Dysregulation of DNA methylation is an established feature of breast cancers. DNA demethylating therapies like decitabine are proposed for the treatment of triple-negative breast cancers (TNBC) and indicators of response need to be identified. For this purpose, we characterized the effects of decitabine in a panel of 10 breast cancer cell lines and observed a range of sensitivity to decitabine that was not subtype specific. Knockdown of potential key effectors demonstrated the requirement of deoxycytidine kinase (DCK) for decitabine response in breast cancer cells. In treatment-naïve breast tumors, DCK was higher in TNBCs, and DCK levels were sustained or increased post chemotherapy treatment. This suggests that limited DCK levels will not be a barrier to response in patients with TNBC treated with decitabine as a second-line treatment or in a clinical trial. Methylome analysis revealed that genome-wide, region-specific, tumor suppressor gene–specific methylation, and decitabine-induced demethylation did not predict response to decitabine. Gene set enrichment analysis of transcriptome data demonstrated that decitabine induced genes within apoptosis, cell cycle, stress, and immune pathways. Induced genes included those characterized by the viral mimicry response; however, knockdown of key effectors of the pathway did not affect decitabine sensitivity suggesting that breast cancer growth suppression by decitabine is independent of viral mimicry. Finally, taxol-resistant breast cancer cells expressing high levels of multidrug resistance transporter ABCB1 remained sensitive to decitabine, suggesting that the drug could be used as second-line treatment for chemoresistant patients.

DNA methylation is essential for gene regulation in normal cells (1). In cancer, DNA methylation is largely dysregulated with global demethylation contributing to genomic instability and the hypermethylation of CpG islands in the promoters of tumor suppressor genes causing their aberrant silencing (1). DNA methyltransferases (DNMT) are required for both de novo methylation and maintenance of existing DNA methylation; DNMT upregulation is associated with both cancer and aberrant methylation. As such, DNMT inhibitors like decitabine and azacytidine are commonly used to treat hematologic disorders [myelodysplastic syndrome (MDS) and some acute myeloid leukemias (AML)], where patients often share common epigenetic perturbations (2). This has generated interest in using demethylating agents to treat solid tumors, including breast cancers (3).

Among breast cancers, triple-negative breast cancers (TNBC) have poorer outcomes and represent the 15%–20% of tumors that lack hormone receptors and targeted therapies (4, 5). The possibility of treating TNBCs with DNMT inhibitor decitabine is currently being investigated in clinical trials (i.e., NCT02957968 and NCT03295552). Therefore, assessing the factors that determine the response to DNMT inhibitors in breast cancer, specifically TNBC, is both timely and critical if decitabine is to be used successfully. To date, there are few studies which examine breast cancer cells exclusively and profile response to decitabine across many cell lines.

A cytosine analog, decitabine is incorporated into DNA during synthesis, which imparts some specificity of the drug for rapidly dividing cells. DNMTs bind DNA-integrated decitabine leading to protein/DNA adduct formation, DNMT degradation, and subsequent reduction of DNA methylation. This inhibits tumor growth by a number of potential mechanisms including demethylation and reexpression of aberrantly silenced tumor suppressor genes (6), induction of the DNA damage response by protein/DNA adduct formation (7), cytotoxicity induced by global demethylation (8), and demethylation of silenced tumor-associated antigens increasing antitumor immune responses (9, 10). Recently, a novel mechanism has been proposed as a key determinant for the response of decitabine—demethylation of endogenous retroviral elements resulting in dsRNA/antiviral responses (11, 12). In addition, decitabine's predominant mode of action is possibly dose dependent: lower doses cause reexpression of silenced genes with minimal DNA damage, while higher doses cause more pronounced DNA damage responses and apoptosis (12, 13). In various cancer models, a number of potential treatment response biomarkers have been investigated including expression or mutation of nucleoside transporters, decitabine metabolism genes deoxycytidine kinase (DCK), and DNA methylation regulating enzymes such as DNMTs and tet methylcytosine dioxygenase 2 (TET2), the methylation of known tumor suppressor genes (e.g., CDH1, BRCA1, RASSF1, and RUNX3), and global methylation levels (14–20). It is unclear which mechanism of decitabine is most important for successful treatment of patients with breast cancer or if the effects of decitabine differ in the breast cancers of different subtypes (i.e., TNBCs vs. hormone-expressing subtypes).

Herein we performed transcriptome, methylome, in vitro growth assays, gene knockdown studies, tumor xenograft assays, and patient dataset analyses to assess the proposed potential anticancer mechanisms of decitabine in breast cancer. Our analyses of a panel of 10 breast cancer cell lines revealed a range of sensitivity in breast cancer to decitabine that was not based on hormone receptor status, genome-wide methylation, demethylation of tumor suppressor genes, or induction of viral mimicry responses. Instead our analyses demonstrated the requirement for expression of the decitabine-processing enzyme DCK and the induction of pathways of genes enriched in apoptosis, cell cycle, stress, and immune pathways. Finally, unlike the commonly used breast cancer drug paclitaxel, decitabine efficacy in breast cancer is not negatively impacted by increased expression of ATP-binding cassette drug efflux transporter ABCB1, which is often seen as a mechanism of multi-drug resistance.

Cell culture

Cancer cell lines were obtained from ATCC, with the exception of SUM149 and SUM159 cells that were obtained from BioIVT (previously Asterand). MDA-MB-231, MDA-MB-468, MCF7, SKBR3, T47D, and HEK293T cells were grown in DMEM (Invitrogen) supplemented with 10% FBS (Invitrogen) and 1× antibiotic antimycotic (AA; Invitrogen). MDA-MB-436 cells were grown in Leibovitz Medium (L-15; Invitrogen) supplemented with 10% FBS, 1× AA, 10 μg/mL human insulin (Sigma), and 16 μg/mL l-glutathione (Invitrogen); MDA-MB-453 cells were cultured in L-15 medium supplemented with 10% FBS and 1× AA; Hs578T cells were cultured in DMEM supplemented with 10% FBS, 1× AA, and 0.01 μg/mL human insulin. SUM149 and SUM159 cells were cultured in F-12 Ham Nutrient Mix Medium supplemented with 5% FBS, 1× AA, 1 μmol/L 4-(2-Hydroxyethyl) piperazine-1-ethanesulfonic acid (HEPES; Invitrogen), 0.01 μg/mL human insulin, and 0.05 μg/mL hydrocortisone (Invitrogen). Cells were cultured in a humidified 37°C incubator with 5% CO2, except for MDA-MB-436 and MDA-MB-453, which were cultured without the addition of CO2.

Colony-forming assay

Cells were seeded at low density and allowed to adhere for 24 hours in 24-well cell culture plates: MDA-MB-231 (30 cells/well), MDA-MB-468 (120 cells/well), MCF7 (60 cells/well), SKBR3 (60 cells/well), T47D (120 cells/well), MDA-MB-436 (200 cells/well), MDA-MB-453 (60 cells/well), Hs578T (60 cells/well), SUM149 (60 cells/well), and SUM159 (30 cells/well). Cells were treated with 0.122 μmol/L–2 μmol/L decitabine (5-aza-2′-deoxycytidine, Sigma-Aldrich) for 72 hours with media refreshed every 24 hours. Alternatively, cells were treated with azacytidine (5-azacytidine, Sigma-Aldrich). Cells were then grown in appropriate media (lacking decitabine or azacytidine) for 7–10 days with media refreshed every other day. Colonies were then fixed in methanol for 10 minutes and stained using 0.05% crystal violet (Sigma); colonies were defined as >50 cells. An IC50 value for each cell line was determined using percent colony-forming efficiency (with no treatment wells representing 100% colony-forming efficiency) and the GraphPad Prism equation: log(inhibitor) versus normalized response standard slope analysis (⁠$Y = 100/(1+10)^{(x - LogIC50)} )$⁠.

Tumor growth studies

All animal studies detailed in this article have been conducted in accordance with the ethical standards and according to the Declaration of Helsinki and the Canadian Council on Animal Care standards. Eight-week-old NOD/SCID female mice were injected with 2 × 106 MDA-MB-468 or MDA-MB-231 cells, or 3.5 × 106 SUM159 cells admixed 1:1 with Matrigel-HC (BD Biosciences) into the mammary fat pad. Once palpable tumors formed, the mice were treated with 0.5 mg/kg decitabine or vehicle control (PBS) by intraperitoneal injection for 3/5 day cycles as described previously (13). Throughout the experiment, tumor volume was quantified (mm3, length × width × depth/2).

