Aberrant expression and regulatory network of splicing factor-SRSF3 in tumors

Alternative splicing facilitates the splicing of precursor RNA into different isoforms. Alternatively spliced transcripts often exhibit antagonistic functions or differential temporal or spatial expression patterns. There is increasing evidence that alternative splicing, especially by the serine-arginine rich (SR) protein family, leads to abnormal expression patterns and is closely related to the development of cancer. SRSF3, also known as SRp20, is a splicing factor. Through alternative splicing, it plays important roles in regulating various biological functions, such as cell cycle, cell proliferation, migration and invasion, under pathological and physiological conditions. Deregulation of SRSF3 is an essential feature of cancers. SRSF3 is also considered a candidate therapeutic target. Therefore, the involvement of abnormal splicing in tumorigenesis and the regulation of splicing factors deserve further analysis and discussion. Here, we summarize the function of SRSF3-regulated alternative transcripts in cancer cell biology at different stages of tumor development and the regulation of SRSF3 in tumorigenesis.


Introduction
Posttranscriptional modification is an essential component of the genetic code of eukaryotic genomes. Alternative splicing is an important process involved in posttranscriptional modification [1]. It is known that ~95% of human genes can be spliced [2]. Through this process, the noncoding introns are removed, and coding exons containing the correct translational information are left to maintain the stability of eukaryotic organisms. Alternative splicing plays important roles in cell proliferation and differentiation, and thus participates in multiple diseases, including cancer [3,4].
Alternative splicing refers to the process by which different splicing isomers are produced from a pre-mRNA by different splicing methods, leading to final protein products with different phenotypes, different functional characteristics or antagonistic effects on tumors [5]. Alternative splicing is accomplished via the spliceosome. The spliceosome is mainly composed of five kinds of small nuclear ribonucleic proteins (snRNPs) and three kinds of non-snRNPs [6][7][8]. Among them, non-snRNPs include the hnRNP protein family, serine/arginine-rich splicing factor (SR) protein family and SR-related protein family. SR protein can bind to pre-mRNA and influence the interaction between snRNP splicing factors and pre-mRNA, thus promoting or inhibiting the recognition and utilization of splicing sites [9].
Serine/arginine-rich splicing factor 3 (SRSF3) is the smallest member of the SR protein family. SRSF3 is relatively conserved and binds to specific pre-mRNAs. It contains one RRM (RNA receptor Ivyspring International Publisher binding domain) and one RS domain ( Figure 1). The RRM domain, located at the N-terminus, can specifically recognize and bind to pre-mRNA. Thus, it is the key factor by which SR proteins determine the splicing site [17]. The RRM domain can interact with pre-mRNA cis-elements in a concentration-dependent manner and then regulate the alternative splicing of these genes [18,19]. For example, exogenous noncoding SRSF3 (exon 4 inclusion) can increase the expression of endogenous SRSF3 and p53 in a dose-dependent way [20]. In addition, the high concentration of SR proteins can even replace U1 snRNP and restore splicing activity to pre-mRNA [21,22]. The RS domain, located at the C-terminus, is composed of alternately arranged arginine and serine, which mainly mediate the interaction between protein and mRNA or protein and protein. Through interaction, the SR protein can bind to other splicing factors and aggregate near splicing sites of pre-mRNA. Therefore, the RS domain determines the splicing activity of the SR protein and is also a necessary region for splicing pre-mRNA [23].

SRSF3-regulated alternative splicing in tumors
Overexpression of SRSF3 in cancer cell lines has been found through database analysis and tissue sample detection. SRSF3 is highly expressed in many tumor tissues, such as the thyroid, ovary, lung, colon, breast, stomach, skin and bladder [15,[24][25][26]. Thus, the expression of SRSF3 might be a useful biomarker for diagnosis and prognosis.

