The role of alternative pre-mRNA splicing in cancer progression

The mitotic cell cycle is a process in which duplicated chromosomes are separated into two cells to obtain two genetically identical daughter cells (Fig. 2A). Chromosomes are replicated through two stages, interphase and M phase. The cell cycle is controlled by complex mechanisms such as cell cycle checkpoints that monitor the appropriate cell size, complete chromosome replication, and correct cell segregation for timely and accurate cell division [26]. The main mechanism of action of the cell cycle checkpoint is to regulate the activity of cyclin-dependent kinases (CDKs), a family of serine/threonine protein kinases. Cyclin/CDK complexes are formed and activated at various stages of the cell cycle, and they control cell cycle progression through phosphorylation of downstream targets such as the tumor suppressor retinoblastoma (Rb) [27].

In keratinocytes, increased eIF2α phosphorylation by UVB enhanced translation of cyclin dependent kinase inhibitor 1 A (CDKN1A, also called p21) transcript variant 4 through a mechanism mediated in part by upstream open reading frames (ORFs) situated in the 5′-leader of CDKN1A mRNA [28]. eIF2α phosphorylation protects cells from stress, such as UVB irradiation, by regulating cell cycle control and determining cell fate through a specific splicing variant of CDKN1A that promotes G1 arrest and subsequent DNA repair. In addition, a mutation that destroys the splice site of CDKN1A has been found to induce the skipping event of exon 2 and, interestingly, cause transcript fusion between CDKN1A and RAB44, which are different genes [29].

Cyclin dependent kinase inhibitor 2 A (CDKN2A), located on human chromosome 9, encodes the p16^INK4A and p14^ARF proteins. A splice site mutation in CDKN2A, found in a family with melanomas, neurofibromas, and multiple dysplastic nevi, was reported to cause the skipping of exon 2, which encodes more than 50% of the p16^INK4A and p14^ARF proteins [30]. In addition, splice site variation of CDKN2A was found by sequencing of 167 melanoma-prone families [31]. The AGgt-to-AGtt and AGgt-to-ATgt mutations in CDKN2A disrupted the normal splicing of exon 2. The AGgt-to-CGgt mutation of exon 1 disrupted splicing and affected the function of p16^INK4A, confirming that the splice site mutation plays an important role in melanoma susceptibility.

Cyclin E1 encoded by CCNE1 controls the G1/S transition. In the Ccne1 splicing variant IN3 (ORF shifted from exon 2 to exon 3 due to alternative 5′ss selection), exon 4 skipping and exon 5 skipping are associated with retarded proliferation in murine hepatocellular carcinoma [32]. Isoforms with partial loss of exons 3 and 8 and complete loss of exons 4, 5, 6, and 7 are localized in the cytoplasm due to the lack of a nuclear localization signal. This Ccne1 isoform is predominantly expressed in G0 in hepatocytes and forms inactive complexes to sequester CDK2 in the cytoplasm, delaying cell cycle re-entry of hepatocytes after G1 arrest.

Human cyclin D1 (CCND1) splicing occurs at the exon 4/intron 4 boundary. When splicing occurs at this site, the cyclin D1a protein is produced; when splicing fails, the cyclin D1b protein, which contains a part of intron 4 instead of exon 5 and terminates prematurely, is produced. Cyclin D1b splicing is increased by G/A870 polymorphism and SRC-associated in mitosis of 68 kDa (SAM68) [33, 34]. Cyclin D1b is upregulated in various carcinomas, including prostate cancer, and exhibits a higher carcinogenic potential than cyclin D1a because of its stronger nuclear localization [35‐37]. The cyclin D2 (Ccnd2) splice variant found in the mouse heart contains exon 1 and the longer exon 2, which generate 20 novel residues in the C-terminal [38]. This isoform aggregates in the endoplasmic reticulum (ER), Golgi, and lysosomal compartments and functions as a negative cell cycle regulator by aggregating and sequestering several cell cycle proteins. The 17 kDa truncated cyclin D2 isoform localizes to the cytoplasm and interacts with CDK4; however, the complex is unable to phosphorylate the target pRb [39].

The long isoform of CDK2 with 48 residues inserted owing to partial insertion of intron 5 has approximately half the specific activity compared to the normal form of CDK2 when forming a complex with cyclin A [40]. Poly(C)-binding proteins (PCBPs) bind to C-rich PPT in intron 4 and promote inclusion of CDK2 exon 5 [41]. Exon 5 exclusion of CDK2 dramatically reduces CDK2 protein expression, affecting cell cycle kinetics.

E2F transcription factor 1 (E2F1) is a transcription factor that regulates the S phase transition from G0/G1. Intron 5 retention and exon 6 skipping variants were identified in the central nervous system of rats, and these were terminated early, resulting in C-terminal truncation [42]. E2F transcription factor 3 (E2F3) produces isoforms called E2F3a and E2F3b owing to an alternative promoter [43, 44]. E2F3a is tightly regulated by cell growth and is expressed only at the G1/S boundary, whereas E2F3b is detected throughout the cell cycle and expressed in both quiescent and proliferating cells. Additionally, in quiescent cells, the E2F3b protein binds to Rb, resulting in a predominant E2F-Rb complex [45].

