- Open Access
Depletion of DNMT1 in differentiated human cells highlights key classes of sensitive genes and an interplay with polycomb repression
- Karla M. O’Neill†1, 5,
- Rachelle E. Irwin†1,
- Sarah-Jayne Mackin1,
- Sara-Jayne Thursby1,
- Avinash Thakur1, 6,
- Ciske Bertens1, 2,
- Laura Masala1, 3,
- Jayne E. P. Loughery1,
- Darragh G. McArt4 and
- Colum P. Walsh1Email authorView ORCID ID profile
© The Author(s) 2018
- Received: 4 January 2018
- Accepted: 21 March 2018
- Published: 29 March 2018
DNA methylation plays a vital role in the cell, but loss-of-function mutations of the maintenance methyltransferase DNMT1 in normal human cells are lethal, precluding target identification, and existing hypomorphic lines are tumour cells. We generated instead a hypomorphic series in normal hTERT-immortalised fibroblasts using stably integrated short hairpin RNA.
Approximately two-thirds of sites showed demethylation as expected, with one-third showing hypermethylation, and targets were shared between the three independently derived lines. Enrichment analysis indicated significant losses at promoters and gene bodies with four gene classes most affected: (1) protocadherins, which are key to neural cell identity; (2) genes involved in fat homoeostasis/body mass determination; (3) olfactory receptors and (4) cancer/testis antigen (CTA) genes. Overall effects on transcription were relatively small in these fibroblasts, but CTA genes showed robust derepression. Comparison with siRNA-treated cells indicated that shRNA lines show substantial remethylation over time. Regions showing persistent hypomethylation in the shRNA lines were associated with polycomb repression and were derepressed on addition of an EZH2 inhibitor. Persistent hypermethylation in shRNA lines was, in contrast, associated with poised promoters.
We have assessed for the first time the effects of chronic depletion of DNMT1 in an untransformed, differentiated human cell type. Our results suggest polycomb marking blocks remethylation and indicate the sensitivity of key neural, adipose and cancer-associated genes to loss of maintenance methylation activity.
- Body mass
- Cancer/testis antigen
DNA methylation is an important mechanism for epigenetic regulation of genes in both mouse and human . It occurs mainly at the CpG dinucleotide, and methylation at this symmetrical site is efficiently maintained during replication by the action of the DNA methyltransferase 1 (DNMT1) enzyme . Methylation is known to play an important role in regulating imprinted loci , genes on the inactive X chromosome  and germline-specific genes  in mouse.
Where methylation occurs at the promoter of a gene, it is strongly associated with the silencing of transcription, particularly if there is a high density of CpGs, a so-called CpG island (CGI). However, studies have shown that most CGI are intrinsically protected from methylation [6, 7] and only a small number shows dynamic changes during development, mostly in the three classes mentioned above [5, 8], though there may be others which have not yet been clearly defined. As you move outward from an island, the shores and shelves show higher levels of methylation and greater dynamic response , though here the link to changes in gene activity is less clear . Methylation is also associated with larger regions of inert chromatin, such as the inactive X, pericentromeric repeats and regions rich in transposable elements , generally consistent with a repressive role. Recent genome-wide surveys have also indicated that high levels of methylation are found in the bodies of active genes, where they may facilitate transcription [11, 12]. In keeping with this, we and others recently showed that artificially decreasing intragenic methylation levels reduced steady-state transcript levels, consistent with a positive role for methylation in the gene body [11–13].
Another major system for epigenetic repression is via histone modification, particularly by the polycomb group of proteins, with EZH2 being one of the main enzymes involved . A number of studies suggest an interplay between polycomb- and DNMT-mediated repression, with a generally negative correlation between DNA methylation and the H3K27me3 mark deposited by EZH2 [15, 16]. Supporting this, a loss of DNA methylation caused a reshaping of the histone landscape and derepression of some polycomb targets in mouse ES cells , suggesting that DNA methylation helps to determine where polycomb marks are deposited.
While DNMT1 is the main maintenance methyltransferase, there also appears to be an important role for the de novo enzymes DNMT3A and DNMT3B in complementing that activity at some loci [18, 19]. In order to clarify which genes are most sensitive to DNMT1 loss in human, a number of studies have been carried out using mutations within the gene to assess the effects of loss of methylation [19–22]. While this has been a fruitful approach in mouse embryonic stem (ES) cells, where null mutants are tolerated, differentiation of the mouse cells leads to cell death [20, 22, 23], whereas DNMT1 disruption in human ES cells is not tolerated even in undifferentiated cells . Genetic ablation in adult differentiated cells also leads to cell death within a few cell cycles, before passive demethylation of the genome can occur [23, 25]. One of the best-studied systems in humans consists of HCT116 colon cancer cells carrying a hypomorphic allele in the DNMT1 gene together with a DNMT3B knockout (HCT116 DKO cells) [26–28]. Blattler et al.  found that there was widespread and relatively uniform demethylation across the genome in the DKO cells, with small effects at CGI (most of which are normally unmethylated anyway) and relatively few genes showing derepression. There was no enrichment by gene ontology (GO) analysis, but some effect at enhancers: however, this is complicated by the presence of the DNMT3B knockout alleles. Acute depletion of DNMT1 using an siRNA-mediated approach in embryonal carcinoma cells also found regions of low CpG density (open sea, shelf) to be the most affected by loss of methylation . Among the small number of dysregulated genes, there was some enrichment for cell morphogenesis and phosphorylation pathways.
Neither of these cancer cell lines, however, are a good model for the normal differentiated cell as they are transformed, aneuploid, hypermethylated, and contain a number of different mutations in key regulatory genes. Additionally, acute depletion of DNMT1 results in cell cycle delay, triggering of the DNA damage response and increased rates of cell death [24, 25, 30], making it difficult to separate acute and chronic effects.
To circumvent some of the difficulties outlined above, we generated a series of isogenic human cell lines derived from the hTERT-immortalised normal fibroblast line hTERT1604 as previously described . These are normosomic and non-transformed, and by using a stably incorporated plasmid with an shRNA targeting DNMT1 we were able to isolate a number of clonally derived lines to allow identification of any cell line-specific effects. While these showed initially the range of shared features indicative of a global response to the loss of this critical regulator, including cell cycle delay, demethylation of imprinted genes and others, they could be cultured for longer under selection , allowing identification of loci with particular sensitivity for decreased maintenance methyltransferase activity. Here we set out to completely characterise the methylation changes seen in the cell lines using the Illumina Infinium HumanMethylation450 BeadChip (450k) array platform  and subsequent analysis using the RnBeads pipeline . These approaches were chosen due to their high reproducibility and low inter-operator variability, ensuring the reliable and sensitive detection of alterations in methylation. A sample of the observations was then further verified using locus-specific assays. In addition and for the same reasons, we used the HT-12 Expression v4 BeadChip array, to assay changes in transcription in our cell lines.
