Although the prime outcome of AS is the fold increase in functionally distinct proteins compared to the number of genes, AS also plays a significant role in the most fundamental biological processes: evolution, differentiation, and adaptation. AS is a source of evolutionary development, a determinant of organ, tissue, and cell characteristics, and part of cellular adaptation to a changing environment (2).

The concerted manner by which protein production is shifted from one isoform to another yields the regulatory characteristics of AS. Its preferences differ among tissues and developmental states and respond to extracellular signals in a dynamic manner that precedes or synchronizes with gene transcription. In parallel to the dependency of intracellular processes on transcriptional activation, cellular events emerge from the shift in protein isoform ratios. How splicing events are concerted and what network cascades occur is a field of active research emphasizing health disorders and malignancies (7). The regulation of splicing events depends on both cis-acting regulatory sequences, located in introns or exons, and trans-acting splicing factor proteins that can strengthen or weaken the spliceosome’s recognition of the splice sites (8). These regulatory proteins belong to families of RNA-binding proteins, such as arginine–serine-rich (SR), heterogeneous nuclear ribonucleoprotein (hnRNP), and RNA-binding motif (RBM) proteins (9). They recognize specific regulatory sequences and enhance or inhibit the recognition of neighboring splice sites by the core splicing machinery (7). The expression level of the regulatory proteins is tissue- and state-specific (10), and they are subjected to regulatory splicing themselves (11).

From the evolutionary point of view, alternative splicing varies significantly among species. The insertion of multiple introns that separate exons has derived from ancestral genes and predated AS in eukaryote development. The option to skip exons was enabled by DNA mutations that may have resulted in splice sites with weaker binding affinity for spliceosomal components such as U1 small nuclear ribonucleoprotein (snRNP) (12). While the emergence of alternative splice sites contributed to protein diversity and is partly unique to a species, particularly when changing the reading frame, comparative genomics indicates that sequences that regulate RNA binding proteins are conserved and shared (13, 14).

Over 200,000 identified isoforms are reported in genome databases, and the majority of them lack functional annotation. Some well-studied examples show how events of retention or exclusion of specific domains may change protein cell localization, membrane anchorage, shedding of ectodomains, mRNA stability, and translational efficiency. The molecular alterations that emerge from AS may occur without any change in the level of the general gene’s transcript or before a change (15). Furthermore, the translational changes in reading frames may produce diverse translation outputs (13) and even insert poison exons, resulting in nonsense-mediated mRNA decay (NMD) and diminished protein levels (16). We can conclude that alternative splicing is timed and regulated in a manner that is not necessarily dependent on active simultaneous gene transcription.

A prevalent splicing pattern observed in immune receptors gives rise to soluble isoforms that lack membrane anchorage and are secreted into the extracellular space. These soluble receptors may regulate signaling cascades, which differ from those initiated by their parental receptor (Table 1). Most prominently, the soluble receptors can function as a decoy of their corresponding ligands and compete with their constitutive, membrane-bound forms (26–35). The ratio between the membranal and the soluble isoforms of a receptor can remain fixed (26, 36). However, it might change depending on the cell’s metabolic, functional, or differentiation state (37, 38). Diverting the pre-mRNA splicing towards the soluble isoforms results in reduced expression of the membrane-bound receptor and can even negate its cellular effect. In a different context, the soluble receptors may have agonistic effects (28, 39) and initiate reverse signaling by binding to other receptors (40–42). For example, glucocorticoid-induced tumor necrosis factor receptor (GITR) ligand that is expressed on plasmacytoid dendritic cells prompts a reverse signal that initiates noncanonical NF-kappaB-dependent induction of indoleamine 2,3-dioxygenase upon binding to soluble GITR. This leads to the tryptophan catabolism immunoregulatory pathway (43). In addition, soluble isoforms have been documented to bind with their ligand to distinct membranal partners, activating trans-signaling pathways (44, 45). Furthermore, some soluble receptors stabilize their ligand configuration or alter their biodistribution (31, 45–47). Another typical example of important alternative splicing of immune cells is the removal of the hydrophobic transmembranal segment of the B-cell receptor to form a secreted immunoglobulin (48). Interestingly, soluble receptors may exert different functions (i.e., agonistic and antagonistic) depending on their concentration (45, 46, 49) ( Figure 2 ).

