Pre-mRNA splicing and its co-transcriptional connections

Abstract

The transcription of eukaryotic genes by RNA polymerase II (Pol II) yields RNA precursors containing introns that must be spliced out and the flanking exons ligated together. Splicing is catalyzed by a dynamic ribonucleoprotein complex called the spliceosome. Recent evidence has shown that a large fraction of splicing occurs co-transcriptionally as the RNA chain is extruded from Pol II at speeds up to 5kb/min. Splicing is more efficient when it is tethered to the transcription elongation complex and this linkage permits functional coupling of splicing with transcription. Here we discuss recent progress that has uncovered a network of connections linking splicing with transcript elongation and other co-transcriptional RNA processing events.

Splicing and synthesis of mRNA precursors are functionally coupled

Almost all protein-coding genes and many long non-coding RNA genes are transcribed by RNA polymerase II (Pol II; see Glossary) to make precursors containing non-coding intron sequences that are spliced out and the flanking exons ligated with exquisite accuracy. The average human gene has about 10 introns that make up about 90% of its length. Splicing of precursors in alternative ways generates multiple mRNAs from over 95% of human genes [1]. Alternative splicing entails the inclusion or skipping of alternative cassette exons, intron retention, and selection of alternative 5' and 3' splice sites (SSs) that extend or truncate exons [2].

Splicing is catalyzed by the spliceosome, a dynamic complex comprising 5 non-coding U snRNAs and over 170 proteins [3-5]. Classic electron microscopy (EM) showed that splicing can occur on growing transcripts still attached to Pol II [6,7] and subsequent nascent RNA sequencing showed that a large fraction of splicing occurs co-transcriptionally [8-11]. This means that the substrate for the spliceosome is a “work in progress”; a transcript that is being extruded through the Pol II exit channel at rates varying from <0.5kb/min to >5kb/min [12].

Co-transcriptional splicing occurs close to the transcription elongation complex (TEC), which comprises RNA polymerase II, the nascent transcript and the chromatin template as evidenced by the genome-wide association of splicing U snRNPs specifically with intron-containing genes [13]. Other factors that “travel” with the TEC include RNA binding proteins (RBPs), elongation factors that control growth of the RNA chain, and processing factors that carry out co-transcriptional capping, base modification and 3’ end formation. Splicing is more efficient when it is physically tethered to the TEC [14-17] and this linkage permits functional coupling of splicing with transcription. These coupling mechanisms are of two classes: spatial and kinetic [18]. Spatial coupling is mediated by contacts between splicing factors and the transcription machinery, while kinetic coupling links splicing outcomes to the rate of RNA chain synthesis. Coupling mechanisms can work in both directions; meaning transcription can affect splicing and vice versa. Here we review recent research on the connections linking splicing with elongation of the transcript and other co-transcriptional RNA processing events.

Heterogeneous pathways of co-transcriptional splicing

Spliceosome assembly and activation are highly regulated and vary greatly in efficiency between introns in different biological contexts. In addition to splice site and branch point sequences recognized by the core spliceosome [19,20], splicing is modulated by diverse RNA binding proteins (RBPs) including SR proteins and hnRNPs. These factors exert context-dependent positive and negative effects on spliceosome function through co-transcriptional binding to exonic and intronic splicing enhancer and silencer elements [2,21,22]. How interaction of splicing factors with the TEC regulates spliceosome assembly and remodeling is an important unresolved question.

In vitro splicing reactions and spliceosomal CryoEM structures are reconstituted with synthetic RNA substrates that often have a single short intron and flanking exons. Splicing complexes assembled on these substrates are uncoupled from transcription. In contrast, natural pre-mRNA substrates are far more complex and dynamic, as they comprise growing RNA chains still attached to Pol II, with multiple exons and introns that can be many kilobases long. Commitment to pairing of 5’ and 3’ SSs is made during formation of A complex (Figure 1) [23], but how proper pairing is achieved co-transcriptionally, when multiple 5’ and 3’ splice sites may be present on a nascent transcript, remains poorly understood. At the commitment step, bridging occurs between U1 snRNP at the 5’ SS and U2 snRNP and U2AF at the 3’ SS through both direct and indirect contacts mediated by a network of SR proteins [24-27]. Bridging between 5’ and 3’ SSs can be made in one-step across an intron to make an intron definition complex, or in a two-step process where initial cross-exon contacts form an exon definition complex that transitions to the intron definition complex [27,28]. U1 snRNP occupancy at both 5’ SSs flanking a 3’ SS synergistically stimulates splicing, suggesting that cross-exon and cross-intron cooperation can promote a single splicing event [29]. Structural modeling suggests yeast A complex is equally compatible with cross-exon and cross-intron contacts between U1 and U2 snRNPs [30] (Figure 2).

