ALKBH5/YTHDF2‐mediated m6A modification of circAFF2 enhances radiosensitivity of colorectal cancer by inhibiting Cullin neddylation

The human tissue samples used in the current study were collected from the pre‐treatment colonoscopy specimens of patients with neoadjuvant chemoradiotherapy for rectal cancer in the Third Affiliated Hospital of Soochow University. Samples were stored in liquid nitrogen following collection. The rectal cancer specimens were collected from patients without any anti‐tumour therapies, including immunotherapy, targeted therapy, biological therapy and surgery. Tumour regression grading (TRG) was used for the pathological analysis of tumour specimens after neoadjuvant chemoradiotherapy to evaluate the therapeutic effect.30, 31 The TRG score was based on the National Comprehensive Cancer Network criteria.31 TRG grade 0 represents absence of residual cancer cells, that is, pCR. TRG grade 1 represents only a single cancer cell or cancer cell cluster and TRG grade 2 is characterised by fibrosis overgrowing cancer cells. TRG grade 3 indicates almost no fibrosis, and a large amount of residual cancer can be seen. Two independent pathologists generated the pathological diagnosis report of the specimen. We informed patients during specimen collection and pre‐treatment interviews, and informed consent was signed. The Ethics Committee of the Third Affiliated Hospital of Soochow University approved this study in accordance with the Declaration of Helsinki (Approval No. 190 [2018]).

Cell lines used in the study, including RKO, SW480, HT‐29, LoVo, HCT‐116, SW620 and NCM460, were obtained from the American Type Culture Collection. The human colorectal cancer cell line was cultured in Dulbecco's modified Eagle's medium (Gibco), supplemented with 10% foetal bovine serum (Gibco), 100 units/mL penicillin and 100 μg/mL streptomycin (Gibco). The cells were cultured in 5% CO₂ at 37°C. The medium was changed every 3 days. When the cells reached 80%−90% confluency, they were passaged with .05% Trypsin + .53 mM Ethylene Diamine Tetraacetic Acid (EDTA) (Gibco).

Plasmids overexpressing circRNAs, short hairpin RNA (shRNA) of circRNAs, and negative control plasmids were purchased from Generbiol (Shanghai, China). Sequences of siRNAs and shRNAs used in the plasmid construction are shown in Table. S1

Details about in situ hybridisation (ISH), immunohistochemistry (IHC), RNA preparation, real‐time Quantitative Polymerase Chain Reaction (qPCR), immunoblotting, colony formation assay, apoptosis detection, γ‐H2AX quantification, RNase R resistance assay, nuclear‐cytoplasmic fractionation, fluorescence in situ hybridisation (FISH), RNA m6A methylation quantification, dot blot, RNA immunoprecipitation (RIP), Me‐RIP‐qPCR, luciferase reporter assay, co‐immunoprecipitation (co‐IP) assay, RNA pull‐down analysis and mass spectrometry and animal experiments are described in Supporting Information. The primers are shown in Table. S2 and S3

The mean ± SD of data from three independent experiments was presented, and the difference between two groups was analysed using Student's t‐test. The survival outcome of patients with different levels of circAFF2 (high/low groups) was evaluated by Kaplan–Meier analysis (log‐rank test). Statistically significant differences were defined as those with p < .05. Otherwise, there was no statistical difference. All statistical analyses were conducted using GraphPad Prism 9.0 with bilateral statistical tests.

