m(6)A mRNA Methylation Regulates Ferroptosis in HPSCC by Targeting NFE2L2/NRF2

Emerging as the most abundant posttranscriptional internal mRNA modication in eukaryotes, N6-methyladenosine (m 6 A) modication has gathered tremendous scientic interest in recent years. However, no study addresses the role of m 6 A modication in ferroptosis. Here, we showed that m 6 A modications are decreased in RSL3-induced ferroptosis in hypopharyngeal squamous cell carcinoma (HPSCC). We found that AlkB homolog 5 (ALKBH5), one of the m 6 A demethylases, is the primary factor involved in aberrant m 6 A modication. to evaluate the correlation between and as well as molecular mechanism of and and and candidate


Background
Emerging as the most abundant posttranscriptional internal mRNA modi cation in eukaryotes, N6methyladenosine (m 6 A) modi cation has gathered tremendous scienti c interest in recent years.
However, no study addresses the role of m 6 A modi cation in ferroptosis. Here, we showed that m 6 A modi cations are decreased in RSL3-induced ferroptosis in hypopharyngeal squamous cell carcinoma (HPSCC). We found that AlkB homolog 5 (ALKBH5), one of the m 6 A demethylases, is the primary factor involved in aberrant m 6 A modi cation.

Methods
Bioinformatics analysis, sample analysis, cell biological analyses and transcriptome sequencing were performed to evaluate the correlation between m 6 A modi cation and ferroptosis as well as molecular mechanism of ALKBH5 function. Transcriptome-wide m 6 A-seq and RIP-seq data and following m 6 A dot blot, MeRIP-qPCR, RIP-qPCR and dual luciferase reporter assays were mapped to screen and validate the candidate targets of ALKBH5.

Results
ALKBH5-knockdown impaired ferroptotic cell death in HPSCC. However, overexpression of ALKBH5 has an opposite effect, suggesting that ALKBH5 is a positive regulator of ferroptosis. Mechanistically, ALKBH5-mediated m 6 A demethylation led to a post-transcriptional inhibition of NFE2L2/NRF2, the central player in the regulation of antioxidant molecules in cells, at two m 6 A residues in the 3 -UTR. Therefore, knocking down ALKBH5 subsequently increases the expression levels of NFE2L2/NRF2 and increased cell resistance to ferroptosis. In addition, m 6 A-mediated NFE2L2/NRF2 stabilization relied on the m 6 A reader IGF2BP2. Conclusion ALKBH5 functions as a tumor suppresser through ferroptosis in HPSCC. ALKBH5 destabilizes NFE2L2/NRF2 expression in HPSCC through an m 6 A-IGF2BP2-dependent mechanism. Together, our work uncovers a critical link between ALKBH5-NFE2L2/NRF2 and ferroptosis, providing insight into the functional importance of the reversible mRNA m 6 A methylation and its modulators in HPSCC.

