Reprogramming of Nucleotide Metabolism Mediates Synergy between Epigenetic Therapy and MAP Kinase Inhibition Free

Small cell carcinoma of the ovary, hypercalcemic type (SCCOHT) is an aggressive malignant tumor with a dismal prognosis (1). The mean age at diagnosis is approximately 24 years, and most patients die within 2 years of diagnosis. For SCCOHT, treatment generally involves surgery and adjuvant chemotherapy, most commonly platinum-based compounds. Despite combination chemotherapy approaches, however, the prognosis still remains poor with overall 5-year survival rates of only 16% (2). Moving forward, personalized therapies for SCCOHT will require proper diagnosis and the identification of oncogenic drivers of these carcinomas. Differentiating SCCOHT from morphologically similar tumors is challenging. The “small” and “large” variants are fairly analogous (3), and SCCOHT also needs to be distinguished from other primary and metastatic tumors that may be found within the same tissue including nonepithelial ovarian neoplasms and metastases from small cell lung carcinoma among others (3). The identification of a central role for SMARCA4 mutations in the pathogenesis of this tumor (4, 5), and subsequent use of SMARCA4 (BRG1) IHC, with or without using antibodies raised against SMARCA2 (BRM) has greatly facilitated the diagnosis (6).

Most SMARCA4 mutations in SCCOHT are deleterious resulting in a complete loss of protein expression, being confirmed by IHC in almost 100% of cases (4, 6). SMARCA4 and its paralog SMARCA2 are essential ATPase components of the multisubunit SWI/SNF (SWItch/Sucrose Non-Fermentable) chromatin remodeling complex and modify histone–DNA interactions by shifting or evicting nucleosomes to change the landscape of accessible regions on chromatin, thereby affecting transcriptional activation (7). SMARCA2 is concomitantly lost with SMARCA4 in almost all SCCOHT cases, and this profile now constitutes a molecular signature of the disease (6).

Previously, it was shown that SMARCA4 and another bromodomain protein BRD4 independently associate with distal enhancer elements of c-MYC in order to activate oncogene transcription, suggesting some redundancy between these two proteins in gene regulation and tumorigenesis (8). This led us to hypothesize that in the absence of SMARCA4 and SMARCA2, BRD family members might represent essential proteins for driving transcriptional networks involved in proliferation and survival (9). Thus, targeting SMARCA4-deficient cancers with bromodomain inhibitors (BETi), that target multiple BRD proteins, might effectively shut down this BRD-driven oncogenic network. We demonstrated that SMARCA4/A2-deficient SCCOHT and non–small cell lung cancer (NSCLC) models were acutely sensitive to BET inhibitors in vitro and in mouse models at very low doses (9). Notably, this work revealed that BETi efficacy correlated with repression of PI3K and MAPK pathways. This is consistent with other studies suggesting that intrinsic resistance to BETi is conferred by constitutive signaling through receptor tyrosine kinase pathways including PI3K and MAPK (10–12). Recently, it was shown that NRAS-mutant melanoma models displayed resistance to BETi. In turn, combining BETi with MEKi led to decreased cell proliferation in vitro and was also effective against melanomas carrying NRAS mutations in vivo (13). The precise mechanism through which this combination works, and whether this approach may be applicable to additional tumor types, remains unclear.

Here, we screened a range of inhibitors targeting PI3K and MAPK pathways for potential synergy with BETi against a panel of SMARCA4-deficient and SMARCA4-expressing ovarian cancer cell lines. Among all the tested compounds, we found a strong synergy between BETi (OTX015) and MEK inhibitors (cobimetinib and trametinib) at suboptimal doses in both ovarian cancer models. This combination also proved highly effective against orthotopic xenograft models of ovarian cancer. Using mass spectrometry to assess changes in protein content after exposure to combination therapy revealed that concurrent treatment with BETi/MEKi represses key enzymes involved in nucleotide metabolism. A concomitant decrease in nucleotide pools was validated using metabolomics profiling. This culminates in a reduced pool of nucleotide precursors and cell-cycle arrest. Overall, the combination of BETi/MEKi highlights a potential new therapeutic approach to treat multiple subtypes of ovarian tumors.

The OVK18 cell line was received from the RIKEN cell bank. The SCCOHT1 cell line was a gift from Dr. Ralf Hass (Hannover Medical School, Hannover, Germany). OVCAR4, OVCAR3, SKOV3, IGROV1, and HEK293T were purchased from the ATCC. The cell lines were grown in RPMI-1640 medium supplemented with 10% FBS. The culture medium for HEK293T cells was DMEM with 10% FBS. The cell lines were cultured for a maximum of 3 weeks. The cell lines were maintained at 37°C in an atmosphere of 5% CO₂. The cells were cultured for a maximum of 10 passages from initial stock thawing. All cell lines were tested for Mycoplasma contamination by DAPI staining and not otherwise authenticated.

