Coupling pre-mRNA processing to transcription on the RNA factory assembly line

Abstract

It has been well-documented that nuclear processing of primary transcripts of RNA polymerase II occurs co-transcriptionally and is functionally coupled to transcription. Moreover, increasing evidence indicates that transcription influences pre-mRNA splicing and even several post-splicing RNA processing events. In this review, we discuss the issues of how RNA polymerase II modulates co-transcriptional RNA processing events via its carboxyl terminal domain, and the protein domains involved in coupling of transcription and RNA processing events. In addition, we describe how transcription influences the expression or stability of mRNAs through the formation of distinct mRNP complexes. Finally, we delineate emerging findings that chromatin modifications function in the regulation of RNA processing steps, especially splicing, in addition to transcription. Overall, we provide a comprehensive view that transcription could integrate different control systems, from epigenetic to post-transcriptional control, for efficient gene expression.

Introduction

In eukaryotes, RNA polymerase II (pol II) catalyzes the synthesis of mRNAs (mRNAs), long non-coding RNAs and several different types of small RNAs, such as the spliceosomal small nuclear RNAs and microRNAs. The primary transcripts of these RNAs are immediately bound by ribonucleoproteins (RNPs) and processed into mature and functional forms via several distinct but coordinated steps.

The vast majority of precursor mRNAs (pre-mRNAs) undergo 5′ capping, splicing and 3′ polyadenylation before export to the cytoplasm for subsequent translation on ribosomes. Although each of these pre-mRNA processing steps can be recaptured independently in vitro, a number of studies have indicated that all these steps are integrated with each other and coupled to transcription.¹ Indeed, trans-acting factors involved in RNA processing can be recruited to emerging transcripts at different stages of transcription. Here, we discuss our current understanding of the molecular connection between transcription and pre-mRNA processing and reciprocal influences between different RNA biogenesis steps. Our discussion also includes recent evidence that nucleosome structure and DNA methylation constitute an additional means by which pre-mRNA splicing may be regulated.

Coordination of transcription and RNA processing by the RNA polymerase II CTD

RNA pol II is a large protein complex. The yeast RNA pol II consists of a 10-subunit core complex and the heterodimeric subcomplex Rbp4/Rbp7.² The catalytic subunit Rbp1 contains an unstructured carboxyl-terminal domain (CTD) that is not present in its functional analogs in RNA pol I and pol III. The pol II CTD consists of heptapeptide (YSPTSPS) repeats, the number of which varies from 26 in Saccharomyces cerevisiae to 52 in mammals.³ These repeats undergo several different post-translational modifications; most essential is phosphorylation of tyrosine (Tyr1), serine (Ser2, Ser5 and Ser7) and threonine (Thr4) residues (Fig. 1A). Moreover, some lysine and arginine residues in certain non-consensus repeats are subject to modifications such as acetylation, methylation, ubiquitylation and sumoylation. The dynamic change of the phosphorylation status of the CTD correlates well with the progression of transcription. Reversible phosphorylation involves several specific kinases and phosphatases.³ At the transcription initiation stage, CDK7, a component of the general transcriptional factor TFIIH, phosphorylates Ser5 and perhaps Ser7 of the CTD.^4,5 Subsequently, the elongation factor P-TEFb (CDK9/cyclin T) phosphorylates Ser2 while Ser5 is then dephosphorylated by several different phosphatases (for a review, see ref. 6). Most notably, the sequential phosphorylation and dephosphorylation of Ser5 and Ser2 is important not only for transcription but also for co-transcriptional processing of pre-mRNAs (for a review, see ref. 3 and 7). In addition, cellular signaling kinases such as ERK can also phosphorylate specific serine residues of the CTD and may thereby affect transcription and/or RNA processing.⁸

Figure 1. The C-terminal domain of RNA pol II coordinates nuclear pre-mRNA processing. (A) Differential phosphorylation of the pol II CTD during transcription. The pol II CTD coordinates 5′ capping (B), 3′-end polyadenylation (C), transcription termination of short transcripts (D) and spliceosome assembly and splicing activation (E). All abbreviations are defined in the text.

