Reflections on the history of pre-mRNA processing and highlights of current knowledge: A unified picture

PolyA—first strong indication of pre-mRNA processing

In 1971–1972, a 3′ unitary-sized polyA segment was discovered in polyribosomal mRNA (Darnell et al. 1971b; Edmonds et al. 1971; Lee et al. 1971). Mary Edmonds, with Richard Abrams, had discovered a polyA polymerase in lymphocyte extracts that needed no primer (Edmonds and Abrams 1960) and George Brawerman had found a polyA segment in rat liver cytoplasm (Hadjivassiliou and Brawerman 1966). However, without a look into the RNA found in polyribosomes, this unusual sequence had not found a home.

Proof that this unit of pure adenylic acid residues was added first to nuclear RNA by enzymes, as Edmonds and Abrams reported, and then the same-sized unit appeared in polyribosomal mRNAs within a few minutes, was the first strong suggestion of processing of large nuclear molecules in the formation of mRNA (Darnell et al. 1971a; Jelinek et al. 1973; Nakazato et al. 1973).

The discovery of polyA addition was followed by the discovery in 1974–1975 of the addition of the 5′-methylated GpppX_mp cap at the 5′ end of animal virus mRNA (Shatkin 1974; Wei and Moss 1974; Furuichi et al. 1975a). That same structure was soon found also in both cellular nuclear RNA and polyribosomal mRNA (Perry and Kelley 1974; Furuichi et al. 1975b; Wei et al. 1975; Salditt-Georgieff et al. 1976).

Discovery of splicing and spliceosomes

Finally, in the spring of 1977, electron microscopic (EM) analysis of late (18 h of infection) adenovirus mRNAs (synthesized from adenovirus DNA by host nuclear enzymes and containing both caps and polyA) (Philipson et al. 1971; Sommer et al. 1976) uncovered splicing (Berget et al. 1977a,b; Broker et al. 1977; Chow et al. 1977a,b; Chow and Broker 1978). Many different completed late cytoplasmic adenovirus mRNAs all bound to adenovirus DNA at the same three short regions close to the transcription start site and to individual, more distant downstream genomic segments. The EM photos revealed that, most likely, RNA splicing—a revolutionary idea—reduced the large (∼28 kb) adenovirus primary RNA transcripts defined by labeled nascent RNA studies (Bachenheimer and Darnell 1975; Goldberg et al. 1977; Weber et al. 1977) to smaller cytoplasmic mRNAs. Not only did these groundbreaking electron micrographs first illustrate the event of splicing, but they also revealed that one of ∼15 different mRNAs emerged (one per primary transcript) from each single large adenovirus pre-mRNA (Broker et al. 1977; Chow et al. 1977a; Berget and Sharp 1978; Chow and Broker 1978), portending regulated differential splicing. Very quickly, many other animal virus RNAs, followed by individual cellular RNAs, were shown to be spliced using the EM and other molecular techniques. The first genomic sequence that confirmed the necessity for splicing was for the mouse β-globin gene, described in December of 1978 (Konkel et al. 1978).

In the three and a half decades that have followed the discovery of splicing, intense interest and effort have established intricate details of, first, the surprising RNA chemistry involved in splicing (transesterification mediated by RNA:RNA interactions) (summarized in Cech 1983, 2009). At about the same time came the discovery of a set of small nuclear RNAs contained in nuclear ribonucleoprotein structures—spliceosomes (Hinterberger et al. 1983; Brody and Abelson 1985; for review, see Steitz 1987; Sharp 1991). With the aid of dozens of processing factors, spliceosomes orchestrate the site-specific transesterification reaction. By the early1990s, the spliceosome was shown to be assembled at each splice event and to act in a dynamic stepwise process that links exons in mRNAs (for review, see Wahl et al. 2009).

³H nucleosides and pre-mRNA synthesis in HeLa cells

The earliest biochemical experiments that ultimately had a direct bearing on pre-mRNA processing utilized the flow of ³H nucleosides (first uridine and later adenosine) into large nuclear RNA and, with a short delay, into much smaller polyribosomal RNA, reasonably assumed to be mRNA.

