New classification systems focus on AS events that directly influence therapeutic response, particularly those enabling drug resistance. Understanding these mechanisms requires precise definition of ‘therapy-resistance-associated isoforms’, alternatively spliced mRNAs experimentally demonstrated to confer resistance by modifying drug targets, activating compensatory pathways, or enhancing survival (10–12). To qualify, isoforms must satisfy three criteria: i) functional validation through drug sensitivity assays; ii) association with a specific treatment; and iii) identification in human tumors or clinically relevant models.

Numerous examples demonstrate this concept. In PARP inhibitor-resistant ovarian cancer, BRCA1 exon 11 skipping generates Δ11q isoforms that restore homologous recombination repair (HRR) capacity, circumventing nonsense-mediated decay (10). Similarly, NT5C2 exon 6a inclusion yields a hyperactive nucleotidase conferring 6-mercaptopurine resistance in leukemia (36). NSCLC with KRAS G12 mutations exhibits MET exon 14 skipping that activates RAS/ERK signaling to evade MET inhibitor effects (30). Splicing alterations also perturb epigenetic regulation: PRMT5 inhibition in CDK4/6 inhibitor-resistant breast cancer induces FUS-dependent IR, impairing DNA synthesis proteins such as PCNA and MCM2 (37). Given their clinical relevance, these mechanisms are emerging as therapeutic targets themselves, for instance, CRISPR/dCasRx-mediated exclusion of TIMP1 exons counteracts SRSF1-driven splicing and attenuates metastasis in CRC (38). Such subclassifications are transforming precision oncology, with clinical trials currently evaluating splice-switching antisense oligonucleotides (ASOs) targeting PTBP1 in glioblastoma (39). Additionally, APA events in BCL2L1 produce anti-apoptotic isoforms in hepatocellular carcinoma (HCC), while poison exon inclusion in MAP3K7 disrupts TGF-β signaling to foster chemoresistance (27,35). Isoforms are excluded from this category if evidence is purely computational or derived from correlative expression without direct functional proof, or if the resistance mechanism is indirect (for example, via general proliferation effects) (10–12).

Representative isoforms that satisfy these criteria, highlighting their drug contexts, functional validation methods, and clinical evidence levels are presented in Table I.

This systematic compilation underscores the diversity of splicing-mediated resistance mechanisms, which often coexist with genetic mutations and contribute substantially to treatment failure. Functional validation through targeted assays (for example, CRISPR and ASOs) is essential to establish causality, while clinical evidence ranges from preclinical models to direct patient data, highlighting the translational relevance of these isoforms (10,12,40).

Cis-regulatory elements and trans-acting factors

Cis-acting elements, including exonic/intronic splicing enhancers (ESE/ISE) and silencers (ESS/ISS), are nucleotide sequences within pre-mRNA that recruit trans-acting regulators (for example, SR proteins, hnRNPs) to govern splice site selection (41,42), Cell-type specific gene expression patterns are regulated by cis-regulatory elements such as promoters and enhancers (43) (Fig. 2). Moreover, the regulation of AS entails recognition of cis-acting RNA elements in the pre-mRNA by trans-acting protein splicing factors (44). These elements operate combinatorially to enhance or repress spliceosome assembly, with mutations or dysregulated expression commonly observed in cancer (12). For example, BCL2L1 ESE mutations drive pro-survival BCL2L1-L isoform expression, conferring apoptosis resistance (40). Genome-wide analyses indicate that ~15% of cancer-associated mutations map to splicing regulatory elements, with ESE/ESS regions in oncogenes for example, MET) exhibiting recurrent alterations (12,45).

