The role of splicing factors in deregulation of alternative splicing during oncogenesis and tumor progression

Abstract

In past decades, cancer research has focused on genetic alterations that are detected in malignant tissues and contribute to the initiation and progression of cancer. These changes include mutations, copy number variations, and translocations. However, it is becoming increasingly clear that epigenetic changes, including alternative splicing, play a major role in cancer development and progression. There are relatively few studies on the contribution of alternative splicing and the splicing factors that regulate this process to cancer development and progression. Recently, multiple studies have revealed altered splicing patterns in cancers and several splicing factors were found to contribute to tumor development. Studies using high-throughput genomic analysis have identified mutations in components of the core splicing machinery and in splicing factors in several cancers. In this review, we will highlight new findings on the role of alternative splicing and its regulators in cancer initiation and progression, in addition to novel approaches to correct oncogenic splicing.

Keywords:

• alternative splicing
• cancer
• hnRNPs
• SR proteins

Introduction

Approximately 95% of human genes encode more than one product.^1,2 This phenomenon is achieved by the process of alternative splicing, which is regulated by both cis and trans acting elements. The cis elements in the pre-mRNA sequence are recognized by a large family of factors called splicing factors, the trans-acting factors that recruit or repel the spliceosomal machinery to catalyze splicing at specific splice sites (for a detailed review see^3,4).

In the past 15 years, multiple studies have revealed altered splicing patterns in cancers^5-10 and several splicing factors were found to contribute to tumor initiation.^11-15 It is therefore not surprising that recent studies using high-throughput genomic and exomic analysis of a variety of cancers have revealed mutations in components of the core splicing machinery and in splicing factors.^16-22 Although identification of mutations in splicing factors does not necessarily indicate a role for aberrant splicing in tumor initiation, these findings have given rise to a growing field of research on the role of alternative splicing factors in tumor initiation and progression, as well as in invasiveness, metastasis, and drug resistance. In this review we will focus mainly on the roles of splicing factors in these processes. Other levels of regulation, which will not be discussed here, include transcriptional regulation in splicing, the role of RNA polymerase II kinetics,^23-25 and histone modifications.^26,27

Genomic Evidence for the Role of Splicing Factors in Cancer

Mutations in splicing factors

In recent years many genome-wide studies have been performed to screen for genomic mutations in tumor samples.^28-31 Some of these screens have unveiled mutations in components of the splicing machinery. The first study to identify mutations in splicing factors was that of Yoshida et al.¹⁶ In this study, 29 patients with myelodysplastic syndrome (MDS) were screened by whole exome sequencing and mutations in splicing components were identified in 16 out of 29 patients. These mutations were in the genes for U2AF1 (also known also as U2AF35), ZRSR2, SRSF2 (also known as SC35), SF3A1, SF3B1, and PRPF40B, although only one case was found for each of the last 3 genes. This initial study was expanded to include 582 patients with myeloid neoplasms, this time specifically searching for mutations in the aforementioned genes as well as in U2AF65, SF1, and SRSF1. Once again, mutations were found in genes for the splicing factors U2AF1, SRSF2, ZRSR2, SF3B1, SF3B1, PRPF40B, and U2AF65 with differential occurrence but mostly in U2AF35, SRSF2, ZRSR2, and SF3B1.¹⁶ This discovery led to multiple follow-up studies that have been performed on a variety of hematological malignancies.^{19,21,22,32-37} To date, mutations in splicing factors have been found mainly in hematological cancers, and to a lesser extent in solid tumors.³⁸ Data from these sequencing studies have been reviewed recently;³⁹ here, we will highlight the most commonly mutated genes and their relevance to tumor development.

SF3B1, a gene encoding one of the U2 snRNP components, was found to be mutated in 75% of patients with MDS,¹⁶ as well as in patients with chronic lymphocytic leukemia (CLL)³⁵ and to a much lesser extent in solid tumors.²¹ However, mutation in SF3B1 did not correlate with a negative prognosis in patients with MDS.^33,34 Analysis of SF3B1 mutation profiles in hematopoietic neoplasms and their biological and clinical importance have been recently reviewed.^37,40 SF3B1 was also found to be mutated in lung adenocarcinoma¹⁸ and uveal melanoma.¹⁹

U2AF1, the gene encoding the 3´ splice site binding protein U2AF35, was found to be mutated in 6% of patients with myeloid neoplasms,¹⁶ and in patients with lung adenocarcinoma.¹⁸

SRSF2, which encodes the SR protein SRSF2 (SC35), was found to be mutated in 28% of patients with chronic myeloid leukemia (CML).¹⁶ Mutations were also detected in patients with MDS, acute myeloid leukemia (AML), myeloproliferative neoplasms, and juvenile myelomonocytic leukemia.^16,36 Another study by Lasho and colleagues³⁴ found that 32 out of 187 (17%) patients with primary myelofibrosis (PMF) had a mutation in SRSF2. Most of these mutations were point mutations at position P95, although some patients had deletions (delP95-R102). Interestingly and importantly, survival of patients with mutations in SRSF2 was worse than that of patients with mutations in SF3B1.³⁴

Gene amplifications and copy number variation

Alterations in gene copy number of several splicing factors have also been detected in various cancers Table 1.

