METTL3 promotes glycolysis and cholangiocarcinoma progression by mediating the m6A modification of AKR1B10

We characterized the expression profile of CCA RNA-seq datasets downloaded from TCGA-Cholangio carcinoma (CHOL) dataset and then the differential expression of eight m6A methylation-related genes (FTO, HNRNPA2B1, HNRNPC, METTL3, WTAP, YTHDC1, YTHDC2, and YTHDC2) between CCA and normal control samples were evaluated using R package and plotted into heatmap using R package.

The GEPIA2 database contained of 36 tumor tissues of CHOL and 9 adjacent tissues samples. We analyzed the differential expression of METTL3 and AKR1B10 between the CCA and adjacent tissues.

A CCA tissue microarray (No. LVC1202) was generated from 60 cancer tissues and paired pericarcinomas that purchased from Boster Biological Technology co.ltd. IHC staining for METTL3 and AKR1B10 were performed using the above microarray tissue blocks of CCA. Briefly, paraffin-embedded tissues were made into 6 μm sections following deparaffinization and hydration. Sections were repaired by high-pressure following incubated with 0.33% H₂O₂ in methanol to block endogenous peroxidases and incubated with 10% normal horse serum in TTBS to block non-specific binding. Next, sections were incubated with anti-METTL3 (1:100, Proteintech, 15073-I-AP) or anti-AKR1B10 (1:500, Abcam, ab192865) overnight at 4 °C and then incubated with horse anti-mouse biotinylated antibody. Finally, sections were with colored by chromogen of DAB and counterstained with hematoxylin. Pictures were captured and exported using NDP.view 2.0. IHC staining results of METTL3 were assigned 1–3 scores and AKR1B10 were assigned 0–3 scores based on staining intensity of positive cells and percentage of positive cells. The section with strong staining intensity and diffuse of positive cells was assigned 3 score; strong staining intensity and focal distribution of positive cells was assigned 2 score; weak- medium staining intensity of positive cells was assigned 1 score; no staining or non-specific staining was assigned 0 score. METTL3 scores of 1 and 2 were categorized as low expression group and 3 as high expression group. AKR1B10 scores of 0 and 1 were categorized as low expression group, scores of 2 and 3 as high expression group.

Human liver bile duct carcinoma cell RBE and HCC9810 were purchased from Procell (China) and were maintained in RPMI 1640 with L-Glutamine (CORNING, China) containing 10% FBS (GIBCO, China) and 1% penicillin/ streptomycin (GIBCO, China) at 37 °C and 5% CO₂.

To construct METTL3 stably overexpressed stable RBE cell line, the full-length of METTL3 was inserted into the lentiviral vector pLenti-EF1a-EGFP-P2A-Puro-CMV-3 × FLAG-WRPE (OE-METTL3 group) and then harvested-lentiviruses were infected with RBE cells using polybrene (hexadimethrine bromide, Sigma 107689-100MG). Blank lentiviral vector was served as negative control (Vector group).

To transient knockdown expression of METTL3 in HCC981 cells and AKR1B10 in RBE cells, we used small interfering RNA (siRNA) method. The synthesized sequence of siRNA targeted METTL3 (siMETTL3) or AKR1B10 (siAKR1B10) by GenePharma (Shanghai, China) were shown in Additional file 1: Table S1.

Besides, for stable knockdown of METTL3 expression used in animal study, lentiviruses vector pLKO.1 puro containing METTL3 shRNA (shMETTL3) and non-targeting scrambled shRNA (shNC) were purchased from GenePharma (Suzhou, China).

According to the manufacturer’s instructions, 5 μL of siMETTL3, siAKR1B10, shMETTL3 and shNC were diluted in 45 μL OPTI-MEM and then mixed with 10 μL Lipofectamine 2000 reagent pre-diluted with 45 μL OPTI-MEM. The mixture was added into cells and cultured for 24 h before further efficiency verification experiments.

