Abstract
The DNA methylation status of the X-chromosome in cancer cells is often overlooked because of computational difficulties. Most of the CpG islands on the X-chromosome are mono-allelically methylated in normal female cells and only present as a single copy in male cells. We treated two colorectal cancer cell lines from a male (HCT116) and a female (RKO) with increasing doses of a DNA methyltransferase 1 (DNMT1)-specific inhibitor (GSK3685032/GSK5032) over several months to remove as much non-essential CpG methylation as possible. Profiling of the remaining DNA methylome revealed an unexpected, enriched retention of DNA methylation on the X-chromosome. Strikingly, the identified retained X-chromosome DNA methylation patterns accurately predicted de novo DNA hypermethylation in colon cancer patient methylomes in the TCGA COAD/READ cohort. These results suggest that a re-examination of tumors for X-linked DNA methylation changes may enable greater understanding of the importance of epigenetic silencing of cancer related genes.
Similar content being viewed by others
Introduction
The methylation of CpG dinucleotides has been known to be essential for mammalian development for many years1. In addition, alterations in methylation patterns, particularly the de novo methylation of CpG islands, are a feature of almost all human cancers2,3. While some proportion of the methylation in both normal and cancer cells is necessary for cellular viability, much of it seems to be dispensable and plays no known function. Cells in culture can exist and divide with drastically reduced methylation levels4. Distinguishing which sites require methylation for cell survival is difficult without genetic or pharmacological interventions which have limitations in the clarity of the data they provide. We previously used colon cancer cell lines obtained from the Vogelstein and Baylin laboratories5,6, which have genetic knockdowns and/or knockouts of two of the three known DNA methyltransferases (DNMTs), DNMT1 and DNMT3B, to discover driver methylation sites for viability7. These experiments identified several key genes which required silencing by promoter methylation for survival, in that their re-expression was associated with cell death.
The three known DNMTs, which all show preference for CpG sites, are expressed as multiple isoforms and seem to have different but overlapping functions in the cell8,9. DNMT1 performs a vital role in copying methylation patterns by converting hemi-methylated sites to full methylation following DNA replication, and also shows de novo activity10. DNMT3A1 and 3A2 are considered to be de novo enzymes in that they can apply methyl groups to either unmethylated or hemimethylated CpGs and do not have known preferences for particular genomic locations. DNMT3B on the other hand, is expressed in multiple catalytically active and inactive splice variants and is required for efficient methylation of gene bodies11,12,13. The catalytically inactive isoform (DNMT3B3) can also act as an accessory protein to bind to the nucleosomal acidic patch and anchor DNMT3A2 to nucleosomes14. Deciphering the relative contributions of these enzymatic functions to the structure and function of the epigenome has been difficult and has not been successful at a global level.
In the current experiments, we used a DNMT inhibitor, GSK3685032 (abbreviated GSK5032) which is specific for the so-called maintenance enzyme DNMT1, to obtain cells which were resistant to continuous exposure to increasing concentrations of the compound. The residual patterns of methylation, which presumably are mostly due to the activities of DNMT3A and DNMT3B isoforms, showed an unexpected dependency on chromosome location. In particular, promoter probes on the active X chromosome which had become de novo methylated in the cancer cells were preferentially resistant to DNMT1 inhibition compared to other chromosomes. Surprisingly, several of these probes were associated with genes strongly expressed in normal colonic tissue yet were frequently methylated in colorectal and other tumors in both males and females. The differential retention of DNA methylation on different chromosomes suggests that DNMT3A/3B isoforms may have chromosome-specific functions. Additionally, we have uncovered several new, to the best of our knowledge, cancer related genes which have been overlooked because of their location on the X chromosome.
