v ϩ ) cells displayed a higher cytoplasmic to nuclear p53 ratio as compared with M21L (␣v Ϫ ) cells (Fig. 3, B  and C). This integrin ␣v-regulated relocalization of p53 was also observed in M21 (␣v ϩ ) and M21L (␣v Ϫ ) tumors grown in vivo in the chick chorioallantoic membrane (CAM) (Fig. 3, E AUGUST 24, 2012 • VOLUME 287 • NUMBER 35

JOURNAL OF BIOLOGICAL CHEMISTRY 29339
and F). Studies using in vivo FM88sicon and FM88si␣v tumors, also grown in the chick CAM, showed similar results with integrin ␣v knockdown resulting in elevated p53 nuclear levels (data not shown). In addition, lysates from cells cultured in three-dimensional collagen showed elevated p53 levels in the nuclear fraction of M21L (␣v Ϫ ) cells as compared with M21 (␣v ϩ ) cells (Fig. 3, G and H). However, analysis of p53 localization using lysate fractions of FM88 cells grown in three-dimensional collagen proved difficult, possibly due to the comparatively low p53 expression levels observed. To examine whether up-regulation of PKC␣ was important for p53 and integrin ␣v function, PKC␣ levels were modified by stable knockdown and overexpression in M21 (␣v ϩ ) and M21L (␣v Ϫ ) cells, respectively (Fig. 2C). PKC␣ can phosphorylate p53 in vitro (16), although its effect upon p53 activity remains unclear (17,20). Importantly, under three-dimensional collagen conditions, M21 (␣v ϩ ) cells with stably knocked down PKC␣ exhibited a reduced p53 relocalisation with a greater proportion of p53 within the nucleus compared with control M21 cells stably expressing GFP knockdown sequences (Fig. 3, B and D). However, PKC␣ overexpression in M21L (␣v Ϫ ) cells, was not capable of rescuing the integrin ␣v-mediated p53 relocalization (Fig.  3, B and D), suggesting that PKC␣ is necessary but not sufficient to promote p53 relocalization and that one or more other cellular responses to integrin ␣v are also required.
PKC␣ Is Required for Integrin ␣v-dependent Melanoma Survival-To identify apoptotic cells within three-dimensional collagen, cleaved caspase-3 levels were analyzed (Fig. 4A). As shown previously (22), a larger fraction of M21L (␣v Ϫ ) cells undergo apoptosis compared with M21 (␣v ϩ ) cells in threedimensional collagen (Fig. 4B). Furthermore, M21 cells with stably knocked down PKC␣ showed increased levels of apoptosis compared with control M21 cells (Fig. 4B). Addition of the PKC␣ inhibitor, Gö6983 also induced apoptosis in M21 cells in three-dimensional collagen (Fig. 4C), suggesting PKC␣ knockdown-mediated induction of apoptosis is not an artifact or a  Table S1 to identify the patterns. B, genes significantly altered between any conditions were detected using the significance analysis of microarrays procedure (38), using a False Discovery Rate cut-off of 0.001, resulting in 262 genes identified. Genes were grouped based upon which of the 12 categories their regulation falls within. Each row is color-coded according to its corresponding pattern in A. Individual categories and the genes that constitute each category can be seen in supplemental Table S1. The expression profiles for each gene and condition shows the levels determined in three separate experiments. C, to select genes that follow a specific pattern of expression, the significant genes were range-normalized to the 0 -1 range and ranked by similarity to a pattern of regulation (see "Experimental Procedures"). Shown are the gene expression patterns of the genes that were most similar to the pattern shown in the top panel A with the corresponding color code: pale yellow (1), low expression in M21L (␣v Ϫ ) and M21 (␣v ϩ ) cells in two dimensions, but high relative expression in M21L cells in three dimensions; purple (2), high expression in M21L (␣v Ϫ ) and M21 (␣v ϩ ) cells in two dimensions, but low relative expression in M21 cells in three dimensions. Relative expression is indicated by color, where green represents low expression and red indicates high expression. D, gene expression patterns (as detailed in C) showing genes with low expression in M21L (␣v Ϫ ) and M21 (␣v ϩ ) in two dimensions but high relative expression in M21 cells in three dimensions (corresponding to the light red pattern in A). E, RT-PCR verification of microarray data showing sample gene transcriptional modification upon growth for M21 (␣v ϩ ) versus M21L (␣v Ϫ ) in three-dimensional collagen with or without integrin ␣v. Graphs shown are mean log10 relative change of three separate experiments. Microarray data are shown alongside the graphs for comparison and are prepared in the manner detailed in C.  