The screening process of m6A‐modified circRNAs in rectal cancer radiosensitivity is shown in Figure 1A. We collected tissue specimens from 16 locally advanced rectal cancer cases before neoadjuvant chemoradiotherapy. The treatment plan for these patients was a combination of neoadjuvant chemoradiotherapy and surgical removal. Of these, eight cases with postoperative pathology of TRG 0 and TRG 1 were selected as the radiosensitive group, while eight cases of TRG 2 and TRG 3 were selected as the radioresistant group. Using a genome‐wide investigation, 14 213 new circRNAs with a total length of 35 039 505 nt and an average length of 2465.31 nt were detected. We identified 100 differentially expressed circRNAs between radiosensitive and radioresistance groups (fold change > 2, p < .05, Figure 1B,C). We further excluded circRNAs that were not annotated in circBase (http://www.circbase.org/↗) and those that were only expressed in one or two specimens to obtain 16 circRNAs for further study. Of these, nine were highly expressed in the radiosensitive group (hsa_circ_0000690, hsa_circ_0005736, hsa_circ_0000734, hsa_circ_0001947, hsa_circ_0003322, hsa_circ_0000023, hsa_circ_0005310, hsa_circ_0001277 and hsa_circ_0000228) and seven had low expression levels in the radiosensitive group (hsa_circ_0003448, hsa_circ_0000271, hsa_circ_0042231, hsa_circ_0008967, hsa_circ_0004568, hsa_circ_0008498 and hsa_circ_0008518). To investigate whether these circRNAs have m6A methylation modification, we performed a Me‐RIP assay in HCT‐116 CRC cells and found that hsa_circ_0000023, hsa_circ_0001947 and hsa_circ_0000734 were m6A methylated (Figure 1D). To further investigate the association of these three circRNAs with the radiosensitivity of CRC, CRC cells with downregulated hsa_circ_0000023, hsa_circ_0001947 or hsa_circ_0000734 were constructed. The expression levels of circRNAs in the transfected HCT‐116 and HT‐29 are shown in Figure S1A. Colony formation assays revealed that only hsa_circ_0001947 was associated with the radiosensitivity of CRC cells (Figure 1E). Therefore, hsa_circ_0001947 was selected to further investigate the association between circRNAs and the radiosensitivity of CRC.

The circBase database revealed that hsa_circ_0001947 is located on chromosome X (147 743 428–147 744 289) and formed by the circularisation of the single exon 3 of the AFF2 gene with a spliced length of 861 bp (hereinafter referred to as circAFF2). The circular structure of circAFF2 was further verified by Sanger sequencing (Figure 2A). A study by Zhou et al.19 suggested that long single‐exon circRNAs are more susceptible to m6A methylation. Both the cDNA from the reverse transcription of circAFF2 and the directly extracted gDNA were amplified by PCR with the convergent primers. However, only cDNA from reverse transcription of circAFF2 can amplify the circRNA fragment with the divergent primer, which indicated that circAFF2 is a circular structure (Figure 2B). Furthermore, RNase R treatment of the HT‐29 and HCT‐116 cells revealed that circAFF2 was more stable than linear RNA (Figure 2C). Nucleocytoplasmic separation and FISH assays demonstrated that circAFF2 was primarily located in the cytoplasm of CRC cells (Figure 2D,E). Together, our findings suggest that circAFF2 is a circular structure and is stably expressed in the cytoplasm of CRC cells.

Our data revealed that among RKO, SW480, HT‐29, LoVo, HCT‐116, SW620 and NCM460 cell lines, HCT‐116 is the most radiosensitive, while HT‐29 is the least radiosensitive (Figure S2). Therefore, HCT‐116 and HT‐29 were used in the following in vitro experiments. We observed that the expression levels of circAFF2 in HCT‐116 were higher than in HT‐29 (Figure S1B). As demonstrated in our colony formation assay, downregulation of circAFF2 significantly increased the cell survival fraction (Figure 1E). γ‐H2AX is an indicator of cellular response to DNA damage, in cells 6 h after 4 Gy irradiation. We used western blotting to detect the expression of γ‐H2AX in HCT‐116 and HT‐29 cells at different time points (0, 1, 2, 4, 6, 8 and 20 h) after being exposed to 4 Gy radiation. The expression of γ‐H2AX peaked at 1 h and then rapidly decreased afterwards. We found that the expression of γ‐H2AX in HCT‐116 and HT‐29 cell lines overexpressing circAFF2 was significantly different from that in the control group at 4−20 h after irradiation (Figure S3). This demonstrates that it is feasible to detect the expression of γ‐H2AX between 4 and 20 h after radiation therapy. Combined with the work and rest arrangement of experimental personnel, as well as the convenience and rationality of the operation process from irradiation room to laboratory, we chose to detect γ‐H2AX at 6 h post‐irradiation. We found that γ‐H2AX levels were dramatically upregulated in cells overexpressing circAFF2, while γ‐H2AX expression was significantly decreased in circAFF2‐depleted cells (Figure 3A,B). Consistently, overexpression of circAFF2 in CRC cells significantly increased cell apoptosis, while cell apoptosis was attenuated in circAFF2‐depleted CRC cells (Figure 3C), suggesting that circAFF2 could increase the radiosensitivity of CRC cells.

