Circular RNA AKT3 upregulates PIK3R1 to enhance cisplatin resistance in gastric cancer via miR-198 suppression

In total, 149 GC tissues (cohorts 1, 2) were obtained from the First Affiliated Hospital of Nanjing Medical University. All samples were collected in accordance with HIPAA guidelines and approved institutional protocols. Patients received treatment with standard CDDP-based therapeutic regimens after surgery. Disease-free survival (DFS) was defined as the time interval between gastrectomy (R0 excision) and the time of either disease recurrence or disease-associated death. CDDP resistance was defined as tumor relapse during CDDP-based chemotherapy after R0 excision, and CDDP sensitivity was defined as no tumor recurrence during CDDP-based therapy; both definitions followed standard CDDP response definitions published elsewhere [14]. Forty-four samples (Cohort 1) were used for circRNAs validation, and another 105 samples (Cohort 2) were used to quantify circAKT3 levels and to analyze the correlation between circAKT3 expression and outcomes after R0 excision in patients undergoing CDDP-based chemotherapy. The samples from cohorts 1 and 2 were obtained in 2013–2016 and 2007–2011, respectively. The grouping of the ROC curve was based on the median relative expression of circAKT3. Detailed information is listed in Additional file 1: Table S1.

The CDDP-sensitive cell lines SGC7901 and BGC823 as well as their CDDP-resistant strains (SGC7901CDDP and BGC823CDDP, respectively) were maintained in RPMI 1640 medium (Wisent, Shanghai, China) supplemented with 10% fetal bovine serum (FBS) (Wisent, Biocenter, China) (Additional file 2: Figure S1A). 293 T cells were cultured in DMEM with high glucose (Gibco-BRL, Carlsbad, CA, USA) supplemented with 10% FBS. 293 T, SGC7901CDDP, BGC823 and SGC7901 cells were purchased from the Cell Bank of Type Culture Collection of Chinese Academy of Sciences, and BGC823CDDP cells were established as previously described [15].

To predict the miRNA-binding sites of circAKT3, we used the bioinformatic databases miRanda, PITA and RNAhybrid. Filtering restrictions were as follows: (1) total score ≥ 140, total energy < 17 kcal/mol; (2) combined interaction energy (△△G) < 10; and (3) minimum free energy (MFE) ≤ 20 kcal/mol. Detailed information is listed in Additional 3: Dataset S1.

Total RNA was extracted from GC cells or tissues using TRIzol Reagent (Invitrogen, 15,596,018). RNase R treatment was carried out for 15 min at 37 °C using 3 U/mg RNase R (Epicenter). For Quantitative real-time PCR (RT-PCR), 500 ng of treated RNA was directly reverse transcribed using Prime Script RT Master Mix (Takara, Japan) and either random or oligo(dT) primers. Reverse transcription of miRNA was performed using a New Poly(A) Tailing Kit (ThermoFisher Scientific, China). mRNA was reverse transcribed into cDNA with a PrimeScript RT Master Mix Kit (Takara, RR036A, Japan). cDNA was amplified using Universal SYBR Green Master Mix (4,913,914,001, Roche, Shanghai, China). The CT value was measured during the exponential growth phase. Relative gene expression levels were determined using the 2^-△△CT method. The primers used are listed in Additional file 1: Table S2.

SGC7901CDDP and BGC823CDDP cells were lysed on ice for 10 min in 0.3% NP-40/NIB-250 buffer (15 mM Tris–HCl (pH 7.5), 60 mM KCl, 15 mM NaCl, 5 mM MgCl2, 1 mM CaCl₂ and 250 mM sucrose) supplemented with protease inhibitors. After centrifugation at 600 × g for 5 min at 4 °C, the resultant supernatant was collected as the cytoplasmic fraction and mixed with an equal volume of TRIsure reagent. After the pellet was washed with NIB-250, the nuclei were lysed in TRIsure reagent.

