Reader-dependent functional duality of FTO: a context-switching node at the intersection of immune evasion and therapeutic resistance.

The convergence of mechanistic and preclinical pharmacologic evidence outlines a plausible, but as yet clinically untested, pathway toward FTO-targeted immunotherapy. We emphasize that no FTO inhibitor or FTO-degrader has, to our knowledge, completed Phase I evaluation in oncology indications, and the framework outlined below should therefore be understood as a preclinical-evidence-based design proposal rather than a near-term clinical roadmap. The translational utility of any FTO-directed strategy is unlikely to be uniform across tumors. Rather, the emerging data support a biomarker-guided model in which the clinical utility of FTO inhibition is likely to depend on molecular context, particularly the composition of m 6 ;A reader networks. In melanoma, FTO-mediated m 6 ;A erasure preserves protumorigenic and immune-resistant transcripts, including PDCD1, CXCR4, and SOX10, by counteracting reader-dependent RNA decay, thereby reducing sensitivity to anti-PD-1 blockade ( 23 ). Parallel work indicates that FTO also reinforces glycolytic reprogramming and lactate accumulation, creating a metabolic barrier to effective T-cell function that can be relieved by FTO inhibition ( 109 ). In hematologic settings, FTO further sustains stemness-associated and immune-evasive programs, including checkpoint-related pathways such as LILRB4, which are reversible with pharmacologic inhibition ( 44 , 97 ). Taken together, these findings argue less for single-agent deployment than for rational combination strategies that integrate FTO inhibition with checkpoint blockade. A plausible near-term clinical design would therefore test FTO inhibitors in combination with anti-PD-1/PD-L1 agents while prospectively incorporating biomarker stratification. The strongest conceptual candidate is m 6 ;A reader composition. Because the biological consequence of FTO inhibition depends on whether methylated transcripts are preferentially routed toward degradation or stabilization, tumors enriched for YTHDF2-like decay programs may be more likely to derive benefit than those dominated by IGF2BP-associated transcript preservation. Early-phase trials could therefore incorporate YTHDF2/IGF2BP2 balance as an exploratory stratification variable rather than an established selection criterion, allowing prospective assessment of whether reader architecture defines a therapeutic window. Pharmacogenomic markers may provide an additional, though still preliminary, layer of enrichment. Retrospective studies suggest that germline FTO polymorphisms are associated with differential ICI outcomes across solid tumors ( 112 ), while other variants appear to influence cancer susceptibility and FTO-linked expression states ( 88 ). At present, these data are insufficient for standalone clinical eligibility decisions, but they could be incorporated alongside tumor transcriptomic signatures as exploratory biomarker modules in basket-style trials. The evidentiary base is still largely preclinical or retrospective, with no prospective validation of reader-based stratification, no clinically standardized assay framework, and unresolved questions regarding dose, toxicity, and tissue-specific dependency. Even so, early-phase FTO inhibitor plus anti-PD-1/PD-L1 trials with mandatory biomarker co-enrollment represent a scientifically justified and clinically testable next step toward context-aware epitranscriptomic immunotherapy. Resistance to FTO inhibition appears to arise less from single-pathway escape than from network-level compensation intrinsic to the m 6 ;A regulatory system. A proximate mechanism is eraser redundancy: FTO and ALKBH5 co-operatively sustain m 6 ;A hypomethylation across overlapping oncogenic loci, such that selective FTO inhibition may be functionally buffered by residual ALKBH5 activity ( 39 ). This redundancy is not merely passive. Combined disruption of FTO/ALKBH5 rebalances writer expression, increasing METTL3 while reducing METTL14, thereby restructuring transcript-specific m 6 ;A deposition patterns rather than uniformly elevating the methylome and, in some contexts, preserving oncogenic signaling ( 39 ). FTO inhibition may therefore reshape the epitranscriptomic landscape rather than simply extinguish it. Transcript-level rescue further reinforces this architecture. IGF2BP2 can engage hypermethylated targets, including HK2 mRNA, in a stabilizing, m 6 ;A-dependent manner, preserving glycolytic flux and FOXO-associated proliferation despite demethylase suppression ( 39 ). In this framework, FTO inhibition does not uniformly destabilize oncogenic transcripts; instead, it may shift selective pressure toward reader-mediated rescue, converting hypermethylation into a pro-survival output through altered RNA fate determination. A parallel challenge emerges at the immune interface. FTO is embedded within a broader regulatory network that includes METTL3, YTHDF1/2, and HNRNP-family proteins, all of which are linked to PD-L1/PD-1 expression and immune infiltration patterns in gastric cancer ( 98 ). FTO inhibition alone may therefore be insufficient to fully dismantle checkpoint-supportive circuitry, as residual writers and alternative readers may sustain immune-regulatory transcripts through parallel routes. At the chromosomal level, FTO also regulates centrosome integrity and mitotic fidelity via NuMA and KIFC1 localization ( 45 ). While this creates a therapeutic vulnerability in chromosomally unstable tumors, it may also impose selective pressure favoring mitotic stress-tolerant subpopulations with greater adaptive capacity. Collectively, these observations support a model of epitranscriptomic plasticity–driven resistance, in which the m 6 ;A network compensates for single-node inhibition through eraser redundancy, reader-mediated transcript rescue, and pathway-level rewiring. FTO is therefore not an isolated demethylase target but a dynamic node within a compensatory regulatory system. Durable therapeutic responses will likely require systems-level co-targeting combining FTO inhibition with reader blockade or writer modulation to restrict the adaptive escape routes that single-agent FTO inhibition is likely to expose or select for. Among the multiple regulatory nodes that govern ferroptotic execution, FTO has been identified in several tumor contexts as one m 6 ;A-dependent contributor to the post-transcriptional stability of SLC7A11 and GPX4 mRNAs, the two principal effectors of the anti-ferroptotic axis. Through limitation of m 6 ;A deposition and consequent attenuation of YTHDF2-mediated decay, FTO can sustain SLC7A11/GPX4 transcript levels and thereby help restrain lipid peroxide accumulation and iron-dependent cell death in tumor types where this circuit is operative. In colorectal cancer, high FTO expression sustains SLC7A11 and GPX4 transcript stability by limiting their m 6 ;A methylation and thereby preventing YTHDF2-mediated decay, establishing a ferroptosis-resistant transcriptional state ( 35 ). This checkpoint is upstream-regulated by AKT signaling: AKT inhibition transcriptionally downregulates FTO, elevating m 6 ;A methylation at a defined GPX4 regulatory site and directing YTHDF2-dependent degradation a ferroptotic signal further amplified when co-occurring protective autophagy is suppressed ( 151 ). This positions FTO inhibition combined with system Xc⁻ antagonism or direct GPX4 targeting as a mechanistically coherent synthetic lethal strategy: FTO inhibition primes the anti-ferroptotic transcriptome for collapse, while ferroptosis inducers trigger irreversible lipid peroxidation execution. Critically, this model is not universal. In papillary thyroid carcinoma, FTO paradoxically functions as a tumor suppressor, downregulating SLC7A11 in an m 6 ;A-independent manner and thereby promoting ferroptotic sensitivity ( 70 ). This apparent contradiction reflects context-dependent transcript targeting, in which the directionality of FTO’s effect on ferroptosis is determined by tissue-specific m 6 ;A target repertoire, reader composition, and site-specific methylation accessibility. The IDO1–AhR–FTO axis in glioblastoma extends this regulatory architecture: tumor-derived immunosuppressive tryptophan catabolism modulates FTO expression via AhR-dependent transcriptional control, with downstream effects on SLC7A11 mRNA stability that reduce ferroptotic vulnerability in this context ( 159 ). This finding directly couples immunometabolic signaling to epitranscriptomic ferroptosis control, nominating IDO1 inhibition as a candidate upstream strategy to derepress ferroptotic sensitivity in tumors that depend on this axis. Together, these findings support a model that may be described as “epitranscriptomic modulation of ferroptotic vulnerability”: the m 6 ;A landscape, and FTO’s activity within it, can contribute to the molecular barrier to ferroptotic execution in a tumor-type-specific and signaling-context-dependent manner. Ferroptosis is not a universal consequence of FTO inhibition but a conditional synthetic