Ribosome biogenesis and selective translational control as oncogenic vulnerabilities in hematologic malignancies

Abstract

The translation of the transcriptome into a functional proteome is not a passive readout of mRNA abundance but a primary regulatory layer that actively shapes the oncogenic landscape. In hematologic malignancies, convergent oncogenic signaling pathways, including MYC, PI3K/AKT/mTOR, and tyrosine kinase fusions, directly reprogram the translational apparatus, driving a qualitative shift toward the selective synthesis of pro-survival, proliferative, and stemness-associated proteins. This deregulated program is orchestrated through multiple interconnected axes: the accelerated biogenesis of ribosomes within the nucleolus, the assembly of the eIF4F initiation complex, the activation of stress-adaptive programs like the integrated stress response (ISR), and a sophisticated layer of epitranscriptomic, tRNA, and RNA-binding protein-mediated regulation. Hematologic cancers provide a uniquely informative framework for deconstructing these processes, illustrated by inherited ribosomopathies, such as Diamond-Blackfan anemia syndrome and Shwachman-Diamond syndrome, which resolve the paradox of how ribosomal insufficiency can evolutionarily transition into malignant transformation. Beyond pathogenesis, the translational machinery represents a rich landscape of actionable vulnerabilities. Pharmacological targeting of RNA Polymerase I, eIF4A, and the ISR has demonstrated selective antitumor activity and synergistic potential with established therapies, including BCL-2 inhibition. This review synthesizes mechanistic and clinical evidence to position the ribosome as an active regulatory node rather than a passive decoding machine. We further examine the challenges facing the field, including translational plasticity under therapeutic pressure, the lack of validated companion biomarkers, and the emerging role of translational control in modulating anti-tumor immunity. Ultimately, we argue that defining the disease-specific translational states of leukemia and lymphoma is essential for the next generation of precision oncology, moving beyond genomic landscapes toward a proteomic-centered understanding of malignancy.

The mRNA translation paradigm in molecular oncology

The tumor proteome is not a faithful mirror of the transcriptome. Genome-wide quantitative analyses in mammalian cells have demonstrated that protein abundance is predominantly determined at the level of translation, with translational rates accounting for a greater proportion of variance in protein levels than mRNA abundance alone [1]. This finding reframes the classical view of gene expression in cancer: oncogenic programs are not solely defined by which genes are transcribed, but critically by which transcripts are selected for translation and how efficiently. In the context of neoplasia, this distinction is not merely semantic, as it carries direct mechanistic and therapeutic implications.

The recognition that translational control constitutes a primary regulatory layer in oncogenesis has a well-defined historical trajectory. The demonstration that overexpression of the cap-binding protein eIF4E in rodent fibroblasts was sufficient to induce tumorigenic transformation established, for the first time, that a single translation initiation factor could function as a bona fide oncogene [2]. Subsequent work showed that eIF4E cooperates with MYC to accelerate B-cell lymphomagenesis in vivo. In this model, eIF4E antagonizes MYC-driven apoptosis while MYC overrides eIF4E-induced senescence, a functional complementation that directly implicates the translational apparatus as a critical effector of oncogenic signaling [3]. In parallel, the identification of the PI3K/AKT/mTOR axis as a master regulator of cap-dependent translation initiation through phosphorylation of the 4E-binding proteins and S6 kinases connected the most frequently mutated signaling pathway in human cancer to the core translational machinery [4].

A key conceptual advance in the field has been the realization that deregulated translation in cancer cells is not merely a quantitative increase in global protein synthesis, but a qualitative shift in the repertoire of translated mRNAs. Malignant cells preferentially enhance the translation of transcripts encoding cyclins, anti-apoptotic factors (such as MCL-1 and BCL-X_L), receptor tyrosine kinases, and stemness regulators. These mRNAs characteristically harbor long, structured 5’ untranslated regions (5’UTRs) enriched in secondary structures and specific sequence motifs that render them acutely sensitive to changes in eIF4F complex activity [5, 6]. This selectivity provides the molecular basis for a translational therapeutic window: pharmacological inhibition of components such as eIF4A or eIF4E disproportionately suppresses the translation of these oncogenic transcripts while relatively sparing the translation of housekeeping mRNAs with simple, unstructured 5’UTRs [5].

Beyond its role as an oncogenic effector of translation initiation, ribosome biogenesis has emerged as an active participant in tumorigenesis rather than a passive consequence of accelerated proliferation. The nucleolus, as the site of rRNA transcription, processing, and ribosome assembly, functions as a sensor of cellular growth capacity. It integrates signals from oncogenes such as MYC, which directly coordinates RNA Polymerase I, II, and III to match ribosome output with proliferative demand [7]. Perturbation of ribosome biogenesis triggers the nucleolar stress response, in which free ribosomal proteins (notably RPL5 and RPL11) sequester MDM2, thereby stabilizing p53 and enforcing cell-cycle arrest. This is a surveillance mechanism with direct relevance to both the pathophysiology of ribosomopathies and the therapeutic targeting of the nucleolus in cancer [8, 9]. Critically, seminal studies demonstrate that inhibition of RNA Polymerase I also engages p53-independent checkpoints mediated by ATM/ATR signaling, substantially expanding the potential therapeutic applicability of ribosome biogenesis inhibitors beyond tumors with intact TP53 [10].

Hematologic malignancies represent a uniquely informative system for dissecting the oncogenic role of translational control for three convergent reasons. First, the major oncogenic drivers in leukemia and lymphoma, including MYC amplification or translocation, FLT3-ITD, BCR::ABL1, and activating mutations in PI3K/AKT/RAS, converge upon the translational machinery with a directness and intensity that is arguably unmatched in solid tumors [6, 11–18]. Second, the existence of both inherited and acquired ribosomopathies, including Diamond-Blackfan anemia syndrome (DBAS), Shwachman-Diamond syndrome (SDS), X-linked dyskeratosis congenita, cartilage-hair hypoplasia, and the acquired 5q− myelodysplastic syndrome (5q − MDS), provides natural models in which primary defects in ribosome biogenesis or function lead to both hematopoietic failure and increased risk of malignant transformation. These conditions establish a causal link between ribosomal dysfunction and leukemogenesis in humans [19, 20]. Third, the translational approach to therapy has advanced significantly in the hematologic setting. The first-in-human trial of a selective RNA Polymerase I inhibitor (CX-5461) was conducted in patients with advanced hematologic cancers, providing proof-of-concept and pharmacodynamic evidence of on-target ribosomal DNA (rDNA) transcription inhibition in patient tumor cells [21]. Furthermore, eIF4A inhibitors such as zotatifin have demonstrated preclinical synergy with venetoclax in acute myeloid leukemia, defining rational combinatorial strategies that are now approaching clinical evaluation [22].

This review examines the translational apparatus as a causally implicated and therapeutically actionable layer of gene regulation in hematologic malignancies. We organize the discussion around three narrative axes: (I) ribosome biogenesis and the nucleolus as an active oncogenic program, not a mere reflection of proliferative rate; (II) selective translation of pro-survival, pro-stemness, and pro-resistance transcripts as a primary determinant of tumor cell identity; and (III) the translational machinery as a pharmacological vulnerability with a demonstrable therapeutic window in hematologic neoplasms. By integrating mechanistic, translational, and clinical evidence and addressing the limitations that must be resolved for the field to advance, we aim to provide a framework for understanding why the ribosome biogenesis-selective translation axis merits recognition as a first-order regulatory layer in hematologic oncology.

Core concepts in mRNA translation control for the oncology reader

This section provides the mechanistic vocabulary required for the disease-oriented discussion that follows. It is intentionally selective; rather than offering a comprehensive textbook account of translation, we highlight the regulatory nodes that are directly targeted by oncogenic signaling and are pharmacologically actionable in hematologic malignancies.

Ribosome biogenesis and the nucleolus

The synthesis of ribosomes is among the most resource-intensive processes in proliferating cells, consuming a substantial fraction of total transcriptional and metabolic output. Ribosome biogenesis begins in the nucleolus with the transcription of a single polycistronic 47 S pre-rRNA precursor by RNA Polymerase I (Pol I). This 47 S transcript is subsequently processed through a series of endonucleolytic and exonucleolytic cleavages to yield the mature 18 S, 5.8 S, and 28 S rRNAs, which form the catalytic and structural core of the 40 S and 60 S ribosomal subunits, respectively. In parallel, a separate 5 S rRNA is transcribed by RNA Polymerase III, while the mRNAs encoding the approximately 80 ribosomal proteins are transcribed by RNA Polymerase II, requiring tight coordination among all three polymerases for the stoichiometric assembly of functional ribosomes (Fig. 1A) [23–25].

Fig. 1.

Ribosome biogenesis, eIF4F assembly, and the cis-regulatory logic of selective translation. A MYC coordinates the tri-polymerase program of ribosome production: RNA Polymerase I (Pol I) transcription of the 47S pre-rRNA precursor, initiated by the SL1–UBF–RRN3/TIF-IA pre-initiation complex at the rDNA promoter and pharmacologically blocked by CX-5461; Pol II transcription of ribosomal protein mRNAs; and Pol III transcription of 5S rRNA and tRNAs. Mature 40S and 60S subunits are exported from the nucleus via the exportin XPO1/CRM1, targeted by the clinical inhibitor selinexor. B Oncogenic signalling (FLT3-ITD, BCR::ABL1, NOTCH1) converges on PI3K/AKT/mTORC1, which phosphorylates 4E-BP1, releasing eIF4E, and S6K1, which activates eIF4B and targets PDCD4 for degradation, enabling assembly of the eIF4F complex (eIF4E–eIF4G–eIF4A) into a closed-loop mRNA topology with PABP. mTOR kinase inhibitors (TORKi) and the eIF4A inhibitor zotatifin act at the nodes indicated. (C) mTORC1-mediated phosphorylation of LARP1 de-represses translation of 5’ terminal oligopyrimidine (TOP) mRNAs. Oncogenic mRNAs with GC-rich, structured 5’UTRs (MYC, MCL1, BCL2) are disproportionately dependent on eIF4A helicase activity for ribosome scanning, while m⁶A-modified 5’UTRs permit eIF3-mediated cap-independent ribosome recruitment as an alternative initiation route. 4E-BP1, eIF4E-binding protein 1; BCR::ABL1, breakpoint cluster region–Abelson fusion; eIF4F, eukaryotic initiation factor 4 F complex; FLT3-ITD, FMS-like tyrosine kinase 3 internal tandem duplication; LARP1, La-related protein 1; PABP, poly(A)-binding protein; PDCD4, programmed cell death 4; S6K1, p70 ribosomal S6 kinase 1; SL1, selectivity factor 1; TAF, TBP-associated factor; TBP, TATA-binding protein; TIF-IA, transcription initiation factor IA; TOP, terminal oligopyrimidine; TORKi, active-site TOR kinase inhibitor; UBF, upstream binding factor; XPO1, exportin 1

The initiation of rDNA transcription by Pol I is the rate-limiting step in ribosome production and is subject to direct regulation by oncogenes and tumor suppressors. The selectivity factor SL1 (a complex of TBP and Pol I-specific TBP-associated factors [TAFs]) binds to the rDNA core promoter and directs pre-initiation complex formation. UBF (upstream binding factor) facilitates chromatin remodeling at rDNA repeats and stimulates promoter escape, while RRN3/TIF-IA bridges SL1 with the Pol I holoenzyme to enable transcription initiation [26, 27]. MYC functions as a central coordinator of ribosome biogenesis: it binds directly to rDNA, recruits the SL1 complex, and increases Pol I loading at rDNA promoters [28, 29], transcriptionally activates ribosomal protein genes via RNA Polymerase II (Pol II), and induces Pol III transcription of 5 S rRNA and tRNAs. Through this tri-polymerase program, MYC calibrates the cell’s ribosome production capacity to its proliferative demands (Fig. 1A) [7].

The assembly of pre-ribosomal particles requires approximately 200 trans-acting assembly factors in addition to the ribosomal proteins themselves. Key nucleolar chaperones include nucleophosmin (NPM1), which participates in rRNA processing, pre-ribosome export, and centrosome duplication, and fibrillarin (FBL), which catalyzes 2’-O-methylation of rRNA guided by C/D box snoRNAs [30]. A second major class of rRNA modification, pseudouridylation, is catalyzed by dyskerin (DKC1), the catalytic subunit of the H/ACA box small nucleolar ribonucleoprotein (snoRNP) complex [31], while base modifications such as N4-acetylcytidine are deposited by dedicated base-modifying enzymes including NAT10 [32]; these modification systems, together with their guide snoRNAs, are described in detail in “Modifications of rRNA as regulators of hematopoiesis and leukemogenesis” section. The export of nascent 40 S and 60 S subunits from the nucleus to the cytoplasm is mediated by the exportin XPO1/CRM1. XPO1 is the target of the clinically approved inhibitor selinexor, which is approved for relapsed or refractory multiple myeloma in combination with dexamethasone and for relapsed or refractory diffuse large B-cell lymphoma as monotherapy, acting through forced nuclear retention of tumour suppressor proteins and reduced translation of oncoprotein mRNAs [33, 34]. Export of the 60 S subunit additionally requires release of the anti-association factor eIF6, a step catalyzed by the SBDS (Shwachman-Bodian-Diamond syndrome protein)/EFL1 (elongation factor-like GTPase 1). GTPase system that is notably disrupted in Shwachman-Diamond syndrome (Fig. 1A) [35, 36].

Cap-dependent translation initiation and the eIF4F complex

The recruitment of ribosomes to mRNA is the rate-limiting step in eukaryotic translation and the principal locus of translational control. The process begins with the recognition of the 5’ m⁷GpppN cap structure by the eIF4F complex, composed of three subunits, eIF4E (the cap-binding protein), eIF4A (a DEAD-box RNA helicase), and eIF4G (a scaffold protein that bridges the mRNA cap to the ribosome; Fig. 1B) [37, 38].

eIF4G functions as a scaffold protein that simultaneously engages four partners. It binds eIF4E, anchoring the complex to the cap. It binds eIF4A, positioning the helicase near the 5’UTR. It binds eIF3, connecting the complex to the 43 S pre-initiation complex. It binds PABP (poly(A)-binding protein), generating a closed-loop mRNA topology that enhances translational efficiency [38, 39]. Cryo-EM structures of the human 48 S initiation complex have revealed that eIF4F binds upstream of the scanning 40 S subunit, near the mRNA exit channel, and that a second, independent molecule of eIF4A occupies the mRNA entry site, resolving a long-standing question about how mRNA secondary structures ahead of the ribosome are unwound during scanning [40].

eIF4E availability is the critical determinant of eIF4F assembly and is regulated by two convergent mechanisms. First, the 4E-binding proteins (4E-BP1/2/3) compete with eIF4G for a shared binding site on eIF4E, sequestering it in a translationally inactive complex when hypophosphorylated [41]. Second, the MNK1/2 kinases phosphorylate eIF4E at Ser209, modulating its affinity for capped mRNA and its oncogenic activity through mechanisms that remain incompletely resolved but include the enhanced translation of specific pro-survival mRNAs [42].

The PI3K/AKT/mTORC1 signaling axis and mRNA translation regulation

mTORC1 (mechanistic target of rapamycin complex 1) is the principal integrator of growth factor signaling, nutrient availability, and energy status with the translational apparatus. When activated by PI3K/AKT-mediated inhibition of the TSC1/TSC2 complex and the consequent activation of the GTPase Rheb, mTORC1 phosphorylates 4E-BP1 at multiple sites (Thr37/46, Ser65, Thr70), triggering its dissociation from eIF4E and enabling the assembly of the active eIF4F complex [43, 44]. In parallel, mTORC1 phosphorylates S6K1 (p70 ribosomal protein S6 kinase), which in turn activates eIF4B (a cofactor that stimulates the helicase activity of eIF4A) and phosphorylates PDCD4 (an endogenous inhibitor of eIF4A that prevents its incorporation into the eIF4F complex), targeting it for proteasomal degradation [45]. Two landmark studies using ribosome profiling demonstrated that the translational program controlled by mTORC1 via the 4E-BPs consists predominantly of mRNAs bearing 5’ terminal oligopyrimidine (TOP) motifs, a class that encodes ribosomal proteins and translation factors. This revealed that mTORC1 governs not only global translational capacity but also the production of the translational machinery itself [46, 47]. The identification of LARP1 as a direct mTORC1-regulated repressor that binds simultaneously to the m⁷G cap and the adjacent TOP motif has provided a molecular mechanism for this selectivity [48].

Understanding this regulatory hierarchy is essential for interpreting the therapeutic limitations of mTOR-targeting agents. Rapalogs (rapamycin, everolimus, temsirolimus) incompletely inhibit mTORC1-mediated phosphorylation of 4E-BP1, particularly at Thr37/46, and fail to suppress mTORC2. This often leads to the paradoxical reactivation of AKT via removal of the S6K1/IRS-1 negative feedback loop [44]. Active-site mTOR kinase inhibitors (e.g., Torin1, PP242, INK128/sapanisertib) achieve more complete suppression of 4E-BP1 phosphorylation and dual mTORC1/mTORC2 inhibition, with a correspondingly greater impact on cap-dependent translation and oncogenic mRNA translation [46, 47]. The regulatory logic of 4E-BP1 phosphorylation and the pharmacological distinction between rapalogs and active-site TOR kinase inhibitors are depicted in Fig. 1B.

Selective translation: mechanisms of transcript discrimination

The concept that altered translational activity does not affect all mRNAs equally is central to understanding the therapeutic window of translation-targeting agents. Several cis-regulatory features in the 5’UTR determine the sensitivity of individual transcripts to changes in eIF4F activity: (i) TOP and TOP-like motifs: mRNAs encoding ribosomal proteins and translation factors characteristically begin with a 5’ cytidine followed by an uninterrupted tract of pyrimidines. Translation of these TOP mRNAs is acutely sensitive to mTORC1 activity and is repressed by LARP1 when mTORC1 is inactive [46–48]. (ii) Complex secondary structures: Transcripts with long, GC-rich 5’UTRs containing extensive secondary structures, such as those encoding MYC, cyclin D1, MCL-1, BCL-X_L, VEGF, and FGF2, are disproportionately dependent on the helicase activity of eIF4A for efficient ribosome scanning. Increased eIF4F activity preferentially enhances translation of these transcripts, which are largely dispensable at basal translation rates but become critical under oncogenic conditions [49, 50]. (iii) Upstream open reading frames (uORFs): Short ORFs in the 5’UTR can attenuate translation of the main downstream coding sequence under normal conditions but, paradoxically, permit or enhance translation under stress. This is exemplified by ATF4, whose translation increases when eIF2α is phosphorylated and ternary complex availability is reduced [51]. (iv) Internal ribosome entry sites (IRES): Certain cellular mRNAs can recruit ribosomes independently of the cap through structured RNA elements in their 5’UTRs that directly engage ribosomal subunits or initiation factors. IRES-dependent translation becomes particularly relevant under conditions of cellular stress, mitosis, or apoptosis, when cap-dependent initiation is suppressed [52]. (v) m⁶A modifications: N6-methyladenosine (m⁶A) modifications in the 5’UTR can directly recruit eIF3 to facilitate cap-independent translation initiation, independent of canonical YTHDF reader proteins. This is an epitranscriptomic layer with growing implications for leukemogenesis, in which methyltransferase-like 3 (METTL3)-mediated m⁶A deposition on specific transcripts promotes their translation [53–55] (Fig. 1C). The nucleolar machinery of ribosome production, MYC-driven tri-polymerase coordination, eIF4F assembly and regulation, and the cis-regulatory features that confer transcript-selective translation are integrated in Fig. 1.

Elongation and translation fidelity

Translational fidelity refers to the accuracy with which the ribosome selects the correct aminoacyl-tRNA for each codon during elongation, discriminating against near-cognate tRNAs that differ from the correct match by as little as a single nucleotide. Fidelity is enforced kinetically, through proofreading steps within the ribosomal decoding centre that reject incorrectly paired tRNAs before peptide bond formation, and is mechanistically coupled to elongation speed through a well-described trade-off in which faster decoding generally comes at the cost of reduced accuracy [56]. Errors in this process, including amino acid misincorporation, ribosomal frameshifting, and stop-codon readthrough, can generate aberrant or mistranslated proteins that are directly oncogenic or that activate proteotoxic stress responses; conversely, cancer cells can exploit controlled reductions in fidelity, such as stop-codon readthrough enabled by rRNA pseudouridylation (“Modifications of rRNA as regulators of hematopoiesis and leukemogenesis” section), to generate functionally novel protein products. While initiation receives the most attention in the context of translational control in cancer, the elongation phase is increasingly recognized as a regulated step with oncological relevance [57, 58].

While initiation receives the most attention in the context of translational control in cancer, the elongation phase is increasingly recognized as a regulated step with oncological relevance [57, 58]. Elongation is driven by the coordinated action of eEF1A, which delivers aminoacyl-tRNAs to the ribosome in a GTP-dependent manner, and eEF2, a translocase that catalyzes ribosomal translocation along the mRNA [59]. eEF2 kinase (eEF2K), a calcium/calmodulin-dependent kinase, phosphorylates eEF2 to slow elongation under conditions of energy stress and is itself inhibited by S6K1 downstream of mTORC1, providing a link between mTOR signaling and elongation rate [60].

Emerging evidence implicates tumor-specific alterations in tRNA abundance, modifications, and codon usage as an additional layer of translational selectivity. The tRNA methyltransferases NOP2/Sun RNA methyltransferase 2 (NSUN2; which deposits m⁵C) and methyltransferase-like 1 (METTL1) (which deposits m⁷G at position 46, in complex with WD repeat domain 4 [WDR4]) modify specific tRNAs to enhance their stability and decoding efficiency; alterations in these enzymes create codon-usage biases that favor translation of mRNAs enriched in codons matching the most abundant tRNAs in the tumor cell. This mechanism operates in parallel with eIF4F-mediated control at initiation and may contribute to proteomic diversification and drug resistance [61, 62].

Ribosome biogenesis as an oncogenic program and the nucleolus as a central vulnerability

The enlarged and morphologically altered nucleolus has been recognized as a cytological hallmark of malignancy for over a century, although this feature is not universal: tumours retaining wild-type TP53 often display comparatively normal nucleolar morphology, reflecting intact p53-dependent surveillance of ribosome biogenesis (“The RPL5/RPL11–MDM2–p53 pathway and broader nucleolar stress signaling” section) [63]. Yet, the conceptual shift from regarding elevated ribosome biogenesis as a passive consequence of rapid proliferation to understanding it as a causally required oncogenic program is recent. This shift rests on three lines of genetic and pharmacological evidence developed primarily in hematologic models.

Hyperactivation of RNA polymerase I in cancer: requirement, not byproduct

The definitive demonstration that elevated Pol I transcription is required for, rather than merely correlated with, the malignant phenotype came from experiments in the Eµ-Myc lymphoma model. Genetic reduction of Pol I transcription to levels observed in normal B cells, via shRNA-mediated knockdown of UBF or RRN3, was rapidly selected against in competitive culture, thereby establishing that lymphoma cells depend on hyperactivated rDNA transcription for survival [10]. This dependency could be pharmacologically exploited using the small molecule CX-5461, which prevents the association of SL1 with the rDNA promoter with > 200-fold selectivity for Pol I over Pol II, and selectively killed B-lymphoma cells in vivo while sparing the normal B-cell compartment. Critically, tumor cell death occurred within hours, before any measurable reduction in total ribosome content, indicating that the therapeutic effect is a direct consequence of nucleolar disruption rather than an indirect consequence of ribosome insufficiency (Fig. 1A) [10].

The generality of this dependency across hematologic malignancies has since been confirmed. In multiple myeloma, CX-5461 induced growth inhibition and apoptosis in both TP53 wild-type and mutant cell lines, with nanomolar IC50 values, and significantly prolonged survival in the Vκ*MYC myeloma mouse model without severe toxicity to normal tissues [64]. In KMT2A-rearranged acute myeloid leukemia (AML), CX-5461 extended the survival of Kmt2a/ENL-Nras and Kmt2a/AF9-Nras models with the concurrent activation of p53 and cell-cycle arrest [65]. These findings establish that dependence on elevated Pol I transcription is a shared feature of diverse hematologic malignancies, creating a therapeutically exploitable vulnerability.

The RPL5/RPL11–MDM2–p53 pathway and broader nucleolar stress signaling

The molecular mechanism linking Pol I inhibition to tumor cell death proceeds through the nucleolar stress response, a surveillance pathway with increasing mechanistic resolution. When rRNA synthesis is acutely perturbed, newly synthesized ribosomal proteins that cannot be incorporated into assembling ribosomes, specifically RPL5 (uL18) and RPL11 (uL5), together with the 5 S rRNA, form the 5 S RNP complex that binds to MDM2 and inhibits its E3 ubiquitin ligase activity toward p53 [66, 67]. Beyond the canonical 5 S RNP, additional ribosomal proteins extend this surveillance paradigm: RPL23 binds the C-terminal acidic domain of MDM2, distinct from the central domain engaged by RPL5, and is required for p53-mediated tumour suppression downstream of oncogenic RAS signalling, indicating that the RP-MDM2-p53 axis responds to multiple oncogenic inputs beyond MYC-driven ribosome biogenesis [68]. The resulting p53 stabilization activates a transcriptional program encompassing p21 (cell-cycle arrest), p53 upregulated modulator of apoptosis (PUMA), and NOXA (apoptosis), establishing the classical nucleolar stress circuit [8].