Figure 1: Co-transcriptional spliceosome assembly on nascent pre-mRNA attached to RNA Pol II.

U1 snRNP binds the 5'SS (GU), SF1 binds the branch point A (red) and U2AF^65/35 subunits bind the polypyrimidine tract (pY) and 3’ SS (AG) to form E complex on the nascent transcript. A star indicates the 3’OH growing end of the RNA. Note interactions between Pol II and splicing components including helicases that mediate rearrangements within the spliceosome are poorly understood and many are not shown in this simplified depiction. U2 snRNP replaces SF1 to form A complex. Recruitment of the U4/U6.U5-tri-snRNP forms the pre-catalytic B complex. Then structural rearrangements release U1 and U4 snRNPs to form the activated B-complex (B^Act). B^Act catalyzes step 1 in which the 2' OH of the bulged branch point A attacks the 5'SS creating an upstream exon with a free 3'OH and an intron lariat attached to the second exon. Further rearrangements form C-complex that catalyzes Step 2, in which exons are ligated and the intron lariat structure (ILS) is excised and subsequently debranched and degraded. Putative protein:protein contacts between Pol II and splicing factors are not shown for clarity.

Figure 2:

Models for recognition of 5’ and 3’ SS by intron and exon definition with U1 snRNP bound to Pol II. Top panel: An intron looping/scanning model of co-transcriptional spliceosome assembly and intron definition [53]. Tethering of the 5’SS by U1 snRNP bound to Pol II causes intron looping that facilitates recognition of the BP and 3'SS by U2 snRNP and U2AF^65/35 by “scanning” thereby promoting formation of cross intron contacts (green arrows) and “ultrafast” splicing [33,34] before synthesis of exon 2 is complete. Pol II associated U2AF^65/35 could participate in scanning. Bottom panel: Pol II bound U1 snRNP could function in exon definition through the formation of cross-exon contacts (red arrows) with U2 snRNP and U2AF. An exchange of U2 snRNP and/or U2AF contacts between the downstream U1 snRNP (blue) to the upstream U1 snRNP (stippled) on the Pol II surface could accomplish the transition to an intron definition complex (top panel). In an alternative pathway, this transition may occur by direct binding of U6 to the upstream 5’SS [27,28]. Cap binding complex (CBC) is bound to the 5’ 7meG cap that is added co-transcriptionally.

It is usually thought that intron definition predominates when introns are short (<250 bases) whereas exon definition predominates when introns are long, which is common in mammals [31,32]. Recent studies of co-transcriptional splicing have challenged this view, however. Sequencing methods have been developed for chromatin-associated RNA that can simultaneously identify the 3’ OH end of the RNA, that defines the position of the RNA polymerase, as well as the splicing status indicated by the presence of exon:exon junctions or lariat branch points located upstream of the 3’ end. The surprising result of these studies is that both lariat formation and exon ligation can be completed when the RNA polymerase has extended as few as 18-26 bases past the 3’ SS (Figure 2, top panel) [33-35]. Using co-transcriptional lariat sequencing (CoLa-seq) to map the 3’ end and branch point of lariat-exon2 splicing intermediates, Zeng and colleagues showed that in human cells with long introns, step 1 often occurs before transcription of the downstream exon is complete, thereby precluding cross exon interactions required for exon definition. This work suggests that exon definition, which predominates in splicing of long introns in vitro uncoupled from transcription, may be less favored when splicing occurs co-transcriptionally in vivo. Additional work is required to establish exactly how important exon definition is in the context of co-transcriptional splicing.

While “ultrafast” splicing can occur at many yeast and human introns, the average co-transcriptional splicing event is probably much slower, and is not completed until Pol II has transcribed hundreds or thousands of bases beyond the end of the intron [34,36,37]. Indeed live cell imaging of individual genes indicates that before they are removed, unspliced introns can persist for many minutes after transcription of the intron is complete by which time the polymerase has presumably travelled far downstream [38]. That most splicing occurs after polymerase has transcribed well past the 3’ SS is also supported by metabolic labelling of RNA for short periods in budding yeast followed by sequencing to measure the kinetics of accumulation of unspliced precursors, step 1 lariat intermediates and fully spliced products [37]. This study showed rates of splicing vary from 30 seconds to 15 minutes for different introns and that the median half-life for conversion of a precursor to fully spliced product is around 2 minutes (about 80 seconds for step 1 and 40 seconds for step 2) during which time polymerase, at its normal speed, would reach the end of most yeast genes. Interestingly splicing is faster for introns with longer distances between the 3’ SS and the end of the gene consistent with the idea that it is facilitated by longer engagement of the spliceosome with the TEC [37].