For ectopic expression of KRAS, pLenti-PGK-KRAS4B(G12D) (Addgene, #35633) was employed. BRAF plasmid was a gift from Ian R. Watson (Goodman Cancer Center, McGill University) and was generated by subcloning HA-BRAF(V600E) into backbone vector pHAGE-EF1a. pReciever-LV120 (GeneCopeia, #EX-EGFP-Lv120) was used as a control vector. pReceiver-LV242 control vector (#EX-LV242), pReceiver-DUT-Flag (#EX-E1984-LV242), and pReceiver-TYMS-Flag (#EX-T0406-LV242) were purchased from GeneCopeia. Ubiquitin-HA-pcdna 3.1 plasmid was given by Dr. James (Zhijian) Chen (Southwestern Medical Center). shRNA plasmids against dUTPase (DUT), thymidylate synthase (TYMS), and RRM1 (Ribonucleotide Reductase, subunit 1) were purchased from Sigma (Supplementary Table S1).

Cobimetinib, trametinib, copanlisib, lapatinib, patritumab, and cisplatin were purchased accordingly: Selleckchem (#S8041), Selleckchem (#S2673), Selleckchem (#S2802), Selleckchem (#S2111), Creative Biolabs (#TAB-189), and Accord (#DIN 02355183). OTX015 was purchased at MedChem Express (#HY-15743), and its chemical structure is available at the PubChem database (14).

OVK18 cells were cultured for six passages in light (R0K0), medium, (R6K4) and heavy (R10K8) isotope containing media provided by François-Michel Boisvert (Department of Anatomy and Cell Biology, Université de Sherbrooke). Next, the cells were exposed to 0.01% DMSO, OTX015 (200 nmol/L), and cobimetinib (200 nmol/L) alone or in combination. After 24 hours of treatment, the cell pellets were frozen in liquid nitrogen and sent for the stable isotope labeling by amino acids in cell culture (SILAC) analysis to the Department of Anatomy and Cell Biology, Université de Sherbrooke. The analysis was performed precisely as previously described (15). Perseus program was used for statistical analysis. The data were obtained from 3 replicates with log₂ fold change > 0.5 and P value ≤ 0.05. Volcano plots were generated with Instant Clue. The data are available at the PRIDE database (accession #PXD017581).

Co-immunoprecipitation (co-IP) experiments were performed as described previously (16). The HEK293T cells were transfected, using polyethylenimine, with cDNAs encoding TYMS-Flag, Ubiquitin-HA, and subsequently treated for 20 hours with MG132 (7 μmol/L), OTX015 (200 nmol/L), and/or Cobimetinib (200 nmol/L). Subsequently, cells were collected, lysed, and immunoprecipitation carried out using antibodies against either FLAG or HA tags.

Animal experiments were performed following guidelines of the Canadian Council of Animal Care and approved by the Animal Resources Centre at McGill University. Five-week-old female NOD/SCID mice were injected with 5 × 10⁶ of OVK18 cells or 1 × 10⁷ of OVCAR4 cells in 1xPBS into left ovary via laparotomy. Treatments with vehicle, 20 mg/kg/day of OTX015 and 5 mg/kg/day of cobimetinib alone or in combination, were initiated after 3 weeks of tumor cell injection. Cohorts were comprised of 5 mice. The compounds were administered by oral gavage for a period of 3 weeks. After this period of time, mice were euthanized and resultant tumors were surgically obtained from the left ovary and surrounding ovarian bursa. A single primary tumor was collected and measured. The calculation of tumor volume was performed by using following formula: ½ × [(length in mm) × (width in mm)²]. IHC analysis was performed at the Segal Cancer Centre Research Pathology Facility (Jewish General Hospital) as previously described (17). Xenograft tumor sections were incubated with TYMS (at 1:50 dilution), DUT (at 1:25 dilution), and RRM1 (at 1:25 dilution) rabbit antibodies. Slides were scanned in low power field to choose the most stained area. TYMS and DUT protein expression was distributed homogeneously within the tumor tissues. RRM1 expression was mostly localized at the tumor border representing approximately 10% of the tissue. Sections were analyzed, and the final score was obtained by normalizing to the average of the total intensity. The staining intensity was quantified by ImageJ and assigned by using five-tiered system (0 = negative, 1 = very weak, 2 = weak, 3 = moderate, 4 = strong, and 5 = very strong staining).