Table 1. Altered splicing factor levels in cancer tissues

Name	Cancer	Expression	Species	Method	Ref
HnRNP A1	HCC	High RNA levels	Human tissue	qPCR	^91
		High RNA levels	Mouse tissue	qPCR	^96
		High protein levels	Mouse tissue	WB	^96
	Lung	High protein levels	Human tissue	WB+ IH	^90
HnRNP A2	Breast	High protein levels	Human tissue	IH	^98
	HCC	High RNA levels	Mouse tissue	qPCR	^96
		High protein levels	Mouse tissue	WB	^96
	Brain	High RNA levels	Human tissue	qPCR	^13
	Lung	High protein levels	Human tissue	WB + IH	^90
		High protein levels	Human tissue	IH	^100
		High RNA levels	Human tissue	NB	^101
	Pancreas	High RNA levels	Human cell lines	PCR	^102
	GI	High protein levels	Human tissue	WB	^99
HnRNP H	Brain	High RNA levels	Human tissue	qPCR	^14
		High protein levels	Human tissue	IH	^14
	Head and neck	High protein levels	Human tissue	IH	^114
		High protein levels	Human tissue	IH	^193
	Colon	High RNA levels	Human tissue	PCR	^114
HnRNP I	Ovary	High protein levels	Human tissue	WB	^82
		High protein levels	Human tissue	IH	^118
	Brain	High protein levels	Human tissue	WB	^119
	Breast	High protein levels	Human Cell lines	WB	^120
HnRNP M	Colon	High protein levels	Human tissue	iTRAQ	^121
SRSF1	Breast	DNA amplification	Human tissue	qPCR	^12
		High protein levels	Human tissue	protein MA	^12
	Lung	High protein levels	Human tissue	WB + IH	^66
	Colon	High protein levels	Human tissue	protein MA	^12
	Thyroid	High protein levels	Human tissue	protein MA	^12
	Small intestine	High protein levels	Human tissue	protein MA	^12
	Kidney	High protein levels	Human tissue	protein MA	^12
SRSF2	Myeloma	DNA mutation	Human tissue	Exome seq	^16
SRSF3	Ovary	High protein levels	Human tissue	WB	^82
	Lung	DNA amplification	Human tissue	PCR	^15
	Cervix	High protein levels	Human tissue	WB + IH	^15
SRSF6	Lung	DNA amplification	Human tissue	qPCR	^11
	Colon	DNA amplification	Human tissue	qPCR	^11
SRSF9	Brain	High protein levels	Human tissue	IH	^86
	Colon	High protein levels	Human tissue	IH	^86
	Lung	High protein levels	Human tissue	IH	^86
	Skin	High protein levels	Human tissue	IH	^86

qPCR: quantitative real time PCR.

WB: western blot analysis.

IH: immunohistochemistry.

NB: northern blot analysis.

protein MA: protein microarray.

Exome seq: whole-exome sequencing.

iTRAQ: g isobaric tags for relative and absolute quantitation.

SRSF1, the gene encoding for SR protein SRSF1 (SF2/ASF), was found to be overexpressed in lung, colon, and breast tumors, and amplified in breast cancer.^12,41

SRSF3, the gene encoding the SR protein SRSF3 (also known as SRp20), has been found to be amplified in lung cancer and cervical cancer.¹⁵

SRSF6, encoding the SR protein SRSF6 (SRp55), has been found to be amplified in 12% of lung and breast tumors, and in 37% of colon tumors samples.¹¹ SRSF6 has also been shown to be amplified in skin cancer.⁴²

HNRNPA2B1, which encodes the hnRNP protein hnRNP A2/B1, was found to be amplified and overexpressed in glioblastoma tumors.¹³

RBFOX1 (A2BP1), the gene encoding the splicing factor RBFOX1, was found to be deleted in 10% of glioblastoma cases and to act as a tumor suppressor in this cancer.⁴³

The accumulation of data showing that genes encoding splicing factors are mutated, amplified, or deleted in cancer has strengthened the link between splicing factors and cancer initiation. However, only limited evidence exists for a direct role of these mutations or copy number variations as driver mutations in cancer. In addition, post-transcriptional modifications, such as phosphorylation on the splicing factor itself, can affect the function or localization of the protein,^44,45 suggesting another layer of regulation. This will not be discussed here but examples are provided elsewhere.^46-49

Mechanisms of Transformation by Splicing Factors

Two major families of alternative splicing factors have been studied in depth: SR proteins and hnRNPs. Although it is generally thought that the SR proteins promote alternative splicing whereas hnRNPs inhibit it, there is accumulating evidence that both SR proteins and hnRNPs work through a combinatorial effect of positive and negative regulation to control alternative splicing. Additional RNA binding proteins such as SAM68,^50,51 members of the RBM family,^52-54 and HuR,^55-57 will not be discussed here.