The parental or wild-type (WT) adherent hTERT1604 lung fibroblast cell line  was cultured in 4.5 g/l glucose DMEM (Thermofisher, Loughborough, UK) supplemented with 10% FBS and 2× NEAA (Gibco/Thermofisher). Generation of the hTERT1604 cell lines stably depleted of DNMT1 using a pSilencer construct (Thermofisher) has been previously described . Knockdown (KD) cells were maintained as for WT, but medium was supplemented with 150 μg/ml hygromycin B (Invitrogen/Thermofisher, Paisley, UK), which was removed at least 48 h before any experimental procedure. Treatment of cells with siRNA for 24 h was as previously described : for the pulse-chase experiment cells were afterwards allowed to recover in normal media and passaged as required for up to 36 days. The siRNA (Dharmacon ON-TARGETplus SMARTpool) for DNMT1 and DNMT3B, as well as scrambled control, was obtained from Invitrogen/Thermofisher. HCT116 and double knockout (DKO) cells  were cultured in 1 g/l glucose DMEM (Gibco) supplemented with 10% FBS and 1× NEAA (Gibco). DZNeP (Sigma-Aldrich, Dorset, UK) was used at a final concentration of 1 μM.
DNA extraction and bisulphite conversion
Genomic DNA was harvested from cells in log phase of growth. Samples were incubated overnight at 55 °C in lysis buffer [50 mM Tris pH 8, 0.1 M EDTA (both Sigma-Aldrich), 0.5% SDS, 0.2 mg/ml proteinase K (Roche, West Sussex, UK)], with rotation, and DNA was subsequently isolated using the standard phenol/chloroform/isoamyl alcohol (25:24:1 pH8, Sigma-Aldrich) extraction method. DNA quality was verified using gel electrophoresis and UV absorbance measurements at 260/280 and 260/230 nm using a Nanodrop UV spectrophotometer (Labtech International, Ringmer, UK). Bisulphite conversion of 500 ng of DNA was carried out using the EpiTect bisulphite kit (Qiagen, Crawley, UK) according to the manufacturer’s instructions.
Hybridisation to 450K array and bioinformatic analyses
Three samples from each cell line were used to prepare DNA, with at least one biological repeat in each set. DNA was assessed for purity and integrity as above prior to quantification using the Quant-iT PicoGreen dsDNA assay kit (Thermo Fisher Scientific) as per manufacturer’s instructions. In total, 500 ng of high-quality bisulphite-converted (Zymo Research) DNA was checked for purity and fragmentation on a bioanalyser and then loaded on the Infinium HumanMethylation450 BeadChip  and imaged using an Illumina iScan (Cambridge Genomic Services). Output files in IDAT format were processed using the RnBeads  methylation analysis package (v1.0.0) which carries out all the analysis from import to differential methylation within the R platform (3.2.0). Briefly, quality control used the built-in probes on the array and included filtering out of probes containing SNPs, and checking for hybridisation performance. Normalisation was then carried out using the SWAN method in minfi  after background subtraction with methylumi.noob. The exploratory analysis module was used to generate probe density distributions and scatter graphs. The differential methylation analyses was based on a combined ranking score, which combined absolute effect size, relative effect sizes and p-values from statistical modelling into one score where rank is computed as the most conservative value among mean difference in means, mean in quotients and combined p value across sites in the region: the enrichment analysis used the combined rank among the 1000 best-ranking regions and a hypergeometric test to identify GO terms in the AmiGO 2 database . Pairwise comparison of triplicate samples from each cell line against WT hTERT was also made to determine change in beta value and associated combined p-value, adjusted for multiple comparison using false discovery rate (FDR). Some tailored analyses were also carried out using custom scripts in R. Additional GO studies were performed using DAVID (v6.7) .
We used the GALAXY platform  to map sites showing highly reproducible changes (FDR < 0.05) against the locations of RefSeq genes or ChromHMM regions on the UCSC genome browser  for each cell line. GO category genes which showed changes in methylation at multiple sites in more than one KD cell line were scored as true hits (Yes in the FDR column), while GO categories with few or no sites reproducibly altered across replicates (FDR > 0.05) or where methylation changes were small (< 0.1 β), inconsistent in direction, or not found in more than one KD cell line, were scored as false positives. Absolute β levels were used to measure median methylation across genes of interest using custom workflows in GALAXY, with further statistical analyses in Statistical Package for the Social Sciences software (SPSS) version 22.0 (SPSS UK Ltd).
Locus-specific methylation analysis
Amplification was carried out using the PyroMark PCR kit (Qiagen) with 2 μl bisulphite-converted DNA, 12.5 μl MasterMix, 2.5 μl CoralLoad Concentrate, 1.25 μl each primer (10 μM) and 5.5 μl nuclease-free H2o using the following conditions: 15 min at 95 °C followed by 45 cycles of 94 °C for 30 s, 56 °C for 30 s, 72 °C for 30 s and a final elongation step of 72 °C for 10 min. Pyrosequencing was carried out on the PyroMark Q24 System, according to the manufacturer’s instructions (Qiagen). Most assays were designed in-house using the PyroMark Assay Design software 2.0 (LEP, MAGEA12, OR10J5, OR51E2, OR2AG1, PCDHA2, PCDHC4, UGT1A1, UGT1A4) prior to synthesis (Metabion, Germany): see Additional file 1: Table S1 for details: DAZL, SYCP3, D4Z4 and NBL2 were as described [34, 40]. In some cases, pre-designed pyrosequencing primers were obtained from Qiagen (GABRQ PM00133483, GHSR PM00014350, SNRPN PM00168252). Clonal analysis was carried out as previously described .
Hybridisation to HT-12 microarray and bioinformatic analyses
Total RNA was extracted using the RNeasy minikit (Qiagen) as per manufacturer’s instructions, including a DNase step. RNA integrity was verified via gel electrophoresis, and quality and quantity were verified using a SpectroStar (BMG Labtech, Aylesbury, UK) and a bioanalyser (Agilent Technologies, Cheadle, UK). Two hundred nanograms of total RNA underwent linear amplification using the Illumina TotalPrep RNA Amplification Kit (Life Technologies/Thermofisher, Paisley, UK) following the manufacturer’s instructions. Microarray experiments were performed at Cambridge Genomic Services, University of Cambridge, using the HumanHT-12 v4 Expression BeadChip (Illumina, Chesterford, UK). After scanning the data were loaded in GenomeStudio (Illumina) and then processed in R (version 3.2.2). The data were filtered to remove any non-expressed probes using the detection p-value from Illumina, transformed using the variance stabilization transformation (VST) from lumi and normalised using the quantile method. Comparisons were made using the limma package with results corrected for multiple testing using false discovery rate (FDR) testing.