In RB1, the G1/S checkpoint regulator, 27% of coding mutations disrupted splicing [46]. Of the RB1 exon splicing mutations, 58% were blocked primarily at the A complex converting to the B complex, and 33% at the B complex. Mutation-induced RB1 exon loss or intronic sequence exonization induced premature termination, resulting in the restriction of RB1 expression [47, 48]. Inhibition of RB1 expression induced a high expression of p16^INK4A. In addition, the RB1 splicing mutation is associated with prognosis and low penetrance in patients with non-small cell lung cancer (NSCLC) or retinoblastoma [47, 49].

The tumor suppressor p53 induces cell cycle arrest at multiple stages, including G/S and G2/M checkpoints, in situations such as DNA damage. The use of an alternative 3′ss located in tumor protein P53 (TP53) intron 6 results in the p53Ψ isoform lacking major parts of the DNA-binding domain, nuclear localization sequence, and tetramerization domain [50]. The p53Ψ isoform is not capable of DNA binding and transactivation, but it induces mesenchymal-like characteristics and improves motility and invasion of normal and malignant cells. TP53, a gene encoding the tumor suppressor protein p53, called “guardian of the genome”, has been reported to have splice site mutations in patients with various cancers, including colorectal cancer [51, 52]. Exon 4–intron 4 junctions mutations identified in pediatric adrenocortical tumors induce erroneous splicing, resulting in protein instability, altered intracellular localization, and loss of function [53]. Exon 6-truncating mutants, similar to the p53Ψ isoform, lack transcriptional activities and the ability to respond to DNA damage, as well as pro-tumorigenic functions that promote cancer cell proliferation, survival, and metastasis [54].

The MYC oncogene family consists of three members: MYC, MYCN, and MYCL, which encode c-Myc, N-Myc, and L-Myc, respectively (Fig. 2B). Myc is a basic helix-loop-helix and leucine zipper (bHLH-LZ) transcription factor that regulates approximately 15% of the total transcriptome [55, 56]. Thus, Myc proteins mediate various biological processes including cell growth and proliferation, apoptosis, differentiation, cell cycle, and metabolism [57]. Myc promotes transcription by forming a heterodimer with Myc-associated factor X (MAX) through the bHLH-LZ domain required for DNA-protein interaction. Unlike Myc, which forms a heterodimer only with MAX, MAX forms a homodimer or binds to the MXD, MGA, and MNT proteins [58]. This process induces an antagonistic effect on the Myc family by sequestering MAX from Myc.

MAX with exon 2 inclusion was considerably more effective at binding homodimeric DNA than exon 2 skipping of MAX [59]. In addition, this isoform induced a decrease in Myc expression, slowed growth, and accelerated apoptosis during growth factor deprivation. In another MAX isoform, the basic region, helix 1, and the loop of the helix-loop-helix region were deleted [60]. Since this MAX isoform cannot bind to E-box Myc site DNA because the basic region does not exist, it appears to function as a dominant negative regulator. The MAX protein with a C-terminus truncated by intron 4 retention was shown to retain its ability to bind to the CACGTG motif in complex with c-Myc; however, it was located in the cytoplasm [61]. In addition to the wild type of MAX, hypoxia induced intron 4 retention and alternative splicing including cassette exon located within intron 4 [62]. The isoforms produced by intron 4 retention are highly destabilized by 36 isoform-specific amino acids, which destabilize heterologous proteins. The variant containing the cassette exon located within intron 4 is degraded by nonsense-mediated mRNA decay (NMD), and both isoforms play a role in downregulating the wild-type MAX isoform.

MAX dimerization protein 3 (MXD3) has a variant that uses exon 6 as the last exon and a variant that uses exon 7 instead of exon 6 as the last exon [63]. These two splice variants are most likely a result of alternative polyadenylation. According to an analysis of The Cancer Genome Atlas (TCGA) data, exon 7 mRNA was expressed at higher levels in normal cells than in cancer cells, whereas exon 6 mRNA was expressed at higher levels in cancer cells. The exon 7 inclusion isoform containing a considerably longer 3′UTR has a greater reduction in protein expression compared to the exon 6 inclusion isoform. In addition, the exon 7 inclusion isoform undergoes phosphorylation, and it is localized throughout the nucleus; in contrast, the exon 6 inclusion isoform does not appear to be phosphorylated, and is mainly confined to the nuclear foci.

The NRF2 pathway regulates transcriptional responses of genes important for oxidative and electrophilic stress responses (Fig. 2C) [64‐66]. Kelch-like ECH-associated protein 1 (Keap1), a substrate adaptor for a cullin-3 (Cul3)-based ubiquitin ligase, inhibits the transcriptional activity of nuclear factor-erythroid factor 2-related factor 2 (Nrf2) by promoting ubiquitination and proteasomal degradation of the transcription factor Nrf2 under basal conditions [67‐70]. Thiols of Keap1 cysteines are modified by oxidants, electrophiles, and Nrf2 activators, and Nrf2 is dissociated from Keap1 [65, 71]. The dissociated Nrf2 translocates to the nucleus without ubiquitination and proteasome degradation, forming a heterodimer with small musculoaponeurotic fibrosarcoma oncogene homologue (sMAF) [72, 73]. The Nrf2/sMaf heterodimer binds to the antioxidant response element (ARE) and induces the transcription of a battery of antioxidant and detoxification genes [74‐76].

