RETRACTED ARTICLE: The circular RNA 001971/miR-29c-3p axis modulates colorectal cancer growth, metastasis, and angiogenesis through VEGFA

The present study was approved by Xiangya Hospital, Central South University. Colorectal cancer (T) and matched non-cancerous tissues (N) were obtained from 70 CRC patients who underwent primary surgical resection at Xiangya Hospital with written consent and stored at − 80 °C until further use. All clinicopathologic features of the patients are listed in Tables 1 and 2.

The normal colon epithelial cell line FHC was purchased from ATCC (Manassas, VA, USA). Five CRC cell lines, HT29, HCT116, LoVo, SW480 and SW620, were characterized in detail and were provided by ATCC. Cell lines were cultured in Dulbecco’s modified Eagle medium (DMEM) supplemented with 10% fetal bovine serum (FBS) in a humidified incubator under 5% CO₂ at 37 °C.

The tissue samples were excised, washed with PBS, arrested in diastole with 10% potassium chloride solution, weighed, placed in 10% formalin, and embedded in paraffin. Several sections (4–5 mm thick) were prepared, stained with hematoxylin and eosin (H&E) for histopathology by standard procedures and then visualized by light microscopy. For IHC staining, the content of VEGFA was detected by IHC following the previously described method [36]. After deparaffinization, hydration and blockage of endogenous peroxidase, the sections were incubated for 20 min with 10% nonfat milk in PBS in order to block specific sites and then were incubated at 4 °C overnight with primary anti-VEGFA antibody (ab1316, abcam, USA). Then, sections were washed and incubated with HRP-conjugated goat anti-mouse secondary antibody (SV-0004, Boster, China) for 30 min at 37 °C. Finally, the sections were washed and incubated with a DAB kit (AR1022, Boster) for 10 min.

The MTT assay was used to detect the viability of CRC cells. Cells were cultured in 96-well plates (5 ×  10⁴ cells/well). Twenty-four hours after plating the cells were attached, and 10 μL MTT was added to the fresh medium to replace the old medium. Four hours after incubation with the medium containing MTT at 37 °C, the medium was discarded, and DMSO (100 μL) was added to each well after discarding the medium. After that, the absorbance was measured at 490 nm. The results are presented as percentage inhibition compared to the untreated control.

The ability to synthesize DNA was detected by 5-ethynyl-2′-deoxyuridine (EdU) assays using an EdU assay kit (Guangzhou RiboBio Co., Ltd.) according to the manufacturer’s instructions. Briefly, cells were incubated with EdU for 24 and 48 h in the dark. Then, the cells were fixed with 4% paraformaldehyde for 30 min and stained with Apollo 567 solution for 30 min at room temperature in the dark. The inflorescence absorbance value of each well was determined by a multifunction microplate reader (Bio-Rad, USA).

The human umbilical vein endothelial cell line HUVEC was purchased from ATCC and seeded in 24-well plates coated with Matrigel (90 μL), incubated at 37 °C for 30 min to polymerize, and then seeded in wells under different SW620 cell-conditioned media (CM) as described. The 24-well plates were incubated for 6 h, and tube formation was then imaged under an inverted microscope at a × 200 magnification.

Transwell assays were performed to detect the invasion capacity of target cells. Cells (5 × 10⁵) were plated on the upper chamber of a polycarbonate Transwell filter (Cell Biolabs, Inc. Santiago, USA) coated with Matrigel using serum-free medium. In the bottom chamber, serum-containing medium was used as a chemoattractant. After incubating at 37 °C for 48 h, the noninvasive cells in the top chambers were removed and the invaded cells on the lower membrane surface were fixed with 100% methanol, stained with DAPI (Beyotime Institute of Biotechnology, Haimen, China),and counted under a microscope.

For miRNA, circRNA and mRNA expression determination, we extracted total RNA using TRIzol reagent (Invitrogen) following the manufacturer’s instructions. For miRNA expression determination, reverse transcription was conducted using miRNA-specific primers and a miScript Reverse Transcription kit (Qiagen, Germany) taking RNU6B expression as an endogenous control. For mRNA expression determination, SYBR green PCR Master Mix (Qiagen) was used, using GAPDH expression as an endogenous control. For circRNA expression determination, the total RNA was digested by Ribonuclease R (Epicentre, USA), and then the RNA was reverse transcribed and subjected to RT-PCR using SYBR green PCR Master Mix (Qiagen). RT-PCR assays with divergent or convergent primers were used for circRNA identification (Fig. S1). The relative fold-change was calculated by the 2^−ΔΔCT method. The primer sequences are listed in Table S1.

