Circular RNA circFGFR1 promotes progression and anti-PD-1 resistance by sponging miR-381-3p in non-small cell lung cancer cells

The seven human NSCLC cell lines (NCI-H358, NCI-H1299, A549, HCC827, NCI-H1650, NCI-H838, and NCI-H292), the LLC mouse lung cancer cell line, and the HEK-293 T cell line were purchased from the Shanghai Institute of Cell Biology, Chinese Academy of Sciences (Shanghai, China). The cells were cultured in Dulbecco’s modified Eagle’s medium (DMEM, HyClone, Logan City, UT) supplemented with 10% fetal bovine serum (FBS, Gibco, Carlsbad, CA, USA), 100 units/ml penicillin and 100 μg/ml streptomycin (Gibco, Carlsbad, CA, USA) in a humidified atmosphere containing 5% CO₂ at 37 °C.

A total of 210 NSCLC tissues and matched nontumor tissues were collected to construct the tissue microarray (TMA), which was analyzed by immunohistochemistry (IHC) and quantitative real-time polymerase chain reaction (RT-qPCR). All patients underwent curative resection, as confirmed by pathological examination at East Hospital of Tongji University, the Second Affiliated Hospital of Nanchang University, and Hwa Mei Hospital of University of Chinese Academy of Sciences. Clinicopathological information was collected from 1 January 2010 to 31 December 2013. The Ethics Committee of the East Hospital of Tongji University Biomedical Research Department provided ethical approval, and informed consent for collecting and preserving samples and documenting details was obtained from every patient.

RT-qPCR, western blotting analysis, IHC, and FISH assays were performed as described in previous studies [7, 20] and described in Additional file 1: Supplementary Materials and Methods. The RT-qPCR primers used in this study are listed in Additional file 2: Table S1. The antibodies used in this study are listed in Additional file 3: Table S2.

Cell proliferation, clonal formation, wound-healing cell migration, and Matrigel invasion assays were performed as described in our previous studies [7, 21] and in Additional file 1: Supplementary Materials and Methods.

In vivo circRNA precipitation (circRIP), RIP, and luciferase reporter assays were performed as described in previous studies [7, 20, 22]. Biotin-labeled circFGFR1 and negative control probes were synthesized by the GeneChem Company. In brief, A549 cells were washed with cold phosphate-buffered saline, fixed with 1% formaldehyde, lysed in co-IP buffer, sonicated and centrifuged. Then, the supernatant was cultured with M280 streptavidin Dynabeads (Invitrogen) mixture and incubated at 30 °C for 12 h. Subsequently, the mixture was washed and incubated with lysis buffer and proteinase K. RNA was extracted from the mixture using TRIzol Reagent (Invitrogen).

The RIP assay was carried out using a Magna RIP RNA-binding protein immunoprecipitation kit (Millipore). In brief, cell lysates were cultured with Dynabeads coated with AGO2 antibody or IgG antibody at 4 °C for 12 h, and total RNA was extracted for the detection of enriched circFGFR1 and miRNA by RT-qPCR.

For the luciferase reporter assay, potential binding sites were predicted using StarBase v3.0. HEK-293 T cells were cotransfected with pGL3-LUC-circFGFR1, pGL3-LUC-UBE2C/FGF7/FGFR2/FGFR1/IGF2BP1/MET/CXCR4 3′ UTR or mutant pGL3-LUC-circNT5E, pGL3-LUC-UBE2C/FGF7/FGFR2/FGFR1/IGF2BP1/MET/CXCR4 3′ UTR and miR-381-3p mimics or negative control mimics. Forty-eight hours later, the cells were harvested, and the luciferase activity was measured with the dual-luciferase reporter assay system (Promega). The argonaute 2 (AGO2) and IgG antibodies used in this study are listed in Additional file 3: Table S2.

Small hairpin RNAs (shRNAs) targeting the junction region of the circFGFR1 sequence and circFGFR1-overexpressing lentivirus were synthesized by Hanbio Company (Shanghai, China). NSCLC cell lines were transfected with circFGFR1 shRNA or the circFGFR1-overexpressing lentivirus according to the manufacturer’s instructions.

In vivo tumor growth and metastasis experiments in nude mice were approved by the Animal Experimentation Ethics Committee of East Hospital, Tongji University. The experiments were performed as described in our previous studies [21, 23] and in Additional file 1: Supplementary Materials and Methods.

CRISPR/Cas9 double-vector lentivirus was used to establish stable CXCR4-knockout NSCLC cell lines. Double-vector lentivirus was purchased from the GeneChem Company and transfected as described in reference [24].

