The DNA damage response cannot account for low-dose Aza induced growth arrest

We first assessed whether the DNA damage response is the basis for growth arrest in response to low-dose Aza treatment in our panel of melanoma cell strains, because Aza at even 0.1 µM can induce DNA damage in human lung cancer cell lines [20]–[23], and concentrations of ~1 µM and above also activate p53, resulting in p21^Cip1 induction and cell cycle arrest [20]–[22]. We performed the Comet assay which measures DNA damage at the level of individual cells. This test revealed that 0.5 µM and 1.0 µM, but not 0.2 µM Aza induced DNA damage in the Aza sensitive YUMAC, but not the resistant YURIF melanoma cells ( Table 2 ). Furthermore, additional tests excluded the induction of double-strand break DNA repair and activation of cell-cycle checkpoints after low-dose Aza, because: a) there were no changes in the levels of phosphorylation of proteins known to transmit the ATR (ataxia telangiectasia mutated (ATM) and ATM and Rad-3 related) response, CHK1, and gamma-H2AX, reported to be activated in response to high dose Aza (1–10 µm) [20]–[22] (data not shown); b) there was no accumulation of p53 phosphorylated forms (Ser37 and Ser20); and c) there was no induction of BAX, that was expressed at equal levels in these cells, or additional p53 signature genes, such as GADD45.

We concluded that 0.2 µM did not cause DNA damage. We therefore, explored gene reactivations that lead to changes in specific signaling pathways as the mechanism of Aza cellular responsiveness, and attempted to identify markers by interrogating specific genes revealed by the bioinformatic analyses.

Whole-genome gene expression profiling in response to Aza

Unsupervised hierarchical clustering based on similarity of genome-wide expression profiles of the eight melanoma cell strains confirmed low variability between replicate experiments, indicating high quality of results ( Figure 2A ). The clustering of an Aza treated cell strain with its untreated counterpart shows that relatively few genes were affected by Aza treatment. The dendogram also suggested that pre-treatment gene expression by itself harbors important information with respect to Aza responsiveness because sensitive YUMAC, YULAC and YUSIT1, clustered separately from the Aza resistant YURIF and 501 mel melanoma cell strains

Open in a separate window

Figure 2

Bioinformatic analysis of whole genome expression arrays.

Panel A. Unsupervised hierarchical clustering of absolute intensity values. The vertical scale indicates 1-pearson's correlation coefficients as a measure of similarity. Panel B. Heatmap of differentially expressed sequences after treatment with low-dose Aza. Panel C. DNMT1 expression at the end of 3-days treatment with Aza (0.2 µM). Cell extracts were subjected to Western blot with anti-DNMT1 antibodies. The same membrane was successively blotted with anti-b-actin antibodies as a measure for protein load in each well. Panel D. Pie chart of the most over-represented Gene Ontology terms (p-value<1e-3); the size is relative to the number of represented genes, and the color represents the enrichment p-value. Panel E. SFRP1 transcripts in melanoma cell strains as assessed by the oligonucleotide array hybridization. The data represent one of two sequence IDs with similar results. The error bars represent the Standard Deviations (SD). One, two, three stars refer to p-value less than 0.05, 0.01 or 0.001, respectively. We determined p-values by unpaired t-test (Aza vs. Untreated). The broken line in this and all subsequent figures separates sensitive (left hand side) from resistant (right hand side) cell strains.

We identified 396 sequence ids, representing 292 genes that were differentially expressed across the cell strains following treatment with low-dose Aza ( Figure 2B ; all microarray data will be deposited in GEO). At least 50 genes in our list are already known to be regulated by DNA methylation, such as those encoding cancer antigens (a set of MAGE and GAGE), H19, S100A4, IGFBP4, UCHL1, COL1A2, CLU, FN1, and TGFBI ( Figure 2B ). We did not see consistent re-expression of genes that have been previously reported to be under epigenetic control and/or reactivated by low concentration of decitabine (0.1 µM) in established uveal melanoma cell lines, such as S100A2 [24], and two melanoma cell lines, such as Apo2L/TRAIL and XAF1 [18], or in response to higher concentrations of decitabine, such as PTEN, HOXB13, APC, RASSF1A, RARB and MGMT [10], [25], [26]. Surprisingly, the Aza sensitive YUSIT1 displayed the lowest intensity of absolute change in gene expression compared to the other melanoma cell strains ( Figure 2B ). This difference might be attributed to higher levels of DNMT1 post Aza treatment in these cells compared to the other cell strains ( Figure 2C ).