Exon 2 or exon 2 + 3 skipping of NFE2 like BZIP transcription factor 2 (NFE2L2), the gene coding for Nrf2, was found in some NSCLCs and patients with head and neck squamous carcinoma (HNSC) [77]. These abnormal transcript variants lack the Keap1 interaction domain, resulting in loss of interaction with Keap1. Thus, Nrf2 is stabilized, resulting in an Nrf2 transcriptional response and Nrf2 pathway dependence. NFE2L2 can be spliced using a second alternative promoter (P2) that exists downstream. Because these P2 transcripts start translation using a different AUG than those that use the upstream promoter (P1), they miss the part of exon 1 that encodes the 16 amino acids and thus have a shorter N-terminus. The ΔN-Nrf2, a protein of the NFE2L2-P2 transcript, is more stable and abundant in cells in the absence of stress owing to its impaired binding to Keap1 [78]. Tert-butylhydroquinone (tBHQ) induces electrophilic stress, and the full-length protein isoform of the NFE2L1-P1 transcript was strongly detected in the nucleus, whereas ΔN-Nrf2 translocation from the cytoplasm to the nucleus was not observed. Further studies are needed to determine the function of ΔN-Nrf2 in the cytoplasm, in addition to its transcriptional activity, and whether this function is related to carcinogenesis.

Keap1 splice variants were found in human highly-metastatic hepatoma (MHCC97H) cells and other cell lines. This Keap1 splice variant (Keap1ΔC) lacks exons 4 and 5, resulting in the loss of 180 amino acid residues [79]. Since the missing amino acid residues are located in most of the double glycine-repeat (DGR) domain and the adjacent C-terminal region essential for interaction with Nrf2, these Keap1 isoforms retain no or little ability to inhibit Nrf2. Therefore, the isoform in which exons 4 and 5 are skipped acts as a dominant-negative competitor of intact Keap1 because of its antagonist effect on Keap1-mediated turnover of Nrf2.

The PI3K/Akt pathway is involved in several cellular physiological processes such as cell cycle, growth, proliferation, survival, and autophagy (Fig. 2D). PI3K activation is stimulated by various oncogenes and growth factor receptors, resulting in the conversion of phosphatidylinositol (3,4)-bisphosphate (PIP2) to phosphatidylinositol (3,4,5)-trisphosphate (PIP3) [80]. PIP3 serves as a plasma membrane docking site for the recruitment and activation of several pleckstrin homology (PH) domain-containing proteins such as AKT, SGK, and PDK1 [81]. Phosphorylation of Akt Thr308 by PDK1 at the membrane leads to partial activation of Akt, and Ser473 phosphorylation by a complex involving mTOR/Rictor (TORC2) results in full activation of this enzyme. Activated Akt phosphorylates downstream effectors, including GSK3β, FoxO, MDM2, and BAD, in the cytoplasm and nucleus, triggering various biological processes such as cell cycle, apoptosis, and glucose metabolism. Thr308 and Ser473 of Akt are dephosphorylated by PP2A and PHLPP1/2, respectively, and PTEN dephosphorylates PIP3 to inhibit Akt activity [82‐84]. Akt causes TSC2 phosphorylation and inactivation and induces RHEB to activate the mTOR/Raptor (TORC1) complex, which affects mRNA translation, nucleotide and lipid synthesis, cell growth, and autophagy.

PI3K is a heterodimeric enzyme composed of a p110 catalytic subunit and a regulatory subunit. The phosphoinositide-3-kinase regulatory subunit 1 (PIK3R1) encodes the PI3K regulatory subunits p85α, p55α, and p50α, with the same C-terminal but different N-terminals as a result of alternative splicing. The p85α protein with the longest N-terminal region can form a homodimer through SH3 domain-PR1 (proline-rich motif 1) and BH-BH domain interactions [85‐87]. Additionally, p85α binds to PTEN through the N-terminal SH3-BH region, which is absent in p55α and p50α, and enhances enzymatic activity. Through this process, p85α plays a role in maintaining the balance of PI3K signaling by regulating not only p110-PI3K but also PTEN-PI3K. Stat3 directly binds to the p55α and p50α promoters in vivo to induce their expression, and the overexpression of p55α or p50α reduced the level of activated Akt [88]. PIK3R1 splice acceptor and donor sites mutations in exon 11 induce exon 11 exclusion, resulting in a protein lacking a portion (Δ434_475) of the iSH2 domain [89, 90]. The protein thus produced has been shown to interfere with the regulation of p110δ, resulting in increased p110δ signaling and triggering activated phosphoinositide 3-kinase δ syndrome (APDS) type 2, which causes primary antibody deficiency, senescent T cells, developmental abnormalities within the T lymphocyte compartment, and immune dysregulation [91‐95].

