FBW7 suppresses ovarian cancer development by targeting the N⁶-methyladenosine binding protein YTHDF2

Totally 120 Chinese patients diagnosed with high-grade serous ovarian carcinoma and 10 Chinese patients with ovarian cyst were involved in this study. All patients had surgical resection at Fudan University Shanghai Cancer Center (FUSCC). The tissue samples were collected immediately after surgery and stored in preservation buffer at − 80 °C. Informed consent was obtained from all patients, and the use of clinical samples in this study was approved by the ethics committee of FUSCC. Ovarian cancer cell lines used in this study, SKOV3, OVCA420, OVCA429 and OVCAR8 were obtained from ATCC.

The specific shRNAs targeting FBXW7 or YTHDF2 (Supplementary Table S1) were cloned into the lentiviral vector pLKO.1 (Sigma-Aldrich, St louis, MO, USA). The specific plasmid overexpressing FBW7 or YTHDF2 was generated by inserting the full-length cDNA amplified by PCR into the lentiviral vector pCDH (Sigma-Aldrich). The 293 T cells were transfected with pCDH or pLKO.1 carrying the specific sequences, along with the packaging plasmids, psPAX2 and pMD2.G. The virus particles were generated and collected 48 h post-transfection and used to infect ovarian cancer cell lines which were then selected by puromycin (Sigma-Aldrich) for 3 days. Knockdown or overexpression efficiency was confirmed at both mRNA and protein levels.

Total RNA was extracted from ovarian cancer samples or cell lines by using TRIzol reagent (Life Technologies, Waltham, MA, USA). Complementary DNA (cDNA) synthesis was performed with PrimeScript RT reagent Kit (Takara, Japan) using ~ 1 μg RNA each sample. RT-qPCR reactions were conducted using TB Green Premix (Takara, Japan). The primers for FBXW7, YTHDF2, BMF, KCNK3, MYBL2, LEMD1, SIRPD, LY6K, SEMA3C, C1QTNF1, ZDHHC14 and MOCS3 are listed in Supplementary Table S1.

Cells were harvested and lysed with RIPA buffer. Proteins were extracted and loaded in SDS-PAGE, and transferred onto PVDF membrane (Millipore, Billerica, MA, USA). After blocking with Bovine Serum Albumin (Beyotime, China) and sequential incubation with the primary and secondary antibodies, the blots were detected using the ECL detection kit (Millipore). The anti-FBW7 antibody was purchased from Bethyl Laboratories (Montgomery, TX, USA), and the anti-YTHDF2 and anti-GAPDH antibodies were purchased from Proteintech (Wuhan, Hubei, China).

Cells were harvested and lysed with mild RIPA buffer directly on plate for 30 min. Meanwhile, 50 μl dynabeads protein G (Life Technologies) were incubated with 3 μg antibody at room temperature for 1 h. Then mix the protein lysate with the beads-antibody complex and incubate overnight at 4 °C. Beads were washed three times with lysis buffer. Bound proteins and 10% inputs were detected by IB.

2 × 104 HeyA8 or OVCAR8 cells were seeded on coverslips in a 24-well dish and further incubated for 12 h. Cells were fixed with 4% paraformaldehyde for 15 min, then incubated with 0.5% Triton X-100 in PBS for 15 min, and washed three times in PBS. After blocking in 1% BSA for 30 min, the cells were incubated with rabbit anti-YTHDF2 antibody Proteintech (Wuhan, Hubei, China) and mouse anti-FBW7 antibody (Santa Cruz Biotechnology, CA, USA) overnight at 4 °C. After washing several times in PBS, the coverslips were incubated with anti-mouse secondary antibody conjugated with Alexa 488 and Cy3-conjugated anti-rabbit secondary antibody for 30 min. After washing several times in PBS, the Slides were mounted and sealed with DAPI Staining Solution. The images were acquired using a confocal microscope (Leica, Wetzlar, Germany).

Immunohistochemistry was performed using a primary antibody against YTHDF2 antibody Proteintech (Wuhan, Hubei, China) and FBW7 antibody (Bethyl Laboratories, Montgomery, TX, USA). The immunostaining results were examined by two researchers independently. Immunohistochemical evaluation was based on the intensity and percentage of membranous and cytoplasmic reactivity. A 10% expression threshold was defined as positive. Then, no positive cells, less than 10% of positive cells, 10–50% of positive cells, more than 51% of positive cells were defined as negative, weak, moderate, strong expression respectively. Tumors with moderate and strong staining were regarded as high expression.

