Abstract
Tumor progression upon treatment arises from preexisting resistant cancer cells and/or adaptation of persister cancer cells committing to an expansion phase. Here, we show that evasion from viral mimicry response allows the growth of taxane-resistant triple-negative breast cancer (TNBC). This is enabled by an epigenetic state adapted to taxane-induced metabolic stress, where DNA hypomethylation over loci enriched in transposable elements (TE) is compensated by large chromatin domains of H3K27me3 to warrant TE repression. This epigenetic state creates a vulnerability to epigenetic therapy against EZH2, the H3K27me3 methyltransferase, which alleviates TE repression in taxane-resistant TNBC, leading to double-stranded RNA production and growth inhibition through viral mimicry response. Collectively, our results illustrate how epigenetic states over TEs promote cancer progression under treatment and can inform about vulnerabilities to epigenetic therapy.
Drug-resistant cancer cells represent a major barrier to remission for patients with cancer. Here we show that drug-induced metabolic perturbation and epigenetic states enable evasion from the viral mimicry response induced by chemotherapy in TNBC. These epigenetic states define a vulnerability to epigenetic therapy using EZH2 inhibitors in taxane-resistant TNBC.
See related commentary by Janin and Esteller, p. 1258.
This article is highlighted in the In This Issue feature, p. 1241
Introduction
Intrinsic or acquired therapeutic resistance represents a major barrier in the treatment of cancer and results in drug-tolerant cell populations with the capacity to evade selective drug pressure (1). Resistance to cytotoxic drugs in diverse cancer types can be enabled by metabolic rewiring. Differential metabolite channeling and dependencies, rewired energy metabolism, and improved cellular antioxidant capacity can provide metabolic states allowing cancer cells to withstand drug-imposed stress (2). Metabolic adaptations can impose long-lasting effects in cancer cells because metabolites serve as substrates or cofactors for epigenetic enzymes (3). In particular, changes in the levels of S-adenosylmethionine (SAM), the universal methyl donor synthesized from the essential amino acid methionine, can alter DNA and histone methylation in tumors (3, 4). Epigenetic modifications such as DNA methylation and histone modifications directly regulate cell identity by controlling the accessibility of chromatin to the transcriptional machinery (5). Metabolic perturbations can thus favor transitions in epigenetic states, a feature of cancer initiation (6) and progression (7).
Reprogramming and dysregulation of epigenetic processes affecting cell identity have been implicated as drivers of carcinogenesis (8, 9). Alterations in histone modifications reflective of new cis-regulatory landscapes also typify cancer progression to drug resistance (10). This is exemplified by changes in the distribution of H3 lysine 4 dimethylation (H3K4me2) and lysine 36 trimethylation (H3K36me3) in endocrine therapy–resistant luminal breast cancer cells (11) and by the formation of repressed chromatin states characterized by increased methylation of H3 lysines 9 and 27 (H3K9 and H3K27) in drug-tolerant cancer cells (12).
Epigenetic regulation of repetitive DNA sequences is required to control the activation of transposable elements (TE) that can create new oncogene cis-regulatory units in transformed cells (13, 14) or generate double-stranded RNA (dsRNA) inducing an IFN-mediated viral mimicry response (15, 16). Epigenetic regulation contributes to the tight balance between TE expression and repression in cancer cells to confer fitness advantage while preventing deleterious effects of uncontrolled TE activation (13). As a consequence, TE repression was shown to allow for immune escape in leukemic and colorectal cancer cells (17) and survival of stress-tolerant lung cancer cell subpopulations (12).
Triple-negative breast cancer (TNBC) is an aggressive form of breast cancer accounting for 15% to 20% of all yearly diagnosed cases. Cytotoxic chemotherapy, including taxanes, remains the only approved systemic therapy for this disease (18). Despite initial benefits of cytotoxic chemotherapy, clinical drug resistance is a major problem for patients with TNBC, causing early and aggressive metastatic dissemination leading to rapid disease progression. Here we report on the interplay between drug-induced metabolic perturbation and epigenetic states that enables evasion from the viral mimicry response induced by chemotherapy, and we show that it creates a targetable epigenetic vulnerability in taxane-resistant TNBC.
Results
Methionine Metabolism Is Altered in Taxane-Resistant TNBC
We generated six independent paclitaxel-resistant models from MDA-MB-436 and Hs 578T TNBC cell lines (resistant populations ResA, ResB, and ResC) upon chronic exposure to increasing concentrations of taxanes (Fig. 1A and B; Supplementary Fig. S1A and S1B). We first examined the metabolic changes in taxane-resistant TNBC cells by metabolomics profiling using liquid chromatography-high resolution mass spectrometry (LC-HRMS) of intracellular metabolite extracts. We identified 69 and 62 metabolites significantly decreased in taxane-resistant MDA-MB-436 and Hs 578T cells, respectively, while 67 and 29 metabolites showed significant increase in taxane-resistant MDA-MB-436 and Hs 578T cells respectively, relative to their corresponding parental cells (Fig. 1C). Significant decrease in metabolites involved in pyrimidine, pentose phosphate pathway, glutathione, and cysteine/methionine metabolism characterized taxane-resistant cells compared with parental cells based on metabolic pathway enrichment analysis using the human Kyoto Encyclopedia of Genes and Genomes (KEGG) compound database and the Small Molecule Pathway Database (SMPDB; Fig. 1D). In contrast, we observed an increase in the abundance of metabolites involved in the purine/pyrimidine pathway and alanine metabolism across all taxane-resistant models relative to parental cells. Integration of metabolomic and gene expression profiles from taxane-resistant and taxane-sensitive TNBC models demonstrates significant concomitant alteration of metabolic and gene expression pathways relating to purine/pyrimidine and amino acid metabolism, including cysteine/methionine metabolism (Supplementary Fig. S1C and S1D).
