ALKBH4 promotes tumourigenesis with a poor prognosis in non-small-cell lung cancer

The human AlkB homolog family (ALKBH) of proteins play a critical role in some types of cancer. However, the expression and function of the lysine demethylase ALKBH4 in cancer are poorly understood. Here, we examined the expression and function of ALKBH4 in non-small-cell lung cancer (NSCLC) and found that ALKBH4 was highly expressed in NSCLC, as compared to that in adjacent normal lung tissues. ALKBH4 knockdown significantly induced the downregulation of NSCLC cell proliferation via cell cycle arrest at the G1 phase of in vivo tumour growth. ALKBH4 knockdown downregulated E2F transcription factor 1 (E2F1) and its target gene expression in NSCLC cells. ALKBH4 and E2F1 expression was significantly correlated in NSCLC clinical specimens. Moreover, patients with high ALKBH4 expression showed a poor prognosis, suggesting that ALKBH4 plays a pivotal tumour-promoting role in NSCLC.


ALKBH4 knockdown reduced cell proliferation by inducing G 1 phase arrest in NSCLC cells.
To investigate the function of ALKBH4 in NSCLC cells, we first evaluated the expression of ALKBH4 in 22 NSCLC cell lines (adenocarcinoma: 15 cell lines; squamous cell carcinoma: 7 cell lines). Compared to the adjacent normal lung tissues obtained from postoperative tissues of NSCLC patients, all 22 NSCLC cell lines expressed high levels of ALKBH4 ( Supplementary Fig. S1). A549 and II-18 cells with a high ALKBH4 expression were used for the subsequent knockdown and overexpression experiments. ALKBH4 knockdown using siRNAs significantly suppressed the proliferation of A549 ( Fig. 2A) and II-18 cells (Fig. 2B). Conversely, ALKBH4 overexpression using pEB-ALKBH4 vector significantly promoted the proliferation of A549 cells (Fig. 2C). Also, in the case of NCI-H23 cells, the overexpression of ALKBH4 led to increased proliferation. However, overexpression of a catalytically inactive mutant ALKBH4 (H169A/D171A) had no significant effect on cell proliferation (Fig. 2D,E). Moreover, ALKBH4 knockdown inhibited the effect of overexpression of wild-type ALKBH4 on cell proliferation in NCI-H23 cells (Fig. 2D,E), suggesting that ALKBH4 promotes cell growth via its enzymatic activity. To www.nature.com/scientificreports/ clarify the tumour-promoting potential of ALKBH4 expression in vivo, A549 cells stably expressing control or ALKBH4 shRNA were constructed. We confirmed the decreased proliferation of A549 cells transfected with ALKBH4 shRNA, compared with those transfected with control shRNA, in vitro (Fig. 2F). These cells were used as an in vivo xenograft model. Suppression of tumour volume, as well as of tumour weight, was observed in mice xenografted with ALKBH4-knockdown A549 cells (shALKBH4), compared with the control cells (shControl) in Fig. 2G,H, respectively. These results suggest that ALKBH4 might function as a critical tumour promoter in NSCLC cells.
To clarify the function of ALKBH4 on cell proliferation, we conducted gene array analysis using the total RNA of A549 cells transfected with either ALKBH4 siRNA #1 or control siRNA. ALKBH4 knockdown affected the . The cells which were transfected for 48 h were reseeded on a xCELLigence E-plate and their proliferation was detected using a xCELLigence DP system (lower panels). Degrees of cell proliferation expressed as cell index by the system are the means ± S.D. of three independent experiments (A-C). (D) Lysates of NCI-H23 cells co-transfected with pEB Multi-Neo-ALKBH4 wild-type (WT), pEB Multi-Neo-ALKBH4 mutant (mt), or mock vector and ALKBH4 siRNAs were subjected to western blot analysis with anti-ALKBH4 and anti-β-actin antibodies. Uncropped Western blot data are shown in Supplementary Fig. S9. Representative pictures of three independent experiments are shown (upper pictures). NCI-H23 cells were transfected for 48 h and reseeded in 96-well plates, and their proliferation was measured using the WST-8 assay (lower panels). The relative cell growth data on day 4 are presented in (E). The values are presented as the mean ± S.D. for each group. *p < 0.05; **p < 0.01 vs. wild-type ALKBH4 (one-way ANOVA with Bonferroni post-hoc tests). (F) A549 cells stably expressing ALKBH4 shRNA or control shRNA were subjected to Western blot analysis with anti-ALKBH4 and anti-β-actin antibodies. Uncropped Western blot data are shown in Supplementary Fig. S9. Representative pictures of three independent