Circular RNA Cdr1as sensitizes bladder cancer to cisplatin by upregulating APAF1 expression through miR‐1270 inhibition

Circular RNAs (circRNAs) have recently emerged as essential regulators in carcinogenesis and cancer progression. Previous studies have shown that Cdr1as functions as a microRNA (miRNA) sponge in various cancer types. However, the role of Cdr1as in cisplatin chemosensitivity in bladder cancer remains unclear. Here, we used real‐time PCR to examine miRNA and gene expression in bladder cancer tissues and cell lines. The abilities of Cdr1as and its downstream regulatory molecules to induce apoptosis and promote cisplatin‐induced chemosensitivity of bladder cancer cells were determined by flow cytometry and cell counting kit. Bioinformatic analysis was utilized to predict potential miRNA target sites, and biotin‐coupled miRNA capture, biotin‐coupled probe pull‐down assay, and RNA fluorescent in situ hybridization were used to study the interaction between Cdr1as and target miRNAs. Dual‐luciferase reporter assay was also used to validate the target genes of miRNAs. The expression level of apoptotic protease‐activating factor 1 (APAF1) in bladder cancer cells was identified via western blot. Finally, the sensitivity of Cdr1as to cisplatin chemotherapy in nude mice xenografts was evaluated in terms of the size, volume of tumors, and the survival of mice. We report that Cdr1as induced the apoptosis and enhanced the cisplatin chemosensitivity of bladder cancer cells both in vitro and in vivo. Silencing of APAF1 reduced the sensitivity of bladder cancer cells to cisplatin chemotherapy. Furthermore, Cdr1as could directly sponge miR‐1270 and abolish its effect on APAF1. Our study verified that Cdr1as exerts a cisplatin‐chemosensitization effect on bladder cancer cells through the Cdr1as/miR‐1270/APAF1 axis. This newly identified axis may be a potential therapeutic target for bladder cancer patients.


Introduction
Bladder cancer is one of the most common tumors in the human genitourinary system. Approximately 70% of all bladder cancer cases are of the non-muscle-invasive type, of which approximately 10-20% cases progress to muscle-invasive bladder cancer (MIBC; Antoni et al., 2017). MIBC can easily infiltrate other tissues and metastasize; thus, this cancer type is highly lethal (Dobruch et al., 2016). Cisplatin-based neoadjuvant or adjuvant chemotherapy plays a key role in the comprehensive treatment of MIBC and the main treatment for metastatic bladder cancer (Alfred Witjes et al., 2017). Cisplatin chemotherapy can effectively reduce the recurrence rate of bladder cancer patients, inhibit tumor progression, and improve the prognosis of patients. The therapeutic regimen can increase the 5-year survival rate of patients with MIBC by up to 58-68% (Meeks et al., 2012). Nevertheless, the overall response rate of cisplatin chemotherapy is only approximately 35% (Sagalowsky, 2013). Cisplatin chemoresistance is the main factor restricting the longterm survival of bladder cancer patients.
Cisplatin damages cells through a variety of mechanisms, including damage to the cell DNA, intracellular calcium overload, damage to the mitochondria interfering with cell metabolism, activation of the apoptosis pathway, and mediation of the oxidative stress reaction (Chang, 2015). Apoptosis protease-activating factor 1 (APAF1), as a major factor regulating apoptosis, is a key molecule of the core apoptosis mechanism that forms a ring-like apoptosome that activates caspase-3, caspase-6, or caspase-7 to execute cell death (Bratton et al., 2001). The free radicals produced by cisplatin can attack the cardiolipin on the mitochondrial membrane, leading to the release of cytochrome C, which, when combined with APAF1, eventually activates the apoptosis pathway and cell death (Hu et al., 1999). Nevertheless, many cancers demonstrate chemoresistance toward cisplatin treatment. Numerous cisplatin-treated patients must tolerate various side effects without actually benefiting from chemotherapy due to chemoresistance. Interestingly, some studies have demonstrated that APAF1 participates in regulating cisplatin sensitivity (Del Bello et al., 2003;Kamarajan et al., 2001;Kuwahara et al., 2003;Zang et al., 2012). This observation suggests that APAF1 may be a target for regulating the sensitivity of bladder cancer to cisplatin. Circular RNAs (circRNA) is a covalently closed RNA molecule without a 5 0 terminal cap structure and 3 0 terminal polyadenylated tail (Jeck et al., 2013). cir-cRNA can function as a competitive endogenous RNA to regulate the biological activity of microRNA (miRNA) and completely or partially restore the inhibitory function of miRNA in the target gene (Memczak et al., 2013). circRNA plays a key role in many biological and pathological processes, including cell proliferation, migration and invasion, metastasis, cell cycle, and apoptosis (Du et al., 2016;Guarnerio et al., 2016;Zhao and Shen, 2017;Zheng et al., 2016). Cdr1as, also known as ciRS-7 or CDR1NAT, is formed by reverse splicing of the antisense strand of the cerebellar degeneration-associated antigen 1 (CDR1) gene, which contains 1500 nucleotides (Hansen et al., 2013a(Hansen et al., , 2013b. It was found that Cdr1as can sponge miR-7 and regulate downstream target gene expression (Memczak et al., 2013). Other studies have shown that Cdr1as plays oncogenic roles in many tumors, such as hepatocellular carcinoma (Yu et al., 2016) and colorectal cancer (Tang et al., 2017). In our previous study , however, we found that Cdr1as sponges miR-135a instead of miR-7, restores p21 activity, and mediates cell cycle arrest in bladder cancer cells. A recent study revealed that circPVT1 mediates docetaxel and cisplatin resistance in osteosarcoma by regulating the drug resistance-associated gene ABCB1 (Kunpeng et al., 2018). However, the role of Cdr1as in the development of cisplatin chemoresistance in bladder cancer still remains unknown. In the present study, we investigated the relationship between Cdr1as and the sensitivity of bladder cancer to cisplatin and explore the potential miRNA and downstream target gene APAF1 mechanistically. Our research provides novel insights into approaches to reverse cisplatin resistance in bladder cancer and improve the clinical efficacy of the drug.