Integrin ␣v and PKC␣ regulate p53 localization in three-dimensional collagen. A, immunofluorescence with an anti-p53 antibody was utilized to analyze p53 localization in M21 (␣v ϩ ) and M21L (␣v Ϫ ) cells grown in two dimensions. F-actin and nuclear staining (DRAQ5) were used to highlight the cell periphery and the nucleus. Scale bar, 10 m. B, p53 localization in three-dimensional collagen was analyzed by immunofluorescence in M21, M21L, M21shGFP, M21shPKC␣, M21LshGFP, and M21LPKC␣ cells. F-actin and nuclear (DRAQ5) staining were determined to highlight the cell periphery and the nucleus. Scale bar, 10 m. C, image quantification of M21 and M21L cells in two dimensions (2D; 0 days) and after 1, 2, and 3 days in three-dimensional (3D) collagen. Representative areas of a uniform size were selected within the cell cytoplasm and nucleus, and the mean p53 intensity was determined. A cytoplasm/nucleus intensity ratio was determined to allow comparison across samples and plotted. Graph shows mean p53 ratio Ϯ S.E. among three separate experiments (n ϭ 24 -45 cells/condition; n.s., not significant; *, p Ͻ 0.05; ***, p Ͻ 0.001 using t test). D, quantification of p53 localization in cells with PKC␣ levels manipulated and control cells after 3 days in three-dimensional collagen. Images for M21shGFP, M21shPKC␣, M21LshGFP, and M21LPKC␣ were analyzed as in C. Graph shows mean p53 ratio Ϯ S.E. among three separate experiments (**, p Ͻ 0.01 using t test). E, representative images of an in vivo M21L tumor section as analyzed in F. M21L cells were stained by immunofluorescence with an anti-p53 antibody to establish the p53 localization. ␤-Catenin and Hoechst were utilized to determine the cell periphery and nucleus, respectively. Scale bar, 10 m. F, quantification of p53 localization from images of in vivo M21 and M21L tumors grown in a chick CAM assay. Images were analyzed as in C, and the graph shows the mean p53 ratio Ϯ S.E. among three separate experiments (***, p Ͻ 0.001 using t test). G, upper: lysate fractionation was assessed by immunoblotting the soluble (sol.) and pellet (pel.) fractions of M21 and M21L lysates for phosphohistone H3 and ␤-tubulin. Immunoblots are representative of two separate experiments. Lower: soluble and pellet levels of p53 were determined by immunoblotting lysates from M21 and M21L cells. The immunoblot is representative of three separate experiments. H, quantification of soluble and pellet p53 levels as determined by immunoblot. Graph shows mean p53 levels Ϯ S.E. among three independent experiments with soluble p53 levels normalized to 1.
consequence of off-target effects. In addition, a second M21shPKC␣ stable cell line, expressing an alternative PKC␣ shRNA sequence, also exhibited increased apoptosis levels compared with M21shGFP cells as determined through visualization of the blebbing phenotype (data not shown). FM88 cells with stably knocked down integrin ␣v also showed higher levels of apoptosis compared with control FM88 (␣v ϩ ) cells (Fig. 4D), indicating that survival in the FM88 melanoma cell line is also controlled by integrin ␣v in three-dimensional-collagen. In addition, inhibition of PKC␣ with Gö6983 also increased the observed incidence of apoptosis in FM88 cells (Fig. 4E). However, M21L (␣v Ϫ ) cells stably overexpressing PKC␣ were unable to promote melanoma survival (data not shown). This correlates with the inability of PKC␣ overexpression to rescue integrin ␣v-mediated relocalization of p53 in three-dimen-sional collagen and indicates it is insufficient to promote melanoma survival alone in the absence of integrin ␣v.