Not only is the timing of co-transcriptional splicing highly variable, but so are the actual splice sites used to remove a given intron. This surprising conclusion came from imaging splicing in live cells where individual introns were marked by insertion of hairpin loops at known positions permitting visualization with fluorescently tagged RNA binding proteins. Loss of the fluorescent signal at the site of transcription acts as a surrogate for excision of the intron. Remarkably, the labelled intron elements often only persisted at the site of transcription for short periods, that were not long enough for transcription of the whole intron to be completed [38]. The explanation of this conundrum is that introns can be excised piecemeal by stochastic splicing to different zero length pseudoexons comprising juxtaposed 3’ and 5’ splice sites (AGGT) within introns. This process regenerates splice sites at the junctions that are subsequently used in another round of recursive splicing (RS) until the entire intron is removed [39]. Live cell imaging suggests that co-transcriptional intron removal by RS occurs far more frequently than previously suspected [38] probably because RS intermediates are unstable and therefore rarely detected by RNA-seq. On the other hand if RS is in fact a common event, it remains unclear how it evades the suppression of 5’ SS’s by the exon junction complex that is rapidly deposited at splice junctions [40,41]. In summary, splicing of individual introns does not follow a unique pre-determined or hard-wired pathway. Instead splicing is achieved in multiple ways: co-transcriptionally or post-transcriptionally, co-transcriptionally immediately after 3’ SS synthesis or with a delay, using either exon or intron definition and by classical two-step splicing or multiple recursive steps. Thus co-transcriptional splicing appears to be highly non-uniform, begging the question of whether regulatory mechanisms modulate the choices between alternative pathways that can be used to remove a given intron.

The U1 snRNP-Pol II connection and spatial coupling

Functional coupling of splicing with transcription likely involves physical interaction of Pol II with splicing factors including U2AF, Prp19 [42], and SR proteins [43,44]. Such interactions involve contacts with both the Pol II body and its C-terminal domain (CTD) a unique feature of Pol II, not found on other RNA polymerases. The CTD is an intrinsically disordered region (IDR) comprised of conserved heptad repeats (Y₁S₂P₃T₄S₅P₆S₇) that are reversibly phosphorylated at multiple positions during transcription [45]. Truncation of the CTD inhibits splicing [46] in vivo, and in vitro the phosphorylated CTD can enhance splicing [47]. Furthermore inhibition of the Ser5 CTD kinase CDK7 causes widespread disruption of splicing [48]. Splicing factors including U2AF and U1 snRNP co-purify preferentially with Pol II bearing the Ser5 phosphorylated isoform of the CTD [42,49,50] which can form phase separated condensates where splicing factors are concentrated [51]. Ser5 phosphorylation is strongly enriched at the 5’ ends of genes but how this bias affects splicing is unknown. The CTD could serve as a sort of antenna that traps splicing factors at the site of transcription or it could allosterically stimulate specific steps in spliceosome assembly or both. While the inhibition of splicing by CTD truncation and other Pol II mutations demonstrates that splicing co-transcriptionally is actually functionally important, it remains to be determined what fraction of introns normally require co-transcriptional splicing for proper expression of the genes in which they reside.

The best characterized interaction of a splicing factor with the TEC is between U1 snRNP and RNA Pol II [52] described in a landmark CryoEM structure [53]. U1 snRNP contacts the back side of Pol II, near the RNA exit channel where the U1-70K subunit approaches conserved residues in the Pol II Rpb2 and Rpb12 subunits. Genetic studies are required to establish the physiological significance of the protein:protein contacts between Pol II and U1 snRNP in the structure, but notably it predicts that U1 snRNA can base pair with a newly synthesized 5’ SS (Figure 2). Formation of the U1 snRNP-Pol II complex does not require an RNA transcript however. Direct contact of U1 snRNP with Pol II is consistent with the recently reported stimulation of transcription elongation by this snRNP [54]. The U1 snRNP-Pol II complex is also consistent with localization of U1, unlike other snRNPs, at transcribed genes independent of splicing [55-57] both at start sites and within gene bodies [57-60]. Furthermore, if Pol II is stalled part way through an intron by using targeted Cas9 as a roadblock, then the 5’SS of that intron co-purifies with Pol II [58] consistent with tethering of the 5’ SS-U1 snRNP complex to Pol II as in the cryo-EM structure of Zhang et al [53].

