CircFAM114A2 Inhibits Bladder Cancer Proliferation and Promotes Cisplatin Sensitivity via the miR-222-3p/P27 and miR-146a-5p/P21 Cascades

Background: Circular RNAs (circRNAs) are noncoding RNAs that have the structure of a covalently closed loop. Increasing data has proved that circRNA can inuence the development and progression of tumors. CircFAM114A2 is generated from several exons of FAM114A2. However, the function and mechanisms of circFAM114A2 in bladder cancer (BCa) remain unclear. This research aimed to reveal that circFAM114A2 inhibits bladder cancer progression and improves sensitivity of cisplatin chemotherapy by inducing G1/S cell cycle arrest via novel miR-222-3p/P27 and miR-146a-5p/P21 cascades. Methods: Here, to elucidate the potential roles of circFAM114A2 in BCa, we conducted RNA-sequencing on 5 pairs of BCa samples and screened for circRNAs. CircRNAs, microRNAs (miRNAs) and mRNAs, as well as levels of P27 and P21, in human cells and tissues were detected by qRT-PCR and western blot, respectively. CircRNA-miRNA interactions and miRNA-downstream mRNAs interactions were investigated by RNA pull-down assay and uorescence in situ hybridization (FISH) or luciferase reporter assays, respectively. Then, the function of circFAM114A2 in BCa was explored using cell proliferation, cell cycle and tumorigenesis assays in nude mice. Finally, the function of circFAM114A2 in cisplatin chemosensitivity in BCa was detected by IC50 and tumor formation of xenograft in cisplatin-treated nude mice. Results: We discovered that circFAM114A2 levels were decreased in BCa cell lines and tissues. According to follow-up data, BCa patients with higher circFAM114A2 expression had better survival. Importantly, the levels of circFAM114A2 were associated with the histological grade of BCa. Overexpression of circFAM114A2 inhibited cell proliferation and increased sensitivity to cisplatin chemotherapy. Mechanistically, circFAM114A2 directly Results were assessed utilizing SPSS version 22.0 and depicted as mean±standard deviation. The P value was statistically signicant when less than 0.05. Two-tailed Student’s t-test and one-way ANOVA was conducted to assess variation among the groups. Correlation was assessed through Pearson’s correlation and Spearman’s rank correlation test. Survival curves were imaged utilizing the Kaplan-Meier method, and variation was assessed through log-rank test.

circRNAs from introns, intergenic circRNAs, exonic circRNAs, exon-intron circRNAs, and antisense circRNAs (7), which are generated from diverse mechanisms including exon skipping, intron pairing and RNA-binding proteins (8). Recently, studies have discovered that circRNAs are intimately related to the incidence and establishment of tumors (7,(9)(10)(11). CircRNAs can perform as competitive endogenous RNAs (ceRNAs) and protect the target mRNA from degradation by sponging microRNA (miRNA) (12). MiRNAs are non-coding RNAs that are 19-25 nucleotides in length and have the capability to in uence various ranges of biological functions through regulating the expression of mRNAs (13). MiRNAs play an important function in facilitating or repressing carcinogenesis, development and metastasis in BCa (14,15). CircRNAs can directly bind to miRNAs, which play a role as 'miRNA sponges' and cease the regulation of target mRNAs (11,16). Yang et al. reported that circITCH could inhibit progression of BCa through a circITCH/miR-17 or miR-224/P21 or PTEN pathway (17). CircSLC8A1 was shown to also inhibit progression of BCa via circSLC8A1/miR-130b or miR-494/PTEN pathway (18). Nevertheless, the biological properties of circRNAs in BCa remain mainly unidenti ed and need further research. To assess the role of circRNAs in the tumorigenesis and formation of BCa, our group conducted RNA sequencing on ve pairs of BCa tissues in 2017 (19) and found that circFAM114A2 was one of three most downregulated circRNAs. CircFAM114A2 has a much higher tissue average read count than the other two most down-regulated circRNAs. CircFAM114A2 is formed by several exons of the FAM114A2 gene. The underlying function and mechanisms of circFAM114A2 in BCa are still unclear.
Cisplatin has always been a major part of the chemotherapy regimen among patients with BCa. The longterm survival of patients is greatly hindered through chemotherapy resistance to cisplatin (20). Therefore, there is a need for more studies to reveal the mechanisms of cisplatin resistance. According to various studies, cell cycle regulation and cell apoptosis are the theoretical basis of cisplatin action and play important roles in cisplatin chemotherapy resistance (21,22). P27 and P21 are classic tumor suppressors as they play a crucial function in regulating cell cycle transition (23,24). Ripani et al. reported that P27 induces cell cycle arrest and increases the sensitivity of colon carcinoma cell to cisplatin (25). Wang et al. reported that P21 could promote cell cycle arrest and increase the cisplatin sensitivity of BCa (26).
In our study, we discovered that: 1) circFAM114A2 was decreased in BCa versus normal tissue and related to prognosis of BCa patients; 2) circFAM114A2 could inhibit BCa cells proliferation and sensitize BCa cells to cisplatin; 3) circFAM114A2 affected the expression of P27 and P21 by sponging miR-222-3p and miR-146a-5p. On the whole, circFAM114A2 can function as a possible biomarker and potential target for BCa diagnosis and therapy.

