Targeting SLC3A2 subunit of system XC- is essential for m6A reader YTHDC2 to be an endogenous ferroptosis inducer in lung adenocarcinoma.

The m6A reader YT521-B homology containing 2 (YTHDC2) has been identified to inhibit lung adenocarcinoma (LUAD) tumorigenesis by suppressing solute carrier 7A11 (SLC7A11)-dependent antioxidant function. SLC7A11 is a major functional subunit of system XC-. Inhibition of system XC- can induce ferroptosis. However, whether suppressing SLC7A11 is sufficient for YTHDC2 to be an endogenous ferroptosis inducer in LUAD is unknown. Here, we found that induction of YTHDC2 to a high level can induce ferroptosis in LUAD cells but not in lung and bronchus epithelial cells. In addition to SLC7A11, solute carrier 3A2 (SLC3A2), another subunit of system XC- was equally important for YTHDC2-induced ferroptosis. YTHDC2 m6A-dependently destabilized Homeo box A13 (HOXA13) mRNA because a potential m6A recognition site was identified within its 3' untranslated region (3'UTR). Interestingly, HOXA13 acted as a transcription factor to stimulate SLC3A2 expression. Thereby, YTHDC2 suppressed SLC3A2 via inhibiting HOXA13 in an m6A-indirect manner. Mouse experiments further confirmed the associations among YTHDC2, SLC3A2 and HOXA13, and demonstrated that SLC3A2 and SLC7A11 were both important for YTHDC2-impaired tumor growth and -induced lipid peroxidation in vivo. Moreover, higher expression of SLC7A11, SLC3A2 and HOXA13 indicate poorer clinical outcome in YTHDC2-suppressed LUAD patients. In conclusion, YTHDC2 is believed to be a powerful endogenous ferroptosis inducer and targeting SLC3A2 subunit of system XC- is essential for this process. Increasing YTHDC2 is an alternative ferroptosis-based therapy to treat LUAD.