Clinical specimens
Bladder cancer tissues and their paired normal tissues were obtained from patients who were diagnosed with bladder cancer and undergone surgery in the First Affiliated Hospital with Nanjing Medical University (Jiangsu Province Hospital) between 2010 and 2013. The follow-up deadline was January 2018. All patients signed informed consent before using clinical materials. The use of tissues for this study has been proved by the ethics committee of the First Affiliated Hospital with Nanjing Medical University (Jiangsu Province Hospital). The study methodologies confirmed to the standards set by the Declaration of Helsinki.

Cell culture
Dulbecco's modified Eagle's medium containing 10% fetal bovine serum (Gibco, Grand Island, NY, USA) and 1% penicillin-streptomycin were used to culture the human bladder cancer cell lines TCCSUP, 5367, T24, and EJ (Type Culture Collection of the Chinese Academy of Sciences; Shanghai, China) in an incubator with humidified 5% CO 2 at 37°C.

Transfection
The Cdr1as plasmid was donated by T. Hansen (Aarhus University, Denmark) and used to construct Cdr1as-overexpressing adenovirus (Hanbio, Shanghai, China). Cdr1as knockdown lentivirus was also constructed by Hanbio. All miRNA mimics/inhibitors, small-interfering (si) APAF1, and their controls were synthesized by GenePharma (Shanghai, China) for cell transfection. The transfection procedure was carried out according to the manufacturer's protocol, and the cells were tested after 48 h of transfection.

IC50 determination
The transfected cells were trypsinized and seeded into a 96-well plate at a density of 5000 cells per well. Three parallel wells were set up, and the plate was incubated overnight in an incubator. Then, the transfected cells were treated with a series of dilute concentrations of cisplatin (128,64,32,16,8,4, 2, and 1 µM, Sigma) for 24 h. Afterward, cell viability was measured by the Cell Counting Kit-8 (CCK-8) method (Dojindo, Kumamoto, Japan) according to the manufacturer's instructions. IC50 was calculated by the probit regression model (Tao et al., 2011). All experiments were independently performed thrice.

Cell apoptosis assays
The transfected cells were stained by an Annexin V-Alexa Fluor 647/Propidium Iodide Apoptosis Kit (FCmacs, Nanjing, China) and detected by flow cytometry. The ratio of early to late apoptotic cells was counted and compared between the experimental and control groups.

Protein extraction and western blot
The transfected cells were lysed with radioimmunoprecipitation assay buffer (Beyotime, Shanghai, China). Total protein was extracted from cell lysates and quantified by BCA (Beyotime). The equivalent amount of protein extract was loaded onto 10% sodium dodecyl sulfate/polyacrylamide gels, separated by electrophoresis, and transferred onto polyvinylidene fluoride membranes (Sigma-Aldrich, Burlington, MA, USA). The membranes were blocked in 5% nonfat milk in Tris-buffered saline and Tween-20 at room temperature for 2 h and then incubated with anti-APAF1 antibody (1 : 1000; Abcam, Cambridge, UK), anti-GAPDH (1 : 2000; Cell Signaling Technology, Danvers, MA, USA), or anti-bactin antibody (1 : 2000; Cell Signaling Technology) overnight at 4°C. Then, the membranes were incubated with a secondary antibody (1 : 5000; Cell Signaling Technology). After washing, the blots were developed with a chemiluminescence system (Bio-Rad, Hercules, CA, USA) and analyzed by Image Lab Software.