Results
DNMT1 inhibition or knockdown cause genome wide demethylation
Our previous work demonstrated a global loss of DNA methylation with a very small proportion of promoter CpGs maintaining methylation in the absence of DNMT1 and DNMT3B7. Importantly, the DNA methylation that was maintained was associated with silencing of key genes whose repression is required for cancer cell survival in cell culture. However, our analysis of retained DNA methylation in this study was restricted to promoter CpGs due to the available technology at the time (Infinium 27k). Additionally, the two clones with concurrent knockdown of DNMT1 and knockout of DNMT3B used in the original study, HCT116 DKO8 and DKO1, demonstrated different degrees of DNA methylation loss and retention. In the present study, we aimed to expand on our initial work by profiling DNA methylation across the genome (Infinium MethylationEPIC array) and generating an additional stable cell model of severely impaired DNMT function in order to identify additional DNA methylation patterns that are essential for cancer cell survival. To generate such a model, we treated two colorectal cancer cell lines with increasing concentrations of GSK5032, a DNMT1 specific inhibitor, over several months. We first determined the IC50s for the HCT116 and RKO colorectal cancer cell lines before (parental lines) and after long term exposure (GSK5032-resistant lines) to increasing concentrations of the GSK5032 inhibitor (Fig. 1a). The parental cells, which were derived from a male and female respectively, showed IC50s of between 50 and 135 nM and both cell types became markedly resistant to the drug as shown by the IC50s increasing almost 1000-fold to over 30,000 nM for HCT116 and 13,000 nM for RKO. The doubling times of the treated (GSK5032-resistant) and HCT116 DKO1 cells, which have a hypomorphic version of the DNA methyltransferase 1 (DNMT1) and a complete knockout of DNMT3B4,6, were markedly increased (Supplementary Fig. 1a). The levels of RNA expression of several DNA methyltransferase isoforms showed minor changes as a result of drug treatment, with the exception of increased expression of DNMT3A2 in treated or knockdown cells (Supplementary Fig. 1b). The data therefore show that the resistance is not due to a strong upregulation of DNMT1 or the other two methyltransferases.
Next, we used Infinium MethylationEPIC arrays to analyze the patterns of DNA methylation represented as beta values (β-value), which are defined as the ratio of methylated probe intensity to the overall intensity. HCT116 parental cells showed a strong bimodal distribution of β-values in which the majority of CpG probes queried are concentrated in either the unmethylated density peak (β-value \(\cong\) 0.0) or the fully methylated density peak (β-value \(\cong\) 1.0) (Fig. 1b, left). Conversely, HCT116 DKO1 and GSK5032-resistant cells showed no density of CpG probes that were fully methylated, but rather an enriched density of CpG probes that were unmethylated (Fig. 1b). Interestingly, the drug treated cells (HCT116 GSK5032-resistant) showed a shoulder of intermediate levels of methylation which was not present in the HCT116 DKO1 cells. Since the HCT116 GSK5032-resistant cells, unlike the HCT116 DKO1 cells, express DNMT3B it is possible that the shoulder is due to targeting of these CpGs by DNMT3B. The inhibitor also strongly reduced the methylation of highly methylated probes in RKO cells (Fig. 1b, right). The average β-values of all probes in both the HCT116 and RKO parental cell lines were reduced from about 0.66 to 0.29 and 0.14 respectively following long-term treatment (GSK5032-resistant) and 0.16 in HCT116 DKO1 cells (Fig. 1c).
Probes on the X Chromosome are preferentially resistant to demethylation
Our next goal was to determine the extent of retained DNA methylation in our globally demethylated cancer cell culture models. First, we profiled the β-value of each individual CpG probe broken down by chromosome location for HCT116 (Parental, DKO1, and GSK5032-resistant) and RKO (Parental and GSK5032-resistant) cell line populations (Fig. 2a (Chr7,9,20,X); Supplementary Fig. 2 (all chromosomes)). The genetic knockdown (HCT116 DKO1) or pharmacological inhibition of DNMT1 (GSK5032-resistant) showed the extensive loss of DNA methylation seen in Fig. 1b, c, but a minority of CpG probes retained some DNA methylation (probes that demonstrate warmer heatmap colors) even after exhaustive drug treatment. Surprisingly, the X-chromosome showed more CpG probes that retained high levels of DNA methylation (as evident by increased warmer heatmap colors in the HCT116 DKO1 and both GSK5032-resistant lines) in comparison to the autosomal chromosomes (Fig. 2a and Supplementary Fig. 2). Similar but distinct patterns of CpG methylation in parental HCT116 or RKO cells were apparent.
To further dissect the distribution and location of retained DNA methylation in our DNMT disrupted cell lines (HCT116 DKO1, HCT116 and RKO GSK5032-resistant), we first used three different β-value cut-offs ranging from 0.6 to 0.85 to categorize highly methylated CpGs that retained DNA methylation (Fig. 2b). Next, we calculated the percentage of highly methylated CpG probes that retained DNA methylation on each chromosome in the cell lines with different β-value cutoffs for categorizing highly methylated probes. Finally, the data for each chromosome were ranked by increasing percentages in HCT116 GSK5032-resistant cells at a β-value of 0.7. Individual chromosomes differed markedly in their abilities to retain a high level of methylation. The patterns of retention were very similar under all conditions tested and showed a strong enrichment on the X chromosome in all demethylated cell lines (HCT116 DKO1, HCT116 GSK5032-resistant, and RKO GSK5032-resistant). Interestingly, the retention patterns on individual autosomes were similar in the three cell lines with chromosomes 19, 22, 17, and 16 being less likely to retain DNA methylation in all cases.