TRIzol (Invitrogen Life Technologies) was applied for isolating total RNA from RBE cells and HCCC-9810 cells. Quality qualified RNA reverse-transcription into cDNA was carried out using High Capacity cDNA Reverse Transcription kit (Applied Biosystems) and then mRNA expression of METTL3, AKR1B10, and GAPDH were measured by QuantStudio 6 Flex Real-Time PCR System (Thermo Fisher Scientific) with FastStart Universal SYBR Green Master mix (Takara, China) according to the product’s protocol. Relative expressions of genes were normalized to GAPDH using 2^−ΔΔCq method. The primers were listed in Additional file 1: Table S1.

Total protein concentration was measured by the BCA protein assay kit (Thermo scientific, USA). Then, 20 μg proteins were resolved by 10% SDS-PAGE and transferred onto PVDF membranes following blocking for nonspecific binding with 5% nonfat milk at 25 °C for 2 h. The membranes were incubated with anti-METTL3 (1:2000, Proteintech, 15,073-I-AP), anti-AKR1B10 (1:1000, Abcam, ab192865), and anti-GAPDH (1:1000, Proteint, 60004-1-Lg) at 4 °C overnight. After that, membranes were incubated with Goat Anti-Mouse IgG H&L (HRP) (1: 10,000, Abcam, ab205719) at 25 °C for 1 h. Immunore-activity was imaged by Bio-Rad ChemiDoc XRS system and quantified by Image J.

One hundred microliter cells with a density of 1 × 10⁵ cells/well were seeded in 96-well plates and cultured for 24 h. Then, at each indicated times (0 h, 24 h, 48 h, 72 h, 96 h), 10 μL CCK-8 solution (Dojindo, Japan) was added and was incubated for another 1 h at 37 °C. The optical density was read at 450 nm using a Multiskan FC microplate reader (Thermo Fisher Scientific).

Cell apoptosis was detected by TUNLE assay using a One Step TUNEL apoptosis kit (red Tunnelyte™ CY3 fluorescence detection) (C1089, Beyotime, China) according to the product instruction. Briefly, adherent CCA cells were washed with PBS and then fixed in immunostaining fixative solution (P0098, Beyotime, China) followed by permeabilized in immunostaining strong permeable solution (P0097, Beyotime, China). After that, cells were incubated with TUNEL solution for 1 h and sealed with anti-fluorescence quenching sealing tablets. Lastly, cells were photographed on a fluorescence microscope.

Cell migration and invasion were accessed by using a Transwell assay. The Transwell chamber was coated with 0.8 μm Matrigel (354480, BioCoat) for cell invasion assay, otherwise for migration assay. CAA cells were seeded into the upper chamber containing serum-free medium and complete medium was added to the lower chamber as a chemoattractant. The cells were cultured 24 h at 37 °C. The migrated cells to the lower chamber was photographed and calculated in three randomly fields under an inverted light microscope. The invaded-cells arriving at the lower chamber were fixed in 10% formaldehyde for 15 min and stained with 0.1% crystal violet for 10 min, and finally photographed in three randomly fields.

Relative glucose uptake required by tumor cells was measured by Glucose Uptake Fluorometric Assay Kit (MAK084, Sigma-Aldrich, USA), and relative lactate production was measured by Lactic Acid Content Assay Kit (D799851-0050, Sangon, China) according to the technical bulletin provided by manufacturer.

A total of twelve 4-week-old female BALB/c nude mice were purchased from Shanghai SLAC Laboratory Animal Company and were randomly divided into two groups: ShMETTL3 group (n = 6) and Vector group (n = 6). Mice were single housed at room temperature (21–26 °C) on a nature light cycle for one week before experiments to adapt laboratory environment. All mice were provided free access to diet and water. HCCC-9810 cells with METTL3 knockdown were digested by trypsin and made into single cell suspension. Next, 2 × 10⁶ HCCC-9810 cells were subcutaneously injected into axillary of mice. After injection, mice were continued to be raised normally for 3 weeks. Tumor volume (length × width × width × 0.5) was measured every 3 days using caliper. Mice were euthanized using CO₂ inhalation after the last measurement, tumor was collected and weighted.

