Aberrant DNA Methylation in Keratoacanthoma

Background Keratoacanthoma (KA) is a self-limiting epidermal tumor for which histopathological examination sometimes suggests malignancy. Based on inconsistent clinical views, KA can be regarded as both a benign tumor and a variant of squamous cell carcinoma (SCC). Aberrant DNA methylation frequently occurs in malignant tumors but it scarcely occurs in benign tumors. Whether aberrant methylation occurs in KA has not been previously examined. Objective The aim is to elucidate whether aberrant methylation of CpG islands (CGI) containing a high density of cytosine-guanine dinucleotide (CpG) sites occurs in KA. Methods Five SCC cell lines, two cultured samples of normal human epidermal keratinocytes (NHEKs), 18 clinical SCC samples, and 21 clinical KA samples were analyzed with Infinium HumanMethylation450 BeadChips, quantitative real-time methylation-specific PCR (RT-MSP) and/or bisulfite sequencing. Results Genome-wide analyses of NHEK, KA, and SCC indicated that there was a greater number of aberrantly hypermethylated CGIs in SCC than in KA and there were aberrantly hypermethylated CGIs which are common in both. Among the common hypermethylated CGIs, RT-MSP and bisulfite sequencing targeting CGIs located on CCDC17, PVR, and MAP3K11 gene bodies also showed that methylation levels were significantly higher in KA than in normal epidermis. Statistical analyses suggested that the methylation level of CGI located on PVR in SCC might be correlated to lymph node metastasis (P = 0.013, Mann-Whitney U test) and that the methylation level of CGI in MAP3K11 in KA might be correlated to age (P = 0.031, linear regression analysis). Conclusion Aberrant DNA methylation occurs in KA.

Introduction obtained from reachable donors or their legal guardians. The ethics committee of The Jikei University School of Medicine waived the requirement for consent from unreachable donors.

Clinical samples and nucleic acid extraction
SCC cell lines HSC-1 and HSC-5 were provided from the Japanese Collection of Research Bioresources (Tokyo, Japan). SCC cell lines A431 and DJM-1 were provided from the Riken BioResources Center (Tsukuba, Japan). SCC cell line A388 was purchased from the American Type Culture Collection (Manassas, VA). Two normal human epidermal keratinocytes (NHEKs) derived from an adult and a neonate were obtained from ScienCell Research Laboratories (Carlsbad, CA). Eighteen paraffin-embedded SCC samples and 21 paraffin-embedded KA samples were obtained from each patient (Tables 1 and 2). Seven normal skin samples were obtained by shaving the margin of the excised epidermal cysts ( Table 3). The TNM classification of SCC was evaluated according to the Union for International Cancer Control TNM Classification of Malignant Tumours (7 th edition). The diagnoses of SCC, KA, and normal skin were made histopathologically by at least two experienced board-certified pathologists. To extract DNA from paraffin-embedded samples, the samples were sliced into 4-10 μm-thick sections, deparaffinized, and then dissected with a fine needle. Genomic DNA was extracted by using the QIAamp DNA mini kit (Qiagen, Valencia, CA). Total RNA was isolated by using ISOGEN (Nippon Gene, Tokyo, Japan).

Genome-wide methylation analysis
Infinium HumanMethylation450 BeadChips (Illumina, San Diego, CA) consisting of 485,577 probes with single-nucleotide resolution were used for genome-wide methylation analysis performed by the contract research service of TAKARA (Otsu, Japan). Output data from the GenomeStudio/Methylation module were analyzed with statistical processing software, R (version 2.15.1). The methylation rate was scored on a range of 0.0 (completely unmethylated) to 1.0 (completely methylated). After filtering through the software's algorithms, analyzed regions were classified according to the relative proximity to a gene as follows: TSS1500, sites −1500 to −200 nucleotides from the respective transcriptional start site (TSS); TSS200, sites −200 to 0 nucleotides from the TSS; 5'UTR, sites from the TSS to the translational start site; 1 st exon, sites from the TSS to the end of the 1 st exon; gene body, sites from the start of the first intron to the translation end site; and 3'UTR, sites from the translation end site to the end of the gene.

