UVB irradiation does not directly induce detectable changes of DNA methylation in human keratinocytes

Cells and biological materials

Normal human keratinocytes (Clonetics; San Diego) were grown in EpiLife Medium (Invitrogen; Carlsbad, CA). The restriction enzyme used for combined bisulfite restriction analysis (COBRA), Taq^αI (5′-TCGA-3′), was obtained from New England Biolabs (Ipswich, MA).

UVB radiation treatment

The UVB source we used consisted of three fluorescent light tubes (Philips TL 20 W/12R) filtered through a cellulose acetate sheet, which eliminates wavelengths below 295 nm. The source has a peak spectral emission at 312 nm. The keratinocytes were grown in 150 mm cell culture dishes. The cells were irradiated with doses of 260 J/m² (high dose) and 130 J/m² (low dose) of UVB after the medium had been removed and cells had been washed three times with phosphate buffered saline. After UVB exposure, the cells were grown in new culture medium for three days. They were then irradiated again, and then nine more times with a two or three day recovery time between each irradiation cycle. After the final irradiation dose was delivered, the cells were grown for a recovery period of eight days or 18 days.

DNA isolation

The cells were trypsinized and collected by centrifugation. After a proteinase K treatment, DNA was isolated with a standard phenol/chloroform extraction method followed by ethanol precipitation²⁵.

MIRA and microarray

To detect potential genome-wide changes in DNA methylation patterns after the UVB treatments, the methylated-CpG island recovery assay (MIRA) combined with microarray analysis was used as described previously^26,27. Nimblegen’s Signalmap program was used to visualize the DNA methylation data and for generation of profiling snapshots.

Bioinformatics analysis

Loess normalization was applied to the raw intensity files of each array to correct intensity-dependent dye bias and obtain log2 ratios between MIRA and input samples. Then the log2 ratios across all the samples were quantile-normalized. Probes were considered positive if their normalized log2 ratios were above 2-fold. Peaks in each sample were called if a minimum of four consecutive positive probes were present with either one gap or no gaps. To identify hypermethylated and hypomethylated targets in UVB-treated samples vs. untreated control samples, the average probe log2 ratio signals within the peaks identified in each UVB-treated sample were compared to the untreated samples. Only the peaks with an average log2 ratio signal difference of more than log2(3) were considered hypermethylated or hypomethylated peaks. These differential peaks were annotated to the Refseq transcript database downloaded from the UCSC genome database. The microarray data have been deposited into the GEO database (accession number GSE42943).

DNA methylation analysis by COBRA

Candidate loci were investigated by combined bisulfite restriction analysis (COBRA)²⁸. PCR was performed with primers and conditions listed in Supplementary Table S1 and Supplementary Table S2. Briefly, COBRA-PCR was performed with bisulfite-converted DNA-specific primers using 50 ng of bisulfite- modified genomic DNA as template for 60 cycles after a 15 min incubation at 95°C, then 30 s at the T_A (see Table S2) and 30 s at 72°C in a 25 µl volume containing 5 nmol dNTPs, 20 pmol primers, and 1.25 units of Hot start Taq DNA polymerase (Qiagen, Valencia, CA). Five microliters of the PCR product was analyzed on a 2% agarose gel. Equal amounts of PCR product were digested with the restriction enzyme Taq^αI (5′-TCGA-3′).

Bisulfite sequencing

COBRA PCR products were created as described above. The PCR products were ligated into a cloning vector (TOPO^® Cloning Kit; Invitrogen, Grand Island, NY, or the pGEM^®-T-Easy Kit, Promega, Madison, WI) and transformed into competent cells. Different clones were picked randomly and the plasmids were isolated and sequenced. For the analysis of methylated and unmethylated cytosines, the free software program Bioedit was used.

Genome-scale DNA methylation analysis of UVB-irradiated keratinocytes

We irradiated human keratinocytes with two different doses of UVB, 130 J/m² and 260 J/m². These doses were well tolerated by the cells and did not produce overt losses in cell viability. Cells were irradiated chronically (11 times total) with these doses with a 2–3-day recovery period between each irradiation cycle. Controls included cells that were not irradiated and cells irradiated once with 130 or 260 J/m² of UVB, but then harvested immediately after the irradiation.

After the irradiation and final recovery times of eight or eighteen days, DNA was isolated from the cells. DNA was sonicated and the methylated fraction of the genome was enriched by use of the methylated CpG island recovery assay (MIRA) technique²⁹. The methylated fraction was hybridized relative to input DNA onto NimbleGen CpG island plus promoter arrays. These arrays cover all ~28,000 CpG islands of the human genome and all Refseq gene promoters from -2.4 kb to +0.6 kb relative to the transcription start site. Analysis of methylation patterns using NimbleGen’s Signalmap display software indicated the excellent reproducibility of the data (Figure 1). The patterns of methylation peaks were remarkably similar between all four controls and all three UVB-irradiated samples. Bioinformatics analysis was used to identify potential differences between the control and treatment groups. Methylation peaks were identified as described in Materials and Methods. Peaks with an average log2 ratio signal difference of more than log2(3) were considered as hypermethylated or hypomethylated peaks, respectively. Table 1 summarizes the number of differences identified between UVB and control treatment groups. Most comparisons revealed only a handful of differences and the numbers of differential peaks were generally below 100 for each comparison. In fact, comparison between two controls, non-irradiated cells grown for 18 days or 8 days (N18con vs. N8con), respectively, showed a greater number of differences than any comparison between a UVB-irradiated and a control sample (Table 1). Nonetheless, a small number of differential peaks could clearly be detected on SignalMap profiles (Figure 2). We show examples for the genes CXXC5, PPP3CB, IL17C, CCDC40 and C21orf29, where either hypermethylation or hypomethylation in a UVB-irradiated sample was observed (Figure 2).

