The epithelial-mesenchymal transition (EMT) is an important process in both normal embryonic development and wound healing, enabling the synthesis of human tissues and organs by converting cells from the epithelial phenotype into the mesenchymal phenotype [1]. However, EMT has also been found to be involved in the invasion and metastasis of tumours, among other important features of carcinogenesis [2]. More than 90% of all malignant tumours originate from epithelial cells, and cancer cells are able to spread to distant organs and infiltrate nearby tissues through the dysregulation of EMT ^[1][3][1,3]. EMT is therefore one of the primary drivers in the process of development of epithelial tumours into more aggressive phases. EMT is regulated at various levels, including transcriptional control, post-translational, differential splicing, and non-RNA regulation [4]. An essential component in the modulation of EMT is the transcriptional regulation, which is mediated by a group of transcription factors (TFs), including members of the snail family zinc-finger transcription factor (SNAIL), twist family BHLH transcription factor (TWIST), TWIST, and zinc-finger E-box binding homeobox (ZEB)ZEB families, that induce the expression of genes necessary for mesenchymal properties and inhibit the expression of epithelial-related genes [5].

The process known as alternative splicing (AS), which produces numerous mRNAs from a single transcript, is one of the main factors that drives the variety of the proteome. During the development of cancer, the dysregulation of AS contributes to many aspects of the tumour cell biology, including EMT. A number of EMT-related molecules have been shown to be able to be regulated by multiple kinds of cancer-specific AS isoforms, which consequently promote the EMT process [6].

2. The Epithelial-Mesenchymal Transition

2.1. Mechanism of EMT

The main features of EMT are downregulation of epithelial cell characteristics and acquisition of mesenchymal properties (Figure 1). The first step of EMT is a disassembly of the epithelial cell–cell contacts, which include the tight junction, adherens junction, desmosomes, and gap junction, as well as the loss of cell polarity. The key cell–cell adhesion molecules—epithelial cadherin (E-cadherin) and cytokeratin—are downregulated while neural cadherin (N-cadherin), vimentin, and fibronectin, are upregulated. Cells acquire the mesenchymal phenotype with motility and invasive capacities by forming lamellipodia, filopodia, and invadopodia, followed by expressing matrix metalloproteinases which have the ability to degrade extracellular matrix proteins [7].

2.2. Classification of EMT

Type 1 EMT (Figure 2A) is encountered during embryogenesis. The primitive epithelium transforms to the primary mesenchyme through EMT, then the primary mesenchyme is re-induced to generate secondary epithelia through the reverse MET. These secondary epithelia can then develop into diverse kinds of epithelial tissues. It is related to embryonic development, such as implantation, embryonic gastrulation, and organ development. Type 2 EMT (Figure 2B) has been correlated to repair mechanisms that include wound healing, tissue regeneration, and organ fibrosis [9]. When tissues are chronically damaged, Type 2 EMT can stimulate a large number of fibroblasts to persistently secrete a large amount of collagen, resulting in tissue and organ fibrosis. This pathological mechanism is common in the development of many fibrosis diseases, such as liver fibrosis, idiopathic pulmonary interstitial fibrosis, and renal fibrosis. Type 3 EMT (Figure 2C) is related to cancer invasion and metastasis, and is crucial for the ability of cancer cells to cross the basement membrane, enter the blood circulation, and colonize target organs [9].

2.3. Regulation of EMT

The mechanisms regulating EMT are quite complex. Mechanical signals and biochemical signals are both playing important roles in this regulation. There are many regulators at several different molecular levels, including transcriptional and post-transcriptional levels like mRNA processing or microRNAs [1]. Snail family zinc-finger transcription factor (SNAIL) is one of the most important transcription factors (TFs) involved in EMT; it was first found as a transcriptional repressor of E-cadherin and originally identified as a repressor of E-cadherin homologs in Drosophila [10]. Proteins within the SNAIL family in vertebrates have been subdivided into two subfamilies: Snail1 (known as SNAI1) and Slug (known as SNAI2) [11]. They both can induce EMT by directly inhibiting the transcription of E-cadherin [12]. Twist family BHLH transcription factor (TWIST) is a basic helix-loop-helix protein which is an important TF for EMT during metastasis and embryonic morphogenesis ^[13][14]. TWIST represses E-cadherin and induces EMT and cell migration by directly or indirectly suppressing the E-cadherin transcription via target E-boxes in the E-cadherin promoter ^[14][15]. Zinc-finger E-box binding homeobox (ZEB) family of TFs, similar to SNAIL and TWIST, bind the E-boxes and repress epithelial junctions and polarity genes as well as activate mesenchymal genes during EMT ^[15][16]. Two subfamilies of proteins from the ZEB family in vertebrates have been reported: ZEB1 and ZEB2. They can both inhibit or stimulate the transcription process by attaching to E-boxes on regulatory gene sequences ^[16][17]. MicroRNAs (miRNAs) are also key regulators of epithelial phenotype during EMT. They can regulate E-cadherin expression either directly or indirectly and regulate the expression of TFs during EMT, as well as other target genes, to assist in identifying the epithelial or mesenchymal phenotype ^[17][18]. Additionally, long-noncoding RNA (lncRNAs) as a post-transcriptional regulator are also involved in EMT regulation, while miRNAs regulate EMT-TF ^[17][18][19][18,19,20].