Background
Lung cancer is the most prevalent cancer worldwide, with an approximate 5-year survival rate of 16.6% [1]. LUAD is the most frequent histological manifestation of lung cancer [2]. Despite the curative methods for treating LUAD have been largely improved in the past few decades [3], the overall survival of LUAD patients still far from satisfactory and this might due to the incomprehensive discovery of the speci c molecular triggers/targets to improve speci c treatments.
Ferroptosis is a newly de ned form of regulated cell death resulting from iron-dependent overwhelming lipid peroxidation [4,5]. Ferroptosis-based therapy is attractive because many types of cancer cells are more metabolic active and sensitive to ferroptosis as compared to normal cells [6,7]. However, intracellular antioxidant system can protect cancer cells against ferroptosis [8,9]. Emerging studies have developed strategies to induce ferroptosis by breaking redox balance in cancer cells. Chemical synthesized small molecules such as erastin and sorafenib can induce ferroptosis via targeting the cystine/glutamate antiporter system X C − . Such effects impair cystine uptake to block subsequent synthesis of glutathione (GSH), the major intracellular antioxidant and prevent activation of its user, such as peroxidase glutathione peroxidase (GPX4) [10,11]. Moreover, the direct inhibition to GPX4 by RASselective lethal 3 (RSL3) can also trigger ferroptosis [12]. Although induction of ferroptosis through small molecules is a promising strategy to treat cancer because of their high e cacy [13,14], some studies also pointed out that these chemicals are highly toxic and hard to metabolize in vivo [15,16]. Hence, it is urgent to explore alternative ways to induce ferroptosis at genetic levels. Unfortunately, to the best of our knowledge, the endogenous ferroptosis inducers in LUAD cells are still unknown. However, Prominin2 has been recently identi ed as a novel endogenous ferroptosis inhibitor to facilitate ferroptosis resistance and prevent cancer cell death [17]. This exciting study hints that we might also have endogenous ferroptosis inducers in our body yet just not been discovered thus far.
N6-methyladenosine (m 6 A) is a WER system consisted by the writers (W), erasers (E) and readers (R).
Writers are methyltransferases while erasers are enzymes catalyze demethylation [18,19]. Unlike writers and erasers, readers belong to the category of RNA binding proteins, recognizing and guiding m 6 A modi ed RNAs for subsequent processing [20,21], thereby they are the executors of m 6 A. YTHDC2 is such an m 6 A reader identi ed in 2017 [22] and its YTH domain is essential for recognizing and binding with the m 6 A-modi ed target RNAs [23,24]. Dysregulation of YTHDC2 has been linked with tumorigenesis and progression [25,26]. Our previous study also demonstrated that YTHDC2 exhibits a tumor suppressor activity to suppress antioxidant function of LUAD cells via accelerating SLC7A11 mRNA decay [24].
SLC7A11 is the major functional subunit of system X C − . As mentioned above that inhibiting system X C − can cause ferroptosis [10,11]. However, whether YTHDC2 can act as an endogenous ferroptosis inducer and also whether suppressing SLC7A11 is su cient for YTHDC2 to trigger system X C − -dependent ferroptosis in LUAD cells still unknown.
SLC7A11 and SLC3A2 are two subunits of system X C − . Despite SLC7A11 is predominantly important for the function of system X C − [27,28], growing evidences have demonstrated that SLC3A2 is equally important for preventing cells from over lipid peroxidation [29,30]. Additionally, SLC3A2 is also required for erastin to induce ferroptosis [10]. SLC3A2 acts as a chaperone protein to support the function of SLC7A11 [31,32]. However, unlike SLC7A11 mRNA, we didn't observe any m 6 A modi cation within the SLC3A2 mRNA via methylated RNA immunoprecipitation sequencing (meRIP-seq) from our previous study [24], demonstrating that SLC3A2 mRNA might not be directly in uenced by YTHDC2 in LUAD cells. SLC3A2 can be transcriptionally regulated under IFN-γ stimulation in the intestinal epithelium [33], indicating that YTHDC2 could indirectly regulate SLC3A2 at transcription level. However, whether and how YTHDC2 regulates SLC3A2 and such effects could co-ordinately induce ferroptosis is not clear.
Thereby, we investigated whether the m 6 A reader YTHDC2 is an endogenous ferroptosis inducer in LUAD cells. We uncover that induction of YTHDC2 to a high level can induce ferroptosis in LUAD cells. In addition to SLC7A11, another subunit of system X C − , i.e. SLC3A2 is also required for this process.
Transcription factor HOXA13 stimulates SLC3A2 transcription; however, YTHDC2 m 6 A-dependently reduces HOXA13 mRNA stability. The data from the current study supplement our knowledge that YTHDC2 is a powerful endogenous inhibitor of system X C − because YTHDC2 can not only suppress SLC7A11 directly but also suppress SLC3A2 indirectly. Thereby, increasing YTHDC2 will be an attractive and alternative ferroptosis-based therapy for LUAD in the future.

Clinical specimen and animal experiments
Clinical samples of paired adjacent and LUAD tissue specimens were the same ones obtained from our previous study [24]. The LUAD patients of cohort#1, #2 and #4 were recruited from Shanghai Chest  with 10% FBS (HyClone, Logan, UT, USA) and 1% penicillin/streptomycin (Invitrogen, Carlsbad, CA, USA). For 3D culture, spheroids were generated in Cultrex® Basement Membrane Extract (BME)-based culture system, which is described previously [24].