Page 3/27
The Head and neck squamous cell carcinoma (HNSCC) are considered as one of the malignities with the most severe impact on quality of life of patients. HNSCC is a heterogeneous group of tumors arising from the mucosal surfaces of the nasal and oral cavity, oropharynx, larynx, and hypopharynx 1 .
Hypopharyngeal squamous cell carcinoma (HPSCC) accounts 3%-4% of HNSCC, however, it is the most invasive cancer among HNSCC 2 . Although diagnosis and treatment of HPSCC have considerably improved, the survival of recurrent HPSCC can hardly be improved owing to the inadequate effective interventions and precise biomarkers 3 . Therefore, it is quite essential to understand the biological mechanisms of HPSCC malignancy aiming to develop more effective therapeutic strategies.
Triggering apoptotic cell death with anti-cancer drugs is one of the principal approaches for killing cancer cells 4 . However, the effectiveness of apoptosis induction in HNSCC is limited, due to the acquired or intrinsic resistance of cancer cells to apoptosis [5][6][7] . Ferroptosis is being explored as alternative ways to eradicate apoptosis-resistant cancer cells 4 . It is de ned as an iron-catalyzed form of regulated necrosis occurring through excessive peroxidation of polyunsaturated fatty acids (PUFAs) 8 . Recent studies found that the cystine/glutamate antiporter inhibitors erastin and sulfasalazine inhibited head and neck cancer (HNC) cell growth and accumulated lipid reactive oxygen species (lipid ROS), which highlights induction of ferroptotic cell death for killing HNSCC 6 . However, the molecular regulation for ferroptosis in HPSCC remain unclear.
Emerging as the most abundant posttranscriptional internal mRNA modi cation in eukaryotes, N6methyladenosine (m 6 A) modi cation has gathered tremendous scienti c interest in recent years. The and Wilms tumor 1-associated protein (WTAP) 9,10 . The demethylases can remove the m 6 A modi cation which are consist of fat-mass and obesity-associated protein (FTO) and AlkB homolog 5 (ALKBH5) 10 . In addition, m 6 A modi cation functions in condition of its recognition by m 6 A reader proteins. YT521-B homology (YTH) domain-containing family proteins (YTHDF1/2/3) are responsible for mRNA decay 11 .
Accumulating proofs con rmed that the m 6 A modi cation regulates multiple biological functions. Besides embryonic development disorders, tumourigenesis, disordered homeostasis and differentiation of immune cells, and nervous system diseases, acquired chemoradio-resistance, in ammatory, autophagy, dominate the m 6 A-related actions in various cancers [12][13][14] .
Serving as the primary demethylase of m 6 A modi cation, ALKBH5 acts as a biological and pharmacological role in human cancers or non-cancers. ALKBH5 plays a dual role in various cancers through regulating kinds of biological processes. The uncovered regulatory mechanisms of ALKBH5dependent m 6 A modi cation are implicated with long non-coding RNA, cancer stem cell, autophagy, and hypoxia 15 . ALKBH5 is also characterized as a tumor suppressor in various cancers [16][17][18][19][20][21] . In current study, we proved that ALKBH5 directly targeted NFE2L2/NRF2 transcripts and mediated their expression in an m 6 A-dependent manner. Further study identi ed that NFE2L2/NRF2 were the targets of IGF2BP2. Upon ALKBH5 knockdown, NFE2L2/NRF2 transcripts with higher m 6 A levels were captured by IGF2BP2, which resulted in enhanced mRNA stabilization and increased protein expression, thus alleviating ferroptosis.
Together, these ndings demonstrated the functional importance of the m 6 A methylation machinery in ferroptosis regulation, which expands our understanding of such interplay that is essential for development of therapeutic strategies in HPSCC.

Methods
Patient specimens HPSCC patients from individual medical centers who underwent surgery and regular medical surveillance between 2012 and 2020 were used in this study. Para n-embedded surgical specimen was stored permanently in Sir Run Run Shaw hospital pathology department.

Global m 6 A measurement
The global m 6 A levels in mRNA were measured with EpiQuik m 6 A RNA Methylation Quanti cation Kit (Colorimetric) (Epigentek, Farmingdale, NY) following the manufacturer's protocol. m 6 A-sequencing (m 6 A-seq, MeRIP-seq) and RNA-sequencing (RNA-seq) Total RNA was extracted using Trizol reagent (Invitrogen, CA, USA) following the manufacturer's procedure. The total RNA quality and quantity were analysis of Bioanalyzer 2100 and RNA 6000 Nano LabChip Kit (Agilent, CA, USA) with RIN number > 7.0. Approximately more than 200 ug of total RNA was subjected to isolate Poly (A) mRNA with poly-T oligo attached magnetic beads (Invitrogen). Following puri cation, the poly(A) mRNA fractions is fragmented into ~ 100-nt-long oligonucleotides using divalent cations under elevated temperature. Then the cleaved RNA fragments were subjected to incubated for 2h at 4℃ with m 6 A-speci c antibody (No. 202003, Synaptic Systems, Germany) in IP buffer (50 mM Tris-HCl, 750 mM NaCl and 0.5% Igepal CA-630) supplemented with BSA (0.5 µg µl − 1). The mixture was then incubated with protein-A beads and eluted with elution buffer (1 × IP buffer and 6.7mM m 6 A). Eluted RNA was precipitated by 75% ethanol. Eluted m 6 A-containing fragments (IP) and untreated input control fragments are converted to nal cDNA library in accordance with a strand-speci c library preparation by dUTP method. The average insert size for the paired-end libraries was ~ 100 ± 50 bp. And then we performed the paired-end 2×150bp sequencing on an Illumina Novaseq™ 6000 platform at the LC-BIO Biotech ltd (Hangzhou, China) following the vendor's recommended protocol.

