Where splicing joins chromatin

Abstract

There are numerous data suggesting that two key steps in gene expression—transcription and splicing influence each other closely. For a long time it was known that chromatin modifications regulate transcription, but only recently it was shown that chromatin and histone modifications play a significant role in pre-mRNA splicing. Here we summarize interactions between splicing machinery and chromatin and discuss their potential functional significance. We focus mainly on histone acetylation and methylation and potential mechanisms of their role in splicing. It seems that whereas histone acetylation acts mainly by alterating the transcription rate, histone methylation can also influence splicing directly by recruiting various splicing components.

Pre-mRNA Splicing Regulation

The vast majority of our genes has a puzzle-like structure where short coding sequences (exons) must be correctly identified and joined together while surrounding much longer non-coding sequences (introns) have to be removed before a mature mRNA is exported to the cytoplasm and translated. A typical human gene contains 8 introns of average length ∼3.4 kb but much larger examples were found (e.g., the last intron of glypican 5 gene is 721 kb in length).¹ In contrast to long introns, exons are short having an average length of 145 bp and represent only a small fraction of primary pre-mRNA transcripts. An extreme example is the largest human gene CNTNAP2 encoding Caspr2 protein which spans over 2.3 Mb (!) but its spliced mRNA is only 10 kb long because 99.6% of the CNTNAP2 gene sequence corresponds to introns.²

To further complicate matters, exons and introns are not always recognized identically and, for example, a particular exon can be skipped in a fraction of mature mRNAs. This process, called alternative splicing, produces different mRNA isoforms from one gene which significantly increases the coding potential of our genome. It is estimated that in human cells almost 95% of genes are alternatively spliced^3,4 and that there are on average 7 different alternative splicing events per single gene leading to different mRNA variants.⁴ The complexity of the process puts pressure on cells to regulate splicing precisely. Then how is alternative splicing regulated?

Introns contain consensus splice sites at both ends. However, compared to yeasts, most splice site sequences in higher metazoans are degenerate and additional regulatory sequences in pre-mRNA are needed to help basal splicing machinery (the spliceosome) to recognize correct splice sites.⁵ These regulatory sequences are bound by many different splicing regulatory proteins which are able to interact with the splicing complex. The combinatorial interplay among splicing factors results in the usage (or skipping) of individual splice sites. Some of the splicing regulatory proteins are widely expressed in different tissues (such as PTB or SR proteins) while the others are highly tissue specific (for example, Nova proteins are expressed almost exclusively in neurons). The differences in the expression of splicing regulatory proteins are believed to be responsible for tissue-specific alternative splicing.⁶

The splicing code—the sum of all splicing related features in a pre-mRNA sequence—can explain a majority of differences in alternative splicing between individual tissues (e.g., 74% out of 97 tested alternative splicing events specific for the central nervous system and muscle tissue were properly predicted based on a pre-mRNA sequence).⁷ Although the splicing code model was effective to estimate whether the alternative exon was included or skipped it was less precise in the prediction of the level of inclusion/exclusion when tested by Barash et al. which indicates a presence of additional regulatory signals.⁷

Splicing is Cotranscriptional and Occurs in the Vicinity of Chromatin

Intron recognition and removal likely occur in the cell nucleus very shortly after intron transcription. Splicing complexes associate with pre-mRNA immediately after the target sequences are synthesized and splicing of many introns is completed before pre-mRNA transcription termination.^8–15 This was shown not only for long human genes but also for many yeast genes, which are much shorter than human genes and are spliced cotranscriptionally as well.^16–21 Moreover, splicing can induce pausing of RNA polymerase II during transcription²¹ and RNA polymerase II was shown to pause in terminal exons, which increases the time window for cotranscriptional splicing.¹⁶

This suggests that transcription and splicing are coupled not only in time, but also functionally. It was shown that the promoter composition and the speed of transcription can influence splicing.^22–33 The speed of transcription affects the dynamics of presentation of individual splice and regulatory sites to the splicing machinery, which then results in using different splice sites not only in transfected minigenes but also in endogenous genes.^34–43 The carboxyterminal domain of RNA polymerase II itself can also contribute to splicing factors recruitment which then might influence alternative splicing.^44,45

