Influence of transcriptional variants on metastasis

ABSTRACT

Cancer metastasis is defined as the dissemination of malignant cells from the primary tumor site, leading to colonization of distant organs and the establishment of a secondary tumor. Metastasis is frequently associated with chemoresistance and is the major cause of cancer-related mortality. Metastatic cells need to acquire the ability to resist to stresses provided by different environments, such as reactive oxygen species, shear stress, hemodynamic forces, stromal composition, and immune responses, to colonize other tissues. Hence, only a small population of cells has a metastasis-initiating potential. Several studies have revealed the misregulation of transcriptional variants during cancer progression, and many splice events can be used to distinguish between normal and tumoral tissue. These variants, which are abnormally expressed in malignant cells, contribute to an adaptive response of tumor cells and the success of the metastatic cascade, promoting an anomalous cell cycle, cellular adhesion, resistance to death, cell survival, migration and invasion. Understanding the different aspects of splicing regulation and the influence of transcriptional variants that control metastatic cells is critical for the development of therapeutic strategies. In this review, we describe how transcriptional variants contribute to metastatic competence and discuss how targeting specific isoforms may be a promising therapeutic strategy.

KEYWORDS:

• Alternative splicing
• transcriptional isoforms
• cancer-specific isoforms
• cancer
• metastasis
• metastasis-initiating potential

1. Introduction

In recent years, intensive studies have contributed to advances in early cancer detection and in the treatment of surgically resected primary tumors and adjuvant therapy [1,2]. These efforts increased survival rates for the majority of patients diagnosed during the initial stages of cancer [1,2]. However, metastatic cancer treatment faces resistance to chemotherapeutic compounds, making it necessary to understand the underlying mechanisms of the metastatic cascade [1,2]. In this sense, metastasis is responsible for 90% of cancer-related mortality and morbidity in patients [2].

Metastasis is a multistage process in which malignant cells disseminate from the primary site and grow on distant organs, leading to new tumors [3]. Nonetheless, malignant cell transformation is not sufficient for the acquisition of metastatic competence [3]. Metastasis mainly occurs due to genome instability, the loss of effective DNA repair, checkpoint control failure and gene expression regulation [4]. Hence, genome comparisons between primary tumor and distant metastasis show distinct mutations in metastatic cells compared with the primary tumors [4].

Recently, several studies have established the importance of alternative transcription and alternative splicing in promoting the initiation and maintenance of cancer cells [5–10]. The use of alternative promoters and/or alternative termination sites can result in multiples pre-mRNA that undergoes splicing, generating a huge diversity of transcripts from a single gene [6]. During splicing, premature mRNA (pre-mRNA) is processed and exons are assembled together to compose a mature mRNA, while the remaining introns are subsequently degraded [11]. Additionally, some sequences can be alternatively included (or excluded) and differentially combined to give rise to different mature mRNA isoforms via a process named alternative splicing [11]. This event may have different consequences in the final product: (i) the alternative splicing does not alter the reading frame, and the resulting in-frame protein may have (or not) a different function from canonical protein, or (ii) alternative splicing causes a frameshift, which leads to the generation of a premature termination codon (PTC) that is directed to nonsense-mediated decay (NMD), a different protein arrangement or a truncated protein production [12,13]. It is estimated that reading frame alteration occurs approximately in 50% of all alternative splicing events [14].

In this sense, alternative splicing provides proteomic diversity through the production of multiple mRNA transcripts from protein-coding genes [11]. More than 95% of human genes are differentially spliced depending on the tissue/cell type or signal transduction in response to developmental stages and environmental stimuli [11,15,16]. Currently, the role of alternative splicing has been associated with different diseases, and disturbances in alternative splicing regulation are considered a hallmark of cancer [10,15–17].

This review is focused on describing the influence of alternative splicing during the metastatic cascade, focusing on current knowledge about alternative transcripts that may be intrinsically related to metastatic competence and how targeting specific isoforms that contribute to metastatic progression may be a promising therapeutic strategy.

2. General view of splicing

Alternative RNA isoforms are regulated by genetic and epigenetic mechanisms and can be formed by means of alternative promoters or by undergoing alternative splicing [6]. In this section, we will briefly summarize these mechanisms, focusing on alternative splicing. For more information, please refer to [6,11,18–20].

The alternative transcription mechanism is defined by the use of alternative promoters and transcriptional termination sites, resulting in multiple pre-mRNA [6]. Its regulation involves the presence of cis-regulatory elements, DNA methylation, histone modifications and chromatin remodeling [6]. Transcription and RNA processing are highly interdependent events and occur simultaneously [6,11]. During transcription, RNA polymerase II (RNA Pol II) elongation produces pre-mRNA, which undergoes cotranscriptional splicing, resulting in mature mRNA, during which pre-mRNA processing introns are removed and exons are joined together [6,19]. However, some exons and introns can be alternatively spliced, generating different isoforms with distinct activities, and such transcriptome variability is observed at the protein level [10]. This process is catalyzed by the spliceosome, a large complex of RNA composed of five small nuclear ribonucleoproteins (snRNPs; U1, U2, U4, U5 and U6) and several associated proteins cofactors, resulting in a highly dynamic conformation [5,21].

The spliceosome recognizes short consensus sequences, called splice sites, in nascent transcripts located at the intron-exon junctions, which are 5ʹ splice sites, 3’splice sites, and a branch point region near the 3’splice site [5,21]. These splice sites can be classified as ‘strong’ when they are similar to consensus sequences and, thus, more efficiently recognized [21]. In general, cognate splicing factors tend to more efficiently recognize ‘strong’ splice sites, which leads to constitutive splicing [21]. However, neighboring ‘strong’ and ‘weak’ splice sites compete with themselves for affinity with the splicing factors, and different positions of these splice sites can cause different alternative splicing outcomes [21]. Thus, additional splicing regulation is performed by trans-acting factors and cis-regulatory sequences [21]. Cis-regulatory sequences include exonic and intronic splicing silencers (ESSs and ISSs) and enhancers (ESEs and ISEs) (Figure 1(a)) [21]. It is common for alternative exons to have weaker splice sites and ESE, indicating an evolutionary advantage of the use of alternative exons [22].

Figure 1. (a) Sequences located at the intron-exon junctions, termed 5ʹ and 3ʹ splice sites, together with branch point positioned near the 3ʹ splice site, are responsible for defining the intron and exon limits. These sequences are recognized by the spliceosome machinery and associated factors. Additional regulation of the site splice choice can be mediated by cis-regulatory sequences, including exonic and intronic splicing silencers (ESSs and ISSs) and enhancers (ESEs and ISEs), and trans-acting factors, such as serine- and arginine-rich (SR) proteins and heterogeneous nuclear ribonucleoprotein (hnRNP). These splicing factors are able to inhibit or promote the recognition of nearby splice sites affecting the activity of the spliceosome components, such as the snRNPs U1 and U2. (b) Representation of the five main alternative splicing events.

Trans-acting factors bind to cis-regulatory sequences to either promote or prevent spliceosome assemblement at an adjacent splice site [5,11,21]. Most of these factors are serine- and arginine-rich (SR) proteins and heterogeneous nuclear ribonucleoprotein (hnRNP), and numerous tissue-specific factors and additional proteins have been described to stimulate or inhibit splicing (Figure 1(a)) [5,11,21]. SR proteins and hnRNPs are described to act as proto-oncogenes and tumor suppressors. For example, the RNA-binding protein hnRNP I is a pre-mRNA splicing regulator in alternative splicing and is related to biological processes, such as cell structure and motility, cell cycle, immunity and others [23]. Its overexpression is observed in colorectal, brain, ovarian and breast cancers, and it has been related to tumor aggressiveness, especially in glioma and ovarian tumors [23]. By contrast, hnRNP I inhibition decreases cell proliferation, migration and invasion in colorectal cancer cells [23].

Another example is the SR splicing factor SRSF1, which is related to malignant transformation and is overexpressed in lung, breast and colon cancers [24]. This protein leads to activation of the mTOR pathway, inducing the missplicing of MNK2 and RPS6KB1 and the alternative splicing of BCL2L11 and BIN1 to produce isoforms without pro-apoptotic activity [24]. In addition, overexpression of SRSF3, SRSF6 and SRSF10 is also related to tumorigenesis [24]. Although SR proteins generally act as oncogenes, hnRNPs can exert both roles, suppressing and promoting tumorigenesis [24].

Alternative splicing regulation is typically influenced by three different aspects of co-transcriptional RNA processing: (i) CTD (RNA Pol II carboxy-terminal domain)-mediated recruitment, (ii) kinetic competition and (iii) chromatin structure [21,25]. In CTD-mediated recruitment, splicing factors are recruited to transcription sites by associating splicing factors with the CTD [21,25]. In kinetic competition, the RNA Pol II elongation rate influences the output timing of splice sites and regulatory sequences from the nascent pre-mRNA [21,25]. In this sense, a fast RNA Pol II elongation rate limits kinetic competition and favors the inclusion of exons flanked by strong splice sites, while weaker splice sites are favored by a slow elongation rate [25]. The coupling of transcription and splicing processes is suggested to be associated with the effects of histone modification and nucleosome positioning [25]. In fact, histone post-translational modifications are related to active transcription, such as histone H3K36me3, which is enriched in exonic compared with intronic regions and is implicated in the alternative splice site choice [21,25]. In human kidney cancer, mutations in SETD2, a histone methyltransferase that specifically catalyzes H3K36me3, promote increased chromatin accessibility and lead to widespread changes in RNA processing and aberrant splicing [26]. In this tumor, chromatin accessibility is pronounced in the proximal region of misspliced exons, where the deficiency of H3K36me3 is related to decreased nucleosome occupancy of these exons [26]. In fact, evidence provides a key role for nucleosomes in splice site recognition and exon definition [25]. Moreover, the roles of nucleosome are enriched in exons, preferentially exons flanked by weaker splice sites [25]. Thus, nucleosome positioning contributes not only to exon definition but also to splicing regulation [25].