Luciferase assay
The fragments of NRF2-3 -UTR containing the wild-type m6A motifs as well as mutant m 6 A motifs (m 6 A was replaced by C) were synthesized at Gene Chemistry(Shanghai, China). The re y luciferase and Renilla luciferase activities in each well were calculated by a dual-luciferase reporter assay system. The ratios between the 3 -UTR reporter of NRF2 and Renilla control were determined 48 h after siRNA treatment. The relative luciferase activity was further normalized to that in cells transfected with the re y luciferase vector control under the same treating conditions. Fire y luciferase activity and Renilla luciferase activity were measured using FLUOstar Omega (BMG LABTECH, Offenburg, Germany). The results were shown in the form of relative re y luciferase activity normalized to Renilla luciferase activity. All the experiments were repeated for three times, and three replicates were conducted for each group.
Mouse treatment and tumor biology studies HPSCC FaDu or Detroit 562 cell lines (1×10 6 cells) stably expressing vector control and construct lentivirus were injected subcutaneously into the right ank of nude mice. Tumor sizes and mice weight in all groups were measured every 3 days for 1-6 weeks. For NRF2 inhibitors treatment, male nude mice at 4 weeks of age were injected subcutaneously into the right ank with Detroit 562 and FaDu cell lines (1×10 6 cells). After 12 days, mice were divided into control and NRF2 inhibitors groups (n = 5 per group). NRF2 inhibitors concentration: 30mg/kg intraperitoneal injection.
All mice were maintained in standard cages in a light and temperature-controlled room and were given standard chow and water. All animal studies were carried out in accordance with Institute of Laboratory Animal Resources guidelines and approved by the University Committee on the Use and Care of Animals at the Zhejiang University.

Statistics methods
Bioinformatic analysis including GO, KEGG, GSEA,circos plot analysis were performed using the OmicStudio tools at https://www.omicstudio.cn/tool. TCGA HNSCC data analysis were performed using GEPIA 36 . NFE2L2/NRF2 functional association data including the protein and genetic interactions, pathways, co-expression, co-localization is summarized by GeneMANIA (http://genemania.org). All statistical analysis was carried out using GraphPad Prism version 7 (GraphPad Software, CA) for Windows. Statistical signi cance was assessed by unpaired two tailed Student's-tests, analysis of variance tests (ANOVA) or Spearman rank correlation. The data are expressed as mean ± SD. *P < 0.05; **P < 0.01; ***P < 0.001. Each experiment was repeated independently at least three times.
More detailed materials and methods are in the Supplementary Materials and Methods.

