POU2F3 (OCT11, Skn-1a) is a keratinocyte-specific POU transcription factor whose expression is tied to squamous epithelial stratification. It is also a candidate tumor suppressor gene in cervical cancer (CC) because it lies in a critical loss of heterozygosity region on11q23.3 in that cancer, and its expression is lost in more than 50% of CC tumors and cell lines. We now report that the loss of POU2F3 expression is tied to the hypermethylation of CpG islands in the POU2F3 promoter. Bisulfite sequencing analysis revealed that methylation of specific CpG sites (−287 to −70 bp) correlated with POU2F3 expression, which could be reactivated with a demethylating agent. Combined bisulfite restriction analysis revealed aberrant methylation of the POU2F3 promoter in 18 of 46 (39%) cervical tumors but never in normal epithelium. POU2F3 expression was downregulated and inversely correlated with promoter hypermethylation in 10 out of 11 CC cell lines. Immunohistochemical analysis on a cervical tissue microarray detected POU2F3 protein in the epithelium above the basal layer. As the disease progressed, expression also decreased, especially in invasive squamous cell cancer (70% loss). Thus, aberrant DNA methylation of the CpG island in POU2F3 promoter appears to play a key role in silencing this gene expression in human CC. The results suggested that POU2F3 might be one of the CC-related tumor suppressor genes, which are disrupted by both epigenetic and genetic mechanisms.
POU transcription factors help in regulating viral transcription, keratinocyte differentiation and other cellular events (Andersen and Rosenfeld, 2001). POU2F3, also known as Skn-1a and OCT11, is a keratinocyte-specific POU transcription factor expressed in stratified squamous epithelia, including the epidermis, cervix and foreskin (Andersen et al., 1993; Goldsborough et al., 1993; Yukawa et al., 1993). Its expression pattern parallels those of the E6 and E7 oncoproteins of human papillomavirus (HPV), showing primary expression in differentiating suprabasal cells and very low expression in proliferating basal keratinocytes (Yukawa et al., 1996). Skn-1a activates genes encoding cytokeratin 10 and SPRR2A, two major protein markers for differentiating keratinocytes (Andersen et al., 1993; Fischer et al., 1996). hSkn-1a contributes to epidermal stratification by promoting keratinocyte proliferation and enhancing subsequent keratinocyte differentiation (Hildesheim et al., 2001). In addition, hSkn-1a can stimulate HPV transcription by activating E6/E7 promoters (Yukawa et al., 1996; Andersen et al., 1997; Kukimoto and Kanda, 2001).
POU2F3 lies in chromosome 11q23.3, a region that often displays loss of heterozygosity in invasive cervical cancer (CC) (Zhang et al., 2005). We previously screened the entire coding region of POU2F3 in CCs with 11q23.3 deletions and found no mutations (Zhang et al., 2005). However, we found that POU2F3 was expressed in normal cervix but absent in 50% of CCs and CC cell lines.
To determine if the downregulation of POU2F3 results from epigenetic silencing, we evaluated the methylation status of the CpG island in the gene's promoter and first exon. We also treated CC cell lines with 5-aza-2′-deoxycytidine (5-aza-dC) to determine whether the demethylating agent could reactivate POU2F3 expression. To evaluate the clinical relevance of POU2F3 at the various stages of CC, we developed and used a tissue microarray to examine the expression of POU2F3 protein by immunohistochemical analysis in normal cervical tissues, precancerous cervical intraepithelial neoplasias (CIN) and invasive CCs.
POU2F3 mRNA expression in cervical cancer cell lines and clinical samples
We initially screened a panel of CC cell lines and a series of clinical samples for POU2F3 mRNA expression by reverse transcription–polymerase chain reaction (RT–PCR). Expression of POU2F3 was observed in all normal epithelia, whereas the expression was only detected in four of 13 (30.8%) CIN (three of six CIN 1, none of three CIN 2 and one of four CIN 3), five of 17 (29.4%) CCs and five of 11 (45.5%) CC cell lines. Figure 1 shows representative expression results of RT–PCR for POU2F3 (Figure 1).