SKOV3 cells were transfected with different combinations of plasmids encoding FBW7, YTHDF2 and His-Ub. After 48 h, cells were harvested and divided into two parts, one for IB and the other for the ubiquitination assay. Briefly, cell pellets were lysed with buffer I (8 M urea, 0.1 M Na2HPO4/ NaH2PO4 (pH 8.0), 10 mM Tris-HCl (pH 8.0), 10 mM β-mercaptoethanol, 5 mM Imidazole) and incubated with Ni-NTA beads (Takara, Japan) for 6 h at room temperature. Beads were washed twice with buffer I, and twice with buffer II (8 M urea, 0.1 M Na2HPO4/NaH2PO4 (pH 6.3), 10 mM Tris-HCl (pH 6.3), 10 mM β-mercaptoethanol). The bound protein complex was eluted and analyzed by IB.

For the cell proliferation assay, 2000 cells were seeded into 96-well plates, cell viability was assessed for 5 consecutive days by the Cell Counting Kit-8 (CCK-8) (Dojindo, Japan). For the colony formation assay, 1000 cells were seeded into 6-well plates for 2 weeks, and colonies were stained with crystal violet and counted. For the anchorage-independent cell growth assay, 20,000 cells were seeded into 6-well plates coated with soft agar for 4 weeks, and the growing colonies were counted under the microscope. The colony formation assay was used to evaluate the long-term cell growth on a solid surface, while the anchorage-independent cell growth assay was employed to assess the ability of transformed cells to grow independently of a solid surface.

In vivo experiments were all conducted using 5-week old Female BALB/c nude mice (Shanghai SLAC Laboratory Animal Co., Ltd). 5 × 10⁵ cells were suspended in 100 μl PBS and then were inoculated subcutaneously. After the tumors were formed, the tumor volumes were measured every 3 days. Mice were sacrificed when the tumor diameter reached about 15 cm, and the tumors were excised and weighted. Tumor volume was calculated based on the formula: volume = length × width² × 0.5. Animal experiments were all performed according to the Control of Department of Laboratory Animal Science in Shanghai Medical College of Fudan University and the animal ethic principles.

RNA m⁶A quantification by LC-MS/MS was conducted as described previously [31]. Briefly, total RNAs were isolated using TRIzol reagent (Life Technologies). 200 ng mRNA was incubated with nuclease P1 (Sigma-Aldrich) in 20 μl buffer containing 25 mM NaCl, 2.5 mM ZnCl₂ at 37 °C for 2 h, then added 2.2 μl NH₄HCO₃ (1 M) and alkaline phophatase (Sigma-Aldrich), and incubated at 37 °C for 2 h. Following centrifugation at 13000 rpm for 10 min at 4 °C, 10 μl of the solution was analyzed by LC-MS/MS at Mass Spectrometry Application Research Center of the Institutes of Biomedical Sciences in Fudan University. The expression level of m⁶A was dichotomized for OS before the log-rank test according to optimal cutoff values calculated by the “surv_cutpoint” function of the “survminer” R package.

Cells were incubated with actinomycin D (5 μg/ml) for 0 h, 3 h or 6 hoursfollowed by RNA extraction. The half-life of BMF mRNA was analyzed by quantitative RT-PCR. For the 3-deazaadenosine (DAA) treatment assay, cells were incubated with 50uM DAA followed by RNA extraction. The level of BMF mRNA was analyzed by quantitative RT-PCR.

Polyadenylated RNAs were prepared and sonicated into fragments of 100–200 nt. A small portion of the RNA fragments was saved as input samples. MeRIP was performed as previously described [32]. Briefly, 4 μg fragmented RNAs were incubated with 2 μg anti-m⁶A antibody (Synaptic Systems, Goettingen, Germany) in 1 x IP buffer (10 mM TrisHCl, pH 7.4, 150 mM NaCl, 0.1% NP-40) for 2 h at 4 °C. The m⁶A-IP complex was then incubated with Dynabeads protein A (Life Technologies) for 2 h at 4 °C. The bound RNAs were washed and eluted through competition with N6-methyladenosine (Santa Cruz Biotechnology, CA, USA) and then purified by the RNA cleanup kit (Zymo Research, CA, USA). The NEBNext® Ultra II Directional RNA Library Prep Kit (New England Biolabs, MA, USA) was used for library construction. Library sequencing was performed on an illumina Hiseq instrument with 150 bp paired-end reads by Cloudseq Biotech Inc. (Shanghai, China).

