CircPLPP4 is upregulated in CDDP-resistant OC cells and tissues
To identify the crucial circRNAs that induce CDDP resistance of OC, RNA-Seq analysis was conducted in five CDDP-resistant and five CDDP-sensitive OC tissues. Unsupervised hierarchical clustering and volcano plot showed that 3218 upregulated and 1891 downregulated circRNAs in CDDP-resistant OC tissues compared with CDDP-sensitive tissues according to the Log (fold-change) ≥ 2 and P-value < 0.05 (Fig. 1A, B). Consistent with the RNA-Seq results, qRT-PCR identified hsa_circ_0008896 as markedly upregulated in CDDP resistant OC cells and OC tissues (Fig. 1C). qRT-PCR analysis also showed that circPLPP4 significantly upregulated in two CDDP resistant OC cells and its corresponding parent cells (Fig. 1D). According to circBase annotation[21], hsa_circ_0008896 is generated from exon 2 to 4 of the PLPP4 transcript (named as circPLPP4) by back-splicing, which was validated by Sanger sequencing (Fig. 1E).
qRT-PCR analyses using divergent and convergent primers confirmed that circPLPP4 was formed by back-splicing rather than genomic rearrangements or transsplicing. RT-PCR revealed that circPLPP4 could only be detected in cDNA, whereas PLPP4 was amplified from both cDNA and gDNA, which also excluded the possibility of circPLPP4 formation from genomic rearrangements or trans-splicing (Fig. 1F, G). Moreover, the linear form of PLPP4, but not circPLPP4, was easily digested by RNase R (Fig. 1H). Actinomycin D was used to inhibit transcription and examine the halflife of circPLPP4 in A2780 CDDP and SKOV3 CDDP cells, which showed that that circPLPP4 was more stable than PLPP4 mRNA (Fig. 1I, J). Subsequently, qRTPCR assays showed that circPLPP4 is located mainly in the cytoplasm of OC cells (Fig. 1K, L). FISH analysis also demonstrated that circPLPP4 is predominately distributed in the cytoplasm of A2780 CDDP and SKOV3 CDDP cells (Fig. 1M, N).
CircPLPP4 expression correlates with poor prognosis in patients with OC
The clinical significance of circPLPP4 expression was further evaluated in 166 OC specimens (Table S1). ISH staining revealing that circPLPP4 expression was upregulated significantly in OC specimens (Fig. 2A, B). Moreover, correlation analysis showed that high circPLPP4 expression was markedly associated with CDDP resistance and patient vital status (Fig. 2C, Figure S1A). Importantly, patients with OC with high circPLPP4 expression experienced significantly shorter overall/relapse-free survival among patients receiving platinum-based treatment (Fig. 2E. Figure S1B). High circPLPP4 expression could serve as an independent prognostic factor for the prognosis of tumor CDDP resistance and patient survival (Fig. 2F). Notably, ISH staining showed that circPLPP4 expression was increased significantly in patients with CDDP resistant OC (Fig. 2G-I).
Downregulation of circPLPP4 enhances CDDP sensitivity of CDDP-resistant OC cells in vitro
Next, we assessed the effect of silencing circPLPP4 using a antisense oligonucleotides (ASOs) targeting the junction site of circPLPP4 on CDDP resistant cells (A2780 CDDP and SKOV3 CDDP cells) with high circPLPP4 expression. qRT-PCR showed that circPLPP4-ASO treatment markedly reduced circPLPP4 expression, but not PLPP4 mRNA expression (Fig. 2J, Supplemental Fig. 2A). CircPLPP4 inhibition decreased the viability of A2780 CDDP and SKOV3 CDDP cells under cisplatin treatment (Fig. 2K, Supplemental Fig. 2B). Additionally, circPLPP4 silencing reduced the number of cell colonies and increased the proportion of apoptotic cells after cisplatin treatment significantly (Fig. 2L, M, Supplemental Fig. 2C-E). Western blotting was used to explore the underlying mechanisms of these functions. Under cisplatin treatment, circPLPP4 silencing in A2780 CDDP and SKOV3 CDDP cells enhanced the protein level of cleaved caspase-3 and cleaved PARP (activated form) (Fig. 2N, Supplemental Fig. 2F). Cisplatin induces DNA crosslinking and promotes H2AX phosphorylation. γ-H2AX is used as a sensitive marker of DNA damage and is known to bind to the BRCT (BRCA1 car-boxyl-terminal) domain of the BRCA1 gene, which predicts sensitivity to cisplatin treatment[22, 23]. Importantly, circPLPP4-ASO treatment markedly increased the protein level of γ-H2AX compared with that in ASO-control cells (Fig. 2O, Supplemental Fig. 2G). Knockdown of circPLPP4 in A2780 CDDP and SKOV3 CDDP cells decreased BRCA1 expression (Fig. 2O, Supplemental Fig. 2G). Moreover, circPLPP4 overexpression in OC cells had the opposite effects (Supplemental Fig. 3A-K).