RNA binding protein immunoprecipitation assay
RNA binding protein immunoprecipitation (RIP) assay was conducted using the Magna RIP Kit (Millipore, Danvers, MA, USA) and Ago2 antibody (Cell Signaling Technology) in accordance with the manufacturer's instructions. The transfected cells were washed with ice-cold PBS and then mixed with an equivalent volume of RIP lysis buffer. Next, the lysis products were incubated with 5 µg of primary antibodies for 2 h at 4°C. Subsequently, each sample was mixed with 50 µL of prepared magnetic beads and incubated at 4°C overnight. The beads were briefly washed (five times in total) with RIP buffer and resuspended in 500 µL of TRIzol LS (Life Technology, Carlsbad, CA, USA). Finally, the RNA of the mixture was extracted and then detected by qRT-PCR.

Biotin-coupled miRNA capture
Approximately 3 9 10 6 cells were transfected with 50 µM of the biotinylated miRNA mimic or nonsense control (NC) (GenePharma, Shanghai, China) for 24 h in an incubator and then lysed in 500 µL of lysis buffer. Subsequently, 50 µL of washed streptavidin magnetic beads (Invitrogen) was blocked for 2 h, added to reaction tubes, and incubated in a rotator (10 rÁmin À1 , 4 h, 4°C) to pull down the biotin-coupled RNA complex. Subsequently, the beads were briefly washed with lysis buffer (five times in total) and resuspended in 500 µL of TRIzol LS (Life Technology) to recover RNAs specifically interacting with miRNAs. Finally, the binding circRNAs were detected by qRT-PCR and agarose gel electrophoresis.

Biotin-coupled probe pull-down assay
The specific biotinylated probes designed to bind to the junction area of Cdr1as were synthesized by Gongsi, China, whereas the oligo probe was taken as the control. The sequences were as follows: Cdr1as sense: 5 0 -GGTGCCATCGGAAACCCTGGATATTG-CAGACA-3 0 -Biotin and Oligo: 5 0 -TGTCTGCAA-TATCCAGGGTTTCCGATGGCACC-3 0 -Biotin. The transfected cells (approximately 1 9 10 7 ) were washed with ice-cold PBS, lysed in 500 µL of lysis buffer, and then incubated with 3 lg biotinylated probes at room temperature for 2 h. Subsequently, 50 µL of blocked streptavidin magnetic beads (Invitrogen) was added to the biotin-coupled RNA complex, which was incubated for another 4 h at 4°C. The beads were briefly washed with lysis buffer (five times in total), and the binding miRNAs in the pull-down products were extracted using TRIzol LS (Life Technology) and detected using qRT-PCR.

Agarose gel electrophoresis
Exactly 0.8 g of agarose powder was weighed in a conical bottle. Then, 40 mL of 0.59 TAE solution was weighed in a measuring cylinder and poured into the same conical bottle. The conical bottle was placed into an induction cooker to heat for 3-5 min. The agarose gel was completely boiled, and 1 µm nucleic acid dye was added to it. The liquid agarose gel was quickly poured into the mold, inserted with a 10-hole comb, and solidified for 30 min at room temperature. The gel was placed into the electrophoresis tank, and 0.59 TAE solution was poured over it until the gel was just immersed. About 5 µL of the PCR product was mixed with 1 µL of the DNA ladder, added to the agarose gel hole, and subjected to constant-pressure 90 V electrophoresis for 20-30 min. After electrophoresis, the gel was placed under an ultraviolet lamp to observe the position and grayscale of the DNA strip.

Luciferase reporter assay
T24 cells were cultured in 24-well plates and then cotransfected with plasmid containing psiCHECK-APAF1-wt or psiCHECK-APAF1-mut together with Firefly and Renilla luciferase and miR-1270 mimics or the control by using Lipofectamine 3000 (Invitrogen) in accordance with the manufacturer's protocol. After transfection for 24 h, Firefly and Renilla luciferase activities were measured using the dual-luciferase reporter assay system (Promega, Madison, WI, USA). The ratios of luminescence from Firefly to Renilla luciferase were normalized through three independent experiments.

Xenografts in mice
The animal studies were approved by the Animal Management Committee of Nanjing Medical University. T24 cells were transfected with the Cdr1as adenovirus and GFP vector. Female athymic BALB/c nude mice (4 weeks old, n = 20) were used. The nude mice were randomly divided into four groups, namely the Cdr1as + cisplatin group, the CTL + cisplatin group, the Cdr1as + saline group, and the CTL + saline group, and five rats were assigned to each group. Approximately 10 7 -transfected T24 cells and control cells were injected subcutaneously into the axilla of the mice. The width (W) and length (L) of the developed tumors were measured using calipers every 3 days, and tumor volume (V) was calculated using the formula V = (W 2 9 L)/ 2. A week after injection, cisplatin (10 mgÁkg À1 ) was intraperitoneally injected into mice in the experimental groups every 3 days, while mice in the control group were injected with the same amount of normal saline. At 3 week after injection, the mice were sacrificed, and tumor weights were measured. The tumors were fixed in 4% formalin for immunohistochemical analysis. In mouse survival assay, 4-week-old nude mice were randomly divided into four groups, namely as before, and each group had ten mice. All mice were injected with T24 cells (6 9 10 6 cells/mouse) through tail vein and two weeks later injected with saline (with or without cisplatin 10 mgÁkg À1 ) intraperitoneally every 3 days for 3 weeks. Mouse survival was monitored throughout the assay, during which in vivo tumor growth and metastases were partly detected with the IVIS 100 Imaging System (Xenogen, Los Angeles, CA, USA).