Next, we investigated the effect of circAFF2 on the radiosensitivity of CRC in vivo. To achieve this, we utilised a subcutaneous tumour grafting mouse model by injecting CRC HCT‐116 cells transfected with Lv‐NC, Lv‐OE‐circAFF2, Lv‐Scramble or Lv‐sh‐circAFF2. Ten days after tumour grafting (tumour volume reaches 200 mm³), mice were exposed to 10 Gy radiation (single dose). Tumour sizes were recorded every 5 days following radiation. Mice were anaesthetised and tissues were collected 30 days after irradiation. Representative images of collected tumours are presented in Figure 3D. The circAFF2‐overexpressed tumour weight was significantly reduced, while the weight of circAFF2‐depleted tumours was significantly increased in comparison to control tumours (Figure 3E). Additionally, the growth rate of circAFF2‐overexpressed tumours was slower, and the tumour volume was smaller than that of the control tumours. On the other hand, circAFF2‐depleted tumours exhibited a faster growth rate and larger tumour volume (Figure 3F). Moreover, we performed TUNEL assay to evaluate the cell apoptosis and observed that apoptosis was significantly increased in circAFF2‐overexpressed tumours and reduced in circAFF2‐depleted tumours following radiation compared to the control tumours (Figure 3G). Collectively, our data suggest that circAFF2 can increase the radiosensitivity of CRC both in vitro and in vivo.

We then assessed m6A levels in radiation‐exposed CRC cells using Dot blot and m6A methylation assay. The m6A levels of HT‐29 and HCT‐116 cells were significantly increased after radiation, suggesting that m6A methylation was induced upon radiation (Figure 3A,B). The proteins involved in the regulation of m6A methylation mainly include methyltransferases (METTL3, METTL14, RMB15, WTAP, VIRMA), demethylases (FTO, ALKBH5) and RNA‐binding proteins (YTHDF1–3, YTHDC1–2, IGF2BPs). The level of ALKBH5, an m6A‐related demethylase, was reduced in both HCT‐116 and HT‐29 after radiation, while no differences in METTL3, METTL14, RBM15, WTAP and VIRMA were observed (Figure 4C). FTO was statistically different in HT‐29, but not in HCT‐116 (Figure 4C). Therefore, ALKBH5 could be related to radiotherapy‐induced m6A methylation in CRC cells. We then constructed a lentiviral vector for ALKBH5 overexpression and downregulation. Transfection efficiency is shown in Figure S1C. As expected, we observed a significant decrease in the m6A level in ALKBH5‐overexpressed CRC cells and an increase in the m6A level in ALKBH5‐downregulated CRC cells (Figure S1D,E). Interestingly, colony formation assay showed that the survival fraction of cells after radiation decreased significantly in ALKBH5‐overexpressed CRC HCT‐116 and HT‐29 cells. At the same time, the survival fraction of cells after radiation increased in ALKBH5‐downregulated CRC HCT‐116 and HT‐29 cells (Figure S4A). Moreover, we observed a significant increase in the expression of γ‐H2AX in the ALKBH5‐overexpressing cells, while the expression of γ‐H2AX was significantly decreased in ALKBH5‐downregulated cells (Figure S4B,C). Moreover, apoptosis of ALKBH5‐overexpressed CRC cells increased, whereas apoptosis was significantly reduced in ALKBH5‐downregulated CRC cells compared to control cells (Figure S4D). Together, these results indicate that ALKBH5 could increase the radiosensitivity of CRC cells.