3. Alternative Splicing

3.1. Splicing

The basic process of splicing contains two major steps (Figure 3): spliceosome assembly and pre-mRNAs actual splicing. These two processes are catalysed by the cooperation of five core small nuclear ribonucleoprotein (snRNP) particles (U1, U2, U4, U5 and U6) and numerous auxiliary proteins. The spliceosome is a macromolecular ribonucleoprotein complex composed of many proteins and small nuclear RNA molecules ^[20][26]. The process of spliceosome assembly is directed by consensus sequences at splice sites and results in the sequential binding and release of snRNPs, protein factors, and the disruption and formation of RNA–RNA, RNA–protein, and protein–protein interactions. Pre-mRNA splicing is a process of removing introns from pre-mRNA through two consecutive transesterification reactions and the ligation of exons in the mRNA ^[21][27].

3.2. Basic Modes of Alternative Splicing

There are five major types (Figure 4) ^[22][28]:

• Cassette exons (Exon skipping)—the most prevalent pattern in which exons are spliced out of the gene or retained in the transcript.
• Mutually exclusive exons—only one of two consecutive exons is retained in the mature transcript.
• Alternative 3′ splice site—when the splice junction at the 3′ end is changed.
• Alternative 5′ splice site—when the splice junction at the 5′ end is changed. Both alternative 3′ and 5′ splice sites can cause changes in the coding sequence.
• Intron retention—when an intron is retained in the final transcript.

3.3. Regulation of Alternative Splicing

The regulation of AS is a highly dynamic combinatorial process that relies on complex coordination between intracellular and post-transcriptional processes ^[23][29]. Regulation of AS can be triggered by the interaction of multiple RNA-binding proteins (RBPs), which bind to cis-regulatory elements around the splice sites resulting in the utilization of the regulated splice site being enhanced or repressed ^[20][26]. The cis-elements, which can be located in either exons or introns, are bound by regulatory proteins to enhance or silence splicing of adjacent regulated exons, to therefore determine whether the mature transcript includes or skips certain exons ^[24][30]. Splicing regulatory elements can be divided into four categories: exonic/intronic splicing enhancer/silencers ^[25][31]. Precise control of AS is vital for normal cells, as aberrant expression of splicing is a common cause of diseases including cancer. The trans-acting splicing factors belong mostly to two large families—SR proteins, which normally bind to splicing enhancers and promote spliceosomes assembly, and hnRNPs opposing SR protein function, which generally bind to splicing silencers to inhibit exon recognition and promote exon skipping ^[26][27][32,33].

4. Manipulation of Alternative Splicing in EMT as a Potential Therapy for Cancer

4.1. Switching Specific Alternative Splicing Patterns in EMT

There are a series of alternative spliced events in EMT which play essential roles during cancer progression. Previous studies show that changing the splicing patterns of these gene isoforms can change phenotypes in both EMT and cancer. Thus, targeting splicing patterns is a potential method of cancer management.