Currently, five main types of alternative splicing are recognizable:(i) exon skipping; (ii) alternative 5ʹ splice site; (iii) alternative 3ʹ splice site; (iv) mutually exclusive exons; and (v) intronic retention (Figure 1(b)) [11]. In vertebrates and invertebrates, the most prevalent alternative splicing type is exon skipping, which represents approximately 30% of all alternative splicing events [11]. However, mutations in splice sites and regulatory sequences may result in aberrant intron retention with pathological consequences [13]. Actually, intron retention is enriched in tumor-suppressor genes when somatic single-nucleotide variants (SNVs) are present [27]. Consequently, these transcripts undergo proteins truncation or NMD by the generation of the PTC [27].

A subset of retained introns has been investigated recently, which was called exitrons because they have features of both introns and protein-coding exons [28]. Exitrons are a class of alternatively spliced introns defined as a protein-coding intron sequence within protein-coding exons, whose regulation is made by internal splice site motifs in the exon [12,28,29]. They have different characteristics from others introns, such as the absence of stop codons, high GC content, a preferential size that corresponds to multiples of three, and prevalence of synonymous substitutions, also frequently observed in exons [29]. Unlikely of retained conventional introns transcripts that are inclined to stay in the nucleus, retained exitrons transcripts are transported to the cytoplasm and translated, being enriched in disordered regions of the codified proteins [12,28,29]. Marquez et al. [28] investigated the role of exitrons in ERBB2-positive breast cancer and identified 29 cases with differential exitron splicing between breast cancer and normal tissue, including the suggested tumor suppressors CIZ1, ZNF238 and EXT1.

Alternative splicing precision, efficiency and plasticity generate the vast transcriptomic and proteomic diversity required for eukaryotic cell biology. However, disturbances in this regulation can lead to incorrect decisions and, consequently, disease development.

3. Metastasis and implications of alternative splicing for the metastatic cascade

Cancer metastasis is described as a dissemination of malignant cells from a primary tumor site to different organs through the blood and lymphatic circulation. In 1889, Stephen Paget described metastasis not as a random process, but rather, he proposed that cells of certain tumors have a preferential affinity to specific organs [2,30]. Paget hypothesized the ‘seed-and-soil’ theory of metastatic outgrowth, where the establishment and growth of malignant cell (‘seeds’) are dependent on particular organs (‘soil’) to survive and proliferate [2,30].

Successful metastasis consists of achieving the five basic steps of the invasion-metastasis cascade: local invasion, intravasation, circulation survival, extravasation and colonization (Figure 2). In epithelium-derived cancers, the cells are maintained together by cell-to-cell adherence and by adherence to the extracellular matrix (ECM) [2].

Figure 2. Cell spreading from the primary tumor involves the acquisition of capabilities to disrupt the basement membrane, invade the surrounding tissue (stroma) and enter the blood circulation. After that, the spreading cells in the circulation must survive in the hostile environment until they can reach distant organs and establish a niche to grow and form a new lesion (secondary tumor).

Once normal cells have lost their adhesion to other cells or the basement membrane, the detached cells undergo anoikis, a mechanism of apoptosis induced by cell detachment [2]. However, metastatic cells are able to overcome this phenomenon by altering the signaling pathways responsible for anoikis regulation, including NF-κB, small GTPases and its effectors, some receptor tyrosine kinases and cytoplasmic kinases, and epithelial-to-mesenchymal transition (EMT) factors [31].

Overcoming anoikis is followed by the invasion of surrounding tissues of the primary tumor, which is initiated by metastatic cells migrating through the basement membrane and ECM [32]. In this sense, malignant cells need to acquire invasive features that are frequently associated with the suppression of epithelial properties and the acquisition of mesenchymal markers to disrupt cell-cell and cell-ECM adhesion [33]. This plastic program, called EMT, is reversible and may provide the ability to the tumor cell to travel to metastatic sites and establish the growth of a secondary tumor [33]. During EMT, epithelial cells acquire a mesenchymal cell phenotype that consists of a (i) loss of adhesion structures, (ii) increased cell death resistance, (iii) altered polarity, (iv) cytoskeleton reorganization, and (v) enhanced migratory capacity [33,34]. However, the reversion of disseminated tumor cells into an epithelial phenotype is required for seeding in distant organs and the formation of a secondary tumor through a process called mesenchymal-epithelial transition (MET) [35]. Some studies using breast and pancreatic cancer cell lines have shown that EMT is not indispensable for metastasis establishment but contributes largely to its aggressiveness by increasing resistance to anticancer therapy [36].

Invasive cells undergo actin cytoskeleton reorganization and the formation of membrane protrusions enriched in filamentous actin (F-actin) that, along with adhesive interactions, are crucial for invasion and migration through the stroma [2,36]. Malignant cells exit the tumor as single cells or as clusters and enter the vasculature, spreading throughout the whole body (Figure 2) [36]. Additionally, platelets and granulocytes are recruited to form niches that contribute to metastatic progression by forming a shield around tumor cells, promoting adhesion to endothelial cells and increasing endothelial permeability [37].

These tumor cells halt their circulation in capillaries at distant sites and start to extravasate into the target organ parenchyma, where these cells adopt strategies to survive and initiate colonization (Figure 2) [3,36].

Although metastasis itself is believed to be an inefficient process, the dissemination of cells from a one-centimeter-sized primary tumor corresponds to an infiltration rate of one million cells/day in the circulatory system [32]. However, due to the limited compatibility of distal sites (‘seed-and-soil’ theory), only < 0.1% of tumor cells reach an appropriate site and are able to survive in the new ‘soil’ and avoid immune surveillance [32]. This low metastatic competence can be explained by the death of tumor cells due to their inability to resist the many types of stresses, such as reactive oxygen species, shear stress, hemodynamic forces, stromal composition, detachment, and immune response [38,39]. For these reasons, only a small population of tumor cells are able to exert an adaptive response to the environmental stress, thus carrying a metastasis-initiating potential [38].

Many splice events can be used to distinguish normal and tumor tissue and contribute to an adaptive response of tumor cells and the success of the metastatic cascade. These contributions frequently alter normal biological processes, such as cell cycle control, cytoskeleton organization, migration, cell-cell adhesion and insulin signaling [40]. Splicing regulatory proteins have been the main investigated targets because the abnormal activity of these effectors could be the key related to pathological splicing alterations. Many articles have provided excellent reviews of master changes affecting the splicing program [24,41,42]. Moreover, cellular features acquired by tumor cells drive a splice program shift that could be exerted even beyond the sole aberrant function of splicing factors. In the next section, we will address the most relevant genes related to metastasis with reported transcriptional variants that contribute to tumor cell adaptation. A brief summary is represented in Table 1.

Table 1. Summary of the main mentioned isoforms and representative examples of therapeutic approaches to targeting them. M.C., metastasis contribution (↑ high contribution; ↓ low contribution).

