AP-2α Induces Epigenetic Silencing of Tumor Suppressive Genes and Microsatellite Instability in Head and Neck Squamous Cell Carcinoma

Background Activator protein 2 alpha (AP-2α) is involved in a variety of physiological processes. Increased AP-2α expression correlates with progression in various squamous cell carcinomas, and a recent publication found AP-2α to be overexpressed in ∼70% of Head and Neck Squamous Cell Carcinoma (HNSCC) patient samples. It was found to repress transcription of the tumor suppressor gene C/CAAT Enhancer Binding Protein alpha (C/EBPα), and its binding site correlated with upstream methylation of the C/EBPα promoter. Therefore, we investigated the potential for AP-2α to target methylation to additional genes that would be relevant to HNSCC pathogenesis. Principal Findings Stable downregulation of AP-2α stable by shRNA in HNSCC cell lines correlated with decreased methylation of its target genes' regulatory regions. Furthermore, methylation of MLH1 in HNSCC with and without AP-2α downregulation revealed a correlation with microsatellite instability (MSI). ChIP analysis was used to confirm binding of AP-2α and HDAC1/2 to the targets. The effects of HDAC inhibition was assessed using Trichostatin A in a HNSCC cell line, which revealed that AP-2α targets methylation through HDAC recruitment. Conclusions These findings are significant because they suggest AP-2α plays a role not only in epigenetic silencing, but also in genomic instability. This intensifies the potential level of regulation AP-2α has through transcriptional regulation. Furthermore, these findings have the potential to revolutionize the field of HNSCC therapy, and more generally the field of epigenetic therapy, by targeting a single gene that is involved in the malignant transformation via disrupting DNA repair and cell cycle control.


Introduction
The Activator Protein-2 family is composed of five members (a, b, c, e, d). These genes are transcription factors that play roles in apoptosis, cell cycle regulation, tissue differentiation, and development [1]. Activator protein 2 alpha (AP-2a) is involved in a variety of processes, including adipogenesis [2], neuronal development [1], and is suspect in cancer progression [3]. AP-2a binds as a homo or heterodimer to targets that contain the motif ''GCCNNNGGC'' in their regulatory regions [4]. After binding, AP-2a can either function as a transcriptional activator or repressor [5,6]. It is normally expressed at very low levels in keratinocytes, and it is elevated in proliferating, malignant keratinocytes of squamous cell carcinoma of the skin [7]. Furthermore, UVA irradiation increases AP-2a expression and results in malignant transformation [3,8]. Because AP-2a transcriptionally regulates such genes as p21 and CDH1, it is feasible to propose its relevance in tumorigenesis [9]. In fact, AP-2a has been shown to act as both a tumor suppressor and an oncogene in different cancer types [3,5,8]. It appears this is determined by the type and quantity of AP-2 isoforms and AP-2 modulating factors present in the cell [9].
Histone and promoter DNA methylation have been revealed in the past decade to be a frequent cause of gene silencing of tumor suppressors in a variety of malignancies [10,11]. DNA methylation sterically inhibits transcription factor binding, thereby inhibiting gene activation [12]. Conversely, decreased gene transcription has been shown in some cases to precede and propogate DNA methylation, resulting in more stable gene downregulation. For example, a transcriptional suppressor protein that competes with a transcriptional activator for the same binding site could result in decreased activation through increased methylation [13].
AP-2a is one such example, providing transcriptional repression to some genes which are also epigenetically modified [14,15]. It competes with the activator SP1 for binding to these target promoters, thereby repressing transcription [2,16]. AP-2a has previously been shown to recruit HDACs to specific gene promoters following retinoic acid induction [15], and a recent study observed that regions of frequent methylation in Acute Lymphoblastic Leukemia often contained AP-2a binding sites [17]. Therefore, it is reasonable to hypothesize that in cancer cells AP-2a may bind to specific target genes, silence transcription, and propagate DNA methylation in order to stably silence tumor suppressors and DNA repair genes.
The following study investigates 8 target genes in HNSCC for the possibility of AP-2a-induced targeted methylation and subsequent gene silencing. Decreased AP-2a expression was found to correlate with decreased target methylation, increased target gene expression, and decreased microsatellite instability. ChIP analysis revealed an association between AP-2a and HDAC binding to target promoters. Also, the effect of HDAC inhibition on target methylation mirrored that of AP-2a downregulation. Therefore, the mechanism of AP-2a-induced methylation is likely explained by HDAC recruitment.

