Loss of RBMS3 Confers Platinum Resistance in Epithelial Ovarian Cancer via Activation of miR-126-5p/β-catenin/CBP signaling

Among the various therapeutic modalities currently used for epithelial ovarian cancer (EOC), cytoreductive surgery combined with platinum-based chemotherapy is still acknowledged to be of the most effective treatments (1, 2). Despite the fact that more than 80% of patients with EOC exhibit encouraging initial clinical responses to chemotherapy with platinum-based drugs, the responses are often temporary. Approximately 25% of patients with EOC display multiple recurrences within 6 months after platinum therapy and more than 75% patients suffer relapse and progression years later, leading to a 5-year survival rate of only about 30% (3–5). Thus, the development of resistance to platinum-based chemotherapy is a serious limitation of long-lasting, effective treatments for patients with EOC.

Diverse range of intrinsic (de novo) mechanisms underlying platinum resistance, such as escape from platinum-induced apoptosis, increased platinum efflux, and enhanced DNA adduct repair, have been identified and usually coexist within the same tumor (6, 7). Meanwhile, intracellular signaling pathways, activated by heritable genetic/epigenetic alterations and the tumor microenvironment, also play critical roles in chemoresistance (8, 9). For instance, constitutively active Wnt/β-catenin signaling frequently occurs in various cancer types, confers resistance to oxaliplatin and gemcitabine through inhibition of apoptosis via upregulation of the antiapoptotic proteins Survivin and MMP-7, and contributes to cancer therapy failure through the repair of damaged DNA by inducing the DNA repair proteins MRE11 and LIG4 (10–12). Moreover, the drug transporters ABCB1 and ABCG2 can be directly upregulated by Wnt/β-catenin signaling, resulting in the extrusion of drugs out of cancer cells and multidrug resistance (13, 14). These results indicate that pharmacologic targeting of the Wnt/β-catenin signaling cascade may increase the clinical efficacy of chemotherapy (10, 15).

The acetyltransferase CREB-binding protein (CBP) is essential for Wnt/β-catenin signaling–mediated chemoresistance via diverse mechanisms (16–18). Thus, a specific small-molecule inhibitor of the CBP/β-catenin interaction, ICG-001, was screened and tested, and appears to have distinctive antitumor effects in multiple cancer types via inactivation of the Wnt/CBP/β-catenin signaling (19–24). In addition, the potential therapeutic value of PRI-724, which was derived from ICG-001 and is a second-generation CBP/β-catenin antagonist, is currently being tested in multiple clinical trials for the treatment of cancers, such as advanced colorectal and pancreatic cancers and myeloid malignancies (ClinicalTrials.gov ID: NCT02413853, NCT01764477, and NCT01606579). Hence, identifying key regulator(s) of Wnt/β-catenin/CBP signaling may reveal how PRI-724 can be successfully utilized for cancer treatment.

Genetic aberrations, such as chromosomal deletions, were recently reported to be associated with the cancer therapy response and may serve as potent predictive markers. For instance, loss of chromosome 8p contributes to taxane resistance and may server as a biomarker for microtubule inhibitor–based chemotherapy in breast cancer, and loss of PTEN predicts trastuzumab (Herceptin) resistance (25, 26). Anaplastic oligodendroglioma lacking 1p and 19q alleles are more responsive to radiochemotherapy (27). Heterozygous chromosomal deletions on the 3p arm are one of the most frequent allelic imbalances and a common early event in the formation of multiple types of cancers (28, 29). Herein, we found that loss of RNA-binding protein RBMS3, located on chromosome 3p, significantly enhanced the resistance of EOC to cisplatin (CDDP) via activation of Wnt/β-catenin/CBP signaling. Cotherapy of CDDP with PRI-724 exhibited remarkable therapeutic efficacy in preclinical models of EOC containing an RBMS3 deletion. These findings substantiate the importance of RBMS3 deletion in chemoresistance and may represent a new therapeutic strategy for EOC treatment.

Signed informed consent was obtained from all patients, and the investigation has been performed in accordance with the ethical standards according to the Declaration of Helsinki, national and international guidelines, and that the studies were approved by an authors' institutional review board.

