Epigenetic regulation of alternative splicing

Abstract

Alternative splicing (AS) serves as an additional regulatory process for gene expression after transcription, and it generates distinct mRNA species, and even noncoding RNAs (ncRNAs), from one primary transcript. Generally, AS can be coupled with transcription and subjected to epigenetic regulation, such as DNA methylation and histone modifications. In addition, ncRNAs, especially long noncoding RNAs (lncRNAs), can be generated from AS and function as splicing factors (“interactors” or “hijackers”) in AS. Recently, RNA modifications, such as the RNA N6-methyladenosine (m6A) modification, have been found to regulate AS. In this review, we summarize recent achievements related to the epigenetic regulation of AS.

Introduction

It was said by the director Eisenstein Sergei that the sum of frame A and frame B will produce C, which contains an absolutely new concept in the film field. This concept is “montage”, which is also known as “splicing”. Similarly, in molecular biology, splicing means the editing of precursor messenger RNA (premRNA) transcripts to yield mature messenger RNAs (mRNAs) or noncoding RNAs (ncRNAs). Splicing removes introns and joins exons together through a series of complicated reactions catalyzed by spliceosomes. Generally, RNA splicing includes constitutive splicing (CS) and alternative splicing (AS). AS refers to the process that selectively splices a set of sites in a premRNA to form variable mature mRNAs, thereby producing proteins with different structures and functions [1,2]. This process occurs for ~95% of transcripts and creates diverse proteins by varying exon composition from a single mRNA [2]. AS effectively increases transcriptomic and proteomic diversity and significantly influences many kinds of cellular processes, as well as tissue specificity or development and disease development [3]. In this review, we will summarize recent findings and provide an overview of approximately 40 years of insights related to AS, focusing mainly on the mechanisms of epigenetic regulation of AS.

History of alternative splicing

AS is important for generating proteome complexity from a limited number of genes. The primary transcripts produced from gene transcription undergo AS, with introns being removed and particular exons included or excluded, resulting in the maturation of mRNAs or ncRNAs (Figure 1A). Mainly based on this process, the concept of “one gene-many polypeptides” has been proposed. The generation of ncRNAs from AS was recently reported, further confirming AS as an efficient and economical mode of regulating gene expression.

Figure 1.

Process of alternative splicing and traditional classification of basic types. A. The brief process of AS that results in the production of multiple proteins from a single gene transcript. In detail, after DNA replication begins, particular exons of primary transcribed RNAs may be alternatively included or excluded under some conditions, generating variant isoforms. As a result, the translated proteins contain differences in amino acid sequence or functional structures. B. The five basic patterns of AS, including 1, exon skipping or cassette exon, in which single or multiple exons are skipped; 2, mutually exclusive exons, in which only one of two exons is retained; 3 & 4, alternative donor/acceptor site or 5’/3’ splice junction is used to alter the boundary of exons; and 5, intron retention indicates that a region may be spliced as an intron or simply retained in the mature RNA. Constitutive exons are represented as dark wide blocks, and alternatively spliced exons are represented as light narrow block.

Generally, AS has five discriminative modes that result from a complicated regulation network, including exon skipping (SE, cassette exon), mutually exclusive exons (MXEs), alternative donor sites (A5’SS), alternative acceptor sites (A3’SS), and intron retention (RI) (Figure 1B). SE is the most prevalent mode, with exons included in mRNAs under certain conditions but omitted from mRNAs under different conditions, similar to the differential frame choice between “TRAILER” and “FEATURE” in films. For example, single exon skipping occurs in CAPN3 (E6) [4], Fas (E6) [5] and PXN-AS1 (E4) [6], and multiple exon skipping occurs in exons 8-12 of TGM2 during leukemia tumorigenesis [7]. MXE indicates a process in which two exons are competitively retained in mRNAs, which occurs in some well-known genes, such as PKM1/2 (E9/10) [8], TPM1 (E2a/b) [9] and FGFR2 (EIIIb/c) [10]. A5’SS involves an alternative 5’ splicing junction due to the uncertain splicing of the 3’ boundary of one upstream exon. Both BCL [11,12] and XIST [13], which remarkably regulate cell growth, have two isoforms, including short (BCL-XS and XIST-S) and long (BCL-X and XIST-L) patterns, which are mediated by the A5’SS splicing mode. On the other hand, A3’SS generates an alternative 3’ splice junction by selectively changing the 5’ boundary of one downstream exon; for example, the PNUTS (also known as PPP1R10) preRNA can develop into an mRNA or a lncRNA through A3’SS in Exon 12 [14]. In the uncommon mode of RI, introns may be completely or partially retained in spliced RNAs, such as ARGLU1, which was found to have a retained intron between Exon 2 and Exon 3 of its preRNA [15]. Notably, multiple splicing modes may exist in one preRNA; for example, SE, MXE and A3’/5’SS together contribute to the various isoforms of RUNX1 [16,17].