OVK18 and OVCAR4 cells were seeded in 6-well plates and exposed to DMSO, OTX015 (200 nmol/L), cobimetinib (200 nmol/L), and OTX015/cobimetinib for 3 days. The cell-cycle analysis was performed via propidium iodide (PI) staining as described previously (18). For bromodeoxyuridine (BrdU) proliferation assay, the cells were incubated with 7 μmol/L of BrdU for 24 hours prior the analysis. The BrdU analysis was assessed by using commercial kit (BD FITC BrdU Flow kit, #51–2354AK). Early and late S phases were subdivided by low and high 7AAD content. ModFit and FlowJoe were used for the analysis.

Note that 2.5 × 10⁶ of OVK18 cells were plated in 10 cm petri dishes and treated for 48 hours with DMSO or OTX015/cobimetinib combination at 200 nmol/L concentration for each compound. Media were washed from adherent cells (4.3 million per plate) using ice-cold 150 mmol/L ammonium formate pH 7.4. Cells were then scraped into 380 μL of 50% methanol/water to which 220 μL of ice-cold acetonitrile (ACN) was added. Cells were then subjected to bead beating for 2 minutes at 30 Hz (Eppendorf Tissue-lyser). Lipids were partitioned through the addition 600 μL of cold dichloromethane and 300 μL of cold H₂O. The upper aqueous layer was then removed and dried using a vacuum centrifuge with sample temperature maintained at −4°C (LabConco). Samples were resuspended in 25 μL of water and subjected to LC-MS analysis. For nucleotide analysis, a 10x dilution was prepared by adding 3 μL of sample to 27 μL of water. The relative concentrations of the targeted nucleotides and deoxynucleotides were measured using a triple quadrupole mass spectrometer (QQQ 6470) equipped with a 1290 ultrahigh-pressure liquid chromatography system (Agilent Technologies). Chromatographic separation was achieved using a Scherzo SM-C18 column 3 μm, 3.0 × 150 mm (Imtakt Corp). The chromatographic gradient started at 100% mobile phase A (5 mmol/L ammonium acetate in water) with a 5-minute gradient to 100% B [200 mmol/L ammonium acetate in 20% ACN/80% water) at a flow rate of 0.4 mL/min. This was followed by a 5-minute hold time at 100% mobile phase B and a subsequent re-equilibration time (6 minutes) before next injection. In order to ensure proper instrumental duty cycle, samples were injected twice: nucleotide analysis followed by deoxynucleotide analysis. A sample volume of 5 μL was injected for each run. Multiple reaction monitoring (MRM) transitions were optimized on standards for each metabolite measured. MRM transitions and retention time windows are summarized in Supplementary Table S3. An Agilent JetStream electrospray ionization source was used in positive ionization mode with a gas temperature and flow were set at 300°C and 5 L/min respectively, nebulizer pressure was set at 45 psi, and capillary voltage was set at 3,500 V. Relative concentrations were determined from external calibration curves prepared in water. Ion suppression artifacts were not corrected; thus, the presented metabolite levels are relative to the external calibration curves and should not be considered as absolute concentrations. Data were analyzed using MassHunter Quant (Agilent Technologies). All LC-MS grade solvents and salts were purchased from Fisher: water, ACN, methanol (MeOH), formic acid, ammonium acetate, and ammonium formate. The authentic metabolite standards were purchased from Sigma-Aldrich Co.

Antibodies for Western blotting were purchased as follows: anti-DUT (Abcam, #ab229122), anti-TYMS (Abcam, #ab108995), anti-RRM1 (Cell Signaling Technology, #8637), anti-BRAF V600E (Abcam, #200535), anti-KRAS G12D (Cell Signaling Technology, #14429), anti–α-tubulin (DSHB, #12G10-s1ea), anti–β-actin (Sigma, #A5316), anti–c23-MS3-nucleolin (Santa Cruz Biotechnology, #sc8031), anti-AKT (Cell Signaling Technology, #2920s), anti-pAKT (Cell Signaling Technology, #4060s), anti-S6 (Cell Signaling Technology, #2217s), anti-pS6 (Cell Signaling Technology, #2215s), anti-ERK1/2 (Cell Signaling Technology, #4695s), anti-pERK1/2 (Cell Signaling Technology, #9101s), anti-Flag (Sigma, #F1804), and anti-HA (Abcam, #ab1424).