AP-2a downregulation decreases target gene methylation
Using a previously generated HNSCC cell line (SCC22B) in which AP-2a is stably downregulated [14] (Figure 1a), DNA methylation was analyzed in 16 potential AP-2a target genes' promoters (list shown in Table S1). These genes were selected based on previous evidence for methylation in a variety of malignancies and potential regulation by AP-2a (i.e. the presence of consensus AP-2a binding sites in the target's promoter). A preliminary methylation screen was performed in wildtype SCC22B cells, which yielded methylation in 8/16 of the targets tested: RARB2, DCC, DAPK, PTEN, MLH1, CDH1, P73, and RASSF1A (Figure 1b and Figure S2). Bisulfite sequencing was then performed on these eight genes in SCC22B cells with and without AP-2a downregulation (detailed gene diagrams are displayed in Figure S1). Of these eight genes, five genes (RARB2, PTEN, MLH1, CDH1, and RASSF1A) showed some level of demethylation in the SCC22B cells with AP-2a downregulation compared to the wildtype cells (Figure 1b; graphical representation in Figure 1d).

Decreased target gene methylation correlates with decreased expression
In order to assess the significance of the target methylation, western blot analysis was performed on lysates from the wildtype and AP-2a-downregulated SCC22B cells. This revealed a visible increase in protein expression in 3/5 of the genes (i.e. PTEN, MLH1, and RASSF1A) and no detectable change in RASGRF2 (Figure 1c). Not surprisingly, CDH1 showed a decrease in protein expression in response to AP-2a-downregulation, which correlates with previous findings that AP-2a is required for activation, not repression, of this gene (Figure 1c) [18]. AP-2a has previously been shown to regulate PTEN [19], but the findings of AP-2a regulation of MLH1 and RASSF1A are novel.
AP-2a was also downregulated in additional HNSCC cell lines (SCC11B and SCC17as) (Figure 2a). The methylation status of the 3 genes with decreased methylation and increased expression in AP-2a-downregulated SCC22B cells (i.e. RASSF1A, MLH1, and PTEN) was analyzed. SCC17as displayed subtle decreases in PTEN and MLH1 methylation in the AP-2a-downregulated cells (Figure 2b and 2d), and both SCC11B and SCC17as revealed increased target gene expression after AP-2a downregulation (Figure 2c). AP-2a-targeted methylation was also investigated in malignancies other than HNSCC, and colon cancer was selected based on the association of MLH1 methylation [20,21,22]. AP-2a was downregulated in two colon cancer cell lines, HCT116 and HT29 ( Figure S3a). The region analyzed for MLH1 showed insignificant methylation levels in the vector-only cancer cells for both HT29 and HCT116 ( Figure S3b). However, both PTEN and RASSF1A demonstrated slightly higher levels of methylation, which were decreased in correlation with AP-2a downregulation in both HT29 and HCT116 ( Figure S3c).
In order to confirm AP-2a actually binds to the MLH1 promoter, ChIP analysis was performed in SCC22B with and without AP-2a downregulation. This revealed AP-2a does bind to the MLH1 promoter, and this binding is decreased in the AP-2adownregulated cells (Figure 3b). Methylation screening was performed for additional DNA repair genes (i.e. MSH3, PMS2, and RAD50) that contained both CpG islands and potential AP-2a binding sites within their promoter elements. This bisulfite sequencing analysis revealed relatively no methylation of these genes (Figure 3c).
Decreased AP-2a and HDAC1/2 binding to target promoters following AP-2a downregulation In a previous publication, AP-2a binding to target promoters was shown to precede recruitment of HDAC proteins to the region for more stable silencing [15]. Therefore, to address the mechanism by which AP-2a may affect gene methylation and silencing in these HNSCC cells, ChIP analysis was performed on SCC22B cells with and without downregulation of AP-2a. As expected, this semi-quantitative analysis revealed AP-2a and HDAC1/2 binding to all of the target promoters tested, except PTEN, in the wildtype SCC22B cells (Figure 4a). Furthermore, both AP-2a and HDAC binding was reduced in the AP-2a downregulated cells. HDAC1 was found to be more frequently associated with the targets compared to HDAC2, which did not pulldown RASSF1A. Since AP-2a and SP1 bind to the same consensus sequence, SP1 binding was also investigated in order to assess whether increased target gene expression was due to the restored binding of the activator SP1 following removal of AP-2a. With the exception of CDH1, decreased AP-2a binding did not allow increased SP1 binding as would have been expected.