Results:
CircFAM114A2 was down-regulated in BCa and associated with prognosis of BCa patients RNA sequencing on ve pairs of BCa tissues was carried out and differentially expressed circRNAs were screened out by certain conditions (Additional Fig. 1A). The down-regulated circRNAs were sorted by fold change of more than 2 and P value of less than 0.05 and then screened them with an average read count of normal tissue more than 150. We found that circFAM114A2 was one of three most down-regulated circRNAs (Additional Fig. 1B). Among them, circFAM114A2 had a much higher tissue average read count than other two circRNAs. Then, we validated head-to-tail splicing through RT-PCR of circFAM114A2 using an estimated size established by Sanger sequencing (Fig. 1A). To study levels of circFAM114A2 in BCa tissue, levels of circFAM114A2 in 46 paired BCa and matched adjacent healthy bladder tissue were detected. Next, results showed that circFAM114A2 was signi cantly reduced in BCa tissues (Fig. 1B) and 8 BCa lines (BIU87, TCC, RT4, 5637, T24, J82, UMUC, and 253J) in comparison to SV-HUC, a healthy urothelial cell line (Fig. 1C). Additionally, we also found that FAM114A2 mRNA was down-regulated in BCa tissue (Additional Fig. 2A) and cell lines (Additional Fig. 2B). Then, the convergent primers to increase FAM114A2 mRNA and divergent primers to amplify circFAM114A2 were designed. CircFAM114A2 could only be ampli ed by divergent primers in cDNA rather than gDNA (Fig. 1D). Through utilizing qRT-PCR, we additionally showed that FAM114A2 mRNA was drastically decreased post-RNase R addition, though circFAM114A2 was impervious to RNase R (Fig. 1E). Patients with different circFAM114A2 expression value were divided into low-and high-circFAM114A2 groups. The mean value was 0.000125103. And we found levels of circFAM114A2 were negatively associated with histological grade (P = 0.02), but not with the use of additional clinicopathological parameters, such as age, gender, tumor node metastasis stage (TNM) and tumor size (Table 1). We predicted recurrence and progression in NMIBC patients and strati cate the risk group by EORTC method (29). We found that recurrence scores were lower in high circFAM114A2 expression group (Table 1). In addition, BCa patients with higher circFAM114A2 levels had better overall survival (Fig. 1F). Taken together, these data indicate that circFAM114A2 was down-regulated in BCa and related to prognosis.  To examine the role of circFAM114A2 in BCa, we separately chose a high malignant BCa cell line T24 and low malignant BCa cell line 5637, and conducted further functional assays. We transfected the circFAM114A2 overexpression vector and knock-down lentivirus into T24 and 5637 cells. We discovered that levels of circFAM114A2 were substantially changed while levels of FAM114A2 mRNA had no signi cant change ( Fig. 2A,E). Furthermore, CCK-8 assays show that viability of T24 and 5637 cells was increased in the circFAM114A2 knockdown group in comparison to the controls (Fig. 2B), while viability was signi cantly reduced in the circFAM114A2-overexpression group in comparison to the vectors (Fig. 2F). The colony numbers of circFAM114A2 knockdown group were signi cantly increased (Fig. 2C), while the colony numbers of circFAM114A2 overexpression group was substantially decreased (Fig. 2G).
In order to reveal the role of circFAM114A2 in cisplatin chemosensitivity, we designed 10 gradient cisplatin diluted concentration (0, 2, 4, 8, 16, 32, 40, 50, 64, 128 µmol/L). The cell viability curve of the cisplatin treatment assays for calculating the IC50 presented in the Additional Fig. 3. CCK-8 assay indicated that circFAM114A2 overexpression may drastically constrain growth of T24 and 5637s across different diluted concentrations of cisplatin compared to the vector group. In the circFAM114A2 knockdown group, proliferation of T24 and 5637 cells was signi cantly increased across the different diluted concentrations of cisplatin in comparison to the relative control group. The IC50 value of circFAM114A2 knockdown group was substantially higher than the controls (Fig. 2D). Furthermore, the IC50 of cisplatin to T24 and 5637 cells in the circFAM114A2 overexpression cohort was meaningfully reduced compared to the vector group (Fig. 2H). Additionally, apoptosis experiments indicate that circFAM114A2 induced cell apoptosis in BCa cells (Additional Fig. 4). Overexpression circFAM114A2 induced higher apoptosis rate compared to vector group with cisplatin treatment (Additional Fig. 4). All the data validated that circFAM114A2 could inhibit the progression and enhance the cisplatin chemosensitivity of BCa cells in vitro.