experiments are shown (upper pictures). Cell proliferation was detected using a xCELLigence DP system (lower panels). Degrees of cell proliferation expressed as cell index by the system are the means ± S.D. of three independent experiments (lower panels). (G) A549 cells stably expressing ALKBH4 shRNA (shALKBH4) and control shRNA (shControl), which are shown in (F), were subcutaneously injected into nude mice. The upper picture shows a xenograft tumour from mice inoculated with shControl-or shALKBH4-expressing A549 cells (shControl: n = 8; shALKBH4: n = 7). White scale bar, 1 cm. Tumour volume was calculated by measuring the tumour size every four days (lower panel  Table S2). Gene ontology enrichment analysis using several databases (Bioplanet, KEGG, and Reactome) revealed that these 2011 genes were mostly associated with cell cycle processes (Supplementary  Table S3). Upregulated 1561 genes were associated with p53 signaling pathway and ECM-receptor interaction (Supplementary Table S4). To examine whether the suppression of cell proliferation induced by ALKBH4 knockdown was due to cell cycle arrest, we performed cell cycle analysis. ALKBH4 knockdown elevated the ratio of G 1 phase cells and decreased the ratio of S phase cells in A549 (Fig. 3A) and II-18 cells (Fig. 3B). Moreover, cyclin-dependent kinase 2 (CDK2), cyclin-dependent kinase 4 (CDK4), and cyclin D3, the key proteins for the progression of the G 1 phase, were markedly downregulated after ALKBH4 knockdown in A549 cells (Fig. 3C). Conversely, overexpression of wild-type ALKBH4 increased the ratio of S phase cells in NCI-H23 cells (Fig. 3D). Apoptosis analysis showed that ALKBH4 knockdown significantly induced apoptosis in A549 cells compared to control siRNA transfection (Fig. 3E). To further clarify the function of ALKBH4 in the cell cycle, we conducted G 0 -marker assays through β-galactosidase staining. ALKBH4 knockdown had no significant effect on senescenceassociated β-galactosidase expression in A549 cells ( Supplementary Fig. S2). These results suggest that highly expressed ALKBH4 promotes cell proliferation via regulation of the progression of the G 1 phase in NSCLC cells. Using Gene Expression Profiling Interactive Analysis (GEPIA) 27 , an interactive web server for analysing RNA sequencing expression data of tumours and normal samples from The Cancer Genome Atlas (TCGA) and The Genotype-Tissue Expression (GTEx) database (https:// gtexp ortal. org/ home/), we found that a broad range of cancer tissues, including lung adenocarcinoma and lung squamous cell carcinoma, express higher levels of ALKBH4 mRNA than each corresponding normal tissue ( Supplementary Fig. S3A). To examine whether ALKBH4 knockdown downregulated E2F1 signalling in NSCLC cells. Since ALKBH4 knockdown induced G 1 phase arrest in NSCLC cells, we focused on E2F1 being known as a critical regulator of G 1 /S phase transition. ALKBH4 knockdown significantly reduced E2F1 expression in A549 and II-18 cells (Fig. 4A,B, respectively). Enrichment analysis using ChEA database 28 showed that 215 E2F1-target genes were also downregulated via ALKBH4 knockdown (Supplementary Table S5). Since phospho-Ser/Thr phosphatase cdc25A (CDC25A), cyclin E1 (CCNE1), and Myb-like protein 2 (MYBL2) have been reported as tumour promoters [29][30][31] and are known as cell cycle regulators in NSCLC, we focused on these genes. To verify the results of gene array analysis, the expression of E2F1-target genes was determined using qPCR. ALKBH4 knockdown significantly downregulated the target genes of E2F1 both in A549 and in II-18 cells (Fig. 4C-H). The relationship between ALKBH4 and E2F1 expression or those of E2F1-target genes was further confirmed in NSCLC specimens. Compared to the normal adjacent lung tissues, the expression of E2F1 (Fig. 5A) and of its target genes ( Fig. 5B-D) was significantly high in tumour tissues. Importantly, there was a positive correlation (r = 0.46, p = 0.001) between ALKBH4 and E2F1 expression in NSCLC specimens (Fig. 5E). A significant positive correlation was also observed between ALKBH4 and each of the E2F1-target genes ( Fig. 5F-H). The positive correlation between the expression of ALKBH4 and E2F1, as well as between that of ALKBH4 and of each of the E2F1-target genes was observed, regardless of the presence or absence of EGFR gene mutation (Supplementary Fig. S4A-H). In addition, only stage I, but not late stage (≥ stage II) NSCLC, showed