The method for overexpressing circRNAs was reported previously [16]. For the construction of circAKT3 overexpression plasmids, human circAKT3 cDNA was amplified using PrimerSTAR Max DNA Polymerase Mix (Takara, RR036A, Japan) and inserted into the pCD5-ciR vector (Greenseed Biotech Co, Guangzhou, China). The pCD5-ciR vector contains a front circular frame and a back circular frame. Transfection was carried out using Lipofectamine 2000 (Invitrogen) according to the manufacturer’s instructions. The luciferase reporter containing the circAKT3 sequence in the 3′-UTR was constructed by subcloning the circAKT3 fragment into the region directly downstream of a cytomegalovirus promoter-driven firefly luciferase (FL) cassette in a pCDNA3.0 vector. Mutations of each miRNA-binding site in the circAKT3 sequence were created using a Mut Express II Fast Mutagenesis Kit (Vazyme, Nanjing, China). The mutations were introduced in both the circAKT3-expressing vector and the luciferase reporter containing the circAKT3 sequence.

siRNA and miRNA mimics and inhibitors were synthesized by GenePharma (Shanghai, China). The sequences used are listed in Additional file 1: Tables S3 and S4. Transfection was carried out using Lipofectamine RNAiMAX (Life Technologies) according to the manufacturer’s instructions.

A pull-down assay was performed as described previously [17, 18]. The biotin-labeled circAKT3 probe was synthesized by RiboBio (Guangzhou, China). In brief, 1 × 10⁷ circAKT3-overexpressing GC cells were harvested, lysed, and sonicated. The circAKT3 or oligo probe was incubated with streptavidin-coupled Dynabeads (Invitrogen) at 30 °C overnight to generate probe-bound Dynabeads. After the treated beads were washed with wash buffer, the RNA complexes bound to the beads were eluted and disrupted with lysis buffer and proteinase K prior to RT-PCR or RT-qPCR. Biotinylated probes sequences used in this study (see Additional file 1: Table S5).

293 T, SGC7901CDDP and BGC823CDDP cells were seeded in 24-well plates and cotransfected with corresponding plasmids and miRNA mimics in triplicate. At 48 h after transfection, luciferase reporter assays were conducted using a dual-luciferase reporter assay system (Promega, Madison, WI) according to the manufacturer’s instructions. Relative luciferase activity was normalized to Renilla luciferase activity.

The double FISH assay was performed in SGC7901CDDP cells and GC tissues as previously described [16, 19]. Biotin-labeled probes specific to circAKT3 and Dig-labeled locked nucleic acid miR-198 probes were used in the hybridization (Exiqon, Vedbaek, Denmark). The sequences are listed in Additional file 1: Table S6, FISH probes sequences used in this study. The signals of the biotin-labeled probes were detected using Cy5-conjugated streptavidin (Life Technologies), and the signals of the Dig-labeled miR-198 probes were detected using a tyramide-conjugated Alexa 488 fluorochrome TSA kit. Nuclei were counterstained with 4,6-diamidino-2-phenylindole. Images were acquired on a Leica TCS SP2 AOBS confocal microscope (Leica Microsystems, Mannheim, Germany). CircAKT3 and miR-198 expression levels were evaluated by the proportions and intensities of the positive cells detected within 5 fields of view on every slide (400-fold magnification). Proportion scores were assigned as follows: < 10% = 0, 10–25% = 1, 26–50% = 2, 51–75% = 3 and > 75% = 4. Intensity scores were assigned as follows: 0 = no staining, 1 = weak, 2 = moderate, 3 = strong and 4 = significantly strong.

For western blot analysis, cells were extracted using a protein extraction kit (Key Gene, KGP9100). Lipid proteins were added into 8, 10, 12% or 15% gels, subjected to 120 V to promote migration, and then transferred onto nitrocellulose membranes. The membranes were blocked with 5% BSA in TBST buffer and incubated with specific primary antibodies at 4 °C overnight. The next day, membranes were washed 3 times for 15 min in TBST and incubated with secondary antibodies for 2 h at room temperature. HRP substrate (WBKL0100, Millipore, USA) was used to detect the protein bands (Molecular Imager, ChemiDoc XRS+, BIO-RAD, USA), and the band intensities were quantified using Image-Pro Plus software (Mediacy, USA). Detailed information of antibody used in this study (see Additional file 1: Table S7).