In the past forty decades, a number of breakthroughs have been made (Figure 2), and these advances will continue to shape the future [3,18-24]. Splicing was primarily discovered in 1977, followed by basic chemistry research to verify the existence of splicing genes [25,26]. Small nuclear ribonucleoproteins (snRNPs) are RNA-protein complexes that combine with premRNA to form spliceosomes, which are key to the AS process. snRNPs were recognized in 1985 and included at least 5 snRNAs and more than 200 proteins. Then, a decade of studies on RNA binding proteins (RBPs), including the SR and hnRNP proteins, arrived. Nuclear speckles and the interchromatin granule distribution pattern of snRNPs were reported in the early 1990s as the foundation for understanding AS and exploring its precise mechanisms in vitro or in vivo. One landmark event was the 1993 Nobel Prize, which was awarded for the discovery of split genes, i.e., the splicing process. Increasingly, due to GFP technology, the dynamic nature of splicing factors and nuclear speckle organization was visualized in living cells, clarifying transcriptional and splicing activity. LncRNAs such as H19, XIST, SRA1, Evf2 and MALAT1 turned out to be closely linked to AS in 2004 [27], with additional research advances emerging. To better understand AS, the systematic structural elucidation of individual spliceosome components and subcomplexes was reported in 2007 [28]. Research on the N6-methyladenosine (m6A) modification emerged in large numbers in the early 2010s and was considered as potential signal for RNA splicing [29], bringing about a new initiative to study AS in more detail. Deciphering the mechanisms of AS is important for understanding the regulatory network in AS and developing applications for therapeutic approaches in the future.

Figure 2.

The history of AS over the past forty decades.

AS coupled to transcription

AS is a dynamic process that occurs simultaneously with and is inseparable from transcription. AS can be affected by the particular sequence or structure of the promoter [30-32]. For example, the C-terminal major domain (CTD) of RNA polymerase II plays a central role in coordinating transcription and AS, and it could assist SR proteins in binding primary transcripts [33,34].

Emerging research has placed AS in a key position in genetic information flow, which indicates that transcription regulators could also subsequently influence AS decisions. For example, TCERG1 is a transcription factor that is associated with the RNA pol II holoenzyme and influences the elongation efficiency of transcribed RNAs, which was found to be a crucial component of spliceosome regulation in AS [35]. Furthermore, Spi-1 tends to interact with spliceosome proteins and modulates AS of RNAs through splice site selection [36]. Moreover, chromatin conformation [32,37,38], the H3K36me3, H3K4me3 and H3K9me3 histone modifications [39,40] and DNA methylation [41-43] go hand in hand with the regulation of Brm [44], MRG15, Pisp1/p52 [45], HP1γ [46] and CTCF/MeCP2 [47], respectively. In detail, the speed bump model shows that nucleosomes can influence transcriptional elongation through hindering RNAPII and promoting exon inclusion, providing evidence that the chromatin structure can reflect underlying recognition of introns and exons. SWI/SNF, a subfamily of chromatin remodelers, not only modulates gene transcription by altering nucleosome organization but also controls AS through interacting with spliceosome components to determine the splicing events of genes such as CD44 [44,48,49].

DNA methylation and histone modifications determine eukaryotic chromatin structures under conditions in which certain histone marks recruit splicing factors through chromatin-binding proteins. For example, H3K36me3 can recruit MRG15 or Psip1 to promote hnRNP or SR splicing factors binding to specific sites of RNAs, respectively, directing the AS process [39,45,50]. DNA methylation frequently occurs in 5’SS and mediates AS through methyl-binding domain proteins (MBDs), such as HP1, which could interact with the histone chaperone complex FACT to bind H3K9me3 containing methylated DNA [51-53].

