Squamous cell carcinoma (SCC) is a treatment-refractory subtype of human cancer arising from stratified epithelium of the skin, lung, esophagus, oropharynx, and other tissues. A unifying feature of SCC is high-level expression of the p53-related protein p63 (TP63) in 80% of cases. The major protein isoform of p63 expressed in SCC is ΔNp63α, an N-terminally truncated form which functions as a key SCC cell survival factor by mechanisms that are unclear. In this study, we show that ΔNp63α associates with histone deacetylase 1 (HDAC1) and HDAC2 to form an active transcriptional repressor complex that can be targeted to therapeutic advantage. Repression of proapoptotic Bcl-2 family member genes including p53 upregulated modulator of apoptosis (PUMA) by p63/HDAC is required for survival of SCC cells. Cisplatin chemotherapy, a mainstay of SCC treatment, promotes dissociation of p63 and HDAC from the PUMA promoter, leading to increased histone acetylation, PUMA activation, and apoptosis. These effects are recapitulated upon targeting the p63/HDAC complex selectively with class I/II HDAC inhibitors using both in vitro and in vivo models. Sensitivity to HDAC inhibition is directly correlated with p63 expression and is abrogated in tumor cells that overexpress endogenous Bcl-2. Together, our results elucidate a mechanism of p63-mediated transcriptional repression and they identify the ΔNp63α/HDAC complex as an essential tumor maintenance factor in SCC. In addition, our findings offer a rationale to apply HDAC inhibitors for SCC treatment. Cancer Res; 71(13); 4373–9. ©2011 AACR.
Understanding the biochemical basis for tumor maintenance is critical to the rational application of targeted therapeutic agents. In squamous cell carcinoma (SCC), the p53 family member p63 is a key survival factor whose inhibition by RNA interference induces apoptosis, and whose degradation by cisplatin-based chemotherapy is thought to be important for the therapeutic response to this agent (1–4). The p63 gene is expressed through two promoters as two distinct isoform classes which either contain (TAp63) or lack (ΔNp63) an N-terminal transactivation domain. Additional isoform heterogeneity is generated through alternative C-terminal splicing (5). Consistently, the major p63 isoform overexpressed in SCC is ΔNp63α, a protein which has been shown to function as a positive and negative transcriptional regulator of different target gene subsets (5, 6).
Given its potential therapeutic relevance, precisely how ΔNp63α mediates tumor-cell survival is under intensive investigation. We previously showed that ΔNp63α functions in part by binding and suppressing the proapoptotic activity of the related p53 family member p73 (1, 7). Whether binding to p73 is sufficient for tumor-cell survival in this setting is unresolved. In addition, we and others have observed localization of p63 to the promoters of proapoptotic Bcl-2 family members including p53 upregulated modulator of apoptosis (PUMA), raising the possibility that ΔNp63α functions as an active transcriptional repressor (1, 8). Here, we use biochemical approaches to identify an endogenous repressor complex involving ΔNp63α, histone deacetylase 1 (HDAC1), and HDAC2, and we show the potential relevance of p63/HDAC-mediated transcriptional repression in the response to cisplatin chemotherapy and HDAC inhibitor therapy in SCC.
Materials and Methods
Cell lines and xenograft assays
Cell lines JHU-029, JHU-011 (1), and HO1N1 (9); KYSE-30, KYSE-150 (10); and FaDU (11) were the generous gifts of David Sidransky (Johns Hopkins University), S. Michael Rothenberg (MGH), and James Rocco (MGH), respectively. Each line was maintained by the MGH Center for Molecular Therapeutics cell line bank and underwent high-density single nucleotide polymorphism (SNP) typing, revealing that each was unique compared with > 800 other banked lines. Xenograft tumors were generated by subcutaneous injection of 2 × 106 JHU-029 tumor cells and 106 NIH 3T3 cells suspended in 1:1 matrigel (BD Biosciences): RPMI.
Lentiviral and retroviral production, luciferase assays, and mRNA quantitative reverse transcriptase–PCR
Production of virus, luciferase assays, and mRNA analysis were carried out as described (1). Primers used for QRT-PCR are shown in Supplementary Table S1.
Preparation of nuclear extracts and glycerol density gradient fractionation
Nuclear extracts were prepared by suspending cells in hypotonic buffer (10mmol/L Tris-HCl pH 7.5, 1.5 mmol/L MgCl2, 10 mmol/L KCl) for 20 minutes, followed by douncing. Pelleted nuclei were suspended in 1 volume 20 mmol/L KCl nuclear buffer (20 mmol/L Tris-HCl pH 7.5, 1.5 mmol/L MgCl2, 0.2 mmol/L EDTA, 25% glycerol). One volume 1.2 mol/L KCl nuclear buffer was added dropwise then incubated for 30 minutes at 4°C with rotation. Cleared supernatant was dialyzed against BC-100 buffer (100 mmol/L KCl, 20 mmol/L Tris-HCl pH 7.5, 0.2 mmol/L EDTA, 20% glycerol). Glycerol density gradient fractionation was carried out as previously described (12).