RNA and protein analysis
Transcriptional assays at individual loci using RT- and RT-qPCR were carried out essentially as in : primer sequences are listed in Additional file 1: Table S1. Protein was extracted from cells growing in log phase using protein extraction buffer (50 mM Tris–HCl, 150 mM NaCl, 1% Triton-X, 10% glycerol, 5 mM EDTA; all Sigma-Aldrich) and 0.5 µl protease inhibitor mix (Sigma-Aldrich). For Western blotting, 30 μg protein was denatured in the presence of 5 μl 4× LDS sample buffer (Invitrogen) and 2 μl 10× reducing agent (Invitrogen) in a total volume of 20 μl nuclease-free water (Qiagen) via incubation at 70 °C. Proteins were separated by SDS-PAGE and then electroblotted onto a nitrocellulose membrane (Invitrogen) and blocked in 5% non-fat milk for 1 h at room temperature (RT). Membranes were incubated with anti-DNMT1 (a kind gift from Guoliang Xu) and anti-β-actin (Abcam ab8226) overnight at 4 °C, followed by HRP-conjugated secondary antibody incubation at RT using ECL (Invitrogen).
Statistical analysis was performed by the RnBeads package, or separately in Excel (Microsoft Office Professional Plus 2013), Prism (Graphpad) or SPSS (v22.0). Experiments were carried out in triplicate and included at least one biological replicate. PCR results were analysed using Student’s paired t-test. Pyrosequencing results were analysed by ANOVA within representative runs and using Student’s t-test on the average of multiple runs. Error bars on all graphs show standard error of the mean (SEM) or in the case of HT12 array data, 95% confidence interval (CI), unless otherwise stated. Asterisks are used to represent probability scores as follows: *p < 0.05; **p < 0.01; ***p < 0.005 or n.s. not significant.
Generation of isogenic hTERT1604 fibroblast cell lines
Characterisation of overall changes in absolute methylation levels in depleted lines
Using the 450K array  and processing in RnBeads  to assess methylation levels across the genome (Fig. 1c), there was still a wide range of methylation values (given for the array as a value β ranging from 0 to 1) in KD lines as compared to WT, but the median values were decreased as expected in all three with d8 being comparable to d10, while d16 was lower (arrowheads at right). Principle components analysis and examination of the sites showing greatest differences in methylation between the stable lines confirmed that d8 and d10 were most similar (Additional file 2: Fig. S1). Probes on the array were annotated by location relative to genomic features, and while all regions showed a decrease in methylation, the difference in median values was smallest for CGI, which were unmethylated anyway in parental cells (β < 0.1 in WT), while the separation in medians was greatest at shelves and shores, where methylation levels were higher (Additional file 3: Fig. S2B). This can most clearly be seen by plotting the difference in medians (Fig. 1d). Both WT and the KD cell lines showed the typical bimodal probe density distribution pattern reported in most cell types  (Fig. 1e). Overall, there was an increase in the numbers of less methylated probes (β < 0.25) in the KD cell lines and a decrease in the numbers of highly methylated probes (β > 0.65). For individual regions CGI again showed the smallest change, while gene bodies (genes) appeared most altered (Additional file 3: Fig. S2C).
To determine whether methylation was lost stochastically in each KD cell line given the variation seen (Additional file 2: Fig. S1), or was more targeted, we determined the degree to which affected sites were shared between the three cell lines (Fig. 1f). The largest set of sites losing methylation (17.2 × 104) was that shared between all three KD lines, supporting a non-random loss. A spike in numbers of probes showing intermediate levels of methylation (β ~ 0.50) in KD cell lines in the density profile plot (Fig. 1e, arrow) had indicated that a possible gain in methylation might also be occurring at some sites. Analysis showed that a substantial number (9.1 × 104) of sites gaining methylation are shared between all three KD lines, indicating reproducible gains in methylation at particular CpGs.
Overall pattern of sites showing significant differential methylation on DNMT1 depletion
We compared WT cells to all three KD lines using the RnBeads package in R and combined rank scoring (see methods). This confirmed that d16 has the greatest number of demethylated sites using a false discovery rate (FDR) cut-off of p < 0.05, but at p < 0.001 all three lines have comparable numbers of hypo- and hypermethylated sites (Additional file 4: Fig. S3A), with more sites losing than gaining. An analysis of the 1000 best-ranking sites highlights sites common to all three KD lines (Additional file 4: Fig. S3B), confirming that there are large numbers of sites which respond in the same way in each KD, with an excess of probes showing loss over gain.
Gene ontology analysis for differentially methylated sites
Homophilic cell adhesion via plasma membrane
Regulation of high-density lipoprotein particle clearance
Transmembrane signalling receptor activity
Calcium ion binding
Signal transducer activity
Gonadal mesoderm development
Negative regulation of cholesterol transport
Negative regulation of interleukin-6 biosynthetic process
Growth factor activity
Olfactory receptor activity
Signalling receptor activity
Histone H4-K20 demethylation
Negative regulation of megakaryocyte differentiation
Histone demethylase activity (H4-K20 specific)
Histone H4-K20 demethylation
Negative regulation of megakaryocyte differentiation
DNA replication-dependent nucleosome assembly
Histone demethylase activity (H4-K20 specific)
Olfactory receptor activity
Loss of methylation at the protocadherin gamma gene cluster particularly affects the A and B class variable genes
Some demethylation of other neuroepithelial genes in this GO category was also seen from the array, such as S100P, ROBO1 and PAX6, with significant (p < 0.05) demethylation of S100P in two-thirds of KD cell lines confirmed by pyrosequencing (not shown).
Loss of methylation at other targets including fat homoeostasis/body mass (FBM) genes
Olfactory receptor (OR) genes appeared in a number of GO categories as having lost methylation, though some gains in the gene body were also indicated (Table 1). ORs encode G protein-coupled receptor proteins and are members of a large gene family, many of which are grouped into major clusters, particularly on chromosome 11 . To buffer against stochastic effects due to the large gene family involved, we carried out a second analysis starting instead with sites in promoters showing reliable methylation loss compared to WT (FDR < 0.05) in the triplicates of each KD line and then overlapping these (Fig. 3d) to see which sites were common to all three KD cell lines (Additional file 5: Table S2). Ontology analysis of these common sites using DAVID independently highlighted signalling receptor genes and more particularly olfactory receptors (n = 21). This group of OR genes also showed significant demethylation compared to WT (Kruskal–Wallis, p < 0.05) across the genes when median methylation at all available probes was analysed (Additional file 4: Fig. S3D). We chose three of these genes—OR10J5, OR51E2 and OR2AG1—located on different chromosomes and could verify loss of methylation in all KD lines (Fig. 3e).
The final GO category of genes (GOFMID:0007506) showing loss of methylation (Table 1) consists largely of the TSPY gene family (TSPY1-4, 8 and 10) located on the Y chromosome and thought to be implicated in both normal gonadal development and in gonadoblastoma . These also showed clear evidence of demethylation (Fig. 3f).
