DNA methylation profiles of DME genes in human tissues
We examined DNA methylation profiles of two adult liver tissues (NLA and NL2), fetal
liver tissue (NLF), adult small intestinal tissue (NSI), and three hepatoma cell lines
(HepG2, HuH7, and JHH1) by HumanMethylation450 Bead Chip. These data were combined
with GEO-registered datasets of 18 healthy adult liver tissues. The levels of DNA
methylation for more than 480,000 CpG sites were determined as the ? values (0?????1). Cluster analysis demonstrated that methylation levels of DME
genes, including 55 CYP genes and 62 phase II DME genes, differed markedly among distinct tissue groups.
The representative profiles of six CYPs (CYP1A2, CYP1B1, CYP2C9, CYP2C19, CYP2D6, and CYP3A4), and two controls (ACTB and BMP4) were shown in Fig. 1. Although the 20 adult liver tissues were derived from two different race populations
(2 Chinese and 18 German), they exhibited similar DNA methylation profiles. However,
some DME genes such as CYP1A2, CYP2C9, CYP2C19, CYP2D6, and CYP3A4 showed relatively variable methylation statuses, compared to CYP1B1 and two control genes. To examine the landscape of DNA methylation in more detail,
we performed the DNA methylation mapping of individual DME genes (Fig. 2 and Additional file 1: Figure S1). For each CpG site, the range of variation was defined as a ?R
value, which was calculated from the difference between the highest and lowest ? values among the 20 livers. Next, for all CpG sites located in the 5? regulatory
region (defined as TSS1500, TSS200, or 5?UTR in the array platform), the maximum ?R
value was used for estimation of the degree of inter-individual differences in DNA
methylation. The distribution of maximum ?R
values (degrees of variation) of the genes is shown as a histogram in Fig. 3. We expected that the control genes would show the least variation because the expression
of these genes should be tightly regulated by DNA methylation on their promoter CpG
islands, as reported by Edgar et al. 19]. Therefore, these genes were used as negative controls of inter-individual variation.
We finally identified 37 (32Â %) DME genes with significantly variable DNA methylation
statuses, for which the maximum ?R
values were more than those of all eight control genes (ACTB, B2M, GAPDH, TBP, BMP4, IGFBP3, MLH1, and MGMT) (maximum ?R
??0.296).
Fig. 1. Heat maps of the DNA methylation profiles of the representative 6 CYP and 2 control genes by cluster analysis. The top of the map indicates samples from
20 normal adult livers (L1 to L81, NLA and NL2), 1 fetal liver (NLF), 1 adult small
intestine (NSI), and 3 hepatoma cell lines. The right of the map indicates the target
ID and gene name (CYP1A2, CYP1B1, CYP2C9, CYP2C19, CYP2D6, CYP3A4, ACTB, and BMP4) examined for individual CpG sites. The representative methylation profiles of individual
genes were selected by excluding profiles showing the similar patterns each other.
Therefore, the hierarchical similarity tree was not shown in the map. Higher DNA methylation
levels are shown in red, while lower DNA methylation levels are shown in black
Fig. 2. Representative results of DNA methylation mapping for 20 liver tissues. In each panel,
the horizontal axis indicates the positions of CpG sites arranged in the 5? to 3?
direction. The 5? regulatory region (TSS1500, TSS200, or 5?UTR) is underlined. The
vertical axis indicates the ? value, with the variations expressed as box-and-whisker plots. The maximum ?R
value within the 5? regulatory region is indicated by a circle for each gene. aCYP1B1, bCYP3A4, c ?-actin (ACTB), and d bone morphogenetic protein 4 (BMP4)
Fig. 3. Distribution of maximum ?R
values of the DME and control genes. The maximum ?R
values of the eight control genes colored in red were relatively low (?0.296). We defined highly variable DNA methylation status as
a ?R
value of more than 0.296. The ?R
values of representative DME genes are also shown in the histogram. ACTB actin, beta, B2M ?2 microglobulin, BMP4 bone morphogenetic protein 4, GAPDH glyceraldehyde-3-phosphate dehydrogenase, IGFBP3 insulin-like growth factor binding protein 3, MGMT O-6-methylguanine-DNA methyltransferase, MLH1 mutL homologue 1, TBP TATA box-binding protein
DME genes regulated by DNA methylation in adult livers
Transcript datasets were also registered for 10 of the 18 liver tissues examined for
DNA methylation analysis. We analyzed the mRNA expression profiles of DME genes in
these 10 adult livers and detected the highest expression of CYP2E1 and CYP3A4 genes, consistent with a previous report 20] (Additional file 2: Figure S2A). In addition, the majority of the DME genes exhibited variable levels
of mRNA expression compared to the housekeeping genes (with coefficients of variation
(CVs) of more than 17.1 %; Additional file 2: Figure 2B). We found inverse correlations between mRNA expression and DNA methylation levels
in seven DME genes (CYP1A2, CYP2C19, CYP2D6, GSTA4, GSTM5, GSTT1, and SULT1A1). In particular, the CYP2C19, GSTA4, and GSTM5 genes had CpG sites that simultaneously showed inverse correlations and highly variable
methylation statuses (?R
??0.296; Fig. 4 and Additional file 3: Figure S3).
