RNA demethylase ALKBH5 promotes colorectal cancer progression by posttranscriptional activation of RAB5A in an m6A‐YTHDF2‐dependent manner

Paired tumour and adjacent normal tissues from 35 CRC patients after curative‐intent resection were collected at the Sixth Affiliated Hospital of Sun Yat‐sen University (SAH‐SYSU). To determine the correlation between ALKBH5 expression and clinic‐pathological characteristics of CRC, a total of 201 CRC patients combined with their fresh frozen tumour specimens originating from the institutional database program of colorectal disease (IDPCD) at our institute were also included in this study. The IDPCD cohort integrated patient data, including demographic and clinic‐pathological characteristics and follow‐up data, which was described in previous publications.20, 21, 22

All cell lines were obtained from the American Type Culture Collection (ATCC), and cultured in ATCC‐recommended media (Gibco, Grand Island, NY, USA) supplementing with FBS (Gibco; 10%, v/v) and 1% penicillin‐streptomycin at a 37°C incubator with 5% CO₂. The cells also routinely underwent PCR evaluation for Mycoplasma‐free using a Mycoplasma detection kit (TaKaRa). The ALKBH5 plasmid, shALKBH5 plasmid, shRAB5A plasmid, and their control plasmids as well as siYTHDF1/2/3 were purchased from TSINGKE (Tsingke Biotechnology Co., Ltd. China). The constructs were confirmed by direct sequencing in an independent third party (Sangon) before use. HCT116 and SW480 cells were transfected using corresponding plasmids with the reference to manufacturer's instructions, and qPCR and western blotting were applied to assess transfection efficiency. Details of the target sequences used in this study were listed in Table S1.

Genomic RNA was extracted from tissues and CRC cells using RNeasy Mini Kit (Qiagen, Germany) following its recommended protocols. Then RNA was reverse transcripted into cDNA according to the recommended instructions (TaKaRa, Japan). PCR using custom‐designed primers and cDNA templates was conducted using an SYBR Green Mix system (Applied Biosystems, USA), and detailed methods and data analysis were reported in our previous study.23 The primers used are listed in Table S2.

The proteins isolated from cell lysates were then electrophoresed by SDS‐PAGE and immunoblotted following classic protocols as we described previously.24, 25 The primary antibodies were used as follows: anti‐ALKBH5 (Abcam, ab195377), anti‐RAB5A (Abcam, ab218624), anti‐YTHDF2 (Proteintech, 24744‐1‐AP) and anti‐GAPDH (Proteintech, 10494‐1‐AP).

All tissues were embedded in paraffin wax and sectioned into 3 µm slices. Standard protocols of IHC have been described previously.26 The deparaffinised sections were incubated with anti‐ALKBH5 (Abcam, ab195377), anti‐RAB5A (Abcam, ab218624), and anti‐Ki67 (Cell Signaling Technology, 9449).

A total of 2 × 10⁵ cells were cultured in a glass‐bottom cell culture dish for 24 h and then fixed with paraformaldehyde, and the membrane was permeated with Triton X‐100, followed by blocking using goat serum (Thermo Fisher Scientific). RAB5A antibody (Abcam, ab218624) and fluorescent secondary antibody were used to develop the expression of RAB5A. The nuclei of the cells were stained with DAPI. The fluorescence was photographed by a Leica TCS SP8 microscope.

Cell proliferation was assessed by Cell Counting Kit‐8 (APExBIO, USA) (3000 cells/well were seeded in triplicate at a 96‐well plate) and colony formation (1000 cells/well were seeded at a 6‐well plate). Tanswell chamber assays (1 × 10⁵ cells in 200 µL serum‐free media were seeded in the top chamber) were applied to evaluate cell migrative and invasive abilities. Standard procedures have been described previously.27

All animal experiments were approved by the animal care Committee of Sixth Affiliated Hospital of Sun Yat‐sen University (No. IACUC‐2022022401). BALB/c‐nude mice (4−6 weeks old) were obtained from the Vital River. Briefly, 4 × 10⁶ transfected HCT116 cells or 8 × 10⁶ transfected SW480 cells in 0.1 mL PBS were injected into mice subcutaneously, and these mice were randomly divided into the experimental and control group. The tumours were monitored, calculated, removed and weighed as well as further fixed and embedded following corresponding protocols as described previously.27

Total RNA was isolated from CRC cells using RNeasy Mini Kit (Qiagen, Germany) following its recommended protocols. Then RNAs were serially diluted to 50, 100 and 200 ng/µL using RNase free water, and 100, 200 and 400 ng RNAs were spotted to dot‐blot assays as introduced by previous publication.28 The primary antibody used was anti‐m6A (Abcam, ab151230).