All of the patients received standardized platinum-based chemotherapy. Platinum resistance or sensitivity was defined as relapse or progression within 6 months or after 6 months from the last platinum-based chemotherapy, respectively. Informed consent was obtained from all patients. Approvals from Sun Yat-sen University Cancer Center Institutional Research Ethics Committee were obtained for this study. A total of 150 paraffin-embedded, archived EOC specimens and 15 freshly collected EOC tissues were histopathologically and clinically diagnosed at Sun Yat-sen University Cancer Center (Guangdong, China) between 2005 and 2016. The clinical information of the samples is shown in Supplementary Tables S1 and S2. Detailed description provided in Supplementary Materials and Methods. Patient-derived EOC cells were prepared from fresh EOC tissues as described previously (30). EOC cell lines HEY were purchased from Bioleaf and SKOV3 were purchased from ECACC, which were cultured in completed DMEM (Gibco). All the cell lines been tested for Mycoplasma contamination and authenticated by short tandem repeat (STR) fingerprinting at Medicine Lab of Forensic Medicine Department of Sun Yat-Sen University (Guangzhou, China).

In the subcutaneous patient-derived xenografts (PDX) tumor model, freshly isolated clinical EOC patient tissues were subdivided into 2–3 mm³ pieces, coated with Matrigel (BD Biosciences) and media in a 1:1 ratio, and embedded within the subcutaneous space underneath the skin of female NOD/Shi-scid/IL-2Rγ^null (NOG) mouse (6–8 weeks old; CREA Japan Inc.). Tumor growth was monitored by measuring the tumor diameters, and the tumor volume was calculated using the equation (L × W²)/2. When the tumor became palpable, mice were intraperitoneally treated with vehicle (control) or CDDP (5 mg/kg) three times per week (as per cycle) for up to 6 weeks. At the end of treatment, the mice were sacrificed and the tumors were excised and weighed, and confirmed by histology. In the intraperitoneal tumor model, the indicated luciferase-expressing cells (1 × 10⁶) were injected intraperitoneally into female NOD/SCID or NOG mice. When the luminescence signal reached 2 × 10⁷ p/sec/cm²/sr, mice were intraperitoneally treated with vehicle (control), CDDP (5 mg/kg), or a combination of CDDP (5 mg/kg) and ICG-001 (5 mg/kg) three times per week (as per cycle) for up to 6 weeks. Mice were sacrificed when moribund as determined by an observer blinded to the treatment, and tumors were excised and paraffin-embedded. The indicated CDDP-resistant cells were isolated from tumors and reinjected intraperitoneally into a new cohort of mice to examine relapse within 4 months. Serial 4.0-μm sections were cut and subjected to IHC and hematoxylin and eosin (H&E) staining. After deparaffinization, sections were H&E-stained with Mayer hematoxylin solution, or IHC stained using antibodies of RBMS3 (1:100; Novus Biologicals, LLC), Ki-67 (Dako), or stained with TUNEL (In Situ Cell Death Detection Kit, TMR red, Roche Applied Science), and counterstained with Phalloidin (Alexa Fluor 488, Invitrogen) and DAPI (Sigma-Aldrich) according to manufacturer's protocols. The images were captured using the AxioVision Rel.4.6 computerized image analysis system (Carl Zeiss). All of the animal procedures were approved by the Sun Yat-sen University Animal Care Committee.

In brief, HEY cells expressing FLAG-RBMS3 were rinsed, resuspended in ice-cold 1 × PBS, and irradiated at 400 mJ/cm² and 200 mJ/cm² subsequently using a Stratalinker 2400. Cells were then scraped in ice-cold PBS and RNA-IP performed as described previously (31) using FLAG-agarose beads (Sigma-Aldrich). Furthermore, RBMS3-associated RNAs were immunoprecipitated and subjected to adapter ligation and RT-PCR to generate libraries for high-throughput sequencing, done on an Illumina Hi-Seq. Sequencing reads were preprocessed and adapter sequences were removed, and trimmed reads were mapped to the human genome (hg19). Identical alignments were collapsed in each sample to remove PCR replicates. Strand-specific read coverage was then calculated using the alignments from the sample.

Statistical tests for data analysis included Fisher exact test, log-rank test, χ² test and Student two-tailed t test. Bivariate correlations between study variables were calculated by Spearman rank correlation coefficients. Survival curves were plotted by the Kaplan–Meier method and compared by the log-rank test. The significance of various variables for survival was analyzed by univariate and multivariate Cox regression analyses. Statistical analyses were performed using the SPSS 11.0 statistical software package. Data represent mean ± SD. P values of 0.05 or less were considered statistically significant.

