ALKBH5‐mediated m6A modification of lncRNA KCNQ1OT1 triggers the development of LSCC via upregulation of HOXA9

Abstract It has been shown that N6‐methyladenosine (m6A) modification is involved in the development of complex human diseases, especially in the development of cancer. Our research investigated the role and mechanism of the m6A modification of lncRNA KCNQ1 overlapping transcript 1 (KCNQ1OT1) in Laryngeal squamous cell carcinoma (LSCC) progression. Microarray analysis was used to quantitatively detect the m6A apparent transcriptional modification level of lncRNA in LSCC tissue. Methylated RNA immunoprecipitation‐qPCR (MeRIP‐qPCR), in situ hybridization (ISH) and quantitative real‐time PCR (qRT‐PCR) were used to examine the m6A modification and expression of KCNQ1OT1. In addition, in vivo and in vitro experiments have tested the effects of KCNQ1OT1 knockdown on the proliferation, invasion and metastasis of LSCC. Mechanically, we found the N6‐methyladenosine (m6A) demethylase ALKBH5 mediates KCNQ1OT1 expression via an m6A‐YTHDF2‐dependent manner and KCNQ1OT1 could directly bind to HOXA9 to further regulate the proliferation, invasion and metastasis of LSCC cells. In general, our research indicates that ALKBH5‐mediated m6A modification of KCNQ1OT1 triggers the development of LSCC via upregulation of HOXA9.

methyltransferases (writers), demethylases (erasers) and m6Abinding proteins (readers). 11,12 The core components of RNA methyltransferase complex include methyltransferase-like 3 (METTL3), methyltransferase-like 14 (METTL14) and Wilms tumour 1-associated protein (WTAP). 13 In turn, fat mass and obesity-associated protein (FTO) and AlkB homolog 5 (ALKBH5) can remove the m6A modification of RNA. 14 In addition, m6A-binding proteins, including YTHDF1, YTHDF2, YTHDF3 and YTHDC1, have been identified as m6A modified 'readers' and regulate the processing, translation and degradation of mRNA. 15 Long non-coding RNA (lncRNA) is an endogenous transcription RNA molecule with a length of more than 200 nucleotides. 16 More and more studies have proved that lncRNAs play a crucial role in the multi-steps of tumour development. 17 For instance, lncRNA growth arrest-specific 5 (GAS5) suppresses LSCC progression through the negative regulation of miR-21 and its targets involved in cell proliferation and apoptosis. 18 Our previous research has demonstrated that lncRNA nuclear paraspeckle assembly transcript 1 (NEAT1) regulates CDK6 expression mediated by miR-107 in LSCC, which may be a potential target for therapeutic intervention of LSCC. 19 It has been proved that m6A-methylated level can affect the expression of lncRNA, and regulate the development of tumour. [20][21][22] Here, we found that KCNQ1OT1 was a valuable prognostic predictor of LSCC patients and revealed the increase of ALKBH5 can promote the development of LSCC in a m6A-YTHDF2-dependent manner by promoting KCNQ1OT1-HOXA9 signalling. ALKBH5-KCNQ1OT1-HOXA9 may be a promising target for the treatment of LSCC.

| Patients and specimens
About 86 pairs of LSCC and adjacent non-tumour tissues were collected from patients who received partial or total laryngectomy in the otolaryngology department of the Second Affiliated Hospital of Harbin Medical University from December 2017 to June 2019. These patients in the study did not receive any cancer treatment before admission.
Differentially m6A-methylated lncRNAs in three pairs of LSCC and nontumour tissues were detected by microarray assay, and the level of m6A modification in the top four low m6A methylation modified lncRNAs was detected by MeRIP-qPCR. In addition, we also detected the expression of ALKBH5 and HOXA9 in 86 pairs of LSCC and non-tumour tissues by IHC and qRT-PCR. This study was approved by the ethics committee of Harbin Medical University and obtained informed consent.

