m6A demethylase ALKBH5 promotes tumor cell proliferation by destabilizing IGF2BPs target genes and worsens the prognosis of patients with non-small-cell lung cancer

Resected NSCLC samples from Hamamatsu University School of Medicine and Seirei Mikatahara General Hospital were collected and named as the HUSM cohort. Tissue microarray (TMA) sections were analyzed using immunohistochemistry (IHC) as previously described [45]. Cores of insufficient quality or quantity were excluded from the analysis. Antibodies for ALKBH5 (HPA007196, Atlas Antibodies, Stockholm, Sweden) and FTO (Ab124892↗, Abcam, Cambridge, UK) were diluted at 1:400, whereas those specific for EGFR E746-A750 deletion (#2085, D6B6, Cell Signaling Technology [CST], Danvers, MA, USA) and EGFR L858R mutant (#3197, 43B2, CST) were diluted at 1:100, followed by incubation at room temperature for 0.5 h. Protein expression levels were then assessed using the H-score, which was calculated by multiplying the percentage of stained tumor area (0%–100%) by the staining intensity (scored on a scale of 0–3) to yield a value ranging from 0 to 300.

We used the lung cancer database in the Kaplan–Meier plotter (http://kmplot.com/analysis/index.php?p=service&cancer=lung↗) to analyze the association between prognosis and ALKBH5 and FTO mRNA expression in NSCLC cohorts. Data were downloaded on December 10, 2020. Kaplan–Meier curves for overall survival (OS) were generated and stratified according to the median expression of each mRNA. To assess the mRNA expression of ALKBH5 and FTO, data from the Cancer Genome Atlas (TCGA) (NSCLC, Provisional) were downloaded from cBioPortal (http://www.cbioportal.org/↗) on November 11, 2019. Expression data were obtained in the form of RNA-seq by Expectation Maximization (RSEM).

Cells grown on coverslips were fixed with 4% paraformaldehyde and permeabilized with 0.1% Triton X-100. After blocking with 5% bovine serum albumin in PBS (−) at room temperature for 1 h, the cells were probed with primary antibodies against ALKBH5 (HPA007196, Atlas Antibodies) and then incubated with a Goat anti-Rabbit IgG (H + L) Cross-Adsorbed Secondary Antibody, Alexa Fluor 546 (#A-11010, Thermo Fisher Scientific, Waltham, MA, USA). Nuclei were stained with ProLong® Gold Antifade Reagent with DAPI (#8961, CST), after which the cells were imaged via fluorescence microscopy using z-stack image reconstructions (BZ-9000; Keyence, Osaka, Japan).

The human lung cancer cell lines H1299, H460, H2087, A549, ABC1, and H358 and the human immortalized cell lines BEAS2B and HEK293 were obtained from Health Science Research Resources Bank (Osaka, Japan) or the American Type Culture Collection (Manassas, VA, USA). PC3 and PC9 lung cancer cells were purchased from the Japanese Collection of Research Bioresources Cell Bank (Osaka, Japan) and RIKEN BioResource Center (Tsukuba, Japan), respectively, whereas ACC-LC176 cells were a kind gift from Dr. Takashi Takahashi (Nagoya University). RERF-LC-MS, HLC-1, and LC-2/ad were kind gifts from Dr. Toshiro Niki (Tokyo University). Lung cancer cell lines were cultured in RPMI1640 medium (R8758, Thermo Fisher Scientific), whereas HEK293 cells were cultured in DMEM (D5796 MERCK, Darmstadt, Germany) containing 10% (vol./vol.) fetal bovine serum (FBS), 100 IU/mL penicillin G, and 100 µg/mL streptomycin. LHC9 (12680013, Thermo Fisher Scientific) was also used as a medium for BEAS2B cells. Cells were maintained in a 5% CO₂ and 95% air incubator at 37 °C. Silencer Select Pre-designed siRNA for ALKBH5 (siALKBH5: s29743, s29744, s29745, Thermo Fisher Scientific), FTO (siFTO: s28147, s28148, s28149, Thermo Fisher Scientific), IGF2BP1 (siIGF2BP1: s20916, s20917, Thermo Fisher Scientific), IGF2BP2 (siIGF2BP2: s20922, s20923, Thermo Fisher Scientific), IGF2BP3 (siIGF2BP3: s20919, s20920, Thermo Fisher Scientific), YTHDF2 (siYTHDF2: s28147, s28148, Thermo Fisher Scientific), CDKN1A (siCDKN1A: s145, s147, Thermo Fisher Scientific), TIMP3 (siTIMP3: s14147, s14148, Thermo Fisher Scientific), and Silencer Select Negative control (siNC: 4390843, Thermo Fisher Scientific) were purchased for transient knockdown. More than two different sequences were used for one target gene to minimize off-target effects. Cells were cultured for 24 h before transfection, after which they were transfected with 15 nM of final siRNA concentrations using Opti-MEM (31985070, Gibco, Dublin, Ireland) and Lipofectamine® 2000 (11668019, ThermoFisher). The cells were then used for further assays at 48–96 h after transfection. When no siRNA sample number was available, siRNA no. 1 (#1) and siRNA no. 3 (#3) were pooled for ALKBH5 unless otherwise specified. siIGF2BP1, siIGF2BP2, siIGF2BP3, siYTHDF2, siCDKN1A, and siTIMP3 were pooled for all transfections.