Knockdown generation

Lentiviral short hairpin RNA (shRNA) knockdown clones of ABCB1, DCK, DDX58 (RIGI), IFIH1 (MDA5), SLC28A1, and SLC29A1, were generated using the pGipZ vector (Dharmacon) packaged in HEK293T cells following standard protocols and listed in Supplementary Table S1. Clones were selected by adding 1.5 μg/mL puromycin and subsequently maintaining 0.25 μg/mL puromycin media. For all knockdown clones created, a GIPz vector control clone (containing a scrambled nonspecific sequence in place of a shRNA) was generated simultaneously. Verification of knockdown was done through qRT-PCR and Western blot analysis (anti-DCK, Abcam, ab151966; anti-MDA5, Cell Signaling Technology, clone D74E4).

Reverse transcriptase quantitative PCR

RNA was extracted from untreated and decitabine-treated cells (1 μmol/L decitabine for 72 hours with media refreshed daily). Cells were collected in TRIzol (Invitrogen) and RNA was purified using a PureLink RNA Kit (Invitrogen, Thermo Fisher Scientific) following the manufacturer's instructions. Equal amounts of purified RNA were then reverse transcribed to cDNA using iScript (Bio-Rad) as per the manufacturer's instructions. Diluted cDNA was used in qRT-PCR reactions with gene-specific primers (Supplementary Table S2) and SsoAdvanced Universal SYBR Supermix (Bio-Rad) as per the manufacturer's instructions with a CFX96 or CFX384 Touch Real-Time PCR Detection System (Bio-Rad). Standard curves were generated for each primer set and primer efficiencies were incorporated into the CFX Manager Software (Bio-Rad). Relative expression for decitabine inducible genes was quantified using the ΔΔCt method of the CFX Manager Software (Bio-Rad), where gene-of-interest quantification was normalized to reference genes RPL29 and TBP and then made relative to no-treatment control mRNA levels. For gene expression comparisons made between cell lines, the ΔCt expression method was used to quantify gene expression.

Patient dataset analysis

Breast cancer (METABRIC, Nature 2012 and Nat Commun 2016; n = 2,509) and breast invasive carcinoma [The Cancer Genome Atlas (TCGA), Cell 2015; n = 817] clinical data (PAM50 subtype, hormone receptor status), and microarray z-score gene expression data were accessed via cBioPortal (21, 22). Microarray-based gene expression of matched pre- and postneoadjuvant chemotherapy biopsies from patients with breast cancer was acquired from GSE28844 (23); histopathologic response was based on Miller and Payne grading system.

Human methylation 450K analysis

Genomic DNA was extracted from untreated and decitabine-treated cells (1 μmol/L decitabine for 72 hours with media refreshed daily) using the PureLink DNA Kit (Invitrogen). Methylation analyses using the human methylation 450K (HM450) bead chip array (Illumina) was performed by The Centre for Applied Genomics at the Hospital for Sick Children (Toronto, Ontario, Canada), including bisulfite conversion, hybridization, background subtraction, and normalization. Data are accessible from GSE78875 (24) and summarized in Supplementary Data S1. Methylation β-value of probes from promoter-associated regions for BRCA1, CDH1, RASSF1, and RUNX3 were extracted. CpGs from RUNX3 promoter–associated CpG island: cg19270505, cg11018723, cg13629563, cg22737001, cg26421310, cg26672794, and cg02970551.

Gene array analysis

RNA purified from SUM159 cells that had been treated with 1 μmol/L decitabine or vehicle for 72 hours (3n) was sent to The Centre for Applied Genomics at the Hospital for Sick Children (Toronto, Ontario, Canada) for sample preparation, amplification, hybridization to the Affymetrix HuGene 2.0 ST array, and data collection. The raw data for MDA-MB-231 and MDA-MB-468 is accessible from GSE103427 (25), and GSE from uploaded new data (GSE133987). The Transcriptome Analysis Console Software (Thermo Fisher Scientific) was used to normalize the data and calculate fold changes in expression (Supplementary Data S2). Transcripts with confirmed gene annotations (not blank in the “Gene Symbol”) category that were 1.5-fold up- or downregulated significantly (based on ANOVA P < 0.05) in at least one cell line were plotted on the basis of the average fold-change across all three cell lines (1,390 transcripts/1,284 genes; Supplementary Fig. S1). Not all microarray gene expression “hits” contained HM450 annotated CpG sites; unannotated transcripts were discarded. Only genes with two or more different regions annotated were plotted. All figures depict the mean methylation β-value of CpGs annotated for a given region (e.g., TSS1500, TSS200 etc.).

Gene set enrichment analysis

Using the HM450 and gene array data, we identified genes that were unmethylated in TSS1500 + TSS200 + Exon1 (≤0.5 β-value in all three regions) and for which decitabine upregulated expression ≥1.5-fold (Supplementary Data S3, Supplementary Table S3). Gene set enrichment analysis (GSEA) was performed using available online software (http://software.broadinstitute.org/gsea/index.jsp) to compute overlaps. The top 100 overlapping gene sets identified by GSEA for each cell line were identified. k, # genes in overlap; K, # genes in gene s; n, # genes in comparison; N, # genes in universe; enrichment = $k/n/K/N$⁠. Gene sets with similar functions were enriched across the three cell lines and were color-coded for easier visualization.

Taxane-resistant breast cancer

MDA-MB-231 cells with acquired ABCB1-mediated taxane resistance (taxol-res) and the matched taxane-sensitive cells (control) were previously generated and characterized (26) and maintained in the same MDA-MB-231 growth conditions described above. Cells were treated with increasing doses of paclitaxel (Corporation Biolyse Pharma Corporation) or decitabine ± 10 μmol/L verapamil (verapamil hydrochloride; Sigma-Aldrich) and a colony forming IC50 was calculated.

Breast cancer cell lines have a broad range of sensitivity to decitabine, independent of subtype

To first explore the range of sensitivity to decitabine in breast cancer, we treated a panel of 10 cell lines representing estrogen receptor–positive breast cancers (MCF7 and T47D), HER2+ (SKBR3), and TNBC (MDA-MB-231, MDA-MB-468, SUM149, SUM159, MDA-MB-436, MDA-MB-453, and Hs578T). Given the current clinical interest in treating TNBCs with decitabine (NCT02957968 and NCT03295552), the panel has overrepresentation of TNBCs. In a subsequent colony-forming assay, the cell lines exhibited IC50s ranging from 1 to 74 nmol/L of decitabine (Fig. 1A). The in vitro sensitivity to decitabine appeared to be independent of hormone receptor status (Fig. 1A).

Figure 1.

Sensitivity of breast cancer cell lines to decitabine treatment. A, Colony-forming assay to determine in vitro decitabine IC50 for 10 breast cancer cell lines after 72-hour treatment. B, Xenograft determination of in vivo decitabine sensitivity. MDA-MB-468, MDA-MB-231, and SUM159 xenografts treated with 0.5 mg/kg decitabine over 3–4 weeks; hatch marks on x-axis indicate when mice were injected; error bars, SEM; one-way t test (**, P < 0.01).

Figure 1.

Sensitivity of breast cancer cell lines to decitabine treatment. A, Colony-forming assay to determine in vitro decitabine IC50 for 10 breast cancer cell lines after 72-hour treatment. B, Xenograft determination of in vivo decitabine sensitivity. MDA-MB-468, MDA-MB-231, and SUM159 xenografts treated with 0.5 mg/kg decitabine over 3–4 weeks; hatch marks on x-axis indicate when mice were injected; error bars, SEM; one-way t test (**, P < 0.01).

Close modal

With the focus of ongoing clinical application in TNBCs, we also assessed the effect of decitabine on tumor growth of TNBC MDA-MB-468, MDA-MB-231, and SUM159 cells. Using the low-dose treatment protocol established by Tsai and colleagues (13), NOD/SCID mice bearing palpable tumors of the TNBC cell lines were treated intermittently with 0.5 mg/kg decitabine. This resulted in tumor growth suppression that mimicked the colony assay, with MDA-MB-468 tumors being the most sensitive and SUM159 tumors being the most resistant to decitabine (Fig. 1B).