CTD and 5′ capping

The 5′ end of each RNA pol II transcript acquires the unique 7-methylguanosine cap structure.⁹ Capping occurs coincidently with promoter-proximal pausing of pol II, which is mediated by 5,6-dichloro-1-β-D-ribofuranosylbenzimidazole (DRB) sensitivity-inducing factor (DSIF) and the negative elongation factor (NELF), while the nascent transcript of ~20–40 nucleotides is being generated.¹⁰ The formation of the 5′ cap requires three enzymatic activities, including RNA 5′-triphosphatase (RT), guanylyltransferase (GT) and methyltransferase.^9,10 The RT removes the 5′-end γ-phosphate of the nascent transcript to generate a diphosphate-ended RNA. Subsequently, GT transfers a guanosine monophosphate to the 5′ diphosphate terminus of the RNA, followed by guanine methylation by methyltransferase. The capping enzyme is a heterodimer of the RT (Cet1) and GT (Ceg1) subunits in yeast, whereas it is a bi-functional single protein with both RT and GT activities in higher eukaryotes.^11,12 Nevertheless, both mammalian and yeast GT domains preferentially recognize CTD phosphorylated at Ser5, which leads to the allosteric activation of GT activity (Fig. 1B).^11,13 In addition to Ser5 phosphorylation, the Spt5 subunit of DSIF also interacts with the capping enzyme and activates mRNA capping.^14,15 After cap formation, P-TEFb phosphorylates Ser2 of the CTD as well as DSIF and NELF;^16,17 the dissociation of DSIF and NELF releases the promoter-proximal pausing of pol II. Interestingly, fission yeast P-TEFb also interacts with the cap-methyltransferase Pcm1 and may thus facilitate the completion of the capping process.¹⁸ Therefore, capping can serve as a gatekeeper for the productive elongation stage of transcription. Thus, pre-mRNA capping and transcription are tightly coupled and influence each other.

CTD and 3′-end processing

Pre-mRNA 3′-end processing also shows strong interdependence with the pol II CTD. The vast majority of mRNAs contain a non-templated poly(A) tail. In metazoans, polyadenylation involves several trans-acting protein complexes and corresponding cis-acting RNA elements (for a review, see refs. 19 and 20) Initially, the polyadenylation signal AAUAAA and its downstream GU-rich element are recognized by the cleavage/polyadenylation-specificity factor (CPSF) and cleavage stimulation factor (CstF), respectively. The CPSF-73 subunit of CPSF is responsible for cleaving the mRNA at the CA dinucleotide located ~10–30 nucleotides downstream of the polyadenylation signal prior to polyadenylation catalyzed by poly(A) polymerase.²¹ During transcription initiation, CPSF is brought to the pre-initiation complex by TFIID, and it remains associated with pol II during elongation (Fig. 1C).²² In mammalian cells, RNAs transcribed by CTD-truncated pol II are not efficiently polyadenylated, indicating the importance of CTD in polyadenylation.²³ In addition, the finding that RNA pol II containing Ser2-phosphorylated CTD predominates at the 3′ end of genes during transcription suggests that Ser2 phosphorylation plays a regulatory role in 3′-end pre-mRNA processing.²⁴ Coincidently, deletion of the yeast Ser2 kinase Ctk1 impairs polyadenylation without obvious effects on transcription elongation.²⁵ However, unlike capping enzymes, most of the 3′-end processing factors have no strong preference for specific phospho-isoforms of the CTD.²⁵

For certain histones required for DNA replication, their pre-mRNAs undergo a unique endonucleolytic cleavage reaction at the 3′ end instead of polyadenylation. This cleavage requires the stem-loop binding protein SLBP and the U7 small nuclear RNP (snRNP) that bind to corresponding cis-elements on histone pre-mRNAs.¹⁹ A recent study showed that CDK9-mediated phosphorylation of CTD Thr4 may facilitate the recruitment of SLBP and CPSF-100, a factor common to both histone mRNA processing and general mRNA polyadenylation, to promote histone mRNA maturation.²⁶ Nevertheless, whether this CTD phosphorylation is modulated by the cell cycle remains an interesting question.

CTD and transcription termination

A large amount of evidence from yeast studies has indicated that the interaction between the pol II CTD and the polyadenylation machinery is important for transcription termination. The Pcf11 subunit of the yeast cleavage/polyadenylation factor binds to the CTD phosphorylated on Ser2 via its CTD-interacting domain.^1,27 Notably, a recent report shows that Tyr1 phosphorylation of the pol II CTD excludes Pcf11 during elongation and thereby suppresses premature termination.²⁸ However, the decline of Tyr1 phosphroylation prior to the polyadenylation site enables phsopho-Ser2 CTD to recruit Pcf11 and other termination factors. Pcf11 mediates transcription termination through its cooperation with another CTD-interacting domain-containing factor, Rtt103 and the 5′–3′ exonuclease Rat1, which facilitates release of the transcription elongation complex from the transcript.^29,30 Another CTD-interacting domain-containing RNA-binding protein, Nrd1, also contributes to transcription termination.³¹ In contrast to Pcf11 and Rtt103, Nrd1 preferentially interacts with the CTD phosphorylated on Ser5; this interaction requires the cis conformation of the upstream phospho-Ser5-Pro6 peptidyl-prolyl bond of the CTD.³² Nrd1 forms a complex with Nab3 and Sen1 to participate in transcription termination and 3′-end processing of certain non-coding RNAs (Fig. 1D).³³ The Nrd1/Nab3/Sen1 complex initially associates with pol II at the promoter region where phospho-Ser5 CTD is predominant and then moves along the gene with pol II toward the 3′ end. A recent report showed that disruption of the Sen1-Ser2 CTD interaction prevents transcription termination.³⁴ Therefore, the Nrd1/Nab3/Sen1 complex may be handed over from phospho-Ser5 to phospho-Ser2 via a sequential interaction of Nrd1 and Sen1 with the CTD. Interestingly, the Nrd1-CTD interaction is also impaired by phospho-Tyr1, suggesting that different termination decisions may depend on phospho-Tyr1.²⁸ Taken together, all evidence has suggested that sequential phosphorylation of the CTD is likely important for proper processing of pre-mRNAs as well as transcription elongation and termination.