The sequence ids have been evaluated for over-representation of functional families based on analysis of enrichment of GeneOntology (GO) terms. Among the highest scoring GO terms were genes associated with the extracellular region, matrix, response to wounding and external stimulus, protease inhibitor activity, and genes associated with acute inflammatory and immune responses ( Figure 2D ). However, we could not find any particular category that could segregate the relatively Aza sensitive from the resistant cell strains.

To better understand cellular responsiveness to the drug, we compared the pre-treatment gene expression levels between sensitive and resistant melanoma cell strains. This analysis uncovered 94 genes (141 sequences) with differential pre-treatment expression. Among them was SFRP1 (Secreted frizzled-related protein 1), the negative regulator of Wnt signaling, whose basal transcript levels were particularly low in the resistant compared to the sensitive melanoma cell strains ( Figure 2E ). Because SFRPs are known to be regulated by DNA methylation and to be reactivated by Aza [27], [28], and Wnt signaling plays an important role in melanoma biology [29], we further explored the role of this pathway in Aza differential responsiveness.

Activated Wnt/β-catenin/MITF pathway confers resistance to Aza

The known role of Wnt downstream target β-catenin compelled us to assess the involvement of β-catenin activating mutation in Aza growth-arrest resistance. Re-sequencing of CTNNB1 exon 3 revealed D32H, and not S37F substitution reported before [30], in our 501 mel cells, and S33C mutation in YURIF melanoma tumor and cultured cells, whereas the gene in all the other cell strains was normal ( Figure 3A ). YURIF and 501 mel melanoma cells also expressed high constitutive levels of β-catenin ( Figure 3B, β-catenin ), as expected from the increased resistance to proteasomal degradation conferred by the activating mutations. In all the others, β-catenin levels were low but re-appeared in the presence of the proteasomal inhibitor MG132. Treatment with Aza upregulated β-catenin in YUGEN8, and to a lesser extent in YUMAC melanoma cells, in agreement with the arrays hybridization intensities.

Open in a separate window

Figure 3

Activated Wnt/β-catenin/MITF pathway confers resistance to Aza.

Panel A. Chromatograms showing CTNNB1 activating mutations in 501 mel cells (GAC/CAC) and YURIF tumor (TCT/TGT) (marked by brackets and arrows), which lead to D32H and S33C mutations, respectively. The same results were obtained with YURIF short term cultured cells. Panel B. Expression of β-catenin and MITF in melanoma cell strains relative to β-actin. CTNNB1 mutation status for each cell strain is indicated at the top. Panel C. siRNA knockdown of β-catenin and downstream targets. Parallel cultures were untreated or treated with Aza (0.2 µM) for 2 days followed by transient transfection with three different CTNNB1 siRNA or with Alexa Fluor as a control for one day. Cell extracts were subjected to successive Western blotting with β-catenin, MITF, BCL2, Myc, and β-actin. Panel D. The same cultures as in panel C were assessed for apoptosis employing the Caspase 3/7 assay. Bars indicate SD of 3 replicate wells.

Among the β-catenin target genes is MITF, encoding a melanocyte-specific transcription factor that can interact with β-catenin to modulate the Wnt signaling and cell growth [29]. MITF transcripts were similar or slightly lower than those in normal melanocytes (data not shown), suggesting that the gene was not amplified. However, basal levels of MITF proteins varied among the cell strains, being particularly high in 501 mel and WW165 cells ( Figure 3B, MITF ).

