Metformin Reduces Histone H3K4me3 at the Promoter Regions of Positive Cell Cycle Regulatory Genes in Lung Cancer Cells


 BackgroundThis study aimed at understanding the effect of metformin on histone H3 modifications at the promoters of cell cycle regulatory genes in lung cancer cells. MethodsHistone H3 modifications were analyzed using ChIP-seq, and changes in DNA methylation and chromatin accessibility were evaluated using the SureSelect Methyl-Seq Target Enrichment System and ATAC-seq, respectively. MLL2 overexpression was analyzed in tumor and matched normal tissues from 42 non-small cell lung cancer (NSCLC) patients. ResultsMetformin showed little effect on DNA methylation or chromatin accessibility at the promoter regions of cell cycle regulatory genes in lung cancer cells but significantly downregulated histone H3K4 methyltransferase MLL2 and reduced H3K4me3 levels at the promoters of positive cell cycle regulatory genes such as CDK1, CDK6, and E2F8. Eighty-eight genes involved in cell cycle showed reduced H3K4me3 levels in response to metformin, and 27% of them showed mRNA downregulation. The siRNA-mediated knockdown of MLL2 significantly reduced the H3K4me3 levels at the promoters of positive cell cycle regulatory genes. MLL2 overexpression was found in 14 (33%) of 42 NSCLC patients, with a higher prevalence in females (P = 0.04). A Cox proportional hazards analysis showed that recurrence-free survival of adenocarcinoma patients with MLL2 overexpression was approximately 1.32 (95% CI = 1.08–4.72; p = 0.02) times poorer than in those without it after adjusting for sex and pathologic stage. ConclusionsThe present study suggests that metformin might reduce H3K4me3 levels at the promoters of positive cell cycle regulatory genes through MLL2 downregulation in lung cancer cells. And, MLL2 may be a potential therapeutic target for reducing the recurrence of lung adenocarcinoma.

streptavidin beads. Eluted DNA was treated with sodium bisul te, ampli ed by PCR, and sequenced on an Illumina Genome Analyzer. Mapping of the sequenced reads was performed with the Bowtie2 program. After duplicate removal, the number of mapped reads from cells treated with metformin and untreated controls was 47,803,761 and 54,906,440, respectively. Differentially methylated regions were de ned as regions with more than 5 consecutive methylated CpGs and a ratio of observed to expected CpG of greater than or equal to 0.6. Regions with a p-value < 0.001 (Wilcoxon rank sum test) and a fold change |FC| ≥ 1.5 were considered signi cant.
The captured histones were detected with a detection antibody, followed by a color development reagent. Absorbance intensity was measured at 450 nm with a microplate reader. The amount of a particular histone modi cation was calculated as a percentage of the total H3 signal.

ChIP-seq assay
Chromatin immunoprecipitation (ChIP) was performed with a SimpleChIP Enzymatic Chromatin IP Kit (Magnetic Beads) (#9003, Cell Signaling Technology, Danvers, MA) according to the manufacturer's instructions. Brie y, cells were cross-linked with 1% formaldehyde (Sigma-Aldrich) and lysed. The crosslinked chromatin was digested with micrococcal nuclease to a length of 150-900 bp of DNA/protein fragments. The fragments were immunoprecipitated with antibodies against H3K4me3 (C42D8, Cell Signaling Technology), H3K27me3 (C36B11, Cell Signaling Technology), or normal rabbit IgG with protein G magnetic beads. The protein-DNA cross-links were reversed, and puri ed DNA was then subjected to next-generation sequencing (NovaSeq 6000 System). Raw data were checked using the FastqQC program and trimmed with Trimmomatic 0.38. The sequence library was mapped to reference genome UCSC hg19 using the Bowtie1 program. Peaks were identi ed using the MACS2 algorithm. Comparisons between paired samples were performed in the MAnorm program, and the M, A, and P values were calculated [14]. The M-value indicates the fold change in normalized read densities, the A-value denotes the average signal strength of the normalized read densities, and the p-value shows the signi cance of read intensity differences between two samples. Regions with p-values < 0.001 (-log10 p-value greater than 3) were considered statistically signi cant.