SRSF3 and exon skipping (ES)
SRSF3 promotes exon skipping CCDC50 (long isoform including alternative exon 6, CCDC50L) has been recognized as a tyrosinephosphorylated protein. Overexpression of SRSF3 can promote CCDC50 exon 6 skipping to form CCDC50S (short isoform lacking alternative exon 6, CCDC50S) ( Table 1). Compared with CCDC50L, CCDC50S has a higher expression and tissue specificity in hepatocellular carcinoma (HCC). SRSF3 can also improve the stability of CCDC50S in the cytoplasm and further increase the activation of the Ras/Foxo4 signaling pathway. Further, upregulated expression of CCDC50S can increase the proliferation and metastasis of HCC [28]. CCDC50S can be used as a new therapeutic target and prognostic biomarker for HCC. SRSF3 can also mediate some receptors. CD44, an important membrane receptor of hyaluronic acid (HA), is also involved in various processes, including tumor proliferation and metastasis [29][30][31]. Overexpression of SRSF3 promotes skipping of CD44 exons 12-14 to form the CD44E variant, which enhances tumorigenesis (Table 1) [32].
Some kinase activity is also related to SRSF3. Only Iso2 (exon 16 inclusion) of mitogen-activated protein 4 kinase 4 (MAP4K4) can interact and phosphorylate c-Jun N-terminal kinase (JNK) [41]. Overexpressed SRSF3 binds to the CU element in exon 16 and promotes the formation of MAP4K4 Iso2, which further activates the JNK signaling pathway [42]. JNK signaling pathways widely promote the migration and invasion of cancer cells [43]. SRSF3 increases the expression of the EMT-related genes N-cadherin and vimentin, thus promoting the invasion and migration of colorectal cancer (CRC) cells [42]. SRSF3 can also promote migration in TNBC. Cortisol-induced SRSF3 overexpression can promote exon 9 inclusion of GR and subsequently form GRα. In contrast to GR, GRα can respond to glucocorticoid actions [44]. Erica et al. found that GRα promotes the transcriptional regulation of receptor for activated C kinase 1 (RACK1), which can further increase the migration of MDA-MB-231 cell lines [45]. TNBC lacks effective treatment strategies, but approximately 25% of invasive TNBC is known to be glucocorticoid receptor (GR)-positive [46]. SRSF3 might be a novel therapy target for GR-positive TNBC.
Forkhead box transcription factor M1 (FoxM1) is also a transcription factor related to tumor proliferation [47]. It has been reported that SRSF3 knockdown promotes FoxM1 exon 9 skipping and downregulates the expression of FoxM1 and its two transcriptional targets PLK1 and Cdc25B, both of which are cell cycle-related proteins. As a result, decreased expression of FoxM1, PLK1 and Cdc25B results in inhibition of the G2/M phase transition. Therefore, it hinders the progression of the cell cycle and promotes the apoptosis of U2OS cells [15].
In addition to the G2/M phase transition, SRSF3 plays an important role of in regulating the progression of the G1/S phase transition in colon cancer cells. Knockdown of SRSF3 can reduce the expression of cyclin D1, cyclin D3, cyclin E1, E2F1 and E2F7, which can impair the G1-to-S-phase progression. In this way, SRSF3 induces G1 cell cycle arrest and apoptosis [48].   Knockdown of SRSF3 causes apoptosis in ovarian cancer cells [49]. The oncogene MDM4 can effectively promote the occurrence and proliferation of melanoma [50,51]. Overexpressed SRSF3 binds to MDM4 exon 6, promotes the inclusion of exon 6 and forms normal function of MDM4 (MDM4-L), thereby inhibiting p53-mediated target gene transcription and promoting cell cycle arrest and melanoma cell apoptosis [52]. Strikingly, the regulation of exon 6 inclusion is also a key determinant of MDM4 protein abundance in normal tissues and cancer cells. Exon 6 skipping can reduce MDM4 protein abundance [52].
Homeodomain-interacting protein kinase-2 (HIPK2) is also related to the apoptosis of cancer cells, and the deletion of HIPK2 increases the proliferation potential of cancer cells [53][54][55]. In addition, full-length HIPK2 can promote the E3 ligase activity of Siah-1 [56]. Knocking down SRSF3 leads to the skipping of exon 8 of HIPK2, thus promoting the formation of the Δe8 isoform. Because the Δe8 isoform lacks 27 amino acids that mediate its binding to Siah-1, it cannot participate in protein degradation [48]. In this way, depletion of SRSF3 promotes the apoptosis of colon cancer cells.
In addition to these solid tumors, significant progress has been made in the treatment of pediatric B-cell acute lymphoblastic leukemias (B-ALL), but the mortality and recurrence rates are still high in childhood cancer [57]. CD19 is a cell surface signaling protein. Because it can recruit intracellular kinases, CD19 is considered to play an indispensable role in promoting B cell proliferation and differentiation [58][59][60][61]. Elena et al. found that knockdown of SRSF3 resulted in enhanced abundance of the CD19 Δex2 isoform. Consistent with this, in relapsed leukemia, low expression of SRSF3 is correlated with high expression of the CD19Δex2 variant. At present, the effective treatment strategy combats CD19 through the application of T cells expressing chimeric antigen receptors (CAR-T) to improve the cure rate of B-ALL in pediatrics [62,63]. However, because of the existence of alternative splicing, a sequence of CD19Δex2 is missing, and the CAR-T cannot recognize CD19Δex2, which reduces the survival rate of pediatric B-ALL [64]. It is worth mentioning that the role of the CD19 Δex2 variant in the differentiation and proliferation of normal B cells is unknown.