CTD and pre-mRNA splicing

Pre-mRNA splicing is catalyzed by the spliceosome, a large RNP complex composed of five spliceosomal small nuclear RNAs and nearly 200 proteins, including members of the SR protein family.³⁵ SR proteins typically contain RNA-binding motifs for cis-element recognition and one or more arginine-serine dipeptide-rich (RS) domains essentially for protein-protein interactions.³⁶ Yeast two-hybrid experiments have unveiled interactions between the pol II CTD and SR proteins, suggesting that pol II recruits SR proteins to nascent transcripts.³⁷ Accordingly, subnuclear localization of SR proteins is profoundly influenced by pol II activity (for a review, see ref. 38), which suggests that splicing is coupled to transcription in cells.^38,39 Moreover, mass spectrometry analysis of pol II-associating proteins identified a number of SR proteins as well as the U1 snRNP.⁴⁰ Recently, an in vitro functional assay revealed that RNA pol II also interacts with the splicing factor U2AF65 and the multiprotein PRP19 complex via its CTD to promote splicing (Fig. 1E).⁴¹ The PRP19 complex joins the spliceosome only at the late stage of splicing and activates the catalytic reaction, suggesting that the CTD may also help to coordinate spliceosome activation.⁴¹ Collectively, these results suggest a recruitment model in which pol II affects pre-mRNA splicing via the interaction of the CTD with splicing factors.⁴² In fact, the impact of transcription elongation on pre-mRNA splicing has been unveiled for a decade. An early report showed that a mutant Drosophila RNA pol II with decreased elongation activity influences alternative splicing of reporter transcripts.⁴³ More lines of evidence then indicated that the RNA pol II processivity affects splicing, suggesting kinetic coupling between transcription and splicing (for a review, see refs. 44 and 45). Recent high-throughput analysis provides a comprehensive view on the impact of pol II elongation on splicing. Reduced elongation rate causes higher levels of pol II occupancy in introns.⁴⁶ Moreover, pol II often accumulates or stalls in the introns that have suboptimal downstream splice sites, and thereby causes regulated exon inclusion.⁴⁶ Therefore, it is assumed that a reduced elongation rate may provide ample time for splicing factors to recognize weak splice signals.

Physiological significance of transcription rate-modulated splicing regulation has also been revealed. It is noteworthy that pol II elongation-dependent splicing regulation is particularly involved in the expression of a set of RNA processing factors in response to changes of growth conditions, which then downregulates those factors via nonsense-mediated mRNA decay (NMD; see below) or causes their exon inclusion.⁴⁶ Conceivably, such a splicing change has a profound effect on the transcriptome. A more detailed mechanism has been shown on the alternative splicing of proapoptotic Bcl-x and caspase 9 caused by UV-induced DNA damage, which acts via hyperphosphorylation of both Ser2 and Ser5 of RNA pol II and, thus, inhibits pol II elongation.⁴⁷ A more recent report shows that a pausing site embedded within the exonic cis-regulatory element of Bcl-x slows down the elongating pol II and, hence, a putative repressive complex forms. Overexpression of the elongation regulator 1 TCERG1/CA150, which accelerates pol II elongation, reactivates the suppressed 5′ splice site of Bcl-x.⁴⁸ Another recent report showing that a mRNP assembled on a large set of regulated exons embedded in A/T-rich DNA can interact with pol II and modulates transcription elongation, also indicates the impact of chromatin-induced transcription pausing on alternative splicing.⁴⁹

So far, we discuss how transcription elongation affects splicing. However, certain aspects of pre-mRNA splicing may influence the rate of transcription elongation, especially on the activity of P-TEFb. One report showed that depletion of the SR protein SRSF2/SC35 impairs P-TEFb recruitment to the pol II elongating complex and subsequent phosphorylation of CTD-Ser2 and, thus, attenuates elongation.⁵⁰ Another finding showed that the 7SK snRNP, which inhibits P-TEFb kinase activity, modulates not only transcription elongation but also alternative splicing.⁵¹ Therefore, these observations infer that transcription elongation and alternative splicing are coupled, and that regulation of one process affects the rate of the other.