Fig. 4. Representative results of correlation analysis. A significant inverse correlation
between DNA methylation level (? value) and mRNA expression level was detected for the CYP2C19 (a) and GSTA4 (c) genes (p??0.05, Spearman’s rank correlation test). The level of mRNA expression, shown as
the vertical axis, was normalized to the expression level of ACTB (set as 100Â %). CpG sites with significant correlations are indicated by asterisks in methylation mapping for the CYP2C19 (b) and GSTA4 (d) genes. Additionally, CpG sites with ?R
values of more than 0.296 are indicated by circles
DME genes regulated by DNA methylation in three hepatoma cell lines
Cluster analysis revealed that the three hepatoma cell lines showed clearly distinct
methylation profiles compared to normal liver tissues (Fig. 1). Aberrant DME gene methylation detected in hepatoma cells is summarized in Additional
file 4: Figure S4. We defined a hypermethylated or hypomethylated CpG site as having a ? value of more than 0.5 for comparisons between hepatoma cell lines and normal liver
(NL2). This cutoff value was determined by validation analysis, as described below.
Using this criterion, 36 DME genes had hypermethylated CpG sites within the 5? regulatory
region in at least one hepatoma cell line. To determine whether the hypermethylation
observed in hepatoma cells was associated with downregulation of the DME gene, DNA
methylation was reversed by treatment with 5-aza-2?-deoxycytidine (DAC), and mRNA
expressions were examined using a SurePrint G3 Human Gene Expression 8?×?60K v2 microarray.
As a result, 44 (38Â %) DME genes showed upregulation by more than twofold following
DAC treatment (Additional file 5: Figure S5). Therefore, these DME genes were downregulated by specific DNA methylation
events. Among these DME genes, the downregulation of CYP1B1, CYP8B1, GSTM2, GSTP1, UGT2B15, and UGT3A2 in hepatoma cells could be explained by DNA methylation status. The results of array-based
DNA methylation and mRNA expression analyses were examined for validity using combined
bisulfite restriction analysis (COBRA) and quantitative real-time PCR assays, respectively.
The criterion for judging hypermethylation was determined by the results of the tumor
suppressor genes used as positive controls (Additional file 6: Figure S6). By using a cutoff value of ???0.5, we identified five DME genes (CYP1B1, CYP8B1, GSTM2, GSTP1, and UGT3A2) that were regulated by DNA methylation in hepatoma cells. Representative results
for the CYP1B1 gene are shown in Fig. 5.