DCK is required for decitabine response in breast cancer cells and tumors

To assess the potential factors which may dictate breast cancer sensitivity to decitabine, we considered the cell-specific factors required for the cytosine analog's incorporation into DNA. Decitabine is imported into cells by sodium/nucleoside cotransporters solute carrier family 28 member 1 (SLC28A1) and SLC29A1 (18, 27). Higher expression of either transporter was not associated with increased sensitivity (Supplementary Fig. S2). Furthermore, knockdown of SLC28A1 in decitabine-sensitive MDA-MB-468 cells and -resistant SUM159 cells did not alter the sensitivity of either cell line to decitabine (Supplementary Fig. S3). Knockdown of SLC29A1 in intermediately sensitive MDA-MB-231 (Supplementary Fig. S4), which had high expression of the transporter (Supplementary Fig. S2), also did not alter the sensitivity of the cells to decitabine. Together, this data suggests that assessing levels of the importers in patient tumors will not predict decitabine sensitivity or resistance in breast cancer.

Once imported, decitabine is sequentially phosphorylated by DCK, cytidine/uridine monophosphate kinase 1 (CMPK1), and finally nucleoside diphosphate kinases 1 and 2 (NME1 and NME2; ref. 28). DCK is a rate-limiting step for the incorporation of decitabine in MDS and AML (28, 29) and could possibly be a rate-limiting step for decitabine response in breast cancer. Consistent with its requirement for decitabine processing, knockdown of DCK significantly decreased the sensitivity of both the decitabine-sensitive MDA-MB-468 cells and decitabine-resistant SUM159 cells (Fig. 2A). This was also observed in vivo, where MDA-MB-468 tumors with DCK knockdown were comparably resistant to decitabine (Fig. 2B). The reduction in DCK levels also hampered expression of decitabine-inducible genes (Supplementary Fig. S5) that were identified as upregulated by decitabine in the gene array (Supplementary Data S2). It is noteworthy that MDA-MB-468 cells with DCK knockdown remained sensitive to the ribonucleoside analog and DNMT inhibitor azacytidine (Fig. 2C), which does not require DCK for DNA incorporation and activity (29).

Figure 2.

DCK is an important mediator of decitabine response and is present in patients with aggressive subtypes of breast cancer (BrCa). A, shRNA-mediated knockdown (KD) of DCK conferred in vitro resistance to decitabine in MDA-MB-468 and SUM159 cells. B, shRNA-mediated knockdown of DCK conferred in vivo resistance to decitabine to MDA-MB-468–treated tumors. C, DCK knockdown in MDA-MB-468 cells does not confer resistance to azacytidine. Error bars, SD; ANOVA with Dunnett post hoc test. D, qRT-PCR of nucleotide kinases (DCK, CMPK1, NME1, and NME2) across breast cancer cell lines, error bars, SD. E and F, Expression of DCK via microarray based on PAM50 subtype and triple-negative receptor status in METABRIC and TCGA (Cell, 2015) breast cancer patient cohorts. Straight lines note groups that are not significantly different, ANOVA with (Tukey post hoc test); ***, P < 0.001; ns, not significant. G, DCK expression via microarray in the GSE28844 dataset of patients with breast cancer pre- and post-anthracycline/taxane treatment; paired two-tailed t test (*, P < 0.05; **, P < 0.01; ***, P < 0.001).

Figure 2.

DCK is an important mediator of decitabine response and is present in patients with aggressive subtypes of breast cancer (BrCa). A, shRNA-mediated knockdown (KD) of DCK conferred in vitro resistance to decitabine in MDA-MB-468 and SUM159 cells. B, shRNA-mediated knockdown of DCK conferred in vivo resistance to decitabine to MDA-MB-468–treated tumors. C, DCK knockdown in MDA-MB-468 cells does not confer resistance to azacytidine. Error bars, SD; ANOVA with Dunnett post hoc test. D, qRT-PCR of nucleotide kinases (DCK, CMPK1, NME1, and NME2) across breast cancer cell lines, error bars, SD. E and F, Expression of DCK via microarray based on PAM50 subtype and triple-negative receptor status in METABRIC and TCGA (Cell, 2015) breast cancer patient cohorts. Straight lines note groups that are not significantly different, ANOVA with (Tukey post hoc test); ***, P < 0.001; ns, not significant. G, DCK expression via microarray in the GSE28844 dataset of patients with breast cancer pre- and post-anthracycline/taxane treatment; paired two-tailed t test (*, P < 0.05; **, P < 0.01; ***, P < 0.001).

Close modal

We assessed expression levels of DCK and the other nucleotide kinases involved in processing deoxynucleotides/decitabine in the 10 breast cancer cell lines and noted similar levels of DCK, CMPK1, NME1, and NME2 across the cell lines (Fig. 2D). The lack of a correlation between DCK expression and decitabine IC50 of the cell lines was likely due to the small range of differences in DCK levels between the cell lines (Fig. 2D). However, given the importance of DCK in mediating decitabine response in breast cancer as demonstrated by our knockdown experiments (Fig. 2A and B; Supplementary Fig. S5), we assessed the expression of the DCK in two breast cancer patient cohorts (METABRIC, Fig. 2E and TCGA Cell 2015, Fig. 2F). DCK mRNA levels in the breast cancer patient tumors followed a normal distribution and was overall more abundant in the luminal B, basal-like (which consists predominately of TNBCs), and TNBC subtypes. This suggests that for most patients (and in the patients with TNBC that are currently being treated by decitabine in clinical trials), limited DCK expression will not be a barrier to successful decitabine therapy in these patients. Since the patients that will be treated with decitabine will likely have had anthracycline and/or taxane chemotherapy prior to enrollment in a clinical trial, it is important to determine whether DCK levels are altered posttreatment with standard chemotherapy drugs. Microarray-based gene expression from 56 matched pre- and post-chemotherapy breast tumor samples showed that DCK expression was elevated after chemotherapy in patients who did not show a histopathologic response to treatment (Fig. 2G). This suggests that these patients would be good candidates for decitabine treatment, at least with respect to the essential decitabine processing enzyme DCK.

Genome-wide and region-specific methylation of breast cancer cells is not predictive of decitabine response

Concurrent genome-wide hypomethylation and hypermethylation of promoter regions is observed in cancer cells (30); therefore, we wondered whether any correlations with genome-wide or region-specific methylation could predict response to decitabine in breast cancer. For this purpose, we performed HM450 and gene expression analysis of decitabine-treated cells. The colony assay (Fig. 1A) is a long-term 2-week plus assay and hence results in an enhanced decitabine sensitivity in the nmol/L range in IC50s. While informative for determining relative sensitivities, this assay is insufficient for harvesting cell samples for HM450 and RNA analysis, which requires >300,000 cells. Therefore, we determined a concentration of decitabine that when applied to subconfluent monolayers of >300,000 cells seeded in 6-well plates would result in approximate 50% growth inhibition of cells after 72 hours (0.6–9 μmol/L; Supplementary Fig. S6). A dose of 1 μmol/L decitabine (which represents a median IC50 in the cell line panel) was applied to subconfluent monolayers for 72 hours (Supplementary Fig. S6). Furthermore, this dose causes minimal apoptosis after 72 hours of treatment (day 4 post-seeding), but in subsequent days (decitabine treatment has ceased but cell culture continued) causes dramatic cell death and growth inhibition (Supplementary Fig. S7). This is further evidence of the long-term effects of decitabine, which are also captured by the ultra-sensitive colony assay (Fig. 1A).

We analyzed HM450 data from 10 breast cancer cell lines (not treated with decitabine) and found that total genome-wide DNA methylation was not associated with response to decitabine and was overall similar among the cell lines (Fig. 3A and B). In agreement with this, the levels of DNMT1A, DNMT3B, and TET 1, 2 and 3 (which act in the demethylation of DNA; ref. 31), were also similar across the cell lines, consistent with the overall similar genome-wide methylation (Supplementary Fig. S8).

Figure 3.

Breast cancer sensitivity to decitabine is not associated with global methylation or promoter methylation. A, DNA methylation as determined via HM450 array among breast cancer cell lines as a frequency distribution of CpG sites with a given methylation β-value. B, Distribution of HM450 methylation β-values for promoter-associated regions TSS1500, TSS200, 5′ UTR, and Exon 1 among the 10 breast cancer cell lines. C, The average methylation β-values for promoter-associated regions TSS1500, TSS200, 5′ UTR, and Exon 1 among the 10 cell lines are plotted against the relative change in methylation of those promoter regions (methylation z-score) based on decitabine responsiveness. Sites located in top left zone of dotplot represent sites that are hypermethylated in (e.g., resistant) cell lines compared with the rest of the cell lines. D, The hypermethylated regions identified in C are represented by the methylation β-value in each cell line.

Figure 3.