To ensure that the observed enrichment of retained DNA methylation on the X chromosome was not due to an inherent bias in the EPIC array composition, we analyzed our data in a number of ways. The probes on the array are distributed much as to be expected based on chromosome size increasing by their numerical designation (Supplementary Fig. 3a). The distribution of probes on the most demethylated chromosome (19) and the most resistant chromosome (X) were very similar with respect to their genomic annotations (Supplementary Fig. 3b). Therefore, the marked differences in retention of DNA methylation were not due to a skewed probe distribution. Also, these observations were not due to different basal levels of methylation of chromosomes in the parental cells, which showed a uniform percentage of probes methylated (β-value ≥ 0.7) in untreated cells, which was around 60% of probes for each chromosome (Supplementary Fig. 3c, d). Finally, we calculated the enrichment bias for DNA methylation retention on individual chromosomes (compared to the number of methylated probes (β-value \(\ge\) 0.7) on each individual chromosome in the respective parental cell lines) and performed hypergeometric testing to determine the level of significance of the enrichment (Fig. 2c). While chromosomes 6, 3, and 4 also demonstrated significant enrichment of retained DNA methylation, the X chromosome consistently had the strongest enrichment score across all queried cell populations (HCT116 DKO1, GSK5032-resistant; RKO GSK5032-resistant) (Fig. 2c). Consistent with the lack of DNA methylation retention observed in our previous analysis (Fig. 2b), chromosomes 19, 22, 17, and 16 demonstrated negative enrichment scores indicating that they are selected against for retention of DNA methylation (Fig. 2c). Indeed, the fact that the same results were obtained in two cell lines treated with the DNMT1 inhibitor (HCT116 and RKO GSK5032-resistant) and in one cell line with a partial DNMT1 knockout (HCT116 DKO1) suggests that particular methyltransferases may differentially act on the various chromosomes.
CpGs that retain DNA methylation on the X-chromosome in the HCT116 GSK5032-resistant cells are either unmethylated or methylated in non-cancer human tissues
To simplify further investigation, we narrowed our analysis to the HCT116 cells before (parental) and after long-term DNMT1 inhibition (GSK5032-resistant) and selected probes resistant to demethylation (β-value ≥ 0.7) which were located in the vicinity of promoters. On a genome scale, we found that 112 of them (9%) had become de novo methylated in HCT116 parental cells (i.e., unmethylated in non-cancer tissues) whereas 1145 were already methylated in non-cancer uncultured human tissues (Fig. 3a). When we further narrowed our inquiry to the X chromosome, it was immediately apparent that a greater percentage (26%) of the resistant probes had acquired methylation de novo in HCT116 parental cells as these probes demonstrate typical methylation patterns of female X chromosome inactivation (unmethylated in males, monoallelic methylation (50%) in females) in non-cancer human tissue (Fig. 3b). Because the X chromosome probes are included in the analysis in Fig. 3a, the actual enrichment is 28% in that 31 of a total of 112 demethylation-resistant probes are located on the active X chromosome present in the HCT116 cell line. Moving forward, we continued to characterize the probes that retained DNA methylation in the HCT116 GSK5032-resistant cells based on their methylation status in non-cancer human tissues.
Next, we extended our analysis to examine the DNA methylation environment flanking the highly resistant CpG sites we had identified (Fig. 3c–f and Supplementary Fig. 4a). Two resistant probes located in the multiple transcription start sites (TSSs) of XIST were methylated in both female and male non-cancer tissues and in the HCT116 parental cells (Fig. 3c). Since the XIST gene is transcribed on the inactive but not the active X chromosome15, the functions of these CpGs, although highly resistant to demethylation, have unknown significance. On the other hand, the 4 probes downstream showed the expected behavior for an X-linked gene being monoallelically methylated in normal female cells and methylated completely in male and HCT116 parental cells. This region became less methylated after drug treatment, but the retained methylation may not be sufficient for silencing given that the treated cells do express XIST (see Table 1). Figure 3d shows regional analysis of a resistant probe in the region of the TSS of PNMA5 demonstrating that retention of methylation was quite focal. The methylation patterns in non-cancer male and female cells were very similar suggesting that this gene is not subject to X inactivation.