The role of β-catenin in conferring resistance to Aza was further explored by knockdown experiments. Transient CTNNB1-directed siRNA knockdown caused about 70% reduction in β-catenin levels compared to Alexa fluor treated cells ( Figure 3C ) or control siRNA (data not shown). In addition, there was repression of MITF, as well as the anti-apoptotic BCL2. Myc, the other known β-catenin target gene, was downregulated by CTNNB1 knockdown but only in control cells and not in those treated with Aza ( Figure 3C ). Furthermore, downregulation of β-catenin sensitized the 501 mel resistant cells to Aza mediated apoptosis ( Figure 3D ), suggesting that β-catenin signaling shields these resistant melanoma cells from undergoing apoptosis. The results are in agreement with the observations that the MITF promoter is responsive to Wnt signaling in melanocytes, that β-catenin binds and trans-activates MITF, and that β-catenin induced melanoma growth requires MITF [31]. Unfortunately, similar experiments could not be conducted with YURIF melanoma cells because transfection with CTNNB1-directed siRNA failed to produce any reduction in β-catenin protein (data not shown).

We went on to explore the role of individual reactivated genes and their protein products in Aza responsiveness, focusing on pathways known to induce growth arrest and/or apoptosis, and further examination of known Aza effects other then DNA demethylation.

Activation of p21^Cip1 in a p53 independent manner

Although the DNA damage response was ruled out, close examination of the oligonucleotide gene expression data showed two-fold increases in CDKNA1 (encoding p21^Cip1) transcripts in some melanoma cell strains ( Figure 4A ). Therefore, we assessed p21^Cip1 levels in cells treated with low-dose Aza, in the absence and presence of MG132, supplemented to the medium 6 hr before harvest in order to prevent proteasomal degradation known to affect the stability of this protein in melanoma cells [32]. Western blotting revealed strong induction of p21^Cip1 in response to Aza in YUMAC, YUSAC, and YUGEN8 melanoma cells, and less so or not at all in the other cell types. The gene was induced in YUMAC cells null for TP53 ( Figure 4 panels A and B , compare p53 to p21^Cip1), conforming our previous conclusion that p53 does not mediate growth arrest. All the p53 expressing melanoma cell strains possessed the P72R variant but none carried an inactivating mutation in exon 4 of p53. P72R is a common allele in Caucasians, from which these melanoma tumors were isolated. Therefore, the different levels of p53 in this panel of melanoma cells could not be explained by TP53 genetic variation.

Open in a separate window

Figure 4

TP53-independet CDKN1A reactivation and promoter methylation.

Panel A. Reactivation of CDKN1A in melanoma cells in response to Aza (0.2 µM) as revealed by oligonucleotide array hybridization. The data represent one of two sequence IDs with similar results. All other details as in Figure 2 Panel E. Panel B. Expression of p21^Cip1 and p53, with β-actin serving as a control. Parallel cultures of melanoma cells were untreated (−) or treated (+) with Aza (0.2 µM). MG132 (20 µM) was added 6 h prior to harvesting the cells where indicated (+). The levels of p53 protein were in agreement with gene transcript levels showing that TP53 was inactivated in YUMAC (absolute hybridization intensities values of ~220, compared to 8,000–12,000 in the melanoma cell strains). Here and in all other Western blots numbers on the left mark the location of prestained protein markers in KDa, heavy and light frames designate Aza resistant and sensitive cells, respectively. Panel C. BS sequencing results of CDKN1A proximal promoter (−214 to +20 relative to TSS). Melanoma cells were untreated (control), and Aza (0.2 µM) treated as described in Panel B. Symbols: Arrows indicate the TSS; open and black circles, unmethylated and methylated CG pairs, respectively; dark and light grey circles indicate about 50% and 10% methylated CG, respectively. Numbers on the bottom indicate bp location relative to TSS.