Regulation of AS by lncRNAs

The spliceosome comprises at least five kinds of snRNAs (U1, U2, U4/U6 and U5) and more than 200 snRNPs as the AS machinery [54]. The functional spliceosome complex is formed step by step and orchestrates with GU and AG dinucleotides. Splicing factors play a significant role in regulating the interaction between AS and exon splicing enhancer (ESE)/silencer (ESS) or intron splicing enhancer (ISE)/silencer (ISS) to promote binding to the SR or hnRNP protein, respectively. Consequently, alternative splicing occurs step by step with the assistance of SR-ESE or inhibition by hnRNP-ESS [1,55] (Figure 3). To obtain a comprehensive understanding of the AS regulation network, we propose some causes to explain various physiological and pathological changes. Elucidating the mechanism of AS is important for understanding gene expression and the pathogenesis of disease.

Figure 3.

A comprehensive overview of the AS machinery and the epigenetic role of lncRNAs. At least five kinds of snRNAs (U1, U2, U4/U6 and U5) and more than 200 snRNPs are included in spliceosomes. The functional spliceosome complex is formed step by step from the A complex to the P complex and recognizes GU and AG dinucleotides. Splicing factors play a significant role in regulating AS, and exon splicing enhancer (ESE)/silencer (ESS) or intron splicing enhancer (ISE)/silencer (ISS) can bind to SR or hnRNP proteins, respectively. LncRNAs can effectively modulate AS in various ways. A & B. Epigenetically, lncRNAs interacted with SR or hnRNP proteins as “interactors” or “hijackers”. C. Many lncRNAs recruited AS-associated RBPs to splicing sites and changed the AS pattern under specific conditions. D. Some lncRNAs, such as Linc-HELLP, could fine-tune AS through direct binding to ribosomal and splicing complexes. E. Partial lncRNAs, especially antisense lncRNAs, tend to form RNA-RNA duplexes with sense transcripts and block spliceosome interactions. F. LncRNAs effectively changed epigenetic modifications or chromatin modeling to influence the overall AS decision.

Vast numbers of transcripts in humans cannot be further translated into proteins. However, ncRNAs seem to affect diverse biological processes by modulating mRNA transcription, splicing, stability, transportation and epigenetic modification [56]. A plethora of long noncoding RNAs (lncRNAs) have been shown to pivotally regulate AS during disease, and the fine-tuning mechanisms extend across all steps of gene expression, including epigenetic sensu lato, co/posttranscriptional control, miRNA maturation and protein stability [27,57,58]. Herein, we summarized the principle that lncRNAs influence AS [5-14,16,59-70] (Figure 3). First, lncRNAs such as Malat1 [71], PVT1 [60], Xist [13], and PANDAR [11] could interact with specific splicing factors (SFs) and function as “interactors” or “hijackers” of splicing regulators, such as SR or hnRNP proteins (Figure 3A, 3B). For example, Malat1 (Neat2) [71,72] is able to interact with SR proteins as an “interactor” and influences the nuclear speckle distributions of SR proteins through modulating their phosphorylation status or reshaping the 3D genome organization to induce AS-related nuclear speckle formation, leading to changes in the whole AS profile. Gomafu exists in the nucleus as a “hijacker” by binding to multiple SFs (SF1, SRSF1, QKI, etc.), and Gomafu is downregulated under stimulation; thus, SFs can freely direct the AS process [66]. In addition, there are two derivative mechanisms: recruitment of RBPs to regulated splicing sites (Figure 3C) and binding to ribosomal or splicing complexes (Figure 3D). Specifically, SPA, a type of lncRNA that is 5’ snoRNA capped and 3’ polyadenylated, could accumulate to alter the AS pattern in PWS disease through binding to RBPs, such as TDP43, RBFOX2 and hnRNP M [73]. INXS could bind the Sam68 splicing modulator and cause a shift in splicing from BCL-XL to BCL-XS downstream, leading to caspase-related apoptosis in cancer cells [12]. LINC-HELLP has been confirmed to bind spliceosome components or ribosomes as a crucial AS regulator [74]. Second, lncRNAs can sometimes form RNA-RNA duplexes with pre-RNA molecules to affect AS events. One example is Fas-antisense (Saf), which can interact with the Fas premRNA and facilitate the exclusion of exon 6 in Fas through combining with SPF45, resulting in the accumulation of FasΔE6, (sFas) which prevents tumor cells from undergoing FasL-induced apoptosis [75]. Zeb2 NAT [76] and lincRNA-uc002yug.2 [17] modulate AS in a similar manner by interacting with the Zeb2 5’UTR to mask the 5’ splice site or covering the SRSF1/MBNL1 recognition site to facilitate AS of the RUNX1 premRNA, respectively. Third, lncRNAs could also recruit chromatin remodelers to the specific locus to regulate the AS process of target genes (Figure 3E). In addition to FGFR2, the FGFR2 locus encodes an evolutionarily conserved nuclear antisense lncRNA (asFGFR2) to promote epithelium-specific alternative splicing of FGFR2. asFGFR2 can recruit Polycomb-group proteins and the histone demethylase KDM2a to create a chromatin environment that is inaccessible for a repressive chromatin-splicing adaptor complex, which is much more important for mesenchymal-specific splicing [10].