RNA-seq assay
Total RNA was isolated from cultured H1299 cells using a TruSeq Stranded mRNA LT Sample Prep Kit (Illumina, San Diego, CA). The RNA was broken into short fragments after removing genomic DNA contamination with DNase I. The RNA fragments were reverse-transcribed into cDNA, ligated to adaptors, ampli ed by PCR, and sequenced using a next-generation sequencing device (NovaSeq 6000 System). The quality of the raw sequence data was checked using the FastQC program; low-quality reads and PCR duplicates were removed using Trimmomatic 0.38. Trimmed sequences were aligned to the reference transcripts using the HISAT2 program. The number of mapped reads was 83,118,285 and 73,808,022 for cells treated with metformin and untreated controls, respectively. Finally, the expression level of each transcript was estimated by counting the reads mapped to the transcripts. The read count was normalized into fragments per kilobase of transcript per million mapped reads value, and reads with a value of zero were excluded. Of the 27,685 genes, 10,654 were removed, and the remaining 17,031 genes were further analyzed. Signi cantly up-or downregulated genes were identi ed using a cutoff value of |FC| ≥ 1.5.

Gene-set enrichment analysis
A Gene Ontology search and Gene Set Enrichment Analysis were performed using the Database for Annotation, Visualization, and Integrated Discovery (DAVID) and a Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analysis [15,16].

ATAC-seq assay
The assay for transposase-accessible chromatin using sequencing (ATAC-seq) was performed by Active Motif (Carlsbad, CA) as previously described [17]. Peaks were identi ed using the MACS2 algorithm at a cutoff of p-value < 1E-7 with the non-model option. Signi cantly different regions were generated by applying the cutoffs MaxTags ≥ 100 and log2 ratio > 1 or < -1.
Quantitative real-time PCR (qRT-PCR) Total RNA was isolated using a PureLink RNA Mini Kit (Invitrogen, Carlsbad, CA) and reverse-transcribed using a SuperScript VILO cDNA Synthesis Kit (Invitrogen, Carlsbad, CA). qRT-PCR was performed with SYBR green dye (4385614, Applied Biosystems, Foster City, CA) under the following conditions: initial denaturation for 5 min at 95 °C, followed by 40 cycles of 5 s at 95 °C and 30 s at 60 °C. The PCR primer sequences are listed in Supplementary Table S1.

Western blot analysis
Total protein was extracted from cultured cells using lysis buffer containing a protease inhibitor cocktail (Roche Applied Science, Indianapolis, IN). The lysates were heated for 5 min at 95 °C, loaded on 10% sodium dodecyl sulfate-polyacrylamide gels, and transferred to a PVDF membrane (Immobilon-P, Millipore, Medford, MA). After blocking with a 3% solution of fetal bovine serum, the membranes were probed with the antibodies listed in Supplementary Table S2. The membranes were then incubated with horseradish peroxidase-conjugated secondary antibodies (Cell Signaling Technology) and visualized with an Immun-Star Western Kit (Bio-Rad, Hercules, CA).
Small interfering RNA (siRNA)-mediated gene silencing To knock down target genes, cells were transiently transfected with 40 nM gene-speci c siRNA (BioNeer, DaeJeon, Korea) or non-targeting siRNA (BioNeer) as a negative control.
The siRNA sequences or their commercial IDs are listed in Supplementary Table S3. Lipofectamine 2000 (Invitrogen, Carlsbad, CA) was used to transfect the siRNA into cells according to the manufacturer's protocol. At 48 hours post-transfection, gene expression was measured using quantitative real-time PCR or western blot analyses.

Statistical analysis
Univariate analyses were performed using the Student's t-test or Wilcoxon rank sum test and Pearson's chi-square test or Fisher's exact test for continuous and categorical variables, respectively. Linear relationships between two continuous variables were analyzed using the Pearson correlation coe cient. The effect of MLL2 overexpression on survival was estimated using Kaplan-Meier survival curves, and the difference between the survival times of two independent groups was evaluated by the log-rank test.
A Cox proportional hazards analysis was conducted to estimate the hazard ratios for MLL2 overexpression on survival after controlling for potential confounding factors. Statistical analyses were conducted using R software (version 3.6.1).