Gene	Isoforms	Splicing	M.C.	Consequence	Therapeutic potential	Ref.
E-cadherin	E-cadherin exon 11 skipping	exon 11 skipping	↑	NMD	HDAC inhibitor MS-275	[44,173]
	E-cadherin exon 8 1054del83	exon 8 partial deletion of the final 83 base pairs	↑	NMD	HDAC inhibitor TSA	[45]
CEACAM1	CEACAM1-L	exon 7 inclusion	↑	Lower survival rate; Accelerated metastasis progression	-	[48,49]
	CEACAM1-S	exon 7 skipping	↓	Normal phenotype	-	[48,50]
CD44	CD44s	exon 6–15 (v1-v10) skipping	↑	Associated to mesenchymal phenotype	-	[54,57]
	CD44v3–10	Standard exons + variable exons 3-10	↑	Chemotaxis induced by OPN	Anti-CD44v antibodies (non-specific to v3-10) CD44v-xCT system (non-specific to v3-10) Soluble CD44v3–v10–Fc fusion protein	[61,174,175]
	CD44v6	Standard exons + variable exons 6	↑	Induce migration and expression of mesenchymal markers; Related to drug resistance	CD44v-xCT system (non-specific to v6) Anti-CD44v6 antibodies Package CD44v6-shRNA/nanoparticles Competitive peptide targeting aa in exon 6	[174–177]
Mena	Mena^11a	inclusion of exon 11	↑	Epithelial phenotype	-	[68]
	MenaΔv6	exclusion of exon 6	↑	Increase invasiveness and drug resistance in the absence of Mena^11a	-	[71]
	Mena^INV	inclusion of exon ‘INV’	↑	Discohesive cell morphology; Cell motility	-	[69]
Rac1	Rac1b	Exon 3b inclusion	↑	Increase cell motility; Cell transformation	Sanguinarine; Ibuprofen	[79,83,178,179]
CTTN	CTTN-A	Exon 11 inclusion	↑	Cell migration	-	[23]
CPEB2	CPEB2B	Exon 4 inclusion	↑	Anoikis resistance	-	[89]
KAI1	KAI1-SP	Exon 7 exclusion	↑	Increase cell motility and cell-ECM adhesion	-	[90]
CD147	CD147/Bsg-2	-	↑	Promote cell proliferation and invasiveness	-	[94]
	CD147/Bsg-3	-	↓	Inhibit proliferation and invasion	-	[94]
RON	RONΔ155	Exon 5, 6 and 11 deletions	↑	Cell transformation and tumor growth	-	[97,98]
	RONΔ160	In-frame deletion of exon 5 and 6	↑	Cell transformation and tumor growth	-	[97,98]
	RONΔ165	In-frame deletion exon 11	↑	Cell motility and invasion	-	[97,98]
	SF-RON	Alternative promoter	↑	Increase cell proliferation and invasiveness	ASLAN002	[95,96,180]
IL-7	IL-7δ5	Exon 5 deletion	↑	Promote cellular proliferation and EMT	-	[107]
CXCR3	CXCR3-A	Formed by exon 1 and 3	↑	Increase proliferation, invasion and migration	AMG487	[109,110]
	CXCR3-B	Formed by exon 2 and 3	↓	Inhibits proliferation and migration	-	[110]
FAK	FAKΔ26	Exon 26 deletion	↑	Inhibit apoptosis	-	[119]
OPN	OPN-a	Full-length	-	Promote cell migration and growth	-	[124]
	OPN-b	Exon 5 deletion	↑	Increase cell migration; Pro-survival	-	[125]
	OPN-c	Exon 4 deletion	↑	Decreased cell migration; Increase detachment	-	[122,125]
FN	EDA-FN	Exon EDA inclusion	↑	Related to angiogenesis, migration and proliferation	Vaccine targeting EDA-FN antigen	[129,131,181]
	EDB-FN	Exon EDB inclusion	↑	Tumor growth and angiogenesis	peptide CTVRTSADC (ZD2); monoclonal antibody BC-1	[129,182,183]
	MSF	Intron 12 retention followed by cleavage events	↑	Highly chemotactic	-	[135]
TP53	Δ133p53β	Initiation of transcription in intron 4 and alternative intron 9 splicing	↑	Promote cell invasion, chemoresistance and expression of pluripotency factors	-	[139,140]
Prrx1	Prrx1a	-	↑	Contribute to MET	-	[35,146]
	Prrx1b	-	↑	Promote invasion and EMT	-	[35,146]
EGFR	EGFRvIII	Exon 2–7 deletion	↑	Contribute to malignant transformation	peptide vaccine rindopepimut, CAR-T targeting EGFRvIII	[150,151,169]
	de4 EGFR	Exon 4 deletion	↑	Contribute to malignant transformation, increase invasiveness	-	[147,150,151]
Id1	Id1a	-	-	Increase proliferation	-	[152]
	Id1b	-	↑	Contribute to stem cell-like phenotype and cell quiescence	-	[152]
XAF1	XAF1F	Lacks exon 5 and contain an exon 3-ext	↑	Related to cell invasion	-	[155]
KLF6	KLF6-SV1	-	↑	Associated to mesenchymal phenotype	polyethylenimine/si-SV1 intratumoral injection	[159,160,184]
Exo70	Exo70-M	Lacks part of exon 8 relative to Exo70-E	↑	Promote invasion	-	[161]
Merlin	Merlin^Δ2−4	Exons 2–4 deletion	↑	Promote migration and invasion	-	[163]
CD99	CD99wt	Full-length	↓	Associated to less migrating phenotype and higher adhesion	-	[164]
	CD99sh	Deletion of a fragment and insertion of 18 bp at the boundary of exons 8 and 9, introducing an in-frame stop codon	↑	Stimulate cell motility	-	[164]

3.1. E-cadherin

E-cadherin is an adhesion regulator that is also known as an important tumor suppressor, which, along with catenins, maintains cell-cell adhesion and cell-matrix interactions [30]. During EMT, cadherin type switching occurs, in which epithelial E-cadherin is downregulated and the mesenchymal marker N-cadherin is upregulated [43]. Nonetheless, a study has demonstrated that E-cadherin downregulation can result from missplicing [44]. Sharma et al. [44] showed that E-cadherin exon 11 is skipped, resulting in a frameshift and PTC that will direct the transcript to NMD. The splicing factor SRSF2 has been suggested to be responsible for the aberrant splicing of E-cadherin, since in cancer cells with skipped exon 11, SRSF2 is upregulated, and its silencing increases wild-type E-cadherin RNA levels [44].

In gastric cancer, the lack of the final 83 base pairs of exon 8 (1054del83) results in a nonfunctional transcript [45]. Although it does not result from abnormal activity of SRSF2, this phenomenon is related to decreased acetylation of histones H3 and H4K16Ac on the E-cadherin gene surrounding the alternative exon 8 [45]. H3K36me3 surrounding the exon 8 region is correlated with abnormal splicing of this same region, which indicates that epigenetic modifications patterns can be critical in gastric cancer [45].

3.2. CEACAM1

Carcinoembryonic antigen-related cell adhesion molecule-1 (CEACAM1) is a transmembrane glycoprotein involved in cell-cell adhesion, and its abnormal expression is associated with a variety of human malignancies [46,47]. CEACAM1 alternative splicing generates 12 isoforms, resulting in three C2-like domains and forming isoforms that differ in the length of the extracellular region [48]. Furthermore, CEACAM1 splicing is able to encode two major cytoplasmic domains that differ by the inclusion (termed long (-L) tail) or exclusion (termed short (-S) tail) of exon 7, respectively, forming CEACAM1-L and CEACAM1-S [47]. Dery et al. [47] observed that stress-induced cytoplasmic accumulation of hnRNP A1 leads to the formation of CEACAM1-L in breast cancer and, consequently, leads to an inappropriate deregulation of CEACAM1-S expression. The prevalence of CEACAM1-L isoform expression is associated with accelerated metastasis progression and a lower survival rate in patients with colon tumors [49]. By contrast, transfection of CEACAM1-S has been described to mediate the reversion of mammary tumor cells to a normal phenotype, as noted by acini structures with the central lumen formation via the apoptosis mechanism [50].

3.3. CD44

The cell surface glycoprotein CD44 also plays an important role in cell adhesion and EMT, in addition to being involved in proliferation, migration and invasion during metastasis, in an extracellular matrix ligand-dependent manner [51,52]. CD44 pre-mRNA has 20 exons and shows several isoforms encoded by standard exons 1–5 (c1-c5) and 16–20 (c6-c10), whereas exons 6–15 (v1-v10) are considered variant exons that are regulated by alternative splicing [52]. In this sense, the inclusion and exclusion of a variant exon represent tissue and environment-specific-dependent factors, for which a single or combined exons can be selected to compose the final transcript [52]. Through these variable combinations, CD44 can interact with different cognate receptors and cooperate in the activation of several signaling pathways that contribute to tumor progression and dissemination [53]

In mesenchymal cells, variant exons are skipped, and only standard exons are included (termed standard isoform or CD44s) [52]. Nonetheless, proliferating and epithelial cells predominantly show variant-exon-encoded sequences (termed the variant isoform or CD44v) [52]. Both CD44s and CD44v may be associated with cellular malignancy depending on their partners, type and tumor stage. For example, CD44v is enriched in normal breast tissue and in lower-grade breast tumors; however, a switch in the expression of CD44v rather than CD44s accelerates the EMT process and contributes to cancer progression [34,54]. Furthermore, CD44s has been reported to be an integral component of invadopodia, which contribute to extracellular matrix degradation and the dissemination of metastatic cells from the primary site to distant organs [55]. In pancreatic cancer, CD44s is related to the EMT phenotype, collaborating to participate in invasiveness and chemoresistance in vivo [56].

In addition, the switch from CD44v to CD44s has been correlated to the mesenchymal phenotype and is prevented by the splicing factors ESRP1 and ESRP2 [41,54,57]. ESRP1 and ESRP2 are involved in alternative splicing events of epithelial cell phenotypes, and their targets are frequently involved in cell-cell adhesion, cell motility, actin cytoskeletal organization and EMT [41,58]. These splicing factors mediate the variable exon inclusion and, consequently, the generation of CD44v transcripts [41,54,57]. Nonetheless, ESRP1 and ESRP2 downregulation by TGF-β regulates CD44 alternative splicing and also leads the phosphorylation of the T179 residue of SMAD3, mediating an interaction of this latter gene product with PCBP1, an RNA-binding protein [59]. This SMAD3-PCBP1 complex interacts with the variable exon region of CD44 pre-mRNA, inhibiting spliceosome assembly and favoring the expression of the mesenchymal isoform CD44s [59]. Finally, another splicing factor, hnRNP M, competes with ESRP1 to bind to the cis-regulatory element in the CD44 pre-mRNA, leading to the generation of CD44s by skipping the variable exons and, thus, promoting EMT and contributing to breast cancer metastasis [60].

Although CD44s is more frequently associated with cellular malignancy, CD44v has also been described in tumorigenesis. A study revealed that a subpopulation of breast cancer stem cells with increased expression of ESRP1 and CD44v (mainly CD44v3-10) had enhanced metastatic capacity to invade lung tissue [61]. Furthermore, in the lung microenvironment, OPN interacts with CD44v through the V3-V7 exon region, but not CD44s, promoting cancer cell invasion in an ESRP1-dependent manner and lung colonization [61]. In fact, CD44-dependent chemotaxis is induced by OPN, while CD44-dependent cell aggregation can be stimulated by hyaluronan, a glycosaminoglycan that belongs to the extracellular membrane [51,62].