lethal state gated by lineage identity, upstream kinase activity, and immune-metabolic inputs. Current evidence remains largely confined to in vitro models with limited in vivo validation and no integration with the tumor immune microenvironment. Realizing this therapeutic opportunity will require biomarker-guided patient stratification incorporating FTO expression, YTHDF2/SLC7A11 co-expression, AKT pathway activity, and IDO1-AhR status to identify tumors in which the synthetic lethal window between FTO inhibition and ferroptosis induction is maximally and durably open. A substantive evaluation of FTO biology requires explicit consideration of the methodological limitations that constrain mechanistic inference across the field. The strength of any claim about FTO function, including most of the mechanisms surveyed in this review, depends on assays whose underlying technical performance is itself a subject of ongoing debate. Four limitations are particularly consequential and merit explicit discussion. substrate identifications rely on m 6 ;A-immunoprecipitation followed by sequencing (MeRIP-seq) or its derivatives ( 168 ). Independent benchmarking studies have shown that peak overlap between technical replicates within the same laboratory is frequently below 50%, with inter-laboratory reproducibility lower still ( 169 , 170 ). Antibody-based enrichment is inherently semi-quantitative and confounded by RNA fragment length, GC content, and local secondary structure. Stoichiometric m 6 ;A mapping methods including miCLIP, GLORI, m 6 ;A-SAC-seq, and DART-seq, partially address these limitations but remain unevenly adopted, and the m 6 ;A-mark stoichiometry at most “FTO target” sites identified in prior work has not been independently validated by an orthogonal quantitative platform ( 171 , 172 ). Caution is therefore warranted before treating any single MeRIP-seq–derived FTO target list as the definitive substrate set. Commercially available anti-m 6 ;A antibodies vary substantially in epitope specificity and cross-react with chemically related modifications. The most important cross-reactivity is with m 6 ;Am at the mRNA 5′ cap (N6,2′-O-dimethyladenosine), since FTO demethylates both modifications and m 6 ;Am sits within the same sequence context that anti-m 6 ;A antibodies recognize ( 13 , 41 ). Cross-reactivity with m¹A has also been reported under specific conditions, and sequence-context binding bias has been documented across multiple antibody clones. The practical consequence is that a substantial fraction of literature signal attributed to “m 6 ;A demethylation by FTO” may, in cap-proximal contexts, reflect m 6 ;Am demethylation instead, a distinction with direct functional implications for translation initiation and mRNA stability that is not equivalent to internal m 6 ;A regulation. FTO demethylates both internal m 6 ;A and cap-proximal m 6 ;Am, but with different kinetics and substrate preferences that depend on subcellular localization: nuclear FTO acts predominantly on internal m 6 ;A, whereas cytoplasmic FTO shows substantial activity toward m 6 ;Am ( 13 ). These two activities have non-overlapping functional consequences, internal m 6 ;A typically regulates transcript stability via reader-mediated decay or stabilization, while m 6 ;Am demethylation influences translation efficiency and cap-dependent mRNA fate. The majority of cancer studies cited in this review do not formally distinguish these two activities, and conclusions framed simply as “FTO-mediated m 6 ;A demethylation” may therefore conflate mechanistically distinct events. The colorectal cancer cytoplasmic m 6 ;Am study ( 25 ) is one of the few that explicitly resolves this question; broader application of m 6 ;Am-specific mapping (e.g., m 6 ;Am-Exo-seq) across cancer contexts is an unmet need. Despite the size of the FTO–cancer literature, the criteria required to establish a transcript as a direct FTO substrate, physical interaction with FTO, m 6 ;A or m 6 ;Am demethylation at a defined site following FTO perturbation, demonstrable dependence of the functional output on that specific site, and exclusion of indirect transcriptional or trans effects, are met for only a small fraction of reported “FTO targets.” Many proposed substrates derive from a combination of (a) altered transcript abundance following FTO knockdown, (b) changes in MeRIP-seq peak intensity at the transcript, and (c) RIP-PCR for FTO–transcript interaction, none of which, individually or together, formally establishes direct catalytic regulation. The recent application of site-specific mutagenesis of m 6 ;A consensus motifs in target transcripts (e.g., GAS5 ( 38 ), FLAD1 ( 17 )) represents the current methodological gold standard but has been applied to only a subset of published targets. Readers should therefore interpret the FTO substrate landscape as a working hypothesis under active methodological refinement rather than a settled catalog. We have attempted, where possible, to flag claims whose underlying mechanistic evidence rests primarily on MeRIP-seq, on antibody-dependent assays without orthogonal validation, or on m 6 ;A/m 6 ;Am-undifferentiated approaches. We acknowledge, however, that complete propagation of these technical caveats through every cited mechanism would render the manuscript unreadable; some level of inference from imperfect data is unavoidable in a synthetic review. The technical limitations outlined above should therefore be treated as a baseline constraint on the certainty of all mechanistic claims in this field, including those discussed in the preceding sections, rather than as a critique of any individual study. The context-dependence of FTO biology — that its functional output depends on tumor lineage, transcript context, and m 6 ;A reader composition — is at this point a well-established observation in the epitranscriptomics field, and we do not claim it as a novel contribution. What this review attempts is something more modest and, we believe, more useful: an evidence-calibrated, multi-dimensional synthesis that the FTO literature has not yet received. Specifically, we (i) annotate every mechanistic claim with its underlying evidence tier and apply consistent language calibration accordingly; (ii) systematically separate tumor-intrinsic from immune-cell-intrinsic FTO functions across CD8 + , CD4 + , and NK lymphoid lineages and across the myeloid compartment; (iii) integrate the post-translational regulation of m 6 ;A reader proteins as a previously underdeveloped layer of the reader-dependent logic; (iv) explicitly catalog the technical limitations of epitranscriptomic methods (MeRIP-seq reproducibility, m 6 ;A antibody specificity, m 6 ;A versus m 6 ;Am discrimination, definition of direct FTO substrates) as constraints on mechanistic inference; and (v) translate reader-network biology into a biomarker-stratified clinical trial design logic for FTO inhibitor–ICI combinations. The conceptual building blocks are largely drawn from prior work; the contribution lies in their integration, calibration, and translational orientation. Among the most urgent priorities for advancing the field is improved spatial resolution. Bulk transcriptomics, by conflating spatially segregated compartments, inevitably generates contradictory mechanistic conclusions. Single-cell dissection further demonstrates that FTO governs cancer stem cell maintenance through the Wnt10b/β-catenin axis in ways invisible to population-level analyses ( 105 ), underscoring that coherent interpretation of FTO function requires cell-type resolution mapped within its spatial niche. Equally urgent is the recognition that FTO biology is intrinsically non-cell-autonomous. Resistance to targeted therapy is propagated intercellularly through exosomal FTO transfer: drug-resistant cells export FTO-laden vesicles that reprogram the m 6 ;A landscape of naïve recipient cells, transmitting chemoresistance via the FTO/YTHDF2/ABCC10 axis ( 148 ). Senescent neutrophils deliver piRNA-17560 that transcriptionally induces FTO in tumor cells, stabilizing ZEB1 and driving EMT and chemoresistance ( 149 ), while BM-MSC-derived exosomal FTO amplifies stemness and cytarabine resistance through LncRNA GLCC1–c-Myc activation ( 147 ). These findings suggest that FTO inhibition may need to be considered in the context of non-cell-autonomous resistance mechanisms: suppressing FTO in tumor cells while leaving donor compartments, senescent immune cells, stromal mesenchymal populations, intact may be insufficient to prevent resistance re-seeding through intercellular m 6 ;A transfer. Improved functional modeling will be needed to capture these dynamics. Organoid–immune co-cultures that recapitulate FTO-dependent macrophage polarization ( 111 ), combined with exosomal m 6 ;A profiling and spatial multi-omics, would help establish whether m 6 ;A regulation operates predominantly as a cell-intrinsic switch or as a spatially organized, intercellularly propagated property of the tumor microenvironment. Translating these models into clinically informative findings remains an open methodological challenge.