Chronic myeloid leukemia (CML) cells exhibit ESS defects in SRSF1-targeted transcripts (for example, PRKCH), sustaining oncogenic signaling despite BCR-ABL1 inhibition (46). Glioblastoma-associated SF3B1 inhibition alters BCL2L1 splicing, favoring anti-apoptotic isoform expression (47). High-throughput studies identify recurrent ESE/ESS mutations in drug resistance genes (for example, ABCB1), which disrupt repressive elements to circumvent therapy (44). For example, aberrant ESS motifs in TP53 exon 6 promote ES, generating dominant-negative isoforms that compromise chemotherapy-induced apoptosis (40). SR proteins (for example, SRSF1) and hnRNPs (for example, hnRNP A1) are frequently overexpressed in therapy-resistant cancers. In CML, SRSF1 upregulation enhances BCL2L1 exon 2 inclusion, elevating pro-survival BCL2L1-S isoforms and undermining imatinib-induced apoptosis (46). SF3B1 dysfunction in glioblastoma drives BCL2L1-L overexpression via preferential 3′ splice site selection, facilitating temozolomide evasion (47). Post-transcriptional modifications further modulate splicing factor activity: METTL3-mediated m6A methylation stabilizes SRSF1 mRNA in pancreatic cancer, potentiating cisplatin resistance (48). Epigenetic regulation, encompassing DNA methylation and RNA epitranscriptomics (for example, m6A), modulates splicing by altering RNA-protein interactions (49–51). In gastric cancer, SNORA37 promotes CMTR1-ELAVL1 interaction to drive CD44 AS and tumor progression (52). Dysregulation of the epigenome drives aberrant transcriptional programs that promote cancer onset and progression (53,54). METTL3 deposits m6A on QSOX1 exon 14, recruiting YTHDC1 to enhance exon inclusion and β-catenin activation, thereby conferring gefitinib resistance in NSCLC (55). Conversely, ALKBH5 erases m6A from FOXO1 mRNA, enabling SRSF3 binding and anti-apoptotic isoform generation (55). Chromatin remodelers (for example, SWI/SNF) also influence splicing: ARID1A loss induces BRD9 ES, activating Wnt signaling in ovarian cancer (44). These findings collectively define a dynamic ‘splicing code’ that integrates cis-regulatory mutations, trans-factor dysregulation and epigenetic modifications, a multilayered system that enables tumor cells to rapidly adapt and develop resistance to targeted therapies, chemotherapy, and epigenetic drugs (Fig. 2).

Altered spliceosome proteins due to somatic mutations, overexpression or post-translational modifications are key drivers of aberrant splicing that promotes tumour survival and therapy resistance. Mutations in SF3B1 in glioblastoma lead to ES such as BRCA1-Δ11q that results in a hypomorphic isoform that provides resistance to PARP inhibition (12). In CML, SRSF1 is overexpressed in leukemic blasts and IL-3 maintains SRSF1 expression following BCR-ABL1 inhibition, leading to imatinib resistance via PKC- and PLC-dependent mechanisms (46). These events promote proliferation and inhibit apoptosis providing a selective advantage in cancer.

DNA cis-element mutations (for example, in exonic splicing silencers) disrupt trans-factor binding, leading to aberrant splicing and drug resistance. In acute lymphoblastic leukemia (ALL), NT5C2 exon 6a inclusion creates a gain-of-function isoform with elevated nucleotidase activity, driving resistance to 6-mercaptopurine (36). These mutations are enriched in relapse samples and represent promising targets for ASOs due to their sequence-specific nature; ASOs can be designed to bind mutant cis-elements or aberrant splice sites, blocking incorrect splicing and restoring normal isoform expression. This re-sensitizes cells to therapy, as demonstrated in leukemia models where ASOs reversed chemoresistance by correcting NT5C2 splicing (Figs. 3B and 4) (36).

AS-mediated exclusion of ABL1 exon 2 results in an in-frame deletion of SH3 domain residues. This 31-amino-acid deletion in BCR-ABL1 weakens asciminib binding by 18-fold and impairs the autoinhibitory conformation required for drug efficacy, conferring resistance in CML (58).

In BRCA1, secondary SSMs induce exon 11 skipping, producing truncated Δ11 and Δ11q isoforms that restore HRR capacity. These isoforms allow tumor cells to bypass PARPi-induced synthetic lethality (10).

In relapsed ALL, the inclusion of exon 6a introduces a novel phosphorylation site that promotes NT5C2 dimerization, leading to a 15-fold increase in enzymatic activity of the protein compared with the wild-type protein (36).

IR and the more recent concept of ‘detained introns’ (DIs), a type of IR, are critical splicing events that drive resistance to cancer therapies (59,60). DIs are introns that are retained in polyadenylated transcripts owing to a stalled co-transcriptional splicing process. These splicing events act as rapid and reversible regulators of gene expression, whereas classical IR, is defined by persistent IR in mature mRNAs.

In glioblastoma, multiple of splicing mechanisms cooperate to mediate both acute and chronic therapy resistance. Chemoradiation with agents such as temozolomide stalls RNA polymerase II elongation and induces DIs in MLH1 through H3K9 lactylation-dependent upregulation of LUC7L2 (62), which introduces a frameshift mutation truncating the protein and leading to mismatch repair (MMR) dysfunction. The functional loss of MMR leads to acute chemoresistance. These DIs are capable of acting as rapid stress sensors, such that inhibition of lactylation rescues MLH1 splicing and re-sensitizes the tumors to treatment (63). In addition, NONO-induced IR in GPX1 results in the loss of antioxidant activity and an increase in reactive oxygen species which leads to apoptosis (63). Radiation further stalls splicing progression, converting the acute DI response into chronic IR in an epigenetically dependent manner, highlighting how stress duration mediates resistance plasticity (61).