SRSF3 and retained intron (RI)
EIF2B5, a major regulator of translation initiation, has been shown to be a marker protein in various cancers [65,66]. EIF2Bε is the largest subunit of EIF2B5, which has two isoforms, truncated eIF2Bε (65 kDa) and full-length eIF2Bε (80 kDa). Under hypoxic conditions, SRSF3 promotes the intron 12 retention of eIF2Bε and produces a premature termination codon (PTC) that inhibits the overall translation and forms truncated eIF2Bε in head and neck carcinoma (HNC) cells (Table 1). This truncated eIF2Bε lacks the binding region of eIF2Bα and an activating guanine exchange factor (GEF) domain [67]. Therefore, it cannot bind eIF2Bα (a subunit of EIF2B5) and guanosine triphosphate (GTP) to initiate the translation process. Finally, truncated eIF2Bε improves the survival rate of HNC cancer cells under prolonged or acute hypoxia [68].

SRSF3 and alternate terminators (A3SS)
Serine/arginine-rich splicing factor 5 (SRSF5) is another important member of the SR protein family that can promote the proliferation of lung cancer cells [69,70]. SRSF5 has recently been discovered as a novel oncogene that promotes oral squamous cell carcinoma (OSCC) tumor formation and proliferation [71]. When alternative splicing occurs at the distal 3' splice site of exon 6, it results in a normal functional long-SRSF5 (SRSF5-L). However, when alternative splicing occurs at the proximal 3' splice site of exon 6, a short fragment (approximately 956 bp) is generated due to the presence of an in-frame stop codon (Table 1). Finally, the short fragment either causes RNA degradation induced by nonsense-mediated mRNA decay (NMD) or produces a SRSF5-L lacking the RS domain [72]. In general, SRSF3 is positively correlated with SRSF5 protein expression in OSCC. Moreover, overexpression of SRSF3 promotes the utilization of the distal 3' splice site of SRSF5 exon 6, resulting in an increase in SRSF5-L, thereby promoting OSCC proliferation [71].

SRSF3 and mutually exclusive exons (MEX)
NudE neurodevelopment protein 1 (NDE1) is a dynein-binding protein that regulates various microtubule-mediated processes, including mitosis [73]. Because NDE1 terminal exon 9 and exon 9' are mutually exclusive, there are two isoforms in GSCs: NDE1_KMLL (exon 9 inclusion) and NDE1_SSSC (exon 9' inclusion). NDE1_SSSC exists in normal brain tissue and participate in mitosis. However, NDE1-KMLL has the specific function of promoting the growth of GBM cells in the process of mitotic spindle formation. Knockdown of SRSF3 can promote a transition from NDE1_KMLL to NDE1_SSSC (Table 1) [40].
By alternative splicing, the pyruvate kinase muscle (PKM) gene can produce two isoforms: PKM1 (exon 10 skipping) and PKM2 (exon 9 skipping). In addition, exon 9 and exon 10 are mutually exclusive. Most cancer cells mainly express PKM2 to maintain glycolysis-dominant energy metabolism [74]. Three splicing factors, SRSF3, polypyrimidine tract binding protein 1 (PTBP1) and heterogeneous nuclear ribonucleoprotein A1 (hnRNPA1), can bind to exon 10 and promote the inclusion of exon 10, reducing the PKM1/PKM2 ratio (Table 1). Three splicing factors, PTBP1, SRSF3 and hnRNPA1, are highly expressed in colorectal cancer compared with adjacent tissue [75]. In addition, this splicing process inhibits the proliferation of colon cancer tumors and alters their metabolic patterns, from glycolysis to oxidative phosphorylation [75].

