Nascent RNA and the Coordination of Splicing with Transcription

SUMMARY

At each active protein-encoding gene, nascent RNA is tethered to the DNA axis by elongating RNA polymerase II (Pol II) and is continuously altered by splicing and other processing events during its synthesis. This review discusses the development of three major methods that enable us to track the conversion of precursor messenger RNA (pre-mRNA) to messenger RNA (mRNA) products in vivo: live-cell imaging, metabolic labeling of RNA, and RNA-seq of purified nascent RNA. These approaches are complementary, addressing distinct issues of transcription rates and intron lifetimes alongside spatial information regarding the gene position of Pol II at which spliceosomes act. The findings will be placed in the context of active transcription units, each of which—because of the presence of nascent RNA, Pol II, and features of the chromatin environment—will recruit a potentially gene-specific constellation of RNA binding proteins and processing machineries.

1. INTRODUCTION

Gene expression begins with the synthesis of RNA from the DNA template by RNA polymerase. The steady state level of any RNA in the cell reflects the balance between its synthesis and degradation by specific enzymes; yet, many reactions take place between these two end points. One implicit goal of next-generation RNA sequencing (RNA-seq) is to characterize all possible RNAs expressed in cells from their 5′ ends—indicative of transcription start sites (TSSs)—to their 3′ ends—indicative of 3′-end formation. RNA-seq also reveals massive variation throughout transcript bodies. In the case of eukaryotic messenger RNA (mRNA), this variation is the result of splicing, editing, nucleotide modification, and 3′-end cleavage and polyadenylation, each of which is subject to regulation. A second goal of RNA-seq is to quantify RNA levels in cells. Standard analysis of mRNA reports on steady state levels of the products of all of these reactions and can generate hypotheses about what may have occurred during an mRNA’s lifetime. Because each step of mRNA processing at least begins co-transcriptionally (Fig. 1), sequencing of nascent RNA provides evidence of the precursors, intermediates, and products. In the past decade, these next-generation sequencing applications have complemented live-cell imaging approaches to provide key insights into the coordination of precursor messenger RNA (pre-mRNA) processing with transcription.

Figure 1.

Co-transcriptional RNA processing of eukaryotic protein-coding genes. Schematic diagram depicting a eukaryotic gene with its transcription start site (forward arrow), two annotated exons (thick boxes), and one intron (line). The 5′ end of the annotated intron, or 5′ splice site, and the 3′ end, the 3′ splice site, are marked 5′SS and 3′SS, respectively. RNA polymerase II (Pol II) (oval) moves from left to right with varying speeds indicated by the forward arrowheads, including pausing marked with a pause symbol. The nascent RNA (exons depicted with dashed lines and intron with solid line) emerges from Pol II as elongation proceeds. The nascent RNA 5′ end immediately receives a 7-methylguanosine cap (black ball), and cleavage at the polyadenylation site releases the nascent RNA from Pol II. Many splicing events occur after capping and before poly(A) cleavage. The 5′SS and 3′SS are depicted in the nascent RNA as open circles, and spliceosome assembly is indicated by looping out of the intron. The intron lariat and spliced nascent RNA products are shown.

What proportion of introns are removed concurrently with transcription? Global studies have approached this question by purifying nascent RNA from a chromatin fraction and preparing RNA-seq libraries or probes for high-density tiling arrays. Co-transcriptional splicing frequencies were similarly high in budding yeast (75%), fly (83%), and human (74%–85%) cell lines and tissues, although a lower proportion (45%) was detected by analysis of mouse liver nascent RNA (Carrillo Oesterreich et al. 2010; Ameur et al. 2011; Khodor et al. 2011; Khodor et al. 2012; Tilgner et al. 2012). Validation of RNA-seq and array data by quantitative reverse transcription polymerase chain reaction (RT-qPCR) strengthened and extended these findings (Carrillo Oesterreich et al. 2010; Ameur et al. 2011). General agreement with these numbers came from metabolic labeling (Windhager et al. 2012) and protein biochemistry and immunofluorescence, showing that most active spliceosomes are associated with chromatin (Girard et al. 2012). That said, not all intron removal is co-transcriptional. In particular, terminal introns are least well removed co-transcriptionally, in agreement with RT-PCR studies of individual endogenous genes (Schmidt et al. 2011; Tilgner et al. 2012). Furthermore, 20% of activated spliceosomes in the cell are not chromatin-associated (Girard et al. 2012), and constitutive splicing appears to be more co-transcriptional than alternative splicing (Ameur et al. 2011; Khodor et al. 2011; Tilgner et al. 2012). Moreover, particular cell types use intron retention or detention as a regulatory step (Pandya-Jones et al. 2013; Wong et al. 2013; Braunschweig et al. 2014; Boutz et al. 2015; Pimentel et al. 2016). Taken together, global studies indicate that most introns are removed co-transcriptionally despite differences in species, cell types, and analysis methods (Brugiolo et al. 2013).