In CRC, SRSF5-induced IR in PKM intron 9 leads to the upregulation of PKM2, which increases glycolysis and enables survival under 5-fluorouracil-induced oxidative stress (56).

In AML, loss of RBM39 leads to widespread IR promoting the nonsense-mediated decay (NMD) of pro-leukemic transcripts including HOXA9 targets (64). IR in spliceosome mutant cells mediates their acute vulnerability to this mechanism, providing a measurable biomarker of therapeutic response. Chronic RBM39 inhibition leads to the consolidation of the IR state promoting continual NMD.

In myelodysplastic syndromes (MDS), loss of ZRSR2 causes IR in a subset of minor introns in LZTR1 inducing NMD and subsequent RAS pathway activation, which drives leukemogenesis in these patients (65). Taken together, chemoradiation stalls co-transcriptional splicing, inducing DIs as an acute, reversible response to therapy (61). Chronic stress promotes the conversion of DIs into stable IR states, allowing chronic resistance to be established. This dichotomy explains how tumors adapt rapidly (via DIs) and persist (via IR) under therapeutic pressure (Fig. 3A and B).

Back-splicing and circRNA-driven drug resistance.

Back-splicing is a non-canonical RNA processing mechanism distinct from linear splicing, wherein a downstream splice donor site covalently joins an upstream splice acceptor site, forming a closed-loop RNA structure called circular RNA (circRNA) (66). Numerous protein-coding genes in higher eukaryotes are capable of producing circRNAs through back-splicing of exons (67–69). This process competes with canonical splicing for splice sites (70–72), resulting in exonuclease-resistant circRNAs that lack 5′ caps and 3′ poly(A) tails, thereby enhancing their stability and allowing secretion via extracellular vesicles (73,74). Experimental evidence underscores the remarkable stability of circRNAs; for instance, Enuka et al (75) demonstrated, using metabolic labeling that the median half-life of circRNAs in mammary cells ranges from 18.8–23.7 h, significantly longer than the 4.0–7.4 h observed for linear RNAs. Their exceptional longevity makes circRNAs well-suited as biomarkers, a feature demonstrated across multiple studies to stem from their covalently closed circular architecture which confers intrinsic resistance to exonuclease degradation (76–78). CircRNAs are broadly classified into three subtypes: Exonic circRNAs (ecircRNAs), intronic circRNAs and exon-intron circRNAs (EIciRNAs), with ecircRNAs being the most prevalent and functionally characterized in cancer (79) (Fig. 3A). Their aberrant expression in cancers provides insights into roles in tumorigenesis (80), and they can act as miRNA sponges, protein scaffolds or peptide templates (74,81). For example, circCDYL2 in HER²⁺ breast cancer stabilizes GRB7 to sustain HER2/AKT/ERK signaling and confers trastuzumab resistance (73), while circPARD3 in laryngeal squamous cell carcinoma sponges miR-145-5p to activate PRKCI/AKT/mTOR signaling and suppress autophagy, promoting chemoresistance (79). The conservation of back-splicing and its dysregulation in tumors highlight its role in therapy evasion, with circRNAs like circRNA-CREIT in TNBC serving as biomarkers for poor prognosis under doxorubicin treatment (74).

CircRNAs may act as competing endogenous RNAs to sponge miRNAs and relieve repression of oncogenic targets. In PDAC, cancer-associated fibroblasts secrete exosomal circFARP1, which sponges miR-660-3p to upregulate LIF, activating STAT3 signaling to induce stemness and gemcitabine resistance (81). In this way, circFARP1 may bind miR-660-3p to derepress the cytokine LIF, which in turn promotes IL-6/STAT3 survival signaling pathways (81). Likewise, circPARD3 acts as a sponge for miR-145-5p, leading to upregulation of PRKCI to activate AKT/mTOR signaling and inhibit autophagy to enhance cisplatin resistance in laryngeal squamous cell carcinoma (79). Such circRNA/miRNA/protein signaling axes are often amplified in tumor cells that are resistant to therapy, for example, circCDYL2 sponges miR-199a-5p to upregulate growth factor adaptor protein GRB7 and focal adhesion kinase signaling in HER²⁺ breast cancer (73). CircRNA sponging can also be blocked in tumor cells to sensitize them to chemotherapy such as with anti-circFARP1 small interfering RNA (siRNA) to to restore miRNA activity (Fig. 3C) (81).