A significant decrease (P < 0.05) in the steady-state levels of most metabolites from the methionine cycle is observed in the LC-HRMS results from taxane-resistant relative to parental cells (Fig. 1E; Supplementary Fig. S1E). Accordingly, decreased incorporation of labeled isotopes into SAM (M+5) and S-adenosylhomocysteine (SAH; M+4) is detected by l-13C5-methionine tracer analysis in taxane-treated parental cells and in taxane-resistant cells compared with parental MDA-MB-436 cells (Fig. 1F; Supplementary Fig. S1F). This observation suggests decreased flux of l-methionine–derived carbons and altered rate of SAM synthesis in taxane-resistant TNBC. In support, the rate of incorporation of 13C-labeled isotopes in 5-methylthioadenosine (5-MTA) (M+1), a metabolite that requires SAM for its synthesis, is significantly decreased in taxane-resistant compared with parental MDA-MB-436 cells (Supplementary Fig. S1G). Furthermore, the viability of taxane-resistant TNBC cells is significantly impaired compared with that of parental cells upon complete deprivation of l-methionine for 48 hours (Supplementary Fig. S1H), indicating the increased dependency of taxane-resistant TNBC cells on exogenous l-methionine. We next exposed the cells to long-term (5 weeks) deprivation of l-methionine using growth media containing 3 mg/L l-methionine (10% of normal levels). Although partial deprivation of l-methionine had no significant effect on the proliferation of parental MDA-MB-436 cells, it significantly affected the growth of taxane-resistant TNBC cells (Fig. 1G). The methionine deprivation was not sufficient to alter paclitaxel sensitivity in MDA-MB-436 cells (Supplementary Fig. S1I). Finally, l-methionine can affect antioxidant capacity by affecting glutathione pools (19). Although the levels of reduced glutathione (GSH) and reduced/oxidized glutathione (GSH/GSSG), indicative of antioxidant capacity, decreased upon taxane treatment in parental cells, they are significantly increased in taxane-resistant relative to taxane-treated TNBC cells (Fig. 1H; Supplementary Fig. S1J). Thus, the increased dependency of taxane-resistant cells on exogenous l-methionine for proliferation is accompanied by an improved glutathione-mediated antioxidant capacity in taxane-resistant TNBC cells, despite cytotoxic drug exposure.
Reduced SAM Aligns with DNA Hypomethylation in Taxane-Resistant TNBC
We next examined whether the observed alterations in methionine cycle and decreased rate of SAM synthesis induced a metabolic stress affecting the epigenetic state of taxane-resistant TNBC. We observed a significant decrease in global DNA methylation levels in taxane-resistant MDA-MB-436 and Hs 578T cells relative to parental cells using an ELISA system measuring DNA methylation at long interspersed nuclear elements (LINE1; Fig. 2A). Paclitaxel treatment for 96 hours in parental cells also led to a significant decrease in LINE1 methylation relative to control-treated cells. Likewise, treatment with the DNA methyltransferase inhibitor 5-aza-2′-deoxycytidine (5-AZA) induced a significant decrease of LINE1 methylation in parental MDA-MB-436 cells using the ELISA system (Fig. 2B). No significant effect of 5-AZA was observed in taxane-resistant TNBC cells, supporting DNA hypomethylation at LINE1 elements in taxane-resistant models. Using the methylationEPIC array, we next quantified the levels and distribution of DNA methylation signals in taxane-resistant and parental TNBC cells. A significant change in the distribution of DNA methylation signal across the genome of taxane-resistant compared with parental MDA-MB-436 cells was observed as illustrated by decreased Pearson correlation score over all probes (Fig. 2C). In particular, 29,247 significantly hypomethylated versus 12,283 significantly hypermethylated CpGs were identified in taxane-resistant relative to parental TNBC cells (Δß-value cutoff ± 0.3 and P < 0.05; Fig. 2D). The gains and losses in DNA methylation events identified using this Δß-value cutoff are representative of changes between low and high methylation states and dismiss small variations that could originate from technical issues (Fig. 2E). The significantly higher number of hypomethylated CpGs relative to hypermethylated CpGs reveals an overall global decrease in DNA methylation in taxane-resistant MDA-MB-436 cells relative to parental cells (Fig. 2F). Identification of differentially methylated regions (DMR) using Probe Lasso method reveals similar results, with 1,559 significantly hypomethylated DMRs versus 374 hypermethylated DMRs in taxane-resistant versus parental TNBC cells (Supplementary Fig. S2A and S2B).
To investigate whether hypomethylation of DNA is a common feature of acquired taxane resistance in breast cancer, we generated taxane-resistant tumors (GCRC2080-PACRes and GCRC1882-PACRes) from paclitaxel-naïve human breast cancer patient-derived xenografts (PDX; GCRC2080 and GCRC1882) engrafted in immunocompromised mice (Supplementary Fig. S2C). Maintenance of paclitaxel resistance in subsequent PDX passages was confirmed (Supplementary Fig. S2D). Using the LINE1 methylation ELISA system, we showed that paclitaxel-resistant breast cancer PDXs display significantly lower levels of LINE1 methylation compared with their matched parental PDXs (Fig. 2G). Accordingly, significant changes in DNA methylation distribution were observed using bisulfite conversion and hybridization to methylationEPIC arrays, with a higher number of hypomethylated CpGs and an overall decrease in CpG methylation ß-value observed in paclitaxel-resistant compared with sensitive PDXs (Fig. 2H and I; Supplementary Fig. S2E).
The hypomethylated CpGs in taxane-resistant cells and PDXs are enriched at intergenic regions (IGR) relative to the distribution of all CpGs assessed, whereas hypermethylated CpGs in taxane-resistant models are mostly observed at promoter regions (TSS1500; Supplementary Fig. S2F and S2G). Downregulated genes harboring hypermethylated promoters significantly associate with IFNα/β and cytokine signaling gene sets (Supplementary Fig. S2H).
H3K27me3 Compensates for DNA Hypomethylation in Taxane-Resistant TNBC
We further examined whether changes in methionine metabolism in taxane-resistant TNBC affect the levels and genomic distribution of histone methylation modifications. Immunoblotting of histone acid extracts showed no major changes in global levels of methylation on histone 3 between taxane-resistant and taxane-sensitive TNBC (Fig. 3A, left; Supplementary Fig. S3A and S3B). However, the H3K27me3 modification was markedly decreased in taxane-resistant relative to parental TNBC cells when using a mild modified RIPA lysis buffer that excludes the insoluble proteins presumably linked to heterochromatin (ref. 20; Fig. 3A, right; Supplementary Fig. S3C), whereas ELISA-based quantification of H3K27me3 from acid extracts shows no significant change between parental and taxane-resistant cells (Supplementary Fig. S3D). This observation suggests that changes in methionine metabolism observed in taxane-resistant TNBC cells are not associated with changes in global histone lysine methylation but instead associate with reallocation of H3K27me3 away from soluble toward insoluble chromatin compartments.
Considering the propensity of insoluble fractions to populate condensates associated with heterochromatin (21, 22), we assessed the genome-wide distribution of H3K27me3 using chromatin immunoprecipitation sequencing (ChIP-seq) assays in all taxane-resistant and parental MDA-MB-436 cells, and included ChIP-seq assays for H3K4me1, H3K4me3, and H3K27ac for comparison (Fig. 3B and C; Supplementary Fig. S3E–S3G). Pearson correlation between the signal intensity of each cell population for each histone modification shows that the H3K27me3 ChIP-seq signal is less correlated between taxane-resistant and parental MDA-MB-436 cells, reflecting a reallocation of H3K27me3 across the genome of taxane-resistant TNBC cells (Fig. 3B). Accordingly, between 49% and 53% of the H3K27me3 enriched genomic regions observed in the parental MDA-MB-436 cells are lost in the taxane-resistant cells (Fig. 3C). The reallocation of H3K27me3 is further characterized by an increase in signal intensity in taxane-resistant MDA-MB-436 cells at genomic regions shared with parental cells (Fig. 3C). A large proportion of these H3K27me3 enriched genomic regions cluster close to each other in taxane-resistant TNBC cells to form Large Organized Chromatin Lysine (K) domains (LOCK), as reported previously for H3K9me2 (refs. 23, 24; Fig. 3D; Supplementary Fig. S3H). We characterized the H3K27me3 signal that comprises LOCKs using clustering of cis-regulatory regions method (CREAM; ref. 25) in both taxane-resistant and parental MDA-MB-436 cells. The proportion of H3K27me3 signal comprised within LOCKs rises from an average of 57% of total H3K27me3 genomic regions in parental cells to an average of 68% in taxane-resistant cells, suggestive of the reallocation of H3K27me3 signal toward LOCKs regions in taxane-resistant TNBC cells.