Co-transcriptional pre-mRNA processing events can be coupled physically or temporally to one another, to chromatin, and/or to RNA polymerase II (Pol II). Early evidence that transcription-splicing coupling is an important determinant of gene expression came from the analysis of mRNA products in response to changes in coupling. For example, the 5′-end capping enzymes physically associate with Pol II carboxy-terminal domain (CTD) heptad repeats that are phosphorylated on serine 5 and are allosterically stimulated by this interaction (Cho et al. 1997; McCracken et al. 1997a; Cho et al. 1998). When the CTD is truncated, capping fails to occur and mRNA becomes degraded (McCracken et al. 1997b). It is currently unclear how the exosome gains access to unspliced RNA and whether some proportion of RNA escapes capping. An example of “kinetic” coupling between splicing and transcription is the observation that global patterns of alternative splicing are dependent on Pol II elongation rates, which are slowed or accelerated on ultraviolet (UV) damage, chromatin modification, and direct mutation of Pol II (Munoz et al. 2009; Luco et al. 2011; Hnilicova et al. 2013; Schor et al. 2013; Dujardin et al. 2014; Fong et al. 2014). In considering how splicing and transcription are coordinated with one another, an obvious goal would be to determine in vivo transcription rates and, independently, in vivo splicing rates. Quantifying the rate of co-transcriptional splicing and the relative position of Pol II when splicing occurs is discussed in the following sections. These splicing rates will inevitably be confounded by the transcription rates.

Pol II transcription elongation rates have been calculated from metabolic labeling data, timed chromatin immunoprecipitation (ChIP) experiments, or fluorescence imaging and are available for specific genes or gene regions. Globally, mammalian Pol II transcribes 1–4.5 kb/min, with faster elongation rates within introns (Fig. 1) (Singh and Padgett 2009; Jonkers et al. 2014; Veloso et al. 2014). Complementary to these approaches, the distribution of Pol II along genes can be determined at relatively low resolution by ChIP or with strand specificity and higher resolution through UV cross-linking of Pol II to nascent RNA or sequencing the 3′ ends of nascent RNA (Churchman and Weissman 2011; Mayer et al. 2015; Nojima et al. 2015; Carrillo Oesterreich et al. 2016; Mayer and Churchman 2016; Milligan et al. 2016; Nojima et al. 2016). Changes in Pol II density inform on relative local rates of elongation. Accumulation of Pol II in gene regions, such as TSSs, transcription termination sites, splice sites, and/or over exons indicates that Pol II can pause locally when capping, splicing, and poly(A) cleavage occur (Harlen and Churchman 2017; Herzel et al. 2017; Mayer et al. 2017). Thus, one cannot assume that transcription rates are constant when measuring splicing. The following sections detail how the field has explored splicing kinetics in vivo using methods that can be divided into two categories: distance-based, which measure the gene location of Pol II when splicing has occurred, and time-based, which measure the time taken for splicing to be completed (Fig. 2).

Figure 2.

Precursor messenger RNA (pre-mRNA) splicing kinetics in living cells can be measured in space or time. Upper panel: Using chromatin-based assays—chromatin spreads, chromatin immunoprecipitation (ChIP), or nascent RNA-seq—the removal of introns can be associated with a gene position, representing the distance RNA polymerase II (Pol II) has traveled by the time the spliceosome has assembled. Lower panel: Live-cell fluorescence imaging of model genes and metabolic labeling followed by RNA-seq report on splicing rates in units of time. Note that co-transcriptional splicing kinetics vary among species, cell types, genes examined, and methods used for analysis. (Modified from Alpert et al. 2017.)