circPLPP4 exerts its function by sponging miR-136
One of the methods by which circRNAs exert their function is by sponging miRNAs. Given that circPLPP4 locates mainly in the cytoplasm, we hypothesized that circPLPP4 promotes cisplatin resistance by binding miRNAs. First, we used an RIP assay using antibodies against argonaute 2 (AGO2) in A2780 CDDP and SKOV3 CDDP cells (Fig. 3A). Then, Circular RNA Interactome algorithm[24] was used to predict circRNA-miRNA interactions. This revealed that hsa-miR-1197, hsa-miR-1231, hsa-miR-1294, hsa-miR-1299, hsa-miR-136, hsa-miR-188-3p, hsa-miR-197, hsa-miR-330-3p, hsa-miR-335, hsa-miR-503, hsa-miR-586, hsa-miR-622, hsa-miR-633 might be associated with circPLPP4 (Fig. 3B). Next, we explored that whether these candidate miRNAs could bind directly to circPLPP4. A biotin-labeled circPLPP4 probe was verified to pull down circPLPP4 in A2780 CDDP and SKOV3 CDDP cells, and the pulldown efficiency increased in cell lines stably overexpressing circPLPP4 (Fig. 3C, D). Furthermore, qRT-PCR was performed to examine whether the 13 candidate miRNAs were pulled down by the circPLPP4 probe. Only miR-136 was abundantly pulled down in both A2780 CDDP and SKOV3 CDDP cells (Fig. 3E, F). Additionally, biotin-labeled miR-136 mimics were used to verify the direct binding of miRNA and circPLPP4: The biotin-labeled miR-136 captured more circPLPP4 than the biotin-labeled negative control (Fig. 3G). A luciferase assay in A2780 CDDP and SKOV3 CDDP cells showed that the overexpression of miR-136 decreased the activity of the wild-type Luc-circPLPP4 reporter gene. By contrast, overexpressed miR-136 had no effect on the activity of a Luc- circPLPP4-mutant reporter gene (Fig. 3H-J).
FISH showed colocalization of circPLPP4 and miR-136 in A2780 CDDP (CDDP-resistant) and A2780 cells (CDDP-sensitive). The expression of circPLPP4 was markedly higher in A2780 CDDP cells than CDDP sensitive A2780 cells; whereas miR136 expression showed the opposite results (Fig. 3K, L). Consistently, we found that circPLPP4 was significantly higher, whereas miR-136 levels were obviously lower, in CDDP-resistant OC tissues than in CDDP-sensitive OC tissues (Fig. 3M).
circPLPP4 enhances PIK3R1 expression by sponging miR-136 in OC cells
In cancer cells, miRNAs mainly exert their function through regulating target genes. RNA-seq was performed on A2780 CDDP cells and A2780 CDDP circPLPP4 -ASO#1 cells to identify the targets gene(s) of miR-136 (Fig. 4A). Next, several algorithms (miRanda, RNAhybrid, miRWalk, and TargetScan) were used to predict the target genes of miR-136 (Fig. 4B). We considered these oncogenes or tumor suppressor genes as miR-136 targets if they met the following three criteria: (1) upregulated in A2780 CDDP cells compared with A2780 cells (Fig. 4D, Supplemental Fig. 4A-C); (2) elevated in CDDP-resistant OC tissues compared with CDDP sensitive tissues (Fig. 4C); (3) predicted as the potential miR-136 target genes by the four algorithms (Fig. 4B). Furthermore, we conducted luciferase reporter assays to determine whether miR-136 directly targets these selected genes in A2780 and SKOV3 cells (Fig. 4E, Figure S5 A). In A2780 CDDP and SKOV3 CDDP cells co transfected with miR-136 mimics, reporter constructs including wild-type miR-136 binding sites from the PIK3R1 3′untranslated region (UTR) showed obviously reduced luciferase activity compared with that of constructs with mutated binding sites (Fig. 4E). Thus, we identified PIK3R1 as the target gene of miR-136. Next, we validated the results by further functional examinations. We found that miR-136 mimics remarkably reduced the RNA levels of PIK3R1 and that ectopic PIK3R1 expression repressed the effect caused by miR-136 overexpression (Fig. 4F-H, Figure S5 B). Additionally, co-transfection of circPLPP4-ASO#1 and anti-miR-136 inhibited the circPLPP4-ASO#1-induced decreased in PIK3R1 expression in A2780 CDDP and SKOV3 CDDP cells (Fig. 4I, Figure S5 C). Notably, co-transfection of circPLPP4 and miR-136 decreased the expression of PIK3R1 compared with transfection of circPLPP4 alone in A2780 and SKOV3 cells (Figure S5 D, E). Furthermore, we found that miR-136 mimics markedly decreased the levels of PIK3R, BRCA1, and canonical PI3K/AKT signaling molecules and increased the CDDP response-related molecules, such as cleaved caspase3 and γ-H2AX (Fig. 4J). Ectopic PIK3R1 expression reduced the effects of miR-136 upregulation (Fig. 4J). In addition, elevation of miR-136 expression repressed cell viability and induced apoptosis in A2780 CDDP and SKOV3 CDDP cells (Fig. 4K, L, Figure S5F, G). However, co-transfection of PIK3R1 and miR-136 abrogated these effects (Fig. 4K, L, Figure S5F, G). Moreover, transfection of circPLPP4-ASO#1 significantly decreased PIK3R1, BRCA1 and p-AKT levels and upregulated the levels of cleaved caspase3 and γ-H2AX. Downregulating of both circPLPP4 and miR-136 abrogated these effects in A2780 CDDP and SKOV3 CDDP cells (Fig. 4M). Furthermore, correlation analysis was performed between circPLPP4 and miR-136 expression levels and PIK3R1 protein levels in 25 OC tissue samples (Fig. 4N, top), Pearson R was used to analyze the correlation between the indicated group (Fig. 4N, bottom).