Fig. 5. Validation of mRNA expression and DNA methylation of the CYP1B1 gene in hepatoma cells. a The level of DNA methylation was mapped on the CYP1B1 gene locus in the three hepatoma cell lines. The gene structure and location of the
CpG sites are shown in the lower panel. b The levels of CYP1B1 mRNA were detected by quantitative real-time PCR in hepatoma cells treated with TSA
and/or DAC. The vertical axis indicates the mRNA level of treated cells relative to
paired control cells without treatment. Each column represents the mean?±?SD (?=?3). c The CYP1B1 gene methylation status of the three control cell lines and adult liver tissues was
examined by COBRA assay. DNA fragments cleaved by TaqI digestion represent methylated
DNA (M), while noncleaved fragments represent unmethylated DNA (U)
DNA methylation and alternative splicing of UGT1A isoforms
Next, the transcript levels of UGT1A isoforms were examined in different tissues using quantitative real-time PCR. The
tissue-specific expression profiles shown in Additional file 7: Figure S7 were similar to the results of a previous study 10]–12]. The UGT1A genes were then classified into two major groups according to the tissues in which
they were dominantly expressed: hepatic type (i.e., UGT1A1, UGT1A3, UGT1A4, UGT1A6, and UGT1A9) and intestinal type (i.e., UGT1A5, UGT1A7, UGT1A8, and UGT1A10). Although the level of UGT1A8 expression in adult livers was somewhat higher than that in small intestines, UGT1A8, unlike the other hepatic-type UGT1As, showed relatively high expression in the small intestine as well. Therefore, we
classified UGT1A8 as an intestinal-type gene. DNA methylation mapping on the UGT1A locus revealed that DNA methylation status was largely variable among different tissues
(Additional file 8: Figure S8). We focused on the first exon of each isoform and found that the hepatic-type
genes UGT1A1, UGT1A4, UGT1A6, and UGT1A9 tended to be methylated at higher levels in the small intestine than in the liver
(Fig. 6). In order to perform statistical analysis, moreover, DNA methylation status in individual
CpG sites were also compared between 18 adult livers and one normal small intestine
(NSI) or one normal fetal liver (NLF). We further focused on the CpG sites most proximal
to each transcription start site (excluding more distal CpG sites of TSS1500). In
hepatic-type UGT1A genes, as a result, we found that ? values of the NSI and NLF were statistically higher than medians of 18 individual
? values of normal adult livers (Additional file 9: Table S1). In contrast, the methylation levels of the intestinal-type genes UGT1A10 and UGT1A8 were relatively higher in the liver than in the small intestine. This suggested that
DNA methylation status around the first exon determined the splicing isoforms of the
UGT1A gene and led to tissue-specific expression. In addition, higher levels of DNA methylation
were observed in fetal livers than in adult livers. This tendency was found in hepatic-type
genes rather than intestinal-type genes, implying that the low levels of expression
observed in the fetal liver may result from downregulation by DNA methylation.
Fig. 6. Representative results of DNA methylation mapping of UGT1A isoforms. The levels of the DNA methylation of hepatic-type (UGT1A1, UGT1A4, UGT1A6, and UGT1A9) and intestinal-type (UGT1A10 and UGT1A8) genes were examined in adult liver tissues (NLA and NL2), fetal liver tissue (NLF),
and adult small intestinal tissue (NSI). In each panel, the open or closed circles located on the horizontal axis indicate the positions of CpG sites arranged in the 5? to 3? direction according
to the relative distance of each CpG site. The closed circles indicate CpG sites in which tissue-specific expression could be explained by the
DNA methylation status
Classification of DME genes based on their DNA methylation landscape
Interestingly, DME genes had unique features of the DNA methylation landscape (Fig. 1), which could be classified into at least three groups, as summarized in Fig. 7. The first group showed highly variable methylation among normal livers (maximum
?R
??0.296) and inverse correlations with mRNA expression. DME genes in this first group
may be candidates for explaining inter-individual variations in drug metabolizing
activity; we classified genes in this group as the highly variable methylation (HVM)
type. Genes in the second group, similar to the CYP1B1 gene, showed stable methylation statuses among normal livers but were hypermethylated
in tumor cells. These DNA methylation features were similar to those of tumor suppressor
genes (i.e., the TSG type). The last group retained low levels of methylation in both
normal and tumor livers, suggesting these genes may act as housekeeping genes. Typical
housekeeping genes, such as ACTB and GAPDH, had similar features of methylation mapping and expression profiles in our study.
For example, the level of DNA methylation was highly stable and low in the 5? regulatory
region (hypomethylated 5?UTR) but considerably high within the gene body (hypermethylated
exons). In addition, the mRNA expression of the housekeeping genes was stable among
normal livers, with CVs of less than 17.1Â %. Although it was difficult to fulfill
all these criteria, we found two DME genes (CYP2R1 and CYP46A1) that had characteristics similar to those of housekeeping genes (i.e., HKG type).
Fig. 7. Classification of DME genes based on the corresponding DNA methylation landscape