Breast cancer sensitivity to decitabine is not associated with global methylation or promoter methylation. A, DNA methylation as determined via HM450 array among breast cancer cell lines as a frequency distribution of CpG sites with a given methylation β-value. B, Distribution of HM450 methylation β-values for promoter-associated regions TSS1500, TSS200, 5′ UTR, and Exon 1 among the 10 breast cancer cell lines. C, The average methylation β-values for promoter-associated regions TSS1500, TSS200, 5′ UTR, and Exon 1 among the 10 cell lines are plotted against the relative change in methylation of those promoter regions (methylation z-score) based on decitabine responsiveness. Sites located in top left zone of dotplot represent sites that are hypermethylated in (e.g., resistant) cell lines compared with the rest of the cell lines. D, The hypermethylated regions identified in C are represented by the methylation β-value in each cell line.

Close modal

To determine whether promoter methylation was predictive of decitabine sensitivity, CpGs identified within 1,500 bp or 200 bp of the transcription start site (TSS1500 and TSS200, respectively), within the 5′ untranslated region (UTR), or within the first exon were evaluated with the HM450 assay. Overall, promoter DNA methylation was absent in most genes across all cell lines (Fig. 3B) and little differential methylation between decitabine-response groups was observed (Fig. 3C). We noted that 434 genes had z-score > 1 for at least one promoter-associated CpG, with the majority of those hypermethylated promoter genes occurring in the decitabine-resistant MDA-MB-453 cell line (Fig. 3D). However, there are only a few genes that are consistently methylated across the three decitabine-resistant cell lines, and in patient tumors the genes are consistently unmethylated across all patient samples (Supplementary Fig. S9). This suggests that the existence of specific CpG sites that can stratify breast cancers for decitabine response based on hypermethylation likely do not exist. We also performed a principal component analysis of TSS200 and gene body CpGs in the 10 cell lines which shows that the cell lines do not separate according to sensitivity (Supplementary Fig. S10).

Genome-wide and gene-specific demethylation of tumor suppressor genes by decitabine in breast cancer cells is not correlated with decitabine response

Although baseline methylation did not reveal any putative biomarkers, we hypothesized that genomic demethylation in the presence of decitabine may correlate with decitabine response. We analyzed HM450 data from six TNBC cell lines treated with 1 μmol/L decitabine for 72 hours. This revealed that decitabine induced a range of demethylation in the cells; however, the extent of demethylation was not reflective of the decitabine sensitivity in the cell lines. For example, SUM149 cells are generally highly sensitive (Fig. 1B), but on a genome-wide scale had comparatively little demethylation (Fig. 4A), while SUM159 are resistant, and yet comparatively lost their methylation. Hence, although demethylation is associated with decitabine treatment in breast cancer cells, it is not correlative of IC50s determined by the colony assay (Figs. 1 and 4A).

Figure 4.

Breast cancer sensitivity to decitabine (DAC) is not associated decitabine-induced demethylation or by induction of common hypermethylated tumor suppressor genes (e.g., RUNX3). A, DNA methylation as determined via HM450 array among breast cancer cell lines after treatment with 1 μmol/L decitabine for 72 hours. Methylation via HM450 of a promoter-associated CpG island in RUNX3 after treatment with 1 μmol/L decitabine for 72 hours (B) and RUNX3 mRNA levels via qRT-PCR after decitabine treatment (C); error bars, SD; one-way t test (**, P < 0.01; ***, P < 0.001; ns, not significant).

Figure 4.

Breast cancer sensitivity to decitabine (DAC) is not associated decitabine-induced demethylation or by induction of common hypermethylated tumor suppressor genes (e.g., RUNX3). A, DNA methylation as determined via HM450 array among breast cancer cell lines after treatment with 1 μmol/L decitabine for 72 hours. Methylation via HM450 of a promoter-associated CpG island in RUNX3 after treatment with 1 μmol/L decitabine for 72 hours (B) and RUNX3 mRNA levels via qRT-PCR after decitabine treatment (C); error bars, SD; one-way t test (**, P < 0.01; ***, P < 0.001; ns, not significant).

Close modal

The extent of genome-wide demethylation was reflective of gene-specific demethylation of promoter regions and of known tumor suppressors RUNX3 (Fig. 4B), BRCA1, CDH1, and RASSF1 (Supplementary Fig. S11), but not associated with response (Fig. 1). Intriguingly, increased expression of RUNX3 (Fig. 4C) and CDH1 (Supplementary Fig. S11) in many cases was not paired with promoter demethylation (Fig. 4B; Supplementary Fig. S11), suggesting that induced expression of these well-characterized hypermethylated tumor suppressor genes (32–34) by decitabine is not necessarily due to promoter demethylation and may be a result of other effects of decitabine treatment. In validation of the HM40 data, we performed bisulfite pyrosequencing of 32 CpGs in the promoter region of RUNX3 (Supplementary Fig. S12). The sequencing data generally agrees with the HM450 data and demonstrated that the SUM159 promoter is hemimethylated and is overall minimally demethylated by decitabine treatment, while the more methylated MDA-MB-231 RUNX3 promoter contains a cluster of CpG sites around the start codon that are demethylated upon decitabine treatment (CpG site 11–17; Supplementary Fig. S12).

Decitabine upregulates gene expression via demethylation of promoters and induces transcriptional programs for stress, cell-cycle arrest, apoptosis, and immune response

Since our analyses of gene-specific demethylation of tumor suppressor failed to identify correlations with expression and decitabine sensitivity (Fig. 4; Supplementary Figs. S11 and S12) we extended our gene expression analyses genome wide. Three cell lines spanning the range of decitabine sensitivity were assessed for expression changes in the presence of decitabine by microarray. Extensive upregulation of gene expression was observed across all three cell lines while concurrently many genes were downregulated (Fig. 5A; Supplementary Fig. S1; Supplementary Data S2). Interestingly, although the magnitude of induction was variable, there was significant overlap in which transcripts were upregulated between the cell lines, with few differentially regulated genes (Fig. 5A). Independent GSEA of the upregulated versus downregulated transcripts showed that there is significant overlap of the breast cancer decitabine up- or downregulated transcripts with existing azanucleoside-mediated gene expression datasets (Supplementary Data S4). The gene set Kim_Response_to_TSA_and_decitabine is from four glioma cell lines treated with combination decitabine and trichostatin A (histone deacetylase inhibitor; ref. 35), dataset Zhong_response_to_azactidine_and_TSA is from four non–small cell lung cancer cell lines treated with a combination of azacitidine and trichostatin A (36), and dataset Heller_HDAC_targets_silenced_by_methylation is from three multiple myeloma cell lines treated with a combination of azacitidine and trichostatin A (37). If these upregulated transcripts are hypermethylated tumor suppressor genes then we would expect the more decitabine-sensitive cell lines to have stronger induction of these gene sets, but this is not the case as SUM159 (decitabine resistant) has the most extensive gene upregulation (Supplementary Fig. S1).

Figure 5.

Decitabine induces expression of methylated genes, drug response pathways, and immune response pathways genes regardless of decitabine sensitivity. A, Gene expression microarray of cell lines treated with 1 μmol/L decitabine for 72 hours showing genes significantly up- or downregulated (1.5-fold) in any cell line; GSEA showing genes from independent datasets of azanucleoside-treated cancer cells. B, HM450 methylation values of genes for which expression was affected by decitabine treatment in MDA-MB-231 cells significantly up- (426 genes) or downregulated (168 genes; left). Proportion of genes with methylated TSS or exon 1 that are up- or downregulated in expression after decitabine treatment (right). C, HM450 methylation on genes for which decitabine affects expression in MDA-MB-231; change in methylation after 72 hours decitabine treatment (DAC-NT) versus predecitabine methylation (NT). D, Top 100 enriched gene sets in the unmethylated (TSS/Exon1) decitabine upregulated genes based on GSEA; gene included in a given pathway indicated by color.

Figure 5.

Decitabine induces expression of methylated genes, drug response pathways, and immune response pathways genes regardless of decitabine sensitivity. A, Gene expression microarray of cell lines treated with 1 μmol/L decitabine for 72 hours showing genes significantly up- or downregulated (1.5-fold) in any cell line; GSEA showing genes from independent datasets of azanucleoside-treated cancer cells. B, HM450 methylation values of genes for which expression was affected by decitabine treatment in MDA-MB-231 cells significantly up- (426 genes) or downregulated (168 genes; left). Proportion of genes with methylated TSS or exon 1 that are up- or downregulated in expression after decitabine treatment (right). C, HM450 methylation on genes for which decitabine affects expression in MDA-MB-231; change in methylation after 72 hours decitabine treatment (DAC-NT) versus predecitabine methylation (NT). D, Top 100 enriched gene sets in the unmethylated (TSS/Exon1) decitabine upregulated genes based on GSEA; gene included in a given pathway indicated by color.