The isoform generated by partial insertion of intron 8 of PTEN lacks part of the C2 domain, C-tail, and PDZ-binding domains [96]. Consequently, it induces reduced cell migration, adhesion, and enhanced apoptosis, which, similar to the full-length isoform, acts as a tumor suppressor. Additionally, studies have reported variants containing segments (3a, 3b, 3c; 5a, 5b, 5c) of different sizes of intron 3 or intron 5 and variants in which a part of exon 5 or all of exon 6 were excluded [97, 98]. The PTEN variant was regulated by p53 and differentially expressed in heritable breast cancer syndrome, sporadic breast cancers, and Cowden syndrome compared with the normal tissues. The PTEN 5a isoform functions similar to full-length PTEN by reducing cyclin D1 promoter activity and Akt phosphorylation; in contrast, PTEN 5b and 5c appear to exert the opposite function by decreasing cyclin D1 promoter activity. Mutations in the splice junction or intron region of the PTEN gene have been identified in patients with Cowden syndrome and Bannayan Zonana syndrome, an autosomal dominant genetic disorder characterized by hamartomas [99, 100]. These mutations caused either out-of-frame skipping of an entire exon or activation of cryptic splice sites, resulting in partial intronic sequence inclusion, premature transcript termination, or polyadenylation within exon 8. Samples with splicing changes due to these mutations exhibited lower PTEN protein expression and higher Akt phosphorylation compared with samples without splicing changes even with mutations; however, p-ERK1/2 did not increase. These findings suggest that PTEN variants may contribute to the pathogenesis of various diseases and may serve as diagnostic markers.

Consisting of the Ras-Raf-MEK-ERK signaling cascade, the mitogen-activated protein kinase (MAPK) pathway is an essential cellular network for inter- and intra-cell communication that transmits, amplifies, and integrates signals from various stimuli, such as growth factors, hormones, tumor-promoting substances, and differentiation factors (Fig. 2E). This pathway activates transcription factors and regulates genes that control cell proliferation, differentiation, angiogenesis, inflammation, development, apoptosis, integrin signaling, and migration. Ras small GTPase, the first component of MAPK signaling, contains three genes, HRAS, NRAS, and KRAS, that encode four RAS proteins (HRas, NRas, KRas4A, and KRas4B) [101]. The Raf family comprises three genes (ARAF, BRAF, and RAF1), and the MEK family, five genes (MEK1, MEK2, MEK3, MEK4, and MEK5). When the ligand binds to the tyrosine kinase receptor, the guanine exchange factor is recruited and activated, leading to the exchange of GDP with GTP in Ras. Activated Ras induces Raf activation to form Ras homo- or heterodimer. Active RAF dimers recruit MEK and subsequently activate ERK. Activated ERK1/2 alters gene expression in cells by phosphorylating multiple substrates and regulating various transcription factors. Conversely, when Ras becomes inactive GDP-bound state by GTPase activating protein, downstream signaling is turned off.

HRAS exon 2 has a relatively weak 3′ss; therefore, inclusion and exclusion are regulated by the balance of positive and negative splicing regulatory factors such as SRSF2 and hnRNPF/H [102]. Therefore, exon 2 mutation identified in attenuated patients with Costello syndrome have been shown to simultaneously interfere with exonic splicing enhancer function and exonic splicing silencer generation, causing exon 2 skipping. These splicing abnormality abrogated the production of HRAS protein and inhibited cancer cell proliferation. Since HRAS exon 2 inclusion may also affect the onset of Costello syndrome and carcinogenic potential, splice switching oligonucleotides that induce exon 2 exclusion can serve as a suitable strategy for therapy.

KRAS generates K-Ras4A and K-Ras4B according to the use of the alternative fourth exon. Exon 4 encodes HVRs, a membrane binding and targeting motif, and K-Ras4A and 4B are differentially expressed in mouse development and adult tissues [103, 104]. K-Ras4A is palmitoylated, whereas K-Ras4B lacks a palmitoylation site [103‐105]. Thus, in addition to the CAAX motif, K-Ras4A has a dual membrane-targeting motif, as a site of palmitoylation and bipartite polybasic region exists at its C-terminus. Therefore, K-Ras4A plays an important role in K-Ras-induced tumors, and the mechanism by which K-Ras4A accesses the plasma membrane differs from that of K-Ras4B [106]. Additionally, unlike K-Ras4B, K-Ras4A does not bind to the cytosolic chaperone δ-subunit of cGMP phosphodiesterase type 6 (PDE6δ); therefore, a significant difference has been observed in their subcellular trafficking. K-Ras4A has been shown to respond to hypoxia and K-Ras4B to ER stress; KRAS4A splicing is controlled by the DCAF15/RBM39 pathway [107]. Cells with a low KRAS4A/KRAS4B ratio have shown higher sensitivity to cancer treatment drugs and an association with high KRAS signaling and a poor patient outcome, suggesting that they could serve as biomarkers of sensitivity to existing cancer treatments [107, 108].

Five splicing isoforms have been reported for NRAS: isoform 1; isoform 2 containing exon 3b; isoform 3 lacking exon 3; and isoform 4 lacking exon 3 and 4 simultaneously; and isoform 5, in which the first 17 codons of exon 2 and 3 codons at the end of exon 5 are fused [109]. Isoforms 1, 2, and 4 are located only in the cytoplasm, whereas isoforms 3 and 5 are located both in the nucleus and in the cytoplasm. Each isoform has different binding affinities for downstream targets, differentially regulating the RAS signaling pathway. In addition, the different protein expression levels of each isoform are different; isoform 5 has shown almost 1000-fold lower expression than isoform 1. However, forced expression of isoform 5, which consists of 20 amino acids, has been reported to increase cell proliferation and transformation by activation of the NRAS target. The splicing isoform of A-Raf, called DA-Raf1, contains a Ras-binding domain, and it can bind to both Ras and M-Ras; however, it lacks a kinase domain and acts as a dominant-negative antagonist by interfering with the ERK pathway [110, 111].