Gains in methylation affect the UGT1A locus
A cluster of loci showing gain of methylation on the X chromosome
Transcriptional changes are enriched at cancer/testis antigen genes on X and Y
Regions hypomethylated in shRNA lines correlate with polycomb repression
These data suggested that polycomb-repressed regions might be more susceptible to demethylation than others. To test whether these regions lost methylation more readily than others, we treated hTERT1604 with siRNA for 72 h, which led to acute depletion of the DNMT1 mRNA (Fig. 7c). We found, however, that there was little difference between polycomb-repressed and other regions in terms of demethylation in the siRNA-treated lines (Fig. 7d), in contrast to the shRNA lines where losses were concentrated at the former (Fig. 7d). This could also be seen at the LEP locus, where MGC27345 and RBM28 now showed loss of methylation following siRNA treatment (Fig. 7b, siRNA track). Also of note, almost no probes showed gains in methylation relative to WT in the siRNA cells (Fig. 7e), indicating that this effect is associated exclusively with chronic treatment. These results suggested that gains of methylation had occurred only in shRNA lines and had effectively restored methylation to near WT levels at most regions outside of those marked as polycomb-repressed.
Since transcriptional analysis did not highlight dysregulation of polycomb regions in shRNA cells (Fig. 6d), we tested to see whether polycomb-mediated repression was being maintained there in the absence of DNA methylation. To do this, we treated with DZNep, an inhibitor of EZH2, and confirmed the upregulation of a positive control gene SLCA4 (Fig. 7f) as previously reported . Likewise, HOXC13—a known polycomb target—showed derepression (Fig. 7f). The FBM genes marked by polycomb including LEP showed reactivation to a comparable degree to SLCA4, whereas the MAGEA12 gene which is in a heterochromatic region not marked by polycomb showed little effect (Fig. 7f).
To further investigate the difference between acute and chronic DNMT1 depletion in these cells, we first examined the effects of acute depletion by siRNA on the loci identified in the stable lines: this confirmed that loci such as the clustered protocadherins and the fat/body mass genes also lose methylation on short-term depletion by siRNA (Fig. 7g). Following treatment, cells were then allowed to recover in the absence of siRNA for an extended period (36 days). DNMT1 levels returned to normal rapidly (Fig. 7h). Examination of the methylation response at various gene classes was very instructive. Germline genes (SYCP3, DAZL), which are known to become de novo methylated to high levels during somatic differentiation [5, 34], showed initial loss versus a scrambled control (Scr), followed by remethylation over time to near WT levels (Fig. 7i), confirming that the hTERT cells possess sufficient de novo activity to remethylate the genome, as already suggested (Fig. 7b–e). Imprinted genes are normally unable to regain methylation somatically , and we could confirm that the SNRPN imprint control region failed to remethylate (Fig. 7i). The polycomb-marked genes LEP and PCDHGA2 were also refractory to de novo methylation, either showing no gain (LEP) or reaching a plateau at an intermediate level of recovery only (PCDHGA2) (Fig. 7i).
Gain in methylation is associated with poised promoters in shRNA lines
Further analysis confirmed that while gains in methylation were seen across all the UGT1A exons in all shRNA lines (Fig. 8e), all of these exons showed a loss of methylation following acute depletion with siRNA. We took advantage of our transient depletion and recovery experiment (Fig. 7h, i) to examine levels of methylation at UGT1A4 using pyrosequencing: this showed that while the region indeed loses methylation on acute depletion, it undergoes steady de novo methylation following recovery and at day 36 was the only gene examined whose methylation exceeded that seen in the scrambled control (32.4 vs. 31.3%), suggesting that these genes are indeed susceptible to hypermethylation.
One possible reason for the gains in methylation seen in the shRNA lines could be over-expression of a de novo enzyme. Previous reports have indicated that between them, DNMT3B and DNMT1 account for the majority of methylation in cultured adult human cells and that there may be a role for DNMT3B in maintenance as well as de novo methylation . We saw little change in DNMT3B levels in the DNMT1 KD lines from the HT12 transcriptional array (Additional file 7: Fig. S4A) or RT-PCR (not shown), indicating that gains in methylation are not due to DNMT3B over-expression. To investigate a possible role in maintenance methylation, we carried out a transient siRNA treatment and could achieve robust knockdown of DNMT3B in the cells (Additional file 7: Fig. S4B). While some germline genes showed little effect, loci previously shown to require DNMT3B including D4Z4 and NBL2 did show loss of methylation (Additional file 7: Fig. S4C), confirming that we had achieved a functional depletion. Examination of the loci identified in our DNMT1 shRNA clones showed that these loci also showed loss of methylation in DNMT3B KD cells (Additional file 7: Fig. S4C), suggesting that loci which remain hypomethylated in the shRNA clones also require input from DNMT3B to retain WT methylation levels.
Summary and model
We and others have previously shown that acute depletion of DNMT1 using siRNA triggered the DNA damage response and cell cycle perturbations in human cell lines, making it difficult to identify genes which are directly controlled by methylation. Here we used isogenic shRNA-containing derivatives of a normosomic lung fibroblast cell line to look at the effects of chronic depletion of the protein. We characterised the alterations in methylation and transcription using microarrays in three different cell lines, processing them using a highly reproducible pipeline, and verified changes using locus-specific pyrosequencing or RT-qPCR assays. Additionally, we compared the effects on methylation of this chronic depletion to the effects of acute depletion using siRNA, as well as investigating possible contributions by DNMT3B. Finally, we investigated the correlations between chromatin state and DNA methylation and showed a role for polycomb-mediated repression at some of the loci.
Our results show that while both siRNA and shRNA-treated cells lose methylation overall as would be expected, only the latter show gains in methylation, most likely reflecting selection against the deleterious effects of hypomethylation during clonal expansion and culture. Figure 8e shows what we propose to have occurred: shRNA treatment gave initial widespread demethylation in all three clonal lines, since each line shows the presence of some highly demethylated sites distributed across the genome, but methylation seems to have recovered at most CpGs (Fig. 8e red line). Comparison to normal chromatin patterns in human lung fibroblasts indicated that remaining hypomethylation in the expanded cells was concentrated at regions normally marked for repression by polycomb (Fig. 8e purple line), while the smaller number of regions becoming hypermethylated relative to the parental cell line are associated with poised promoters (green line). TET expression was not detected, and the cells had little or no 5-hydroxymethylation (5hmC; data not shown), in keeping with other reports , suggesting that the hypermethylation does not represent 5hmC. Likewise, no over-expression of DNMT3B was detected.
In terms of what type of gene was particularly affected by chronic DNMT1 KD, the enrichment analyses and laboratory verification consistently pointed at the same small group of gene categories, namely (1)neuroepithelial genes, and in particular the protocadherins; (2) fat homoeostasis/body mass genes; (3) olfactory receptors; (4) the cancer/testis antigens; and (5) the UGT1A complex.