We previously found that the downregulation of PI3K and MAPK pathways correlates with the antiproliferative effects of BETi in SMARCA4-deficient cells (9). However, SMARCA4-expressing cells were largely unresponsive to BETi exposure. In addition, BETi failed to reduce PI3K and MAPK activity in SMARCA4-proficient cells. We also previously found that re-expression of the RTK, HER3, conferred partial resistance to BETi, indicating RTK signaling is indeed important for BETi-mediating resistance (9). However, it remains unproven that downstream effectors of RTK directly confer resistance. As such, we introduced BRAF (V600E) and KRAS (G12D) into initially BETi-sensitive, SMARCA4-deficient ovarian cancer cells and subsequently compared cell viability after BETi exposure (Supplementary Fig. S1A). As expected, expression of the constitutively active BRAF and KRAS induced ERK phosphorylation (Supplementary Fig. S1B) which was paralleled by decreased sensitivity to BETi (Supplementary Fig. S1A). These data confirmed that both upstream and downstream activators of RAS-MAPK oncogenic pathways may impart resistance to BETi. Thus, we hypothesized that BETi may cooperate with RTK or MAPK inhibitors to repress oncogenic growth in SMARCA4-deficient cells and that such combinations may potentially overcome resistance in SMARCA4-expressing models.

To explore this, we first carried out short-term (5-day) in vitro cell viability assays, with the degree of synergy being calculated using “Excess Over Bliss” (EOB) as previously described (19). These experiments were carried out on a panel of SMARCA4-deficient SCCOHT (SCCOHT1 or OVK18) and SMARCA4-expressing ovarian adenocarcinoma cell lines (OVCAR4, OVCAR3, SKOV3, or IGROV1). The clinically tested BETi, OTX015, was combined with a panel of inhibitors targeting RTK-dependent signaling pathways including cobimetinib or trametinib (MEK inhibitors), copanlisib (PI3K inhibitor), patritumab (anti-HER3 antibody), and lapatinib (HER2/EGFR inhibitor). Only weak cooperativity was observed between BETi and patritumab or BETi/lapatinib (HER2/EGFR inhibitor; Supplementary Fig. S2A, EOB generally < 1). A more pronounced effect was seen between copanlisib and BETi (Supplementary Fig. S2B, EOB generally between 0 and 10). However, of the tested combinations, the most marked synergy was detected between the MEK inhibitors cobimetinib and trametinib with OTX015 (Fig. 1, EOB observed between 5 and 20). Surprisingly, this response was also observed in the SMARCA4-expressing ovarian adenocarcinoma cancer models OVCAR4 and OVCAR3, but to a lesser degree in IGROV1 and SKOV3. This supports the concept that targeting the MAPK pathway represents a means to overcome intrinsic resistance to BETi.

To investigate whether the combination of OTX015/cobimetinib has in vivo relevance, we tested the combination at suboptimal doses against orthotopic ovarian xenograft models of SMARCA4-deficient SCCOHT (OVK18) and SMARCA4-expressing ovarian adenocarcinoma (OVCAR4; Fig. 2A). The cells were injected into the left ovary via laparotomy surgery procedure. For these studies, we used suboptimal doses of both drugs, with OTX015 being given at 20 mg/kg/day and cobimetinib at 5 mg/kg/day, each administered 5 days per week. These concentrations are half of the drug concentrations typically employed in preclinical studies (20, 21). Three weeks after cell implantation, mice received either vehicle, OTX015, cobimetinib, or OTX015/cobimetinib in combination for 3 weeks. At the experimental endpoint, tumors from the left ovary and ovarian bursa surrounding area were surgically isolated, and volume measured. We employed OVK18 as a model of SCCOHT because of its relative resistance to BETi in vivo compared with other SCCOHT models which show acute sensitivity at the concentrations currently employed (9). This relative resistance is likely due to an activating A59G KRAS mutation in OVK18 that may impart partial resistance in vivo. The treatment was well tolerated by mice without obvious signs of gross toxicity or significant body weight loss (Fig. 2B). We observed that the OTX015/cobimetinib combination nearly eradicated tumor growth in the SCCOHT model. We previously found that ovarian adenocarcinomas generally harbor intrinsic resistance to BETi (9). However, the data from these xenograft experiments revealed drastically stronger effects of OTX015/cobimetinib combination against OVCAR4 xenografts, as compared with single-drug treatments (Fig. 2A). This demonstrates that resistance to BETi can be overcome by combining them with MEKi (Fig. 2A). Both compounds have been reported to be associated with strong toxicities (22, 23). Thus, the combinatorial approach might be promising from a clinical perspective due to the suboptimal doses required to prohibit tumor growth.

