A novel protein encoded by a circular RNA circPPP1R12A promotes tumor pathogenesis and metastasis of colon cancer via Hippo-YAP signaling

A total of 20 paired human CC and adjacent non-tumor tissues were collected between 2014 and 2017 at the Third Affiliated Hospital of Soochow University. This study was approved by the Ethical Committee of the Third Affiliated Hospital of Soochow University, and the written informed consent was provided by each participant prior to surgery.

The circRNA microarray (Arraystar Human circRNAs chip, ArrayStar) consisting of more than 5000 probes specific to splicing sites of human circRNAs was used. After hybridization, 20 pairs of CC samples (tumor tissues and paired adjacent non-tumor tissues) were examined using the circRNA micro-array provided by Kangcheng Bio-Tech Inc.

Human CC cell lines HT-29, HCT-116, SW480, SW620, LoVo, SW48, DLD-1, Caco2 and HCT-15) and normal human colon mucosal epithelial cell line NCM460 were obtained from Chinese Academy of Sciences, Shanghai Institutes for Biological Sciences. All these cell lines were maintained in DMEM supplemented with 10% FBS under standard culture conditions (5% CO₂, 37 °C).

RNA extraction and quantitative real-time PCR were performed as previously described [19]. Genomic DNA (gDNA) was extracted from tissues using PureLink Genomic DNA Mini Kit according to the manufacturer’s instructions (Thermo Fisher Scientific, K182001). The real-time PCR was conducted using the ABI Vii7 system (Applied Biosystems, USA), and SYBR Green was used as a DNA-specific fluorescent dye. Human GAPDH was selected as a housekeeping gene. Primers were synthesized as follows, the human circPPP1R12A forward primer: 5′-ACAGCAGCAGGCTAGAAAAG-3′, and reverse primer: 5′-TGTCCTAAGCAGGAAAAACA-3′; GAPDH forward: 5′-TGACTTCAACAGCGACACCCA-3′, and GAPDH reverse: 5′-CACCCTGTTGCTGTAGCCAAA-3′. Relative gene expression was calculated by the comparative CT method (ΔΔCT), where the fold enrichment was determined as: 2^{-[ΔCT (sample) − ΔCT (calibrator)]}.

ISH and FISH were carried out to detect the circPPP1R12A expression in colon tissue microarray (Shanghai Outdo Biotech, Shanghai, China) using digoxigenin- or FITC-labeled probe as previously reported [20]. After stained with an anti-digoxin or anti-FITC mAb (Roche Applied Science), the sections were incubated in the presence or absence of NBT/BCIP in the dark, mounted and examined under microscope.

Human circPPP1R12A over-expressing vector (Lv-circPPP1R12A) and the control plasmid (pLVX-IRES-GFP) were purchased from Genelily Biotechnology Company (Shanghai, China). The expression of circPPP1R12A was transiently silenced by small interference RNAs (siRNAs) specific to human circPPP1R12A, which were generated by GenePharma (GenePharma Corporation, Shanghai, China). The experiment was divided into several groups as follows: si-NC (transfected with siRNA NC vector) group, si-circPPP1R12A (transfected with circPPP1R12A siRNA) group, OE-NC (transfected with over-expression NC vector) group and OE-circPPP1R12A (transfected with circPPP1R12A over-expression vector) group. The above-mentioned vectors were transfected by Lipofectamine® 3000 transfection reagent as previously reported. After transfection for 48 h, the transfection efficiency of cells in each group was assessed by real-time PCR analysis.

Cell growth was monitored by CCK8 assay as previously reported [21]. Colony formation assays were performed to indicate the cloning capability of CC cells. For cell cycle assay, subG1, S and G2 peaks were detected from the propidium iodide-stained CC cells by flow cytometry (BD FACS Canto II, BD Bioscience, NJ, USA) and analyzed using the Modfit software.

The cell migration ability was assessed using wound-healing assay as previously reported. The invasion assay was analyzed by transwell chamber with 8-μm pores (BD Falcon, Franklin Lakes, NJ, USA).