A clinically important function of FTO-mediated m 6 ;A demethylation is its capacity to reshape the tumor cell’s regulated-cell-death network, the balance of apoptotic and regulated-death pathways that determines survival under chemotherapeutic insult. Critically, FTO does not operate as a unidirectional pro-survival enzyme but as a context-sensitive epitranscriptomic switchboard whose polarity is determined by target transcript identity, cancer type, and the specific drug employed ( Table 4 ). Role of FTO in Therapy Response and Resistance Across Cancer Types: Chemotherapy, Targeted Therapy, Radiotherapy, and Intercellular Transmission. Cellular Compartment indicates the cell type in which FTO is acting: Tumor-intrinsic , FTO functions within tumor cells; non-malignant, protective role in healthy tissue (granulosa, cardiomyocyte); Stromal and Myeloid-derived, FTO originating in non-tumor cells delivered to tumor cells via exosomes. Evidence Tier (per the Evidence Calibration framework in Section 2.1): (i) multi-model in vivo with patient-sample correlation; (ii) single in vivo perturbation with mechanistic support; (iii) multi-cell-line in vitro evidence; (iv) single-cell-line or single-model evidence; (v) correlative bioinformatic evidence only. Direction & Mechanistic Category indicates whether FTO promotes therapy resistance (oncogenic in the resistance context) or opposes it (sensitizes tumor cells to therapy), followed by the principal mechanistic axis. Table 4 exposes three patterns that the original literature-summary format obscured. First, FTO’s directional role in therapy resistance is highly context-dependent: across 32 tumor-context studies (excluding the 3 non-malignant cytoprotective rows), FTO promotes resistance in approximately 24 studies (predominantly upregulated, oncogenic) and opposes resistance in 6 studies (FTO loss promotes resistance, or FTO required for drug cytotoxicity). The remaining 2 studies report drug- or context-dependent direction (e.g., neuroblastoma, bladder cancer). Second, evidence-tier distribution is uneven: Tier ii evidence (single in vivo with mechanism) supports 16 studies, while Tier iii (multi-cell-line in vitro only) accounts for 14 meaning that approximately 44% of cited mechanistic claims rest on in vitro evidence alone, a constraint that should temper inference about generalizability. Third, the three non-malignant cytoprotective findings (cardiomyocyte, granulosa cells) raise a clinically important translational concern: any systemic FTO inhibitor used as a chemosensitizer in tumors may concurrently exacerbate cardiotoxicity or gonadotoxicity in non-malignant tissues, since FTO appears to function as a protective node in those tissues, a consideration absent from current trial-design discussions of FTO-targeted therapy.↑, increased expression, activation, or upregulation; ↓, decreased expression, inhibition, downregulation, or loss; →, mechanistic or regulatory relationship between molecular events. A well-documented resistance mechanism operates through FTO-mediated suppression of apoptotic mediators. In 5-FU-resistant colorectal cancer, FTO demethylates SIVA1 mRNA to trigger YTHDF2-mediated pro-apoptotic gene degradation ( 116 ), while simultaneously stabilizing NUPR1 mRNA to establish an iron-homeostasis buffer via LCN2 and FTH1 that insulates tumor cells from cytotoxic stress ( 20 ). Pharmacological FTO inhibition using CS1 in CRC cell lines and patient-derived organoids sensitizes resistant tumors to 5-FU, reverses EMT, impairs stemness, and reduces xenograft tumor formation with clinical relevance reinforced by correlation between FTO overexpression and poor survival in CRC cohorts ( 117 ). In breast cancer, STAT3 binds the FTO promoter to drive its expression under doxorubicin pressure, while FTO reciprocally activates STAT3 signaling, creating a self-reinforcing resistance circuit ( 118 ). Separately, FTO-mediated stabilization of lncRNA LINC01559 via m 6 ;A-YTHDF2-dependent demethylation suppresses miR-1343-3p to enforce docetaxel resistance ( 119 ), establishing that a single demethylase can coordinate resistance to mechanistically distinct agents within the same cancer type. FTO’s resistance program extends into translational control, organellar adaptation, and drug efflux. In gemcitabine-resistant pancreatic cancer, FTO is stabilized through USP7-mediated deubiquitination and demethylates NEDD4 to suppress PTEN and activate PI3K/AKT ( 120 ), while independently maintaining LINC01134 stability to license WNT5A/WNT pathway-driven stemness and resistance ( 121 ). In gastric cancer, FTO demethylates CDKAL1 to induce mitochondrial fusion, conferring 5-FU resistance through organellar metabolic reprogramming ( 122 ). Conversely, omeprazole-mediated FTO suppression activates mTORC1, inhibits pro-survival autophagy, and upregulates the apoptosis-related DDIT3 in an m 6 ;A-dependent manner, chemosensitizing gastric cancer cells to multiple agents ( 123 ) a pharmacologically exploitable strategy with direct translational relevance. In cisplatin-resistant NSCLC, paradoxical FTO downregulation generates m 6 ;A hypermethylation at 5′ UTR regions, excluding eIF3A recruitment and silencing resistance-associated transcripts in an RBM5-dependent manner ( 124 ). In bladder cancer, FTO loss similarly promotes resistance to cisplatin, and pharmacological FTO inhibition using MA2 restores cytotoxicity ( 125 ). In lung cancer, FTO regulates MRP7 mRNA stability to modulate drug efflux, with FTO inhibition via entacapone restoring paclitaxel sensitivity in resistant cells ( 126 ). Equally important is evidence that FTO can function as a chemosensitizer through activation of alternative cell death programs. In cisplatin-resistant ovarian cancer, FTO overexpression restores sensitivity by activating NLRP3/caspase-1/GSDMD-dependent pyroptosis through NLRP3 demethylation ( 127 ), a pro-death function diametrically opposed to its resistance-promoting roles elsewhere. FTO also inhibits doxorubicin-induced ferroptosis in cardiomyocytes via a P53–P21/Nrf2 axis in a HuR-dependent manner ( 128 ), cautioning against indiscriminate FTO inhibition in doxorubicin-based regimens. In granulosa cell models, FTO protects against cisplatin cytotoxicity through both Hippo/YAP1 pro-survival signaling ( 129 ) and FGF2-driven autophagy activation ( 130 ). Neuroblastoma provides a particularly clear illustration of this drug-specificity: FTO sensitizes cells to paclitaxel, has no significant effect on cisplatin response, yet produces opposing effects on etoposide sensitivity depending on expression level ( 131 ) arguing against a uniform pro-resistance model. These findings carry significant limitations. Nearly all mechanistic studies rely on cell lines or xenograft models, without capturing clonal evolution under drug pressure or clinical pharmacokinetic complexity. The temporal dynamics of m 6 ;A reprogramming during acquired resistance remain uncharacterized, and attributing resistance to individual downstream targets risks overinterpretation within a multi-pathway system. Collectively, FTO behaves as a context-dependent epitranscriptomic regulator whose net chemoresistance or chemosensitizing effect is dictated by the functional identity of its m 6 ;A-regulated targets in each cancer–drug context. Blanket FTO inhibition will not uniformly sensitize tumors to chemotherapy and may paradoxically promote resistance or organ toxicity where FTO’s pro-death functions are dominant. Systems-level, cancer-type-resolved mapping of the FTO-dependent m 6 ;A landscape under specific chemotherapeutic pressures is an essential prerequisite for deploying FTO-targeting strategies with the mechanistic precision required for clinical translation. The capacity of tumors to escape kinase inhibition represents one of the defining challenges of precision oncology, and FTO-mediated epitranscriptomic reprogramming has emerged as a recurrent though mechanistically heterogeneous contributor to this resistance landscape. Across EGFR-targeted therapies and beyond, FTO functions less as a static pro-resistance factor than as a dynamic regulator of signaling plasticity that enables tumor cells to rewire their transcriptional and phenotypic state under therapeutic pressure ( Table 4 ). The EGFR-TKI resistance context provides the most mechanistically dissected illustration. In gefitinib-resistant NSCLC, elevated FTO stabilizes PELI3 mRNA in an m 6 ;A-dependent manner, which cooperates with autophagy activation to sustain tumor cell survival against EGFR inhibition; genetic or pharmacological disruption of either PELI3 or autophagy is sufficient to restore gefitinib sensitivity in vitro and in vivo ( 132 ). A parallel but mechanistically distinct resistance axis involves FTO-dependent stabilization of c-Myc, which transcriptionally amplifies the drug efflux transporters BCRP and MRP7, reducing intracellular gefitinib accumulation and maintaining downstream AKT and MAPK signaling ( 133 ). Critically, meclofenamic acid a pharmacological FTO inhibitor restores gefitinib sensitivity by dismantling the FTO/c-Myc/efflux transporter axis, demonstrating that epitranscriptomic and pharmacokinetic resistance mechanisms are functionally coupled through a single m 6 ;A-regulatory node ( 133 ). Together, these recent NSCLC studies ( 132 , 133 ) reveal that FTO-driven TKI resistance is not mechanistically monolithic: autophagy-mediated survival and transporter-mediated drug exclusion represent parallel, independently sufficient resistance programs that share FTO as their upstream coordinator. In contrast, loss-of-function FTO contexts generate distinct resistance vulnerabilities that can be therapeutically exploited. FTO downregulation across epithelial cancers promotes EMT through increased m 6 ;A modification and altered 3′-end processing of Wnt pathway transcripts, enhancing invasion and metastasis while simultaneously sensitizing tumors to Wnt inhibition ( 134 ) a synthetic vulnerability arising directly from the FTO-low, EMT-high phenotypic state. In HIF2α-low clear cell renal cell carcinoma, aberrant FTO activation drives BRD9 mRNA stabilization, enabling SOX17-recruited BRD9 to establish de novo super-enhancers that amplify oncogenic programs including CCND1, VEGFR2, and MAPK6; consequently, BRD9 inhibition selectively suppresses FTO-high tumors with greater efficacy than sunitinib in patient-derived xenografts ( 135 ). These findings collectively establish that FTO’s resistance function is directional and lineage-specific: its overexpression creates kinase inhibitor resistance in lung cancer while simultaneously generating BRD9 dependency in renal cancer, and its loss creates Wnt inhibitor sensitivity in epithelial cancers. Critical limitations pervade