To better define the aberrant structure within chromosome 3 in tumors, we examined the patterns of chromosome 3 across 10,823 tumors spanning 33 human cancer types (TCGA dataset) using high-resolution single nucleotide polymorphism (SNP)-based copy number analysis. Chr3p was markedly deleted in 13 types of cancer, but deletion of Chr3q arm was only observed in pheochromocytoma and paraganglioma (false discovery rate (FDR)-corrected q value < 0.25, Fig. 1A). Copy number alterations of the 3p were significantly correlated with poorer overall and relapse-free survival (P = 0.016 and P = 0.045, respectively) in patients with these 13 cancer types (Supplementary Fig. S1A; Fig. 1B), indicating that deletion of 3p is extensively involved in cancer progression. Furthermore, statistical analysis showed that three deleted regions in 3p, namely Chr3p14.3-21.33, 3p22, and 3p23-24.1, were strongly correlated with shorter relapse-free survival (all P < 0.001), but only the 3p23-24.1 deletion was recognized as an independent prognostic factor for relapse-free survival in patients with cancer (P = 0.035; Fig. 1C and D). These pan-cancer data suggest that loss of 3p23-24.1 may contribute to poor patient outcome and cancer recurrence.

Cancer recurrence is mainly attributed to resistance to chemotherapy (7). Because EOC is the most common recurrent tumor disease, we used the HEY and SKOV3 EOC cell line to determine whether candidate genes in the Chr3p arm are involved in chemoresistance. The MTT cell viability assay showed that among 368 genes in three deleted regions of the Chr3p arm, silencing of RBMS3, one of the genes located in Chr3p23-24.1, had a significant inhibitory effect on CDDP-induced apoptosis both in HEY and SKOV3 EOC cells (Fig. 1E), suggesting that loss of RBMS3 may contribute to resistance of EOC to platinum chemotherapy. Consistent with this hypothesis, the low expression of RBMS3 mRNA was associated with shorter progression-free survival of patients with platinum-treated EOC as well as with gene signature–related CDDP resistance (Supplementary Fig. S1B–S1D).

Furthermore, we found that RBMS3 protein levels were clearly lower in RBMS3-deleted EOC tissues than in RBMS3-nondeleted tissues, as assessed by IHC and FISH methodologies (Fig. 1F). Furthermore, of the 15 freshly collected clinical EOC tissues, 5 RBMS3-deleted EOC tissues had lower RBMS3 mRNA and protein expression than RBMS3-nondeleted EOC (n = 10; Fig. 1G), suggesting that RBMS3 gene copy alterations represent one mechanism of protein deregulation. Importantly, RBMS3 deletion or low RBMS3 expression was significantly correlated with chemoresistance and shorter overall/relapse-free survival in patients with EOC treated with platinum-based therapy (Fig. 1H and I; Supplementary Fig. S1E and S1F; Supplementary Tables S1 and S2).

In addition, statistical analysis revealed that RBMS3 deletion was significantly associated with poorer overall/relapse-free survival in patients with cancer and with signature-related chemoresistance in 13 cancer types (Supplementary Fig. S1G–S1I). These data suggest a widespread correlation between RBMS3 deletion and chemoresistance in human cancers.

To examine the biological role of RBMS3 deletion in chemoresistance, a patient-derived xenograft (PDX) model was employed using the abovementioned 15 collected EOC tissues, of which 5 RBMS3^+/+ and 3 RBMS3^+/− tissues were successfully engrafted in female NOD/Shi-scid/IL-2Rγ^null (NOG) mice (Fig. 2A). Tumor-bearing mice were further treated with vehicle or CDDP, and both groups of xenograft tumors in vehicle-treated mice showed similar tumor growth kinetics (Fig. 2A), which suggest that the loss of RBMS3 did not affect tumor growth. However, compared with vehicle-treated tumors, CDDP chemotherapy dramatically reduced the growth rate of RBMS3^+/+ tumors, but RBMS3^+/− tumors displayed higher CDDP-resistant capability as indicated by their faster growth rate (Fig. 2A).