Results
Metformin reduces histone H3K4me3 at the promoter regions of positive cell cycle regulatory genes in lung cancer cells Global alterations of histone H3 were analyzed using an enzyme-linked immunosorbent assay. Metformin resulted in histone H3 modi cations in lung cancer cells, including methylation at H3K4, but the effect was minimal (Fig. 1a). To more clearly evaluate the effect of metformin on histone modi cations, we analyzed the global levels of H3K4me3, H3K9me2, and H3K27me3, which are all important in lung cancer pathogenesis, using western blotting (Fig. 1b). Metformin reduced H3K4me3 and H3K9me2 in both A549 and H1299 cells and H3K27me3 in H1299 cells. ChIP-seq was performed to nd the DNA regions in which metformin induced the observed alterations in H3K4me3 and H3K27me3. The ChIP-seq data are available at the Gene Expression Omnibus (GEO) with the ID: GSE141053. More than 80% of H3K4me3 was detected at gene promoters, whereas less than 10% of H3K27me3 was observed at promoters. Most (70%) of H3K4me3 was enriched at a transcription start site (TSS) ± 1 kb (Fig. 1c), and H3K27me3 was largely located outside of gene promoters (Fig. 1d). Metformin signi cantly reduced the H3K4me3 levels at the promoter regions of 1113 genes and increased the levels at the promoter regions of 1492 genes.
The Gene Set Enrichment Analysis revealed that H3K4me3 levels decreased at the promoters of 88 cell cycle regulatory genes in response to metformin in lung cancer cells (Supplementary Table S4). At the promoter regions of CDKN1A (p21) and CDKN1B (p27), H3K4me3 was upregulated in response to metformin, but the difference was statistically signi cant only in CDKN1A (Supplementary Table S5). Representative ChIP-seq images of H3K4me3 reduction are shown in Fig. 1E.
H3K4me3 reduction at the promoter regions of positive cell cycle regulatory genes is associated with mRNA downregulation To identify genes whose transcription was signi cantly altered by metformin, we performed RNA-seq in H1299 cells. The RNA-seq data are available at the GEO with the ID: GSE141052. The expression levels of 1,114 genes were altered by 1.5 times or more in cells treated with metformin: the mRNA levels of 499 genes were downregulated, and those of 615 genes were upregulated (Fig. 2a). A KEGG enrichment analysis showed that the genes signi cantly up-and downregulated in response to metformin were involved in the cell cycle, apoptosis, cellular senescence, and p53 signaling pathways (Fig. 2b). Eighty-ve genes involved in the cell cycle are listed in Supplementary Table S6. Positive cell cycle regulators such as cyclin A2, cyclin E2, CDK1, E2F2, E2F6, and E2F8 were downregulated, whereas negative cell cycle regulators such as CDKN2B (p15), CDKN1A (p21), DDIT4, and GADD45 were upregulated. Of the 85 genes, 24 genes (28%), including MKI67, E2F8, CDK1, CDC7, and UHRF1, showed reduction of H3K4me3 at their promoter regions ( Fig. 2c and Supplementary Table S7), and those H3K4me3 levels were positively related to mRNA levels (Fig. 2d).
The effect of metformin on histone H3 methyltransferase is independent of AMPK We investigated whether metformin regulates the expression of histone H3 methyltransferase and demethylase in lung cancer cells. Metformin downregulated the protein (Figs. 3a and 3b) and mRNA ( Fig. 3c) expression of EHMT1, EHMT2, EZH1, EZH2, MLL1, MLL2, and WDR82 in lung cancer cells. Metformin also reduced the expression of H3K4 demethylase KDM5A (JARID1A) and H3K9 demethylase KDM4B (JMJD2B) at the levels of protein (Fig. 3d) and mRNA (Fig. 3e) in A549 and H1299 cells. Because metformin is known to exert its antitumor action partly through AMPK activation, we analyzed whether its effect on the expression of histone H3 methyltransferases depended on AMPK. AMPK activation using 5aminoimidazole-4-carboxamide ribonucleoside (AICAR) did not suppress mRNA expression of MLL2, WDR82 (Fig. 3f), DOT1L, EHMT2, or EZH2 (data not shown), but AMPK inhibition using dorsomorphin suppressed the expression of those genes (Fig. 3g). Thus, the downregulation of histone H3 methyltransferases by metformin might not depend on AMPK.
MLL2 knockdown signi cantly reduces the expression of H3K4me3 and positive cell cycle regulators To nd the H3K4 methyltransferases that were most signi cantly involved in the reduction of H3K4me3 in cells treated with metformin, we knockdowned the H3K4 methyltransferases MLL1, MLL2, WDR5, and WDR82 using siRNA in NSCLC cells. The knockdown of histone H3K4 methyltransferases by siRNA was con rmed by real-time PCR analysis (Fig. 4a). MLL2 knockdown most signi cantly reduced global levels of H3K4me3 (Figs. 4b & 4c) and cell proliferation (Fig. 4d) in both A549 and H1299 cells. The number of cells on the third day after siMLL2 transfection decreased signi cantly by 66% and 51% compared with the siControl condition in A549 cells and H1299 cells, respectively (Fig. 4d). MLL2 knockdown also signi cantly downregulated the expression of both positive cell cycle regulators that showed signi cant H3K4me3 reduction in lung cancer cells (Fig. 4f) and those that did not (Fig. 4f).
Metformin does not signi cantly change DNA methylation or chromatin accessibility at the promoter regions of positive cell cycle regulatory genes To understand whether the observed changes in the mRNA levels of cell cycle regulatory genes in response to metformin are associated with DNA methylation, we analyzed the methylation levels of CpGs at the promoter regions of the genes using the SureSelect Methyl-Seq Target Enrichment System. Some regions at the genome level were found to be hyper-or hypomethylated, but most of the cell cycle regulatory genes did not show signi cantly altered DNA methylation like that seen in CDK1 and E2F8 (Fig. 5a). ATAC-seq was performed to investigate whether the downregulation of positive cell cycle regulators in cells treated with metformin was associated with changes in chromatin accessibility. The ATAC-seq data are available at the GEO with ID: GSE141059. Chromatin accessibility in cells treated with metformin decreased in 159 regions and increased in 124 regions. Most regions with changes in chromatin accessibility were outside of promoter regions, de ned as TSS ± 3 kb. Tag distribution (Fig. 5b) in the promoter regions and the peak tag number in the total merged peak ( Fig. 5c) were found to be similar. The regions with the most severe increases and decreases in chromatin accessibility were located 4,667 bp downstream of CHAC1 ( Fig. 5d) and 15,111 bp downstream of ERRFI1 (Fig. 5e), respectively. We identi ed 54 genes whose chromatin accessibility was signi cantly changed at the promoter regions, but most of those genes were not related to cell cycle regulation (Supplementary Table S8).