These results indicated that circPLPP4 functions as a competing endogenous RNA (ceRNA) via targeting miR-136 to regulate PIK3R1 expression and thus promote CDDP resistance in OC.
circPLPP4 is modulated by m6A methylation
We showed that circPLPP4 is likely formed from pre-mRNA back-splicing of exons 2–4 of the PLPP4 transcript. However, the mechanism controlling circPLPP4 generation is unclear. Recently studies suggested that epigenetic mechanisms are frequently involved in the dysregulation of noncoding RNAs; therefore, we wondered whether epigenetic regulation is responsible for circPLPP4 upregulation in OC cisplatin resistance. Firstly, treatment of OC cells with a DNA methyltransferase inhibitor had no effect on circPLPP4 expression (Figure S6 A), suggesting that DNA methylation does not have a major role in circPLPP4 regulation. Next, we investigated whether histone acetylation participates in circPLPP4 regulation. Treatment of OC cell with broad-spectrum HDAC inhibitors (SAHA and NaB) did not affect circPLPP4 expression (Figure S6 B), indicating that histone acetylation modification does not participate in circPLPP4 regulation.
Studies suggest that m6A is the most prevalent modification of mRNAs and noncoding RNAs; therefore, it might have an important role in circRNA biogenesis[25, 26]. The m6A modification occurs mainly on the consensus motif ‘RRm6ACH’ (R = G or A; H = A, C or U)[27]. SRAMP algorithm[28] predicted an m6A modification site close to the junction region of circPLPP4 (Fig. 5A). m6A-specific immunoprecipitation assays demonstrated an increased m6A level on circPLPP4 in A2780 CDDP and SKOV3 CDDP cells compared with that in their parental cells (Fig. 5B-D), suggesting that m6A modification might be involved in circPLPP4 upregulation. To further identify the molecule that induces m6A modification of circPLPP4, we examined m6A-related gene expression in OC in our SYSUCC cohort (20 platinum resistant OC tissues and 20 platinum sensitive OC tissues; 20 OC tissues and 20 normal ovary tissues) using qRT-PCR. The results showed that several m6Arelated genes were dysregulated in OC (Fig. 5E, Figure S6 C). We validated that METTL3 was markedly upregulated in OC tissues; whereas, we found no significant difference among other m6A-related genes (Figure S6 C). Notably, an RNA pulldown assay revealed that circPLPP4 interacts with the key m6A methyltransferase, METTL3, and a classical m6A reader, IGF2BP1 (Fig. 5F, G). These results encouraged us to explore the role of METTL3 in regulating m6A modification of circPLPP4. We treated A2780 CDDP and SKOV3 CDDP cells with an siRNA targeting METTL3, which showed that METTL3 knockdown markedly decreased the m6A level and the level of circPLPP4 (Fig. 5H-J). Interestingly, METTL3 and circPLPP4 expression were positively associated in our SYSUCC cohort (Fig. 5K), indicating that positive regulation by METTL3 on circPLPP4. Additionally, the circPLPP4 level was reduced when ALKBH5 (an N6-demethylase) was overexpressed (Figure S6 D, E). These results suggested that the m6A modification upregulates the circPLPP4 level. Next, explored the detailed mechanism of m6Amediated upregulation of circPLPP4 in OC cells. Silencing of METTL3 affect the stability and decreased the half-life of circPLPP4 significantly (Fig. 5L), suggesting that METTL3 regulates the circPLPP4 level by modulating its stability. The m6A-mediated regulation of circPLPP4 needs an m6A reader to recognize the m6A modification, we investigated which m6A reader regulates circPLPP4. An RIP assay showed that circPLPP4 was markedly immunoprecipitated by IGF2BP1 rather than other m6A readers (Fig. 5M, Figure S6 F). Moreover, IGF2BP1 inhibition in OC cells decreased the level and half-life of circPLPP4 significantly (Fig. 5N, Figure S6 G). A nucleus-cytoplasm fractionation analysis suggested that METTL3 silencing did not affect the localization of circPLPP4 in OC cells (Figure S6 H). Additionally, we found that the level and half-life of circPLPP4 strongly decreased upon METTL3 and IGF2BP1 knockdown (Figure S6 I, J). Furthermore, when we mutated the putative m6A site in circPLPP4 (Fig. 5O), the direct binding between circPLPP4 and IGF2BP1 was impaired, as assessed using an RIP assay (Fig. 5P). Moreover, METTL3- or IGF2BP1-induced regulation of circPLPP4 was abolished when the m6A sites were mutated (Fig. 5Q). In addition, METTL3 or IGF2BP1-mediated m6A modification of circPLPP4 was repressed upon mutation of both m6A sites (Fig. 5R).
