CircKEAP1 is downregulated in LUAD
The expression profile of circRNAs in seven paired samples of cancer tissues and adjacent normal tissues from LUAD patients was analyzed by Human CircRNA microarray. There were 32 dysregulated circRNAs with p-value < 0.05, of which 15 circRNAs were downregulated and 17 circRNAs were upregulated (Fig. 1A and B and Table s4). Since the downregulated circRNAs in cancer tissues might play tumor suppressive roles and could serve a novel method to treat cancer, the 15 downregulated circRNAs were chosen for furth study. Firstly, we analyzed these downregulated circRNAs in another two circRNA expression profile (GSE112214 and GSE101586) of LUAD in GEO database. The results showed hsa_circRNA_104126 and hsa_circRNA_102442 were significantly downregulated in all the three circRNA expression profile (Fig. 1C). Then, the two circRNAs were determined in the seven paired samples of cancer tissues and adjacent normal tissues from LUAD patients by qRT-PCR. Consistent with the results of circRNA expression profile, hsa_circRNA_104126 and hsa_circRNA_102442 were downregulated in the cancer tissues (Fig. 1D). Compared to the hsa_circRNA_104126, the expression level of hsa_circRNA_102442 in the normal tissues was much higher (Fig. 1D).
Hsa_circRNA_102442 (termed circKEAP1) is located at chr19: 10610070-10610756, and derived from exon2 of KEAP1 gene with a length of 686nt by alternative splicing (Fig. 2A). We amplified the back-spliced junction of circKEAP1by divergent primers and confirmed by Sanger sequencing (Fig. 2A). Then, we performed Northern blot analysis by a probe targeted the back-spliced junction, and the result showed that circKEAP1 could be observed at 686nt (Fig. 2B). Additionally, PCR analysis for reverse-transcribed RNA (cDNA) and genomic DNA (gDNA) showed that divergent primers could amplify products from cDNA of adjacent normal tissues but not from cDNA of cancer tissues and gDNA (Fig. 2C). Next, the expression level of circKEAP1 in another 105 paired LUAD samples was analyzed by quantitative reverse transcription PCR (qRT-PCR). As showed in Fig.2D, the expression of circKEAP1 was obviously decreased in cancer tissues (Fig. 2D).
The same results were revealed in cell experiments. CircKEAP1 was amplified from cDNA of the human normal lung epithelial cell line BEAS-2B, but not from cDNA of LUAD cancer cell line A549 and gDNA (Fig. 2E). Subsequently, we treated the BEAS-2B cells with Actinomycin D (an inhibitor of transcription) to explore the stability of circKEAP1 and KEAP1. As showed in Figure 2F, the half-life of circKEAP1 transcript exceeded 24 h and was much longer than KEAP1. Finally, qRT-PCR analysis was conducted for nuclear and cytoplasmic circKEAP1 to observe cellular localization of circKEAP1. As showed in Fig.1H, circKEAP1 transcript was preferentially located in the cytoplasm (Fig. 2G).
Taken together, circKEAP1 mainly existed in the cytoplasm, and significantly decreased in LUAD cancer tissues.
circKEAP1 inhibits tumor growth in vivo and in vitro
To study the function of circKEAP1 in LUAD progression, we synthesized the full-length cDNA of circKEAP1 and cloned into the pLCDH-ciR plasmid, which contained a front and back circular frame. We found that circKEAP1 plasmid could successfully increase circKEAP1 expression in A549 cells (Fig. 3A). A nude mice xenograft model by implanting A549 cells transfected with control plasmid or circKEAP1 plasmid was established to identify the effect of circKEAP1 on tumor growth in vivo. The A549 cells were subcutaneously injected into the right flank of 6-week-old male BALB/c nude mice. The tumor volumes were monitored from the 5 days after A549 cell injection, and the mice were sacrificed after 20 days. We found overexpression of circKEAP1 drastically suppress tumor growth of A549 cells (Fig. 3B). The tumor weights were remarkably decreased by circKEAP1 (Fig. 3C and D). As showed in figure 3E, both H&E staining and Ki-67 staining of these tumors showed decreased cell mitosis and lower percentage of proliferative cells in the circKEAP1 overexpression group. As the previous results in the cells (Fig. 3A), the expression level of circKEAP1 was much higher in the circKEAP1 overexpression group, compared to the control group (Fig. 3F). These data indicated that circKEAP1 repressed tumor growth in vivo. To identify the effect of circKEAP1 on cancer cell proliferation in vitro, the EDU assay and CCK-8 assay were performed to the A549 cells transfected with control plasmid or circKEAP1 plasmid. As the results showed in Fig. 3G and H, both the EDU assay and CCK-8 assay revealed circKEAP1 significatly suppressed the cancer cell growth in vivo.
