Abstract

   Understanding the regulation of N6-methyladenosine (m^6A), the
   prominent internal modification in mRNA, fosters the development of
   potential therapeutic strategies for human cancers. While the m^6A
   demethylases FTO and ALKBH5 are recognized for their crucial roles in
   various cancers, their impact on lymphoid leukemia remains uncertain.
   Using T cell acute lymphoblastic leukemia (T-ALL) as a model system, we
   identify FTO as a unique vulnerability in T cell leukemia. Knockout of
   FTO, but not ALKBH5, significantly suppresses leukemia initiation and
   progression. Mechanistic analysis reveals that FTO heightens ELK3 mRNA
   stability in an m^6A-dependent manner. Elevated ELK3 in turn
   transcriptionally activates the expression of glycolytic genes.
   Pharmacological inhibition of FTO suppresses ELK3 expression, hampers
   glycolysis and manifests remarkable antileukemia efficacy. Our findings
   unravel the crucial role of FTO in T-ALL and highlight the FTO-ELK3
   axis as a key nodule during leukemogenesis, thereby providing a
   fundamental basis to harness selective FTO antagonist for T-ALL
   therapeutics.
     __________________________________________________________________

   The RNA m^6A demethylase FTO represents a promising therapeutic target
   in T cell leukemia.

INTRODUCTION

   T cell acute lymphoblastic leukemia (T-ALL) is a lethal and aggressive
   hematological malignancy characterized by aberrant proliferation of
   immature thymocytes that afflicts both children and adults ([58]1).
   Although overall survival (OS) of 80% in the pediatric setting has been
   achieved using intensive multi-agent combination chemotherapeutic
   protocols, it is often accompanied by substantial short-term and
   long-term side effects ([59]2, [60]3). Moreover, OS rates for adult
   patients with T-ALL are lower than 50% due to higher treatment-related
   toxicities ([61]4). A considerable proportion of patients with T-ALL
   confront the challenge of disease relapse, with a particularly grim
   prognosis for cases resistant to chemotherapy ([62]5). These clinical
   challenges underscore an urgent need for the implementation of targeted
   therapies. Molecular genetic analyses and sequencing studies have
   identified genetic abnormality involved in T-ALL development ([63]6).
   However, clinically approved targeted treatment options for patients
   with T-ALL remain limited, highlighting the requirement for a more
   in-depth understanding of the molecular pathogenesis that may yield
   potential effective therapeutic strategies.

   N^6-methyladenosine (m^6A) is the most prominent internal modification
   in mRNA that regulates gene expression at the post-transcriptional
   level ([64]7, [65]8). The m^6A modification is deposited on mRNA by the
   methyltransferase-like 3 (METTL3)-methyltransferase-like 14
   (METTL14)-Wilms’s tumor 1-associating protein (WTAP) methyltransferase
   complex and can be reversibly removed by demethylases alkB homolog 5
   (ALKBH5) or fat mass- and obesity-associated protein (FTO) ([66]9).
   This modification plays a crucial role in determining the fate of
   modified RNA molecules, influencing various aspects of RNA metabolism.
   In many cases, it governs mRNA stability and translation, contributing
   to the regulatory landscape of cellular processes ([67]7, [68]8). As
   such, the RNA m^6A modification is considered as a prominent epigenetic
   regulation of gene expression, implicated in many pathophysiological
   processes including cancer ([69]10). Accumulating experimental data
   have demonstrated that the global abundance of m^6A, as well as the
   expression levels of methylases and demethylases, is dysregulated in
   human cancer. These factors are crucial for cancer initiation and
   progression ([70]11–[71]15). Yet, the functional significance of these
   m^6A regulators in T-ALL remains unclear.

   In this study, we systematically analyzed the genes encoding m^6A
   writer and eraser proteins in primary T-ALL samples and revealed that
   the m^6A demethylases, FTO and ALKBH5, are more abundant in transformed
   thymocytes. These results suggest that RNA m^6A demethylation plays a
   vital role in T cell leukemogenesis. To provide proof of concept, we
   used the T cell–specific lineage knockout mouse strains that delete Fto
   or Alkbh5 to establish NOTCH1-driven T-ALL. T cell leukemogenesis is
   strictly dependent on FTO, as ablation of FTO but not ALKBH5
   significantly blocks T-ALL onset and progression. Our multi-omics
   integration revealed ELK3 as a pivotal FTO downstream target.
   FTO-mediated removal of m^6A from ETS domain-containing protein Elk-3
   (ELK3) mRNA enhances its stability, leading to increased levels of
   ELK3, which in turn activates the transcription of glycolysis-related
   genes, such as phosphoglycerate kinase 1 (PGK1), thereby promoting
   glycolysis in T-ALL. With several FTO inhibitors advancing to clinical
   stages, our findings suggest that these small-molecule drugs could be
   exploited for the treatment of T cell leukemia.

RESULTS

Fto is selectively required for murine T-ALL development

   We compared the expression of m^6A modulators in 57 primary T-ALL
   specimens with 21 normal thymocyte samples ([72]16) and observed that
   both m^6A demethylases (FTO and ALKBH5) were significantly up-regulated
   in T-ALL ([73]Fig. 1A). Expression changes of the methyltransferase
   complex components (METTL3 and WTAP) were moderate and insignificant,
   and METTL14 was even decreased in transformed T cells (fig. S1A),
   suggesting that RNA m^6A modifications become less prevalent during
   transformation. In line with these data, m^6A dot blots confirmed
   weaker m^6A signals in murine T-ALL cells as compared with normal
   thymocytes (fig. S1B). Analysis of gene expression profiles from
   additional primary T-ALL samples revealed that elevated FTO and ALKBH5
   expression is not restricted to any specific T-ALL subtype nor is it
   associated with common mutations found in T-ALL (fig. S1, C and D)
   ([74]17–[75]24).

Fig. 1. Fto is required for NOTCH1-induced murine T-ALL development.

   [76]Fig. 1.
   [77]Open in a new tab

   (A) Expression analysis of FTO and ALKBH5 among 57 primary patients
   with T-ALL and 21 normal T cells from [78]GSE33469 and [79]GSE33470.
   The expression values are presented as box-and-whisker plots with a
   log2 scale. (B) Experimental scheme for intracellular NOTCH1
   (ICN1)–induced T-ALL model. Bone marrow (BM) lineage-negative (Lin^−)
   cells were isolated from the mice as depicted and infected with
   MigR1-GFP-ICN1 retrovirus and then transplanted into irradiated
   recipient mice. (C and D) Percentage of GFP^+ leukemia cells in
   peripheral blood (PB) of recipient mice at the indicated time after BM
   transplantation (BMT). Recipient mice were transplanted with
   ICN1-transduced Fto^fl/fl (n = 14), Fto^−/− (n = 16) (C), Alkbh5^fl/fl
   (n = 12), or Alkbh5^−/− (n = 14) (D) BM Lin^− cells as indicated. (E)
   Representative images of Wright-Giemsa staining of PB from recipient
   mice as indicated. Mice were euthanized at the same time when the
   Fto^fl/fl or Alkbh5^fl/fl control became moribund. The scale bar
   represents 20 μm. (F) Spleen image (left) and flow cytometry analysis
   of GFP^+ leukemia cells in the spleen (right) are shown. (G)
   Hematoxylin and eosin (H&E) staining (left) and PCNA immunohistological
   images (right) in spleen sections from indicated mice. The scale bar
   represents 20 μm. PCNA, proliferating cell nuclear antigen. (H and I)
   Kaplan-Meier survival curves for recipient mice transplanted with
   ICN1-transduced BM Lin^− cells from Fto^fl/fl (n = 14), Fto^−/−
   (n = 13) (H), Alkbh5^fl/fl (n = 11), or Alkbh5^−/− (n = 13) (I). Data
   are represented as means (±SD). Data were analyzed by Welch t test (A),
   Mann-Whitney test [(C) and (D)], and log-rank test [(H) and (I)].
   **P < 0.01 and ***P < 0.001. n.s., nonsignificant.