The cytotoxicity assay was performed as previously described [15]. Cell viability was measured using Cell Counting Kit-8(CCK8)following the manufacturer’s directions (Dojindo, Kumamoto, Japan).

A clonogenic assay was performed as previously described [15]. At 48 h after transfection, BGC823CDDP, SGC7901CDDP, BGC823 and SGC7901 cells were cultured with CDDP at the indicated concentrations for 3 h. Then, the cells were harvested, seeded into six-well plates (500 cells per well) and cultured for an additional 2 (BGC823CDDP and SGC7901CDDP cells) or 3 weeks (BGC823 and SGC7901 cells). For scoring the colony-forming units, we fixed cells in 1 ml of methanol for 10 min and then stained the cells with crystal violet for 15 min.

Cell apoptosis was detected using a PI/Annexin V-FITC Apoptosis Detection Kit (BD Pharmingen, 556,547) according to the manufacturer’s instructions. Briefly, after GC cells were treated with CDDP at the indicated concentrations for 48 h in 6-well plates, they were harvested and resuspended in 300 ml of binding buffer. Next, 5 μl of Annexin V-FITC and 5 μl of PI were added to the suspensions, and the cells were incubated in the dark at 4 °C for 15 min. The samples were subsequently analyzed with a flow cytometer (Gallios, Beckman, USA).

The Actinomycin D assay was performed as previously described [16]. SGC7901CDDP and BGC823CDDP cells were seeded in 5 wells in 24-well plates (5 × 10⁴ cells per well). Twenty-four hours later, the cells were exposed to Actinomycin D (2 μg/ml, Abcam, ab141058) for 0 h, 6 h, 12 h, 18 h and 24 h. The cells were then harvested, and the relative RNA levels of circAKT3 and AKT3 mRNA were analyzed by RT-qPCR and normalized to the values measured in the group in the 0 h group (mock treatment).

Cells seeded onto coated cover slips growth for 24 h, then treated with CDDP, and harvested the cells at 0, 2, and 8 h. The cells were fixed with 4% paraformaldehyde at room temperature for 15 min and then permeabilized with PBS containing 0.25% Triton X-100 for 10 min. Next, the cells were blocked with 1% BSA for 20 min before incubation with primary antibodies at room temperature for 2 h. After the cells were washed with PBS, they were incubated with appropriate secondary antibodies (FITC-conjugated goat anti-rabbit, Molecular Probes, USA) at room temperature for 2 h. Following a final wash with PBS, cells were mounted with antifading mounting medium containing DAPI. The images were captured with a Leica DMI3000B (Germany) fluorescence microscope.

SGC7901CDDP cells stably expressing circAKT3 siRNA (si-circ-1) and its negative control siRNA (si-NC) were generated by infection with lentiviruses as previously described [20]. Transfection was carried out according to the manufacturer’s instructions. The lentiviral expressing vectors were purchased from HanBio Co. Ltd. (Shanghai, China).

Six-week-old female BALB/c nude mice were purchased from the Laboratory Animal Center of Nanjing Medical University and maintained under pathogen-free conditions. A total of 5 × 10⁶ SGC7901CDDP cells infected with lentivirus containing si-circ-1 or si-NC (2 μl of 10⁹ viral genomes μl^− 1, HanBio) in 100 μl of PBS were subcutaneously injected into different sides of the groin of each mouse. One week after injection, we intraperitoneally injected mice with cisplatin (5 mg/kg) in PBS or PBS alone three times per week. The xenograft tumors were harvested after 5 weeks. The entire experimental protocol was conducted in accordance with the guidelines of the local institutional animal care and use committee.