this body of evidence. The mechanistic studies are overwhelmingly cell-line-based, without prospective clinical validation or pharmacodynamic assessment in patients receiving TKIs. The temporal dynamics of FTO-mediated resistance evolution whether FTO reprogramming precedes, accompanies, or follows kinase inhibitor exposure remain uncharacterized, and clonal heterogeneity within resistant tumors is entirely unaddressed. Furthermore, the field carries a marked EGFR-centric bias, and whether the FTO–signaling plasticity axis generalizes to ALK, RET, or KRAS-targeted therapies is unknown. Nonetheless, the emerging conceptual model is compelling: FTO functions as an epitranscriptomic regulator of signaling plasticity, enabling tumors to dynamically rewire their dependence on survival pathways under therapeutic pressure. Co-targeting of FTO alongside kinase inhibitors, or exploitation of FTO-status-dependent synthetic lethalities such as BRD9 or Wnt-pathway dependencies, has been proposed as a mechanistically grounded combination strategy. We classify this as a preclinically rationalized hypothesis (readiness level R2): supportive evidence derives from in vitro and single-model in vivo studies, and the context-specificity of FTO’s resistance polarity has not yet been prospectively characterized in human tumors. Clinical evaluation should therefore be biomarker-stratified from the outset and preceded by independent preclinical reproduction across additional disease-relevant models. Resistance to radiotherapy and temozolomide (TMZ) in CNS tumors is not a monolithic phenomenon but a multi-layered adaptive program in which FTO-mediated epitranscriptomic reprogramming plays a convergent coordinating role coupling DNA repair efficiency with stemness maintenance and transcriptional circuit stabilization under therapeutic stress. Integrating evidence across glioblastoma, diffuse midline glioma, and non-CNS tumor models treated with radiation, a unified mechanistic framework emerges in which FTO functions as a master regulator of the resistance niche rather than a modifier of any single pathway ( Table 4 ). At the DNA damage interface, FTO sustains homologous recombination (HR) capacity by maintaining RAD51 focus formation following radiation-induced double-strand breaks. In both HNSCC and glioblastoma stem cells (GSCs), FTO inhibition increases γH2AX persistence and reduces RAD51 availability, impairing HR efficiency and sensitizing cells to ionizing radiation ( 136 , 137 ). This mechanistic convergence across CNS and non-CNS lineages establishes FTO-dependent RAD51 regulation as a general radioresistance program, though it raises the unresolved question of whether FTO acts directly on HR machinery transcripts or indirectly through broader m 6 ;A landscape remodeling. The stemness dimension is uniquely amplified in the CNS context. GSCs are the principal reservoir of radioresistance in GBM, and FTO inhibition with FB23–2 suppresses GSC self-renewal, reduces tumor sphere formation, and extends survival in intracranial mouse models ( 137 ) indicating that FTO supports the stem-like compartment implicated in recurrence. Part of this stemness maintenance operates through FTO-mediated stabilization of VEGFA mRNA; its demethylation sustains VEGFA expression, supporting both vascular programs and autocrine pro-survival signaling within the GSC niche ( 137 ). The FTO/m 6 ;A/SOX9 axis further reinforces stemness and lineage plasticity in TMZ-resistant GBM: SOX9 is markedly upregulated in resistant cells and is sustained by FTO-mediated demethylation, while pharmacological FTO inhibition via the rhein derivative SYSUP007 downregulates SOX9 and restores TMZ sensitivity ( 138 ). The transcriptional circuit layer introduces additional complexity. In glioma, FTO stabilizes MYC mRNA, enabling MYC to suppress its own antagonist MXI1 through miR-155 and the miR-23a cluster in a self-reinforcing feedback loop; FTO inhibition with MA2 dismantles this circuit and synergizes with TMZ to suppress glioma proliferation ( 139 ). Independently, FTO negatively regulates phosducin (PDC), a suppressor of glioma proliferation and TMZ resistance; FTO-mediated PDC repression thus removes a pro-sensitivity factor, and its restoration phenocopies FTO inhibition in enhancing TMZ efficacy ( 140 ). In cervical cancer, a non-CNS parallel is provided by FTO-mediated β-catenin mRNA demethylation, which elevates ERCC1 activity and confers chemo-radiotherapy resistance through Wnt-driven DNA repair augmentation ( 141 ) a mechanistic principle that may operate in CNS tumors where Wnt signaling is equally active. In diffuse midline glioma, FTO inhibition induces S-phase arrest and upregulates CDKN1A and GADD45B, implicating FTO in cell cycle checkpoint suppression through m 6 ;A-dependent regulation of chromosome segregation machinery ( 142 ). Critical limitations constrain interpretation. All studies rely on cell lines or xenograft models without longitudinal patient-derived data capturing resistance evolution under clinical treatment. The spatial architecture of resistant niches and whether FTO activity is enriched within specific GSC subpopulations remains unresolved at single-cell resolution. Crucially, whether FTO is a primary driver establishing resistance programs or a modifier of pre-existing stemness states is mechanistically undetermined across all models, and the integration of radiotherapy and TMZ resistance into a single biological framework rather than parallel independent mechanisms has not been directly demonstrated. Collectively, FTO emerges as an epitranscriptomic integrator that couples DNA repair capacity with stemness maintenance and oncogenic circuit stability under CNS therapeutic stress. Combining FTO inhibitors with radiotherapy or TMZ prioritizing CNS-penetrant agents such as MA2 or SYSUP007 provides a mechanistically grounded rationale for simultaneously targeting the DNA-repair advantage, the stem-like resistant compartment, and the pro-survival transcriptional programs implicated in GBM recurrence, a hypothesis that requires direct preclinical and clinical testing. In hematologic malignancies, FTO has been linked to therapy resistance through two distinct mechanisms: cell-autonomous epitranscriptomic reprogramming within tumor cells, and intercellular transmission of resistance via exosomal cargo. Both axes have been implicated as potential therapeutic vulnerabilities ( Table 4 ). Within tumor cells, FTO drives resistance through convergent stabilization of MYC-centered transcriptional networks across lymphoid and myeloid malignancies. In DLBCL, FTO directly demethylates MYC mRNA to prevent YTHDF2-mediated decay, sustaining MYC protein expression and conferring ibrutinib resistance in a MYC-dependent manner ( 143 ). A parallel but mechanistically distinct axis involves FTO-mediated stabilization of FLOT2 mRNA, which activates PI3K/AKT/mTOR signaling and has been associated with DLBCL aggressiveness ( 144 ) indicating that FTO can engage multiple oncogenic networks in this disease context. In AML, the resistance program is coupled to a differentiation block: FTO overexpression in relapsed samples hypomethylates FOXO3 mRNA, accelerating its degradation and impairing myeloid differentiation capacity, thereby entrenching the undifferentiated, chemoresistant state ( 145 ). Extending beyond hematologic contexts, the FTO/LINK-A/MCM3/HIF-1α axis in esophageal cancer illustrates a mechanistic principle with broad relevance: FTO-stabilized lncRNA LINK-A coordinates cell-cycle progression through MCM3 phosphorylation while simultaneously relieving HIF-1α repression to activate glycolysis and chemoresistance ( 146 ). This raises the possibility, not yet established across multiple lineages, that lncRNA-mediated circuit stabilization is a recurring motif of FTO-dependent resistance. Intercellular (horizontal) transfer introduces an additional regulatory layer of potential therapeutic relevance. Bone marrow mesenchymal stem cell-derived exosomes deliver FTO protein directly into AML cells, where it demethylates lncRNA GLCC1 to stabilize it via HuR binding; GLCC1 then scaffolds the IGF2BP1–c-Myc complex, activating c-Myc-associated survival programs and Ara-C resistance in recipient cells ( 147 ). This finding establishes FTO not merely as a tumor cell-intrinsic factor but as a bone marrow niche-derived signal that reprograms AML epitranscriptomes from outside the tumor cell. In NSCLC, gefitinib-resistant cell-derived exosomes deliver FTO to sensitive recipient cells, reducing global m 6 ;A levels and upregulating ABCC10 drug efflux transporter via the FTO/YTHDF2/ABCC10 axis directly transmitting acquired resistance to previously sensitive cells both in vitro and in vivo ( 148 ). Senescent neutrophil-derived exosomes complete this intercellular resistance network by delivering piRNA-17560 to breast cancer cells, where it upregulates FTO to destabilize ZEB1 m 6 ;A marks, promoting EMT and chemoresistance in a YTHDF2-dependent manner ( 149 ). The reported correlation between exosomal piR-17560 levels and poorer chemotherapy response in patients provides preliminary clinical support for this immune-cell-to-tumor resistance axis, although the evidence remains correlative. Critical limitations require explicit acknowledgment. All horizontal transfer mechanisms are demonstrated primarily in vitro or in xenograft systems, without lineage-tracing or quantitative assessment of the contribution of exosomal FTO transfer relative to cell-intrinsic resistance in patient tumors. The temporal dynamics of resistance propagation whether exosomal transfer precedes or follows intrinsic resistance establishment remain uncharacterized, and the spatial distribution of FTO-transferring niche cells within bone marrow or tumor microenvironments is entirely unmapped. The quantitative sufficiency of exosome-delivered FTO protein to meaningfully reprogram recipient cell m 6 ;A landscapes has not been rigorously established. Collectively, these findings outline a working model in which FTO contributes to a distributed resistance network coupling intrinsic epitranscriptomic reprogramming with intercellular resistance transmission. Whether simultaneous disruption of both dimensions, combining intracellular FTO inhibition with strategies targeting exosomal FTO loading or uptake, would yield additive or synergistic clinical benefit remains a hypothesis-generating proposal rather than an experimentally validated strategy. The proposition is conceptually coherent but has not been formally tested in any in vivo system, and substantial mechanistic groundwork (defining exosomal FTO functional sufficiency, identifying druggable exosome-loading machinery) is required before clinical evaluation would be appropriate.