CD44v has been correlated to malignancy in colorectal cancer stem cells, where CD44v6 expression and its cooperation with c-MET drive a highly tumorigenic population with metastatic potential and have been correlated with a poor prognosis [63]. This expression is regulated by HGF, OPN and SDF-1, which induce the expression of EMT-related genes [63]. Furthermore, CD44v6 acts as a coreceptor and activates c-MET and VEGFR-2 in response to HGF and VEGF-A, inducing endothelial cell migration, sprouting and tubule formation [64]. Some SR proteins (SRSF1, SRSF6 and SRSF9) also regulate CD44 splicing, inhibiting formation of the CD44v6 isoform, while SRSF2 reduces the expression of V6 but increases the expression of isoforms containing exons V6-10 and V6, V8-V10 [65]. CD44v6 has been suggested to be important for drug resistance since its knockdown prevents colony formation in response to several antitumoral drugs [66]. In addition, CD44v6 knockdown reduces the expression of mesenchymal markers, such as vimentin, Snail1/2 and Twist [66]. Another study showed that hnRNP LL is involved in the regulation of CD44 splicing in colorectal cancer since the knockdown of this splicing factor leads to increased expression of CD44v6 [67]. In colon cancer cell lines and clinical samples, the downregulation of hnRNP LL is associated with EMT and contributes to the increased cell invasion ability in these samples [67].

3.4. Mena

Human enabled homolog (Mena), a relative of enable/vasodilator-stimulated phosphoprotein (ENA/VASP), is an actin regulatory protein that influences cell adhesion, protrusion, motility and invasion [68]. The upregulation of this protein has been linked to cell malignancy in breast, pancreas, colon and cervical cancers [68]. In agreement with this statement, Mena deficiency is related to reduced invasion, intravasation and metastatic dissemination [68]. The variant isoforms of Mena that are produced during tumorigenesis are Mena^11a, MenaΔv6 and Mena^INV [68–71]. The first one is an epithelial-associated isoform that has 21 additional amino acids in the EVH2 domain, which is encoded by exon 11a and is described in highly cohesive and well differentiated tumor phenotypes, whereas is downregulated in invasive cells and ones that have undergone EMT [68,69]. In fact, the epithelial regulatory splicing factors ESRP1 and ESRP2 regulate exon 11a inclusion, which contributes to epithelial phenotype and the formation of tumors with a differentiated morphology [68]. Furthermore, Mena^11a contributes to survival and resistance to phosphatidylinositol-3-kinase (PI3K) inhibition therapies [72]. MenaΔv6 lacks exon 6 and is related to increased invasiveness and drug resistance of breast tumors in the absence of Mena^11a [71]. In pancreatic cancer, TGF-β1 differentially regulates MenaΔv6 upregulation along with the downregulation of Mena^11a, promoting EMT [70].

Mena^INV has an additional 19 amino acids codified by the ‘INV’ exon. Its expression is not enriched in highly proliferative cells and reveals an inverse correlation of E-cadherin expression in breast cancer [69]. In primary tumor from breast cancer, this isoform contributes to a discohesive morphology and the loss of epithelial cell-cell contact, as well as increased cell motility [73]. The expression of Mena^INV is responsible for sensitizing tumor cells to EGF, eliciting tumor cells to a more invasive phenotype in low EGF microenvironments [68,69]. Additionally, a subpopulation of highly motile cancer cells can respond to the chemotactic stimuli provided by macrophages through EGF and promotes CSF1 production by tumor cells [74,75]. This motile tumor cells show increased expression of both Mena^classic and Mena^INV and decreased expression of the Mena^11a isoform [75,76]. Mena-expressing tumor cells and macrophages mutually exert chemotaxis to each other so tumor cells can co-migrate with macrophages towards blood vessels and, consequently, intravasate at sites enriched with perivascular macrophages [74,75]. These intravasation sites are called tumor microenvironment of metastasis, and the number of these regions are positively correlated with the Mena^INV abundance [75,76]. Furthermore, macrophages also can facilitate cancer cell intravasation by physically interacting with perivascular macrophages and tumor cells, increasing the formation of invadopodia by inducing RhoA activity and leading to macrophage-dependent transendothelial migration [76]. In this case, Mena^INV was observed to stimulate the maturation of invadopodia by promoting the phosphorylation state of CTTN at the tyrosine 421 residue and inhibit normal dephosphorylation mediated by phosphatase PTP1B [77].

3.5. Rac1

Rac1 belongs to the Rho family of small GTPase and acts by cycling between an inactive GDP-bound state to an active GTP-bound state in response to different stimuli [78]. Rac1 is crucial in the regulation of the cell cycle, cytoskeletal dynamics and formation of filopodia, lamellipodia and stress fibers [79]. Rac1 mRNA is alternatively spliced by the inclusion of exon 3b, generating Rac1b, which shows the insertion of 19 amino acids at the end of the switch II region, leading to a predominantly GTP-bound state, preventing the inactive state [78]. This mRNA isoform is overexpressed in colorectal, lung, pancreatic, thyroid and breast cancers [79,80].

In breast cancer, hnRNP A1 prevents Rac1b expression, whereas the activation of SRSF1 (through phosphorylation mediated by SR kinases (SRPK1)) in colorectal cancer is related to increased expression of Rac1b [81,82]. A study has shown that upon EGF treatment, hnRNP A1 is ubiquitylated and retained in the cytoplasm, preventing its repressive splicing activity [83]. Furthermore, EGF signaling stimulates the action of SR kinases, which in turn leads to the activation of SR proteins that mediate exon 3b inclusion, generating Rac1b and promoting cell migration [83]. Conversely, Rac1b promotes growth transformation, causing a loss of density and anchorage-dependent growth regulation in the NIH3T3 cell line in comparison to Rac1 [79].

However, Rac1b lacks some Rac1 functions and is incapable of forming lamellipodia or interacting with PAK1 and Rho-GDI [78]. Rac1b also induces NF-κB activation, which stimulates proliferation and apoptosis resistance in neoplastic thyroid cells [78,84].

3.6. CTTN

Cortactin (CTTN) is a cytoplasmic F-actin binding protein and acts as a Src substrate, participating in the assembly and cross-linkage of F-actin, both of which are controlled Rac1 activity [85,86]. Branched actin offers structural support for the plasma membrane and, in response to a specific stimulus, provides lamellipodial protrusions by branching of existing filaments, which requires the activity of CTTN to promote Arp2/3 complex activation leading to actin nucleation [85]. CTTN has three transcriptional isoforms: CTTN-A, CTTN-B and CTTN-C [23]. CTTN-A is the dominant isoform and has a region that consists of six motifs containing tandem amino acids repeats, where the 3^rd to 6^th units are codified by exons 8, 9, 10 and 11, respectively [87]. These six units represent the full functional domain responsible for filamentous actin-binding, cross-linking and cell migration [23,87]. CTTN-B and CTTN-C have five and four CTTN repeats, respectively, and bind to F-actin less robustly [87]. In this sense, cells overexpressing CTTN-B and CTTN-C exhibit reduced migration in comparison to cells overexpressing CTTN-A due to the absence of the full F-actin binding domain [23]. In colorectal cancer, hnRNP I promotes the inclusion of exon 11, leading to the overexpression of CTTN-A in comparison to normal tissue and increasing cell migration and invasion [23]. However, when hnRNP I was knocked down, the total mRNA of CTTN remains the same, although CTTN-A levels decline, indicating the preference of hnRNP I in the generation of CTTN-A in colorectal cancer [23]. This isoform is also overexpressed in B cells of patients with chronic lymphocytic leukemia, which suggests that it is related to disease progression [88].

3.7. CPEB2

In triple-negative breast cancer (TNBC), a subpopulation with anoikis resistance has been observed to exhibit altered pre-mRNA splicing of cytoplasmic polyadenylation element binding 2 (CPEB2), a translation regulator [89]. CPEB2 is associated with TNBC resistance by regulating cellular stress response [89]. CPEB2 is localized to RNA stress granules, where it binds to specific transcripts, maintaining them in an inactive state [89]. During the stress response, dissociation of the CPEB2 target transcript occurs, and the transcript is translated [89]. However, altered alternative splicing of CPEB2 leads to CPEB2B isoform expression in a subpopulation of TNBC cells [89]. This alternative isoform includes exon 4, is sufficient to promote anoikis resistance and is associated with higher metastasis and mortality rates [89].

3.8. KAI1

The transmembrane glycoprotein Kangai1 (KAI1, also known as CD82) has been described as a metastasis suppressor, and its loss or downregulation is correlated with a poor prognosis in prostate, lung, gastric, breast, bladder, colon, skin, cervix, endometrium and ovary tumors [90,91]. The expression of KAI1 reduces cell motility and invasiveness through the crosstalk with integrin signaling, adherens and tight junctions components [92]. A splice variant of KAI1 (KAI1-SP) has been detected in gastric cancer and associated with a poor prognosis and increased invasive potential in vivo [90,91]. KAI1-SP lacks exon 7 encoding 28 amino acids from the distal part of the second extracellular loop (ECL2) to the proximal part of the fourth transmembrane region [90,91]. Colon carcinoma cell lines transfected with KAI1-SP display a greater binding capacity to ECM components, such as type IV collagen and fibronectin, compared to those transfected with wild-type KAI1 [91]. In this study, KAI1-SP transfected cells showed increased motility mediated by cell-ECM adhesion components (especially fibronectin) and a weaker interaction with integrin α3β1, the major receptor for laminin [91]. Successful cell migration depends on an intermediate level of cell adhesiveness to the underlying matrix, but stronger cellular attachment may be related to cell immobilization [90]. In this sense, fibronectin and its receptors have been reported to act not only in attachment and cellular adhesion but also to provide signals that regulate cellular locomotion [91]. Therefore, in vitro ovarian cancer cells expressing KAI1-SP display decreased cell adhesiveness to vitronectin and type I collagen and increased cell motility mediated by αvβ3 integrin and vitronectin [90]. Finally, KAI1-SP fails to promote E-cadherin-mediated cell-to-cell contacts, which may prevent cell spreading [90].