The retroviral plasmid pRetroX-TetOne puro (634307, Clontech, Mountain View, CA, USA) was amplified using NEB Stable competent Escherichia coli (high efficiency) (C3040H, NEW ENGLAND BioLabs, Ipswich, MA, USA). The full-length ALKBH5 sequence (NM_017758↗), which was confirmed using Sanger sequencing, was subcloned into pRetroX-TetOne puro vector using EcoRI and BglII restriction sites (pRetroX-TetOne puro-ALKBH5). Retroviral supernatants were produced using the GP2-293 packaging cell line (Clontech), in which pRetroX-TetOne puro empty vector or pRetroX-TetOne puro-ALKBH5 were each cotransfected with the envelope vector VSV-G using Xfect transfection reagent (Clontech). BEAS2B, HEK293, PC9, and A549 cells were transfected for 24 h using 4 µg/mL polybrene (H9268, Sigma-Aldrich, St. Louis, MO, USA). Puromycin selection (0.5–1.5 µg/mL) began 48 h after transfection and lasted for 3 days until all nontransfected cells had died. In subsequent experiments, Retro-X cells were induced with 0.1–100 ng/mL of doxycycline (DOX) diluted in culture media upon cell seeding for 24–96 h. Cells transfected with pRetroX-TetOne puro-ALKBH5 with 100 ng/mL DOX were designated as ALKBH5-overexpressed (OE) cells, whereas those infected without DOX were designated as negative control (NC) unless otherwise noted.

Total RNA was extracted using the RNeasy Plus Mini Kit (#74136, QIAGEN, Hilden, Germany) according to the manufacturer’s instructions, with the total RNA concentration calculated using Nanodrop (NanoDrop1000, Thermo Fisher Scientific). cDNA was synthesized from 1 µg of total RNA using the ReverTra Ace qPCR RT Master Mix (FSQ-201, TOYOBO) or SuperScript III Reverse Transcriptase (1080044, Thermo Fisher Scientific) according to the manufacturer’s instructions. qPCR reactions were performed on a Step One Plus Real-Time PCR System (Applied Biosystems, Thermo Fisher Scientific) using the THUNDERBIRD qPCR Mix (QPS-201, TOYOBO, Osaka, Japan). The relative RNA expression levels were calculated using the ΔΔCt method, with the levels normalized to glyceraldehyde 3-phosphate dehydrogenase (GAPDH) mRNA. All amplicons were confirmed as a single product using agarose gel visualization and/or melting curve analysis. The applied primer sequences are listed in Table. S1

Total protein lysates were extracted from whole cells using 1× sodium dodecyl sulfate (SDS) sample buffer. The Pierce BCA Protein Assay Kit (Cat#23225, Thermo Fisher Scientific) was used to determine the protein concentration. All proteins were separated using SDS-polyacrylamide gel electrophoresis and transferred to PVDF Blotting Membrane (P 0.45, A29532146, GE healthcare Life science, Chicago, IL, USA) using the Trans-Blot Turbo Cassette (Bio-Rad, Hercules, CA, USA). Blocking One (03953, Nacalai, Kyoto, Japan) or 5% skimmed milk were used for blocking. Primary antibodies for ALKBH5 (1:1000 dilution, HPA007196; Atlas Antibodies), FTO (1:1000 dilution, Ab124892↗; Abcam), IGF2BP1 (1:1000 dilution, 22803-1-AP; Proteintech), IGF2BP2 (1:1000 dilution, 11601-1-AP; Proteintech), IGF2BP3 (1:2000 dilution, 14642-1-AP; Proteintech), YTHDF2 (1:400, Ab170118↗, Abcam), TIMP3 (1:3000 dilution, Ab39184; Abcam), p21 (1:1000 dilution, A1483; ABclonal, Woburn, MA, USA), E2F1 (1:500 dilution, A2067; ABclonal), CCNG2 (0.2 μg/mL, Ab251826↗; Abcam), p53 (1:200 dilution, Sc-126; SANTA CRUZ BIO TECHNOLOGY, Dallas, TX, USA), and GAPDH (1:1000 dilution, Ab8245; Abcam) were incubated for overnight at 4°C. Secondary antibodies for rabbit (1:20000 dilution, NA9340; GE healthcare Life science) or mouse (1:20000 dilution, NA9310; GE Healthcare Life Science) were incubated at room temperature with 1%–5% skimmed milk for 1 h. Enhanced chemiluminescence (Pierce ECL Plus Substrate or West Atto Ultimate Sensitivity Substrate, Thermo Fisher Scientific) was used to visualize the protein bands using ChemiDocTouch (Bio-Rad).

Cells were seeded into 96-well plates with 3000 cells per well after 48 h of knockdown or overexpression. Cell proliferation was monitored using Cell Counting Kit-8 (CCK-8; Dojindo, Kumamoto, Japan) according to the manufacturer’s protocol. Thereafter, the cells were incubated with 10% CCK-8 for 1 h, followed by absorbance assessment at 450 nm in each well via spectrophotometry (Synergy HT, BioTek, Winooski, VT, USA) every 24 h.