| Microarray analysis
The Arraystar Human m6A-lncRNA Epitranscriptomic microarray analysis was from Arraystar Arraystar Company (Rockville, MD, USA). Total RNA from each sample was quantified using the NanoDrop ND-1000. Briefly, the total RNAs were immunoprecipitated with anti-N6-methyladenosine (m6A) antibody. The modified RNAs were eluted from the immunoprecipitated magnetic beads as the 'IP'. The unmodified RNAs were recovered from the supernatant as 'Sup'. The 'IP' and 'Sup' RNAs were labelled with Cy5 and Cy3, respectively, as cRNAs in separate reactions using Arraystar Super RNA Labeling Kit. The cRNAs were combined together and hybridized onto Arraystar Human lncRNA Epitranscriptomic Microarray (8x60K, Arraystar). After washing the slides, the arrays were scanned in two-colour channels by an Agilent Scanner G2505C.
Washed with 300 μl 1×IP buffer for three times, then resuspended in the prepared total RNA antibody mixture, and rotated the RNA bound to the antibody beads for 2 h at 4°C. The beads were then washed three times with 500 μl 1×IP buffer and twice with 500 μl washing buffer. The enriched RNA was eluted with 200 μl elution buffer at 50°C for 1 h. RNA was extracted by acidic phenol chloroform and ethanol precipitation. Then qRT-PCR detection was performed.

| Quantitative real-time PCR (qRT-PCR)
According to the manufacturer's protocol, total RNAs was isolated from LSCC samples and cell lines using Trizol Reagents (Invitrogen, Carlsbad, CA, USA). The purified RNA was performed by the Reverse Transcription Kit (Takara, China) to reverse-transcribed into cDNA. The expression level of RNAs was quantified using the SYBR Green Master Mix (Roche, Switzerland) and normalized to internal control GAPDH mRNA, finally, the 2 −ΔΔCt method was executed to detect the relative RNA expression level of RNAs. Three independent experiments were carried out. All primers' sequences used in the qRT-PCR were shown in Table S1.

| Subcellular fractionation location
According to the manufacturer's instructions, the PARIS Kit (Life Technologies, USA) is used to separate cytoplasmic and nuclear components.

| In situ hybridization
ISH was performed to detect the expression of KCNQ1OT1. The RNA scope ® 2.5 Assay and HybEZ™ Hybridization System employing in ISH, as well as KCNQ1OT1 target, positive and negative control probes, were provided by Advanced Cell Diagnostics (ACD). The common housekeeping protein, peptidyl-prolyl isomerase B (PPIB) and the bacterial protein dihydrodipicolinate reductase (DapB) were used as positive and negative control probes, respectively. Slices were scored microscopically according to the manufacturer's instruction.
The low expression of KCNQ1OT1 is represented by 0 and 1 scores, while 2, 3 and 4 scores indicated the high expression of KCNQ1OT1.

| Cell culture and cell transfection
Human oral keratinocytes (HOK) and human laryngeal cancer cells  Table S2.

| Cell proliferation assay
According to the cell counting kit 8 (CCK-8, Sigma-Aldrich, MO, US) manufacturer's instructions, 10 μl CCK-8 solution was added into and incubated with cells. At the appointed time, the absorbance at 450nm was measured. In addition, a 5-ethynyl-20-deoxyuridine assay (EdU) kit (BeyoClick™ EdU Cell Proliferation Kit with Alexa Fluor 555; Shanghai, China) was used for detection of cell proliferation. Every step was carried out in strict accordance with the instructions of the kit.

| Colony formation assay
The differently grouped cells were seeded and cultured in 6-well plate for 2-3 weeks. Then, 4% paraformaldehyde was used to fix cell colony for 15 min, then 0.1% crystal violet was stained for 20-30 min. Images of colony formation were saved and analysed.

| Transwell and wound healing assay
Cells were cultured in serum-free medium for 24 h and resuspended in serum-free medium at a density of 1 × 10 5 cells/ml. About 200 μl of cell suspension was added into each Transwell chamber coated with diluted Matrigel (BD Biosciences), as well as, 600 μl of high glucose medium containing 20% FBS was added into each well of 24-well culture plate. After incubated at 37°C for 24 h, the invaded cells were fixed with 4% paraformaldehyde, stained with 0.1% crystal violet and counted by the microscope in five randomly selected fields. Cells were seeded into a 6-well plate and incubated to 80% confluence. Vertical scratches in the central area of monolayer adherent cells per well were generated using 100 μl pipette tip. Images of wound of same observed location at different time points were captured by the microscope, and wound healing extent was evaluated by Image J.