Reagents and plasmids
For the reagents used in this study, Deferoxamine ( The primer sequences were listed in Supplementary Table S1. IB, ELISA, IHC and RT-qPCR . The SLC3A2 and HOXA13 protein levels in tissues were also determined by ELISA kits from Lichen Biotech Ltd (Shanghai, China). The tissue microarray assay (TMA) was subjected into IHC using conventional protocols. IHC scores were calculated as we described previously [24]. The antibody used for IHC were: anti-SLC3A2 (Abcam, #ab108300), anti-HOXA13 (Abcam, #ab172570) and anti-4-hydroxynonenal (4-HNE) (Abcam, #ab48506). For RT-qPCR, total RNA was isolated using Trizol reagent (Invitrogen). cDNA was synthesized with RT-PCR Kit (Vazyme Biotech, #P611, Nanjing, China). The relative mRNA and RNA fragments expression level was quanti ed by qPCR with a SYBR Kit (Vazyme Biotech, #Q711) and was normalized to the data from GAPDH or IgG. The qPCR primers are listed in Supplementary Table S1.

Luciferase reporter assay
For the promoter study, truncated versions of the SLC3A2 promoter were cloned into the pGL4-basic plasmids (Promega, Madison, WI, USA). For the analysis of mRNA stability, partial 3' untranslated region (3'UTR) of the HOXA13 mRNA was cloned into the pmir-GLO plasmids (Zuorun Biotech Ltd). For mutation of 3' UTR, adenosine (A) in the m 6 A motif was replaced by a cytosine (C). The WT and mutant (Mut) PCR products were synthesised by Generay Biotech Ltd (Shanghai, China). Luciferase activity was detected by the dual luciferase reporter gene assay kit (Promega). The re y luciferase activities were normalized to the Renilla luciferase activities. The primers used for reporter construction were listed in Supplementary  Table S1.
MeRIP-seq, PAR-CLIP, RIP, RNA pull down assay MeRIP-seq data were obtained from previous study and PAR-CLIP analysis was measured as previously described [24]. For RIP experiments, cell lysates were incubated with magnetic beads loaded with 5 μg antibodies overnight at 4 °C. RNA-protein mixture was then digested by proteinase K. The remaining RNA was nally measured by RT-qPCR with normalization to input. For RNA pull down, probes were synthesized with or without m 6 A modi cation at GGAC or mutant GGCC motifs. Brie y, the probes were labelled by biotin (Takara, Dalian, China). Incubated cell lysed with 3 μg biotinylated probes at 4°C overnight. Then biotin-coupled RNA-protein complex were pulled down using streptavidin magnetic beads (Life Technologies, Carlsbad, CA, USA). After washing, the streptavidin beads were boiled and analyzed for the IB analysis. The primer and probe sequences are supplied in Supplementary Table S1.

EMSA and ChIP
For EMSA, the WT and mut short oligonucleotide probes were synthesized by GenePharma (Shanghai, China). The probes sequences are listed in Supplementary Table S1. EMSA was performed using the gel shift kit (Promega). For the super shift assay, 0.5 μg, 1 μg, or 2 μg of the anti-HOXA13 (Abcam, #ab172570) antibody or IgG (CST, #3900) were added to the nuclear extracts, and incubated at RT for 10 min prior to the DNA binding reaction. All DNA-protein mixtures were resolved in electrophoresis on 5% native polyacrylamide. For ChIP, it was performed using a ChIP-IT ® Express kit (Active Motif, Cat #53008, Carlsbad, CA, USA) according to the manufacturer's instructions. Protein-DNA complexes were incubated with anti-HOXA13 (Abcam, # ab172570) or IgG (CST, #3900) antibodies coupled protein G beads at 4°C overnight. After elution and reverse cross-link, DNA was puri ed for subsequent qPCR. The probes and primers used for EMSA and ChIP-qPCR were listed in Supplementary Table S1.
Fluorescence in situ hybridization (FISH) Brie y, samples were depara nized, digested with proteinase K, prehybridized with prehybridization buffer for 1h at 37°C. Slides were incubated with speci c Cy3-labeled probes for HOXA13 mRNA overnight. After washing 3 times with saline sodium citrate buffer, slides were incubated with anti-YTHDC2 antibody (Invitrogen, # PA5-67256) and uorescent secondary antibody (CST, #4412) for 1h at RT. Finally, nuclei were counterstained with DAPI. All images were collected via a confocal microscope (Leica, wetzlar, German). The probes were listed in Supplementary Table S1.