Total RNA was collected from HCT116 cells with stably ALKBH5‐knockdown and control cells (two replications, which were labelled as shALKBH5 and shNC, respectively) with a TRIzol reagent (Invitrogen, CA). RNA‐seq was conducted using Illumina NovaSeq 6000 platform. Paired‐end clean reads were aligned to the human reference genome (Ensemble_GRCh38.104). Differential expression analysis was performed using DESeq2 R package (1.20.0). For m6A‐RIP and MeRIP‐seq, chemically fragmented RNAs were incubated with anti‐m6A antibody for immunoprecipitation according to the standard procedures of the Magna MeRIP™ m6A Kit (Millipore, Germany). Enrichment of the m6A‐containing mRNA was then analysed either through quantitative reverse‐transcription polymerase chain reaction (MeRIP‐qPCR) or by high‐throughput sequencing. Enrichment of m6A mRNAs was then used for library construction with the NEBNext Ultra RNA library Prep kit from Illumina (NEB) and analysed via high‐throughput sequencing using Illumina Hiseq X platform. RNA‐seq, MeRIP‐seq and the sequencing data analysis were largely supported by Novogene (Beijing, China).

The RIP assay was conducted using the RIP kit (BersinBio, China) according to the manufacturer's guidelines. Briefly, magnetic beads were first incubated with 5 µg of anti‐ALKBH5 (Millipore, ABE547), YTHDF2 (Proteintech, 24744‐1‐AP) and normal IgG (Beyotime, A7016) at 4°C overnight. After the incubation in digestion buffer containing proteinase K, RNAs of interest were eluted from RNA‐protein complexes and then purified for further validation by qPCR to assess the enrichment.

Seeding stably transfected CRC cells into a 6‐well plate for 24 h and then treating them with actinomycin D (Sigma‐Aldrich, HY‐17559) at 5 µg/mL for 0, 3, and 6 h, followed by RNA extraction. The remaining RAB5A mRNA was measured by qPCR as described above.

According to the MeRIP‐seq results, the sites of RAB5A transcript where differential enrichment of m6A motif were focused to conduct luciferase reporters containing either wild‐type or mutant (N6‐methylated adenosines were replaced by cytosines) RAB5A 3′UTR sequences into the pmirGLO Basic vector by TSINGKE Biological Technology (Beijing, China). Detailed descriptions of the wild‐type and m6A sites mutated regions of RAB5A were shown in Table. Subsequently, CRC cells were transfected with the constructed luciferase reporter plasmids after being cultivated for 48 h. The luciferase activity was measured through a Firefly & Renilla Luciferase Reporter Assay Kit (Meilunbio, China) following the recommended instructions. S3

The statistical analysis of the differences among the intergroup comparisons were based on a two‐tailed Student's t test or one‐way analysis of variance (ANOVA). The Wilcoxon group signed‐rank test was adopted to compare two group samples, which were not normally distributed. Plot survival curves were plotted by Kaplan–Meier analysis, and the statistical significances were compared through a log‐rank test. A chi‐square test was applied to analyse qualitative data. Univariate and multivariate analysis based on the Cox hazards regression were conducted to estimate the prognostic factors correlated with the clinical variables. Statistical correlation analysis of the expression of ALKBH5 and RAB5A was conducted with linear regression. The statistical analyses were performed through GraphPad Prism 8.0 (Inc., USA), SPSS 22.0 (Inc., USA) and R‐project 3.5.1. Experiments were performed and repeated at least three times independently. *p < .05, **p < .01 and ***p < .001 indicated statistically significant, and ‘ns’ was considered with no statistical significance (p > .05).