Protein domains that mediate the link between transcription and splicing

The interconnection between the transcription machinery, essentially RNA pol II, and different pre-mRNA processing factors, contributes substantially to the coupling of transcription and splicing events. For example, the multifunctional protein p54^nrb forms a heterodimer with the splicing factor PSF to interact with the pol II CTD, and such an interaction has been implicated in early spliceosome assembly during transcription.⁵² In addition, p54^nrb interacts specifically with the transcription factor Sox9 and modulates both the promoter activity and alternative splicing of Sox9 targets.⁵³ Thus, proteins with dual roles in gene-specific transcription and pre-mRNA processing may function in coordinating gene expression. Three domains, i.e. the RS as well as the WW and FF domains (see below), are most representative in splicing factors that associate with pol II. In this section, we discuss the common domains of splicing factors implicated in linking transcription with pre-mRNA splicing.

RS domain

The involvement of the RS domain in pre-mRNA splicing is firmly established (for a review, see ref. 54 and 55). The canonical SR splicing factors contain a typical RS domain that consists of successive RS dipeptides. In addition, SR-like splicing factors, such as U1-70K and SRm160, contain degenerate RS domains, in which dispersed RS or arginine-aspartate/glutamate dipeptides exist. Differential phosphorylation of the RS domain by the SRPK and Clk/Sty families of kinases modulates the cellular localization and activity of SR proteins.⁵⁴ Moreover, in vitro studies have demonstrated that dephosphorylation of SR proteins is also important for splicing.⁵⁶ The spliceosome-associated protein phosphatase 2Cγ and the activity of protein phosphatase 1/2A are required for early spliceosome assembly and the catalytic reaction of splicing, respectively,^57,58 but whether these phosphatases are specific to SR splicing factors remains to be clarified. In addition to splicing, a set of SR proteins also functions in mRNA export; these proteins include the yeast SR-like Npl3 protein and several mammalian shuttling SR proteins.^59,60 Nuclear export of Npl3 is facilitated by arginine methylation of its atypical RS domain by Hmt1 and dephosphorylation by the phosphatase Glc7.^61,62 Analogously, mammalian SR shuttling proteins also undergo dephosphorylation during export, and dephosphorylated SR proteins are engaged in mRNA translation.^63-65 These observations further imply that dynamic phosphorylation of SR proteins, from transcription through translation, governs their activities in controlling gene expression.

Physical interactions between several prototypical SR proteins and phosphorylated pol II CTD support the function of SR proteins in coordinating pre-mRNA splicing and transcription as discussed above.⁴⁰ In general, SR proteins facilitate exon definition and intron bridging during spliceosome assembly, essentially through RS domain-mediated protein-protein interactions.⁵⁴ Many SR proteins also participate in alternative splice site choice by specific recognition of regulatory cis-elements, namely exonic or intronic splicing enhancers or silencers.⁵⁵ Interestingly, in vitro studies have shown that the CTD can substitute for the RS domain of an SR protein to promote early spliceosome assembly, further indicating the significance of recruiting SR proteins to pol II during transcription elongation for subsequent splicing activation.⁶⁶

RS domains were originally thought to bind RNA non-specifically due to their high content of positively charged residues. Experimental evidence has then indicated that the RS domain can promote or stabilize base-pairing of U1 and U2 snRNAs with the 5′ splice site and branch site of introns perhaps through its direct contact with short snRNA-pre-mRNA helices.^67,68 Therefore, SR proteins that are recruited by the RNA pol II CTD may facilitate early spliceosome assembly not only through interaction with proteins but also with RNAs.

WW and FF domain

The WW and FF domains, named for their two spaced signature tryptophan and phenylalanine residues, respectively, also represent protein interaction modules.^69,70 The WW domain is composed of ~40 amino acid residues and folds as a triple-stranded sheet that binds to proline-rich peptides.⁷⁰ WW domain subgroups are distinguished by their similar but slightly distinct ligands. For example, group I and II recognize the PPxY (x denotes variable residues) and PPLP motifs, respectively. The WW domain exists in a large variety of proteins, including several transcriptional co-activators and spliceosomal components.⁷⁰ The FF domain often co-exists with the WW domain in those proteins. The FF domain contains ~50–60 residues and folds into a three-α-helix structure. The first α-helix contains a highly conserved FxxL motif, which can be specifically recognized by nuclear receptors, suggesting that FF domain-containing proteins may be responsible for ligand-dependent transcriptional activation.⁶⁹