Cell viability and cell death analysis
Cell viability was determined using the CellTilter-Glo cell viability assay (Promega, #G9682) according to manufacturer's instructions. Cell death was measured by staining with SYTOX Green followed by ow cytometry.

Electron microscopic analysis
To observe the morphological change of mitochondria following induction of YTHDC2, H1299-iYTHDC2 cells were seeded onto 4-well chambered cover glass (Thermo Fisher Scienti c, #155382, Waltham, MA, USA) at a density of 15,000 cells/well and treated with or without Dox for 24h. Images were captured using the Olympus EM208S transmission electron microscope (TEM, hitachi, Tokyo, Japan).

Bioinformatics
UALCAN was mined to predict the transcriptional expression of SLC3A2. The Kaplan-Meier plotter databases were used to analyze survival information in clinical LUAD patients. The statistics of correlation between SLC3A2 and YTHDC2 were carried out with Starbase database. The HOXA13 binding site within the SLC3A2 promoter was predicted by JASPAR and UCSC database.

Statistical analysis
GraphPad Prism 8 software was used for all statistical analyses. P values were calculated with t-test, one-way ANOVA, two-way ANOVA, pearson analysis and the Chi-squared test. Survival curves were generated using the Kaplan-Meier method and compared using the log-rank test. All the values are presented as means ± SEMs from three independent experiments. The indicated p values (*p < 0.05 and **p < 0.01) were considered statistically signi cant. N.S. means non-signi cance.

YTHDC2 is an endogenous ferroptosis inducer in LUAD cells
Because it's di cult to maintain LUAD cells at a healthy state when YTHDC2 was constitutively expressed, we thereby established Dox-inducible YTHDC2-expressing lung epithelia cell BEAS2B, bronchus epithelia cell 16HBE and LUAD cells H1299 and H441 to investigate whether YTHDC2 induces ferroptosis ( Figure. 1A). Similar to that we reported previously [24], YTHDC2 was downregulated in LUAD cells as compared to lung and bronchus epithelia cells ( Figure. 1A). By adding a serial concentration of Dox for 24h, we con rmed that YTHDC2 could be induced dose-dependently ( Figure. 1A). Under such conditions, cell viability was dose-dependently reduced while cell death and lipid ROS generation was induced in H1299 and H441 cells, and such effects could be prevented by simultaneously administrating with Fer-1, a ferroptosis inhibitor ( Figure. 1B), suggesting that YTHDC2-induced ferroptosis did occur in LUAD cells. However, BEAS2B and 16HBE showed a resistance to YTHDC2-induced ferroptosis ( Figure. 1B). These results demonstrated that YTHDC2-induced ferroptosis might be selectively sensitive to LUAD cells. Meanwhile, we also found that cell proliferation was not affected following induction of YTHDC2 for 24h (Supplementary Figure. 1A). Reasoning that a concentration of 1.6 µM Dox was su cient to induce a signi cant YTHDC2 expression and ferroptosis in H1299 and H441 cells ( Figure. 1A-B), we didn't increase Dox concentration any more and xed this concentration to induce YTHDC2 expression for the following study. To investigate the time upon occurrence of ferroptosis following induction of YTHDC2, we monitored H1299 cells and found that YTHDC2 was induced and sustained at a similar high level began at 24h following administrating Dox ( Figure. 1C). Cell detachment and overt cell death were also observed began at that time ( Figure. 1C). Ferroptosis is characterized morphologically by the presence of smaller than normal mitochondria with condensed mitochondrial membrane densities [35,36], induction of YTHDC2 following treating with 1.6 µM Dox for 24h caused similar results in H1299 and H441 cells (Supplementary Figure. 1B-C). These results demonstrated that ferroptosis could be triggered in LUAD cells once upon the expression of YTHDC2 was induced to a high level.
To exclude other forms of cell death, apoptosis inhibitor Z-VAD-FMK, necrosis inhibitor Necrosulfonamide and autophagy inhibitors 3-MA and Chloroquine were co-treated with Dox. Unlike ferroptosis inhibitors DFO and Fer-1, other cell death inhibitors were unable to prevent YTHDC2-induced cell death ( Figure. 1D-E), suggesting that at least apoptosis, necrosis and autophagy are not involved. Apoptosis and autophagy were further excluded because YTHDC2-induced cell death was not aggravated when MG132 and Bortezomib, two proteasome inhibitors can also induce apoptosis, were co-treated ( Figure. 1D-E), and induction of YTHDC2 was still able to induce cell death when the expression of ATG5 and ATG7, two key players of autophagy, were impaired ( Figure. 1F-G).
The 3D cellular models mimic interactions present in tissues better than conventional monolayer 2D cell culture [37,38]. We found that induction of YTHDC2 was also effective to induce ferroptosis in 3D spheroids that generated from H1299 and H441 cells ( Figure. 1H-I). Also, the number and size of the 3D spheroids were signi cantly suppressed following YTHDC2 induction, by which could be reversed by Fer-1 ( Figure. 1H-I), indicating that induction of YTHDC2 might be useful to reduce tumor burden in a ferroptosis-dependent manner.