RNA sequencing was performed at Yingbio Technology (Shanghai, China) using an Illumina HiSeq 2500. For differentially expressed genes (DEGs) identification, the thresholds was Log2fold change (FC) > 1 or < − 1 and false discovery rate (FDR) < 0.05. Gene onology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) were carried out for the DEGs using R package. Transcriptome sequencing was repeated in three replicates.

Total RNA was extracted from RBE cells-overexpressed METTL3 or not, and then isolated mRNA was purified by Dynabeads mRNA Purification Kit (Invitrogen, USA) according to the manufacturer’s instructions. Next, purified mRNA was fragmented by RNA Fragmentation Reagent (Invitrogen, USA) before immunoprecipitation. After that, the anti-m6A antibodies or anti-METTL3 were conjugated to protein magnetic beads for immunoprecipitation, anti-immunoglobulin G (IgG) was served as negative control. Finally, RNA was eluted from RNA–protein immunocomplexes followed by RT-qPCR analysis.

RBE cells overexpressed METTL3 or NC were seeded in a 12-well plate and cultured for 24 h. Then, 5 μg/mL actinomycin D were added into cells and cultured another 0 h, 3 h, 6 h, 9 h, and 12 h following cell collection. Once collection, the total RNA was extracted from these RBE cells used for RT-qPCR as described above.

For m6A reporter assays, the wild-type of AKR1B10 sequence and the mutated at m6A motif 1 (mut-1, m6A was replaced by G), at m6A motif 2 (mut-2, m6A was replaced by G), and at m6A motif 3 (mut-3, m6A was replaced by C) were inserted into XhoI/NotI site of the psiCHECK-2 luciferase reporter vector. Then, these recombinant plasmids were transfected into RBE cells overexpressed METTL3 and NC using Lipofectamine 2000 reagent as described above.

Data analysis was performed by GraphPad Prism 9.0 and data were presented as mean ± SD. Kolmogorov–Smirnov test was used for evaluating the data normality, and Levene test was used for evaluating homogeneity of the variance of data. One-way ANOVA with Tukey test for three groups and t test for two groups were utilized when the data was normality and homogeneous. The chi-square test was used to evaluate the correlation between molecular expression and clinical data. P value less than 0.05 was considered significant.

To obtain m6A methylation-related genes in CCA, we analyzed TCGA-CHOL data. A total of 8 m6A methylation-related genes were differentially expressed between CCA and normal tissues (Fig. 1A). Among these 8 genes, METTL3 showed the highest fold-change of upregulation in CCA (Fig. 1C). The GEPIA2 dataset supported that METTL3 was significantly accumulated in CCA tissue compared with normal tissues (Fig. 1B). Specifically, TIMER analysis (http://​timer.​comp-genomics.​org/​) showed that METTL3 was upregulated in a variety of tumors, including CHOL (Additional file 2: Fig. S1A). Accordingly, we focused primarily on METTL3. To determine the expression of METTL3 in CCA tissues, we performed IHC staining for METTL3 in a CCA tissue microarray study that enrolled 60 patients. Compared with peritumoral tissues, METTL3 was overexpressed in CCA tissues (Fig. 1D). METTL3 expression was scored at three levels on the basis of IHC staining intensity, and the METTL3 score in CCA tissues was significantly higher than that in peritumoral tissues (Fig. 1E, F). According to the METTL3 score, we found that METTL3 expression was significantly correlated with TNM stage but not with the other clinicopathological characteristics (Table 1). The survival analysis showed that high expression of METTL3 indicated a poor prognosis in patients with CCA (Fig. 1G). Collectively, METTL3 expression is upregulated in CCA and is associated with CCA progression.