PKC␣ Promotes p53 Relocalization and Melanoma Survival in Vivo-To determine whether PKC␣ may be required downstream of integrin ␣v for p53 relocalization and melanoma tumor growth in vivo, control FM88 cells (␣v ϩ ), and control FM88 cells with stably knocked down PKC␣, were injected subcutaneously into the back of balb/c nude mice. After 18 days of growth, FM88 cells with stably knocked down PKC␣ had a significantly reduced mass compared with control cells (Fig. 5A), identifying a role for PKC␣ in melanoma tumor growth. To establish whether the reduction in FM88sicon shPKC␣ tumor growth corresponded with the altered p53 localization observed in vitro, p53 immunofluorescence was performed on tumor sections. FM88sicon shPKC␣ tumor cells had a sig- . PKC␣ regulates melanoma cell survival in three-dimensional collagen. A, apoptosis was measured by immunofluorescence with an anti-cleaved caspase-3 antibody. F-actin and nuclear (DRAQ5) staining was performed to visualize the cell periphery and nucleus, respectively. Representative images of a viable and an apoptotic M21shGFP cell in three-dimensional collagen are shown. Scale bar, 10 m. B, quantification of apoptotic cells cultured in threedimensional collagen for 5 days. Automated identification of nuclear regions was determined by DRAQ5 staining, and the mean nuclear cleaved caspase-3 intensities were measured. M21sip53 cells after 1 h in three-dimensional collagen were used as a control for cleaved caspase-3 intensity in non-apoptotic cells and used to define viable and apoptotic intensities in subsequent experiments. Graphs show mean percentage of apoptotic cells from three independent experiments Ϯ S.E. (*, p Ͻ 0.001 using t test). C, light microscopy was used to manually count healthy versus apoptotic cells cultured for 5 days in threedimensional collagen. M21, M21L, M21shGFP, and M21shPKC␣ cells were cultured with dimethyl sulfoxide or 100 nM Gö6983 and scored as healthy or blebbing. Graph shows mean levels of apoptosis Ϯ S.E. among three independent experiments (n.s., not significant; **, p Ͻ 0.01 using chi squared test on arcsin normalized data). D, quantification of apoptotic cells cultured in three-dimensional collagen for 3 days. FM88sicon and FM88si␣v were stained with PI, and the apoptotic population (sub-G 1 ) was established using FACS. Graphs show mean percentage of apoptotic cells from three independent experiments Ϯ S.E. (**, p Ͻ 0.01 using t test). E, quantification of apoptosis was determined using light microscopy to manually quantify the apoptotic cells cultured in threedimensional collagen for 3 days. FM88sicon cells were cultured with dimethyl sulfoxide or 100 nM Gö6983 and scored as healthy or blebbing. Graph shows mean levels of apoptosis Ϯ S.D. for 170 (dimethyl sulfoxide, DMSO) and 136 (Gö6983) cells (*, p Ͻ 0.05 using a Chi-squared test on arcsin normalized data). nificantly higher nuclear p53 localization compared with FM88sicon shGFP tumor cells (Fig. 5, B and C), suggesting that PKC␣ is required for promoting a cytoplasmic relocalization of p53 both in vitro and in vivo. Furthermore, FM88sicon shPKC␣ tumors had significantly higher levels of cleaved caspase-3 (Fig.  5, D-F), suggesting increased rates of apoptosis within the FM88sicon shPKC␣ tumor. This supports a role for PKC␣ regulating melanoma tumor apoptosis and suggests that PKC␣ is an essential in vitro and in vivo downstream component of integrin ␣v-mediated cell survival.