Clinicopathological characteristics of MLL2 overexpression in primary NSCLC
We used previously reported data to analyze the clinicopathological signi cance of MLL2 overexpression in 42 primary NSCLCs [18]. MLL2 was de ned to be overexpressed when its mRNA level was ≥ 1.5 fold higher in tumor tissues than in matched normal tissue. MLL2 overexpression was found in 14 (32%) of the 42 NSCLCs tested and was not associated with patient age, pathologic stage, or differentiation (data not shown). However, MLL2 overexpression was found to be highly prevalent in females (P = 0.04, Fig. 6a). Although MLL2 overexpression was not statistically associated with smoking status (P = 0.34; Fig. 6b), it tended to occur more frequently in never-smokers: the number of pack-years for patients with and without MLL2 overexpression was 17 ± 21 and 34 ± 28, respectively (P = 0.04; Fig. 6c). MLL2 overexpression was also found in adenocarcinoma at a higher prevalence than in squamous cell carcinoma, though that difference was not statistically signi cant (P = 0.27; Fig. 6d). The effect of MLL2 overexpression on patient survival was analyzed in the 42 NSCLCs. The median follow-up period for the patients was 5.3 years. MLL2 overexpression was signi cantly associated with poor recurrence-free survival in adenocarcinoma (p = 0.02; Fig. 6e) but not in squamous cell carcinoma (p = 0.87). The veyear recurrence-free survival rate of 27 adenocarcinoma patients with and without MLL2 overexpression was 34% and 78%, respectively. A Cox proportional hazards analysis showed that the recurrence-free survival of adenocarcinoma patients with MLL2 overexpression was approximately 1.32 (95% CI = 1.12-4.57; p = 0.01) times poorer than in patients without MLL2 overexpression after controlling for sex and pathologic stage (Supplementary Table S9).