The reader-dependent organizing logic for FTO function has emerged incrementally from a series of studies over the past several years and is now broadly recognized within the m 6 ;A field; it does not, in itself, constitute a novel proposal of this review. What this section attempts is somewhat different: a systematic integration of accumulated evidence that articulates the reader-network logic across cancer contexts and identifies the operational variables that determine FTO output in any given setting. The premise is straightforward: while FTO demethylates m 6 ;A marks, the biological consequence of that demethylation is not encoded in FTO itself but is shaped, in part, by which m 6 ;A reader proteins engage the newly demethylated transcript in a given cellular context. Under this framing, FTO is better understood not as a driver of cancer per se but as a context-integrating directional switch whose output reflects the reader network, and increasingly, the post-translational reader-state, within which it operates. We treat this framing as a useful synthetic interpretation that organizes a fragmented literature, not as a definitively established mechanism, and we have attempted to enrich it by integrating the PTM-dependent reader regulation, transcript-level reader competition, and substrate-specificity considerations developed in the preceding sections. The mechanistic logic of this network is often articulated through the YTHDF2–IGF2BP axis, which provides a useful conceptual heuristic rather than a fixed binary rule. YTHDF2 has been most extensively characterized as recruiting the CCR4–NOT deadenylase complex to drive m 6 ;A-dependent mRNA decay, while IGF2BP1–3 is best known for shielding m 6 ;A-modified transcripts from degradation and stabilizing them. Within this framework, FTO-mediated demethylation removes the shared substrate of both readers, and the net effect on transcript fate depends on which reader was previously dominant. When YTHDF2-mediated decay is the prevailing fate, as with PKM2 in oral cancer ( 150 ), FLAD1 in HCC ( 17 ), and GPX4 in colorectal cancer ( 151 ). FTO demethylation stabilizes these transcripts and produces oncogenic metabolic or immune-evasive outputs. When IGF2BP2-mediated stabilization is the operative reader context, as with HK2 in colorectal cancer ( 39 ) and GAS5 in breast cancer ( 38 ). FTO loss paradoxically increases m 6 ;A methylation, enabling IGF2BP2 to stabilize glycolytic or tumor-suppressive circuits. The GAS5 example is conceptually pivotal: FTO-mediated demethylation of this lncRNA reduces its stability in an IGF2BP2-dependent manner, attenuating a GAS5–IGF2BP2–QKI tumor-suppressive scaffold ( 38 ), demonstrating that FTO can suppress anti-tumor non-coding RNA programs through the same enzymatic mechanism that stabilizes oncogenes elsewhere. However, the YTHDF2–IGF2BP dichotomy should be viewed as a useful heuristic rather than a fixed binary model. Although YTHDF2 is classically linked to CCR4–NOT-dependent decay of m 6 ;A-marked transcripts, it can also promote translation in specific tumor contexts; for example, in ovarian cancer, YTHDF2 cooperates with eIF3F and DDX1 to enhance translation of targets such as CKAP2, contributing to paclitaxel resistance ( 152 , 153 ). Conversely, IGF2BP proteins are best known for stabilizing m 6 ;A-modified RNAs and enhancing translation, but they also participate in RNA localization, storage, and ribonucleoprotein organization ( 154 , 155 ). Reader function is further shaped by post-translational regulation, including YTHDF2 SUMOylation at K571, EGFR/SRC/ERK-dependent stabilization of YTHDF2, and phosphorylation-dependent remodeling of IGF2BP1-containing RNP condensates ( 156 – 158 ). Thus, FTO output cannot be predicted from reader abundance alone, but also depends on transcript identity, competing reader occupancy, cellular signaling state, and reader post-translational status. The colorectal cancer finding that FTO and ALKBH5 co-operatively restrain HK2 expression via IGF2BP2 ( 39 ) while FTO separately stabilizes KCTD15 to activate p53 through HDAC1 suppression ( 34 ) illustrates that within a single cancer type, FTO can simultaneously promote tumor-suppressive and oncogenic transcript fates depending on the specific reader–transcript combination. This is not a contradiction but rather the expected outcome of a regulatory node whose output depends on the specific transcript-reader combination engaged. Critical limitations persist. Quantitative understanding of reader competition, how relative YTHDF2 versus IGF2BP abundance, together with their post-translational activation states, determines transcript fate at individual m 6 ;A sites, is largely absent from the current literature. Studies overwhelmingly characterize single FTO–reader–target axes in isolation, without simultaneous profiling of competing reader activities or multi-transcript network dynamics. The temporal evolution of reader composition under oncogenic stress or therapeutic pressure remains unresolved. These limitations define the conceptual frontier: advancing from FTO-centric, single-target mechanistic models toward reader-network-centric frameworks that treat m 6 ;A regulation as an emergent property of competing reader activities across the transcriptome. Systems-level modeling of reader competition, integrated with context-specific transcriptomic and proteomic profiling, will be essential to predict the functional output of FTO activity in any given tumor and to deploy epitranscriptomic therapeutic strategies with the precision that the biology demands. A useful integrative observation emerges from juxtaposing the metabolic and checkpoint sections above. The pan-tumor glycolytic-immune-evasion axis ( 93 ), the FLAD1–PD-L1 axis in HCC ( 17 ), and the hypoxia–HIF-1α–FTO–PDK1–PD-L1 cascade in breast cancer ( 107 ) together indicate that glycolytic reprogramming and PD-L1 upregulation can be coupled outputs of overlapping FTO-regulated transcript networks, with PDK1 in particular providing a single transcript whose stabilization simultaneously amplifies glycolytic flux and licenses AKT/STAT3-driven PD-L1 transcription ( 107 ). This coupling does not appear to be obligatory: the Liu et al. pan-tumor study indicates that glycolytic competition can be functionally sufficient for CD8 + T-cell suppression independent of direct PD-L1 regulation ( 93 ), suggesting that the two outputs are mechanistically linked but separately tunable across tumor contexts. No published study has, to our knowledge, simultaneously measured glycolytic flux and PD-L1 expression under FTO perturbation in a single experimental system, leaving the relative causal hierarchy unresolved. The translational implication is operational rather than novel: combination strategies pairing FTO inhibition with checkpoint blockade and glycolytic inhibitors warrant evaluation in tumors where both outputs are operative, with reader-network and metabolic profiling potentially identifying patients most likely to benefit. A notable feature of FTO biology is that its functional consequences extend beyond canonical survival signaling into two cellular programs that have been examined largely independently in the literature: regulated ferroptotic cell death and cancer stem cell (CSC) maintenance. While these programs are mechanistically distinct, they exhibit a conceptual parallel that may warrant integrated investigation: in each, FTO has been reported to attenuate fate-committing signals — pro-ferroptotic transcripts in one case, differentiation-promoting transcripts in the other. Whether this parallel reflects a coordinated, shared regulatory logic operative within the same tumor cell, or rather two independent, lineage-specific FTO functions that happen to share a directional signature, remains an open question that current experimental designs cannot resolve. The ferroptosis regulatory axis centers on SLC7A11 and GPX4, the two principal effectors of ferroptotic resistance. In colorectal cancer, FTO stabilizes both transcripts through YTHDF2-dependent m 6 ;A demethylation, and its suppression whether by genetic knockdown or the novel inhibitor Mupirocin, induces ferroptotic cell death and sensitizes tumors to Erastin and RSL3 ( 35 ). In thyroid cancer, the same SLC7A11 axis is operative but directionally inverted: FTO is downregulated and functions as a tumor suppressor by destabilizing SLC7A11 in an m 6 ;A-independent manner, triggering ferroptosis to restrain tumor progression ( 70 ). This lineage-specific polarity FTO promoting ferroptotic resistance in CRC yet enabling ferroptosis in thyroid cancer recapitulates the reader-dependent master model: the biological outcome is determined by transcript-specific m 6 ;A dynamics rather than demethylase activity per se. In glioblastoma, the IDO1–AhR axis transcriptionally represses FTO at the promoter level, reducing m 6 ;A methylation on SLC7A11 mRNA and stabilizing it to suppress ferroptosis coupling an immunosuppressive tryptophan catabolism program directly to ferroptotic resistance through FTO as the epitranscriptomic intermediary ( 159 ). This finding is conceptually significant: it positions FTO not as an autonomous regulator but as a node through which upstream immunometabolic signals are transduced into ferroptotic fate decisions. The stemness dimension reveals a parallel architecture. FTO inhibition attenuates leukemia stem cell self-renewal and suppresses LILRB4-mediated immune checkpoint expression, simultaneously collapsing stem-like and immune-evasive programs ( 97 ). In pancreatic cancer, FTO overexpression sustains CSC markers and EMT, while its depletion arrests cells in G1, induces MET, and impairs tumor formation establishing FTO as a regulator of the epithelial plasticity continuum ( 63 ). In colorectal cancer, cytoplasmic FTO operates through m 6 ;Am demethylation rather than canonical m 6 ;A, suppressing CSC properties; its loss elevates m 6 ;Am levels and enhances in vivo tumorigenicity and chemoresistance, while PCIF1 inhibition fully reverses this phenotype ( 25 ) expanding FTO’s regulatory scope beyond m 6 ;A to a broader cap-proximal epitranscriptomic layer. Critical limitations persist: ferroptosis and stemness studies are conducted in entirely separate experimental systems, and whether FTO coordinates both programs within the same tumor cell population remains untested. The context-specificity of FTO’s ferroptosis polarity across lineages demands systematic transcriptome-wide analysis before any therapeutic generalization. Targeting FTO to induce ferroptosis while simultaneously depleting stem-like populations represents a mechanistically compelling but experimentally unvalidated combination strategy requiring integrated, lineage-aware preclinical modeling.