Since a paucity of evidence had been found on the role of FBW7 in ovarian cancer, we first determined its clinical implication by comparing the expression of FBW7 in both cancerous and non-cancerous ovarian tissues. FBW7 was markedly downregulated in ovarian cancer tissues (Fig. 1a, b and Supplementary Fig. 1A). Consistently, FBW7 was also found to be significantly reduced in the TCGA RNA-Seq dataset of 586 ovarian cancer samples (Supplementary Figure 1B). Next, the FBW7 expression in the ovarian cancer tissue microarrays, containing 120 tumor samples, was analyzed by immunohistochemistry (IHC) staining and as thus the ovarian cancer specimens could be categorized into two groups with low or high expression level of FBW7 (Fig. 1c and d). We then explored the relationship between FBW7 expression and the clinicopathological features of ovarian cancer. However, no significant correlation was observed between FBW7 expression and age, tumor stage, ascites, lymph node metastasis or menopause (Supplementary Table S2; P > 0.05). The Kaplan-Meier survival analysis showed that high level of FBW7 is significantly associated with favorable overall survival in ovarian cancer (Fig. 1e, P = 0.04), which is consistent with the result from the TCGA ovarian cancer cohort (Supplementary Fig. 1C). Additionally, high level of FBW7 was moderately, but not significantly, correlated with better progress-free survival (PFS) (Fig. 1f), probably due to the limited sample size. Therefore, these observations indicate that FBW7 is a favorable prognostic marker that may suppress ovarian tumorigenesis and progression.

FBW7 was a typical tumor suppressor, mainly involved in the process of ubiquitylation and degradation of many onco-proteins. N⁶-methyladenosine (m⁶A) has emerged as an abundant reversible modification, which has been reported to play critical roles in physiological and pathological processes. However, the regulation of the m⁶A modification system remains elusive. Here we measured m⁶A levels in 60 human ovarian cancer tissue samples. According to the optimal cutoff point calculated by “survminer” R package, samples with a m⁶A expression no more than 0.01 were considered as the m⁶A low expression population, while the rest was considered as the m⁶A high expression population. Interestingly, m6A levels were elevated in patients with tumors expressing higher levels of FBW7 (Fig. 1g), and better overall survival was significantly correlated with higher levels of m⁶A (Fig. 1h), which suggested FBW7 might be an upstream regulator for the m⁶A modification system.

To investigate the biological function of FBW7 in ovarian cancer, we generated FBW7-stably overexpressing SKOV3 and OVCAR429 cell lines, both of which sustain low expression level of endogenous FBW7 (Fig. 2a, Supplementary Fig. 2). Overexpression of FBW7 significantly suppressed ovarian cancer cell proliferation (Fig. 2b), colony-forming ability (Fig. 2c), and anchorage-independent cell growth (Fig. 2d). The tumor inhibitory activity could be attributed to the induction of apoptosis, as ectopic FBW7 strikingly augmented the Annexin V-positive cell population (Fig. 2e). Furthermore, we established a xenograft mouse model to test if FBW7 suppresses ovarian tumor growth in vivo. In accordance with the cell-based results, ectopic FBW7 markedly repressed tumor growth in mice (Fig. 2f), as reflected by the significant inhibition of the tumor volume and weight when compared to the control group (Fig. 2g and h). In addition, we also generated FBW7-depleted HeyA8 and OVCAR8 ovarian cancer cell lines using two independent shRNAs (Supplementary Fig. 3A). Consistently, ablation of FBW7 significantly promoted ovarian cancer cell proliferation (Supplementary Fig. 3B), colony-forming ability (Supplementary Fig. 3C), and anchorage-independent cell growth (Supplementary Fig. 3D). Taken together, these results demonstrate that FBW7 plays a tumor suppressive role in ovarian cancer.

To decipher the molecular mechanism behind FBW7-mediated regulation of the m⁶A modification system, we conducted a co-IP assay using the anti-FBW7 antibody coupled with mass spectrometry analysis to identify the FBW7-interacting proteins in the FBW7-stably overexpressing SKOV3 cells (Supplementary Fig. 4A-C, Supplementary table 3). This screening led to identification of YTHDF2 as one of the FBW7-binding proteins. To validate this interaction, we performed the reciprocal co-IP assays and indeed verified that FBW7 binds to YTHDF2 in SKOV3 cancer cells (Fig. 3a). The immunofluorescence (IF) staining suggested that the interaction may occur in the cytoplasm (Fig. 3b). Since FBW7 is a component of the SCF E3-ubiquitin ligase, we sought to determine if FBW7 regulates YTHDF2 protein stability through their physical interaction. By introducing the FBW7-encoding plasmid into SKOV3 and OVCAR429 cells, we found that the protein level of YTHDF2 dramatically declined upon FBW7 overexpression, while this reduction is completely restored by treatment of cells with the proteasome inhibitor MG132 (Fig. 3c). Additionally, overexpression of FBW7 shortened the half-life of YTHDF2 protein as indicated by the cycloheximide-chase experiment (Fig. 3d). In line with these results, we also showed that FBW7 drastically enhances YTHDF2 ubiquitination (Fig. 3e), which was validated by another set of ubiquitination assays using the anti-YTHDF2 antibody (Fig. 3f). Therefore, these findings demonstrate that FBW7 prompts ubiquitination-mediated proteolytic degradation of YTHDF2 in ovarian cancer to regulate the m⁶A modification system.