To understand the role of H3K27me3 LOCKs, we classified them using a 2-fold difference cutoff between average H3K27me3 signal over LOCKs genomic regions identified in parental and taxane-resistant MDA-MB-436 cells. We identified 90 taxane resistant–specific LOCKs and 64 parental-specific LOCKs in addition to 659 shared LOCKs (Fig. 3E). Importantly, taxane resistant–specific LOCKs are wider than parental-specific or shared LOCKs, with an average size of 630 kb versus 29 and 21 kb in parental and shared LOCKs, respectively (Fig. 3F). Interrogation of the genes comprised within LOCKs reveals that taxane resistant–specific LOCKs lie in gene-poor regions, a feature common with shared LOCKs but distinct from parental-specific LOCKs (FDR < 0.05 by permutation testing; Fig. 3G), and do not show significant association with changes in gene expression (Supplementary Fig. S3I). Therefore, the reprogramming of H3K27me3 defines a new epigenetic state characterized by a gain of widespread heterochromatin domains over gene-poor regions in taxane-resistant TNBC cells.
Given the well-described repressive nature of DNA methylation and H3K27me3 modification, we asked whether the reprogramming of both epigenetic modifications in gene-poor genomic regions orchestrates a H3K27me3 compensatory mechanism to maintain a repressive chromatin state over hypomethylated DNA regions in taxane-resistant TNBC cells. More than 26% of the total number of CpG probes covered by LOCKs regions (14,628/55,842) show differential methylation in taxane-resistant cells whereas only 3% of CpG probes located outside of LOCKs (26,901/738,148) display differential methylation (Supplementary Fig. S3J). Notably, assessment of β-value signals of CpGs that fall within LOCKs revealed an overall decrease in global level of DNA methylation over LOCKs (Fig. 3H). Accordingly, more than 95% (13,949 of 14,628) of the differentially methylated CpG probes located inside H3K27me3 LOCKs are significantly hypomethylated in taxane-resistant cells, whereas significantly changed CpG probes located outside LOCKs are as likely to be hypo- or hypermethylated in taxane-resistant MDA-MB-436 cells versus parental cells (Fig. 3I). This supports a model where the reallocation of H3K27me3 modification at LOCKs in taxane-resistant MDA-MB-436 cells provides a compensatory repressive mechanism over DNA-hypomethylated gene-poor regions (Fig. 3J; Supplementary Fig. S3K).
H3K27me3 Reallocation into LOCKs Prevents Activation of the Viral Mimicry Response
We next assessed whether taxane-resistant LOCKs are enriched for genomic features characteristic of intergenic regions such as TEs. Enrichment analysis revealed the preferential overlap between taxane resistant–specific H3K27me3 LOCKs and families of TEs, such as LINE1, long terminal repeat (LTR), and human endogenous retrovirus (HERV) families (P < 0.01; Fig. 4A; Supplementary Fig. S4A). Expression of some HERV3 subfamily members can induce dsRNA accumulation, which activates an IFN response through viral mimicry and blocks cell growth (15). Using immunofluorescence staining with an antibody recognizing dsRNA, we first show that paclitaxel treatment leads to the accumulation of dsRNA in parental MDA-MB-436 cells whereas no dsRNA accumulation is observed in taxane-resistant cells grown in presence of paclitaxel (Supplementary Fig. S4B). To examine whether reprogramming of H3K27me3 provides transcriptional repression of hypomethylated TE in taxane-resistant TNBC cells, we treated TNBC cells using UNC1999 or GSK343, two inhibitors of the EZH2 methyltransferase responsible for H3K27me3 deposition. We show that therapeutic inhibition of EZH2 equally depletes H3K27me3 signal in both parental and taxane-resistant MDA-MB-436 cells relative to vehicle or the control chemical probe UNC2400 (Fig. 4B; Supplementary Fig. S4C). Immunofluorescence staining with the dsRNA-specific antibody shows accumulation of dsRNA specifically in taxane-resistant MDA-MB-436 cells upon UNC1999 treatment compared with the vehicle or control probe UNC2400 (Fig. 4C). No significant induction of dsRNA is observed upon UNC1999 treatment in the parental TNBC cells. Quantification of dsRNA enrichment using RNAseA treatment over total RNA shows significant increase in dsRNA accumulation upon UNC1999 or GSK343-mediated inhibition of EZH2 in taxane-resistant MDA-MB-436 cells, whereas pharmacologic inhibition of EZH2 has no effect on dsRNA accumulation in parental MDA-MB-436 cells (Fig. 4D). In addition, depletion of H3K27me3 following pharmacologic inhibition of EZH2 with UNC1999 or GSK343 induced the expression of a specific subset of HERVs (MER54A, MER21C, and ERVL) in taxane-resistant MDA-MB-436 cells while having no effect in parental cells (Fig. 4E). Yet, both paclitaxel and 5-AZA treatment could induce HERV expression in parental MDA-MB-436 cells (Fig. 4F and G). Although 5-AZA also induced ERVL expression in taxane-resistant MDA-MB-436 cells (Fig. 4G), it had no effect on MER54C or MER21C, suggesting that some TEs remain regulated by DNA methylation in taxane-resistant MDA-MB-436 cells.
We next asked whether accumulation of dsRNA in taxane-resistant MDA-MB-436 cells induces the expression of genes regulating the viral mimicry–induced IFN response. We observed that the gene expression profile of taxane-treated relative to untreated parental MDA-MB-436 cells is positively associated with the induction of the 20-gene viral mimicry–response signature previously reported (ref. 15; P = 0.048, NES = 1.49; Supplementary Fig. S4D, left), whereas that of the taxane-resistant relative to parental TNBC cells negatively associates with the viral mimicry gene signature (P < 1e-04, NES = −2.064; Supplementary Fig. S4D, right). Using gene expression data from a cohort of TNBC tumors from patients treated with a combination of chemotherapeutic agents that include paclitaxel, we observed that the viral mimicry–response gene signature is also significantly downregulated in tumors of patients with TNBC presenting progressive disease upon treatment relative to tumors of patients responding to treatment (P < 1e-04, NES = −3.02; Supplementary Fig. S4E, left). This signature is also significantly downregulated in the corresponding PDXs from these patients' tumors grown in immunocompromised mice (P < 1e-04, NES = −1.96; Supplementary Fig. S4E, right), indicating that the downregulation of viral mimicry–induced IFN response genes is also intrinsically observed in tumor cells derived from patients. Accordingly, individual genes from the viral mimicry gene signature are significantly downregulated in patients with TNBC presenting progressive disease upon treatment relative to those showing therapeutic response (Supplementary Fig. S4F).