Cancer remains a leading cause of mortality worldwide, and despite major advances in targeted therapies and immunotherapy, durable clinical responses are still limited by tumor heterogeneity, adaptive plasticity, and therapeutic resistance ( 1 – 4 ). Accordingly, the discovery of new therapeutic vulnerabilities and a deeper mechanistic understanding of tumor initiation and progression remain essential for advancing both early detection and effective treatment. In this context, RNA modifications have recently emerged as a major area of interest in cancer biology ( 5 – 7 ). RNA modifications comprise a diverse set of chemical alterations affecting RNA nucleotides, with broad consequences for RNA conformation and function. To date, more than 170 distinct RNA modifications have been described across virtually all classes of RNA molecules ( 8 ). These epitranscriptomic marks regulate multiple aspects of RNA biology, including splicing, stability, subcellular localization, translation, and RNA–RNA as well as RNA–protein interactions, thereby shaping a wide range of cellular processes ( 9 – 11 ). Among the enzymes involved in this regulatory landscape, FTO was the first m6A demethylase to be identified and is now recognized as a key modulator of N6-methyladenosine (m6A) dynamics in mRNA ( 12 ). Beyond its function in m6A demethylation, FTO has also been implicated in regulating m6A and m6Am modifications in snRNA, as well as m1A demethylation in tRNA ( 13 ). Available evidence suggests that FTO preferentially acts on m6A sites within nuclear polyadenylated RNA, whereas in the cytoplasmic compartment it can also remove methyl groups from mRNA, with a notable preference for the cap-adjacent m6Am modification ( 13 ). As a reversible and highly dynamic epitranscriptomic mark, m6A has been implicated in fundamental biological processes, including the control of gene expression, RNA turnover, and cell fate decisions ( 14 ). In the context of malignancy, FTO reshapes m6A patterns on selected transcripts, thereby contributing to tumor progression through the regulation of cellular aggressiveness, metastatic capacity, therapeutic resistance, and immune escape ( 15 – 21 ). FTO has emerged as a context-dependent epitranscriptomic regulator in cancer, influencing tumor progression, immune evasion, and therapeutic resistance in a manner that cannot be reduced to a simple oncogene–tumor suppressor dichotomy ( 22 – 25 ). Of particular relevance, its function is context-dependent: in hepatocellular carcinoma and bladder cancer, both oncogenic and tumor-suppressive roles have been reported ( 26 – 32 ), while in prostate and colorectal cancer FTO can either restrain or promote malignancy depending on the transcript context and downstream signaling output ( 33 – 37 ). These observations suggest that FTO functions as a context-switching regulatory node rather than a linear cancer driver ( 22 – 25 ). Mechanistically, the net biological consequence of FTO-mediated demethylation appears to depend on tumor lineage, target-transcript selection, and the balance of competing m6A readers, particularly YTHDF2- and IGF2BP2-dependent pathways ( 17 , 38 , 39 ). However, the field still lacks an integrated framework that explains and predicts why the same demethylase produces opposite biological outcomes across cancers. Moreover, most available evidence remains largely preclinical, underscoring the need for reader-network-based and lineage-aware models to guide translational targeting of FTO ( 17 , 34 , 35 , 38 , 39 ). This review begins with a concise synthesis of FTO’s biological functions and its pivotal role in RNA epitranscriptomic regulation, before delineating the tumor-type-specific landscape of its activity and dissecting how it reshapes the tumor immune contexture through metabolic competition, immune checkpoint modulation, and remodeling of lymphoid and myeloid compartments. We subsequently integrate current evidence implicating FTO in resistance to chemotherapy, targeted therapy, radiotherapy, and immunotherapy, with a particular focus on the shared molecular nodes that mechanistically couple immune evasion to treatment failure. We then survey emerging pharmacological strategies against FTO and revisit the reader-dependent organizing logic that has emerged from recent literature, with particular attention to dimensions that remain underdeveloped in current syntheses: explicit evidence-tier calibration, the post-translational regulation of m 6 ;A reader function itself, and the translation of reader-network biology into biomarker-stratified clinical trial design. By integrating these dimensions, this review aims to address an integration gap in the literature and to consider how context-aware, biomarker-guided targeting of FTO may contribute to more selective therapeutic approaches.