Quantitative real-time methylation-specific PCR (RT-MSP) and bisulfite sequencing
One μg of the BamHI-digested genomic DNA was modified by sodium bisulfite by using the EZ DNA Methylation-Gold Kit (Zymo Research, Irvine, CA) according to the instruction manual and was then dissolved in 40 μl of buffer. For RT-MSP, 1.0 μl of the sodium bisulfite-treated DNA was amplified with the 7500 Real-Time PCR System (Applied Biosystems, Foster City, CA) by using a mixture of primer sets that were specific to the methylated or unmethylated DNA sequence (M or U set, respectively) and the SYBR Green PCR Master Mix I (Toyobo, Osaka, Japan). The primer sequences are shown in S1 Table. The number of molecules of a specific sequence in a sample was measured by comparing its amplification with that of standard samples containing 10 1 to 10 8 copies of the molecule. The methylation level was defined as the number of methylated DNA molecules divided by the total number of methylated and unmethylated DNA molecules. DNA methylated with SssI methylase (New England Biolabs, Beverly, MA) and DNA amplified with the GenomiPhi DNA amplification kit (GE Healthcare Bioscience, Little Chalfont, UK) were used as methylated and unmethylated DNA controls, respectively, under specific amplification conditions.
For bisulfite sequencing, 1.0 μl of the sodium bisulfite-treated DNA was used for PCR with primers common to methylated and unmethylated DNA sequences. The primer set sequences are shown in S1 Table. The PCR products were cloned into a cloning vector, and 12 clones were cycle-sequenced for each sample.

Statistical analysis
Statistical analysis was performed by using SPSS version 18 (SPSS Japan, Tokyo, Japan). Significant differences in data were analyzed by linear regression analysis, the Kruskal-Wallis test followed by the non-parametric multiple comparison test, the Mann-Whitney U test, and the logrank test. P values <0.05 were considered statistically significant.

Genome-wide analysis detects hypermethylated CGIs in both KA and SCC
To screen CGIs with aberrant DNA methylation in KA, two KA samples (#19, #20), two NHEKs (from an adult and a neonate), and two SCC cell lines (A431 and HSC-5) were examined with the genome-wide methylation analysis platform of HumanMethylation450 BeadChip with 485,577 probes. After filtering through the software algorithms, data from 480,844 probes were suitable for analyses.

Bisulfite sequencing data are compatible with RT-MSP data
To confirm the data obtained by RT-MSP, bisulfite sequencing was performed for SCC, KA, and normal epidermis samples with representative data for RT-MSP (Fig 3). The examination for CGI located on the CCDC17 gene body revealed that SCC sample #2 (66.3% methylation level by RT-MSP) and KA sample #38 (86.1%) had densely methylated CpG sites, while normal epidermis sample #45 (19.8%) had sparsely methylated CpG sites. The examination for CGI located on the PVR gene body revealed that SCC sample #8 (100% methylation level by RT-MSP) and KA sample #34 (84.9%) had densely methylated CpG sites, while normal epidermis sample #45 (29.8%) had less densely methylated CpG sites. The examination for CGI located on the MAP3K11 gene body revealed that SCC sample #8 (80.2% methylation level by RT-MSP) and KA sample #38 (88.3%) had densely methylated CpG sites, while normal epidermis sample #41 (11.5%) had sparsely methylated CpG sites. Thus, the bisulfite sequencing data were compatible with the results of RT-MSP.

Methylation level is possibly related to lymph node metastasis and age
To investigate the correlation between methylation levels of CGIs located on the CCDC17 gene body, PVR gene body, and MAP3K11 gene body and clinical parameters of SCC or KA patients, including age, sex, site of occurrence, T-classification, and N-classification, statistical analyses were performed (Tables 4 and 5). The methylation level of the CGI located on the PVR gene body in SCC was correlated to regional lymph node metastasis (P = 0.013, Mann-Whitney U test) (Fig 4a). The methylation level of the CGI located on the MAP3K11 gene body in KA was correlated to age (P = 0.031, linear regression analysis; R 2 = 0.245) (Fig 4b). These methylation levels were not associated with other clinical parameters.