Results
Sensitivity to ferroptosis was associated with m 6 A modi cation level in HNSCC.
RSL3 is an activator of ferroptosis, which inhibits the activity of GPX4 by covalent bonding with GPX4 and leads to lipid peroxides accumulation 22 . Firstly, we performed different HNSCC cell lines exposed to RSL3 with various concentrations for 24 hours. AlamarBlueassay and LDH release assay showed that sensitivity to RSL3 treatment differed considerably between the various HNSCC cell lines. Inhibition of ALKBH5 expression decrease sensitivity toferroptosis in HPSCC.
HPSCC cells were treated with the lentivirus encoding a short hairpin RNA speci c for ALKBH5.The transfection e ciency was validated by qRT-PCR and western blotting (Fig. S2a, b). A noticeable increase in m 6 A levels was observed in polyadenylated RNAs (poly(A) RNAs) of the ALKBH5-knockdown cells, compared to the wild-type (WT)HPSCC cells, asmeasured by m 6 A dot blot (Fig. 2a) and EpiQuik™ m6A quanti cationassay (Fig. S2c).
To elucidate whether ALKBH5 plays a role in HPSCC ferroptosis, we performed western blotting to compare the novel ferroptosis-related gene expression of ALKBH5-knockdown Detroit 562 cells and control cells. Knocking down ALKBH5 remarkedly increased the expression of GPX4, HO-1and NRF2 protein in shALKBH5 cells, but not in control cells, along with the decreased expression of TFRCand Keap1 protein (Fig. 2b). Immuno uorescence staining showed higher positivity of GPX4 in shALKBH5 cells with RSL3 treatment, compared with control cells (Fig. 2c). Inhibition of the ALKBH5 gene signi cantly reversed RSL3-induced cell death and cellular lipid ROS levels ( Fig. 2d-f). Along with that, using electron microscopy we observed much more shrunken mitochondria in shCON cells than shALKBH5 FaDu cells with RSL3 treatment (Fig. 2g). These data suggested that ALKBH5 positively regulates RSL3-induced ferroptosis.
The overexpression of ALKBH5 sensitized HPSCC cells to RSL3 treatment in vitro and vivo We next examined the effect of ALKBH5 overexpression on cell growth and viability of HPSCC cells. The transfection e ciency was validated by qRT-PCR and western blotting (Fig. S2d). Inhibition of NRF2 and GPX4 could be induced by the ALKBH5 transfection. When the ALKBH5 gene was overexpressed by plasmid transfection, NRF2 and GPX4 expression decreased in HPSCC with or without RSL3 treatment (Fig.3a). Meanwhile, ALKBH5 overexpression inhibitsthe growth and cell viability of HPSCC cells (Fig.3bd). The cell death was enhanced by a combination of RSL3 with ALKBH5 plasmid transfection ( Fig.3b-d).
To clarify the effects of ALKBH5 on HPSCCin vivo, ALKBH5-overexpressing and negative control FaDu cells were implanted into BALB/c mice via subcutaneous injection.RSL3 treatment reduced HPSCC tumor growth and tumor weight in xenograft mouse models (Fig. 3e, f).In addition, overexpression of ALKBH5 signi cantly enhanced the RSL3-inducded tumor suppression. The data strongly suggests that ALKBH5 may be a tumor-suppress gene and mediate ferroptosis in HPSCC.
MeRIP-seq combined with RNA-seq identi ed potential targets of ALKBH5.
To identify potential mRNA targets of ALKBH5 whose m 6 A levels were decreased by ALKBH5 in HPSCC cells, ALKBH5-knockdown and control FaDu cells were selected for transcriptome-wide m 6 A-sequencing (m 6 A-seq, MeRIP-seq) and RNA-sequencing (RNA-seq) assays. Principal component analysis (PCA) showed that two repeats (shCON: shCON_1 and shCON_2; shALKBH5: shK5_1 and shK5_2) of each sample clustered together, suggesting good repeatability among the two replicates of each group (Fig.   S3a). Consisted with the previous research, the most common m 6 A motif"GGAC" is signi cantly enriched in the m 6 A peaks (Fig. 4a).RNA-seq uncovered 548 down-regulated and 564 up-regulated transcripts (p < 0.05) upon ALKBH5 overexpression (Fig. S3b). Peaks were located in protein-coding transcripts and enriched in the 5'UTR and 3′UTR, especially near stop codons, which was coincidence with the m 6 A distribution (Fig. 4b, c).