Methylation status of the POU2F3 CpG islands in cervical cancer cell lines
In silico promoter and CpG island prediction identified a putative promoter, which locates −1000 to +50 bp relative to the transcription start site. A typical CpG island with a CpG observed/expected ratio of >0.60 and a GC content of >50% was also identified. This island extends from −613 nucleotides to +319 nucleotides relative to the transcription start site, and it contained 67 CpG sites (Figure 2).
Using bisulfite sequencing, we looked for methylation at all 67 CpG sites in the POU2F3 promoter, comparing two cell lines that do not express POU2F3 (Ca Ski and SW756) with two that do (SiHa and ME-180). The results are shown in Figure 2. DNAs from SiHa and ME180 displayed little or no methylation of the CpG sites. In contrast, DNAs from Ca Ski and SW756 were heavily methylated across the CpG sites. Methylation of the 20th to 49th CpG sites (−287 to −70 bp) correlated with POU2F3 expression in these cell lines.
We previously provided evidence for allele loss in CC cell lines at 11q23.3, as determined by homozygosity mapping-of-deletions (Zhang et al., 2005). Therefore, we decided to correlate these results with POU2F3 gene expression and the methylation status of the CpG islands in the promoter region (Table 1). The data show a good correlation (P<0.05) of POU2F3 expression by RT–PCR with promoter methylation by combined bisulfite restriction analysis (COBRA) and loss of heterozygosity (LOH) (Zhang et al., 2005) in 10 out of 11 cell lines examined. The lack of expression of POU2F3 with promoter methylation and regional LOH was seen in cell line HeLa, C-4 I, Ca Ski and SW756. The expression of POU2F3 was seen in cell line DoTc24510, CCI, ME-180 and SiHa, which show neither promoter methylation nor regional LOH.
Methylation status of the POU2F3 CpG island in cervical tissues
As methylation status of the 20th to 49th CpG sites (−287 to −70 bp) correlated with POU2F3 expression in four CC cell lines. We investigated the methylation status of this region in cervical samples and other CC cell lines using COBRA. The resulting PCR fragment was digested with BstUI or Hpy99I, which cut the fragment only if the CGCG target sites (BstUI) or CGWCG target sites (Hpy99I) had been retained through bisulfite-mediated deamination. Enzymatic cleavage thus indicates methylation, whereas lack of digestion indicates the absence of methylation. As expected, COBRA confirmed the methylation status of the POU2F3 promoter in the CC cell lines. The PCR amplicons generated from SW756 and Ca Ski cells were completely digested (Figure 3), which is consistent with dense methylation from bisulfite sequencing. The SiHa and ME-180 amplicons were totally undigested and were therefore considered to be predominately unmethylated. Partial digestions of the C-33A and C-4 I amplicons were observed, indicating partial methylation in these cells.
The aberrant promoter hypermethylation was detected in 39% (18 of 46) of the cancers tested by COBRA (Figure 3). The methylation status of POU2F3 did not correlate statistically with clinicopathological features such as tumor histology, stage or HPV type (P>0.05). The POU2F3 promoter was not methylated in normal epithelium, suggesting that methylation of the POU2F3 promoter was unique to CC cells (Table 2).