We next show that treatment of taxane-resistant MDA-MB-436 cells with UNC1999, which leads to the accumulation of dsRNA, induces a gene expression profile positively associated with the activation of the viral mimicry signature (P = 0.0033, NES = 1.869; Fig. 4H). Accordingly, induction of a subset of IFN response genes (IFIH1, OASL, ISG15) was confirmed specifically in taxane-resistant TNBC cells following pharmacologic inhibition of EZH2 with UNC1999 or GSK343 (Fig. 4I). Induction of a selected subset of TEs and concomitant induction of IFN response gene expression is also observed upon depletion of EZH2 using siRNAs against the EZH2 gene in taxane-resistant TNBC cells, whereas no effect is observed upon EZH2 perturbation in parental TNBC cells (Supplementary Fig. S4G and S4K). The expression of IFN response genes is specifically induced in parental MDA-MB-436 cells upon 5-AZA treatment, whereas no significant effect in taxane-resistant cells is observed with 5-AZA (Fig. 4J), indicating that DNA methylation regulates the viral mimicry response in parental TNBC cells. 5-AZA–mediated inhibition of DNA methyltransferase does not on its own alter the paclitaxel sensitivity of parental MDA-MB-436 cells (Supplementary Fig. S4L). Altogether, these results indicate that the reallocation of H3K27me3 modification into LOCKs over hypomethylated CpGs in taxane-resistant TNBC cells provides a compensatory epigenetic mechanism for maintaining a repressive chromatin state over regions enriched for hypomethylated TEs, thereby preventing dsRNA accumulation and induction of IFN response.
To investigate whether H3K27me3-mediated repression of TEs and IFN response is a common feature of paclitaxel resistance, we used paclitaxel-resistant (PDXO-1986) and sensitive (PDXO-1915) organoids derived from PDXs of basal subtype breast cancer (Supplementary Fig. S4M). We show that pharmacologic inhibition of EZH2 using UNC1999 significantly induced a subset of TEs and of IFN response genes in paclitaxel-resistant PDXO (PDXO-1986), while having no effect in the sensitive PDXO (PDXO-1915; Fig. 4K and L). Using the paclitaxel-sensitive (GCRC1882) and paclitaxel-resistant (GCRC1882-PACRes) breast cancer PDX models, we assessed the effect of H3K27me3 depletion following pharmacologic inhibition of EZH2 using UNC1999 in vivo (Fig. 4M). Pharmacologic inhibition of EZH2 using UNC1999 induces a significant increase of IFIH2 and DDX58 viral mimicry gene expression in paclitaxel-resistant PDX but not in the taxane-sensitive PDX (Fig. 4N). Together, these results demonstrate that the reallocation of H3K27me3 in taxane-resistant TNBC models maintains repression of TEs and protects against induction of viral mimicry response, thereby mediating viral mimicry response evasion.
Taxane Resistance Creates a Druggable Epigenetic Vulnerability
Because of the known antitumor effect induced by the viral-mimicry state resulting from dsRNA expression (15), we assessed whether alleviating the H3K27me3-mediated repression at LOCKs using EZH2 inhibition creates an epigenetic vulnerability in taxane-resistant TNBC cells. Although pharmacologic inhibition of EZH2 using UNC1999 or GSK343 has no effect on the growth and survival of parental MDA-MB-436 or Hs 578T cells, it significantly antagonized the growth and survival of all taxane-resistant MDA-MB-436 and Hs 578T cells (Fig. 5A and B; Supplementary Fig. S5A–S5C), thereby indicating an epigenetic vulnerability. Dose-response cell viability assay shows that all taxane-resistant TNBC cells gained sensitivity to lower concentrations of EZH2 inhibitors compared with parental cells (Supplementary Fig. S5D). Accordingly, siRNA-mediated depletion of EZH2 expression for 96 hours also specifically impaired the viability of taxane-resistant TNBC cells while not affecting that of the parental cells (Supplementary Fig. S5E). In addition, pharmacologic inhibition of EZH2 using UNC1999 or GSK343 induces the apoptotic marker cleaved caspase-3 specifically in taxane-resistant TNBC cells (Supplementary Fig. S5F). Finally, we show that UNC1999-mediated inhibition of EZH2 does not alter the paclitaxel sensitivity of parental MDA-MB-436 cells at the time tested (Supplementary Fig. S5G).
To test whether EZH2-mediated growth inhibition in taxane-resistant cells depends on the induction of IFN response genes following TE expression, we used siRNA-mediated depletion of the IFIH1 gene coding for MDA5, the IFN response factor responsible for binding dsRNA (Supplementary Fig. S5H). Depletion of IFIH1 expression for 96 hours partly rescued cell growth inhibition observed upon pharmacologic repression of EZH2 by UNC1999 or GSK343 in taxane-resistant MDA-MB-436 cells, while having no significant effect on the growth of parental cells (Fig. 5C).
This epigenetic vulnerability is also observed upon pharmacologic inhibition of EZH2 activity in paclitaxel-resistant TNBC patient–derived models. Pharmacologic inhibition of EZH2 using UNC1999 significantly impairs the viability of paclitaxel–resistant PDX-derived organoids (PDXO-1986) while not affecting the viability of the organoid model with higher paclitaxel sensitivity (PDXO-1915; Fig. 5D). Accordingly, PDXO-1986 shows increased sensitivity for UNC1999 treatment compared with the paclitaxel-sensitive PDXO-1915 (Fig. 5E). Finally, to test whether this epigenetic vulnerability is also observed in vivo, we treated breast cancer PDX mice bearing paclitaxel-sensitive (GCRC1882) and matched resistant (GCRC1882-PACRes) tumors with UNC1999 or control. While paclitaxel treatment induced significant regression of paclitaxel-sensitive GCRC1882 tumors, treatment with UNC1999 did not significantly affect the growth of this paclitaxel-sensitive PDX (Fig. 5F). In contrast, UNC1999 treatment in mice bearing the paclitaxel-resistant GCRC1882-PACRes tumors induced significant delay in tumor growth (Fig. 5G, left) and significantly increased survival time (Fig. 5G, right). Together, these results demonstrate that paclitaxel resistance in TNBC dictates an epigenetic profile that warrants viral mimicry evasion through the maintenance of a repressive heterochromatin state over vulnerable LOCK regions to prevent the induction of TE expression, concomitant accumulation of dsRNA and activation of IFN response, thereby creating an epigenetic vulnerability in paclitaxel-resistant tumors.