It was demonstrated that the 5 S RNP, rather than individual ribosomal proteins, functions as the sensor of ribosome biogenesis defects; disruption of 5 S RNP integrity abolished the p53 response to nucleolar stress, whereas other ribosomal proteins were dispensable [69]. Subsequent cryo-EM structural studies resolved the architecture of the RPL5/RPL11–MDM2 complex, revealing that the 5 S RNP binds to the acidic domain and zinc finger of MDM2; this interaction physically blocks the p53-interaction interface, thereby providing an atomic-resolution explanation for MDM2 inhibition [70].

A series of subsequent studies has substantially complicated the original framing of CX-5461 as a clean Pol I inhibitor. CX-5461 was shown to induce ATM and ATR kinase signaling in the absence of detectable global DNA damage, leading to G1/G2 arrest even in p53-null cells; moreover, combining CX-5461 with ATM/ATR inhibition in Tp53−/− Eµ-Myc lymphoma triggered mitotic catastrophe and significantly enhanced therapeutic efficacy in vivo [71]. This finding was mechanistically elaborated by the discovery that CX-5461 stabilizes G-quadruplex (G4) DNA structures, particularly at rDNA promoter regions, inducing replication stress and activating the DNA damage response independently of its Pol I inhibitory activity [72, 73]. It was further shown that CX-5461-induced replication stress operates through a distinct mechanism involving MRE11-dependent degradation of stalled replication forks, thereby creating synthetic lethality in the context of homologous recombination repair deficiency [74]. Additional evidence has refined the pharmacological profile of CX-5461 by identifying it as a Topoisomerase II poison that induces DNA damage preferentially localized at rDNA loci, positioning it as a locus-specific agent rather than a purely transcriptional inhibitor [75]. Collectively, these findings indicate that CX-5461 is mechanistically composite rather than a selective Pol I inhibitor, and that the relative contributions of Pol I transcription inhibition, G4 stabilisation, and TOP2 poisoning to its clinical activity remain to be disentangled. This recognition has motivated the development of second-generation Pol I inhibitors with improved selectivity. PMR-116, a non-quinolone Pol I inhibitor recently described by the same group that developed CX-5461, exhibits improved drug-like properties (higher solubility, oral bioavailability, blood-brain barrier penetration), inhibits Pol I transcription by stalling polymerase loading at the rDNA promoter, and crucially achieves anti-cancer activity in preclinical haematological and solid tumour models without inducing detectable DNA damage signalling, thereby providing the first pharmacological evidence that selective Pol I transcription inhibition is therapeutically tractable in the absence of genotoxic stress [76].

Nucleolar regulators of ribosome biogenesis in leukemogenesis

Nucleophosmin is the most abundantly expressed nucleolar protein, functioning as a chaperone for pre-ribosomal particle assembly, a shuttle for the nuclear export of pre-60 S subunits, and a regulator of centrosome duplication and genomic stability. In AML, NPM1 is mutated in approximately 30–35% of adult cases, constituting the single most frequent genetic lesion in the disease [77]. These mutations generate a novel nuclear export signal that displaces the NPM1 mutant protein (NPM1c+) from the nucleolus to the cytoplasm, a diagnostic feature detectable by immunohistochemistry [78].

The cytoplasmic dislocation of NPM1c+ deprives the nucleolus of a chaperoning function essential for pre-rRNA processing [79]. However, the relationship between NPM1 mutation and translational output is unexpectedly complex. Rapid degradation of NPM1c + in AML cells had minimal direct impact on the global ribosome footprint, suggesting that the oncogenic effects of NPM1 mutation operate primarily through delocalization of the HOXA/MEIS1 chromatin-regulatory axis rather than through direct modulation of translational efficiency [80]. The HOXA/MEIS1 program is a master regulator of leukemic stem cell (LSC) identity in NPM1-mutant AML, sustaining a stem-like transcriptional state that maintains self-renewal capacity [81]. This oncogenic program likely imposes an elevated demand for ribosome biogenesis, rendering NPM1-mutant cells selectively vulnerable to Pol I inhibition, a vulnerability that is mechanistically distinct from any direct effect of NPM1c + on translational efficiency [82].

Recent work has identified FBL as a direct driver of AML leukemogenesis. FBL depletion causes cell-cycle arrest specifically in AML cells by impairing the 2’-O-methylation of rRNA, a modification that alters the ribosome’s preference for specific oncogenic mRNAs. The small molecule CGX-635 targets FBL by inducing its aggregation and impairing pre-rRNA processing, opening a new therapeutic opportunity [30].

MYC as a major coordinator of ribosome biogenesis in lymphomagenesis

In MYC-driven B-cell lymphomas, the magnitude of ribosome biogenesis increases in proportion to MYC activity. MYC binds to rDNA promoters, recruits SL1 and TRRAP (transformation/transcription domain-associated protein), and increases Pol I loading [28, 29]. Simultaneously, MYC upregulates ribosomal protein mRNAs and nucleolar rRNA-modifying enzymes such as FBL and dyskerin (DKC1) [83].

A genetic demonstration that ribosome biogenesis is required for MYC oncogenesis came from the finding that haploinsufficiency of RPL24 (eL24) dramatically suppressed Myc-driven lymphomagenesis in Eµ-Myc mice without affecting normal development [84]. This establishes that MYC-transformed cells operate at the edge of tolerable translational output. More recently, the concept of ribosome heterogeneity has emerged as a critical layer of MYC-driven control, in which MYC not only increases ribosome abundance but also modulates rRNA modifications to favor the translation of pro-survival transcripts [85, 86].

The conceptual core of the selective translation and remodeling of the tumor proteome

The preceding sections established that cancer cells often exhibit increased ribosome production and activation of signaling pathways that drive translation initiation. While a natural expectation might be that this results in a proportional, genome-wide increase in protein output, one of the most consequential insights in translational oncology over the past decade is that the relationship between translation factor activity and proteomic output is profoundly non-linear. Not all mRNAs respond equally to changes in eIF4F availability or ribosome abundance. Instead, a defined subset of transcripts, enriched for those encoding oncoproteins, anti-apoptotic regulators, cell-cycle drivers, and mediators of stemness, is disproportionately sensitive to the translational state of the cell. This selectivity is not an incidental feature of the system; it is the mechanistic foundation upon which the therapeutic targeting of translation rests. This selectivity predicts that the pharmacological suppression of eIF4F activity will preferentially reduce the expression of the oncogenic proteome while relatively sparing the housekeeping functions necessary for normal cell viability. Without selectivity, a therapeutic window would not exist.

Why does increased global translation not explain the malignant phenotype

The principle of selective translation was anticipated by mathematical modeling long before genome-wide experimental tools existed. In 1974, Lodish proposed that when translation initiation factors are limited relative to the pool of competing mRNAs, even modest changes in the rate of initiation complex assembly will disproportionately affect transcripts that recruit ribosomes inefficiently, while leaving efficiently translated housekeeping mRNAs relatively unchanged [87]. This model predicts that the mRNAs most sensitive to fluctuations in eIF4F activity will be those that present structural or sequence-encoded barriers to ribosome loading, and that oncogenic transformation, by increasing eIF4F availability, will selectively derepress this specific class of transcripts.

The experimental validation of this framework required the development of ribosome profiling (Ribo-seq), which quantifies translational efficiency at nucleotide resolution across the transcriptome. The foundational application of Ribo-seq to the mTOR-translation axis, published simultaneously by Hsieh and colleagues. and Thoreen and colleagues in 2012 produced a surprising and clarifying result. Rather than broadly enhancing the translation of all structured 5’UTR-bearing mRNAs, mTORC1 signaling through the 4E-BPs regulated a discrete translational program dominated by transcripts containing TOP or TOP-like motifs, encoding ribosomal proteins and translation factors themselves [46, 47]. This means that oncogenic mTORC1 activity feeds forward into its own effector system: by promoting the translation of the very machinery required for increased protein synthesis, it amplifies the translational capacity of the cell in a self-reinforcing loop. The finding that loss of 4E-BP1/2 alone was sufficient to render TOP mRNA translation resistant to mTOR inhibition placed the 4E-BP/eIF4E axis, rather than S6K, as the dominant translational effector downstream of mTORC1 [46].

However, the mTOR-sensitive translatome does not fully account for the spectrum of mRNAs whose translation defines the malignant phenotype. Parallel studies demonstrated that eIF4E and eIF4A control partially overlapping but functionally distinct sets of transcripts. Hsieh and colleagues (2012) identified a subset of translationally regulated mRNAs, including vimentin, MTA1, CD44, and Y-box binding protein 1 (YBX1), that are regulated through a pyrimidine-rich translational element (PRTE) in their 5’UTRs and are directly linked to invasion and metastasis in Pten-loss-driven prostate cancer [47]. More recently, genome-wide analysis of the eIF4F-driven translatome using the selective eIF4A inhibitor silvestrol identified 284 genes that depend on eIF4A for efficient translation. These are characterized by long, thermodynamically stable, GC-rich 5’UTRs enriched in G-quadruplex-forming sequences [58]. These eIF4A-dependent mRNAs encode transcription factors, kinases, epigenetic regulators, and oncoproteins, but not housekeeping genes, establishing a clear molecular basis for selectivity [49, 58].

A parallel, proteome-level analysis further refined this picture. Comparative proteomic profiling of eIF4A1 versus eIF4E silencing revealed extensive mechanistic divergence: while both factors promoted proliferation through shared, network-level regulation of cell-cycle proteins, their individual target repertoires were substantially non-overlapping [88]. This indicates that eIF4E and eIF4A do not simply act as interchangeable components of a single complex, but rather serve as independent selectivity filters that together define the oncogenic translatome. The clinical implication is direct: pharmacological inhibitors of eIF4A and eIF4E will have distinct, though partially complementary, therapeutic profiles, and rational combination strategies should take this divergence into account.

A particularly instructive genetic demonstration of the selectivity principle was provided by the evidence that heterozygous reduction of eIF4E to approximately 50% of normal levels had no measurable effect on normal development, global protein synthesis, or organismal viability in mice, but profoundly impaired Ras-driven tumorigenesis [89]. This experiment reveals a fundamental asymmetry: normal cells operate well below the threshold of eIF4E dependence for their essential functions, whereas oncogene-transformed cells have pushed eIF4E demand to a level where even modest reductions become growth-limiting. This differential requirement is the biological basis of the therapeutic window and provides the strongest genetic evidence that targeting eIF4F can achieve cancer selectivity without unacceptable toxicity to normal tissues [89].

Taken together, these data dismantle the naïve model in which translational deregulation in cancer is merely quantitative, a matter of more ribosomes making more of everything. Instead, the emerging picture is one of qualitative reprogramming, in which oncogenic signaling rewires the translational apparatus to selectively amplify the production of a defined set of proteins that drive proliferation, suppress apoptosis, promote invasion, and maintain stemness. The transcripts encoding these proteins share structural features, including long, structured, G-quadruplex-rich or PRTE-containing 5’UTRs, that render them acutely sensitive to changes in eIF4F activity. It is this selectivity that transforms the translational machinery from a passive executor of the transcriptome into an active oncogenic force with its own regulatory logic and vulnerabilities.

eIF4A, eIF4F, and the oncogenic translatome

The identification of eIF4A as the enzymatic selectivity filter within the eIF4F complex has redefined the pharmacological logic of targeting translation initiation. Unlike eIF4E, whose cap-binding function is shared by essentially all capped mRNAs, eIF4A confers selectivity through its helicase activity, which is required in proportion to the thermodynamic stability and structural complexity of the 5’UTR. This means that transcripts with simple, unstructured leader sequences are translated with minimal reliance on eIF4A, whereas those harboring long, GC-rich 5’UTRs with extensive secondary structures, including RNA G-quadruplexes, are acutely dependent on eIF4A-mediated unwinding for ribosome scanning and start-codon recognition [49, 58]. The structural basis of eIF4A inhibition by rocaglates, the sequence-selective mechanism by which zotatifin converts eIF4A into a translational repressor clamped onto polypurine 5’UTR motifs, and the oncogenic translatome that is selectively suppressed by this mechanism, are illustrated in Fig. 1B, C.

The structural basis for this selectivity was resolved by Iwasaki and colleagues, who determined the crystal structure of human eIF4A1 in complex with an ATP analog, the rocaglate rocaglamide A (RocA), and a polypurine RNA strand. The structure revealed that rocaglates function as interfacial inhibitors, occupying a bimolecular cavity formed cooperatively by the eIF4A1 protein and a sharply bent pair of consecutive purine bases in the RNA [90]. This cavity does not exist in the absence of RNA or in the presence of pyrimidine-containing sequences, explaining the remarkable sequence selectivity of these compounds. The practical consequence is that rocaglates do not simply block eIF4A as a helicase; instead, they convert eIF4A from a general-purpose unwinding enzyme into a sequence-specific translational repressor, clamping the factor onto polypurine motifs in the 5’UTR and physically obstructing ribosome scanning at those sites [91]. This gain-of-function mechanism of inhibition has no equivalent among other classes of enzyme inhibitors and defines a unique pharmacological principle.

Zotatifin (eFT226), a synthetic rocaglate optimized for clinical development, exploits this principle with nanomolar potency. Zotatifin promotes the formation of a stable ternary complex between eIF4A, the drug, and polypurine recognition motifs in the 5’UTR of target transcripts, with an IC50 of approximately 2 nM for reporter constructs containing these motifs [92, 93]. The specificity of the interaction was confirmed by surface plasmon resonance, demonstrating increased drug residence time for polypurine-containing RNA oligonucleotides but not for pyrimidine-rich sequences [92]. Among the validated translational targets whose 5’UTR contains zotatifin-sensitive motifs are: (i) cell cycle regulators (CDK4 and cyclin D1); (ii) oncogenic drivers (KRAS, HER2, and FGFR1/2); and survival and stemness factors (MYC, MCL-1, and AKT).

The therapeutic relevance of eIF4A inhibition in B-cell malignancies has been systematically evaluated. It was demonstrated that zotatifin exhibits potent antiproliferative activity (EC50 below 15 nM) and induces apoptosis across a panel of diffuse large B-cell lymphoma, Burkitt lymphoma, and mantle cell lymphoma cell lines. Sensitivity correlated with activation of the PI3K/AKT/mTOR pathway and was associated with dose-dependent suppression of MYC, cyclin D1/D3, BCL-2, and MCL-1 protein levels [94]. In vivo, zotatifin achieved significant tumor growth inhibition in DLBCL and Burkitt lymphoma xenograft models at doses of 1 mg/kg/week or less, with a favorable tolerability profile [94]. Importantly, PTEN loss, which renders lymphoma cells partially resistant to eIF4E-targeting strategies by relieving 4E-BP-mediated repression, was associated with diminished zotatifin efficacy in vivo. This suggests that the integrity of the PI3K/mTOR feedback circuit influences the therapeutic response and should be considered in patient stratification [94].

In parallel, eIF4A inhibition with silvestrol and rocaglamide A has been shown to suppress B-cell receptor (BCR)-driven translational programs in chronic lymphocytic leukemia (CLL). BCR stimulation in CLL cells induces eIF4A expression while simultaneously reducing PDCD4, a natural eIF4A inhibitor, establishing eIF4A as a critical nexus downstream of BCR signaling. Treatment with eIF4A inhibitors blocked anti-IgM-induced global mRNA translation and specifically prevented the accumulation of MYC and MCL-1 without affecting upstream signaling events such as ERK1/2 and AKT phosphorylation [95]. This dissociation between intact upstream signaling and suppressed downstream translational output is a hallmark of on-target eIF4A inhibition and distinguishes it pharmacologically from kinase inhibitors.

The most immediate translational opportunity in hematologic oncology may lie in the combination of eIF4A inhibition with venetoclax in AML. EIF4A1 is highly expressed in leukemia stem and progenitor-like cells compared to their normal counterparts. Ribo-seq analysis revealed that zotatifin treatment induces increased ribosome occupancy in the 5’UTR and around the start codon, consistent with eIF4A clamping, and results in translational downregulation of cell-cycle and survival transcripts. At the protein level, this translates into the coordinated suppression of AKT, STAT5, and MCL-1, three proteins whose sustained expression is central to venetoclax resistance in AML [22, 96]. In patient-derived xenograft models of relapsed and refractory AML, the combination of zotatifin and venetoclax significantly reduced tumor burden and prolonged survival, while the impact on healthy human blood cells was limited [22]. The mechanistic elegance of this combination resides in its dual targeting of complementary anti-apoptotic programs: venetoclax neutralizes BCL-2, while zotatifin suppresses MCL-1 translation. Because LSCs frequently co-express both proteins as a survival strategy, simultaneous inhibition creates a synthetic lethal condition that neither agent achieves alone [22, 96].

A second-generation eIF4A inhibitor was shown to exploit translational reprogramming as a vulnerability in triple-negative breast cancer, achieving in vivo efficacy with oral bioavailability [97]. While this study addressed a solid tumor, it establishes the pharmacological feasibility of next-generation rocaglates with improved drug-like properties, addressing one of the historical limitations of the class and clearing a path toward broader clinical investigation in hematologic contexts.

An important recent advance concerns the recognition that eIF4A controls not only cap-dependent but also certain cap-independent translation events. The translation of the estrogen receptor alpha (ERα) in breast cancer is eIF4A-dependent but eIF4E-independent, proceeding through a non-canonical mechanism mediated by its 5’UTR [98]. Although this finding pertains to breast cancer, it has profound conceptual implications: it reveals that the scope of eIF4A control extends beyond the classical cap-dependent paradigm, suggesting that eIF4A inhibitors may suppress oncogenic programs that escape mTOR/4E-BP/eIF4E-targeting strategies entirely. Finally, resistance mechanisms to rocaglates are beginning to be characterized. A forward genetic screen identified the upregulation of ABCB1/P-glycoprotein as a major mediator of acquired rocaglate resistance, positioning drug efflux as a potential clinical challenge [99]. More remarkably, the identification of a natural amino acid substitution in eIF4A from a parasitic fungus that colonizes rocaglate-producing Aglaia plants demonstrates that the bimolecular drug-binding cavity is subject to evolutionary pressure. This raises the theoretical possibility of target-based resistance through eIF4A mutations in human tumors under prolonged selective pressure [100]. Whether such mutations emerge clinically remains unknown, but their evolutionary precedent warrants prospective monitoring (Fig. 5C).

Fig. 5.

The pharmacological repertoire, rational combinations, resistance mechanisms, and biomarker framework for targeting translation in hematologic malignancies. A Inhibitors act at distinct nodes of the translational apparatus: CX-5461 and BMH-21 block RNA Pol I-mediated rDNA transcription; selinexor blocks XPO1-mediated nuclear export of ribosomal subunits; zotatifin and silvestrol inhibit the eIF4A helicase within the eIF4F complex; rapalogs and active-site mTOR kinase inhibitors (TORKi) block mTORC1-dependent 4E-BP1 phosphorylation; HC-7366 (GCN2 activator) and ISRIB (eIF2B restorer) act at opposite ends of the ISR; STM2457 (METTL3 inhibitor) and Ro 08-2750 (MSI2 inhibitor) represent emerging epitranscriptomic and RNA-binding protein-directed classes. Dark red boxes denote FDA-approved or late-stage clinical agents; hatched pink boxes denote agents in early clinical trials. B Rational combinations exploit synthetic lethality and compensatory signalling: zotatifin-mediated suppression of MCL-1 translation combined with venetoclax-mediated BCL-2 inhibition achieves dual neutralization of the anti-apoptotic shield; concurrent mTOR and MNK1/2 inhibition prevents the compensatory eIF4E phosphorylation that limits mTOR inhibitor monotherapy. C Tumour cells evade translational inhibition through redundant initiation routes (IRES-, m⁶A-dependent) that maintain MYC/BCL-2 production, metabolic rewiring (fatty acid oxidation-driven HRI/eIF2α/ATF4 signalling) that sustains LSC survival under venetoclax pressure, and target evolution through acquired mutations in the eIF4A drug-binding cavity. D Static genomic markers (TP53 status, NPM1 mutation) and functional readouts (ATF4 activity, 4E-BP1 phosphorylation) converge on RNA-based functional translational profiling to guide patient stratification for Pol I-, eIF4A-, and ISR-targeted therapies. HRI, heme-regulated inhibitor; ISR, integrated stress response; LSC, leukemic stem cell; MNK, MAP kinase-interacting kinase; TORKi, active-site TOR kinase inhibitor; XPO1, exportin 1

The integrated stress response and the eIF2α axis as a parallel mRNA translation program

The eIF4F-centered paradigm described above accounts for a substantial fraction of translational deregulation in cancer, but it does not encompass the full landscape. A mechanistically independent program of selective translation operates through the ISR, which converges on eIF2α phosphorylation at serine 51 and engages a distinct set of effector transcripts, with growing relevance to hematologic oncology.

The ISR is initiated by four kinases that sense distinct cellular insults. The protein kinase R-like ER kinase (PERK) responds to endoplasmic reticulum stress and the accumulation of unfolded proteins. The general control nonderepressible 2 (GCN2) detects amino acid deprivation through uncharged tRNAs. PKR is activated by double-stranded RNA, typically of viral origin. The heme-regulated inhibitor (HRI) senses heme depletion and mitochondrial dysfunction [101, 102]. Despite the diversity of upstream signals, all four kinases converge on a single modification: phosphorylation of eIF2α. Phosphorylated eIF2α is a potent non-competitive inhibitor of eIF2B, the guanine nucleotide exchange factor required to regenerate the eIF2-GTP-Met-tRNAi ternary complex, and the resulting reduction in ternary complex availability globally suppresses cap-dependent translation initiation [101, 103]. The structural basis for this inhibition was resolved by two independent cryo-EM and crystallographic studies, which showed that phosphorylated eIF2α locks eIF2B in a nonproductive conformation, preventing the GDP-to-GTP exchange required for translation reinitiation [104, 105] (Fig. 2A). Importantly, the ISR does not operate in isolation from the eIF4F axis but intersects with mTORC1 signaling through metabolic feedback loops. Amino acid deprivation simultaneously activates GCN2 and suppresses mTORC1, reinforcing global translational repression while selectively preserving ISR-dependent translation of ATF4 and its targets. Conversely, mTORC1 inhibition can potentiate ATF4 induction through a mechanism involving LARP1-mediated derepression of TOP mRNAs, establishing a bidirectional regulatory circuit between these two translational control axes [106].

Fig. 2.

The integrated stress response as a bidirectional translational switch in leukemic stem cells. A Four stress-sensing kinases (PERK, GCN2, PKR, HRI), responding respectively to ER stress, amino acid deprivation, double-stranded RNA, and heme/mitochondrial stress, converge on phosphorylation of eIF2α, a non-competitive inhibitor of the guanine nucleotide exchange factor eIF2B, reducing eIF2-GTP-Met-tRNAi ternary complex availability and globally suppressing cap-dependent translation. B Under normal conditions, ribosomes reinitiating after the permissive uORF1 stall at the inhibitory uORF2 of ATF4, precluding translation of the main ORF and its downstream adaptive targets (ASNS, SLC7A11, PSAT1). Under ISR-active conditions, reduced ternary complex availability allows uORF2 bypass, enabling ATF4 translation and induction of these targets. C Leukemic stem cells (LSCs) in the bone marrow niche maintain intrinsically high ISR/ATF4 activity and low global protein synthesis. Venetoclax-induced BCL-2 inhibition and fatty acid oxidation converge on mitochondrial stress, activating HRI and amplifying ISR/ATF4 signalling to sustain LSC survival and venetoclax resistance; the ATF4-LCN2 axis additionally promotes immune evasion through recruitment of immunosuppressive macrophages and T-cell exhaustion. D The ISR displays an intensity-dependent switch between adaptive and apoptotic outcomes. GCN2 activators such as HC-7366 push ISR activity beyond the adaptive threshold, engaging the ATF4-CHOP pro-apoptotic branch, whereas ISRIB restores eIF2B activity and global translation, denying LSCs access to the adaptive program. ASNS, asparagine synthetase; ATF4, activating transcription factor 4; CHOP, C/EBP homologous protein; FAO, fatty acid oxidation; GCN2, general control nonderepressible 2; HRI, heme-regulated inhibitor; ISR, integrated stress response; ISRIB, integrated stress response inhibitor; LCN2, lipocalin 2; LSC, leukemic stem cell; PERK, protein kinase R-like ER kinase; PKR, protein kinase R; PSAT1, phosphoserine aminotransferase 1; SLC7A11, cystine/glutamate antiporter; uORF, upstream open reading frame

The ISR is not merely a brake on protein synthesis; it simultaneously activates a selective translational program that paradoxically enhances the translation of a small group of mRNAs bearing regulatory upstream ORFs in their 5’UTRs. The canonical effector of this program is ATF4, a bZIP transcription factor whose mRNA contains two uORFs that function as a molecular switch [107]. Under conditions of abundant ternary complex, ribosomes that translate the first, short uORF reinitiate efficiently at the second, inhibitory uORF, which overlaps with the ATF4 coding sequence and thereby prevents production of ATF4 protein. When ternary complex availability is reduced by eIF2α phosphorylation, post-termination ribosomes scanning past the first uORF fail to reacquire a new ternary complex quickly enough and bypass the inhibitory second uORF, initiating instead at the ATF4 main ORF [107, 108] (Fig. 2B). Additional transcription factors regulated by this uORF-dependent mechanism include activating transcription factor 5 (ATF5), C/EBP homologous protein (CHOP/DDIT3), and the more recently identified QRICH1 [109]. The identification of non-canonical initiation factors, including eIF2D, DENR/MCT-1, and the cap-binding protein eIF3d, as participants in ATF4 induction has expanded the mechanistic repertoire of ISR signaling beyond the classical ternary complex depletion model [110].

