Suppression of m6A mRNA modification by DNA hypermethylated ALKBH5 aggravates the oncological behavior of KRAS mutation/LKB1 loss lung cancer

We first investigated the association between m6A RNA modification and aggressive KL lung cancer. IHC staining showed the reduced m6A level by loss of LKB1 in patients with KRAS mutation. And, the lowest of m6A level was found in KL tumor tissues when compared to wild type (WT), KRAS^Wt; LKB1 ^Loss (L), or KRAS^Mut; LKB1 ^Wt (K) tumor tissues (Fig. 1A, B). The m6A level was inversely correlated with TNM (Tumor, Node, and Metastasis) and clinical stage, whereas positively with tumor differentiation in KRAS mutant patients, but not in those with Kras wildtype (Fig. 1C). In addition, we found lower m6A level in K specimens that were null for thyroid transcription factor 1 (TTF1), a positive prognostic feature (Fig. 1D)²³. Thus, reduced m6A modification is associated with aggressive KL lung cancer.

To investigate how m6A is regulated in KL lung cancer, we compared the expression of proteins that act as the m6A writer complex (METTL3, METTL14, and WTAP) and erasers (ALKBH5 and FTO). IHC staining showed that only ALKBH5 expression was higher in KL than that of in K tumor tissues (Fig. 2A, B). ALKBH5 protein expression was negatively correlated with m6A level, in contrast to positive correlation with the tumor size, TNM and clinical stage in K lung cancer patients (Fig. 2C). However, LKB1 loss alone was not sufficient to change m6A level or ALKBH5 expression or their relationship with aggressive tumor phenotypes (Fig. S1 A–D).

To confirm observations of clinical specimens, we screened for the differential expressions of m6A mediators and readers based on database queries. Analysis of TCGA, MTB^8,24, and CCLE databases revealed that ALKBH5 mRNA expression was consistently higher in KL cancer tissues or cells, and negatively correlated with LKB1 expression (Figs. 2D and S2A–D). Furthermore, elevated ALKBH5 expression correlated with poor prognosis for patients with KRAS mutation (Fig. 2E). Additionally, ALKBH5 had the highest basal mRNA expression in human and mouse lung cancer tissues and cell lines (Fig. S2A–C). Taken together, these findings indicate that ALKBH5, a major regulator of m6A modification, contributes to aggressive phenotypes of KL lung tumors.

To investigate whether LKB1 loss affects ALKBH5 and m6A modification, we screened several lung cancer cell lines and categorized them based on KRAS mutation and LKB1 expression status. A549^KL (KRAS^mut/LKB1^loss), H1792^K (KRAS^mut/LKB1^high), H1299 (KRAS^wt/LKB1^high), and H1703 (KRAS^wt/LKB1^low) were selected from each category for further analyses (Fig. S3A). We then generated the loss of LKB1 function by siRNA-LKB1-transfection in H1299 and H1792 cells, while gain-of-function by LKB1 overexpression in H1703 and A549 cells. As expected, LKB1 expression positively regulated m6A levels in cells with KRAS mutation, but not with KRAS wild type (Fig. 2F). Consistent with clinical data, western blot and qRT-PCR assays showed that ALKBH5 was also the only m6A modulator negatively regulated by LKB1 in KRAS^mut cells, but was unaffected in KRAS^wt cells (Figs. 2G and S3B–E). These observations were supported by immunofluorescence staining that demonstrated negligible ALKBH5 signal in LKB1 overexpression A549 cells (Fig. 2H). Exogenous ALKBH5 expression blocked m6A staining in the presence of LKB1 (Fig. 2I). Furthermore, LKB1 expression negatively correlated with ALKBH5 expression was also found in KRAS-mutated pancreatic and colorectal cancer cell lines (Fig. S4A–E). Therefore, LKB1 expression directly affects the global m6A levels via ALKBH5 in KRAS-mutated cancer cells.

Next, we explored functional relationships between LKB1 and ALKBH5 in KRAS mutant lung cancer cells. First, we established the transient ALKBH5 and/or LKB1 knockdown models in H1792 cells, as well as overexpression in A549 cells (Fig. 3A). As expected, ALKBH5 could fully release LKB1 repressed m6A levels in both cell lines (Fig. 3B, C). Functionally, ALKBH5 knockdown significantly inhibited cell proliferation, as shown by decreased colony formation in LKB1-silenced H1792 cells (Fig. 3D). We observed similar for H1792 cell migration in the transwell assay (Fig. 3E). Conversely, ALKBH5 overexpression promoted cell proliferation and migration in LKB1-transfected A549 cells (Fig. 3F, G). Thus, ALKBH5 negatively regulated LKB1 role for cell proliferation and migration in KRAS mutant cells.