Discussion
The various mechanisms proposed to underlie the development of resistance to standard-of-care treatment in cancer are diverse, yet most of them involve the engagement of survival pathways that bypass or compensate for the detrimental effects induced by the drugs. Diverse metabolic adaptations support the development of drug-resistant cancer cells under the stressful environment that stems from therapy (1, 26). Here, we show an impairment in methionine metabolism giving rise to decreased SAM synthesis rate and availability in taxane-resistant TNBC cells. The altered methionine metabolism results in increased dependency on exogenous l-methionine across taxane-resistant TNBC cells and parallels previous observations in drug-resistant and metastatic cancer cells. In particular, one-carbon metabolism, trans-sulfuration and glutathione metabolism, which are linked to methionine metabolism, can support metabolic antioxidant capacity to improve the survival of cancer cells exposed to cytotoxic drugs (27, 28). Accordingly, increased dependency on exogenous l-methionine in taxane-resistant TNBC cells may provide substrates to sustain glutathione-mediated antioxidant capacity despite cytotoxic drug exposure (19, 29). Altered methionine metabolism may also affect drug sensitivity by having an impact on folate metabolism, which was reported to affect the expression and activity of proteins involved in breast cancer drug resistance such as the breast cancer resistance protein (BCRP) and the multidrug resistance protein-1 (MRP1; ref. 30). Finally, dietary methionine restriction was shown to induce a therapeutic response in PDX models of chemotherapy-resistant RAS-driven colorectal cancer and in a mouse model of radiation-resistant mutant KRAS soft-tissue sarcoma (31, 32). Therefore, impaired methionine metabolism and increased exogenous methionine dependency observed in taxane-resistant TNBC cells may support advantageous metabolic conditions favoring the survival of taxane-resistant TNBC.
An important consequence of the altered methionine metabolism observed in taxane-resistant TNBC cells is the DNA hypomethylation compensated by the reallocation of H3K27me3 into LOCKs occurring over TEs. This parallels changes in LINE1 DNA methylation observed upon alterations in one-carbon metabolism induced either genetically or through the diet in pancreatic and breast cancer models (4, 33, 34). In bladder cancer cells, exposure to reactive oxygen species (ROS) activates glutathione synthesis via the methionine cycle, depleting SAM, consequently leading to hypomethylation of LINE1 regions (33). Likewise, l-methionine deprivation in breast cancer cells was shown to induce IFN response genes (35). Our observations also align with reports of H3K27me3 and other repressive histone modifications compensating for DNA hypomethylation to ensure genome stability in mouse embryonic stem cells (36) and to maintain transcriptional repression across various cancer types (37). Although the sensitivity of different methyltransferases to changes in SAM availability depends on cell type and developmental stage (38), the fact that global decreases in DNA methylation observed in taxane-resistant TNBC cells are not paralleled with a global decrease in H3K27me3 levels agrees with the higher Michaelis constant (Km) for SAM reported for DNA methyltransferases (DNMT) compared with EZH2, suggesting that DNMT activity is susceptible to changes in SAM levels (39).
Our data suggest that cytotoxic treatment of TNBC cells with paclitaxel may contribute to the metabolic and epigenetic rewiring observed in taxane-resistant TNBC cells. Indeed, exposing sensitive TNBC cells to paclitaxel treatment decreases methionine flux and antioxidant capacity, impairs LINE1 DNA methylation and induces dsRNA, TE, and viral mimicry. Previous studies report that taxane, cisplatin, and doxorubicin treatment induce IFN response genes in colon cancer cells, in breast cancer PDXs, and in isolated malignant cells from breast tumors (40–42). In addition, our data indicate that cytotoxic drug-mediated induction of viral mimicry is suppressed in taxane-resistant cells and in patient-derived tumors. Thus, the metabolic and epigenetic states allowing viral mimicry evasion may be favored in cancer cells upon exposure to cytotoxic drugs and may contribute to the adaptation/selection of drug-resistant TNBC cells upon cytotoxic stress. Whether this process applies to resistance to other cytotoxic drugs remains to be determined. Breast cancer PDX models exposed to various chemotherapeutic agents reveal that despite initial induction of IFN gene expression upon treatment, IFN-related gene expression is repressed upon relapse (41). Moreover, the necessity of maintaining repression over LINE1 has also been observed upon development of drug resistance and tumor progression in lung and pancreatic cancer models (4, 12). Overall, these observations suggest that overexpression of viral mimicry may be a transient event induced upon cytotoxic stress and which can be controlled by epigenetic rewiring upon progression toward drug resistance.
In the absence of repressive histone modification–based compensation, DNA hypomethylation can reactivate the expression of TEs, including HERVs, leading to dsRNA accumulation and further activation of the IFN-mediated viral mimicry response and cell growth inhibition (15). Accordingly, TE repression is required for immune escape in leukemic stem cells (17) and survival of drug-tolerant lung cancer cell subpopulations (12). Collectively, our results further highlight the importance of maintaining transcriptional repression over potentially deleterious TEs to enable cancer progression under taxane treatment. The requirement for evading the viral mimicry response in taxane resistance is further supported by the downregulation of the IFN response genes in taxane-resistant TNBC models. Accordingly, alleviating the H3K27me3-mediated compensation of DNA hypomethylation using EZH2 inhibitors leads to accumulation of dsRNA and induction of the viral-mimicry response to block cell growth in paclitaxel-resistant models. Given that induction of IFN response in tumors can promote an immunologic response at tumor sites, future work investigating how this epigenetically regulated viral mimicry evasion in tumor cells affects the host immune response will be fundamental to better target drug-refractory tumors. Epigenetic regulation of TE expression enabling evasion of viral mimicry response could thus represent an important determinant of cancer cell survival upon drug treatment. Preventing this epigenetic-mediated evasion to reactivate the viral mimicry response may provide a new avenue for targeting drug-resistant cancer.
Methods
Cell Culture and Reagents
For all experiments, MDA-MB-436 and Hs 578T were maintained in high-glucose DMEM (Gibco) supplemented with 10% FBS (Gibco) and penicillin/streptomycin (100 U/mL Penicillium and 100 μg/mL streptomycin), unless otherwise stated. All cell lines used were routinely tested for Mycoplasma (Lonza Mycoalert and EZ-PCR Mycoplasma Detection Kit) and were authenticated using short tandem repeat analysis. Taxane-resistant cells were derived upon long-term exposure to increasing concentrations of paclitaxel at a starting concentration of 0.05 nmol/L, with increments until reaching cytotoxic concentration of 10 to 20 nmol/L (>6 months). Resistant cells were continuously cultured in the presence of 10 or 20 nmol/L paclitaxel (Sigma-Aldrich, T4702). Parental cells were maintained with DMSO as a control. For l-methionine depletion and deprivations, taxane-resistant and parental cells were transferred in custom DMEM deprived of l-methionine (Wisent) and supplemented with dialyzed FBS. l-methionine–deprived DMEM was supplemented with 30 mg/L (complete) or 3 mg/L (10%) l-methionine (Sigma, M8439) for add-back experiments. For acute paclitaxel treatment, parental cells were transferred to DMEM containing 20 nmol/L paclitaxel for the indicated times. The list of antibodies used for ChIP assays, Western blots, IHC, and immunofluorescence is available in the Supplementary Material and Methods. Chemical probes (UNC1999, UNC2400, GSK343) were obtained from the Structural Genomic Contiortium (SGC) group in Toronto. 5-AZA (Sigma) was used at 1 μmol/L.