Discussion
YTHDC2 was previously reported to inhibit LUAD tumorigenesis by suppressing SLC7A11 mRNA in an m 6 A-dependent manner [24]. However, other potential target transcripts of YTHDC2 have not yet been fully veri ed. In the current study, we found that in addition to SLC7A11, another subunit of system X C − , i.e. SLC3A2 is also m 6 A-dependently suppressed by YTHDC2, but rather via an indirect way. Familiar to SLC7A11 mRNA, HOXA13 mRNA can also be destabilized by YTHDC2 because a potential m 6 A recognition site was identi ed within the 3'UTR of HOXA13 mRNA. Thereby, another novel YTHDC2 target transcript is revealed. Interestingly, HOXA13 mRNA encodes HOXA13; a transcription factor has the capacity to stimulate SLC3A2 transcription. Hence, the indirect role of YTHDC2 to suppress SLC3A2 via HOXA13 is also established in this study (Fig. 8).
System X C − is consisted by SLC7A11 and SLC3A2, due to the dual inhibitory roles of YTHDC2 to suppress both of the two subunits; we believe that YTHDC2 is a powerful endogenous inhibitor to system X C − in LUAD cells (Fig. 8). Recently, simultaneous transcriptional downregulation of SLC3A2 and SLC7A11 by interferon gamma (IFNγ) has been revealed to induce ferroptosis in tumour cells [30]. Although IFNγ can e ciently induce ferroptosis in cancer cells, its pharmacological cytotoxicity is still an insurmountable problem [49,50]. In comparison to IFNγ, we report here that elevating the level of YTHDC2 might be a safer and controllable ferroptosis-based therapy to treat LUAD because even induction of YTHDC2 to a high level didn't confer ferroptosis in lung and bronchus epithelia cells.
To the best of our knowledge, others and we have not reported that SLC3A2 mRNA is m 6 A methylated.
This cannot exclude that the SLC3A2 mRNA is truly not to be m 6 A methylated, because at some occasions, failure to identify m 6 A methylation only indicates that the abundance of certain modi cation is not high enough to reach the detection threshold [51]. However, such negative results at least manifest that the direct m 6 A regulation of SLC3A2 is less critical in LUAD cells. As we expected, the indirect m 6 A modulation of SLC3A2 is essential for YTHDC2 to suppress system X C − function, and this is mediated by the m 6 A methylation of the HOXA13 mRNA, which encodes HOXA13, a member of class I homeobox gene family [52]. High expression of HOXA13 is linked with poor survival in cancers [52,53]. HOXA13 also plays pivotal roles in promotion, growth and therapy resistance in a variety of cancers [54][55][56], suggesting that YTHDC2-mediated suppression of HOXA13 is not only effective for the induction of ferroptosis but also has potential functions to impair tumor growth and reverse therapy resistance. Interestingly, overexpression of HOXA13 confers Sorafenib resistance in hepatocellular carcinoma [52], and Sorafenib itself has been well established as an inhibitor to suppress system X C − [10,11]. It's not di cult to understand why this happens because HOXA13 has the potential to stimulate system X C − function via upregulating SLC3A2 expression. Given that HOXA13 and system X C − are closely associated, the m 6 A-dependent indirect inhibition to SLC3A2 via suppressing HOXA13 cannot be ignored for YTHDC2 to act as an endogenous ferroptosis inducer.
YTHDC2 has been previous showed to promote translation of target mRNA by recognizing m 6 A residues located within the protein-coding region [57]. Our present and previous studies [24] provide another two examples, i.e. HOXA13 and SLC7A11 mRNAs to expand our understanding that YTHDC2 also recognizes m 6 A modi cations within the 3' UTR regions. Reasoning that the m 6 A modi cations are highly enriched near the stop codons or within the 3'UTR of the target transcripts [58], YTHDC2 recognition of m 6 A modi cation within 3'UTR might be more important. The fates of the target transcripts, to a certain extent are determined by the location of m 6 A sites [59]. It has been well accepted that the 3'UTR is essential for the stability of variety of mRNAs in eukaryotes because it's also the target site for regulatory molecules, such as microRNAs, or contains regulatory motifs, such as AU-rich elements [60,61]. This also supports that recognizing m 6 A sites located within the 3'UTR might be critical for YTHDC2 to reduce mRNA stability. The mRNA degradation often begins with the removal of the poly(A) tail at the end of 3'UTR, a process named as deadenylation [62]. After this, the mRNAs can be rapidly degraded with RNA exosome [63]. EXOSC10, a catalytic subunit of the RNA exosome, is the major exonuclease responsible for 3'-5' exonucleolytic activity [45,64]. Of note, we found that EXOSC10 is also required for YTHDC2 to destabilize both SLC7A11 and HOXA13 mRNAs. Thereby, the model underlying how YTHDC2 facilitates degradation of target mRNAs following m 6 A methylation might be established; however, how YTHDC2 mediates deadenylation of target transcripts to initiate RNA degradation and how RNA exosome works afterwards are need to be answered in our future work.