Two examples of CpG probes resistant to drug induced demethylation in the HCT116 parental cells (ARMCX2 and MAGEH1), showed the behavior expected for X-linked genes in non-cancer tissues (Fig. 3e, f). They, and other adjacent probes, were roughly 50% methylated in non-cancer female tissues and completely unmethylated in males. On the other hand, they were heavily methylated in the HCT116 parental line and maintained this DNA methylation after long-term DNMT1 inhibition (GSK5032-resistant). As the EPIC array probes provide a snapshot of the DNA methylation status of an individual CpG, we confirmed the EPIC array results by targeted strand specific bisulfite sequencing of the region surrounding the identified retained CpGs (Supplementary Fig. 4b, c). Importantly, targeted bisulfite sequencing of the region demonstrated that many CpGs flanking the EPIC probe retained DNA methylation following long-term DNMT1 inhibition (Supplementary Fig. 4b, c). Consistent with our EPIC array results, the specific CpGs corresponding to the probes in XIST and ARMCX2 showed retention of methylation after drug treatment.
Identified CpGs that retain hypermethylation in promoters on the X chromosome are frequently methylated in TCGA tumors
Probes on the X chromosome are routinely excluded from methylation analysis in the TCGA and other data sets because their inclusion leads to strong clustering based on biological sex rather than on tumor type. We therefore queried TCGA data for the methylation status of CpG probes we had identified as unmethylated in non-cancer human tissues and as being highly resistant to DNA methylation inhibition in the cultured HCT116 parental and GSK5032-resistant cells. First, we examined data obtained using the now-obsolete Illumina 450 K arrays in colorectal and rectal tumors and adjacent normal tissue (Fig. 4a). We were able to do this for 22 probes which were shared on both arrays. With the exception of two probes (found in the promoters of genes DFCAF12L2 and INGX), which were highly methylated in all samples, the remaining 20 probes showed the expected behavior in tumor-adjacent normal tissue in that they were substantially unmethylated in male tissue and about 50% methylated in females (similar pattern to the probes profiled in Fig. 3e, f). They also showed variable, but substantial de novo methylation in the tumors in both females and males (Fig. 4 and Supplementary Fig. 5a, b). This varied from almost universal methylation for the top 10 probes to more sporadic levels in the bottom 10. Indeed, evaluation of the distribution of β-values for individual probes of interest between tumor and adjacent normal tissues demonstrate significant hypermethylation of these CpG sites in COAD/READ tumors (Supplementary Fig. 5c). The de novo methylation of genes on the active X chromosome is therefore not sex-specific.
Interestingly, while the same probes showed similar behaviors in other normal tissues (except for HTR2C which appeared to be unmethylated in some female samples), there was clear evidence for de novo methylation across a panel of diverse cancer types (Fig. 4b). The methylation was not sex-specific but was markedly less evident than in the colorectal/rectal tumors displayed above (Fig. 4a). The methylation changes are therefore related to tumor type. We also queried the status of 62 of the probes categorized as being methylated in non-cancer and tumor tissues (Supplementary Fig. 6). Most of the probes behaved as expected in normal tissues although a minority were less methylated than anticipated particularly in colorectal tumors relative to the other cancer types examined. Our approach of using cultured colorectal tumor cells to isolate these probes therefore validates the value of concentrating on probes highly resistant to demethylation.
Finally, the expression levels of the genes corresponding to the methylated promoter probes were examined in normal colonic tissue reported in the GTEX database16 and compared to expression levels in the HCT116 parental and GSK5032-resistant cells (Table 1). Robust expression of the X-linked housekeeping genes, PGK1 and HPRT1 was seen in all samples as expected for these controls. Eleven genes (indicated by bold type) were expressed in normal colon, and nine of these were not expressed in HCT116 parental cells.
Discussion
The recent development of a specific DNMT1 inhibitor, which does not require incorporation into DNA for its activity17, provided us with a useful tool to probe the role of the enzyme in the maintenance of DNA methylation patterns in cancer cell lines. Long term treatment of the HCT116 cells resulted in the derivation of cells which showed impaired growth rates and were markedly resistant to the compound. Importantly, we did not observe any mutations in the catalytic domain of DNMT1 that would account for the resistance observed and few changes were seen in the levels of mRNA for any of the DNMT isoforms examined; therefore, the mechanism of resistance remains unexplained.