The U2 snRNP can be modeled into the U1snRNP-Pol II complex [53] suggesting that cross-intron or cross-exon interactions could form in close proximity with the TEC (Figure 2). An intriguing possibility suggested by Zhang and colleagues is that as an intron is transcribed, it forms an expanding loop with the growing end of the transcript in the pol II active site at one end, and the 5’ SS captured by U1 snRNP bound to pol II at the other end [53] (Figure 2). In this scenario the TEC might function as an intron “scanner” that promotes formation of cross intron complexes. It would presumably be advantageous for the “scanner” to recognize 3’ SSs as well as 5’ SSs, and U2AF might function in this role. U2AF co-purifies with Pol II and the U2AF65 subunit appears to be handed off from Pol II to the RNA transcript [42,61,62] (Figure 2). Rapid intron definition facilitated by intron looping/scanning may permit “ultrafast” splicing before the exon is fully transcribed (Figure 2). Whether exon and intron definition are equally feasible in the co-transcriptional context and whether U1 snRNP bound to Pol II participates in both recognition pathways remains to be seen. Given that only one U1 snRNP can contact Pol II at one time, co-transcriptional exon definition and subsequent splice site pairing would presumably require an exchange on the pol II surface between U1 snRNP bound to the downstream 5’SS (non-stippled in Figure 2) and U1 snRNP bound to the upstream 5’SS (stippled in Figure 2).

In summary the U1 snRNP-Pol II structures have galvanized thinking about co-transcriptional splicing and raised interesting questions for future investigation. For example: “Is there a mechanism, perhaps a transcriptional pause, that ensures U1 snRNP is in place on the Pol II surface ready to capture a 5’ SS when it emerges from the RNA exit channel?” and “Could U1 snRNP association with Pol II affect its interaction with the 5’ SS?” If so, this could have significant consequences for U1 snRNP’s function in discrimination between genuine and cryptic 5’ SSs, and in regulation of 5’SS recognition by enhancer and silencer elements [63-65].

Determinants of co-transcriptional splicing

Transcription imposes an order and timing on synthesis of splice sites and RBP binding sites whereas on a full-length transcript they are all presented simultaneously. This distinction means that different regulatory mechanisms may operate on co-transcriptional and post-transcriptional splicing. An important challenge is to uncover how co-transcriptional splicing efficiency (i.e. the fraction of transcripts where splicing of a particular intron is completed before release from the polymerase by cleavage at the polyA site) is controlled. Nascent RNA sequencing has identified features that correlate with efficient co-transcriptional splicing including slow transcription which lengthens the window of opportunity for splicing before transcription terminates [66,67], structured RNA at SSs [67], and strong U2AF binding sites [34]. Co-transcriptional and post-transcriptional splicing take place in distinct sub-nuclear environments where regulatory proteins could operate differently. This possibility is demonstrated by recent work on the polypyrimidine tract binding protein PTBP1 best known as a repressor of splicing at specific introns which was discovered by examination of mature mRNAs [68,69]. Analysis of nascent RNA however, revealed that PTBP1 has an opposite function in promoting co-transcriptional splicing of a distinct group of introns [70]. Thus PTBP1 appears to have a predominantly negative effect on splicing when acting post-transcriptionally and a positive effect when acting co-transcriptionally.

Ongoing transcription is an important influence on the order in which introns are removed and on coordination between splicing of different introns in the nascent transcript. In general introns near 5’ ends are removed before those at 3’ ends as predicted by the “first come first served” model [71], however the order in which introns are removed is by no means strictly determined by the order in which they are transcribed [72]. Among adjacent pairs of human introns, upstream introns are preferentially spliced out first only about half the time [36] and frequently within a pair, one is “always first” [72]. Adjacent splicing events are often interdependent resulting in clusters of introns with similar co-transcriptional splicing efficiencies [32,36]. In the most extreme cases, nascent transcripts from a given gene have all-or-none splicing of the introns [73,74]. The all-unspliced transcripts are also poorly processed at polyA sites reflecting the coupling between these processing steps (see below). It is not known whether the all-unspliced transcripts are eventually processed into mRNA or are dead-end products. In summary co-transcriptional splicing efficiency is determined by mechanisms operating on individual introns, on groups of neighboring introns, and even on whole transcripts.