Metabolomic Profiling by LC/MS and Mass Isotopomer Distribution Analysis
Metabolomics analysis was performed with liquid chromatography (LC) coupled to Q Exactive Plus high resolution mass spectrometer (HRMS). The HPLC (Ultimate 3000 UHPLC) was coupled to QE-MS (Thermo Fisher Scientific) for metabolite separation and detection. An Xbridge amide column (100 × 2.1 mm i.d., 3.5 μm; Waters) was employed for compound separation at room temperature. Mobile phase A was water with 5 mmol/L ammonium acetate (pH 6.9), and mobile phase B was acetonitrile. See Supplementary Information for gradients. HRMS acquisition and raw data analysis method was described previously (43). For intracellular metabolites, 5e5 cells were incubated in 6-well plates in 3 mL medium for 96 hours. Cells were briefly washed with 1 mL ice-cold saline (0.9% NaCl in water) twice before metabolite extraction. Metabolites were extracted into 80% methanol, dried in a vacuum concentrator, and stored at −80°C until further analysis. Metabolites were reconstituted into sample solvent (water:methanol:acetonitrile, 2:1:1, v/v) with volume proportional to cell number; 3 μL was injected for metabolomics analysis. The LC/MS peak intensity (integrated peak area) of each metabolite was used to calculate the relative abundance ratio. Peak intensity was quantile-normalized and differential enrichment of metabolites was calculated between taxane-resistant TNBC cells and parental cells using one-way ANOVA. Pathway enrichment and integrated gene expression and metabolite levels were assessed with Metaboanalyst online tools. For Mass Isoproptomer Distribution analysis and LC/MS, parental and taxane-resistant MDA-MB-436 cells were seeded in triplicate in 100-mm plates, grown to 50% confluence and treated with control or paclitaxel for 24 hours before changing media to 7 mL l-Methionine–depleted high glucose DMEM (Wisent) supplemented with 10% dialyzed FBS (Wisent) and either 30 mg/L l-13C5-methionine (uniformly labeled, Cambridge Isotope Laboratories, CLM-893-H-PK, 99%) or unlabeled l-methionine (Sigma-Aldrich Co). After 30-minute or 4-hour pulses, cells were extracted and subjected to LC/MS analysis of methionine flux into SAM, SAH, and 5-MTA, as described previously (2). Matrix corrections for tracer analysis were carried out as described previously (44).
Global DNA Methylation Quantification, DNA Methylation Array, and Analysis
Quantification of 5-mC DNA was assayed using the Global DNA Methylation Assay LINE1 ELISA system (Active Motif) as specified by the manufacturer. Bisulfite conversion of the DNA was performed using the EZ DNA Methylation Kit (Zymo Research) on 500 ng genomic DNA in parental and taxane-resistant cell lines. The Illumina Infinium MethylationEPIC BeadChips were processed as per the manufacturer's recommendations. R package ChAMP v2.6.4 was used to process and analyze data for differentially methylated probes and regions (45). For analysis purposes, probes differentially methylated between parental and taxane-resistant MDA-MB-436 cells with P < 0.05 and delta-beta value < −0.3 or > 0.3 were considered significantly changed. Data has been deposited at GSE111541.
Protein Extraction, Analysis, and Quantification of H3K27me3
Proteins were extracted from 2–3 million cells using specified extraction buffers including RIPA lysis buffer: 25 mmol/L Tris•HCl, pH 7.6, 150 mmol/L NaCl, 1% NP-40, 1% sodium deoxycholate, 0.1% SDS; mild modified RIPA lysis buffer: 50 mmol/L Tris HCl, pH 8, 150 mmol/L NaCl, 1 mmol/L EDTA, 1% NP-40, 0.25% deoxycholate; or acid extraction buffer (0.4 mol/L HCl) and neutralized with sodium dibasic phosphate. Quantification of acid-extracted H3K27me3 in TNBC cells was achieved using the Histone H3 methylated Lys27 ELISA system (Active Motif) as per the manufacturer's instructions.
ChIP-seq
For ChIP-seq analyses, chromatin was prepared from parental and taxane-resistant MDA-MB-436 cells grown at 70% to 80% confluence. For each immunoprecipitation, 3–5 million cells were cross-linked with 1% formaldehyde for 10 minutes at room temperature. Extracted chromatin was sonicated as described previously (46). ChIP was performed using 2–4 μg of antibody directed toward each histone modification reported. The ChIP primers are listed in Supplementary Table S1. For ChIP sequencing, sonicated DNA samples were quantified using Qubit reagent (Life Technologies); 1 ng of DNA per sample was used following ThruPLEX DNA-seq Kit manufacturer's protocol (Takara Bio). See Supplementary Information for sequencing details. Following ChIP-seq, quality control–filtered reads were aligned to the reference assembly hg19 using bwa v.0.5.9. Only uniquely mapped reads were retained for further analysis. Redundant reads were removed using SAMtools v0.1.18. Number of reads were downsampled to the lowest number of reads of the samples to be compared together. Peaks were called using MACS2.0 with default values (mfold[10,30], P = 1e-04). WIG files were generated using MACS2.0 with default parameters. Bed files contain peak calls on merged reads using corresponding inputs as control. Scores represent peak intensity. Data has been deposited at GSE113687. Pearson correlation of signal intensity was obtained and peak annotation, tag directory, bed file generation, and merging was performed with the Homer package v3.1 (http://homer.ucsd.edu/homer/ngs/). Tag heat maps were obtained with Java TreeView (47).
CREAM Analysis
H3K27me3 LOCKs were identified using CREAM (25) considering a cutoff of 0.1 for window size in both parental and taxane-resistant MDA-MB-436 cells using the H3K27me3 peak calls from MACS2.0. The clusters were further annotated as parental-specific, resistant-specific, or shared between parental and taxane-resistant MDA-MB-436 cells considering 2 fold-change differences between H3K27me3 average signal intensity within LOCKs in parental and resistant samples.