Analysis of the patterns of residual methylation in the HCT116 and RKO GSK5032-resistant cells underscored the predominant role played by DNMT1 in maintaining DNA methylation in mammalian cells. Comparison of the patterns generated by inhibitor treatment with those in the HCT116 DKO1 line which has a complete knockout of DNMT3B and a hypomorphic version of DNMT1 allowed us to gain further insight into the potential roles of the so-called de novo enzymes DNMT3A and DNMT3B in contributing to maintenance methylation. An interesting finding was that the HCT116 DKO1 line, which has no detectable DNMT3B, has less global and focal retained methylation than the resistant cells strongly suggesting a role for DNMT3B in maintenance of methylation in addition to its function in its establishment during development. The further delineation of these complementary roles of DNMT3B will require more experimentation than presented here.
The most surprising result of our analysis of the residual methylation patterns were the marked differences in the percentages of probes remaining methylated on individual chromosomes. The differences between the percentages of probes maintaining a β-value \(\ge\) 0.7 varied by as much a 5 to 25 fold when comparing the lowest chromosome 9 to the highest chromosome X. The patterns of retained methylation on a per chromosome basis were remarkably similar at all β-value stringencies tested and in all cell lines. Since a common feature of all cell lines was the presence of DNMT3A, but not DNMT3B or a completely active DNMT1, a possible explanation of the result is that DNMT3A shows chromosome specificity in its ability to methylate DNA. Such specificity has not been noted before and it is not immediately apparent what might be driving such a pronounced variation between the macromolecular structures of chromosomes. Possible explanations include the existence of chromosome territories and differential nuclear organization18. Previously, Raddatz et al.19 showed that DNMT3A was able to maintain DNA methylation in large chromosome domains which had become hypomethylated in mouse tumors. Together with our data, it seems likely that DNMT3A shows methylation specificity over much larger domains than previously recognized.
It was remarkable that many of the same promoter CpGs found to be highly resistant to demethylation in culture were also de novo methylated in a considerable proportion of uncultured human colon and other tumors. Our expression analysis confirmed that many of the genes associated with these promoters were highly expressed in normal colonic tissue and silent in the HCT116 cell line (Table 1). While we have not formally established a link between the acquired methylation and gene silencing this seems to be quite likely. A possible explanation for the functions of these genes is that they are involved in cell viability and are selected for by the stringent requirement for continued growth in the presence of the inhibitor. Indeed, two of the examples we have highlighted, ARMCX2 and MAGEH1 (Fig. 3e, f), have previously demonstrated promoter hypermethylation and gene silencing in ovarian and liver cancer, respectively. Zeller et al identified ARMCX2 as a primary target for acquired promoter DNA hypermethylation in cisplatin-resistant ovarian cancer cell lines, including cell lines derived from patients at relapse after becoming resistant to chemotherapy20. Wang et al. profiled the expression of MAGEH1 across liver cancer cell lines and hepatocellular carcinoma patients and determined that downregulation of MAGEH1 led to increased proliferation and poor prognosis21. Unfortunately, little is known about the functional roles of ARMCX2 and MAGEH1 aside from their expression patterns in non-cancer human tissues22,23. Taken together with our results, our approach has identified potential tumor-suppressor genes that warrant further investigation for their functional roles’ in both non-cancerous and cancerous human cells.
Whatever the mechanism, the retention of DNA methylation in these promoters and their potential functions in cancer have largely been overlooked given the difficulties of deciphering DNA methylation patterns and gene expression changes on X-linked genes. We are therefore reanalyzing databases for their possible functions in cancer cells.
Methods
Cell culture
HCT116 parental and RKO parental cells were obtained fresh from ATCC (ATCC® CCL-247™, ATCC® CRL-2577™; Manassas, VA) and cultured in McCoy’s 5A (Life Technologies, 16600–082; Waltham MA) supplemented with 10% heat-inactivated fetal bovine serum, 1% penicillin/streptomycin, and RPMI (Gibco, 11875–093; Gaithersburg, MD) supplemented with MEM Non-Essential Amino Acids (Life Technologies, 11140–050; Waltham MA), sodium pyruvate (Life Technologies, 11360-070; Waltham MA), 10% heat-inactivated fetal bovine serum, 1% penicillin/streptomycin, respectively. Cells were treated with increasing doses of GSK5032 (GlaxoSmithKline; Collegeville, PA) starting with 200 nM and ultimately reaching 7400 nM over the course of 150 days. Fresh media and inhibitor were given to the cells every 3–4 days, passing the cells as needed. All cell lines tested negative for Mycoplasma contamination throughout the duration of the experiments. HCT116 and RKO GSK5032-resistant cell lines are available upon request. HCT116 DKO1 cells were kindly provided by Dr. Stephen B. Baylin and cultured in McCoy’s 5A (Life Technologies, 16600-082; Waltham MA) supplemented with 10% heat-inactivated fetal bovine serum and 1% penicillin/streptomycin.