HDAC inhibition's effect on target methylation correlates with AP-2a downregulation
In order to investigate whether the mechanism of AP-2atargeted methylation involved HDAC recruitment and activity, Efficacy of TSA treatment was verified via western analysis for acetyl histone H4K5 (Figure 4b). Following TSA treatment, target methylation was then compared with that of the SCC22B cells with intact HDAC activity. Both PTEN and MLH1 demonstrated a significant decrease in methylation subsequent to TSA treatment; however, RASSF1A remained unchanged (Figure 4c; ** P-value = 3.77610 26 and * 0.0452, respectively; graphical quantitation shown in Figure 4d). This correlates with previous findings that TSA alone is insufficient to change the histone acetylation status of RASSF1A [23]. However, no obvious change in target gene expression resulted from the increased acetylation (data not shown). This may be due to the brief time course (24 hours), which may not allow enough time for the methylation changes to reflect their impact on expression [24].

Lack of AP-2a inhibits remethylation of PTEN following 5aza-29 deoxycytidine treatment
In order to exclude the possibility that AP-2a is affecting DNA methylation at the global level, an methylation sensitive digest was performed on DNAs from SCC22B cells with and without AP-2a downregulation. This revealed no substantial difference in DNA methylation levels between the two cell types (Figure 5c), suggesting the DNA methylation changes from AP-2a are genespecific. To determine whether AP-2a acts to maintain or induce this specific target DNA methylation, the HNSCC cells SCC22B (+/2 AP-2a shRNA) were treated with a demethylating drug (5-aza-29deoxycytidine) and then permitted to recover methylation by continued culturing in the absence of the drug. Cells collected at 7 days post-treatment showed no PTEN methylation in both cell types (Figure 5d). However, after 10 days PTEN methylation was returning in the SCC22B cells with AP-2a expression, whereas, PTEN remained unmethylated in the cells lacking AP-2a expression ( Figure 5d). Therefore, this is suggestive of a mechanism by which AP-2a induces target methylation.

Discussion
It has previously been shown that AP-2a is expressed at relatively low levels in normal keratinocytes and is significantly elevated in cancerous and proliferating keratinocytes in squamous cell carcinoma of the skin [7]. Forced expression of AP-2a in keratinocytes causes increased expression of the oncogene Amyloid Precursor Protein (APP), and these two genes have both been found to be increased in squamous cell carcinomas [25]. Furthermore, AP-2a interacts with SMAD2/3 to affect TGFB signaling, which is frequently disrupted in HNSCC [26]. These findings reveal the potential importance of AP-2a's transcriptional regulation of target genes in the pathogenesis of HNSCC.
It has been suggested that AP-2a's ability to repress target genes results in cellular transformation and tumorigenesis [27], and decreased transcriptional suppression of AP-2a targets via abolished target binding has been shown to favor cellular differentiation [2,28]. In agreement, the results of this study in HNSCC reveal that downregulating AP-2a increases target gene expression by decreasing promoter DNA methylation and results in decreased cellular proliferation. In addition to its effects on proliferation, AP-2a also has been shown to potentially impact microsatellite instability through its interaction with p53 and subsequent RAD51 suppression [29]. In this study, we demonstrate that AP-2a expression correlates with increased MSI in HNSCC cell lines, which correlates with MLH1 methylation and decreased expression.
It has been demonstrated that AP-2a may help to recruit HDACs to the promoters to which it is bound [15]. Therefore, we investigated the association between AP-2a and HDAC binding to target promoters. This revealed a correlation between AP-2a and HDAC binding, and HDAC inhibition via TSA yielded similar effects on target demethylation as AP-2a downregulation. In fact, in some cases such as SCC11B and SCC17as, deacetylation via HDAC recruitment may be the primary means of target silencing by AP-2a (since AP-2a downregulation yielded no accountable methylation changes). Therefore, this study reveals that AP-2a acts as a suppressor for certain ''tumor suppressive'' genes in HNSCC by targeting promoter methylation and/or deacetylation via HDAC recruitment.
Perhaps most interesting is the novel discovery that AP-2a is not only involved in the epigenetic silencing of tumor suppressor and DNA repair genes, but also appears to contribute to microsatellite instability in HNSCC. In this manner, AP-2a allows for both increased tumor proliferation and increased genomic instability. Furthermore, this phenomenon in HNSCC is distinct from the colon cancer cell lines used in this study, in which MSI appears to be unaffected by AP-2a expression changes. In light of a recent publication demonstrating AP-2a overexpression in 69% of HNSCC patients [14], these findings emphasize the impact AP-2a has in HNSCC pathogenesis. Therefore, clinical treatment that targets AP-2a expression in the context of the MSI-positive HNSCC tumor would potentially deter tumor proliferation (ie. via PTEN and RASSF1A reexpression) and resolve genomic instability (ie. via MLH1 reexpression).