2. QUANTIFICATION OF SPLICING RELATIVE TO TRANSCRIPTION USING CHROMATIN-BASED ASSAYS

Co-transcriptional splicing was initially discovered by analyzing electron micrographs of chromatin spreads from Drosophila melanogaster, in which shortening of the nascent RNA and concomitant loss of the attached particles provided evidence for co-transcriptional splicing (Osheim et al. 1985). The DNA axis, variable lengths of attached nascent RNA, and electron-dense ribonucleoprotein (RNP) particles attached to the nascent RNA were visualized in these striking images. RNP particles 25 and 40 nm in diameter were separated by an RNA loop and appeared at predictable sites along nascent RNA, suggesting they are components of the assembling spliceosome. These studies measured DNA distance in µm to deduce that splicing occurred when Pol II had traveled 4.5 kb past the 3′SS (Fig. 2) (Beyer and Osheim 1988). Many of the observations made using chromosome spreading methods are consistent with findings in the Balbiani ring of Chironomus tentans, where it has been possible to dissect out chromosomal regions and infer co-transcriptional splicing by RT-PCR and other methods (Bauren and Wieslander 1994; Bjork and Wieslander 2015). Although the harsh conditions of chromatin spreading unfold nascent RNA for analysis, electron tomography images of nascent RNPs in the Balbiani ring system showed the U2 small-nuclear ribonucleoprotein (snRNP) and Pol II in the same electron-dense particle adjacent to the DNA axis (Wetterberg et al. 2001). These early observations of co-transcriptional splicing in model systems raised the possibility that the physical proximity of Pol II, chromatin, the splicing machinery, and the nascent RNP provide opportunities for regulation. There is currently no existing in vitro method that recapitulates splicing in the context of transcription of chromatin, making in vivo approaches obligatory.

2.1. ChIP-Based Detection of Spliceosome Assembly and Splicing within Gene Bodies

Given the proximity of nascent RNA to the DNA axis and chromatin, it was natural to extend ChIP protocols to investigate whether and how spliceosomes assemble co-transcriptionally on individual genes. The key concept is that if RNA binding proteins and higher-order complexes, such as assembling spliceosomes, are present on nascent RNA, it should be possible to preserve interactions with formaldehyde cross-linking and reveal the regions of the DNA template where nascent RNPs contain factors of interest (Bieberstein et al. 2014). The cap binding complex (CBC) and mRNA export factors were among the first to be analyzed. Capping factors and CBC accumulated in promoter-proximal regions (Zenklusen et al. 2002; Listerman et al. 2006; Glover-Cutter et al. 2008), as predicted by the demonstration that nascent RNA is only ∼20 nt long when the 7-methylguanosine cap is added (Rasmussen and Lis 1993; Martinez-Rucobo et al. 2015). In contrast, mRNA export factors were detected in downstream gene regions where nascent messenger ribonucleoproteins (mRNPs) might be expected to mature for export (Lei et al. 2001; Zenklusen et al. 2002), and 3′-end processing factors similarly were detected by ChIP associated with downstream gene regions extending past the poly A cleavage site (Glover-Cutter et al. 2008). Thus, ChIP is capable of identifying molecular components of RNA processing machineries previously inferred from chromatin spreads.

Application of ChIP to spliceosome assembly in yeast revealed that the spliceosomal snRNPs accumulate in distinct patterns along gene regions, consistent with stepwise assembly (Kotovic et al. 2003; Görnemann et al. 2005; Lacadie and Rosbash 2005; Tardiff et al. 2006; Tardiff and Rosbash 2006; Harlen et al. 2016): The U1 snRNP, which recognizes the 5′ splice site (5′SS), was detected over introns, consistent with its recruitment to 5′SSs as soon as they are transcribed. Accumulation of U2 and U5 snRNPs as well as active spliceosomal components downstream from 3′ splice sites (3′SSs) was accompanied by the loss of U1 snRNP, as expected from in vitro biochemical analyses (Wahl 2009). The findings also indicate that U1 snRNP and CBC independently promote later steps of spliceosome assembly, as observed in vitro. In yeast, a ChIP assay was also used to detect the occurrence of splicing relative to transcription of a reporter gene, HZ18, in which the formation of MS2 stem loops either in the intron or after exon–exon ligation would be bound by MS2-binding protein (Lacadie et al. 2006). The MS2 ChIP signal 1.5 kb downstream from the intron suggested that splicing takes place long after intron synthesis in HZ18 (Fig. 2). It appears that splicing delays are present in HZ18, given later findings that indicate more rapid splicing in yeast (see below).