Serous ovarian carcinoma represents the largest proportion (up to 70%) of EOC cases. To further clarify the effects of RBMS3 deletion in EOC on CDDP resistance, two primary serous ovarian carcinoma cells OV-2 and OV-11, isolated from EOC tissues OV-#2 (RBMS3^+/+) and OV-#11 (RBMS3^+/−), were engineered to stably overexpress the luciferase reporter and were then intraperitoneally injected into NOG mice (Fig. 2B and C). When the bioluminescence signal of intraperitoneal tumors reached 2 × 10⁷ p/sec/cm²/sr, NOG mice were treated with CDDP (Fig. 2C). After 6 weeks/cycles of CDDP therapy, the tumors formed by OV-11 cells remained higher growth rate, as shown by stronger Ki67 signals and less terminal deoxynucleotidyl transferase dUTP nick end labeling (TUNEL)-positive cells compared with OV-2/tumors (Fig. 2C–E), indicating that RBMS3-deleted EOC cells possessed higher chemoresistant capability. In contrast, silencing RBMS3 via RBMS3 siRNA incorporated into dioleoylphosphatidylcholine (DOPC) nanoliposomes dramatically abrogated the inhibitory effect of CDDP on OV-2/tumors growth, which exhibited a decreased apoptotic index and increased proliferation, resulting in shorter survival of tumor-bearing mice (Fig. 2C–F).

To further support the contribution of RBMS3 deletion to CDDP resistance, serous ovarian cancer cell line HEY, the major subtype of EOC (32), was utilized to establish HEY/RBMS3 gRNA#1 and gRNA#2 EOC cells via CRISPR/Cas9 genome editing technology (Supplementary Fig. S2A). Concordant with the PDX results, both HEY/RBMS3 gRNA#1 and gRNA#2 tumors treated with CDDP exhibited a stronger bioluminescence signal, a lower percentage of TUNEL-positive apoptotic cells, and a higher proliferative index compare to control tumors, and the mice bearing HEY/RBMS3 gRNA#1 and gRNA#2 tumors survived significantly shorter than the CRISPR-Control mice (Fig. 2G–I; Supplementary Fig. S2B). Strikingly, as few as 100 HEY/RBMS3 gRNA#1 EOC cells, which were isolated from CDDP-treated mice and reinjected into secondary mouse recipients, were able to develop tumors in NOG mice around 3 months, while no tumor was formed by HEY cells after 4 months (Fig. 2J; Supplementary Fig. S2C), demonstrating the vital role of RBMS3 deletion in EOC recurrence.

Next, the effect of RBMS3 restoration on chemosensitivity was assessed in OV-11 and HEY cells with ectopic expression of RBMS3 cDNA (Supplementary Fig. S2D). As expected, upregulation of RBMS3 significantly enhanced the antitumor efficacy of CDDP on growth of OV-11/tumors and HEY/tumors, as indicated by the increased CDDP-induced apoptotic rate and decreased Ki67 signals (Supplementary Fig. S2E and S2F), which demonstrated that RBMS3 restores the sensitivity of EOC cells to CDDP in vivo.

In concordance with the in vivo results, RBMS3-deleted or silenced cells did not show obvious alterations of apoptotic rate or colony-forming ability compared with control cells under vehicle treatment. However, deletion or silencing of RBMS3 dramatically reduced CDDP-induced apoptotic death and promoted colony formation in EOC cells (Supplementary Fig. S3A–S3C). Moreover, we found that the IC₅₀ value of CDDP was significantly increased in RBMS3-deleted or RBMS3-silenced cells (Supplementary Fig. S3D). These results indicate that loss of RBMS3 confers resistance of EOC cells to CDDP in vitro.

There is ample evidence demonstrating that reduced drug accumulation is a major mechanism underlying CDDP resistance (6, 7). Interestingly, we found that the CDDP concentration was much lower in RBMS3-ablated/tumors, and that the content of intracellular and genomic DNA-bound CDDP was reduced in RBMS3–deleted/silenced cells (Supplementary Fig. S3E and F), suggesting that loss of RBMS3 promotes CDDP efflux. This was further confirmed by an efflux kinetic assay, which demonstrated that deletion/silencing of RBMS3 significantly increased the efflux velocity of CDDP (Supplementary Fig. S3G).