Close modal

To determine whether there is direct upregulation of gene expression via demethylation of promoter regions, HM450 methylation β-values from the genes that were up- or downregulated by decitabine in all three cell lines were examined (Supplementary Data S1). More often, the TSS1500, TSS200, 5′ UTR, and Exon1 regions were methylated in upregulated genes compared with downregulated genes in MDA-MB-231 (Fig. 5B), MDA-MB-468 (Supplementary Fig. S13), and SUM159 cells (Supplementary Fig. S14). As expected, there was general demethylation of these regions in upregulated genes after decitabine treatment in MDA-MB-231 (Fig. 5C), MDA-MB-468, and SUM159 cells (Supplementary Fig. S15). Intragenic (especially gene body) DNA methylation may serve as a positive regulator of transcription, where loss of methylation inhibits gene expression (38). This may explain why downregulated genes were more likely to have methylation of the 3′ UTR and gene body compared with upregulated genes (Supplementary Fig. S16).

Genes that were confirmed to have hypermethylated promoter regions only constituted 16%–28% of total decitabine-upregulated genes (Fig. 5B), leaving the majority of upregulated genes without a clear cause for upregulation (Supplementary Table S3). Direct gene expression changes induced by demethylation are well-characterized outcomes of decitabine treatment; however, the indirect gene expression changes that are not readily explained by methylation changes are not as well understood. These indirect changes may be due to the demethylation of transcription factors or the epigenetic resurrection of other upstream signaling pathways. In a GSEA of the decitabine-upregulated genes with unmethylated transcription start site (TSS) and Exon1 in each cell line, all three cell lines showed upregulation of genes associated with cell-cycle arrest, cell death, DNA damage, stress, immune response, and transcriptional machinery (Fig. 5D).

Decitabine induces viral mimicry response in breast cancer cells, but dsRNA sensors are not required for in vitro sensitivity

Among the most consistently upregulated pathways by decitabine in the cell lines were immune-related pathways (Fig. 5D). Consistent with this, IFN-inducible 2′-5′-oligoadenylate synthetase-like (OASL) was upregulated in MDA-MB-231 cells upon decitabine treatment (Supplementary Data S2 and S4). Recent reports have highlighted the requirement for the viral mimicry response (IFN-induced OASL is part of this response) for decitabine sensitivity for in vitro and in vivo growth inhibition of colorectal and ovarian cancer cells (11, 12). In those studies, demethylation by decitabine leads to reexpression of epigenetically silenced endogenous retroviral elements (ERV), leading to a cascade of events that result in induction of the IFN response (Fig. 6A). We assessed the level of induction of previously described mediators of the decitabine-induced viral mimicry response in the breast cancer cell lines. We assessed expression of the ERVs in our cell lines utilizing a panel of previously described and newly designed primers (Supplementary Table S2). We were unable to detect expression of many of these transcripts in the 10 breast cancer cell lines (pooled cDNA sample of all cell lines no treatment and decitabine treated; Supplementary Table S4). Decitabine-induced expression of ERVs MLT1C49 or MLT2B4 was observed in some of the cell lines (e.g., SKRBR3 and MDA-MB-436) treated with 1 μmol/L decitabine for 72 hours; however, induction was inconsistent and did not correlate with sensitivity (Fig. 6B). We further explored the induction of ERVs EnvE, ERVFRD1, and the 5′ UTR region of HERV-K in decitabine-sensitive MDA-MB-468 cells, intermediate-sensitive MDA-MB-231 cells, and resistant SUM159 cells treated with their cell confluency IC50s of decitabine. This again resulted in limited induction of these ERVs, which did not correlate with decitabine sensitivities (Supplementary Fig. S17). Of note, these ERVs were also not induced in the MDA-MB-468 tumors that had been treated with decitabine (Fig. 1B; Supplementary Fig. S17).

Figure 6.

Decitabine induces expression of the dsRNA pathway but it is not essential for decitabine sensitivity. dsRNA “viral mimicry” pathway components (A) show increased gene expression via qRT-PCR in breast cancer cell lines after 1 μmol/L decitabine treatment for 72 hours; error bars, SD; one-sample t test (B). shRNA-mediated knockdown (KD) of MDA5 (IFIH1; C) or RIGI (DDX58; D) in MDA-MB-468 cells does not affect in vitro decitabine sensitivity in colony-forming assay. error bars, SD; ANOVA with Dunnett post hoc test (*, P < 0.05; **, P < 0.01; ***, P < 0.001).

Figure 6.

Decitabine induces expression of the dsRNA pathway but it is not essential for decitabine sensitivity. dsRNA “viral mimicry” pathway components (A) show increased gene expression via qRT-PCR in breast cancer cell lines after 1 μmol/L decitabine treatment for 72 hours; error bars, SD; one-sample t test (B). shRNA-mediated knockdown (KD) of MDA5 (IFIH1; C) or RIGI (DDX58; D) in MDA-MB-468 cells does not affect in vitro decitabine sensitivity in colony-forming assay. error bars, SD; ANOVA with Dunnett post hoc test (*, P < 0.05; **, P < 0.01; ***, P < 0.001).

Close modal

The general response to ERVs is better detected by assessing levels of dsRNA recognition pattern receptors, melanoma differentiation–associated protein (MDA5, encoded by IFIH1), and retinoic acid-inducible gene I (RIG-I, encoded by DDX58) after decitabine treatment. RIG-I and/or MDA5 were induced by decitabine in most of the cell lines (Fig. 6B). Downstream effectors such as mitochondrial antiviral signaling protein (MAVS), transcriptional activator of IFN, interferon regulatory factor 7 (IRF7), anti-viral OASL, and IFN response mediator IFN-stimulated gene 15 (ISG15) were also generally induced in the cell lines, but did not correlate with sensitivity (Fig. 6B). We also utilized our gene expression microarray data to evaluate a larger panel of reported IFN-stimulated genes (Supplementary Table S5) in decitabine-treated MDA-MB-468, MDA-MB-231, and SUM159 cells (Supplementary Fig. S18). These analyses suggest two points; (i) that strong induction of IFN-stimulated genes is not associated with increased sensitivity to decitabine, and (ii) that decitabine-induced expression of IFN-stimulated genes is not due a direct demethylation of the promoter regions of these genes. Please note that the HM450 methylation data revealed that the most strongly induced genes (DDX58, HPSE, IFI6, IFIT1, IFIT3, OAS3, and ZC3HAV1) have an unmethylated promoter region (TSS200/1500; Supplementary Data S1; Supplementary Fig. S19); therefore it is unlikely that the increased expression is a result of decitabine directly demethylating these genes. Regardless, the induction of some of these genes is consistent with a viral mimicry response being induced through increased MDA5/RIG-I in the breast cancer cells.

Via knockdown, MDA5 has been shown to be an essential mediator of the viral mimicry response induced by decitabine in colorectal cancer (12). We therefore knocked down MDA5 in the most decitabine-sensitive MDA-MB-468 cells (Fig. 6C) to determine, whether, like DCK knockdown (Fig. 2), this would render the cells less sensitive to decitabine. However, in contrast to our results with DCK (Fig. 2), knockdown of MDA5 did not make MDA-MB-468 cells resistant to decitabine (Fig. 6C). Similarly, RIG-I knockdown did not make the cells more resistant to decitabine (Fig. 6D). Together these data suggest that while the viral mimicry response is generally induced in breast cancer cells, it does not appear to be the key determinant of decitabine sensitivity in the cells in vitro.

High levels of exporter ABCB1 does not limit decitabine response in breast cancer cells

A consideration for the patients with TNBC receiving decitabine in clinical settings is potential presence of multidrug resistance acquired during initial chemotherapy treatment. Increased expression of exporter multidrug resistance gene ABCB1 is a common resistance mechanism for many drugs (39, 40). Gene expression from 56 matched breast tumor samples pre- and post-chemotherapy treatment showed that ABCB1 expression was elevated post-chemotherapy in patients with at least partial response to treatment (Fig. 7A). This suggests that the effect of ABCB1 expression in breast cancer needs to be assessed for its impact on decitabine response.

Figure 7.