The experiments in the C57BL/6 mice were approved by the Animal Experimentation Ethics Committee of East Hospital, Tongji University. A total of 2 × 10⁶ cells (LLC with or without increased circFGFR1 expression) were implanted subcutaneously in the left flank of the six-week-old C57BL/6 mice. When tumors reached a size of approximately 100 mm³, the mice were randomly assigned to 4 groups. Then, the mice were injected in the tail vein with a PD-1 monoclonal IgG antibody (a gift from Hengrui Medicine Company) or its mouse isotype control at 100 μg per dose three times a week for two weeks. Animals were euthanized when tumors reached a maximum of 1000 mm³ (n = 6). The day that the mice received the first therapy is considered day 1.

Statistical analysis was performed with SPSS software (19.0; SPSS, Inc., Chicago, IL) as in our previous study [7] and described in Additional file 1: Supplementary Materials and Methods. P < 0.05 was considered statistically significant.

In terms of the critical roles of the FGFR1-related pathway in NSCLC progression, FGFR1 amplification has been identified in all types of lung carcinoma [25‐27]. Therefore, we examined 17 circular RNAs derived from FGFR1 genes by analyzing the circular RNA sequencing data from circBase. Among 17 circRNAs, we found that circFGFR1 (hsa_circ_0084003) expression levels were significantly and consistently upregulated in four of the NSCLC tumor tissues compared to the expression levels in the matched adjacent nontumor lung tissues (Fig. 1a). CircFGFR1 was composed of 1844 nucleotides and 13 exons (Fig. 1b). To explore the relationship between circFGFR1 and clinical characteristics, we measured circFGFR1 expression in the NSCLC tissues and adjacent nontumor tissues and found that circFGFR1 was significantly upregulated in the NSCLC tissues (111/210) (Fig. 1c). Next, we explored the relationship between circFGFR1 expression and the clinicopathological characteristics of 210 NSCLC patients, as listed in Table 1. The results showed that NSCLC patients with circFGFR1^high cells had larger tumors (P = 0.010), lymph node metastasis (P = 0.004), and poor cell differentiation (P = 0.019). Then, we explored the prognostic implications of circFGFR1 expression. Importantly, our results showed that patients with circFGFR1^high cells expression had a significantly worse prognosis than those with circFGFR1^low (Fig. 1d and e). The results from the multivariate analysis indicated that circFGFR1 expression is an independent predictor for postoperative recurrence and overall survival (OS; Tables 2 and Table 3). These results indicate that circFGFR1 likely participates in the progression of NSCLC.

To explore the biological functions of circFGFR1 in NSCLC, we measured circFGFR1 expression in seven types of human NSCLC cells (Additional file 4: Figure S1a). Next, we designed two shRNA plasmids to target the unique back-splice junction. The back-splice junction-specific shRNA (shcircF1 and shcircF2) effectively knocked down circFGFR1 expression but had no effect on the level of FGFR1 mRNA in the A549 and HCC827 cells (cell lines with high circFGFR1 expression) (Additional file 4: Figure S1b-c). Using the above-mentioned vector, we succeeded in overexpressing circFGFR1 in NCI-H358 and NCI-H1299 cells (Additional file 4: Figure S1d). In vitro CCK-8, clone formation, wound-healing cell migration, and invasion assays revealed that the NCI-H358 and NCI-H1299 cells (which had low circFGFR1 expression) in which circFGFR1 expression was forced were significantly more likely to exhibit a malignant phenotype than the mock cells (Fig. 2a-d). Conversely, reduced circFGFR1 expression inhibited the proliferation, migration, and invasion abilities of the A549 and HCC827 cells, according to the results from the CCK-8, clonal formation, wound healing, and Matrigel Transwell assays (Additional file 4: Figure S2a-d). To verify the in vitro findings, we examined the biological role of circFGFR1 in mediating in vivo proliferation. NCI-H358 cancer cells with stably forced circFGFR1 expression were subcutaneously implanted into nude mice. Consistent with the above in vitro findings, the overexpression of circFGFR1 dramatically promoted tumor growth and lung metastasis (Fig. 2e and f).