As YTHDF2 is a substrate of the tumor suppressor FBW7 for degradation, we inquired if the inverse correlation between the two proteins also exists in primary ovarian cancer tissues. First, we analyzed the expression of YTHDF2 in the cancerous and non-cancerous ovarian tissues, and found that YTHDF2 is frequently elevated in ovarian cancer (Fig. 4a). Consistently, YTHDF2 was significantly upregulated in ovarian cancer samples versus normal ovarian tissues in the TCGA database (Supplementary Fig. 5A). We then examined YTHDF2 expression using the ovarian cancer tissue microarrays as mentioned above (Fig. 1c) and indeed found a negative correlation between the expression of YTHDF2 and FBW7 (Fig. 4b and c), which is in agreement with the results illustrated in Fig. 3c-f. According to the expression level of YTHDF2, these tumor specimens were categorized into two groups (Fig. 4d) and the clinicopathological features of each group were analyzed. However, no significant correlation was observed between YTHDF2 expression and age, tumor stage, ascites or lymph node metastasis (Supplementary table 4, Supplementary table 5; P > 0.05). The Kaplan-Meier survival analysis displayed that increased expression of YTHDF2 is significantly correlated with worse overall survival (OS) in ovarian cancer (Fig. 4e, P = 0.0013), which is in line with the result from the TCGA ovarian cancer cohort (Supplementary Fig. 5B). Nevertheless, YTHDF2 is not an appropriate prognostic marker for the progress-free survival (PFS) (Fig. 4f). And the OS of the FBW7 high expression /YTHDF2 low expression group was better than that of the BW7 low expression /YTHDF2 high expression group (Supplementary Fig. 13). In multivariate analysis, YTHDF2 expression was adverse independent prognosticators for OS (Supplementary table 6). Thus, these observations strongly suggest that YTHDF2 may promote survival and growth of ovarian cancer, which is addressed as follows.

Although YTHDF2 was found to be involved in the development of several cancers [26‐29, 33], it remains largely elusive if the protein is required for ovarian cancer cell survival and growth. To test this speculation, the YTHDF2-depleted SKOV3 and OVCAR420 cell lines were generated using two independent shRNAs (Fig. 5a). We showed that ablation of YTHDF2 significantly prohibits proliferation (Fig. 5b), colony-forming ability (Fig. 5c), and anchorage-independent growth (Fig. 5d) of the two ovarian cancer cell lines. Also, the flow cytometry analysis indicated that knockdown of YTHDF2 drastically triggers ovarian cancer cell apoptosis (Fig. 5e). Moreover, the xenograft mouse model revealed that YTHDF2 depletion markedly hampers ovarian tumor growth in vivo (Fig. 5f) leading to significant reduction of the tumor volume and weight (Fig. 5g and h). And Knockdown of YTHDF2 significantly restores cell proliferation and colony-forming ability induced by FBW7 depletion (Fig. 5i-k). Thus, these findings indicate that YTHDF2 is essential for ovarian cancer cell propagation. To confirm these results, we also evaluated the biological function of ectopic YTHDF2 in ovarian cancer cell lines. By examining the expression of YTHDF2 in multiple ovarian cancer cell lines, we generated YTHDF2-stably overexpressing lineages using OVCAR433 and OVCAR8 cells that sustain relatively low expression of endogenous YTHDF2 (Supplementary Fig. 6 and 7A). As expected, overexpression of YTHDF2 significantly boosted cell proliferation (Supplementary Fig. 7B), colony-forming ability (Supplementary Fig. 7C), and anchorage-independent growth (Supplementary Fig. 7D). Collectively, we demonstrate that YTHDF2 may drive ovarian cancer development, and that FBW7 restrains ovarian carcinogenesis by inducing proteolytic degradation and thus compromising the oncogenic activity of YTHDF2.