The ChIP approach also revealed co-transcriptional recruitment of spliceosomal snRNPs to mammalian genes, although mapping spliceosome assembly along gene lengths has been less clear-cut in metazoans at least in part because of more complex gene architectures (Listerman et al. 2006; Pabis et al. 2013). Intriguingly, histone posttranslational modifications and the variant histone H2Az modify co-transcriptional spliceosome assembly patterns in yeast, indicating that regulation can be mediated by chromatin as well as by more traditional trans-acting splicing factors (Kress et al. 2008; Gunderson and Johnson 2009; Gunderson et al. 2011; Moehle et al. 2012; Herissant et al. 2014; Patrick et al. 2015; Neves et al. 2017; Nissen et al. 2017). Roles for specific histone posttranslational modifications (e.g., H3K4me3 and H3K36me3) in promoting splicing and responding to splicing activity have also been proposed (Sims et al. 2007; de Almeida et al. 2011; Hnilicova et al. 2011; Bieberstein et al. 2012; Kfir et al. 2015). ChIP and UV cross-linking followed by immunopurification of RNA (CLIP) of members of the SR protein family of splicing factors revealed recruitment to actively transcribed intron-containing genes, consistent with their roles in co-transcriptional splicing (Sapra et al. 2009; Brugiolo et al. 2017). In addition, SR proteins were detectable by ChIP at gene promoters because of the association of SR proteins with the regulator of transcription initiation, 7SK (Ji et al. 2013). Thus, ChIP can reveal splicing-related events in the in vivo context of chromatin; yet, it also carries the caveat that other, splicing-unrelated processes are being measured at the same time.

2.2. Single-Molecule Nascent RNA-seq Relates Splicing to Pol II State and Position

There are two methods for obtaining nascent RNA for the analysis of spliceosome assembly and splicing (Fig. 3A): (1) purification of Pol II and the nascent RNPs bound to it and (2) purification of nascent RNA from a chromatin preparation. Changes in posttranslational modifications (PTMs) of the Pol II CTD mirror and influence the different phases of transcription and nascent RNA processing, owing to the interaction of the CTD with factors that regulate transcription, mRNA processing, and downstream steps (Custodio and Carmo-Fonseca 2016; Saldi et al. 2016; Zaborowska et al. 2016; Harlen and Churchman 2017). The CTD consists of repeats of nearly the same seven amino acids, Tyr₁Ser₂Pro₃Thr₄Ser₅Pro₆Ser₇, mainly modified by phosphorylation of Ser2, Ser5, Ser7, Thr4, and Tyr1. In budding yeast, Ser5 and Ser7 phosphorylation (yielding Ser5P and Ser7P) occurs in promoter-proximal gene regions, whereas Ser2P, Thr4P and Tyr1P appear in downstream regions. PTM transitions have been mapped to pause sites along yeast gene bodies and in particular at 3′SSs, consistent with changes in Pol II elongation rate around intron–exon boundaries (Harlen et al. 2016; Herzel et al. 2017). Although the Ser5P levels are highest at the beginning of transcription units, a link between this PTM and splicing has been found in both yeast and human cells (Fig. 3B), particularly during alternative splicing (Mayer et al. 2015; Nojima et al. 2015; Mayer and Churchman 2016; Nojima et al. 2016). In yeast, nascent RNP protein interactomes obtained by pulldown of the differently phosphorylated forms of Pol II are enriched for splicing factors with the exception of Thr4P (Harlen et al. 2016), consistent with co-transcriptional spliceosome assembly, splicing, and the dissociation of the spliceosome from nascent RNA in downstream gene regions after splicing and before termination.

Figure 3.

Analysis of nascent RNA captures splicing precursors, intermediates, and products. (A) Schematic illustration showing that nascent RNA can be prepared through chromatin purification and/or RNA polymerase II (Pol II) immunoprecipitation. These nascent RNAs vary in length and complexity because of lengthening by Pol II and shortening through splicing. The 3′ end of nascent RNA marks the position of Pol II when the cells were lysed. (B) Alignment of 3′ end reads from mNET-seq shows that immunoprecipitation of Pol II with antibodies specific for phosphorylated Ser5 of the carboxy-terminal domain (CTD) enriches splicing first step intermediates (the 3′ end of exon 1, green asterisk). When splicing is blocked chemically by pladienolide B (Pla-B), this splicing intermediate is lost. (Data from Nojima et al. 2015.) (C) Genome-wide, full-length, long-read nascent RNA sequencing using the Pacific Biosciences or Oxford Nanopore platform reveals the splicing status of individual transcripts relative to the Pol II position (3′ end). (D) Single-molecule intron tracking (SMIT) is a targeted sequencing strategy that detects splicing status relative to Pol II position at base-pair resolution, in which zero indicates Pol II at the 3′SS. A representative budding yeast gene is shown, including the half maximum of splicing (pink dotted line) and the saturation level (green). (Data for panels C and D are from Carrillo Oesterreich et al. 2016.)