Cell viability
IC50 curves were generated using FluoReporter™ Blue Fluorometric dsDNA Quantitation Kit (Invitrogen, F2962; Waltham MA) according to manufacturer’s instructions. Briefly, cells were plated in 96 well white walled plates (Corning, 3601; Corning, NY) and treated with 13 increasing two-fold dilutions of GSK5032. Plates were harvested on day 5 after treatment and processed. Fluorescence was measured with excitation/emission wavelengths were measured at 360 nm/460 nm using a Synergy HT multimode microplate reader (BioTek; Winooski, VT).
RNA isolation, cDNA synthesis, qPCR
Cells were harvested for RNA by adding TRIzol (ThermoFisher, 15596026; Waltham, MA) directly to the plate to lyse the cells. The Direct-zol™ RNA MiniPrep Plus kit (Zymo Research, R2072; Irvine, CA) with DNaseI digestion was used to extract RNA. RNA concentration was determined with a Nanodrop Lite (ThermoScientific; Waltham, MA) then samples were normalized to the same concentration. The High-Capacity cDNA Reverse Transcription Kit (ThermoFisher, 4368814; Waltham, MA) was used to generate cDNA with 2ug input RNA according to the manufacturer’s instructions.
After cDNA was diluted 4 fold, 4 μL of cDNA was used in a 20 μL qPCR reaction with 250 nM primers and 10 μL KAPA SYBR® FAST qPCR Master Mix (2X) (Kapa Biosystems, KK4618; Wilmington, MA). Primers (Supplementary Table 1) were used to measure gene expression. Displayed values are ΔΔCq values normalized to TBP expression. Each graph represents the average of three replicates with standard error plotted.
DNA isolation and EPIC array
DNA was isolated using the DNeasy Blood and Tissue kit (Qiagen, 69504; Germantown, MD) according to manufacturer’s instructions, then submitted to the Van Andel Institute Genomics Core for processing on Illumina’s Infinium MethylationEPIC BeadChip.
EPIC array processing
EPIC array image data (idat) files were read into R (Version > 3.6) and processed using the SeSAMe package (version 1.8.2) using the standard settings24. EPIC array probes with a detection p-value > 0.05 were excluded from the analysis as probes that fail this check are unable to provide reliable β-value calculations.
Enrichment bias calculation and hypergeometic distribution testing
Enrichment bias calculations were done by first determining the following values for each chromosome:
q = Number of CpGs that retained DNA methylation in the DNMT compromised cell lines (DKO1; GSK5032-resistant) (β-value \(\ge\) 0.7)
m = Total number of CpGs on the EPIC array with high DNA methylation (β-value \(\ge\) 0.7) in the parental cells (HCT116; RKO)
n = Total number CpGs on the EPIC array that do not match feature
k = Total number of all CpGs with DNA methylation (β-value \(\ge\) 0.7) in the DNMT compromised cell lines (DKO1; GSK5032-resistant).
Next, the expected number of CpGs that would retain DNA methylation on each chromosome by random chance was determined with the following equation:
Finally, percent enrichment bias was calculated with the following equation:
Where positive or negative enrichment values indicate more or less enrichment for a chromosome than would be expected by random chance, respectively.
Hypergeometric distribution testing for determining significance of enrichment bias was performed using the phyper() function in R (Version > 3.6) with the following values: q, m, n, k.
Categorization of probe methylation status in non-cancer human tissues
Number of probes remaining methylated after long term, high dose GSK5032 treatment were determined using a β-value cut off of \(\ge\) 0.7. Using a panel of non-cancerous male and female EPIC datasets (Supplementary Table 2), probes were called to be unmethylated when they were unmethylated in males or 50% methylated in females for probes on the X chromosome, but were greater than 70% methylated in HCT116 parental and GSK5032-resistant cells. Probes were considered to be methylated when they also had a high beta value (β-value \(\ge\) 0.7) in the non-cancerous male and female tissues.