   We hypothesized that FTO and ALKBH5 play crucial proleukemogenic roles
   in T-ALL, akin to their functions in acute myeloid leukemia (AML)
   ([80]12–[81]14). To test the notion, we obtained Fto^fl/fl and
   Alkbh5^fl/fl mice and bred them with the Lck-Cre strain to generate
   Fto^fl/flLck-Cre or Alkbh^fl/flLck-Cre (hereafter referred to as
   “Fto^−/−” and “Alkbh5^−/−”) mouse strains that allow specific deletion
   of Fto or Alkbh5 in T cell lineage (fig. S2, A to D). To establish
   NOTCH1-induced T-ALL murine model ([82]25), we transduced MigR1–ICN1
   (intracellular NOTCH1) retroviruses, with green fluorescent protein
   (GFP) as a surrogate marker, into bone marrow (BM) lineage negative
   (Lin^−) cells from Fto^fl/fl, Fto^−/−, Alkbh5^fl/fl, or Alkbh5^−/− mice
   and transplanted these cells into irradiated recipient mice ([83]Fig.
   1B). In comparison to the Fto^fl/fl control cohort, deletion of Fto
   resulted in a significantly lower percentage of GFP^+ leukemia cells,
   which were also CD4^+CD8^+ T cell blasts, in the peripheral blood (PB)
   after BM transplantation (BMT) ([84]Fig. 1C and fig. S2E).
   Unexpectedly, the deficiency of Alkbh5 did not suppress the peripheral
   production of GFP^+ leukemia cells ([85]Fig. 1D). Wright-Giemsa
   staining of PB smears corroborated this finding ([86]Fig. 1E). Deletion
   of Fto resulted in less severe splenomegaly and fewer disseminated
   leukemia cells in the spleen. Conversely, Alkbh5 knockout did not
   decrease the spleen size nor the infiltration of leukemia cells in the
   spleen ([87]Fig. 1F and fig. S2F). Hematoxylin and eosin (H&E) and
   immunohistochemical staining of proliferating cell nuclear antigen
   (PCNA) further demonstrated that deficiency of Fto instead of Alkbh5
   restrained leukemia cells infiltration and proliferation ([88]Fig. 1G).
   As a result, only deletion of Fto significantly delayed leukemia onset
   and prolonged survival in the recipient mice ([89]Fig. 1, H and I).
   Deficiency of Alkbh5 or Fto in leukemia cells was confirmed by
   quantitative polymerase chain reaction (qPCR) and immunoblotting (fig.
   S2G), ruling out the possibility that the phenotypic consequence from
   Alkbh5^−/− mice was due to incomplete gene knockout. Of note, Fto
   deficiency in T cell lineage did not appreciably affect thymus size,
   overall thymocyte numbers, or the distribution of T cell progenitor
   subsets (fig. S3), suggesting an insignificant role of FTO in normal T
   cell development. Our findings therefore support that FTO presents a
   unique and specific dependency in T-ALL.

FTO demethylase activity is required for T-ALL progression

   We next assessed the role of FTO in T-ALL progression after leukemia
   onset. Fto^fl/fl mice were crossed with the Mx1-Cre strain (fig. S4A),
   and the resulting mice were used as donors for MigR1-GFP-ICN1
   retroviral transduction and subsequent transplantation. Following the
   procedure, polyinosinic-polycytidylic acid (pIpC) was intravenously
   injected to induce Fto knockout ([90]Fig. 2A). Consistently, ablation
   of Fto hampered the peripheral production of GFP^+ leukemia cells and
   reduced leukemia burden in BM, spleen, and liver ([91]Fig. 2, B to D,
   and fig. S4B). As a result, Fto depletion after leukemia initiation
   considerably prolonged leukemia mice survival and life span ([92]Fig.
   2E). In contrast, conditional Alkbh5 deletion by Mx1-Cre did not affect
   leukemia cell infiltration in vivo, as well as the OS (fig. S4, C to
   F).

Fig. 2. FTO promotes T-ALL in an m^6A-dependent manner.

   [93]Fig. 2.
   [94]Open in a new tab

   (A) Experimental scheme for ICN1-induced T-ALL model using Fto^fl/fl or
   Fto^fl/flMx1-Cre mice as donors. BM Lin^− cells were isolated and
   infected with MigR1-GFP-ICN1 retrovirus, followed by transplantation
   into irradiated recipient mice. pIpC (polyinosinic-polycytidylic acid)
   (10 mg/kg) was injected 3 weeks later. (B) Percentages of GFP^+ cells
   in PB of recipient mice at the indicated time after transplantation.
   Fto^fl/fl (n = 15), Fto^fl/flMx1-Cre (n = 17). (C) Spleen image (left)
   and flow cytometry analysis of GFP^+ leukemia cells in the spleen
   (right) of recipients 8 weeks postengraftment. (D) Representative
   images of H&E staining for BM, spleen, and liver of recipients 8 weeks
   postengraftment. The scale bar represents 20 μm. (E) Kaplan-Meier
   curves of recipient mice transplanted with ICN1-transduced BM Lin^−
   cells from Fto^fl/fl and Fto^fl/flMx1-Cre mice (n = 14 per group). (F)
   The growth curves of KOPTK1, CUTLL1 cells, and two primary T-ALL cells,
   following FTO knockdown by two independent shRNAs. (G) KOPTK1 cell
   growth upon FTO depletion, with or without the restoration of wild-type
   (WT) or m^6A demethylase-inactive mutant (MUT) FTO (left). Protein
   expression was analyzed by immunoblotting and shown on the right. (H)
   In vivo bioluminescence imaging showing leukemia burden in luciferase
   expressing KOPTK1 xenograft (n = 5 per group). The unit of radiance is
   “photons/second/cm^2/steradian”. (I) Kaplan-Meier survival curves of
   KOPTK1 xenografts shown in (H) (n = 5 per group). Data are presented as
   means (±SD) from two biological replicates, each performed with three
   technical replicates (F and G). Data were analyzed by Mann-Whitney test
   (B), two-way ANOVA with Bonferroni’s multiple comparisons [(F) and
   (G)], and log-rank test [(E) and (I)]. **P < 0.01 and ***P < 0.001.
   n.s., nonsignificant.

   We then assessed the functional importance of FTO in human T-ALL cells
   using shRNA-mediated gene silencing. As shown in [95]Fig. 2F and fig.
   S4G, FTO depletion substantially suppressed the growth of human T-ALL
   KOPTK1 and CUTLL1 cells, as well as primary T-ALL cells from patients
   (table S1). Cell growth inhibition was associated with a notable
   increase in global m^6A levels (fig. S4H), leading us to hypothesize
   that the FTO enzymatic activity contributes to its proleukemic
   function. To test this, wild-type (WT) FTO or the enzymatically
   inactive mutant (MUT) FTO (H231A/D233A) ([96]12) was expressed in
   FTO-depleted cells. Only WT FTO expression was sufficient to rescue the
   growth defect ([97]Fig. 2G and fig. S4I), reinforcing the essential
   role of FTO demethylase activity in T-ALL cells.

   We further established a human xenograft using KOPTK1 cells expressing
   firefly luciferase. Equal amounts of FTO-depleted cells with or without
   the exogenous expression of WT/MUT FTO were intravenously injected into
   immunodeficient NOD.Cg-Prkdcscid Il2rgtm1Vst/Vst (NPG) mice (fig. S4, J
   and K). FTO knockdown markedly inhibited leukemia progression in vivo,
   as evidenced by much weaker luciferase signals over time compared with
   the control ([98]Fig. 2H and fig. S2L). As a result, mice transplanted
   with FTO-deficient cells exhibited a significantly extended lifespan
   ([99]Fig. 2I). Supplementation of WT, but not catalytically inactive
   FTO, significantly boosted leukemia cell expansion in vivo ([100]Fig.
   2H and fig. S4L), leading to shortened life span of leukemia mice
   ([101]Fig. 2I). These findings consolidate the indispensable role of
   FTO demethylase activity in T-ALL progression.