| In vivo experiment
Animal experiments were carried out with the approval of the Animal Ethics Committee. 4-5 weeks old, 15-20 g weight Balb/c male nude mice were provided by Vital River Laboratory Animal Technology Co. Ltd. (Beijing, China). 1 × 10 6 (100 µl) cells of infected and uninfected by lentiviral were, respectively, injected subcutaneously into nude mice which divided randomly into scramble group and shKCNQ1OT1-1 group. After tumour formation, tumour volumes were measured twice a week until mice were euthanized and tumours were surgically removed.

| Immunohistochemistry (IHC)
Paraffin-embedded LSCC and non-tumour tissues, as well as xenograft tumours, were cutted into 4μm slices. The antigen retrieval was performed after samples were removed from paraffin for deparaffi- Thereafter, images of IHC staining were obtained by a microscope.
According to the manufacturer's instructions, RIP lysis buffer was prepared to treat cells. The cell lysates were incubated with protein antibodies and normal rabbit IgG overnight at 4°C. The RNAprotein/antibody complexes were then immunoprecipitated with protein A/G magnetic beads. RNA is extracted from the precipitated complex for qRT-PCR.

| RNA pull-down assay and Western blot analysis
In simple terms, biotin (Bio)-labelled lncRNA KCNQ1OT1 were used to incubate with total proteins from AMC-HN-8 and TU212 cell lysates.
The complexes formed bound to Streptavidin-coupled Dynabeads, after which Western blot analysis was carried out to verify the enriched proteins after elution and recovery. After the proteins were extracted and the protein concentration was measured, the proteins were separated electrophoretically in SDS-PAGE gels and transferred to PVDF membranes. Subsequently, the primary and secondary antibodies were incubated with PVDF membranes, respectively, according to conventional methods. Finally, chemiluminescence detection reagent (ECL) (Solaibao, Beijing, China) was used for detection. All antibodies used in Western blot analysis, unless specially stated, were purchased from Aibokang (Shanghai) Trading Co. Ltd.

| RNA stability assays
Medium containing actinomycin D (a9415, Sigma, USA, 5 μg/ml) was used to culture treated cells, and the mRNA expression was calculated. The procedure of isolating total RNA for qPCR analysis was as described previously.

| Luciferase reporter assay
AMC-HN-8 and TU212 cells were inoculated into 12-well plates at 50% density. Luciferase reporter gene plasmid and control plasmid were co-transfected into cells after reaching 70% cell confluency.
Following the co-transfection for 36 h, the Dual-Luciferase Reporter System (Promega, USA) was used to detect the luciferase activities in cell lysates. The experiment was performed in triplicate.

| Statistical analysis
The experiment data were analysed using SPSS version 17.0 software and presented as mean ± SD. All graphs were drawn with GraphPad Prism 5 and 8 software. P-value was calculated by the Student's t tests. P-value with * indicates statistically significant. Three levels of significance (*P < 0.05; ** P < 0.01; and ***P < 0.001) were used for all the tests.

| Overview of the m6A-lncRNA expression profiles in LSCC
Using the method of Epitranscriptomic Microarray, we evaluated the m6A-lncRNA expression profiles in three pairs of LSCC and nontumour tissues. Differentially m6A-methylated lncRNAs based on 'm6A methylation level' and 'm6A quantity' passing fold change and statistical significance cut-offs were identified and compiled. The default thresholds are |FC| ≥ 1.2 and p-values ≤ 0.05. Hierarchical clustering heatmap analysis was performed for differentially m6Amethylated lncRNAs. Results of hierarchical clustering heatmap showed that there was a significant m6A-methylated lncRNAs expression profile between the samples ( Figure 1A). Volcanic map showed the significant difference m6A-methylated lncRNA in LSCC and non-tumour tissues ( Figure 1B). The data showed the significant differential m6A modification of the 42 lncRNAs. 33 lncRNAs showed high m6A methylation and 9 lncRNAs showed low m6A methylation ( Figure 1C and Table 1).  Figure 2E). We also found KCNQ1OT1 was mainly located in the nucleus of LSCC cells ( Figure   S2). ISH was performed in 86 pairs of LSCC and non-tumour tissues to verify the level of KCNQ1OT1 was significantly increased in LSCC ( Figure 2F). Next, we verified KCNQ1OT1 upregulation in 86 pairs of LSCC and non-tumour tissues by qRT-PCR. In this sample cohort, KCNQ1OT1 was significantly upregulated in 42% (36/86) of LSCC patients ( Figure 2G,H). According to the correlation between clinicopathological status and KCNQ1OT1 expression in LSCC patients, it was found that the expression of KCNQ1OT1 was correlated with differentiation, smoking and lymph node status (Table 2). Besides, we compared the difference of survival rate between patients with high and low expression levels of KCNQ1OT1 survival curve (86 patients, 5 patients were lost to follow-up). The results showed that high expression of KCNQ1OT1 was significantly associated with poor prognosis at 60 months (Logrank p = 0.047, Figure 2I). The KCNQ1OT1 expression was detected in TU212 and AMC-HN-8 cells with HOK cell as   Figure 2J). In conclusion, KCNQ1OT1 is highly expressed and low m6A methylation in LSCC.