SRSF3 alone regulates various splicing variants
Interleukin enhancer binding factor 3 (ILF3), a double stranded RNA binding protein, is the main factor regulating the proliferation of cancer cells [76,77]. SRSF3 binds to different motifs on ILF3 pre-mRNA and produces multiple ILF3 isoforms by exclusion/inclusion of ILF3 exon 18 or by selection of an alternative 3' splice site within exon 18. Among them, ILF3 isoform 1 and isoform 2 can promote cell proliferation, while isoform 5 and isoform 7 suppress tumor cell proliferation and the isoform 7 induces cell apoptosis. Rong et al. confirmed that knocking down SRSF3 attenuated the expression of isoform 1 and isoform 2, and thus inhibited in HeLa and U2OS cell proliferation [78]. Meanwhile, when SRSF3 is knocked down, increased ILF3 isoform 5 inhibits the proliferation of HeLa and U2OS cells, and increased ILF3 isoform 7 induces apoptosis (Table 2) [79].
As a tumor suppressor, programmed cell death 4 (PDCD4) is downregulated in most tumors [80]. The PDCD4 gene consists of 12 exons, and the cassette exon involved in alternative splicing is located between exons 2 and 3. Under normal conditions, total PDCD4 exists in major isoform of PDCD4 (PDCD4 AS-1), but there is little PDCD4 AS-2 (cassette exon inclusion) expression in the cytoplasm. Knockdown of SRSF3 increases the expression level of PDCD4 AS-2 mRNA in the cytoplasm and its nuclear export in colorectal cancer cells (Table 2) [81]. Interestingly, SRSF3 can bind directly to PDCD4 mRNA and mediate translational repression of PDCD4 AS-1 [15,27,82]. Taken together, SRSF3 represses PDCD4 AS-1 mRNA at the translational level, and PDCD4 AS-2 mRNA at the alternative splicing and mRNA export levels, finally promoting the proliferation and migration of colorectal tumors.
Interestingly, SRSF3 can interact with the cellular RNA-binding protein, PCBP2, a protein that binds to internal ribosome entry site (IRES) sequences within the genomic RNAs of certain picornaviruses and is required for viral translation. SRSF3 regulates IRES-induced translation initiation [83].  Promote the generation of tumors [109] The regulation of SRSF3 on PDCD4 AS-1 mRNA translation proves that SRSF3 can also play a role in the translation process.

SRSF3 collaboratively regulates various splicing variants
HER2 is overexpressed in 20-30% of breast cancer cases. The increase in HER2 expression is related to invasive tumors and adverse prognosis [84]. At present, three splicing variants (Δ16 HER2/p100/X5) of full-length HER2 have been identified, and these isomers have different functions and roles in cancer development. Knockdown of HnRNP H1 can increase the expression of the X5 variant (exon 15a inclusion) and oncogenic Δ16 HER2 (exon 16 skipping), thus increasing the trend of breast tumor proliferation. Notably, knockdown of SRSF3 results in the transition from the Δ16 HER2 to p100 variant (intron 15 retention), which inhibits the proliferation of breast cancer cells [85]. SRSF3 and HnRNP H1 are the first splicing factors shown to produce HER2 splicing variants with different regulatory functions ( Table 2).

The regulation of SRSF3
SRSF3 is frequently upregulated in many human tumors and therefore plays a role as an oncogene in various cancer processes. Cells maintain stable expression of SRSF3 through the splicing mechanism of SRSF3. However, because of its carcinogenic effect, SRSF3 must be strictly monitored (Table 3).