CircRNA-protein interactions in therapy resistance: CircCDYL2/growth factor receptor-bound protein 7 (GRB7) complex formation and circRNA-CREIT-mediated protein kinase R (PKR) degradation. CircRNAs also participate in direct protein interactions that modulate cellular function. In HER²⁺ breast cancer, circCDYL2 binds GRB7, protecting it from ubiquitination, and thereby maintaining HER2/AKT/ERK signaling in the presence of trastuzumab (73). This leads to HER²⁺ tumors exhibiting resistance to HER²⁻ targeted therapies (73). In TNBC, circRNA-CREIT acts as a molecular scaffold to bring HACE1 (HACE1 ubiquitin protein ligase) and PKR together, leading to degradation of PKR and disruption of stress granules, and thereby allowing the tumor to exhibit resistance to doxorubicin (74). These protein-interaction mechanisms establish circRNAs as versatile regulators of therapeutic resistance, suggesting that targeting such interfaces, for instance through GRB7 inhibition, may represent a promising approach to reestablish treatment sensitivity (Fig. 3C).

ATS generates chimeric mRNAs by joining exons from two distinct pre-mRNA transcripts through spliceosome-mediated ligation (82,83). Unlike conventional cis-splicing, ATS connects splice sites located on different RNA molecules, often mediated by complementary base pairing or trans-acting factors (4,84). The process involves coordinated recruitment of spliceosomal components, including U1 and U2 snRNPs, to donor and acceptor sites on distinct pre-mRNAs, enabling exon joining via transesterification reactions (11). ATS events are classified as intergenic when combining exons from different chromosomal loci or intragenic when linking exons from alternative transcripts of the same gene (4,12) (Fig. 3A). This mechanism is thus expected to substantially increase transcript diversity in cancers and likely underlies oncogenic adaptation in treatment-resistant tumors (84). Although fusion genes such as ABL/BCR are widely considered to arise from translocations such as the Philadelphia chromosome, ATS alone can generate similar chimeric transcripts at the RNA level.

In CML, ATS gives rise to fusion transcripts such as CLEC12A-MIR223HG, which has been found in both malignant and normal cells, demonstrating its capacity to diversify the transcriptome in the absence of DNA rearrangement (85). ATS variants of the BCR-ABL oncogene may escape detection or cause resistance to imatinib due to an altered configuration of the kinase domain (58). By augmenting the functional transcriptome beyond what the genome can encode, ATS can increase the adaptability of tumor cells for oncogenic adaptation under therapeutic pressure (86,87).

ATS confers drug resistance through the production of oncogenic chimeric transcripts or via the disruption of drug-target interactions. In glioblastoma, FBXO7 stabilizes RBFOX2 to induce trans-splicing events that produce FOXM1 variants, which further promote the expression of stemness-related markers such as CD44 and ID1, thereby conferring chemotherapy resistance (3) (Fig. 3D). Apart from resistance to chemotherapy, ATS may also play a role in immune escape mechanisms. Pan-cancer analyses have identified tumor-specific splicing junctions, or ‘neojunctions’, that produce peptides not found in normal tissues, and computational predictions suggest these peptides may serve as ligands for MHC-I proteins and thereby influence the response to immune checkpoint inhibitors (12). Trans-splicing between BCL2L1 and adjacent genes generates truncated BCL2 variants that impair chemotherapy-induced apoptosis (11). These mechanisms highlight how ATS diversifies the transcriptome to promote survival under therapeutic pressure.

Specific ATS events are linked to therapy resistance. In prostate cancer, FOXA1 orchestrates ATS by binding enhancer regions of splicing factors such as SRSF1, leading to exon 30 inclusion in FLNA, which stabilizes isoforms promoting cytoskeletal remodeling and resistance to androgen receptor (AR) inhibitors (88). Additionally, FBXO7-driven trans-splicing in glioblastoma stabilizes RBFOX2, which coordinates splicing of mesenchymal genes such as POSTN to augment temozolomide resistance through integrin-mediated survival signaling (3). Furthermore, cross-strand chimeric RNAs (cscRNAs) produced by ATS between sense and antisense transcripts have been identified in multiple cancer cells. For instance, in prostate cancer PC3 cells and HCC Huh7 cells, specific cscRNAs such as cscR-8-21 and cscR-2-22 are generated through the fusion of transcripts from opposite DNA strands. Knockdown of these cscRNAs using siRNA significantly diminishes cell proliferation, colony formation and migration, as validated by reverse transcription PCR, Sanger sequencing, RNA fluorescence in situ hybridization, and functional assays, supporting their role in promoting cancer cell survival and potentially contributing to therapy resistance (89). In CML, canonical splicing of BCR-ABL produces p210 or p190 transcripts, but ATS can generate truncated isoforms for example, BCR-ABL1/b3a3) that lack ABL1 exon 2-encoded SH3 domain residues, leading to constitutive kinase activity and resistance to allosteric inhibitors such as asciminib by destabilizing the autoinhibited kinase conformation (58). Moreover, EZH2-mediated repression of splicing factors such as CELF2 in CML results in aberrant splicing patterns that promote stem cell-like properties and diminish sensitivity to tyrosine kinase inhibitors (Fig. 3D) (87). These ATS-derived variants activate alternative signaling pathways, such as dynamic phosphorylation events, thereby contributing to treatment failure and relapse (86). These findings demonstrate that ATS actively contributes to the development of aggressive, therapy-resistant cancers.