The influence of co-transcriptional RNA folding on splicing

Folding of a growing RNA chain is a highly dynamic process that has recently been revealed in vivo by chemical probing methods [75]. Nascent RNA folding is sensitive to transcription speed; slow transcription favors base pairing of more proximal elements whereas fast transcription favors pairing of more distal elements to form more open structures [76]. The alternative RNA conformations assumed by a growing transcript are most accurately described as ensembles of multiple local energy minima that govern their probabilities of formation [77,78]. The ensemble of alternative structures for a particular sequence element changes as the RNA chain grows in a way that is sensitive to its rate of growth and to transcriptional pausing [79,80]. Even the addition of a single nucleotide to a growing RNA chain can instigate a dramatic change in structure through co-transcriptional strand displacement [81]. Different RNA structures assumed by the pre-mRNA influence constitutive and alternative splicing by modulating the proximity and accessibility of splice sites and RBP binding sites [82-85]. Splice sites and binding sites for RBPs must be presented for recognition in a suitable conformation that is usually single-stranded [86]. Indeed extensive RNA unfolding of structures around splice sites is predicted to occur within the spliceosome [87] and sequestration of splice sites within structures has long been known to reduce splicing efficiency [88-90]. The relevance of nascent RNA structure was recently highlighted by a study of disease causing mutations in the MAPT gene encoding the microtubule associated Tau protein. These mutations change inclusion of alternative exon 10 in ways that correlate with their effects on the ensemble of structures around the 5’ SS of that exon [87].

Alternative splicing is slower and more frequently completed post-transcriptionally than constitutive splicing [8,9,72] in part due to poorly defined mechanisms that delay splicing of introns that flank alternative exons [72]. However splicing reactions that are completed post-transcriptionally may still be affected by earlier co-transcriptional events. One example occurs at alternative exons that are sensitive to transcription speed. When slow transcription favors inclusion of these exons, it correlates with reduced RNA structure at the 3’ SS downstream of the cassette exon, and conversely when slow transcription favors exon skipping, RNA structure at the downstream 3’ SS is elevated [67]. Hence co-transcriptional RNA folding appears to influence the outcome of alternative splicing that is completed post-transcriptionally. In summary the rapidly changing ensembles of alternative structures assumed by the nascent transcript as it grows are likely to have important effects on splicing outcomes whether the process is completed co-transcriptionally or post-transcriptionally.

Splicing and co-transcriptional RNA modifications

Splicing is one of several interdependent co-transcriptional pre-mRNA processing steps that include 5’ capping, covalent nucleotide modification and cleavage-polyadenylation (CPA). The 5’ cap is added shortly after the 5’ end emerges from the RNA exit channel by capping enzymes that directly contact Pol II [91-93]. The cap is recognized by nuclear cap binding complex (CBC, Figure 2) which promotes recognition of the exon 1 5’ SS by U1 snRNP and subsequent handoff to U6 [94-97].

As the nascent transcript grows, it becomes partially modified at specific nucleotides within introns and exons and some of these covalent marks affect splicing. One of the most widely distributed covalent modifications in nascent RNA is pseudouridine. Knock out of the PUS1 pseudouridine synthase alters the inclusion of over 2000 alternative cassette exons [98]. This widespread effect of pseudouridylation on splicing is mediated at least in part by modification of RBP binding sites [98], but it could also work by enhancing the stability of RNA duplexes [99].

N⁶ Methyladenosine (m6A), which is a modification deposited co-transcriptionally by METTL3/METTL14 [100] affects splicing by multiple mechanisms that affect RBP binding and RNA folding. m6A deposition in exons close to splice junctions is associated with rapid constitutive splicing, while in introns it is associated with slower splicing typical of alternative cassette exons [100]. m6A destabilizes RNA structures in a way that facilitates binding of the splicing regulators hnRNPC and hnRNPG, and thereby affects inclusion of many alternative exons [101,102]. RNA binding by the m6A reader YTHDC1 can also influence splicing by selective recruitment of SR proteins to the transcript [103].

Folding of the nascent transcript generates duplexes, often between oppositely oriented repeat elements in introns, and these structures serve as substrates for ADAR1 and ADAR2 (adenosine deaminase acting on RNA) that co-transcriptionally convert adenosine to inosine [104]. A-I conversions can affect splice sites and more commonly regulatory elements and thereby influence numerous alternative splicing decisions [105]. Conversely, reduced splicing enhances intronic A-I editing probably by lengthening the window of opportunity for attaching these co-transcriptional modifications [106]. Enzymatically inactive ADAR1 also affects splicing presumably by binding to duplexes and stabilizing them or excluding other factors [105].

Connections between splicing and cleavage/polyadenylation

There is a complicated relationship between co-transcriptional splicing and 3’ end processing by cleavage/polyadenylation (CPA). Splicing of the last intron and processing at the polyA site are mutually interdependent though the mechanisms responsible for this coupling are not well understood. PolyA site recognition facilitates splicing of the last intron probably because it defines the 3’ end of the last exon and mutation of the polyA site inhibits terminal intron splicing [107,108]. Conversely recognition of the last intron by splicing factors facilitates processing at the polyA site that defines the end of the last exon [109-112]. Coupling between splicing and 3’ processing at the end of the gene likely involves cooperative interactions between U2snRNP, U2AF and CPA factors [113,114].