Tandem affinity purification
Cells were stably infected with pMSCV-ΔNp63α- FLAG-HA (C-terminal) or pMSCV-GFP-FLAG-HA plasmids, and cleared lysates from nuclear extracts were incubated for 4 hours with α-FLAG conjugated beads. Beads were washed with 100 mmol/L, 250 mmol/L, 500 mmol/L, 250 mmol/L, and 100 mmol/L KCl wash buffer (50 mmol/L Tris-HCl pH 7.5, 5 mmol/L MgCl2, 0.2 mmol/L EDTA, 0.1% NP-40, 10% glycerol). Immune complexes were eluted with 0.5mg/mL FLAG peptide in 150 mmol/L KCl wash buffer. Eluate was incubated 12 hours at 4°C with α-HA conjugated beads. Beads were washed with 100 mmol/L, 200 mmol/L, 250 mmol/L, 200 mmol/L, and 100 mmol/L KCl wash buffer and boiled in Laemmli buffer. Proteins were visualized using the SilverQuest Silver Staining Kit (Invitrogen).
Immunoprecipitation and chromatin immunoprecipitation
Cleared nuclear lysates were incubated with antibody and protein A beads for 3 hours at 4°C, and immunocomplexes were washed with 100 mmol/L, 250 mmol/L, 400 mmol/L, 250 mmol/L, 100 mmol/L KCl wash buffer. For transient transfections, 293T cells were transfected with pcDNA-ΔNp63α–FLAG (C-terminal) mutants and pcDNA3-HDAC1. Forty hours posttransfection, cells were washed with cold PBS and incubated in hypotonic buffer for 20 minutes at 4°C. Following sonication, 3 mol/L KCl was added dropwise to a final concentration of 150 mmol/L and proteins were immunoprecipitated as above. ChIP was carried out as previously described (13) with modifications detailed in Supplementary Methods.
P values were determined using the student's unpaired t test unless indicated otherwise. Pearson's Product-moment Correlation Coefficient (R2) was calculated and a two tailed P-value was generated from a probability table.
Results and Discussion
Interaction between endogenous p63 and HDAC1/2
In order to uncover the biochemical basis for p63-dependent transcriptional regulation, we isolated p63-associated nuclear proteins from JHU-029, a human squamous cell carcinoma (SCC)-derived cell line in which endogenous p63 functions as an essential suppressor of apoptosis (1, 7). Using tandem affinity purification (TAP), we purified complexes from nuclear extracts of cells expressing either ΔNp63α-FLAG/HA or control nuclear GFP-FLAG/HA. Expected p63-associated proteins, including endogenous p63 and p73, were identified on silver-stained gels and subsequently confirmed by mass spectrometry (Fig. 1A) (1, 14). The next most abundant silver-stained band, observed consistently following ΔNp63α but not green fluorescent protein (GFP) purification, contained HDAC1 and HDAC2 proteins (Fig. 1A). To confirm the specificity of their interactions with p63, we conducted western analysis for HDAC1 and HDAC2 following TAP for tagged ΔNp63α or nuclear GFP control. Consistent with our mass spectrometry findings, endogenous HDAC1 and HDAC2 specifically interacted with ΔNp63α but not with nuclear GFP (Fig. 1A).
Using glycerol density gradient fractionation, we observed cofractionation of endogenous ΔNp63α, HDAC1, and HDAC2 in complexes greater than 440 kDa, potentially suggesting the presence of a complex involving these three proteins (Fig. 1B). To confirm the endogenous association we carried out reciprocal coimmunoprecipitations for p63, HDAC1, and HDAC2 in JHU-029 cells (Fig. 1C) and a second head and neck SCC (HNSCC) line, FaDU (Supplementary Fig. S1A), and observed a specific interaction between these three proteins. Finally, in order to examine these interactions in more detail we mapped the domain of p63 required for HDAC association. We transfected a series of FLAG-tagged p63 deletion mutants (Supplementary Fig. S1B) together with HDAC1 into 293T cells, and carried out immunoprecipitations using either α-FLAG (Fig. 1D) or α-HDAC1 (Supplementary Fig. S1C) antibodies. Remarkably, only the transactivation inhibitory domain (TID) of ΔNp63α was required for HDAC binding, whereas the sterile alpha motif (SAM) domain, a putative protein interaction domain, was entirely dispensable (5). Given the well-established association between HDAC1 and HDAC2 (15), our findings collectively suggest that ΔNp63α, HDAC1 and HDAC2 exist in a trimeric complex in SCC cells.