AS events including trans-splicing, are prevalent somatic perturbations in cancer that promote tumor heterogeneity and therapy resistance (90). By contrast, hereditary cancer syndromes are frequently driven by germline chromosomal abnormalities, including RB1 deletions in retinoblastoma, TP53 mutations in Li-Fraumeni syndrome, NF1 deletions in neurofibromatosis type 1, APC mutations in familial adenomatous polyposis, and BRCA1/BRCA2 germline mutations in hereditary breast cancer (91–93). AS predominantly occurs as a somatic event that broadly reprograms gene networks, while chromosomal abnormalities generally underlie inherited cancer syndromes with high penetrance. In neurofibromatosis type 1, mis-splicing of NF1 exon 23a disrupts neurofibromin function in high-grade gliomas, thereby activating the Ras/MAPK pathway to drive tumor formation (94).

Similarly, retinoblastoma exhibits upregulated spliceosome activity regulated by pRB/E2F3a under oncogenic stress (90). Genomic studies establish AS dysregulation as a nearly universal feature of tumorigenesis, whereas chromosomal abnormalities primarily drive ~10–20% of hereditary cancer syndromes (91–93). This predominance establishes AS as a central mechanism in sporadic cancers and a promising therapeutic target, as exemplified by metabolic vulnerabilities conferred by TP53 splicing variants in Li-Fraumeni syndrome (95).

Somatic mutations in core spliceosomal components including SRSF2 and SF3B1 produce recurrent mis-splicing events that create novel immunogenic peptides, these splicing-derived neoantigens are presented by MHC class I molecules and trigger CD8⁺ T-cell responses in myeloid malignancies (96,97). While these neoantigens provide effective targets for immunotherapy, tumors can develop resistance through selection of low-immunogenicity splice variants or impaired antigen presentation through mechanisms such as aberrant HLA-I splicing (98). Pharmacological spliceosome modulation with agents such as indisulam or MS-023 counteracts this resistance by restoring neoantigen presentation, and exhibits synergistic effects with anti-PD-1 therapy to enhance T-cell-mediated tumor clearance in preclinical models (97).

Mis-splicing of microexons, short exons typically regulated in tissue-specific patterns, represents an important disease mechanism that disrupts essential protein interaction networks, in autism spectrum disorder, CPEB4 microexon mis-splicing drives protein aggregation and functional impairment (99). While this mechanism has been primarily characterized in neurological disorders, similar processes likely operate in cancer biology, where microexon alterations may influence oncogenic signaling pathways and immune evasion capabilities, thereby creating opportunities for novel biomarkers and therapeutic strategies (98,99). These observations underscore splicing-derived neoantigens as promising biomarkers of response to immune checkpoint blockade response and as a basis for emerging approaches such as personalized cancer vaccines and combination therapies that target spliceosome activity is targeted (97,98).

Computational methods have been developed to enable systematic study of splicing heterogeneity with both bulk and single-cell RNA sequencing data (scRNA-seq). Deep learning models such as APARENT-Perturb can predict splicing from sequence and uncover key regulatory mechanisms including the CFIm complex, as well as clinically relevant splicing signatures associated with drug resistance (100). Integrating these models with large-scale genomic repositories such as TCGA enables systematic pan-cancer studies that associate specific splicing patterns with clinical outcomes.

Single-cell technologies provide an unprecedented resolution of splicing dynamics among cell populations in the tumor microenvironment (TME). Reference atlases such as Tabula Sapiens show cell-type-specific splicing of MYL6 and other genes, and highlight the required resolution to resolve complex cellular heterogeneity (101). In aplastic anemia, scRNA-seq enables detection of aberrant splicing in the rare population of hematopoietic stem cells, and can provide mechanistic hypotheses for clonal evolution and treatment resistance (102).