| LncRNA KCNQ1OT1 depletion inhibits cell proliferation, migration and invasion of LSCC Cells
Next, we knockdown the expression of KCNQ1OT1 in TU212 and AMC-HN-8 cells. ShKCNQ1OT1-1 was selected for further experiments based on its more effective inhibition ( Figure 3A). CCK-8 and Edu results showed that the growth of LSCC cells was significantly inhibited after KCNQ1OT1 was silenced ( Figure 3B,C). Cell staining with crystal violet showed that TU212 and AMC-HN-8 cell clone numbers were significantly reduced in the shKCNQ1OT1-1 group ( Figure 3D), which showed that KCNQ1OT1 knockdown inhibited cell proliferation. Furthermore, KCNQ1OT1 knockdown greatly inhibited wound closure and invasive abilities ( Figure 3E,F). These results indicate that KCNQ1OT1 promotes the growth of LSCC in vitro.

| The expression of ALKBH5 was positively correlated with that of KCNQ1OT1 in LSCC
The data from GEPIA showed that the expression level of ALKBH5 in HNSCC is higher than in normal tissues ( Figure 5A). Using HOK cells as negative control, the expression of ALKBH5 in TU212 and AMC-HN-8 cells was significantly higher than that in HOK cells ( Figure 5B). Then, the expression of ALKBH5 in 86 pairs of LSCC and non-tumour tissues was detected by qRT-PCR. The results showed that ALKBH5 was highly expressed in LSCC ( Figure 5C). The protein expression levels of ALKBH5 in 86 paired LSCC and non-tumour tissues were detected by IHC. As shown in Figure 5D, ALKBH5 exhibited nuclear localization, and high level of ALKBH5 is observed in LSCC compared with non-tumour tissues. These results suggested that ALKBH5 was overexpressed in LSCC. Notably, we also found ALKBH5 expression was positively correlated with the KCNQ1OT1 expression in 86 pairs of LSCC ( Figure 5E).

| ALKBH5 mediates expression of KCNQ1OT1 in an m6A-YTHDF2-dependent manner
We speculated whether the low m6A methylation of KCNQ1OT1 in LSCC is due to the binding with ALKBH5. The online software RIPseq was used to predict the combination of KCNQ1OT1 and ALKBH5. The results showed that random forest (RF) classifier was 0.85 and support vector machine (SVM) classifier was 0.95 ( Figure 6A), which means that the corresponding RNA and protein  The next question we needed to investigate was how m6A affects KCNQ1OT1 expression. Published studies have shown that m6A 'reader' proteins selectively recognize and mediate the degradation of m6A-containing RNA. 23 In order to detect which reader protein is responsible for ALKBH5-mediated KCNQ1OT1 upregulation, we used YTHDF1, YTHDF2, YTHDF3 and YTHDC2 antibodies to carry out RIP analysis. The results showed that only YTHDF2 was significantly associate with KCNQ1OT1 ( Figure 6J). Furthermore, the interaction between KCNQ1OT1 and YTHDF2 was enhanced after silencing ALKBH5 expression in TU212 and AMC-HN-8 cells ( Figure 6K), while ALKBH5 overexpression in HOK cells inhibited this interaction ( Figure 6L). In order to detect whether YTHDF2 is involved in ALKBH5-mediated KCNQ1OT1 upregulation, we further knockdown YTHDF2 in LSCC cells to detect the change of KCNQ1OT1 expression. The results had identified that knockdown of ALKBH5 decreased the expression of KCNQ1OT1 and increased the m6A methylation level. YTHDF2 and ALKBH5 knockdown at the same time rescued the decrease of KCNQ1OT1 expression ( Figure 6M). We further investigated whether YTHDF2 regulates KCNQ1OT1 expression by regulating its stability. We used the transcription inhibitor actinomycin D to detect the half-life of KCNQ1OT1 transcripts in each group, respectively. Indeed, the halflife of KCNQ1OT1 in both cells with ALKBH5 knockdown was reversed after deletion of YTHDF2 expression ( Figure 6N). Therefore, ALKBH5 mediates KCNQ1OT1 expression via an m6A-YTHDF2dependent manner.