TCGA analysis
DNA methylation array data (Infinium HumanMethylation450 BeadChip; HM450) and clinical data were downloaded from the GDC Data Portal (https://portal.gdc.cancer.gov/repository), which includes 8790 and 714 whitelisted human tumor and adjacent normal samples, respectively, across 32 different tumor types. Among the 119 EPIC array probes, 84 are present on the HM450 array. Heatmaps were generated in RStudio (Version > 3.6), with samples split based on clinical biological sex along with subdivisions into either COAD/READ or non-COAD/READ tumor and their adjacent normal.
TCGA statistical analysis
To test if the amount of methylation across all 22 CpGs of interest differed between the COAD/READ and normal tissues, a beta mixed-effects model with random intercepts for both CpG and patient ID was fit using R (Version > 3.6) via the glmmTMB package (https://cran.r-project.org/web/packages/glmmTMB/index.html). An interaction between biological sex and group were included to allow the mean methylation to vary by biological sex (which was expected due to the number of X chromosomes present in biological males versus females). To investigate differential methylation between COAD/READ samples and normal tissue at a subset of specific loci (cg02626719, cg18124907, cg19355555, cg24401557, and cg12950441) beta mixed-effects regressions with random intercepts for each patient and an interaction for biological sex were used. cg18124907 was an exception due to poor diagnostics assuming a beta distribution and was instead analyzed using a semi-parametric ordinal mixed-effects model (https://cran.microsoft.com/snapshot/2017-08-01/web/packages/ordinal/index.html).
Statistics and reproducibility
Statistical analyses were performed as described in the methods section and figure legends. In all cases, an independent biological replicate indicates that cells were plated and treated with drugs completely separately and on completely different days for a true biological replicate. For Fig. 1a, the biological replicates for the IC50 curves were 5 for HCT116 Parental and 7 for HCT116 GSK5032-resistant. The biological replicates for RKO Parental and RKO GSK5032-resistant were 3 each. The EPIC data was generated from one biological replicate for each cell line and contained a total of 627,029 probes that were included in the analysis for the rest of the figures.
For Supplementary Fig. 1a, the cell counts had the biological replicates as indicated by the number in parentheses for HCT116 Parental (9), HCT116 GSK5032-resistant (6), HCT116 DKO1 (4), RKO Parental (5), and RKO GSK5032-resistant (4). For Supplementary Fig. 1b, the qPCR data and standard error were calculated from three biological replicates for each sample. Biological replicates were from separate preparations of RNA that was extracted and converted to cDNA independently as well.
Reporting summary
Further information on research design is available in the Nature Research Reporting Summary linked to this article.
Data availability
All Illumina Infinium MethylationEPIC array data is deposited in GEO and available under accession GSE182209. Source data for figures are available in Supplementary Data 1.
References
Li, E., Bestor, T. H. & Jaenisch, R. Targeted mutation of the DNA methyltransferase gene results in embryonic lethality. Cell 69, 915–926 (1992).
Jones, P. A. & Laird, P. W. Cancer-epigenetics comes of age. Nat. Genet. 21, 163–167 (1999).
Jones, P. A., Issa, J.-P. J. & Baylin, S. Targeting the cancer epigenome for therapy. Nat. Rev. Genet. 17, 630–641 (2016).
Egger, G. 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 103, 14080–14085 (2006).
Rhee, I. et al. CpG methylation is maintained in human cancer cells lacking DNMT1. Nature 404, 1003–1007 (2000).
Rhee, I. et al. DNMT1 and DNMT3b cooperate to silence genes in human cancer cells. Nature 416, 552–556 (2002).
De Carvalho, D. D. et al. DNA methylation screening identifies driver epigenetic events of cancer cell survival. Cancer Cell 21, 655–667 (2012).
Jones, P. A. & Liang, G. Rethinking how DNA methylation patterns are maintained. Nat. Rev. Genet. 10, 805–811 (2009).
Liang, G. et al. Cooperativity between DNA methyltransferases in the maintenance methylation of repetitive elements. Mol. Cell Biol. 22, 480–491 (2002).
Haggerty, C. et al. Dnmt1 has de novo activity targeted to transposable elements. Nat. Struct. Mol. Biol. 28, 1–10 (2021)
Yang, X. et al. Gene body methylation can alter gene expression and is a therapeutic target in cancer. Cancer Cell 26, 577–590 (2014).