The common splicing patterns that lead to the generation of soluble isoforms of membranal receptors include (1) Transmembrane domain (TMD) skipping: In the process of alternative splicing, skipping of the transmembrane encoding exon results in the creation of a soluble product encompassing both the intracellular and extracellular domains (26, 33, 50–55); (2) Alternative terminator: a shortened soluble isoform is encoded by a sequence that includes a mutually exclusive exon containing an alternative polyadenylation site, or by alternative splicing that results in frameshift and premature stop codon (48, 56, 57). As a result, the translated proteins include only the extracellular domain, lose their membrane anchorage and become soluble (27, 58, 59).

It should be noted that alternative splicing is not the sole mechanism that creates soluble receptors. These segments can also be made by proteolytic cleavage of extracellular domains by proteases in the extracellular matrix ( Figure 3 ). However, unlike AS, the intracellular domain (ICD) of a cleaved receptor remains anchored and theoretically may retain its effect. The impact of a truncated signaling domain is diverse or unknown. Typical examples of receptors that utilize both mechanisms to produce their soluble formats are cytokine receptors, including the TNF and TNFR superfamily (59, 60). In addition, some immune receptor genes lack the transmembrane domain and are, therefore, constitutively expressed as soluble receptors with linked intracellular and extracellular domains. They mainly function as decoy receptors (61). For example, decoy receptor 3 (DcR3, TNFRSF6B) is a secreted TNFR superfamily member that lacks a transmembrane domain. DcR3 can interrupt FAS-FASL interaction by binding FASL and inhibiting FASL-induced apoptosis (62).

The immunoglobulin (Ig) superfamily is a large group of proteins with a common Ig domain. The Ig superfamily is critical in the immune response networks (63). Some Ig superfamily receptors can be translated to a soluble form by an alternative splicing process (64, 65). Among these are CTLA4 (39, 58), CD83 (42, 66), and LAG3 (66). Here, we will focus on two specific examples: The programmed cell death receptor PD-1 and the B cell receptors (BCRs) that convert, after splicing, into immunoglobulins (antibodies).

The TNFRSF comprises trimeric receptors made of three homologous molecules that initiate signaling pathways involved in inflammation, proliferation, differentiation, cell migration, and induction of cell death (72). Like the Ig superfamily, TNFRSF members share similar splicing patterns that result in soluble isoforms.

FAS (CD95, TNFRSF6) is one of the best-known members of the TNF receptors superfamily. It is abundantly expressed in many tissues, including the gastrointestinal tract, the respiratory system, and lymphoid tissues (73–75). FAS is mainly known for its pro-apoptotic pathway activation following FAS ligand (FASL) binding (76). However, it also has other functions, e.g., it takes part in the differentiation of naïve T cells to memory cells (77). FAS is robustly expressed on T-cells and has an apoptosis-inducing role during T-cell development (78) and activation (79–81). A specific alternative splicing event is the skipping of exon 6, which encodes the transmembrane domain of FAS, resulting in a soluble form of the FAS receptor (52, 53). The FAS exon 6 skipping mechanism has been studied extensively. It has been shown that many splicing factors can regulate this event, among them TIA-1 (82), PTB (82), HuR (83), hnRNP A1 (84), SRSF4 (85), SRSF7 (86), and SRSF6 (87). Similarly to PD-1, the soluble FAS receptor competes with membrane-bound FAS for FASL ligation, thereby limiting FAS signaling (30, 52). Bajgain et al. described the ability of secreted FAS extracellular domain to enhance CAR T-cell antitumor activity against a FAS-ligand-expressing tumor (88).

TGFβ (transforming growth factor beta receptor) TGFβ is of special interest because it controls immunity via a rich network of cells and mediators, with the end result being immune evasion of the cancer tissue. The biological functions of TGFβ are mostly mediated by the monomeric, soluble form of the protein. The monomer is cleaved by proteases in the Golgi complex and later released from glycoproteins that ligate it in a non-covalent manner (89). TGFβ enhances the expansion of regulatory T cells (Tregs), the inhibition of NK and effector T cells, and the induction of immune suppressive cytokines including IL-4 and IL-10. Active TGFβ exerts its effect via receptors that activate SMAD transcription factors, a family with hundreds of regulatory elements. Tumors exploit TGFβ to induce a supportive stroma that weakens the immune response by acting as a mechanical barrier and expressing inhibitory membranal ligands, such as PD-L1 (90, 91). In a series of patients with gynecological cancers who received immune checkpoint inhibitors, a high TGFβ expression score correlated with treatment failure and reduced survival (92). The type II receptor for TGFβ has a splicing variant which lacks the transmembrane domain, and exert a higher binding affinity to the three sub-types of TGFβ. By doing so, it competes with the natural ligands and reduces fibrotic pathology (93).