Chromatin fractionation provides an alternative source of nascent RNA (Fig. 3A), which has been analyzed using single-molecule RNA-seq methods that determine the position of Pol II when splicing is completed (Carrillo Oesterreich et al. 2016). Direct, long-read sequencing of nascent transcripts with Pol II positions marked by linker ligation to their 3′ ends provides images reminiscent of chromatin spreads (Fig. 3C). A second method, single-molecule intron tracking (SMIT), uses paired-end sequencing to associate splicing status with Pol II position; when compiled, the data indicate the percentage of molecules spliced at a given Pol II position with approximately 300 reads per position (Fig. 3D). SMIT analysis of 87 endogenous genes and long-read sequencing of nascent RNA from Saccharomyces cerevisiae and Schizosaccharomyces pombe reveal detectable exon–exon ligation when Pol II has transcribed ∼50 bp downstream from 3′SSs (Fig. 2). These findings indicate that the active spliceosome is physically very close to Pol II, harkening back to tomography images of nascent RNPs in the Balbiani ring system, in which the U2 snRNP and Pol II were shown to be in the same electron-dense particle adjacent to the DNA axis (Wetterberg et al. 2001). Because 15 nt of nascent RNA is embedded within the elongating polymerase (Martinez-Rucobo et al. 2015), it seems likely that the active spliceosome is positioned at the exit channel of Pol II similar to the 5′-end capping enzymes (see above). Thus, direct interactions between Pol II and spliceosome components appear feasible (Saldi et al. 2016). In mammalian cells, U2AF and FUS interact directly with Pol II and components of the splicing machinery, potentially bridging the two machineries during splicing (Ujvari and Luse 2004; Yu et al. 2015). The observed association of particular Pol II CTD phosphorylation states with spliceosome assembly and splicing suggests that regulation of such interactions occurs during transcription elongation.

3. QUANTIFICATION OF SPLICING RATES THROUGH RNA METABOLIC LABELING

To measure splicing in terms of time rather than relative to the progress of transcription, some studies have implemented inducible transcription to allow tracking of (pre-)mRNA intermediates over a time course. To date, the highest time resolution achieved with RT-qPCR analysis is 30 sec, during which an integrated, tetracycline-inducible reporter gene in budding yeast yielded detectable spliced transcripts 60 sec after pre-mRNA transcription was induced (Alexander et al. 2010a). In human mammary epithelial MCF10A cells stimulated with EGF, pre-mRNA half-lives of 2–3 min were measured by RT-qPCR (Zeisel et al. 2011). Reversible treatment with the elongation inhibitor 5,6-dichloro-1-β-D-ribofuranosylbenzimidazole (DRB) enables the analysis of RNA undergoing transcription without induction, using RT-qPCR. In a study designed to analyze co-transcriptional splicing in very long genes, all >100-kb introns were spliced within 5–10 min in human Tet21 cells (Singh and Padgett 2009). This study confirmed that splicing is co-transcriptional, even in endogenous human genes with extremely long introns.

In traditional metabolic labeling, nucleotide analogs, such as 4-thiouracil (4tU) or 4-thiouridine (4sU), can be fed to cells and incorporated into newly synthesized RNA (Fig. 4). Thiol-specific biotinylation and purification of labeled RNA on streptavidin-coated magnetic beads can be followed by sample preparation for RNA-seq, microarray analysis, or RT-qPCR. This approach has been used to quantify transcription and degradation dynamics by labeling newly transcribed RNA for minutes to hours. Recent technical advances allow for shorter labeling times from 1.5 to 5 min and more efficient biotinylation (Rabani et al. 2011, 2014; Windhager et al. 2012; Barrass et al. 2015; Duffy et al. 2015; Eser et al. 2016). The amount of metabolically labeled RNA obtained depends on cellular uptake of nucleotide analogs, labeling time, incorporation rate, transcription rate, processing rate, and degradation rate. The first three are predefined or can be determined by measuring the proportion of labeled versus unlabeled RNA. The latter can be inferred by mathematical modeling and/or analyzing the labeled RNA fraction over multiple time points.