Tissue microarray and immunohistochemistry
Tissue microarray (TMA) was constructed using 160 cases of formalin-fixed, paraffin-embedded bladder cancer samples. Immunohistochemistry (IHC) was performed on TMA to evaluate APAF1 protein expression, also on paraffin-embedded tissues from nude mice. Tissue sections and TMAs were dewaxed and then rehydrated in graded ethanol. Tissue sections were placed in sodium citrate buffer (pH = 6), and the antigen was extracted by microwave heating. Tissue sections were soaked in 3% H 2 O 2 for 10 min and then incubated with APAF1 antibody (Abcam Company, 1 : 400) at 4°C overnight. After washing, the tissue sections were incubated with horseradish peroxidaseconjugated rabbit antibody for 30 min at room temperature. After washing once more, the tissue sections were stained with fresh 3,3 0 -diaminobenzidine, and images were obtained under a microscope (Olympus, Tokyo, Japan) with the appropriate magnification.

Bioinformatics analysis
We downloaded the bladder cancer patient data from the TCGA data collection (https://xenabrowser.net/da tapages/). The database is generated by the TCGA Research Network: http://cancergenome.nih.gov/. GO pathway analysis was obtained from http://geneontol ogy.org/. Gene expression and survival data were got from http://ualcan.path.uab.edu/. Median expression of APAF1 and miR-1270 in cancer tissues was used as a cutoff value in survival analysis. Original data were acquired from TCGA.

Statistical analysis
Data were analyzed using SPSS version 22.0 (IBM, Armonk, NY, USA) and are presented as mean AE standard deviation. Two-tailed Student's t-test and oneway ANOVA were performed to analyze differences between the groups. Correlations were analyzed by Pearson's correlation method and Spearman's rank correlation test. Survival curves were plotted using the Kaplan-Meier method, and differences were analyzed by the log-rank test. In all statistical tests, P < 0.05 was considered statistically significant.

Cdr1as enhances the cisplatin chemosensitivity of bladder cancer cells in vitro and in vivo
To investigate the effects of Cdr1as on bladder cancer cells, we infected T24 and EJ cells with the Cdr1asoverexpressing adenovirus (Cdr1as OE) or control GFP adenovirus (Cdr1as CTL). qRT-PCR assay indicated the relative abundance of Cdr1as in all available bladder cancer cell lines (Fig. S1A,B) and in cells infected with adenovirus ( Fig. S1C). CCK-8 assay showed that Cdr1as overexpression could dramatically inhibit the proliferation of T24 and EJ cells in a series of dilute concentrations of cisplatin (Fig. 1A,B). The IC50 of cisplatin to T24 and EJ cells in the Cdr1as OE group was significantly lower than that in the Cdr1as CTL group (Fig. 1C). To confirm the effect of Cdr1as on the chemosensitivity of bladder cancer cells to cisplatin, we also constructed three kinds of Cdr1as knockdown lentivirus (sh-Cdr1as①, ②, and ③, Fig. 1D) and GFP lentivirus (sh-CTL). qRT-PCR confirmed that Cdr1as expression level was significantly downregulated in T24 and EJ cells by sh-Cdr1as① instead of sh-Cdr1as② and sh-Cdr1as③ (Figs 1E and Fig S1D), and the opposite phenomenon was observed ( Fig. 1F-H). These results demonstrate that Cdr1as enhances the cisplatin chemosensitivity of bladder cancer cells.  The Cdr1as + cisplatin group lives longer than the Cdr1as + saline and much longer than the control + saline group. Data represent the mean AE SD from three independent experiments. Student's t-test with two biological independent replicates and log-rank test were used to determine statistical significance; **P < 0.01, ***P < 0.001. We also knocked down the expression of Cdr1as in TCCSUP and 5637 cells by the Cdr1as knockdown lentivirus. qRT-PCR confirmed that Cdr1as expression level was significantly downregulated in TCCSUP by sh-Cdr1as① and sh-Cdr1as②, which in 5637 cells by sh-Cdr1as① (Fig. S2A). CCK-8 assay showed that Cdr1as knockdown could promote proliferation of TCCSUP and 5367 cells in a series of dilute concentrations of cisplatin (Fig. S2B,C). The IC50 of cisplatin in the sh-Cdr1as group was higher than that in the sh-NC group (Fig. S2D). These results in TCCSUP and 5637 cells were consistent with the results in T24 and EJ cells.
T24 cells transfected with Cdr1as or GFP were injected subcutaneously into nude mice to investigate whether Cdr1as overexpression affects the cisplatin chemosensitivity of tumors in vivo. Tumor volumes were measured weekly after injection (Fig. 1I,J), and the volume and quality of xenografts in the cisplatin chemotherapy group were lower than those in the normal saline control group (Fig. 1K,L). Compared with those of the Cdr1as CTL group, the tumor volume and quality of the Cdr1as-overexpressing group were significantly reduced after cisplatin chemotherapy (Fig. 1K,L).
In mouse survival assay, mice were injected with T24 cells through tail vein. No signal was detected in a normal mouse by the IVIS 100 Imaging System, and its normal lungs are shown in Fig. S3A. Signals of metastasis in a mouse's body were detected by the IVIS 100 Imaging System, and metastasis in its lungs was observed after its death (Fig. S3B). The results showed that mice in the Cdr1as + saline group lived longer than those in the CTL + saline group, while mice in the Cdr1as + cisplatin group lived longer than those in the CTL + cisplatin group (Fig. 1M).