Protocadherins are major targets of DNA methylation in human cells
Emerging evidence suggests that the clustered protocadherin genes may be central to specifying individual neural cell identity [47, 48] and they have been shown to become heavily methylated during embryonic development in mouse , suggesting that stable repression of non-transcribing copies is a programmed event during development. Recent work has shown that DNMT3B is important for de novo methylation at these loci and suggested that dysregulated expression may contribute to the phenotype in immunodeficiency, chromosome abnormalities and facial anomalies (ICF) syndrome , where DNMT3B is frequently mutated , and we found that depletion of DNMT3B was accompanied by loss of methylation at PCDHGA2. The PCDHA and PCDHB loci are heterochromatic and show persistent loss of methylation, as does the 5′ end of the PCDHG locus which is polycomb-repressed, but not the 3′ end which shows little loss of methylation and has instead chromatin marks associated with weak transcription (Additional file 3: Fig. S2B). Meehan and co-workers recently showed that long-term loss of DNA methylation in mouse Dnmt1 −/− ES cells led to spreading of polycomb marks (in particular H3K27me3): their analyses singled out the Pcdh genes, which were heavily methylated in WT but not mutant ESC, as also shown by others . Reddington et al.  also showed an increase in H3K27me3. A similar sequence of events in our human cells would cause an increase in H3K27me3 on PCDH genes and potentially help block remethylation. The sensitivity of the protocadherin cluster to methylation changes may explain why these genes are frequently identified in screens for differentially methylated loci in cancer . The lack of derepression in our stable fibroblast cells is unsurprising here since expression of these genes is restricted to neurons : they are also, with the exception of part of the PCDHG complex, heterochromatic rather than polycomb-repressed and may as such be harder to reactivate.
Fat/body mass genes can be repressed by DNA methylation and polycomb
Currently, there is much interest in the possibility that altered diet, folate status or exposure to environmental toxins may lead to stable changes in the human methylome which particularly affect metabolic processes, as this offers an attractive mechanism by which it may be possible to partly explain the foetal origins of adult disease [55, 56]. Enrichment analysis in our cells identified the FBM genes involved in the common processes of lipid storage and body mass homoeostasis, including LEP, GHSR and the APOC cluster. These loci are readily demethylated on acute DNMT1 depletion and remain demethylated in chronically depleted cells where many other loci have recovered methylation. These loci are heavily marked by polycomb in normal fibroblasts, rather than being heterochromatic, which can potentially explain both their resistance to remethylation and their lack of transcriptional depression in the stable lines. In keeping with this, inhibition of the polycomb repressor EZH2 which generates H3K27me3 marks could reactivate these genes, as well as the canonical polycomb targets the HOX genes. These results suggest that in cells which have both DNA methylation and polycomb-mediated repression, both layers of repression must be removed to achieve gene activation. Interestingly a recent report by Hajkova and colleagues showed that reprogramming of germ cells in mouse also required both removal of DNA methylation and alteration of polycomb marks .
Olfactory genes are methylated and largely inert
Olfactory receptors are also involved in specification of neural cell identity, where individual receptors are expressed in only a small group of cells in the olfactory epithelium . They are largely monoallelically expressed, and methylation has been implicated as playing a role in their control [59, 60]. The OR gene family is the largest in the genome, with approx. 380 active members, many organised into “gene factories” where they are flanked by many more pseudogenes and repeats, such as the large cluster on chr11 . These regions are often transcriptionally inert and heterochromatic, which together with the requirement for tissue-specific factors may explain their lack of derepression.
Cancer/testis antigen genes are particular targets for demethylation and activation
The TSPY and MAGE genes fall into a functionally defined group known as the cancer/testis antigen (CTA) genes ([61, 62]; http://www.cta.lncc.br/) which are expressed during testis development normally, but which are aberrantly expressed in some tumours, such as melanoma and gonadoblastoma (e.g. TSPY2). This latter property makes them of particular interest for cancer immunotherapy, and monoclonal antibodies against some CTA members have already gained clinical approval . CTA genes have been shown previously to lose methylation and become derepressed in several cancer cell types after treatment with the methyltransferase inhibitor 5′aza-2-deoxycytidine (Aza) [64–66] and in the HCT116 DNMT1 mutant line [66, 67] using locus-specific approaches. Our study (1) shows in an unbiased genomic screen that CTA genes are the genes most affected by loss of maintenance activity, (2) shows this for the first time in a normal, differentiated cell line and (3) highlights the subset of CTA genes which are particularly dependent on maintenance activity to keep them repressed. It is noteworthy that the majority of these genes are on the X chromosome, which shows major fluxes in methylation in our stable lines. The genes are largely associated with heterochromatin, rather than polycomb repression, and do not respond to EZH2 inhibition, but rather directly to loss of methylation, which may reflect some difference in heterochromatin marking on the X. Strategies to demethylate and turn on these genes in tumour cells (e.g. with Aza) to facilitate cancer vaccine development may be worthwhile to pursue, given that these genes are the most responsive to loss of methylation in our cell lines.
UGT1A genes and other poised promoters are susceptible to hypermethylation
From the enrichment analysis, the UGT1A gene cluster was highlighted in terms of genes gaining methylation. These genes are known to be highly expressed in skin fibroblasts postnatally, and to be repressed in non-expressing tissues by methylation [68, 69]. The WT cells already had substantial levels of methylation but the increased methylation in the stable cell lines led to small but consistent decreases in transcription on the HT12 array, though levels were so low these could not be confirmed by Taqman qPCR (data not shown). It may be that the particular marks associated with a recent inactivation of the UGT1A cluster in the fibroblasts during adaptation to cell culture led to an increased de novo activity here, and in our transient KD experiment we saw the greatest gains in methylation at UGT1A4. Consistent with this, hypermethylation relative to the WT cells was associated with weak and poised promoters genome-wide, and the latter showed the greatest tendency to gain methylation above normal WT levels in the shRNA-containing lines.
Lack of transcriptional changes in part due to polycomb
It is notable that while there was widespread changes in methylation in the KD cell lines, this was not accompanied by large-scale transcriptional derepression, with only a few hundred genes showing dysregulation, and the fold change in transcription being small. Of the four gene classes identified as most affected in terms of methylation, only one—that containing the TSPY and MAGE genes—showed robust transcriptional derepression. A lack of global changes in transcription, also reported by others [29, 70], is likely due to in part to the absence of transcription factors in fibroblasts needed to transcribe neural or adipocyte genes at high levels. However, many of the regions showing most persistent hypomethylation are polycomb-marked and this is likely to be sufficient in itself, as it is for example in Drosophila, to maintain repression of these genes. However, we could show that in the presence of an EZH2 inhibitor, polycomb-marked loci which lacked DNA methylation, such as those involved in fat homoeostasis/body mass regulation, became upregulated, along with canonical polycomb targets such as the HOX genes. Our results therefore indicate both that the polycomb system is sufficient in itself to repress and also that polycomb-repressed regions appear to be refractive to remethylation, which may be due to the action of FBXL10 . It has previously been proposed that the two systems work in parallel, with their own sets of targets and a degree of mutual exclusivity [15–17]: our results would support such a conclusion.