To a large extent, various types of lncRNAs participate in modulating AS patterns, including NATs, UCEs and piRNAs. Historically, efforts to generate small molecular drugs targeting protein activation have been unsatisfactory. Given that lncRNAs could be inhibited much more easily through oligonucleotide-based drugs, such drugs seem preferable to undruggable proteins for cancer therapy. We are hoping to develop lncRNAs as pivotal diagnostic or therapeutic targets in the near future [77-80].

Regulation of AS by the m6A modification

N6-methyladenosine (m6A) is the most prevalent and significant modification of eukaryotic transcripts, affecting various aspects of biological processes, including RNA splicing [81]. This process is catalyzed by methyltransferases, such as METTL3, METTL14, WTAP, KIAA1429 or RBM15/RBM15B, which are called “Writers” in mammals [82,83]. RNA m6A methylation occurs mainly at the “RRACH” consensus sequence motif (R, purine; H, nonguanine base). The methylation process is reversible because of the existence of demethyltransferases, such as FTO and ALKBH5, which serve as “Erasers”. Essentially, m6A modification could contribute to mRNA fate determination by recruiting m6A recognition proteins named “Readers” (Figure 4).

Figure 4.

Schematic diagram of m6A modification, which is essential for AS. m6A methylation of RNAs is catalyzed by the multicomponent methyltransferase complex called “Writers”, which includes three main subunits: METTL3, METTL14 and WTAP, and sometimes also contains KIAA1429 or RBM15/15B. FTO and ALKBH5 are m6A demethylases that catalyze m6A removal as “Erasers”. Some m6A-binding proteins named “Readers” contribute to recognizing m6A modifications and modulating various pathways. Thus, hnRNAPA2B1, hnRNPC and YTHDC1 seem to fine-tune RNA processing or splicing, while YTHDF1/3 and eIF3/4F influence mRNA translation, and YTHDF2, YTHDC2 and HuR play pivotal roles in RNA stability or decay. Perturbation of the dynamic m6A status causes alterations in AS. For example, the m6A removal catalyzed by FTO brings about the failure of SRSF2 binding to methylated RNAs, and exons are skipped in unmethylated transcripts; in contrast, YTHDC1 could effectively bind to m6A methylated RNAs and recruit SRSF3 instead of SRSF10 to pre-RNAs, leading to exon inclusion.

Perturbation of the dynamic status of m6A could affect the levels of a large number of RNAs, since the methylation-related molecules were located in nuclear speckles and confirmed to interact with splicing factors that may be involved in modulating AS [84]. Additionally, dysregulation of methyltransferases, demethyltransferases and readers all remarkably influenced AS. METTL3 knockdown facilitates the MyD88 variant through regulating AS [85] and can alter the whole splicing isoform pattern by combining with WTAP, which was initially identified as the WT1 binding splicing factor through targeting the AS site in multi-isoform genes [86]. Emerging evidence has indicated that METTL16, the U6 snRNA m6A methyltransferase, is required for the AS of MAT2A with retained introns and stabilizes MAT2A mRNA by methylation to modulate SAM synthetase. Furthermore, METTL16 could mediate splicing of diverse cellular RNAs, including ncRNAs [87-89].