3.9. CD147

CD147, called basigin, is a highly enriched glycosylated transmembrane protein found on the plasma membrane of various human cancer cells and is related to the increased tumorigenicity and invasive potential of such cells [93]. The most known activity is related to the expression, secretion, and activity regulation of matrix metalloproteinases (MMPs) [94]. A study has shown that CD147 expression increases cancer cell viability by modulating intercellular contacts through fibronectin assembly and preventing anoikis by downregulating Bim via the MAP kinase pathway [93]. CD147 has four isoforms, termed CD147/Bsg-1 to 4. Among them, the most predominant is CD147/Bsg-2, while CD147/Bsg-1 is retina-isoform specific and CD147/Bsg-3 and −4 were first identified in human endometrial cells and cervical carcinoma cell lines [94]. In hepatocellular carcinoma, the expression levels of CD147/Bsg-2 and −3 were higher compared to adjacent normal tissue [94]. However, these isoforms have contrasting effects, while CD147/Bsg-2 overexpression promotes cell proliferation and invasiveness, CD147/Bsg-3 is responsible for inhibiting the aforementioned processes, possibly by hetero-oligomerization of CD147/Bsg-3 with CD147/Bsg-2, thus preventing the invasive role of the latter [94].

3.10. RON

The recepteur d’origine nantais (RON) proto-oncogene encodes a transmembrane tyrosine kinase receptor and is activated by interacting with MSP, leading to RON kinase activity upregulation [95]. RON downstream signaling involves the regulation of many intracellular processes, such as EMT stimulation, cell proliferation, survival, migration, invasion, chemoresistance and angiogenesis [95]. In normal and malignant cells, two transcripts encoded by different RON gene promoters are observed. These transcripts are the short-form RON (SF-RON) and the full-length RON, although different isoforms can be generated through post-transcriptional modifications (mostly alternative mRNA splicing as well as protein truncation) [95].

SF-RON is generated by an alternative transcription start site controlled by a second promoter located between introns 8 and 10 [95]. This isoform is constitutively active with ligand-independent activity [96]. In ovarian cancer, SF-RON expression is related to increased tumor growth and spreading to the abdominal cavity [96]. Additionally, the RONΔ165 variant is formed by an in-frame exon 11 and results in the accumulation of the single-chain pro-RONΔ165 [97]. Furthermore, RONΔ165 is constitutively phosphorylated by oligomerization due to the deletion of exon 11 and is involved in mobile and invasive phenotypes [97,98].

RONΔ170 is a defective variant, the alternative splicing of which generates the loss of exon 19, which leads to the absence of kinase activity, inhibiting the oncogenic activity mediated by another oncogenic variant, RONΔ160 [97,98]. RONΔ160 results from an in-frame deletion of 109 amino acids encoded by exons 5 and 6 [95,97]. RONΔ160 and RONΔ155 are associated with tumor growth and metastasis, and RONΔ160 leads to the accumulation of β-catenin and the increased gene expression of β-catenin targets [97,98]. RONΔ155 carries deletions of exons 5, 6 and 11, and due to the loss of exon 11, the translated protein is retained in the cytoplasm [95,97,98].

Moreover, RONΔ110 is a constitutively active proteolytic truncated isoform that is present on the cell surface of cancer cells overexpressing RON, but its role in tumorigenicity it is not completely understood [95,97,98]. Among the splicing factors, SRSF1 is responsible for inhibiting the inclusion of exon 11, while hnRNP A2/B1 and hnRNP H are also described to mediate glioma cell transformation through exclusion of exon 11 from the proto-oncogene RON [99–101].

3.11. TPM1

Tropomyosin-1 (TPM1) belongs to a family of actin-binding proteins that provide stability to actin filaments in non-muscle cell cytoskeleton and the contractility of skeletal and smooth muscle cells, as well as affecting integrin expression and activity [102–104]. Previous studies have shown that TPM1 is downregulated in bladder, breast, colorectal, esophageal, oral cavity and oropharynx, ovarian and renal cancers [102,105]. Apoptosis evasion and cell invasiveness may be associated with disrupted cytoskeleton structures and components. Hence, TPM1 seems to have a crucial role as a potential tumor suppressor by mediating cytoskeletal reorganization and preventing metastasis [102,105,106]. In this sense, the restoration of TPM1 expression in breast cancer cells promotes cytoskeletal reorganization and recovery of cell-ECM interactions, reestablishing cell adhesion and promoting anoikis of detached cells [102].

The TPM1 isoforms are classified according to the number of amino acids in length as high molecular weight (HMW) or low molecular weight (LMW) and result from alternative promoter usage [103]. The use of exon 1a and 2a (or 2b) generates HMW, whereas exon 1b is used for the LMW isoform [103]. Furthermore, alternative splicing of exon 6 may occur via mutually exclusive splicing of 6a (or 6b) and exon 9, and 14 isoforms have been identified to date [103,106]. Among them, TPM1δ and TPM1γ are HMW expressed in normal human breast epithelial cells (but downmodulated in human breast cancer), the product of which is incorporated into stress fibers [106]. In fact, the downregulation of HMW isoforms is related to increased anoikis resistance, and tumorigenesis and malignant cells often undergo to a switch from HMV to LMW [103,104]. By contrast, the HMW isoform TPM1λ is upregulated in malignant breast cell lines and shows an inverse correlation with stress fiber formation [106].

Other tropomyosin isoforms, such as the LMW TM5NM1 and TM4, have been related to metastasis progression in melanoma and breast cancer, respectively [103]. This information indicates that their regulation is crucial for the fate of malignant cells that can potentially undergo anoikis when specific isoforms are expressed or possess increased invasiveness through isoform switching.

3.12. IL-7

Interleukin 7 (IL-7) is a cytokine related to tumor development and progression and, together with its receptor IL-7R, has been identified in several cancers, such as colorectal, renal, breast, lung and central nervous system cancers [107]. An IL-7 variant lacking exon 5 (IL-7δ5) has been related to promoting cellular proliferation and cell cycle progression in breast cancer via PI3K/Akt pathway activation [107]. Furthermore, IL-7δ5 is able to induce EMT through the PI3K/Akt pathway, in addition to exogenous IL-7δ5, that promotes downregulation of E-cadherin and upregulation of N-cadherin expression [107].

3.13. Chemokines

Chemokines are chemotactic cytokines that are subdivided into four main classes (CC, CXC, C and CX₃C) according to the position of the first two cysteine residues [108]. Considering the tumor microenvironment, chemokines can be secreted by tumor cells or stromal and immune cells [108]. Specific chemokines drive immune cell homing to the tumor microenvironment, influencing immune-cell infiltration and affecting tumor progression [108]. Furthermore, chemokines can target tumor and stromal cells when these cells express chemokine receptors and are able to respond to different chemokine gradients [108]. In this sense, chemokines affect cell proliferation, invasiveness, cancer stem-like cell properties and metastasis [108]. Several chemokine receptors have been described to undergo aberrant splicing, which may contribute to tumor progression and metastasis.

CXCR3 is a receptor for CXC chemokines and is activated by the ligands CXCL4/PF4, CXCL9/MIG, CXCL10/IP10 and CXCL11/IP9/I-TAC [109]. The activity observed for this receptor is related to its two major splice isoforms: CXCR3-A, formed by exon 1 and exon 3 and considered the classical isoform, and CXCR3-B, which is assembled by exon 2 and exon 3 [110]. Interestingly, these isoforms have opposing functions, CXCR3-A is a pro-proliferative isoform that is related to invasion and migration promotion in several tumors, while CXCR3-B has no chemotactic functions and inhibits migration [109,110]. CXCR3-A, activated by its ligand CXCL9, has been shown to change the levels of ERK1/2 phosphorylation, a member of MAPK signaling, leading to the overexpression of MMP2 and MMP9 and, consequently, contributing to the invasion and metastasis of CD133⁺ liver cancer cells [111]. CXCR3-A is expressed at higher levels than CXCR3-B in malignant cells and supports cell proliferation and migration, invasiveness and metastasis, whereas CXCR3-B reduces the metastatic potential and colonizing capacity [109,110]. However, in bulk populations, CXCR3-B inhibits cell proliferation and shows a supportive role for cells with cancer stem-like properties [110]. In this sense, CXCR3-B may contribute to resistance in therapies that target highly proliferative cells after those stem cells have achieved a low proliferation rate, contrasting with general populations of malignant cells [110].

SDF-1, also known as CXCL12, is a chemokine member of the CXC class and the ligand of the chemokine receptors CXCR4 and CXCR7 [112]. Serum SDF-1 concentration is positively correlated with the occurrence of distant metastasis in patients with gastric cancer [113]. This chemokine has six alternative spliced isoforms, comprising SDF-1α, SDF-1β, SDF-1γ, SDF-1δ, SDF-1ε and SDF-1θ [112]. In tumors, there is an increase in the expression of the SDF-1α G801A polymorphism associated with lymph node metastasis in patients with colorectal cancer, whereas SDF-1β is involved in angiogenesis and, together with SDF-1γ, is related to tumor size in patients with metastasis [112].