Integrative analysis of differential RNA expression and m^6A modification
identifies potential FTO targets in T-ALL

   To comprehensively view FTO-regulated RNA transcripts in T-ALL, we
   performed RNA sequencing (RNA-seq) in KOPTK1 cells with or without FTO
   shRNA expression. A total of 597 genes were significantly up-regulated
   whereas 682 genes were down-regulated ([102]Fig. 3A and fig. S5A).
   Differentially expressed genes were subjected to Kyoto Encyclopedia of
   Genes and Genomes (KEGG) enrichment analysis, and we identified the top
   eight pathways in which the down-regulated genes were enriched.
   Notably, FTO knockdown caused significant suppression of genes involved
   in “transcriptional misregulation in cancer”, which ranked at the top
   ([103]Fig. 3B). We next explored the m^6A methylomes using methylated
   RNA immunoprecipitation sequencing (MeRIP-seq/m^6A-seq) in KOPTK1
   cells. Consistent with prior studies in other tumor cells ([104]26),
   FTO loss led to a noticeable increase in m^6A abundance ([105]Fig. 3C).
   These m^6A modifications were mapped primarily to the classical DRACH
   motif ([106]Fig. 3D) and enriched in coding sequences (CDSs), 3′
   untranslated regions (3′UTRs) with the abundance reaching peak around
   stop codons ([107]27, [108]28) ([109]Fig. 3E and fig. S5B). FTO
   depletion induced 1619 significantly increased peaks (m^6A-hyper) and
   907 decreased signals (m^6A-hypo) (fig. S5C), highlighting a prominent
   gain of m^6A methylation on mRNA transcripts following FTO knockdown.

Fig. 3. Identification of potential key targets of FTO in T-ALL cells.

   [110]Fig. 3.
   [111]Open in a new tab

   (A) Heatmap showing differentially expressed genes in KOPTK1 cells with
   or without FTO shRNA expression. Differential genes are derived from
   RNA-seq data and defined as |fold change| ≥ 1.5, P value <0.05. (B)
   KEGG enrichment analysis of down-regulated transcripts in FTO-depleted
   KOPTK1 cells. (C) Cumulative distribution function representing the
   log2 peak intensity of m^6A-modified sites in KOPTK1 cells with or
   without FTO knockdown. The P value was calculated using two-tailed
   Kolmogorov-Smirnov test. (D) Top consensus sequence motif identified
   within m^6A peaks by hypergeometric optimization of motif enrichment
   (HOMER) in KOPTK1 cells with or without FTO knockdown. (E) Metagene
   profiles of m^6A peak distribution along a normalized transcript
   composed of three rescaled nonoverlapping segments: 5′UTR, CDS, and
   3′UTR in KOPTK1 cells with or without FTO knockdown. (F) Distribution
   of genes exhibiting a significant change in both m^6A level and overall
   transcript level in FTO-depleted KOPTK1 cells. (G) KEGG pathway
   enrichment analysis with the 634 hyperdown genes shown in (F). (H)
   Bubble-rank plot of hyperdown transcripts in (F). The genes encoding
   transcription factors are highlighted in red. (I) Growth curves of
   FTO-depleted KOPTK1 cells with or without the ectopic expression of
   ELK3 (left). Immunoblots of FTO and ELK3 are shown on the right. Data
   are presented as means (±SD) from two biological replicates, each
   performed with three technical replicates. Data were analyzed by
   two-way ANOVA with Bonferroni’s multiple comparisons. ***P < 0.001.

   To pinpoint crucial FTO target RNA transcripts whose altered gene
   expression results from FTO-mediated m^6A demethylation, we merged
   differentially expressed genes with those with m^6A abundance change.
   Combined analysis of RNA-seq and m^6A-seq data from KOPTK1 cells
   identified 656 hypermethylated m^6A peaks. Among these, the
   corresponding mRNA transcripts were predominantly down-regulated
   (634/97%, Hyper-down), while a small fraction were up-regulated (22/3%,
   Hyper-up) ([112]Fig. 3F). Given their potential role as direct FTO
   targets with proleukemogenic functions, we focused on the Hyper-down
   group. KEGG analysis enriched the pathway associated with
   “transcriptional misregulation in cancer” ([113]Fig. 3G), analogous to
   the enrichment of FTO-induced genes shown in [114]Fig. 3B. We further
   conducted a detailed analysis of genes encoding transcription factors,
   identifying 11 transcription factors in the Hyper-down group ([115]Fig.
   3H). Among those, BCL6 has been previously characterized as a
   proleukemogenic transcription factor in T-ALL ([116]29). These results
   support the effectiveness of our integrative analysis in revealing key
   T-ALL associated transcription factors. Down-regulation of these genes
   following FTO knockdown were validated by qPCR (fig. S5D).

   Notably, ELK3 is the most down-regulated transcription factor among all
   identified. ELK3, also known as NET, encodes a member of the ETS-domain
   transcription factor family and the ternary complex factor subfamily
   ([117]30). ELK3 has been identified as a susceptibility locus for
   childhood acute lymphoblastic leukemia (ALL) ([118]31) and several
   studies showed that ELK3 plays a critical oncogenic role in various
   cancers, including glioma, pancreatic, and gastric cancers ([119]30).
   We found ELK3 expression was significantly elevated in human T-ALL
   samples (fig. S5E). Overexpression of ELK3 significantly rescued the
   growth defect of T-ALL cells caused by FTO loss ([120]Fig. 3I and fig.
   S5F), suggesting that ELK3 is a functionally important downstream
   target gene regulated by FTO. Consequently, we focused on ELK3 for
   further regulatory and functional studies.

FTO regulates ELK3 mRNA stability in an m^6A-dependent manner

   We examined genome browser tracks and affirmed that depletion of FTO
   significantly attenuated the expression of ELK3, concomitant with a
   notable elevation of the m^6A peak in the exon 4 ([121]Fig. 4A). The
   decrease in ELK3 RNA and protein expression was confirmed in human
   T-ALL KOPTK1 and CUTLL1 cells following FTO knockdown ([122]Fig. 4B and
   fig. S6A). Consistently, down-regulation of Elk3 was also found in
   primary murine Fto^−/− leukemia cells ([123]Fig. 4C), suggesting that
   the activation of ELK3 expression by FTO is an evolutionarily conserved
   regulatory mechanism in human and mouse. Enforced expression of WT FTO,
   but not catalytically inactive mutant, substantially restored ELK3
   expression in FTO-depleted cells ([124]Fig. 4D and fig. S6B). We
   confirmed an increase in ELK3 m^6A modifications upon FTO knockdown
   ([125]Fig. 4E), and a direct binding of the ELK3 transcript by FTO in
   human T-ALL KOPTK1 and CUTLL1 cells ([126]Fig. 4F). These data suggest
   that the FTO-mediated ELK3 RNA m^6A demethylation is responsible for
   high ELK3 expression.

Fig. 4. FTO regulates ELK3 expression through enhancing its mRNA stability.

   [127]Fig. 4.
   [128]Open in a new tab

   (A) Genome browser tracks showing the mRNA abundance and m^6A peaks of
   ELK3 transcripts in KOPTK1 cells with or without FTO depletion. (B)
   qPCR analysis of ELK3 expression in FTO knockdown T-ALL cells as
   indicated. (C) mRNA (top) and protein (bottom) levels of Elk3 in
   primary murine T-ALL cells from Fto^fl/fl or Fto^−/− BMT as shown in
   fig. S2G (n = 4 per group). (D) qPCR (left) and immunoblot (right)
   analysis of ELK3 expression in FTO-depleted KOPTK1 cells rescued with
   WT or MUT FTO. (E) Gene-specific m^6A-qPCR analysis of the ELK3 mRNA in
   T-ALL cells as indicated. (F) FTO-RIP qPCR validation of FTO
   interaction with the ELK3 mRNA in T-ALL cells. (G) The mRNA half-life
   (t[1/2]) of ELK3 in FTO knockdown KOPTK1 cells with enforced expression
   of WT or MUT FTO. (H) Relative luciferase activity showing the WT or
   MUT ELK3 exon4 firefly luciferase reporter in 293T cells transduced
   with Control or FTO shRNA. (I) YTHDF2-RIP qPCR showing the binding of
   YTHDF2 to ELK3 mRNA in FTO knockdown leukemia cells. (J) The ELK3 mRNA
   level in FTO-deficient T-ALL cells with or without YTHDF2 shRNA
   expression. (K) Scatterplot showing the correlation of FTO and ELK3 in
   108 T-ALL samples from databases of National Omics Data Encyclopedia
   (NODE) (OEP002748). The expression levels were log[2](TPM). Pearson
   correlation coefficient (R) and P value (P) are marked. Data are
   presented as means (±SD) of technical triplicates, and the experiments
   were independently repeated at least twice with the similar results
   [(B) and (D) to (J)]. Data were analyzed by one-way ANOVA with Tukey’s
   multiple comparisons [(B), (D), and (J)], two-way ANOVA with
   Bonferroni’s multiple comparisons (H), and unpaired two-tailed
   Student’s t test [(C), (E), (F), and (I)]. ***P < 0.001, n.s.,
   nonsignificant.