ABCB1 causes resistance to taxanes but has minimal effects on decitabine sensitivity. A, ABCB1 expression via microarray in GSE28844 dataset of patients with breast cancer treated with anthracyclines and taxanes (paired two-tailed t test). B, Expression of ABCB1 across breast cancer cell lines by qRT-PCR. C, shRNA-mediated ABCB1 knockdown (KD) and in vitro decitabine sensitivity; ANOVA with Dunnett post hoc test. ABCB1 expression via qRT-PCR in taxol-resistant MDA-MB-231 cells; one-tailed t test. Colony forming IC50 assay in taxol-resistant MDA-MB-231 treated with ABCB1 inhibitor verapamil and paclitaxel (D) or decitabine (E). ANOVA with Tukey post hoc (*, P < 0.05; **, P < 0.01; ***, P < 0.001).

Figure 7.

ABCB1 causes resistance to taxanes but has minimal effects on decitabine sensitivity. A, ABCB1 expression via microarray in GSE28844 dataset of patients with breast cancer treated with anthracyclines and taxanes (paired two-tailed t test). B, Expression of ABCB1 across breast cancer cell lines by qRT-PCR. C, shRNA-mediated ABCB1 knockdown (KD) and in vitro decitabine sensitivity; ANOVA with Dunnett post hoc test. ABCB1 expression via qRT-PCR in taxol-resistant MDA-MB-231 cells; one-tailed t test. Colony forming IC50 assay in taxol-resistant MDA-MB-231 treated with ABCB1 inhibitor verapamil and paclitaxel (D) or decitabine (E). ANOVA with Tukey post hoc (*, P < 0.05; **, P < 0.01; ***, P < 0.001).

Close modal

We assessed expression of ABCB1 across the 10 breast cancer cell lines, and noted increased expression in decitabine-resistant SUM159 and Hs578T cells, but also decitabine-sensitive MDA-MB-468 cells (Fig. 7B). This may suggest that increased ABCB1 is a potential mechanism of decitabine resistance in breast cancer; however, knockdown of ABCB1 in decitabine-sensitive MDA-MB-468 cells had minimal impact on decitabine sensitivity (Fig. 7C).

To further assess how multidrug resistance impacts decitabine response, we used a variant of the MDA-MB-231 cell line with acquired taxol- and anthracycline-resistance and ABCB1 overexpression (26). We confirmed the resistant cell line had increased ABCB1 expression, was significantly more resistant to paclitaxel than the control cell line, and that paclitaxel resistance in the cells could be reduced by the addition of the ABCB1 inhibitor verapamil (Fig. 7D). Importantly, the taxol-resistant MDA-MB-231 cells were not resistant to decitabine and ABCB1 inhibition did not sensitize cells to decitabine, implying that ABCB1 does not play a major role in decitabine response (Fig. 7E). Therefore, increased expression of ABCB1 in the tumors of patients with breast cancer that will receive decitabine after having received chemotherapy will not be a significant barrier to decitabine response.

Our comprehensive analysis of decitabine response in breast cancer cells screened many of the potential mechanisms of decitabine to determine whether essential factors and/or predictors of sensitivity could be identified that may impact the success of ongoing clinical trials in TNBCs. Together, this study revealed a range of responses to decitabine in breast cancer that could not be predicted on the basis of commonly proposed mechanisms (e.g., demethylation of tumor suppressor genes and viral mimicry response). Our transcriptome and methylome analyses also demonstrated that while decitabine induces genome-wide expression changes and demethylation, these effects are not necessarily paired and do not correlate with sensitivity. Instead, it became apparent that the many gene expression changes induced by decitabine are either an indirect result of its demethylation and/or a consequence of cell death, DNA damage, and immune responses being induced. Notably, we describe a requirement for DCK, the decitabine processing enzyme, for decitabine-induced growth suppression and gene expression changes. DCK is comparatively abundant in TNBC and is increased post-chemotherapy in nonresponding patients. This, combined with our demonstration that multidrug-resistant/ABCB1-overexpressing breast cancer cells remain sensitive to decitabine, suggests that decitabine may be useful as part of a second-line therapy regimen for patients with TNBC.

The initial enthusiasm for hypomethylating agents (e.g., decitabine and azacytidine) to treat MDS has been tempered by years of clinical use, which indicate that fewer than half of patients maintain a response to the therapy (41). This is also true for chronic monomyelocytic leukemia (CMML, for which azacytidine and decitabine are the only approved drugs), where overall response rates hover at 40% and complete response at <20% (42). As an AML therapy, overall response to decitabine is also approximately 40%, although there is some encouraging data to suggest a precision medicine approach could work for treating older patients with AML with decitabine. For example, DNMT3A mutations, TET2 mutations, and IDH1/2 mutations have all been associated with favorable outcomes in AML (and some MDS cases; refs. 43–45). However, similar stratification strategies are unlikely to work in breast cancer and other solid tumors; because mutations in epigenetic machinery genes are far less common (21, 22). Regardless, even for these blood malignancies/disorders there are many other clinical factors such as age, cytogenetics, prior treatment regimen, global or gene-specific methylation, and gene expression patterns that influence response to decitabine (46). While there is compelling evidence for each of these factors individually, most recent studies cannot form a cohesive molecular profile of favorable decitabine response for AML, MDS, or CMML. As such, our gene-specific and genome-wide transcription and methylation results in breast cancer are unsurprising and suggest that predicting decitabine response will be multi-factorial and decitabine likely be used in combination with other drugs.

Potential rational combinations include using DNA hypomethylating agents followed by immunotherapeutics such as checkpoint inhibitors. This strategy should increase the immunogenicity of tumors by inducing expression of hypermethylated cancer testis antigens (and HERVs), and concomitantly inhibit the downregulation of elicited antitumor cytotoxic T-cell responses by blocking PD-1–PD-L1 or CTLA4–CD80/86 interactions. This is the intended strategy for an ongoing clinical trial for TNBC (NCT02957968), where patients are treated sequentially with decitabine plus checkpoint inhibitor pembrolizumab (anti-PD-1), followed by standard chemotherapy.

In these immunotherapeutic-based combination approaches, it is also important to consider how decitabine affects PD-L1 expression, and in this regard the data are somewhat contradictory. In murine solid tumor studies, decitabine reduced PD-L1 levels on tumor cells and subsequently improved infiltration of cytotoxic CD8+ T cells and enhanced anti-PD-1 immunotherapy (47–50). In another study examining DNA methylation and PD-L1 expression patterns in melanoma cell lines and TCGA melanoma patient data, decitabine induced expression of PD-L1 in cells, and in hypomethylated treatment-naïve patient tumors, there was strong expression of “viral mimicry” HERVs and higher PD-L1 (51). However, decitabine treatment also increased PD-L1 levels in patients with MDS, CMML, and AML (52). Importantly, patients with MDS with the most elevated PD-L1 post-decitabine were also the most resistant to decitabine therapy. Together, these studies underlie the importance of determining how decitabine affects immunogenicity in the clinic versus in preclinical animal studies, and that these effects may also be cancer type specific.

On the basis of the mixed results of these well-established immunogenic pathways, we should be hopeful but cautious when interpreting any viral mimicry responses induced by decitabine. We see a clear induction of the viral mimicry response genes (e.g., OASL and ISG15) in breast cancer cells; however, this was not paired with a favorable decitabine response. The direct cytotoxicity of MDA5-mediated viral mimicry that was observed in colorectal cancer cells (12) was not replicated in our MDA5-knockdown breast cancer cells. Importantly, we did not assess the immunogenicity of decitabine treatment; hence, it is possible that while induction of this pathway, while not directly cytotoxic in breast cancer cells, may be important for inducing antitumor effects in immunocompetent models and patients.

No potential conflicts of interest were disclosed.