Six BALB/c nude male mice (4 weeks old) were used in the experiments. All animals were randomly assigned into two groups. SW620 cells, infected with Lv-sh-NC or Lv-sh-circ-001971 (GeneChem, China) were suspended in 200 μL growth medium/Matrigel and hypodermically injected into the left axillaries of mice in the different groups. Twenty-five days after the injection, mice were sacrificed under anesthesia. The length (L) and width (W) of the tumors were measured and the tumor volumes (V) were calculated following the formula V  =  L  ×  W²/2 every 5 days for 25 days. At the end of the animal experiments, the tumor weights in the two groups were determined.

RIPA buffer (Sigma-Aldrich, St. Louis, MO, USA) was used to lyse the target cells and a complete protease inhibitor cocktail (Roche, USA) was used for total protein collection. Then, total proteins were separated using SDS-PAGE and transferred onto PVDF membranes. The antibodies listed below were used to incubate the blots at 4 °C overnight: anti-VEGFA (ab1316) and anti-GAPDH (ab8245, Abcam). The membranes were then incubated with an HRP-conjugated secondary antibody (1:5000). Signals were visualized using ECL (enhanced chemiluminescence) substrates (Millipore, USA) using endogenous GAPDH levels for normalization.

For miR-29c-3p binding to circ-001971 or the 3’UTR of VEGFA, the fragment of circ-001971 or the 3’UTR of VEGFA was amplified by PCR and cloned downstream of the Renilla gene in the psiCHECK-2 vector (Promega, Madison, WI, USA), which was named wt-circ-001971 or wt-VEGFA 3’UTR. To generate the circ-001971, VEGFA 3’UTR mutant reporters, the seed region of the circ-001971 and the VEGFA 3’UTR were mutated to remove the complementarity to miR-29c-3p, and the resulting constructs were named mut-circ-001971 and mut-VEGFA 3’UTR. HEK293 cells (ATCC, USA) were seeded into a 24-well plate. Then, HEK293 cells were cultured overnight and cotransfected with a luciferase reporter vector and miR-29c-3p mimics/inhibitor. After 48 h, the alterations in the luciferase activity were evaluated by the Dual-Luciferase Reporter Assay System (Promega) using firefly luciferase activity for normalization.

RIP assay was conducted to validate the predicted binding. A Magna RIP RNA-Binding Protein Immunoprecipitation Kit (17–700; Millipore, Burlington, MA, USA) according to the product’s instructions. The immunoprecipitated RNAs were reverse transcribed and subjected to RT-PCR using the PrimeScript™ RT reagent Kit (Takara, Japan) and SYBR Green PCR Master Mix (Qiagen). miR-29c-3p, circ-001971, and VEGFA levels in the immunoprecipitates were detected using specific primers.

Each experiment was repeated at least three times. Data were processed by SPSS 17.0 and presented as the mean ± S.D. Differences between two groups were compared using Student’s t-test; differences among more than two groups were compared using one-way ANOVA following Turkey’s test. Pearson Correlation analysis was used to determine the relationships among miR-29c, circ-001971 and VEGFA. Kaplan-Meier analysis of overall survival by the log-rank test was used for overall survival analysis. Receiver operating characteristic (ROC) curve analysis was performed to evaluate the diagnostic and prognostic sensitivity and specificity of circ-001971 for CRC. *P < 0.05; **P < 0.01.

Tissue samples were collected and verified by H&E staining (Fig. 1a). To identify circRNAs that might be related to CRC carcinogenesis, the top 11 upregulated circRNAs previously reported to be highly expressed in CRC tumor tissue samples compared with normal mucosa tissue samples (Table S2) [37], were selected for real-time PCR confirmation. Figure 1b shows that in 13 paired tumor and noncancerous tissue specimens, circ-002739, circ-001971 and circ-004028 were significantly upregulated in tumor tissues, and circ-001971 was the most upregulated (Fig. 1b). Thus, circ-001971 was selected for further experiments.