ERK1 exon 4 encodes a phosphorylation site for ERK1 kinase. SMNDC1, which is highly expressed in patients with pancreatic and ovarian cancers and associated with poor patient outcomes, has been reported to induce ERK1 exon 4 inclusion [112]. Forced exclusion of exon 4 using antisense oligonucleotides has been shown to significantly reduce oncogenic ERK1, survival, and pro-apoptotic signaling and suppress target gene expression and tumor cell growth.

RNA binding motif protein 4 (RBM4), an RNA-binding protein that shuttles between the nucleus and cytoplasm, serves as a splicing factor that modulates alternative splicing by binding to the RNA recognition motifs (RRMs) GTAACG or CGGCGG, functioning as a general splicing inhibitor (Fig. 3A) [113, 114]. Following mRNA-sequencing analysis in RBM4-expressing H157 cells, alternative splicing events regulated by RBM4 were investigated using gene ontology. This analysis revealed that RBM4 target genes were associated with crucial cellular processes including cell proliferation, cell cycle regulation, apoptosis, migration, and tumorigenesis [115]. RBM4 demonstrated the ability to inhibit proliferation and migration in various cancer cell lines and effectively hindered cancer progression in tumor xenograft models [115‐117]. Moreover, RBM4 influenced the utilization of the 5′ss of Bcl-x, an apoptosis regulator, inducing a shift from the anti-apoptotic isoform (Bcl-xL) to the pro-apoptotic isoform (Bcl-xS), thereby promoting apoptosis and impeding tumor progression [118, 119]. In addition, RBM4 downregulated the protein level of serine/arginine-rich splicing factor 1 (SRSF1), a splicing factor that functions as a proto-oncogene, and inhibited mTOR activation [120‐122]. In gastric cancer cell lines, RBM4 has been shown to impede cancer progression by suppressing the expression of MAPK-dependent signaling pathway proteins [117]. Another study provided evidence that RBM4 exerted control over cell proliferation and mediated inflammatory responses by regulating the alternative splicing of transcription factors and co-activators [116]. Consistent with its role as a tumor suppressor, RBM4 expression has been found to be decreased in patients with NSCLC, breast cancer, pancreatic cancer, and gastric cancer [123]. In addition, patients with lung, breast, ovarian, and gastric cancers presenting higher expression of RBM4 exhibited higher survival rates, suggesting that RBM4 could serve be a target for human cancer treatment.

Neuro-oncological ventral antigen 1 (NOVA1) is a well-known brain-specific splicing factor that plays a crucial role in alternative splicing regulation in neurons by binding to the YCAY sequence on target pre-mRNAs (Fig. 3B) [124‐130]. In the context of gastric cancer, increased levels of miR-146b-5p in the surrounding tissues following gastrectomy have been associated with tumor recurrence and poor survival rates. Notably, miR-146b-5p inhibits NOVA1 expression in immune cells and stromal spindle cells within the remnant microenvironment [131]. Decreased NOVA1 levels in T cells within gastric cancer tissues are correlated with a decrease in FOXP3-positive regulatory T cells and are indicative of poor patient prognosis. These findings suggest that NOVA1 may serve as a potential biomarker for predicting the prognosis of gastric cancer patients and the presence of occult residual disease in remnant tissues post-gastrectomy.

Further studies by the same group found that attenuation of NOVA1 expression in T cells, tumor cells, and stromal spindle cells reduced FOXP3 + Treg density. In contrast, the infiltration of CD68 + macrophages and CD163 + M2 macrophages was increased, and NOVA1 expression was found to be associated with changes in immune cell composition [132]. Frequent NOVA1 inhibition was observed in the gastric cancer microenvironment, and decreased NOVA1 expression in tumor cells was strongly associated with tumor progression and poor prognosis. However, in contrast to these findings, NOVA1 mRNA has been reported to be significantly elevated in gastric cancer compared to that in non-tumor tissues, and high NOVA1 expression is associated with poorer prognosis [133]. In gastric cancer, miR-339 expression was downregulated, and mRNA and protein levels of NOVA1 were suppressed in miR-339-overexpressing cells. Overexpression of miR-339 inhibited gastric cancer cell proliferation, cell cycle progression, and invasion, and overexpression of NOVA1 impaired the inhibitory effect of miR-339 on gastric cancer cells. Another negative regulator of NOVA1 is miR-181b-5p, and its overexpression has been shown to suppress proliferation, migration, and invasion while promoting apoptosis in astrocytoma cells [134]. Similarly, downregulation of NOVA1 resulted in reduced cell growth, inhibited migration and invasion, and promoted apoptosis in U87 cells. Notably, patients with astrocytoma presenting high levels of NOVA1 expression exhibited worse survival outcomes compared with those with low NOVA1 expression levels.