To date, most studies aiming to define the downstream effectors of BETi response have focused on RNA sequencing (RNA-seq). However, mRNA expression does not necessarily coincide with, and accurately reflect, protein expression. In order to find potential drivers of the response to the combination of BETi/MEKi, we conducted SILAC mass spectrometry experiments on OVK18 cells treated with DMSO, OTX015, cobimetinib, or OTX015/cobimetinib combination for 24 hours (Fig. 3A and B). This method employs stable isotope labeling of cells for 6 days prior to drug exposure, allowing us to quantify abundance ratios between proteins from treated and untreated conditions. Surprisingly, under these conditions, of the 1,797 detected proteins, only a few were identified as undergoing significant changes in expression of either drug individually. In contrast, 27 proteins underwent significant changes in the combination setting after 24 hours of drug exposure. This relatively early time point was chosen to identify potential effectors of BETi/MEKi that might initiate the antiproliferative effects later observed in these cells. Even though a small number of target proteins were identified, Gene Ontology (GO) term pathway analysis of proteins undergoing altered expression revealed an enrichment for proteins involved in nucleotide metabolism (Fig. 3C, P < 0.05). Importantly, among the key downregulated targets, we discovered proteins involved in DNA synthesis such as TYMS, DUT, and RRM1 (24, 25). TYMS and DUT are enzymes catalyzing thymidine DNA synthesis (24). RRM1 is one of the subunits of the ribonucleotide reductase (RNR) complex which participates in DNA synthesis by converting ribonucleotides to deoxyribonucleotides (dNTPs; ref. 25). Notably, all three enzymes have been reported overexpressed in several cancers and are considered as attractive anticancer therapeutic targets (26, 27).

To validate the key targets from our SILAC screen, we carried out time course treatments with OTX015 and cobimetinib followed by RT-qPCR and Western blotting analysis of putative targets. We did not observe major changes in mRNA levels for TYMS, DUT, and RRM1 after BETi/MEKi combination treatment as compared with control or single treatments (Fig. 3D), suggesting that the decrease to these proteins revealed by mass spectrometry is not transcriptionally regulated. Western blotting validated that the expression levels for all the three proteins were significantly downregulated in sensitive OVK18 and OVCAR4 cell lines but not in more resistant SKOV3 cells (Fig. 3E). This suggests that diminished expression of enzymes involved in nucleotide metabolism plays an important role in mediating the antiproliferative effects of BETi/MEKi.

Our data indicate that BETi and MEKi cooperate to promote reduced protein levels of key enzymes involved in proliferation and survival. Previous proteomic studies (available at PhosphoSitePlus; ref. 28) identified TYMS as a target for ubiquitylation (lysine 169), and thus, it is a likely candidate for proteasome-mediated degradation (29). Toward this end, we treated OVK18 cells with proteasome inhibitor MG132 and subsequently exposed the cells to OTX015, cobimetinib alone, or in combination (Fig. 3F). Interestingly, we observed that the MG132 treatment abrogated protein downregulation stabilizing TYMS levels. These data indicate that BETi/MEKi act to promote proteasome-mediated degradation of TYMS, potentially affecting nucleotide synthesis. To identify whether there is an association between TYMS and ubiquitin, we carried out co-IP experiments and revealed the TYMS/Ubiquitin association after exposure to combination therapy (Fig. 3G; Supplementary Fig. S3A). This association was enhanced in response to MG132 treatment, indicating TYMS is ubiquitinated and undergoes proteasomal degradation in response to the combination treatment.

To substantiate that downregulation of these targets might be responsible for the antiproliferative activity of BETi/MEKi in vivo, we probed for DUT, TYMS, and RRM1, using IHC, within tumors after drug exposure. Consistent with our in vitro findings, we observed a striking synergy in the reduction of TYMS, DUT, and RRM1 within tumors from mice treated with the OTX015/Cobimetinib combination (Fig. 4).

Our study revealed that DUT and TYMS are likely involved in the antitumor response to the combination of OTX015/cobimetinib. TYMS catalyzes thymidine DNA synthesis through conversion of deoxyuridine monophosphate (dUMP) to deoxythymidine monophosphate (dTMP; refs. 30–32). TYMS is found overexpressed in some cancers and appears to promote proliferation and may be correlated with poor clinical outcomes (33, 34). The inhibition of TYMS promotes metabolic imbalance leading to aberrant incorporation of nucleotides during DNA replication resulting in cell-cycle arrest (35). TYMS is a target of the potent, and widely used, chemotherapeutic 5-fluorouracil (5-FU), and targeted TYMS inhibitors have also been explored clinically (36, 37). The enzyme DUT is responsible for the hydrolysis of dUTP to dUMP which is subsequently further metabolized by TYMS (38). It has been shown that DUT overexpression conferred resistance to TYMS inhibitors (27). Thus, concomitant inhibition of both DUT and TYMS might represent a logical approach to anticancer treatment because this will lead to DNA replication arrest and ultimately cell death. Indeed, this approach was found to be synthetically lethal in NSCLC models using siRNAs (24). Consistent with these studies, we observed a significant synergy between shDUT and shTYMS using combinations of four shRNAs, two unique shRNAs targeting each enzyme in OVK18 cells (Fig. 5A). Notably, knockdown of DUT or TYMS alone did not exert substantial antiproliferative effects in this setting, highlighting the limitations of pharmacologic targeting either of these enzymes as monotherapy in ovarian cancer. To further validate that a synthetic lethality between TYMS and DUT represents a potential mechanism whereby OTX015/cobimetinib acts in synergy, we utilized the lentiviral shRNAs targeting DUT and TYMS either alone, or in combination, and carried out cell viability analysis for 5 days in OVCAR4 (Supplementary Fig. S3B). Again, targeting TYMS and DUT displayed stronger antiproliferative effects than targeting a single enzyme. Together, these data strongly suggest that DUT and TYMS downregulation is at least partially responsible for the antiproliferative effects of OTX015/cobimetinib combination. Ectopic expression of TYMS conferred partial resistance to the combination, reducing cell death by over 20%, on its own. DUT was slightly less effective on its own, and surprisingly, coexpression did not enhance the effect (Supplementary Fig. S3D). This indicates that the downstream effectors of BETi/cobimetinib represent a complex network, likely involving multiple modulators of nucleotide metabolism including, but not limited to, TYMS and DUT.