Recent work in computational biology has led to the development of advanced frameworks for integration of multi-omics data and the mapping of splicing regulatory networks. The model APARENT-Perturb relies on deep neural networks to predict the effect of genetic perturbations on splicing outcomes, and has uncovered functional interactions between regulatory elements, as well as sequence features influencing polyadenylation site selection (100). Integration of splicing data with other molecular measurements using tools such as CellPhoneDB enables systematic investigation of networks of cell-cell communication that underlie splicing heterogeneity in tissue microenvironments (101).

Splicing regulation can dynamically respond to metabolic and stress signals through adaptive systems such as the minor spliceosome (MiS), which is activated during cancer progression. In prostate cancer, a progressively dysregulated MiS, as evidenced by elevated U6atac snRNA levels, has been associated with disease progression and splicing reprogramming under metabolic stress (103). Cellular stressors, including androgen deprivation therapy and AR pathway inhibition, have been shown to further enhance the efficiency of minor intron splicing, a clear indication of how treatment-related stress can remodel the spliceosome to facilitate the development of therapeutic resistance (103). This dynamic regulation of splicing in turn selectively reprograms the splicing network to express additional antiapoptotic isoforms, firmly positioning MiS at the center of the adaptive response in cancer.

The U12-type spliceosome functions as a specialized regulatory system for processing stress and oncogenic signals via minor intron splicing. The structural analyses of the MiS point to MiS-specific components, such as SCNM110 and PPIL2, as key factors that ensure a stable catalytic core required for the accurate processing of genes involved in cell-cycle checkpoint and DNA repair pathways (104,105). The evolutionary conservation of this system and the mutational vulnerability of its key components underscore the importance of MiS in preserving cellular homeostasis under metabolic and therapeutic stress (103,104).

In conclusion, splicing serves as an integrator of metabolic and stress signals, and the U12-type spliceosome offers a compelling model of how an ancient system of gene regulation can influence the tumor adaptive response. Further work is needed to elucidate the direct links between metabolism and splicing as well as the epitranscriptomic layer of regulation, both of which still remain to be fully characterized in the current literature.

Small molecule splicing modulators are an emerging class of therapeutics designed to correct or exploit dysregulated pre-mRNA splicing in cancer. These molecules bind to core components of the spliceosome, the macromolecular machinery that mediates intron removal and exon ligation, thereby either directly altering splice site selection or modulating the kinetics of spliceosome assembly. One prominent target is the SF3b complex, a subcomplex of the U2 small nuclear ribonucleoprotein (U2snRNP) that is responsible for recognizing the branchpoint sequences during the early stages of spliceosome assembly. Inhibitors of SF3b, such as pladienolide derivatives including H3B-8800 and E7107, stabilize a conformation that prevents its interaction with pre-mRNA, leading to widespread IR or ES (108,110). The effect of these inhibitors is amplified in cancer cells harboring mutations in spliceosomal genes (for example, SF3B1, U2AF1, or SRSF2) since these cells exhibit heightened dependence on already compromised spliceosome function. Recent studies also highlight the role of nuclear condensates, which are phase-separated compartment enriched in splicing factors, in potentially concentrating splicing modulators selectively in cancer cells (111). In addition to SF3b-targeting molecules, several small molecules have been identified that modulate other components of the splicing machinery, such as SR proteins, or even target upstream RNA modification pathways, such as METTL3, which mediates N6-methyladenosine (m6A) methylation, a post-transcriptional modification known to regulate splicing efficiency and pre-mRNA transcript stability (112,113). Consequently, the diversity of mechanisms underlying the action of splicing modulators suggests their potential efficacy across a wide range of splicing-driven malignancies.

The therapeutic efficacy of SF3b inhibitors is exemplified in MDS and chronic lymphocytic leukemia (CLL), where recurrent SF3B1 mutations drive aberrant splicing and oncogenesis. H3B-8800, an orally bioavailable pladienolide derivative, selectively eliminates spliceosome-mutant cells by causing retention of short GC-rich introns in genes that encode spliceosome proteins and increase splicing stress (108). Preclinical studies in SF3B1-mutated pancreatic cancer and leukemia models have shown that H3B-8800 reduced viability by greater than 50% at nanomolar concentrations, with only mild effects in wild-type cells. Resistance to H3B-8800 was caused by mutations in other components of SF3b, such as PHF5A-Y36C, indicating on-target activity (108). In MDS, the SF3B1 mutations lead to a hypersensitivity to splicing perturbation, as these cells depend on residual activity of the spliceosome to produce transcripts critical for the survival of mutant cells (110). In early clinical trials in MDS, H3B-8800 treatment led to a reduction in blasts and an improvement in other hematologic parameters in patients with SF3B1-mutated MDS (108).