MIR-222-3p/146a-5p interference rescued the progression inhibition and cisplatin chemo-sensitization induced by circFAM114A2 in BCa cells
We found that miR-222-3p and miR-146a-5p had high expression in 46 BCa tissues in comparison to surrounding healthy tissues (Fig. 4A). A negative relationship among levels of circFAM114A2 and miR-222-3p/miR-146a-5p was indicated utilizing Pearson correlation analysis (Fig. 4B). Due to the interaction of circFAM114A2 and miR-222-3p/miR-146a-5p, the potential function of miR-222-3p and miR-146a-5p in BCa cells were studied by transfecting miRNA and NC mimics. Overexpression of miR-222-3p or miR-146a-5p substantially encouraged the proliferation of T24 and 5637 cells (Fig. 4C). In addition, the IC50 values of cells transfected with miR-222-3p or miR-146a-5p mimics were substantially increased in comparison to NC mimic group across both T24 and 5637 cells. (Fig. 4D). The rescue experiment was performed by co-transfecting circFAM114A2 and miR-222-3p/miR-146a-5p mimics into BCa cells. CCK-8 and colony formation assay indicated that high expression of circFAM114A2 was associated with an inhibition in growth. However, this in uence may, in part, be inhibited by ectopic miR-222-3p or miR-146a-5p expression (Fig. 4E,G). More than that, results from the CCK-8 assay suggested that progression of BCa cell growth was appreciably hindered when treated with different diluted concentrations of cisplatin by overexpression of circFAM114A2, while the ectopic expression of miR-222-3p or miR-146a-5p could partly attenuate this phenomenon (Fig. 4F). All these data con rm that miR-222-3p/146a-5p could rescue the inhibition of progression and cisplatin chemosensitization induced by circFAM114A2 in BCa cells.
CircFAM114A2 in uences the expression of P27 and P21 through sponging miR-222-3p and miR-146a-5p In three pairs of circFAM114A2 overexpression and negative control cells, mRNA sequencing and enrichment analysis (GSEA analysis) was performed (Additional Fig. 5). We found that the cell-cycle transition genes, P27 and P21, were signi cantly related to circFAM114A2 expression. Hence, we selected P27 and P21 from the results for additional validation. According to miRanda and TargetScan prediction, we determined that miR-222-3p and miR-146a-5p target the 3′-UTR of P27 and P21, respectively (Fig. 6A). Dual luciferase reporter assay was also carried out to con rm miR-222-3p/miR-146a-5p binding with P27/P21. The wild-type and mutant 3′-UTR of P27 and P21 was added to the construct reporter plasmids, respectively. Co-transfection of miR-222-3p mimic and wild-type P27 reporter plasmid reduced luciferase activity dramatically. Co-transfection of miR-146a-5p and wild type P21 reporter plasmid also reduced luciferase activity. (Fig. 6B). Quite the reverse, co-transfection of miR-222-3p or miR-146a-5p mimics and mutated P27 or P21 vectors indicated no signi cant in uence on luciferase activity (Fig. 6B). The qRT-PCR and western blot results revealed that miR-222-3p/miR-146a-5p could reduce the expression of P27/P21 respectively (Fig. 6C,D). We also preformed qRT-PCR and western blot experiments to validate if miR-222-3p could in uence the expression of P21 and miR-146a-5p could in uence the expression of P27 (Additional Fig. 6). The results showed that these two pathways were two separate pathways. The qRT-PCR and western blot results indicated that knockdown of circFAM114A2 could decrease P27 and P21 mRNA and protein (Fig. 6E,F). Consequently, high levels of circFAM114A2 may heighten mRNA and protein expression of P27 and P21 (Fig. 6G-J). On the other hand, co-transfection of miR-222-3p or miR-146a-5p mimics and circFAM114A2 partly inhibited the increased expression of P27 and P21 induced by circFAM114A2 (Fig. 6G-J). All these ndings con rmed that P27 and P21 were directly regulated by miR-222-3p and miR-146a-5p respectively. Overall, results indicated that circFAM114A2 inhibited BCa advancement and cisplatin chemo-resistance by eliminating miR-222-3p and miR-146a-5p oncogenic effect. In this study, we revealed two axes of circFAM114A2-miR-222-3p-P27 and circFAM114A2-miR-146a-5p-P21 which played an important role in inhibition of BCa progression and cisplatin chemoresistance.