The oncogenic activities of SR proteins

SR proteins are a family of 12 proteins containing 1 or 2 RNA-recognition motifs (RRMs) and a C-terminal RS domain (arginine–serine repeats).⁵⁸ SR proteins are required for constitutive pre-mRNA splicing as well as alternative splicing, and also have other non-splicing functions in the cell including mRNA nuclear export, nonsense-mediated mRNA decay (NMD),⁵⁹ translation,^60,61 genomic stability,⁶² cell cycle progression,⁶³ and miRNA biogenesis.⁶⁴ In recent years several SR proteins were found to play a causative role in several types of cancer. In many cases, SR proteins regulate alternative splicing events leading to enhanced production of pro-oncogenic isoforms and reduced formation of tumor suppressive isoforms, although in other cases the oncogenic activity of SR proteins is independent of splicing regulation.

SRSF1

SRSF1 plays an important role in regulating many splicing events in the cell and is one of the most-studied splicing factors in the context of its biological and pathological effects (for a recent review see ⁶⁵). SRSF1 is overexpressed in different cancer types and acts as a proto-oncogene. Overexpression of SRSF1 in immortalized rodent fibroblasts and human mammary epithelial cells results in oncogenic transformation as evidenced by proliferation, resistance to apoptosis, and ability to form tumors in mice.^12,41 Furthermore, SRSF1 overexpression in lung adenocarcinoma cells results in a more aggressive phenotype and confers resistance to anticancer drugs such as carboplatin and paclitaxel.⁶⁶

SRSF1 is known to have a variety of functions in addition to its role as an alternative splicing factor and its proto-oncogenic potential is most likely due to a combinatorial effect of these functions. Several mechanisms have been identified that contribute to its pro-oncogenic activity. SRSF1 has been shown to activate the mTORC1 signaling pathway by bypassing upstream components such as AKT.^67,68 In addition, blockade of mTORC1 can abolish the transformation effect of SRSF1 in mouse immortal fibroblasts. SRSF1 has also been shown to activate the Ras-MAPK pathway by increasing the expression of B-Raf, and the RRM1 domain of SRSF1 is necessary for this activation.⁶⁹

Some of the splicing functions of SRSF1 have been shown to contribute directly to its pro-oncogenic activity. SRSF1 promotes skipping of exon 4 in the CCND1 transcript to generate the oncogenic cyclin D1b isoform in prostate cancer.⁷⁰ SRSF1 switches the splicing pattern of BIN1 in breast cancer to generate the BIN1+12a isoform that no longer binds MYC.⁴¹ The BIN1 protein interacts with the product of the MYC proto-oncogene, suppressing its oncogenic activity. The inclusion of exon 12a thus abolishes the tumor suppressor activity of BIN1 by interfering with its MYC binding. SRSF1 also regulates alternative splicing of the proapoptotic gene BIM. In breast cancer, SRSF1 generates 2 new splicing isoforms of BIM: BIM γ1 and γ2. These isoforms lack exons 2 and 3, which encode the BH3 domain of BIM, and as a result act as antiapoptotic proteins.⁴¹ SRSF1 also controls the splicing of caspase-9 in lung cancer.⁷¹ Two splice variants, proapoptotic caspase-9a and antiapoptotic caspase-9b, are derived from the CASP9 gene; SRSF1 promotes the generation of caspase-9b. SRSF1 enhances the inclusion of exon 13b of the gene encoding Mnk2.¹² Recently, it was shown that SRSF1 altered the ratio of the Mnk2 isoforms in breast cancer cells, reducing production of the Mnk2a isoform and enhancing Mnk2b. The Mnk2a isoform acts as a tumor suppressor by activating the p38-MAPK stress pathway, whereas the Mnk2b isoform cannot activate the p38-MAPK pathway but activates eIF4E phosphorylation and is pro-oncogenic.⁷² An additional study has found that favored production of the Mnk2b isoform through the action of SRSF1 in pancreatic ductal adenocarcinoma results in resistance to the drug gemcitabine,⁷³ further supporting the contribution of SRSF1 to the cancerous phenotype. Another splicing target of SRSF1 is the RPS6KB1 gene encoding the ribosomal protein S6K1. SRSF1 promotes expression of the short isoform (isoform-2) of S6K1. Whereas S6K1 isoform-1 acts as a tumor suppressor by blocking Ras-induced transformation, the short isoform-2 possesses oncogenic properties by activating mTORC1.⁷⁴ SRSF1 also regulates alternative splicing of the tyrosine kinase receptor MST1R (also known as RON) and enhances generation of the ΔRON isoform, which is constitutively active as a result of skipping of exon 11. This isoform was documented to enhance motility and invasion in several cell lines.⁷⁵

SRSF2

Although SRSF2 (known also as SC35) was found to be mutated in many hematopoietic cancer types, not much is known about its role as a tumor promoter or in tumor progression. Nevertheless, there is some experimental evidence supporting its role in cancer. SRSF2 was found to be overexpressed in a panel of neuroendocrine lung tumors. In these cases, SRSF2 contributed to the cancerous phenotype by causing cells to enter S phase. However, this effect is not achieved through splicing, but rather by cooperation with the transcription factor E2F1. SRSF2 is required for E2F1-mediated transcription of S-phase genes such as cyclin E and p45^SKP2.⁷⁶