   Given the strong link between m^6A modification and mRNA stability
   ([129]32), we treated FTO-depleted KOPTK1 and CUTLL1 cells with the
   transcription inhibitor actinomycin D and measured the half-life of
   ELK3 transcripts. FTO ablation caused a noticeable decrease in the
   half-life of ELK3 mRNA (fig. S6C). Exogenous expression of WT FTO, but
   not inactive mutant, effectively restored ELK3 mRNA stability
   ([130]Fig. 4G and fig. S6D). We also conducted a luciferase reporter
   assay and observed that FTO knockdown significantly decreased the
   activity of the luciferase construct fused with the ELK3 exon 4
   in-frame but not the mutant sequence with m^6A site being disrupted
   ([131]Fig. 4H and fig. S6E). YTH N6-Methyladenosine RNA Binding Protein
   F2 (YTHDF2) is the primary m^6A reader that promotes the degradation of
   the m^6A-containing RNAs ([132]33, [133]34). RIP-qPCR analysis
   confirmed that YTHDF2 interacts with the ELK3 mRNA (fig. S6F). This RNA
   binding of YTHDF2 was more profound upon FTO knockdown ([134]Fig. 4I).
   In addition, YTHDF2 knockdown diminished the effects of FTO deficiency
   on ELK3 mRNA and protein down-regulation ([135]Fig. 4J and fig. S6G),
   demonstrating that YTHDF2 is the m^6A reader for the ELK3 mRNA,
   responsible for its m^6A-modified RNA degradation. We analyzed primary
   patients with T-ALL samples, and observed that ELK3 expression was
   significantly and positively correlated with FTO ([136]17) ([137]Fig.
   4K), further supporting the regulatory role of FTO in sustaining ELK3
   expression. In contrast, ELK3 expression was neither associated with
   ALKBH5 in primary human T-ALL (fig. S6H) nor regulated by Alkbh5 in
   NOTCH1-driven murine T-ALL cells (fig. S6I). On the basis of these
   results, we demonstrate that FTO specifically catalyzes the
   demethylation of ELK3 mRNA, resulting in its detachment from YTHDF2 and
   thereby stabilization of the RNA transcript.

ELK3 serves as a master regulator promoting glycolysis in T-ALL

   To explore the functional significance of ELK3 as an FTO downstream
   effector, we knocked down ELK3 in CUTLL1 and KOPTK1 cells, along with
   two primary T-ALL cells. ELK3 depletion suppressed T-ALL cell growth,
   analogous to the phenotype of FTO deficiency ([138]Fig. 5A and fig.
   S7A). We next conducted RNA-seq following ELK3 knockdown in CUTLL1 and
   KOPTK1 cells. KEGG pathway analysis of significantly down-regulated
   genes revealed top five enriched pathways, with glycolysis showing the
   most remarkable enrichment in both cells ([139]Fig. 5B and fig. S7B).
   Consistently, gene set enrichment analysis (GSEA) also showed
   significant enrichment of glycolysis hallmark genes that were decreased
   in ELK3-deficient cells ([140]Fig. 5C and fig. S7C). Seahorse analysis
   confirmed that ELK3 depletion significantly reduced both basal and
   compensatory glycolytic rates ([141]Fig. 5D and fig. S7D). Moreover, a
   targeted metabolomics analysis of glycolysis metabolites validated the
   reduction of several intermediates, including glucose-1-phosphate,
   fructose-6-phosphate, 3-phospho-glyceric acid, and phosphoenolpyruvic
   acid, upon ELK3 depletion ([142]Fig. 5E). Consistently, ELK3 depletion
   also down-regulated multiple genes encoding glycolytic enzymes
   ([143]Fig. 5F), suggesting that ELK3 acts as a key transcription factor
   orchestrating glycolysis regulation in T-ALL.

Fig. 5. ELK3 acts as a functionally important target of FTO through
regulating glycolysis.

   [144]Fig. 5.
   [145]Open in a new tab

   (A) Growth curves of ELK3-depleted T-ALL cell lines (CUTLL1 and KOPTK1)
   and primary T-ALL cells. Data are presented as the mean ± SD from two
   biological replicates, each with three technical replicates. (B) KEGG
   pathway analysis of down-regulated genes in ELK3-depleted KOPTK1 cells.
   (C) GSEA plot showing enrichment of gene sets of hallmark glycolysis in
   ELK3 knockdown CUTLL1 cells. (D) Glycolytic proton efflux rate
   (glycoPER) analyzed in ELK3-depleted CUTLL1 cells using a Seahorse
   extracellular flux analyzer. (E) Heatmap presentation of
   glycolysis-related metabolites from metabolomic analysis in KOPTK1
   cells (n = 3 per group). (F) qPCR analysis of glycolytic gene
   expression in ELK3-depleted KOPTK1 cells. (G) Density plot of ELK3
   (red) and H3K27ac (blue) ([146]GSE76783) occupancy around the
   transcription start site (TSS) of genes significantly down-regulated in
   ELK3-depleted KOPTK1 cells. ChIP-seq signal normalized with input. (H)
   Genome browser tracks of ELK3 and H3K27ac ChIP-seq signals at the
   promoters of PGK1 (top) and PFKFB4 (bottom). (I) ChIP-qPCR validation
   of ELK3 and H3K27ac binding to PGK1 and PFKFB4 promoters in KOPTK1
   cells. (J) qPCR analysis of glycolysis-related genes in FTO-depleted
   KOPTK1 cells with or without the overexpression of ELK3. (K) Glycolytic
   proton efflux rate (glycoPER) analyzed in FTO-depleted CUTLL1 cells
   with or without the overexpression of ELK3. Data are presented as means
   (±SD) of technical triplicates, and the experiments were independently
   repeated at least twice with the similar results [(D), (F), (I), (J),
   and (K)]. Data were analyzed by one-way ANOVA with Tukey’s multiple
   comparisons (J), two-way ANOVA with Bonferroni’s multiple comparisons
   (A), and unpaired two-tailed Student’s t test [(D), (F), and (I)].
   *P < 0.05, **P < 0.01, and ***P < 0.001.

   To determine ELK3 direct transcriptional targets, we performed
   chromatin immunoprecipitation sequencing (ChIP-seq) in KOPTK1 cells. In
   agreement with a previous report ([147]35), the most prominent
   proportion (87.07%) of the peak signals located in the promoters and
   ELK3 preferentially occupied the CCGGAAG sequence (fig. S7, E and F).
   ChIP-seq analysis revealed a notable enrichment of ELK3 near the
   transcription start sites (TSS), accompanied with the signal of H3K27ac
   ([148]36), the epigenetic hallmark of transcriptional activation
   ([149]Fig. 5G). Genome browser tracks demonstrated that ELK3, along
   with H3K27ac, directly bound to the promoter regions of various
   glycolytic genes, including PGK1 and PFKFB4 ([150]Fig. 5H).
   Gene-specific ChIP-qPCR validation further corroborated these findings
   ([151]Fig. 5I and fig. S7, G and H). Ectopic expression of ELK3
   efficiently rescued the expression of glycolytic genes, as well as the
   glycolysis defect following FTO ablation ([152]Fig. 5, J and K, and
   fig. S7, I and J). Our findings thus highlight the functional
   significance of ELK3 as a downstream effector of FTO, driving T cell
   leukemogenesis by regulating glycolysis.

FTO inhibitor Dac51 exhibits potent anti–T-ALL efficacy in vitro and in vivo

   We have thus far demonstrated that FTO is essential for the onset and
   maintenance of T-ALL. FTO inhibitors have shown remarkable antitumor
   effects ([153]26, [154]37, [155]38). To test whether FTO serves as a
   therapeutic target in T-ALL, we treated various T-ALL cells with
   selective FTO inhibitors Dac51 or FB23-2 ([156]26, [157]37). As shown
   in [158]Fig. 6A and fig. S8 (A and B), both compounds inhibited cell
   growth in a dose-dependent manner. Since Dac51 achieved more profound
   suppression, with 300 nM Dac51 producing an effect equivalent to 1500
   nM FB23-2 ([159]Fig. 6A and fig. S8, A to C), we chose Dac51 for
   further study. It is important to note that Dac51 induced profound
   lethal effects on various T-ALL cells while sparing normal BM cells
   ([160]Fig. 6B), demonstrating its selective cytotoxicity toward
   transformed T cells.