Since YTHDF2 was shown to prompt the decay of the m⁶A-modified transcripts [24], we determined if FBW7 regulates cellular m⁶A enrichment through inhibition of YTHDF2. By employing a LC-MS/MS assay that allows quantification of m⁶A level in cells, we showed that ablation of YTHDF2 significantly increases m⁶A abundance in the ovarian cancer cell lines, OVCAR8 and HEY (Fig. 6a), which is in accordance with a previous study performed in HeLa cells [24]. Remarkably, overexpression of FBW7 led to significant elevation of m⁶A level in SKOV3 and OVCAR429 cell lines (Fig. 6b), which phenocopies the effect of YTHDF2 depletion. In addition, we assessed m⁶A enrichment in ovarian cancer tissues with high or low expression level of FBW7, which again verified that FBW7 expression is positively associated with m⁶A level in tumors (Fig. 1g). Of note, we also ascertained that FBW7-mediated regulation of m⁶A abundance is dependent on YTHDF2, as evidenced by the fact that knockdown of YTHDF2 dramatically neutralized the effect of FBW7 depletion on m⁶A level (Fig. 6c).

Next, we examined if YTHDF2 orchestrates gene expression in ovarian cancer by conducting an RNA-sequencing analysis (Fig. 6d). Ablation of YTHDF2 led to upregulation of 348 genes and downregulation of 320 genes in SKOV3 cells. We also performed the genome-wide mapping of the m⁶A modification in control and YTHDF2-depleted SKOV3 cell lines through a MeRIP (m⁶A)-sequencing assay. The results indicated that the m⁶A peaks mainly locate in the CDS regions (42.3%), then the stop codon sites (32.9%), start codon sites (14.5%), 5′ UTRs (7.7%) and 3′ UTRs (2.6%) (Fig. 6e). The m⁶A consensus sequence identified in SKOV3 cells was GGAC [U/A] (Fig. 6f) that is consistent with previous studies [24, 34], indicating that YTHDF2 may also bind to m⁶A in ovarian cancer. By integrating the RNA-Sequencing and m⁶A-Sequencing results, we found that 25 genes with 39 peaks are consistently elevated upon YTHDF2 knockdown (Fig. 6g and Supplementary table 4). Among these dysregulated genes, BMF was considered an interesting target given that it encodes a pro-apoptotic BH3-only Bcl2 family protein required for inhibiting pro-survival Bcl2 members and triggering apoptosis [35] and that ablation of YTHDF2 exhibited higher peak enrichment compared to the control group (Fig. 6h). We further testified that ablation of YTHDF2 significantly increases BMF mRNA expression level (Fig. 6i) and prolongs its half-life (Fig. 6j), which validates BMF mRNA as a target of YTHDF2 for degradation. Since the m⁶A modification is indispensable for YTHDF2-induced mRNA decay, we determined if inhibition of m⁶A stabilizes BMF mRNA. As shown in Fig. 6k, treatment of SKOV3 cells with the global methylation inhibitor, 3-deazaadenosine (DAA), significantly elevated the mRNA level of BMF, suggesting that methylation is critical for degradation of BMF mRNA (Fig. 6k). The interaction between YTHDF2 and BMF was verified by the RIP assay (Fig. 6l). And overexpression of YTHDF2 significantly reduces the expression of BMF (Supplementary Fig. 8A). The effect of YTHDF2 on cell proliferation and apoptosis can be reversed by the restoration of BMF expression (Supplementary Fig. 8B-E). Moreover, overexpression of FBW7, like knockdown of YTHDF2, strikingly induced BMF mRNA level (Fig. 6m), which is owing to YTHDF2 degradation by ectopic FBW7 (Fig. 3c-f). Further, the overexpression of FBW7 or knockdown of YTHDF2 increased the levels of m6A modification of BMF pre-mRNA respectively, compared with modification levels of control vectors in SKOV3 cells (Fig. 6n).

We also showed that the FBW7-mediated regulation of BMF is dependent of YTHDF2 (Fig. 6o and Supplementary Fig. 12,). Importantly and in line with these in vitro results, we also revealed that the expression of BMF is inversely correlated with YTHDF2 expression (Fig. 6p), whereas is positively associated with FBW7 expression (Fig. 6q, Supplementary Fig. 10), in ovarian cancer tissues. And the mRNA levels of BMF were elevated in patients with tumors expressing higher levels of FBW7, while YTHDF2 mRNA levels demonstrate no significant difference (Supplementary Fig. 11). Collectively, our results clearly demonstrate that FBW7 provokes the expression of pro-apoptotic BMF through an m⁶A-dependent mechanism by interacting with and degrading the m⁶A reader YTHDF2 in ovarian cancer (Fig. 6r).