We next compared the genes with altered-m 6 A modi cations between the control and ALKBH5knockdown cells. The analysis of m 6 A-seq revealed a global hypermethylation of m 6 A in the transcription level after knockdown of ALKBH5 in FaDu cells (Fig. 4d). Considering the role of ALKBH5 in the m 6 A methyltransferase complex, mRNA transcripts carrying hypermethylated m 6 A peaks in FaDu ALKBH5 knockdown cells were likely potential targets. Through analysis of the RNAseq data, we identi ed 967hyper-methylated m 6 A genes whose mRNA transcripts were down-regulated (p < 0.05, Hyper-down) and 49 hyper-methylated m 6 A genes whose mRNA transcripts were up-regulated (p < 0.05, Hyper-up) in ALKBH5-knockdown cells, compared with shCON cells (Fig. 4e). GO enrichment and KEGG pathway analysis revealed that these hyper-methylated m 6 Agene transcripts (p < 0.05) were predominantly enriched in "protein and lipid binding", "lipid metabolic process", "membrane" , "endocytosis" and "ferroptosis" (Fig. S3c). Moreover, GSEA showed that these genes are involved in response_to_oxidized_phospholipids, KEGG_oxidative_phosphorylation, missiaglia _regulated _by_methylation_up (Fig. S3d).It is consisting with the acknowledge that lipid peroxidation of long chain fatty acid and oxidative phosphorylation were correlated with cell membrane oxidative damage in ferroptosis process [23][24][25][26] .182 genes involved in ferroptosis and lipid oxidation andoxidoreductase activity were selected for further validation. Six genes STEAP3, AIFM2, NFE2L2/NRF2, FH(Fumarate hydratase), ZNRF2, PEBP1 whose alterations suppress ferroptosis were nally selected as candidate targets of ALKBH5 (Fig. 4f).
Further qRT-PCR analyses showed the mRNA levels in NFE2L2/NRF2 genes were dramatically increased in ALKBH5-knockdown FaDu cells (Fig. 4g).The Integrative Genomics View (IGV) data peak revealed m 6 A peak enrichment in the 5′-UTR of NFE2L2/NRF2 mRNA that was diminished upon ALKBH5 knockdown ( Fig. 4h,  ALKBH5 abolishes NFE2L2/NRF2mRNA levels and stability depending on its m 6 A methyltransferase activity. MeRIP-qPCR assays with speci c primers aiming at potential m 6 A sites was then applied to con rm the ALKBH5-mediated m 6 A demethylation of NFE2L2/NRF2mRNAin 3' UTR. When compared to the IgG group (pulldown control), an enrichment of NFE2L2/NRF2 mRNA in 3' UTR was obtained by the reaction to m 6 Aspeci c antibody (Fig. 5a).Western blotting and qRT-PCR was then applied to con rm NFE2L2/NRF2expression was upregulated in ALKBH5-knockdown FaDu cells (Fig. S4a).
The m 6 A-modi cation is interpreted by its readers, such as YTH domain-containing proteins, to regulate mRNA fate; insulin-like growth factor 2 mRNA-binding proteins (IGF2BPs; including IGF2BP1/2/3), to promote the stability and storage of their target mRNAs in an m 6 A-dependent manner under normal and stress conditions and therefore affect gene expression output 35 . To analyze the effect of m 6 A modi cation on the stability of ALKBH5 target transcripts, we conducted RNA stability assays. The RNA stability curves showed that knockdown of ALKBH5 prolong the half-life of