a positive correlation between ALKBH4 and E2F1, or each of the E2F1-target genes (Supplementary Fig. S5A-H). Moreover, we recognised the positive correlation between the expression of ALKBH4 and E2F1, as well as between that of ALKBH4 and of each of the E2F1-target genes, regardless of the histologic subtypes, using TCGA database (Supplementary Fig. S6A,B) in NSCLC. These results suggested that ALKBH4 upregulates the expression of E2F1, followed by that of its target genes, in NSCLC. www.nature.com/scientificreports/ High ALKBH4 expression correlates with overall-and recurrence-free survival in NSCLC. Finally, to clarify the relationship between ALKBH4 expression and the prognosis of NSCLC patients, we performed immunohistochemical staining using anti-ALKBH4 antibody on NSCLC specimens. The immunohistochemical analysis showed that ALKBH4 was positive in 35 tumours and negative in 45 tumours (Fig. 6A). There was no significant intergroup heterogeneity regarding tumour size, presence or absence of pleural invasion, intrapulmonary metastasis, and nodal metastasis. During the follow up (median: 68.5 months; range: 2 to 110 months), 35 recurrences and 31 deaths occurred in a total of 80 patients. Kaplan-Meier survival curves are shown in Fig. 6B. The recurrencefree survival rate was significantly lower in the ALKBH4 positive group than in the ALKBH4 negative group (log rank test, p = 0.01). Likewise, the overall survival rate was significantly lower in the ALKBH4 positive group than in the ALKBH4 negative group (log rank test, p = 0.03). Additionally, we performed survival analysis using TCGA   www.nature.com/scientificreports/ database. However, neither overall survival nor disease-free survival was significantly related to ALKBH4 mRNA expression levels in NSCLC (Supplementary Fig. S7).

Discussion
In the present study, we demonstrated that ALKBH4 is highly expressed in cell lines as well as in tumour tissues of NSCLC patients, and that it functions as a tumour promoter via its enzymatic activity. Knockdown of ALKBH4 downregulated the expression of E2F1, a critical regulator of the G 1 /S phase transition in NSCLC cells. E2F1 expression was positively correlated with the expression of ALKBH4 in NSCLC clinical samples. Moreover, the expression of E2F1-target genes (CDC25A, CCNE1 and MYBL2) was positively correlated with the expression of ALKBH4 in NSCLC clinical samples. ALKBH4 knockdown downregulated the expression of CDK2 and cyclin D3 encoding gene in A549 cells. Since CDK2 and CCND3 (cyclin D3) have also been reported as E2F1target genes [32][33][34][35] , the downregulation of CDK2 and CCND3 might be due to the regulation of E2F1 expression by ALKBH4. It is well known that E2F1 is overexpressed in several cancers, including NSCLC [36][37][38] and that an aberrant E2F1 expression is correlated with a lower overall survival of NSCLC patients 26 . Taken together, the upregulated ALKBH4 expression plays a pivotal NSCLC-promoting role, leading to a poor prognosis in NSCLC patients.
Although we have shown that ALKBH4 overexpression promotes cell growth via its enzymatic activity (Fig. 2D,E), the effect of ALKBH4 enzymatic activity on downstream effector molecules such as E2F1 and E2F1-target genes remains unclear and requires further investigation.
Recently, it was reported that ALKBH4 functions as a demethylase for N6-adenosine modification in DNA 39 , and the C. elegans ALKBH4 ortholog (denoted NMAD-1) also has such a demethylase function 40 . Abrogation of ALKBH4 (or NMAD-1) function appears to have strong effects on various phenomena related to cell cycle progression, such as cytokinesis, DNA replication, and meiosis 15,41,42 . As shown in Supplementary Table S3, ALKBH4 knockdown also affected DNA replication and meiosis-related genes in A549 cells. Moreover, the enzymatic domain of ALKBH4 was critical for the upregulation of cell proliferation in NSCLC cells, suggesting that ALKBH4 may function as a tumour promoter by targeting N6-adenosine modification in NSCLC cells.
Although few publications suggest that gene expression is mostly regulated at the mRNA level 43,44 , many studies have reported a discrepancy between mRNA and protein levels 45,46 . Protein levels more closely reflect the cancer phenotype because these proteins execute major intracellular molecular functions. Therefore, we believe that survival analysis using IHC data reflects the true nature of ALKBH4 in NSCLC.