CircFAM114A2 inhibited growth of xenografted tumor in vivo
To examine if high levels of circFAM114A2 affect tumor growth in vivo, T24 cells transfected using circFAM114A2 or vector were inoculated in nude mice subcutaneously. After four weeks, we observed the tumor in nude mice and removed it (Fig. 7A-B). CircFAM114A2 overexpression group had a signi cantly lower average tumor weight (Fig. 7C) and lower mean tumor volume (Fig. 7D) than those of vector group. IHC assessment showed that P27 and P21 levels were promoted by overexpression circFAM114A2 (Fig. 7E). QRT-PCR results showed miR-222-3p and miR-146a-5p expression levels were reduced in 4 pairs of xenograft tumors transfected with circFAM114A2 compared with those of NC group (Fig. 7F-G). Western blot results indicated that expression of P27 and P21 were increased in circFAM114A2 overexpression group (Fig. 7H).

CircFAM114A2 promoted cisplatin sensitivity in vivo
To investigate whether circFAM114A2 could in uence the sensitivity of cisplatin chemotherapy in vivo, T24 cells transfected with si circFAM114A2 or si NC were inoculated in nude mice subcutaneously. One week after inoculation, nude mice in the experimental groups were intraperitoneally injected with cisplatin (5 mg/kg) every three days, while the control groups were injected with the same volume of saline. After four weeks, we observed the tumor in nude mice and removed it (Fig. 8A-B). The volume and weight of tumors in cisplatin treated group were lower than those in saline control group (Fig. 8C-D). The tumor weight and volume of si circFAM114A2 group treated with cisplatin were higher compared si NC group treated with cisplatin ( Fig. 8C-D). All the results indicated that circFAM114A2 could promote cisplatin sensitivity in vivo.

Discussion:
During recent years, circRNAs have been increasingly shown to act as vital factors in the incidence and formation of BCa (17,18,20). We detected circFAM114A2 through circRNA sequencing technology and con rmed that circFAM114A2 was decreased in BCa tissues and cell lines consistent with a study by Liu et al. (30). BCa patients with lower circFAM114A2 levels had shortened survival. Subsequently, we found that circFAM114A2 inhibited progression of BCa and encouraged sensitivity of BCa cells to cisplatin in vitro and vivo. CircFAM114A2 may directly associate with miR-222-3p/miR-146a-5p, and then perform as a miRNA sponge to control levels of miR-222-3p/miR-146a-5p target gene P27/P21, which suppressed progression of BCa. Overall, this data showed that circFAM114A2 performed as a tumor-suppressor in BCa. It is very interesting that the screening of downstream miRNAs of circFAM114A2 in our paper was different from those reported by Liu et al. The differences may be caused by the different selection of prediction databases and the different screening conditions in each database. Except MiRanda and RNAhybrid database, regRNA was also used in our prediction process. The screening conditions like combined score or minimum free energy may also different in these two papers. For the selection of downstream mRNA, mRNA sequencing was performed in circFAM114A2 overexpression and negative control cells. Then target genes P27 and P21 were identi ed through enrichment analysis. While Liu et al.
predicted and selected the downstream target mRNA by using publicly available databases.
CircRNAs may perform as miRNA sponges to carry out their function. Yang et al. showed that circITCH can inhibit the progression of BCa through circITCH/miR-17 or miR-224/P21 or PTEN pathway (17). While some circRNAs can in uence the modulation of transcription, some circRNAs can be translated (31). CircRNAs are playing an increasingly important role in the treatment of tumors. For example, circRNAs can regulate immune function and affect the occurrence and development of tumors, which provides more enlightenment for immunotherapy of different tumors (32). CircMET can lead to immunosuppression and resistance to anti-PD1 therapy in hepatocellular carcinoma (33). In addition, some studies have found that circRNA can enrich in exosomes and further affect the occurrence and development of tumors (26). Studies have shown that detection of circRNA in serum