Above, we showed that the expression of circAFF2 is associated with m6A. In addition, both ALKBH5 and circAFF2 impact radiosensitivity. Next, we sought to investigate whether circAFF2 may regulate ALKBH5 expression or vice versa. First, we overexpressed or downregulated ALKBH5 in CRC cells and examined circAFF2 levels. As shown in Figure 4D, the expression of circAFF2 was significantly increased in ALKBH5‐overexpressed CRC cell, whereas expression of circAFF2 was significantly decreased in ALKBH5‐depleted CRC cells compared to control cells. However, alterations in circAFF2 levels had no effects on the levels of m6A and ALKBH5 in CRC cells (Figure S5A,C), which was consistent in grafted tumours (Figure S5D,E). Furthermore, Me‐RIP‐qPCR showed that potential m6A‐modified segments of circAFF2 could be enriched by anti‐m6A rather than anti‐immunoglobulin G. ALKBH5 downregulation exhibited a significantly increase in the circAFF2 m6A modification levels (Figure 4E).

Next, we sought to explore whether the regulatory effect of ALKBH5 on circAFF2 depends on m6A methylation and identify specific m6A methylation sites of circAFF2. We utilised SRAMP software (http://www.cuilab.cn/sramp/↗) to predict the five possible m6A methylation sites of circAFF2 (Figure S6). Based on the predicted sites, construct luciferase reporter plasmids were constructed by inserting partial sequence of wild‐type circAFF2 or circAFF2 with mutant m6A methylation sites (MUT1: mutation site 367; MUT2: mutation site 404; MUT3: mutation site 707; MUT4: mutation site 719; MUT5: mutation site 823) (Figure 4F). The luciferase activity of CRC cells transfected with the wild‐type plasmid was attenuated, while the luciferase activity of cells transfected with MUT5 circAFF2 remained unchanged upon ALKBH5 downregulation (Figure 4G). Therefore, our data suggest that the regulation of circAFF2 by ALKBH5 depends on m6A methylation, and the main m6A methylation site of circAFF2 is 823.

The main m6A‐related reader proteins are YTHDF1/2/3, YTHDC1/2 and IGF2BP1/2/3; among these, YTHDF2 has been widely involved in regulating the stability of m6A‐containing mRNA.17, 20 Recent studies have also demonstrated that YTHDF2 can initiate the decay of m6A‐containing circRNA.18, 19, 32 We observed a dramatic upregulation of circAFF2 expression in YTHDF2‐downregulated HCT‐116 and HT‐29 cells (Figure 4H). Moreover, we observed that knockdown of YTHDF2 could reverse the downregulation of circAFF2 caused by ALKBH5 knockdown in HCT‐116 and HT‐29 cells (Figure 4I), indicating that the expression of circAFF2 is regulated by ALKBH5‐mediated demethylation, while YTHDF2 mediates circAFF2 degradation in an m6A‐dependent manner. In vitro radiosensitivity rescue experiments showed that the radiosensitivity induced by ALKBH5 or YTHDF2 could be reversed by circAFF2 (Figure 5A,D), which was confirmed in grafted tumours (Figure 6A,D). Collectively, these data suggest that ALKBH5/YTHDF2 regulates the radiosensitivity of CRC through circAFF2.