To explore the expression of ALKBH5 in CRC, we first detected the levels of mRNA and protein in fresh frozen tumours and matched adjacent normal tissues using qPCR, WB and IHC assays, respectively. These results revealed that ALKBH5 was upregulated significantly in CRCs (Figures 1A and B and S1). Then, we also examined ALKBH5 levels in CRC cell lines, ALKBH5 was obviously overexpressed in most CRC cells compared to normal mucosal cell lines (Figure 1C and D). To further investigate the relevance of ALKBH5 expression in CRC clinically, a correlation analysis were conducted among clinicopathologic variables from an in‐house cohort consisting of 201 pathologically confirmed CRC patients and corresponding ALKBH5 expression of fresh frozen tumour tissues by qPCR assays. A high level of ALKBH5 was significantly associated with positive preoperative CEA levels (p = .010), more aggressive differentiation (p = .013) and advanced TNM stages (p = .001) (Figure 1E and Table 1). The Kaplan–Meier survival analysis showed that CRC patients characterised with higher ALKBH5 had extremely worse disease‐free and overall survival (DFS and OS) (Figure 1F and G). In addition, multivariate regression analysis exhibited that overexpressed ALKBH5 was demonstrated as an independent prognostic factor in CRC patients (HR = 4.213, p < .001) (Figure 1H). Collectively, these results implied that ALKBH5 is upregulated as well as has significant clinical relevance in CRC and acted as a prognostic biomarker for CRC patients.

To evaluate the function of ALKBH5 in CRC, ALKBH5 was silenced through two shRNA targeting the ALKBH5 (shA5#1, shA5#2), while stably upregulated through a lentivirus vector in HCT116 and SW480 cells according to the ALKBH5 expression in CRC cells (Figure 1C and D). The transfection efficiency was validated by WB and qPCR (Figure S2). CCK‐8 assays and colony formation indicated that the up‐regulation of ALKBH5 enhanced the proliferative ability of CRC cells (Figure 2A and C), while knockdown of ALKBH5 showed the opposite effect (Figure 2B and D). The Transwell assays suggested that the overexpression of ALKBH5 apparently increased both migration and invasion of CRC (Figure 2E), while attenuation of the expression significantly impaired those phenotypes (Figure 2F). To further examine the carcinogenic effect of ALKBH5 in CRC, we performed the in vivo experiments with subcutaneous xenograft. We found that the volumes, weights and growth rate of xenografted tumours significantly increased compared to control group when ALKBH5‐overexpressed cells were implanted (Figure 2G). On the contrary, silencing ALKBH5 expression drastically inhibited tumour weight and growth (Figure 2H). Moreover, the expression of Ki‐67 was also tested along with the diverse ALKBH5 expression (Figure S3). On the whole, these results revealed that ALKBH5 serves as an oncogene to promote CRC malignancy.

In order to further investigate the precise mechanism of ALKBH5 in the malignancy of CRC, we employed an integrated approach by combining RNA‐seq and MeRIP‐seq using stable ALKBH5‐knockdown HCT116 cells and the corresponding control cells. RNA‐seq data revealed that 427 differential transcripts were significantly regulated upon ALKBH5‐knockdown. We calculated the fold change of gene expression and considered genes with |log₂FC| > 0.5 and padj < .05 as statistically different expression (Supplementary file 2: RNA‐seq data). Meanwhile, MeRIP‐seq revealed that the m6A peaks of 372 transcripts showed increased enrichment upon ALKBH5 silence (log₂FC > 0 and padj < 0.05) (Supplementary file 2: MeRIP‐seq data). In this study, Venn analysis was performed by combining with the differentially expressed genes identified via RNA‐seq and RIP‐seq (Figure 3A), and analysis showed that seven genes were overlapped, namely, DNAJB1, RAB5A, CCSER2, DYNLRB1, PLCD3, RPL14P1 (pseudogene) and HIP1R. Then, they were scheduled for initial validation in ALKBH5‐ overexpressing or ‐silencing cells through qPCR. Intriguingly, only RAB5A was the only gene, which was consistently identified to be positively mediated by ALKBH5 in all cells (Figures 3B and S4). Moreover, as indicated by western blotting and immunofluorescence, the levels of RAB5A were increased in ALKBH5‐overexpressing cells and decreased in ALKBH5‐knockdown cells (Figures 3C and D and S5), which were further confirmed by IHC assay using xenograft tissues in nude mice (Figure S3). Furthermore, the levels of RAB5A mRNA were also increased in tumour tissues (Figure 3E) and showed a positive correlation with those of ALKBH5 in CRC tissues (Figure 3F). As a result, a preliminary conclusion could be drawn that RAB5A might be the direct downstream target of ALKBH5.