Splicing regulation of SRSF3
SRSF3 has two splicing forms: the full-length normal form (lacking exon 4, Iso1) and the splicing isoform (containing exon 4, Iso2). Iso2 contains a preterminated codon, which is degraded by NMD to avoid cell toxicity, or it encodes a nonfunctional truncated SRSF3 protein that lacks the RS domain at the C-terminus [86] (Figure 1). Hassan et al. found that SRSF3 can regulate alternative splicing of its own mRNA. Overexpression of SRSF3 attenuates the production of the exon 4-deleted transcript and activates the production of exon 4-containing transcripts [86].

SRSF1
Serine/arginine-rich splicing factor 1 (SRSF1, ASF/SF2) was the first member of the SR protein family to be found. As a cancer-promoting factor, SRSF1 can promote the proliferation and EMT of breast epithelial cells [68]. SRSF1 has been proven to antagonize the self-splicing effect of SRSF3 exon 4. Overexpressed SRSF3 promotes the recognition of exon 4, whereas SRSF1 inhibits the recognition and antagonizes the inclusion of SRSF3 exon 4, thus producing full-length SRSF3 (Iso1) [17,86].

HnRNP L
Nuclear heterogeneous ribonucleoprotein L (HnRNP L) is a selective pre-RNA splicing factor that maintains translation and RNA stability [36,[87][88][89]. There is a positive correlation between HnRNP L and SRSF3 in OSCC tissues [71,90]. It has been reported that HnRNP L can promote splicing of SRSF3 exon 4 and then promote the proliferation of OSCC cells [90]. This suggests that HnRNP L may be a potential carcinogenic target. Therefore, in OSCC, SRSF3 can be spliced by HnRNP L or used as a splicing factor to act on SRSF5 [71].

SLU7
SLU7, as a key splicing factor of RNA precursors, has an effect on inhibiting the development of HCC and maintaining the homeostasis of normal hepatocytes [91]. Moreover, the downregulation of SLU7 can damage liver glucometabolic and lipid metabolism disorders, and increase the risk of cancer [91]. Jiménez M et al. also found that deletion of SLU7 can promote the production of truncated SRSF3-Iso2 in a variety of different cancer cell lines. SLU7 knockdown not only increases the abnormal expression of truncated SRSF3 but also leads to the altered proliferation and abnormal metabolism of tumor cells [92].

PTBPs and RBM4
PTBPs (PTBP1 and PTBP2) can bind to the CU element of SRSF3 exon 4, impairing the automatic splicing regulation of SRSF3 and promoting SRSF3 exon 4 skipping [93]. Intriguingly, RNA-binding motif protein 4 (RBM4) is an antagonist of PTBP that can bind to intronic CU elements and promote the inclusion of SRSF3 exon 4, decrease full-length SRSF3 expression and inhibit the proliferation of CRC cells [94]. Overall, PTBP1 can promote the skipping of SRSF3 exon 4, and increased SRSF3 enhances the production of MAP4K4 Iso2 and Iso5 in HCT-8 and HCT-116 cells. When the expression of RBM4 is higher than that of PTBP1, RBM4 promotes the inclusion of SRSF3 exon 4 and the production of the MAP4K4 Iso1 variant in Colo205 cells [42].

circSMARCA5
Imbalances in the expression of circular RNAs (circRNAs) have been emphasized in several types of cancers [95][96][97]. Without splicing, circSMARCA5 is a novel tumor suppressor that inhibits the migration of glioblastoma multiforme (GBM) cells [98]. Barbagallo et al. have found that the overexpression of circSMARCA5 increased the inclusion of SRSF3 exon 4 (Iso2) and that the downregulation of circSMARCA5 significantly increased the skipping of SRSF3 exon 4 (Iso1) in the process of splicing SRSF3 by SRSF1. In addition, the ratio of SRSF3 Iso2/ Iso1 is positively correlated with circSMARCA5 expression [98].