A direct role for the splicing function of SRSF2 in cancer has also been demonstrated. SRSF2 was found to interfere with alternative splicing of the KLF6 gene, a tumor suppressor. Expression of SRSF2 results in increased generation of the isoform containing exon1a. This exon has an early termination sequence that leads to the production of a protein that lacks the DNA binding domain and thus, unlike wild type KLF6, cannot act as a tumor suppressor.⁷⁷

SRSF2 has also been shown to have a tumor suppressor role. SRSF2 was found to cooperate with E2F1 to alter VEGF-A splicing. VEGF-A has several splice variants that are proangiogenic and are upregulated in human tumors. However, alternative splicing of exon 8 of VEGFA-A leads to different isoforms of the same length, but with 6 distinct C-terminal amino acids. These isoforms play an antiangiogenic role and are downregulated in some tumors. SRSF2 promotes a shift in the ratio of VEGF isoforms in favor of the antiangiogenic isoforms.⁷⁸ In addition, SRSF2 cooperates with E2F1 to control the splicing of c-flip, caspases-8 and -9, and Bcl-x, favoring the proapoptotic isoforms.⁷⁹

SRSF2 also plays a positive role in the response to cisplatin treatment. In this case, cisplatin stabilizes SRSF2 levels, leading to an increase in the caspase 8L isoform that promotes cell death via apoptosis.⁸⁰ More studies are needed to better understand the pro- and antioncogenic activities of SRSF2.

SRSF3 (SRp20)

The smallest member of the SR family, this protein contributes to cellular proliferation and apoptosis through its RNA processing functions. The role of SRSF3 in disease was reviewed recently;⁸¹ in this review we will focus on its involvement in cancer.

SRSF3 was found to be overexpressed in ovarian cancer.⁸² Knockdown of SRSF3 in ovarian cell lines resulted in inhibition of cell growth and induction of apoptosis through downregulation of Bcl-2.⁸³ SRSF3 overexpression was also detected in other cancer types including lung, breast, stomach, skin, bladder, colon, liver, thyroid, and kidney.¹⁵ In vitro experimental systems indicate that SRSF3 is required in order for cells to enter the G2/M phase. SRSF3 expression is also required for tumor maintenance as knockdown of SRSF3 results in a decrease in tumor size in vivo.¹⁵ SRSF3 was found to be translocated with BCL6 in follicular lymphoma.⁸⁴

A direct role for the splicing function of SRSF3 in cancer is seen in the response of cells to treatment with drugs such as doxorubicin. In ovarian cancer, SRSF3 expression correlates with splicing of MRP1, a member of the ATP binding cassette transporter family that is associated with multidrug resistance.⁸⁵ Some of these spliced isoforms can decrease the effect of treatment with drugs such as doxorubicin, resulting in increased drug resistance.⁸²

On the other hand, a hepatic-specific SRSF3 knockout mouse model developed tumors with aging. This phenomenon was correlated with aberrant splicing patterns of Igf2 and Insr, which led to a more mitogenic profile. In addition, decreased levels of SRSF3 were observed in patients with hepatocellular carcinoma (HCC), strengthening the role of SRSF3 as a tumor suppressor in liver cancer. This intriguing finding suggests distinct roles for the same factor in different tissues, adding more complexity to the already complex field of splicing factors and cancer.

SRSF6 (SRp55)

SRSF6 was found to be amplified in breast, lung, and colon cancer. Overexpression of SRSF6 was able to transform non-cancerous cell lines and cause tumor formation in mice.¹¹ Transgenic mice overexpressing SRSF6 show hyperplasia of the sensitized skin due to dysregulation of tissue homeostasis. This mechanism involves splicing of the extracellular matrix protein tenascin-C, resulting in increased expression of isoforms that are known to play a role in several cancers.⁴²

SRSF9

SRSF9 expression has been found to be elevated in multiple types of cancer such as glioblastoma, colon adenocarcinoma, squamous cell lung carcinoma, and malignant melanoma.⁸⁶ SRSF9 overexpression was able to transform NIH 3T3 cells to form colonies in soft agar in vitro and to form tumors in nude mice, whereas downregulation of SRSF9 in colon cancer cell lines reduced their colony formation ability.⁸⁶ The mechanism by which SRSF9 accomplishes this is not known.