Average eukaryotic RNA polymerase II elongation rate vary from ∼1.5–4.5 kb/min,^8,14,46–49 which means that an average intron is transcribed in 1–3 minutes. An average splicing event takes 30 seconds to accomplish^19,20,50,51 and therefore it is very likely that recognition of splice sites as well as intron excision occur when pre-mRNA is still in close proximity to the chromatin of its own gene, and this makes it possible for chromatin modifications to influence splicing.

Chromatin Marks Affecting Splicing

In cells, DNA is packed together with histones in nucleosomes. Each nucleosome consists of an octamer composed of four core histones (H2A, H2B, H3 and H4) and 147 bp DNA, which is wrapped around this octamer. Histones can be posttranslationally modified on many sites, especially at their N-terminal ends that protrude out of the nucleosomal core. Regulation of gene activity via nucleosome positioning and histone modifications has been well established.^52,53 However, it seems that nucleosome positions and modifications are also important for exon recognition. In the genome, nucleosomes are preferentially found at exons and exon-positioned nucleosomes carry a specific set of histone modifications (see Fig. 1),^54–62 although the latter statement is still a matter of a debate.^63,64 Three histone marks were shown to affect splicing and/or to mediate interaction of histones with splicing factors: methylation, acetylation and phosphorylation (see Table 1). These modifications differ in the charge they introduce into the chromatin. While methylations are electrostatically neutral and their action is mediated via specific protein readers, it is supposed that histone acetylation, which adds a negative charge, can itself affect nucleosomal stability and make chromatin more accessible (reviewed in ref. 61). Histone phosphorylation, which brings a negative charge, can regulate the binding of SR proteins to the chromatin during the cell cycle but its role in splicing regulation is unclear.⁶⁵ The modifications also differ substantially in their turnover: the addition and removal of acetylation is rapid and most acetyl residues are exchanged within minutes, phosphorylations can last up to several hours and methylations, the most stable ones, can be present even for days.⁶⁶

Figure 1.

Different sets of chromatin modifications were identified at exons and at introns. The enrichment of a specific histone mark depends on the position of an exon within a gene (e.g., H2BK5me1 and H4K20me1 are found preferentially at exons toward the 5′ end of genes),⁶¹ or it can depend on gene expression (highlighted in bold), e.g., H3K27me3 is high at exons of low expressed genes, but it was reported to be elevated at introns of highly expressed genes.⁵⁸

Table 1.

Interaction of chromatin and splicing machinery

Histone modification	>Interacting protein	>Link to splicing
H3K4me3	Chd1	Chd1 associates with SRp20 (SRSF3)119 and U2 snRNP (via SF3 subunits) and increases efficiency of pre-mRNA splicing86
H3K4me3	Sgf29	Sgf29 interacts with SF3B5 (SF3b10) and SF3B3 (SF3b130) subunits of U2 snRNP87
H3K9me3	PTB, hnRNP A1, hnRNP A/B, hnRNP A2/B1, hnRNP K, hnRNP L	PTB and most of the hnRNP proteins are direct regulators of alternative splicing but it is not known whether the association of hnRNP proteins with H3K9me387 affects splicing
methylated H3K9	HP1 (HP1a)	HP1 binds to Drosophila hnRNP proteins (PEP, DDP1, HRB87F)114
H3K36me3	MRG15	MRG15 recruits PTB; tethering of PTB to chromatin changes alternative splicing88
methylated H3K79	TP53BP1	TP53BP1 immunoprecipitates U1 and U2 snRNA (but also other small RNAs)120
histone H3 (not phosphorylated at serine 10)	SRp20 (SRFS3), SF2/ASF (SRSF1)	Both SR proteins participate in constitutive and alternative splicing, but the role of interaction with histone H3 in splicing is not known65
DNA methylation	MeCP2	MeCP2 regulates alternative splicing121