Figure 4.

Metabolic labeling measures intron half-lives. Upper panel: Schematic representation of nascent RNA after incorporation of a nucleotide analog, such as 4-thio-UTP (4tU), administered to cells for short, medium and long periods of time as indicated. A short pulse will label nucleotides close to RNA polymerase II (Pol II), whereas longer analog feeds lead to incorporation farther along the nascent transcript. Lower panel: Example of a 4tU-seq experiment in Schizosaccharomyces pombe, showing that at short times (e.g., 2 min) read density across the gene is relatively flat and may approximate Pol II transcription, whereas introns are removed and degraded as time progresses (Eser et al. 2016).

All five of the above-mentioned studies detect substantial amounts of pre-mRNA splicing at the very first labeling time point: 1.5 min in budding yeast, 2 min in S. pombe (with a calculated median intron splicing time of 37 sec), 5 min in human B cells, and 10 min in lipopolysaccharide (LPS)-stimulated mouse dendritic cells (with a median intron splicing time of 14 min) (Windhager et al. 2012; Rabani et al. 2014; Barrass et al. 2015; Eser et al. 2016). Such genome-scale data sets identify broad ranges in splicing duration from 4 seconds to many minutes (Fig. 2). Direct comparison of these values is difficult, given differences in experimental systems and analysis methods. It is also important to note that these time windows reflect the convolution of transcription, processing, and degradation, as is the case for live-cell imaging approaches addressed below. Nevertheless, the global nature of metabolic labeling has enabled the discovery of gene architecture and sequence motif correlations linked to synthesis, degradation, and splicing kinetics with which to analyze the coordination between transcription and splicing (Rabani et al. 2011; Windhager et al. 2012; Barrass et al. 2015; Eser et al. 2016).

4. QUANTIFICATION OF SPLICING RATES THROUGH LIVE-CELL FLUORESCENCE IMAGING

Early confirmation that splicing begins co-transcriptionally came from high-resolution fluorescence imaging experiments, which showed the presence of endogenous spliced transcripts at their sites of transcription in mammalian tissue culture cells (Zhang et al. 1994). Consistent with these findings and those of the ChIP experiments discussed above, splicing factors were also localized to sites of transcription by imaging of fixed and living cells (Jimenez-Garcia and Spector 1993; Huang and Spector 1996; Misteli et al. 1997; Neugebauer and Roth 1997). An important development toward determining the in vivo kinetics of splicing through imaging came from the implementation of fluorescent tags and reporter genes to allow the visualization of transcription and splicing within the three-dimensional space of the living cell nucleus (Fig. 5). One study took a global approach by measuring the residency times of fluorescently labeled spliceosomal snRNPs on pre-mRNA transcripts using fluorescence recovery after photobleaching (FRAP) and fluorescence correlation spectroscopy (FCS) in HeLa cells (Huranova et al. 2010). U1 and U4 snRNPs, which transiently associate with the assembling spliceosome, display shorter residency times; subunits of the active spliceosome—U2 and U5 snRNPs—reside on pre-mRNA for 15–30 sec. This bulk value indicates that the average duration of the splicing process at steady state in HeLa cells lies within a 30-sec window. In contrast, most groups measure specific intron half-lives with fluorescent probes. Results from these experiments vary widely (Fig. 2), which is difficult to explain. Chromosomally integrated reporter genes do not report the diversity of kinetics observed for endogenous genomes, and reporters may not be spliced with the same kinetics as the endogenous, parental gene (Martin et al. 2013). However, the attraction of live-cell imaging is that real-time measurements provide higher time resolution than time points taken in metabolic labeling or gene induction analyses. Single-cell information can also provide insight into cell-to-cell variation within the population.

Figure 5.

Measurement of transcription and splicing times with fluorescent live-cell imaging. Upper panel: Diagram of a representative experimental setup, utilizing a reporter gene in which intronic PP7 stem loops bind to fluorescent PP7 binding protein (red) when the intron is transcribed. MS2 stem loops in the last exon bind to MS2 coat protein (green) and allow quantification of transcripts present at the transcription site. Fluorescence detection of both fluorophores reports on transcription elongation time, splicing, and release after poly(A) cleavage. Lower panel: Example showing overlapping PP7 and MS2 signals in human U2OS cells in three dimensions (Coulon et al. 2014).