Comparison to other recent work
Two recent studies have also examined the effects of DNMT1 mutation on DNA methylation and gene transcription in human, albeit in cancer cells [29, 70]. Acute depletion of DNMT1 using an siRNA-mediated approach found, as we did, regions of low CpG density (open sea, etc.) to be most affected, but differed in finding more evidence for cell morphogenesis and phosphorylation pathways being affected . This might reflect differences between acute and chronic depletion and the high levels of cell death during acute depletion. Blattler and colleagues  also found that relatively few genes were dysregulated in DNMT1/3B double KO HCT116 cells, but some cancer/testis genes (the related GAGE genes) were upregulated, along with Krüppel-associated box genes, while chaperonins figured prominently among down-regulated genes. The latter two gene classes may therefore be more dependent on DNMT3B, or the combination of DNMT1 and 3B, for their maintenance; alternatively the differences may be due to the experiment being carried out in colon cancer cells rather than, as here, in non-transformed fibroblasts.
In conclusion, our study sheds new light on the loci which are most sensitive to sustained loss of maintenance activity in humans and shows an interplay between polycomb and DNA methylation-mediated repression in these differentiated cells.
KON and REI supervised and carried out the majority of wet laboratory work and assembled figures; SJM and SJT carried out the majority of the bioinformatics analyses; AT carried out the DNMT3B work; CB and LM contributed results on specific loci; JL carried out initial KD experiments; DGM supervised and carried out bioinformatics analyses; CPW designed the experiments, carried out bioinformatics analyses, interpreted results and wrote the MS. All authors read and approved the final manuscript.
We thank Julien Bauer, Daniel Harkin and Steven McLoughlin for help with bioinformatics analysis, Andrew Irwin, Lee McCahon, Anna McLaughlin and Bob Goodman for technical help, Paul Thompson for advice and other members of the laboratory for critical comments. We are obliged to Bert Vogelstein for the HCT116 WT and DKO cells.
The authors declare that they have no competing interests.
Availability of data and materials
Data from the 450K and HT-12 arrays have been deposited with the Gene Expression Omnibus database at the National Centre for Biotechnology Information, USA, under the Series number GSE90012. Supplementary Figures and Tables are available in the online version. Cell lines or other materials are available from the corresponding author on request.
Ethics approval and consent to participate
Work in the Walsh laboratory was funded by grants from the Medical Research Council (MR/J007773/1) and the ESRC/BBSRC (ES/N000323/1).
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Open AccessThis article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.
- Smith ZD, Meissner A. DNA methylation: roles in mammalian development. Nat Rev. 2013;14:204–20.View ArticleGoogle Scholar
- Edwards JR, Yarychkivska O, Boulard M, Bestor TH. DNA methylation and DNA methyltransferases. Epigenetics Chromatin. 2017;10:23. https://doi.org/10.1186/s13072-017-0130-8 View ArticlePubMedPubMed CentralGoogle Scholar
- Li E, Beard C, Jaenisch R. Role for DNA methylation in genomic imprinting. Nature. 1993;366:362–5.View ArticlePubMedGoogle Scholar
- Beard C, Li E, Jaenisch R. Loss of methylation activates Xist in somatic but not in embryonic cells. Genes Dev. 1995;9:2325–34.View ArticlePubMedGoogle Scholar
- Weber M, Hellmann I, Stadler MB, Ramos L, Paabo S, Rebhan M, et al. Distribution, silencing potential and evolutionary impact of promoter DNA methylation in the human genome. Nat Genet. 2007;39:457–66.View ArticlePubMedGoogle Scholar
- Krebs AR, Dessus-Babus S, Burger L, Schübeler D. High-throughput engineering of a mammalian genome reveals building principles of methylation states at CG rich regions. Elife [Internet]. 2014;3:e04094.Google Scholar
- Wachter E, Quante T, Merusi C, Arczewska A, Stewart F, Webb S, et al. Synthetic CpG islands reveal DNA sequence determinants of chromatin structure. Elife [Internet]. 2014;3:1–16.Google Scholar
- Walsh CP, Bestor TH. Cytosine methylation and mammalian development. Genes Dev [Internet]. 1999;13:26–34.View ArticleGoogle Scholar
- Irizarry RA, Ladd-Acosta C, Wen B, Wu Z, Montano C, Onyango P, et al. The human colon cancer methylome shows similar hypo- and hypermethylation at conserved tissue-specific CpG island shores. Nat Genet [Internet]. 2009;41:178–86.View ArticleGoogle Scholar
- Jones PA. Functions of DNA methylation: islands, start sites, gene bodies and beyond. Nat Rev Genet [Internet]. 2012;13:484–92.View ArticleGoogle Scholar
- Neri F, Krepelova A, Incarnato D, Maldotti M, Parlato C, Galvagni F, et al. Dnmt3L antagonizes DNA methylation at bivalent promoters and favors DNA methylation at gene bodies in ESCs. Cell. 2013;155:121–34.View ArticlePubMedGoogle Scholar
- Wu H, Coskun V, Tao J, Xie W, Ge W, Yoshikawa K, et al. Dnmt3a-dependent nonpromoter DNA methylation facilitates transcription of neurogenic genes. Science. 2010;329:444–8.View ArticlePubMedPubMed CentralGoogle Scholar
- Irwin RE, Thakur A, O’ Neill KM, Walsh CP. 5-Hydroxymethylation marks a class of neuronal gene regulated by intragenic methylcytosine levels. Genomics [Internet]. 2014;104:383–92.View ArticleGoogle Scholar
- Margueron R, Reinberg D. The Polycomb complex PRC2 and its mark in life. Nature [Internet]. 2011;469:343–9.View ArticleGoogle Scholar
- Lynch MD, Smith AJH, De Gobbi M, Flenley M, Hughes JR, Vernimmen D, et al. An interspecies analysis reveals a key role for unmethylated CpG dinucleotides in vertebrate Polycomb complex recruitment. EMBO J [Internet]. 2012;31:317–29.View ArticleGoogle Scholar