To further verify the direct mechanism that ALKBH5 targets RAB5A mRNA and whether it depends on the m6A catalytic activity. We first detected the global m6A levels in different transfected CRC cells and their control groups through m6A dot‐blot assays. As expected, the m6A levels of total RNA were decreased obviously in both ALKBH5‐overexpression HCT116 and SW480 cells compared with vector cells, while ALKBH5‐deletion had the opposite effects (Figure S6). The MeRIP‐seq analysis showed that m6A‐peak of RAB5A in 3′UTR accreted apparently with the knockdown of ALKBH5 (Figure 4A). RIP assays performed by the anti‐ALKBH5 antibody in HCT116 and SW480 cells validated that RAB5A mRNA was enriched significantly in the ALKBH5 group, but not the IgG group (Figure 4B). Next, we treated cells with actinomycin D to block transcription and assess the effect of ALKBH5 on RAB5A mRNA stability. We found that ALKBH5 sufficiency enhanced the stability of RAB5A, whereas ALKBH5 knockdown induced a faster degradation rate of RAB5A mRNA (Figure 4C). The above results demonstrated that ALKBH5 was bound to RAB5A and increased the stability of RAB5A mRNA via a posttranscriptional mechanism.

Then MeRIP‐qPCR with specific primers was performed to detect the enrichment of m6A in RAB5A. The results demonstrated that the m6A modification of RAB5A mRNA was significantly elevated on ALKBH5 knockdown as well as substantially reduced on activation of ALKBH5 (Figure 4D). To clarify the essential role of m6A modification on RAB5A, luciferase reporters containing either wild‐type or mutant (N6‐methylated adenosines were replaced by cytosines) RAB5A 3′UTR sequences were designed (Figure 4E). The luciferase activities of CRC cells transfected with RAB5A‐WT plasmid, but not the mutant counterpart, obviously decreased in the silence of ALKBH5, and the analogous results would be seen in ALKBH5‐overexpressed cells (Figure 4F), which revealed that the expression of RAB5A mRNA was under the regulation of ALKBH5‐related m6A modification.

It was acknowledged that m6A ‘readers’ were crucially involved in recognising methylated target transcripts; we next planned to investigate the potential reader protein participating in the procedure elucidated above. Given that YTHDFs were reported to extensively mediate translation efficiency and degradation of m6A modified RNAs,29 we hence explored the effect of YTHDF1/2/3 on the stability RAB5A mRNA. Two specific designed siRNAs targeting YTHDF1/2/3 were used in two CRC cells to validated the alterations of RAB5A expression. We found that YTHDF2 knockdown remarkably augmented RAB5A expression (Figure 4G and H), but YTHDF1/3 exhibited no effect on RAB5A in two cells (Figure S7). Moreover, YTHDF2 deficiency could enhance the stability of RAB5A in two CRC cells at indicated time points after actinomycin D treatment (Figure S8). These results were consistent with the knowledge that YTHDF2 tended to facilitate targeted mRNA decay.30 The interaction between YTHDF2 protein and RAB5A mRNA was then verified by RIP and RNA pull down assays (Figures 4I and S9). In addition, although YTHDF2 expression would not be affected by ALKBH5, the knockdown of YTHDF2 could counteract the degradation of RAB5A caused by ALKBH5 loss (Figure 4J). Collectively, overexpressed‐ALKBH5 could enhance the stability of RAB5A mRNA via an m6A‐YTHDF2‐dependent pathway.

To gain insight into the phenotype of RAB5A in CRC, the expression of RAB5A in HCT116 and SW480 cells was silenced using shRNAs targeting RAB5A (shRAB5A), and the transfection efficiency was also validated by WB and qPCR (Figure 5A and B). CCK‐8 assays and colony formation revealed that the silence of ALKBH5 strongly inhibited the growth and viability of CRC (Figure 5C and D). As shown in the Transwell assays, loss‐of‐function of RAB5A markedly impaired the migrative and invasive abilities of CRC cells (Figure 5E). In vivo experiments, we also found that silencing RAB5A expression could drastically diminish tumour weights and volumes of xenografted tumours compared to control group (Figure 5F‐H). In summary, RAB5A was activated during CRC tumourigenicity and promotes the oncogenesis of CRC.

To prove that the above phenotypes induced by ALKBH5 were also regulated by the ALKBH5‐RAB5A axis, we then designed some rescue functional assays in vitro. WB and qPCR assays were conducted to validate that the overexpression of RAB5A mediated by overexpressing ALKBH5 was strikingly reduced by RAB5A knockdown in HCT116 and SW480 cells (Figure 6A and B). Silencing of RAB5A would extensively reverse the cell proliferation, colony formation, migration and invasion promoted by overexpression of ALKBH5 in cells (Figure 6C‐E).