Cdr1as induces cell apoptosis in bladder cancer cells
Flow cytometry was performed to detect the effect of Cdr1as on cell apoptosis. Compared with that of the Cdr1as CTL group, the apoptotic rate (early apoptosis + late apoptosis) of T24 and EJ cells increased significantly after Cdr1as overexpression ( Fig. 2A,B); the opposite trend was observed in the sh-Cdr1as and sh-CTL groups (Fig. 2C,D). The same results were also observed in the knockdown Cdr1as of TCCSUP and 5637 cells (Fig. S4A,B). These results indicate that Cdr1as induces cell apoptosis in bladder cancer cells.

Cdr1as acts as a molecular sponge for miR-1270
Circular RNAs function mainly as miRNA sponges to bind functional miRNAs and then regulate gene expression (Hansen et al., 2013a. Our previous study preliminarily demonstrated that miR-1246, miR-7, and miR-135a could bind to Cdr1as. We further found that miR-1270 could also bind to Cdr1as and three binding sites are shown as illustrated (Fig. 3A). AGO2 RIP assay showed that AGO2 adsorbs more Cdr1as than IG (over 10-fold) in Cdr1as-transfected cells (Fig. 3B). The miR-1270 levels pulled down by Ago2 in Cdr1as OE cells were approximately fivefold greater than those in Cdr1as CTL-transfected cells (Fig. 3C). In Cdr1as OE cells, more miR-1270 was observed in the Ago2 pulldown eluents than in the IG pull-down products (Fig. 3D). We then applied the biotin-coupled probe pull-down assay to further confirm this interaction. As shown in Fig. 3E,F, specific enrichment of Cdr1as and miR-1270 was detected in the Cdr1as pulled down pellet compared with that in the Oligo group. This result indicates that Cdr1as could directly sponge miR-1270. We conducted biotin-coupled miRNA capture and FISH to confirm the sponge effect of Cdr1as. Biotin-coupled miR-1270 analogously captured more Cdr1as than biotin-coupled NC. This phenomenon indicates that miR-1270 could bind to Cdr1as (Fig. 3G). PCR experiments were conducted to find out the relative abundance of miR-1270 in tumors of mice formed by T24 cells infected with Cdr1as or control. And we found less miR-1270 in Cdr1as-overexpressed tumors (Fig. 3H). FISH technology showed that Cdr1as (red fluorescence) and miR-1270 (green fluorescence) could be visualized in the cells, and colocalization was observed in T24 cells (Fig. 3I). Finally, the Cdr1as expression in the RNA pull-down products was verified by agarose gel (Fig. 3J). Taken together, the results indicate that Cdr1as could serve as a sponge for miR-1270 in bladder cancer.