RNA Extraction, RNA Sequencing, Analysis, and Gene Set Enrichment
For expression analysis, parental and taxane-resistant TNBC cells were harvested at 80% confluency and RNA extraction was performed with the RNeasy Mini Kit (Qiagen) and reverse-transcribed using SensiFAST cDNA Synthesis Kit (Bioline) and analyzed by Q-RT-PCR with SensiFAST SYBR-green NO-ROX Kit (Bioline). Two hundred nanograms total RNA from the 30 RNA samples were purified for polyA tail–containing mRNA molecules using poly-T oligo attached magnetic beads; following purification the RNA was fragmented and libraries were prepared as detailed in the Supplementary Material and Methods. Alternatively, for the UNC1999 treatment experiment, 200 ng total RNA from 42 samples was prepared into libraries using TruSeq Stranded Total RNA Kit (Illumina). RNA samples were ribosomal RNA–depleted using Ribo-zero Gold rRNA beads; following purification the RNA was fragmented. All libraries were loaded onto an Illumina NextSeq cartridge for cluster generation and sequencing on an Illumina Nextseq500 instrument (Illumina) using Paired-end 75 bp protocol to achieve approximately 40 million reads per sample. Reads were aligned to hg19 using tophat (2.0.8) with default parameters, transcripts were merged using cufflinks (2.1.1), and differential expression computed for all relevant contrasts with default parameters. Differentially expressed genes were used as input for determining gene set enrichment analysis (GSEA) using the GSEA canonical pathways or hallmarks of cancer pathways. Primer sequences for RT-PCR for gene expression are listed in Supplementary Table S1. Data have been deposited at GSE111920 (UNC1999 treatment) and GSE113687 (steady-state gene expression). For detection of viral mimicry enrichment gene set, the viral mimicry gene signature from Roulois and colleagues (15) was used for gene set enrichment (IRF7, DDX58, PLAC8, OASL, TAP1, IFI44, IFIH1, DHX58, IFIT2, HERC5, IFIT1, ISG15, IFITM2, MX1, STAT2, STAT1, IFI27, IRF9, IRF1, IFI6, IFITM1, IFITM3).
Association of ChIP-seq and Cluster Regions with Gene Expression
Genes in 100 kb proximity of individual peaks or LOCKs are considered as genes associated with those regions. Expression of these genes obtained from RNA sequencing were then compared between parental and taxane-resistant samples using Wilcoxon ranked sum test. The resulting P values were FDR corrected for all the hypotheses (number of genes) tested for each set of elements (set of peaks or LOCKs identified for a given ChIP-seq profile).
Repeat Element Enrichment Score
To identify the enrichment of each family of repeat elements, we first obtained sets of randomly selected chromosome regions with the same distribution of the LOCKs within each chromosome. The actual number of repeats identified in LOCKs for either parental-specific, shared, or resistant-specific H3K27me3 LOCKs were compared with the number of repeats obtained in the 10,000 randomly selected chromosome regions. FDR was calculated as one minus the number of times the random regions contained more (positive enrichment) or less (negative enrichment) repeat elements than the observed H3K27me3 LOCKs and divided by the total number of random sets of regions (10,000).
Cell Proliferation Assays
Cells were maintained as described above, seeded in 24-well plates and treated with the indicated epigenetic probes or cytotoxic drugs or supplemented with or depleted from the indicated metabolites. Cell viability was assessed using crystal violet as described previously (2) or using CellTiter-Glo assay (Promega). For proliferation/cell density assay, cells were seeded in 96-well plates transferred to IncuCyte ZOOM Live cell analysis system (Essen BioScience) and proliferation was monitored using IncuCyte ZOOM phase-contrast quantification software. All results shown represent the average of at least three independent replicate experiments each carried out with at least three technical replicates.
IHC and Immunofluorescence Imaging
For immunofluorescence assay, parental and taxane-resistant TNBC cells were seeded in 6-well plates and treated with UNC1999 (3 μmol/L), UNC2400 (3 μmol/L), or DMSO on the following day for 4 days. Cells were then counted and seeded into 24-well plates at a density of 1 × 105 cells per well and treated again with UNC1999 (3 μmol/L), UNC2400 (3 μmol/L), or DMSO for an additional 24 hours. Cells were fixed in methanol and stained overnight with J2 scion antibody for dsRNA staining (Mouse J2 antibody Product no: 10010200 Scicons 1:500) or for nuclear staining (Hoescht, H1399 Thermo Fisher Scientific, 1:2,000), incubated in secondary antibody (Cell Signaling Technology 4410S anti-mouse IgG Alexa 647 1:1000) or anti-Rabbit secondary antibody (Cell Signaling Technology, Alexa 488 1:1,000), and imaged using a 40× objective. Total and dsRNA-positive cells in a minimum of 4 independent fields (at least 1,000 total cells per condition) were quantified. For IHC, tumors were harvested from mice, and 5-μm sections of formalin-fixed, paraffin-embedded murine mammary tumors were IHC stained with H3K27me3 antibody (Diagenode; 1:100 dilution).
Generation and Maintenance of TNBC Organoids
Biobanking, PDX, and organoid experiments were approved by the McGill University Health Centre (MUHC) Institutional Review Board and all subjects provided written informed consent (SDR-99-780 and SDR-00966). Breast cancer tissue was cut into 1 mm3 pieces and engrafted in the mammary fat pad of 4- to 6-week-old NOD/SCID gamma (NSG) female mice. PDX at passages 2 to 4 were used to generate organoids. PDX tumor tissue was minced and digested in a rotating shaker with collagenase/dispase (1 mg/mL; Sigma) for 2 hours at 37°C in DMEM/F12. The tissue was further digested for 15 minutes at 37°C in Accutase (Sigma). Mouse cells were depleted using magnetic beads (Miltenyi Biotec) and single human tumor cells were plated on Matrigel-coated dishes and cultured in DMEM/F12 GlutaMAX supplemented with B27, EGF, bFGF, insulin, hydrocortisone, prostaglandin E2, R-spondin-1, noggin, Rock inhibitor, and 5% Matrigel. Media was changed every 4 days and organoids were passaged every 1–2 weeks. For cell viability assay upon paclitaxel and UNC1999 treatment, single cell dissociated organoids were seeded at 1 × 104 cells/well in 96-well plate. Drugs were added after 4 days at the indicated concentrations and incubated for 96 hours. Cell viability was assessed using CellTiter Blue with 4-hour incubation (Promega).
In Vivo Studies of TNBC PDX and Generation of Acquired Paclitaxel Resistance
All human participants provided written informed consent for this study (SDR-99-780 and SDR-00966), and tissue was collected at the MUHC in accordance with the protocols approved by the research ethics board (SUR-99-780). All animal studies were approved by the McGill University Animal Care Committee (2014-7514) and conducted in NSG mice from The Jackson Laboratory. In vivo studies were randomized in cohorts of 6–8 mice per arm. 1–2 mm3 tumor fragments were engrafted in the fourth mammary fat pad of 6- to 8-week-old mice. Paclitaxel was administered at 24 mg per kg (mpk) every 7 days, intravenously. The powder of UNC1999 (HPLC- and mass spectrometry–verified) was slowly dissolved and incorporated to vehicle [0.5% of sodium carboxymethylcellulose (NaCMC) and 0.1% of Tween-80 in sterile water] with continuous trituration by a pestle and mortar and kept for < 1 week at 4°C in the dark. UNC1999 (300 mg/kg) or vehicle were administered by oral gavage daily for 5 consecutive days followed by one day off and one day of paclitaxel treatment. Tumors were measured and mice weight was monitored by a blinded animal technician twice a week. Tumor volume was calculated by [(smaller tumor dimension2 × largest tumor dimension)/2].