Effect of demethylating agents and histone deacetylase inhibitor on expression of POU2F3
To determine whether methylation is responsible for the decreased level of POU2F3 mRNA in CC cell lines and to investigate the role of histone deacetylation in transcriptional silencing, we treated the cell lines SW756 and Ca Ski with the demethylating agent 5-aza-dC alone or in combination with trichostatin A (TSA), a potent inhibitor of histone deacetylase. Both real-time PCR and RT–PCR analysis revealed that 5-aza-dC could restore the expression of POU2F3 transcript at all concentrations used, even as low as 1 μM. The addition of TSA did not increase levels of mRNA further (Figure 4). To examine the mechanism of this gene reactivation, we sequenced the 20th–49th CpG sites of the POU2F3 promoter in the cell lines SW756 and Ca Ski before and after treatment. Bisulfite DNA sequencing confirmed that 5-aza-dC alone could partially demethylate most of the CpG sites in SW756. The effect of demethylation of 5-aza-dC on the CpG sites in Ca Ski is less frequent than that in SW756. However, in four of these sites (CpG 21st, 22nd, 32nd and 44th) demethylation was significant at all concentrations used. Furthermore, as we increased the concentration of 5-aza-dC, the demethylation of the CpG sites did not increase. And combination use of TSA did not significantly enhance the demethylation effect of 5-aza-dC in both cell line (Figure 5).
Immunohistochemical pattern of POU2F3 expression in cervical neoplasia
We prepared a tissue microarray to examine the expression of POU2F3 protein in cervical carcinogenesis. The samples on the array came from 76 different women, and represented normal cervix (n=15), koilocytotic atypia (n=7), CIN 1 (n=10), CIN 2 (n=10), CIN 3 (n=15), squamous carcinomas (n=9) and adenocarcinomas (n=10). Each sample was evaluated for staining intensity, location of the cells that stained and percentage of cells stained. Two to three replicate sections were available for 59% of the cases. Three cases were not evaluated owing to missing epithelium. The intra-specimen correlation per case was excellent. No two specimens within a case differed by more than one intensity grade, and only eight of the cases differed in staining by more than 30% (30% in three cases, and 40–50% in five cases).
In normal epithelium, POU2F3 staining was intense throughout the cytoplasm of cells above the basal cells. In one case, 70% of the cells stained at an intensity of 1. In the remaining cases, 70–100% of the cells stained at an intensity of 2–3. No membrane staining was identified. The stroma and basal epithelium were negative in all cases. In the samples of koilocytotic atypia and CIN 1, 65% of the epithelium stained at an intensity of 2–3. There was also glandular epithelium on these sections, although fewer than 20% of its cells showed cytoplasmic staining, which varied in intensity. However, in four of the cases, 100% of the cervical reserve cells, located beneath the glandular epithelium at the transformation zone, stained at an intensity of 2. In the CIN 2 and 3 samples, respectively, 76 and 60% of the epithelium stained at an intensity of 1–2, except for two cases, which had an intensity of 3. Four cases of invasive squamous cancer did not stain, three cases had less than 30% staining at an intensity of 1 and two cases had staining of 80–90% at an intensity of 1–2. The quantity and intensity of staining varied widely in the adenocarcinoma samples. Loss of POU2F3 protein was significant in squamous carcinoma compared with normal cervix (P<0.01). The percentage of cell staining of POU2F3 was significantly decreased in CIN and CC than in normal cervix (P<0.01), and the staining intensity significantly decreased as the disease grade increased (P<0.01). The results are summarized in Table 3. The representative staining sections were shown in Figure 6.
We have shown that expression of POU2F3 mRNA and protein is significantly reduced or absent in both CIN and CC compared to normal cervical epithelium. Loss of expression correlates with the promoter methylation as assessed by bisulfite sequencing and COBRA techniques. Treating CC cell lines with micromolar concentrations of 5-aza-dC for 5 days demethylated many of the CpG sites in the promoter and led to partially transcriptional reactivation of POU2F3. Adding TSA to 5-aza-dC treatment did not enhance the degree of demethylation.