Discussion
A nucleosome is composed of genomic DNA wrapped around an octamer that contains two copies of each of the histone proteins H2A, H2B, H3, and H4. Five lysines in histone H3 (K4, K9, K27, K36, K79) and a lysine in histone H4 (K20) are known to be modulated by methylation. H3K4me3 is present at the promoters of active genes and leads to low nucleosome density at those promoters, and H3K36me3 is associated with elongating RNA polymerase II and occurs in the bodies of active genes. H3K9me2/3 is associated with constitutive heterochromatin formation at promoters and strongly associated with gene repression. H3K27me3 and H4K20me3 are also tightly associated with heterochromatin as clear markers of gene repression. In this study, we investigated the effect of metformin on epigenetic alterations (DNA methylation, histone H3 modi cation, and chromatin accessibility) in lung cancer cells. Although the effects of metformin on DNA methylation and chromatin accessibility were minimal, metformin downregulated H3K4me3, especially at the promoter regions of positive regulatory genes of cell cycle.
H3K4me3 is found in ∼75% of all human active gene promoters in several cell types [19], and the presence of this mark at the promoters of protein coding genes is known to correlate with an active state of gene expression [20]. Many studies have reported the molecular mechanisms that underlie the regulation of gene expression by H3K4me3. H3K4me3 is known to positively regulate gene expression through chromatin remodeling. Recruitment of the chromatin remodeling complex NURF by H3K4me3 allows the transcription factors more access to DNA. Transcriptional pre-initiation complex formation is facilitated by recruiting transcription factor II D (TFIID) via the TATA binding protein associated factor 3 (TAF3) [21]. TAF3, as a component of TFIID, interacts with H3K4me3 and provides the mechanism by which TFIID is recruited to the promoter. TAF3 binding sites are signi cantly enriched at cell cycle regulatory genes [21]. In contrast, the loss of H3K4me3 inhibits transcription by reducing TFIID binding to promoters [22,23]. In this study, approximately 80% of H3K4me3 was found to be enriched in gene promoters, and 24 (27%) of 88 cell cycle genes with downregulated H3K4me3 in response to metformin also showed downregulated mRNA levels. Based on these observations, it is likely that metformin affects the expression of cell cycle regulatory genes at least partially through the reduction of H3K4me3 at their promoter regions. AMPK activation is known to be one of the mechanisms underlying the antineoplastic effect of metformin. Metformin can inhibit cancer cell growth by suppressing mammalian target of rapamycin complex 1 (mTORC1) in AMPK-dependent and AMPK-independent pathways [24]. Several groups have reported that metformin could inhibit mTORC1 activation in the absence of AMPK. Metformin inhibited mTORC1 in a GTPase-dependent manner in HEK293T cells and did not have a pronounced effect on the energy status of AMPK dKO MEFs in the absence of AMPK [25]. AMPK inhibition had a minor impact on the metformin effect, which induced mTOR inhibition and cell cycle arrest through REDD1 in prostate cancer cells [26]. In addition, metformin suppressed the proliferation of glioma cells through PRAS40mediated mTOR inhibition independent of AMPK [27]. In this study, we also found that the effect of metformin on H3K4 methyltransferase expression did not depend on AMPK activation (Figs. 3f & 3 g). Accordingly, the ability of metformin to downregulate histone H3K4 methyltransferase in lung cancer cells could be independent of AMPK activation despite the contribution of the AMPK/mTOR signaling pathway to metformin-induced anticancer therapy.
The transfer of methyl groups to lysine residues on histone H3K4 is known to be principally catalyzed by methyltransferase. MLL1 and MLL2 are members of the SET1/COMPASS complex and control the expression of actively transcribed genes by regulating H3K4me3 [28]. WDR82, as a core subunit of the SET1A complex, mediates recruitment of the SET1A histone methyltransferase complex to transcription start sites [29]. In this study, metformin signi cantly inhibited the expression of MLL1, MLL2, and WDR82 in lung cancer cells (Fig. 3A). H3K4me3 was most signi cantly downregulated in A549 and H1299 cells treated with siMLL2 rather than siMLL1 or siWDR5 (Figs. 4b & 4c), suggesting that MLL2 could function as a core factor regulating