TP53-independent induction of p21^Cip1 in leukemic cells was attributed to decitabine-induced re-expression of the tumor suppressor p73, an upstream regulator of p21^Cip1 [33], [34]. However, p73 is not expressed or induced in melanoma cells by Aza. The suppression of CDKN1A by methylation of the proximal promoter in senescing fibroblasts [35], prompted us to explore direct methylation and demethylation as the cause for p21^Cip1 silencing and reactivation, respectively. Indeed, the CDKN1A promoter was highly methylated only in 501 mel non-expressing cells, and underwent partial demethylation after treatment with low-dose Aza ( Figure 4C, 501 mel, compare control to Aza ). Interestingly, a cluster of seven CpG dinucleotides proximal to the transcription start site (TSS) remained fully methylated after treatment, which may explain the weak reactivation of p21^Cip1 in 501 mel cells ( Figure 4A, 501 mel ). In contrast, this promoter region was not methylated in any of the other melanoma cell strains ( Figure 4C ), in agreement with basal p21^Cip1 gene transcripts and protein ( Figure 4A and B ).

Because low-dose Aza induced CDKN1A in several melanoma cell strains in which the promoter was unmethylated, we attempted to identify additional genes that behave in a similar fashion whose reactivation can lead to growth arrest and further explored the mechanism of Aza activity.

Activation of CLU by DNA demethylation

The global oligonucleotide gene expression data showed reactivation of TGFβ induced genes (CLU and TGFBI), which encode secreted proteins with potential to be markers for Aza-responsiveness. Clusterin levels in Aza treated cells were also assessed in the presence of MG132, because CLU, like p21^Cip1, is sensitive to proteasomal degradation [36]. Western blot confirmed that protein levels corresponded, in general, to CLU transcripts, except for 501 mel cells which displayed reactivated gene transcripts with barely detectable protein, and YUSIT1 cells which expressed very little CLU mRNA but nevertheless exhibited high levels of Clusterin after inhibition with MG132 ( Figure 5A, B , relative mRNA levels confirmed by Real-Time RT-PCR, Figure 6D , and data not shown). Nevertheless, the presence and absence of the protein correlated with the pattern of drug sensitivity and resistance, respectively, enhanced by blocking proteasomal degradation ( Figure 5B ). However, BS modified DNA sequencing showed that promoter methylation could not fully explain Clusterin basal and reactivated expression. CLU proximal promoter was methylated in a CpG rich island about 120 bp downstream of TSS, which underwent demethylation in response to Aza to the same extent in YUMAC, YUGEN8, WW165 and 501 mel cells ( Figure 5C ). Furthermore, the promoter was unmethylated in the non-expressing, untreated YURIF melanoma cells ( Figure 5C, YURIF ), results reminiscent of p21^Cip1 activation.

Open in a separate window

Figure 5

CLU reactivation and promoter methylation.

Panel A. CLU re-expression in melanoma cells in response to Aza (0.2 µM) as assessed by the oligonucleotide array hybridization. The data represent one of three sequence IDs with similar results. All other details as in Figure 2 Panel E. Panel B. Clusterin expression as revealed by Western blots with anti-CLU antibodies. The results are representative of two biological replicas. Panel C. BS DNA sequencing results of the proximal CLU promoter and part of first exon regions. The bar indicated the CG island and the arrows the site of primers used for amplification. All other details as in Figure 4.

Open in a separate window

Figure 6

Reactivation of CDKNA1 and CLU by histone acetylation.

The Western blots show p21^Cip1 (Panel A) and CLU (Panel B) expression in YUMAC melanoma cells treated with increasing concentrations of Trichostatin A (TSA) overnight, as revealed by probing with the respective antibodies using β-actin as a control (left panel). Middle panels show expression of p21^Cip1 and CLU after 2-days treatment with Aza (0.2 µM), where indicated (+) followed by 1-day incubation with PXD101 (1 µm). Left panels: Real-Time PCR data comparing fold-difference in CDKN1A and CLU transcript in YUMAC and 501 mel cells after treatment with low-dose Aza and PXD101, alone and in combination, compared to non-treated cells. PXD101 (1 µM) was added for 24 hrs before harvesting the cells. Notice differences in scale of absolute hybridization intensities in YUMAC and 501 mel cells. However, 501 mel expressed about 370 fold less CDKN1A gene transcripts compared to YUMAC cells, apparently insufficient to lead to detectable p21^Cip1 protein. The basal levels of CLU transcripts in YUMAC were about 50 fold higher relative to 501 mel melanoma cells, in agreement with low protein levels (data not shown). Panel C. Growth response to combination treatment with HDAC inhibitors Trichostatin A (TSA) and PXD101. The sensitive YUMAC and resistant 501 mel cells were incubated in triplicate wells without or with Aza (0.2 µM and 0.4 µM as indicated) followed by one-day recovery in regular medium, or medium supplemented with increasing concentrations of TSA (a), or PXD101 (b). Cell viability was assessed with the CellTiter-Glo® Luminescent Cell Viability Assay. Data are presented as percent of control, non-treated cells.