An increasing number of studies have reported that circRNAs act as miRNA sponges; therefore, we investigated whether circFGFR1 has the ability to bind to miRNAs. Through StarBase v3.0, we found that 26 miRNAs were predicated to be possible targets of circFGFR1. To verify the critical functional miRNAs that may interact with circFGFR1 in NSCLC cells, a circFGFR1-specific probe was used to perform RNA in vivo precipitation (RIP) in A549 cells, which were then screened by the RT-qPCR for the potential miRNAs that had been predicted. Using RIP circFGFR1 pull-down experiments, we purified circFGFR1-associated RNAs and analyzed 26 candidate miRNAs in the complex. The results showed a specific enrichment of circFGFR1 and miR-381-3p compared to the negative control probe, whereas the other miRNAs were slightly enriched or not enriched, indicating that miR-381-3p is one of the critical circFGFR1-associated miRNAs in NSCLC cells (Fig. 3a). Next, we performed RNA immunoprecipitation (RIP) with argonaute 2 (AGO2) antibody in A549 and NCI-H1299 cells. Our results showed that circFGFR1 and miR-381-3p, but not circANRIL (a circular RNA that reportedly does not bind to AGO2) [20], were significantly enriched, as they were precipitated by the AGO2 antibody (Fig. 3b). These results indicated that circFGFR1 may act as a binding platform for AGO2 and miR-381-3p. To verify these results, we performed a luciferase assay using miR-381-3p mimics cotransfected with luciferase reporters (which contained a wild-type or miR-381-3p-target mutant circFGFR1 sequence) into HEK-293 T cells. Compared with the negative control RNA (NC), miR-381-3p decreased the luciferase reporter activity significantly in the cells with the wild-type circFGFR1 sequence but not the cells with either the wild-type- or the miR-381-3p-target mutant circFGFR1 sequence (Fig. 3c and d). Furthermore, using a pull-down assay with biotinylated miR-381-3p mimics, we observed significant enrichment of circFGFR1 compared with the level in the negative controls, while circANRIL was not enriched in the A549 and the NCI-H1299 cells (Fig. 3e). Moreover, the results from the double FISH assay indicated the colocalization of circFGFR1 and miR-381-3p in the NSCLC cells (Fig. 3f). In addition, miR-381-3p did not show significant changes after circFGFR1 was overexpressed or silenced, and circFGFR1 did not show significant changes after miR-381-3p expression was either upregulated or knocked down (Fig. 3g and h; Additional file 4: Figure S3). These findings indicate that circFGFR1 and miR-381-3p are likely not degraded by each other. All of the above experiments confirmed that circFGFR1 may function as a sponge for miR-381-3p in NSCLC cells.

To further explore the function of miR-381-3p in NSCLC cells, StarBase v3.0, the PITA and miRanda algorithms were used to predict potential targets of miR-381-3p. The predicted result showed that several oncogenes, including UBE2C, FGF7, FGFR2, FGFR1, IGF2BP1, MET, and CXCR4 3′ UTR mRNAs, contained target sequences for miR-381-3p (Fig. 4a). To verify whether the 3′ UTR of UBE2C, FGF7, FGFR2, FGFR1, IGF2BP1, MET, and CXCR4 mRNAs were targets of miR-381-3p in the NSCLC cells, a pLG3 luciferase reporter gene assay was used. The wild-type (wt) 3′ UTR sequence and the mutant (mu) 3′ UTR sequence of UBE2C, FGF7, FGFR2, FGFR1, IGF2BP1, MET, and CXCR4 were cloned and placed into a pLG3 luciferase reporter vectors. The luciferase activity was significantly inhibited by the miR-381-3p mimics in the HEK-293 T cells transfected with the wt 3′ UTR sequence. The luciferase activity was not changed by the miR-381-3p mimics in the HEK-293 T cells transfected with the mu 3′ UTR sequence (Fig. 4b; Additional file 4: Figure S4a). In addition, the levels of CXCR4 mRNA and protein were significantly decreased after miR-381-3p expression was upregulated in the A549 and HCC827 cells (Additional file 4: Figure S5a and b), and the levels of CXCR4 mRNA and protein were significantly increased after miR-381-3p expression was knocked down in the NCI-H358 and NCI-H1299 cells (Additional file 4: Figure S5c and d). We also used to RT-qPCR to measure the FGF7, FGFR2, FGFR1, IGF2BP1, and UBE2C mRNA expression in the A549 cells with or without forced miR-381-3p expression. The results showed that the UBE2C/FGF7/FGFR1/IGF2BP1/MET mRNA level was slightly inhibited by miR-381-3p in the A549 cells; however, FGFR2 was not inhibited (Additional file 4: Figure S4b). Moreover, the levels of CXCR4 mRNA and protein significantly were increased after the circFGFR1 expression was upregulated in the NCI-H358 and NCI-H1299 cells (Fig. 4c and d), and the levels of CXCR4 mRNA and protein expression were significantly decreased after circFGFR1 expression was knocked down in the A549 and HCC827 cells (Fig. 4e and f). Furthermore, using IHC, we measured the expression levels of CXCR4 in the 210 NSCLC patient tissues (Fig. 4g and h). A positive relationship between circFGFR1 and CXCR4 was found in the NSCLC patient tissue samples (Fig. 4i). In addition, using RT-qPCR, we measured the expression of miR-381-3p in the 210 NSCLC patient tumor tissues. A negative relationship between circFGFR1 and miR-381-3p was found in the NSCLC patient tissues (Fig. 4j). Finally, we explored the prognostic implications of CXCR4 and miR-381-3p expression. Importantly, our results showed that patients with CXCR4^high or miR-381-3p^low cells had significantly worse prognoses than those with CXCR4^low or miR-381-3p^high cells (Additional file 4: Figure S6).