Several proteins containing a WW or FF domain have been found to link transcription and splicing by their interaction with pol II and splicing factors. The yeast WW domain-containing splicing factor Prp40 interacts directly with the phosphorylated pol II CTD via its WW domain.⁷¹ Prp40 associates with the U1 snRNP and interacts with the U5-associated protein Prp8 and the branch point-binding protein (yeast BBP; human SF1).^72,73 Thus, Prp40 may mediate intron bridging early during spliceosome assembly. It is notable that the human orthologs of Prp40, i.e., FBP11 and FBP21, also interact with the phosphorylated CTD and associate with the 17S U2 snRNP.⁷⁴ Importantly, both the Prp40-BBP and FBP11/21-SF1 interactions engage their respective WW domain and proline-rich motif, indicating that the interactions among these splicing factors are evolutionarily conserved.^73,75,76 Moreover, early reports have shown that the WW and FF domain-containing TCERG1 interacts with the phosphorylated CTD and the splicing factors that associate with the 3′ end of the intron and that TCERG1 can activate splicing via its WW and FF domains.^77,78 More recent evidence indicates that TCERG1 regulates alternative splicing via modulating pol II transcription rate⁴⁸ (see above) and, thus, supports previous findings. Therefore, yeast WW-containing proteins may promote early spliceosome formation, whereas mammalian analogs have an additional function in alternative splicing.

Co-transcriptional assembly of mRNA-containing ribonucleoproteins

RNA binding proteins can be loaded onto a nascent transcript when it just begins to emerge from RNA pol II but function in following nuclear RNA processing steps. Some of the proteins may remain associated with mature mRNAs and participate in post-splicing events such as mRNA export, stability control and translation. We discuss two co-transcriptionally loaded protein complexes, namely the transcription/export (TREX) complex and exon-junction complex (EJC), which escort mRNAs to the cytoplasm and how these complexes functionally connect different steps of RNA biogenesis.

The TREX complex links transcription and mRNA export

The evolutionarily conserved TREX complex is composed of the THO complex, which functions in transcription elongation, and the mRNA export factors UAP56 (Sub2 in yeast) and Aly/REF (Yra1 in yeast). The yeast THO complex contains four subunits (Tho2, Hpr1, Mft1 and Thp2), whereas the mammalian complex contains only two yeast-equivalent THO components (Tho2 and Hpr1) and three other proteins.⁷⁹ Although a recent study showed that the PRP19 splicing complex could recruit the TREX complex to actively transcribed genes,⁸⁰ it is generally believed that yeast THO is recruited to activated genes by RNA pol II.⁸¹ Chromatin immunoprecipitation has revealed that THO shows biased localization toward the 3′ end of genes and interacts with the 3′-end processing factor Pcf11.^82,83 These results suggest that THO also participates in 3′-end processing and/or transcription termination.⁸⁴ Pcf11 recruits Yra1 to THO. The mutually exclusive interaction of Yra1 with Pcf11 and Sub2 suggests the handover of THO from the 3′-end processing complex to export-competent mRNPs.⁸⁴ Once 3′-end processing is completed, TREX is released from the transcription site along with its associated mRNA. Unlike yeast Yra1 and Sub2, which associate with mRNPs independent of splicing, mammalian UAP56 has a role in promoting spliceosome assembly during pre-mRNA splicing (for a review, see ref. 85) and Aly/REF is loaded onto mature mRNAs once splicing is completed.^86,87 Nevertheless, UAP56, in conjunction with Aly/REF, can facilitate nuclear export of both spliced and unspliced mRNAs by recruiting the mRNA export receptor, i.e., the TAP/p15 heterodimer (yeast Mex67/Mtr2).^86,88 Alternatively, mammalian TREX can be recruited to the 5′ end of mRNAs by the CBP80 subunit of the cap binding complex and may, thus, promote the export of intronless mRNAs.⁸⁹ Thus, TREX may promote efficient mRNA synthesis and ensure that correctly formed mRNPs are exported.

TREX-2 is another complex implicated in linking transcription to mRNA export and essentially functions via its association with the nuclear pore.⁹⁰ The TREX-2 complex consists of Sac3, Thp1, Cdc31 and the Spt-Ada-Gcn5 acetyltransferase (SAGA) complex.^91,92 Sac3 interacts with Mex67/Mtr2 and a component of the nuclear pore complex, Nup1, and facilitates docking of mRNPs to the nuclear pore.^93,94 Moreover, the yeast SAGA complex, via its intrinsic component Sus1, localizes actively transcribing genes close to the nuclear pore and may also help mRNA export. Furthermore, the role of the SAGA complex and Thp1 in transcription elongation suggests that TREX-2, analogous to TREX, may define a pathway connecting transcription elongation with mRNA export.⁹⁰ The Drosophila homolog of Sus1, ENY2, is also co-transcriptionally recruited to nascent mRNAs, indicating the evolutionarily conserved role of Sus1/ENY2 in mRNA export.⁹⁵ However, the association of ENY2 with components of THO instead of TREX-2 suggests that Sus1/ENY2 may participate in mRNA export, in conjunction with similar but somewhat different complexes, in higher eukaryotic cells.⁹⁶ Nevertheless, as compared with yeast, the detailed function of TREX and related complexes in higher eukaryotes is less clear and requires much more investigation.