Synergistic reactivation with HDAC inhibitor

Aza reactivation of the unmethylated CDKN1A and CLU promoters suggested de-repression by methylation-independent mechanism. Because Aza can reverse histone-mediated silencing of unmethylated CDKN1A and other promoters [37], and the histone deacetylase (HDAC) inhibitor Trichostatin A (TSA) acts synergistically with Aza to reactivate hypomethylated promoters [38], we explored reactivation by TSA and the clinical HDAC inhibitor PXD101 [39], [40]. Indeed, TSA induced p21^Cip1 and Clusterin in YUMAC cells in which the respective promoters are un- and hypo-methylated, respectively, in a dose dependent manner, even in the absence of MG132 ( Figure 6A, B, left panels ), but not p21^Cip1 in 501 mel cells in which the promoter is fully methylated (data not shown). Likewise, p21^Cip1 and Clusterin levels were increased in YUMAC and YURIF, but not in 501 mel cells, after overnight treatment with PXD101 ( Figure 6A, B, middle panels ). However, there was synergistic reactivation of p21^Cip1 and Clusterin in 501 mel cells when PXD101 was combined with Aza ( Figure 6A, B middle panels ). These results, confirmed at the mRNA level by Real-Time PCR ( Figure 6A, B, right hand panels ), are consistent with the notion that the un- and hypomethylated promoters of these two genes are suppressed in melanoma cells by acetylated histone H3, and that Aza can release HDAC1 suppression and can act in synergy with HDAC inhibitors, as reported for AML and colorectal carcinoma cells [37].

The growth inhibitory effect of combination treatment of Aza with TSA and PDX101 were further explored and shown in Figure 6Ca, b . In YUMAC melanoma cells, CI-isobol analysis showed that TSA acted synergistically while PXD101 acted at most additive when combined with Aza (Figure S2) This suggests that PXD101 may induce cell arrest by other mechanisms, independent of gene re-expression [41].

Reactivation of TGFBI

We were interested in reactivation of TGFBI (transforming growth factor, beta-induced, 68 kDa), because it is one of the novel genes that was not previously reported to be controlled by DNA methylation and it belongs to a set of ~11 genes active in normal human melanocytes, silenced in melanoma cells, and reactivated by low-dose Aza (COL1A2, CTSK, GLB1L, IL11RA, MMP1, RND2, SERINC2, STC1, TNFRSF10D, FLJ22662) ( Figures 2B and 7A ), and thus has the potential to serve as a marker for melanoma progression and responsiveness to Aza. The basal and Aza induced transcript levels of TGFBI were confirmed at the protein level ( Figure 7B ). However, unlike CLU, there was no complete separation between resistant and sensitive cells. Although TGFBI was not reactivated in two resistant cell strains (501 mel and WW165), it was induced in the third one, YURIF, to levels similar to those in sensitive cells ( Figure 7B ).

Open in a separate window

Figure 7

TGFBI reactivation by promoter demethylation.

Panel A. TGFBI re-expression in response to Aza (0.2 µM) as assessed by the oligonucleotide array hybridization. The heavy line on the ordinate represent the levels of TGFBI- transcript levels in adult melanocytes. The data represent one sequence ID. All other details as in Figure 2 panel E. Panel B. Validation of TGFBI re-expression at the protein level by Western blots with anti-TGFBI antibodies employing b-actin as a control. The results are representative of two biological replicas. Panel C. BS sequencing results of TGFBI proximal promoter and first exon in normal human melanocytes (NBMel) and melanoma cells untreated (control), and Aza (0.2 µM) treated cells. Panel D. Chromatograms of the distal promoter about -50 to −100 bp downstream of TSS as shown in C. Boxed nucleotide pairs indicate position of intact (CG), partially BS modified (C/TG) and deaminated (TG) CG pairs. All other details as in Figure 4.