Western blotting analysis was performed as previously described [19] with antibodies against flag and GAPDH (Abcam, Cambridge, MA, USA). GAPDH was used as an endogenous control to normalize the protein loading.

The protein identification by LC-MS/MS was performed according to previously reported method [22]. Briefly, equal amounts of proteins (50 μg) were subjected to 12% SDS-PAGE. The protein bands near 10 kDa were excised according to size and chopped into 1 mm³ pieces. Finally, peptide mixtures were extracted from the gels and dried prior to LC-MS/MS analysis.

Firstly, circRNA microarray was employed to characterize the expression profiles of circRNAs in paired CC tissues and adjacent non-tumor tissues from 10 patients. A total of 126 circRNAs (P < 0.05 and fold change > 1.5) were differentially expressed between the CC tissues and paired adjacent non-tumor tissues. Among the 126 differentially expressed circRNAs, 110 circRNAs were up-regulated, while 16 ones were down-regulated in CC tissues compared with the adjacent non-tumor tissues (Fig. 1 a). Additional file 1: Table S1 lists the detailed information about these dysregulated circRNAs. These circRNAs were mostly located at exonic regions (Fig. 1 b). As the most up-regulated circRNA, hsa_circ_0000423 (termed as circPPP1R12A) was back-spliced of exons 24/25 of PPP1R12A gene located at 12q21.2 (Fig. 1 c). Next, we re-examined the expression of circPPP1R12A in CC and paired non-tumor tissue samples from 20 patients by quantitative real-time PCR to confirm its elevated expression (Fig. 1 d). We further found that the circPPP1R12A expression was consistently and significantly increased in CC tissues compared with the matched controls, while the expression of PPP1R12A (linear transcript of PPP1R12A gene) was comparable in CC tissues and matched controls (Additional file 2: Figure S1a). Moreover, the expression of circPPP1R12A was significantly up-regulated in a series of cultured CC cell lines (HT-29, HCT-116, SW480, SW620, LoVo, SW48, DLD-1, Caco2 and HCT-15) compared with a normal human colon mucosal epithelial cell line NCM460 cells. The highest expression of circPPP1R12A was found in HCT-116 cells, followed by LoVo cells (Fig. 1 e). Therefore, our subsequent experiments focused on the role of circPPP1R12A in CC progression.

In the present study, we designed two sets of primers to characterize circPPP1R12A. One pair (divergent primers) was used to amplify the circular transcripts, while the other pair (convergent primers) was used to detect the linear transcripts. The results suggested that the circular form could be amplified using the convergent primers from both cDNA and gDNA, while it was only amplified from cDNA by divergent primers (Fig. 2 a). To further confirm the existence of circPPP1R12A, the RNase R degradation assay was used to evaluate the resistance of circPPP1R12A to RNase R treatment. Figure 2 b shows that the linear transcripts of PPP1R12A were degraded by RNase R treatment, while such treatment failed to degrade the circular transcripts of circPPP1R12A. Nuclear mass separation assay (Fig. 2 c) and FISH analysis (Fig. 2 d) reveled that over 93% of circPPP1R12A appeared in the cytoplasm of HCT-116 and LoVo cells. We also detected the expression of circPPP1R12A in CC tissues by ISH using TMA consisting of 100 pairs of CC and adjacent non-tumor tissues (Fig. 2e). Table 1 lists the detailed clinical parameters of these patients. Among the clinicopathological variables, pathological stage and circPPP1R12A ISH score were identified as risk factors for predicting overall survival based on univariate analysis, while multivariate analysis with Cox regression model further confirmed that pathological stage III and circPPP1R12A ISH score 3–4 were the independent poor prognostic factors (Table 2). Kaplan–Meier survival curves showed that patients with higher expression of circPPP1R12A had a shorter overall survival [HR = 1.886; 95% confidence interval (CI), 1.129–3.1529; P = 0.0154; Fig. 2 f].