Histone Acetylation Induces Changes in the Alternative Splicing

Histone acetylation can be easily increased in cells by histone deacetylase (HDAC) inhibitors and the treatment with HDAC inhibitors influences an alternative splicing pattern of transiently expressed minigene splicing reporters.³⁶ Recently, we showed that HDAC inhibition changes alternative splicing of ∼700 endogenous human genes. In addition, HDAC1 depletion affected the splicing pattern of the fibronectin gene.³⁸ The correlation between histone acetylation and alternative splicing was also observed with the NCAM gene, whose alternative splicing is regulated by depolarization of neuronal membranes.⁶⁷ In this instance, it is worthy to note that many genes that changed alternative splicing after HDAC inhibition were encoding various ion channels and regulators of cell cycle or apoptosis.³⁸ Thus it seems that acetylation preferentially modulates alternative splicing of genes that must react quickly to changing conditions.

The mechanism by which acetylation modulates splicing has been connected to RNA polymerase II processivity. Increased RNA polymerase II processivity, which is induced by histone acetylation, modulates the association of splicing regulators with the nascent RNA as was shown in the case of the fibronectin gene and the splicing regulator SRp40 (SRSF5).³⁸ However, in yeast the deletion of histone deacetylases or the histone acetyltransferase Gcn5 influenced cotranscriptional recruitment of spliceosomal proteins without obvious changes in RNA polymerase II distribution along the gene.^68,69 It is then possible that histone acetylation also affects splicing directly via the recruitment of spliceosomal subunits, without altering transcription elongation.

Mutual Influence between Splicing Machinery and Chromatin

The spliceosome assembles from five small ribonucleoprotein particles (snRNPs) that assemble on the pre-mRNA in a stepwise manner.^{17,18,50,70–75} Each snRNP consists of several proteins and one small nuclear RNA (snRNA) according to which the snRNPs are named.⁷⁶ First, the U1 snRNP binds to the premRNA and recognizes the 5′ splice site. Then, the U2 snRNP binds to the vicinity of the 3′ splice site, closely followed by the U4/U6•U5 tri-snRNP. After tri-snRNP incorporation the spliceosome undergoes extensive rearrangements that result in the formation of an active complex that catalyzes both steps of the splicing reaction. The U1 and U2 snRNP binding is important for the accuracy of splicing as this defines the intron. U1 and U2 snRNPs associate with pre-mRNA quickly and they are already present when RNA polymerase transcribes only hundreds of nucleotides downstream of splice sites.^11,17,74,75 This implies that snRNPs are in close vicinity to chromatin and their association with pre-mRNA could be influenced by chromatin and vice versa, snRNPs themselves might influence the chromatin state. One example of plausible mutual influence is discussed below.

5′ Splice Site Influences Transcription and Chromatin

U1 snRNA interacts with chromatin⁷⁷ and it has been shown that the presence of functional 5′ splice site in the gene can affect chromatin state.⁷⁸ However, U1 snRNA—chromatin interaction might reflect only the role of U1 snRNA in transcription. The U1 snRNA associates with proteins from two general transcription factor complexes, TFIID and TFIIH.^77,79 In the TFIIH complex U1 snRNA enhances transcription presumably through the interaction with cyclin H and regulation of the cyclin H associated kinase CDK7,^79,80 which phosphorylates carboxyterminal domain of RNA polymerase II. It was shown that the functional 5′ splice site near the promoter enhances the assembly of general transcription factors at this promoter.⁷⁸ If the presence of U1 snRNA (and U1 snRNP) stimulates transcription then introncontaining genes should be transcribed more than intron-less genes. This is true in Saccharomyces cerevisiae where almost half of all introns found in the genome are located in highly expressed genes coding ribosomal proteins (95 out of 132 of these genes contain introns).^70,81 Moreover, deletion of introns from two nonribosomal yeast genes decreases the expression of these genes.⁸² In mammalian cells, the presence of intron with functional splice sites also increases gene expression.^78,82–85