Figure 1. DNA methylation peaks in control and UVB-irradiated keratinocytes.

A. A random segment of chromosome 1 is shown to indicate the reliability of the method and to show the uniform peaks between control and UVB-irradiated samples. The positions of transcripts and CpG islands are indicated. The samples labeled ‘low UVB control’ and ‘high UVB control’ represent DNA immediately harvested following a one-time irradiation of cells with either 130 J/m² (low) or 260 J/m² (high) of UVB. Samples labeled ‘8 days control’ or ’18 days control’ were grown for the same time periods as the irradiated cells but were never irradiated. Samples labeled ‘18 days low UVB’ or ‘18 days high UVB’ were chronically irradiated with 130 J/m² (low) or 260 J/m² (high) of UVB followed by an 18 day recovery period. The sample labeled ‘8 days high UVB’ was chronically irradiated with 260 J/m² of UVB followed by an 8 day recovery period. B. Methylation peaks are shown for randomly selected genomic regions on chromosomes 1, 3, 5, 7, 11, 17 and X. The positions of transcripts and CpG islands are indicated. The chromosomal coordinates are shown above each snapshot.

Figure 2. Apparent DNA methylation differences in control and UVB-irradiated keratinocytes.

Methylation peaks are shown for the genes CXXC5, PPP3CB, IL17C, CCDC40, and C21orf29. The positions of transcripts and CpG islands are indicated. The chromosomal coordinates are shown above each snapshot. The signals framed by the red rectangles are significantly (*) differentially methylated pairs of a control sample and a UVB-treated sample, as determined by bioinformatics analysis. The description of the samples is shown in Figure 1.

Bisulfite-based DNA methylation assays

To verify the apparent methylation differences observed by microarray analysis, we performed sodium bisulfite conversion-based DNA methylation assays. COBRA analysis is shown for the genes CXXC5, PPP3CB, IL17C, CCDC40 and C21orf29 in Figure 3. Cleaved molecules in these assays indicate methylated restriction sites that are resistant to bisulfite conversion and remain cleavable by the CpG-targeting restriction enzyme after PCR. Uncut molecules represent unmethylated DNA fragments. The COBRA assays indicated that the methylation patterns were the same or very similar between DNA isolated from control cells and DNA from UVB-irradiated cells.

Figure 3. DNA methylation analysis by COBRA of candidate differentially methylated genes.

COBRA analysis of restriction enzyme sites indicates no substantial difference in DNA methylation in UVB-irradiated cells versus controls. Samples labeled ‘8 d high UVB’ and ‘18 d high UVB’ were chronically irradiated with 260 J/m² of UVB followed by an 8 day or 18 day recovery period, respectively. Samples labeled ‘8 d control’ or ‘18 d control’ were grown for the same time periods as the irradiated cells but were never irradiated. The samples labeled ‘high UVB control’ represent DNA immediately harvested following a one-time irradiation of cells with 260 J/m² of UVB. The genes CXXC5, PPP3CB, IL17C, CCDC40 and C21orf29 were analyzed by Taq^αI digestion. The PCR products were cleaved with the enzyme (E) or were left uncleaved (-). The methylated and unmethylated control samples at the right side of the gel panels were fully CpG-methylated and unmethylated control DNAs.

COBRA assays, although generally indicative of the methylation status of a genomic target, can score only a limited number of CpG sites. Therefore, we performed sodium bisulfite sequencing to provide the methylation status of all CpG within the amplified target fragments. These assays also indicated no substantial difference between control and UV-irradiated cells (Figure 4). Similar methylation patterns were observed for the five different gene targets, regardless of whether the cells were UVB-irradiated or not. A somewhat lower frequency of methylated CpG sites was observed for the IL17C gene in UVB-irradiated cells (18.4% versus 29%). However, these results may be biased by the few molecules in the population that had almost every CpG methylated. Taken together, our results suggest that the rather small number of differential peaks observed by bioinformatics analysis were false positives. Such small numbers of false positive differences can be expected when comparisons are made for over 28,000 CpG islands and about 20,000 Refseq promoters.

Figure 4. DNA methylation analysis by bisulfite sequencing of candidate differentially methylated genes.

Bisulfite sequencing data are shown for the genes CXXC5 (A), PPP3CB (B), IL17C (C), CCDC40 (D), and C21orf29 (E). The chronic UVB dose was 260 J/m². Open circles represent unmethylated CpG sites and black circles indicate methylated CpG sites. The percentage of methylated sites is indicated below the sequence data for individual cloned molecules.