Cell migration was evaluated using a 24-well plate with cell culture inserts (353097, Falcon, Mexico City, Mexico) containing a filter with 8 μm-diameter pores. Briefly, after serum starvation for 24 h with 0.1% FBS-containing RPMI1640 medium, 1 × 10⁵ cells resuspended in 500 μL of RPMI1640 medium (Gibco) were seeded into the upper chamber, after which RPMI1640 medium containing 10% FBS was placed in the lower compartment of the chamber. After incubation for 16 h, the upper surface of the membrane was wiped with a cotton-tipped applicator to remove nonmigrating cells, whereas the migrating cells on the lower surface were fixed with cold methanol and stained with 0.5% crystal violet. Migrating cells were automatically counted in three random microscopic fields using the Hybrid Cell Count software (BZ-II Analyzer, Keyence, Osaka, Japan).

To assess cell migration, 2 × 10⁵ cells were seeded into 6-well plates. Thereafter, cells were incubated in 5% CO₂ at 37 °C for 48 h and an additional 24 h with 0.1% FBS-containing RPMI1640 medium. A wound was scratched into the cells using a 200-μL plastic tip and washed with PBS (−). The cells were then incubated in RPMI1640 containing 10% FBS. The relative distance of the scratches was observed under an optical microscope (IX53, Olympus, Tokyo, Japan) at 3–6 time points after wounding and assessed using the Image J software.

Cell Cycle Assay Solution Blue (C549, Dojindo) was used to measure the cell cycle according to the manufacturer’s instructions. Briefly, treated cells were synchronized at the G1 phase through serum starvation with 0.1% FBS-containing medium for 48 h. At 24 h after the release of serum starvation, the treated cells were collected, washed with PBS (−), and incubated with 5 μL cell cycle assay solution for 15 min at 37 °C. Thereafter, DNA content was determined based on staining intensity using a Gallios flow cytometer (Beckman Coulter, Miami, FL, USA). The annexin V-FITC Apoptosis Detection Kit (15342-54 Nacalai) was used to detect apoptosis by measuring annexin V and propidium iodide (PI)-positive cells following the manufacturer’s instructions. Briefly, cells were incubated for 96 h after siRNA transfection. To induce apoptosis, the cells were exposed to either 7.5 μM of gefitinib (078-06561, FUJIFILM) or 10 μM of cisplatin (P4394, Sigma-Aldrich) alone for 48 h after siRNA transfection. The treated cells were collected, washed with PBS (−), and incubated with 5 μL of annexin V-FITC solution and 5 μL of PI solution for 15 min. Thereafter, apoptotic cells were determined using a Gallios flow cytometer. Results were analyzed using the FlowJo software (Becton, Dickinson, Franklin Lakes, NJ, USA), after which the extent of apoptosis and cell cycle distribution were determined.

Cancer cells were incubated for 48 h after siRNA transfection. Cells were treated with actinomycin D at a final concentration of 5 μg/mL. Total RNA was extracted at 0, 2, 4, and 6 h after adding actinomycin D. The remaining CDKN1A and TIMP3 mRNA was measured through quantitative real-time PCR and normalized to RPL32 mRNA, which has a half-life of 25 h.

PolyA-enriched RNA was extracted using PolyATract mRNA isolation systems (#Z5310 Promega, Madison, WI, USA) according to the manufacturer’s instructions. PolyA-enriched RNA concentration was calculated using Qubit 2.0. The polyA-enriched RNA was enzymatically hydrolyzed using 8-OHdG Assay Preparation Reagent Set (292-67801, FUJIFILM Wako Pure Chemical Corporation, Tokyo, Japan). Technically, 100 ng of polyA-enriched RNA was digested using 5.7 μL of acetic acid buffer and 3 μL of Nuclease P1 included in the 45-μL sample containing nuclease-free water at 37 °C for 30 min, followed by incubation with 6 μL of Tris Buffer and 0.3 μL of alkaline phosphatase at 37 °C for 30 min. After digestion, the sample was centrifuged at 14,000 × g and 4 °C for 20 min using a Nanosep 3 K Omega centrifugal device (Pall Corporation, Port Washington, NY, USA) according to a previously published method [46].

As an internal standard, N⁶-methyladenosine-d3 (m⁶A-d3; M275897, Toronto Research Chemicals, Toronto, Canada), which is a stable isotope of N⁶-methyladenosine labeled with three deuterium atoms on the N⁶-methyl group, was added to the nucleosides obtained via digestion of polyA-enriched RNA. These nucleosides were separated using an Acquity UPLC HSS T3 column (2.1 mm × 100 mm; Waters, Milford, CT, USA) with 0.1% (vol./vol.) formic acid in water as mobile phase A and methanol as mobile phase B at a flow rate of 200 μL/min in a linear gradient elution of 5%–60% B from 0 to 7 min. Standard compounds of adenosine (A; A9251, Sigma-Aldrich), N⁶-methyladenosine (m⁶A; A170736, Sigma-Aldrich), and m⁶A-d3 were used to confirm the nucleoside-to-base ion mass transitions of 268.1–136.4 (A), 282.2–150.2 (m⁶A), and 285.2–153.2 (m⁶A-d3). Peak areas of A, m⁶A, and m⁶A-d3 in the nucleosides digested from polyA-enriched RNA were calculated using the column retention time of the standard compounds using Analyst 1.6.1 software (AB SCIEX, Foster City, CA, USA). The m⁶A level was quantified as the ratio of m⁶A to A or m⁶A-d3 based on the calibrated concentrations.