Another potential mediator of the antiproliferative activity of BETi/MEKi identified from our mass spectrometry experiments is RRM1. This is the largest subunit of the RNR complex, formed with the smaller subunit, RRM2 (25). RRM1 participates in DNA synthesis by catalyzing the production of mature dNTPs from ribonucleotides. High RRM1 expression is associated with dismal prognosis in lung and pancreatic cancer (25, 39, 40), underscoring its potential as an anticancer target. We again employed a lentiviral shRNA approach to target RRM1, and we observed that 5 days of RRM1 depletion effectively reduced the viability of OVK18 and OVCAR4 cells without the depletion of additional factors (Fig. 5B; Supplementary Fig. S3C). Next, we performed concomitant knockdown for either TYMS and RRM1 or DUT and RRM1. However, we did not observe a synergy between these shRNA combinations, likely due to effects of RRM1 alone being quite strong. Again, these data suggest that altered nucleotide metabolism plays a critical role in compromising cell proliferation after exposure to BETi/MEKi. We observed that TYMS, DUT, and RRM1 were expressed at high levels across the cell lines tested, including OVCAR3, OVCAR4, SKOV3, IGROV1, and OVK18 (Supplementary Fig. S3E). In SCCOHT1, TYMS and DUT levels were quite low. Our findings imply that these enzymes are highly expressed and may frequently drive proliferation, and that the combination therapy we tested will compromise cell viability even in tumors where TYMS, DUT, or RRM1 is highly expressed.

The enzymes DUT, TYMS, and RRM1 are responsible for nucleotide synthesis and potentiate DNA replication and progression through S phase of cell cycle. To examine the impact of OTX015/cobimetinib treatment on cell-cycle profiles, we treated OVK18 cells with DMSO, OTX015, or cobimetinib alone or in combination for 3 days and analyzed the cell cycle using two complementary techniques, PI staining and BrdU incorporation (Fig. 6A and B; Supplementary Fig. S4). Both approaches showed a decrease in S phase in the combination setting. This profile was also observed in OVCAR4 cells (Supplementary Fig. S4). BrdU incorporation offered a more refined S-phase profile and suggested that the reduced number of cells detected in S phase is primarily due to a lack of progression to late S phase. This supports a model where DNA replication is initiated, but quickly stalls due to either the misincorporation of nucleotides or lack of available nucleotides.

Our cell-cycle and protein profiling data strongly suggested that nucleotide pools might be altered within the cells in response to combination treatment. To explore this hypothesis, we utilized LC/MS metabolomics profiling to examine the steady-state levels of 35 metabolites, including nucleotides, nucleosides, and their metabolic precursors in OVK18 cells treated with DMSO, or OTX015/cobimetinib, for 48 hours (Fig. 6C). This analysis revealed a marked decrease to a spectrum of both pyrimidines and purines, beyond the expected changes such as reduced dTMP, a product of TYMS activity. DUT, TYMS, and RRM1 are all involved in de novo nucleotide synthesis, and the profile we observed likely reflects an attempted compensation through the nucleotide salvage pathway at this early time point (see Discussion). Metaboanalyst pathway analysis of significantly regulated metabolites also predicted the enrichment for changes in purine and pyrimidine metabolism (Fig. 6D). These data strengthen the concept that the antiproliferative effects of OTX015/cobimetinib in combination are mediated through extensive metabolic reprogramming and downregulation of nucleotide synthesis.