Targeting circPLPP4 in vivo retards CDDP-resistant OC
To further verify the in vitro results and to discover potential clinical therapy targets, we constructed intraperitoneal implantation models to explore the role of circPLPP4 in CDDP resistance in vivo. Intraperitoneal implantation of A2780 CDDP cells transfected with ASO#1 resulted in tumors that were significantly more sensitive to cisplatin than those formed from the control cells, while the two groups showed a similar response to PBS treatment (Fig. 6A-C). IHC analysis of tumor xenograft samples also showed that the levels of γH2AX and cleaved caspase-3 were obviously upregulated, whereas PIK3R1 and BRCA1 levels were decreased upon circPLPP4 inhibition (Fig. 6D). Moreover, tail vein injection of the in vivo-optimized circPLPP4 inhibitor (circPLPP4-ASO#1) in the CDDP-resistant A2780-CDX models obviously enhanced the sensitivity of OC cells to CDDP treatment (Fig. 6E-H).
circPLPP4 is a therapeutic target in OC pre-clinical models
Next, we assessed the therapy efficacy of the in vivo-optimized circPLPP4 inhibitor (circPLPP4-ASO#1) in two OC Patient-derivedxenografts (PDX) models (Fig. 7A). Based on the ISH scores of patient tumor sections, one PDX (named as PDX-1) was defined as a circPLPP4Low patient, and the other (PDX-2) was regarded as a circPLPP4High patient (Fig. 7B). When the circPLPP4-ASO#1 was used in the PDX-2 model, tumor growth was inhibited significantly compared with that in the controls (Fig. 7C-E). As expected, circPLPP4-ASO#1 failed to repress tumor growth in the PDX-1 model (Fig. 7F-H). In addition, mouse body weights showed no significant change the PDX-1 and PDX-2 models, with or without circPLPP4-ASO#1 treatment (Fig. 7I, J). An IHC assay indicated that PIK3R1 and γ-H2AX levels were decreased, accompanied by tumor growth repression, after circPLPP4-ASO#1 application in the PDX-2 model (Fig. 7K, L). Moreover, the apoptotic response (TUNEL positive cells) increased markedly in the PDX-2 models treated with circPLPP4-ASO#1 (Fig. 7M). These results suggested that circPLPP4 is a promising therapeutic target.
Clinical relevance of the m6A/circPLPP4/PIK3R1 axis in OC
To further explore the clinical relevance of the m6A/circPLPP4/PIK3R1 axis in OC, we detected the expression of circPLPP4, METTL3, IGF2BP1, and PIK3R1 in OC tissues from the SYSUCC cohort via ISH and IHC assays. The results revealed that circPLPP4 was positively associated with METTL3, IGF2BP1, and PIK3R1 expression (Fig. 7N, O). Moreover, correlation analysis indicated a positive association between circPLPP4 expression and METTL3, IGF2BP1, or PIK3R1 expression, as detected using qRT-PCR (Figure S7 A-C). Notably, western blotting and qRT-PCR analysis confirmed the positive association between circPLPP4 expression and METTL3, IGF2BP1, or PIK3R1 in five OC tissues (Fig. 7P). In summary, METTL3-mediated m6A modification increases circPLPP4 level in an IGF2BP1-recognized manner, which activates the circPLPP4/miR-136/PIK3R1 axis, which subsequently contributes to CDDP resistance in OC (Fig. 7Q).