exosomes can distinguish cancer patients from healthy controls, and circRNA in exosomes can be used as a biomarker for liquid biopsy (34). In our study, we identi ed that circFAM114A2 was generally down-regulated in BCa tissues through circRNA sequencing technology. We further con rmed that circFAM114A2 was decreased in BCa tissues and cell lines by qRT-PCR. Afterward, we found that circFAM114A2 inhibited the proliferation of BCa and encouraged the sensitivity of BCa cells in vitro and in vivo through the CCK-8 assay, clone formation assay, IC50 determination and xenografts. Then, by RNA pull down assay and RNA FISH, we found that circFAM114A2 may sponge miR-222-3p and miR-146a-5p. Next, by luciferase reporter assay, we discovered that miR-222-3p and miR-146a-5p can be bound to P27 and P21, respectively. The rescue experiments con rmed that circFAM114A2 inhibited BCa proliferation and increased cisplatin sensitivity by circFAM114A2-miR-222-3p-P27 and circFAM114A2-miR-146a-5p-P21 axes. We also explored the relationship between these two pathways. And the results of qRT-PCR and western blot showed that these two pathways are more inclined to two independent pathways. However, P27 and P21 play a synergistic role in promoting cell cycle arrest (35). And they are consistent in inhibiting tumor proliferation and promoting cisplatin chemotherapy sensitivity in bladder cancer (36). So we think that these two pathways are independent in mechanism but synergistic in inhibiting tumor development. However, circRNAs can also promote the progression of cancer. For instance, circTP63 could promote progression of lung squamous cell carcinoma via up-regulating FOXM1 expression (37). Advanced RNA sequencing (RNA-seq) and bioinformatics technologies have helped us to identify various circRNAs and reveal their important functions. With more attention and research, the biological functions of circRNAs have been brought to light.
Cisplatin has always had a vital function in the chemotherapy of BCa. The long-term survival of patients is greatly hindered due to cisplatin resistance (38). Therefore, there is a need for more attention to reveal mechanisms of cisplatin resistance. Previously, Yuan et al. reported that circCDR1as could improve the sensitivity of BCa to cisplatin through increasing APAF1 levels by inhibiting miR-1270 (20). Our research also revealed a chemo-sensitization function of circFAM114A2 in BCa treatment. CCK-8 assay indicated that circFAM114A2 could drastically block the growth of BCa cells across different diluted concentrations of cisplatin. In the circFAM114A2 knockdown group, the IC50 was higher than the relative control group.
On the other hand, in the circFAM114A2 overexpression group, the IC50 was lower when compared with relative control group. During the BCa cell lines veri cation, we found that cell lines (TCC, T24 and 253J) which had very low expression of circFAM114A2 are all relative high malignant degree BCa cell lines and had mutations in different oncogenes like P53, H-ras or TREK2. So these cell lines may intrinsic resistant to chemotherapy because of high malignant degree or relative oncogene mutation. There are indeed some mutations in oncogenes like P53, H-ras or TREK2. But we did not nd any association between these mutation oncogenes and expression of circFAM114A2. It is need to do more research on this area in the future. As reported, cell cycle regulation and cell apoptosis are the theoretical basis of cisplatin action and play important roles in cisplatin chemotherapy resistance (21,22). It has been reported that tumors are most sensitive to cisplatin in G1 phase (39,40). Therefore, blocking tumor cells in the G1 phase can enhance their sensitivity to cisplatin. In our study, we found that high expression of circFAM114A2 could block more BCa cells in the G1 phase. Furthermore, we discovered that high expression of circFAM114A2 may heighten the apoptosis rate of BCa cells. Many studies have shown that circRNAs can in uence the apoptosis level of gastric cancer, urothelial carcinoma and non-small cell lung cancer cells (41)(42)(43). Hence, our results suggest that improved sensitivity induced by circFAM114A2 may also relate to a high apoptosis rate.