The oncogenic activities of hnRNP splicing factors

The heterogeneous nuclear ribonucleoproteins (hnRNPs) are a large family of proteins containing more than 20 members with common structural domains.⁸⁷ hnRNPs have roles in various cellular processes such as RNA metabolism, DNA repair, telomere biogenesis, cell signaling, and regulation of gene expression at transcriptional, RNA processing, and translational levels. Emerging evidence suggests their involvement in tumor development and progression. More specifically, hnRNPs have been shown to function in proliferation, apoptosis, angiogenesis, and cell invasion.⁸⁸ In this review we will elaborate on recent studies implicating hnRNP proteins in cancer.

hnRNP A1

hnRNP A1 has been found to be deregulated in various types of cancers, including colon,⁸⁹ lung,⁹⁰ and liver,⁹¹ usually leading to overexpression of hnRNP A1 mRNA and protein. Expression levels are correlated with increased proliferation^92,93 and tumor metabolism.^94,95 hnRNP A1 was found to be overexpressed in HCC.^91,96 Knockdown of hnRNP A1 in metastatic HCC cells caused a decrease in cell invasion, whereas upregulation of hnRNP A1 in poorly metastatic HCC cells led to a significant increase in their invasive ability. This effect correlated with CD44v6 expression.⁹¹ Another study found that hnRNP A1 functions as a tumor promoter in model systems of overexpression in non-tumorigenic liver progenitor cell lines, in part due to increased proliferation.⁹⁶

hnRNP A1 is required for RON alternative splicing; expression of hnRNP A1 decreases the formation of the ΔRON isoform, which is known to drive epithelial-to-mesenchymal transition (EMT). As a result the cells undergo mesenchymal-to-epithelial transition (MET), which leads to the establishment of secondary tumors.⁹⁷ The fact that hnRNP A1 and SRSF1 act in opposite manners on RON alternative splicing supports the view that tight regulation of splicing factors is necessary for metastasis of cancer cells.

hnRNP A2/B1

Like hnRNP A1, hnRNP A2/B1 is also deregulated in various types of cancers. hnRNP A2 was found to be overexpressed in breast,⁹⁸ brain,¹³ liver,⁹⁹ lung,^100,101 pancreas,¹⁰² and GI⁹⁹ cancers. Some controversy exists with regard to the role of hnRNP B1, an isoform of hnRNP A2, in cancer. These isoforms are identical except for 12 amino acids in the N-terminal region.¹⁰³ hnRNP B1 has also been reported to be overexpressed in some cancers, such as esophagus¹⁰⁴ and lung.^105-108 However, in contrast to hnRNP A2, forced overexpression of hnRNP B1 in normal mouse liver cells was unable to transform these cells.⁹⁶

Forced overexpression of hnRNP A2 was able to transform both normal mouse fibroblast¹³ and normal mouse liver progenitor cells,⁹⁶ whereas knockdown of hnRNP A2 reduced the cancerous phenotype of brain¹³ or liver⁹⁶ cancer cell lines injected into mice. In addition, knockdown of hnRNP A2 induced apoptosis in breast cancer cell lines¹⁰⁹ or reduced proliferation in Colo16 and HaCaT cells.¹¹⁰

hnRNP A2 can activate the RAS-MAPK pathway in liver cancer cell lines. This is achieved by favoring formation of the wild type A-Raf isoform and decreasing formation of the dominant negative short A-Raf isoform that lacks the kinase domain.⁹⁶ In brain cancer, hnRNP A2 is required for development and maintenance of tumors and regulates splicing of the RON tyrosine kinase receptor. hnRNP A2 promotes the formation of the ΔRON isoform, which acts as a constitutively active receptor.¹³

The hnRNP alternative splicing factors also play a role in the regulation of cancer metabolism. hnRNP A2, hnRNP A1, and polypyrimidine tract-binding protein (PTB, also known as hnRNP I) control splicing of PKM enzyme mRNA,^94,111 resulting in 2 mutually exclusive isoforms. The PKM1 isoform, which is generated by the inclusion of exon 9, promotes oxidative phosphorylation. Inclusion of exon 10 generates the PKM2 isoform that promotes aerobic glycolysis and the “Warburg effect." The PKM2 isoform is crucial for rapidly growing cells and is highly expressed during embryogenesis. However, cancer cells were also found to express this isoform.¹¹²

Other members of the hnRNP protein family have also been studied in the context of cancer but not much is known about their function.

hnRNP C

hnRNP C has been shown to increase translation of c-myc mRNA via use of an internal ribosome entry site (IRES).¹¹³

hnRNP H

This splicing factor was found to be overexpressed in glioblastoma.¹⁴ Moreover, hnRNP H was shown to act as an oncogene in this context by controlling the splicing of 2 important genes; hnRNP H controls the splicing of RON to form the ΔRON isoform which is constitutively active, and the splicing of IG20/MADD,¹⁴ an adaptor protein that is involved in apoptosis through caspase-8 activity. hnRNP H promotes formation of the MADD product, which mediates cell survival, over the IG20 product, which triggers apoptosis.¹⁴ hnRNP H is also overexpressed in head and neck carcinomas. One possible role for hnRNP H in cancer is to block apoptosis, as knockdown of hnRNP H increases cell death via caspase-3 activity.¹¹⁴

One mechanism of transformation by hnRNP H is by regulating the alternative splicing of ARAF, reducing the formation of a short isoform of ARAF that lacks the Ras binding domain and acts as an inhibitor of the Ras-MAPK pathway.¹¹⁵

hnRNP I (PTB1)

This family member was shown to bind the pyrimidine-rich region in introns. PTB1 and its variant nPTB are important regulators of neuronal development and their ratio changes during development.^116,117

hnRNP I was found to be overexpressed in ovarian,^82,118 brain¹¹⁹ and breast cancers.¹²⁰ In brain cancer it was found to promote the skipping of an exon in FGF1, leading to the production of a high-affinity receptor.