In 70 paired tissue specimens, circ-001971 expression was significantly upregulated, consistent with the previous small sample size examination (Fig. 1c). Next, we analyzed circ-001971 expression according to the TNM stage and found that circ-001971 expression was higher in specimens at advanced TNM stages (III + IV) than in specimens at early stages (I + II; Fig. 1d). To further validate the effect of circ-001971 on CRC, the association between the expression of circ-001971 and the clinical indicators of CRC patients was detected. Seventy cases were assigned into two groups, high and low circ-001971 expression groups, by the median value of circ-001971 expression. Table 1 shows that high circ-001971 expression was dramatically related to advanced TNM stages and lymphatic metastasis. Moreover, we employed a Cox proportional hazards regression model to detect the overall survival rate and pathologic features of the 70 CRC patients. Univariate analysis revealed that the TNM stage, lymph node metastasis and circ-001971 expression caused remarkable differences in the overall survival rate. Mutivariable analysis demonstrated that the TNM stage and circ 001971 expression led to the differences in overall survival rate with statistical significance (Table 2). To determine circ-001971 as a biomarker of CRC, we performed ROC curve analysis to evaluate the diagnostic (Fig. 1e) and prognostic (Fig. 1g) sensitivity and specificity of circ-001971 for CRC. The area under the curve was 0.788 (Fig. 1e) or 0.792 (Fig. 1g), which shows that circ-001971 has a potential diagnostic and prognostic capability (Fig. 1e and g). Kaplan-Meier curves of overall survival by log-rank test demonstrated that the overall survival rate of patients in the low circ-001971 expression group was higher (Fig. 1f), suggesting the underlying effect of circ-001971 on CRC.

To investigate the specific functions of circ-001971 in CRC carcinogenesis, we examined its expression in five CRC cell lines (HT29, HCT116, LoVo, SW480, and SW620) and a normal cell line, FHC. Consistent with its expression in tissues, the expression of circ-001971 was significantly increased within CRC cells compared with FHC cell lines (Fig. 2a) and was more highly increased in HCT116 and SW620 cells (Fig. 2a). We transfected si-circ-001971#1 and si-circ-001971#2 to conduct circ-001971 knockdown in HCT116 and SW620 cells, and performed real-time PCR to verify the transfection efficiency. Then, we selected si-circ-001971#1 as the siRNA targeting circ-001971 because of its greater efficiency (Fig. 2b).

Next, we transfected HCT116 and SW620 cell lines with si-circ-001971 and determined the proliferation and invasion of cells. After knocking down circ-001971, the DNA synthesis capacity (Fig. 2c), cell viability (Fig. 2d-e), and cell invasion (Fig. 2f-g) were all significantly inhibited. To further investigate the effects of circ-001971 on HUVEC tube formation through the tumor microenvironment, we collected conditioned medium from circ-001971-knockdown or control SW620 cells (si-circ-001971 CM and si-NC CM), cultured HUVECs in these two CM, respectively, and examined HUVEC angiogenesis in different CM. Figure 2h-i shows that culture with si-circ-001971 CM significantly suppressed HUVEC tube formation compared with culture in si-NC CM.

To further confirm the tumor suppressor roles of circ-001971 knockdown, we constructed lentivirus containing shRNA targeting circ-001971 (Lv-sh-circ-001971) and performed real-time PCR to verify the knockdown efficiency (Fig. 3a). Next, six animals were randomly assigned into two groups (n = 3), and SW620 cells infected with Lv-sh-NC (negative control) or Lv-sh-circ-001971, were hypodermically injected into the left axillaries of the mice in these two groups (Fig. 3b). As shown in Fig. 3c and d, circ-001971 knockdown significantly reduced the tumor volumes at every measurement time point and reduced the tumor weight determined at the end of the experiment. These data indicate that circ-001971 knockdown inhibits CRC tumor growth in vivo.

CircRNAs act as miRNA sponges to offset miRNA-induced inhibition of downstream target genes [38]. To investigate the molecular mechanism of circ-001971 functioning in CRC growth and metastasis, we screened for miRNAs that may be related to CRC metastasis and may target VEGFA, a key regulator of tumor angiogenesis [39]. According to the online microarray expression profile GSE126093, a total of 440 miRNAs were upregulated and 74 were down-regulated in CRC tissue samples than that within normal control tissue samples (p < 0.05, │logFC│ > 0.6) (Fig. 4a). In circ-001971 knocked-down HCT116 and SW620 cells, the protein levels of VEGFA were also significantly decreased (Fig. 4b-c), suggesting the potential crosstalk between circ-001971 and VEGFA. Next, we employed TargetScan to predict miRNAs that might target VEGFA. Among the 627 predicted miRNAs, 16 miRNAs were down-regulated in CRC reported by GSE126093. Among the 16 miRNAs, 3 were predicted to target circ-001971, namely, miR-29c-3p, miR-497-5p, and miR-943 (Fig. 4d).