In summary, these findings suggest that circKEAP1 could repress the tumor growth in vivo and in vivo.
circKEAP1 act as a sponge for miR-141-3p
Recently, circRNA has been increasingly found to be involved in cancer by releasing downstream molecules by competing with miRNA 5, 6. As showed in Fig.2G, circKEAP1 was mainly located in the cytoplasm (Fig. 2G). CircKEAP1 was speculated to be severed as a miRNA sponge. In order to verify the hypothesis, we firstly analyzed the miRNA expression profiles of the seven paired LUAD cancer tissues and non-tumor tissues by Human miRNA Expression Assay. The volcano plot showed significantly different profiles of miRNAs between the seven paired samples of LUAD and adjacent normal tissues (Fig. 4A and Table s6). Then, we predicted the potential binding sites of miRNAs in the circKEAP1by Targetscan18, and found a total of 135 miRNAs were identified as potential targets of circKEAP1 (Table s7). As the circKEAP1 has been found to be downregulated in LUAD cancer tissues, compared to adjacent normal tissues (Fig. 2D), the miRNA which was upregulated in the LUAD tissues and had the binding site with circKEAP1 was choose for further study. The resulted showed 13 miRNAs, including miR-141-3p, miR-106b-3p, miR-93-3p, miR-21-3p, miR-19b-3p, miR-29b-5p, miR-3445-5p, miR-375 and miR-200c-3p, contained at least one binding site with circKEAP1 and significantly upregulated in LUAD cancer tissues (Fig. 4B). A dual-luciferase reporter assay was performed to further confirm the interaction between circKEAP1 and its predicted 13 miRNAs in A549 cells. Among the 13 predicted miRNAs, all the miRNAs significantly attenuated the luciferase activity of A549 cells (Fig. 4C), and the miR-141-3p was more obvious than other miRNAs and selected for deeper investigation. It has been proved that circRNAs bind with miRNAs through AGO2 (Argonaute 2). Therefore, we conducted the anti-AGO2 RNA immunoprecipitation (RIP) assay in BEAS-2B cells to pull down the RNA transcripts which bind to AGO2 with anti-AGO2 antibody. Interestingly, both circKEAP1 and miR-141-3p were efficiently pulled down by anti-AGO2 antibodies (Fig. 4D). Thus, it’s supposed that circKEAP1 might act as a sponge for miR-141-3p. Subsequently, we examined the expression levels of miR-141-3p in the 105 pairs of LUAD cancer tissues and adjacent non-cancerous tissues, and found miR-141-3p was markedly upregulated in LUAD cancer tissues compared with adjacent non-tumor tissues (Fig.4E). As showed in Fig.4F, there are two binding sites between miR-141-3p and circKEAP1 (Fig. 4F). We performed the dual-luciferase reporter assay to confirm the bioinformatics prediction analysis in A549 cells. The full-length of circKEAP1 and muted circKEAP1 with muted miR-141-3p binding sites were subcloned into luciferase reporter pGL3 plasmid. We found miR-141-3p mimic could effectively increase the expression level of miR-141-3p in A549 cells (Fig S1), and significantly decreased the luciferase activity of wildtype group, however it had no effect on the mutant group (Fig. 4G). These results suggested there might be a direct interaction between circKEAP1 and miR-141-3p. Then, the miRNA pull-down assay with specific biotin-labeled miR-141-3p was performed to further verify the binding of circKEAP1 and miR-141-3p. As expected, circKEAP1was significantly enriched in the biotin-labeled miR-141-3p group compared with control (Fig. 4H). Additionally, we also found the expression level of miR-141-3p in both A549 cells (Fig. 4I) and tumors (Fig. S2A) was markedly decreased when we overexpressed the circKEAP1 by circKEAP1 plasmid. The decrease could be attenuated by miR-141-3p mimic in A549 cells (Fig. 4J).
These results reveal that circKEAP1 could server as a sponge for miR-141-3p in LUAD cancer cells.