In ovarian cancer, changes in splicing of MRP1, a member of the ATP binding cassette transporter family that is associated with multidrug resistance, correlated with expression of hnRNP I and SRSF3, suggesting a role for hnRNP I in cancer drug resistance.⁸²

hnRNP M

The hnRNP M gene has been found to be overexpressed in colon carcinoma.¹²¹ In addition, hnRNP M was found to promote EMT in breast cancer through alternative splicing of CD44; hnRNP M binds to GU-rich elements in the transcript to favor skipping of exon 8. It has been shown that knockdown of hnRNP M can block EMT in vitro and in vivo. Indeed, hnRNP M levels correlated with a more aggressive breast cancer phenotype and its overexpression promoted metastasis of breast cancer cells into the lungs in an in vivo mouse model.¹²² Importantly, hnRNP M can compete with another splicing regulator, ESRP1 (epithelial splicing regulatory protein 1), which also binds to GU-rich elements and promotes exon inclusion¹²³ (see below).

hnRNP K

This hnRNP family member was found to be overexpressed in lung¹²⁴ and liver¹²⁵ cancers. Moreover, elevation of hnRNP K protein levels in head and neck or oral squamous cell carcinomas can be used as a biomarker for poor prognosis.¹²⁶

Role of Splicing Factors in EMT and Metastasis

Epithelial-to-mesenchymal transition (EMT) is a key step in tumor progression and metastasis.¹²⁷ Several studies have shown that this process is regulated at the transcriptional level and several transcription factors that control and execute the EMT program have been identified.^128,129 However, it is now becoming clear that regulation at the post-transcriptional level also occurs in EMT, specifically at the level of alternative splicing.¹³⁰ Several alternative splicing events have been characterized and the roles of the different splice variants in the process of EMT have been studied.^123,131-133 RNA deep sequencing methods were used to determine the EMT-driven alternative splicing program in breast cancer.¹³⁴ In addition to changes in the splicing pattern, the trans-acting proteins responsible for this pattern were identified. One of the best-studied splicing factors in this context is RBFox2. Expression of RBFox2 was shown to be higher during EMT and this factor is responsible for executing splicing of Cttn, Pard3, and DNM2, all of which are known to play a role in EMT.¹³⁵ Deletion of RBfox2, however, did not prevent EMT, suggesting redundancy in the regulation of the alternative splicing signature. As mentioned above, SRSF1 regulates splicing of the proto-oncogene tyrosine kinase receptor RON, which is involved in several physiological processes such as migration and invasion. Elevated expression of the constitutively active ΔRON splice variant results in enhanced EMT.⁷⁵ In recent years, additional splicing factors have also been found to regulate the splicing of RON. hnRNP H and hnRNP A2 were found to increase the production of the ΔRon variant in glioblastoma, resulting in a more aggressive phenotype.^13,14 In contrast, hnRNP A1 was found to compete with SRSF1 for binding to the same site on RON and prevent skipping of exon 11. In this way, hnRNP A1 contributes to the reverse process of MET.⁹⁷ MET is thought to be involved in later stages of metastasis, when cells begin to form secondary foci in metastatic organs. In some cases a splicing switch in EMT is controlled by the ESRP proteins, important mediators of the epithelial tissue identity that regulate many epithelial-specific alternative splicing events. The splicing activity of these factors is inhibited during EMT, and they might be considered tumor suppressors.^123,136,137 ESRP1 and ESRP2 control the splicing of CD44, FGFR2, CTNND1, and ENAH.^138,123

Role of Alternative Splicing in Drug Resistance

One of the greatest problems in cancer treatment is recurrence of the disease and resistance of the tumors to anticancer drugs. In some cases this is due to acquired resistance to chemotherapeutic drugs that initially worked well. Drug resistance has been ascribed to several mechanisms, alternative splicing being one of them. A classic example is prolonged exposure of cancer cells to the BRAF inhibitor vemurafenib, which results in the loss of sensitivity to the drug.¹³⁹ In approximately 30% of vemurafenib-resistant melanoma tumors the identified mechanism is alternative splicing of BRAF;^140,141 skipping of exons 3–8 in the resistant tumors results in increased production of a BRAF isoform that lacks the Ras binding domain and can enhance constitutive dimerization, resulting in activation of the MAPK pathway in a Ras-independent manner. The trans-acting factor responsible for this alternative splicing has yet to be identified and how prolonged exposure to the drug results in this alternative splicing remains unknown.¹⁴² Another striking example of the role of alternative splicing in drug resistance is alternative splicing of the androgen receptor (AR) gene in prostate cancer.¹⁴³ As this cancer is androgen/AR-dependent a first line of treatment is androgen depletion therapy, and indeed a response can be achieved by this method. However, resistance to this treatment may occur and one mechanism by which this is achieved involved alternative splicing of the AR gene that generates a product lacking the ligand binding domain.¹⁴⁴

Modulating Alternative Splicing as a Novel Therapeutic Approach

Given that aberrant splicing is one of the characteristic features of cancer it is obvious that the development of therapeutic approaches based on splicing modulation would be the focus of intense research.¹⁴⁵ The aim of these approaches is to either alter or inhibit the splicing machinery to convert it to the non-cancerous state, or to degrade the cancer-specific splicing isoforms that contribute to tumor development, maintenance, or progression. Several approaches will be discussed and are outlined in Figure 1.