Although 5-aza-dC treatment could induce expression of POU2F3 mRNA, many CpG sites remained partially methylated, and the combination of TAS subsequent to 5-aza-dC treatment did not extend the degree of demethylation seen with 5-aza-dC alone. The possible explanation is that the CpG sites that were demethylated following the treatment of 5-aza-dC are critical in maintaining the transcriptional activity of POU2F3. Also, the reactivation of the gene could emanate from only a small number of alleles, not detected by bisulfite sequencing analysis, which become extensively demethylated following 5-aza-dC treatment (Cameron et al., 1999). Another possibility is that 5-aza-dC might activate upstream genes that are responsible for the activation of POU2F3, and the methylation status of the POU2F3 CpG island is not directly linked to the expression of this gene. The treatment of 5-aza-dC causes a variety of changes in cells, including the decondensation of chromatin and global genomic hypomethylation (Christman, 2002; Suzuki et al., 2002). Alternatively, the pattern of the DNA demethylation after 5-aza-dC treatment suggests that DNA methylation of only specific CpG is linked with expression. The reproducible demethylated CpG sites after the treatment in Ca Ski might be important in the transcriptional regulation of POU2F3 in this cell line. The distance between the CpG sites (21st, 22nd, 32nd and 44th) is 133–151 bp, which is around one turn of double-strand DNA (146 bp) wrapped around a nucleosome. This raises a possibility that methylated CpGs cause the lack of capability for nucleosomal histones to bind to the specific DNA sequence. CpG methylation indeed affects histone–DNA interaction by reducing DNA backbone flexibility and dynamics (Pennings et al., 2005). The most positive evidence of CpG methylation repositioning a nucleosome was found on the promoter sequence of the chicken adult β-globin gene. A strong histone octamer positioning site was largely abolished by CpG methylation. The methylation-sensitive nucleosome position was proposed to act as a switch between methylated and unmethylated overlapping CpG positions, with implications for access to promoter elements (Davey et al., 2004).
We did not observe the dose-dependent demethylation of CpG sites of 5-aza-dC treatment on the treated cell lines. This same observation was also reported by a resent microassay study, which revealed that 5-aza-dC and TSA effects on gene expression patterns were independent of dose and duration of exposure (Gius et al., 2004).
Interestingly, we previously identified a somatic mutation in this region of the promoter in one of 12 CCs. The C > A mutation occurs −221 base pairs (bp) upstream of the transcription initiation site in a predicted Sp1 binding site. (Zhang et al., 2005). Whereas transcription factors that include CpG in their binding site, such as the CCCTC binding factor (CTCF) and the CpG-binging protein, are sensitive to the presence of the 5-methyl group in the DNA major groove and are prevented from binding to these sites when methylated. This is the case even for the ubiquitous transcription factor SP1, previously thought to be insensitive to the DNA methylation state of a promoter (Pennings et al., 2005). This warrants further experiment to investigate the relationship between SP1 binding site mutation in the promoter region and POU2F3 expression in CC.
Cell cycle disruption and cellular transformation are known to occur when the HPV viral oncoproteins E6 and E7 target tumor proteins p53 and pRB, respectively (Howley, 1991). However, chromosome complementation experiments suggest that alteration of additional tumor suppressor proteins likely contributes to cervical carcinogenesis. Microcell transfer of a partial copy of chromosome 11 suppresses tumorigenicity in CC cell lines (Saxon et al., 1986; Koi et al., 1989). POU2F3 lies within a significant region of LOH at 11q23.3 in CC, and is therefore a candidate tumor suppressor gene that could be inactivated by both chromosomal loss (Rader et al., 1996; Huettner et al., 1998; O’Sullivan et al., 2001; Zhang et al., 2005) and aberrant methylation of its promoter. Our data in CC cell lines show a good correlation of lack of POU2F3 expression with promoter methylation and LOH. It supported the hypothesis that abnormal promoter hypermethylation can have the same effect as a coding-region mutation in one copy of the gene (the first hit); often, loss of the other copy serves as the second hit (Herman and Baylin, 2003). Further evidence that POU2F3 functions as a CC tumor suppressor gene was provided by Enomoto et al. (2004), who showed that growth is repressed and differentiation is induced in CC cell lines expressing exogenous hSkn-1a.