Conception and design: M.L. Dahn, B.M. Cruickshank, I.C.G. Weaver, P. Marcato

Development of methodology: M.L. Dahn, I.C.G. Weaver, P. Marcato

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): M.L. Dahn, B.M. Cruickshank, A.J. Jackson, C. Dean, R.W. Holloway, S.R. Hall, K.M. Coyle, H. Maillet, D.M. Waisman, K.B. Goralski, I.C.G. Weaver, P. Marcato

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): M.L. Dahn, B.M. Cruickshank, A.J. Jackson, C. Dean, S.R. Hall, P. Marcato

Writing, review, and/or revision of the manuscript: M.L. Dahn, S.R. Hall, K.M. Coyle, D.M. Waisman, K.B. Goralski, P. Marcato

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): M.L. Dahn, B.M. Cruickshank

Study supervision: M.L. Dahn, C.A. Giacomantonio, P. Marcato

Support was provided by grant funding to P. Marcato by the Cancer Research Society in partnership with the Institute of Cancer Research of the Canadian Institutes of Health Research (CIHR; grant number 22185). In addition, support was provided by grant funding to I.C.G. Weaver from the Natural Sciences and Engineering Research Council of Canada (RGPIN-2013-436204). M.L. Dahn was supported by CGS-D award from the CIHR, a Nova Scotia Health Research Foundation studentship, an NS graduate scholarship, and a Killam Laureate scholarship. B.M. Cruickshank was supported by Nova Scotia Research and Innovation Graduate and Killam Laureate scholarships. K.M. Coyle was supported by a CGS-D award from CIHR and by the DeWolfe Graduate Award from the Dalhousie Medical Research Foundation, and a studentship from the Beatrice Hunter Cancer Research Institute and the Canadian Imperial Bank of Commerce. The results published here are in part based upon data generated by the TCGA Research Network: https://www.cancer.gov/tcga. The shRNA knockdown clones were obtained by accessing Dalhousie University's Faculty of Medicine Gene Analysis & Discovery Core Facility.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