Sequencing of BS modified DNA revealed that TGFBI promoter was unmethylated and partially methylated in expressing normal human melanocytes and YUSIT1 melanoma cells, and completely methylated in non-expressing melanoma cells WW165, YUGEN8, YUMAC and YUSAC2 ( Figure 7C and D ). Furthermore, Aza caused demethylation in the three cell strains examined, YUSAC2, YUGEN8 and YUMAC, in which TGFBI was reactivated ( Figure 7C and D ). TGFBI promoter methylation was not restricted to metastatic cells or to cells in culture, because it was also methylated in primary melanoma cells freshly isolated from a 2.2 mm lesion (passage 1) and in five independent snap-frozen metastatic tumors (data not shown). These results suggest that TGFBI is indeed controlled by DNA methylation in melanoma cells and that promoter methylation may serve as a marker for malignant transformation.

We assessed the contribution of the two TGFβ-pathway genes to the Aza apoptotic response the relatively sensitive YUMAC cell strain by short-term knockdown with gene-specific siRNA. Clu and TGFBI siRNA reduced the targeted protein to almost undetectable levels ( Figure 8A ). On the other hand, the apoptotic response of parallel cultures was reduced by 30% and 50% in Clu and TGFBI knockdown, respectively compared to Alexa fluor control transfectants, without any further increase in double knockdown cells ( Figure 8B ). These results indicated that Clu and TGFBI can account for some, but not all the apoptotic effect of Aza and that the two may act on the same pathway.

Open in a separate window

Figure 8

Validation of CLU and TGFBI apoptotic activity.

Panel A. Reduction of Clu and TGBFI proteins by gene specific siRNA knockdown assessed by Western blotting. YUMAC melanoma cells were untreated or treated with Aza (0.2 µM) for 2 days followed by transient transfection with siRNA directed to Clu, TGFBI, or a mixture of the two, employing Alexa Fluor as a control, as indicated. Cells were harvested the following day and extracts subjected to successive Western blotting with the respective antibodies, and anti-β-actin as a control. Panel B. Parallel cultures were tested for apoptosis with the Caspase 3/7 assay. Values are given as percent of control, i.e., non-transfected cultures.

Apoptotic assays

Apoptosis was measured with the Caspase-Glo 3/7 assay kit from Promega following the manufacturer instructions. In addition, we used immunofluorescence with affinity-purified rabbit anti-caspase-3 active antibodies (AF835, R&D Systems) to assess the number of apoptotic cells after Aza treatment compared to controls. DAPI (4′,6-diamido-2-phenylindole dihydrochloride, Sigma Chemicals) was used to visualize nuclear DNA.

Single cell DNA damage assay

The CometAssay Single Cell Gel Electrophoresis Assay kit (CometSlideTM, R&D Systems, Minneapolis, MN) was used to assess DNA damage in response to Aza following the manufacture's instructions. Briefly, melanoma cells were untreated or treated with Aza (0.2 µM, 0.5 µM and 1.0 µM) for 2 days followed by one day recovery as described above. As a positive control we used melanoma cells suspended in PBS and treated with 100 µM of hydrogen peroxide for 10 minutes on ice. Cells were harvested, re-suspended in PBS at 125,000 cells/ml and 50 µl portions were processed, subjected to electrophoreses and stained with SYBR Green I. Single fluorescing cells (100–120 cells from each treatment) were photographed, and analyzed with CASP software (http://casp.sourceforge.net). Quantitative and statistical data were generated by analysis of the results using the commercially available image analysis software packages that calculates tail length and tail moment termed CASP software (http://casp.sourceforge.net).