H3K4me3 in lung cancer cells. MLL2 is located at 12q12-13 and is also known as KMT2D/ALR/MLL4. MLL2 is known to function as an oncogene or a tumor suppressor, depending on the cancer type. Several groups have reported that MLL2 plays an oncogenic role in human cancer. MLL2 overexpression compared with adjacent benign epithelium was reported in invasive carcinomas of the breast and colon [30]. MLL2 knockdown resulted in a signi cant decrease in the migration of esophageal squamous cell carcinoma cells. Clinically, a high level of MLL2 is signi cantly associated with early-stage ESCC lymph node metastasis [31]. shRNA-mediated knockdown of KMT2D attenuated neoplastic cell migration [32]. Knockdown of MLL2 using siRNA or lentiviruses suppressed the proliferation of gastric cancer cells [33]. In this study, MLL2 knockdown also signi cantly inhibited the proliferation of lung cancer cells (Fig. 4d) and downregulated positive cell cycle regulators (Figs. 4e & 4f). These observations suggest that MLL2 might function as an oncogene in some kinds of human cancer and play a role in reducing H3K4me3 at the promoter regions of positive cell cycle regulatory genes in response to metformin in lung cancer cells.
MLL2 is known to be involved in tumor progression and is associated with poor prognosis in a variety of cancers. MLL2 knockout suppressed cell cycle progression by inducing cell cycle arrest at the G1 stage in esophageal squamous cell carcinoma cells in vitro and inhibited cell migration [34]. In addition, high MLL2 expression predicts poor prognosis and promotes tumor progression by inducing the EMT in esophageal squamous cell carcinoma [34]. MLL2 maintains overall H3K4me2/me3 levels in MLL-AF9 leukemia cells, and deletion of MLL2 reduces the survival of those cells in vitro and in vivo [35]. MLL2 overexpression positively affected cell migration in pheochromocytoma and was negatively associated with patient survival in gastrointestinal diffuse large B-cell lymphoma and breast cancer [36][37][38]. NSCLC patients with an MLL2 mutation showed reduced overall and progression-free in NSCLC compared with those who had wild-type MLL2 [39]. Consistent with those results, we found that MLL2 overexpression was signi cantly associated with poor recurrence-free survival in lung adenocarcinoma.
The human lysine demethylase 5 (KDM5) subfamily consists of four family members designated KDM5A-D and speci cally catalyzes the demethylation of H3K4. KDM5A, also known as retinoblastoma binding protein (RBP2), was initially identi ed as a binding partner of retinoblastoma protein, and its overexpression has been observed in cancers such as glioblastoma, gastric cancer [40], hepatocellular carcinoma, and lung cancer [41]. Increasing evidence suggests that KDM5A plays an oncogenic role in tumorigenesis and the progression of human cancer. KDM5A promotes small cell lung cancer tumorigenesis by repressing NOTCH signaling and sustaining ASLC1 expression [42]. Wang et al. [43] reported that KDM5A downregulated the expression of E-cadherin and that KDM5A overexpression induced the epithelial-mesenchymal transition in NSCLC cells. KDM5A promotes cell cycle progression by repressing p27 and activating cyclins D1 and E1 in lung cancer cells [41]. In lung cancer cells, KDM5A binds directly to the promoter regions of cyclin D1 and cyclin-dependent kinase inhibitor p27 (KIP1) and promotes G 1 -S progression by activating cyclins D1 and E1 and suppressing p27 (CDKN1B) [41]. KDM5A also promotes the angiogenesis of NSCLC cells by upregulating VEGF and activating HIF-1α via PI3K/Akt signaling [44]. KDM5A depletion enhanced the promoter activities of p16 (INK4A), p21 (CIP1), and p27 (KIP1) and upregulated H3K4me3 levels in gastric cancer cells [40]. In this study, metformin suppressed KDM5A expression and upregulated H3K4me3 at the promoter regions of p21 and p27 (Supplementary  Table S5). Thus in lung cancer cells, metformin might increase p21 and p27 expression by upregulating H3K4me3 at the promoter region of each gene through the downregulation of KDM5A.
This study was limited by several factors. First, the clinical signi cance of MLL2 expression was retrospectively studied in only a small number of samples. Those results need to be con rmed in a prospective study with a large number of patients. Second, the effects of MLL2 and KDM5A on histone H3K4me3 were analyzed separately. H3K4me3 alteration needs to be analyzed after double knockdown of MLL2 and KDM5A.