Xenografts and GC tissues exposed to the indicated concentrations of CDDP were prepared for IHC as previously described [21]. Sections were identified by IHC Imager (DM4000B, LEIKA, Germany), and target protein expression levels were evaluated by the proportions and intensities of positive cells detected within 5 fields of view on every slide (400-fold magnification). Proportion scores were assigned as follows: < 10% = 0, 10–25% = 1, 26–50% = 2, 51–75% = 3 and > 75% = 4. Intensity scores were assigned as follows: 0 = no staining, 1 = weak, 2 = moderate, 3 = strong and 4 = significantly strong.

All experiments were performed in triplicate. Data were analyzed with SPSS 19.0 software (IBM, USA) and presented as the mean ± SEM. The statistical significance of the results was calculated using an unpaired Student’s t-test. DFS analysis was performed using the Kaplan-Meier method and log-rank test. Clinicopathological features were analyzed by a χ² test. A Cox proportional hazards regression model was used to identify independent prognostic factors associated with DFS. Linear correlation analyzes were performed to determine correlations between circAKT3, miR-198 and PIK3R1 expression levels. A P value< 0.05 was defined as statistically significant.

To characterize circular RNA transcripts, we conducted RNA-Seq analysis of CDDP-resistant SGC7901 and BGC823 cells (i.e., SGC7901CDDP and BGC823CDDP) and their corresponding parental strains (i.e., SGC7901 and BGC823), which are sensitive to CDDP. The sequencing statistics are not shown. The analysis indicated that a series of circRNAs were differentially expressed in CDDP-resistant GC cells compared with the sensitive parental GC cells. We then chose the top 20 significantly upregulated circRNAs and verify their expression levels. Detailed information of 20 candidate circRNAs in Additional file 1: Table S8 (including location, genomic and spliced length). Using divergent primers to specifically target the circular junction as well as combined quantitative reverse transcription PCR (RT-qPCR) analysis and sequencing validation, we found that only 10 of these circRNAs had confirmed differences in expression and that circAKT3 was the most obviously upregulated circRNA in CDDP-resistant patients of cohort 1 (Fig. 1a and Additional file 2: Figure S1b-c). circAKT3 (hsa_circ_0000199) has been mapped to exons 8, 9, 10, and 11 of the AKT3 gene (555 bp) (Additional file 2: Figure S1d). Consistent with the RNA-Seq results, the expression of circAKT3 was obviously increased in CDDP-resistant GC cells (Fig. 1b). Subsequently, we verified the head-to-tail splicing of the RT-PCR product of circAKT3 by Sanger sequencing (Fig. 1c). Meanwhile, to exclude possibilities such as genomic rearrangements or trans-splicing, several experiments were employed. First, we designed convergent primers to amplify AKT3 mRNA and divergent primers to amplify circAKT3. Using cDNA and genomic DNA (gDNA) from SGC7901CDDP and BGC823CDDP cell lines as templates, the circAKT3 amplification product was only observed in cDNA by divergent primers but not in gDNA (Fig. 1d). In addition, the fragment of the linear form of AKT3 was digested by RNase R, but circAKT3 remained after RNase R treatment (Fig. 1e). Then, the relative expression levels of circAKT3 were detected in the cytoplasm and nucleus of SGC7901CDDP and BGC823CDDP cells (Fig. 1f and Additional file 2: Figure S1e). The RT-qPCR results demonstrated that circAKT3 was enriched in the cytoplasm. Moreover, we used Actinomycin D to suppress transcription and measure the half-life of circAKT3 in SGC7901CDDP and BGC823CDDP cells; we found that circAKT3 was more stable than AKT3 mRNA (Fig. 1g and Additional file 2: Figure S1f). Additionally, the FISH results displayed a dominantly cytoplasmic distribution of circAKT3 (Fig. 1h).