NFE2L2/NRF2 mRNA in HPSCC cells (Fig. 5f). As YTHDF2, YTHDF3 and YTHDC2 are responsible for mRNA decay, they were hardly the potential effectors participating in the current process.Nevertheless, qRT-PCR data showed that knockdown of IGF2BP2, but not IGF2BP1/3 signi cantly reduced the mRNA level of NFE2L2/NRF2 in HPSCC cells (Fig. S4d,e). Furthermore, the half-life of NFE2L2/NRF2 mRNA was signi cantly decreased when IGF2BP1 was impaired (Fig. S4f). Notably, NFE2L2/NRF2 expression was positively correlated with the IGF2BP2 level in the TCGA HNSCC dataset (Fig. S4g) 36 . RIP-assay showed that IGF2BP2 directly bound to the 3' UTR of NFE2L2/NRF2 mRNA (Fig. 5g). RIP-PCR indicated that knockdown of ALKBH5 signi cantly enhanced the binding e ciency of IGF2BP2 to the 3' UTR of NFE2L2/NRF2 mRNA (Fig. 5h).In addition, overexpression of ALKBH5-WTbut not ALKBH5-Mut signi cantly reduced the binding e ciency of IGF2BP2 to the 3' UTR of NFE2L2/NRF2 mRNA (Fig. 5i).
NFE2L2/NRF2 is functionally essential target gene of ALKBH5 in HPSCC.
We next performed rescue-experiments to illustrate whether NFE2L2/NRF2 participated in the biological function of ALKBH5 in HPSCC. Detroit 562 control and ALKBH5-knockdown cells were transfected with control and siNFE2L2/NRF2 (siNRF2), while FaDu control and ALKBH5-knockdown cells were transfected with control and NFE2L2/NRF2-WT plasmids. The transfection e ciency of were con rmedby qRT-PCR and Western blot analysis (Fig. S5a, b). NFE2L2/NRF2-overexpression downregulated protein levels of GPX4 and FTH1 of ALKBH5-knockdowncells (Fig. S5c). Cell viability assay, LDH release assay and Colonyformation assays indicated that silencing of NFE2L2/NRF2 impairedcell growth and viability of ALKBH5-knockdown cells (Fig. 6a-c). Silencing of NFE2L2/NRF2 also signi cantly increased cellular lipid ROS levelsof ALKBH5-knockdown cells (Fig. 6d). NFE2L2/NRF2 overexpression also led to signi cantly restored cell growth and viability of ALKBH5-overexpression cells (Fig. S5d, e). These results strongly indicated that NFE2L2/NRF2 is a critical target gene of ALKBH5 in HPSCC cells. Inhibition of GPX4 renders HNC cells susceptible to ferroptosis, while activation of the Nrf2-ARE system caused the resistance to GPX4 inhibition 22,34 . Both the light chain and heavy chain of ferritin (FTL/FTH1), stores iron in ferritin, as well as ferroportin (SLC40A1), which is responsible for iron e ux out of the cell, are controlled by NRF2 37,38 . Furthermore, western blot analysis showed that NFE2L2/NRF2 silencing restored the protein levels of GPX4 and FTH1 of ALKBH5-knockdown cells (Fig. 6e). Analysis of TCGA HNSCC datasets also revealed high GPX4 levels positively correlated with NFE2L2/NRF2 expression (Fig. S5f) 36 .
So far, we have proved that ALKBH5 may function as a tumor suppresser through ferroptosis by abolishing NFE2L2/NRF2 mRNA stability and expression in HPSCC cells. Several small molecular NRF2 inhibitors have been developed as anti-tumor drug candidates 39,40 .We treated the mice baring shALKBH5 and control xenograft tumor with one NRF2 inhibitor-ML385 41,42 . ML385 is e cient to inhibit tumor growth and tumor weight in nude mice bearing shALKBH5 cells (Fig. 6f-h). We next performed IHC analyses of HPSCC and paracancerous tissues from patients, to further explore the correlation between expression of ALKBH5 and NFE2L2/NRF2 in HPSCC tissues. As expected, ALKBH5 negatively interrelated with NFE2L2/NRF2 expression (Fig. 6i, j). Taken together, the data show that NFE2L2/NRF2 mediates the ferroptosis regulation of ALKBH5 in HPSCC cells.