Epithelial cancers arise when the normal program of differentiation is disrupted and the tightly regulated gene expression program is modified. The innermost basal layer of the cervix is made of undifferentiated keratinocytes that have a high proliferation potential. The cells committed to terminal differentiation migrate upwards where they undergo distinct morphological and structural changes. In the cervix, most cancers begin at the transformation zone, where the glandular epithelium undergoes squamous metaplasia by ingrowth from adjacent squamous epithelium or by squamous differentiation of reserve cells. The product of the POU2F3 gene promotes keratinocyte proliferation, leading to stratification. It subsequently influences differentiation by upregulating the expression of terminal differentiation genes (Hildesheim et al., 2001). The immunohistochemical staining pattern seen in our human tissue microarray was consistent with the gene's purported role in vitro. POU2F3 was highly expressed in the cytoplasm of the epithelium above the basal cells and reserve cells. We also observed loss of POU2F3 expression as the grade of CIN increased confirming the finding of Hietala et al. (1997), who found 20% less cytoplasmic staining in CIN 3 samples than in normal epithelium. However, we observed the most marked decrease in expression in invasive squamous CC, where cytoplasmic expression in the epithelium was 70% less than in normal epithelium.
Two characteristic features of HPV infection are tissue specificity and a requirement for advanced differentiation of epithelial tissues. Thus, the cell layers above the basal layer support increased levels of viral DNA replication and synthesis of capsid proteins. Skn-1a has been linked to the regulation of HPV transcription via the activation of E6/E7 promoters (Yukawa et al., 1996; Andersen et al., 1997; Kukimoto and Kanda, 2001). However, cellular transformation does not occur as a consequence of this productive infection because differentiated cells that express the oncoproteins have lost the ability to proliferate and are sloughed off. In contrast, viral and host changes within basal cells lead to viral integration and a malignant phenotype. It is unclear at this time whether methylation of the POU2F3 promoter relates to the gene's role in viral regulation. POU transcription factors respond to a wide variety of extracellular and intracellular signals, and they exercise important regulatory control over the maintenance and differentiation of cells via specific interactions with DNA and other proteins (Andersen and Rosenfeld, 2001). Therefore, POU2F3 may have two distinct roles in the cervix: differentiation-dependent regulation of HPV transcription and the development of invasive cancer by epigenetic mechanisms. Alternatively, regulation of viral expression by POU2F3 may affect transcriptional co-activators that indirectly affect methylation of POU2F3, as it does with E7 binding when the acetylation activity of pCAF is inhibited (Avvakumov et al., 2003). Blocking pCAF reduces its acetyltransferase activity and alters cellular gene expression such as the differentiation-specific function of pRb (Nguyen et al., 2004).
In summary, our findings demonstrate that CpG sites in the POU2F3 promoter are often aberrantly methylated in CCs and that such hypermethylation of the promoter associates with transcriptional silencing of the POU2F3 gene. Thus, epigenetic mechanisms may mediate POU2F3 downregulation in CC cells. The tissue specificity of POU2F3 and its loss of function in more than 40% of squamous CCs make it a target for transcriptional therapy aimed at restoring differentiation as a component of cancer treatment.
Materials and methods
Identification of the POU2F3 promoter, CpG island and transcriptional binding sites
We used the complete cDNA sequence of POU2F3 (GenBank Accession No. AF133895) to search the human genome database (May 2004 Assembly) at http://genome.UCSC.edu. The genomic sequence was located from 119 616 256 to 119 695 863 bp (a span of 79 608 bp) on chromosome 11. An interval of 3 kb upstream and downstream from the transcription initiation site was used for in silico promoter region and CpG island identification. The putative promoter region was predicted by PromoSer software (http://biosulf.buledu/cgi-bin/zlab/getprom.pl). The CpG plot program (http://www.ebi.ac.uk/emboss/cpgplot) was used to identify the CpG island, using the following parameters: length >200 bp, C+G% >50.00 and observed/expected ratio >0.60. Transcriptional binding sites were identified by TESS: Transcription Element Search Software at http://www.cbil.upenn.edu/tess.