What makes the ISR particularly consequential for hematologic oncology is the recognition that it is not a sporadic, stress-induced phenomenon in tumor cells but a constitutively active program in both normal hematopoietic cells and LSCs. The ISR is uniquely active in human HSCs, with ATF4 expression significantly higher in CD34 + CD38− HSC/MPPs than in committed progenitors [111]. Mechanistically, HSCs display intrinsic scarcity of the eIF2 ternary complex, creating constitutive low-level eIF2α phosphorylation that facilitates their survival under amino acid deprivation [111]. Primitive AML cells recapitulated this phenotype: mass spectrometry revealed lower eIF2α, eIF2B, and eIF2C protein levels in CD34 + CD38− AML cells compared to their CD34 + CD38+ counterparts, and ATF4 reporter expression marked LSC in primary patient samples [111]. These findings position the ISR not as a generic stress response but as a defining molecular feature of the stemness state in both normal and malignant hematopoiesis (Fig. 2C).

Subsequent studies have extended this observation into the domain of therapeutic resistance. Firstly, it was demonstrated that venetoclax resistance in AML stem cells involves metabolic reprogramming with increased fatty acid oxidation (FAO), a process that generates mitochondrial stress and activates the ISR as a survival mechanism [112]. The inhibition of amino acid metabolism selectively targets LSCs, and LSCs isolated from relapsed AML patients display unique metabolic properties, including elevated nicotinamide metabolism, that sustain resistance to venetoclax through mechanisms intimately connected to mitochondrial fitness and ISR engagement [113, 114]. A consolidated view emerged, in which epigenetic plasticity and ISR-driven translational reprogramming are convergent mechanisms sustaining LSC persistence within minimal residual disease, framing the ISR as an intrinsic feature of therapy-resistant AML biology rather than a secondary phenomenon [115] (Fig. 2C).

Beyond its cell-intrinsic roles in LSC survival and metabolic adaptation, the ISR has recently emerged as a critical regulator of tumor-immune interactions. It was demonstrated that intra-tumoral hypoxia imposes chronic ATF4 activity in CD8 + tumor-infiltrating lymphocytes, inducing metabolic polarity, mitochondrial oxidative stress, and a terminal exhaustion state that is independent of T cell receptor stimulation. Genetic or pharmacologic attenuation of ATF4 in CD8 + T cells reduced mitochondrial oxidative stress and restored viability, enabling response to anti-PD-1 therapy and conferring durable tumor control [116]. In a complementary finding, lipocalin 2 (LCN2) was identified as the principal ATF4-dependent effector of immune evasion through an unbiased genetic screen of ATF4-regulated genes. Loss of ATF4 decreased tumor progression in immunocompetent but not immunocompromised mice, and LCN2 promoted T cell exclusion by favoring infiltration with immunosuppressive interstitial macrophages. Anti-LCN2 antibodies promoted robust anti-tumor T cell responses in preclinical models of aggressive solid tumors [117]. Although these data originate primarily from solid tumor models, the high constitutive ISR activity documented in AML’s LSCs suggests that the ATF4-LCN2 axis may similarly contribute to immune evasion in the bone marrow microenvironment, a hypothesis with direct implications for the design of immunotherapeutic combinations in hematologic malignancies (Fig. 2C). Verginadis and colleagues (2022) further expanded the cell-extrinsic scope of the ISR by showing that stromal ATF4 regulates collagen biosynthesis in perivascular cancer-associated fibroblasts, supporting angiogenesis and tumor progression in melanoma and pancreatic cancer models [118].

The ISR also exerts direct control over mitochondrial proteostasis and bioenergetics, a dimension of particular relevance to AML. ATF4 is now recognized as a common downstream mediator of the metabolic reprogramming that follows diverse mitochondrial stressors, including inhibition of mitochondrial translation, disruption of OXPHOS complex stability, and impairment of mitochondrial protein import [119, 120]. In AML, CD39-mediated activation of the P2RY13-cAMP-PKA pathway triggers ATF4-dependent mitochondrial ROS production and enhanced OXPHOS activity, directly contributing to intrinsic chemoresistance [121]. Also, it was demonstrated that venetoclax suppresses mitochondrial respiration and the TCA cycle partially through an ISR-mediated, BCL-2-independent mechanism involving ATF4, raising the possibility that part of venetoclax’s therapeutic efficacy in AML derives from ISR-dependent metabolic disruption rather than solely from apoptosis induction [122]. In a related but distinct discovery, the identification of Ras-responsive element-binding protein 1 (RREB1) as a RAS effector that functions as a selective translational regulator of nuclear-encoded mitochondrial respiratory proteins in AML, coupling translational control to succinate metabolism and chemoresistance [123]. ATF4 also occupies a central node in the regulation of one-carbon metabolism through mitochondrial folate cycle enzymes, particularly MTHFD2, and supports nucleotide biosynthesis under conditions of mitochondrial stress [119]. In DLBCL, it was demonstrated that translational activation of ATF4 through mitochondrial anaplerotic pathways is essential for lymphoma growth and survival, with SIRT3 functioning as a key upstream regulator of TCA cycle flux that sustains amino acid-dependent ATF4 induction [124].

The therapeutic implications of constitutive ISR activation in LSCs are profound but inherently dualistic. The ISR exhibits a well-documented intensity-dependent switch in which moderate activation promotes adaptation through ATF4-driven amino acid biosynthesis, redox management, and autophagy, whereas severe or sustained activation engages the ATF4-CHOP axis, transcriptionally inducing pro-apoptotic genes and suppressing survival factors [125, 126]. ATF4-driven upregulation of amino acid transporters and biosynthetic enzymes, including asparagine synthetase (ASNS), serine biosynthesis pathway components (PHGDH, PSAT1, PSPH), and the cystine/glutamate antiporter SLC7A11, enables metabolic flexibility under nutrient limitation, supporting leukemia cell survival in amino acid-poor niches such as the bone marrow microenvironment [127, 128]. In T-ALL, ATF4 and ZBTB1 are recruited to the ASNS promoter to sustain asparagine biosynthesis under L-asparaginase pressure, providing a mechanistic basis for the clinical observation that ISR activation mediates resistance to this therapy [129].

This dose-response relationship creates a pharmacological opportunity that has recently been translated into a clinical-stage compound. HC-7366 simultaneously overcomes multiple known venetoclax resistance mechanisms through several concerted actions: ISR hyperactivation, increased levels of the pro-apoptotic protein PUMA, reduced mitochondrial respiration and glycolysis, and suppressed S100A8/A9, a protein pair whose elevation correlates with venetoclax resistance [130]. In the MV4-11 FLT3-ITD/KMT2A-rearranged model, the combination of HC-7366 with venetoclax produced 26% tumor regression, and a genome-wide CRISPR screen in KG-1 cells revealed that BCL2 sgRNAs were significantly depleted in HC-7366-treated cells, suggesting a synthetic lethal relationship between GCN2 activation and BCL-2 inhibition that provides a rational basis for the therapy combination [130]. The in vivo efficacy of HC-7366 was independent of TP53 mutational status, with 100% complete response in the TP53-mutant MOLM-16 model, a finding of considerable clinical significance given the dismal outcomes of TP53-mutant AML under current regimens [130]. The bidirectional therapeutic logic of ISR modulation in LSCs, encompassing GCN2 activation as a strategy to push ISR intensity beyond the adaptive threshold and ISRIB-mediated restoration of global translation as a strategy to deny adaptive survival programs, is schematized in Fig. 2D.

Conversely, pharmacological inhibition of the ISR can also serve as a therapeutic strategy by denying tumor cells access to the adaptive programs that sustain survival under microenvironmental stress. ISRIB enhances eIF2B activity, overcoming the inhibitory effect of phosphorylated eIF2α and restoring global translation even during stress conditions [131, 132]. By preventing the ISR-mediated translational rewiring that supports leukemia cell survival under hypoxia and nutrient deprivation, ISRIB and related compounds may sensitize tumor cells to chemotherapy and targeted agents. This bidirectional therapeutic logic, where both hyperactivation and blockade of the ISR can be cytotoxic depending on context, mirrors the dual nature of the pathway itself and demands careful patient selection (Fig. 2D).

It was recently described as a “split ISR” in which reduced eIF2B activity engages a distinct ISR program with altered translational control and gene expression that is separable from the canonical eIF2α phosphorylation-dependent module [133]. In trastuzumab-resistant HER2 + breast cancer cells, ER stress-induced ATF4 expression was shown to depend primarily on eIF4E-mediated non-canonical ISR rather than the canonical PERK-eIF2α axis, revealing cross-talk between the two major translational control circuits [133]. Although this observation originates from a solid tumor model, it has immediate implications for hematologic malignancies, where the relative contributions of canonical versus non-canonical ISR in sustaining LSC survival remain unexplored. Complementing this, ATF4 has recently been shown to occupy hundreds of genomic loci under basal conditions, functioning not merely as a stress-induced transcription factor but as a constitutive chromatin-associated regulator, poised for rapid transcriptional amplification upon ISR activation. Selective transcriptional activation of ATF4-bound genes upon stress does not require increased H3K27 acetylation, chromatin accessibility changes, or enhancer-promoter loop reorganization, but instead depends on the redistribution of CEBPγ from non-ATF4 sites to ATF4-bound regions, forming ATF4/CEBPγ heterodimers [134]. This pre-wired chromatin architecture explains why cells with constitutively elevated ATF4, such as HSCs and LSCs, possess a transcriptional program that can be activated instantaneously under therapeutic stress, conferring a survival advantage that is embedded in the epigenetic landscape rather than requiring de novo regulatory events.

The ISR also intersects with DNA replication and genome integrity. ISR activation rapidly reduces DNA replication fork progression through suppression of histone synthesis and accumulation of DNA-RNA hybrids (R-loops) that obstruct the replication machinery [135]. This coupling between translational stress and replication stress creates vulnerabilities that could be exploited in combination with replication-targeting agents such as ATR inhibitors, which are already under clinical investigation in hematologic malignancies.

Non-canonical regulators with direct translational functions

The models of translational control discussed in the preceding sections are centered on the canonical initiation factors (eIF4F, eIF2) and the signaling pathways that regulate them. However, an expanding body of evidence demonstrates that proteins historically classified as transcription factors, RNA-binding proteins, or epitranscriptomic modifiers operate directly on the translational machinery, often through mechanisms that are entirely independent of the eIF4F complex. In hematologic malignancies, these non-canonical regulators are particularly relevant because many of them are genetically altered, functionally linked to stemness programs, or required for LSC maintenance. Their identification expands the repertoire of druggable nodes in translational control beyond what eIF4F-targeting or mTOR-targeting strategies can reach.

Among the best-characterized non-canonical translational regulators in myeloid leukemia is Y-box binding protein 1 (YBX1), a member of the cold-shock protein superfamily. Although YBX1 was originally studied as a transcription factor, its cytoplasmic fraction functions as a direct translational regulator by binding to structured 5’UTRs and 3’UTRs of target mRNAs and modulating their recruitment to polysomes. YBX1 is an essential driver of AML development and maintenance, and its genetic inactivation by CRISPR-Cas9 leads to selective loss of oncogenic transcripts, including MYC, from the polysomal fraction without affecting their total mRNA abundance [136]. This distinction is critical, establishing that YBX1 controls MYC expression at the level of translational efficiency rather than transcription or mRNA stability, as confirmed by parallel RNA-seq of total versus polysome-associated mRNAs. Approximately 20% of transcripts lost from polysomes upon YBX1 deletion were identified as functional dependencies by CRISPR screening, and many of these, including cell-cycle mediators and ribosomal subunits, showed correspondingly decreased protein levels by global proteomics correspondingly [136]. In a complementary study, Feng and colleagues showed that YBX1 cooperates with insulin-like growth factor 2 mRNA-binding proteins (IGF2BPs)via its cold-shock domain to stabilize m⁶A-tagged transcripts of BCL-2 family members (BCL2L1, MCL1) and MYC in leukemia-initiating cells, and that conditional deletion of Ybx1 in murine AML models induced apoptosis and impaired leukemic capacity without affecting normal hematopoiesis [137]. Together, these findings suggest that YBX1 and IGF2BP proteins form cooperative ribonucleoprotein assemblies that function as integrated translation-stability modules, simultaneously enhancing mRNA half-life and translational efficiency of oncogenic transcripts, enabling sustained protein output even under conditions of global translational repression such as metabolic stress or therapeutic pressure [136, 137] (Fig. 3D).

Fig. 3.

Ribosome compositional heterogeneity, rRNA and tRNA epitranscriptomic modification, and non-canonical m⁶A-linked regulators. A Substoichiometric enrichment of individual ribosomal proteins generates functionally distinct ribosome populations: RPS25-enriched ribosomes preferentially translate Wnt-pathway transcripts, whereas RPL10A-enriched ribosomes favour metabolic transcripts. The somatic RPL10-R98S mutation in T-ALL enhances BCL-2 translation. Post-translational modifications, including RPS6 phosphorylation by S6K1 and RPS2/RPS3 ubiquitylation, further fine-tune ribosome function. B Fibrillarin (FBL)-guided 2’-O-methylation of rRNA is repressed by p53; p53 loss elevates FBL and rRNA hypermethylation, enhancing IRES-dependent translation of MYC and IGF1R. Dyskerin (DKC1)-catalysed pseudouridylation maintains IRES-dependent translation of the tumour suppressors p27 and XIAP; DKC1 loss, as in dyskeratosis congenita, impairs this translation. C The METTL1/WDR4 complex deposits m⁷G on specific tRNAs (e.g., Arg-TCT), enhancing decoding efficiency and preferentially promoting translation of AGA-codon-enriched cell-cycle transcripts. MYC-driven RNA Polymerase III activity expands the modified tRNA pool, synchronizing tRNA supply with oncogenic translational demand. D The METTL3 writer complex deposits m⁶A, which is interpreted by IGF2BPs and YTHDF1 to enhance translation of MYC and BCL2. YBX1 and MSI2 bind the 3’UTR and 5’UTR, respectively, of target transcripts, directly recruiting ribosomes and enhancing translation of stemness-associated programs in AML. DKC1, dyskerin pseudouridine synthase 1; FBL, fibrillarin; IGF2BP, insulin-like growth factor 2 mRNA-binding protein; IRES, internal ribosome entry site; METTL1, methyltransferase-like 1; METTL3, methyltransferase-like 3; MSI2, Musashi RNA-binding protein 2; WDR4, WD repeat domain 4; XIAP, X-linked inhibitor of apoptosis protein; YBX1, Y-box binding protein 1; YTHDF1, YTH domain-containing family protein 1

A related non-canonical translational regulator with emerging relevance in hematologic malignancies is DAP5 (also known as EIF4G2, p97, or NAT1). DAP5 is a structural paralog of eIF4G that lacks the eIF4E- and PABP-binding domains, rendering it incapable of supporting canonical cap-dependent translation initiation. Instead, DAP5 interacts with eIF4A and eIF3 to facilitate ribosome recruitment to mRNAs with long, structured 5’UTRs or upstream open reading frames, a class that is substantially enriched for regulatory and pro-survival transcripts, including BCL2 and CDK1, whose cap-independent translation DAP5 sustains during mitosis and cellular stress [138]. The direct relevance of DAP5 to myeloid leukemia is supported by the observation that miR-139-5p, a microRNA whose promoter is transcriptionally silenced across multiple cytogenetically defined AML subgroups, including t(8;21), inv(16), KMT2A-rearranged, and complex karyotype AML, directly targets EIF4G2, and that restoration of miR-139-5p expression induces cell cycle arrest and apoptosis in vitro and in xenograft models in vivo [139]. While a comprehensive characterization of DAP5’s translational targets in LCSs remains to be performed, its ability to maintain selective translation of pro-survival mRNAs under conditions of eIF4F suppression positions it as a candidate mediator of translational resilience in hematologic malignancies [138].

A mechanistically distinct but functionally adjacent layer of translational regulation is provided by nonsense-mediated mRNA decay (NMD), a translation-coupled mRNA quality control pathway that recognizes and degrades transcripts harboring premature termination codons through the exon-junction complex (EJC) and its central RNA helicase eIF4A3 [140, 141]. Although mechanistically distinct from the cytoplasmic eIF4A1/eIF4A2 paralogues that drive cap-dependent initiation discussed in “eIF4A, eIF4F, and the oncogenic translatome” section, eIF4A3 operates as a nuclear DEAD-box helicase that coordinates pre-mRNA splicing, mRNA export, and NMD-mediated transcript surveillance, with direct downstream consequences for the translational output of the cell. Kanellis and colleagues demonstrated that eIF4A3 controls cell fate through coordinated regulation of ribosome biogenesis and translational output, positioning eIF4A3 as a node that integrates Pol I and Pol II post-transcriptional events [142]. In acute myeloid leukemia, eIF4A3 has emerged as a tractable therapeutic target, with pharmacological or genetic inhibition impairing AML cell proliferation and survival [143]. In parallel, NMD itself shapes the translational landscape of leukemic cells through degradation of aberrantly spliced transcripts generated by recurrent mutations in splicing factors such as SF3B1, SRSF2, U2AF1, and ZRSR2, and through regulation of physiological NMD substrates that include uORF-containing mRNAs critical for the ISR. The splicing factor RBM17 was shown to mark and sustain LSCs by repressing the inclusion of poison exons in pro-leukemic transcripts and in EIF4A2, thereby evading NMD and preserving translation of these factors; RBM17 depletion triggers NMD-mediated degradation of these transcripts and recapitulates the effect of EIF4A2 inhibition in AML stem cells [144]. These observations position NMD and the EJC not as passive surveillance modules but as active determinants of the translatable transcriptome in hematologic malignancies, expanding the repertoire of translation-coupled vulnerabilities beyond canonical cap-dependent initiation.

This convergence between RNA-binding proteins and epitranscriptomic modifications is anchored in the m⁶A system, which has emerged as one of the most consequential regulators of translation in hematologic malignancies. The writer complex, centered on the methyltransferase METTL3 in association with methyltransferase-like 14 (METTL14), Wilms tumor 1-associated protein (WTAP), and the cofactor RBM15, deposits m⁶A at consensus DRACH motifs on nascent mRNAs. METTL3 is bound to the promoters of active genes in AML cells, and its methyltransferase activity is required to maintain the translation of key oncogenic transcripts, including MYC, BCL2, and PTEN, through enhanced ribosome engagement [53]. It was also shown that METTL3 is essential for the maintenance of AML and that its depletion induces differentiation and apoptosis in leukemia cells and delays leukemogenesis in vivo [145]. In addition to its nuclear role as a methyltransferase, METTL3 has been shown to localize to the cytoplasm, where it promotes translation independently of its catalytic activity by directly interacting with eIF3H at the 5’UTR-proximal region, facilitating mRNA circularization and enhancing the translational efficiency of oncogenic transcripts [146]. This non-canonical cytoplasmic function expands the scope of METTL3 as a dual regulator of gene expression at both the epitranscriptomic and translational levels. Notably, RBM15, a component of the METTL3 complex, is also the 5’ fusion partner in the t(1;22)(p13;q13) translocation that generates the RBM15-MKL1 oncoprotein in acute megakaryoblastic leukemia, providing a direct genetic link between the m⁶A machinery and leukemogenesis [147]⁶A writer-reader-eraser network and its convergence with cap-independent translation initiation, RBP-mediated translational stabilization, and the mechanistic basis for the apparently paradoxical anti-leukemic effects of both METTL3 and FTO inhibition, are illustrated in Fig. 3D.

The functional output of m⁶A deposition depends on reader proteins that interpret the modification and direct the marked transcript toward distinct fates. The classical model assigned specific functions to individual YTHDF proteins, with YTH domain-containing family protein 1 (YTHDF1) promoting translation, YTHDF2 mediating decay, and YTHDF3 acting as a facilitator of both processes [148]. However, this framework has been substantially revised. YTHDF1, YTHDF2, and YTHDF3 act largely redundantly in regulating mRNA degradation, and translational enhancement is not the exclusive domain of any single reader but rather emerges from context-dependent interactions with additional RNA-binding proteins, initiation factors, and RNA structural features [149]. This revised model implies that the translational output of m⁶A-modified transcripts is not hardwired to a single reader protein but arises from combinatorial assemblies of readers and co-factors, further increasing the plasticity of epitranscriptomic regulation in leukemia cells. A conceptually distinct class of m⁶A readers, the insulin-like growth factor 2 mRNA-binding proteins (IGF2BP1/2/3), stabilize and enhance translation of m⁶A-containing transcripts through a mechanism that increases both mRNA half-life and polysome association [150]. IGF2BPs were identified as a new family of m⁶A readers whose target repertoire includes MYC, MYB, and CDK6 [150]. In AML, IGF2BP2 is selectively required for leukemia cell survival and to cooperate with YBX1 in stabilizing m⁶A-modified transcripts of pro-survival genes [137]. Importantly, m⁶A-dependent translation can occur independently of eIF4E through direct recruitment of eIF3 to modified 5’UTRs, providing an alternative cap-independent initiation mechanism that becomes particularly relevant under conditions of eIF4F inhibition or mTOR suppression [151]. This observation connects the epitranscriptomic axis to the broader theme of translational resilience in cancer cells, where multiple redundant pathways maintain oncogenic protein synthesis even when the canonical eIF4F route is pharmacologically blocked (Fig. 3D).

The RNA-binding protein Musashi-2 (MSI2) represents perhaps the most thoroughly characterized non-canonical translational regulator in LSC biology. MSI2 is the predominant Musashi family member expressed in hematopoietic stem cells; its overexpression cooperates with BCR::ABL1 to induce aggressive leukemia, and its expression correlates with decreased survival in patients with AML [152]. Building on this, conditional deletion of Msi2 in the hematopoietic compartment was shown to delay leukemogenesis and impair LSC function in KMT2A-AF9-driven AML [153]. Using HITS-CLIP combined with RNA-seq, the same study demonstrated that MSI2 directly binds to and maintains efficient translation of HOXA9, MYC, and IKZF2, three transcripts central to the KMT2A self-renewal program, through binding to consensus motifs in their 3’UTRs, competing with translational repressors and promoting ribosome engagement [153, 154]. Vu and colleagues extended the MSI2 interactome through proteomic analysis, identifying synaptotagmin-binding cytoplasmic RNA-interacting protein (SYNCRIP) as a functionally required RBP partner of MSI2 that maintains HOXA9 translation in LSCs [155]. More recently, DEP domain-containing mTOR-interacting protein (DEPTOR), a negative regulator of mTOR, was identified as transcriptionally induced by ATF4 and post-transcriptionally stabilized by MSI2 binding to its mRNA, creating a regulatory circuit that links the ISR, mTOR signaling, and MSI2-mediated translational control in AML [156]. This finding extends the role of MSI2 beyond stemness maintenance to include metabolic adaptation and stress responses, suggesting that MSI2-dependent translational control contributes to LSC survival under the fluctuating nutrient and oxygen conditions of the bone marrow niche. The small molecule Ro 08-2750 was identified as a direct MSI2 inhibitor that competes for RNA binding, reduces MYC expression, and demonstrates in vivo efficacy in a murine AML model, providing proof-of-concept that RNA-binding proteins can be pharmacologically targeted in leukemia [157] (Fig. 3D).

The DEAD-box helicase DDX3X occupies a distinct niche in translational regulation. Unlike eIF4A, which unwinds secondary structures within 5’UTRs to facilitate ribosome scanning, DDX3X participates in translation initiation through interactions with eIF3 and the 43 S pre-initiation complex, particularly for mRNAs with complex 5’UTR structures [158]. DDX3X is recurrently mutated with loss-of-function in T-ALL, Burkitt lymphoma, and NK/T-cell lymphoma [159, 160]. It was also demonstrated that DDX3X loss-of-function mutations alter the translation of specific programs involved in differentiation and cell-cycle control, contributing to MYC-driven lymphomagenesis through a translational mechanism rather than through the transcriptional or splicing functions that had previously been attributed to DDX3X [161]. More broadly, DDX3X appears to function as a sensor of RNA secondary structure that selectively enables translation of transcripts with highly structured 5’UTRs while restricting aberrant initiation events. In this framework, DDX3X acts as a translational checkpoint that preserves proteome integrity, and its loss may promote oncogenesis not by reducing global translation but by reshaping the translational landscape to favor proliferative over differentiation-associated programs [54, 161].

Finally, emerging evidence indicates that even factors traditionally classified as general translation initiation components can exert transcript-selective functions. The multi-subunit eIF3 complex, long regarded as a passive scaffold for 43 S pre-initiation complex assembly, has been shown to directly bind specific 5’UTR elements in a subset of mRNAs, including those encoding MYC and cell-cycle regulators, and to regulate their translation independently of eIF4F [162]. The eIF3D subunit possesses an alternative cap-binding activity that enables cap-dependent but eIF4E-independent translation, a mechanism that may become particularly relevant under conditions where mTOR/4E-BP signaling suppresses eIF4E availability, such as during nutrient stress or pharmacological mTOR inhibition [163]. Although the specific role of eIF3-mediated selective translation in hematologic malignancies remains to be fully characterized, these findings challenge the assumption that general initiation factors function identically across all mRNAs and suggest that transcript-selective roles for individual eIF3 subunits may represent additional vulnerabilities in leukemia cells that depend on non-canonical translation.