miR-1270 plays an oncogenic role in bladder cancer cells and reduces their cisplatin chemosensitivity
To investigate whether miR-1270 expression is altered in bladder cancer, we examined the miR-1270 expression in 32 pairs of bladder cancer tissues and matched adjacent normal tissues via qRT-PCR. The clinicopathological characteristics of the samples are presented in Table S1. As shown in Fig. 4A, the expression level of miR-1270 in bladder cancer tissues was significantly increased compared with that in adjacent nontumor tissues. Correlation analysis demonstrated that the expression of Cdr1as showed a moderately negative correlation with miR-1270 (R = À0.5965, P < 0.001, Fig. 4B). We also confirmed that miR-1270 showed higher expression in seven BCA cell lines (T24, EJ, J82, UMUC, 253J, RT4, and BIU-87) compared with that in SV-HUC-1, which is considered the normal urothelial cell line (Fig. 4C). Compared with that in the NC mimic group, the apoptosis rate of T24 and EJ cells decreased significantly after miR-1270 overexpression (Fig. 4D,F). The opposite trend was observed when the miR-1270 and NC inhibitor groups were compared (Fig. 4E,G). We explored the effects of four potential miRNAs (miR-1246, miR-7, miR-135a, and miR-1270) on the cisplatin chemosensitivity of bladder cancer cells and found that only miR-1270 induced significant changes in this chemosensitivity (data of other miR-NAs are shown in Fig. S5A-D). As shown in Fig. 4H,I, the cell growth inhibition rate of the miR-1270 overexpression group was significantly lower than that of the control group. The probit regression model showed that the IC50 of cisplatin in the miR-1270 overexpression group was higher than that in the control group (Fig. 4J). These results could be reversed after knocking down miR-1270 ( Fig. 4K-M). Taken together, the findings thus far indicate that miR-1270 plays an oncogenic role in bladder cancer cells and reduces their cisplatin chemosensitivity.

miR-1270 can partially reverse the effect of Cdr1as on cisplatin chemosensitivity and cell apoptosis in bladder cancer cells
We first constructed T24 and EJ cells cotransfected with the Cdr1as adenovirus and miR-1270 mimics to further explore the functional relationship between Cdr1as and miR-1270. The results showed that miR-1270 mimics could partially attenuate the Cdr1as overexpression-mediated inhibition of cisplatin chemosensitivity in bladder cancer cells (Fig. 5A-C). Moreover, cotransfection of Cdr1as and miR-1270 partially reversed the apoptosis induced by overexpressing Cdr1as (Fig. 5D,E).
We revalidated this result by building T24 and EJ cells cotransfected with the sh-Cdr1as lentivirus and miR-1270 inhibitor. After knocking down Cdr1as and miR-1270 simultaneously, the cisplatin chemosensitivity induced by sh-Cdr1as alone was partially reversed Cdr1as and miR-1270 were colocalized in T24 cell by FISH. Cdr1as was stained red, miR-1270 was stained green, and nuclei were stained blue (DAPI) (Scale bar, 2.5 lm). (I) Cdr1as in pull-down products from G was amplified by PCR and then detected by agarose gel electrophoresis. Data represent the mean AE SD from three independent experiments. Student's t-test with two biological independent replicates was used to determine statistical significance; **P < 0.01. in T24 and EJ cells (Fig. 5F-H). Similarly, the decrease in T24 and EJ cell apoptosis induced by knocking down Cdr1as was partially reversed by cotransfection of the sh-Cdr1as lentivirus and miR-1270 inhibitor (Fig. 5I,J). Thus, miR-1270 can partially reverse the effect of Cdr1as on the cisplatin chemosensitivity and apoptosis of bladder cancer cells.

APAF1 is the direct target gene of miR-1270, which could enhance the cisplatin chemosensitivity of bladder cancer cells and induce their apoptosis
We found 189 miR-1270-related potential target genes through 4 database [starBase, miRWalk, miRDB, and TargetScan (TS)] (Fig. 6A). All targets of miR-1270 are listed in the Tables S2-S3. Then, we did GO analysis and found apoptotic signaling pathway was one of the eight pathways. The apoptotic signaling pathway contained three genes INHBB, TPD52L1, and APAF1 (Fig. 6B). We chose APAF1 to continue our research because it was considered as the most correlated with tumors and chemotherapy. As shown in Fig. 6C,D, miR-1270 overexpression could decrease the mRNA level of APAF1 and decrease its protein expression levels. We performed dual-luciferase reporter assay and found that cotransfection of miR-1270 mimics and reporter plasmids strongly reduced luciferase activity. Conversely, cotransfection of miR-1270 mimics and mutated vectors showed no obvious effect on the luciferase activity (Fig. 6E,F). Consequently, the results prove that APAF1 is the direct target gene of miR-1270.
To investigate the effects of APAF1 on bladder cancer cells, we transfected si APAF1 or si-CTL into T24 and EJ cells and subsequently detected protein levels of APAF1 by western blot. The results showed depressed APAF1 expression (Fig. 6G). Compared with those in the si-CTL group, cells transfected with si-APAF1 revealed significantly increased viability and less sensitivity to cisplatin in a series of dilute concentrations ( Fig. 6H-J). In addition, APAF1 blockage also markedly inhibited cell apoptosis in T24 and EJ cells (Fig. 6K,L). The same results were found in cells overexpressing miR-1270, which suggests that miR-1270 regulates cisplatin chemosensitivity in bladder cancer by targeting APAF1. Furthermore, we investigated the expression of APAF1 in TMA. APAF1 was expressed in the nucleus and the cytoplasm (Fig. S6). The result showed that the expression of APAF1 was related to tumor histological grade by IHC analysis in TMA. The histological grade was higher in the high expression of APAF1 group (Table 1). Finally, Kaplan-Meier survival curves showed patients with low APAF1 expression had worse prognosis and shorter survival time, compared to those with high APAF1 expression (Fig. 6M), which was consistent with the results from tumor samples with detailed clinical information which was downloaded from TCGA database (APAF1 low n = 197, APAF1 high n = 198, P = 0.0101, Fig. S7A). But no significant difference of overall survival was observed according to miR-1270 level in bladder tumor samples with detailed clinical information which was downloaded from TCGA database (miR-1270 low n = 197, miR-1270 high n = 198, P = 0.3089, Fig. S7B).