Cervical tumors and cervical cancer cell lines
A total of 46 snap-frozen, 22-optimum cutting temperature OCT embedded CC biopsies (five adenocarcinomas, nine adenosquamous carcinomas and 32 squamous cell carcinomas), 13 normal cervical epithelia, 13 CINs and 11 CC cell lines were analysed for POU2F3 expression or POU2F3 methylation. Normal cervix was defined as stratified squamous epithelium histopathologically absent of cellular atypia. In this study, the normal cervix was from ectocervix. DNAs/RNAs for eight of the cancers, all normal epithelia and all the CIN specimens were obtained by laser capture microdissection (LCM, PixCell® IIe LCM System, Arcturus Engineering, Mountain View, CA, USA). All other tumor DNAs used for methylation analysis only were sectioned from the OCT-embedded tumor biopsies. Cervical cancer cell lines were developed either by JR (CCI) (Rader et al., 1990) or purchased from American Type Culture Collection (HeLa, SiHa, Ca Ski, C-4 I, HT-3, ME-180, C-33A, SW756, Hs588.T and DoTc2 4510) (American Type Culture Collection (ATCC), Manassas, VA, USA). Cells were grown in the media as recommended by the ATCC.
Analysis of POU2F3 expression
RNAs were extracted from epithelia using the PicoPure™ RNA Isolation Kit (Arcturus Biosciences Inc., Mountain View, CA, USA). RNA from CC cell lines was extracted using TRIZOL® Reagent (Invitrogen, Carlsbad, CA, USA). Reverse transcription–PCR was performed with POU2F3-specific primers (forward primer, 5′-gagccaggaaatgatcgaaa-3′; reverse primer, 5′-ctggcatttagcccagacat-3′), which were designed to cross an intron to avoid false-positive results. Amplification of glyceraldehyde-3-phosphate dehydrogenase (GAPDH) from the same cDNA samples provided an internal control of RNA loading.
Sodium bisulfite treatment and bisulfite sequencing
Genomic DNA was obtained by digesting cervical tissues with proteinase K. It was extracted with phenol/chloroform and treated with sodium bisulfite to convert unmethylated cytosine to uracil.
One microgram genomic DNA was treated with sodium bisulfite at 50°C for 18–20 h, using the CpGENOME™ universal DNA modification kit (Chemicon International Inc., Temecula, CA, USA) (Li et al., 2005). Bisulfite-treated DNA was amplified by nested PCR, using overlapping primer sets to amplify both methylated and unmethylated DNA in the 933 bp CpG island region that extends from the 5′-end of the POU2F3 CpG island into the first exon (Table 4). The resulting PCR products include all 67 CpG dinucleotides in the island. PCR was performed in a 10 μl volume containing 1 × PCR buffer, 1 × GC-rich buffer, 2.0 mM of MgCl2, 200 nM of dNTPs (Amersham Biosciences Corp., Piscataway, NJ, USA), 0.2 mM of each primer and 0.4 U of FastStartTaq (Roche Diagnostics Co., Indianapolis, IN, USA). The PCR conditions were: 95°C for 3 min, 94°C for 30 s, 60°C for 30 s and 72°C for 1 min with annealing temperature decreasing 0.5°C every cycle for 14 cycles, and 94°C for 30 s, 53°C for 30 s, 72°C for 1 min for 26 cycles for both reactions. Amplified products were purified using AMPure (Agencourt Bioscience Corp., Beverly, MA, USA), and were sequenced directly. DNA sequencing reactions were performed using BigDye® Terminator v3.1 Sequencing Reagents according to the manufacturer's instructions and an AB 3100 sequencer (Applied Biosystems, Foster, CA, USA) was employed to examine the DNA fragment.