Next, we detected the expression level of circAKT3 in tissues of patients from cohort 2. Consistent with the RNA-Seq results, circAKT3 was significantly more highly expressed in the CDDP-resistant GC tissues than in the sensitive tissues (Fig. 2a). Compared with GC patients expressing low levels of circAKT3, GC patients receiving CDDP therapy and exhibiting upregulation of circAKT3 showed a significant association with decreased five-year DFS (Fig. 2b). To further verify that circAKT3 may be a therapeutic target for CDDP-resistant patients, we calculated the area under the receiver operating characteristic curve (AUC) using the expression levels of circAKT3. The area under the curve is 91% (Fig. 2c), suggesting that the expression level of circAKT3 is a good predictive biomarker of CDDP resistance for GC patients. Analysis of the clinicopathological characteristics in cohort 2 showed that circAKT3 expression was positively related to tumor size, histological grade, clinical stage, T classification and CDDP resistance (Table 1). A univariate analysis showed that DFS was obviously related to tumor size, histological grade, clinical stage and circAKT3 expression level (Additional file 1: Table S9). Subsequently, multivariate analysis indicated that circAKT3 expression, along with tumor size and clinical stage, was an independent risk factor for DFS (Additional file 1: Table S9 and Fig. 2d).

First, we designed two siRNA oligonucleotides (si-circ-1 and si-circ-2) to target the unique back-splice junction of circAKT3 (Fig. 3a); si-circ-1 successfully knocked down circAKT3 expression but had no effect on the levels of endogenous linear AKT3 transcript in SGC7901CDDP and BGC823CDDP cells (Fig. 3b and Additional file 4: Figure S2a). Additionally, to further assess the role of circAKT3, circAKT3 was overexpressed in SGC7901 and BGC823 cells via transfection of pCD5-ciR-AKT3 e8–11 (Fig. 4a). Importantly, elevated expression of circAKT3 had no effect on the levels of linear AKT3 mRNA, as confirmed by RT-qPCR (Fig. 4a). circAKT3 inhibition reduced the viability of SGC7901CDDP and BGC823CDDP cells (Fig. 3c and Additional file 4: Figure S2b). In addition, knockdown of circAKT3 significantly decreased the number of cell colonies (Fig. 3d and Additional file 4: Figure S2c) and promoted apoptosis (Fig. 3e and Additional file 4: Figure S2d). The phosphorylated histone family member X (γH2AX) forms discrete nuclear foci and acts as a platform to recruit additional factors and enhance the DNA repair pathway [22]. Meanwhile, circAKT3-knockdown cells showed significantly more γH2AX foci than the control cells at 2 h after CDDP treatment (Fig. 3g and Additional file 4: Figure S2e). circAKT3-knockdown cells had a higher percentage of active foci relative to that in control cells from 0 to 8 h after CDDP treatment (Fig. 3h and Additional file 4: Figure S2f). However, compared with the negative control, ectopic circAKT3 expression significantly increased cell viability and the number of cell colonies and inhibited apoptosis and the formation of γH2AX foci in SGC7901 and BGC823 cells (Fig. 4b-g and Additional file 5: Figure S3a-e). We used western blotting to investigate the underlying mechanism of these activities. In the presence of CDDP, knockdown of circAKT3 in SGC7901CDDP and BGC823CDDP cells increased cleaved caspase-3 protein levels, while the levels of the inactivated form of caspase-3 protein was decreased (Fig. 3f). In contrast, cleaved and inactivated caspase-3 protein levels were observed when circAKT3 was overexpressed (Fig. 4e). These data are consistent with a previous study reporting that CDDP-induced increases in Breast cancer type 1 susceptibility protein(BRCA1) expression leads to enhanced DNA damage repair (DDR) in breast cancer cells [23]. After CDDP treatment, knockdown of circAKT3 in SGC7901CDDP and BGC823CDDP cells increased γH2AX but decreased BRCA1 protein levels. circAKT3 overexpression also inhibited γH2AX and promoted BRCA1 protein levels compared with the levels in the controls (Figs. 3i and 4h).