We then explored how LKB1 regulates ALKBH5 through DNA methylation. Consistent with previous studies in pancreatic ductal epithelial cells¹¹, LKB1 alterations also negatively regulated the global 5mC DNA methylation in KRAS mutant lung cancer cells (Fig. 4A) but had no effect in KRAS wild-type cells. Specifically, ALKBH5 gene DNA methylation positively correlated with ALKBH5 mRNA expression in KRAS mutant lung, and colorectal cancer cell lines, based on the CLLE database (Fig. S5A–D). Subsequent treatment with 5-azacytidine (5-aza), an inhibitor of DNA methylation, reduced ALKBH5 mRNA and protein expression in a dose-dependent manner in H1792 and A549 cells, independent of LKB1 status (Fig. 4B–D). Based on ChIP-sequence analysis from the ENCODE database, we identified a single putative transcriptional repressor, CTCF (CCCTC-binding factor) and several activators, including H3K4me1, H3K4me2, H3K4me3, H3K9ac, and H3K27ac, on the ALKBH5 gene core promoter in A549 cells (Fig. S6). Interestingly, the CTCF peak region was colocated in the CpG island of ALKBH5 promoter (Fig. 4E). Further analysis by MeDIP assay indicated 5mC-methylation on the CTCF peak region was decreased with 5-aza treatment or exogenous LKB1-transfection in A549 cells (Fig. 4F). Lastly, bisulfite sequencing confirmed decreased DNA methylation on the CTCF peak region in 5-aza-treated or LKB1 overexpressed A549 cells (Fig. 4G, H). Thus, the data indicate that LKB1 loss induces DNA hypermethylation, thereby controlling ALKBH5 expression in KRAS mutant cancer cells.

To determine whether CTCF directly represses ALKBH5, we used western blot and RT-qPCR. We found that silencing CTCF increased ALKBH5 protein and mRNA expression, and also rescued ALKBH5 repression by LKB1 overexpression (Fig. 5A–C). We next generated serial deletions, including CTCF motif deletion constructs based on the human ALKBH5 promoter, for the luciferase reporter assay. The CTCF-containing construct had lower ALKBH5 promoter activity than constructs that lacked the motif (Fig. 5D). Notably, CTCF motif deletion increased ALKBH5 promoter activity to levels similar to those in constructs without the CTCF motif. This evidence suggests that CTCF directly represses ALKBH5 transcriptional activity.

Next, we found that inhibition of DNA methylation by 5-aza treatment or by LKB1 overexpression reduced ALKBH5 promoter activity on the CTCF motif-containing construct, but had no effect on CTCF motif-deficient construct (Fig. 5E). Further ChIP-qPCR analysis indicated that histone activators, such as H3K4me1, H3K4me2, H3K4me3, H3K9ac, and H3K27ac, occupy the ALKBH5 gene promoter (Fig. 5F), consistent with ChIP-seq results (Fig. S6). Interestingly, 5-aza treatment or LKB1 overexpression promoted CTCF enrichment on ALKBH5, and also partially prevented enrichment of H3K4me3, H3K9ac, and H3K27ac (Fig. 5H–J). Therefore, loss of LKB1 promoted DNA hypermethylation in the CTCF peak region, thus preventing CTCF binding and releasing repression of ALKBH5 in KRAS mutant cells.

To identify downstream targets of ALKBH5-mediated m6A modification in lung cancer, we overlapped 2605 genes using the canonical m6A motif-enriched gene stop codon region based on m6A-sequence in A549 cells (16) (Fig. S7A). Gene ontology analysis revealed that the 2605 genes were significantly enriched in gene transcription regulation, mRNA splicing, cell cycle, and mRNA stability (Fig. S7B). KEGG pathway analysis revealed that the overlapping genes were closely associated with Hippo and TGF-β pathways (Fig. S7C). Using qRT-PCR, we confirmed that 45.2% (14/31) of Hippo-Yap pathway genes were directly regulated by ALKBH5 (Fig. S7D, E). Western blot assay further confirmed that ALKBH5 regulated SOX2, SMAD7, and MYC proteins in A549 and H1792 cells (Fig. 6A, B).