Ribosome heterogeneity and onco-ribosomes

From canonical homogeneity to functional specialization

For much of the twentieth century, the ribosome was regarded as a molecular machine of invariant composition, universally decoding mRNAs without selectivity. This view rested on the assumption that the approximately 80 ribosomal proteins and four rRNAs that constitute the eukaryotic ribosome are assembled in fixed stoichiometries, producing functionally interchangeable particles. The conceptual challenge to this framework originated with the ribosome filter hypothesis, which proposed that ribosomes could vary in composition and thereby preferentially translate specific subsets of mRNAs [164]. An early and influential functional demonstration of this principle came from mice harbouring hypomorphic mutations in Rpl38, in which global protein synthesis was unaffected but translation of a specific subset of Hox mRNAs was selectively impaired through loss of RPL38-dependent recognition of IRES-like elements, producing pronounced homeotic transformations of the axial skeleton and establishing that a single ribosomal protein can confer transcript-specific translational control in vivo [165]. Although initially speculative, this proposal has been substantiated over the past decade by quantitative proteomic, genetic, and functional studies that collectively establish ribosome heterogeneity as a bona fide regulatory layer in gene expression. More recent frameworks have further refined this model by proposing that ribosome heterogeneity exists along a continuum rather than as discrete subpopulations, with variations in ribosomal protein stoichiometry, ribosomal protein and rRNA sequence (including paralog selection, such as RPL22 versus RPL22L1, and rDNA sequence polymorphisms), rRNA modification patterns, and ribosome-associated factors generating a spectrum of translationally distinct ribosomes whose functional output is probabilistic and context-dependent [166, 167].

The first genome-wide demonstration that compositionally distinct ribosomes preferentially translate different mRNA subpools came from selective ribosome profiling, which showed that ribosomes containing or lacking specific ribosomal proteins, notably RPS25/eS25, exhibit measurable differences in translational output across the transcriptome [168]. RPS25-containing ribosomes preferentially translated mRNAs involved in cell signaling and vitamin B12 metabolism, while RPS25-depleted ribosomes showed no such enrichment, establishing that substoichiometric variation in a single ribosomal protein is sufficient to alter the translational landscape [168] (Fig. 3A).

A critical extension of this concept into stem cell biology and developmental regulation came from a work that employed quantitative mass spectrometry to map ribosome composition changes during human embryonic stem cell differentiation along the mesoderm lineage [169]. The study identified 31 ribosomal proteins whose stoichiometry in actively translating polysomes changed significantly during cell fate specification. Among these, RPL10A/uL1 was upregulated in polysomes of primitive streak cells and progressively declined during differentiation toward the sclerotome [169]. An Rpl10a loss-of-function allele in mice produced striking mesodermal phenotypes, including posterior trunk truncations, and ribosome profiling in these mice revealed decreased translation of Wnt pathway regulators, which were also preferentially enriched on RPL10A-containing ribosomes [169]. These findings reveal unexpected compositional modularity in the ribosome that is attuned to the differentiation state of the cell, suggesting that ribosome heterogeneity is not a static property but a dynamically regulated feature that changes in concert with cell fate transitions. Notably, the 31 heterogeneous ribosomal proteins identified in this study were statistically enriched at solvent-accessible positions on the ribosome surface, near mRNA entry and exit channels, consistent with a model in which these proteins modulate interactions with specific mRNAs or regulatory factors rather than altering core catalytic function [169] (Fig. 3A).

Rather than representing fixed subpopulations, specialized ribosomes are increasingly understood as dynamic entities whose composition and function are continuously remodeled in response to cellular context, stress, and oncogenic signaling [166, 170]. This remodeling is tightly coupled to the cellular metabolic state and to the signaling pathways discussed in preceding sections. MYC-driven transcriptional programs do not increase all ribosomal protein mRNAs uniformly but differentially upregulate specific components, thereby altering the stoichiometric balance of the resulting ribosomes and contributing to compositional heterogeneity [7, 84]. Similarly, stress-responsive pathways can reshape ribosome composition and function through selective regulation of ribosomal protein translation, as TOP mRNAs encoding different ribosomal proteins exhibit varying sensitivities to mTORC1 inhibition [46]. The P-stalk proteins provide a striking recent example of how quantitatively minor variations in ribosome-associated proteins can exert outsized effects on specific translational programs. Dopler and colleagues (2024) demonstrated that P-stalk-containing ribosomes act as master regulators of cytokine-mediated processes, selectively translating transcripts encoding inflammatory mediators [171].

The relevance of ribosome heterogeneity to cancer was formalized with the concept of “onco-ribosomes,” defined as ribosomes that have undergone cancer-specific alterations in their protein composition, rRNA modifications, or associated factors [172, 173]. Rather than representing a fixed entity, onco-ribosomes are increasingly viewed as context-dependent assemblies that emerge from the convergence of oncogenic signaling, altered ribosome biogenesis, and RNA modification programs. This dynamic definition better reflects the plasticity observed in cancer cells and accommodates the coexistence of multiple ribosome populations within the same tumor [172]. Onco-ribosomes are proposed to arise through two classes of mechanisms: programmed changes driven by oncogenic signals, such as MYC-dependent upregulation of specific ribosomal proteins and rRNA-modifying enzymes, and spontaneous somatic mutations in ribosomal protein genes that alter translational selectivity [172, 173].

The RPL10 (uL16) R98S mutation, found in approximately 8% of T-cell acute lymphoblastic leukemia, provides a paradigmatic example. This mutation was identified through exome sequencing, and subsequent functional studies demonstrated that RPL10-R98S alters IRES-dependent translation, enhances the translation of the anti-apoptotic protein BCL-2, and increases JAK-STAT signaling, collectively promoting leukemia cell survival [160, 174]. Notably, RPL10-R98S has also been shown to enhance overall translational fidelity alongside translation levels, indicating that its oncogenic selectivity does not reflect a general loss of decoding accuracy but rather a qualitative reprogramming of ribosome function [175]. The effect of RPL10-R98S on global protein synthesis is highly context-dependent: the mutation impairs ribosome maturation and proliferation under standard growth conditions, while paradoxically enhancing translation levels in other cellular contexts, including nutrient-replete conditions favouring recombinant protein production [175]. Rather than reflecting a uniform change in global translational output, the oncogenic contribution of RPL10-R98S is better explained by its capacity to shift the relative efficiency of translation toward specific oncogenic programs, such as IRES-dependent BCL-2 translation, illustrating the selectivity principle that distinguishes onco-ribosomes from mere translational hyperactivation [160, 174] (Fig. 3A). More recent functional studies have extended these findings by demonstrating that cancer cells can become selectively dependent on specific ribosomal proteins or ribosome-associated factors. Systematic CRISPR-based screens and cancer dependency mapping have revealed context-specific essentiality of individual ribosomal components that varies across tumor types and genetic backgrounds, indicating that not all ribosomes are equally dispensable in tumor cells and that specific ribosomal protein dependencies may represent exploitable vulnerabilities [24, 176]. A direct illustration of this principle comes from the RPL5-I60V mutation identified in T-ALL, which increases global ribosome output and cellular proliferation while conferring differential sensitivity to a panel of translation-targeting compounds, including increased sensitivity to silvestrol, homoharringtonine, and cytarabine, providing a concrete example of how a single oncoribosome-generating mutation can be pharmacologically exploited [177].

Post-translational modifications of ribosomal proteins add another dimension to functional heterogeneity. Ubiquitylation of RPS2 (uS5) and RPS3 (uS3) regulates ribosome quality control and translational fidelity, while UFMylation of RPL26 (uL24) is essential for ER-targeted translation and is required for hematopoietic stem cell function [178, 179]. Phosphorylation of RPS6 (eS6) by S6K1, long used as a readout of mTORC1 activity, is now recognized to modulate ribosome selectivity for specific mRNA classes rather than merely serving as a marker of global translational activation [180]. A systematic proteomic survey of the mammalian ribosome interactome catalogued the full landscape of ribosomal protein post-translational modifications alongside heterogeneous ribosome-associated proteins, revealing extensive and combinatorial modification patterns, including acetylation, methylation, and additional phosphorylation and ubiquitylation sites beyond those described above, that vary across cell types and developmental contexts [181, 182]. These modifications do not alter the core catalytic function of the ribosome but fine-tune its interaction with specific mRNAs, initiation factors, and regulatory proteins, expanding the combinatorial potential for translational regulation (Fig. 3A).

The scaffold protein RACK1, which stably associates with the 40 S subunit near the mRNA exit channel, exemplifies how ribosome-associated factors can modulate translational selectivity. RACK1 recruits kinases, including PKC and Src, to the ribosome, integrating signal transduction with translation [183]. In cancer, RACK1-containing ribosomes have been shown to mediate IRES-dependent translation of viral and cellular mRNAs and to preferentially translate oncogenic transcripts, including MYC and survivin [183, 184]. RACK1 exhibits a dynamic association with ribosomes, meaning that its presence or absence on a given ribosome population can shift the translational output of the cell without requiring changes in ribosome biogenesis rate [184]. Similarly, eIF6, a factor that prevents premature 60 S–40 S subunit joining and whose release is catalyzed by the SBDS/EFL1 system disrupted in Shwachman-Diamond syndrome, has been shown to function as a regulator of translational efficiency rather than an assembly factor. Cytoplasmic eIF6 restricts translational output, and its haploinsufficiency enhances translation and cooperates with oncogenic signaling, revealing that the control of subunit joining is itself a regulatory node with oncogenic potential [185].

The implications for hematologic oncology are substantial. If LSCs maintain a ribosome composition that is qualitatively distinct from their normal counterparts, as suggested by the differential expression of ribosomal proteins and rRNA-modifying enzymes in primitive AML populations [186], then the onco-ribosome represents a potential therapeutic target whose specificity derives from the composition of the translational machinery itself rather than from any single upstream signaling pathway. This possibility is particularly relevant in the context of LSC, which may rely on specialized ribosomes to sustain the translation of stemness-associated transcripts under conditions of metabolic and therapeutic stress, a hypothesis supported by the observation that HSCs require a precisely regulated protein synthesis rate and that even modest perturbations in ribosome composition or output can compromise stem cell function [186]. Although no onco-ribosome-specific drugs have yet reached clinical development, the identification of compounds that differentially affect modified versus unmodified ribosomes represents an active area of investigation that could fundamentally alter the therapeutic landscape.

Modifications of rRNA as regulators of hematopoiesis and leukemogenesis

The ribosomal RNAs that constitute the catalytic and structural core of the ribosome are not chemically uniform. Human rRNAs carry over 200 post-transcriptional modifications, including 106 sites of 2’-O-methylation, approximately 100 pseudouridine residues, and a smaller number of base methylations and acetylations [187, 188]. Collectively, these modifications define a “ribosome epitranscriptome” in which combinatorial patterns of chemical marks on rRNA generate functionally distinct ribosome populations [189, 190]. This layer of regulation parallels mRNA epitranscriptomics but operates directly at the level of the catalytic core of the translational machinery. Advances in high-resolution mapping techniques, including RiboMeth-seq for quantitative assessment of 2’-O-methylation; HydraPsiSeq, RBS-Seq, and Pseudo-Seq for pseudouridine profiling; and nanopore-based RNA sequencing, have enabled measurement of rRNA modification stoichiometry at single-nucleotide resolution, revealing unexpected levels of heterogeneity across cell types and disease states [191–195]. RBS-Seq has since been applied directly to human tumour tissue, revealing wide inter-patient variability in site-specific rRNA pseudouridylation in breast cancer, with dysregulated sites associated with clinicopathological features [196]. A key insight from these technologies is that modification at specific rRNA sites is not binary but fractional, with individual positions exhibiting methylation or pseudouridylation levels that vary between 20% and 100% across tissues and between normal and malignant cells [193, 197]. This fractional modification is not merely stochastic but appears to be tightly regulated, enabling cells to fine-tune translational output without altering ribosome abundance [198, 199].

The most abundant rRNA modification, 2’-O-methylation of the ribose moiety, is catalyzed by the methyltransferase FBL, which assembles into the core C/D box small nucleolar ribonucleoprotein (snoRNP) complex together with NOP56, NOP58, and the 15.5 kDa protein (SNU13), scaffolded on C/D box small nucleolar RNAs (snoRNAs) that guide site-specific modification through base-pairing complementarity with the target rRNA [187, 188]. Marcel and colleagues (2013) demonstrated that p53 directly represses FBL expression, and that p53 inactivation in cancer cells leads to FBL overexpression, altered rRNA 2’-O-methylation patterns, impaired translational fidelity, and increased IRES-dependent translation of key oncogenic mRNAs, including IGF1R and MYC [200] Fig. 3B. This finding established a direct link between the most frequently inactivated tumor suppressor and the chemical modification state of the ribosome. Structurally, 2’-O-methylation sites cluster in functionally critical regions of the ribosome, including the decoding center of the 18 S rRNA and the peptidyl transferase center of the 28 S rRNA, where they influence ribosome conformation, tRNA accommodation, and codon-anticodon interactions [187, 201]. Using cell-free in vitro translation assays, it was demonstrated that the intrinsic translational capabilities of human ribosomes are governed by their rRNA modification status, providing direct evidence that modulation of rRNA 2’-O-methylation affects translation initiation of IRES-containing mRNAs [201]. This connects rRNA chemistry to the selective translation programs discussed in “The conceptual core of the selective translation and remodeling of the tumor proteome” section, because IRES-dependent translation of mRNAs encoding anti-apoptotic proteins, growth factors, and cell-cycle regulators is precisely the class of translation most sensitive to ribosome modification state. In addition to their effects on IRES-dependent initiation, rRNA modifications have been increasingly linked to codon-dependent translation and elongation dynamics, as altered modification patterns in the decoding center can influence ribosome dwell time on specific codons, thereby modulating translation efficiency of transcripts enriched for particular codon signatures [187, 201, 202].

The clinical relevance of rRNA 2’-O-methylation heterogeneity was established by using RiboMeth-seq to map the 2’-O-methylation landscape in 195 primary human breast tumors, generating the first atlas of rRNA modification in human cancer tissue [198]. The study identified a subset of variable sites whose modification levels distinguished tumor subtypes and were associated with clinical features. Among these, 18 S-Gm1447, located in proximity to the ribosomal protein uS10 that participates in HCV IRES recognition and resolution of stalled ribosomes, displayed the highest inter-tumor variability, with some tumors completely lacking methylation at this position [198]. In AML, the connection between FBL and leukemogenesis was determined by the demonstration of FBL as a driver of AML through a CRISPR screen, demonstrating that its depletion causes cell-cycle arrest and death specifically in AML cells while sparing normal hematopoietic progenitors, with its phase-separation domains being essential for pre-rRNA processing and the subsequent translation of oncogenic proteins [30]. A PRC2-independent function for the histone methyltransferase EZH2 further illustrates how rRNA modification intersects with epigenetic regulation in cancer. EZH2 localizes to nucleoli independently of its polycomb function, where it bridges FBL and nucleolar protein 56 (NOP56), facilitating rRNA 2’-O-methylation. EZH2 deficiency impaired global translation and reduced IRES-dependent translation initiation in cancer cells, revealing that a protein canonically associated with transcriptional repression exerts a direct and separable function in ribosome modification and translational control [203].

The second major class of rRNA modification, pseudouridylation (the isomerization of uridine to pseudouridine), is catalyzed by the H/ACA box snoRNP complex, whose catalytic subunit is dyskerin. DKC1 is mutated in X-linked dyskeratosis congenita, a ribosomopathy characterized by bone marrow failure, mucocutaneous abnormalities, and increased risk of MDS and AML [204]. DKC1 deficiency impairs IRES-dependent translation of anti-apoptotic mRNAs, including those encoding p27 and XIAP, establishing a mechanistic link between pseudouridylation defects and tumor suppressor function [205] (Fig. 3B). Also, loss of DKC1 function perturbs p27 translation and contributes to pituitary tumorigenesis in mice [206]. More recent studies have expanded the functional impact of pseudouridylation beyond IRES-dependent translation to include regulation of decoding fidelity, stop codon recognition, and ribosome pausing [207, 208]. A single pseudouridine in the decoding center of the ribosome can alter stop codon recognition, converting a termination codon into a sense codon and enabling translational readthrough [207]. Loss of pseudouridylation at specific sites can increase translational errors and proteotoxic stress, thereby activating protein quality control pathways that are frequently upregulated in cancer cells and linking rRNA modification defects to both genome instability and malignant transformation [209]. Notably, DKC1 also participates in telomerase RNA (TERC) stabilization, and its dual role in ribosome function and telomere maintenance complicates the interpretation of its tumor-suppressive versus oncogenic effects in hematologic disorders, as both ribosome dysfunction and telomere attrition can independently contribute to bone marrow failure and cancer predisposition [204, 208].

A third, less characterized modification is acetylation of cytidine at position 1842 of the 18 S rRNA (ac4C1842), catalyzed by the acetyltransferase N-acetyltransferase 10 (NAT10). This modification is located within the decoding center and affects translational fidelity by modulating codon-anticodon interactions [210]. Beyond its role in rRNA acetylation, NAT10 has been implicated in DNA damage response, nucleolar integrity, and cellular senescence, and pharmacological inhibition of NAT10 with remodelin has been shown to impair ribosome biogenesis and selectively reduce cancer cell viability [32]. Although data specifically connecting NAT10 to hematologic malignancies remain limited, the positioning of its catalytic product within the decoding center suggests functional consequences for translational accuracy that are relevant to any system in which codon-level selectivity shapes the oncogenic proteome.

Small nucleolar RNAs, which guide both 2’-O-methylation and pseudouridylation to specific rRNA positions, are themselves subject to dysregulation in hematologic malignancies. The RUNX1::RUNX1T1 fusion oncoprotein requires enhanced C/D box snoRNA/snoRNP formation to induce self-renewal and leukemia, directly linking snoRNA-mediated rRNA modification to the molecular mechanism of a recurrent leukemia-associated translocation [211]. Alterations in specific snoRNAs, including SNORD14D, SNORD44, and SNORD42A, have been documented in MDS and AML, with functional consequences for the modification of rRNA positions that influence decoding fidelity and IRES-dependent translation [212]. Emerging evidence indicates that snoRNAs can exert functions beyond guiding rRNA modifications, including direct regulation of alternative splicing, mRNA stability, and translation [213]. In some contexts, individual snoRNAs have been shown to function as oncogenic drivers independent of their canonical role in rRNA modification, directly promoting proliferation and survival pathways [211, 213].

rRNA modification patterns are also sensitive to cellular metabolic state, as the methyl donor S-adenosylmethionine (SAM, required for 2’-O-methylation and base methylation) and the acetyl donor acetyl-CoA (required for NAT10-catalyzed acetylation) are derived from central metabolic pathways whose flux is altered in cancer cells [214]. This metabolic coupling suggests that the metabolic rewiring characteristic of LSCs, as discussed in “The integrated stress response and the eIF2α axis as a parallel mRNA translation program” section, may indirectly shape ribosome function through modulation of rRNA chemistry, creating an additional layer of integration between metabolism and translational control.

Using developmental models, it was demonstrated that changes in specific 2’-O-methylation sites during differentiation are functionally required for proper lineage commitment, and that perturbation of these modifications alters the translational output of the ribosome in a transcript-selective manner [199]. This finding has immediate implications for hematopoietic differentiation, where lineage commitment requires precise translational control of transcription factors with structured 5’UTRs. In LSCs, which are characterized by low translational output and high reliance on selective protein synthesis, fine-tuning of rRNA modification patterns may provide a mechanism to preserve stemness-associated translation while limiting proteotoxic stress, a hypothesis supported by the observation that HSCs require a precisely regulated protein synthesis rate and that even modest perturbations in ribosome composition or output can compromise stem cell function [186].

Modifications of tRNA and tumor codon usage

The preceding sections have established that ribosome heterogeneity at the level of protein composition (“From canonical homogeneity to functional specialization” section) and rRNA modifications (“Modifications of rRNA as regulators of hematopoiesis and leukemogenesis” section) generates translationally specialized ribosomes with distinct functional properties. A third and complementary layer of translational regulation resides at the interface between the ribosome and the mRNA it decodes, determined by the availability, modification state, and charging levels of tRNAs. The efficiency with which a given codon is translated depends not only on the abundance of its cognate tRNA but also on tRNA post-transcriptional modifications, aminoacylation levels, and codon-dependent elongation kinetics, collectively defining the decoding landscape of the cell [215, 216]. This landscape is further shaped by ribosome dwell time at specific codons and the kinetics of elongation-coupled quality control pathways [217]. Importantly, this decoding landscape is not fixed but is dynamically remodeled in cancer cells to favor the translation of transcripts with specific codon signatures, creating a selective advantage that operates at the level of elongation rather than initiation, and that is invisible to therapeutic strategies targeting eIF4F or mTOR.

The most extensively characterized tRNA modification with oncogenic relevance is the N⁷-methylguanosine (m⁷G) modification at position 46, catalyzed by the METTL1/WDR4 methyltransferase complex. Orellana and colleagues (2021) demonstrated that METTL1 is frequently amplified and overexpressed across multiple cancer types, and that its depletion causes decreased abundance of m⁷G-modified tRNAs, altered cell cycle progression, and inhibition of oncogenicity [62]. Conversely, METTL1 overexpression induced oncogenic cell transformation through a mechanism that is resolved at the level of individual codons: m⁷G modification increases the stability and abundance of a subset of tRNAs, most notably Arg-TCT-4-1, and this increased tRNA availability enhances translation of mRNAs enriched in the corresponding AGA codon, including cell-cycle regulators [62]. Strikingly, ectopic overexpression of the single tRNA Arg-TCT-4-1 was sufficient to phenocopy METTL1-driven malignant transformation in mouse embryonic fibroblasts and to accelerate leukemia progression in MOLM-13 xenograft models, establishing that a single tRNA species can function as an oncogene by altering the codon-level translational landscape [62] (Fig. 3C). Subsequent studies confirmed METTL1-mediated m⁷G modification as a driver of tumor progression and therapy resistance across lung cancer, cholangiocarcinoma, esophageal squamous cell carcinoma, and osteosarcoma, consistently demonstrating that tRNA modification-dependent codon-biased translation represents a general oncogenic mechanism [218–221]. The structural basis for this modification was resolved by two independent cryo-EM studies that revealed how the METTL1/WDR4 complex recognizes the tRNA elbow structure and catalyzes methylation, providing a framework for the rational design of inhibitors targeting this [222, 223].

Additional tRNA modifications contribute to decoding efficiency and fidelity through distinct mechanisms. The TRMT6/TRMT61A (tRNA methyltransferase 6/61A) complex adds m¹A (1-methyladenosine) at position 58 of several tRNAs, influencing tRNA tertiary structure, stability, and charging efficiency, with emerging evidence linking TRMT61A overexpression to enhanced proliferation in cancer cells [224, 225]. Modifications at the wobble position (position 34) of the anticodon, including 5-methoxycarbonylmethyl-2-thiouridine (mcm⁵s²U), play a particularly critical role in decoding efficiency for specific codon families. The wobble U34 modifications catalyzed by the Elongator complex are essential for the translation of codon-biased transcripts in melanoma, and the loss of these modifications selectively impairs translation of mRNAs enriched in the corresponding codons, including those encoding HIF1α and other stress-adaptive proteins [226]. The cytosine-5 methyltransferase NSUN2 modifies multiple tRNAs and has been shown to regulate tRNA stability and fragment generation, with NSUN2 loss leading to accumulation of tRNA-derived stress-induced fragments that inhibit global translation [227]. The functional consequences of NSUN2 dysregulation thus extend beyond individual codon decoding to encompass broader translational reprogramming through tRNA fragment biology.

The concept of codon optimality has undergone considerable refinement in recent years. The classical view held that codon optimality reflects elongation speed, with optimal codons translated more rapidly because of higher cognate tRNA availability. While this remains valid, it is now recognized that codon optimality also influences mRNA stability, ribosome density, co-translational protein folding, and nascent peptide folding trajectories, making it a central determinant of gene expression output that operates through multiple intersecting mechanisms [228, 229]. Non-optimal codons can induce ribosome pausing and, when ribosomes accumulate, ribosome collisions that activate quality control pathways, including zinc finger protein 598 (ZNF598)-mediated ubiquitylation and the ribosome-associated quality control complex, linking codon usage to proteostasis and cell survival [228, 229]. These processes are increasingly recognized as key determinants of cellular fitness in cancer, where elevated translational flux sensitizes cells to perturbations in elongation dynamics [230]. In cancer cells, this creates a dual vulnerability: the dependence on rapid translation of codon-biased oncogenic transcripts requires an optimized tRNA pool, while the high translational demand simultaneously increases the risk of ribosome collisions and proteotoxic stress if that optimization is perturbed. The systems-level demonstration that proliferation-associated and differentiation-associated genes possess distinct codon usage signatures and are decoded by correspondingly distinct tRNA pools establishes that the codon composition of transcriptional programs and the tRNA repertoire of the cell are co-evolved regulatory modules [231].