As unlimited cell proliferation, migration and invasiveness are hallmarks of malignant tumor, we next explored the role of circPPP1R12A in progression of CC by gain- and loss-of-function approaches. We overexpressed circPPP1R12A by artificial plasmids in two (DLD-1 and Caco-2) CC cell lines with low circPPP1R12A expression, and the overexpression efficiency was confirmed by real-time PCR (Fig. 3 a). CCK8 assay (Fig. 3 b) and colony formation assay (Fig. 3 c) indicated that circPPP1R12A overexpression accelerated the cell proliferation of DLD-1 and Caco-2 cells. Subsequently, we explored the effect of circPPP1R12A on motility of CC cells. Wound-healing assay and invasion assay showed that circPPP1R12A overexpression also enhanced the migration (Fig. 3 d) and invasion (Fig. 3 e) abilities of DLD-1 and Caco-2 cells. Moreover, we overexpressed circPPP1R12A by artificial plasmids in two (HCT-116 and LoVo) CC cell lines with high circPPP1R12A expression. Similarly, circPPP1R12A overexpression accelerated the cell proliferation of HCT-116 and LoVo cells, evidenced by CCK8 assay (Additional file 3: Figure S2a) and colony formation assay (Additional file 3: Figure S2b). However, circPPP1R12A overexpression failed to alert the migration (Additional file 3: Figure S2c) and invasion (Additional file 3: Figure S2d) abilities of DLD and Caco-2 cells in wound-healing assay and invasion assay.

We next investigated whether silencing of circPPP1R12A affected cell proliferation, migration and invasiveness. The silencing efficiency was confirmed by real-time PCR (Fig. 4 a). As expected, silencing of circPPP1R12A affected the proliferation (Fig. 4 b) and colony forming abilities (Fig. 4 c) of both cell lines. Furthermore, in wound-healing assay and invasion assay, silencing of circPPP1R12A significantly affected the migration (Fig. 4 d) and invasion (Fig. 4 e) abilities of HCT-116 and LoVo cells. Collectively, these results suggested that circPPP1R12A played a critical role in proliferation, migration and invasion of CC cells.

As circRNA circPPP1R12A is originally annotated in circRNA Db database, we found a short 216-nt small ORF with the potential to encode a conserved 73-aa peptide (Fig. 5 a). Especially, in human circPPP1R12A, the circularization created the tandem start codon ‘AUG’, which started the translation in combination with overlapping genetic codes. The ORF of circPPP1R12A was supposed to be translated into a putative 73-aa protein (we termed ‘circPPP1R12A-73aa’), in which the unique aa sequences were formed (Fig. 5 a). To dissociate the role of circPPP1R12A-73aa from circPPP1R12A, we constructed several flag-labeled vectors for circPPP1R12A (Fig. 5 b). Transfection with the Lv-flag-circPPP1R12A vector and Lv-flag-circPPP1R12A-Mut vector both successfully resulted in the overexpression of circPPP1R12A, while transfection with the control vector or Lv-circPPP1R12A-73aa did not (Fig. 5 c). Moreover, the level of flag-labeled circPPP1R12A-73aa protein was elevated after Lv-OE-flag-circPPP1R12A-73aa or Lv-circPPP1R12A-73aa transfection (Fig. 5 d). We further confirmed that circPPP1R12A-73aa was translated from circPPP1R12A by LC-MS/MS, and the unique aa sequence of circPPP1R12A-73aa was identified. More than 76% of the circPPP1R12A-73aa sequences were identified in the LC-MS/MS results from the 10-kDa band (Fig. 5 e). In addition, one unique aa sequence formed by the circPPP1R12A “GRLRHVNCLSPGVQD” was identified (Fig. 5 f). Taken together, circPPP1R12A, which was annotated as a circRNA, actually encoded an uncharacterized protein.