In contrast, at cryptic polyA sites within introns, there is strong antagonism between splicing and CPA that plays a major role in preventing premature truncation of mRNAs [115,116]. There are two models to account for repression of intronic polyA sites by splicing: kinetic competition and direct interference. According to the kinetic competition model, if splicing removes an intron harboring a cryptic polyA site(s) before that site is processed, then premature 3’ end formation will be averted. This model is supported by the finding that lengthening of an intron and hence increasing the delay until it is spliced out can enhance use of an intronic alternative polyA site [117]. By the same token accelerating splicing by strengthening of a 5’ SS can suppress use of an intronic polyA site [118]. Similarly, inhibition of splicing by Pladienolide B (Plad B) which blocks U2 snRNP function can enhance intronic polyadenylation and coupled transcription termination [35]. The interference model postulates that U1 snRNP bound to a 5’ SS downstream of a polyA site can prevent cleavage at that site via inhibitory interactions with the CPA machinery [119,120]. Depletion of U1 snRNA stimulates premature CPA globally suggesting that U1 snRNP normally safeguards transcripts against truncation of their 3’ ends in a process dubbed “telescripting”[116]. It is also possible that U1 depletion might de-repress premature CPA by slowing down splicing as predicted by the kinetic competition model. Another potential interference mechanism might be that U1 snRNP binding to Pol II competes with the recruitment of CPA factors that are thought to perform poly A site cleavage in very close proximity to Pol II [121,122].

Whether a polyA site is subject to suppression by the splicing machinery appears to depend on whether it is situated in an intron or not. The importance of intronic versus exonic context for polyA site recognition is demonstrated by examples in which an entire gene is embedded within an intron of a host gene. In these cases the same polyA site is processed when it is transcribed as the last exon of the nested gene but when it is transcribed as one of the host gene introns, processing is prevented [123]. This example suggests that the TEC somehow “knows” whether it is in an intron and unable to support CPA or in an exon where CPA is supported. The distinction might correspond to different factors associated with the TEC when it is transcribing an exon compared to an intron which might depend on its co-transcriptional splicing history.

Splicing connections with transcription initiation, elongation and chromatin modification

The average speed of transcript elongation by RNA Pol II varies from <0.5kb/min to >5.0kb/min both between genes and within genes [12]. Transcript elongation comprises short intervals of relatively rapid growth punctuated by frequent pauses where the polymerase may even backtrack. Pausing is more frequent, and transcription is slower, in exons that have higher GC content than introns [124,125]. Slower elongation in exons, or pausing at specific positions in the vicinity of splice sites, could affect co-transcriptional splicing possibly by facilitating timely recruitment of splicing factors to the TEC [126]. High resolution mapping of Pol II show that pausing does not generally occur at 3’ SSs, at least in metazoans [74,127]. Whether pausing occurs near 5’ SSs remains unresolved in part because contaminating Step1 intermediates with a free 3’ OH end at the splice site confound the analysis. It is intriguing to note however that the consensus sequence of mammalian Pol II pause sites is GT at the pause and +1 positions in common with the first two nucleotides of most introns [127].

Changes in average transcription speed can have profound effects on alternative and constitutive splicing [66,128-130]. Pol II mutants that slow down or speed up elongation affect numerous alternative splicing decisions but we cannot predict whether decelerating or accelerating transcription will promote exon inclusion or skipping [128,131,132]. It is likely that transcription speed affects splicing in multiple ways that include changing the time delay between synthesis of competing splice sites or RBP binding sites, altering how the transcript folds (Figure 3) and how it is covalently modified. Both co-transcriptional N⁶A methylation and A-I editing are sensitive to transcriptional speed with slow elongation increasing the level of these modifications [67,133].

Figure 3: Potential mechanisms by which transcription speed affects co-transcriptional splicing.

Pol II speed influences co-transcriptional RNA folding, which can in turn promote or repress RBP binding. The RNA structural landscape and RBP interactome in turn influence constitutive and alternative splicing decisions. Fast elongation (top panel) favors less local RNA folding and could thereby favor binding of RBPs with single-stranded binding sites. In this example, the RBP enhances skipping of the alternative exon (red). Slow elongation (bottom panel) favors more local RNA structure, that can reduce RBP binding resulting in exon inclusion.