Total RNA was extracted from ALKBH5-knockdown or control PC9 cells 96 h after transfection. RNA samples were used for global gene expression profiling on human Clariom S Assay microarrays (Thermo Fisher Scientific, Wilmington, DE, USA), which include 24351 genes. All microarray analyses were entrusted to Filgen Inc. (Aichi, Japan). A total RNA quality control check was performed using a NanoDrop ND-1000 (Thermo Scientific) and an Agilent 2100 Bioanalyzer. Using the Gene Chip TM WT PLUS Reagent Kit, fragmented and labeled cDNA samples were prepared from 250 ng of total RNA according to the manufacturer’s instructions (Gene Chip TM WT PLUS Reagent Kit User Manual). Thereafter, 100 μL of hybridization solution was prepared using 73 μL of Hybridization Master Mix and 2.3 μg of fragmented and labeled cDNA. The array was incubated using the Gene Chip TM Hybridization Oven 645 at 45°C for 16 h (60 rpm). The array was cleaned using the Gene Chip TM Fluidics Station 450 and scanned using the Gene Chip TM Scanner 3000 7G according to the manufacturer’s instructions [Gene Chip TM Command Console (AGCC) 4.0 User Manual]. The Microarray Data Analysis Tool version 3.2 (Filgen, Aichi, Japan) was used for data normalization and subsequent processing. Differentially expressed mRNAs were identified using a set cutoff (fold change > 1.5 or <0.67; P < 0.01). Gene set enrichment analysis (GSEA) was performed to examine the gene sets regulated by ALKBH5 knockdown (http:/software.broadinstitute.org/gsea/omdex.jsp). For analysis, the false discovery rate (FDR) based on gene set permutation was used. Microarray data has been deposited in the Gene Expression Omnibus (GEO) at the National Center for Biotechnology Information (NCBI) (accession number GSE165453↗).

Unfragmented total RNA was extracted from ALKBH5-knockdown or control PC9 cells at 96 h after transfection and quantified using the NanoDrop ND-1000. RNA samples were used for global m⁶A expression profiling on an Arraystar Human mRNA&lncRNA Epitranscriptomic Microarray (8 × 60 K; Arraystar), which includes 44,122 protein-coding mRNAs and 12,496 long noncoding RNAs. Microarray analyses were entrusted to Arraystar Inc. (Rockville, MD, USA). Sample preparation and microarray hybridization were performed based on Arraystar’s standard protocols. Briefly, total RNAs were immunoprecipitated with an anti-m⁶A antibody (Synaptic Systems, 202003). The “immunoprecipitated (IP)” and “supernatant (Sup)” RNAs were labeled with Cy5 and Cy3, respectively, as cRNAs in separate reactions using the Arraystar Super RNA Labeling Kit. The cRNAs were combined and hybridized onto Arraystar Human mRNA&lncRNA Epitranscriptomic Microarray (8 × 60 K, Arraystar). After washing the slides, the arrays were scanned in two-color channels using an Agilent Scanner G2505C. Agilent Feature Extraction software (version 11.0.1.1) was used to analyze acquired array images. Raw intensities of IP (Cy5-labeled) and Sup (Cy3-labeled) were normalized with an average of log2-scaled Spike-in RNA intensities. The “m⁶A methylation level” was calculated to determine the percentage of modification based on the IP (Cy5-labeled) and Sup (Cy3-labeled) normalized intensities. “m⁶A quantity” was calculated to determine the amount of m⁶A methylation based on the IP (Cy5-labeled) normalized intensities. Differentially m⁶A-methylated RNAs between both comparison groups were identified by filtering with a fold change of >1.5 or <0.67 (P < 0.01) through the unpaired t-test. Microarray data had been deposited in the GEO at the NCBI (accession number GSE165454↗).