- Weinhofer I, Hehenberger E, Roszak P, Hennig L, Köhler C. H3K27me3 profiling of the endosperm implies exclusion of polycomb group protein targeting by DNA methylation. PLoS Genet. 2010;6:1–14.View ArticleGoogle Scholar
- Reddington JP, Perricone SM, Nestor CE, Reichmann J, Youngson NA, Suzuki M, et al. Redistribution of H3K27me3 upon DNA hypomethylation results in de-repression of Polycomb target genes. Genome Biol. 2013;14:R25.View ArticlePubMedPubMed CentralGoogle Scholar
- Goll MG, Bestor TH. Eukaryotic cytosine methyltransferases. Annu Rev Biochem [Internet]. 2005;74:481–514.View ArticleGoogle Scholar
- Lei H, Oh SP, Okano M, Juttermann R, Goss KA, Jaenisch R, et al. De novo DNA cytosine methyltransferase activities in mouse embryonic stem cells. Development. 1996;122:3195–205.PubMedGoogle Scholar
- Li E, Bestor TH, Jaenisch R. Targeted mutation of the DNA methyltransferase gene results in embryonic lethality. Cell. 1992;69:915–26.View ArticlePubMedGoogle Scholar
- Okano M, Xie S, Li E. Cloning and characterization of a family of novel mammalian DNA (cytosine-5) methyltransferases. Nat Genet. 1998;19:219–20.View ArticlePubMedGoogle Scholar
- Okano M, Bell DW, Haber DA, Li E. DNA methyltransferases Dnmt3a and Dnmt3b are essential for de novo methylation and mammalian development. Cell. 1999;99:247–57.View ArticlePubMedGoogle Scholar
- Jackson-Grusby L, Beard C, Possemato R, Tudor M, Fambrough D, Csankovszki G, et al. Loss of genomic methylation causes p53-dependent apoptosis and epigenetic deregulation. Nat Genet. 2001;27:31–9.View ArticlePubMedGoogle Scholar
- Liao J, Karnik R, Gu H, Ziller MJ, Clement K, Tsankov AM, et al. Targeted disruption of DNMT1, DNMT3A and DNMT3B in human embryonic stem cells. Nat Genet. 2015;47:469–78.View ArticlePubMedPubMed CentralGoogle Scholar
- Chen T, Hevi S, Gay F, Tsujimoto N, He T, Zhang B, et al. Complete inactivation of DNMT1 leads to mitotic catastrophe in human cancer cells. Nat Genet. 2007;39:391–6.View ArticlePubMedGoogle Scholar
- Rhee I, Jair KW, Yen RW, Lengauer C, Herman JG, Kinzler KW, et al. CpG methylation is maintained in human cancer cells lacking DNMT1. Nature. 2000;404:1003–7.View ArticlePubMedGoogle Scholar
- Rhee I, Bachman KE, Park BH, Jair KW, Yen RW, Schuebel KE, et al. DNMT1 and DNMT3b cooperate to silence genes in human cancer cells. Nature. 2002;416:552–6.View ArticlePubMedGoogle Scholar
- Egger G, Jeong S, Escobar SG, Cortez CC, Li TW, Saito Y, et al. Identification of DNMT1 (DNA methyltransferase 1) hypomorphs in somatic knockouts suggests an essential role for DNMT1 in cell survival. Proc Natl Acad Sci USA. 2006;103:14080–5.View ArticlePubMedPubMed CentralGoogle Scholar
- Blattler A, Yao L, Witt H, Guo Y, Nicolet CM, Berman BP, et al. Global loss of DNA methylation uncovers intronic enhancers in genes showing expression changes. Genome Biol [Internet]. 2014;15:469.View ArticleGoogle Scholar
- Loughery JE, Dunne PD, O’Neill KM, Meehan RR, McDaid JR, Walsh CP. DNMT1 deficiency triggers mismatch repair defects in human cells through depletion of repair protein levels in a process involving the DNA damage response. Hum Mol Genet. 2011;20:3241–55.View ArticlePubMedGoogle Scholar
- Bibikova M, Barnes B, Tsan C, Ho V, Klotzle B, Le JM, et al. High density DNA methylation array with single CpG site resolution. Genomics. 2011;98:288–95.View ArticlePubMedGoogle Scholar
- Assenov Y, Muller F, Lutsik P, Walter J, Lengauer T, Bock C. Comprehensive analysis of DNA methylation data with RnBeads. Nat Methods. 2014;11:1138–40.View ArticlePubMedPubMed CentralGoogle Scholar
- Ouellette MM, McDaniel LD, Wright WE, Shay JW, Schultz RA. The establishment of telomerase-immortalized cell lines representing human chromosome instability syndromes. Hum Mol Genet. 2000;9:403–11.View ArticlePubMedGoogle Scholar
- Rutledge CE, Thakur A, O’Neill KM, Irwin RE, Sato S, Hata K, et al. Ontogeny, conservation and functional significance of maternally inherited DNA methylation at two classes of non-imprinted genes. Development. 2014;141:1313–23.View ArticlePubMedPubMed CentralGoogle Scholar
- Aryee MJ, Jaffe AE, Corrada-Bravo H, Ladd-Acosta C, Feinberg AP, Hansen KD, et al. Minfi: a flexible and comprehensive Bioconductor package for the analysis of Infinium DNA methylation microarrays. Bioinformatics. 2014;30:1363–9.View ArticlePubMedPubMed CentralGoogle Scholar
- Carbon S, Ireland A, Mungall CJ, Shu S, Marshall B, Lewis S, et al. AmiGO: online access to ontology and annotation data. Bioinformatics. 2009;25:288–9.View ArticlePubMedGoogle Scholar
- Huang DW, Sherman BT, Lempicki RA. Systematic and integrative analysis of large gene lists using DAVID bioinformatics resources. Nat Protocol. 2009;4:44–57.View ArticleGoogle Scholar
- Giardine B, Riemer C, Hardison RC, Burhans R, Elnitski L, Shah P, et al. Galaxy: a platform for interactive large-scale genome analysis. Genome Res. 2005;15:1451–5.View ArticlePubMedPubMed CentralGoogle Scholar
- Karolchik D, Baertsch R, Diekhans M, Furey TS, Hinrichs A, Lu YT, et al. The UCSC genome browser database. Nucleic Acids Res. 2003;31:51–4.View ArticlePubMedPubMed CentralGoogle Scholar
- Kondo T, Bobek MP, Kuick R, Lamb B, Zhu X, Narayan A, et al. Whole-genome methylation scan in ICF syndrome: hypomethylation of non-satellite DNA repeats D4Z4 and NBL2. Hum Mol Genet. 2000;9:597–604.View ArticlePubMedGoogle Scholar
- Glusman G, Yanai I, Rubin I, Lancet D. The complete human olfactory subgenome. Genome Res. 2001;11:685–702.View ArticlePubMedGoogle Scholar
- Lau YFC, Li Y, Kido T. Gonadoblastoma locus and the TSPY gene on the human y chromosome. Birth Defects Res Part C Embryo Today Rev. 2009;87:114–22.View ArticleGoogle Scholar
- Ernst J, Kheradpour P, Mikkelsen TS, Shoresh N, Ward LD, Epstein CB, et al. Mapping and analysis of chromatin state dynamics in nine human cell types. Nature [Internet]. 2011;473:43–9.View ArticleGoogle Scholar