Not only does transcription affect splicing but recent studies suggests ways that splicing feeds back on transcription at the level of elongation and initiation. U1 snRNP contacts with Pol II clash with Rtf1, an important allosteric activator of elongation [134,135] suggesting that U1 snRNP binding might slow down transcription. Remarkably, inhibition of U2 snRNP by Plad B results in accumulation of Pol II near the 5’ end of the gene at the promoter proximal pause site consistent with impaired elongation [136,137]. It is unclear how inhibition of U2 snRNP engaged at the BP of an intron transcribed by one Pol II could affect elongation by another Pol II near the TSS [136], but notably the TAT-SF1 (Cus2 in yeast) subunit of U2 can regulate transcription elongation [138,139]. One possibility is that Plad B affects transcription by inhibiting release of TAT-SF1 from the U2 snRNP when it engages the BP [140].

Remarkably, inclusion of an alternative exon near the 5’ end of a gene can activate transcription initiation at an alternative start site within a few kb upstream [141] thereby creating an alternative first exon which is a major generator of transcript diversity [142]. This phenomenon of exon mediated activation of transcription start sites or EMATS is thought to be mediated through splicing dependent recruitment of transcription initiation factors [141,143,144]. The link between alternative splicing and transcription initiation may have significant consequences for therapeutics that affect alternative exon inclusion like the Spinal Muscular Atrophy drug Nusinersen, an antisense oligonucleotide that increases inclusion of SMN2 exon 7 by masking a splicing silencer. Increased exon inclusion may therefore not only alter mRNA coding capacity but also stimulate transcription through EMATS [144].

Splicing factors operate in the vicinity of chromatin where they have the opportunity to engage in cross-talk with histones, and indeed changes in splicing correlate with deposition of certain histone variants [145] and covalent histone marks [146,147]. In particular, histone H3 K36me3 is co-transcriptionally deposited by SETD2 preferentially on nucleosomes within exons, in a way that correlates with splicing activity [146,148]. An unintentional side effect of altering splicing with antisense oligonucleotides like Nusinersen can be the establishment of a repressive chromatin mark, H3 K9me2, close to the target exon. This observation has suggested new therapeutic strategies that target both splicing and chromatin modification in a synergistic way [149]. In summary, recent research has revealed a web of interactions that link splicing of the nascent transcript with chromatin modification and transcription initiation and elongation. Working out exactly how these interactions function in regulated mRNA biogenesis remains an important challenge for the future.

Concluding remarks and Future Perspectives

Our understanding of co-transcriptional splicing has been advanced significantly in recent years by the development of sophisticated methods to capture and sequence nascent pre-mRNAs as they are being synthesized and processed by the spliceosome. These studies suggest some new rules that apply specifically to co-transcriptional splicing of nascent RNAs which were not revealed by work on splicing uncoupled from transcription, or by analysis of mature mRNAs. The relative importance of exon definition for splicing long introns and intron definition for splicing short introns has been questioned in light of nascent RNA studies that uncovered ultra-fast splicing [33,34]. Hard-wired or pre-determined pairing of 5’ and 3’ splice sites has also been questioned in light of evidence that it can be stochastic in the co-transcriptional context and that the timing of co-transcriptional splicing is very heterogeneous [37,38]. We are also beginning to learn about splicing regulatory mechanisms that work differently in co-transcriptional versus post-transcriptional contexts [70]. Structural probing of nascent transcripts is starting to yield information about the relationship between RNA folding and co-transcriptional splicing [67,87]. Nascent RNA studies are currently limited by sequencing read length and depth and by our ability to relate different data sets with one another. When these limitations are overcome many new insights will be revealed by relating splicing of different introns to one another, to RNA structures, to binding of RBPs, and to transcriptional pauses. The structures of complexes between components of the splicing machinery and the transcription machinery promise to revolutionize understanding of co-transcriptional splicing. Indeed this revolution has already been sparked by the suggestion that U1 snRNP binds Pol II in a way that is compatible with early 5’SS recognition [53](Figure 2).

Box 1: Spliceosome assembly and catalysis.

Splicing begins with base pairing of the 5’ SS with the U1 snRNA in U1 snRNP and recognition of the branch point (BP) Adenosine and 3’ SS by SF1 and U2AF respectively to make the E complex (Fig. 1). U2 snRNP then binds the BP in an ATP dependent step through base pairing with U2 snRNA and establishes contact with U1 snRNP to form the A complex where 5’ and 3’ SSs are selected. Subsequent recruitment of the U4/6 U5 tri snRNP and rearrangement ejects the U1 snRNP which is replaced at the 5’ SS by U6 snRNA. Base pairing between U6 and U2 snRNAs creates an active site which catalyzes the first trans-esterification reaction, Step 1 (Fig. 1). Step 1 creates two products: exon1 cleaved at the 5’ SS and the intron/exon2 lariat intermediate. Further remodeling converts the active site to the exon ligation conformation that catalyzes Step 2 trans-esterification. This reaction joins the exons and releases the intron as a lariat (Figure 1). Reviewed in [3,5,150].