With respect to the demethyltransferases, re-cent studies indicated that the downregulation of ALKBH5 could affect the distribution of splicing factors, such as SC35, ASF/SF2 and SRPK1 [90]. Additionally, ALKBH5-mediated m6A erasure effectively ensured the appropriate splicing event to produce mRNAs with longer 3’UTR and prevent degradation, mainly due to the predominant colocalization of ALKBH5 and SC35 in the nucleus [91]. Moreover, FTO could mediate AS of RUNX1 through recruiting SRSF2 (recognizes the ESE (Exon Splicing Enhancer) sequence). Specifically, FTO reduces the capacity of SRSF2 to bind to RUNX1, while exon skipping occurs in the unmethylated premRNA to form the short isoform and promote preadipocyte differentiation. In contrast, m6A-modified premRNA (FTO deficiency) could effectively recruit SRSF2 and induce exon inclusion, while METTL3 could antagonize this effect [92] (Figure 4). However, apart from the abovementioned results, AS is tightly coupled to 3’ end processing, and m6A sites are markedly enriched in the 3’UTR [93]. FTO could potentially influence alternative poly(A) site (APA) usage and 3’UTR length [94], possibly through influencing the U2AF 65-CF I_m-PAP interaction or altering the miRNA binding pattern [95,96].

Ultimately, the classical m6A readers include YTHDF-3 and YTHDC1-2 of the YTH domain family [97]. The functions of YTH domain family readers are well established, and they effectively accelerate RNA processing, translation, decay or transportation/localization and even affect viral life cycles [98,99]. YTHDC1, which is a nuclear m6A reader, has been verified to modulate splicing through maintaining a dynamic regulation network [29,100]. In detail, YTHDC1 recognizes m6A modification and effectively recruits SRSF3 to pre-RNAs, accompanied by exon inclusion. Conversely, the pre-RNAs are unmethylated, and YTHDC1 fails to recognize the RNA; therefore, SRSF10 binds to and masks the region near the m6A sites, ultimately resulting in exon skipping owing to the failure of m6A modification [98,100]. Above all, m6A modification could mediate fine-tuning by decreasing the cognate combination of SFs and changing the secondary RNA structure to alter the interaction capacity between SFs and m6A-methylated RNAs [101]. The second class of m6A readers includes heterogeneous nuclear ribonucleoproteins (hnRNPs), such as hnRNPC, hnRNPG and hnRNPA2B1 [99,101]. In addition to eliciting AS in a manner similar to that described for METTL3 by directly binding nuclear transcripts, hnRNPA2B1 plays a pivotal role in AS and in premiRNA processing by interacting with the DGCR8-DROSHA complex [102] (Figure 4). Overall, m6A modification is closely linked to the AS process and may develop into overriding signaling for mRNA processing and splicing.

Conclusions and perspectives

By reprogramming genome expression to confer proteomic and transcriptomic diversity, AS is becoming one of the most important dynamic regulatory processes for adaptation to a changing microenvironment. Therefore, it is not surprising to see so many mechanisms of AS regulation by epigenetic networks, such as the modification of DNA, RNA and histone proteins. In this review, we aim to retrospectively evaluate new research on alternative splicing, which will facilitate the process from discovery to future therapeutic use. The main topics for the study of AS are as follows: history of mRNA splicing, components of the spliceosome, introns/exons in alternative splicing, spliceosome complex regulation, diseases related to splicing and potential therapeutic approaches to splicing-induced diseases [103,104].

With regard to therapeutic options for AS-relevant disease, especially for cancer, there are quite a few achievements that could be used as a reference. Many small molecules that can modulate AS have been developed as potential treatment targets for disease through reversing cancer or drug resistance [105-107]. For instance, BaxΔ2, a novel isoform of Bax with exon 2 alternative splicing, could induce the mutation of Bax, and this AS event effectively increased the sensitivity of colon cancer cells to caspase-8-targeted chemotherapeutic drugs [108]. In addition, clinical trials (phase III) of antisense oligo-targeted therapies have been reported [109].

In conclusion, further exploration of AS is needed to elucidate the full landscape of AS, the complete regulatory networks that control AS, such as ncRNAs or m6A modifications, and the function of the vast majority of AS events associated with diseases or disorders.

Disclosure of conflict of interest

None.