The EJC complex links splicing with mRNA surveillance and translation

The EJC is deposited ~20 nucleotides upstream of ligated exon junctions of spliced mRNAs,⁹⁷ and influences several downstream mRNA metabolic events, including mRNA export, surveillance and translation (for a review, see refs. 98–101). A recent comprehensive EJC protein-RNA interaction analysis has indicated that the EJC occupies the majority of exon junctions of mRNAs in cells.¹⁰² The EJC comprises a four-subunit core (Y14, Magoh, eIF4AIII and MLN51) that is assembled prior to exon ligation.¹⁰⁰ Assembly of the EJC is indeed splicing-dependent and requires an intron-binding protein, IBP160.¹⁰³ Besides, a number of peripheral components of the EJC join the core after splicing. The core remains on the mRNA along the export pathway until translation commences.¹⁰⁴ The observation that defective splicing or 3′-end processing results in accumulation of EJC factors at transcription sites suggests that formation of the EJC occurs co-transcriptionally.¹⁰⁵

The EJC interacts with mRNA export factors UAP56, Aly/REF and TAP/p15.^99,106 However, the observation that depletion of EJC factors does not significantly impair bulk mRNA export suggests that the EJC may be dispensable for efficient mRNA export.^107,108 On the other hand, the EJC has been implicated in NMD, which governs quality control of mRNAs that have been newly exported and bound by the nuclear cap-binding complex CBP20/80.¹⁰⁹ NMD destroys transcripts containing a premature termination codon (PTC) during the initial round of translation and thus prevents expression of truncated proteins.^99,101 The EJC recruits the essential NMD factor Upf3 in the nucleus and subsequently Upf2 in the cytoplasm.¹¹⁰ A translating ribosome is stalled when it encounters a PTC and subsequently recruits additional NMD factors, including Upf1, to trigger the decay of PTC-containing transcripts.^111,112 Knockdown of individual EJC factors inhibits NMD, indicating the importance of the EJC in NMD.¹¹³

The EJC participates in NMD during the pioneer round of translation.¹⁰⁰ An activated mTOR kinase, ribosomal protein S6 kinase 1 (S6K1), can be recruited by the EJC-associated RNA-binding protein SKAR to CBP20/80-bound mRNPs, suggesting that the EJC can mediate cellular signaling pathways to regulate the pioneer-round translation and hence NMD.¹¹⁴ Perhaps EJC-recruited S6K1 may further promote subsequent rounds of translation to increase protein yield particularly under favorable cellular conditions. Indeed, a tethering experiment has already indicated that some of the EJC components, such as the core factor Y14 and peripheral factor RNPS1, can enhance productive translation by promoting mRNA polysome association.¹¹⁵ The Drosophila EJC factors are particularly involved in localization and/or regulated translation of some maternal mRNAs in the developing oocyte, indicating an evolutionarily conserved role of the EJC in translational control.¹¹⁶

In addition to interacting with factors that function in mRNA export or NMD, the EJC multimerizes with itself and forms large complexes with SR proteins to promote mRNA packaging and protect mRNAs from degradation; therefore, the EJC provides a function in constitutive mRNP remodeling during mRNA maturation.¹⁰² Moreover, recent evidence also indicates that individual EJC factors may have specific mRNA targets or functions. For example, eIF4AIII selectively binds to a set of dendritic mRNAs and controls their expression perhaps via NMD and, thus, modulates synaptic activities.¹¹⁷ Interestingly, eIF4AIII interacts with an eIF4G-like protein, NOM1, which mimics the interaction between the translation initiation factors eIF4A and eIF4G, and the eIF4AIII/NOM1 participates in rRNA processing.¹¹⁸ We have recently reported that Y14 itself can inhibit the dcapping activity of Dcp2 and prevents mRNA decay.¹¹⁹ Therefore, EJC or NMD factors may exert different activities and provide a wide range of cellular functions.

Together, nuclear RNA processing, particularly pre-mRNA splicing, imprints mRNAs in the nucleus with a specific set of RNA binding proteins, which influences the fate of mRNAs once they reach the cytoplasm.

Beyond transcription: Epigenetic control of splicing

It has long been known that chromatin structure has a tremendous impact on transcriptional regulation.¹²⁰ Recently, evidence of epigenetic effects on pre-mRNA splicing has begun to emerge. In fact, an early finding that the Gcn5-related histone acetyltransferase interacts with the U2 snRNP has suggested a possible role for chromatin structure in splicing (Fig. 2A).¹²¹ Nevertheless, recent analyses of data on large-scale chromatin precipitation and high-throughput sequencing have largely unveiled the close relationship between chromatin signatures and various aspects of splicing (for a review, see refs. 122–124). In this section, we discuss emerging evidence and underlying mechanisms of the effects of epigenetic modifications on pre-mRNA splicing.