Fig. 6. The FTO inhibitor Dac51 shows antileukemia efficacy in vitro and in
vivo.

   [161]Fig. 6.
   [162]Open in a new tab

   (A) Growth curves of indicated T-ALL cells treated with varying doses
   of Dac51. (B) CCK8 viability assessment of various T-ALL cells and
   normal BM cells (BM #1 and #2) after Dac51 treatment for 24 hours. (C)
   qPCR (left) and immunoblots (right) showing the ELK3 mRNA and protein
   levels upon Dac51 (150 or 300 nM) treatment for 36 hours. (D) qPCR
   analysis of glycolysis-related genes in CUTLL1 cells following Dac51
   treatment (300 nM, 36 hours). (E) Assessment of glycolytic proton
   efflux rate (glycoPER) in KOPTK1 cells treated with 300 nM Dac51 for 36
   hours. (F) Growth curves of 300 nM Dac51 treated CUTLL1 cells with or
   without ectopic ELK3 expression. (G) Schematic of Dac51 treatment in
   vivo. Primary human or murine T-ALL cells were transplanted into NPG or
   WT C57BL/6J mice, respectively, followed by Dac51 (4 mg/kg) treatment.
   (H) Assessment of leukemia cells in the PB from T-ALL mice shown in
   (G). Percentages of hCD45^+ leukemia cells were analyzed at day 20
   post-transplantation (patient #1 and #2), and Kras^G12D/Notch1^L1601P
   GFP^+ murine leukemia cells were analyzed at day 14. (I) H&E and
   immunohistochemistry (PCNA) staining of spleens from T-ALL mice shown
   in (H) at 28, 20, and 14 days post-transplantation, respectively. Scale
   bar, 20 μm. (J) Kaplan-Meier survival curves of T-ALL mice shown in
   (H). Data are presented as the mean (±SD) from two biological
   replicates, each with three technical replicates [(A) and (F)], and
   from technical triplicates with experiments repeated at least twice
   [(B) to (E)]. Data were analyzed by one-way ANOVA with Tukey’s multiple
   comparisons (C), two-way ANOVA with Bonferroni’s multiple comparisons
   [(A) and (F)], unpaired two-tailed Student’s t test [(D) and (H)] and
   log-rank test (J). *P < 0.05, **P < 0.01, and ***P < 0.001.

   Similar to FTO knockdown, Dac51 treatment decreased ELK3 RNA and
   protein levels ([163]Fig. 6C and fig.S8D), suppressed multiple
   glycolytic genes ([164]Fig. 6D and fig. S8E), and inhibited glycolysis
   ([165]Fig. 6E and fig. S8, F and G). Notably, enforced expression of
   ELK3 antagonized Dac51-mediated growth defect ([166]Fig. 6F and fig.
   S8, H and I), further corroborating the role of the FTO-ELK3 axis in
   promoting T-ALL cell growth.

   To evaluate the antileukemia efficacy of Dac51 in vivo, we established
   patient-derived T-ALL xenografts (PDX) using primary T-ALL cells from
   various subtypes. Primary human T-ALL cells from two T-ALL subtypes
   (TAL1 and LMO2) were transplanted into NPG mice, respectively, followed
   by vehicle or Dac51 injection once per day ([167]Fig. 6G). As shown in
   [168]Fig. 6H, Dac51 treatment significantly reduced leukemia cell
   expansion in PB. H&E and immunohistochemical staining revealed fewer
   leukemia cell infiltration in the spleen following Dac51 treatment
   ([169]Fig. 6I and fig. S8J). Consequently, administration of Dac51
   substantially delayed T-ALL onset and prolonged the survival of
   recipient mice compared with the vehicle group ([170]Fig. 6J and fig.
   S8K). We also set up allograft models using primary murine T-ALL cells
   harboring NOTCH1 mutation (L1601P) in the Kras^G12D genetic background
   ([171]39) or overexpressing ICN1 alone. Similarly, Dac51 exhibited
   strong antileukemia activity in these animal models ([172]Fig. 6 H to
   J, and fig. S8, J and K). In addition, minimal body weight differences
   were detected between treated or untreated groups (fig. S8 L). These in
   vitro and in vivo data provide strong evidence supporting Dac51 as a
   promising anti–T-ALL candidate drug with tolerable toxicity.

DISCUSSION

   T-ALL is a heterogeneous disease resulting from a multistep
   transformation process in which genetic alterations in epigenetic
   regulators and chromatin modifiers are frequently found ([173]1).
   Accumulating data demonstrate that aberrant activity or expression of
   these epigenetic factors plays a critical role in the initiation and
   development of T-ALL, leading to increasing number of small-molecule
   epigenetic drugs being evaluated in clinical trials ([174]40). Despite
   large amount of work focuses on targeting epigenetic regulators that
   alter DNA methylation and post-translational modifications on histones,
   little is known about how the RNA modification and the regulators
   contribute to T-ALL progression. The RNA m^6A modification represents a
   unique epigenetic mechanism in regulation of gene expression and a
   potential therapeutic target in human cancers, given that multiple m^6A
   modifiers are deregulated during tumorigenesis ([175]10, [176]41). Both
   the m^6A writers and erases are often found up-regulated in tumors as
   compared to normal tissue ([177]10, [178]41), suggesting that
   extraordinary m^6A dynamics occurs in transformed cells. This is
   probably associated with swift epigenetic adaption of tumor cells to
   survive in arduous environment. Notably, the m^6A demethylases FTO and
   ALKBH5 have been reported to be functionally essential in a variety of
   tumors such as AML and breast cancer ([179]42, [180]43). Disruption of
   either one leads to tumor growth suppression, most likely due to the
   regulation of respective key protumorigenic gene expression by each
   demethylase ([181]12–[182]14). We reveal that FTO, but not ALKBH5, is a
   unique dependency factor in T-ALL. Multi-omics studies and experimental
   validations identify transcriptional factor ELK3 as a key and selective
   downstream effector regulated by FTO in an m^6A-dependent manner. We
   further demonstrate ELK3 as a major transcription factor promoting
   glycolysis, providing additional mechanistic insights in molecular
   pathogenesis of T-ALL. These findings lay a theoretical foundation for
   therapeutic targeting of FTO in T-ALL ([183]Fig. 7).

Fig. 7. A working model depicting the crucial role of FTO and targeted
strategy in T-ALL.

   [184]Fig. 7.
   [185]Open in a new tab

   Elevated FTO expression in T-ALL maintains low levels of m^6A
   methylation on many transcripts including the ELK3 mRNA, which prevents
   the YTHDF2-mediated degradation. Augmented ELK3 expression in turn
   up-regulates glycolysis-related genes like PGK1 and PFKFB4, thereby
   enhancing glycolysis. Selective FTO inhibitor Dac51 leads to increased
   m^6A modification on ELK3 mRNA, resulting in decreased ELK3 expression,
   ultimately lowering glycolytic capacity and suppressing T-ALL
   progression. See more details in the text.

   Whereas both m^6A demethylases are more abundant in T-ALL in comparison
   to normal thymocytes, T-ALL onset and progression selectively depend on
   FTO. Although ALKBH5 may regulate m^6A levels in T-ALL cells, the
   affected genes could be nonessential for the initiation and maintenance
   of T-ALL. The oncogenic or tumor-suppressive feature of demethylases
   may not be determined by overall m^6A levels, but rather by m^6A
   dynamics on specific key target genes driving leukemogenesis, such as
   ELK3, which is evidently not a shared target with ALKBH5. The
   differential dependency of FTO and ALKBH5 in the same tumor context has
   also been reported in bladder cancer ([186]44, [187]45) and B cell
   lymphoma ([188]46). It is also interesting to note that ALKBH5, not the
   FTO, plays a crucial role in the function and development of T cells
   ([189]47, [190]48). We therefore reason that normal and transformed T
   cells exploit distinct m^6A machinery to exert its physiological and
   pathological functions.