Interestingly, the presence of a functional 5′ splice site can increase gene specific H3K9 acetylation and H3K4me3,⁷⁸ which are chromatin marks associated with active transcription. In another study, the level of H3K4me3 at the minigene decreased when both splice sites were mutated compared to the minigene containing intron with functional splice sites, however the reported difference was quite small.⁸³ It is not known whether these changes in the chromatin state are coupled generally with splicing or whether they are specifically related to U1 snRNP recruitment. Moreover, it is difficult to distinguish between the cause and the consequence in these observations. Chromatin modifications might reflect only changes in transcription and the increase of H3K4 trimethylation and histone H3 acetylation occurs independently of splicing. An example of mutual dependence is H3K4me3: the level of H3K4me3 correlates with transcription activity and is enhanced by the presence of a functional 5′ splice site within the gene but in turn this chromatin modification affects splicing (see below).

H3K4me3 Associates with U2 snRNP

Association of the U2 snRNP with chromatin is mediated by trimethylated lysine 4 histone H3.^86,87 In addition, the functional significance of H3K4me3 for splicing was shown.^86,88 Several complexes that methylate lysine 4 of histone H3 in human cells share the same structural subunits.⁸⁹ The depletion of one of these subunits, Ash2, leads to the decrease of overall H3K4me3.^90,91 Ash2 depletion causes a decrease of splicing efficiency immediately after induction of transcription⁸⁶ and Ash2 overexpression affects alternative splicing.⁸⁸

U2 snRNP binds to chromatin via the chromatin remodeling protein Chd1 that binds H3K4me3 and enhances pre-mRNA splicing efficiency within a short time window after activation of gene transcription.⁸⁶ U2 snRNP subunits also interact with another H3K4me3 reader, Sgf29 protein.⁸⁷ Considering that H3K4me3 is frequently elevated around transcription start sites,^92–97 it is surprising that this particular histone modification interacts with U2 snRNP. However, the level of H3K4me3 positively correlates with gene expression indicating that U2 snRNP might be recruited preferentially to highly expressed genes.^93,94,97

As mentioned before, U2 snRNP recruitment to the transcription unit is also regulated by the histone acetyltransferase Gcn5.⁶⁸ Although Gcn5 participates in a number of protein complexes the most important for U2 snRNP recruitment seems to be the SAGA complex.^68,98 Gcn5 in the SAGA complex acetylates predominantly histones at promoters⁶⁸ that subsequently serve as an anchor stabilizing the binding of the whole SAGA complex to the promoter.⁹⁹ Interestingly, yeast Msl1 (a homologue of mammalian U2B”, one of the U2 snRNP proteins) was also found at promoters.⁶⁸ Sgf29 and Chd1, both of the proteins interacting with U2 snRNP in human cells, are conserved in yeasts and they are components of the yeast SAGA complex.⁹⁸ It is possible that Gcn5 plays a role in the recruitment of the SAGA complex to promoters and that SAGA complex subunits are bringing the U2 snRNP. In human cells, the STAGA complex (homologue of the yeast SAGA complex) interacts with the U2 snRNP specific protein SF3B3 (SF3b130).¹⁰⁰