Figure 1. Splicing-based therapeutic approaches. The diagram shows a cassette exon splicing event. Three exons (rectangles) and 2 introns (black lines) are shown. The black box in the middle exon represents an exonic splicing enhancer (ESE) sequence that is recognized by the enhancer splicing factor (black SF). The orange box in the middle exon represents an exonic splicing silencer (ESS) sequence that is recognized by the splicing factor silencer (orange SF). SF3B binds to the branch point sequence (BPS) in the intron. (A) No treatment. In this scenario 2 mRNAs can be generated: one in which the middle exon is included in the mRNA (mRNA A) and a second in which the middle exon is skipped (mRNA B). The width of the black arrows indicates that formation of mRNA A is favored. (B) SF3B inhibitor (SSA) prevents binding of SF3B to the BPS, resulting in inhibition of splicing. In this case the intronic sequences are retained in the mRNA and may contain a premature stop codon. This mRNA will be recognized by the NMD machinery and degraded. Alternatively, the premature stop codon will not lead to degradation of the mRNA, but rather to the production of a truncated protein. Another outcome is that the transcript will be retained in the nucleus. SF3B inhibitors act as general inhibitors of splicing and do not target a specific mRNA. (C) Inhibition of specific splicing factors. Small molecules that block activity (e.g., phosphorylation) of a specific SF (in this example a SF that binds to the ESE) will prevent its binding to the cis elements in the pre-mRNA and result in reduced inclusion of the exon. Other silencer SFs will still bind to the pre-mRNA to favor skipping of this exon. As a result, more mRNA B will be generated. (D) ASOs/siRNAs directed against the SF mRNA will lead to degradation of the SF mRNA, and in this case to decreased binding to the ESE resulting in decreased inclusion of the middle exon. Other SFs can still bind to the pre-mRNA and favor skipping and production of mRNA B. (E) Specific ASOs designed to prevent binding of the SF to the cis elements in the mRNA (ESE) will lead to decreased inclusion of the middle exon. Other SFs can still bind to the pre-mRNA and favor production of mRNA B. (F) Specific ASOs designed to prevent binding of the splicing machinery result in skipping of the exon and production of more mRNA B.

General inhibition of splicing by small molecules

Recent studies have attempted to directly target the spliceosome machinery as a global anticancer treatment.¹⁴⁶ Treatment of cells with amiloride, a diuretic agent that affects sodium transport and fluid homeostasis, causes alterations in alternative splicing. Amiloride changed the alternative splicing pattern of several oncogenic genes such as BCL-X, HIPK3, and RON/MISTR1,¹⁴⁷ most likely due to hypophosphorylation of the splicing factor SRSF1. Another small molecule used to interfere with splicing is spliceostatin A (SSA). SSA is a derivative of FR901464, a fermentation product of Pseudomonas bacteria. It has been shown that treatment with SSA prolongs the life of tumor-bearing mice.¹⁴⁸ SSA binds to splicing factor SF3b ¹⁴⁹ and thus inhibits splicing in vitro and in cultured cells.^149-151 Another closely related molecule, sudemycin E, also has been shown to alter splicing.¹⁵² Cells treated with sudemycin E were arrested in G2 and showed increased cell death. It has been suggested that this is due to changes in alternative splicing as well as changes in chromatin condensation.

Inhibition of splicing factor activity by small molecules

Although general inhibition of splicing by small molecules may be effective in reversing aberrant splicing in cancer, non-specific inhibition can lead to cellular toxicity. Therefore, there is a need for small molecules that inhibit only specific alternative splicing events. SR proteins are known to undergo both phosphorylation and dephosphorylation events that control the precise stage of splicing assembly¹⁵³ and the splicing reaction.¹⁵⁴ Several kinases have been identified that phosphorylate SR proteins, including SRPK,^155,156 HhPRP4,¹⁵⁷ topoisomerase I,¹⁵⁸ Clk,¹⁵⁹ and DIRK1A (for a more detailed review see ⁴⁴). Specific inhibition was achieved by inhibition of the kinase activity of topoisomerase I; as a result, phosphorylation of SRSF1 and other SR proteins is reduced and spliceosome assembly does not occur.¹⁶⁰ In addition, small molecules have been identified that can inhibit the phosphorylation of specific SR proteins.¹⁶¹

As in the case of topoisomerase I inhibition, reduction of phosphorylation of SR proteins by Clk inhibition has been shown to interfere with the alternative splicing pattern of several target genes.¹⁶² Inhibitors of other SR kinases, the SRPK proteins, were shown to reduce phosphorylation of SRSF4 and as a result interfere with viral replication.¹⁶³ This more specific approach may open up new windows for splicing-based therapeutics. However, more studies need to be performed to ascertain whether these small molecules can be effective, but not toxic.