Outstanding Questions.

• Are there functional differences between mRNA RNPs produced by co-transcriptional versus post-transcriptional splicing?

• What is the general significance of stochastic recursive splicing in removal of long introns and more generally how is pairing of 5’ and 3’ splice sites determined in a co-transcriptional setting?

• What is the significance of ultrafast splicing with intron definition versus delayed splicing with exon definition for regulated gene expression?

• Does a pol II-U1 snRNP-5’SS complex facilitate scanning of a looped intron for candidate 3’ SSs?

• How does regulation of pol II transcription elongation and pausing affect co-transcriptional splicing?

• How does coordinated splicing of neighboring introns work?

• How do the splicing and cleavage polyadenylation machineries communicate to achieve positive and negative regulation of 3’ end processing?

• What are the mechanisms that make transcription initiation and elongation sensitive to splicing?

Highlights.

• The pol II transcription elongation complex (TEC) makes contacts with the splicing apparatus to facilitate co-transcriptional splicing.

• Splicing can occur in an ultra-fast way soon after the 3’ splice site is synthesized even for long introns, suggesting that exon definition is not essential.

• Splicing of individual introns is highly plastic, with variation in kinetics, mode of exon:intron recognition, and the use of recursive splice sites.

• Nascent RNA folding into alternative structures is sensitive to the speed of transcription and may exert widespread effects on splicing.

• Exon mediated activation of transcription starts (EMATS) is a means by which splicing of an alternative exon feeds back on transcription of the same gene to activate weak promoters and select alternative mRNA 5’ ends.

Glossary

5' splice site (SS):

Site of RNA cleavage between the end of an exon and beginning of an intron which usually starts with the sequence GU. A loosely conserved 9 base sequence around the 5' SS is recognized by base pairing with U1 snRNA.

3' splice site (SS):

Site of RNA cleavage usually after an AG at the end of the intron. The AG is preceded by a polypyrimidine rich sequence (Py). The Py is recognized by U2AF⁶⁵ and die AG is contacted by U2AF³⁵.

5' Cap:

The 5’ triphosphate end of the primary transcript is converted co-transcriptionally to a N⁷-meG(5’) ppp(5’)-2’O meN cap structure by removing the terminal phosphate, GMP transfer, N⁷-methylation of the terminal G, and 2’O-methylation of the first transcribed base, N. Cap binding complex (CBC) facilitates splicing of the first exon.

Branch point (BP):

A short sequence usually containing an A base located 15-50 nucleotides upstream of the 3' SS that is recognized by base pairing with U2 snRNA. The 2’ OH of the BP A base attacks the 5’SS in step 1 to create an intronic lariat intermediate.

Cleavage/polyadenylation (CPA):

mRNA 3’ end formation by endonucleolytic cleavage (by CPSF73) of the nascent transcript about 20 bases 3’ of a polyadenylation signal (AAUAAA) and subsequenct addition of a non-templated polyA tail by polyA polymerase. CPA is functionally coupled to transcription termination.

Exon:

A sequence (often coding) within the pre-mRNA that is included in the mRNA. Exons are ligated together in step 2 of splicing facilitated by the spliceosome. Exons are ligated together via a transesterification reaction facilitated by the spliceosome.

Intron:

A transcribed non-coding sequence between exons that is excised by the spliceosome as a lariat RNA.

hnRNP:

Heterogeneous nuclear ribonucleoproteins are a class of RBPs that can affect alternative splicing by binding exonic splicing silencers and modulating spliceosome assembly.

PTBP:

The two polypyrimidine tract binding proteins are RBPs that bind upstream of 3’SSs. They can compete with U2AF and regulate alternative splicing

RNA polymerase II (pol II):

One of the three nuclear RNA polymerases; it transcribes all protein coding and many non-coding coding genes. It has a signature C-terminal domain (CTD) on its large subunit comprising intrinsically disordered heptad (YSPTSPS )Repeats.

SR proteins:

A class of conserved RNA binding proteins that facilitate early spliceosome assembly and other biological processes. SR proteins have RNA recognition motifs (RRMs) and phosphorylated RS domains with Arginine (R) Serine (S) dipeptide repeats.

U2AF:

U2 auxiliary factor is a heterodimer of 35 and 65kd subunits that serves to recognize functional 3’SSs early in spliceosome assembly. U2AF65 contacts the Py tract and U2AF35 contacts the AG of the 3’SS.

U snRNP:

U1, U2, U4, U5, U6 small nuclear ribonucleoproteins are core splicing factors that contain unique non-coding Uridine-rich RNAs and a set of common and unique protein subunits.