To evaluate the role of METTL3 in CCA, we first detected endogenous mRNA expression of the METTL3 gene in two CCA cell lines. We selected HCCC-9810 cell lines for knockdown and RBE cell lines for overexpression of METTL3 due to their mRNA (Fig. 2A) and protein (Fig. 2B) expression patterns. Next, the efficacy of the two siRNAs in knocking down METTL3 expression was confirmed by RT‒qPCR (Fig. 2C) and western blotting (Fig. 2D). This deletion strategy resulted in impaired proliferation and elevated TUNEL-positive apoptotic cells in HCCC-9810 cells (Fig. 2E, F). Transwell assays showed similar impairments in migration and invasion in HCCC-9810 cells as a consequence of METTL3 knockdown (Fig. 2G). To determine whether METTL3 is necessary for aerobic glycolysis for the energy supply of CCA, we subjected siMETTL3 HCCC-9810 cells and siNC controls to glucose uptake and lactate production assays. As expected, METTL3 knockdown reduced the glycolytic response in HCCC-9810 cells, as evidenced by the decreased glucose uptake and lactate production (Fig. 2H, I). Therefore, METTL3 knockdown inhibits glycolysis and the malignant phenotype of CCA cells in vitro.

Following METTL3 knockdown in CCA cells, we next sought to determine the effects of METTL3 overexpression. For this, a lentivirus plus METTL3 plasmid (OE-METTL3) or empty vector controls (Vector) were transfected into RBE cells. The overexpression efficiency of METTL3 was confirmed at both the mRNA (Fig. 3A) and protein levels (Fig. 3B). The functionality of METTL3 overexpression produced precisely the opposite effect of METTL3 knockdown, as evidenced by the fact that cell proliferation, survival, migration, and invasion were boosted in the OE-METTL3 group relative to the Vector group (Fig. 3C–E). Moreover, high METTL3 expression was accompanied by spontaneous increases in glucose uptake and lactate production (Fig. 3F–G). Therefore, METTL3 overexpression promotes glycolysis and the malignant phenotype of CCA cells in vitro.

Demonstrating the promotion of CCA by METTL3 in vitro raised the question of whether METTL3 was also effective in vivo. For this, HCCC-9810 cells containing shMETTL3 or Vector were subcutaneously injected into the axilla of mice, and tumor growth was observed (Fig. 4A). Compared with Vector mice, shMETTL3 mice showed smaller tumor volumes and lighter tumor weights (Fig. 4B, C). Thus, METTL3 promotes CCA tumor growth in vivo.

Next, to explore whether METTL3 can drive a spontaneous molecular alteration in CCA cells, we performed whole-transcriptome sequencing in RBE cells with or without METTL3 overexpression. Enforced expression of METTL3 resulted in changes in the expression profiles of 240 genes, of which 103 were upregulated, including AKR1B10, and 137 were downregulated (Fig. 5A). Interestingly, these DEGs were extensively involved in glycolytic metabolism-related pathways, including 2-oxocarboxylic acid metabolism, galactose metabolism, pyruvate metabolism, and carbon metabolism (Additional file 2: Fig. S1B). These results echo those of our previous conclusion that METTL3 promotes glycolysis. Given that METTL3 promotes the expression of downstream target genes through m6A modification [20], we focused on those DEGs that were upregulated upon METTL3 overexpression. All of the upregulated DEGs were subjected to KEGG analysis, and the results showed that the DEGs were primarily preferentially enriched in metabolic pathways, including galactose metabolism, butanoate metabolism, and riboflavin metabolism (Fig. 5B). To further determine the major affected molecules mediated by METTL3, we overlapped the genes upregulated in the METTL3-overexpression group from our whole-transcriptome sequencing and the genes upregulated in CCA tissues from TCGA, and then the intersection DEGs were analyzed by KEGG. The results showed that these intersection DEGs were mainly involved in glycolysis metabolism-related pathways, such as galactose metabolism, fructose and mannose metabolism, and glycerolipid metabolism (Additional file 2: Fig. S1C). Thus, 4 intersection DEGs (AKR1C1, AKR1C2, cyclin D2 (CCND2), and AKR1B10) related to energy metabolism were selected for RT‒qPCR validation. Except for CCND2, the remaining 3 DEGs were upregulated in the OE-METTL3 group compared to the Vector group (Fig. 5C). Moreover, MeRIP-qPCR results showed that a significant increase was observed only in AKR1B10 mRNA, which was precipitated by the m6A antibody, in the OE-METTL3 group compared with the Vector group, suggesting that AKR1B10 mRNA contains a METTL3-m6A modification site and that its m6A level was enhanced by METTL3 (Fig. 5D). Knockdown of METTL3 significantly reduced the m6A level of AKR1B10 (Additional file 2: Fig. S1D). The protein expression of AKR1B10 was also elevated upon METTL3 overexpression (Fig. 5E). Interestingly, according to SRAMP (http://​www.​cuilab.​cn/​sramp), there were indeed 5 potential m6A modification sites on AKR1B10 mRNA (Fig. 5F). Taken together, given our results above, we consider AKR1B10 to be a target of METTL3 and chose to further evaluate the role of AKR1B10 in CCA progression.