Discussion
The present study demonstrated for the first time that aberrant hypermethylation occurs in KA and that aberrant hypermethylation less frequently occurs in KA than in SCC based on the data from genome-wide analysis. Methylation analyses using a genome-wide analysis platform, RT-MSP, and bisulfite sequencing clearly showed aberrant DNA methylation of some CGIs in KA, as is the case with SCC.
Results of a comparison of the genome-wide analysis between in vivo KA samples (probably including tumor-surrounding tissues such as fibroblasts, inflammatory cells, and vascular  In genome-wide analysis, different cut-off values were also set between SCC samples and KA samples when CpG sites with aberrant methylation were detected in both KA samples and SCC samples. The value to detect hypermethylated CpG sites was set at a methylation level of 0.5 or 0.8 in SCC samples, while it was set at a methylation level of 0.5 in KA samples. The value to detect hypomethylated CpG sites was set at a methylation level of 0.5 or 0.2 in SCC samples, while it was set at a methylation level of 0.5 in KA samples. The in vivo clinical KA samples probably included various types of cells, while the in vitro SCC cell lines and the in vitro NHEKs contained only SCC cells and only normal keratinocytes, respectively. The contamination of non-tumor tissues in KA samples could make the methylation level of the in vivo sample closer to that of normal tissue in some CGIs. Therefore, the cut-off value in KA samples should be lower or higher than that in SCC samples to detect CGIs with DNA hypermethylation or hypomethylation, respectively, in KA samples when compared to NHEKs and/or SCC cell lines.
Bisulfite sequencing was performed to confirm the results of RT-MSP. The bisulfite sequencing data were largely consistent with the RT-MSP data. These results suggest that RT-MSP in the present study is a reliable method to assess the methylation levels of CGIs located on CCDC17, PVR, and MAP3K11 gene bodies.
Statistical analysis unexpectedly suggested that the methylation level of the CGI located on the PVR gene body might be related to regional lymph node metastasis; low methylation levels in the primary lesion might be related to regional lymph node metastasis. To confirm that the methylation level can predict regional lymph node metastasis, a cohort study is needed. We will perform careful follow up on the patients with low methylation levels in the CGI but no lymph node metastasis.
The statistical analysis unexpectedly suggested that the methylation level of the CGI located on the MAP3K11 gene body might be related to age; a higher methylation level in KA might be related to increased age. The coefficient of determination (R 2 ) of 0.245 indicates a weak association between the methylation level and age. Although the data is compatible with the previous report that DNA methylation is affected by aging [21], a larger study is needed to confirm the correlation.
The methylation levels of CGIs on CCDC17, PVR, and MAP3K11 gene bodies in KA and SCC samples ranged widely from low levels, which are comparable to levels in normal skin, to high levels of 100%, or almost 100%. These wide ranges could be caused by considerable contamination with tissues surrounding the tumor that reduced the methylation levels in in vivo  (a) The association between the methylation level of the CGI located on the PVR gene body and regional lymph node metastasis in SCC.
The vertical and horizontal axes indicate methylation level (%) and regional lymph node metastasis, respectively. Plus (+) and minus (-) indicate presence and absence of regional lymph node metastasis, respectively. An asterisk represents a significant difference (P < 0.05). (b) The association between the methylation level of the CGI located on the MAM3K11 gene body and age in KA. The vertical and horizontal KA or SCC samples. Alternatively, the CGIs might be less methylated in some KAs or SCCs, suggesting that the hypermethylation of CGIs is not essential for tumorigenesis in KA and SCC. Previous literature describes the functions of proteins encoded by PVR and MAP3K11, although the function of the protein encoded by CCDC17 has not been studied in detail. PVR encodes CD155 of a transmembrane glycoprotein belonging to the immunoglobulin superfamily. Soluble isoforms of CD155 are increased in various cancers [22]. CD155 plays a key role in cell motility during tumor cell invasion and migration [23]. These findings might be compatible with our data of PVR gene body methylation perhaps leading to gene transcription promotion in KA and SCC. On the other hand, Knackmuss et al. reported that MAP3K11 protein might function as an important tumor suppressor neutralized by oncomiR-125b in B-cell leukemia [24], although the function of MAP3K11 in epidermal tumors is unknown.
A limitation of this study is that the results of other cut-off values in the genome-wide analysis were not studied. The analyses might be enormous when various cut-off values are challenged. Since the main aim of the present study is to determine if aberrant DNA methylation occurs in KA, we selected only a few cut-off values in each genome-wide analysis of KA samples, SCC cell lines, or NHEKs, although the values may be subjective.

Conclusions
The present study demonstrated that aberrant DNA methylation occurs in KA.

Resources: YN.
axes indicate methylation level (%) and age (years old), respectively. The methylation level of the CGI located on the MAP3K11 gene body in KA is correlated to age (P = 0.031, linear regression analysis; R 2 = 0.245).