4.1.1. Fibroblast Growth Factor Receptor 2 (FGFR2)

FGFR2 is one of the well-recognized genes spliced differently in EMT which is essential in embryogenesis and organ regeneration ^{[28][29][30][31]}[46,47,48,49]. The FGFR family functions through the binding of FGF ligands to their receptors, activation via dimerization and, therefore, prompting the tyrosine kinase domains to a set of subsequent signal activations ^[32][50]. Following this cell division, growth and differentiation are induced ^{[33][34][35][36]}[51,52,53,54]. There are two mutually exclusive FGFR2 splice isoforms—FGFR2 IIIb and FGFR2 IIIc. Splicing switch from FGFR2 IIIb to FGFR2 IIIc has been implicated in pathological EMT ^[37][55]. Epithelial phenotype involves exon IIIb inclusion while, in EMT, exon IIIc inclusion induces a mesenchymal phenotype. It has been shown that ESRP1 and 2 are regulating splicing of FGFR2, switching to the epithelial phenotype—FGFR2 IIIb isoform ^[38][56]. ESRP expression is higher in epithelial cells and decreases in EMT. The switch from exon IIIc to IIIb of FGFR2 in mesenchymal cells may be promoted by ESRP overexpression.

4.1.2. Recepteur d’Origine Nantais (RON, MST1R)

The proto-oncogene RON, also known as macrophage stimulating 1 receptor (MST1R), is a member of the receptor tyrosine kinase (RTK) family. The mature RON protein is a 180 kDa heterodimer composed of a 40 kDa α chain and a 150 kDa β chain. The transmembrane β chain has tyrosine kinase activity, and the precursor (pro-RON) exists as a single chain [6]. The extracellular sequence of RON includes amino-terminal semaphoring (Sema), plexin-semaphorin-integrin (PSI), and four immunoglobulin-like IPT domains. Variants are mainly generated through mechanisms such as alternating pre-mRNA splicing, protein truncation and alternative transcription. RON and its alternatively spliced variants, including RONΔ85, Δ110, Δ155, Δ160, Δ165, and Δ170, play essential roles in numerous tumour biological activities in cancer, such as cell–cell adhesion, proliferation, apoptosis, and EMT ^[39][59]. The production of the ΔRon165 isoform during EMT is generated by exclusion of exon 11, which results in lacks of 49 amino acids (aas) in the extracellular β-chain [6].

4.1.3. CD44

The protein encoded by CD44 is part of a group of integral membrane glycoproteins which mediate cell–cell interactions and the interaction between cells and the extracellular matrix, together with cell adhesion and migration. The CD44 gene transcript undergoes complex alternative splicing, resulting in many functionally different protein subtypes. These different isoforms have diverse tissue-specific functions and are involved in various cellular processes, including tumour progression and metastasis. It has been found that expression of CD44v (CD44 novel splice variant) isoforms is often associated with initiation, progression, and metastasis of colon, prostate, intestinal, gastric, and breast cancers ^{[40][41][42][43][44][45]}[72,73,74,75,76,77].

4.1.4. Catenin Delta 1 (CTNND1)

The CTNND1 gene encodes a member of the Armadillo protein family, also known as p120, which plays a role in both oncogenic and tumour suppressor functions, and its alternative splicing is related to cell proliferation, migration, invasion, and EMT ^[46][47][48][81,82,83]. Two isoforms of CTNND1 generated by AS, p120-1 and p120-3, are associated with distinctive functions and different interactions in various cell types. A long mesenchymal-specific splice variant with exons 2 and 3, p120-1 is normally mostly expressed in mesenchymal cells, whereas p120-3 is a shorter isoform with lack of these exons and is mostly predominant in epithelial cells ^[49][50][84,85]. p120 isoform expression evaluation within a panel of breast cancer cell lines revealed that more invasive cells show higher expression of the large isoforms p120-1 and p120-2; however, p120-3 is expressed in all cell lines ^[51][86]. As p120-3 often has high expression levels in epithelial cells, a switch from isoform p120-1 to p120-3 might be a possible way to revise EMT and inhibit tumour growth.

4.1.5. Other Genes Spliced Differently during EMT

ARHGEF11 (Rho guanine nucleotide exchange factor 11) is the guanine nucleotide exchange factor (GEF) for the RhoA small GTPase protein. It has been revealed to promote tumour metastasis in glioblastoma and ovarian carcinoma and promotes proliferation and EMT of hepatocellular carcinoma by activating β-catenin ^[52][87].