DISCUSSION
A role for PKC␣ in regulating p53 localization and melanoma survival is of particular interest given its strong association with melanoma (6). Although PKC may regulate integrin ␣v␤5 function in melanoma cells in two-dimensional culture (42), our results here link PKC␣ into signaling downstream of integrin ␣v in melanoma cells. The PKC protein family may regulate melanoma survival, as PMA treatment promoted anchorageindependent survival of melanoma cells (43). To this end, PKC␣ has been suggested to be anti-apoptotic (12), and differing FIGURE 5. PKC␣ regulates p53 localization and apoptosis in vivo and promotes melanoma tumor growth. A, FM88sicon, shGFP (control) and FM88sicon, shPKC␣ (PKC␣-) melanoma cells were subcutaneously injected into Balb/c nude mice. Tumors were removed from mice after 18 days of growth. Removed tumors were weighed on electronic scales. Graph shows mean tumor wet weight Ϯ S.D. (n ϭ 7, *, p Ͻ 0.05 using t test). B, localization of p53 was observed in FM88 tumor sections with an FITC-conjugated anti-p53 antibody. ␤-Catenin and nuclear (DRAQ5) staining was performed to visualize the cell periphery and nucleus, respectively. Representative images of FM88sicon, shGFP and FM88sicon, shPKC␣ tumor sections are shown. Scale bar, 10 m. C, quantification of p53 localization in FM88 tumor sections. Representative areas of a uniform size were selected within the cell cytoplasm and nucleus, and the mean p53 intensity was determined. Graph shows mean p53 ratio Ϯ S.D. (n ϭ 60 -120 cells/tumor, five tumors per cell type; **, p Ͻ 0.01 using t test). D, apoptosis in FM88 tumor sections was analyzed by immunofluorescence using an anti-cleaved caspase-3 antibody. Nuclear (DRAQ5) staining was performed to visualize the nucleus. Tumor sections stained with no primary antibody added were used as a background fluorescence control. E, representative areas of a uniform size were selected using the DRAQ5 channel, and the mean intensity of cleaved caspase-3 was determined. Graph shows mean cleaved caspase-3 intensity Ϯ S.E. (n ϭ 50 cells/tumor, three tumors/cell type; ***, p Ͻ 0.001 using t test). F, data from E plotted as a histogram showing frequency of cleaved caspase-3 intensity for each tumor type. mechanisms have been described, including PKC␣-mediated phosphorylation of bcl2 (13,44), although data in glioma cells suggested that the mechanism of PKC␣ was kinase-independent (4). However, our data indicate that PKC␣ kinase activity is critical for melanoma cell survival, as use of the PKC kinase inhibitor Gö6983 resulted in elevated levels of apoptosis. PKC␣ has also been associated with melanoma metastasis (8), possibly through the regulation of adhesion turnover and Rac activity (9). In addition, PKC␣ inhibition reduced ERK1/2 activity and was also implicated as a possible mechanism to inhibit melanoma invasion (10). This finding corresponds with our previous data implicating integrin ␣v in the regulation of MEK1 and ERK1/2 upon melanoma growth in three-dimensional collagen (22) and supports the finding that PKC␣ is a downstream effector of integrin ␣v.