Comparison among experiments points to several sources of variability. Two studies used stably integrated β-globin reporter genes with MS2 or PP7 stem loops inserted into intronic or exonic sequences to track pre-mRNA transcription and splicing (Fig. 5). One reported splicing within 20 and 30 sec for the two introns in HEK293 cells; their analysis identified short bursts of fluorescence caused by transcription, and the number and lifetime of introns in each burst yielded median intron lifetimes that were considered to be the time window in which splicing occurred (Martin et al. 2013). The second study detected splicing 267 sec after transcription of the last intron in U2OS cells, applying an autocorrelation function to the fluorescent fluctuations to determine how the intronic and exonic signals correlated after a time delay (Coulon et al. 2014). An advantage of this system was the second label placed in the downstream exon, which reports on the amount of RNA at the transcription site before poly(A) cleavage and release. Other possible explanations for the differences between these two values include differences in the cell lines, model genes, and/or data analysis methods. A third study in U2OS cells measured the fluorescent half-life of the MINX reporter intron labeled with MS2-GFP as 105 sec, an intermediate value (Schmidt et al. 2011). It is important to note that measurements of intron half-lives by fluorescence microscopy or metabolic labeling encompass the transcription of intron and exon elements, spliceosome assembly, splicing, spliceosome disassembly, and/or intron release, intron debranching and degradation. Thus, the overall time range observed—0.5 to 3 min—could include variation in the rates of any one or more of these processes.

5. THE COORDINATION OF SPLICING WITH TRANSCRIPTION

It would seem that co-transcriptional RNA processing should aid the efficiency of eukaryotic gene expression by allowing processing to take place within the same time frame and protecting against degradation of the product. In higher metazoans, splicing is a burden. 5′-end capping and poly(A) cleavage happen only once per transcript in all species; yet, splicing occurs typically seven times per human transcript. Moreover, the spliceosome assembles de novo each time an intron is spliced out. Thus, introns hiding within the body of the gene have the time provided by transcription elongation to work out which splice sites will be used and how efficiently splicing will occur. One possibility is that splicing might be facilitated by high local concentrations of splicing, transcription, and chromatin factors, which tend to be intrinsically disordered and may even phase separate at active genes (Harlen and Churchman 2017; Herzel et al. 2017). From the stand point of timing, typical human genes (∼30 kb) will require ∼15 min for transcription elongation given the known range of transcription elongation rates. As we have seen, in vivo distance-based or time-based measurements of splicing support the possibility that introns “can” be completely removed before poly(A) cleavage (Fig. 2). Perhaps the challenge of co-transcriptional splicing is greatest for terminal introns, which will have the least amount of time for splicing before poly(A) cleavage. Interestingly, last exons are on average 10 times longer than internal exons (940 bp and 120 bp, respectively) in humans, providing an extra ∼30 sec for splicing to finish. Indeed, gene architecture may have evolved to take advantage of co-transcriptional splicing and vice versa (Davis-Turak et al. 2015; Hollander et al. 2016). Although splicing can continue posttranscriptionally, the nuclear exosome competes with splicing co-transcriptionally and additional degradation mechanisms await intron-containing mRNAs after transcription (Bousquet-Antonelli et al. 2000; Vargas et al. 2011; Braunschweig et al. 2014). Taken together, attaining high mRNA levels may rely on the ability of the splicing to keep up with transcription.

Transcription and splicing rates seem remarkably well-matched. Although numbers vary (Fig. 2), the spliceosome is capable of acting on the 3′SS shortly after it emerges from the exit channel of Pol II. When transcription is made faster through mutagenesis of Pol II, splicing lags behind in budding yeast (Braberg et al. 2013; Carrillo Oesterreich et al. 2016; Aslanzadeh et al. 2018), suggesting that splicing and transcription rates are matched in wild-type conditions. Naïvely, we would expect slower transcription to favor both fidelity and orderly splice site choices predicated by the order of intron synthesis. However, analyses of a series of slow and fast Pol II variants have revealed that when the rate of transcription elongation is increased or decreased, alternative splicing patterns are disrupted in human cells and splicing fidelity is compromised in yeast (Fong et al. 2014; Aslanzadeh et al. 2018). These recent findings are consistent with an earlier study that altered transcription rates without mutating Pol II (Howe et al. 2003) and with the reported global correlation between intron retention and Pol II pausing (Braunschweig et al. 2014), implicating transcription rate as a general determinant of splice site choice and fidelity. There are many other factors that can impinge on co-transcriptional splicing efficiency, and how these are integrated in a gene-specific manner is less well understood. Included are RNA modifications, editing, and folding. For example, slowing transcription elongation rate inhibits proper nascent RNA folding, thereby preventing 3′-end processing of histone transcripts (Saldi et al. 2018). Considering the diversity of gene architecture and sequences involved, these observations hint that the kinetics of transcription, spliceosome assembly, and splicing catalysis likely evolved together and with the chromatin landscape.