- Fujiwara T, Saitoh H, Inoue A, Kobayashi M, Okitsu Y, Katsuoka Y, et al. 3-Deazaneplanocin A (DZNep), an inhibitor of S-adenosylmethionine-dependent methyltransferase, promotes erythroid differentiation. J Biol Chem. 2014;289:8121–34.View ArticlePubMedPubMed CentralGoogle Scholar
- Tucker KL, Beard C, Dausmann J, Jackson-Grusby L, Laird PW, Lei H, et al. Germ-line passage is required for establishment of methylation and expression patterns of imprinted but not of nonimprinted genes. Genes Dev. 1996;10:1008–20.View ArticlePubMedGoogle Scholar
- Nestor CE, Ottaviano R, Reddington J, Sproul D, Reinhardt D, Dunican D, et al. Tissue type is a major modifier of the 5-hydroxymethylcytosine content of human genes. Genome Res. 2012;22:467–77.View ArticlePubMedPubMed CentralGoogle Scholar
- Yagi T. Molecular codes for neuronal individuality and cell assembly in the brain. Front Mol Neurosci [Internet]. 2012;5:45.Google Scholar
- Rubinstein R, Thu CA, Goodman KM, Wolcott HN, Bahna F, Mannepalli S, et al. Molecular logic of neuronal self-recognition through protocadherin domain interactions. Cell. 2015;163:629–42.View ArticlePubMedPubMed CentralGoogle Scholar
- Borgel J, Guibert S, Li Y, Chiba H, Schübeler D, Sasaki H, et al. Targets and dynamics of promoter DNA methylation during early mouse development. Nat Genet [Internet]. 2010;42:1093–100.View ArticleGoogle Scholar
- Toyoda S, Kawaguchi M, Kobayashi T, Tarusawa E, Toyama T, Okano M, et al. Developmental epigenetic modification regulates stochastic expression of clustered Protocadherin genes, generating single neuron diversity. Neuron. 2014;82:94–108.View ArticlePubMedGoogle Scholar
- Xu GL, Bestor TH, Bourc’his D, Hsieh CL, Tommerup N, Bugge M, et al. Chromosome instability and immunodeficiency syndrome caused by mutations in a DNA methyltransferase gene. Nature. 1999;402:187–91.View ArticlePubMedGoogle Scholar
- Otani J, Kimura H, Sharif J, Endo TA, Mishima Y, Kawakami T, et al. Cell cycle-dependent turnover of 5-hydroxymethyl cytosine in mouse embryonic stem cells. PLoS ONE. 2013;8:e82961.View ArticlePubMedPubMed CentralGoogle Scholar
- van Roy F. Beyond E-cadherin: roles of other cadherin superfamily members in cancer. Nat Rev Cancer [Internet]. 2014;14:121–34.View ArticleGoogle Scholar
- Chen WV, Maniatis T. Clustered protocadherins. Development [Internet]. 2013;140:3297–302.View ArticleGoogle Scholar
- Barker DJP. The developmental origins of chronic adult disease. Acta Paediatr Suppl. 2004;93:26–33.View ArticlePubMedGoogle Scholar
- Irwin RE, Pentieva K, Cassidy T, Lees-Murdock DJ, McLaughlin M, Prasad G, et al. The interplay between DNA methylation, folate and neurocognitive development. Epigenomics [Internet]. 2016;8:863.View ArticleGoogle Scholar
- Hill PWS, Leitch HG, Requena CE, Sun Z, Amouroux R, Roman-Trufero M, et al. Epigenetic reprogramming enables the transition from primordial germ cell to gonocyte. Nature [Internet]. 2018;555:392–6.View ArticleGoogle Scholar
- McClintock TS. Achieving singularity in mammalian odorant receptor gene choice. Chem Senses. 2010;35:447–57.View ArticlePubMedPubMed CentralGoogle Scholar
- MacDonald JL, Gin CSY, Roskams AJ. Stage-specific induction of DNA methyltransferases in olfactory receptor neuron development. Dev Biol. 2005;288:461–73.View ArticlePubMedGoogle Scholar
- Colquitt BM, Markenscoff-Papadimitriou E, Duffie R, Lomvardas S. Dnmt3a regulates global gene expression in olfactory sensory neurons and enables odorant-induced transcription. Neuron. 2014;83:823–38.View ArticlePubMedPubMed CentralGoogle Scholar
- Simpson AJG, Caballero OL, Jungbluth A, Chen Y-T, Old LJ. Cancer/testis antigens, gametogenesis and cancer. Nat Rev Cancer. 2005;5:615–25.View ArticlePubMedGoogle Scholar
- Almeida LG, Sakabe NJ, de Oliveira AR, Silva MCC, Mundstein AS, Cohen T, et al. CTdatabase: a knowledge-base of high-throughput and curated data on cancer-testis antigens. Nucleic Acids Res. 2009;37:D816.View ArticlePubMedGoogle Scholar
- Gjerstorff MF, Andersen MH, Ditzel HJ. Oncogenic cancer/testis antigens: prime candidates for immunotherapy. Oncotarget. 2015;6:15772–87.View ArticlePubMedPubMed CentralGoogle Scholar
- Samlowski WE, Leachman SA, Wade M, Cassidy P, Porter-Gill P, Busby L, et al. Evaluation of a 7-day continuous intravenous infusion of decitabine: inhibition of promoter-specific and global genomic DNA methylation. J Clin Oncol. 2005;23:3897–905.View ArticlePubMedGoogle Scholar
- Karpf AR. A potential role for epigenetic modulatory drugs in the enhancement of cancer/germ-line antigen vaccine efficacy. Epigenetics. 2006;1:116–20.View ArticlePubMedPubMed CentralGoogle Scholar
- Koslowski M, Bell C, Seitz G, Lehr HA, Roemer K, Müntefering H, et al. Frequent nonrandom activation of germ-line genes in human cancer. Cancer Res. 2004;64:5988–93.View ArticlePubMedGoogle Scholar
- James SR, Link PA, Karpf AR. Epigenetic regulation of X-linked cancer/germline antigen genes by DNMT1 and DNMT3b. Oncogene. 2006;25:6975–85.View ArticlePubMedGoogle Scholar
- Belanger AS, Tojcic J, Harvey M, Guillemette C. Regulation of UGT1A1 and HNF1 transcription factor gene expression by DNA methylation in colon cancer cells. BMC Mol Biol. 2010;11:9.View ArticlePubMedPubMed CentralGoogle Scholar
- Sumida K, Kawana M, Kouno E, Itoh T, Takano S, Narawa T, et al. Importance of UDP-glucuronosyltransferase 1A1 expression in skin and its induction by UVB in neonatal hyperbilirubinemia. Mol Pharmacol. 2013;84:679–86.View ArticlePubMedPubMed CentralGoogle Scholar
- Tiedemann RL, Putiri EL, Lee J-H, Hlady RA, Kashiwagi K, Ordog T, et al. Acute depletion redefines the division of labor among DNA methyltransferases in methylating the human genome. Cell Rep [Internet]. 2014;9:1554–66.View ArticleGoogle Scholar
- Boulard M, Edwards JR, Bestor TH (2015) FBXL10 protects polycomb-bound genes from hypermethylation. Nat Genet 47(5):479–85.View ArticlePubMedGoogle Scholar