To investigate m6A modifications on SOX2, SMAD7, and MYC mRNA, we used firefly luciferase assay (Fig. 6C). We found their activities were significantly less than those of their mutant reporters, indicating that m6A represses gene activity. Moreover, overexpression of ALKBH5 increased luciferase activities of three m6A WT gene reporters, and also rescued the repressive activity mediated by LKB1-transfection. However, we observed no such effects for m6A mutant gene-fused reporters. Our m6A-RIP-qPCR assay showed significantly higher accumulation of m6A at mRNA fragments of all three genes than for the IgG control. However, this m6A enrichment was reduced by ALKBH5 overexpression. Also, ALKBH5 upregulation inhibited the increased m6A occupancy caused by LKB1 overexpression (Fig. 6D). Given that ALKBH5 overexpression or knockdown did not change the pre-mRNAs of SMAD7 and MYC (Fig. S8A–D), we predicted m6A reader protein YTHDF2 regulated their mRNA stability. Interestingly, actinomycin-D induced degradation of SMAD7, SOX2, and MYC mRNAs was partially prevented by silencing YTHDF2 in both A549 and H1792 cells. (Fig. S8E–J). RIP-qPCR assay showed that YTHDF2 also occupied SMAD7, SOX2, and MYC m6A regions, and this enrichment increased in the absence of ALKBH5 (Fig. 6E). Furthermore, YTHDF2 silencing rescued their gene expression upon LKB1 overexpression or ALKBH5 knockdown (Fig. 6F). Thus, ALKBH5-m6A-YTHDF2 signaling prevented SOX2, SMAD7, and MYC mRNA decay.

Next, we validated the underlying regulation of m6A by LKB1 using a panel of clinically relevant lung adenocarcinoma specimens. IHC staining showed that LKB1 loss increased the global 5-mC DNA methylation in lung cancer tissues with KRAS mutations, but not in tissues with WT KRAS (Fig. 7A). Similar pattern was found for both DNA methylation and mRNA expression of ALKBH5 (Fig. 7B, C). ALKBH5 protein expression negatively correlated with global m6A level, as well as the m6A levels for SOX2, SMAD7, and MYC genes only in KRAS mutant tissues (Fig. 7D–F). Furthermore, m6A levels of these three genes were negatively correlated with their mRNA expression (Fig. 7G–I). Notably, the protein levels of SOX2, SMAD7, and MYC were significantly increased by loss of LKB1 and correlated with ALKBH5 expression in KRAS mutant tissues (Fig. 7J, K). Thus, our clinical data further confirmed that LKB1 loss increased the DNA methylation and expression level of ALKBH5, which resulting in the demethylation of m6A modification on SOX2, SMAD7, and MYC, and maintaining their expression for KRAS mutant lung cancer.

We randomly obtained fresh and paraffin-embedded lung adenocarcinoma specimens from 72 patients who underwent tumor resection surgery between January 2016 and September 2019 at the Cancer Hospital of Harbin University Medical College, China. Two pathologists performed blinded histological confirmation by hematoxylin and eosin (H&E) staining. KRAS mutations were detected by Cobas KRAS Mutation Test, which could detect 19 mutation sites based on a real-time PCR method. Exclusion criteria were patients who had received preoperative chemotherapy, radiation therapy, or had multiple metachronous or metastatic lesions. Clinical characteristics of patients were retrospectively analyzed and summarized in Table. We collected all clinical samples with informed consent according to Health Insurance Portability and Accountability Act (HIPAA)-approved protocols. This study was approved by the ethical review board of Cancer Hospital of Harbin University Medical College on October 22, 2018. All of the patients were given and accepted informed consent form prior to their enrollment. S1

MRC-9, H1299, H1650, H1703, H1795, H1792, A549, DLD-1, SW480, Panc-1, and MIA PaCa-2 cell lines were purchased from ATCC, USA. Cells were cultured in medium according to the manufacturer’s instructions and grown in a humidified incubator at 5% CO₂. All cell lines were authenticated and confirmed negative for mycoplasma contamination by providers.

To knockdown endogenous gene expression, we purchased si-RNAs targeting human LKB1 (sc-35816), ALKBH5 (sc-93856), YTHDF2 (sc-78661), and CTCF (sc-35124) from Santa Cruz and transiently transfected these si-RNAs into lung cancer cells for 48 h using Lipofectamine RNAiMAX (Invitrogen). We ordered LKB1 (#8590) and ALKBH5 (#38073) overexpressing vectors from Addgene and transfected them into human cancer cells using Lipofectamine 2000 (Invitrogen) for 24 h. Si-RNA-A (sc-37007, Santa Cruz) and c-Flag pcDNA3 (Plasmid #20011, Addgene) were served as transfect controls for genes knockdown and overexpression, respectively.