Intratumoral Nova1 has been reported to be strongly correlated with hepatocellular carcinoma (HCC) poor survival and increased early recurrence [135]. HCC cell lines overexpressing Nova1 promoted cell proliferation, invasion, and migration. Nova1 plays a crucial role in promoting the inclusion of exon 9 in the inhibitory neurotransmitter receptor subunits GABA_ARγ2 [136‐138]. Moreover, it interacts with the GABA_ARγ2 protein and demonstrates an inverse relationship with the expression levels of GABA_ARγ2 and GABA. Additionally, the upregulation of Nova1 expression promoted the growth of subcutaneous HCC in nude mice, indicating a potential oncogenic function for Nova1. Compared with normal lung tissue, NSCLC tissue showed significantly elevated NOVA1 expression, which correlated with indicators of poor differentiation, TNM stage, T stage, and lymph node metastasis [139]. Additionally, patients with NSCLC exhibiting high NOVA1 expression experienced shorter survival compared to those with low expression. NOVA1 facilitated the proliferative and invasive capacities of NSCLC cells by regulating the activation of the Wnt/β-catenin signaling pathway. Moreover, NOVA1 expression promoted the inclusion of exons 7 and 8 of human telomerase reverse transcriptase (hTERT), generating enzymatically active telomerase and influencing telomere length [140]. Notably, NOVA1 knockdown significantly reduced tumor growth in a xenograft model.

Additionally, NOVA1 has been associated with an unfavorable prognosis in patients with colorectal cancer (CRC) [141]. It upregulates MMP-2, MMP-7, and MMP-9 by regulating JAK2/STAT3 signaling through binding to and stabilizing IL6 mRNA. Consequently, NOVA1 has been identified as a novel regulator that influences the proliferation and metastasis of CRC cells. NOVA1 has been reported to exhibit a tumor-suppressive effect depending on the specific environment; however, studies conducted till date collectively indicate that NOVA1 appears to have a dominant tumor-promoting effect affecting proliferation, invasion, migration, and telomerase activity.

The SF3B complex regulates splicing of pre-mRNA by binding to the U2 snRNP and recognizing BPS [142]. The SF3B complex consists of seven proteins (SF3B1–7) with a molecular weight of 10 ~ 155 kDa. SF3B components, including SF3B1, have been implicated in cancer and various genetic disorders. Splicing factor 3B subunit 1 (SF3B1) is the largest protein within the SF3B complex, and it contains a HEAT [Huntingtin, elongation factor 3, subunit A of protein phosphatase 2 A, phosphatidylinositol 3-kinase (PI3K) target of rapamycin 1] domain consisting of 22 tandem repeats (Fig. 3C) [143]. The HEAT domain serves as the central region for RNA and protein binding in the SF3B complex. In human diseases, most SF3B1 mutations are localized in the HEAT 4–12 region. SF3B1 mutation has been detected in 30% of patients with myelodysplastic syndrome (MDS) and in 80% of patients presenting MDS subtype with ringed sideroblasts (RARS) [144‐147]. Additionally, mutations have been detected in 20% of patients with MDS/myeloproliferative neoplasms and 15% of patients with chronic lymphocytic leukemia (CLL) [146, 148‐150]. Furthermore, mutations in SF3B1 have been reported at low frequencies in patients with acute myeloid leukemia, breast cancer, prolactinomas, uveal melanoma, leptomeningeal melanoma, blue nevus-like cutaneous melanoma, pancreatic ductal adenocarcinoma, and prostate cancer [151‐159].

Although some reports have suggested no significant effect, SF3B1 mutations in patients with MDS have been associated with a favorable prognosis and long survival rates [146, 160‐162]. However, in most diseases except MDS, but especially cancer, mutations in SF3B1 have been associated with poor prognosis and survival [163]. Furthermore, SF3B1 is overexpressed in glioblastoma, hepatocellular carcinoma, prostate cancer, and endometrial cancer [164‐167]. Similar to its mutations, SF3B1 overexpression has been directly correlated with adverse patient prognosis, lower survival rates, and drug resistance. Dysregulation caused by SF3B1 mutations or changes in expression modulates oncogenic splicing variants such as BCL2L1-xL, KLF6-SV1, AR-v7 (androgen receptor variant 7), and In1-ghrelin [163‐166, 168, 169]. The blockade or silencing of SF3B1 has been demonstrated to regulate the expression levels of essential components involved in mRNA homeostasis, including spliceosome, splicing factors, exon-junction complex (EJC) and SMG-1Upf1–eRF1–eRF3 (SURF) components, and NMD factors [164, 166]. Furthermore, it has been shown to modulate the AKT/mTOR/ß-catenin, JNK, PDK1, GSK3b, ERK, and AMPK signaling pathways [164‐166]. Additionally, SF3B1 inhibition has been found to suppress proliferation, migration, apoptosis, and the formation of tumor spheres and colonies, as well as angiogenesis [164‐166].

Capitalizing on the oncogenic properties of SF3B1 and its potential for suppressing tumor growth through blockade or silencing, several drugs have been developed to target SF3B1. These drugs include pladienolide B, spliceostatins, herboxidiene, sudemycins, and H3B-8800 [170‐174]. Notably, H3B-800 entered phase I clinical trials in 2016, further affirming SF3B1 as a promising biomarker and a target for pharmacological treatment [175, 176].

Splicing factor 3b subunit 4 (SF3B4) is a major subunit of the SF3B complex, consisting of two N-terminal RRMs and a C-terminal proline-rich (PR) domain (Fig. 3D) [177]. In addition to its role in pre-mRNA splicing, SF3B4 is implicated in cell signaling, transcription, and translation processes [178].