ALKBH5-knockdown or control lung cancer cells were used for methylated RNA immunoprecipitation assay. The Magna MeRIP m⁶A kit (catalog no.17-10499, Millipore, Burlington, MA, USA) was used according to the manufacturer’s protocol. Briefly, the polyA-enriched RNA was fragmented into 100–200 nucleotides incubated with RNA fragmentation buffer for 55 s (CS220011, Millipore). The size of polyA-enriched RNA fragments was optimized using the Agilent 4200 TapeStation (Agilent Technologies, Santa Clara, CA, USA). We used 0.5 μg of fragmented polyA-enriched RNA as input control and 5 μg of fragmented polyA-enriched RNA for m⁶A mRNA immunoprecipitation, followed by incubation with m⁶A antibody (MABE1006, Millipore)- or mouse IgG-conjugated Protein A/G Magnetic Beads in 500 μL 1× IP buffer supplemented with RNase inhibitors at 4 °C overnight. Methylated RNAs were immunoprecipitated with beads, eluted via competition with free m⁶A, and purified using the RNeasy kit (Qiagen). Moreover, modification of m⁶A toward particular genes was determined using qPCR analysis with specific primers [primers for the positive control region (stop codon, EEF1A1+) or NC region (exon 5, EEF1A1−) of human EEF1A1 was included in the Magna MeRIP m6A kit]. To design primers for MeRIP qPCR, m⁶A sites of specific genes were predicted using the sequence-based RNA adenosine methylation site predictor algorithm (http://www.cuilab.cn/sramp↗) [47]. We focused on the potential m⁶A sites in the 3′ UTRs near the stop codon and designed primers to ensure that the target sequences were present in these sites with a limited length of 120 nt. Self-designed primers for MeRIP qPCR are listed in Table S1.

Discrete variables were expressed as numbers (percentages), whereas continuous variables were expressed as means ± standard deviations (SDs) unless otherwise specified. The Mann–Whitney U test was used to compare continuous individual samples, whereas Student’s t-test was applied to compare continuous experimental data. Fisher’s exact test for independence was used to compare categorical data between groups. The Wilcoxon matched-pairs signed-rank test was used to compare two corresponding groups. Spearman’s correlation coefficient was used for correlation analysis. Kaplan–Meier curves with log-rank tests were used to analyze survival. Accordingly, OS was defined as the duration from baseline to the date of death, whereas recurrence-free survival (RFS) was defined as the duration from baseline to the recurrence date. Univariate and multivariate Cox proportional hazards models were applied to generate hazard ratios (HRs) for death while adjusting for other potential confounding factors. Cell proliferation and RNA stability assays were analyzed using two-way analysis of variance. Statistical analyses were performed using GraphPad Prism Version 8 (GraphPad Software, San Diego, CA, USA) and EZR software (Saitama Medical Center, Jichi Medical University, Saitama, Japan), with P values of <0.05 indicating statistical significance.

To investigate the impact of ALKBH5 and FTO in NSCLC, we examined the mRNA expression levels of ALKBH5 and FTO in non-cancerous lung tissues and NSCLC tissues using TCGA data. Accordingly, our results showed no significant difference in ALKBH5 mRNA expression between non-cancerous and cancerous tissues. By contrast, our findings showed that NSCLC had a significantly lower FTO mRNA expression than non-cancerous tissues (Fig. 1A). We subsequently investigated the protein expression levels of ALKBH5 and FTO in non-cancerous lung alveolar tissue and corresponding NSCLC tissues using TMA of patient samples. Furthermore, our results showed that cancerous tissues had significantly higher H-scores for ALKBH5 and FTO than non-cancerous tissues (Fig. 1B). ALKBH5 and FTO expression were evaluated in immortalized bronchial epithelial cells (BEAS2B) and lung cancer cell lines. Consequently, qPCR analysis demonstrated that ALKBH5 mRNA expression was higher in lung cancer cell lines except for LC-2/ad and RERF-LC–MS, whereas FTO mRNA expression was lower in lung cancer cell lines except for HLC-1, ABC1, and PC3 (Fig. 1C). Western blot analysis demonstrated that ALKBH5 was endogenously expressed in all lung cancer cell lines and FTO was expressed in almost all lung cancer cell lines except HLC-1 (Fig. 1D). IHC analysis showed that ALKBH5 and FTO were mainly localized in the nucleus of the cells (Fig. 1E). Furthermore, immunofluorescence analysis showed that ALKBH5 was localized in the nucleus of PC9 cells overexpressing ALKBH5 (Fig. 1F). We analyzed the clinical characteristics of 627 NSCLC cases used in IHC of TMA in the context of ALKBH5 or FTO expression in tumors of the HUSM cohort (Table S2). The median age was 68 (range, 23–88) years; 430 (68.6%) patients were male and 184 (29.3%) had never smoked. The tumors were histologically classified as adenocarcinoma (n = 413, 65.9%), squamous cell carcinoma (n = 170, 27.1%), or other histological types (n = 44, 7.0%). A total of 395 (63.0%) patients had stage Ι disease, whereas 127 (20.3%) cases had EGFR mutations. Postoperative adjuvant chemotherapy was prescribed to 258 (41.1%) patients. The median H-score values for ALKBH5 and FTO expression were 110 (0–225, range) and 65 (0–281), respectively, with high ALKBH5 protein expression being correlated with high FTO protein expression (r = 0.41) (Fig. S1A). Based on the median value, cases were divided into “high” and “low” expression groups, after which their association with clinical data as well as prognostic significance was examined. Lymph node metastasis, chemotherapy, and EGFR status significantly differed depending on ALKBH5 expression, whereas tumor status, lymph node metastasis, pathological stage, chemotherapy, and EGFR status significantly differed depending on FTO expression. Kaplan–Meier curves showed that patients with high ALKBH5 expression had significantly worse survival than those with low ALKBH5 expression (Fig. 1G: log-rank p = 0.0009 for OS; Fig. S1B: log-rank p = 0.0008 for RFS). Conversely, Kaplan–Meier curves showed no significant difference in survival between the low and high FTO expression groups (Fig. 1G: log-rank p = 0.20 for OS, Fig. S1C: log-rank p = 0.07 for RFS). Univariate analysis revealed high ALKBH5 expression as a predictor of unfavorable OS (HR, 1.675; 95% CI, 1.230–2.521). Moreover, multivariate analysis of age, sex, smoking status, histology, pathological stage, and ALKBH5 expression revealed that ALKBH5 expression was an independent prognostic factor associated with unfavorable OS (HR, 1.468; 95% CI, 1.039–2.073) (Table S3). To validate the prognostic value of ALKBH5 and FTO in other cohorts of patients with NSCLC, the lung cancer database in the Kaplan–Meier plotter was used. Accordingly, Kaplan–Meier curves showed that patients with high ALKBH5 expression had a significantly worse survival than those with low ALKBH5 expression (Fig. S1D: log-rank p = 0.014 for OS). In contrast, Kaplan–Meier curves showed that patients with high FTO expression had significantly favorable survival compared with those with low FTO expression (log-rank p < 0.0001 for OS) (Fig. S1E). These observations suggested that ALKBH5 played a critical role in the poor prognosis of patients with NSCLC.