To study the downstream signaling pathways of circAFF2 in CRC cells, we performed RNA pull‐down experiments followed by mass spectrometry analysis and found 843 proteins that may bind to circAFF2 in HCT‐116 CRC cells (Figure 7A and Table S4). According to the RNA–protein interaction prediction website (http://pridb.gdcb.iastate.edu/RPISeq/↗) and literature review, CAND1, RRAS2, CKB and KNG1 were selected as candidate circAFF2‐binding proteins (Figure S7). We then performed RIP assay to evaluate the binding capacity of CAND1, RRAS2, CKB and KNG1 to circAFF2. As shown in Figure 7B, CAND1 was the only protein that bound to circAFF2. CAND1 has been shown to inhibit cellular neddylation modification by binding to CRL.26, 27, 28 Additionally, in the circAFF2 RNA pull‐down–liquid chromatography/mass spectrometry analysis, detectable enrichment of Cullin1 was found in the circAFF2 pull‐downed components, whereas no other Cullin subtypes were found (Table S4). Overexpression or downregulation of circAFF2 expression in HCT‐116 and HT‐29 CRC cells had no significant effects on CAND1 mRNA and protein expression both before and after irradiation (Figure S8). However, Cullin‐neddylation levels in radiation‐exposed cells were increased, which was significantly inhibited by circAFF2 overexpression and increased by circAFF2 knockdown (Figure 7C), suggesting that the interaction between circAFF2 and CAND1 can promote neddylation in CRC cells. Immunoprecipitation data showed that overexpression of circAFF2 could promote the binding of Cullin1 and CAND1, while downregulation of circAFF2 could inhibit the binding of Cullin1 and CAND1 (Figure 7D). Taken together, our data suggest that circAFF2 inhibits Cullin neddylation by promoting the binding of CAND1 and Cullin1.

Neddylation is a post‐translational protein modification that exerts a critical role in tumour development and radiosensitivity.22, 23, 24 To investigate the effect of neddylation on the radiosensitivity of CRC, we treated CRC cells with MLN4924, a neddylation inhibitor. We found that inhibition of neddylation strongly increased the radiosensitivity of CRC cells (Figure S9). We further performed rescue experiments to investigate whether circAFF2 enhances the radiosensitivity of CRC by promoting the binding of CAND1 to Cullin1. The circAFF2 overexpression reduced Cullin‐neddylation levels were restored by CAND1 silence (Figure 8A). Cloning formation assays, western blotting, ELISA assays and flow cytometry were used to demonstrate that CAND1 knockdown can restore the enhanced radiosensitivity caused by circAFF2 overexpression to the level of the control group (Figure 8B,E). The effects of circAFF2 on inhibiting neddylation and increasing radiosensitivity were eliminated in CAND1‐knockout CRC cells. Together, these data demonstrated that circAFF2 increased the radiosensitivity of CRC via promoting the binding between CAND1 and Cullin1, which led to the inhibition of Cullin neddylation. With the knockdown of CAND1, the Cullin neddylation was activated, which resulted a decreased radiosensitivity of CRC.

Preoperative tissues were collected from 58 patients with locally advanced rectal cancer who underwent neoadjuvant chemoradiotherapy (radiation dose range of 4500–4800 cGy supplemented with single oral administration of capecitabine) and surgical removal. ISH assay was adopted to examine the expression of circAFF2 and IHC was used to evaluate ALKBH5 levels in the collected tissues. For all patients, the median age was 60 years, 79% were male and 91% had low‐to‐moderate rectal cancer. Twenty‐five (43%) and 33 (57%) patients were clinically stage II and III before neoadjuvant chemoradiotherapy, respectively. A total of 12 patients showed pCR by postoperative pathology. Compared with the clinical staging before neoadjuvant chemoradiotherapy, 36 cases were pathologically staged down after the operation. Postoperative pathology showed 12 cases (20.6%), 18 cases (32.8%), 17 cases (29.3%) and 11 cases (19.0%) in TRG = 0, 1, 2 and 3, respectively. The findings of ISH and IHC demonstrated that the expression of circAFF2 and ALKBH5 was high in radiosensitive CRC tissues and low in radioresistant tissues (Figure 9A). Moreover, the expression of circAFF2 positively correlated with ALKBH5 (Figure 9B), while negatively correlated with TRG score (Figure 9C). Specifically, patients with TRG 0−1 exhibited a notably higher expression of circAFF2 than those with TRG 2−3 (Figure 9D). An ISH score of circAFF2 ≥6 was defined as high expression, and Kaplan–Meier curve analysis showed that the disease‐free survival of patients with elevated circAFF2 levels was significantly higher than that of patients with relatively low circAFF2 expression (Figure 9E). Collectively, our data suggest that circAFF2 can be used as an indicator for screening and efficacy evaluation of sensitive populations before concurrent chemoradiotherapy for rectal cancer.

RNA‐seq data were uploaded in the NCBI GEO database with accession number GSE186940 (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE186940↗).