A high expression of ALKBH4 is correlated with an overall-and recurrence-free survival in NSCLC. On the contrary, Supplementary Fig. S4 shows a positive correlation between the expression of ALKBH4 and E2F1, as well as between that of ALKBH4 and E2F1-target genes, which were only observed in early stage NSCLC tumour tissues, but not in those of late stage NSCLC. Among the Polo-like kinase (PLK) family, PLK1 and PLK4 are highly expressed in NSCLC 47,48 and promote metastasis via epithelial-mesenchymal transition (EMT) induction 49,50 . Since ALKBH4 knockdown downregulated the expression of PLK1 and PLK4 (Supplementary Table S2), ALKBH4 may function as a tumour promoter by accelerating cancer cell proliferation via upregulation of E2F1 signalling in early stage, and by promoting metastasis via upregulation of PLK signalling in late-stage NSCLC.
We found that ALKBH4 functions as a tumour promoter in NSCLC. However, Shen et al. recently reported that ALKBH4 expression is decreased in colon cancer tissues, compared to adjacent normal tissues, and functions as a tumour suppressor by decreasing H3K4me3 levels by competitively binding to methyltransferase WDR5 17 . WDR5 is overexpressed in colon cancer and the depletion of WDR5 reduces cell viability of colon cancer cell lines 57 . On the contrary, an overexpression of WDR5 was reported to induce G 1 arrest in A549 cells, independently of the H3K4me3 status, and WDR5 was suggested to possibly have different functions in different cancers 58 . Therefore, although the underlying mechanism should be analysed, the tumour promoting role of ALKBH4 in NSCLC might partly rely on WDR5.
Cancer cells adapt their metabolism to promote tumour growth. One important metabolic feature of cancer cells is the Warburg effect, which leads to high rates of glucose utilisation and lactate production 59 . It has been reported that E2F1 promotes this effect by enhancing glycolysis and repressing glucose oxidation in the mitochondria 60,61 . Moreover, E2F1-mediated repression of oxidative metabolism results in the self-renewal of cancer stem cells 62 , suggesting that ALKBH4 may confer the Warburg effect through an increased expression of E2F1, leading to efficient recurrence in NSCLC patients.
The specific types of mutations that confer drug sensitivity to EGFR-targeted drugs are present in the tyrosine kinase domain of the EGFR gene, corresponding to exons 19 and 21 with 5-20% of overall incidence 63 . In contrast, upregulated ALKBH4 expression was independent of the mutation status of the EGFR gene in NSCLC specimens (Fig. 1G,H). Moreover, an increased ALKBH4 expression was shown in broad types of cancer (Supplementary Fig. S3A). We showed that ALKBH4 knockdown induced the downregulation of cell proliferation and G 1 arrest in breast cancer cell line MCF-7 cells, as well as in NSCLC cells (Supplementary Fig. S3D www.nature.com/scientificreports/ respectively). Therefore, we propose that ALKBH4 may be an important molecular target for, not only NSCLC, but also for a wide range of cancers.

Materials and methods
Clinical specimens. Specimens of NSCLC tissues and adjacent non-cancerous tissues were obtained from the patients who had undergone primary curative resection of a lung tumour at Kagoshima University Hospital (Japan) as described before 64   Western blotting. Western blotting was conducted as described before 65 . Protein samples were separated on a 7.5-15% sodium dodecyl sulphate (SDS)-polyacrylamide gel electrophoresis (PAGE) gel, and then transferred to a polyvinylidene difluoride (PVDF) membrane by using the Bio-Rad semidry transfer system (1 h, 12 V) (Hercules, CA, USA). Immunoreactive proteins were visualised by treatment with a detection reagent (ECL Prime western blotting detection reagent, GE Healthcare), using the antibodies described above, in an ImageQuant LAS 4000 mini system (GE Healthcare). Densitometric analysis was performed using the National Institute of Health (NIH) Image J software. G 0 -marker analysis. A549 cells transfected with ALKBH4 or control siRNAs were cultured for 72 h, and G 0 -marker assay was conducted using senescence-associated β-galactosidase (Senescence Detection Kit, BioVision, Milpitas, CA, USA) according to the manufacturer's protocol. β-galactosidase-positive cells were counted in randomly selected 25 fields.