MYC occupies a central position in this regulatory axis. As discussed in “Ribosome biogenesis and the nucleolus” and “MYC as a major coordinator of ribosome biogenesis in lymphomagenesis” section, MYC globally activates RNA polymerase I to increase rRNA production and ribosome biogenesis. In parallel, MYC directly activates RNA polymerase III to increase tRNA transcription, but this activation is not uniform across all tRNA genes [28, 232]. MYC-driven upregulation of Pol III not only increases tRNA abundance and coordinates tRNA expression with the codon usage of proliferation-associated transcripts, but has also been linked to the regulation of tRNA-modifying enzymes, suggesting that MYC may coordinate not only tRNA abundance but also their functional maturation [231, 232]. This integrated regulation synchronizes the decoding machinery with oncogenic transcriptional programs and connects tRNA supply with ribosome biogenesis, reinforcing MYC’s role as a master regulator of the translational apparatus at every level, from ribosome production to initiation factor activation to the optimization of the decoding landscape itself. The demonstration that MYC-driven cancers exhibit elevated levels of specific tRNA species whose anticodons correspond to codons enriched in proliferation-associated mRNAs provided direct evidence for this co-adaptive model [233] (Fig. 3C). The implication is that therapeutic targeting of MYC’s Pol III activity, or of specific tRNA-modifying enzymes that are essential for the stability of MYC-induced tRNAs, could selectively compromise the decoding capacity required for oncogenic translation.

tRNA-derived fragments (tRFs and tiRNAs) are now recognized as active regulators of translation with functional consequences for cancer biology. These fragments, generated by specific endonucleolytic cleavage of mature or precursor tRNAs under stress conditions, can inhibit translation initiation by displacing eIF4G or eIF4A from mRNAs, modulate ribosome assembly, or selectively repress subsets of mRNAs through sequence-specific interactions [234, 235]. In addition, some tRFs associate with Argonaute proteins and function in a microRNA-like manner to repress translation of specific targets, expanding the regulatory reach of tRNA-derived species beyond direct ribosome interference [236]. Specific tRNA-derived fragments suppress breast cancer metastasis by competitively displacing the RNA-binding protein YBX1 from its pro-metastatic mRNA targets, directly connecting tRNA fragment biology to the translational regulators discussed in section [237]. In hematologic malignancies, the functional characterization of tRNA fragments remains an area of active investigation, but their generation is intimately linked to the tRNA modification state, as loss of modifications by NSUN2 or DNMT2 increases tRNA susceptibility to endonucleolytic cleavage and promotes fragment accumulation [227, 238]. Furthermore, METTL1 promotes tumorigenesis through regulation of tRNA-derived fragment biogenesis, establishing a mechanistic link between tRNA modification, fragment generation, and oncogenic translation [239].

Aminoacylation levels of tRNAs provide an additional layer of control over elongation that connects tRNA biology directly to cellular metabolism. tRNA charging is dynamically regulated by nutrient availability, amino acid biosynthesis, and metabolic flux, and uncharged tRNAs activate GCN2 to initiate the integrated stress response, closing the circuit between tRNA biology and the ISR pathway discussed in “The integrated stress response and the eIF2α axis as a parallel mRNA translation program” section [103]. This establishes tRNA charging as a real-time sensor of metabolic stress that directly feeds into translational control. Aminoacyl-tRNA synthetases (ARSs), which catalyze tRNA charging, have also emerged as regulators of cancer biology beyond their canonical enzymatic roles, as several ARSs exhibit non-canonical signaling functions and are differentially expressed in malignancies, linking amino acid sensing, translation elongation, and oncogenic signaling pathways [240].

Ribosomopathies as genetic models of mRNA translation deregulation

The ribosomopathy paradox and the paradigm of Diamond-Blackfan anemia syndrome

Ribosomopathies are a group of inherited disorders caused by loss-of-function mutations, whether heterozygous (as in Diamond-Blackfan anemia syndrome), hemizygous (as in X-linked dyskeratosis congenita), or homozygous or compound heterozygous (as in Shwachman-Diamond syndrome), in genes encoding ribosomal proteins, ribosome assembly factors, or rRNA-modifying enzymes. They present a fundamental paradox for the field of translational oncology, one that recapitulates at the molecular level a clinical riddle first posed by Dameshek in 1967: why do disorders of bone marrow hypoplasia and failure, such as aplastic anemia and the ribosomopathies, share a paradoxical predisposition to leukemic transformation with disorders of marrow hyperplasia [241]. In contemporary molecular terms, this riddle is restated as follows: these diseases are defined by ribosomal insufficiency, yet they confer a markedly elevated risk of cancer, particularly hematologic malignancies, in which translation is typically hyperactivated [242, 243]. The resolution of this paradox has reshaped the understanding of how translational perturbations drive malignant transformation, revealing that ribosomopathies should be viewed not simply as disorders of reduced ribosome function but as conditions that impose chronic translational stress, creating a selective landscape in which clones with adaptive translational rewiring and impaired quality control are positively selected [244, 245]

DBAS is the prototypical ribosomopathy, and the disease in which the mechanistic basis for this paradox has been most thoroughly investigated. Pathogenic variants, predominantly heterozygous loss-of-function alleles but also including missense variants, in more than 20 genes encoding ribosomal proteins account for approximately 75% of DBAS cases, most commonly RPS19 (eS19; approximately 25% of cases), RPL5 (uL18), RPL11 (uL5), RPS26 (eS26), RPS10 (eS10), and RPS24 (eS24) [246, 247] (Fig. 4A). The remaining cases are attributable to variants in non-ribosomal genes, including GATA1 and HEATR3, indicating that the DBAS phenotype can arise through disruption of erythroid-specific translation or ribosome assembly chaperoning independently of core ribosomal protein haploinsufficiency [248]. These mutations cause ribosomal protein haploinsufficiency, which impairs ribosome biogenesis and activates the nucleolar stress pathway described in “The RPL5/RPL11–MDM2–p53 pathway and broader nucleolar stress signaling” section: unassembled RPL5 and RPL11, together with 5 S rRNA, form the 5 S ribonucleoprotein particle (5 S RNP), which accumulates in the nucleoplasm, binds MDM2, and stabilizes p53, triggering cell-cycle arrest and apoptosis preferentially in erythroid progenitors (Fig. 4B) [249, 250]. This p53-dependent mechanism accounts for the bone marrow failure that characterizes DBAS and explains the selective vulnerability of the erythroid lineage, which demands exceptionally high rates of cell division and ribosome production during terminal differentiation [249, 250] (Fig. 4B).

Fig. 4.

From primary molecular defects to malignant transformation in ribosomopathies. A Distinct primary lesions reduce the pool of functional ribosomes: ribosomal protein haploinsufficiency (RPS19 in Diamond-Blackfan anemia; RPS14 in del(5q) MDS), impaired 60 S maturation (SBDS deficiency blocking eIF6 release, in Shwachman-Diamond syndrome), and impaired rRNA pseudouridylation (DKC1 mutations, in dyskeratosis congenita). B The resulting ribosomal stress activates the nucleolar surveillance pathway: unassembled RPL5 and RPL11, together with 5 S rRNA, form the 5 S RNP, which inhibits MDM2 and stabilizes p53, driving cell-cycle arrest and apoptosis. Reduced ribosome availability selectively impairs translation of GATA1 mRNA, whose complex 5’UTR confers heightened sensitivity to ribosome scarcity, producing erythroid-restricted bone marrow failure and cytopenias. C Clonal escape from p53-dependent surveillance, most commonly through somatic TP53 mutations or deletions, removes this checkpoint. In Shwachman-Diamond syndrome, this transforming trajectory is mechanistically distinct from an alternative, non-transforming adaptive trajectory in which compensatory EIF6 mutations restore ribosome function. TP53-mutant clones retain the altered translational selectivity imposed by the primary lesion, biasing translation toward pro-survival and proliferative programs and driving malignant transformation to AML or MDS. DKC1, dyskerin pseudouridine synthase 1; EIF6, eukaryotic translation initiation factor 6; MDM2, mouse double minute 2 homolog; MDS, myelodysplastic syndrome; RNP, ribonucleoprotein; SBDS, Shwachman-Bodian-Diamond syndrome protein

However, p53 activation alone does not fully explain the pathophysiology of DBAS. A quantitative framework has emerged in which transcripts with complex 5’UTRs exhibit higher sensitivity to reductions in ribosome availability, providing a concentration-dependent basis for the selective translational defects observed in ribosomopathies [251, 252]. Ribo-seq analysis in ribosomal protein-haploinsufficient human hematopoietic stem and progenitor cells identified a class of mRNAs whose translation is disproportionately sensitive to ribosome availability, including transcripts encoding key lineage-specifying transcription factors [251]. Among these, GATA1 emerged as a prototypical example of lineage-specific translational vulnerability; ribosomal protein haploinsufficiency reduces translation of GATA1 mRNA, likely because this transcript has a higher threshold for translation initiation than most other mRNAs, and the defective hematopoiesis in DBAS patients could be partially rescued by increasing GATA1 protein levels, establishing that the erythroid defect is driven by a selective translational deficit rather than by a global reduction in protein synthesis [252, 253] (Fig. 4B). Single-cell transcriptomic analyses have further refined this picture by demonstrating that the erythroid defects in DBAS originate in immature GATA1-negative BFU-E progenitors, with replication stress and p53 activation occurring already at this early stage, and a preferential compromise of the erythroid trajectory, while myeloid lineages are relatively preserved [254]. This lineage skewing at the level of multipotent progenitors indicates that the translational vulnerability is embedded in the earliest stages of hematopoietic hierarchy.

More recently, ribosomal protein deficiencies in DBAS were shown to cause a p53-independent loss of ATF4 by reducing the pool of actively translating ATF4 mRNAs, with this defect more pronounced in RPS deficiencies because the loss of 40 S subunits specifically destabilizes ATF4 transcripts [255]. ATF4 downregulation occurred in early hematopoietic progenitors and correlated with the severity of erythroid differentiation defects in patients, directly connecting ribosomopathy pathophysiology to the ISR axis discussed in “The integrated stress response and the eIF2α axis as a parallel mRNA translation program” section [255]. This finding suggests that ribosomal protein-haploinsufficient cells undergo chronic ISR adaptation rather than transient stress activation, resulting in a rewired stress response that becomes permissive for survival despite persistent translational defects. The concept that ribosomes generated under conditions of ribosomal protein haploinsufficiency may themselves be compositionally and functionally altered adds another dimension: as established in “From canonical homogeneity to functional specialization” section, substoichiometric variation in individual ribosomal proteins generates ribosomes with distinct translational selectivity [168, 169]. In DBAS, the obligate reduction in the haploinsufficient protein creates a population of ribosomes that either lack this component entirely or incorporate it at reduced stoichiometry, effectively constituting a form of stress-induced specialized ribosomes whose translational preferences differ from those of wild-type ribosomes [168].

The cancer predisposition in DBAS is substantial, with an approximately 5-fold increased risk of developing malignancies, including AML, MDS, and osteosarcoma [242, 256]. The transition from ribosomal insufficiency to malignancy is understood as an evolutionary process in which chronic translational stress imposes selective pressure favoring clones that escape p53-dependent growth restriction, most commonly through somatic TP53 mutations or deletions, and that simultaneously acquire compensatory mechanisms restoring translational capacity [244, 245]. Recent genomic studies indicate that individuals with ribosomopathies exhibit increased rates of clonal hematopoiesis, often involving mutations in TP53, DNMT3A, and other epigenetic regulators, suggesting that ribosomal stress creates a selective environment that accelerates clonal evolution even before overt malignancy [257, 258]. These escaped clones possess a doubly dangerous phenotype: they have lost the p53-dependent surveillance mechanism that normally constrains cells with defective ribosomes, and they retain the altered translational selectivity imposed by ribosomal protein haploinsufficiency, creating a permissive environment for oncogenic translation [244, 245] (Fig. 4C). It remains largely unexplored whether ribosomal protein haploinsufficiency cooperates with epitranscriptomic mechanisms, such as m⁶A-dependent translation, to reshape the translatome in ribosomopathies, an intersection that could reveal additional layers of translational adaptation not captured by current models. Notably, ribosomopathies may also represent an accelerated model of hematopoietic aging, in which chronic translational stress recapitulates features of age-related clonal hematopoiesis and functional decline, a parallel with implications for understanding the translational basis of both congenital and acquired bone marrow failure [259]. The mechanistic continuum from ribosomal insufficiency to malignant transformation, spanning Diamond-Blackfan anemia syndrome, del(5q) MDS, and Shwachman-Diamond syndrome, and the divergent clonal trajectories that emerge under translational stress, are depicted in Fig. 4.

Dyskeratosis congenita and the dual role of dyskerin in rRNA modification and telomere maintenance

Dyskeratosis congenita (DC) is a clinically heterogeneous ribosomopathy characterized by the classical triad of abnormal skin pigmentation, nail dystrophy, and oral leukoplakia, together with bone marrow failure, pulmonary fibrosis, and a substantially increased risk of MDS and AML [204, 260]. DC occupies a unique position among the ribosomopathies because it is caused by mutations in genes that participate in two mechanistically distinct cellular processes: (i) rRNA pseudouridylation and (ii) telomere maintenance. The most common X-linked form is caused by mutations in DKC1, encoding dyskerin, the catalytic subunit of the H/ACA box snoRNP complex that catalyzes pseudouridylation of rRNA, as discussed in “Modifications of rRNA as regulators of hematopoiesis and leukemogenesis” section [204, 261] (Fig. 4A). However, DKC1 is also an essential component of the telomerase holoenzyme, where it stabilizes the telomerase RNA component (TERC), and clinical studies have linked DKC1/TERC expression levels to patient survival independently of telomerase activity, as shown in non-small cell lung cancer [262]. Autosomal dominant and recessive forms of DC are caused by mutations in TERC, TERT, NHP2, NOP10, WRAP53, and other genes, further underscoring the dual involvement of telomere and ribosome biology [260, 263].

The dual role of DKC1 in rRNA pseudouridylation and telomerase function has historically complicated the mechanistic attribution of disease phenotypes. Recent studies suggest that translational defects and telomere dysfunction may contribute to disease pathogenesis in a context-dependent and partially independent manner, with ribosome-mediated defects predominating in early hematopoietic dysfunction and telomere attrition contributing to long-term stem cell exhaustion and genomic instability [204, 205, 208]. Functional studies have demonstrated that DKC1 deficiency impairs IRES-dependent translation of specific mRNAs, including those encoding the anti-apoptotic factors p27 and XIAP, establishing a ribosome-intrinsic translational defect that is independent of telomere length [205, 206]. This IRES-dependent defect extends to the tumour suppressor p53 itself, and is target-specific rather than uniformly repressive: whereas dyskerin loss impairs IRES-mediated translation of p27, p53, and XIAP, it paradoxically enhances IRES-mediated translation of the pro-angiogenic factor VEGF, indicating that pseudouridylation deficiency reshapes rather than simply diminishes the IRES-dependent translatome [264]. More recent work has extended the functional impact of pseudouridylation defects beyond IRES-dependent translation to include regulation of decoding fidelity and stop codon recognition, with a single pseudouridine in the decoding center capable of converting a termination codon into a sense codon and enabling translational readthrough [207, 209, 262]. These findings suggest that pseudouridylation defects selectively impair translation of transcripts involved in stress responses and genome maintenance, and may alter cellular responses to stress by modulating the translation of ISR-regulated transcripts [209]. This creates a translational environment that is permissive for the accumulation of oncogenic lesions, with the relative contributions of ribosome dysfunction and telomere attrition likely varying across disease stages and hematopoietic compartments.

Shwachman-Diamond syndrome and defective ribosome maturation

Shwachman-Diamond syndrome (SDS) is an autosomal recessive ribosomopathy characterized by exocrine pancreatic insufficiency, skeletal abnormalities, bone marrow failure, and a markedly elevated predisposition to myeloid malignancies [265, 266]. Approximately 90% of cases are caused by biallelic mutations in the SBDS gene, which encodes a highly conserved protein that functions in the final cytoplasmic step of 60 S ribosomal subunit maturation [267]. Unlike DBAS and del(5q), in which the primary defect involves the stoichiometry of ribosomal proteins within assembled subunits, SDS disrupts the functional activation of otherwise structurally assembled ribosomes, making it a mechanistically distinct ribosomopathy that illuminates a different node of translational control.

The molecular function of SBDS has been elucidated over the past two decades through a combination of yeast genetics, cryo-EM structural studies, and patient-derived cell analyses. SBDS cooperates with the GTPase EFL1 to catalyze the removal of eIF6 from the intersubunit face of nascent 60 S ribosomal subunits in the cytoplasm [35, 268]. eIF6 is an essential assembly factor that accompanies the pre-60 S subunit during nuclear export but must be evicted before the 60 S subunit can join the 40 S subunit to form a translationally competent 80 S ribosome [35, 268]. The binding site of eIF6 on the 60 S subunit overlaps with the 40 S subunit interface, explaining its function as an anti-association factor whose retention sterically blocks subunit joining [269]. Cryo-EM structures of the human SBDS-EFL1 complex bound to 60 S subunits carrying endogenous eIF6 revealed a cofactor-dependent conformational switching mechanism in which SBDS couples GTP hydrolysis by EFL1 to eIF6 displacement [36, 269]. In SDS, biallelic SBDS mutations result in reduced SBDS protein levels, impaired eIF6 eviction, accumulation of 60 S subunits that cannot participate in translation, and a consequent reduction in the pool of active 80 S ribosomes available for protein synthesis [35, 270] (Fig. 4A).

The hematologic consequences of this ribosome maturation defect parallel those of other ribosomopathies. Impaired translational capacity reduces HSPC fitness and activates a p53-dependent checkpoint that drives cellular senescence and apoptosis, clinically manifesting as bone marrow failure [271, 272]. SBDS deficiency has also been shown to preferentially affect translation re-initiation of specific mRNAs, including CCAAT/enhancer-binding protein alpha (CEBPA) and beta (CEBPB) essential for myeloid differentiation, revealing that the translational consequences of defective subunit joining exhibit transcript-selective features, with preferential impact on mRNAs encoding myeloid differentiation factors [273, 274]. Defective subunit joining leads to the accumulation of immature ribosomal subunits that cannot engage in productive translation, and this imbalance activates quality control pathways that contribute to proteotoxic stress and may amplify the fitness defect in hematopoietic progenitors [270, 275].

A landmark discovery in SDS biology has been the identification of somatic mutations in EIF6 as the most frequent clonal event in SDS hematopoiesis. Deep whole-exome sequencing of paired marrow and fibroblast samples from patients with biallelic SBDS mutations revealed that somatic EIF6 mutations are present in 61% of patients and frequently arise at an early age [276]. The majority are missense substitutions that reduce eIF6 protein stability or impair its binding to the 60 S subunit, functionally mimicking partial eIF6 loss [276]. Introduction of the recurrent EIF6^N106S mutation into SBDS-deficient human CD34 + cells improved myeloid and erythroid colony formation and enhanced subunit joining, confirming that these mutations are compensatory and restore translational capacity [276]. This represents a remarkable example of somatic genetic adaptation to a germline ribosome assembly defect, in which the hematopoietic system selects for mutations that directly alleviate the molecular lesion. Longitudinal sequencing studies have established SDS as a paradigmatic model of genetically constrained clonal hematopoiesis, in which a limited repertoire of somatic mutations, primarily in EIF6 and TP53, emerges under strong selective pressure imposed by defective ribosome maturation [276, 277]. This restricted evolutionary landscape contrasts with the broader mutational diversity observed in age-related clonal hematopoiesis and highlights the deterministic role of translational stress in shaping clonal trajectories.

Critically, the compensatory EIF6 mutations and the transforming TP53 mutations define distinct and largely non-overlapping clonal trajectories [276, 277]. EIF6-mutant clones represent a pre-malignant but adaptive pathway that restores ribosome function and improves cellular fitness without promoting leukemic transformation, while TP53-mutant clones represent a transforming pathway that removes the p53 checkpoint and predisposes to MDS and AML [276, 277] (Fig. 4C). These findings reveal a fundamental evolutionary trade-off: mutations that restore translational capacity improve cellular fitness but maintain regulatory integrity, whereas mutations that abrogate checkpoint control permit unchecked proliferation at the cost of genomic stability. Biallelic TP53 alterations, which are strongly associated with complex karyotype and dismal prognosis, are found in 45% of SDS patients and represent the principal driver of leukemic transformation [276, 277]. The coexistence of compensatory and transforming clones within the same patient creates a complex clonal architecture in which the therapeutic challenge is to preserve or expand the adaptive EIF6-mutant clones while eliminating or preventing the emergence of TP53-mutant clones. The dependence of SBDS-deficient cells on residual ribosome function and stress adaptation pathways suggests potential therapeutic vulnerabilities, including heightened sensitivity to translation inhibitors, proteostasis disruptors, or ISR-modulating agents that could selectively target the transformed clone while sparing the compensatory compartment [272, 277]. Recent studies have expanded the genetic spectrum of SDS-like syndromes by identifying biallelic mutations in EFL1, DNAJC21 (DNAJ heat shock protein family member C21), and SRP54 (signal recognition particle 54) as causes of phenotypically overlapping disorders, reinforcing the concept that defects in late stages of ribosome maturation and subunit joining represent a convergent pathogenic mechanism with shared hematologic and oncologic consequences [278–280].

The del(5q) syndrome and other acquired ribosomopathies

The interstitial deletion of the long arm of chromosome 5, del(5q), is the most common cytogenetic abnormality in MDS and defines a distinct clinicopathological entity characterized by macrocytic anemia, hypolobulated megakaryocytes, thrombocytosis, and a relatively favorable prognosis when present as the sole cytogenetic abnormality [281, 282]. The molecular basis of the 5q- syndrome was resolved through an RNA interference screen targeting each gene within the commonly deleted region, which identified RPS14 (uS11), encoding the 40 S ribosomal protein S14, as the gene whose haploinsufficiency recapitulates the erythroid differentiation defect and the megakaryocytic dysplasia characteristic of the disease [283] (Fig. 4A). This discovery established del(5q) MDS as an acquired ribosomopathy and connected it mechanistically to the congenital ribosomopathies.

The pathophysiology of the 5q- syndrome converges on p53-dependent erythroid failure, paralleling DBAS. Haploinsufficiency of RPS14 activates p53 through the nucleolar stress pathway, and genetic inactivation of p53 rescues the erythroid defect in murine models of 5q- syndrome [249]. Independently, a p53-dependent mechanism was confirmed to underlie the macrocytic anemia in 5q- mice, with the critical observation that p53 loss restores erythropoiesis while simultaneously accelerating leukemic transformation, providing direct evidence for the paradox in which removal of the p53 checkpoint resolves the cytopenia but promotes malignancy [284]. Integrated translatome and proteome analyses in RPS14-deficient cells subsequently identified specific rules of translation selectivity that govern which transcripts are most sensitive to ribosomal protein haploinsufficiency, including transcripts with highly structured 5’UTRs and specific codon usage patterns [285]. These data reinforce the principle, established in DBAS, that ribosomal protein haploinsufficiency does not produce a uniform reduction in translation but instead reshapes the translational landscape in a transcript-selective manner.

The standard-of-care therapy for lower-risk del(5q) MDS is lenalidomide, whose mechanism of action illustrates the complexity of translational regulation in this disease. Lenalidomide induces selective degradation of casein kinase 1α (CK1α) through the cereblon (CRBN) E3 ubiquitin ligase complex [286]. CK1α is encoded by CSNK1A1, which maps within the commonly deleted region on 5q and is therefore haploinsufficient in del(5q) cells. The additional pharmacological reduction of CK1α protein by lenalidomide-induced degradation selectively kills the del(5q) clone by activating p53 beyond the threshold tolerable by cells with monoallelic CSNK1A1 expression, creating a synthetic lethal interaction between the chromosomal deletion and the drug [286, 287]. However, this therapeutic mechanism carries an inherent risk: because lenalidomide’s efficacy depends on p53 activation, it simultaneously exerts selective pressure that favors expansion of TP53-mutant subclones that are resistant to both the drug and the p53-dependent growth restraint [257, 288]. Longitudinal sequencing studies have demonstrated that TP53-mutant subclones in del(5q) MDS often pre-exist at low variant allele frequencies and undergo selective expansion under both intrinsic ribosomal stress and lenalidomide therapy, with their detection predicting disease progression to AML and resistance to treatment [288]. TP53 mutations are strongly associated with transformation to complex karyotype and dismal prognosis, establishing TP53 mutational status as the single most important prognostic variable in del(5q) MDS [257, 288].