However, the Phase I first-in-human dose-escalation study of H3B-8800 (NCT02841540) completed dose escalation with defined dose-limiting toxicities, including ocular toxicity observed with prior pladienolide derivatives (for example, E7107). The trial identified reversible QTcF prolongation and gastrointestinal disturbances (diarrhea, nausea and vomiting) as common treatment-emergent adverse events, with the maximum tolerated dose established at 30 mg for a 5-days-on/9-days-off schedule and 14 mg for a 21-days-on/7-days-off schedule (107). Despite safety considerations, these findings validate SF3b as a therapeutically tractable node in spliceosome-mutant cancers and provide a roadmap for targeting splicing vulnerabilities in other malignancies.

Aberrant splicing also underlies resistance to targeted therapies, as observed in BRAF-mutant melanomas treated with vemurafenib. ~30% of resistant tumors express BRAF splice variants (for example, BRAF3-9), which lack the RAS-binding domain and dimerize independently of upstream signals (113). A C-to-G mutation in intron 8 of BRAF creates a cryptic branchpoint, promoting BRAF3-9 expression through enhanced recognition by mutant SF3B1 (113). The splicing modulators, including spliceostatin A (SSA), reverse the shift in isoform expression by inhibiting SF3b and thus re-establishing sensitivity to BRAF inhibitors. In vemurafenib-resistant melanoma xenografts, SSA treatment decreased BRAF3-9 levels by 60% and slowed tumor growth by 40%, showing that splicing modulation can overcome drug resistance (113). The splicing modulators and BRAF inhibitors are enriched in nuclear condensates containing MED1 and BRD4, which enhance the efficacy of the combination therapy (111). Parallel studies implicate RNA m6A methylation in regulating BRAF splicing, as METTL3 knockdown reduces BRAF3-9 expression and synergizes with vemurafenib (112). These insights highlight the dual utility of splicing modulators: As monotherapies in spliceosome-mutant cancers and as adjuvants to reverse adaptive splicing in drug-resistant tumors.

ASOs are short synthetic nucleic acids. Specifically, ASOs are single-stranded, chemically modifiable short nucleic acid sequences usually in the range of 18–25 nucleotides in length (114). They are designed to bind pre-mRNA and modulate splicing events by blocking splice-site recognition or altering exon inclusion. These molecules are chemically modified (for example, phosphorothioate backbones, 2′-MOE/cEt modifications) to enhance stability and binding affinity. For example, in HCC, ASO1-cEt/DNA targeting PKM pre-mRNA induced a splice switch from the oncogenic PKM2 isoform to tumor-suppressive PKM1, suppressing glycolysis and tumor growth in xenografts (115). Similarly, in KRAS Q61K-mutant cancers, ASOs disrupting exonic splicing enhancer motifs near cryptic splice sites induced ES, generating non-functional KRAS transcripts and sensitizing tumors to osimertinib (116). These studies underscore ASOs' ability to target splicing-dependent oncogenic drivers with high specificity.

A key design strategy is to target variants that mediate resistance to therapy. In gallbladder cancer, PTBP3 mediates the skipping of IL-18 exon 3, leading to the expression of a tumor-specific ΔIL-18 isoform that in turn is responsible for the suppression of the tumour-fighting activity of CD8⁺ T-cells. ASO4, designed to block the splice site of the ΔIL-18 isoform, was shown to restore expression of full-length IL-18 and also to induce improved tumor clearance by T-cells in vivo (117). Likewise, in B-cell tumors, two CD20 mRNA isoforms with different 5′-UTRs (V1/V3) have been shown to control protein translation. In these isoforms, splice-switching morpholinos selecting for the translation-competent V3 isoform raised CD20 levels, the increased CD20 expression resensitized rituximab-resistant cell lines (118). These examples demonstrate how ASOs can reverse resistance by reprogramming splicing toward nonpathogenic isoforms.

Clinical progress in solid tumors mirrors advances observed with nusinersen (a spinal muscular atrophy therapy). In ovarian cancer, BUD31 sustains antiapoptotic BCL2L12 expression by promoting exon 3 inclusion. BUD31-targeting ASOs induced ES, triggering nonsense-mediated decay of BCL2L12 mRNA and apoptosis in xenograft models (119). Meanwhile, in HCC, systemic delivery of a mouse-specific ASO (mASO3-cEt/DNA) redirected PKM splicing, inhibiting tumorigenesis without toxicity (115). Similarly, in lung cancer models, the addition of the MERTK ASO to XRT⁺anti-PD1 and XRT⁺ anti-CTLA4 profoundly slowed the growth of both primary and secondary tumors and significantly extended survival (120). Challenges remain in tumor-specific delivery, but chemical conjugates (for example, GalNAc for hepatic targeting) and lipid nanoparticles (LNPs) show promise.