To examine whether circFGFR1 promotes NSCLC cell proliferation, migration, and invasion via the miR-381-3p/CXCR4 axis, we established CXCR4-knockout NCI-H358 and NCI-H1299 cells (NCI-H358-CKO and NCI-H1299-CKO cells) through CRISPR/Cas9 (Additional file 4: Figure S7a). The CXCR4 protein did not show significant change after circFGFR1 expression was upregulated in the H358-CKO and NCI-H1299-CKO cells (Additional file 4: Figure S7b and c). Furthermore, results from the in vitro CCK-8, clone formation, wound-healing cell migration, and invasion assays revealed that the H358-CKO and NCI-H1299-CKO cells with forced circFGFR1 expression had no significantly change in malignant phenotype compared with that of the mock cells (Fig. 5a-d).

It has been reported that increased CXCR4 expression promotes epithelial-mesenchymal transition (EMT), invasion, and migration of NSCLC cells [28, 29]. Therefore, we examined EMT marker expression, including that of E-cadherin, N-cadherin, Snail, and Twist, in NSCLC cells with forced circFGFR1 expression. The results showed that circFGFR1 significantly promoted N-cadherin, Snail, and Twist expression compared with the expression in the mock groups (Additional file 4: Figure S8).

CXCR4 is a potential mediator that induces cytotoxic T-lymphocyte exclusion and participates in resistance to anti-PD-1 therapy of cancer [30, 31]. To further investigate the relationship between the circFGFR1/CXCR4 axis and immune escape, we measured the infiltration of CD8+ T cells in tissues from the 210 cases of NSCLC and matched nontumor tissues. The number of CD8+ T cells in the NSCLC tissues was significantly lower than that of the adjacent nontumor tissues (155/210; 2-fold) (Fig. 6a and b). The results from a scatter plot analysis revealed a negative correlation between circFGFR1/CXCR4 expression and CD8+ T cell frequency in the NSCLC tissues (Fig. 6c and d). Moreover, the results from a scatter plot analysis also revealed a positive correlation between miR-381-3p expression and CD8+ T cell frequency in the NSCLC tissues (Fig. 6e). These results suggest that circFGFR1 may exert its immunosuppressive effects by upregulating CXCR4 via sponging miR-381-3p. We then analyzed retrospective data from 20 patients with recurrent NSCLC who had undergone lung resection 2–60 months before receiving PD-1 antibody immunotherapy. After four therapy cycles, an efficacy assessment was performed using enhanced CT. In terms of RECIST1.1, the results showed that there were 3 patients with a partial response (PR), 5 patients with stable disease (SD), and 2 patients with progressive disease (PD) in the circFGFR1^low group; while 4 patients with stable disease (SD) and 6 patients with progressive disease (PD) were in the circFGFR1^high group (Fig. 6f). Overall, the above results indicated that circFGFR1 promoted NSCLC cell resistance to anti-PD-1 agents.

Interestingly, humans and mice have the same miR-381-3p sequence, and mouse CXCR4 contains a speculated seed-sequence of miR-381-3p (Additional file 4: Figure S9a). We completed a luciferase assay using miR-381-3p mimics that were cotransfected with luciferase reporters (containing mouse wild-type or the mutant miR-381-3p seed-sequence of CXCR4) into HEK-293 T cells. Compared with the negative control RNA (NC), miR-381-3p decreased the luciferase reporter activity significantly in the cells with the wild-type mouse CXCR4 sequence but not the cells with the mutant mouse CXCR4 sequence (Additional file 4: Figure S9b). Furthermore, our results showed that CXCR4 was significantly upregulated when circFGFR1 was overexpressed in LLC mouse lung cancer cells (Additional file 4: Figure S9c-e). To further assess the effects of circFGFR1 expression on anti-PD-1 therapy resistance, we analyzed the anti-tumor effects of the PD-1 antibody in xenograft C57BL/6 mice that received LLC-circFGFR1 cells or the respective mock cells. Compared to that of the mock cell group, the tumor growth in the LLC-circFGFR1 cell recipient xenograft mice showed an obvious phenotype of resistance to anti-PD-1 therapy, and the xenograft mice had a shorter survival time (Fig. 6g-i). In addition, our results showed that the expression of PD-L1 was not changed significantly when circFGFR1 was overexpressed in the LLC mouse lung cancer cells (Additional file 4: Figure S10a). Moreover, in the xenograft mice, the number of PD-1-positive LLC cells was significantly reduced when circFGFR1 was forcibly expressed (Additional file 4: Figure S10b). Consistent with the in vitro experiment, the PD-L1 expression in the LLC cells with forcibly expressed circFGFR1 was not significantly changed in the xenograft mice (Additional file 4: Figure S10c).