Formation and persistence of DNA adducts of CDDP are vital in inducing apoptosis (33). Concomitantly, we found that the number of CDDP-induced γH2AX foci was drastically decreased in RBMS3–deleted/silenced cells (Supplementary Fig. S3H). Furthermore, a comet assay showed that deletion or silencing of RBMS3 dramatically abrogated CDDP-induced DNA damage (Supplementary Fig. S3I). These results suggest that loss of RBMS3 may also promote DNA damage repair.

Consistent with our results that the loss of RBMS3 contributes to chemoresistance, apoptosis, drug efflux, and DNA repair, Gene ontology (GO) enrichment analysis in public datasets has shown that genes with GO biological process terms “Response to drug,” “Apoptosis,” “Transporter activity,” and “DNA repair” were enriched in RBMS3-downregulated cells (enrichment cut-off P < 0.05; Fig. 3A). Similar enrichments were also observed in our analysis results obtained from mRNA microarray and high-throughput sequencing crosslink immunoprecipitation (HITS-CLIP; Fig. 3B). Strikingly, we found that genes with GO terms, “Wnt signaling pathway,” “Negative regulation of canonical Wnt signaling pathway,” and “Regulation of Wnt signaling pathway” were also enriched (Fig. 3A and B). Meanwhile, GSEA analysis showed that RBMS3 levels were inversely associated with related gene signatures of Wnt signaling in EOC (Fig. 3C). These results suggest that RBMS3 may be involved in modulating Wnt signaling.

This hypothesis was further confirmed by several lines of evidence, which showed that deletion or silencing of RBMS3 increased, but overexpression of RBMS3 decreased, the transcriptional activity and DNA-binding activity of β-catenin, as well as the expression of multiple Wnt/β-catenin downstream genes and nuclear β-catenin levels (Fig. 3D–H). The inverse correlation of RBMS3 expression and activation of Wnt/β-catenin signaling, as indicated by nuclear β-catenin levels, was further verified in a cohort of clinical EOC tissues (Fig. 3I). Consistently, the reverse correlation of RBMS3 expression and activation of Wnt/β-catenin signaling was also found in multiple types of cancer via GSEA analysis (Supplementary Fig. S4A). Moreover, the duration of Wnt/β-catenin signaling, as indicated by the mRNA expression of Survivin, was dramatically prolonged in RBMS3-deleted cells but was much shortened in RBMS3-transduced cells (Supplementary Fig. S4B), suggesting that loss of RBMS3 sustains Wnt/β-catenin signaling in EOC cells. Importantly, inactivation of Wnt/β-catenin signaling in RBMS3-ablated cells by silencing β-catenin or TCF4 drastically decreased the levels of multiple Wnt/β-catenin signaling–regulated chemoresistance-related genes, such as ABCB1, Survivin, and MRE11, and cisplatin IC₅₀ and CDDP content, but increased CDDP-induced apoptotic rates (Supplementary Fig. S4C–S4F), demonstrating that Wnt/β-catenin signaling is required for RBMS3 deletion–mediated chemoresistance.

Next, we identified the molecular basis for the regulation of RBMS3 in Wnt/β-catenin signaling. Our HITS-CLIP data indicated that RBMS3 may associate with 22 genes involved in the negative regulation of Wnt/β-catenin signaling (Fig. 4A; Supplementary Fig. S5A). Furthermore, RNA immunoprecipitation (RIP) assay revealed that RBMS3 most significantly interacted with the mRNA of DKK3, AXIN1, BACH1, and NFAT5 in both OV-2 and HEY EOC cells (Fig. 4B). A concomitant decrease in DKK3, AXIN1, BACH1, and NFAT5, at both the mRNA and protein levels, was also observed in RBMS3-deleted/silenced cells, which were increased in RBMS3-transduced cells (Fig. 4C and D; Supplementary Fig. S5B). Importantly, loss of RBMS3 reduced, while overexpression of RBMS3 increased the mRNA half-life of these transcripts (Fig. 4E and F). Taken together, these results suggested that RBMS3 mediated the stabilization of these transcripts at the posttranscriptional level through association with their 3′UTR.