To further explore the biological function of circPPP1R12A-73aa, we generated several cell lines with stable transfection of those above-mentioned four vectors (Fig. 5 b). Similar with the previous results (Fig. 3 b), overexpression of wild-type circPPP1R12A with flag labeling resulted in increased proliferation ability, evidenced by CCK8 (Fig. 6 a), colony formation (Fig. 6 b) and cell cycle (Fig. 6 c) assays in DLD-1 and Caco-2 cells. Indeed, only overexpression of circPPP1R12A-73aa increased the proliferation ability too. However, overexpression of mutant circPPP1R12A with flag labeling, which was unable to translate circPPP1R12A-73aa, failed to increase the proliferation ability of DLD-1 and Caco-2 cells. For the motility of CC cells, overexpression of circPPP1R12A-73aa increased the migration (Fig. 6 d) and invasion (Fig. 6 e) abilities too, while overexpression of mutant circPPP1R12A failed to increase the migration and invasion abilities of DLD-1 and Caco-2 cells. These data suggested that circPPP1R12A did not regulate the cell proliferation of CC cells without its encoding protein circPPP1R12A-73aa.

Next, we further investigated the role of circPPP1R12A-73aa in growth of CC cells in vivo using a xenograft tumor model. Figure 7 shows that the tumors generated from cells transfected with Lv-flag-circPPP1R12A and Lv-circPPP1R12A-73aa had a bigger size (Fig. 7 a) and larger weight (Fig. 7 b) compared with those generated from vector control cells. Similar with the in vitro data, the tumors generated from cells transfected with Lv-flag-circPPP1R12A-mut had a smaller size (Fig. 7 a) and lower weight (Fig. 7 c). As the liver metastasis of CC is the most common metastasis in vivo, we then assessed the effects of circPPP1R12A-73aa on tumor liver metastasis by injecting four stably transfected DLD-1 cells (Fig. 5 b) into the tail veins of nude mice. Subsequently, in vivo bioluminescent imaging (BLI) of mice showed that overexpression of circPPP1R12A-WT and circPPP1R12A-73aa significantly changed the detectable metastasis (red arrow indicates metastasis) in the liver compared with the controls and the circPPP1R12A-Mut group (Fig. 7 d and e).

Collectively, these data showed that circPPP1R12A-73aa encoded by circPPP1R12A, not circPPP1R12A, promoted the cell proliferation and metastasis of CC cells.

RNA-Seq analyses were carried out to identify the possible signaling pathways, by which circPPP1R12A-73aa promoted the proliferation, migration and invasion. RNA from the four stably transfected DLD cells (Fig. 5 b) was isolated and analyzed. Differentially expressed genes in the four groups were analyzed and enrichened in KEGG pathway database. To elucidate the regulatory mechanism of the signaling influenced by circPPP1R12A-73aa, we focused on the sub-category of signal transduction pathway from the KEGG pathway database. The pathways in this category were ranked, and Fig. 8 a lists the top 10 pathways. As the most enriched pathway, the Hippo-YAP signaling pathway was further proved to be affected by circPPP1R12A-73aa. We next conducted Western blotting analysis to evaluate the dysregulated genes in Hippo-YAP signaling (Fig. 8 b). YAP1 is a transcriptional co-activator in the Hippo signaling pathway, and YAP1-induced transcriptional responses are essential in proliferation and metastasis of cancer cells. To illustrate whether YAP1 activation was critical for the circPPP1R12A-73aa-induced growth and metastasis of cancer cells, we then treated the circPPP1R12A-73aa-overexpressing DLD-1 and Caco-2 CC cells with YAP1 specific inhibitor Peptide 17. The results revealed that the Peptide 17 significantly alleviated the promotive effect of circPPP1R12A-73aa overexpression on the proliferation (Fig. 8 c), migration (Fig. 8 d) and invasion (Fig. 8 e) abilities of CC cells. Therefore, we confirmed that circPPP1R12A-73aa promoted the proliferation, migration and invasion abilities of CC via activating Hippo-YAP signaling pathway (Fig. 9).