   ELK3 has been reported to act as an oncogene in various types of cancer
   ([191]30) and reside in a risk locus for childhood ALL ([192]31).
   However, the precise mechanism by which ELK3 participates in
   tumorigenesis remains undefined. Our findings help decipher the
   oncogenic role of ELK3 and the mechanism of action contributing to
   T-ALL progression. ELK3 acts as a master regulator in controlling the
   expression of glycolysis genes by directly binding to their promoters.
   Intriguingly, ELK3 was identified as a glucose-dependent gene in murine
   hypothalamic cells through analysis of the glucose-dependent
   transcriptome ([193]49), suggesting a potential reciprocal regulation.
   FTO has been reported to directly regulate PFKP and LDHB expression to
   affect glycolysis in AML ([194]50, [195]51). Alternatively, we found
   that in T-ALL, FTO regulates glycolysis-related genes through ELK3.
   These data suggest that FTO reinforces glycolysis to drive leukemogenic
   program through multiple mechanisms and in a context-dependent manner.
   Metabolic rewiring in T-ALL is predominantly regulated by major
   transcriptional factors, such as NOTCH1, MYC, and RUNX2
   ([196]52–[197]54), and our findings identify a previously unrecognized
   transcriptional factor ELK3 contributing to glycolytic capacity.

   Our study thus highlights the indispensable role of FTO as m^6A
   demethylase involved in glycolytic metabolism in T-ALL and provides a
   promising therapeutic strategy for T-ALL through the use of selective
   FTO inhibitor Dac51. Of note, FTO also mediates mRNA or snRNA m^6A[m]
   demethylation, in addition to m^6A demethylation ([198]55). Whether the
   m^6A[m] demethylase activity of FTO contributes to T-ALL progression
   remains unresolved. Considering that the overall abundance of m^6A[m]
   is significantly lower than that of m^6A ([199]37, [200]55), the potent
   antileukemia efficacy triggered by Dac51 treatment is likely driven
   primarily by m^6A demethylation. However, we cannot entirely rule out
   the possibility that FTO also exerts protumor effects through
   demethylation of m^6A[m]. Given the multifunctional feature of FTO and
   its extraordinary protein stability (half-life >24 hours), designing
   agents that specifically degrade FTO, such as proteolysis-targeting
   chimeras (PROTAC), may offer a more effective strategy for achieving
   prolonged and durable antileukemia efficacy.

MATERIALS AND METHODS

Cell culture

   T-ALL cell lines KOPTK1, CUTLL1, G4A2, and T6E were provided by W. Pear
   (University of Pennsylvania, USA). 293T cell line was purchased from
   American Type Culture Collection. Human T-ALL KOPTK1 and CUTLL1 cells,
   murine T-ALL G4A2, and T6E cells were cultured in RPIM-1640 with 10%
   fetal bovine serum (FBS; Hyclone), 1% penicillin-streptomycin (Gibco,
   15140122), 1% nonessential amino acids (Gibco, 11140050), 2
   mM l-glutamine (Gibco, 35050061), and 1 mM sodium pyruvate (Gibco,
   11360070). 293T cells were cultured in DMEM (Hyclone) supplemented with
   10% FBS (Gemini) and 1% penicillin-streptomycin. All cell lines are
   authenticated with short tandem repeats analysis and generally cultured
   at 37°C under 5% CO[2] for fewer than 6 months after resuscitation and
   regularly tested for mycoplasma contamination using MycoAlert (Lonza).

   Human BM cells were isolated from healthy donors and cultured in
   StemSpan SFEM II (StemCell Technologies, 09605) medium supplemented
   with rhSCF (100 ng/ml), rhFLT3L (100 ng/ml), rhTPO (50 ng/ml), and 1%
   penicillin-streptomycin. Primary human T-ALL cells used in this study
   were obtained from participants who provided informed written consent,
   and these cells were cocultured with MS5-DL1 feeder cells as previously
   described ([201]56). All the cytokines were purchased from PeproTech.
   The study was conducted in accordance with protocols approved by the
   Ethics Committees of Soochow University (SUDA20201208H06) and Zhongnan
   Hospital of Wuhan University (2024063K).

Mice

   The Alkbh5^fl/fl and Fto^fl/fl mice were constructed as previously
   reported ([202]14, [203]48). Lck-Cre and Mx1-Cre strains were described
   as before ([204]25). Both male and female mice at 8 to 10 weeks old
   were used for experiments, with age- and gender-matched littermates as
   control. NOD.Cg-Prkdc^scidIl2rg^tm1Vst/Vst (NPG) mice were purchased
   from Beijing Vitalstar. All mice were bred and maintained under
   specific pathogen–free conditions at the animal facility of Medical
   Research Institute at Wuhan University. All animal husbandry and
   experiments comply with the protocol approved by the Animal Care and
   Use Committee of Medical Research Institute (MRI2020-LAC40), Wuhan
   University.

Plasmid construction

   The shRNAs targeting human FTO, ELK3, or YTHDF2 were cloned in pLKO.1.
   The coding regions FTO and ELK3 were amplified from the cDNA of KOPTK1
   cells by PCR, and then cloned into the pCDH lentiviral vector. The
   enzymatic inactive mutant of FTO (H231A D233A) and the FTO construct
   that is resistant to the shFTO #1 by introducing silent mutations were
   generated using Q5 Site-Directed Mutagenesis Kit (NEB, E0554). For
   luciferase reporter assay, the WT or the mutant Exon4 of ELK3 was
   inserted into pGL3-Promoter Vector (Promega) using ClonExpress Ultra
   One Step Cloning Kit (Vazyme, C115). Escherichia coli DH5α and Stbl3 E.
   coli were used in the transformation. shRNA targeting sequences are
   listed in table S2.

Lentivirus preparation and infection

   Lentiviruses from the pLKO.1 and pCDH-CMV-Pruo constructs were packaged
   with helper plasmid pMD2.G, psPAX2 (Addgene) in 293T cells. In brief, 9
   μg of lentiviral constructs, 6 μg of psPAX2, and 3 μg of pMD2.G were
   cotransfected into 293T cells in 100-mm cell culture dish with
   Lipofectamine 2000 (Invitrogen, 11668019). Viral supernatants were
   generally collected 48 and 72 hours after transfection and concentrated
   with UltracelRegenerated Cellulose (Millipore). The appropriate amount
   of lentivirus was directly added into T-ALL cells in the presence of
   polybrene (8 μg/ml; Santa Cruz Biotechnology, sc-134220) and incubated
   at 37°C for 15 min, followed by centrifugation at 32°C, 1000g for 90
   min. For cell lines, transduced cells were selected with puromycin (3
   μg/ml; InvivoGen, ant-pr-1) 48 hours post-transduction. For primary
   T-ALL cells, GFP^+ cells were isolated by flow cytometry 72 hours after
   transduction. These freshly transduced cells were subsequently prepared
   for functional analysis.

RNA isolation and quantitative RT-PCR analysis

   Total cellular RNA was extracted using TRIzol reagent (Takara, 9109).
   One microgram of RNA was used to synthesize cDNA, using the ReverTra
   Ace qPCR RT Kit according to the manufacturer’s instructions (TOYOBO,
   FSQ-101). qPCR was performed with ChamQ SYBR qPCR Master Mix (Vazyme,
   Q311-02) in CFX Connect Real-Time PCR System (Bio-Rad). Relative
   expression of the mRNA was calculated by the 2^−ΔΔCt method and
   normalized to ACTIN. All the specific PCR primer sequences are listed
   in table S2.

Generation and analysis of murine ICN1-driven T-ALL model

   Mouse BM transplantation was conducted as previously described
   ([205]25). ICN1 expression retrovirus was generated by cotransfection
   of 9 μg of MigR1-ICN1 and 6 μg of packaging plasmid (pCgp and pHIT)
   into 293T cells in 100-mm cell culture dish with Lipofectamine 2000
   (Invitrogen, 11668019). BM cells were collected from 6- to 8-week-old
   mice, and lineage negative (Lin^−) cells were enriched with the Mouse
   Direct Lineage Cell Depletion Kit according to manufacturer’s
   instructions (Miltenyi Biotec, 130-110-470). The Lin^− cells were
   cultured in the presence of mTPO (50 ng/ml), mSCF (100 ng/ml), and
   mFLT3-L (50 ng/ml) and infected with concentrated MigR1-ICN
   retroviruses twice in the presence of polybrene (8 mg/ml; Santa Cruz
   Biotechnology, sc-134220). All the cytokines were purchased from
   PeproTech. GFP^+ Lin^− cells (1 × 10^5) were transplanted via tail vein
   injection into irradiated [6 grays (Gy)] 6- to 8-week-old WT C57BL/6
   mice. For the T-ALL progression analysis shown in [206]Fig. 2, all
   recipients were intraperitoneally injected with pIpC (10 mg/kg;
   InvivoGen, tlrl-pic-5) in physiological water (NaCl 0.9%) three times
   with a 1-day interval, while the GFP^+ leukemia cells were detected
   after 3 weeks of transplantation. GFP^+ leukemia cells in PB were
   analyzed periodically after transplantation.