3.14. Integrin

Integrins are the major cell surface receptors and are responsible for mediating cell adhesion to the ECM [114]. They are crucial for the maintenance of homeostatic balance in the cell, but are also involved in inflammation, leukocyte trafficking, tumor metastasis and wound repair [114]. Integrins can transmit an ‘outside-in’ signal through the plasma membrane, as well as an ‘inside-out’ signal, functioning as an important intermediate a bidirectional communication between cells and their microenvironment [114]. In tumors, integrins regulate different processes through the tumor cell-microenvironment interface, influencing the cancer response and progression [114].

The α6 integrin subunit (CD49f) can heterodimerize with either β1 or β4, forming α6β1 or α6β4 integrins and then acting as the main laminin receptor [115,116]. The α6 subunit undergoes alternative splicing, forming two cytoplasmic isoforms: α6A, which is enriched in epithelial populations, and α6B, which is enriched in mesenchymal populations, where the latter has an important role in slow-cycling stem cell functions and is directly involved in the stem cell phenotype; it is related to a poor outcome in TNBC patients [116]. In this sense, VEGF signaling can affect the alternative splicing of the α6 subunit, promoting the repression of α6A and favoring the expression of α6B via induction of the oncogene GLI1 and transcriptional inhibition of the splicing factor ESRP1 [116]. However, isoform α6A knockdown reduces tumor growth, although it does not affect α6B levels [115]. In fact, α6A has been described to have a positive proliferative effect in colorectal cancer, which is involved in the Wnt/β-catenin pathway, where the knockdown of α6A leads to enhanced phosphorylation of β-catenin by GSK3β [115].

Integrin β1 alternative splicing leads to the formation of four isoforms, called β1A, β1B, β1C and β1D [117]. The expression levels of isoforms β1B, β1C and β1D are downregulated in several tumors, whereas β1A is related to focal adhesion formation, cell spreading and motility [117]. A study has shown that β1A expression is related to metastasis in pancreatic cancer, while Fyn inhibition and hnRNP E1 overexpression promote the splicing of β1C, which reduces the migration and invasiveness of tumor cells [117].

3.15. FAK

Focal adhesion kinase (FAK) mediates cell communication with the extracellular environment, triggering intracellular signaling in response to ECM remodeling, nutrient availability and growth factors [118]. Furthermore, its action is involved in cell adhesion, motility, survival, vasculogenesis and proliferation [118]. During tumorigenesis, FAK overexpression is frequently observed in breast, colon, cervical, ovary and prostate cancers, and it is associated with metastasis [118,119]. When detached cells undergo anoikis, FAK is proteolyzed by caspase-3 and −7, and anoikis evasion is recurrently associated with increased FAK levels or mutations [119]. FAK shows four traditional alternatively spliced exons: 13, 14, 16 and 31 [119,120]. However, in human breast cancers, an aberrant alternative isoform that lacks exon 26, a region that contains the caspase-3/caspase-7-like cleavage motif, was observed [119,120]. Thus, although this isoform has the same kinase activity as wild-type FAK, the missing segment causes aberrant FAK to become resistant to caspase-mediated proteolysis, promoting tumor cell survival [119].

3.16. OPN

Osteopontin (OPN) is a glycophosphoprotein that is secreted by several cells and has been broadly overexpressed in plasma and tumor tissues, as well as described to promote tumor progression and metastasis [121]. OPN interacts with different integrins and CD44, activating signal transduction pathways involved in cell proliferation, adhesion, migration and invasion [121].

The primary transcript of OPN can be alternatively spliced in the upstream region of the central integrin binding domain and the C-terminal CD44 binding domain to form (i) the full-length OPN-a, which consists of 7 exons; (ii) the isoform OPN-b, which lacks exon 5; (iii) and OPN-c, which is generated by the lack of exon 4 [121].

A study has shown that OPN-a and OPN-b both act through integrin signaling, while OPN-c has an integrin-independent function [122]. Furthermore, OPN-b augments cell migration in vitro compared with OPN-c [122]. In the same study, stable cells transfected with OPN-b showed increased cell adhesion, proliferation and invasion, while OPN-c transfected cells exhibited decreased cell invasion and migration but enhanced cell detachment [122]. OPN-c is present in 75% of breast cancers, and it is directly related to anchorage-independent growth and metastasis [123]. In non-small cell lung cancer, OPN-a is highly expressed compared to OPN-b and OPN-c [124]. However, both OPN-b and OPN-c are overexpressed in gastric cancer and showed anti-apoptotic activity when induced by H₂O₂, but not OPN-a [125]. This pro-survival effect is promoted by OPN-b, possibly through the interaction with CD44v6 and CD44v7, which triggers downstream anti-apoptotic signaling [125].

In addition, OPN-a overexpression has been related to the promotion of proliferation, migration, adhesion and invasion in A549 cells through the interaction with ανβ3 integrin [124]. This interaction allows lung cancer cells to invade the bone matrix during metastasis, and once OPN-a is silenced, significant inhibition of bone matrix attachment occurs [124]. In thyroid cancer cell lines, increased levels of OPN-a have been related to cell growth and enhancement of cell migration and motility, possibly through the induction of MMP2 and MMP9 activity, leading to extracellular matrix degradation [126]. In papillary thyroid carcinoma, OPN-a is correlated with extrathyroid extension, vascular invasion and BRAF^V600E mutation, which are significantly related to a poor prognosis [126].

Finally, in breast cancers, all tumor-derived OPN transcriptional isoforms increase the tumor expression of the immunosuppressive cytokines TGF-β1 and MCP-1 [127]. Additionally, these tumor-derived OPN transcripts lead to the inhibition of pro-inflammatory cytokines TNF-α and increase anti-inflammatory IL-10 production by monocytes [127]. This mechanism may be related to alternative immune cell activation, which is associated with tumor-associated macrophages (TAMs) and possibly contributes to immunological evasion mediated by OPN [127].

3.17. FN

Fibronectin (FN) is a glycoprotein present in the plasma and ECM that has cell adhesiveness and migration activities [128]. Fibronectin is formed by the combination of three repeating domains, called Types I, II and III [128]. The Type III domain is alternatively spliced by the inclusion of extra domain-A (EDA) or extra domain-B (EDB) and a variable type III connecting segment (IIICS), which can be excluded (or included) completely or partially [128,129]. FN can also be found either as a plasmatic FN (pFN) or as a cellular FN (cFN), and only the latter comprises variable proportions of EDA or EDB [128].

In normal tissues, these isoforms are not detected by antibodies, but they were highly representative in ECM components of aggressive tumors and other diseases that exhibit considerable tissue remodeling, serving as markers for tumor angiogenesis [128,130]. EDA-FN has been related to the promotion of neo-lymphangiogenesis in colorectal carcinoma cells through integrin α9 activation, in addition to the stimulation of angiogenesis, cell adhesion, migration and proliferation in malignant tumors [131]. EDB-FN can be regulated by ECM stiffness, which mediates splicing through PI3K/Akt signaling activation, promoting the regulation of SR protein phosphorylation, such as SRp40 [132]. The high level of proteolytic degradation and remodeling of the stroma surrounding the tumors alters the stiffness, enabling the invasion of migrating cells [133]. This remodeling leads to the release of activated growth factors stored in the ECM, influencing cell signaling [133]. Finally, in urinary bladder cancer, IIICS shows superexpression by 21.5-fold in invasive cells compared with noninvasive cells, as well as increased de novo O-linked glycosylation [129,134].

Another alternatively spliced FN isoform known as migration stimulating factor (MSF) is overexpressed by TAMs [133,135]. The MSF isoform consists of intron 12 retention followed by a two-step series of intra-intronic cleavage events, which results in a 70-kDa truncated protein [133,136]. Furthermore, MSF is highly chemotactic for monocytes and tumor cells, stimulating cell migration and possibly contributing to metastasis, which it is not observed for full-length FN [135]. In established tumors, M2-like macrophages become TAMs and show poor cytotoxicity against tumor cells, acting in the production and secretion of chemotactic molecules that stimulate tumor cell migration to distant sites and metastatic niche preparation [135]. MSF has been reported to be upregulated in several tumors, and in addition to TAMs, MSF is also expressed by neoplastic cells and vascular endothelial cells [133,136].

3.18. Tp53

The tumor suppressor TP53 is well known for its multiple roles in cell cycle arrest, genomic stability, DNA repair, apoptosis, autophagy, differentiation, stemness, migration, drug resistance, metabolism and angiogenesis [137]. The relationship of TP53 with cancer is frequently associated with mutations that are located in the TP53 DNA binding domain (DBD) or its conformational mutants [137]. TP53 encodes twelve isoforms through different mechanisms, which involve the usage of the alternative promoter and splicing: TAp53α, TAp53β, TAp53γ, Δ40p53α, Δ40p53β, Δ40p53γ, Δ133p53α, Δ133p53β, Δ133p53γ, Δ160p53α, Δ160p53β and Δ160p53γ [137,138]. TAp53 and Δ40p53 are long TP53 isoforms that contain the transactivation domain, which it is not present in either of the short isoforms Δ133p53 or Δ160p53 [137]. Furthermore, β and γ have an additional domain with an unknown function and do not contain a canonical C-terminal oligomerization domain [137,138]. The isoform TAp53α is the canonical isoform expressed and found in normal tissues, but Δ133p53β has been shown to increase the risk of recurrence by eight times and the risk of death by three times in breast cancer [139]. High levels of Δ133p53 positively regulate the cancer stem cell potential in the MCF-7 cell line, stimulating mammosphere formation and the expression of pluripotency and stemness regulators, such as SOX2, OCT3/4, NANOG and CD24/CD44 [140]. Concomitantly, the metastatic capacity, invasiveness and chemoresistance are associated with increased Δ133p53 expression in different cancer cell lines, which is not associated with TP53 mutations; the depletion of Δ133p53 in wild-type TP53 eliminates the scattering and invasive activities, as well cancer stem cell formation [139,140].