Conclusions:
Based on our data, we con rmed that circFAM114A2 was decreased in BCa tissue and cell lines. To investigate the apoptotic rate of BCa cells treated with cisplatin, BCa cells at 70% con uence were treated with 4 µmol/ml cisplatin for 24h and then collected for follow-up apoptosis testing.
Biotin-coupled miRNA capture About 2×10 6 BCa cells at 50% con uence were transfected using 50 μM biotinylated miRNA mimics or nonsense control (NC) (Ribobio Guangzhou, China). After 24h, cells were collected and washed with phosphate-buffered saline (PBS) twice. Streptavidin magnetic beads (Thermo Fisher, USA) were placed in blocking buffer for 2h and put into every reaction to extract the biotin-coupled RNA compound. Tubes were placed in incubation while rotating at low speed (10rpm /min) for 4h. Beads were washed by lysis buffer ve times. Then, Trizol reagent was utilized to retrieve the RNAs that interacted with miRNAs.
Quantity of circFAM114A2 in bound portions was assessed by qRT-PCR and agarose gel electrophoresis.
Biotin-coupled probe pull down assay Biotinylated probe was targeted to bind the junctional region of circFAM114A2. Oligo probe served as control. Approximately 1×10 7 cells were washed with pre-cooled PBS, and then lysed using lysis buffer.
The lysed products were placed in incubation using 3 μg biotinylated probes at room temperature for 2h. Then, cell lysates were incubated with 50 μl streptavidin magnetic beads for an additional 4h. The beads were cleaned using lysis buffer ve times and miRNAs in the pull-down material were isolated utilizing Trizol reagent and assessed with qRT-PCR assay.