CCND1 has two isoforms derived by splicing, named cyclin D1a and cyclin D1b, with inclusion of intron 4 in D1b mRNA. Both D1a and D1b are often upregulated in human cancers; however, cyclin D1b alone can induce cellular transformation, and is correlated with cancer progression and poor prognosis in prostate, colon, colorectal, and urinary bladder cancer ^{[53][54][55][56][57]}[89,90,91,92,93]. Androgen receptor (AR) variants, especially AR3, could contribute to prostate cancer progression through inducing EMT, achieving stem cell characteristics, and regulating stem cell related pathways ^[58][59][60][94,95,96]. Research revealed that the expression of MBNL1 isoforms lacking exon 7 (MBNL1 Δex7) proteins plays a role as a tumour suppressor, as cancer cells tend to downregulate in the presence of MBNL1 isoforms containing exon 7 ^[61][97]. Zinc-finger antiviral protein (ZAP) is an important antiviral factor that specifically inhibits the replication of a variety of viruses by binding to the target RNA sequence of the virus and interfering with the translation initiation of the target mRNA. ZAP has two major isoforms that result from alternative splicing at their C terminus: ZAPL (long) encodes a poly (ADP-ribose) polymerase (PARP)-like domain that is missing from ZAPS (short).

4.2. Targeting Splicing Regulators Involved in EMT

4.2.1. Epithelial Splicing Regulatory Proteins (ESRPs)

ESRPs drive splicing events of about 200 genes, including FGFR2, NUMB, EXOC1, MAPK14, SCRIB, and GSK3, switching these genes splicing towards isoforms characteristic of the epithelial phenotype ^[62][63][99,100]. It has been shown that ESRPs play important roles in regulating EMT and are crucial in maintaining epithelial properties and therefore reducing tumour transformation towards cells with mesenchymal phenotype ^[62][99].

4.2.2. RNA-Binding Fox Protein 2 (RBFOX2)

RBFOX2 contains an evolutionarily highly conserved RNA recognition structural motif (RNA recognition motif, RRM), which can specifically bind to the (U)GCAUG sequence. The activation or inhibition of splicing by RBFOX proteins depends on the location of their binding sites. The binding sites are located upstream of the exons. RBFOX proteins can inhibit splicing, but when located downstream of the exons, they usually activate splicing. Several studies implicated the splicing factor of RBFOX2 in splicing regulation in EMT during cancer progression ^[64][65][66][35,36,105].

4.2.3. Other Splicing Factors Involved in EMT

Expression of one of the serine/arginine (SR) proteins—serine/arginine-rich splicing factor 1 (SRSF1)—is upregulated in several cancers and is positively correlated with tumour growth and lymph node metastasis; additionally, it is negatively correlated with the sensitivity of cancer cells to anti-tumour therapy. The SRSF1 gene may regulate the occurrence of EMT in cancer cells by positively regulating the expression of the transcription factor SNAIL ^[67][68][110,111]. Polypyrimidine tract-binding protein 1 (PTBP1) is another hnRNP that sometimes acts as a splice factor. The expression of PTBP1 in the highly metastatic liver cancer cell line HCCLM3 is significantly increased, and the expression level of PTBP1 in liver cancer tissues is significantly higher than that in normal tissues. The neuro-oncological ventral antigen (NOVA) protein family, NOVA1 and NOVA2, consists of RNA-binding proteins associated with alternative splicing and transport of some target mRNAs. Recent studies demonstrate that NOVA1 and NOVA2 promote EMT through β-catenin in breast cancer cells; NOVA proteins induced an increase in epithelial and decrease in mesenchymal markers expression by restoring β-catenin expression, proposing that NOVA proteins are potential targets in breast cancer management ^[69][124]. Muscleblind-like 1 (MBNL1), a gene implicated in myotonic dystrophy that also acts as a splice factor, is identified as a suppressor of multiorgan breast cancer metastasis. MBNL1 suppresses cell invasiveness by improving the stability of these genes’ transcripts. Consistently, increased MBNL1 expression is related to reduced breast tumour metastasis ^[70][126]. 

5. Conclusions

EMT is regulated by many factors in cancer progression. During dynamic change between EMT and MET, tumour cells can migrate to neighbouring organs or metastasize at a distance and can also develop resistance to traditional chemotherapy. However, studies have shown that EMT is not an irreversible process. Reversing or inhibiting EMT may be an effective way to inhibit tumour cell migration or distant metastasis. Among different levels of modulation of EMT, alternative splicing plays an important role. The in-depth study of AS and EMT during cancer progression not only helps to better understand the occurrence and regulation of EMT in cancer, but also provides a new direction for understanding the mechanism of tumour metastasis and recurrence and provides new treatment avenues. The development of new anti-tumour drugs targeting AS and EMT will hopefully become a new research direction. Multiple genes, splicing factors, and biological processes involved in AS in EMT during cancer progression provide various targets for novel therapeutic strategies in cancer management.