Generation of Resistant PDX Model
All human participants provided informed written consent for this study (SDR-99-780 and SDR-00966) and tissue was collected at the MUHC in accordance with the protocols approved by the research ethics board (SUR-99-780). GCRC1882 single-mouse was treated with paclitaxel 24 mpk every 7 days intravenously until a reduction of 100% in tumor volume was achieved. Treatment was stopped and initiated again when tumor relapse reached a volume of > 120 mm3. Treatment was given on and off until the tumor stopped responding to the drug (GCRC1882PACRes). To confirm acquired resistance, GCRC1882PACRes was collected and 2 mm3 fragments were retransplanted into the fourth mammary fat pad of 4 mice. Three mice were treated with paclitaxel 24 mpk every 7 days intravenously, and 1 mouse was left untreated as comparison.
Disclosure of Potential Conflicts of Interest
D.W. Cescon reports grants from Terry Fox Research Institute and grants from Stand Up To Cancer Canada/Canadian Cancer Society Research Institute during the conduct of the study; personal fees from Agendia, AstraZeneca, GlaxoSmithKline, Merck, Novartis, Pfizer, Puma, Roche, Exact Sciences, and Dynamo Therapeutics; grants from Pfizer; other funding from Merck (Research Funding to Institution), and other funding from GlaxoSmithKline (Research Funding to Institution) outside the submitted work. In addition, D.W. Cescon has a patent for Biomarkers for TTK inhibitors pending. D.D. De Carvalho reports grants from Pfizer and grants from Nektar Therapeutics outside the submitted work. J.W. Locasale reports grants from NIH and grants from American Cancer Society during the conduct of the study and personal fees from Restoration Foodworks outside the submitted work. M. Park reports grants from FRQS-Reseau Cancer, grants from FRQS-Oncopole, Quebec Breast Cancer Foundation, and SU2C during the conduct of the study. No potential conflicts of interest were disclosed by the other authors.
Authors' Contributions
G. Deblois: Conceptualization, data curation, software, formal analysis, supervision, validation, investigation, writing-original draft, writing-review and editing. S.A. Madani Tonekaboni: Data curation, software, formal analysis, validation. G. Grillo: Validation, investigation. C. Martinez: Investigation. Y.I. Kao: Investigation. F. Tai: Investigation. I. Ettayebi: Investigation. A.-M. Fortier: Investigation. P. Savage: Investigation. A.N. Fedor: Investigation. X. Liu: Investigation. P. Guilhamon: Data curation, formal analysis. E. Lima-Fernandes: Investigation. A. Murison: Data curation, formal analysis. H. Kuasne: Investigation. W. Ba-alawi: Formal analysis. D.W. Cescon:Supervision, funding acquisition. C.H. Arrowsmith: Supervision, funding acquisition. D.D. De Carvalho: Supervision, funding acquisition. B. Haibe-Kains: Supervision, funding acquisition. J.W. Locasale: Supervision. M. Park: Resources, supervision, funding acquisition. M. Lupien: Conceptualization, resources, supervision, funding acquisition, project administration, writing-review and editing.
Acknowledgments
We thank Ken Kron, Aislin Treloar, Christopher Arlidge, and Sarina Cameron for assistance in the maintenance of the various cell lines. The authors wish to thank the Princess Margaret Genomics Core for sequencing of ChIP and RNA and for methylEPIC array hybridization, Daina Avizonis and Shawn McGuirk at the Goodman Cancer Research Centre Metabolomics Core Facility for metabolite measurements, and Cian Monnin for technical assistance. We thank Annie Monast for mouse husbandry and for technical assistance. We acknowledge the Princess Margaret Bioinformatics group for providing the infrastructure assisting us with analysis presented here. This study was conducted with the support of the Terry Fox Research Institute (New Frontiers Research Program PPG-1064, to M. Lupien, B. Haibe-Kains, C.H. Arrowsmith, and D.W. Cescon), the Canadian Cancer Research Society, and the Ontario Institute for Cancer Research through funding provided by the Government of Ontario. This work was also supported by the Canadian Institute for Health Research (CIHR; Funding Reference Number 136963 and 158225, to M. Lupien; 363288, to B. Haibe-Kains), The Princess Margaret Cancer Centre, The Princess Margaret Cancer Foundation, and the Ontario Ministry of Health (to M. Lupien, B. Haibe-Kains, D.D. De Carvalho, C.H. Arrowsmith, and D.W. Cescon), the Réseau de Recherche en Cancer of the FRQS (to M. Park), Québec Breast Cancer Foundation (to M. Park), Oncopole (to M. Park), Gattuso-Slaight Personalized Cancer Medicine Fund at Princess Margaret Cancer Centre (to B. Haibe-Kains), and the SU2C Canada - Canadian Cancer Society Breast Cancer Dream Team Research Funding (SU2C-AACR-DT-18-15), with supplemental support from the Ontario Institute for Cancer Research, through funding provided by the Government of Ontario (to M. Park, D.W. Cescon, and B. Haibe-Kains). Stand Up To Cancer Canada is a Canadian Registered Charity (Reg. # 80550 6730 RR0001). Research Funding is administered by the American Association for Cancer Research International - Canada, the Scientific Partner of SU2C Canada. G. Deblois is a recipient of fellowships from the CIHR, of the Fonds de Recherche en Santé du Québec (FRQS) postdoctoral research award, and of a Cancer Research Society Next-Generation of Scientists transition award, and is a Research Scholars - Junior1 of the Fonds de Recherche en Santé du Québec. M. Lupien holds an Investigator Award from the Ontario Institute for Cancer Research, a CIHR New Investigator Award, and the Bernard and Francine Dorval Award for Excellence from the Canadian Cancer Society. M. Park is a James McGill Professor and holds the Diane and Sal Guerra Chair in Cancer Genetics at McGill University. P. Guilhamon is a recipient of a CIHR fellowship. S.A. Madani Tonekaboni was supported by Connaught International Scholarships for Doctoral Students. C. Martinez is a recipient of a doctoral award from FRQS. The SGC is a registered charity (number 1097737) that receives funds from AbbVie, Bayer Pharma AG, Boehringer Ingelheim, Canada Foundation for Innovation, Eshelman Institute for Innovation, Genome Canada through Ontario Genomics Institute (OGI-055), Innovative Medicines Initiative (EU/EFPIA; ULTRA-DD grant no. 115766), Janssen, Merck KGaA, Darmstadt, Germany, MSD, Novartis Pharma AG, Ontario Ministry of Research, Innovation and Science (MRIS), Pfizer, São Paulo Research Foundation-FAPESP, Takeda, and Wellcome.
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.