SCCOHT is an extremely aggressive cancer where conventional therapy has proved to be largely ineffective, with only about one third of those diagnosed at an early stage experiencing long-term survival (2). Recent studies of SCCOHT resulted in the discovery that it is a virtually monogenic disorder, with germline and/or somatic mutation in the SMARCA4 gene accounting for nearly all cases (4). These mutations are usually deleterious leading to a complete loss of the protein detected by IHC analysis in almost 100% of cases (6). Coupled with this loss, the SMARCA4 paralog, SMARCA2, is invariably lost as well. This unique genetic signature of SCCOHT led to the rational development of new approaches to treat SCOOHT in preclinical studies. We previously found that SCCOHT and NSCLC SMARCA4/A2-deficient tumors are highly sensitive to epigenetic drugs targeting bromodomain proteins (BETi; ref. 9). In parallel with the significant antitumor BETi response observed in the SMARCA4-deficient cancer models, we also found that SMARCA4-expressing ovarian and lung cancer cells are intrinsically resistant to BETi.

One of the antiproliferative signatures of BETi response in sensitive cells, based on our RNA-seq data, was the downregulation of RTK-dependent PI3K-AKT and RAS-MAPK signaling. In particular, inactivation of the MAPK pathway was strongly repressed by BETi in sensitive cells (9), but not observed in the SMARCA4-expressing, BETi-resistant cells. Thus, MAPK repression may act as a pharmacodynamic biomarker predicting sensitivity to BETi, and constitutive MAPK activity may represent a possible mechanism whereby cells resist the antiproliferative effects of BETi. These findings are consistent with additional reports that indicate signaling through PI3K and MAPK imparts both intrinsic and acquired resistance to BETi (41, 42).

BET inhibitors have been widely studied in preclinical settings in numerous types of cancer, but the drug concentrations used in many of these studies were far above the nanomolar concentrations required to inhibit BRD4 functions. As a result, the biomarkers identified in preclinical studies have not proven robust in a clinical setting, and these clinical trials have not met expectations (22, 43, 44). Thus, there is an urgent need to find biomarkers predicting sensitivity to BETi and to uncover therapeutic avenues to overcome intrinsic resistance to BETi. Combination treatment of BETi with inhibitors of RTK-dependent signaling potentially represents a rational approach to more effectively switch off oncogenic signaling and overcome resistance to BETi. Here, among all the tested inhibitors of RTK, PI3K, and MAPK pathways, we found the most significant synergy between BETi and MEK inhibitors in both SMARCA4-deficient and SMARCA4-expressing in vitro and in vivo ovarian tumor models. Our results are consistent with previous reports showing synergistic efficiency between these types of compounds in NRAS-mutant melanoma (13). Thus, our data, and those of others, highlight that this approach might hold a broad clinical relevance for diverse, aggressive cancers.

Currently, the mechanism through which BETi and MEKi act in synergy is uncertain. Our SILAC-based mass spectrometry experiments pinpointed rate-limiting enzymes involved in nucleotide metabolism including TYMS, DUT, and RRM1 are being strongly repressed by BETi/MEKi. These targets are necessary for DNA replication, and their depletion is expected to prevent progression through S phase due to a lack of available nucleotides or misincorporation of nucleotides, both leading to DNA damage and stalled replication forks (38, 45, 46). Our data suggest TYMS may be regulated through proteasomal degradation. TYMS is known to be a target for ubiquitylation (29, 30), which we suggest is precipitated by the combination therapy. Future work will aim to identify the ubiquitin ligase targeting TYMS after cotreatment with BETi/MEKi, but a candidate ligase of TYMS, UBE2K, has already been identified (47). In addition, it is currently unclear whether DUT and RRM1 are targeted in the same manner as TYMS. However, previous mass spectrometry reports demonstrated that both DUT and RRM1 are targets for ubiquitination (48, 49). We expect that the degradation of these enzymes may be cell-cycle dependent (50), or alternative mechanisms may be at play, such as loss of mRNA translation.

Clinically approved pyrimidine analogs that inhibit TYMS, such as 5-FU, are commonly used in the treatment of multiple cancers. However, overexpression of TYMS and DUT results in the resistance to this therapy (26, 38). Because DUT and TYMS function within the same essential pathway, dual TYMS/DUT inhibition might hold promise for anticancer treatments through effective dampening of nucleotide synthesis. Consistent with this, it has been shown that concomitant downregulation of TYMS and DUT leads to a synthetic lethality in NSCLC (24). Recently, a new DUT inhibitor, TAS-114, in combination with a novel TYMS inhibitor, capecitabine, which is selectively converted to 5-FU in tumors, entered to clinical trials against solid tumors (51). This approach is supported by our data, where dual knockdown of TYMS and DUT achieved synergy in killing ovarian cancer cells. Our data also indicate that BETi or MEKi might be combined with already clinically approved TYMS inhibitors or novel DUT inhibitors in order to elicit antitumorigenic effects. RRM1 represents another repressed target of OTX015/cobimetinib that participates in DNA synthesis (25). The significant loss of proliferation upon depletion of RRM1 suggests that targeted RRM1 inhibitors hold great potential as anticancer therapy, either as solo agents, or perhaps in combination with DUT or TYMS inhibitor. It has been shown that ERK inhibition along with the mTOR inhibitor Everolimus overcame resistance in renal cell carcinoma by reducing nucleotide pools as a result of RRM1 inhibition (52). This is consistent with our data indicating that nucleotide reprogramming is critical determinant of the growth inhibitory properties of BETi/MEKi.