Conclusions
The present study suggests that metformin may suppress the expression of positive cell cycle regulatory genes by reducing H3K4me3 at their promoter regions through MLL2 downnregulation in lung cancer cells. And, MLL2 may be a therapeutic target for reducing recurrence after surgery in lung adenocarcinoma.

Declarations
Ethics approval and consent to participate: This study was approved by the Institutional Review Board Effects of metformin on histone H3 modi cations H1299 and A549 cells were treated with 5mM metformin or left untreated as controls. a. Global alterations in histone H3 modi cations were analyzed using an enzyme-linked immunosorbent assay. The Y-axis indicates the amounts of histone proteins in metformin-treated cells relative to untreated cells. Error bars indicate the standard deviation (n = 3). b.
Protein levels of H3K4me3, H3K9me2, and H3K27me3 were analyzed by western blotting. The bar graphs show the expression of three proteins in cells treated with 5 mM metformin relative to untreated cells.
Error bars indicate the standard deviation (n = 3, *P < 0.05). c. The results from the ChIP-seq analysis show DNA regions in which H3K4me3 or H3K27me3 modi cation were enriched in response to metformin. d. The read count frequency of H3K4me3 or H3K27me3 within the transcription start site (TSS) ± 3kb is shown. e. The images show ChIP-seq peaks of two representative genes whose H3K4m3 was reduced by metformin. Peaks indicate read count frequency.

Figure 1
Effects of metformin on histone H3 modi cations H1299 and A549 cells were treated with 5mM metformin or left untreated as controls. a. Global alterations in histone H3 modi cations were analyzed using an enzyme-linked immunosorbent assay. The Y-axis indicates the amounts of histone proteins in metformin-treated cells relative to untreated cells. Error bars indicate the standard deviation (n = 3). b.
Protein levels of H3K4me3, H3K9me2, and H3K27me3 were analyzed by western blotting. The bar graphs show the expression of three proteins in cells treated with 5 mM metformin relative to untreated cells.
Error bars indicate the standard deviation (n = 3, *P < 0.05). c. The results from the ChIP-seq analysis show DNA regions in which H3K4me3 or H3K27me3 modi cation were enriched in response to metformin. d. The read count frequency of H3K4me3 or H3K27me3 within the transcription start site (TSS) ± 3kb is shown. e. The images show ChIP-seq peaks of two representative genes whose H3K4m3 was reduced by metformin. Peaks indicate read count frequency. Protein levels of H3K4me3, H3K9me2, and H3K27me3 were analyzed by western blotting. The bar graphs show the expression of three proteins in cells treated with 5 mM metformin relative to untreated cells.
Error bars indicate the standard deviation (n = 3, *P < 0.05). c. The results from the ChIP-seq analysis show DNA regions in which H3K4me3 or H3K27me3 modi cation were enriched in response to metformin. d. The read count frequency of H3K4me3 or H3K27me3 within the transcription start site (TSS) ± 3kb is shown. e. The images show ChIP-seq peaks of two representative genes whose H3K4m3 was reduced by metformin. Peaks indicate read count frequency.       siMLL2 to siControl (n = 3, *P < 0.05). The error bars in all bar graphs indicate the standard deviation (n = 3, *P < 0.05). siMLL2 to siControl (n = 3, *P < 0.05). The error bars in all bar graphs indicate the standard deviation (n = 3, *P < 0.05).