According to a TCGA analysis, only SF3B4 has shown high expression in the SF3B complex in cervical squamous cell carcinoma and endocervical adenocarcinoma [179]. SF3B4 has been reported to enhance the proliferation and invasion of cervical cancer cells, thereby promoting their malignant behavior. RNA-seq analysis conducted in SF3B4-knockdown HeLa cells revealed that differentially expressed genes (DEGs) were enriched in cellular processes such as regulation of cell proliferation, transcription, apoptotic process, and cell adhesion. Among the downstream targets of SF3B4, the gene exhibiting the most significant change in mRNA expression was sperm-associated antigen 5 (SPAG5). SF3B4 knockdown reduced SPAG5 expression by inducing the retention of SPAG5 intron 21, subsequently causing premature termination of the transcript. SPAG5 is a mitotic spindle-binding protein involved in regulating mitosis [180]. It has been reported to promote the proliferation and progression of not only cervical cancer but also hepatocellular cancer and breast cancer by modulating the cell cycle [179, 181‐184]. Notably, the presence of SPAG5 intron retention transcripts correlated with extended survival times in patients. Collectively, these findings underscore the oncogenic role of SF3B4 in cervical cancer by virtue of its regulatory influence on SPAG5 splicing.

SF3B4 has been reported to be up-regulated in hepatocellular carcinoma [185, 186]. High SF3B4 expression has been found to be associated with intrahepatic metastasis and poor prognosis [187]. Overexpression of SF3B4 has been shown to trigger the SF3b complex, which induced exon skipping of the tumor suppressor Kruppel-like factor 4 (KLF4), resulting in non-functional KLF4 transcripts [188]. Consequently, cyclin-dependent kinase inhibitor 1B (CDKN1B; p27Kip1) became transcriptionally inactive, disrupting cell cycle regulation, while the activation of the Snail family transcriptional repressor 2 (SNAI2) gene promoted epithelial–mesenchymal transition (EMT). These events contributed to the malignant transformation and proliferation of liver cells. Furthermore, SF3B4 outperformed the existing diagnostic markers for hepatocellular carcinoma, namely glypican 3 (GPC3), glutamine synthetase (GS), and heat-shock protein 70 (HSP70) and demonstrated its potential as a reliable diagnostic marker for early-stage hepatocellular carcinoma.

Mutations in SF3B4 mostly lead to reduced expression due to frameshifts, which cause acrofacial dysostosis, Nager syndrome, and Rodriguez syndrome [189, 190]. The SF3B4 mutation affects the regulation of gene expression and abnormal splicing of crucial genes involved in skeletal development within growth plate chondrocytes [191, 192]. Consequently, this disruption ultimately manifests as defects in craniofacial and limb development, which are observed in acrofacial dysostosis.

SF3B4 has exhibited significant upregulation in ovarian cancer and a correlation with unfavorable patient prognosis [193]. The expression of SF3B4 has been found to be negatively regulated by miR-509-3p. RAD52, involved in DNA damage repair, assumes an oncogenic role in various tumors [194‐196]. Loss of SF3B4 has been reported to reduce RAD52 expression by inducing retention of intron 8 of RAD52 and generating premature termination codons. SF3B4 has been shown to act as an oncogene by modulating the alternative splicing of RAD52, thereby facilitating the proliferation, migration, and invasion of ovarian cancer cells. Additionally, SF3B4 has been reported to function as an oncogene in esophageal squamous cell carcinoma (ESCC) [197]. Conversely, in pancreatic cancer, the protein level of SF3B4 has shown a reduction compared with that in adjacent symptomatic tissue, and this diminished expression of SF3B4 has been found to facilitate the proliferation and migration of pancreatic cancer cells, indicating an inhibitory role for SF3B4 in pancreatic cancer [198]. These findings suggest that SF3B4 may have different biological functions, depending on the tumor type.

The serine/arginine-rich splicing factor (SRSF) protein family comprises RNA-binding proteins that regulate various RNA biological processing such as mRNA transport and polyadenylation, as well as constitutive and alternative splicing [199‐206]. Currently, 12 members (SRSF1–12) of the SRSF family have been identified in humans, with SRSF3 being the smallest member (Fig. 3E). SRSF3 has been observed to exhibit high expression levels in a wide range of tumors, including breast cancer, cervical cancer, colorectal cancer, gastric cancer, glioblastoma, head and neck squamous cell carcinoma, hepatocellular carcinoma, non-small cell lung cancer, oral squamous cell carcinoma, ovarian cancer, and retinoblastoma [203, 207‐225].

SRSF3, a potential exonic splicing enhancer that is upregulated in glioblastoma, binds to the CA(G/C/A)CC(C/A) motif and alters more than 1,000 alternative splicing events [215]. Particularly, the knockout of SRSF3 leads to the exclusion of exon 7 in the ETS variant 1 (ETV1) gene and the replacement of the terminal exon 9 in the nudE neurodevelopment protein 1 (NDE1) gene with a mutually exclusive exon 9′. The ETV1 isoform with exon 7 inclusion and the NDE1 isoform with terminal exon 9 have been confirmed to be important for mitosis and cell proliferation of tumor cells, significantly increasing oncogenic activity. In addition, SRSF3 has been shown to inhibit PDCD4 protein expression by participating in alternative splicing, cytoplasmic export and translation of PDCD4, a tumor suppressor gene involved in antiproliferation, apoptosis, and antimetastasis [203, 226, 227].