Requirement for p63 promoter association in p63-mediated repression
We hypothesized that p63 mediates direct transcriptional repression in SCC cells through recruitment of HDACs to the promoters of proapoptotic genes including PUMA. This hypothesis requires that p63 and HDACs are localized to this promoter, and that promoter binding by p63 is essential for its ability to repress transcription. We therefore carried out chromatin immunoprecipitation (ChIP) for p63 and HDAC1 in SCC cells, and observed specific binding of both endogenous proteins to the PUMA locus (Fig. 2A). Binding of p63 and HDAC1 was also observed within the regulatory regions of other p63-repressed genes (Supplementary Fig. S2A). To address the functional contribution of promoter binding by ΔNp63α we first used a PUMA promoter reporter assay (1). We co-expressed either wild-type ΔNp63α (WT) or a naturally-occurring DNA binding-deficient point mutant, ΔNp63α (R304W) (5), together with TAp73β or p53 and examined luciferase activity. Wild-type ΔNp63α was a potent suppressor of both p73 and p53-dependent PUMA reporter activation, while the non-DNA binding mutant ΔNp63α (R304W) was defective in suppressing activation (Fig. 2B). Of note, the p63 mutant was expressed at similar levels as the wild-type (Fig. 2B) and exhibits comparable binding to p73 (Supplementary Fig. S2B) but not to p53 (Supplementary Fig. S2C).
Because transient reporter assays lack chromatin context, we next tested whether suppression of endogenous PUMA transcription required DNA-bound p63. We expressed retroviral FLAG-tagged wild-type or mutant ΔNp63α (R304W) in SCC cells, then carried out ChIP using an anti-FLAG antibody. As expected, wild-type ΔNp63α showed significant binding to the PUMA promoter, whereas the mutant showed little or no binding over background (Fig. 2C). As a functional test we then ablated expression of endogenous p63 in these cells, having engineered the ectopic ΔNp63α constructs to contain silent point mutants, which made them resistant to the lentiviral shRNA (Supplementary Fig. S2D). Ectopic wild-type ΔNp63α nearly completely suppressed PUMA induction following endogenous p63 knockdown, whereas mutant ΔNp63α-expressing cells showed dramatic PUMA induction (Fig. 2D and Supplementary Fig. S2E) and cell death (Supplementary Fig. S2F) in this setting. Taken together, these data show the requirement for promoter-bound p63 in suppression of endogenous PUMA transcription and cell death in SCC cells.
HDAC and p63-dependent regulation of PUMA and chemotherapy response in squamous cell carcinoma cells
Having documented the presence of endogenous HDAC1 and p63 at the PUMA promoter (Fig. 2A) we wished to test the biochemical requirement for HDAC activity in PUMA regulation. Treatment with the potent class I/II HDAC inhibitor trichostatin A (TSA) caused a dose-dependent induction of PUMA mRNA in three different SCC cell lines (Fig. 3A). A similar dose-dependent induction of PUMA was observed following treatment with vorinostat (SAHA), a second generation inhibitor which is currently Food and Drug Administration–approved for treatment of cutaneous T-cell lymphoma (CTCL) (Supplementary Fig. S3A) (16). PUMA mRNA induction by TSA corresponded temporally with increased histone H4 acetylation at the p63 binding site within the PUMA promoter (Fig. 3B and S3B), consistent with a direct effect of the inhibitor at this promoter. In order to show directly a connection between the presence of p63 and HDAC activity at the PUMA promoter we assayed histone acetylation following p63 knockdown in SCC cells. Indeed, histone H4 acetylation was significantly induced following ablation of p63, concurrent with endogenous PUMA upregulation (Fig. 3C and Supplementary Fig. S3C). Thus, HDAC activity controls PUMA expression in SCC cells in a p63-dependent manner.
Cisplatin-based chemotherapy, a mainstay for treatment of advanced HNSCC, promotes degradation of ΔNp63α and induction of PUMA, which have been linked to the therapeutic response in this disease (3, 4). We found that disruption of the p63/HDAC complex contributes to the response to cisplatin, as PUMA expression induced by cisplatin (Supplementary Figs. S3D and S3E) was accompanied by a loss of endogenous p63 and HDAC1 at the PUMA promoter, and by an increase in histone acetylation (Fig. 3D and Supplementary Fig. S3D). Thus p63/HDAC-mediated PUMA transcriptional repression is mitigated in the physiological response to cisplatin chemotherapy.