Inhibition of apoptosis by integrin ␣v is achieved through the suppression of the DNA-binding activity of p53 (22). PKC␣ can phosphorylate the C-terminal region of p53 at Ser-376 and Ser-378 (16) and inhibit p53 in yeast (20), whereas PKC␣ overexpression promoted murine p53 transcriptional activity (17). The observed differences in the role of PKC␣ in p53 regulation could be a consequence of differences between the model systems used. The induction of apoptosis upon PKC␣ knockdown in this study, together with previous data (22), suggests that PKC␣ may function as an inhibitor of p53 in melanoma. However, we were unable to detect direct PKC␣ regulation of p53. The anti-p53 antibody epitope of pAb421 covers the region potentially phosphorylated by PKC␣, and the binding of the antibody is inhibited by phosphorylation in this region (18). However, immunoblotting with pAb421 showed no clear difference between M21 and M21L cells after growth in threedimensional collagen (data not shown). On the other hand, addition of Gö6983 promoted apoptosis, suggesting that PKC␣ kinase activity is required for melanoma cell survival. PKC␣ phosphorylation of p53 has been reported to promote p53 ubiquitination and degradation in unstressed fibroblasts (45). However, PKC␣ knockdown did not increase p53 levels in our melanoma cells, but we did observe an elevated level of nuclear p53. Interestingly, phosphorylation of p53 by GSK3␤ at Ser-315 and Ser-376, a PKC␣ phosphorylation site, promoted a p53 cytoplasmic localization and an inhibition of p53-mediated apoptosis during endoplasmic reticulum stress (41). This finding is in concert with our results in melanoma cells upon growth in three-dimensional collagen. Although it appears likely that PKC␣ does regulate p53 in a kinase-dependent manner, this may be an indirect regulation via other effectors or through phosphorylation at p53 sites other than Ser-376/378. An alternative to directly regulating p53 transcriptional activity is to influence its cellular localization. Nuclear export signal sequences within p53 can drive export upon p53 monomerization (38) or after DNA damage-dependent phosphorylation of Ser-15 (46). In addition, p53 nuclear localization sequences and a cytoplasmic sequestration domain have been identified (47,48). We found that in two-dimensional culture, p53 was almost entirely nuclear located in both M21 (␣v ϩ ) and M21L (␣v Ϫ ) cells. However, within three-dimensional collagen and in vivo, p53 became progressively more cytoplasmic in a PKC␣-and integrin ␣v-dependent manner. The development of three-dimensional tissue culture has furthered our understanding of tissue (30) and cancer development, including melanoma (25). Integrin ␣v␤3 expression is critical for the development of invasive melanoma (21,24). However, it was the use of three-dimensional tissue culture that allowed unraveling of the role of integrin ␣v in promoting melanoma survival (25) through inhibition of p53 (22). The role of integrin ␣v in inhibiting apoptosis through blocking p53 function suggested an explanation to why p53 mutations are rarely observed in malignant melanoma (32). The translocation of p53 into the cytoplasm was identified as a three-dimensional microenvironment-, integrin ␣v-, and PKC␣-dependent event. We propose that melanoma cells promote a cytoplasmic p53 localization to overcome apoptosis, avoiding the requirement for p53 mutation (Fig. 6). Our results also point out a pathway with multiple potential targets for development of melanoma therapy, including integrin ␣v (26), p53 (49), and PKC␣.
Sequestration of p53 in the cytoplasm has also been implicated in oxaliplatin-resistant cervical carcinoma cell survival (50), supporting the principle of cytoplasmic sequestration of p53 in the inhibition of apoptosis. However, it is unclear whether the shift in p53 localization in melanoma may be a consequence of increased nuclear export and/or decreased nuclear import/tethering within the cytoplasm.
In conclusion, we have identified PKC␣ as a downstream effector of integrin ␣v that promotes a p53 cytoplasmic relocalization and inhibits melanoma cell apoptosis both in vitro and in vivo, resultantly promoting melanoma tumor growth. This pathway is dependent upon cellular growth in a three-dimensional environment and is not apparent under standard twodimensional conditions, stressing the importance of using as physiological conditions as possible. Although the role for Increased PKC␣ levels promote a relocalization of endogenous p53 protein from a chiefly nuclear to a cytoplasmic localization. PKC␣ signaling could promote the relocalization of p53 by either 1) enhancing the nuclear export of p53 or 2) by inhibiting the translocation of p53 into the nucleus from the cytoplasm. PKC␣-mediated relocalization of p53, in conjunction with other integrin ␣v-mediated signaling pathways, results in increased melanoma survival.
PKC␣ in this process is apparent, further studies are needed to delineate in detail the signaling cascade that results in p53 cytoplasmic shuttling.