Duymich, C. E., Charlet, J., Yang, X., Jones, P. A. & Liang, G. DNMT3B isoforms without catalytic activity stimulate gene body methylation as accessory proteins in somatic cells. Nat. Commun. 7, 11453 (2016).
Baubec, T. et al. Genomic profiling of DNA methyltransferases reveals a role for DNMT3B in genic methylation. Nature 520, 243–247 (2015).
Xu, T.-H. et al. Structure of nucleosome-bound DNA methyltransferases DNMT3A and DNMT3B. Nature 586, 151–155 (2020).
Loda, A. & Heard, E. Xist RNA in action: Past, present, and future. PLoS Genet. 15, e1008333 (2019).
Carithers, L. J. et al. A novel approach to high-quality postmortem tissue procurement: The GTEx project. Biopreserv. Biobank. 13, 311–319 (2015).
Pappalardi, M. B. et al. Discovery of a first-in-class reversible DNMT1-selective inhibitor with improved tolerability and efficacy in acute myeloid leukemia. Nat. Cancer 2, 1002–1017 (2021).
Parada, L. A. & Misteli, T. Chromosome positioning in the interphase nucleus. Trends Cell Biol. 12, 425–432 (2002).
Raddatz, G., Gao, Q., Bender, S., Jaenisch, R. & Lyko, F. Dnmt3a protects active chromosome domains against cancer-associated hypomethylation. PLoS Genet. 8, e1003146 (2012).
Zeller, C. et al. Candidate DNA methylation drivers of acquired cisplatin resistance in ovarian cancer identified by methylome and expression profiling. Oncogene 31, 4567–4576 (2012).
Wang, P.-C. et al. Downregulation of MAGE family member H1 enhances hepatocellular carcinoma progression and serves as a biomarker for patient prognosis. Future Oncol. 14, 1177–1186 (2018).
López-Doménech, G. et al. The Eutherian Armcx genes regulate mitochondrial trafficking in neurons and interact with Miro and Trak2. Nat. Commun. 3, 814 (2012).
Florke Gee, R. R. et al. Emerging roles of the MAGE protein family in stress response pathways. J. Biol. Chem. 295, 16121–16155 (2020).
Zhou, W., Triche, T. J. Jr, Laird, P. W. & Shen, H. SeSAMe: Reducing artifactual detection of DNA methylation by Infinium BeadChips in genomic deletions. Nucleic Acids Res. 46, e123–e123 (2018).
Acknowledgements
We thank the Van Andel Institute Core Technologies and Services including the Genomics Core for performing the Illumina Infinium methylation EPIC array, and Benjamin Johnson for helpful discussions and technical expertize. This work was supported by the VAI, by R35CA209859 from the National Cancer Institute to P.A.J., and by PF-16-245-01-DMB from the American Cancer Society Michigan Cancer Research Fund to R.L.T.
Author information
Authors and Affiliations
Contributions
A.K.W. and P.A.J. conceived and designed the study. A.K.W. performed the experiments. R.L.T. conceived the genomics studies. A.K.W. and R.L.T. performed the EPIC array data analysis. H.F. and H.S. consulted and performed the TCGA analysis. Z.M. performed statistical analysis on the TCGA dataset. M.B.P. and M.T.M. provided critical discussion and review of the study. P.A.J., A.K.W., and R.L.T. wrote the original draft, and all authors edited the final manuscript.
Corresponding author
Ethics declarations
Competing interests
M.B.P. and M.T.M are employees and shareholders of GlaxoSmithKline (GSK). All other authors declare no competing interests.
Peer review
Peer review information
Communications Biology thanks Gary Stein and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. Primary Handling Editors: Erik Knudsen and Eve Rogers.
Additional information
Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Rights and permissions
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as 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 images or other third party material in this article are included in the article’s Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/.
About this article
Cite this article
Wiseman, A.K., Tiedemann, R.L., Fan, H. et al. Chromosome-specific retention of cancer-associated DNA hypermethylation following pharmacological inhibition of DNMT1. Commun Biol 5, 528 (2022). https://doi.org/10.1038/s42003-022-03509-3
Received:
Accepted:
Published:
DOI: https://doi.org/10.1038/s42003-022-03509-3
Comments
By submitting a comment you agree to abide by our Terms and Community Guidelines. If you find something abusive or that does not comply with our terms or guidelines please flag it as inappropriate.