In addition to the Ig and TNFR superfamilies, members of other immune receptor families can generate soluble forms through alternative splicing, including type I cytokine receptors IL6Rα (54, 94), IL-4Rα (57), and IL-7Rα (33). Another example is the common γ chain IL-2, 7, and 15 cytokine receptors, for which skipping exon 6 encoding the transmembrane domain results in a frameshift and a premature stop codon. The resultant proteins contain only the extracellular domain. The soluble IL-2 and IL-7 receptors were reported to impair T-cell signaling and function (27).

Neurodegenerative diseases are a group of disorders caused by the gradual loss of neuronal cell function or structure. Strikingly, splicing-related mutations are one of the leading causes of many neurodegenerative diseases (119). The most investigated neuronal disorder instigated by splicing is spinal muscular atrophy (SMA). Nonetheless, most neural pathologies, such as early-onset Parkinson’s (119–121), Alzheimer’s disease (122), familial dysautonomia (123), and Amyotrophic Lateral Sclerosis (ALS), could evolve from splicing mutations (124, 125).

Given the significant number of neurodegenerative disorders caused by mutations impacting RNA splicing, it is not surprising that there have been numerous efforts to investigate the use of splicing-editing techniques as a treatment option for these conditions. For example, antisense oligonucleotides (ASOs) are being widely researched for their potential use in treating SMA, DMD, and ALS (126). The use of ASOs to manipulate alternative splicing is further discussed elsewhere in this review. The use of CRISPR-Cas9 to affect alternative splicing has been suggested for Duchenne muscular dystrophy (127), and spliceosome-mediated RNA trans-splicing (SMaRT) (128) has been tested in Huntington’s disease (129), DMD (130), and Alzheimer’s disease (131).

Splicing-related mutations in cancer can be grouped into three categories: 1-those affecting the core spliceosome complex, resulting in new isoforms; 2-those impacting splicing factors, affecting the expression levels of multiple isoforms; and 3-those affecting splicing recognition sites, altering the expression level of a single gene or creating new isoforms ( Table 2 ; Figure 4 ).

Splicing factors mutations are particularly prevalent in myeloid neoplasms; for example, SF3B1, that increases anti-apoptotic isoforms, enhances tumor proliferation and progression, and is associated with poor survival of patients (134, 140, 141, 180, 181). U2AF1 is another splicing factor mutated in myeloid malignancies that drives altered splicing preferences. Intronic mutations are more frequent than exonic, and a third of somatic mutations in the exon-intron boundary are associated with splicing changes. If a mutation occurs in a 5’ or 3’ splicing site, there is a greater than 50% chance of it leading to a splicing shift (182).

Splicing can be employed as a cancer treatment approach in various forms: using single-stranded oligonucleotides to change the splicing of specific genes and switch between oncogenic and tumor-suppressing forms, as has been demonstrated for the BCL gene (67); regulating specific splicing factors through drugs that directly impact them, such as blocking SF3B1 (68); or by attacking the pathway which the mutant splicing factor exploits. Thus, tumors with driver mutations in SF3B1 or U2AF1 may be vulnerable to NMD inhibition (68–72). Some widely used therapies, such as camptothecin and cisplatin, have been found to impact RNA splicing, potentially contributing to their efficacy (73–75).

As discussed in the following paragraph, recent attention has focused on the generation of neo-antigens by including erroneous transcripts. However, altered splicing and the emergence of usually unexpressed isoforms independently impact tumor immunogenicity. These effects often hinder the anticancer immune response. For example, HLA tumor-enriched alternative splicing events occur in 10-30% of lung and breast cancers, affecting MHC expression. When HLA expression is inconsistent, the ability of tumor epitopes to be presented and recognized is diminished or completely lost (183). In ovarian cancer, certain splicing factors, such as BUD31, SF3B4, and CTNNBL1, may indirectly support immune evasion (184). This immune escape may involve increased PD-L1 expression and primary resistance to PD-1 inhibitors. Such mechanisms are seen in clear renal cell carcinoma, where an exon-including splicing event in the chromatin remodeling gene PBRM1 contributes to immune evasion (185).