To complement our assessment of the effects of RBMS3 deletion on chemoresistance, the clinical relevance of these RBMS3-associated regulators was examined in EOC tissues. Integrating Kaplan–Meier plot and IHC analyses revealed that low levels of the mRNA or protein of these regulators coincided with failure of platinum therapy in EOCs (Supplementary Fig. S5C; Fig. 4G and H), which further supported the notion that RBMS3 deletion results in chemoresistance and recurrence of EOC. Moreover, individual force expressing DKK3, AXIN1, BACH1, and NFAT5 in RBMS3^+/− significantly inhibited TOP/FOP flash reporter activity and reduced the CDDP IC₅₀ value, but increased CDDP-induced apoptosis (Supplementary Fig. S5D; Fig. 4I–K), demonstrating that these regulators are functional effectors of RBMS3 loss–induced activation of Wnt/β-catenin signaling and chemoresistance.

Interestingly, we found that the RNA-binding site of RBMS3, AAUAAU, was complementary to the “seed” sequences of miR-126-5p (UUAUUA), suggesting that miR-126-5p may be involved in regulation of DKK3, AXIN1, BACH1, and NFAT5 mRNAs (Fig. 5A and B). As expected, miR-126-5p could associate with 3′UTR of DKK3, AXIN1, BACH1, and NFAT5 mRNA, resulting in a consistent reduction in luciferase activity of target mRNA 3′UTRs (DKK3, AXIN1, BACH1, and NFAT5; Supplementary Fig. S6A and S6B). However, mutation of the miR-126-5p “seed” sequence abrogated the suppressive ability of miR-126-5p (Supplementary Fig. S6B). Meanwhile, the expression of these negative regulators, at both the mRNA and protein levels, was decreased in miR-126-5p–transduced cells, but increased in miR-126-5p–silenced cells (Supplementary Fig. S6C; Fig. 5C).

We then further investigate the competitive effect between RBMS3 and miR-126-5p upon target mRNAs. As shown in Fig. 5D, the upregulated effect of RBMS3 on expression of DKK3, AXIN1, BACH1, and NFAT5 was dramatically decreased upon miR-126-5p overexpression, while blocking miR-126-5p enhanced the expression of these transcripts in RBMS3^+/− cells. Furthermore, RIP assays indicated that enforced expression of miR-126-5p significantly reduced the binding activity of RBMS3 with these target mRNAs, and upregulation of RBMS3 tremendously impaired the association of miR-126-5p with mRNA of DKK3, AXIN1, BACH1, and NFAT5 (Fig. 5E), indicating that RBMS3 and miR-126-5p exerted competitive regulation in these mRNAs. Accordingly, miR-126-5p overexpression suppressed the RBMS3-mediated stabilization of target mRNAs, resulting in Wnt/β-Catenin signaling activation, while downregulation of miR-126-5p significantly increased these negative regulators and inhibited Wnt pathway (Fig. 5F–H; Supplementary Fig. S6D and S6E). These observations suggest that miR-126-5p abrogates RBMS3-mediated stabilization of the mRNAs of DKK3, AXIN1, BACH1, and NFAT5, which activates Wnt/β-catenin pathway in ovarian cancer.

Acetyltransferase CBP is reported to play crucial roles in Wnt/β-catenin signaling–mediated chemoresistance (16–18). Multiple RBMS3-associated regulators, such as BACH1, KLF4, NFAT5, and WT1, modulate Wnt/CBP/β-catenin signaling via blocking the interaction of β-catenin and CBP (34–36), which raises the possibility that blocking CBP/β-catenin signaling confers chemosensitivity to RBMS3-deleted cells. To test this hypothesis, we first examined the effects of RBMS3 on the association of β-catenin with CBP. As shown in Supplementary Fig. S7A, the level of β-catenin–associated CBP, through pulling down the same amount of nuclear β-catenin, was greatly increased in RBMS3 knockdown cells, but was decreased in RBMS3-transduced cells compared with control cells, suggesting that RBMS3 modulated the binding efficiency of β-catenin and CBP. Furthermore, we found that deletion or silencing of RBMS3 significantly increased, but overexpressing RBMS3 decreased levels of β-catenin, CBP, and acylated-histone 3 in promoter of ABCB1, Survivin, and MRE11 (Supplementary Fig. S7B).