A recurring and conceptually significant observation in the FTO literature is that, within the same cancer type, FTO has been reported to exert diametrically opposite biological effects. In hepatocellular carcinoma, bladder cancer, prostate cancer, and colorectal cancer, peer-reviewed studies employing comparable methodologies arrive at contradictory conclusions about whether FTO promotes or suppresses tumor progression. Dismissing one set of findings as artifactual is unlikely to constitute an adequate explanation. These contradictions appear robust within their respective experimental contexts, and resolving them represents a substantial conceptual challenge for epitranscriptomic oncology. In HCC, the contradiction is particularly sharp. Several studies associate elevated FTO with poor prognosis and converge on a pro-tumorigenic model in which FTO stabilizes oncogenic programs linked to TGF-β signaling, immune evasion, exosomal suppression of CD8 + T cells, and glycolytic rewiring ( 17 , 26 , 53 , 94 ). Rather than any single target, the recurring principle is that FTO preserves transcripts that reinforce aggressive, immune-resistant tumor states. Yet an equally rigorous body of HCC evidence positions FTO as a tumor suppressor: its downregulation correlates with metastatic HCC and poor prognosis ( 27 , 28 ), its loss promotes invasion through dysregulated VEGFA ( 27 ), and a circGPR137B/miR-4739/FTO feedback loop establishes FTO as a suppressor of HCC tumorigenesis and metastasis ( 29 ). Both bodies of evidence are clinically grounded. The resolution lies less in methodological discrepancy than in the identity of the m 6 ;A-regulated transcript network operative in each cellular context: when FTO protects oncogenic RNAs from reader-mediated decay, it promotes progression; when it preserves tumor-suppressive transcripts, the same enzymatic activity becomes growth-restraining. The net biological output depends entirely on which transcripts are most m 6 ;A-methylated and reader-accessible in a given HCC subtype ( Figure 5 ). Opposing reported roles of FTO in hepatocellular carcinoma. In oncogenic contexts (left), FTO stabilizes transcripts contributing to TGF-β signaling, metabolic reprogramming, and immune evasion; in tumor-restraining contexts (right), FTO destabilizes pro-invasive transcripts. The directional output appears dictated by the m 6 ;A target landscape of distinct HCC subtypes rather than by FTO activity per se. A similar interpretive logic applies in bladder cancer. FTO stabilizes STAT3 mRNA to promote proliferation and migration in one context ( 30 ), while in other bladder cancer models FTO is downregulated, its loss promotes proliferation and cisplatin resistance, and its overexpression suppresses invasion by demethylating MALAT1 and NOTCH1 transcripts ( 31 , 32 , 125 ). In prostate cancer, the contradiction is equally stark and better mechanistically resolved: FTO deletion enhances PC-3 cell motility and EMT through IGF2BP2/3-mediated stabilization of DDIT4 ( 33 ), while FTO overexpression consistently suppresses PCa proliferation, invasion, and metastasis across multiple studies by stabilizing tumor-suppressive targets including FOXO3a ( 68 ), miR-139-5p ( 67 ), CLIC4 ( 69 ), and EGR2 ( 36 ) and FTO is transcriptionally upregulated by ZFHX3 loss to drive tumor-promoting m 6 ;A reprogramming ( 37 ). Collectively, the prostate cancer data suggest that FTO is predominantly tumor-suppressive in this lineage under baseline conditions, yet can become pro-tumorigenic when upstream alterations, such as ZFHX3 loss, redirect its activity toward a different transcript network. In colorectal cancer, FTO simultaneously stabilizes both oncogenic mediators NUPR1 ( 20 ), ZNF687 ( 77 ), SLC7A11/GPX4 ( 35 ) and tumor-suppressive ones KCTD15 to activate p53 ( 34 ) with net oncogenic dominance in most experimental systems ( 117 ). The same YTHDF2-dependent mechanism that produces ferroptotic resistance via SLC7A11 stabilization also produces tumor suppression via KCTD15 stabilization; FTO’s oncogenic or tumor-suppressive identity in CRC is determined by the relative m 6 ;A stoichiometry and reader availability across these competing target transcripts. These contradictions expose a fundamental limitation in the field’s analytical framework: the single-target mechanistic study, however rigorous, captures only a slice of FTO’s regulatory output and cannot predict net biological direction. The literature’s frequent characterization of FTO as uniformly oncogenic may partly reflect publication bias toward high-FTO, aggressively growing tumor contexts rather than the full biological range of FTO function. Quantitative assessment of competing reader abundance—particularly YTHDF2 versus IGF2BP family members—together with transcriptome-wide m 6 ;A mapping would help define which transcript networks are functionally dominant in each subtype. The conceptual advance demanded by this contradiction catalogue is a transition from gene-centric to network-centric models of epitranscriptomic regulation: FTO is neither oncogene nor tumor suppressor but a context-switching regulatory node whose biological polarity is an emergent property of the reader–transcript network it operates within. Therapeutic targeting of FTO should therefore be preceded by systematic characterization of the dominant m 6 ;A target landscape in each tumor subtype; otherwise, FTO inhibition may inadvertently disrupt tumor-suppressive demethylation programs in contexts where its net function is restraining, rather than driving, malignant progression. A candid appraisal of the FTO literature reveals that many published claims, while mechanistically plausible, remain provisional rather than definitive not primarily because the experiments are poorly designed, but because orthogonal validation strategies have been applied inconsistently across the field, limiting the advance from mechanistic association to robust causal inference. Three interrelated methodological limitations combine to produce a body of evidence that is suggestive at the individual study level yet structurally limited when evaluated in aggregate. The most pervasive weakness is causal insufficiency rooted in incomplete rescue architecture. Across studies spanning cervical cancer, ovarian cancer, and glioma, the dominant experimental design pairs FTO overexpression or knockdown with phenotypic readouts proliferation, apoptosis, invasion and identifies a correlated downstream target. This establishes that FTO perturbation and target expression change together, but does not demonstrate that the identified target is necessary or sufficient for the observed phenotype. In glioma, PDC is proposed as a functional FTO target because PDC overexpression attenuates FTO-driven proliferation, yet whether PDC restoration fully rescues FTO-knockdown phenotypes the bidirectional rescue standard required for pathway necessity is incompletely established ( 140 ). Similarly, FGF2 is nominated as the mechanistic effector of FTO in cervical cancer based on co-expression and knockdown correlations without full rescue validation ( 51 ). Without symmetric rescue experiments in both gain- and loss-of-function directions, the distinction between a pathway mediator and a co-regulated bystander cannot be established, and the risk of identifying incidental rather than causal targets is substantial. The second limitation concerns model-system reductionism. The overwhelming majority of FTO studies employ two-dimensional cell line models or immunodeficient xenografts, systems that lack functional immune components, stromal architecture, and intratumoral heterogeneity. Studies reporting FTO effects on autophagy and AKT signaling in ovarian cancer ( 59 ) or drug sensitivity in neuroblastoma ( 131 ) are conducted in systems that cannot capture the conditional logic reader competition, metabolic state, immune context that the contradiction data across HCC, bladder, and prostate cancer demonstrate to be functionally decisive. When in vivo validation is included, it typically confirms cell-line findings rather than interrogating the tumor microenvironmental determinants of FTO function. The third and most underappreciated limitation involves pharmacologic ambiguity. Meclofenamic acid is deployed as an FTO inhibitor across multiple studies ( 133 ), yet its primary pharmacological identity is that of a cyclooxygenase-inhibiting NSAID with well-characterized anti-inflammatory and prostaglandin-suppressing activities that are entirely independent of m 6 ;A demethylation. Attributing the sensitization effects of meclofenamic acid to FTO inhibition without demonstrating that an FTO-null background occludes the drug’s activity or that catalytically inactive FTO mutants phenocopy it inflates confidence in FTO-specific interpretations. The same concern applies to omeprazole, whose m 6 ;A-elevating effects in gastric cancer ( 123 ) are presented as FTO-mediated without excluding proton pump-dependent or FTO-independent transcriptional mechanisms. By contrast, structure-validated inhibitors such as FB23-2, developed through direct FTO binding confirmation and selectivity profiling ( 44 ), and entacapone, validated biochemically against purified FTO with substrate-level evidence ( 162 ), represent the methodological standard that repurposed agents rarely meet. The key methodological gap is therefore not a shortage of data but the absence of a validation triad genetic perturbation, biochemical target engagement, and rescue design applied consistently within the same experimental system. Until this triad becomes the field’s methodological baseline, FTO’s causal contributions to specific oncogenic phenotypes will remain mechanistically compelling but epistemically provisional. The translation of preclinical FTO-targeting strategies into clinical interventions confronts a fundamental biological problem that the field has not yet adequately addressed: FTO is not a cancer-specific vulnerability but a broadly expressed epitranscriptomic regulator whose activity is required for normal tissue homeostasis across multiple organ systems. Framing FTO inhibition as a straightforward anti-oncogenic strategy may not adequately reflect the complexity of organism-level FTO biology and may convey a degree of therapeutic confidence that current evidence does not support. The most direct challenge to systemic FTO inhibition is its cardioprotective function. FTO expression in cardiomyocytes is downregulated by doxorubicin, and its restoration activates a P53–P21/Nrf2 ferroptosis-suppressive circuit that preserves cardiac function under chemotherapy-induced oxidative stress ( 128 ). This finding reveals that FTO inhibition in a patient receiving doxorubicin-based chemotherapy would simultaneously suppress tumor epitranscriptomic programs and dismantle a cardioprotective mechanism an on-target toxicity that no current FTO inhibitor program has systematically evaluated. The risk is not hypothetical; it is mechanistically defined. Beyond cardiotoxicity, FTO’s role in non-malignant proliferative disease extends the safety concern. In endometriosis, FTO is overexpressed in ectopic lesions in response to estrogen and inflammatory signals, driving stromal cell invasion via GEF-H1/RhoA in a YTHDF1-dependent manner, and FTO inhibition suppresses lesion growth in vivo ( 167 ). While therapeutically appealing in isolation, this observation illustrates that FTO inhibitors would exert biologically meaningful effects in hormonally and immunologically active non-malignant tissues—with unpredictable consequences for normal endometrial function, immune regulation, and metabolic homeostasis. Across the FTO literature, claims of therapeutic promise are built almost exclusively on short-term tumor response endpoints in immunodeficient models, without organism-level safety assessment, tissue-specific pharmacodynamic profiling, or long-term toxicity evaluation. The context-dependence of FTO function tumor-suppressive in prostate and thyroid cancer yet oncogenic in AML and CRC means that a systemic inhibitor could simultaneously suppress one tumor while promoting progression in a second malignancy or protecting critical normal tissues. What is required is not merely more preclinical data but a conceptual reorientation toward a selective vulnerability model: FTO targeting should be pursued only in tumor contexts where FTO’s net function is demonstrably pro-tumorigenic, where inhibitor delivery can be restricted to the tumor compartment, and where biomarker-guided stratification identifies patients whose normal tissue FTO programs are least likely to be disrupted. Integration of safety endpoints into early-stage mechanistic design is not optional—it is the prerequisite for responsible clinical translation. The proliferation of FTO-related biomarker studies has generated models with high statistical performance in retrospective analyses whose clinical utility remains, upon inspection, largely undemonstrated. The central problem is a systematic conflation of predictive performance measured by AUC values, hazard ratios, and survival stratification with clinical and biological validity, which requires mechanistic anchoring, prospective verification, and independence from confounding tumor-state variables. These are not equivalent, and the field has pursued the former while neglecting the latter. Pan-cancer analyses correlating FTO expression with immune infiltration, tumor mutational burden, microsatellite instability, and checkpoint-gene expression across cancer-genomic cohorts ( 90 ) yield biologically plausible associations but cannot resolve whether FTO is a driver of these immune phenotypes or a surrogate readout of broader tumor-state differences, including tumor purity, stromal composition, and proliferative index, that co-vary with FTO expression without being causally downstream of it. The same limitation applies to machine-learning risk models in colorectal cancer that incorporate FTO alongside seven other m 6 ;A regulators ( 114 ): when FTO’s independent predictive contribution cannot be isolated from the ensemble signal, and when explanatory analyses (e.g., Shapley additive explanations) identify other regulators as the primary contributors, labeling the composite an “FTO-related biomarker” overstates FTO’s specific role. In acute myeloid leukemia, risk models integrating m 6 ;A and immune features ( 115 ) demonstrate external validation in retrospective transcriptomic datasets but not prospective clinical validation in patients receiving defined treatment regimens, a critical distinction between a statistical classifier and a clinically actionable tool. Three structural gaps undermine the translational potential of these models collectively. First, all rely on retrospective bulk transcriptomic datasets whose tumor-purity and immune-composition confounders are computationally estimated rather than experimentally controlled. Second, the gene signatures selected are algorithm-dependent and dataset-sensitive, raising reproducibility concerns that internal cross-validation cannot resolve. Third, and most fundamentally, none of the models establishes that FTO perturbation rather than co-regulated m 6 ;A network state causally determines the immune or prognostic phenotype being predicted. FTO-based biomarkers currently function as integrative readouts of epitranscriptomic tumor state rather than as independent predictive variables. Elevating them to clinical tools requires a fundamental methodological shift: prospective cohort validation, functional dissociation of FTO-specific from m 6 ;A-network contributions, and spatially resolved immune profiling that distinguishes FTO-driven immune exclusion from confounded stromal signals.