Cell lines, antibodies, and plasmids
The established human HNSCC cell lines used in the study (SCC22B, SCC11B, and SCC17as) were kindly provided by Dr.  AP-2a overexpression construct was made by PCR amplification of the complete AP-2a cDNA sequence, using primers with EcoRV and SalI ends. Blunt-end ligation was performed using pBABE following BamHI digestion and Klenow polymerase reaction.

DNA isolation
DNA was isolated using the DNA QiaAmp kit (Qiagen) according to the manufacturer's protocol.

Trichostatin A (TSA) treatment
The SCC22B cell line was treated with either 300 nM or 5 mM TSA for 24 hours. The cells were then collected for western blot analysis and DNA isolation. The 5 mM treated cells were used for the bisulfite sequencing analysis.

5-aza-29deoxycytidine treatment
Cells were treated with 1 mM of the drug (Sigma-Aldrich) for 96 hours while maintaining ,50% confluency and changing the drug daily.

Western blot analysis
Whole cell lysates were suspended in lamaelli buffer and boiled for 15 minutes, followed by centrifugation. Lysates were loaded onto a 7.5% acrylamide gel and run for 2 hours at 110 volts followed by semidry transfer of the proteins to nitrocellulose membranes. The membranes were then blocked in 5% milk for 1 hour. 1:1,000 dilution of the primary antibodies were incubated with the blot overnight shaking at 4 degrees Celsius. A one hour incubation with the secondary antibodies (1:5,000 dilution) followed. The blots were then washed in 1x TBS-Tween for 1 hour and films exposed using ECL solutions. Western blots are representative of three independent experiments.

Combined Bisulfite Restriction Analysis (COBRA)
COBRA was performed as previously described [34]. Primer sequences will be provided upon request.

Bisulfite sequencing analysis
Bisulfite sequencing was performed as previously described [35]. Primer sequences will be provided upon request.

Chromatin immunoprecipitation analysis
ChIP analysis was done as previously described [34], according to the Upstate Cell Signaling Solutions protocol. Sequences of the primers used for the quantitative ChIP PCRs can be provided upon request.

Hha I digest
Global methylation status was analyzed in SCC22B cells +/2 AP-2a shRNA. 1ug of DNA was incubated with 1 unit HhaI (New England Biolabs) at 37 degrees Celsius for 3 hours. The digests were then loaded onto a 1% agarose gel for comparison.

Supporting Information
Table S1 Found at: doi:10.1371/journal.pone.0006931.s001 (0.08 MB RTF) Figure S1 Diagram of 5 genes displaying the amplified region in respect to the corresponding gene and associated CpG island. Bisulfite sequencing analysis was performed on five genes which yielded differences in methylation following AP-2a downregulation in HNSCC. The diagram shows the location of the bisulfite PCR product, the CpG island, and the gene layout. The numbers provided are in respect to the transcription start site.