In vivo bioluminescence imaging

   KOPTK1 cells were infected with pWPXLd-GFP-Luciferase lentivirus
   packaged with pMD2.G and psPAX2, and then GFP^+ cells were sorted by BD
   FACS Aria III. GFP^+ KOPTK1 cells were transduced with the appropriate
   lentivirus, and subsequently infected cells were subjected to selection
   using puromycin (3 μg/ml) to establish stable expression of luciferase.
   Seven- to 8-week-old NPG mice were irradiated (1.5 Gy) before tail vein
   injection of one million cells. For bioluminescence imaging, mice were
   given the substrate d-luciferin (150 mg/kg) by intraperitoneal
   injection, then anesthetized with isoflurane and imaged with Ami HTX
   (Spectral Instruments Imaging). The bioluminescent signals were
   quantified using Aura imaging software (Spectral Instruments Imaging).

Dac51 treatment in PDX and murine allograft

   To establish PDX model, primary T-ALL cells from patients were injected
   into the tail veins of 7- to 8-week-old irradiated (1.5 Gy) NPG mice at
   appropriate cell numbers: 1 × 10^5 cells for patient #1 and 2 × 10^6
   cells for patient #2 and patient #3. For allograft model, 1 × 10^6
   GFP^+ murine T-ALL cells driven by Kras^G12D/Notch1^L1601P ([207]39) or
   1 × 10^5 GFP^+ ICN1-driven murine T-ALL cells were injected into the
   irradiated (4.5 Gy) WT C57BL/6J via tail injection. These mice were
   randomly grouped and treated with Dac51 (4 mg/kg per day) or vehicle
   control via daily intraperitoneal injection starting on day 3
   post-transplantation.

RNA-seq and data analysis

   For RNA-seq, total RNA was extracted using TRIzol. Genomic DNA was
   removed by in-solution deoxyribonuclease (DNase) I treatment (Thermo
   Fisher Scientific, 89836). Poly(A) mRNA was subsequently purified from
   1 μg of total RNA using Dynabeads mRNA purification kit (Invitrogen,
   61006). RNA libraries were constructed using NEB Next Ultra Directional
   RNA Library Prep Kit (NEB, E7760) and sequenced as 150-bp paired-end
   reads by Illumina NovaSeq 6000 (Beijing Annoroad Co. Ltd). Sequencing
   quality was evaluated using FastQC and reads were mapped to human
   genome version GRCh38 by STAR ([208]57). Differential expressed genes
   were analyzed by DESeq2 ([209]58) and defined as |fold change| > 1.2
   and P value <0.05 unless otherwise specified. We used KOBAS ([210]59)
   for KEGG pathway analysis and used clusterProfiler ([211]60) for GSEA
   analysis.

m^6A-seq and data analysis

   For m^6A-seq, m^6A-IP was performed according to the protocol with some
   modifications ([212]61). Total RNA was extracted from KOPTK1 cells
   using TRIzol reagent and then treated with DNase I (Thermo Fisher
   Scientific, 89836). mRNA was purified with Dynabeads mRNA Purification
   Kit (Invitrogen, 61006). One microgram of mRNA was fragmented into ~150
   bp with RNA Fragmentation Reagents (Invitrogen, AM8740) at 90°C for 1
   min. Twenty microliters of Dynabeads Protein A (Invitrogen, 10002D) was
   washed twice with IP buffer [10 mM tris-HCl (pH 7.5), 150 mM NaCl, and
   0.5% Igepal CA-630] and mixed with 1 μg of m^6A antibody (Abcam,
   ab151230) before orbital rotation at 4°C for 2 hours. The prepared mRNA
   was subsequently added for immunoprecipitation. The beads were washed
   twice with IP buffer, twice with low-salt reaction buffer [10 mM
   tris-HCl (pH 7.5), 50 mM NaCl, and 0.5% Igepal CA-630], and twice with
   high-salt reaction buffer [10 mM tris-HCl (pH 7.5), 500 mM NaCl, and
   0.5% Igepal CA-630] before eluted with m^6A elution buffer. The eluates
   and mRNA of input were purified by the RNA Clean & Concentrator-5 kit
   (Zymo Research, R1013) to construct libraries with the NEBNext Ultra
   Directional RNA Library Prep Kit (NEB, E7760) and sequenced as 150-bp
   paired-end reads by NovaSeq 6000 (Beijing Annoroad Co. Ltd). Reads were
   mapped to human genome version GRCh38 by STAR. m^6A peaks were called
   by exomePeak version 2.16.0 ([213]62), and 100-bp sequences around the
   peak center were extracted for motif enrichment with hypergeometric
   optimization of motif enrichment (HOMER) ([214]63).

Gene-specific m^6A qPCR

   m^6A-qPCR was performed according to the m^6A-seq procedure. Rabbit
   immunoglobulin G (IgG; Abclonal, AC005) was used as control. Purified
   mRNA fragments from IP and input samples were reverse transcribed using
   ReverTra Ace qPCR RT Kit, followed by qPCR with the primers listed in
   table S2.

Metabolomics analysis

   Targeted analysis of glycolysis metabolites was performed with the
   assistance of Shanghai Biotree Biotech according to a previous
   publication ([215]64). In brief, ten million KOPTK1 cells were
   collected and quenched 30 s in liquid nitrogen. Metabolites were
   extracted using extraction solvent
   (V[methanol]:V[acetonitrile]:V[water] = 2:2:1, −20°C) and dried in a
   vacuum concentrator. Metabolites were analyzed by TSQ Quantiva (Thermo
   Fisher Scientific) and quantified according to calibration curves.

Chromatin immunoprecipitation sequencing

   ChIP-seq was performed as previously with some modifications ([216]65).
   Ten million KOPTK1 cells were cross-linked in phosphate-buffered saline
   (PBS) containing 1% formaldehyde (Sigma-Aldrich, F8775) for 10 min and
   then quenched with 0.125 M final concentration of glycine for 5 min at
   room temperature. Chromatins were sheared to an average size of ~300 bp
   using the Bruptor (Diagenode). Soluble chromatin containing supernatant
   was used for immunoprecipitation in IP buffer [10 mM tris-HCl (pH 8.0),
   100 mM NaCl, and 1 mM EDTA] with the presence of 7 μg of ELK3 antibody
   (Novus, NBP1-83960) rotating overnight at 4°C. The DNA-antibody mixture
   was incubated with 30 μl of Protein A/G (Invitrogen, 80105G) for 4
   hours at 4°C. The immunoprecipitate was washed three times with washing
   buffer I [20 mM tris-HCl (pH 8.0), 150 mM NaCl, 2 mM EDTA, 1% Triton
   X-100, and 0.1% SDS], washing buffer II [20 mM tris-HCl (pH 8.0), 500
   mM NaCl, 2 mM EDTA, 1% Triton X-100, and 0.1% SDS], and washing buffer
   III [50 mM Hepes-KOH (pH 7.5), 1 mM EDTA, 250 Mm LiCl, 1% NP40, and
   0.7% Na-deoxycholate], then eluted with 300 μl of elution buffer [50 mM
   tris-HCl (pH 8.0), 10 mM EDTA, and 1.0% SDS]. All buffers used were
   added with protease inhibitors. Cross-link reversal of
   immunoprecipitated DNA was performed overnight at 65°C and proteins
   were digested with 2 μl of proteinase K (Invitrogen, AM2546) at 55°C
   for 4 hours. DNA was purified using the phenol-chloroform extraction
   and ethanol precipitation.