Figure 2. Chromatin modifications affect alternative splicing via adaptors. (A) Via respective adaptors, different types of histone modification (methylation and acetylation) and DNA methylation exert their effects on the regulation of alternative splicing. CpG methylation antagonizes the activity of CTCF in CD45 exon 5 inclusion. (B) H3K36me3-modified histones recruit the adaptor MRG15 to regulate PTBP1-mediated exon selection of the FGFR2 transcript.

Nucleosome positioning and exon-definition

Histones undergo extensive post-translational modifications, and different types of modifications even have combinatory effects on transcriptional regulation. Studies using chromatin precipitation coupled with deep sequencing have mapped specifically modified histones over the entire genome.^122,124 The first identified histone signature enriched in exons is histone H3 trimethylation at lysine 36 (H3K36me3). H3K36me3 exhibits a biased distribution toward downstream exons rather than the promoter region and is less prominent in alternatively spliced exons than the constitutive ones.¹²⁵ The H3K36me3 signature is established by splicing-dependent recruitment of the methyltransferase HYPB/Setd2 to the elongating pol II,^126,127 indicating that certain types of histone modification modulate pol II processivity and hence influence RNA processing. Moreover, the exon-definition model has also been suggested for the role of nucleosome positioning in pre-mRNA splicing.^45,128 Two independent reports showed that exons, with an average size of 140~150 base-pairs and a higher GC content, are coincidently occupied by the nucleosome that wraps 147-bp long DNA.^129,130 The transcription-independent but stably positioned nucleosomes are enriched in exons with weak splice sites but excluded from the immediate upstream/downstream intron regions and the pseudoexons flanked by functional splice sites. Such a chromatin-splice site configuration is indeed evolutionarily conserved.¹³⁰ Finally, an extensive mapping of 38 different histone modifications revealed a set of histone modification signatures specific to exons, which also correlates with exon selection.¹³¹ These observations underscore the impact of the exon-positioned nucleosome as well as the global histone methylation signature on pre-mRNA splicing and exon utilization.

Histone methylation: Kinetic coupling vs. recruitment

It is conceivable that the nucleosome landscape affects chromatin accessibility and hence transcription processivity, which, in turn, modulates splicing.¹²³ For example, transcriptional gene silencing (TGS) can be achieved by small interfering RNA (siRNA)-mediated heterochromatin formation. Transfection of siRNAs against the downstream region of the alternatively selected extra domain I (EDI) exon of fibronectin induces locally assembled heterochromatin, which inhibits pol II elongation and hence causes EDI inclusion.¹³² Specific modification of nucleosomes facilitates exon inclusion possibly also through a similar mechanism.¹³³ An example is given by the observation that the H3K9me3-binding transcriptional repressor HP1γ slows down the elongation rate on the variant exons of CD44 and, accordingly, causes exon inclusion (Fig. 2A). In general, this kinetic coupling mechanism involves chromatin adaptors that read histone marks and modulate pol II elongation rate. Moreover, readers of histone modifications can interact with the splicing factors that bind to the pre-mRNA to influence splicing. This is evidenced by the mutually exclusive selection of exon IIIb or IIIc of the fibroblast growth factor receptor 2 (FGFR2) transcripts (Fig. 2B).¹³⁴ Profiling of histone modifications over the FGFR2 gene revealed a tight correlation between exon inclusion and histone signatures.¹³⁵ When the chromatin encompassing the FGFR2 gene contains abundant H3K36me3, the exon IIIc-containing isoforms are predominantly expressed. It has been previously shown that PTBP1 suppresses exon IIIb inclusion via binding to the silencing elements around exon IIIb, which leads to a concomitant increase of the exon IIIc-containing mRNA product.¹³⁴ Present data indicate that the chromatin-associated protein MRG15 specifically recognizes H3K36me3 and recruits PTBP1 to prevent exon IIIb inclusion.¹³⁵ Moreover, it has also been reported that H3K36me3 and H3K4me3 can respectively interact with the SRSF1-interacting protein Psip1/p52 and the chromatin remodeling protein CHD1 to further recruit corresponding splicing factors to target transcripts (Fig. 2A).^136,137 A growing body of evidence supports the recruitment model, in which modified histones can recruit splicing factors through chromatin adaptors.