The concept of ribosomopathy has traditionally been restricted to inherited disorders, but accumulating evidence indicates that ribosomopathy-like states can arise during leukemogenesis through somatic mechanisms. Somatic mutations in ribosomal protein genes, including RPL5, RPL10 (uL16), RPL22 (eL22), and RPS15 (uS19), have been identified through exome sequencing of T-cell acute lymphoblastic leukemia, chronic lymphocytic leukemia, and other hematologic malignancies [160, 289, 290]. A systematic pan-cancer screen of somatic mutations and copy-number alterations across 81 ribosomal protein genes in more than 16 cancer types extended this observation beyond hematologic malignancies, identifying recurrent heterozygous deletion or mutation of RPL5, RPL11, RPL23A, RPS5, RPS20, and RPSA in a substantial fraction of solid tumours, including glioblastoma, melanoma, and breast cancer, and establishing RPL5 haploinsufficiency as a tumour suppressor mechanism operating broadly across cancer types [291]. The RPL10-R98S mutation in T-ALL, discussed in “From canonical homogeneity to functional specialization”, represents the most thoroughly characterized example; rather than causing haploinsufficiency, it alters ribosome function qualitatively, enhancing IRES-dependent translation and JAK-STAT signaling [174, 292]. These somatic alterations do not merely impair ribosome function but generate qualitatively altered ribosomes with distinct translational preferences, reinforcing the concept that oncogenesis can be driven by ribosome specialization rather than simple hyperactivation of translation. Somatic mutations in RPL22 are recurrent in T-ALL and promote leukemogenesis through activation of the NF-κB and Notch pathways, with the translational consequences of RPL22 loss extending to selective impairment of p53 induction [289]. In CLL, recurrent mutations in RPS15 impair p53-dependent nucleolar stress surveillance, allowing CLL cells to evade the ribosome quality control checkpoint that would normally constrain cells with defective ribosomes [290]. Structural and translatome analyses of an isogenic cell line panel modeling the most recurrent ribosomal protein mutations in cancer showed that, in contrast to ribosomal protein haploinsufficiency, RP point mutations produce disproportionate translational rewiring: cryo-EM structures of RPS15-mutant ribosomes revealed destabilization of the C-terminal RPS15 domain and deregulated accommodation of aminoacylated tRNAs at the ribosomal A-site, generating a codon-specific elongation defect (“Modifications of tRNA and tumor codon usage” section) that selectively reduces translation of transcripts enriched for the affected codons, including the transcriptional regulator RUNX3 [293]. These acquired ribosomal protein mutations do not produce the classical ribosomopathy phenotype of bone marrow failure but instead generate a state of qualitatively altered translation that favors oncogenic programs, consistent with the onco-ribosome concept.

Beyond point mutations, dysregulation of ribosome biogenesis pathways through transcriptional or epigenetic mechanisms can generate functional ribosomopathy-like states in the absence of genetic lesions in ribosomal protein genes. MYC-driven overactivation of rRNA synthesis, discussed in “MYC as a major coordinator of ribosome biogenesis in lymphomagenesis” section, generates quantitative ribosome excess, but the stoichiometric imbalances that arise when ribosome biogenesis is forced beyond its normal regulatory constraints may simultaneously produce qualitatively altered ribosomes [84]. Similarly, NPM1 mutations in AML disrupt nucleolar architecture and ribosome assembly, potentially generating ribosomes with altered composition and function [77]. The concept of acquired ribosomopathies thus extends the spectrum from inherited germline disorders to somatically acquired translational defects that contribute to leukemogenesis through mechanisms that are conceptually continuous with those operating in the classical ribosomopathies.

Ribosomopathies, both inherited and acquired, represent constrained evolutionary systems in which a defined primary defect in the translational machinery restricts the spectrum of adaptive mutations, thereby providing a uniquely tractable model for understanding how translational stress shapes clonal evolution and cancer predisposition. Together, these disorders demonstrate that perturbations of the translational apparatus do not merely reduce protein synthesis but reshape its selectivity, generating a proteomic landscape that can either impair tissue homeostasis or promote malignant transformation depending on the cellular context and evolutionary trajectory.

Resolution of the ribosomopathy paradox and therapeutic implications

The ribosomopathies discussed in the preceding sections share a common paradox: germline or acquired defects that reduce ribosome quantity or quality predispose to malignancies in which translational output is characteristically elevated. The resolution of this paradox requires abandoning the simplistic model in which ribosome insufficiency and oncogenic translation are viewed as opposite ends of a single axis. Instead, a multistep evolutionary framework has emerged in which chronic translational imbalance imposes selective pressure that drives clonal evolution through a series of adaptive and maladaptive genetic events, ultimately generating cells whose translational apparatus is not merely restored but qualitatively reprogrammed [244, 245, 252]. As elaborated below, this reprogramming reflects a change not only in ribosome quantity but in ribosome quality, a distinction that runs through every step of this framework [209, 245].

The first step in this evolutionary trajectory is the imposition of translational stress by the primary ribosomopathy lesion. Whether the defect involves ribosomal protein haploinsufficiency (DBAS, del(5q)), impaired subunit maturation (SDS), or defective rRNA modification (DC), the immediate consequence is a reduction in the pool of functional ribosomes available for translation. This reduction is not merely quantitative: as established in “From canonical homogeneity to functional specialization” section, substoichiometric variation in individual ribosomal proteins generates ribosomes with measurably different translational selectivity, and ribosomes produced under conditions of haploinsufficiency or assembly defects may themselves be compositionally and functionally altered, effectively constituting stress-induced specialized ribosomes with distinct translational preferences [168, 169]. In ribosomopathies, the obligate reduction in the affected protein creates a ribosome population whose translational output may be biased toward or against specific classes of mRNAs, including those encoding oncoproteins, tumor suppressors, and differentiation factors [168, 251]. Consistent with this compositional remodeling, this reduction does not affect all transcripts equally but preferentially compromises translation of mRNAs with high initiation thresholds, complex 5’UTR structures, or specific codon compositions, generating a selectively altered proteome that impairs lineage commitment and differentiation while preserving basal cellular functions [205, 251, 285]. The resulting cytopenias, particularly the erythroid failure that is a hallmark of most ribosomopathies, reflect the vulnerability of hematopoietic lineages that demand exceptionally high rates of ribosome production and protein synthesis during terminal differentiation. While translational rewiring can buffer the effects of ribosome insufficiency, it simultaneously creates dependencies on specific translational programs that may be therapeutically targetable.

The second step is the activation of quality control mechanisms, most notably the p53-dependent nucleolar stress pathway. The 5 S ribonucleoprotein particle (5 S RNP), formed by unassembled RPL5 and RPL11 together with 5 S rRNA, binds MDM2 and stabilizes p53, triggering cell-cycle arrest and apoptosis in cells with defective ribosome biogenesis (Fig. 4B) [249, 250]. This checkpoint functions as a tumor suppressor by eliminating cells with aberrant translational capacity, but it simultaneously constrains hematopoietic output and is the proximate cause of the bone marrow failure that characterizes the clinical phenotype. The ISR likely plays a complementary role in this phase, as evidenced by the p53-independent loss of ATF4 in DBAS and the chronic ISR adaptation that enables survival under conditions of persistent translational insufficiency [111, 255]. Together, p53 activation and ISR engagement define a state of constrained hematopoiesis in which the translational defect is partially compensated but the proliferative potential of stem and progenitor cells is markedly restricted. Translational rewiring during this phase is constrained not only by ribosome availability but also by the metabolic capacity to sustain decoding fidelity, as the cofactors required for rRNA and tRNA modifications, including S-adenosylmethionine and acetyl-CoA, are derived from metabolic pathways whose flux is itself altered under conditions of hematopoietic stress [128, 214].

The third step, and the critical transition toward malignancy, is the clonal escape from these quality control mechanisms. The genetic pathways of escape are remarkably convergent across the ribosomopathies. TP53 mutations or deletions are the most frequent transforming events in DBAS, del(5q) MDS, and SDS, and their acquisition removes the checkpoint that constrains cells with defective ribosomes [242, 245, 276, 294]. In SDS, longitudinal genomic studies have revealed that TP53 mutations often arise early and expand gradually under the selective pressure of decoding stress and, in del(5q), under additional therapeutic pressure from lenalidomide [276, 284, 288]. The resulting TP53-mutant clones possess a doubly dangerous phenotype: they have lost the surveillance mechanism that would normally eliminate cells with translational defects, and they retain the selectively altered translational landscape imposed by the primary ribosomopathy lesion. This combination creates a cellular state that is simultaneously permissive for the accumulation of oncogenic lesions and biased toward the translation of survival and proliferation programs (Fig. 4C). The compensatory pathway identified in SDS, in which somatic EIF6 mutations restore ribosome function without disabling quality control, provides a striking counterpoint and illuminates the evolutionary logic of the system [272, 277].

These insights point toward several emerging therapeutic strategies that span multiple nodes of the translational apparatus. RNA polymerase I inhibitors such as CX-5461 directly target ribosome biogenesis and have entered clinical trials in hematologic malignancies, with their efficacy potentially modulated by p53 status, given the dependence of the nucleolar stress response on intact p53 signaling [21]. Homoharringtonine (omacetaxine), a translation elongation inhibitor already approved for CML, demonstrates that pharmacological targeting of the translational machinery is clinically feasible in hematologic oncology and may have particular relevance in contexts where elongation dynamics are altered by ribosome composition or tRNA pool changes [295]. ISR modulators, including ISRIB and the GCN2 activator HC-7366, could alter the balance between adaptive and maladaptive stress responses in ribosomopathy-affected cells, potentially shifting the evolutionary trajectory away from TP53-dependent escape [130, 131]. The SDS paradigm suggests that promoting compensatory mutations through gene therapy or pharmacological mimicry, for example, by reducing eIF6 expression or activity in SBDS-deficient cells, could restore ribosome function and reduce leukemic risk without the dangers of checkpoint abrogation [272, 296]. A complementary pharmacological strategy, applicable to the subset of SDS patients harboring SBDS nonsense mutations, is translational readthrough induction: ataluren, a readthrough-inducing drug already approved for Duchenne muscular dystrophy, restores full-length SBDS protein synthesis, improves 80 S ribosome assembly, and reduces p53 levels in patient-derived cells, and has recently shown clinical benefit in a compassionate-use case series, improving both hematopoietic and exocrine pancreatic manifestations in SDS patients [297]. Emerging inhibitors of tRNA-modifying enzymes, including METTL1, could exploit the altered decoding landscape discussed in “Modifications of tRNA and tumor codon usage” section, and the identification of specific translational vulnerabilities in ribosomopathy-derived malignancies, including dependence on IRES-mediated translation or selective epitranscriptomic programs, opens additional avenues for targeted intervention [62].

mRNA translation reprogramming by hematologic entity

The preceding sections have dissected translational control along mechanistic axes, from initiation and the ISR to ribosome heterogeneity and ribosomopathies. In clinical practice, however, these mechanisms do not operate in isolation but converge within specific disease contexts, where their relative contributions, interactions, and therapeutic implications are shaped by the genetic architecture, differentiation state, and microenvironmental dependencies of each malignancy. This section reorganizes the translational landscape by disease entity, demonstrating how the principles established in “Ribosome biogenesis as an oncogenic program and the nucleolus as a central vulnerability” through “Ribosomopathies as genetic models of mRNA translation deregulation” section are deployed in AML, ALL, aggressive lymphomas, and multiple myeloma.

Acute myeloid leukemia

AML is a hematologic malignancy in which the convergence of translational control mechanisms is most extensive and in which the translational dependency of LSC has been most rigorously characterized. The disease is defined by a heterogeneous genetic landscape in which driver mutations in NPM1, FLT3, DNMT3A, IDH1/2, and RAS pathway components converge upon the translational apparatus through multiple intersecting routes, creating a state of translational dependency that is qualitatively distinct from that of normal hematopoietic cells [298, 299].

The central observation that frames translational control in AML is the paradox of LSC protein synthesis. HSCs and LSCs share a fundamental property: they maintain a lower global rate of protein synthesis than their downstream progeny, yet they depend critically on the selective translation of a restricted set of stemness-associated transcripts for self-renewal and survival [111, 186]. This is not merely a state of reduced translational output but an actively maintained and evolutionarily selected decoding landscape, a low-output, high-selectivity, stress-adapted translational state that is sustained by the constitutive activity of the ISR. eIF2 ternary complex scarcity and elevated ATF4 expression mark functionally validated LSCs in primary patient samples [111]. The implication is that LSCs do not tolerate perturbations in translational selectivity, even when their global translational output is already reduced, creating a therapeutic window that does not exist in bulk leukemia cells with higher translational rates. This translational identity is further reinforced by the dependence of LSCs on mitochondrial oxidative phosphorylation, which requires selective translation of nuclear-encoded mitochondrial respiratory proteins through the RREB1-dependent system [123]. The intimate coupling between mitochondrial translation, OXPHOS, and ISR adaptation creates a metabolic-translational circuit that sustains LSC fitness and contributes to therapy resistance.

This therapeutic window is accessible through multiple pharmacological nodes. At the level of translation initiation, eIF4A is highly expressed in primitive AML cell populations compared to their normal counterparts, and its inhibition by zotatifin reduces translation of cell-cycle and survival transcripts, including AKT, STAT5, and MCL-1 [22, 96]. MCL-1 is particularly central to AML translational biology because it is a short-lived protein whose steady-state levels are acutely dependent on ongoing translation, making it exquisitely sensitive to perturbations in eIF4A-mediated initiation [22, 300]. The synergy between zotatifin and venetoclax exploits this dependency with mechanistic precision. Venetoclax neutralizes BCL-2, while zotatifin suppresses MCL-1 translation, creating a synthetic lethal condition because LSCs co-express both anti-apoptotic proteins as a survival strategy [22, 96] (Fig. 5B). In patient-derived xenograft models of relapsed and refractory AML, this combination significantly reduced tumor burden and prolonged survival while sparing healthy hematopoietic cells [22]. The ISR provides a complementary axis, in which the GCN2 activator HC-7366 potently activates the ISR in AML cells at nanomolar concentrations, inducing ATF4 targets ASNS and PSAT1 and reducing cell viability through a mechanism that simultaneously overcomes multiple venetoclax resistance pathways, including reduction of S100A8/A9, increased PUMA expression, and suppression of mitochondrial respiration (“The integrated stress response and the eIF2α axis as a parallel mRNA translation program” section) [130].

The epitranscriptomic layer adds further specificity to the AML proteome. METTL3-mediated m⁶A deposition at oncogenic transcripts, including MYC, BCL2, and PTEN, maintains their translational output in AML, and METTL3 depletion induces differentiation and apoptosis in leukemia cells while delaying leukemogenesis in vivo [53, 145]. The m⁶A readers IGF2BP1/2/3 and the RNA-binding protein YBX1 cooperate to stabilize and enhance translation of m⁶A-modified transcripts in LSCs, with YBX1 specifically recruiting MYC mRNA to polysomes without affecting its transcription [136, 137]. MSI2, through its direct binding to HOXA9, MYC, and IKZF2 mRNAs, sustains the KMT2A self-renewal program at the translational level and has recently been linked to metabolic adaptation through stabilization of DEPTOR mRNA, connecting MSI2-dependent translation to mTOR regulation and amino acid metabolism in LSCs [153, 156]. At the level of elongation, the METTL1/WDR4-mediated m⁷G modification of tRNAs and its impact on codon-biased translation of cell-cycle regulators provides an additional decoding-level mechanism through which AML cells may optimize their translational output architecture, as demonstrated by the acceleration of leukemia progression upon overexpression of a single m⁷G-modified tRNA (Arg-TCT-4-1) in MOLM-13 xenograft models [62]. These parallel epitranscriptomic, RBP-mediated, and decoding-level programs operate in concert with the eIF4F and ISR axes to define the LSC translatome, and their pharmacological targeting by Ro 08-2750 (MSI2 inhibitor), zotatifin (eIF4A inhibitor), and HC-7366 (GCN2 activator) represents orthogonal therapeutic strategies that converge on the same biological outcome: disruption of the selective translational programs that sustain LSC identity.

Ribosome biogenesis itself constitutes an independent vulnerability in AML. RNA Pol I inhibition by CX-5461 extends survival of KMT2A-rearranged AML models with concurrent p53 activation and cell-cycle arrest [65]. The recent identification of FBL as a driver of AML leukemogenesis through phase-separation-dependent regulation of pre-rRNA processing and MYC translation opens an additional node within ribosome biogenesis that may be pharmacologically accessible [30]. NPM1 mutations, present in approximately 30% of adult AML, displace the nucleolar chaperone to the cytoplasm and disrupt ribosome assembly, although the translational consequences of NPM1c+ appear to operate primarily at the epigenetic and transcriptional level rather than through direct modulation of translational efficiency [77, 78].

The resistance landscape in AML further underscores the centrality of translational control and reveals that resistance emerges not from restoration of normal translation but from adaptive rewiring of the decoding landscape. As established in the context of the ISR, venetoclax resistance in LSCs is sustained by ISR activation downstream of fatty acid oxidation-driven mitochondrial stress, a survival mechanism that, in this context, operates not merely as a stress response but as a hardwired feature of the resistant LSC state, tightly coupled to the metabolic dependencies that define it [112, 114]. Nicotinamide metabolism mediates resistance in relapsed AML LSCs through mechanisms connected to mitochondrial fitness and oxidative phosphorylation [114]. This translational plasticity as a resistance mechanism is further exemplified by the ATF4-dependent transcriptional programs that modulate immune interactions, with the ATF4-LCN2 axis recently identified as a cell-extrinsic mediator of immune evasion that promotes T cell exclusion through recruitment of immunosuppressive macrophages [116, 117]. These observations suggest that translational reprogramming in AML extends beyond cell-intrinsic survival to reshape the immune microenvironment, connecting the ISR to immunotherapy resistance.

Acute lymphoblastic leukemia

ALL presents a translational landscape that differs fundamentally from AML in two respects: the oncogenic drivers that feed into the translational apparatus are distinct, and the disease offers unique examples of how somatic ribosomal protein mutations and amino acid metabolism intersect with translational control. T-ALL and B-ALL will be discussed in sequence, as they engage different translational vulnerabilities despite sharing the common feature of PI3K/AKT/mTOR pathway hyperactivation [301, 302].

In T-ALL, the NOTCH1-MYC signaling axis constitutes the dominant route through which oncogenic signaling converges on the translational machinery. Activating mutations in NOTCH1 are present in more than 60% of T-ALL cases and result in ligand-independent, sustained NOTCH1 receptor signaling [303, 304]. NOTCH1 activation positively regulates mTORC1 activity, as demonstrated by the observation that pharmacological inhibition of NOTCH1 by gamma-secretase inhibitors (GSIs) recapitulates the phosphorylation changes induced by rapamycin on 4E-BP1, S6K, and S6 ribosomal protein, establishing a direct connection between the most frequently mutated pathway in T-ALL and cap-dependent translation initiation [305]. The mechanism involves both direct and indirect routes: NOTCH1 transcriptionally upregulates MYC, which in turn activates mTORC1 and drives ribosome biogenesis through the tri-polymerase program described in “MYC as a major coordinator of ribosome biogenesis in lymphomagenesis” section, while NOTCH1 also upregulates IL7Rα and other surface receptors that feed into PI3K/AKT signaling [304–306]. Enforced expression of MYC can fully rescue mTOR effectors from NOTCH withdrawal in a subset of T-ALL cell lines, implicating MYC as the critical intermediary that connects NOTCH to translational control [305]. This NOTCH1-MYC-mTOR circuit renders T-ALL cells particularly dependent on protein translation for growth, proliferation, and survival, and the eIF4A inhibitor silvestrol has been shown to preferentially suppress translation of MYC and other oncogenic transcripts with complex 5’UTRs in T-ALL, consistent with the selectivity principle discussed in “eIF4A, eIF4F, and the oncogenic translatome” section [306, 307].

T-ALL also provides the most compelling examples of acquired ribosomopathy in leukemia. The RPL10-R98S mutation, present in approximately 8% of pediatric T-ALL cases, alters ribosome function qualitatively rather than quantitatively, enhancing IRES-dependent translation, increasing BCL-2 protein levels, and activating JAK-STAT signaling (“From canonical homogeneity to functional specialization” section) [160, 174]. DDX3X loss-of-function mutations, recurrent in T-ALL and Burkitt lymphoma, alter the translation of differentiation and cell-cycle programs through a mechanism that is specifically translational rather than transcriptional or splicing-mediated (“Non-canonical regulators with direct translational functions” section) [159, 161]. These somatic ribosomal and helicase mutations coexist with NOTCH1-driven translational hyperactivation, creating a multilayered translational deregulation in which both the quantity and the selectivity of protein synthesis are simultaneously altered. The dual inhibition of mTOR and MNK1/eIF4E signaling has emerged as a rational combinatorial strategy, as mTOR inhibition by rapalogs paradoxically activates MNK1-dependent eIF4E phosphorylation, and concurrent MNK1 blockade overcomes this compensatory mechanism to achieve synergistic killing of T-ALL cells [308].

In B-ALL, the translational landscape is shaped by different genetic drivers. Ph-positive B-ALL, driven by BCR::ABL, constitutively activates the PI3K/AKT/mTOR axis, and the eIF4F complex inhibitor 4EGI-1 preferentially inhibits the growth of Philadelphia chromosome-transformed cells compared to non-transformed counterparts [309, 310]. Ph-like B-ALL, which carries alternative kinase-activating lesions (cytokine receptor-like factor 2 (CRLF2) rearrangements, JAK mutations, ABL-class fusions), similarly converges on mTORC1 activation. Importantly, mTORC1 activity is largely maintained in Ph-like B-ALL even when oncogenic tyrosine kinase signaling is pharmacologically suppressed, because nutrient and cytokine inputs sustain mTORC1 independently of the primary oncogenic driver [311]. This residual mTORC1 activity preserves cap-dependent translation of survival transcripts and limits the cytotoxicity of tyrosine kinase inhibitors, providing a mechanistic rationale for combining TKIs with selective mTORC1 inhibitors. Phosphorylation of 4E-BP1, the gatekeeper of eIF4F assembly, correlates with poor prognosis in B-ALL, directly linking the translational output of the eIF4F complex to clinical outcome [311, 312]. Second-generation ATP-competitive mTOR kinase inhibitors achieve more complete suppression of 4E-BP1 phosphorylation than rapalogs and demonstrate superior preclinical activity in B-ALL, consistent with the principle that effective translational inhibition requires blocking the 4E-BP1/eIF4E axis rather than merely suppressing S6K [311, 312].

The amino acid metabolism axis provides a translational vulnerability in ALL that is clinically exploited and mechanistically connected to the ISR. L-asparaginase, a cornerstone of ALL therapy for over five decades, depletes extracellular asparagine and, to a lesser extent, glutamine, triggering amino acid deprivation that activates GCN2, phosphorylates eIF2α, and globally suppresses translation [313]. However, ALL cells can acquire resistance by upregulating asparagine synthetase (ASNS) through an ATF4-dependent mechanism, directly connecting L-asparaginase resistance to the ISR axis [129]. In T-ALL, both ATF4 and zinc finger and BTB domain-containing protein 1 (ZBTB1)are recruited to the ASNS promoter to sustain asparagine biosynthesis under therapeutic pressure [129]. This pharmacological-molecular circuit, in which the therapeutic agent activates a stress response that the tumor cell exploits for adaptive survival, exemplifies the dual nature of the ISR in leukemia and suggests that combining L-asparaginase with ISR modulators (either GCN2 activators to push beyond the adaptive threshold, or ATF4 inhibitors to block the compensatory response) could enhance therapeutic efficacy.

A distinctive feature of ALL biology relative to AML is the importance of glucocorticoid sensitivity, which intersects with translational control at multiple levels. Glucocorticoids transcriptionally downregulate MYC in T-ALL, reducing ribosome biogenesis and global translational capacity [304, 306]. The combination of gamma-secretase inhibitors with glucocorticoids is synergistic precisely because both agents converge on MYC suppression through complementary routes, establishing the MYC-translation axis as the critical convergence point [306, 314]. Glucocorticoid resistance has also been associated with sustained mTORC1 activity and preserved eIF4F-dependent translation, suggesting that pharmacological disruption of cap-dependent translation could restore glucocorticoid sensitivity in resistant disease [301, 314].

Aggressive B-cell lymphomas and mantle cell lymphoma

Diffuse large B-cell lymphoma and mantle cell lymphoma (MCL) occupy a distinctive position in the translational landscape of hematologic malignancies because their defining oncogenic lesions directly implicate the translational apparatus. In both diseases, B-cell receptor (BCR) signaling functions as a master upstream regulator of the translational machinery, simultaneously activating mTORC1, sustaining eIF4F assembly, and driving MYC expression, thereby positioning eIF4A as the convergence node through which oncogenic signaling is converted into translational output [95, 315, 316]. In DLBCL, MYC rearrangement or overexpression drives ribosome biogenesis and global translational capacity through the mechanisms described in “MYC as a major coordinator of ribosome biogenesis in lymphomagenesis” section, while constitutive activation of the PI3K/AKT/mTOR pathway, occurring through PTEN loss, PIK3CA mutations, or tonic BCR signaling, sustains eIF4F-dependent cap-dependent translation of oncogenic transcripts [315, 316]. In MCL, the defining t(11;14)(q13;q32) translocation places CCND1 under immunoglobulin promoter control, and cyclin D1 mRNA possesses a highly structured 5’UTR that renders its translation acutely dependent on eIF4A helicase activity [317, 318]. These molecular features create a direct mechanistic link between the primary genetic lesion and dependence on translation initiation.