Cdr1as plays a regulatory role through the Cdr1as/miR-1270/APAF1 axis
To explore whether Cdr1as exerts its cisplatin chemosensitivity effect by modulating APAF1 expression, we first constructed T24 and EJ cells cotransfected with the Cdr1as adenovirus and miR-1270 mimics. In the cotransfected cells, increased APAF1 mRNA and protein expression levels induced by overexpressing Cdr1as were partly reversed (Fig. 7A). We then cotransfected the sh-Cdr1as lentivirus and miR-1270 inhibitor in T24 and EJ cells and observed the opposite phenomenon (Fig. 7B). Moreover, APAF1 Overexpression of miR-1270 could decrease the sensitivity of T24 and EJ cells to cisplatin, while the weakening the expression of miR-1270 could increase the sensitivity of T24 and EJ cells to cisplatin. Data represent the mean AE SD from three independent experiments. Student's t-test with two biological dependent or independent replicates was used to determine statistical significance; **P < 0.01.  expression in the same 32 pairs of bladder cancer tissues and adjacent normal tissues was measured, and correlation analysis revealed a moderate negative correlation between the expression of miR-1270 and APAF1 (R = À0.6209, P < 0.001, Fig. 7C), as well as a moderate positive correlation between the expression of Cdr1as and APAF1 (R = 0.4098, P < 0.01; Fig. 7D). When it came back to xenografts, we found less miR-1270 in Cdr1as OE tumors (Fig. 7E). Immunohistochemical staining and western blot also revealed that APAF1 levels were elevated in Cdr1as OE tumors (Fig. 7F,7G). Overall, the data demonstrate that Cdr1as plays a regulatory role through the Cdr1as/miR-1270/APAF1 axis.

Discussion
In this study, we explored the effect of circRNA Cdr1as on the cisplatin chemosensitivity of bladder cancer and demonstrate the related regulatory mechanism, that is, the novel Cdr1as/miR-1270/APAF1 axis. Our results indicate that upregulated Cdr1as could increase the cisplatin sensitivity of bladder cancer T24 and EJ cells accompanied by increased cell apoptosis and vice versa. Cdr1as can function as a molecular sponge of miR-1270, which weakens the inhibitory effect of miRNA on the downstream target gene APAF1. Moreover, Cdr1as enhances the expression level of APAF1. Silencing APAF1 decreases cisplatin chemosensitivity and inactivates the apoptotic pathway. These findings demonstrate that Cdr1as dysregulation can contribute to the increased sensitivity of bladder cancer cells to cisplatin by cell apoptosis.
Cisplatin is one of the core drugs of many chemotherapeutic schemes, and resistance to this drug has drawn considerable research attention. Numerous bladder cancer patients are insensitive to cisplatin chemotherapy, which seriously restricts their long-term survival (Godwin et al., 2018). As a classic noncoding RNA, miRNA has been quite mature in chemoresistance. In our previous study, we found that miR-218 could target Glut1, activate the oxidative stress pathway in bladder cancer, and enhance the sensitivity of bladder cancer cells to cisplatin chemotherapy . In recent years, the growing popularity of high-throughput sequencing technology has enabled researchers to further investigate the expression and action mechanism of circRNA (Gao et al., 2017), identified 18 differentially expressed circRNA in breast cancer by screening microarrays of drug-resistant cell lines and sensitive cell lines, and then explored the potential chemoresistance mechanism of hsa_-circ_0006528 (Kunpeng et al., 2018). Cdr1as had been revealed to play important roles in esophageal cancer (Sang et al., 2018), hepatocellular carcinoma Yu et al., 2016), colorectal cancer (Tang et al., 2017), and non-small-cell lung cancer . In our previous study, we found that Cdr1as was significantly downregulated in bladder cancer tissues and that Cdr1as could inhibit bladder cancer cell proliferation, migration, and invasion by sponging miRNA-135a . In the present study, we found that Cdr1as overexpression in T24 and EJ cells could significantly reduce the survival rates of cells at different cisplatin concentrations. This finding suggests that Cdr1as enhances the sensitivity of bladder tumor cells to cisplatin and vice versa.
The molecular sponge mechanism of miRNA, as the classic regulatory pathway of circRNA, has been confirmed in many diseases, such as hepatocellular carcinoma (20), colorectal cancer (Tang et al., 2017), and esophageal cancer (Sang et al., 2018). In the present study, bioinformatics analysis, RNA pull-down, and FISH experiments confirmed that Cdr1as could adsorb miR-1270 in bladder cancer. Although miR-1270 in tumors is rarely reported (Knox et al., 2018;Zhong et al., 2018), our study found that miR-1270 expression is higher in tumor tissues than in adjacent normal tissues. We also discovered that upregulated miR-1270 inhibits cell apoptosis and significantly reduces the cisplatin chemosensitivity of T24 and EJ cells. Therefore, Cdr1as may indirectly enhance the expression level of downstream target genes by adsorbing miR-1270, thereby enhancing the sensitivity of cells to cisplatin chemotherapy. Cell apoptosis is an important mechanism in the anticancer effect of cisplatin. Mitochondrial apoptosis is the endogenous pathway of apoptosis and the center of apoptosis regulation (Petrosillo et al., 2003). During apoptosis, the BCL2 family is activated by apoptotic members (Park et al., 2002). Then, BAX transfers to the mitochondria, and the formation of ion channels promotes the release of cytochrome C from the mitochondrial matrix to the cytoplasm (Park et al., 2002). APAF1, which is the core molecule necessary to form ring-like apoptosomes, forms a complex with cytochrome C, thereby activating the caspase-9 precursor, which can activate caspase-3 (Park et al., 2002). Finally, the caspase cascade reaction is initiated to induce apoptosis (Park et al., 2002). Previous studies have shown that APAF1 overexpression in cisplatin-resistant Hela cell lines can partly reverse cisplatin chemoresistance, thus suggesting that APAF1 is involved in the regulation of cisplatin resistance (Kamarajan et al., 2001). In the present study, we detected apoptotic proteins in T24 and EJ cells overexpressing Cdr1as and found significantly increased APAF1 expression. Bioinformatics software and experimental results confirm that APAF1 is the direct target of miR-1270. The cell apoptosis rates and cisplatin chemosensitivity of bladder cancer T24 and EJ cells were also significantly increased by knocking down APAF1.