3.19. CSF-1

Colony stimulating factor 1 (CSF-1) is a hematopoietic growth factor found in the bloodstream and has been related to the survival, differentiation and proliferation of mononuclear phagocytic lineages, as well as to macrophage spreading and motility [141,142]. CSF-1 is composed of at least 10 exons, and its mRNA undergoes alternative splicing, which it involved in the alternative use of a 3ʹ non-coding region represented by exon 9 and 10 [142]. Exon 6 is also alternatively spliced, during which a large portion of the 5ʹ end is removed in one of the alternative transcripts [142]. In addition, exon 6 has a variable length, depending on which of the two spliced acceptor sites is used, forming different transcripts with different roles [142,143]. After alternative splicing and differential post-translational proteolytic processing, CSF-1 generates three mature homodimeric proteins: a cell surface-associated glycoprotein (csCSF-1), a secreted extracellular matrix-anchored proteoglycan (spCSF-1) and a secreted soluble glycoprotein (sgCSF-1) [141,142]. spCSF-1 and sgCSF-1 are secreted isoforms and regulate cells nearby or at a distance, while cell surface isoforms act on local regulation [141,142].

CSF-1 has an important role in the chemotaxis and activation of macrophages, which are important for innate immunity and for tissue morphogenesis and/or remodeling [141]. CSF-1 is related to cancer, and its activity has been associated with the recruitment and re-education of macrophages, which are also known as TAMs [133,144]. In fact, blood circulating monocytes are recruited locally and differentiate into TAMs in response to signals from the tumor microenvironment, such as chemokines and growth factors secreted by stromal and tumor cells [133,144]. TAMs are regarded to have an M2-like phenotype and are suggested to have pro-tumor effects due to their higher levels of anti-inflammatory cytokines, scavenging receptors, angiogenic factors and proteases, promoting an immunosuppressive microenvironment and contributing to tumor progression [144,145]. This ability results from to secretion of molecules that directly affect cell growth, motility, migration, angiogenesis, intravasation, invasion, inflammatory response, ECM remodeling and proliferation [135,145].

Different CSF-1 isoforms have been described to be associated with TAM activation [145]. Elevated levels of CSF-1 have been detected in breast, ovarian, endometrial, cervical, and colorectal cancers, leukemia, lymphoid malignancies and myelodysplastic syndromes [145]. Among the isoforms, sgCSF-1 is related to poor survival outcomes in patients with colorectal and breast cancer [145]. One study detected more infiltrating TAMs in the microenvironment containing csCSF-1 and sgCSF-1 isoforms [145]. Both isoforms were able to promote macrophage re-education, but they induced different spectra of M2 phenotype-related genes [145]. In this sense, csCSF-1 is more related to the stimulation of TAM proliferation in the tumor microenvironment, whereas sgCSF-1 is more prone to the recruitment and induction of TAM differentiation [145].

3.20. Prrx1

Alternative splicing of the paired-related homeobox transcription factor (Prrx1) results in two isoforms, termed Prrx1a and Prrx1b, which differ from each other at the C-terminus but are identical in the 199 N-terminal amino acids [35,146]. Both isoforms maintain the homeobox domain, but only Prrx1a has a conserved OAR domain with potentially important functions in transactivation [35,146].

In pancreatic carcinogenesis, Prrx1 isoform switching regulates epithelial and mesenchymal phenotypes in primary and metastatic cancer: Prrx1b is upregulated in primary tumor cells, where it promotes EMT, tumor invasion and differentiation [35]. In addition, Prrx1b facilitates tumor cell invasion by upregulating HGF through signaling mediated by its receptor (c-MET), driving metastasis by promoting the invasion of tumor cells and EMT [35].

However, higher levels of Prrx1a expression have been found in metastatic cells, contributing to MET, which is crucial to the colonization of distant organs after extravasation, as well as an increased capacity for anchorage-independent growth and self-renewal [35]. Furthermore, knockdown of both isoforms in the KPC2 cell line decreased the invasiveness and self-renewal capacity, and in vivo transplantation generated a smaller primary tumor volume and a reduced number of circulating tumor cells [35].

3.21. EGFR

Epidermal growth factor receptor (EGFR) is a transmembrane glycoprotein and a member of the ErbB receptor tyrosine kinase family of growth factor receptors [147]. It is described as providing a substantial contribution to cell survival and proliferation motility [147]. Its overexpression indicates a poor prognosis in breast, lung, liver, prostate and ovarian cancers [147].

EGFR is composed of 30 exons that can be spliced to form EGFRv1, EGFRv2, EGFRv3, EGFRv4, EGFRvIII and de4 EGFR [148–150]. The EGFRvIII isoform results in a gain-of-function mutation that leads to genomic deletion of exons 2–7, and its expression is restricted to tumors, such as lung carcinoma, glioma, glioblastoma and head and neck squamous cell carcinoma [150]. de4 EGFR forms a truncated EGF variant due to a deletion of exon 4, and it has been detected in glioma, prostate and ovarian cancers, and like EGFRvIII, de4 EGFR is also found exclusively in tumoral tissues [150,151]. Both de4 EGFR and EGFRvIII have minimal EGF-binding activity and ligand-independent autophosphorylation, exhibiting similarities in their biological activities and extracellular structures [151]. Some de4 EGFR tyrosine residues are constitutively phosphorylated even without ligand activation, which leads to downstream signaling pathway activation [147,150]. In this sense, de4 EGFR promotes the upregulation of FAK, Src, ERK, Akt and MMP9, whereas E-cadherin is downregulated in ovarian cancer cells [147]. Furthermore, de4 EGFR contributes to the increase in cell proliferation and malignant transformation, stimulating metastasis-promoting activities and higher cisplatin tolerance [147].

3.22. Id1b

The inhibitor of differentiation (Id1) belongs to the bHLH family consisting of three other members, Id2 to Id4 [152]. Its role in tumorigenesis, angiogenesis and metastasis has been described, and it is associated with a poor prognosis in patients with glioblastoma and breast cancer [152]. Id1 has an important role in maintaining self-renewal and pluripotency in embryonic stem cells, and its role extends to cancer stem cells [152]. The alternative splicing of Id1 forms two isoforms: the classical isoform, known as Id1a, containing 155 amino acids; and the alternative isoform, referred as Id1b, containing 149 amino acids [152]. The difference between the two isoforms is found in the C terminal region, which alters their capacity to interact with specific transcription factors [152]. Id1a is related to increased cell proliferation, whereas overexpression of Id1b reduces proliferation, promotes self-renewal, and leads to a stem cell-like phenotype, increasing the expression of cancer stem cell markers, such as ALDH1A1, Notch-1, Sca-1, Tert, Sox-2 and Oct-4 [152]. Thus, Id1b may contribute to cell quiescence, a strategy adopted by many primary tumor or metastatic cells [152,153]. These quiescent cells are maintained in a non-proliferative state for long periods, and the awakening of these dormant cells leads to tumor progression and metastatic relapse [153,154].

3.23. XAF1

XIAP-associated factor 1 (XAF1) is a nuclear protein that when overexpressed leads to apoptosis and tumor growth inhibition [155]. XAF1 splice variants have been detected in several cancer cell lines and are regulated by aberrant epigenetic modifications [155]. The aberrant variants frequently contain a PTC, which leads to NMD [155]. However, suppression of the NMD pathway may occur due to cellular stresses influenced by the tumor microenvironment, stabilizing the NMD-target mRNA [155]. In this sense, the upregulated XAF1 variant was found to lack exon 5 and contain an extended region from exon 3, termed XAF1F [155]. XAF1F harbors a PTC and was found in highly metastatic cell lines compared to parental cells and in peripheral blood containing circulating tumor cells from patients with gastric cancer, indicating a suppression of the NMD pathway and, consequently, an accumulation of XAF1F [155]. Furthermore, although the function of XAF1F has not yet been reported, its mRNA expression levels in circulating tumor cells is correlated with venous invasion, lymph node metastasis, and the depth of tumor invasion [155].

3.24. FAM3B

Family with sequence similarity 3 member B (FAM3B) is a member of the FAM3 family of cytokine-like genes and has been described to have at least seven alternatively spliced isoforms [156]. Among them, FAM3B-258 is a non-secretory variant with 258 amino acids that is highly expressed in colon adenocarcinomas. It is related to poor cancer cell differentiation and has been shown to promote invasion and migration in vitro and metastasis in nude mice [156]. FAM3B-258 promotion of invasion is regulated by Slug-dependent activation, which inhibits the expression of JAM and E-cadherin, leading to the breakdown of cell-cell interactions by downregulating cellular adhesion proteins [156].

3.25. KLF6

The transcription factor Krüppel-like factor 6 (KLF6) encodes a zinc finger protein that regulates the cell cycle, differentiation, and apoptosis [157]. KLF6 acts as a tumor suppressor and is inactivated in many cancers [157]. KLF6 alternative splicing generates three cytoplasmic isoforms: KLF6-SV1, KLF6-SV2 and KLF6-SV3 [158]. In prostate cancer, a germline single nucleotide polymorphism located in an intronic region of KLF6 is responsible for disruption of their own splicing regulation by creating a novel functional SRp40 DNA binding site and removing two overlapping SR protein sites [158]. This process leads to increased expression of three alternatively spliced isoforms from KLF6 and an augmented risk of prostate cancer [158].