We extracted total cellular and tissue RNA using TRIzol Reagent (Thermo Fisher Scientific, USA) and used 1 µg total RNA for reverse transcription using the iScript^™ cDNA Synthesis Kit (Bio-Rad). qRT-PCR was performed with 2× SYBR Green qPCR Master Mix (Bimake, USA) with a CFX96 Touch^™ Real-Time PCR detection system (Bio-Rad Inc. USA). We calculated the relative gene expression using the comparative CT method and β-actin RNA sequences as a control. Primer sequences were listed in Table S2.

We lysed total protein from treated cells using RIPA buffer (Cell Signaling Technology) supplemented with Protease Inhibitor Cocktail (Thermo Fisher; 78430). Western blotting was performed as previously described⁴¹, and the primary antibodies included GAPDH (Santa Cruz, sc-137179), LKB1 (Santa Cruz, sc-32245), ALKBH5 (Proteintech, 16837-1-AP), CTCF (Santa Cruz, sc-271474), METTL3 (Abcam, ab195352), METTL14 (Abcam, ab220030), FTO (Abcam, ab126605), WTAP (Proteintech, 60188-1-Ig), SOX2 (Cell signaling, #3579), SMAD7 (Santa Cruz, sc-11392), and MYC (Cell signaling, #13987). We performed densitometric analyses of band intensity using ImageQuant TL 8.2 image analysis software (GE Healthcare Life Sciences, USA) and GAPDH was as an internal control.

IHC analysis of paraffin-embedded lung cancer tissues containing primary tumors and matched normal lung tissues as previously described^42,43. In brief, we de-paraffinized and rehydrated human tissue sections to retrieve antigens, and then incubating in 1% hydrogen peroxide. After blocking, we applied primary antibodies to the slides at 1:500 (anti-LKB1, METTL3, METTL14, WTAP, and FTO antibodies), 1:200 (anti-ALKBH5 antibody), and 1:1000 (anti-m6A and 5-mC antibody) dilutions and incubated at 4 °C overnight. We stained slides with EnVision+ Dual Link System-HRP (Dako) for 1 h at room temperature. Images of stained cells in four random fields were captured by using an optical microscope (Olympus, Japan). Relative protein expression was evaluated by a Histoscore (H-score) system. The results were evaluated by two independent pathologists.

We performed immunofluorescence staining on cultured A549 cells transfected with siRNA or pcDNA. After treatment, we fixed cells with 4% paraformaldehyde (PFA), permeabilized with 0.3% Triton/phosphate-buffered saline (PBS), blocked in 1% bovine serum albumin (BSA), and then incubated with the indicated primary antibody for LKB1 (Santa Cruz, sc-32245), ALKBH5 (Proteintech, 16837-1-AP), or m6A (Synaptic Systems, 202111) overnight at 4 °C. The next day, cells were washed and incubated with secondary antibody conjugated to Alexa Fluor 555 (donkey anti-rabbit, Invitrogen) for 1 h at room temperature. We used DAPI (Sigma) staining to label nuclei. Fluorescence was observed under a Leica SP8 confocal laser scanning microscope.

For the ALKBH5 promoter reporter assay, we generated serial ALKBH5 promoter reporters by PCR amplification and inserted into pGl3-basic plasmid, as we previously reported⁴¹. We deleted the CTCF region using the Q5 Site-Directed Mutagenesis Kit (NEB) per the manufacturer’s protocol. Firefly luciferase activity was used to evaluate the effect of m6A modification on SOX2, SMAD7, and MYC activation. We used the pmirGLO Dual-Luciferase miRNA target expression vector from Promega to construct the reporter plasmid, which contained both a firefly luciferase and a Renilla luciferase. We constructed mutant three reporter plasmids by replacing the adenosine bases within the m6A consensus sequences with cytosine. All constructs were confirmed by Sanger sequencing. Nucleotide sequences of primers were in Table S2. A549 cells grown in 96-well plates were transfected with reporter vectors and SV-40-Renilla-Luc in the presence of Lipofectamine 2000 Reagent (Invitrogen). Luminescence was measured with the Dual-Luciferase Reporter Assay System (Promega). We performed experiments for each vector as biological triplicates with six technical repeats.

We measured total cellular and tissue 5-mC in DNA and m6A in mRNA levels by the MethylFlash Methylated DNA 5-mC Quantification Kit (Colorimetric) and EpiQuik m6A RNA Methylation Quantification ELISA kit (Colorimetric) (Epigentek Group Inc.), respectively. Total genomic DNA and RNA were isolated using the PureLink Genomic DNA Mini Kit and TRIzol Reagent (Thermo Fisher Scientific). We used 200 ng of DNA or RNA for additional global 5-mC and m6A measurement according to the manufacturer’s instructions. Measurements were performed in triplicate.