While reports indicate that altered splicing isoforms contribute to tumor immune evasion, splicing alterations are now attracting significant interest as a source of cancer antigenicity. This interest stems from the potential of AS to drive isoforms that include retained intronic sequences. These intronic transcripts, in turn, may form neoantigens—peptide sequences that have not had the opportunity to tolerize the immune system. Such newly transcribed sequences hold the potential for generating protective immunity and improving clinical responses to immune checkpoint inhibitors (61–64).

Several pharmacological compounds have been used that either degrade splicing factors, disrupt spliceosome assembly, or inhibit nonsense-mediated decay (186, 187). One example is indisulam, an anticancer sulfonamide that generates aberrant transcripts. Interestingly, indisulam does not directly inhibit cancer growth; instead, it triggers a T-cell response against cryptic sequences from abnormal RNA, which impedes tumor progression. Other splicing-disruptive compounds, such as pladienolide B and H3B-8800, are currently being evaluated in experimental systems and clinical trials for myeloid neoplasms. Predicting the effect of splice manipulation on the tumor microenvironment is challenging, but as will be discussed, induction of soluble ectodomains from immune-modulatory receptors may interfere with immune checkpoint inhibitors. Soluble PD1, for instance, may saturate PD-1 blocking antibodies and reduce their availability to rescue exhausted antitumor T cells (188).

The advancement of chemical modifications of nucleic acids marked a significant milestone in the clinical application of this compound class. A key outcome was the development of splice-switching antisense oligonucleotides (SSOs), designed to modify alternative splicing patterns and enhance exon skipping. Specifically, chemical modifications that reduce RNaseH activity form a stable DNA-mRNA hybrid, preventing subsequent RNA degradation (206). These SSOs can be directed towards splice-site sequences, hindering and redirecting the spliceosome to an alternative splice site in the subsequent exon (207, 208) ( Figure 6 ).

In addition to splicing alterations via complementation to splice sites, SSO can modify splicing by targeting splicing enhancers (ESE, ISE) or silencers (ESS, ISS) (209, 210). These interventions may interrupt splicing by inhibiting linkage to splicing factors, leading to exon exclusion or inclusion. In addition to whole exon skipping, the pre-mRNA splicing modulation can result in intron retention, alternative 5′ and 3′ splice sites, alternative promoter, or alternative polyadenylation sites (209).

Finding SSO-targetable splicing motifs is not trivial. A systematic scan of the exon of interest is necessary to spot the precise sequence, which the SSO should complement to alter the wild-type splicing pattern.

Two types of chemical modification are currently used for FDA-approved drugs: 2′-O-methoxyethyl (2′-MOE) nucleosides with phosphorothioate (PS) backbone and phosphorodiamidate morpholino oligomers (PMO) with a N, N-dimethylamino phosphorodiamidate backbone ( Figure 7 ).

2’-MOE belongs to a group of modifications in the 2’O of the furanose ring of the nucleic acid. Other prevalent modifications are 2′-O-methyl (2′-OMe), locked nucleic acid (LNA), and SSOs containing 2′-constrained ethyl (2′-cEt). Alongside the RNaseH1 resistance, the 2’O modifications increase the SSO affinity (211). High affinity is attributed to higher potency, longer half-life, and less immune-provoking properties (212, 213). 2’MOE modifications are usually accompanied by switching Oxygen in the backbone to Sulfur (PS). This switch decreases the SSO affinity but improves the resistance to nuclease activity (214, 215) and molecular binding to proteins – resulting in reduced kidney clearing (216) and improved uptake by target cells (217–219).

In PMO (220), a morpholino ring replaces the furanose ring. In addition, the negatively charged backbone is replaced by a N, N-dimethylamino phosphorodiamidate backbone. As a result of these changes, the SSOs have higher in vivo tolerance but faster kidney clearance, which requires a higher dosage (221, 222).

Although these are the main modifications currently used for SSO drugs, recent publications have shown how additional chemical modifications can further improve splicing modulation. For example, Langner et al. synthesized a hybrid that combines PMO modification with a PS backbone, which exhibits higher efficiency than 2’-MOE modification with the same backbone (223).