Ubiquitination
DARPP-32, also named dopamine and cAMP-regulated phosphoprotein (Mr 32000) [99]. DARPP-32 is overexpressed in 2/3 early stage gastric cancer patients and can increase the survival rate, drug resistance and invasive activity of gastric cancer cells [100][101][102]. DARPP-32 can enhance the SRSF3 protein expression. The reason is that DARPP-32 can significantly reduce the ubiquitination of SRSF3 in digoxin-induced AGS cells. Conversely, the elimination of endogenous DARPP-32 in MKN-45 cells significantly increases the ubiquitination of SRSF3. DARPP-32 can enhance the stability of SRSF3 and regulate SRSF3-induced CD44 splicing, thus promoting CD44E expression and gastric growth [103]. However, the detailed ubiquitination mechanism is still elusive.

Neddylation
Deepak et al. found that the expression of SRSF3 decreased in early human liver diseases, including nonalcoholic fatty liver disease (NAFLD), nonalcoholic steatohepatitis (NASH), or cirrhosis [104]. Neddylation is a posttranslational modification that is similar to ubiquitination [105]. NEDD8 is a member of the ubiquitin-like protein family and covalently binds to Lys residues of substrates [106]. Under oxidative stress induced by palmitic acid, NEDD8 promotes the neddylation of SRSF3 at 11 lysine sites, resulting in SRSF3 degradation in liver diseases. When lysine 11 mutation (K11R) occurs, SRSF3 neddylation is reduced, protecting SRSF3 from degradation and reducing the risk of liver disease and further progression to HCC [104]. SRSF3 is regulated in a neddylation-dependent manner in the liver, which provides an effective therapeutic strategy for suppressing liver diseases and hepatocellular carcinogenesis.

Phosphorylation
Similar to SR proteins, SRSF3 is mainly located in the nucleus. In the absence of splicing, the SRSF3 protein mainly aggregates into irregular granular nuclei and becomes a typical structure in the nucleus, which is called a "speckle" [107].
SRSF3 is also a nuclear shuttle protein. Its shuttling property is related to the phosphorylation of the RS domain [108]. When serine (residue 105 relative to the C-terminus) in the RS domain is phosphorylated by serine-arginine protein kinase 2 (SRPK2), the SR protein can be released from the speckles in the nucleus. Then, the N-terminal RRM domain specifically binds to the pre-mRNA to complete splicing with other splicing factors [109]. Therefore, the phosphorylation of SR protein is very important to the shuttling property of SR protein. SRSF3 is hypophosphorylated in cell homeostasis. Enhanced phosphorylation of SRSF3 promotes the occurrence of splicing events mediated by carcinogenic SRSF3 and the generation of GBM tumors [40]. In addition, Wang et al. found that hypothalamic gonadotropin-releasing hormone (GnRH) has normal reproductive function and can induce phosphorylation of SRSF3 in a time-dependent way [110].

Conclusions
SRSF3 is upregulated in many kinds of tumors and functions as a proto-oncogene. However, SRSF3 expression is decreased in >50% of human HCCs [111]. Specific loss of SRSF3 in hepatocytes impairs the differentiation and metabolism of hepatocytes in early adulthood [112]. Hepatocyte-specific SRSF3 knockout mice are at risk of liver fibrosis, while SRSF3 can protect mice from CCL4-induced fibrosis and carcinogenesis by inhibiting the inclusion of EDA fibronectin 1 exon [111]. In addition, deletion of SRSF3 promotes the skipping of Git2 exon 15/Slk exon 13, the inclusion of Myo1B exon 23/Ctnnd1 exon 18, and the development of HCC metastasis with aging [113].
In alternative splicing, most of the isomers produced by SRSF3 have different phenotypes, which are manifested in promoting and suppressing cancer. SRSF3 can regulate the cell cycle, apoptosis, cell metastasis and proliferation according to different splicing variations ( Figure 2B).
At present, although it is known that SRSF3 can splice related proteins, the mechanism by which SRSF3 splices related proteins are still unclear. In addition, the splicing role of SRSF3 in tumorigenesis and its clear molecular mechanism deserve further investigation.
Elucidation of the molecular mechanism of abnormal splicing of SRSF3 caused by tumorigenesis could be of great significance in clarifying the occurrence and development of tumors.