The eIF4A inhibitor zotatifin exhibits potent antiproliferative activity across a panel of DLBCL (both ABC and GCB subtypes) and Burkitt lymphoma cell lines, with a mean GI50 of 9 nM [94]. At the molecular level, zotatifin treatment induces coordinated suppression of MYC, cyclin D1/D3, MCL-1, and BCL6 protein levels in lymphoma cells, consistent with the selective inhibition of mRNAs whose 5’UTRs contain polypurine and G-quadruplex recognition motifs [94]. In vivo, zotatifin achieves significant tumor growth inhibition in DLBCL and Burkitt lymphoma xenograft models at doses of 1 mg/kg/week or less, with a favorable tolerability profile [94]. Sensitivity correlates with activation of PI3K/AKT/mTOR signaling, and PTEN loss was associated with diminished zotatifin efficacy in vivo, possibly reflecting reduced dependence on eIF4A-mediated unwinding when eIF4E availability is constitutively elevated through derepressed mTORC1 signaling, although alternative mechanisms, including upstream pathway rewiring, cannot be excluded [94]. Novel eIF4A inhibitors with non-rocaglate scaffolds have demonstrated picomolar biochemical potency and robust cellular activity in DLBCL, indicating that the pharmacological space for eIF4A targeting extends beyond the rocaglate backbone [319].

In CLL, BCR stimulation induces eIF4A expression while simultaneously reducing PDCD4 (a natural eIF4A inhibitor), establishing eIF4A as a critical translational nexus downstream of BCR signaling [95]. Treatment with eIF4A inhibitors blocks anti-IgM-induced global mRNA translation and specifically prevents accumulation of MYC and MCL-1 without affecting upstream signaling events such as ERK1/2 and AKT phosphorylation [95]. This dissociation between intact signaling and suppressed translational output is pharmacologically significant because it demonstrates that translational inhibition can circumvent the compensatory pathway reactivation that limits kinase inhibitor efficacy. This principle has direct implications for resistance to BTK inhibitors such as ibrutinib, where disease progression is frequently associated with maintenance of MYC and MCL-1 translation through alternative signaling inputs that bypass BTK but converge on the same eIF4F-dependent translational programs [320, 321]. Translation inhibition may therefore represent a strategy to overcome BTK inhibitor resistance by targeting the downstream translational output rather than the increasingly diverse upstream signaling routes that sustain it.

In DLBCL, the ISR provides an additional translational axis with disease-specific relevance. ATF4 translational activation through mitochondrial anaplerotic pathways is essential for DLBCL growth and survival, with SIRT3 functioning as a key upstream regulator of TCA cycle flux that sustains amino acid-dependent ATF4 induction independently of ER stress [124]. This metabolic-translational circuit suggests that DLBCL cells with high mitochondrial metabolism may be particularly dependent on ATF4 for maintenance of amino acid homeostasis, creating a vulnerability that could be exploited by ISR modulators or metabolic inhibitors that disrupt the TCA cycle-ATF4 connection.

Double-hit lymphomas (DHL), which carry concurrent MYC and BCL2 rearrangements, represent perhaps the most extreme example of translational dependency in lymphoid malignancies. MYC directly drives ribosome biogenesis and translational capacity, while BCL-2 provides the anti-apoptotic shield that permits cells to tolerate the proteotoxic consequences of this elevated translational output [322]. The combination of eIF4F disruption with BCL-2 inhibition by venetoclax mirrors the logic demonstrated in AML: disruption of the eIF4G-eIF4E interaction with SBI-756 sensitizes B-cell lymphomas to venetoclax, with the combination inducing apoptosis more effectively than either agent alone [323]. In MCL specifically, silvestrol-mediated eIF4A inhibition produces rapid depletion of cyclin D1, MCL-1, and cyclin E, whose short protein half-lives make them acutely dependent on ongoing translation, with concurrent activation of caspase-dependent apoptosis [317, 324]. The translational vulnerability of MCL thus resides in the structural dependence of its defining oncoprotein on eIF4A-mediated 5’UTR unwinding.

MYC-driven Burkitt lymphoma provides the paradigm in which the integration of ribosome biogenesis and selective translation was first demonstrated to be pharmacologically exploitable. Haploinsufficiency of RPL24 suppresses Myc-driven lymphomagenesis in Eµ-Myc mice [84], and active-site mTOR inhibitors suppress MYC-dependent translation with superior efficacy compared to rapalogs [86]. These data frame the MYC-ribosome biogenesis-eIF4F axis as a linear vulnerability pathway in aggressive lymphomas with multiple druggable nodes, from Pol I inhibition (CX-5461) through mTOR inhibition (sapanisertib) to direct eIF4A inhibition (zotatifin), each operating at a different level of the same translational cascade.

Multiple myeloma

Multiple myeloma (MM) occupies the end of the translational dependency spectrum among hematologic malignancies. The defining biology of MM, the clonal expansion of plasma cells that secrete massive quantities of monoclonal immunoglobulin, imposes an extraordinary burden on the protein synthesis and secretory machinery that is unmatched by any other hematologic malignancy [325, 326]. This maximal translational demand creates both the defining vulnerability of the disease and the mechanistic basis for its most effective therapies, establishing MM as the paradigmatic example of translation-proteostasis coupling in cancer, in which translational output and protein quality control are obligately co-dependent and jointly determine cell survival.

The secretory commitment of myeloma plasma cells requires sustained ribosome biogenesis, an expanded endoplasmic reticulum, and constitutive activity of the unfolded protein response (UPR) to manage the proteotoxic stress inherent in high-volume immunoglobulin production [325, 327]. The PERK-eIF2α arm of the UPR is constitutively engaged in myeloma cells, providing a baseline level of ISR activity that is essential for adaptation to ER stress but that simultaneously positions these cells near the threshold beyond which ISR activation becomes cytotoxic [327, 328]. This proximity to the ISR cytotoxic threshold is the molecular basis for the extraordinary clinical sensitivity of MM to proteasome inhibitors. Bortezomib, carfilzomib, and ixazomib disrupt the proteasome-mediated degradation of misfolded proteins, leading to rapid accumulation of polyubiquitinated substrates, exacerbation of ER stress, hyperactivation of PERK and the ISR, and engagement of the ATF4-CHOP pro-apoptotic axis [327–329]. Myeloma cells, already operating at maximal ER stress under homeostatic conditions, cannot buffer this additional proteotoxic insult, whereas normal cells with lower secretory loads possess sufficient proteostatic reserve to survive proteasome inhibition [329]. This differential vulnerability explains the therapeutic window of proteasome inhibitors in MM and connects directly to the intensity-dependent ISR toggle between survival and death discussed in “The integrated stress response and the eIF2α axis as a parallel mRNA translation program” section. Resistance to proteasome inhibitors involves adaptive mechanisms that include reduction of immunoglobulin secretory load, rewiring of UPR signaling from the PERK-CHOP pro-apoptotic branch toward the IRE1α-X-box binding protein 1, spliced isoform (XBP1s) pro-survival branch, and upregulation of alternative proteostatic pathways, including autophagy, effectively lowering the baseline proteotoxic stress and increasing the distance from the cytotoxic threshold [330, 331].

The dependence of MM on ribosome biogenesis has been demonstrated pharmacologically. CX-5461 induces growth inhibition and apoptosis in both TP53 wild-type and mutant myeloma cell lines at nanomolar concentrations, and significantly prolongs survival in the Vκ*MYC myeloma mouse model [64]. The mechanism involves both the classical nucleolar stress response through the RPL5/RPL11-MDM2-p53 axis and p53-independent pathways, including ATM/ATR signaling and G-quadruplex stabilization at rDNA loci, ensuring activity even in the TP53-mutant setting that characterizes high-risk and relapsed MM [21, 64]. MYC, which is deregulated through translocation or amplification in a substantial fraction of MM cases, coordinately drives rRNA synthesis, ribosomal protein production, and translational capacity, rendering MYC-high myeloma particularly dependent on ribosome biogenesis and sensitive to its inhibition [332].

The mTORC1/eIF4F axis in MM is activated by multiple upstream inputs, including aberrant PI3K/AKT signaling driven by cytokine stimulation (IL-6, IGF-1), FGFR3 activation, and loss of negative regulators [85]. The resulting eIF4F-dependent translation sustains production of critical oncoproteins, including MYC, cyclin D1/D2, IRF4, and MCL-1, each of which possesses a structured 5’UTR amenable to eIF4A-dependent regulation [85, 94]. Zotatifin demonstrates activity against myeloma cell lines, and the translational logic for combining eIF4A inhibition with proteasome inhibitors in MM is compelling: eIF4A inhibition would suppress production of the short-lived oncoproteins that drive proliferation, while proteasome inhibition would prevent degradation of the misfolded proteins whose accumulation triggers apoptosis, creating convergent proteotoxic and translational stress [94]. Whether ribosome composition or rRNA/tRNA modification patterns in myeloma plasma cells differ from those in normal plasma cells or other B-cell malignancies remains to be defined, but the extreme translational demand of these cells suggests that any alteration in ribosome quality or decoding efficiency would be amplified and potentially exploitable.

The epitranscriptomic and RBP layers add further nuance. m⁶A modifications are pervasive in myeloma cells, where they regulate the stability and translation of transcripts encoding IRF4 and MYC [333]. The immunomodulatory drugs lenalidomide and pomalidomide, which function through CRBN-mediated degradation of the transcription factors Ikaros family zinc finger transcription factors IKZF1 and IKZF3, induce downstream effects that include suppression of IRF4 and MYC at the translational level, connecting the primary pharmacological mechanism to a translational output [334]. The UPR sensor IRE1α and its downstream effector XBP1s, which are essential for plasma cell differentiation and myeloma cell survival, regulate a transcriptional program that includes ribosomal protein genes and ER-associated translation factors, reinforcing the coupling between the secretory phenotype and translational capacity [327, 335]. The disease-specific translational vulnerabilities discussed in this section, and the combinatorial therapeutic strategies they rationalize, are consolidated in Table 1.

Table 1.

Disease-specific translational vulnerabilities and rational combinatorial strategies

Disease entity	Key driver	Translational node	Rational therapeutic strategy	Reference
AML (LSCs)	NPM1, FLT3, RAS	ISR/ATF4 axis; eIF4A-dependent MCL-1	Zotatifin + Venetoclax (Synergy via MCL-1/BCL-2)	[22, 96, 111, 186, 300]
T-ALL	NOTCH1-MYC-mTOR	eIF4F assembly; ribosome biogenesis	TORKi + MNK inhibitors (Overcome feedback)	[303–306, 308]
B-ALL	BCR::ABL1, L-asparaginase stress	eIF4F; ISR-mediated ASNS upregulation	L-Asparaginase+GCN2 activators	[129, 309, 310, 313]
Aggressive B-cell Lymphoma	MYC rearrangement; BCR signaling	eIF4A (Cyclin D1); SIRT3-ATF4 circuit	Zotatifin (Targets structured 5’UTRs)	[94, 124, 317, 318]
Multiple Myeloma	High Ig secretory load	UPR/ISR (PERK-eIF2α); Ribosome Biogenesis	CX-5461 or Proteasome Inhibitors (ISR toggle)	[64, 327–329]
Ribosomopathies	RPS19, RPS14, SBDS mutations	Selective translation (GATA1); p53 activation	Gene Therapy (eIF6 reduction) or ISR modulators	[246, 247, 252, 253, 267, 272, 283, 296]

Therapeutic landscape, biomarkers, and current limitations

Therapeutic agents by class and development stage

The preceding sections have discussed individual therapeutic agents within their mechanistic or disease-specific contexts. This subsection consolidates the pharmacological landscape into a unified framework organized by the targeted translational node, highlighting the development stage and key pharmacological principles for each class.

The most advanced class of agents targeting the translational machinery in hematologic oncology comprises inhibitors of ribosome biogenesis. CX-5461, a first-in-class RNA Polymerase I transcription inhibitor, has completed a phase I dose-escalation study in patients with advanced hematologic cancers, demonstrating pharmacodynamic evidence of on-target rDNA transcription inhibition and manageable toxicity at biologically active doses [21]. Its mechanism extends beyond Pol I inhibition to include G-quadruplex stabilization and topoisomerase II poisoning at rDNA loci, thereby expanding its potential activity to TP53-mutant disease but also creating a fundamental interpretive ambiguity regarding whether clinical responses are driven by translational suppression, DNA damage, or both [72, 73]. BMH-21, a structurally distinct Pol I inhibitor that induces degradation of the RPA194 catalytic subunit, was initially considered a cleaner Pol I-selective agent than CX-5461 because it does not produce the same DNA damage signaling profile [336]. More recent evidence, however, has complicated this framing: BMH-21, like other DNA intercalators, also induces time-dependent degradation of POLR2A, the catalytic subunit of RNA Pol II, with concomitant reduction of Pol II Ser5 and Ser2 phosphorylation, indicating that its anti-cancer activity is not strictly confined to Pol I inhibition [337]. As discussed in “Limitations and unresolved challenges” section, this finding is consistent with the broader recognition that the three eukaryotic RNA polymerases share core subunits and functional dependencies, making truly selective single-Pol inhibition mechanistically challenging. The chemotherapeutic agent CGX-635, which targets FBL phase separation and impairs pre-rRNA processing, represents an emerging approach within ribosome biogenesis that has shown preclinical activity specifically in AML [30] (Fig. 5A).

eIF4A inhibitors constitute the most pharmacologically developed class of direct translation initiation inhibitors. Zotatifin forms a stable ternary complex with eIF4A and polypurine RNA motifs in 5’UTRs, has been evaluated in a phase I/II clinical trial in solid tumors (NCT04092673), and is the most clinically advanced eIF4A inhibitor [94, 338]. Its evaluation in hematologic malignancies remains preclinical but has generated compelling data in AML (synergy with venetoclax), DLBCL (single-agent efficacy at sub-milligram dosing), and Burkitt lymphoma [22, 94, 96]. Second-generation eIF4A inhibitors with non-rocaglate scaffolds and oral bioavailability are in development, potentially addressing drug-like property limitations of the rocaglate class [97, 319]. The eIF4G-eIF4E interaction disruptor SBI-756 and the eIF4E-cap interaction competitor ribavirin represent alternative approaches to eIF4F disruption that have demonstrated preclinical activity in lymphoma and AML, respectively [323, 339] (Fig. 5A).

MNK1/2 inhibitors target the phosphorylation of eIF4E at Ser209, a modification that modulates eIF4E oncogenic activity through mechanisms that are incompletely resolved but include enhanced translation of pro-survival mRNAs. Tomivosertib (eFT508), an oral MNK1/2 inhibitor, has been evaluated in clinical trials and has shown preclinical activity in lymphomas and AML [340]. The principal rationale for MNK inhibition lies in its role as a compensatory pathway, in which mTOR inhibition paradoxically activates MNK1-dependent eIF4E phosphorylation, and concurrent targeting of both mTOR and MNK overcomes this resistance mechanism to achieve synergistic cytotoxicity in T-ALL and lymphoma models [308, 341] (Fig. 5B). This compensatory circuit underscores a general principle in translational therapeutics: single-node inhibition frequently elicits adaptive responses through parallel pathways.

mTOR inhibitors occupy a complicated position in the therapeutic landscape. Rapalogs (everolimus, temsirolimus) have been evaluated in clinical trials across multiple hematologic malignancies, with modest single-agent activity attributable to their incomplete suppression of 4E-BP1 phosphorylation and consequent failure to fully block eIF4F assembly [342, 343]. Active-site mTOR kinase inhibitors (sapanisertib/INK128, CC-214) achieve more complete 4E-BP1 dephosphorylation and superior preclinical activity, but clinical development has been limited by toxicity concerns arising from dual mTORC1/mTORC2 inhibition [47, 85]. Bi-steric mTORC1 inhibitors (RapaLinks), which combine the binding specificity of rapamycin with the catalytic completeness of active-site inhibitors, represent a third-generation approach that achieves full 4E-BP1 suppression with selective mTORC1 targeting, potentially improving the therapeutic window [344].

ISR modulators are the most conceptually innovative class, exploiting the intensity-dependent toggle between adaptive survival and cytotoxic ISR engagement. HC-7366, a first-in-class GCN2 activator, is in a phase I clinical trial in solid tumors (NCT05121948) and has demonstrated preclinical efficacy in AML models as monotherapy and in combination with venetoclax [130]. Its mechanism, which pushes cells beyond the adaptive ISR threshold while simultaneously reducing mitochondrial respiration and overcoming venetoclax resistance mechanisms, is pharmacologically distinct from any existing anti-leukemic agent [130]. ISRIB, which enhances eIF2B activity to override the ISR, represents the opposite therapeutic logic: denying tumor cells access to the adaptive programs that sustain survival under stress [131]. This bidirectional pharmacological approach, in which both activation and inhibition of the same pathway can be cytotoxic depending on context, demands biomarker-guided patient stratification, though validated predictive biomarkers, likely including baseline ISR activity signatures, ATF4 target expression, and mitochondrial stress indices, remain to be prospectively defined (Fig. 5A).

Translation elongation inhibitors have established clinical precedents in hematology. Omacetaxine mepesuccinate (homoharringtonine), which inhibits the initial elongation step by preventing aminoacyl-tRNA positioning in the ribosomal A-site, is FDA-approved for CML resistant to tyrosine kinase inhibitors and represents the only translation-targeting agent currently in routine clinical use for a hematologic malignancy [295]. Its mechanism of action suggests potential broader applicability in diseases with altered elongation dynamics, although systematic evaluation in other hematologic contexts remains limited.

Epitranscriptomic drugs represent a rapidly expanding pharmacological frontier, and inhibitors of the m⁶A methyltransferase METTL3 are the most advanced within this class. STM2457 and its clinical derivative STC-15 have demonstrated potent anti-leukemic activity in preclinical AML models by reducing m⁶A deposition on oncogenic transcripts and impairing their translation [345]. Inhibitors of the m⁶A eraser FTO have also shown activity in AML, where FTO is overexpressed and promotes leukemogenesis by removing m⁶A from differentiation-associated and tumor-suppressive transcripts, effectively silencing them and sustaining the oncogenic program [346]. Notably, the apparently paradoxical observation that inhibition of both the m⁶A writer METTL3 and the m⁶A eraser FTO yields anti-leukemic effects reflects the transcript-selectivity and reader-dependency of m⁶A function, rather than a contradiction. METTL3 deposits m⁶A on oncogenic transcripts such as MYC and BCL2, where the modification is interpreted by IGF2BPs and YTHDF1 to enhance mRNA stability and translational efficiency; [53, 145] its inhibition, therefore, selectively destabilizes this oncogenic program. FTO, by contrast, removes m⁶A from a distinct set of targets that includes differentiation-associated and tumor-suppressive transcripts, effectively silencing them; its inhibition restores m⁶A deposition on these targets and partially rescues their expression [347]. The directional effect of m⁶A on any given transcript is thus determined not by the modification per se, but by the identity of the transcript, the genomic position of the modified site, and the repertoire of reader proteins available in that cellular context, parameters that differ systematically between METTL3 and FTO target sets in AML cells. Targeting of m⁶A readers (IGF2BP family) and tRNA-modifying enzymes (METTL1/WDR4) remains at an earlier preclinical stage but is supported by strong genetic validation [62, 150]. The broader concept of decoding-level targeting, encompassing inhibition of tRNA modification enzymes, modulation of aminoacyl-tRNA synthetases, and pharmacological interference with Pol III-dependent tRNA transcription, constitutes a new therapeutic frontier that is mechanistically orthogonal to all existing anti-cancer agents. Its clinical development will require novel pharmacological approaches, including biomarker strategies capable of identifying tumors with selective dependencies on specific decoding components [215].

RNA-binding protein inhibitors complete the therapeutic repertoire. Ro 08-2750, a small molecule that directly binds MSI2 and competes for RNA binding, has demonstrated in vivo efficacy in murine AML models [157]. The pharmacological targeting of RNA-binding proteins has historically been considered challenging because these proteins lack traditional enzymatic pockets, but the MSI2 paradigm demonstrates that competitive inhibition of RNA-binding interfaces is achievable and therapeutically relevant. The pharmacological agents discussed in this section, spanning ribosome biogenesis inhibitors, translation initiation inhibitors, ISR modulators, elongation inhibitors, and epitranscriptomic drugs, are summarized in Table 2, together with their molecular targets and current stage of clinical development. The principal pharmacological nodes, rational combinatorial strategies, mechanisms of translational plasticity and acquired resistance, and the framework for biomarker-guided patient stratification are integrated in Fig. 5.

Table 2.

Pharmacological landscape of translation-targeting agents in hematologic oncology

Class	Agent	Molecular target	Development stage	References
Ribosome biogenesis	CX-5461	RNA Pol I inhibitor; G-quadruplex stabilizer	Phase I (Hematologic)	[21, 72, 73]
	CGX-635	Targets Fibrillarin (FBL) phase separation	Preclinical	[30]
Translation initiation	Zotatifin (eFT226)	eIF4A inhibitor (Rocaglate); clamps eIF4A on RNA	Phase I/II (Solid tumors)	[94, 338]
	Tomivosertib	MNK1/2 inhibitor; blocks eIF4E phosphorylation	Clinical trials	[340]
	Sapanisertib (INK128)	Active-site mTOR inhibitor (TORKi); blocks 4E-BP1	Preclinical/Clinical	[47, 85]
Integrated stress response	HC-7366	GCN2 activator; triggers cytotoxic ISR/ATF4 levels	Phase I	[130]
	ISRIB	eIF2B activator; bypasses ISR to restore translation	Preclinical	[131]
Elongation	Omacetaxine	Prevents aminoacyl-tRNA positioning in the A-site	FDA-approved (CML)	[295]
Epitranscriptomics/RBP	STM2457 / STC-15	METTL3 (m6A writer) inhibitor	Preclinical	[345]
	Ro 08-2750	MSI2 inhibitor; competes for RNA binding	Preclinical	[157]

Biomarkers and patient stratification

The clinical translation of translational therapeutics is critically dependent on the identification of biomarkers that predict response and enable rational patient selection. At present, no validated companion diagnostic exists for any translation-targeting agent in hematologic oncology, and this gap represents one of the most significant barriers to clinical development.

A fundamental distinction that should guide biomarker development in this field is between static and functional biomarkers. Genomic biomarkers, such as TP53 mutational status, PTEN loss, or NOTCH1 activating mutations, are static indicators that identify genetic contexts in which specific translational dependencies are likely but do not directly measure translational activity. Functional biomarkers, by contrast, report on the dynamic state of the translational machinery and are more likely to predict response to translation-targeting agents, as they capture the actual output of the translational network rather than inferring it from upstream genetics [111, 348]. ATF4 activity, for example, is a more direct predictor of ISR-dependent vulnerability than any individual ISR kinase mutation, because it integrates the net effect of all upstream inputs on the eIF2α-ATF4 axis [348] (Fig. 5D).

Several candidate biomarkers have emerged from preclinical studies. For eIF4A inhibitors, activation status of the PI3K/AKT/mTOR pathway, assessed by phosphorylation of 4E-BP1 or S6K, correlates with zotatifin sensitivity in lymphoma xenograft models, while PTEN loss is associated with diminished response [94]. For Pol I inhibitors, TP53 mutational status influences but does not determine response, as CX-5461 retains activity through p53-independent mechanisms in TP53-mutant disease, albeit at reduced sensitivity [21, 72]. For ISR modulators, baseline ISR activity, potentially assessed by ATF4 reporter expression or eIF2α phosphorylation levels, could stratify patients likely to respond to GCN2 activation (those near the adaptive threshold) versus those who might benefit from ISR inhibition (those with high constitutive ISR activity sustaining survival) [130, 348]. Phosphoproteomics, measuring phosphorylation of 4E-BP1, S6K, and eIF2α by mass spectrometry or quantitative immunoassay, represents a more clinically deployable approach to assessing translational state than Ribo-seq, and could provide a pharmacodynamic readout for mTOR inhibitors, ISR modulators, and eIF4F-targeting agents [1].

Translational profiling technologies, including Ribo-seq and polysome profiling, provide the most comprehensive assessment of translational state but are currently restricted to research settings by their requirement for fresh tissue, specialized equipment, and bioinformatic expertise [349]. RNA-based translational signatures, derived from expression levels of genes with TOP-motif, ISR target genes (ASNS, PSAT1, ATF3), or MYC translational programs, offer a potential surrogate that could be applied to standard RNA-seq data from diagnostic specimens, effectively inferring translational state from transcriptomic data without the need for ribosome profiling. However, this inference carries an inherent limitation: mRNA abundance and translational efficiency are only modestly correlated at the genome-wide level, and transcript-selective regulation, including mTORC1-dependent control of TOP mRNAs and ISR-dependent reprogramming, can alter protein output without proportional changes in mRNA levels, potentially reducing the sensitivity of transcriptomic proxies for detecting clinically relevant translational states [5]. In AML, EIF4A1 expression in primitive leukemic cell populations, detectable by single-cell transcriptomics, has been proposed as a biomarker for zotatifin sensitivity [22]. In del(5q) MDS, TP53 variant allele frequency is already used clinically to predict lenalidomide response and disease progression risk, demonstrating that translational stress-related biomarkers can inform therapeutic decisions [288] (Fig. 5D).