Previous advances in delivery platforms have significantly enhanced the in vivo efficacy of ASOs by improving targeted delivery and splice-switching efficiency. GalNAc conjugates, which facilitate hepatocyte-specific uptake through asialoglycoprotein receptor-mediated endocytosis, have demonstrated a 10-fold improvement in gene silencing potency in murine liver models, highlighting their specificity for hepatic applications (121). LNPs, engineered to encapsulate splice-switching oligonucleotides, have achieved over 60% splice correction in tumor xenografts, as measured by functional recovery of target genes, underscoring their utility in oncology settings (122). Exosomes allow delivery of nucleic acids to neurons by exosomes, which can be functionalized with RVG peptides to target delivery to the brain. It has led to ~50% reduction in models of neurodegenerative disorders (123). These approaches allow for delivery to be more precise, and for the systemic levels of agents to be lower, reducing off-target effects and improving therapeutic indices.

However, the translation of ASO medicines into the clinic for the treatment of solid tumors is significantly hampered by challenges that arise during systemic delivery. This includes poor tissue penetration, limited uptake and poor transport of ASOs into and within cells. A key clinical failure for systemic delivery of ASOs was the SYNERGY Phase 3 trial of custirsen, a clusterin-targeting ASO, in combination with docetaxel and prednisone in patients with metastatic castration-resistant prostate cancer. It was observed that custirsen had no impact on overall survival in these patients, a finding that underscores the need to overcome delivery barriers through means such as ligand conjugates and nanocarriers (109).

Circular RNAs (circRNAs) are covalently closed RNA molecules formed by back-splicing of precursor mRNAs. CircRNAs are highly stable molecules that are resistant to exonucleases. CircRNAs are involved in the development of cancers by regulating the expression of target genes through microRNA sponging, protein interactions and transcriptional regulation (124,125). CircRNA-directed nanotherapeutics is a preclinical proof-of-concept approach that employs RNA interference technology in combination with a delivery system that integrates nanoparticles to target oncogenic circRNAs. CircRNA-directed nanotherapeutics utilizes siRNAs or ASOs encapsulated in biocompatible nanoparticles to enhance stability, targeted delivery and mitigate off-target effects relative to free RNA agents. This strategy holds promise to overcome issues such as rapid renal clearance and enzymatic degradation of free RNA, and has been shown to enable specific targeting of dysregulated circRNAs in tumors (126,127).

Recent studies have provided proof-of-concept evidence that circRNA-targeted nanomedicines can be effective in different cancer types in mouse models. In gastric cancer, lipid-polymer hybrid nanoparticles carrying siRNAs to knockdown circ_0008315 resensitized tumors to cisplatin by disrupting miR-3666/CPEB4 signaling. This nano-formulation exhibited 2.3-fold higher tumoral accumulation than free siRNA and significantly decreased tumor volumes in a patient-derived xenograft model (65.8 mm³ vs. 132.4 mm³ in free siRNA-treated mice; with no effects on organ function) (124). In HCC, poly(β-amino ester)-based nanoparticles carrying circMDK siRNA achieved (78% target gene knockdown) inhibited tumor growth by upregulating ATG16L1 via an m6A-associated mechanism. The nanoparticles were designed with pH-sensitive components, so that 92% of the siRNA payload could be released specifically in the TME, suggesting potential superiority over conventional chemotherapy treatment in an orthotopic tumor model (126). These examples underscore the potential, at a proof-of-concept stage, for nanocarriers to improve the pharmacokinetic properties of circRNA-targeting agents in a preclinical model.

There are other examples of preclinical validation of circRNA nanotherapeutics, whereby multifunctional nanoparticle designs provide supporting evidence (Table III). For instance, delivery of circRHBDD1 siRNA in PLGA-PEG nanoparticles effectively blocked the the IGF2BP2/PD-L1 axis and increased infiltration of CD8⁺ T cells by 4.1-fold in murine gastric cancer, achieving complete tumor regression in 40% of treated animals when combined with anti-PD-1 antibody (125). PLGA nanoparticles were also used to deliver si-cSERPINE2 to breast cancer cells in orthotopic mouse models. This treatment attenuated IL-6-mediated crosstalk between breast cancer cells and tumor-associated macrophages, and inhibited lung metastasis by 72% (127). Current early-phase clinical trials are beginning to validate such alternative strategies. For instance, a Phase I clinical trial of LNP-formulated circRNA inhibitors was conducted in solid tumors, and indicated a favorable safety profile, but was undertaken as an exploratory study and was not further validated in large animal studies.