Nucleotides for DNA synthesis may be formed through either de novo synthesis or the nucleoside salvage pathway (53). For the de novo pathway, glucose is converted to phosphoribosyl diphosphate which subsequently leads to the synthesis of purine and pyrimidine precursors, IMP and UMP, respectively. This leads to NDPs' (ribonucleotides) production, that are in turn converted by RRM1 to dNDPs utilized for the synthesis of dNTPs. All three enzymes, RRM1, DUT, and TYMS, are involved in de novo DNA synthesis (24, 25, 38). The salvage pathway utilizes early precursor nucleosides (dNs) as substrates to synthetize DNA through multiple enzymatic reactions. These two pathways are interconnected, and the salvage pathway has been shown to compensate for the de novo pathway in case of its inhibition (54). As TYMS and DUT were downregulated in the response to combination therapy, we expected to see a substantial depletion of dUMP and dTMP, the products of TYMS- and DUT-dependent activity. At the timepoint we chose, there is indeed a tendency toward dUMP and dTMP decrease, though they are not as significant as the changes observed for the early precursors. This suggests that the de novo pathway might be inhibited, but a later timepoint would be required to see the fully executed metabolomic reprogramming. It is likely that flux through the salvage pathway is enhanced to compensate for suppressed de novo DNA synthesis. In this situation, cells are slowly depleting their nucleotide precursors to maintain dNTP levels, resulting in the stark loss of the precursor pools that we observed in our metabolic analysis.

Overall, this work demonstrates that the synergy between BETi/MEKi in ovarian cancer models is mediated by the downregulation of multiple, critical regulators of nucleotide metabolism. Considering the suboptimal concentrations of drugs required to elicit antitumor effects, this combination appears to hold great promise for the treatment of poor prognosis cancers.

W.D. Foulkes reports grants from Astra Zeneca outside the submitted work. No disclosures were reported by the other authors.

T. Shorstova: Conceptualization, data curation, software, formal analysis, validation, investigation, visualization, methodology, writing-original draft, project administration, writing-review and editing. J. Su: Formal analysis, investigation, visualization, methodology. T. Zhao: Formal analysis, investigation, visualization, methodology. M. Dahabieh: Resources, software, formal analysis. M. Leibovitch: Formal analysis, investigation, visualization, methodology. M. De Sa Tavares Russo: Software, formal analysis, investigation, visualization, methodology, writing-review and editing. D. Avizonis: Software, investigation, visualization, methodology, writing-original draft, writing-review and editing. S. Rajkumar: Resources, investigation, visualization, methodology. I.R. Watson: Resources, investigation, visualization, methodology, writing-review and editing. S.V. del Rincón: Software, methodology, writing-review and editing. W.H. Miller Jr: Software, methodology, writing-review and editing. W.D. Foulkes: Supervision, funding acquisition, writing-original draft, project administration, writing-review and editing. M. Witcher: Conceptualization, resources, data curation, supervision, funding acquisition, validation, investigation, methodology, writing-original draft, project administration, writing-review and editing.

We would like to thank Dr. François-Michel Boisvert and the Department of Anatomy and Cell Biology, Université de Sherbrooke, for LC/MS SILAC mass spectrometry analysis. Metabolites were analyzed at the metabolomic facility of the Goodman Cancer Research Centre (GCRC), McGill University. We thank Dr. Russell Jones of the Van Andel Research Institute for insightful discussions regarding our metabolomics data. We would also like to thank Dr. Ivan Topisirovic for useful discussions and insights.

This study was supported by grants from the Canadian Institute for Health Research PJT-159759 and PJT-159663 to M. Witcher and FDN-148390 to W.D. Foulkes. M. Witcher is a research scholar of the FRQS. Goodman Cancer Research Centre's (GCRC) metabolomics facility at McGill University is supported by the Terry Fox Foundation, Quebec Breast Cancer Foundation, The Dr. R. John Fraser and Mrs. Clara M. Fraser Memorial Trust, GCRC, and McGill University.

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.