Targeted modulation of splicing by antisense oligonucleotides

In contrast to global alterations in splicing, a more subtle approach has been established to target only specific splicing events. One method in development is to use complementary sequences to the splice site to redirect the splicing machinery from the specific splice site.¹⁶⁴ This approach is based on the fact that trans-acting factors recognize and bind to specific cis elements on the mRNA to execute splicing. Antisense oligonucleotides (ASO) are synthetic nucleic acids (15–25 bases) that can bind cis elements on the mRNA through base pairing.^145,165 ASOs can be designed to either block factor binding to a cis element, preventing splicing from occurring at that specific site and resulting in enhanced inclusion of a specific exon, or to target splicing enhancers or silencers to either block or promote splicing. This approach has been used with some success, mainly in the treatment of neurodegenerative diseases such as Duchenne muscular dystrophy¹⁶⁶ and spinal muscular atrophy.^167,168

Targeting the cis elements by ASOs has major advantages over targeting the splicing factors directly. When protein expression of a specific factor is inhibited all of its biological functions are affected, including splicing, transport, or translation. Introduction of ASOs into the cells only affects the targeted mRNA and other cell functions remain unaffected. However, there are some disadvantages to this approach. One disadvantage is that the ASOs have to be absorbed by the target tissue/cells at a very high efficiency. This difficulty has been overcome in the treatment of neurodegenerative disease with ASOs.¹⁶⁹ Cancer treatment based on splicing modulation has also been attempted. To date, several targets have been studied, such as RON,¹⁷⁰ PKM,¹⁷¹ Mdm,¹⁷² mcl-1,¹⁷³ FGFR1,¹⁷⁴ and BCL-X.¹⁷⁵ For a further review on ASO-based treatment please see.¹⁴⁵

Summary and Future Scope

Splicing is a complex process requiring a combinatorial network of cis-acting elements and trans-acting factors. Elucidating the roles of the trans-acting splicing factors in cancer development and progression is challenging; these factors are involved in almost every layer of cellular function, they are ubiquitously expressed in all tissues with differential roles in different tissues,^87,176,177 and they regulate many steps of RNA processing including splicing, mRNA transport, mRNA stability, nonsense-mediated decay, and miRNA biogenesis.^{60,62-64,178,179} In the past few years, it has become clear that splicing factors are major drivers of cancer initiation and progression. The involvement of splicing factors in multiple steps of RNA processing and the redundancy of activities among some of these factors makes it difficult to decipher the precise constellation of factors necessary for a specific splicing decision and for cancer development. Another hurdle in understanding the role of splicing factors in normal physiological conditions is the lack of good animal model systems. Mice that lack the alternative splicing factors SRSF1, SRSF2, or SRSF3 are non-viable^180-182 and therefore there is a need for more sophisticated mouse model systems such as inducible and/or tissue specific knockouts.¹⁸³ There is also a need for transgenic mice model systems that overexpress splicing factors. One recent example is the inducible transgenic SRSF6 mouse model, in which overexpression of this oncogenic SR protein was shown to induce skin hyperplasia.⁴² In vitro modulation, such as overexpression or knockdown, is more common and has been used extensively, but has limitations. For example, the expression of some splicing factors is tightly regulated, with some factors auto-regulating their own expression. New techniques, such as cross-linking immunoprecipitation (CLIP) and modifications of CLIP, have been established to identify mRNA targets of splicing factors and characterize their cis-acting sequences.^184-192 Results of these studies will surely be applied to cancer research. The recent identification of recurrent mutations in spliceosomal components^16-22 reinforces the recognition of splicing factors as important drivers of cancer development and progression and as promising targets for the development of anticancer drugs. In this regard, newly identified alternative splicing events that contribute to cancer initiation and/or progression present promising targets for splicing modulation using modified antisense RNA oligos as described above. Resources such as The Cancer Genome Atlas (TCGA), which contains RNA-seq data from hundreds of tumors and corresponding normal tissues, will contribute to the identification of new alternative splicing events that drive tumor formation and maintenance. Modulation of these splicing events by antisense oligonucleotides or small molecules presents a new approach for cancer therapy (Fig. 1). Moreover, over the past few years increasing lines of evidence have suggested that splicing factors can act as driving oncogenes in several types of cancer, and their inhibition might be a new avenue in cancer therapy (Fig. 1). Although less well established, certain splicing factors can probably act as tumor suppressors through mechanisms that are yet to be elucidated. The field of alternative splicing is entering an exciting new era in which its implication in human diseases, and specifically in cancer research and treatment, is becoming increasingly important.

Disclosure of Potential Conflicts of Interest

No potential conflicts of interest were disclosed.

Acknowledgments

The authors wish to thank the Karni lab for helpful discussions.

Funding

This work was supported by the Israeli Science Foundation (ISF Grant no. 1290/12 to R.K.), Israel-US Bi-national Science Foundation (BSF Grant no. 2009026) and the Len & Susan Mark Initiative for Ovarian and Uterine/MMMT Cancers from the Israel Cancer Research Fund (ICRF).