1.
Sharma
S
,
Kelly
TK
,
Jones
PA
. 
Epigenetics in cancer
.
Carcinogenesis
2010
;
31
:
27
36
.
2.
Figueroa
ME
,
Abdel-Wahab
O
,
Lu
C
,
Ward
PS
,
Patel
J
,
Shih
A
, et al
Leukemic IDH1 and IDH2 mutations result in a hypermethylation phenotype, disrupt TET2 function, and impair hematopoietic differentiation
.
Cancer Cell
2010
;
18
:
553
67
.
3.
Connolly
R
,
Stearns
V
. 
Epigenetics as a therapeutic target in breast cancer
.
J Mammary Gland Biol Neoplasia
2012
;
17
:
191
204
.
4.
Peddi
PF
,
Ellis
MJ
,
Ma
C
. 
Molecular basis of triple negative breast cancer and implications for therapy
.
Int J Breast Cancer
2012
;
2012
:
217185
.
5.
Shipitsin
M
,
Campbell
LL
,
Argani
P
,
Weremowicz
S
,
Bloushtain-Qimron
N
,
Yao
J
, et al
Molecular definition of breast tumor heterogeneity
.
Cancer Cell
2007
;
11
:
259
73
.
6.
Gore
SD
,
Baylin
S
,
Sugar
E
,
Carraway
H
,
Miller
CB
,
Carducci
M
, et al
Combined DNA methyltransferase and histone deacetylase inhibition in the treatment of myeloid neoplasms
.
Cancer Res
2006
;
66
:
6361
9
.
7.
Lyko
F
,
Brown
R
. 
DNA methyltransferase inhibitors and the development of epigenetic cancer therapies
.
J Natl Cancer Inst
2005
;
97
:
1498
506
.
8.
Costa
G
,
Barra
V
,
Lentini
L
,
Cilluffo
D
,
Di Leonardo
A
. 
DNA demethylation caused by 5-Aza-2′-deoxycytidine induces mitotic alterations and aneuploidy
.
Oncotarget
2016
;
7
:
3726
39
.
9.
Goodyear
O
,
Agathanggelou
A
,
Novitzky-Basso
I
,
Siddique
S
,
McSkeane
T
,
Ryan
G
, et al
Induction of a CD8+ T-cell response to the MAGE cancer testis antigen by combined treatment with azacitidine and sodium valproate in patients with acute myeloid leukemia and myelodysplasia
.
Blood
2010
;
116
:
1908
18
.
10.
Almstedt
M
,
Blagitko-Dorfs
N
,
Duque-Afonso
J
,
Karbach
J
,
Pfeifer
D
,
Jäger
E
, et al
The DNA demethylating agent 5-aza-2′-deoxycytidine induces expression of NY-ESO-1 and other cancer/testis antigens in myeloid leukemia cells
.
Leuk Res
2010
;
34
:
899
905
.
11.
Chiappinelli
KB
,
Strissel
PL
,
Desrichard
A
,
Li
H
,
Henke
C
,
Akman
B
, et al
Inhibiting DNA methylation causes an interferon response in cancer via dsRNA including endogenous retroviruses
.
Cell
2015
;
162
:
974
86
.
12.
Roulois
D
,
Loo Yau
H
,
Singhania
R
,
Wang
Y
,
Danesh
A
,
Shen
SY
, et al
DNA-demethylating agents target colorectal cancer cells by inducing viral mimicry by endogenous transcripts
.
Cell
2015
;
162
:
961
973
.
13.
Tsai
HC
,
Li
H
,
Van Neste
L
,
Cai
Y
,
Robert
C
,
Rassool
FV
, et al
Transient low doses of DNA-demethylating agents exert durable antitumor effects on hematological and epithelial tumor cells
.
Cancer Cell
2012
;
21
:
430
46
.
14.
Abáigar
M
,
Ramos
F
,
Benito
R
,
Díez-Campelo
M
,
Sánchez-del-Real
J
,
Hermosín
L
, et al
Prognostic impact of the number of methylated genes in myelodysplastic syndromes and acute myeloid leukemias treated with azacytidine
.
Ann Hematol
2013
;
92
:
1543
52
.
15.
Traina
F
,
Visconte
V
,
Elson
P
,
Tabarroki
A
,
Jankowska
AM
,
Hasrouni
E
, et al
Impact of molecular mutations on treatment response to DNMT inhibitors in myelodysplasia and related neoplasms
.
Leukemia
2014
;
28
:
78
87
.
16.
Shen
L
,
Kantarjian
H
,
Guo
Y
,
Lin
E
,
Shan
J
,
Huang
X
, et al
DNA methylation predicts survival and response to therapy in patients with myelodysplastic syndromes
.
J Clin Oncol
2010
;
28
:
605
13
.
17.
Qin
T
,
Castoro
R
,
El Ahdab
S
,
Jelinek
J
,
Wang
X
,
Si
J
, et al
Mechanisms of resistance to decitabine in the myelodysplastic syndrome
.
PLoS One
2011
;
6
:
e23372
.
18.
Damaraju
VL
,
Mowles
D
,
Yao
S
,
Ng
A
,
Young
JD
,
Cass
CE
, et al
Role of human nucleoside transporters in the uptake and cytotoxicity of azacitidine and decitabine
.
Nucleosides Nucleotides Nucleic Acids
2012
;
31
:
236
55
.
19.
Rius
M
,
Stresemann
C
,
Keller
D
,
Brom
M
,
Schirrmacher
E
,
Keppler
D
, et al
Human concentrative nucleoside transporter 1-mediated uptake of 5-azacytidine enhances DNA demethylation
.
Mol Cancer Ther
2009
;
8
:
225
31
.
20.
Treppendahl
MB
,
Kristensen
LS
,
Grønbæk
K
. 
Predicting response to epigenetic therapy
.
J Clin Invest
2014
;
124
:
47
55
.
21.
Cerami
E
,
Gao
J
,
Dogrusoz
U
,
Gross
BE
,
Sumer
SO
,
Aksoy
BA
, et al
The cBio cancer genomics portal: an open platform for exploring multidimensional cancer genomics data: figure 1
.
Cancer Discov
2012
;
2
:
401
4
.
22.
Gao
J
,
Aksoy
BA
,
Dogrusoz
U
,
Dresdner
G
,
Gross
B
,
Sumer
SO
, et al
Integrative analysis of complex cancer genomics and clinical profiles using the cBioPortal
.
Sci Signal
2013
;
6
:
pl1
.
23.
Vera-Ramirez
L
,
Sanchez-Rovira
P
,
Ramirez-Tortosa
CL
,
Quiles
JL
,
Ramirez-Tortosa
M
,
Lorente
JA
. 
Transcriptional shift identifies a set of genes driving breast cancer chemoresistance
.
PLoS One
2013
;
8
:
e53983
.
24.
Coyle
KM
,
Murphy
JP
,
Vidovic
D
,
Vaghar-Kashani
A
,
Dean
CA
,
Sultan
M
, et al
Breast cancer subtype dictates DNA methylation and ALDH1A3-mediated expression of tumor suppressor RARRES1
.
Oncotarget
2016
;
7
:
44096
112
.
25.
Coyle
KM
,
Maxwell
S
,
Thomas
ML
,
Marcato
P
. 
Profiling of the transcriptional response to all-trans retinoic acid in breast cancer cells reveals RARE-independent mechanisms of gene expression
.
Sci Rep
2017
;
7
:
16684
.
26.
Hall
SR
,
Toulany
J
,
Bennett
LG
,
Martinez-Farina
CF
,
Robertson
AW
,
Jakeman
DL
, et al
Jadomycins inhibit type II topoisomerases and promote DNA damage and apoptosis in multidrug-resistant triple-negative breast cancer cells
.
J Pharmacol Exp Ther
2017
;
363
:
196
210
.
27.
Pastor-Anglada
M
,
Pérez-Torras
S
. 
Nucleoside transporter proteins as biomarkers of drug responsiveness and drug targets
.
Front Pharmacol
2015
;
6
:
13
.
28.
Momparler
RL
. 
Pharmacology of 5-Aza-2′-deoxycytidine (decitabine)
.
Semin Hematol
2005
;
42
:
S9
16
.
29.
Ebrahem
Q
,
Mahfouz
R
,
Durkin
L
,
Sekeres
M
,
Advani
AS
,
Carraway
H
, et al
Mechanisms of resistance to 5-azacytidine/decitabine in MDS-AML and pre-clinical in vivo proof of principle of rational solutions to extend response
.
Blood
2015
;
126
:
678
.
30.
Esteller
M
,
Herman
JG
. 
Cancer as an epigenetic disease: DNA methylation and chromatin alterations in human tumours
.
J Pathol
2002
;
196
:
1
7
.
31.
Kohli
RM
,
Zhang
Y
. 
TET enzymes, TDG and the dynamics of DNA demethylation
.
Nature
2013
;
502
:
472
9
.
32.
Song
X-Y
,
Li
B-Y
,
Zhou
E-X
,
Wu
F-X
. 
The clinicopathological significance of RUNX3 hypermethylation and mRNA expression in human breast cancer, a meta-analysis
.
Onco Targets Ther
2016
;
9
:
5339
47
.
33.
Esteller
M
,
Silva
JM
,
Dominguez
G
,
Bonilla
F
,
Matias-Guiu
X
,
Lerma
E
, et al
Promoter hypermethylation and BRCA1 inactivation in sporadic breast and ovarian tumors
.
J Natl Cancer Inst
2000
;
92
:
564
9
.
34.
Shargh
SA
,
Sakizli
M
,
Khalaj
V
,
Movafagh
A
,
Yazdi
H
,
Hagigatjou
E
, et al
Downregulation of E-cadherin expression in breast cancer by promoter hypermethylation and its relation with progression and prognosis of tumor
.
Med Oncol
2014
;
31
:
250
.
35.
Kim
T-Y
,
Zhong
S
,
Fields
CR
,
Kim
JH
,
Robertson
KD
. 
Epigenomic profiling reveals novel and frequent targets of aberrant DNA methylation-mediated silencing in malignant glioma
.
Cancer Res
2006
;
66
:
7490
501
.
36.
Zhong
S
,
Fields
CR
,
Su
N
,
Pan
Y-X
,
Robertson
KD
. 
Pharmacologic inhibition of epigenetic modifications, coupled with gene expression profiling, reveals novel targets of aberrant DNA methylation and histone deacetylation in lung cancer
.
Oncogene
2007
;
26
:
2621
34
.
37.
Heller
G
,
Schmidt
WM
,
Ziegler
B
,
Holzer
S
,
Müllauer
L
,
Bilban
M
, et al
Genome-wide transcriptional response to 5-aza-2′-deoxycytidine and trichostatin a in multiple myeloma cells
.
Cancer Res
2008
;
68
:
44
54
.
38.
Yang
X
,
Han
H
,
De Carvalho
DD
,
Lay
FD
,
Jones
PA
,
Liang
G
. 
Gene body methylation can alter gene expression and is a therapeutic target in cancer
.
Cancer Cell
2014
;
26
:
577
90
.
39.
Trock
BJ
,
Leonessa
F
,
Clarke
R
. 
Multidrug resistance in breast cancer: a meta-analysis of MDR1/gp170 expression and its possible functional significance
.
J Natl Cancer Inst
1997
;
89
:
917
31
.
40.
Burger
H
,
Foekens
JA
,
Look
MP
,
Meijer-van Gelder
ME
,
Klijn
JGM
,
Wiemer
EAC
, et al
RNA Expression of Breast Cancer Resistance Protein, Lung Resistance-related Protein, Multidrug Resistance-associated Proteins 1 and 2, and Multidrug Resistance Gene 1 in Breast Cancer: Correlation with Chemotherapeutic Response
.
Clin Cancer Res
2003
;
4
:
1533
42
.
41.
Santini
V
. 
How I treat MDS after hypomethylating agent failure
.
Blood
2019
;
133
:
521
9
.
42.
Coston
T
,
Pophali
P
,
Vallapureddy
R
,
Lasho
TL
,
Finke
CM
,
Ketterling
RP
, et al
Suboptimal response rates to hypomethylating agent therapy in chronic myelomonocytic leukemia; a single institutional study of 121 patients
.
Am J Hematol
2019
;
94
:
767
79
.
43.
Du
M
,
Zhou
F
,
Jin
R
,
Hu
Y
,
Mei
H
. 
Mutations in the DNA methylation pathway predict clinical efficacy to hypomethylating agents in myelodysplastic syndromes: a meta-analysis
.
Leuk Res
2019
;
80
:
11
8
.
44.
Bejar
R
,
Lord
A
,
Stevenson
K
,
Bar-Natan
M
,
Pérez-Ladaga
A
,
Zaneveld
J
, et al
TET2 mutations predict response to hypomethylating agents in myelodysplastic syndrome patients
.
Blood
2014
;
124
:
2705
12
.
45.
Voso
MT
,
Niscola
P
,
Piciocchi
A
,
Fianchi
L
,
Maurillo
L
,
Musto
P
, et al
Standard dose and prolonged administration of azacitidine are associated with improved efficacy in a real-world group of patients with myelodysplastic syndrome or low blast count acute myeloid leukemia
.
Eur J Haematol
2016
;
96
:
344
51
.
46.
Wang
H
,
Li
Y
,
Lv
N
,
Li
Y
,
Wang
L
,
Yu
L
. 
Predictors of clinical responses to hypomethylating agents in acute myeloid leukemia or myelodysplastic syndromes
.
Ann Hematol
2018
;
97
:
2025
38
.
47.
De Beck
L
,
Melhaoui
S
,
De Veirman
K
,
Menu
E
,
De Bruyne
E
,
Vanderkerken
K
, et al
Epigenetic treatment of multiple myeloma mediates tumor intrinsic and extrinsic immunomodulatory effects
.
Oncoimmunology
2018
;
7
:
e1484981
.
48.
Yu
G
,
Wu
Y
,
Wang
W
,
Xu
J
,
Lv
X
,
Cao
X
, et al
Low-dose decitabine enhances the effect of PD-1 blockade in colorectal cancer with microsatellite stability by re-modulating the tumor microenvironment
.
Cell Mol Immunol
2019
;
16
:
401
9
.
49.
Stone
ML
,
Chiappinelli
KB
,
Li
H
,
Murphy
LM
,
Travers
ME
,
Topper
MJ
, et al
Epigenetic therapy activates type I interferon signaling in murine ovarian cancer to reduce immunosuppression and tumor burden
.
Proc Natl Acad Sci U S A
2017
;
114
:
E10981
90
.
50.
Lucarini
V
,
Buccione
C
,
Ziccheddu
G
,
Peschiaroli
F
,
Sestili
P
,
Puglisi
R
, et al
Combining type I interferons and 5-Aza-2′-deoxycitidine to improve anti-tumor response against melanoma
.
J Invest Dermatol
2017
;
137
:
159
69
.
51.
Chatterjee
A
,
Rodger
EJ
,
Ahn
A
,
Stockwell
PA
,
Parry
M
,
Motwani
J
, et al
Marked global DNA hypomethylation is associated with constitutive PD-L1 expression in melanoma
.
iScience
2018
;
4
:
312
25
.
52.
Yang
H
,
Bueso-Ramos
C
,
DiNardo
C
,
Estecio
MR
,
Davanlou
M
,
Geng
QR
, et al
Expression of PD-L1, PD-L2, PD-1 and CTLA4 in myelodysplastic syndromes is enhanced by treatment with hypomethylating agents
.
Leukemia
2014
;
28
:
1280
8
.