Figure 4
Effects of MLL2 on H3K4me3 and the mRNA levels of positive regulators of cell cycle A549 and H1299 cells were transfected with siMLL1, siMLL2, siWDR5, siWDR82, and the off-target control siRNA (siControl). a. The relative mRNA levels of the targeted genes were measured by qRT-PCR (n = 3, *P < 0.05). b. The expression level of H3K4me3 was analyzed by western blotting. c. Levels of H3K4me3 in the siRNAs relative to the siControl. The western blot analysis was performed three times, and the average value was calculated. Error bars indicate the standard deviation (n = 3, *P < 0.05). d. Cells were transfected with the indicated siRNAs, and the number of cells was calculated using the WST-8 assay on the zero, second, and third days after transfection (n = 4, *P < 0.05). e-f. The mRNA levels of positive regulatory genes of cell cycle whose H3K4me3 levels were reduced signi cantly (e) or were not changed (f) in response to metformin were measured by qRT-PCR. Values are expressed as fold changes from siMLL2 to siControl (n = 3, *P < 0.05). The error bars in all bar graphs indicate the standard deviation (n = 3, *P < 0.05). clusters by the bigWIG metrics program. c. The peak tag number in the total merged peak regions was compared between cells treated and untreated with metformin. d-e. The regions with signi cantly up-or downregulated chromatin accessibility in response to metformin were identi ed using the cutoff value |Log2 Ratio| ≥ 1. The images show that regions with severely downregulated (d) or upregulated (e) chromatin accessibility in response to metformin are located outside the promoter areas of the CHAC1 and ERRFI1 genes. Peaks indicate read count frequency. clusters by the bigWIG metrics program. c. The peak tag number in the total merged peak regions was compared between cells treated and untreated with metformin. d-e. The regions with signi cantly up-or downregulated chromatin accessibility in response to metformin were identi ed using the cutoff value |Log2 Ratio| ≥ 1. The images show that regions with severely downregulated (d) or upregulated (e) chromatin accessibility in response to metformin are located outside the promoter areas of the CHAC1 and ERRFI1 genes. Peaks indicate read count frequency. clusters by the bigWIG metrics program. c. The peak tag number in the total merged peak regions was compared between cells treated and untreated with metformin. d-e. The regions with signi cantly up-or downregulated chromatin accessibility in response to metformin were identi ed using the cutoff value |Log2 Ratio| ≥ 1. The images show that regions with severely downregulated (d) or upregulated (e) chromatin accessibility in response to metformin are located outside the promoter areas of the CHAC1 and ERRFI1 genes. Peaks indicate read count frequency.

Figure 6
The effect of MLL2 overexpression on recurrence-free survival in lung adenocarcinoma a-d.
Clinicopathological signi cance of MLL2 overexpression in 42 primary NSCLCs. MLL2 overexpression was compared according to sex (a), smoking status (b), pack-years (c), and histology (d). e The effect of MLL2 overexpression on the recurrence-free survival of 27 patients with adenocarcinoma was analyzed using Kaplan-Meier survival curves. The p-value was calculated using the log-rank test. Figure 6