   Library preparation was done using the VAHTS Universal DNA Library Prep
   Kit for Illumina V4 (Vazyme, ND610-01), followed by sequencing on a
   NovaSeq 6000 (Beijing Annoroad Co. Ltd). Reads were aligned to the
   human genome (GRCh38) using bowtie2 ([217]66). MACS2 was used to call
   peaks and an initial threshold q value of 0.01 was set as the cutoff
   ([218]67). The IP signal was normalized to input and bigwig files were
   visualized in the IGV software. Peak annotation and motif analysis were
   performed on ChIPseeker and HOMER, respectively ([219]63, [220]68).

m^6A dot blot assay

   m^6A dot blot assays were performed as described previously with minor
   modifications ([221]12, [222]46). In brief, total RNA was extracted
   from cell pellets using TRIzol. The concentration of the total RNA
   sample was determined by Nanodrop 2000 (Thermo Fisher Scientific).
   Indicated amounts of total RNA were denatured at 65°C for 5 min before
   samples were loaded onto the Amersham Hybond-N+ membrane (GE
   Healthcare, RPN303B). After cross-linking under 254-nm ultraviolet for
   2 min, membranes were washed for 5 min at room temperature, blocked in
   5% milk in PBS with Tween 20 (PBST) buffer for 1 hour, and then
   incubated with the m^6A antibody (CST, 56593S) at 4°C overnight. The
   membranes were washed three times with PBST and incubated with
   horseradish peroxidase (HRP)–conjugated goat antirabbit IgG (Jackson
   ImmunoResearch, JAC-111-035-003) for 1 hour at room temperature before
   developing with Clarity Western ECL Substrate (Bio-Rad, 1705061) and
   imaged on the Chemidoc Touch imaging system (BioRad). Equal loading of
   RNA was confirmed by staining of the membrane with 0.02% methylene blue
   in 0.3 M sodium acetate (pH 5.2).

RNA immunoprecipitation

   Ten million cells were collected and washed with ice-cold PBS before
   resuspended in 1.1 ml of RIP lysis buffer [50 mM tris-HCl (pH 7.0), 150
   mM NaCl, 2 mM EDTA, and 0.5% Igepal CA-630], containing RNasin (100
   U/ml; Promega, N2111) and protease inhibitors (Roche, 04693159001).
   Lysates were centrifuged at 4°C to remove cell debris after incubating
   on ice for 30 min. Supernatants were collected into a new tube and 50
   μl was kept for input control. The remaining lysates were divided into
   equal volumes (500 μl) to incubate with either 4 μg of antibody
   (Proteintech; FTO, 27226-1-AP; YTHDF2, 24744-1-AP) or rabbit IgG
   (Abclonal, AC005) rotating overnight at 4°C. Thirty microliters of
   Dynabeads Protein G (Invitrogen, 10004D), prewashed three times with
   RIP lysis buffer, was added to the samples and rotated at 4°C for 4
   hours. Beads were washed with RIP lysis buffer five times before RNA
   was purified by TRIzol. cDNA synthesis was performed using the ReverTra
   Ace qPCR RT Kit according to the manufacturer’s instructions. qPCR was
   used to measure bound RNA abundance.

RNA stability assay

   For RNA stability assay, two million cells were treated with
   actinomycin D (Sigma-Aldrich, A9415) at a final 5 μg/ml concentration
   and collected at indicated time points. Total RNA was extracted,
   reverse transcribed, and used for reverse transcriptase qPCR. ACTIN was
   used as endogenous control. mRNA’s degradation rate was estimated
   according to the previously published paper ([223]69).

Dual-luciferase reporter and mutagenesis assays

   The WT or mutant ELK3-Exon4 was subcloned into the downstream of
   firefly luciferase of pGL3-Promoter Vector in-frame (Promega). For
   dual-luciferase reporter assay, 400 ng of WT or mutant ELK3-Exon4
   constructs and 20 ng of pRL-TK Renilla were cotransfected into 293T
   cells using Lipofectamine 2000 (Invitrogen, 11668019). Luciferase
   activities were measured 24 hours later using Dual-Luciferase Reporter
   Assay System (Promega, E1960). The relative luciferase activities were
   normalized to Renilla luciferase control under the same treatment
   conditions.

Seahorse extracellular flux analysis

   Glycolytic capacity of T-ALL cells was measured as their glycolytic
   proton efflux rate (glycoPER) with Seahorse XFe24 Extracellular Flux
   Analyzer and a Seahorse XF Glycolytic Rate Assay Kit (Agilent,
   103344-100) according to the manufacturer’s instruction. T-ALL cells (1
   to 1.5 × 10^5) per well were plated in XFe24 cell culture microplates
   precoated with Cell-Tak (22.4 μg/ml; Corning, 354240). The cells were
   then incubated in Seahorse XF RPMI medium supplemented with 2
   mM l-glutamine, 1 mM sodium pyruvate and 10 mM d-glucose (Agilent,
   103681-100) for 45 min in CO[2]-free incubator before the assay. Data
   were analyzed using Wave software and Seahorse Glycolytic Rate Assay
   Report Generator (Agilent).

Measurement of glucose uptake and lactate secretion

   Glucose uptake or lactate secretion was measured using respective
   Colorimetric Assay Kits (ab136955, K6027-100, BioVision). In brief,
   1 × 10^6 cells were seeded at 12-well plates and culture medium were
   then collected to quantify glucose and lactate according to the
   manufacturer’s instructions. The values of consumed glucose and
   released lactate were normalized to the same total protein.

Immunoblotting

   Proteins were extracted using RIPA buffer [50 mM tris-HCl (pH 7.4), 150
   mM NaCl, 1 mM EDTA, 0.1% SDS, 1% sodium deoxycholate, and 1% NP-40],
   supplemented with a protease inhibitor cocktail (Roche, 4693132001) and
   protein concentrations were determined using BCA assay kit (ZOMANBIO,
   ZD301). Equal amounts of protein were fractionated by various
   concentrations of SDS–polyacrylamide gel electrophoresis and
   transferred to a polyvinylidene difluoride membrane (Bio-Rad).
   Membranes were blocked for 1 hour at room temperature in TBST
   supplemented with 5% nonfat dried milk and incubated overnight at 4°C
   with primary antibody diluted in the same blocking buffer. After
   washing three times in TBST, membranes were incubated for 1 hour at
   room temperature with appropriate HRP-conjugated secondary antibodies
   (Jackson ImmunoResearch JAC-111-035-003 or JAC-115-035-003). The signal
   was detected using ECL Western Blotting Substrate (Bio-Rad) on the
   Chemidoc Touch imaging system (Bio-Rad). Antibodies used in the
   experiments include FTO (Proteintech, 27226-1-AP), ELK3 (OriGene,
   TA503603; NOVUS, NBP2-01264), ALKBH5 (Abcam, ab195377), and β-ACTIN
   (Abclonal, AC026).

H&E staining and immunohistochemistry

   The tissue samples were fixed in a 4% formaldehyde solution, followed
   by processing and embedding in paraffin wax. Serial sections (3 μm in
   thickness) were obtained and subjected to H&E staining (C0105,
   Beyotime). For immunohistochemical analysis, paraffin-embedded spleen
   sections were incubated with a PCNA antibody (Santa Cruz Biotechnology,
   sc-56) at a 1:200 dilution. Staining results were visualized using the
   DAB substrate kit from Vector Laboratories, and images were captured
   using CaseViewer (3DHISTECH).

Wright-Giemsa staining

   PB obtained from T-ALL mice was diluted with an equal volume of PBS and
   then subjected to a smear. Following air drying, the slides were
   incubated in Giemsa stain for 30 s and subsequently stopped with buffer
   B following the manufacturer’s instructions (BaSO, BA4017). After a
   gentle rinse with ddH[2]O, the slides were dried for imaging using
   Aperio VERSA 8 (Leica).

Flow cytometry analysis

   Percentage of GFP^+ leukemia in the PB or spleen of T-ALL mice was
   analyzed by Accuri C6 (BD Biosciences) or CytoFLEX (Beckman) after
   excluding red blood cells using Red Blood Cell Lysis Buffer (BioLegend,
   420301). Live cells were gated on the basis of FSC-A and SSC-A
   characteristics.

Statistical analysis

   All quantitative data are presented as means ± SD. Statistical analyses
   were performed using GraphPad Prism8 and R software. Statistical
   analyses were performed using two-tailed, unpaired Student’s t test,
   Mann-Whitney test, or Welch’s t test for comparisons between two
   groups, while one-way analysis of variance (ANOVA) with Tukey’s
   multiple comparisons or two-way ANOVA with Bonferroni’s multiple
   comparisons was used for experiments involving multiple groups.
   Significance of Kaplan-Meier survival curves was estimated by log-rank
   test. Unless otherwise stated in the text, differences were considered
   significant when P < 0.05.

Acknowledgments