As expected, silencing of CBP resulted in significant decreased β-catenin transactivity and expression of ABCB1, Survivin, and MRE11 in RBMS3-deleted cells compared with control cells (Supplementary Fig. S7C–S7E), indicating that CBP is essential for RBMS3 loss–induced chemoresistance. Similar to the effect of CBP silencing in RBMS3-deleted cells, treatment with ICG-001, a specific inhibitor of CBP/β-catenin signaling via blockage of β-catenin/CBP interaction, in RBMS3-deleted cells also overtly reduced expression of ABCB1, BIRC5, and MRE11 compared with RBMS3-nondeleted cells (Supplementary Fig. S7F). Importantly, ICG-001 treatment of RBMS3-deleted/silenced cells significantly enhanced the inhibitory effects of CDDP on clonogenic growth and CDDP-induced apoptosis compared with RBMS3-nondeleted cells (Supplementary Fig. S7G and S7H). The superiority of the combination effects was further confirmed by the drug combination index, which indicated a synergistic effect of ICG-001 and CDDP in RBMS3-deleted cells, but not in RBMS3-nondeleted cells (Supplementary Fig. S7I). Collectively, these results demonstrate that ICG-001 chemosensitizes RBMS3-deleted EOC to CDDP.

To assess whether the in vitro findings could be recapitulated in vivo, the therapeutic efficacy of CDDP combined with PRI-724, a second-generation ICG-001 that has been tested in clinical trials in multiple cancers, was examined in a mouse model. As shown in Fig. 6A–F and Supplementary Fig. S8A and S8B, the curative effects of CDDP were significantly augmented by PRI-724 cotherapy in RBMS3-ablated/tumors, in a striking reduction in tumor volume, higher apoptotic index, and longer survival of tumor-bearing mice. In contrast, PRI-724 treatment only moderately increased the antitumor effects of CDDP on RBMS3^+/+/tumors (Fig. 6D–F). From these experiments, we concluded that the combination of PRI-724 and CDDP produced synergistic cytotoxicity in RBMS3-deleted EOCs. To validate our findings in patients with cancer, the PDX model in Fig. 2A was employed, and showed that cotherapy of PRI-724 with CDDP considerably impaired the tumor growth of RBMS3^+/− xenografts, whereas no additional antitumor effects of PRI-724 were observed in CDDP-treated RBMS3^+/+-implanted tumors (Fig. 6G). Therefore, our results demonstrate that RBMS3 loss enhances platinum chemoresistance via activating Wnt/β-Catenin/CBP pathway, suggesting cotherapy of PRI-724 and CDDP might be a valuable therapeutic strategy for RBMS3-deleted EOC (Fig. 6H).

Platinum-based chemotherapy, the cornerstone of modern treatment for EOC, can induce considerable clinical remission. However, most patients with advanced EOC invariably experience relapse and eventually die from this disease (1–4). In this scenario, understanding the molecular basis of platinum resistance may aid in the stratification of patients with EOC who are most likely to benefit from platinum-based chemotherapy, as well as the delineation of new targets for the pharmacologic intervention of patients with poor outcomes. Herein, we demonstrated that loss of RBMS3 significantly enhanced the resistance of EOC to cisplatin through activation of Wnt/β-catenin/CBP signaling, and that combined CDDP with PRI-724 exhibited significant therapeutic efficacy in preclinical models of RBMS3-deleted EOC. These results provided mechanistic and clinical insights into platinum-based resistance and suggested that the combination of platinum-based therapy with PRI-724 may serve as a potential tailored treatment for patients with RBMS3-deleted EOC. Considering that the action mechanisms of carboplatin and cisplatin are similar, loss of RBMS3 may also confer the resistance of EOC to carboplatin therapy. Meanwhile, because the gold-standard chemotherapeutic regimen for EOC is carboplatin combined with paclitaxel, the resistant effect of RBMS3 deletion on paclitaxel–carboplatin chemotherapy in EOC is investigated currently in our laboratory.