RNA isolation and hybridization to DNA microarrays

Approximately 20–30 million cells (normal human melanocytes and melanoma cells/each) were used for mRNA extraction. The melanoma cells were treated with low-dose Aza (0.2 µM) for 2 days, followed by one day recovery, total RNA was extracted with the TRIzol reagent (Invitrogen Life Technologies, Inc., Invitrogen Corp., Carlsbad, CA), and Poly(A) mRNA was further isolated using the PolyATtract mRNA isolation system IV (Pro-mega, Madison, WI) following the manufacturer's instructions, and reversed transcribed to double stranded cDNA.

NimbleGen human whole genome expression microarrays (array 2005-04-20_Human_60mer_1in2) were used for hybridization. The same chip was hybridized with Cy3/Cy5 labeled polyA-selected cDNA from untreated and Aza treated melanoma cells. Each hybridization was repeated with dye swapping. The array hybridizations and data captures were performed by personnel at NimbleGen Systems Iceland LLC. Vínlandsleið 2–4, 113 Reykjavik, Iceland (currently Roche Applied Science, Basel, Switzerland).

Bioinformatic analysis of global gene expression

Validation of gene expression

Protein levels were assessed by Western blots as described [55]. The membranes were probed with the following antibodies: anti-Clusterin (C-18, goat, sc-6419), anti-DNMT1 (K-18, goat, sc-10221), anti-p53 (pS20 sc-18079R rabbit), anti-p53 (pSer37, sc-28464-R), all from Santa Cruz Biotechnology CA; anti-MITF (clone D5) and anti-BCL2 (clone 124) mouse monoclonal antibodies from DAKO; anti-p53 (AF1355, goat) from R&D Systems; anti-p21^Cip1 mAb (C24-4420) from BD Transduction Laboratories, Canada; anti-β-Actin mAb (A1978) from Sigma-Aldrich, St. Louis, MO 63103; anti-β-catenin (rabbit polyclonal from Dr. David Rimm, Pathology department, Yale University) [56]; and anti-TGFBI (rabbit polyclonal from Dr. Jan Johannes Enghild, diluted 110,000).

Quantitative real-time RT-PCR was carried out in triplicate employing cDNA, using ABI 7500 Fast Real-Time PCR Systems and Power SYBRGreen (Applied Biosystems, Foster City, CA). The genes and primers used for probing are listed in Table S2 . The expression of ACTB was used as a reference to normalize for input cDNA. The relative expression values were computed by the comparative Ct method.

Downregulation of gene by siRNA

CTNNB1 (β-catenin) was knock-downed with three different gene specific siRNA purchased from Qiagen, Valencia, CA as follows: CTCGGGATGTTCACAACCGAA (Hs_CTNNB1_5); CAGCGGCTTCTGCGCGACTTA (Hs_CTNNB1_8); CAGGATGATCCTAGCTATCGT (Hs_CTNNB1_9). Alexa Fluor 488 siRNA was used to monitor transfection efficiency as well as a control. An additional control was Allstars negative control siRNA from Qiagen (Cat number: 1027280). Melanoma cells were treated with Aza (0.2 µM) for 2 days and siRNAs were added at 10 nM employing the HiPerFect transfection reagent kit following the manufacturer instructions (Qiagen). The cells were harvested the following day and were assessed in parallel for protein expressions and apoptosis (triplicate wells). The extent of target gene knockdown (β-catenin), as well as downstream targets, MITF, Myc (β-catenin target genes), BCL2 (MITF target gene) [45], were assessed at protein levels by successive Western blotting with antibodies to β-catenin as a control.

Five different siRNA purchased from Qiagen were tested for Clusterin and TGFBI knockdown as revealed by Western blot analysis of the respected protein and one from each group was chosen for further experiments (CLU: ACAGACCTGCATGAAGTTCTA, and TGFBI: CGGGAAGGCGATCATCTCCAA) as described for CTNNB1 knockdown.

We thank the Cell Culture Core facility of the Yale Skin Disease Research Core Center (YSDRCC) supported by NIAMS grant 5 P30 AR 041942-12 (Dr. Robert Tigelaar, PI) for providing normal human melanocytes and melanoma cells, Mr. Paul Hively for biostatistical analyses, Ms. Pavithra Shivakumar, Mrs. Kathleen Hoyt and Dr. Yasuo Koga for technical assistance.