MDM4 is an oncogene that suppresses the p53 tumor suppressor [228, 229]. SRSF3 is necessary for the inclusion of exon 6 (exon 7 in mice) in the human MDM4 gene [230]. When this exon is excluded, it produces an unstable transcript containing a premature termination codon, which is subject to NMD. Consequently, this splicing event affects MDM4 protein levels. The in vitro and in vivo induction of MDM4 exon 6 skipping using an antisense oligonucleotide have been reported to inhibit MDM4 protein abundance and melanoma growth and to increase sensitivity to MAPK-targeting therapeutics.

TAR DNA-binding protein (TDP43) is overexpressed in triple-negative breast cancer (TNBC) and is a major regulator of unique alternative splicing in TNBC [209]. SRSF3 interacts with these TDP43 to control specific splicing events, including that of PAR3 and NUMB. SRSF3/TDP43 knockdown has been shown to inhibit cell proliferation by inducing NUMB exon 12 exclusion and inhibition of cell migration and invasion by inducing PAR3 exon 12 inclusion. SRSF3 knockdown has been shown to induce G1 arrest and apoptosis by promoting downregulation of G1/S transition-related genes, BCL2 protein reduction, and homeodomain-interacting protein kinase-2 (HIPK2) exon 8 exclusion [214].

The oncogene Erb-B2 receptor tyrosine kinase 2 (ERBB2) is overexpressed in 20–30% of invasive breast cancer and is associated with poor prognosis [231]. In breast cancer cells, SRSF3, along with hnRNP H1, has been identified as a regulator responsible for controlling the production of distinct splice variants of ERBB2 with different functionalities [232]. In particular, the knockdown of SRSF3 converted the oncogenic variant (exon 16 skipping) to a cell proliferation suppressive variant (premature stop codon generated by intron 15 inclusion). Additionally, SRSF3 binds to exon 18 of the interleukin enhancer-binding factor 3 (ILF3), leading to the production of an ILF3 isoform that facilitates cell growth [233]. Furthermore, SRSF3 has been demonstrated to be strongly related to the PI3K-AKT signaling pathway [201]. Cumulative evidence from various studies supports the role of SRSF3 in tumorigenesis, proliferation, and anti-apoptosis. Consequently, targeting the downregulation of SRSF3 holds potential as a therapeutic strategy for anticancer treatment.

Polypyrimidine tract binding protein (PTB) shuttles between the nucleus and cytoplasm, and it is involved in various mRNA metabolic pathways, such as polyadenylation, mRNA stability, and initiation of translation, as well as regulation of pre-mRNA splicing (Fig. 3F) [234, 235]. PTB has a high affinity for binding to CU rich sequences including UCUU and CUCUCU [236, 237].

PTB is found to be overexpressed in epithelial ovarian tumors, glioma, and various cancer cell lines [238‐240]. Knockdown of PTB has been shown to decrease cell proliferation, anchorage-independent growth, and invasiveness in these cell lines. Specifically, PTBP1 knockdown enhances the inclusion of exon 3 in reticulon 4 (RTN4) [241]. The presence of exon 3 in RTN4 isoforms is associated with reduced cell proliferation, suggesting that PTB-induced cell proliferation in glioma cells is partly mediated by RTN4 splicing. Another study reported that in glioblastoma, PTBP1 recognizes an alternative 5′ss within ubiquitin specific peptidase 5 (USP5) exon 15, resulting in the generation of USP5 isoform 2 with a shorter exon length [242]. Enforced expression of USP5 isoform 1 through antisense targeting in glioblastoma cell lines has been demonstrated to inhibit cell growth and migration. These findings demonstrate that both the regulation of RTN4 splicing by PTBP1 and the splicing of USP5 play important roles in gliomagenesis.

Keloid is a fibrotic skin disease characterized by excessive accumulation of extracellular matrix due to the proliferation of dermal fibroblasts [243, 244]. Its pathological features are similar to that of tumors, including tissue invasion and recurrence. PTB is overexpressed in keloid tissues and fibroblasts, and as in glioblastoma, alternative splicing changes in RTN4 and USP5 by PTB have been observed [245]. Furthermore, suppressing PTB has been shown to decrease the expression of fibronectin 1 (FN1) in transplanted keloid tissues and TGF-β1-treated keloid fibroblasts. FN1 is an important protein involved in cell adhesion, migration, and differentiation and is closely associated with cancer and fibrosis. Additionally, inhibition of PTB led to a reduction in excessive deposition of collagen type III alpha 1 chain (COL3A1), demonstrating that PTB siRNA promoted regression of keloid tissue in vivo by regulating both dermal cell proliferation and extracellular matrix accumulation.

PTB induces the exclusion of α-exon in fibroblast growth factor receptor 1 (FGFR1), which is associated with proliferation, and regulates the mutually exclusive splicing of exons IIIb and IIIc in fibroblast growth factor receptor 2 (FGFR2) [246, 247]. Moreover, PTB inhibits the inclusion of caspase 2 (CASP2) exon 9 and FAS exon 6, which are involved in apoptosis [248‐250]. The binding of PTBP1 to pyruvate kinase M1/2 (PKM) intron 8 leads to the skipping of exon 9 in PKM, resulting in the generation of a PKM2 isoform that promotes the Warburg effect [251, 252]. This increased expression of PKM2 confers drug resistance in pancreatic ductal adenocarcinoma, suggesting that PKM2 and PTBP1 could serve as potential therapeutic targets to enhance the response to chemotherapy [253].