Conclusions
Based on our previous study and the present findings, we conclude that Cdr1as can adsorb miR-1270 in bladder cancer, restore its suppressed APAF1 expression level, induce apoptosis, and increase cisplatin chemosensitivity. Targeting the Cdr1as/miR-1270/ APAF1 axis presents a new strategy to enhance the cisplatin chemosensitivity of bladder cancer patients. Cdr1as is potentially playing role in predicting and monitoring cisplatin chemosensitivity in patients with bladder cancer.
cisplatin. (D) Knocking down Cdr1as could reduce the sensitivity of TCCSUP and 5637 cells to cisplatin. Data represent the mean AE SD from three independent experiments. Student's t-test with two biological independent replicates were used to determine statistical significance; *P < 0.05, **P < 0.01. Fig. S3. The nude mice metastasis models. (A) Almost no signal detected in a normal nude mouse (Scale bar, 10mm) by IVIS 100 Imaging System and its lungs (Scale bar, 5mm). (B) High signal detected in a mouse with metastasis (Scale bar, 10mm) by IVIS 100 Imaging System and its lungs (Scale bar, 5mm). Fig. S4. Knockdown of Cdr1as could decrease cell apoptosis in TCCSUP and 5637 cells. (A, B) Cell apoptosis analyzed by flow cytometry in TCCSUP and 5637 cells after the knockdown of Cdr1as. Data represent the mean AE SD from three independent experiments. Student's t-test with two biological independent replicates were used to determine statistical significance; *P < 0.05, **P < 0.01 Fig. S5. MiR-7, miR-135a and miR-1246 showed no influence on the sensitivity to cisplatin in T24 cells. (A-C) Cells viability was determined by the CCK-8 method to be of no significant differences among miR-7, miR-135a and miR-1246 over-expressed cells in a series of dilute concentrations of cisplatin. (D) Overexpression of miR-7, miR-135a or miR-1246 could not affect the IC50 of T24 cells to cisplatin. Data represent the mean AE SD from three independent experiments. Student's t-test with two biological independent replicates were used to determine statistical significance; *P < 0.05, **P < 0.01 Fig. S6. Representative IHC analysis of APAF1 protein in bladder cancer tissues. Magnification: 200 9 (top) and 4009 (bottom) (Scale bar, 10lm). Fig. S7. Kaplan-Meier survival analysis for OS of patients in an aggregate bladder cancer dataset according to APAF1 and miR-1270 expression status. The P value was determined using the log-rank test. Original data was obtained from TCGA. Table S1. Demographic and clinical features of 32 patients with bladder cancer. Table S2. List of all potential targets of miR-1270 predicted by 4 databases. Table S3. List of 189 common potential targets of miR-1270 predicted by 4 databases.