KLF6-SV1 and KLF6-SV2 are described to antagonize the activity of wild-type KLF6 during p21 upregulation and to suppress cell proliferation [158,159]. Furthermore, KLF6-SV1 promotes metastasis, and its upregulation is considered to be a poor prognostic factor for patients with prostate, lung, ovarian and breast cancers [159,160]. Its expression is detected in both healthy and tumor tissues, but it is frequently upregulated in the latter [159].

In breast cancer, KLF6-SV1 expression in primary tumors is correlated with an increased metastatic potential and EMT-like phenotype [159]. Furthermore, overexpression of KLF6-SV1 leads to increased colony formation in tumorigenic and non-tumorigenic cell lines, promoting cell survival, migration and invasion [159]. KLF6-SV1 knockdown decreases the metastatic capacity via diminished invasion and reduced expression of N-cadherin with concomitant E-cadherin overexpression, reverting the mesenchymal-like phenotype [159].

3.26. Exo70

Exocyst is an octameric protein complex composed of Sec3, Sec5, Sec6, Sec8, Sec10, Sec15, Exo70 and Exo84 [161]. This complex is crucial for exocytosis and is related to plasma membrane expansion during migration [161,162]. Furthermore, Exo70 is crucial for cell invasion by mediating MMP secretion at focal degrading sites [162]. However, during the transition of the EMT state, the epithelial Exo70 isoform (Exo70-E) is switched to a mesenchymal Exo70 isoform (Exo70-M) by the splicing factor ESRP1 [161]. Exo70-E differs from Exo70-M by an insertion of a portion of exon 8 generated by ESRP1-mediating alternative splicing [161]. This switch does not affect exocyst complex function, but it is implicated to have different functions according to the assembled Exo70 isoform [161].

When Exo70 knockdown cell lines are transfected with Exo70-M, these cells recover their invasive ability, while Exo70-E transfection fails to do the same [161]. These observations were explained by the high invadopodial activity levels in cells expressing Exo70-M, whereas Exo70-E cells had lower invadopodial activity [161]. Additionally, Exo70-M interacts with the Arp2/3 complex and stimulates actin branching, which is crucial for membrane protrusion formation during migration and invasion [161].

3.27. Merlin

Moesin-ezrin-radixin-like protein (Merlin) is a tumor suppressor that participates in the regulation of cell proliferation, motility and survival [163]. Merlin is codified by the Nf2 gene, which is heterozygous in mice and is related to the development of highly metastatic malignant tumors [163]. Merlin is important for the prevention of tumorigenesis by promoting adherens junctions stabilization and controlling cell-cell contact mediated by cadherin and activation of Rac [163]. Splicing variants of Merlin lacking exon 2 (Merlin^Δ2), exon 3 (Merlin^Δ3), or both of them (Merlin^Δ2/3) are related to their inactivity as a tumor suppressor [163]. In addition, another alternative splicing variant lacking exons 2, 3 and 4 (Merlin^Δ2−4) was identified in hepatocellular carcinoma and in portal vein tumor thrombus [163]. Merlin^Δ2−4 overexpression promotes migration and invasion by upregulating Twist1, and it impairs the ability of wild-type Merlin to bind to β-catenin and Ezrin/Radixin/Moesin proteins [163]. This isoform also leads to stemness activity by increasing spheroid formation and inducing the expression of stem-related genes, including Sox2, OCT4, Klf4, c-MYC and Nanog [163].

3.28. CD99

CD99 is a transmembrane glycoprotein that is involved in the cell adhesion, migration, apoptosis, differentiation, activation and proliferation of lymphocytes, monocyte migration, transport of several transmembrane proteins, and intercellular adhesion between lymphocytes and endothelial cells [164,165].

CD99 produces two distinct transcripts: the full-length CD99 isoform (CD99wt) and a truncated shorter isoform (CD99sh) resulting from alternative splicing [164,165]. CD99sh carries a deletion of the intracytoplasmic fragment and an 18-bp insertion at the boundary of exons 8 and 9, which introduces an in-frame stop codon [164]. Studies in breast cancer cell lines have revealed that CD99sh stimulates MMP9 activity and has an FN-augmented binding potential, in addition to increasing cell motility and invasiveness [165]. The higher invasion capacity can result in diminished cell-cell adhesion in cells expressing CD99sh and an augmented interaction between FN-CD99sh [165]. Furthermore, CD99sh upregulates the AP-1 signaling pathway, which mediates cell motility and MMP9 expression [165]. The higher activity of proteolytic enzymes (such as MMP) allows ECM degradation, contributing to cell movement and invasion [165].

In osteosarcoma and prostate cancer cells, CD99 isoforms have opposing effects: CD99wt-expressing cells show a less migrating phenotype, with organized actin filaments and greater collagen type I adhesion, laminin and fibronectin and sensitivity to anoikis [164]; CD99sh activates c-Src kinase activity, which can act as a mediator of CD99 functions in stimulating cell motility [164].

4. Therapy

The differences observed between the tumor and normal transcriptome are vast as a consequence of several changes in genomic and transcriptomic regulation. In summary, the major genomic and transcriptomic changes are as follows: (i) the presence of mutations in transcription factor binding and splice sites; (ii) mutations that generates aberrant proteins; (iii) different DNA arrangements and histone modifications; (iv) changes in chromatin accessibility; and (v) altered expression of splicing factors and transcription factors [6]. In the past, drugs that act on specific cellular targets were used without regard for the aberrant isoforms generated during oncogenesis, mostly due to missing information regarding the isoform variants in a gene [6,166]. The differential expression of several isoforms encoded by the same gene might be a contributing factor to the observed variation in patient responses during drug treatment [6,166].

Studies have shown that the analysis of isoform-level expression offers an effective discrimination of cancer and non-cancer cells, providing a better identification of different subgroups within a cancer-type and allowing a robust signature to classify patients in molecular subgroups [6,167]. Furthermore, during tumor progression, alternative splicing occurs in a disease-specific manner and contributes to the disease outcome [168]. The observation that some isoforms are detected only in tumor cells suggests that cancer-specific isoforms may be good diagnostic markers and therapeutic targets.

Validation of the isoform-specific expression and role in tumor cells highlights a therapeutic potential and an opportunity to develop drugs that target specific isoforms. However, it is important to maintain deep large-scale studies that provide information concerning whether an isoform is really cancer-specific or has a pro- or anti-oncogenic role, before developing a therapeutic design to ensure the correct drug-to-isoform (or drug-to-isoform product) assignment.

Some studies addressing cancer-specific isoforms as targets have shown promising results and are listed in Table 1. A successful example is EGFRvIII, which is the target of several immunotherapies for patients with glioblastoma, such as peptide vaccines, dendritic cell vaccination therapy, monoclonal antibodies and modified T cells [169]. Among them, the peptide vaccine rindopepimut was approved by the US FDA based on phase I and II clinical trials showing a significantly higher survival rate in patients with glioblastoma expressing EGFRvIII [169]. Advances have also been realized with cell therapy using the chimeric antigen receptor T (CAR-T) as an antitumor treatment, and CAR-T targeting the EGFRvIII isoform efficiently eliminates glioma cells [169].

The modulation of alternative splicing may reduce the severity of disease and prevent metastasis and its incidence. In this sense, different approaches to perform this manipulation have been studied, including targeting protein isoforms, isoform expression, alternative splicing through trans-acting elements, oligonucleotides specific to specific cis-acting elements in the transcript and inhibition of the specific isoform by RNAi [168]. One example is the use of compounds that bind to SF3B (a component of the U2 snRNP), leading to disruption of the early stages of spliceosome assembly [5]. These compounds are E7107, spliceostatin A and sudemycins, which promote unspliced (or incompletely spliced) pre-mRNA accumulation in the nucleus and show anti-proliferative activity in vitro [5]. E7107 was used in phase 1 clinical trials, but its use was suspended due to its high toxicity in patients [5,170]. Moreover, the spliceosome modulation in MYC-driven cancers has been suggested as an alternative to therapeutic intervention. MYC is a transcription factor that is frequently amplified in human cancers. It is highly related to cell transformation, and studies have revealed that the role for MYC as an oncogene is dependent on spliceosome activity [171,172].

5. Conclusive remarks and future directions

Although great advances have been achieved, understanding all of the mechanisms that contribute to tumor formation and progression is a great challenge. The presence or absence of specific isoforms may affect the metastatic potential of malignant cells, and variations in isoform expression from different genes are observed in cancer compared to normal tissues. Currently, the growth of large-scale studies has allowed us to integrate information and fill knowledge gaps in all areas of molecular biology – including information regarding splicing mechanisms and isoform variety. Certainly, further investigations are still necessary to understand the links between the splicing mechanisms, isoform diversity and the context of the disease. Finally, comprehension of the functional characteristics of abnormal splicing and their relationship with genomic instability is crucial for the development of novel therapeutic approaches.

Acknowledgments

This work was supported by CNPq under Grant 302969/2016–0 and CAPES.

Disclosure statement

No potential conflict of interest was reported by the authors.

Additional information

Funding

This work was supported by the Conselho Nacional de Desenvolvimento Científico e Tecnológico [302969/2016–0];