The evidence synthesized in this review indicates that FTO functions as a context-dependent regulator in cancer rather than a classical oncogene or tumor suppressor. Its divergent roles across tumor types, oncogenic in AML, HCC, and gastric cancer, but tumor-suppressive in prostate, thyroid, and lung adenocarcinoma, are partly explained by the predominant m 6 ;A reader proteins engaged and the downstream fate of its target transcripts. When YTHDF2-mediated decay predominates, FTO demethylation tends to promote oncogenic programs, immune evasion, and therapeutic resistance. Conversely, when IGF2BP-dependent stabilization is prevalent, the same enzymatic activity can support tumor-suppressive pathways. This context-dependent biology has important therapeutic implications. FTO inhibition should not be assumed to be universally beneficial; in certain contexts, it may inadvertently promote tumor progression or impair normal tissue function, such as cardioprotection during doxorubicin treatment or CD8 + T-cell fitness. Future therapeutic development would benefit from moving beyond broad inhibitor application toward precision strategies, including patient stratification based on tumor lineage, m 6 ;A reader expression profiles, and key target transcript landscapes. Prospective clinical trials will be necessary to validate proposed combination approaches with immune checkpoint inhibitors, ferroptosis inducers, or metabolic modulators. Overall, while mechanistic insights into FTO biology have advanced considerably, translation into clinical benefit remains limited by the lack of robust biomarkers and prospective validation. Progress in this field will require the same context-aware approach that its underlying biology necessitates.

The pharmacological targeting of FTO has evolved through three conceptually distinct phases, each advancing the field while simultaneously exposing limitations that compound rather than resolve. The foundational phase established structural druggability: rational, binding-pocket-directed design produced the first inhibitors that suppressed FTO m 6 ;A demethylase activity, phenocopied genetic depletion in AML by inducing differentiation and apoptosis, and demonstrated in vivo efficacy in xenograft systems ( 44 ). These probes validated the target but were constrained by suboptimal potency and pharmacokinetic liabilities. The subsequent phase pursued systematic chemical diversification thiophene-based scaffolds achieved improved selectivity for FTO over the closely related eraser ALKBH5, and acylhydrazone derivatives delivered sub-micromolar antiproliferative activity with in vivo tumor growth inhibition ( 160 , 161 ) while structure-guided screening of approved drug libraries identified entacapone as an FTO inhibitor operating through a metabolic FTO–FOXO1 axis ( 162 ). The repurposing approach broadened the chemical landscape but introduced pharmacological ambiguity, given entacapone’s established catechol-O-methyltransferase activity and attendant off-target liabilities. A notable mechanistic development involved a shift in mode of action: targeted FTO degradation elevated m 6 ;A methylation on ribosome biogenesis transcripts, routing them to YTHDF2-mediated decay and impairing protein translation with efficacy exceeding catalytic inhibition ( 163 ). This established that FTO’s oncogenic functions are not fully reducible to its enzymatic activity, introducing a critical distinction between active-site occupancy and full biological loss-of-function. Across hematologic models, inhibitor-mediated m 6 ;A elevation converges on suppression of MYC-driven transcriptional programs, upregulation of RARA and ASB2, and disruption of ribosome biogenesis ( 160 , 164 , 165 ). In solid tumor contexts, FTO inhibition attenuates ERBB3/AKT-mTOR signaling in hepatocellular carcinoma ( 166 ) and synergizes with doxorubicin in B-ALL through nucleolar stress and mitochondrial dysfunction ( 165 ), demonstrating that inhibitor phenotypes are context-dependent and not uniformly predictable from AML-derived mechanistic models. Critically, the entire inhibitor landscape remains anchored to hematologic malignancies, with only nascent expansion into solid tumors ( 166 ). Selectivity claims against ALKBH5 are routinely made at the enzymatic level but rarely substantiated transcriptome-wide, and no compound has been evaluated within a stratified biological framework accounting for reader composition, tumor lineage, or upstream signaling context. Current FTO inhibitors therefore function as mechanistic probes rather than context-aware therapeutics. Advancing the field requires next-generation compounds integrated into biomarker-defined deployment frameworks pairing inhibitor selection with YTHDF2/IGF2BP2 reader profiling and tumor-type-specific m 6 ;A target landscapes to ensure that chemical FTO inhibition is applied where its transcriptomic consequences are both coherent and clinically decisive. The synergy observed when FTO inhibition is combined with mechanistically distinct therapies appears, in several preclinical models, to exceed simple additivity. A plausible explanation is that FTO inhibition concurrently affects multiple oncogenic outputs, epitranscriptomic stability, stress responses, and immune-cell function — increasing tumor vulnerability to partner agents that target related axes. Whether this combinatorial logic generalizes across tumor types remains to be tested. Along the epitranscriptomic axis, FTO inhibition elevates m 6 ;A methylation across oncogenic transcripts, directing them through YTHDF2-mediated decay and collapsing downstream proliferative programs. In breast cancer, this destabilization of MYC and E2F1 mRNAs is substantially amplified by BTK inhibition, whose convergent suppression of the same transcriptional network produces synergy because both agents attack the same oncogenic hub through independent molecular inputs ( 21 ). Along the stress amplification axis, FTO loss impairs homologous recombination repair by reducing RAD51 focus formation and compromising homology-directed repair efficiency, rendering cancer cells acutely sensitive to radiation-induced DNA damage in head and neck squamous cell carcinoma ( 136 ). Simultaneously, FTO sustains aerobic glycolysis and intratumoral lactate accumulation that suppresses immune cell activation; relieving this metabolic barrier is further compounded by sonodynamic therapy-driven ROS generation and immunogenic cell death, which activate dendritic cell and T lymphocyte infiltration ( 109 ). The immune reactivation axis is most directly engaged by anti-PD-1 blockade: FTO suppression increases m 6 ;A methylation and YTHDF2-mediated decay of PDCD1 , CXCR4 , and SOX10 , dismantling cell-intrinsic immunosuppressive circuitry and restoring IFNγ sensitivity in an adaptive immunity-dependent manner ( 23 ). Critically, synergy across this framework is not guaranteed, it is axis-dependent and tumor-context-specific. The dominant vulnerability axis varies by lineage: DNA repair deficiency predominates in HPV-negative HNSCC ( 136 ), metabolic-immune coupling in melanoma ( 23 , 109 ), and transcription factor network collapse in breast cancer ( 21 ). No systematic comparison across these axes currently exists. Realizing the full potential of this model will require biomarker-guided combination selection matching partner therapies to the dominant FTO-dependent vulnerability axis operative in each tumor context rather than empirical combination screening.