Cell proliferation assay. Cell proliferation was examined using the xCELLigenceReal Time Cell Analyzer
Dual Purpose (RTCA DP) system (Roche, Basel, Switzerland). A549 or II-18 cells were transfected with the ALKBH4 siRNA or control siRNA. After 24 h of incubation, cells were reseeded on an E-plate 16 (A549 cells: 1 × 10 3 cells/well; II-18 cells: 3 × 10 3 cells/well) and incubated for the indicated times. Water-soluble tetrazolium salt-8 (WST-8) reagent (Dojindo) was used for the cell proliferation assay (Fig. 2D), which was conducted as described previously 66 . NCI-H23 cells transfected with the pEB Multi-Neo-ALKBH4 vector (wild-type or mutant) and ALKBH4 siRNAs were reseeded in a 96-well plate and incubated for the indicated times. After incubation for 2 h with WST-8 reagent at 37 °C and 5% CO 2 , the optical density was determined at 450/630 nm (Ex/Em).  Gene array analysis. Total  Establishment of ALKBH4 shRNA stable cells. A549 cells were seeded on the day before lentivirus infection and cultured in DMEM (Wako) supplemented with 10% foetal bovine serum and 100 mg/L kanamycin at 37 °C under a 5% CO 2 atmosphere. Lentiviral particles, which were purchased from Sigma Aldrich (control shRNA: SHC002V and ALKBH4 shRNA: SHCLNV-NM_17621), were added to the culture medium to 4 multiplicity of infection (MOI), and polybrene (Thermo Fisher Scientific) was added at a final concentration of 8 ng/ µL. After selection by 5 µg/mL puromycin (Sigma Aldrich), the expression levels of ALKBH4 were confirmed using qPCR, and cell lines with a high ALKBH4 knockdown efficiency were used for the next experiments.
Establishment of ALKBH4 shRNA stable cell-xenografted mice. Female BALB/c nude mice were obtained from Oriental Yeast Corporation (Tokyo, JAPAN). Five-week-old mice were used for ALKBH4 shRNA stable cell-xenograft experiments. Animals were kept under 12 h light-dark cycles at 22-24 °C. A549 cells, which had been stably transfected with ALKBH4 shRNA (A549-shALKBH4) and control shRNA (A549-shControl), were both adjusted to a concentration of 0.6 × 10 7 cells in 100 μL of serum-free DMEM. The cell suspensions, together with 100 μL of Matrigel Matrix High Concentration (Corning, New York, USA) were then injected subcutaneously into the right flanks of BALB/c nude mice (A549-shALKBH4, n = 8; A549-shControl, n = 8). One xenografted mouse, which was inoculated with A549-shALKBH4, died during the experiment. The tumour volume was calculated as follows: (tumour length × tumour width 2 )/2. All procedures were performed under a protocol approved by the Animal Experimentation Committee at Osaka University. All methods were carried out in accordance with relevant ARRIVE guidelines and regulations. We confirmed that all methods were carried out in accordance with relevant guidelines and regulations. Developed tumours were resected 52 days after xenografts.
ALKBH4 immunohistochemistry in clinical cases. Eighty patients who underwent radical operation for primary lung adenocarcinoma at Kagoshima University Hospital from January 2001 to December 2007 were subjected to immunohistochemistry for ALKBH4. Immunohistochemistry was conducted as described before 64 . Paraffin-embedded section (3 μm of thickness) were deparaffinised and dehydrated. The endogenous peroxidase activity of specimens was blocked using a 0.3% hydrogen peroxide solution in methanol. The sections were blocked with 1% bovine serum albumin and were incubated with the rabbit polyclonal antibody against human ALKBH4 (1:200; Novus Biologicals, NBP2-14737) overnight at 4 °C, followed by staining with a streptavidinbiotin peroxidase kit (Vector Laboratories, CA, USA). The immune complex was visualised by incubating the sections with diaminobenzidine tetrahydrochloride. The sections were counterstained with haematoxylin and www.nature.com/scientificreports/ mounted. Non-cancerous colon samples were used as positive controls for ALKBH4. ALKBH4 expression was determined by counting the number of cancer cells in which the cytoplasm was stained with the anti-ALKBH4 antibody. Two investigators evaluated ALKBH4 expression via immunohistochemistry within each tumour by assessing a total of 1000 cancer cells in 10 selected fields (100 cells/field) using high-power (× 200) microscopy, in an independent manner. The average labelling index of ALKBH4 was assessed according to the proportion of positive cells present in each field. Tumours with an average labelling index of 20% or more were defined as ALKBH4-positive. The specificity of the anti-ALKBH4 antibody was verified through immunofluorescence staining using A549 cells transfected with shControl and shALKBH4 ( Supplementary Fig. S8).
Statistics. The results were expressed as the mean ± standard deviation (S.D.). Differences between the values were statistically analysed using the Student's t-test, paired t-test or one-way analysis of variance (ANOVA) with Bonferroni post-hoc tests (GraphPad Prism 6.0, GraphPad software). Pearson correlation analysis was used for the correlation analysis. A p-value < 0.05 was considered statistically significant.
Ethical approval. The