To investigate the impact of ALKBH5 and FTO deficiency in lung cancer cell function, ALKBH5 and FTO were knocked down in PC9 and A549 cells using small interfering RNA (siRNA). Based on knockdown efficacy, ALKBH5 siRNA no. 1 (siALKBH5#1) and siRNA no. 3 (siALKBH5#3) and FTO siRNA no.1 (siFTO#1) and siRNA no. 3 (siFTO#3) were used in subsequent knockdown experiments (Fig. 2A, B, and S2A). ALKBH5 knockdown significantly suppressed the proliferation of PC9 and A549 cells (Fig. 2C). By contrast, FTO knockdown showed no significant suppressive effects on the proliferation of PC9 and A549 cells (Fig. 2D). Thereafter, we assessed migration abilities in ALKBH5-knockdown cells. Accordingly, the transwell migration assay showed no significant reduction in the migratory PC9 and A549 cells (Fig. 2E). Moreover, the wound-healing assay showed that ALKBH5 knockdown promoted no significant reduction in the migration ability of PC9 and A549 cells (Fig. 2F, G). Together with the prognostic value of ALKBH5 in NSCLC, these observations suggested that ALKBH5 played a cancer-promoting role by regulating cell proliferation.

To subsequently examine the mechanism by which ALKBH5 knockdown suppressed cell proliferation, cell cycle, and apoptosis analyses were performed using flow cytometry. Accordingly, ALKBH5 knockdown significantly increased the number of PC9 cells in the G1 phase and reduced the number of PC9 cells in the G2/M phase with a consistent result of two different sequences of siRNAs (Fig. 3A, B). On the other hand, the number of A549 cells in the G1 phase did not show significant differences by ALKBH5 knockdown with siALKBH5#1 and significantly increased by ALKBH5 knockdown with siALKBH5#3. The number of A549 cells in the G2/M phase did not show significant differences by ALKBH5 knockdown with a consistent result of two different sequences of siRNAs (Fig. 3C, D). ALKBH5 knockdown increased the number of apoptotic PC9 cells (Fig. 3E, F). Furthermore, under drug-induced apoptosis via cisplatin and gefitinib administration, ALKBH5 knockdown also increased the number of apoptotic PC9 cells (Fig. 3G, H). ALKBH5 knockdown also increased the number of apoptotic A549 cells (Fig. 3I, J). Moreover, ALKBH5 knockdown increased the number of apoptotic A549 cells with cisplatin (Fig. 3K, L). Overall, the aforementioned data showed that ALKBH5 knockdown suppressed cell proliferation through G1 phase arrest and/or apoptosis induction in NSCLC cell lines.

To analyze whether ALKBH5 overexpression in immortalized cells promoted malignant changes in cell function, BEAS2B and HEK293 cells, which are immortalized cells, were infected with a doxycycline-inducible vector, pRetroX-TetOne puro-ALKBH5. ALKBH5 overexpression was confirmed in HEK293 and BEAS2B cells (Fig. 4A) and significantly enhanced HEK293 and BEAS2B cell proliferation (Fig. 4B, C). In contrast, ALKBH5 overexpression showed no significant effects on the migration ability of HEK293 (Fig. 4D, E) and BEAS2B cells (Fig. 4F, G). The aforementioned results provided further evidence that ALKBH5 played a cancer-promoting role by regulating cell proliferation.