MRG15: Chromatin Binding Protein that Recruits Splicing Regulator PTB

PTB is an example of a splicing regulator that was shown to associate with chromatin and, more importantly, to modulate through this association alternative splicing of a number of genes.⁸⁸ PTB binds chromatin via MRG15 protein and this interaction preferentially regulates splicing of pre-mRNAs with weak PTB binding sites. MRG15 contains chromodomain that is able to bind H3K36me3 ¹⁰¹ and H3K36me3 level on a particular gene correlates with PTB recruitment to pre-mRNAs transcribed from this gene.⁸⁸ Interestingly, MRG15 does not only bind H3K36me3, but also interacts with H3K4me3.⁸⁷ A simple MRG15 knock-down influences alternative splicing of more than 180 genes independently of PTB.⁸⁸ This suggests that there might be an alternative mechanism that does not depend on PTB, although many of these alternative splicing changes might be due to the indirect effects caused by altered expression of splicing regulators after MRG15 knockdown. MRG15 is a component of several chromatin modifying complexes. These complexes contain histone acetylases Tip60 ^102,103 and hMOF,¹⁰⁴ histone deacetylases such as HDAC1 or HDAC2 ^105,106 and histone demethylases.^107,108 Eaf3, a yeast orthologue of MRG15, associates with RNA polymerase II¹⁰⁹ and is important for histone deacetylation during transcription elongation that inhibits the internal transcription initiation inside coding sequences of genes.^110,111 In mammalian cells, a similar complex that consist of both HDAC1 and MRG15 was identified recently and it was shown to affect progression of RNA polymerase II through transcribed regions.¹⁰⁶ Thus, MRG15 significantly participates in regulation of chromatin acetylation and transcription, processes that were shown to modulate splicing outcome (see above) and it would be interesting to know whether MRG15 alone is able to affect alternative splicing directly through modulation of chromatin.

hnRNP Proteins and H3K9 Methylation

H3K9 methylation, which recruits the HP1 protein, is important for the formation and maintenance of heterochromatin. Although average levels of H3K9me3 and H3K9me2 correlate with gene silencing, both histone modifications can be also detected in some actively transcribed genes.^{61,93,112,113} Recently, PTB, together with other splicing regulators hnRNPs (hnRNP A1, hnRNP L, hnRNP K, hnRNP A2/B1, hnRNP A/B), were identified as proteins able to recognize H3K9me3.⁸⁷ In Drosophila, several other hnRNP proteins interact with HP1 and heterochromatin.¹¹⁴ What can be the reason for such an association? One of the H3K9me3 recognizing proteins, hnRNP K, binds to large intergenic noncoding RNAs (linc-RNAs) and the complex of linc-RNA and hnRNP K is an effector of p53 signaling and represses transcription of hundreds of genes.¹¹⁵ When PTB-RNA interactions were analyzed genome-wide by CLIP-seq method, a substantial fraction of tags (almost 30%) was mapped to intergenic regions, implying that PTB may also bind to noncoding RNAs.¹¹⁶ PTB was also detected at the promoter of the HMGA2 gene containing a very specific ∼60 bp polypyrimidine/polypurine element but it is unclear whether binding to this promoter is mediated by chromatin.^117,118 It is intriguing to speculate that the association of hnRNP proteins with the chromatin might regulate transcription. In addition, hnRNP proteins binding to HP1 are involved in heterochromatin formation in Drosophila.¹¹⁴

H3K9me2 and H3K9me3 were also reported to be enriched at introns in contrast to exons.^58,59 hnRNP proteins associated with chromatin then might help to define intronic sequences. In human cells, the increase of heterochromatin marks (H3K9me2, H3K27me3) together with the presence of HP1α leads to changes in alternative splicing of the fibronectin gene.⁴⁰ Recently, it was shown that association of HP1γ with H3K9me3 modulates alternative splicing of several genes including CD44.¹¹³ Although the induced heterochromatin formation near the alternative exons reduces the processivity of RNA polymerase II which causes the increased inclusion of alternative exon, it is still possible that methylation of H3K9 influences splicing of certain genes directly via tethering of splicing factors. Interestingly, exons enriched in the H3K9me3 are preferentially excluded from mature mRNAs indicating again the role of H3K9me3 in the regulation of splicing.⁶¹

Perspectives

The major task in the near future will be to reveal molecular details and mechanism(s) through which chromatin modifications affect pre-mRNA splicing. This knowledge is important not only from a theoretical point of view but might be essential for practical medical usage. Many small molecules that influence chromatin modifications serve as therapeutical agents and thus there is a hope that these compounds might help to treat splicing related diseases as well. The better we understand the connection between splicing and chromatin the higher chance to identify such a drug.

Abbreviations

HDAC

histone deacetylase

snRNP

small ribonucleoprotein particle

hnRNP

heterogeneous nuclear ribonucleoprotein