To address whether circAKT3 could sponge miRNAs in GC cells, we selected 11 candidate miRNAs by overlapping the prediction results of the miRNA recognition elements in the circAKT3 sequence using miRanda, PITA, and RNAhybrid (Fig. 5a-b). Next, we investigated whether candidate miRNAs could directly bind circAKT3. A biotin-labeled circAKT3 probe was designed and verified to pull down circAKT3 in SGC7901CDDP and BGC823CDDP cell lines, and the pull-down efficiency was significantly enhanced in cells with stable circAKT3 overexpression (Fig. 5c-d). The miRNAs were extracted after pull-down, and the levels of the 11 candidate miRNAs were detected by RT-qPCR. As shown in Fig. 5e-f, in both SGC7901CDDP and BGC823CDDP cells, miR-198 was abundantly pulled down by circAKT3. Furthermore, using the RNAhybrid bioinformatics prediction tool, we calculated the secondary conformation of circAKT3 and miR-198 and found that there were 8 predicted binding domains (largest combined with a MFE > − 20 kcal/mol) (Additional file 3: Dataset S2). Next, the results of the luciferase reporter assays showed that miR-198 expression significantly reduced the luciferase activity of the reporter containing the complete circAKT3 sequence appended to the 3′-UTR of luciferase (luc-wt) compared to that of the reporter containing circAKT3 with mutated miR-198 binding sites (luc-m1, m2 and m8) (Fig. 5g-h). Moreover, RNA FISH assays revealed that circAKT3 and miR-198 were colocalized in the cytoplasm (Fig. 5i).

A microarray assay was further performed with SGC7901CDDP, BGC823CDDP, SGC7901 and BGC823 cells to validate the results of the ceRNA analysis. We analyzed the top 20 upregulated genes according to four algorithms (miRanda, RNAhybrid, miRWalk and TargetScan) prediction, and miR-198 could target the 3’UTRs of PIK3R1, CHRM3, HIPK2 and MAFB (Additional file 3: Dataset S3). We performed luciferase reporter assays to determine whether miR-198 directly targets these 4 genes in 293 T cells (Fig. 6c and Additional file 6: Figure S4c and e). In 293 T, SGC7901CDDP and BGC823CDDP cells cotransfected with miR-198 mimic, reporter constructs containing wild-type miR-198 binding sites at the PIK3R1 3′UTR exhibited decreased luciferase activity relative to that of reported constructs with mutated binding sites (Fig. 6d and Additional file 6: Figure S4f and g). PIK3R1 protein (p85α, encoded by PIK3R1) is the regulatory subunit of PI3K. A functional study demonstrated that PIK3R1 was highly expressed in CDDP-resistant ovarian cancer cells, and downregulated PIK3R1 resensitized the abovementioned cells to platinum-based treatment, which reveals the promising involvement of p85α in secondary CDDP resistance [24]. Compared with parental CDDP-sensitive cells, CDDP-resistant cells showed obvious increases in the expression of PIK3R1 mRNA and protein levels (Fig. 6f and g). Furthermore, we found that miR-198 mimics significantly inhibited PIK3R1 mRNA and protein levels and that ectopic PIK3R1 expression abolished the influence caused by miR-198 overexpression (Fig. 6h and i). Subsequently, the data showed that overexpression of miR-198 inhibited cell viability and induced apoptosis in SGC7901CDDP and BGC823CDDP cells. However, cotransfection of PIK3R1 and miR-198 abrogated these effects (Fig. 6j and k and Additional file 6: Figure S4h-j and Additional file 7: Figure S5a-b).