To assess the amount of m⁶A in cells, a quantitative evaluation of m⁶A was performed via LC–MS/MS using polyA-enriched RNA extracted from PC9 cells with altered ALKBH5 gene expression (Fig. 5A). We investigated the technical variability that occurs when adenosine and N⁶-methyladenosine-d3 (m⁶A-d3) are used as internal standards. Although both adenosine (A) and m⁶A-d3 showed a strong positive correlation with m⁶A (r = 0.92 and r = 0.90), the measurement with m⁶A-d3 as the internal standard showed less technical variability than that with A as the internal standard (Fig. S3A–C). Hence, we used m⁶A-d3 as the internal control for subsequent experiments. ALKBH5 knockdown increased m⁶A modification in PC9 and A549 cells (Fig. 5B, C), whereas ALKBH5 overexpression reduced m⁶A modification in a doxycycline concentration-dependent manner in PC9 cells (Fig. 5D and S3D, E). Moreover, ALKBH5 overexpression reduced m⁶A modification regardless of the time that had elapsed after doxycycline addition (Fig. 5E). Furthermore, ALKBH5 overexpression reduced m⁶A modification in BEAS2B and HEK293 cells (Fig. 5F). The aforementioned results presented evidence suggesting that ALKBH5 alters the global m⁶A abundance in cells.

An expression microarray analysis was herein performed to investigate gene expression profiles in ALKBH5-knockdown PC9 cells with two different sequences of siRNA (ALKBH5#1 and ALKBH5#3). Differentially expressed genes (DEGs) were defined as those with a fold change of >1.5 or <0.67 (P < 0.01). A total of 697 DEGs were detected for ALKBH5#1 comprising 392 upregulated and 305 downregulated genes (Fig. 6A), whereas 1394 DEGs were detected for ALKBH5#3 comprising 803 upregulated and 591 downregulated genes (Fig. 6B). Moreover, 82 upregulated genes (Table S4) and 47 downregulated genes (Table S5) overlapped between ALKBH5#1 and ALKBH5#3 (Fig. 6C). Except for ALKBH5, genes associated with m⁶A modification described in a previous review [48] were not included in the overlapped DEGs (Table S6). GSEA with the hallmark gene set revealed that the PC9 cells transfected with siALKBH5#1 and siALKBH5#3 had a more enriched expression of genes involved in cell cycle, such as MYC_TARGETS_V2, P53_PATHWAY, and G2/M_CHECKPOINT, than those transfected with siNC (Fig. S4A, B). We selected 10 DEGs associated with cell proliferation or apoptosis based on their description in the NCBI gene summary [49–54] or by referring to previous literature [55–58] and confirmed the upregulation of E2F1, GADD45A, TIMP3, and CDKN1A and downregulation of CASP14 and CCNG2 by qPCR (Fig. 6D). The aforementioned results revealed that ALKBH5 regulated the expression of genes associated with cell proliferation.

We performed m⁶A-specific methylated RNA immunoprecipitation microarray analysis in PC9 cells on an Arraystar Human mRNA&lncRNA Epitranscriptomic Microarray to comprehensively examine whether differentially regulated genes were associated with m⁶A modification using unfragmented total RNA. The median methylation level in unfragmented total RNA was 50.4% (6.8%–94.5%) (Fig. S5A). A positive correlation was observed between the methylation level in unfragmented total RNA and the RNA length of each transcript (r = 0.35) (Fig. S5B). Moreover, a negative correlation was noted between the rate at which ALKBH5 knockdown increased m⁶A modification (methylation level in siALKBH5—methylation level in siNC) and methylation level at baseline (methylation level in siNC) (r = −0.35) (Fig. S5C). The volcano plot showed that 1 RNA was hypermethylated by ALKBH5#1 knockdown (fold change > 1.5, P < 0.01) (Fig. S5D), whereas 28 RNAs were hypermethylated by ALKBH5#3 knockdown (fold change > 1.5; P < 0.01) (Fig. S5E). No hypermethylated genes overlapped between ALKBH5#1 and ALKBH5#3 knockdown with a fold change threshold of >1.5 (P < 0.01) (Fig. S5F). GSEA showed no common hallmark gene set with an FDR q-value of < 0.25 for PC9 cells transfected with siALKBH5#1 and siALKBH5#3 compared with the control group (Fig. S5G).

Considering that the m⁶A levels of unfragmented RNAs regulated by ALKBH5 are affected by the baseline RNA length and endogenous m6A level, we performed methylated RNA immunoprecipitation (MeRIP) with m⁶A antibody using fragmented polyA-enriched RNA in ALKBH5-knockdown PC9 cells to investigate focal m⁶A alterations in the mRNA. The fragmentation conditions were optimized (Fig. S6A), and the m⁶A changes in the positive and NC were confirmed through qPCR using the primers included in the Magna MeRIP m⁶A Kit (Millipore) (Fig. S6B). To verify the accuracy of the MeRIP experiment, we selected MFAP5 out of the 11 hypermethylated genes (>1.5 fold change and P < 0.05) in the human mRNA&lncRNA Epitranscriptomic Microarray (Table S7) and analyzed the m⁶A target site via qPCR. Specific primers were designed for the predicted m⁶A-harboring regions, and MeRIP qPCR confirmed that ALKBH5 knockdown increased m⁶A levels in the 3′ UTR of MFAP5 (Fig. S6C). Thereafter, MeRIP qPCR was performed in six DEGs verified using qPCR with ALKBH5-knockdown PC9 cells. Our results showed increased m⁶A levels in the 3′ UTRs of CDKN1A, TIMP3, E2F1, and CCNG2 in ALKBH5-knockdown PC9 cells (Fig. 6E–G). ALKBH5 knockdown with two different sequences of siRNAs also increased the m⁶A level in the 5’UTR of CDKN1A (Fig. 6F). The aforementioned results of the MeRIP qPCR suggest that ALKBH5 targeted the 3′ UTRs of m⁶A in these four transcripts (Fig. 7A).