Targeting p63/HDAC-dependent transcriptional repression in squamous cell carcinoma
We have shown previously that some SCC cell lines are able to bypass the requirement for ΔNp63α as a survival factor through overexpression of endogenous Bcl-2 itself (1, 17). Consistent with this observation, we found that SCC lines which exhibit low expression of ΔNp63α showed high-level expression of Bcl-2, and vice-versa [Fig. 4A and reference (1)]. We therefore hypothesized that lines with high ΔNp63α expression are “addicted” to ΔNp63α/HDAC function and therefore would be sensitive to HDAC inhibition, whereas lines with low ΔNp63α would exhibit HDAC inhibitor resistance. Indeed, we observed a direct correlation between ΔNp63α protein levels and sensitivity to TSA in SCC cells (Fig. 4B and Supplementary Fig. S4A). In addition, we found that ectopic Bcl-2 expression was sufficient to confer remarkable in vitro TSA resistance in the TSA-sensitive line JHU-029 (Supplementary Fig. S4B). Thus, although multiple pathways may contribute to effects of HDAC inhibition in SCC cells (18), these data support a prominent role for the p63-dependent pathway we describe here.
Finally, we sought to model HDAC inhibition for treatment of SCC in vivo and to determine the contribution of Bcl-2 expression in this setting. Notably, we recently showed that Bcl-2 expression in primary HNSCC is an intrinsic resistance factor and a powerful predictor of relapse following cisplatin-based therapy (17). We established a xenograft assay using JHU-029 cells, which form tumors in 100% of Nude mice when injected subcutaneously. Mice bearing palpable tumors, derived from JHU-029 cells expressing either a retroviral control (GFP) or Bcl-2 vector, were treated by IP injection with vorinostat or vehicle control. Vorinostat treatment substantially and consistently blocked tumor progression in mice with GFP-expressing tumors (Fig. 4C). Remarkably, however, expression of Bcl-2 induced complete resistance to vorinostat treatment (Fig. 4C). To determine the physiological basis for the response to HDAC inhibition in vivo we assayed markers of proliferation and apoptosis in these tumors. We observed no difference in proliferation following vorinostat treatment in any tumors, as assessed by Ki67 staining (Supplementary Figs. S4C, D). In contrast, control vorinostat-treated tumors showed substantial cleaved PARP-1 (Fig. 4D) and activated Caspase 3 (Fig. 4D), which were completely absent in Bcl-2-expressing tumors. All together, our findings show the presence of a functional p63/HDAC complex which serves as a direct repressor of the apoptotic transcriptional program in SCC. HDAC inhibitors target this complex to induce tumor cell killing through upregulation of proapoptotic Bcl-2 family members, whereas sensitivity to these drugs can be abrogated in tumor cells that overexpress Bcl-2.
These findings reveal a tumor-specific context for HDAC function in SCC, which will inform the rational and effective application of these agents. For example, a recent clinical trial of late-stage, chemotherapy-refractory HNSCC patients treated with vorinostat did not show clinical responses (19). This finding is consistent with our data showing that a common transcriptional and apoptotic response pathway involving p63 and HDAC1/2 seems to participate in the response to both cisplatin and HDAC inhibitors. Conceivably, treating patients earlier in the course of disease may improve the efficacy of HDAC inhibition in SCC. Our study also provides insight into a specific resistance mechanism, suggesting that HDAC inhibitors may not be useful as single agents in Bcl-2 positive SCCs. An attractive approach for these tumors might instead include Bcl-2 inhibitors, which are currently in clinical trials, either alone or in combination with HDAC inhibitors (20). If successful, such a stratified and targeted approach based on an understanding of tumor-selective biology would represent a significant advance against this disease.
Disclosure of Potential Conflicts of interest
No potential conflicts of interest were disclosed.
We thank Jonathan Whetstine, David Sweetser, and Anders Naar for helpful advice and reagents; Kristine Torres-Lockhart, Mary Lynch, Zachary Nash, and Catherine Wilson for technical assistance; and the Taplin Mass Spectrometry facility (Harvard Medical School) for protein identification.
NIH R01 DE-015945 (L.W. Ellisen and L. He); ACS/Mass Biotech Council Cancer Research Challenge-AstraZeneca Pharmaceuticals LP Fellowship PF-09–100-01 MGO (M.R. Ramsey); Tracey Davis Memorial Fund (N. Forster); and Fondation pour la Recherche Medicale (B. Ory).