As described in the Introduction, accumulating evidence that elongation rates can vary in vivo suggests that local regulation of transcription might control co-transcriptional splicing, leading to different alternative splicing outcomes. The higher density of Pol II over internal and terminal exons in mammalian cells and correspondingly slower elongation rates may be due to the preferential exonic position of nucleosomes, which can be obstacles to transcription elongation (Hodges et al. 2009; Schwartz et al. 2009; Tilgner et al. 2009). In addition, nascent RNA folding and the formation of RNA–DNA hybrids (R-loops) within the transcription bubble can promote or impede transcription elongation (Huertas and Aguilera 2003; Klopper et al. 2010). These fluctuations in Pol II elongation rate relative to intron–exon architecture within mammalian genes could be directly related to splicing activity or indirectly related to sequence biases in introns and exons. On the other hand, evidence that splicing directly influences Pol II elongation behavior comes from budding yeast, in which transcriptional pausing within terminal exons is specifically detected in short genes that undergo efficient co-transcriptional splicing (Carrillo Oesterreich et al. 2010). In addition, interference with splicing stimulates pausing at 5′ and 3′SSs (Alexander et al. 2010b; Chathoth et al. 2014). A recent study has shown that the large subunit of Pol II (Rbp1) in yeast is modified by ubiquitination on lysines 452, 695, and 1246 and that Bre5, an RNA binding protein, recruits de-ubiquitinase activity to remove this modification on K1246 (Milligan et al. 2017). A viable deletion of Bre5 or mutation of K1246 to arginine alters Pol II density in intron-containing gene bodies, raising the possibility that de-ubiquitination releases Pol II from splicing-associated pausing and affects the efficiency of splicing. These findings suggest that regulation of Pol II by Bre5 may provide an important link between splicing and transcription.

6. CONCLUDING REMARKS

To bring the concepts discussed here into even better focus and to facilitate biomedical understanding, more fundamental knowledge is needed. First, we need specific, comprehensive knowledge of potential direct contacts between spliceosomal complexes and chromatin components and between spliceosomal complexes and the transcriptional machinery. Second, we need to determine comprehensively the rates of intron removal in higher metazoans in which alternative splicing is common. There are many reasons to suspect that these splicing rates will vary enormously, likely depending on the many factors discussed above and possibly factors yet to be discovered. Third, we need more information about splicing intermediates to obtain a fuller picture of in vivo kinetics. In general, current assays exploit either detection of exon–exon junctions or determination of intron half-lives. Because each has biases, new methods for detecting splicing intermediates would enhance accuracy (Wallace and Beggs 2017). Related to this, it will be important to identify steps in spliceosome assembly that are rate limiting in vivo, in which the multiple, potentially competing interactions between nascent RNA and RNA binding proteins and spliceosomal components may be modulated by the chromatin environment, transcription rates, RNA editing and modification, and the rates of co-transcriptional nascent RNA folding.

These steps forward will bring us closer to the unexplored frontier of how the coordination of splicing and transcription impacts disease and can be exploited for therapeutics. Cis mutations that affect gene splicing, trans mutations that affect splicing factors, and the overexpression of splicing factors can be drivers of cancer, diseases like myelodysplastic syndrome, and neurodegenerative conditions like spinal muscular atrophy. Today, standard RNA-seq approaches can reveal faulty products of splicing but fail to tell us how those products arose and what additional downstream consequences may accompany these mutations in living cells. Does a given mutation cause splicing rates to slow down or accelerate? Will kinetic changes in splicing lead to changes in transcription elongation and/or output? Will chromatin signatures respond to such changes and drive broader effects? Clearly, the spectacular advances in methodologies and the insights they have already provided point us in numerous, exciting future directions.