Discussion
The m 6 A modi cation exerts direct control over the RNA metabolism including mRNA processing, mRNA exporting, translation initiation, mRNA stability and the biogenesis of long-non-coding RNA (LncRNA), thereby can in uence various aspects of cell function. Accumulating evidence demonstrates that m 6 A modulates gene expression, thereby regulating cellular processes ranging from differentiation, apoptosis, therapeutic resistance, and immune response 43,44 .
ALKBH5-directed m 6 A demethylation is involved in splicing and the production of longer 3'-UTR mRNAs.
Regulatory mechanisms of ALKBH5 including m 6 A-dependent modi cation implicated with long noncoding RNA, cancer stem cell, autophagy, and hypoxia.In epithelial ovarian cancer (EOC), up-regulated ALKBH5 impairs the autophagy and facilitates cell proliferation and invasion 45 .METTL3 relieved autophagic ux in hypoxia/reoxygenation treated cardiomyocytes, however ALKBH5 could reverse this effect 16 . Here our work rstly uncovered a critical link between ALKBH5-NFE2L2/NRF2 and ferroptosis in HPSCC (Fig.7). Knockdown ALKBH5 dramatically suppress the cell response to RSL3-induced ferroptosis. Overexpression ALKBH5 remarkably enhanced the ferroptotic cell death and lipidperoxidation.
Ferroptosis is an iron-dependent, lipid peroxidation-driven cell death cascade that shows great potentials during the cancer therapy 38, 46 . Several small molecules have been identi ed to be functioning in ferroptosis initiation and ferroptosis sensitivity in different cancers 38,46 . A clear understanding of ferroptosis sensitivity in cancers will bene t the clinical practice in applying ferroptosis to cancer therapy.
Importantly we found ALKBH5 expression enhanced sensitivity to ferroptosis in HPSCC in m 6 A-dependent manner.
By RNA-seq and m 6 A-seq analysis, we identi ed NFE2L2/NRF2 as a downstream target of ALKBH5mediated m 6 A modi cation, which was further convinced by the MeRIP-qPCR analysis, luciferase and qRT-PCR assays in control and ALKBH5-knockdown FaDu cells. Tremendous studies highlightedthatNFE2L2/NRF2 is a critical mitigator of both lipid peroxidation and ferroptosis, established NRF2 target genes that mitigate these pathways, as well as the relevance of the NRF2-lipid peroxidationferroptosis axis in disease 38 . Keap1-NRF2 antioxidant system is involved in the RSL3-induced resistance mechanism 47,48 . NFE2L2/NRF2 inhibition or p62 silencing sensitized chemoresistant HNC cells to RSL3. Activation of the NRF2-ARE pathway contributed to the resistance of HNC cells to GPX4 inhibition, and that inhibition of this pathway reversed the resistance to ferroptosis in HNC cells.As discussed in detail in prior studies, GPX4 is an established NFE2L2/NRF2 transcriptional target, is an integral anti-ferroptotic reducer of lipid peroxides 31,38,49 . NFE2L2/NRF2 functional association data including the protein and genetic interactions, pathways, co-expression, co-localization is summarized byGeneMANIA (http://genemania.org). Therefore, inthe current study we did not go into details to explore the potential mechanisms by which NFE2L2/NRF2 regulates the GPX4-dependent lipid peroxidation.
There is also an essential link between NFE2L2/NRF2 function and iron homeostasis 38 . Our previous work revealed that YTHDF1 is closely correlated with iron metabolism and tumor progression in HPSCC.
Mechanistically, YTHDF1 enhanced transferrin receptors (TFRC) expression through an m 6 A-dependent mechanism 50 . To determine whether YTHDF1 recognized the m 6 A modi cation on NFE2L2/NRF2, we reanalyzed the sequence data, in which YTHDF1-knockdown and control FaDu cells were subjected. The m 6 A-seqdata indicated m 6 A peaks in the mRNA and gene expression change. but MeRIP-qPCR failed to validate this nding (data not shown).Then we speculated that ALKBH5 stabilizes NFE2L2/NRF2 expression through RNA stabilization. The de ned m 6 A site is located around the m 6 A enrichment regions in the 5′ UTR. We then observed a signi cantly augmented NFE2L2/NRF2 mRNA stability in HPSCC cells upon the knockdown of IGF2BP2. RIP assay con rmed the binding between IGF2BP2 and NFE2L2/NRF2 mRNA. Functional rescue experiments were additionally performed. Silencing of NFE2L2/NRF2 impaired cell growth and upregulated cellular lipid ROS levels of ALKBH5-knockdown cells (Fig. 6a-d).
In addition, NRF2 inhibitor has anti-cancer effect on ALKBH5-knockdown xenograft tumor (Fig. 6f-h). As ALKBH5 functions as a tumor suppresser through ferroptosis in HPSCC, we speculated that targeting NRF2 may be more e cient in low-expression ALKBH5 tumor via inhibiting the ferroptosis-resistance. These hypotheses need to be validated in vivo experiments in future.
The prognostic value role of ALKBH5 in HPSCC progress may not be fully unveiled in present study.
Actually, the role of ALKBH5 in human cancers was still controversial. It had been found to play an oncogenic or tumor suppressive role in different kinds of cancers 51 . In our study,ALKBH5 was characterized as a critical regulator linking ferroptosis and m 6 A modi cation. Therefore, at least we proved ALKBH5 could be functioning in ferroptosis sensitivity in HPSCC. Inducing ferroptosis would be more e cient in cancer therapy among high-expressed ALBH5 HPSCC.
Taken together, ALKBH5 leads to impaired NFE2L2/NRF2 mRNA stability and decreased expression in HPSCC through an m 6 A-IGF2BP2-dependent mechanism. Our work uncovers a critical link between ALKBH5-NFE2L2/NRF2 and ferroptosis, providing insight into the functional importance of the reversible mRNA m 6 A methylation and its modulators in HPSCC.

Conclusions
Here, for the rst time, we found that m 6 A modi cation is closely correlated with ferroptosis pathway in HNSCC cells. NFE2L2/NRF2 mRNA were found to be regulated by m 6 A-modi cation. The ALKBH5-NFE2L2/NRF2-IGF2BP2 axislikely plays a critical role in the ferroptosis of HPSCC. Ferroptosis as a anticancer therapy may bepromising for treating HPSCC patients withhigh ALKBH5 expression.       NFE2L2/NRF2 is functionally essential target gene of ALKBH5 in HPSCC. a. b. Cell viability assay (a), LDH release assay (b) were performed with control and shALKBH5 Detroit 562 cells transfected with a control vector (siCON) or NFE2L2/NRF2 siRNA (siNRF2) and subsequently treated with 5μm RSL3. c. A schematic illustration was proposed to summarize the relationship among ALKBH5,m6A modi cation, cell death and ferroptosis.

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download.