Cotransfection of si-circ-1 and anti-miR-198 could counteract the si-circ-1-induced downregulation of PIK3R1 in SGC7901CDDP cells (Fig. 7a). Notably, cotransfection of circAKT3 and miR-198 attenuated the expression of PIK3R1 compared to transfection of circAKT3 alone in SGC7901 cells (Fig. 7b). The CCK8 and flow cytometry analyses indicated that transfection with si-circ-1 inhibited cell viability and promoted apoptosis after CDDP treatment while cotransfection with si-circ-1 and anti-miR-198 significantly promoted cell viability and inhibited apoptosis compared with the controls (Fig. 7c and d, Additional file 7: Figure S5d). Furthermore, upregulation of circAKT3 led to enhanced cell viability, but this effect could be significantly abolished by ectopic expression of miR-198 (Additional file 8: Figure S6a). Additionally, overexpression of circAKT3 induced a reduction in apoptosis. However, cotransfection of circAKT3 and miR-198 mimics led to enhanced apoptosis (Additional file 8: Figure S6b). Notably, circAKT3 upregulation could inhibit γH2AX expression, as indicated by the reduced fluorescence of γH2AX, cotransfection of circAKT3 and miR-198 mimics led to abolish this effect (Additional file 8: Figure S6c and d). Transfection of si-circ-1 significantly reduced PIK3R1 expression and inhibited the expression of canonical PI3K/AKT signaling molecules, as shown by western blot, and downregulating of both circAKT3 and miR-198 abrogated these effects in SGC7901CDDP cells (Fig. 7e). Similar results are presented in Additional file 9: Figure S7a-f. Meanwhile, Transfection of circAKT3 significantly increased PIK3R1 expression and induced the expression of canonical PI3K/AKT signaling molecules, as shown by western blot, and concomitant overexpression of both circAKT3 and miR-198 abrogated these effects in SGC7901 cells (Fig. 7f). Transfection of si-PIK3R1 significantly inhibited PIK3R1 and levels of phosphorylated PI3K/AKT signaling molecules in SGC7901 cells with circAKT3 overexpression. We employed a specific p110α inhibitor, BKM20, and performed western blotting to determine whether deactivation of PI3K/AKT signaling can overcome the changes caused by circAKT3 overexpression. The results indicated that si-PIK3R1 and BKM20 significantly inhibited PIK3R1 and p110α levels, respectively as well as inhibited p-AKT, reduced BRCA1 levels, increased cleaved caspase-3 levels, and promoted γH2AX (Fig. 7g). These results suggest that circAKT3 functions by targeting miR-198 as a ceRNA to regulate PIK3R1 expression, activate the PI3K/AKT signaling cascade and facilitate CDDP resistance.

To investigate the potential clinical relevance of circAKT3 in vivo, we subcutaneously injected SGC7901CDDP cells with or without stable circAKT3 knockdown (Additional file 8: Figure S6e) into the dorsal flanks of female BALB/c nude mice and allowed the cells to proliferate for 5 weeks. Tumor xenograft data indicated that circAKT3 inhibition in CDDP-resistant cells can significantly decrease xenograft tumor growth and sensitize cells to CDDP treatment (Fig. 8a and b). IHC analysis of tumor xenograft samples further indicated that the protein levels of γH2AX and cleaved caspase-3 were notably increased, but BRCA1 was decreased upon circAKT3 inhibition (Fig. 8c). FISH showed the colocalization of circAKT3 and miR-198 in tissues from patients with CDDP-resistant or CDDP-sensitive GC (Fig. 8d). The FISH score data showed that the expression of circAKT3 was significantly higher in CDDP-resistant GC tissues than CDDP-sensitive GC tissues; however, miR-198 expression showed the opposite result (Fig. 8d). Similarly, IHC scores and western blot analyses indicated that PIK3R1 protein expression was obviously increased in CDDP-resistant GC tissues compared to CDDP-sensitive GC tissues (Fig. 8e and Additional file 10: Figure S8a and b). Furthermore, correlations were identified between circAKT3 and miR-198 expression levels and PIK3R1 protein levels in these 44 GC tissue samples (Fig. 8f). Based on these data, we conclude that circAKT3 increases the tolerance of GC cells to CDDP by targeting PIK3R1 through miR-198.