Next, the protein expression levels of the potential target transcript of ALKBH5 were quantified by western blot analysis. Accordingly, our results showed that CDKN1A (p21) expression increased independent of p53 in ALKBH5-knockdown PC9 cells, whereas TIMP3 expression increased in ALKBH5-knockdown A549 cells (Fig. 7B).

To investigate the effect of both CDKN1A in PC9 and TIMP3 in A549 on cell proliferation in ALKBH5-deficient lung cancer cell lines, we confirmed the reduced expression using siRNA (Fig. 7C, D, S7A, B). The knockdown of CDKN1A partly rescues the decreased cell proliferation in ALKBH5-deficient PC9 cells (Fig. 7E). The knockdown of TIMP3 also partly rescues the decreased cell proliferation in ALKBH5-deficient A549 cells (Fig. 7F). These results suggest that CDKN1A and TIMP3 are important targets for cell proliferation and are required for cell number reduction by repression of ALKBH5.

Additionally, we evaluated the expression of ALKBH5 target mRNAs CDKN1A and TIMP3 in lung cancer using the TCGA dataset. Accordingly, cancerous tissues (high ALKBH5 expression) had lower CDKN1A and TIMP3 expression than noncancerous tissue (low ALKBH5 expression; Fig.). The results were consistent with the experimental results of CDKN1A and TIMP3 knockdown described above. S7C, D

IGF2BP1, IGF2BP2, IGF2BP3 (IGF2BPs), and YTHDF2 are well-known m⁶A-recognizing RNA-binding proteins and readers of m⁶A that have been known to stabilize or destabilize mRNA. The expression of proteins of IGF2BPs in a series of cell lines was analyzed by western blotting, and the results showed differences in their expression in lung cancer cells and immortalized bronchial epithelial cells according to the cell lines (Fig. 8A).

Thereafter, ALKBH5 and IGF2BPs were knocked down with siRNA to investigate the association between ALKBH5 and IGF2BPs and the expression of CDKN1A (p21) or TIMP3 (Fig. S8A–D). Western blot analysis revealed that IGF2BPs knockdown sufficiently reduced the expression of IGF2BPs (Fig. 8B), although some of them appeared to be weakly expressed as shown in Fig. 8A. Accordingly, ALKBH5 knockdown increased the mRNA expressions of CDKN1A in PC9 cells and TIMP3 in A549 cells. The knockdown of IGF2BPs alone did not significantly change the expression of CDKN1A and TIMP3 (Fig. S8E), but that of both ALKBH5 and IGF2BPs offset the increased expression of CDKN1A and TIMP3 (Fig. 8C). We further analyzed the protein expression of YTHDF2 in lung cancer cell lines (Fig. 8D) and conducted transfection of siRNA for YTHDF2 as previously described [59], which confirmed that upregulated expression of CDKN1A by ALKBH5 knockdown was increased by YTHDF2 knockdown in PC9, which suggested that m⁶A of CDKN1A may be affected by both IGF2BPs and YTHDF2 reversibly (Fig. 8E and S8F). Actinomycin D assay showed that ALKBH5 knockdown stabilized CDKN1A mRNA in PC9 cells, and this stabilization was offset by the knockdown of IGF2BPs. ALKBH5 knockdown also stabilized TIMP3 mRNA in A549 cells, although not statistically significant, and this stabilization was decreased by IGF2BP3 knockdown (Fig. 8F). These results suggest that these alterations in mRNA expression were offset by a double knockdown of both ALKBH5 and one of the IGF2BPs, and the decline of mRNAs was, at least partly, owing to the destabilization of these mRNAs by one of the IGF2BPs.

Considering that the interaction between ALKBH5 and IGF2BPs was found to regulate the expression of genes associated with cell proliferation, cell proliferation assays were conducted using ALKBH5- and IGF2BPs-knockdown cells. The knockdown of IGF2BPs did not cause a significant change in cell proliferation in PC9 cells, but the knockdown of IGF2BPs decreased cell proliferation in A549 cells (Fig. S8G). Notably, ALKBH5 knockdown reduced cell proliferation in PC9 and A549 cells and this reduction of cell proliferation was offset by IGF2BPs knockdown (Fig. 8G). These results suggest that the increased effect of cell proliferation upon simultaneous knockdown of IGF2BPs and ALKBH5 cannot be explained by the effect of IGF2BPs alone. Overall, our results support the hypothesis that IGF2BPs are required for ALKBH5 regulation of target mRNA expression and cell proliferation.

All data and supplementary information within the article are available from the published article (including Supplementary information files) or available on published databases (TCGA or GEO). GEO accession numbers of our microarray data are GSE165453↗ and GSE165453↗.