Beyond single biomarkers, machine learning approaches are increasingly being applied to derive multigene prognostic signatures of ribosome biogenesis activity directly from transcriptomic data. Using LASSO-penalized Cox regression across multiple AML patient cohorts (ClinSeq, BEAT-AML, FINN-AML, and TCGA), a 31-gene ribosome biogenesis (RiBi) signature was developed that stratifies patients into low, moderate, and high RiBi groups inversely correlated with overall survival, retaining independent prognostic value after adjustment for NPM1 and TP53 mutational status and the ELN2022 risk classification; the signature also carried predictive value, identifying patients with moderate-to-high RiBi activity as those most likely to benefit from allogeneic hematopoietic stem cell transplantation over chemotherapy alone [350]. Mil Single-cell deconvolution of this signature revealed a redistribution of high RiBi activity from mature immune populations in healthy bone marrow to leukemic stem and progenitor cells and monocytic blasts in AML, consistent with the RiBi-stemness coupling discussed throughout this review. A complementary approach integrated single-cell and bulk transcriptomic data using a random survival forest algorithm interpreted through SHapley Additive exPlanations (SHAP), generating a ribosome biogenesis signature in which EXOSC2 emerged as the top predictive feature; low-signature patients displayed an immune-activated microenvironment and improved predicted immunotherapy response, and molecular docking identified the cardiac glycosides ouabain and digoxin as high-affinity EXOSC2 ligands, illustrating how AI-interpretable models can nominate drug-repurposing candidates alongside prognostic stratification [351]. These studies illustrate the growing role of explainable machine learning in translating the ribosome biogenesis axis discussed throughout this review into clinically deployable, multigene prognostic tools.”

The integration of translational biomarkers with existing molecular classification systems represents an opportunity to refine risk stratification. The ELN 2022 genetic risk classification in AML does not currently incorporate translational parameters, yet the translational state of LSCs (ISR activity, eIF4A expression, METTL3/m⁶A status) may provide prognostic and predictive information that is orthogonal to genetic risk [299]. Similarly, in DLBCL, the cell-of-origin classification (ABC versus GCB) correlates with differential BCR signaling and mTOR activation, suggesting that translational biomarkers could complement existing molecular subtypes for therapeutic decision-making [352].

Limitations and unresolved challenges

Despite the compelling mechanistic rationale and promising preclinical data, the clinical translation of therapies targeting the translational apparatus faces several fundamental challenges that must be addressed for the field to advance.

The first and most significant challenge is the therapeutic window. Translation is a universal cellular process, and the concern that its pharmacological inhibition will produce unacceptable toxicity to normal tissues has been a persistent barrier to clinical development [24]. The evidence reviewed in this article suggests that the therapeutic window exists not because cancer cells translate more than normal cells in absolute terms, but because they translate differently, a principle of qualitative rather than quantitative selectivity [6, 89]. The selectivity of rocaglates for polypurine-containing 5’UTRs, the dependence of LSCs on constitutive ISR activity, and the compositional differences between onco-ribosomes and normal ribosomes all provide mechanistic bases for differential sensitivity. This framing positions translational targeting within the broader concept of non-oncogene addiction, in which cancer cells become dependent on stress support systems that are not themselves oncogenes but are essential for tolerating the consequences of oncogenic transformation [353]. Nonetheless, the clinical validation of this therapeutic window remains incomplete, and dose-limiting toxicities, including cytopenias, mucositis, and gastrointestinal effects, have been observed with mTOR inhibitors and in the early clinical evaluation of zotatifin [338, 342].

The second challenge is the distinction between on-target pharmacological effects and mechanism-divergent activities. CX-5461 exemplifies this problem: its clinical activity may derive from Pol I inhibition, G-quadruplex stabilization, topoisomerase II poisoning, or a combination of these, and disentangling their relative contributions is essential for rational patient selection and combination design [72, 73]. Similarly, rocaglates may exert effects beyond eIF4A clamping, and the pharmacodynamic consequences of translation inhibition extend to cellular processes (stress granule dynamics, autophagy, immune signaling) that are not captured by simple measurements of translational output. This complexity demands pharmacodynamic monitoring strategies that can distinguish between intended translational effects and mechanism-divergent activities in clinical settings, yet measuring translation in vivo remains technically challenging, as no validated clinical assay currently exists for real-time assessment of translational state in patient tumor cells [348].

A related and increasingly recognized challenge is the development of inhibitors that selectively target a single RNA polymerase among the three eukaryotic Pols. Although Pol I, Pol II, and Pol III transcribe distinct classes of RNA and are conventionally viewed as functionally segregated, recent evidence has substantially blurred this division of labor. Pol II operates within human nucleoli near rRNA genes, where it generates an R-loop shield at intergenic spacers that protects nucleolar organization and indirectly supports ribosome biogenesis, revealing a previously unappreciated nucleolar function for Pol II [354]. Conversely, agents nominally classified as Pol I inhibitors exhibit cross-polymerase activity: CX-5461 functions as a Topoisomerase II poison and G-quadruplex stabilizer independently of Pol I transcription (“The RPL5/RPL11–MDM2–p53 pathway and broader nucleolar stress signaling” section), and BMH-21 induces degradation of the Pol II catalytic subunit POLR2A in addition to its established effects on Pol I [337]. The three Pols also share five core subunits and additional regulatory factors, and their activities are coordinately regulated by oncogenic drivers such as MYC. These observations collectively indicate that achieving genuine pharmacological selectivity for a single RNA polymerase is mechanistically more difficult than originally assumed, and that the clinical activity of agents in this class may reflect composite effects on transcriptional machinery rather than isolated inhibition of their nominal target. This recognition reinforces the need for rigorous pharmacodynamic monitoring and for the development of next-generation compounds with empirically validated selectivity profiles.

The third challenge is redundancy and compensatory mechanisms. The translational apparatus possesses multiple parallel pathways for maintaining oncogenic protein synthesis, as demonstrated throughout this review, such as eIF4F-dependent and eIF4F-independent initiation, canonical and non-canonical ISR, m⁶A-dependent and RBP-mediated translation, and codon-level optimization through tRNA modifications. Inhibition of any single node may be compensated by upregulation of parallel pathways, as exemplified by the MNK1-dependent eIF4E phosphorylation that is activated upon mTOR inhibition in T-ALL [308] and by the eIF4A-independent, cap-independent translation mechanisms that may sustain oncogenic protein production when the canonical eIF4F route is blocked [98] (Fig. 5C). This redundancy provides a strong rationale for combinatorial approaches but also increases the complexity of clinical trial design.

The fourth challenge is the absence of validated biomarkers for patient stratification, as discussed in “Biomarkers and patient stratification” section. Without reliable predictive biomarkers, clinical trials of translation-targeting agents risk enrolling heterogeneous patient populations in which the translational vulnerability is present in only a subset, diluting the treatment effect and potentially leading to premature abandonment of agents with genuine but context-specific activity.

The fifth challenge is the temporal and spatial heterogeneity of translational states, compounded by translational plasticity under therapeutic pressure. The translational landscape of a tumor is not static but evolves during disease progression, under treatment, and across different anatomical niches [115]. LSCs in the hypoxic bone marrow niche may exhibit different translational dependencies than leukemia blasts in the peripheral blood, and the translational state of residual disease after therapy may differ fundamentally from that at diagnosis. Critically, tumor cells can shift their translational state in response to pharmacological perturbation, as exemplified by the metabolic and ISR rewiring that mediates venetoclax resistance in AML [112, 114] (Fig. 5C). This translational plasticity means that a therapeutic strategy designed to exploit a translational vulnerability at diagnosis may encounter a fundamentally different translational landscape at relapse, demanding adaptive therapeutic approaches that can track and respond to evolving translational states.

The sixth challenge concerns the interaction between translational perturbations and the immune microenvironment. The ISR modulates anti-tumor immunity through ATF4-dependent secretory programs, and translational inhibition may either enhance or suppress anti-tumor immune responses depending on context [116, 117]. The integration of translation-targeting agents with immunotherapy, an increasingly important component of hematologic oncology, requires systematic characterization of their immunomodulatory effects, which has not been undertaken for most agents in this class.

Finally, the translational control network is sufficiently complex that rational combinatorial design requires computational modeling and systems biology approaches that are not yet fully developed. Predicting which combinations of translational inhibitors will be synergistic, additive, or antagonistic in specific genetic and disease contexts demands a quantitative understanding of translational network topology that exceeds current capabilities. The future of the field lies not in identifying a single dominant translational target, but in defining disease-specific translational states and matching them with rationally designed combinations that collapse the adaptive capacity of the translational network.

Perspectives and priority gaps

The mechanistic, translational, and clinical evidence assembled in this review establishes the translational apparatus as a causally implicated and pharmacologically accessible layer of regulation in hematologic malignancies. Yet, for every question that has been answered, several remain open. This section identifies what we consider to be the highest-priority gaps in current knowledge and the most promising directions for investigation over the next decade.

The first and arguably most consequential priority is the systematic mapping of disease-specific translational states at single-cell resolution. The translational identity of LSCs, therapy-resistant clones, and microenvironmentally adapted subpopulations remains defined primarily by bulk Ribo-seq and polysome profiling, technologies that average across heterogeneous cell populations and cannot resolve the translational programs of functionally distinct subclones [349, 355]. The development of single-cell ribosome profiling (scRibo-seq) and related approaches that measure translational output at the resolution of individual cells is technically challenging but would transform the field by enabling the identification of translational dependencies that are masked in bulk analyses [356]. Computational methods that infer translational efficiency from single-cell RNA-seq data, using signatures derived from known translationally regulated gene sets, represent a more immediately deployable alternative, although their accuracy as surrogates for direct translational measurement remains to be fully validated [348].

The second priority is the elucidation of how translational states evolve under therapeutic pressure. The evidence reviewed in “The integrated stress response and the eIF2α axis as a parallel mRNA translation program” and “Acute myeloid leukemia” section indicates that venetoclax resistance in AML involves metabolic and translational rewiring, but the temporal dynamics of this process, the specific translational programs that emerge during resistance, and the degree to which they are reversible remain poorly characterized [112, 114]. Longitudinal translatomic profiling of patients during treatment, from diagnosis through remission, minimal residual disease, and relapse, would provide the empirical foundation for adaptive therapeutic strategies that target the translational state of the disease at each phase rather than relying on a fixed regimen designed for the diagnostic translational profile. The integration of such data with clonal tracking by whole-exome or whole-genome sequencing would reveal whether translational adaptation is genetically encoded (through selection of clones with specific mutations) or epigenetically and metabolically driven (through plastic rewiring of existing translational circuits), a distinction with direct implications for therapeutic design.

The third priority is the systematic characterization of how translation-targeting agents interact with the immune microenvironment. The discovery that ATF4 drives immune evasion through LCN2-mediated macrophage recruitment and that chronic ISR activation in CD8 + T cells promotes exhaustion opens the possibility that translational inhibitors could function as immunomodulators [116, 117]. However, the immunological consequences of pharmacological translation inhibition are likely to be context-dependent and bidirectional, and suppressing translation in tumor cells may enhance antigen presentation or reduce immune checkpoint expression, while simultaneously impairing effector T cell function if translation is inhibited in immune cells [116]. Preclinical studies systematically comparing the immunological effects of translation-targeting agents in syngeneic and humanized mouse models, combined with correlative immune profiling in clinical trials, are needed to define the immunotherapeutic potential and risks of this class of agents.

The fourth priority is the development of clinically deployable biomarkers for translational state. As discussed previously, the field currently lacks validated companion diagnostics for translation-targeting agents. The most promising near-term approaches include immunohistochemical or flow cytometric panels for phospho-4E-BP1, phospho-eIF2α, and ATF4, which could be implemented on standard clinical platforms, and RNA-based translational signatures that infer translational state from transcriptomic data [5, 111]. The validation of such biomarkers will require prospective collection of translational measurements alongside clinical outcomes in trials of translation-targeting agents, an investment that should be incorporated into trial design from the outset rather than pursued retrospectively.

The fifth priority is the exploration of translational control in the context of minimal residual disease and clonal hematopoiesis. The evidence from ribosomopathies demonstrates that chronic translational stress drives clonal evolution and cancer predisposition, and the observation that clonal hematopoiesis of indeterminate potential (CHIP) involves mutations in genes with translational relevance (TP53, DNMT3A, splicing factors) suggests that translational perturbations may contribute to the earliest stages of leukemogenesis [257, 258]. Whether targeting translational vulnerabilities at the CHIP stage could prevent progression to overt malignancy is a provocative but untested hypothesis with significant implications for cancer prevention. Similarly, the translational state of minimal residual disease cells after therapy, which persist in the bone marrow niche under conditions of metabolic and therapeutic stress, may differ fundamentally from that of bulk disease and could represent a particularly tractable therapeutic target if appropriately characterized.

The sixth priority is the integration of translational control with the epitranscriptomic and epigenetic layers that regulate gene expression. The evidence that mRNA, rRNA, and tRNA modifications collectively shape translational output reveals a multi-layered regulatory architecture in which the chemical modification state of RNA at every level (from mRNA to rRNA to tRNA) contributes to the final proteomic output. The interactions between these layers are only beginning to be characterized: whether m⁶A modifications on mRNA influence their translation by modified versus unmodified ribosomes, whether tRNA modification state affects the decoding of m⁶A-marked transcripts, and whether rRNA modifications alter the response of ribosomes to ISR-driven translational reprogramming are all open questions with fundamental implications for understanding how the translational apparatus generates proteome diversity in cancer [53, 62, 190].

The seventh priority concerns the application of computational and systems biology approaches to translational network modeling. The redundancy of translational control mechanisms, demonstrated throughout this review, implies that rational combinatorial design cannot be achieved through intuition alone but requires quantitative models that predict network behavior under pharmacological perturbation [244]. Such models would integrate data on ribosome biogenesis rates, eIF4F assembly dynamics, ISR activation thresholds, tRNA pool composition, and mRNA feature analysis to predict which translational dependencies are most vulnerable in specific genetic and disease contexts. The development of these models will require close collaboration between computational biologists, structural biochemists, and clinical investigators, representing a genuinely interdisciplinary frontier.

Finally, the field would benefit from a conceptual reframing that positions the ribosome not as a passive executor of transcriptional programs but as an active regulatory node that filters, selects, and modulates gene expression with a degree of autonomy that rivals transcriptional control [167, 168]. The evidence assembled in this review, from ribosome heterogeneity and specialized translation through rRNA and tRNA modifications to the ISR and epitranscriptomic regulation, consistently supports this view. Recognizing the ribosome as a regulatory device, rather than merely a decoding machine, has implications that extend beyond oncology to developmental biology, immunology, and regenerative medicine, and positions translational control as a fundamental principle of gene regulation whose full therapeutic potential remains to be realized.

Conclusions

The translational apparatus occupies a unique position in the regulatory architecture of hematologic malignancies. It is the obligatory intermediary between the genome and the proteome, and its deregulation is not a secondary consequence of oncogenic signaling but a primary determinant of tumor cell identity, survival, and therapeutic resistance. The evidence assembled in this review supports three principal conclusions.

First, translational control in hematologic malignancies operates through multiple mechanistically distinct but functionally interconnected layers. Cap-dependent translation initiation through the eIF4F complex, the ISR through the eIF2α-ATF4 axis, non-canonical regulators including RNA-binding proteins and epitranscriptomic modifiers, ribosome heterogeneity and onco-ribosomes, rRNA and tRNA modifications, and codon-level decoding optimization collectively define the translational output of malignant cells. No single layer is sufficient to explain the oncogenic proteome, and no single therapeutic strategy targeting a single layer is likely to achieve durable responses. The disease-specific configurations of these layers, ranging from the low-output, high-selectivity, stress-adapted state of AML LSCs to the maximally engaged, minimally buffered translational apparatus of myeloma plasma cells, define the translational vulnerability profile of each malignancy and should guide therapeutic design.

Second, the ribosomopathies provide genetically defined demonstrations that translational perturbation is causally linked to malignant transformation. The paradox of ribosomal insufficiency predisposing to cancer is resolved by an evolutionary framework in which chronic translational stress selects for clones that escape quality control checkpoints while retaining altered translational selectivity, generating a cellular state permissive for oncogenesis. The compensatory versus transforming clonal trajectories identified in Shwachman-Diamond syndrome epitomize the evolutionary trade-offs that govern the relationship between translational fitness and malignant transformation, and the principles that emerge from these natural experiments, that translational stress is a selective force and that ribosome composition determines proteomic output, apply broadly to acquired hematologic malignancies.

Third, the therapeutic targeting of the translational apparatus has advanced from concept to clinical reality. RNA Polymerase I inhibitors, eIF4A inhibitors, GCN2 activators, mTOR kinase inhibitors, elongation inhibitors, epitranscriptomic drugs, and RNA-binding protein inhibitors collectively define a pharmacological repertoire that covers every major node of translational control. The challenge is no longer whether translation can be therapeutically targeted but how to match the right combination of translational inhibitors with the right disease context, guided by biomarkers that capture the dynamic translational state of tumor cells.

The path forward requires the convergence of translational biology, computational modeling, single-cell technologies, and clinical investigation. The translational apparatus is not merely a downstream effector of oncogenic signaling. It is a regulatory system with its own logic, its own vulnerabilities, and its own capacity for adaptation. Understanding and exploiting that logic, with the precision that the complexity of the system demands, is the central challenge and the central opportunity of translational oncology in hematologic malignancies.

Abbreviations

4E-BP

eIF4E-binding protein

5'UTR

5’ untranslated region

ABC

Activated B-cell (subtype of DLBCL)

AKT

Protein kinase B

ALL

Acute lymphoblastic leukemia

AML

Acute myeloid leukemia

ASNS

Asparagine synthetase

ATF4

Activating transcription factor 4

ATF5

Activating transcription factor 5

ATM

Ataxia telangiectasia mutated

ATR

Ataxia telangiectasia and Rad3-related

BCL-2

B-cell lymphoma 2

BCL-X_L

B-cell lymphoma extra-large

BCR

B-cell receptor

BCR-ABL

Breakpoint cluster region-Abelson fusion

BRD4

Bromodomain-containing protein 4

CaMKII

Calcium/calmodulin-dependent kinase II

CDK

Cyclin-dependent kinase

CEBPA

CCAAT/enhancer-binding protein alpha

CHIP

Clonal hematopoiesis of indeterminate potential

CHOP

C/EBP homologous protein (DDIT3)

CLL

Chronic lymphocytic leukemia

CML

Chronic myeloid leukemia

CRBN

Cereblon

CRLF2

Cytokine receptor-like factor 2

DBAS

Diamond-Blackfan anemia syndrome

DC

Dyskeratosis congenita

DDX3X

DEAD-box helicase 3 X-linked

DEPTOR

DEP domain-containing mTOR-interacting protein

DHL

Double-hit lymphoma

DKC1

Dyskerin pseudouridine synthase 1

DLBCL

Diffuse large B-cell lymphoma

DNAJC21

DNAJ heat shock protein family (Hsp40) member C21

DNMT3A

DNA methyltransferase 3 A

eEF1A

Eukaryotic elongation factor 1 A

eEF2

Eukaryotic elongation factor 2

eEF2K

Eukaryotic elongation factor 2 kinase

EFL1

Elongation factor-like GTPase 1

eIF2

Eukaryotic initiation factor 2

eIF2α

Eukaryotic initiation factor 2 alpha subunit

eIF2B

Eukaryotic initiation factor 2B

eIF3

Eukaryotic initiation factor 3

eIF4A

Eukaryotic initiation factor 4 A

eIF4B

Eukaryotic initiation factor 4B

eIF4E

Eukaryotic initiation factor 4E

eIF4F

Eukaryotic initiation factor 4 F complex

eIF4G

Eukaryotic initiation factor 4G

eIF6

Eukaryotic initiation factor 6

ELN

European LeukemiaNet

ER

Endoplasmic reticulum

ERK

Extracellular signal-regulated kinase

FAO

Fatty acid oxidation

FBL

Fibrillarin

FGFR

Fibroblast growth factor receptor

FLT3

fms-like tyrosine kinase 3

FLT3-ITD

FLT3 internal tandem duplication

FTO

Fat mass and obesity-associated protein (m⁶A demethylase)

GCB

Germinal center B-cell (subtype of DLBCL)

GCN2

General control nonderepressible 2

GI50

50% growth inhibition concentration

GSI

Gamma-secretase inhibitor

GTP

Guanosine triphosphate

HRI

Heme-regulated inhibitor

HSC

Hematopoietic stem cell

IC50

Half-maximal inhibitory concentration

IDH

Isocitrate dehydrogenase

IGF2BP

Insulin-like growth factor 2 mRNA-binding protein

IKZF

Ikaros family zinc finger

IL-6

Interleukin 6

IL7Rα

Interleukin 7 receptor alpha

IRES

Internal ribosome entry site

IRF4

Interferon regulatory factor 4

IRE1α

Inositol-requiring enzyme 1 alpha

ISR

Integrated stress response

ISRIB

Integrated stress response inhibitor

JAK

Janus kinase

LARP1

La-related protein 1

LCN2

Lipocalin 2

LSC

Leukemic stem cell

m¹A

N1-methyladenosine

m⁶A

N6-methyladenosine

m⁷G

7-methylguanosine

MCL

Mantle cell lymphoma

MCL-1

Myeloid cell leukemia 1

MCT-1

Multiple copies in T-cell malignancy 1

MDM2

Mouse double minute 2 homolog

MDS

Myelodysplastic syndrome

METTL1

Methyltransferase-like 1

METTL3

Methyltransferase-like 3

METTL14

Methyltransferase-like 14

MM

Multiple myeloma

MNK

MAP kinase-interacting kinase

MPP

Multipotent progenitor

MRD

Minimal residual disease

mRNA

messenger RNA

MSI2

Musashi RNA-binding protein 2

mTOR

Mechanistic target of rapamycin

mTORC1

mTOR complex 1

mTORC2

mTOR complex 2

MYC

MYC proto-oncogene

NAT10

N-acetyltransferase 10

NOP56

Nucleolar protein 56

NPM1

Nucleophosmin 1

NPM1c+

Cytoplasmic nucleophosmin mutant

NSUN2

NOP2/Sun RNA methyltransferase 2

OXPHOS

Oxidative phosphorylation

PABP

poly(A)-binding protein

PARP

poly(ADP-ribose) polymerase

PDCD4

Programmed cell death 4

PDX

Patient-derived xenograft

PERK

Protein kinase R-like ER kinase

PI3K

Phosphatidylinositol 3-kinase

PKR

Protein kinase R

Pol I

RNA polymerase I

Pol II

RNA polymerase II

Pol III

RNA polymerase III

PSAT1

Phosphoserine aminotransferase 1

PTEN

Phosphatase and tensin homolog

PUMA

p53 upregulated modulator of apoptosis

RBP

RNA-binding protein

rDNA

ribosomal DNA

Ribo-seq

Ribosome profiling (sequencing)

RNP

Ribonucleoprotein

RP

Ribosomal protein

RPL

Ribosomal protein of the large subunit

RPS

Ribosomal protein of the small subunit

RREB1

Ras-responsive element-binding protein 1

rRNA

Ribosomal RNA

S6K

Ribosomal protein S6 kinase

SAM

S-adenosylmethionine

SBDS

Shwachman-Bodian-Diamond syndrome protein

scRibo-seq

Single-cell ribosome profiling

SDS

Shwachman-Diamond syndrome

SL1

Selectivity factor 1

snoRNA

small nucleolar RNA

SRP54

Signal recognition particle 54

STAT

Signal transducer and activator of transcription

SYNCRIP

Synaptotagmin-binding cytoplasmic RNA-interacting protein

TAF

TBP-associated factor

T-ALL

T-cell acute lymphoblastic leukemia

TBP

TATA-binding protein

TCA

Tricarboxylic acid (cycle)

TKI

Tyrosine kinase inhibitor

TOP

Terminal oligopyrimidine

TP53

Tumor protein p53

tRNA

transfer RNA

TRMT6

tRNA methyltransferase 6

TRMT61A

tRNA methyltransferase 61 A

TSC

Tuberous sclerosis complex

UBF

Upstream binding factor

uORF

Upstream open reading frame

UPR

Unfolded protein response

VAF

Variant allele frequency

WDR4

WD repeat domain 4

WHO

World Health Organization

WTAP

Wilms tumor 1-associated protein

XBP1

X-box binding protein 1

XBP1s

Spliced X-box binding protein 1

XPO1

Exportin 1

YBX1

Y-box binding protein 1

YTHDF

YTH domain-containing family protein

ZBTB1

Zinc finger and BTB domain-containing 1

ZNF598

Zinc finger protein 598

Authors’ contributions

Conception and design: JL Coelho-Silva and JA Machado-Neto; Administrative support: JA Machado-Neto; Provision of study materials or patients: None; Collection and assembly of data: JL Coelho-Silva and JA Machado-Neto; Data analysis and interpretation: JL Coelho-Silva and JA Machado-Neto; Manuscript writing: JL Coelho-Silva and JA Machado-Neto; Final approval of manuscript: JL Coelho-Silva and JA Machado-Neto.

Funding

JLC-S received a fellowship from FAPESP (2024/21624-7).

This study was supported by grants 2023/12246-6, 2023/08735-1, and 2025/26187-7 from the São Paulo Research Foundation (FAPESP) and grant 305758/2021-7 from the Conselho Nacional de Desenvolvimento Científico e Tecnológico (CNPq).

Data availability

No datasets were generated or analysed during the current study.

Declarations

Ethics approval and consent to participate

This Review did not involve new studies in human participants or animals and did not use identifiable patient data; therefore, ethics approval and consent were not required.

Competing interests

The authors declare no competing interests.

Generative AI statement

During the preparation of this work, the authors utilized Gemini to assist with figure generation. Following its use, the author thoroughly reviewed and edited the content as necessary, taking full responsibility for the accuracy and integrity of the publication.