Aberrant activation of Wnt/β-catenin signaling has been observed in multiple tumors and contributed to tumor cell properties characteristic of the malignant phenotype, which raises the possibility that targeting the members of this signaling cascade may be an attractive therapeutic approach for the treatment of cancer (10, 15). Preclinical studies have underscored the potential value of inhibiting Wnt/β-catenin activation in multiple tumor types such as colorectal cancer, advanced myeloid malignancies, and prostate cancer (NCT01931046). Under this circumstance, rational combination with Wnt/β-catenin signaling inhibitors may improve the curative effects and reduce the side effects of platinum-based therapies. Recently, it has been reported that CBP plays essential roles in Wnt/β-catenin pathway–mediated chemoresistance via its association with β-catenin (16–24). Herein, we demonstrated that the inactivation of Wnt/CBP/β-catenin signaling via PRI-724 significantly enhanced the therapeutic efficacy of platinum therapy in RBMS3-deleted EOC, which further supports the notion that Wnt/β-catenin signaling is a druggable target for preventing chemoresistance and provides an attractive clinical strategy for treating RBMS3-deleted EOC. Because the potential therapeutic value of PRI-724 is currently tested in clinical trials for the treatment of multiple cancer types, such as advanced colorectal and pancreatic cancers, a clinical trial that incorporates PRI-724 in patients with ovarian cancer is also worthy to design.

Aberrant RNA metabolism induced by RNA-binding proteins has emerged as an important determinant of oncogenesis and chemotherapy resistance via the global rewiring of pathways. For instance, the RNA-binding protein hnRNPA0 was recognized as the “successor” to p53 for checkpoint control, which drives chemoresistance of p53-mutant tumor cells via p27/Gadd45a mRNAs (37). The RNA-binding protein LARP1, acts as a central posttranscriptional regulator of survival and promotes chemotherapy resistance by differentially regulating the stability of a selection of mRNAs (38). These studies suggest that dysregulation of RNA-binding proteins contribute to chemotherapy failure. RBMS3, a newly identified RNA-binding protein with two RNA-binding motifs, plays roles in craniofacial and pancreas development and liver fibrosis by modulating the half-life of mRNA via binding to 3′UTR (39–41). Although its downregulation has been found in multiple cancers (42–45), RBMS3 reduction–mediated cancer progression was found to be through the transcriptional repression c-Myc (45). In contrast, the RBMS3 protein is predominantly restricted to the cytoplasm, suggesting that its major role is in RNA metabolism rather than transcription. In this study, we demonstrated that RBMS3 directly associated and stabilized the mRNA of multiple negative regulators in Wnt/β-catenin cascade via competitively preventing the repression ability of miR-126-5p, and that the loss of RBMS3 provided refuge protection for EOC cells from platinum therapy via activation of Wnt/β-catenin/CBP signaling. These results present reveal a novel RNA metabolic function of RBMS3 and may shed light on the breadth and extent of aberrant RNA metabolism in the context of chemoresistance.

It was recently reported that approximately 25% of the genome is affected by arm-level chromosomal deletions in cancers (46), and analysis of TCGA data in pan-cancer revealed that heterozygous chromosomal deletions on the 3p arm, among copy number alterations, is most associated with the poor survival of cancer patient (29). We found that the loss of Chr3p23-24.1 in 13 cancer types was significantly correlated with shorter relapse-free survival, and as such, may serve as an independent prognostic factor. We further demonstrated that loss of RBMS3, one of the genes in Chr3p23-24.1, correlates with the poor response and resistance of platinum therapy in EOC. Therefore, future studies are needed to determine the molecular mechanisms underlying chemoresistance induced by loss Chr3p23-24.1. The results of these studies will not only contribute to our understanding of 3p23-24.1-contanining genes to cancer progression and responses to platinum-based therapy but will also lead to the characterization of specific tumor patients' groups who may benefit from personalized therapies.

No potential conflicts of interest were disclosed.

Conception and design: L. Song, J. Li

Development of methodology: G. Wu, L. Cao, J. Zhu

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.):

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): G. Wu, L. Cao, J. Zhu, Z. Tan, M. Tang, Z. Li, Y. Hu, R. Yu, S. Zhang

Writing, review, and/or revision of the manuscript: J. Li

Study supervision: L. Song, J. Li

This work was supported by Natural Science Foundation of China [no. 81830082 (to J. Li), 91740119 (to J. Li), 91529301 (to J. Li), 81621004 (to J. Li), 91740118 (to L. Song), 81773106 (to L. Song), and 81530082 (to L. Song)]; Guangzhou Science and Technology Plan Projects [201803010098 (to J. Li)]; Guandong Natural Science Foundation [2018B030311009 (to J. Li), 2016A030308002 (to L. Song)]; The Fundamental Research Funds for the Central Universities [no. 17ykjc02 (to J. Li)].

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.