| KCNQ1OT1 directly interacts with HOXA9 in LSCC
Starbase 3.0 was used to predict mRNA as a target for KCNQ1OT1.
The results showed that there was a direct binding site between KCNQ1OT1 and HOXA9. The potential complementary binding sites of HOXA9 and KCNQ1OT1 were shown in Figure 7A. Previous studies have shown that HOXA9 was significantly upregulated in LSCC tissues compared to control tissues. 24,25 The analysis of TCGA database showed that HOXA9 was highly expressed in HNSCC, and its expression level was related to tumour grade and lymph node metastasis status ( Figure S3A-C). qRT-PCR verified that HOXA9 was highly expressed in LSCC tissues compared with non-tumour tissues in 86 pairs of LSCC sample cohort, and its expression was positively correlated with the expression level of KCNQ1OT1 ( Figure S4A,B). Besides, we found that high expression of HOXA9 was significantly associated with poor prognosis ( Figure S4C). Using HOK cells as a negative control group, it was found that the expression of HOXA9 in TU212 and AMC-HN-8 was significantly higher ( Figure S4D). IHC showed HOXA9 was expressed significantly higher in LSCC tissues and mainly located in the nucleus ( Figure S5).
Next, we changed the expression of HOXA9 in TU212 and AMC-HN-8 ( Figure S6A,B). CCK-8, Edu, colony formation, transwell and wound healing assays demonstrated that the silencing of HOXA9inhibited LSCC cells proliferation, migration and invasion ability ( Figure S7A-E). The results of qRT-PCR showed that the expression of HOXA9 decreased significantly after KCNQ1OT1 knockdown ( Figure 7B). Luciferase reporter assays were used to further verify that HOXA9 was the target of KCNQ1OT1. The results showed that luciferase activity increased after transfection with wild-type vector. However, the luciferase activity of the mutant type vector was not affected ( Figure 7C)

| DISCUSS ION
The classification of head and neck tumours includes tumours of nasal cavities, paranasal sinuses, oral cavity and larynx, nasopharynx, oropharynx and hypopharynx. 26 More than 90% of head and neck tumours are squamous cell carcinoma 27 ; therefore, they are classified as HNSCC. 28 LSCC is one of the largest subtypes. 26,27 Increasing evidences show that lncRNAs play an important role in the development of tumour and have been proved to be important regulators of signal pathways in the carcinogenesis. 29  For the first time, we integrally analysed differential expression of m6A-lncRNA in LSCC and non-tumour tissues, revealing a tight relationship between m6A methylation and the emergence of LSCC.
We found ALKBH5 increase can promote the development of LSCC by promoting KCNQ1OT1-HOXA9 signalling via a m6A-YTHDF2dependent manner. However, whether the combined expression of ALKBH5-KCNQ1OT1-HOXA9 can be used as a diagnosis and treatment biomarker of LSCC needs more sample data. This study enriched the research on the pathogenesis of LSCC and provided a new idea for the diagnosis and treatment of LSCC. In future, our team will continue to explore the potential clinical significance of lncRNA methylation and its methylation modification in LSCC.

ACK N OWLED G EM ENT
This study was supported by the National Natural Science

CO N FLI C T O F I NTE R E S T
The authors declare that they have no competing interests.

DATA AVA I L A B I L I T Y S TAT E M E N T
The authors declare that the data in this article are available.