Histone acetylation

Besides histone methylation, altered acetylation status of histones and nucleosome structure can also influence splicing patterns. The histone deacetylase inhibitor, Trichostatin A (TSA), could trigger exon skipping in a minigene system.¹³⁸ A screen using splicing-sensitive microarrays revealed that alternative splicing of as many as 700 transcripts was perturbed upon inhibition of histone deacetylases.¹³⁹ Experimental evidence then indicated that inhibition of histone deacetylation prevents alternative exon inclusion of fibronectin likely by increasing RNA pol II processivity as well as reducing SRSF5/SRp40 binding to the alternatively spliced exon.¹³⁹ Moreover, the chromatin remodeling complex SWI/SNF modulates alternative exon selection via the activity of its catalytic subunit Brg1/Brm that reduces the rate of pol II elongation and thereby facilitates spliceosome assembly.¹⁴⁰ The physiological significance of histone acetylation on splicing has been revealed by the finding that TSA treatment mimics neuronal excitation as stimulated by KCl and induces exon 18 skipping of the neural cell adhesion molecule (NCAM) in murine neuroblastoma cells, which correlates with an increased abundance of H3K9 acetylation around exon 18.¹⁴¹ Therefore, changes in nucleosome modification or structure can exert their effect on splicing via similar mechanisms.

DNA modification

In addition to histone modifications, DNA methylation also impacts pre-mRNA splicing. The finding that the methyl-CpG-binding protein MeCP2 can regulate pre-mRNA splicing has initially suggested the potential effect of CpG distribution and cytosine methylation on alternative splicing.¹⁴² A recent bisulfite sequencing of CpG island-enriched genomic regions revealed that DNA methylation is abundant in exonic regions and correlates well with H3K36me3 levels, providing a global picture of DNA methylation and even its combinatory effect with histone methylation in exon selection.¹⁴³ Another study showed that the methylation-sensitive DNA binding transcriptional regulator CTCF facilitates the inclusion of CD45 exon 5 by inducing RNA pol II pausing at this exon and, interestingly, that CpG methylation antagonizes the effect of CTCF on exon 5 inclusion (Fig. 2A).¹⁴⁴ This result provides evidence that DNA methylation influences splicing. Therefore, it is conceivable that changes in DNA methylation patterns during development or in disease, particularly in cancer, could determine alternative splicing outcome and subsequently affect transcriptome makeup.

Splicing influences chromatin modifications

Finally, the reciprocal influence of splicing on chromatin modifications has also emerged recently. A genome-wide analysis revealed that the exon-dominant positioning of H3K36me3 is enriched in intron-containing genes relative to intronless genes.¹⁴⁵ Further analysis showed that splicing inhibition impairs the recruitment of the H3K36 methyltransferase HYPB/Setd2 and then reduces the level of H3K36me3 as well as repositions H3K36me3 over the gene.^127,145 Another evidence indicates the reaching-back mechanism mediated by the splicing regulator Hu protein. In neuronal cells, Hu targets the pre-mRNA along with the elongating pol II and subsequently induces local histone acetylation surrounding the regulated exons via inhibiting the histone deacetylase, HDAC2. This may result in increased elongation rate of subsequent rounds of transcription, which, thus, promotes exon skipping.¹⁴⁶ A very recent report reveals the influence of the first exon length in promoter-proximal histone modifications and transcription output.¹⁴⁷ Short first exons appear to increase the levels of H3K4me3 and acetylated H3K9 at promoters and, accordingly, increase the accuracy and efficiency of transcription initiation. More importantly, intron deletion reduces H3K4me3 levels at promoters, indicating splicing-dependent chromatin modification.

It has been evident that histone modifications provide another layer of alternative splicing modulation in a tissue/development-specific manner. Evidence also begins to emerge for the influence of splicing on histone modification profiles along the gene. Perhaps splicing can feedback control transcription through histone modifications.

Conclusion

We have reviewed studies of the past decade that unveil the molecular mechanisms of nuclear pre-mRNA processing and their functional interconnection with transcription. We have also discussed new results obtained by using global sequencing approaches that provide a genome-wide view of the function of individual trans-acting RNA processing factors. Recently, advanced crosslinking-immunoprecipitation and sequencing techniques enable fine-mapping of in vivo interactions between trans-factors and mRNAs even at single-nucleotide resolution.^148,149 We may thus gain insights into how trans-factors and mRNAs constitute a highly dynamic ribonucleoproteins, and how RNA regulons are possibly assembled during mRNA processing and act in response to cellular signaling. Moreover, recent studies uncover a previously unrecognized role for chromatin modifications on pre-mRNA processing, particularly splicing regulation. We must further explore how epigenetic signals may coordinately control both transcription and alternative splicing through respective regulatory factors, and how the nucleosome modifying factors act hierarchically to optimize gene expression outcome. A further question is how environmental or heritable cues control alternative splicing by changing chromatin signatures. Finally, it is of fundamental importance to understand how the fidelity of each RNA biogenesis step is maintained and whether checkpoint mechanisms exist to control the progression from transcription to subsequent steps of RNA processing. The advent of techniques and bioinformatics would certainly unveil more comprehensive regulatory mechanisms of post-transcriptional regulation and elucidate the regulation of RNA metabolism in the context of normal physiology and disease.