For DNA methylation analysis, we used BGS method, as previously described^31,44. Briefly, 500 ng of genomic DNA was bisulfite converted using the BisulFlash DNA Modification Kit (Epigentek) following the manufacturer’s instructions. We amplified the fragment containing CTCF peak region of ALKBH5 promoter using primers listed in Table S2. PCR products were purified and cloned into a pCR4 TOPO vector using the TOPO TA Cloning Kit (Thermo Fisher Scientific, Rockford, IL, USA). We isolated and sequenced plasmid DNA from ten randomly selected clones (Genewiz, Piscataway, NJ, USA).

To confirm BGS results, we performed MeDIP using the Methylamp Methylated DNA Capture Kit (EpiGentek) according to the manufacturer’s instructions. Briefly, we extracted cellular and tissue chromatin DNA and digested to ~150–700 bp using a micrococcal nuclease (CST). The fragmented DNA was immunoprecipitated with anti-5-mC (Abcam, ab10805) at room temperature for 2 h. After washing and purifying the DNA, we quantified the methylation status using qPCR. Primers were listed in Table. S2

We extracted total RNA from treated A549 cells or tumor tissues, and incubated with DNase according to the TURBO DNA-free TM Kit (Thermo Fisher) protocol to avoid DNA contamination. Then, we chemically fragmented 1 µg/µl RNA and incubated with m6A antibody to immunoprecipitate according to the standard protocol for the EpiMark N6-Methyladenosine Enrichment Kit (NEB). Enrichment of m6A containing mRNA was then analyzed using qRT-PCR. Primers targeting m6A-enriched regions of SOX2, SMAD7, and MYC were listed in Table S2.

For cell colony formation assay, A549 or H1792 cells were transfected with siRNA or pcDNA for 24 h and seeded into 6-well plates (500/well). After 1 week, we fixed formed colonies and stained with 0.1% crystal violet in 20% methanol, and counted colonies consisting of at least 50 cells. For the migration assay, we resuspended 5 × 10⁵ cells in Opti-MEM Reduced Serum Media (Invitrogen) and seeded into the upper chamber of a transwell apparatus (8.0 μm, BD Biosciences), and added complete medium to the bottom chamber to provide chemoattractants for migration. To count migrated cells, we captured images of stained cells in five random fields by using an optical microscope (Olympus, Japan) and counted samples in triplicate.

The differentially expressed genes (DEGs) of m6A modulators (METTL3, METTL14, WTAP, FTO, and ALKBH5) and readers (YTHDF1, YTHDF2, YTHDF3, YTHDC1, and YTHDC2) between KL and K-lung cancer cell lines or tissues from the databases of the Cancer Genome Atlas (TCGA, https://portal.gdc.cancer.gov/↗), Mouse Tumor Biology (MTB, http://tumor.informatics.jax.org/mtbwi/index.do↗), and Cancer Cell Line Encyclopedia (CCLE, https://portals.broadinstitute.org/ccle/home↗).

Based on the NCBI-GEO database (GEO: GSE76367↗), we determined the number of m6A-seq fragments mapped to each gene using HT-Seq. We overlapped the core m6A peaks based on functional m6A signal enriched around the stop codon (−5 to +5 kbp) of mRNAs and canonical m6A peaks^45,46. Then, core m6A peaks were subjected to functional enrichment analysis by Ingenuity Pathway Analysis (IPA) (http://www.qiagen. com/ingenuity) and KEGG pathway (https://www.genome.jp/kegg/↗).

All experiments were repeated at least four times, unless otherwise stated in the figure legend. We performed statistical analyses using the SPSS v20.0 software (SPSS Inc., Chicago, IL), and used Student’s t-test (two-tailed) or one-way ANOVA analysis followed by Tukey’s, Sidak’s, or Bonferroni test to assess statistical significance between or among groups. We calculated the survival rate using the log-rank (Mantel–Cox) test. Normality was assumed and variance was compared between or among groups. All numerical data were presented as mean ± standard deviation (SD) and a p value of <0.05 was considered significant.

The study was performed in accordance with the Declaration of Helsinki. All of the patients were given and accepted an informed consent form prior to their enrollment. The use of human patient samples was approved by the Cancer Hospital of Harbin University Medical College Ethics Committee.