Conclusion
The m 6 A reader YTHDC2 is identi ed as an endogenous ferroptosis inducer via targeting system X C − in LUAD cells. In addition to the direct suppression of SLC7A11, the indirect m 6 A-dependent suppression of SLC3A2 via HOXA13 expands our knowledge that both two subunits of system X C − are dispensable for YTHDC2 to induce ferroptosis. Increasing YTHDC2 might be an alternative strategy to supplement small chemical molecules-based ferroptosis therapy to improve treatment of LUAD in the future.

Availability of data and materials
The data sets used and/or analyzed during the current study are available from the corresponding author on reasonable request.
Ethics approval and consent to participate Our study draft was approved by the institutional review boards of Shanghai Chest Hospital and the second Hospital of Shandong University and written informed consents have been received from each patient.

Consent for publication
Written informed consent for publication was obtained from all the participants.       Schematic presentation of the study Brie y, the m6A reader YTHDC2 is an endogenous ferroptosis inducer and its function is to suppress system XC-. In addition to what we've previously reported that the mRNA stability of SLC7A11 is suppressed by YTHDC2, in the present study we also found that SLC3A2, another subunit of system XC-can also be suppressed by YTHDC2 via targeting HOXA13-mediated transcription of SLC3A2, and the m6A methylation and subsequent destabilization of HOXA13 mRNA is a prerequisite for this process.

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