To further confirm the miRNA that might mediate the crosstalk between circ-001971 and VEGFA, we transfected miRNA mimics to overexpress the three candidate miRNAs in HCT116 and SW620 cells, and performed real-time PCR to verify the transfection efficiency (Fig. 4e). Next, we transfected HCT116 and SW620 cell lines with miRNA mimics and examined the expression of circ-001971 and VEGFA. As shown in Fig. 4f and g, all three miRNAs inhibited the expression of VEGFA within both CRC cells, however, only miR-29-3p strongly inhibited circ-001971 expression within both CRC cell lines. Thus, we selected miR-29c-3p for further experiments.

We transfected miR-29c-3p inhibitor to conduct miR-29c-3p inhibition within HCT116 and SW620 cell lines, as confirmed by real-time PCR (Fig. 4h). In miR-29c-3p inhibited HCT116 and SW620 cells, circ-001971 expression was significantly upregulated (Fig. 4i); in circ-001971 knockdown HCT116 and SW620 cell lines, miR-29c-3p expression was dramatically upregulated (Fig. 4j). Moreover, within HCT116 and SW620 cell lines, miR-29c-3p overexpression dramatically decreased, while miR-29c-3p inhibition increased VEGFA protein levels (Fig. 4k-l).

To investigate the putative miR-29c-3p binding to circ-001971 and VEGFA, we performed luciferase reporter and RIP assays. We constructed two kinds of reporter vectors, wild-type and mutant, which named wt-circ-001971/mut-circ-001971 and wt-VEGFA 3’UTR/mut-VEGFA 3’UTR, respectively. In the mutant reporter vectors, 4 bases were mutated in the predicted miR-29c-3p binding site of (Fig. 5a and c). Next, we cotransfected HEK293 cells with these vectors and miR-29c-3p mimics/inhibitor, and determined the luciferase activity. Wild-type vector (wt-circ-001971 and wt-VEGFA 3’UTR) luciferase activity was dramatically reduced via the overexpression of miR-29c-3p, and it was increased via the inhibition of miR-29c-3p. Once the putative binding site was mutated, the luciferase activity was reduced to almost the original value (Fig. 5b and d). To further confirm the binding, RIP assays were performed. Based on the results, circ-001971, VEGFA and miR-29c-3p were associated with AGO2; in RNA extracted from the protein precipitate, circ-001971, VEGFA and miR-29c-3p levels were twice as high as those in the IgG sample (Fig. 5e). Moreover, endogenous circ-001971 and VEGFA pull-down by AGO2 was enriched after miR-29c-3p overexpression (Fig. 5f). These data confirmed that circ-001971 and VEGFAbind to miR-29c-3p in a AGO2-associated manner.

After confirming circ-001971 binding to miR-29c-3p, the study continued to investigate the dynamic effects of circ-001971 and miR-29c-3p on HUVEC tube formation. First, HCT116 and SW620 cells were cotransfected with miR-29c-3p inhibitor and si-circ-001971, and the protein levels of VEGFA were examed. As shown in Fig. 6a-b, in both cell lines, miR-29c-3p inhibition increased, while circ-001971 knockdown decreased the protein levels of VEGFA, and the effects of circ-001971 knockdown on VEGFA protein levels were significantly reversed by miR-29c-3p inhibition. Then, different CM samples were collected for HUVEC culture. As shown in Fig. 6c-d, miR-29c-3p inhibition significantly enhanced, while circ-001971 knockdown inhibited the tube formation ability of HUVECs and the effects of circ-001971 knockdown on HUVEC tube formation were significantly reversed by miR-29c-3p inhibition. Overall, these data indicate that the circ-001971/miR-29c-3p axis could modulate VEGFA protein levels and HUVEC tube formation.

As a further confirmation, we performed real-time PCR to examine miR-29c-3p and VEGFA expression in tissue samples. Figure 7a-b shows that the expression of miR-29c-3p was obviously inhibited while the expression of VEGFA was enhanced within tumor tissues compared with non-cancerous tissue samples. Consistent results were revealed obtained by IHC and immunoblotting assays that VEGFA protein contents within tumor tissue samples showed to be increased as compared to those within non-cancerous tissue samples (Fig. 7c-d). In addition, miR-29c-3p was negatively related to circ-001971 and VEGFA (Fig. 7e-f); circ-001971 was positively related to VEGFA (Fig. 7g).