miR-141-3p repress the KEAP1 expression in tumors
It’s well known that circRNAs can act as miRNAs sponge to regulate downstream targets 5, 6, by sharing the same miRNAs with mRNA 7, 19. Three bioinformatics software TargetScan 18, miRDB 20 and miRanda 21 were used to predict the target genes of miR-141-3p. Bioinformatics analysis showed that the 3′-UTRs of ZEB1, ZEB2, and KEAP1 contained miR-141-3p complementary sequences by all the three algorithm (Fig. 5A). In order to confirm the repression of miR-141-3p, we firstly performed dual luciferase reporter assay in A549 cells. We found miR-141-3p remarkably suppressed the luciferase activity of luciferase reporter vector containing the KEAP1 3’UTR sequence, while it has no effect on ZEB1 and ZEB2 in A549 cells (Fig. 5B). As predicted, miR-141-3p have one binding sites in the 3’-UTR of KEAP1 (Fig.5C). In order to further confirm the repression of miR-141-3p on KEAP1, we firstly performed dual luciferase reporter assay in A549 cells. We found miR-141-3p remarkably suppressed the luciferase activity of luciferase reporter vector containing the KEAP1 3’UTR wild type sequence (Fig. 5D), while had no influence on the muted vector containing the muted KEAP1 3’UTR sequence (Fig. 5D). Then, we performed the miRNA pull-down assay with specific biotin-labeled miR-141-3p, and the result showed KEAP1 mRNA was also specific enrichment in the biotin-labeled miR-141-3p group, like circKEAP1 (Fig. 5E). The IHC staining for KEAP1 protein in 105 paired LUAD cancer tissues and adjacent normal tissues and western blotting for KEAP1 protein in 12 paired LUAD cancer tissues and adjacent normal tissues was performed to analysis the protein level of KEAP1. The results showed the protein level of KEAP1 was remarkably downregulated in LUAD cancer tissues, compared to the adjacent non-cancerous tissues (Fig. 5F and G). Pearson correlation analysis revealed a significant negative correlation between KEAP1 protein level and miR-141-3p. These results suggested miR-141-3p might regulat the protein level of KEAP1. To determine whether miR-141-3p could inhibit KEAP1 expression, BEAS-2B cells were transfected with mimics of miR-141-3p to increase the level of miR-141-3p, while A549 cells were transfected with inhibitors of miR-141-3p to downregulate the cellular levels of miR-141-3p. As anticipated, mimics of miR-141-3p could inhibit KEAP1 expression, while inhibitors increased KEAP1 levels (Fig. 5H). KEAP1 is confirmed to bind to NRF2 (nuclear factor erythroid 2-related factor 2) and promote its degradation by the ubiquitin proteasome pathway, which could in turn repress the HDAC4 (histone deacetylase 4) expression in LUAD 22.We found miR-141-3p mimics significantly increased the NRF2 and HDAC4 expression by repressing the KEAP1 expression, and miR-141-3p inhibitors remarkably suppressed the NRF2 and HDAC4 expression by reliving the suppression of miR-141-3p for KEAP1.
Actually, KEAP1 has been proved to be a target gene of miR-141-3p in LUAD by many previous studies 23, 24, 25, 26, 27, 28. Our results confirmed again that miR-141-3p can inhibit KEAP1 by binding to its 3’-UTR.
circKEAP1 relieved repression of miR-141-3p for KEAP1 expression
To confirm circKEAP1 could regulate the repression of miR-141-3p for KEAP1, the luciferase reporters containing KEAP1 3’UTR wild type sequence or muted sequence was co-transfected with circKEAP1 plasmid or miR-141-3p mimic in A549 cells. The results showed circKEAP1 significantly increased the luciferase activity of KEAP1 wild type reporter, while it had no effect on the mutated reporter. Moreover, the increase by the circKEAP1 could be abolished by miR-141-3p overexpression through miR-141-3p mimic (Fig. 6A). Subsequently, the protein level of KEAP1 was evaluated in cells with overexpressing circKEAP1, and resulting to decrease the NRF2 and HDAC4 expression, which was reported as the downstream genes of KEAP1 pathway 29, but it was rescued by miR-141-3p overexpression (Fig. 6B-C). Additionally, we also found the protein level of KEAP1 was significantly upregulated in the tumor of circKEAP1 overexpression group, which in turn downregulated the protein level of NRF2 and HDAC4 (Fig. S2B).
Several studies had been confirmed KEAP1 could suppress the expression of NRF2, and reduce its downstream gene HDAC4 expression, result to increase the tumor suppressor miRNAs miR-1/206 level to regulates glucose metabolism in cancer cells 29. In our study, we found circKEAP1 could significantly increase the expression of miR-1 and miR-206 both in vitro (Fig. 7A) and in vivo (Fig. S2C), however this increase could be attenuated by miR-141-3p overexpression. Additionally, in accordance with the results before, the circKEAP1 significantly repressed the cancer cells proliferation (Fig. 7B-D), and the suppression could be rescued by miR-141-3p overexpression using miR-141-3p mimic (Fig. 7B-D).
Taken together, our results found circKEAP1 could server as a sponge for miR-141-3p to regulate KEAP1 and activate the KEAP1/NRF2/HDAC4 signal pathway via the ceRNA mechanism to repress tumor growth (Fig. 8).