Combined bisulfite restriction analysis
The promoter region encompassing 30 CpG islands (−287 to −70 bp) was amplified and purified as described above. Purified PCR products were digested with methylation-sensitive restriction enzymes (BstUI and Hpy99I; New England BioLabs Inc., Beverly, MA, USA). Digested PCR products were identified by ethidium bromide staining on 3% agarose gels. The restriction enzyme BstUI cleaves (CGCG) sequences, although failing to cut (TGTG) sites that result from bisulfite conversion of unmethylated CpG sites. Hpy99I cleaves (CGWCG) sequences, although failing to cut (TGWTG) sites (W=A or T). Variation in methylation is indicated by the creation of new or the retention of pre-existing restriction sites.
Treatment of cell lines with demethylating agents and histone deacetylase inhibitor
Ca Ski and SW756 cell lines in which the POU2F3 promoter had been methylated were seeded at a density of 1 × 105 cells/100 mm dish and treated with 1, 3, 5 or 10 μM 5-aza-dC (dissolved in dimethyl sulfoxide (DMSO)) (Sigma, St Louis, MO, USA) for 5 days. For the combination study, the cells were treated for 4 days with 5 or 10 μM 5-aza-dC, and 100 nM TSA (dissolved in ethanol) (Wako Chemicals USA Inc., Richmond, VA, USA) was added for the last 24 h. The media and drug additives were replaced daily. DNA and RNA were isolated after 5 days of treatment. The control cultures were treated with equal amount of vehicle DMSO and/or ethanol.
Constructing the tissue microarray
Hematoxylin- and eosin (H&E)-stained slides from representative paraffin-embedded blocks of CC, CIN and normal epithelium were reviewed and the location of the disease process was marked by two investigators (PCH and MF). The Siteman Cancer Center Tissue Procurement Facility used a Beecher tissue puncher/array system (Beecher Instruments, Silver Spring, MD, USA) to assemble the array, as described previously (Kononen et al., 1998). It contained triplicate sections from 15 normal cervixes, seven koilocytotic atypia, 10 CIN 1, 10 CIN 2, 15 CIN 3, nine squamous cancers and 10 adenocarcinomas.
Immunohistochemical localization and grading of POU2F3
Sections (5 mm thick) were cut from the tissue microarray. Initial sections were stained for H&E to verify histology. Owing to the small size of the lesions used in array, not all samples were adequate for analysis; 73 samples proved adequate with 43 of these available in duplicate or triplicate. Avidin–biotin staining was used as described previously (Hietala et al., 1997).
Briefly, POU2F3 antigen was retrieved with Antigen Retrieval Citra Solution (BioGenex, San Ramon, CA, USA). Affinity-purified rabbit polyclonal antibody raised against a peptide mapping to the carboxyl-terminus of rat Skn-1a (Skn-1a/i (C-20): sc-330; Santa Cruz Biotechnology Inc., Santa Cruz, CA, USA) was diluted to 1:500. The antibody was incubated with the array section overnight in a humidified chamber at 4°C. After they had reacted with the primary antibody, the sections were treated with a biotinylated anti-rabbit IgG (H+L) (Vector laboratories, Burlingame, CA, USA) for 30 min. The sections were then developed with three, 3′-diaminobenzidine (DAKO, Carpinteria, CA, USA), counterstained with hematoxylin, dehydrated in graded alcohols and placed under coverslips. Three investigators (PCH, ZZ, MB) independently assessed the patterns and percentages of stained cells in each section of the tissue microarray. Staining intensity was graded as: 3=strongly staining, 2=moderately staining and 1=weakly staining.
Statistical analyses of tumor characteristics and promoter methylation/POU2F3 protein expression were performed using the χ2 test. Fisher's exact test was used when the expected frequencies were too low for the χ2 test to be used reliably.
This research was supported by National Institutes of Health (NIH) Grant CA94141-04. We thank Dr FT Kraus for his generous advice on immunohistochemistry. Margo C Funk was supported by the Howard Hughes Medical Institute. We also thank the Alvin J Siteman Cancer Center at Washington University School of Medicine and Barnes-Jewish Hospital in St Louis, MO, for the use of the Tissue Procurement Core that provided the tissue array. The Siteman Cancer Center is supported in part by an NCI Cancer Center Support Grant No. P30 CA91842.