Agarose gel electrophoresis
The agarose gel was made of agarose powder, 1×TAE solution and nucleic acid dye. About 2 µl DNA ladder was added to 8 µL of the PCR product, and loaded to agarose gels. The constant-pressure 200 V was set up for 20 min electrophoresis. When electrophoresis ends, the gel was viewed by ultraviolet lamp to identify the region of DNA.

Northern blot
Northern blot was conducted using northern blot kit. In short, approximately 30 μg of total RNA was denatured by formaldehyde and ran under electrophoresis in a 1% agarose-formaldehyde gel. Electrophoresed RNAs were moved onto a Hybond-N+nylon membrane. The biotinlabeled DNA probes were used for hybridization with relative RNAs. Biotin chromogenic detection was utilized to identify the bound RNAs. Lastly, membranes were assessed by Image Lab software (Bio Rad, USA).

Fluorescence in situ hybridization (FISH)
Cy3-labeled probes speci c to circFAM114A2 and fam-labeled probes speci c to miR-222-3p/miR-146a-5p were constructed and produced through Genepharma (Shanghai, China) and probe signals were identi ed through a Fluorescent In Situ Hybridization Kit (Genepharma, Shanghai, China) as per established protocol. Pictures were attained on a Zeiss LSM880 NLO confocal microscope system (Leica Microsystems, Germany).
Luciferase reporter assay HEK293 cells were co-transfected with mutant or wild 3′-UTR fragments of P27 or P21 and miRNA mimics utilizing Lipofectamine3000 (Invitrogen, CA) as per established instructions. After 24h posttransfection, re y and renilla luciferase activities were quanti ed consistently by utilizing dual luciferase reporter assay system (Promega, USA). Lastly, the proportion of luminescence through luciferase was measured and every experiment was conducted three times.

Immunohistochemistry (IHC)
Para n-embedded tumors from mice were sliced to 4 μm slides. Tissue slides were rehydrated with different grade ethanol, and then placed in sodium citrate buffer (pH=6). Antigen was isolated utilizing heating through microwave. Slides were dipped in 3% H2O2 for 10 min and then treated with P27 or P21 antibody at 4°C overnight. After washed, the slides were incubated with HRP-conjugated antibody and standard avidin biotinylated peroxidase complex method. The images were viewed with a microscope. The degree of positivity was assessed by at least two pathologists based on the proportion of positive tumor cells.

Xenografts in mice
Approximately 1×107 cells underwent a stable transfection using si circFAM114A2, si NC, circFAM114A2 and vector were inoculated into axilla of the BALB/C nude mice (4-6 weeks old, 18-22 g, 4 mice/group) subcutaneously. We divided nude mice inoculated with si circFAM114A2 and si NC transfected cells into four groups and named them si circFAM114A2+cisplatin group, si NC+cisplatin group, si circFAM114A2+saline group and si NC+saline group. One week after inoculation, nude mice in the experimental groups were intraperitoneally injected with cisplatin (5mg/kg) every three days, while the control groups were injected with the same volume of saline. Tumor evolution was followed each week, and the width (W) and length (L) were quanti ed utilizing calipers. The volume (V) of tumor was measured utilizing the formula V= (W 2 ×L)/2. After four weeks, the mice were sacri ced and tumor bulk was quanti ed. Animal experiments were conducted according to ethics guidelines for animal studies and granted approval through the animal ethics board of Nanjing Medical University.