To further confirm the binding relationship between METTL3 and AKR1B10, RIP assays were carried out. RIP results showed that AKR1B10 protein was specifically enriched by the METTL3 antibody (Fig. 5G). Ctinomycin D, a transcription inhibitor, was used to assess the stability of AKR1B10 mRNA upon METTL3 overexpression. We found that METTL3 overexpression significantly enhanced the stability of AKR1B10 mRNA compared with the control in the presence of actinomycin D (Fig. 5H). To explore the m6A modification site on AKR1B10, three AKR1B10 mutants with a single mutation were constructed. As shown in Fig. 5I, METTL3 significantly elevated the luciferase activity of the AKR1B10 wild-type luciferase reporter, as well as AKR1B10 mut-2 and mut-3, but not AKR1B10 mut-1. The results suggested that the binding between METTL3 and AKR1B10 was dependent on the m6A modification site of mut-1. Collectively, METTL3 promotes mRNA stability and increases the expression of AKR1B10 by directly binding in an m6A modification manner.

We assessed AKR1B10 expression in CCA tissues using the GEPIA 2 database, and the results revealed that AKR1B10 was upregulated in CHOL tissue compared with normal tissue (Fig. 6A). A CHOL tissue microarray containing 60 samples also concluded that AKR1B10 was upregulated in CHOL tissues (Fig. 6B). Moreover, the overall IHC staining of AKR1B10 expression was scored according to 4 levels: no staining (0), weak (1), moderate (2), and strong (3) (Fig. 6D). Based on the IHC score, AKR1B10 expression in CHOL tissue was significantly higher than that in peritumoral tissue (Fig. 6C) and was not correlated with any clinicopathological characteristics (Table 1). In addition, IHC of the CHOL tissue microarray showed a highly significant positive correlation between the AKR1B10 score and the METTL3 score (Fig. 6E). Therefore, these results indicated that AKR1B10 is highly expressed in CCA.

To inquiry the function of AKR1B10 in CCA, we knocked down AKR1B10 expression in REB cell lines. The reduced expression of AKR1B10 in RBE cells was confirmed both at the mRNA (Fig. 7A) and protein levels (Fig. 7B). The CCK8 assay showed that siAKR1B10 led to impaired proliferation (Fig. 7C). Knockdown of AKR1B10 expression also resulted in significant declines in the invasive and migratory capabilities (Fig. 7D, E). Given that METTL3 regulates glycolysis and AKR1B10 mediates glycolysis, as reported in the literature [21], we examined the effect of AKR1B10 knockdown on glycolysis. As shown in Fig. 7F, G glucose uptake and lactate production were significantly reduced after AKR1B10 knockdown in RBE cells. In conclusion, AKR1B10 could promote glycolysis and the malignant phenotype of CCA cells in vitro.

As confirmed in the preceding text, AKR1B10 is a target of METTL3, so we wondered whether METTL3 exerts an oncogenic role in CCA through AKR1B10. To this end, rescue assays of AKR1B10 silencing in METTL3-overexpressing cells were performed. As expected, METTL3 overexpression accelerated the proliferation, migration, and invasion in RBE cells compared to the Vector control group, but knockdown of AKR1B10 led to a partial reversal exhibited by the capacities for proliferation, migration, and invasion (Fig. 8A–C). Moreover, knockdown of AKR1B10 also reversed the glycolysis phenotypes of glucose uptake and lactate production caused by METTL3 overexpression (Fig. 8D–E). Therefore, METTL3 exerts an oncogenic role in CCA through AKR1B10.