Label times of 10, 20, and 45 sec with ³H uridine (Derman and Darnell 1974; Derman et al. 1976) established that it was possible to label only nascent RNA in the large nonribosomal nuclear RNA (hnRNA, heterogeneous nuclear RNA) (Soeiro et al. 1966; Perry and Kelley 1970). Accumulation of label at these three short times was not linear but in the ratio of 1:4:15, indicating the nucleoside triphosphate pool was not yet at peak radioactivity even by 45 sec (Fig. 1A). Further, the sedimentation profile of these briefly labeled growing RNA chains was virtually identical at the three times, and the labeled molecules had a maximum at ∼4–5 kb but ranged in size up to at least 40 kb. However, after a 30-min exposure to ³H uridine, the nonribosomal nuclear RNA sedimentation profile had a maximum in molecules >14 kb. Thus, the molecules labeled for 45 sec or less were growing nascent RNA chains, and all transcripts of whatever eventual size would be equally labeled only on their growing 3′ ends. Numerical analysis of the size of the labeled nascent RNA chains showed 50%–60% would eventually be larger than the 5-kb marker (28S rRNA), and >30% of the eventually finished RNA chains would be larger than the 14-kb marker (45 pre-rRNA) (Derman et al. 1976). These conclusions gibe well with the transcript lengths now determined by whole-genome sequence analysis of DNA and by sequence of transcribed nonribosomal RNA (RNA_seq).

Total nuclear molecules briefly labeled with ³H adenosine gave a very similar sedimentation profile to the ³H uridine-labeled molecules. PolyA⁺ nuclear RNA also became labeled in 2 min and ∼50% of the labeled ∼230-nt polyA segment was added to molecules larger than 6.7 kb (32S) (Fig. 1B; Derman and Darnell 1974; Sawicki et al. 1977).

Even in steady state, >60% of ³²P-labeled polyA⁺ hnRNA molecules were larger than 6.7 kb (32S pre-rRNA), while 80% of polyribosomal mRNA was smaller than 2 kb (18S rRNA) (Derman et al. 1976; Salditt-Georgieff et al. 1976). So, if processing delivered most or all of the polyA to the cytoplasmic mRNA (see Puckett et al. 1975), polyadenylation generally preceded the completion of processing, whatever that entailed.

Another result pertinent to the timing of polyA addition was the flow of ³H uridine into total hnRNA and into uridine residues next to polyA. Limited alkaline treatment of extracted nuclear nonribosomal RNA after brief label times of ³H uridine incorporation (30 sec to 4 min) reduced all labeled molecules to an average size of ∼500 nt, and polyA⁺ fragments and polyA⁻ fragments were separated. The polyA⁻ ∼500-nt hnRNA fragments and the polyA⁺ fragments accumulated label at almost the same rate, with labeling of polyA+ fragments lagging by, at most, 30 sec (Salditt-Georgieff et al. 1980). PolyA addition, therefore, occurred within, at most, 30 sec to newly synthesized nuclear RNA (Fig. 1C).

Altogether, these early experiments provided very strong evidence (prior to the discovery of splicing) that much of the polyA was quickly added in the nucleus to molecules much larger than mRNA and that, beginning ∼5–10 min later, the same ∼230-nt 3′ polyA unit (Jelinek et al. 1973; Sawicki et al. 1977; Salditt-Georgieff et al. 1980) began to accumulate in increasing amounts in the much smaller cytoplasmic polyribosomal mRNA. As we shall see, the early short-label experiments are consonant with the latest RNA_seq studies on nascent, still chromatin-associated molecules (Bhatt et al. 2012; A Pandya-Jones, DM Bhatt, C-H Lin, ST Smale, and DL Black, in prep.)

Brief labeling of specific nuclear RNAs

The experiments discussed above were carried out on total RNA samples. A few early studies were also performed with briefly labeled, specific pre-mRNA transcripts from both viral and cellular DNA. Globin precursor to mRNA (Kinniburgh and Ross 1979) and immunoglobin nuclear RNA precursors to heavy chain (11 kb) and light chain (3.3 and 5.3 kb) mRNAs (Schibler et al. 1978) became labeled and polyadenylated more prominently than labeled mRNA-sized molecules. The same was true for briefly labeled SV40 (Lai et al. 1978) and early adenovirus (Weber et al. 1980) primary transcripts and mRNA molecules. Obviously, polyA was added to all these relatively short precursor molecules before they were spliced.

Detailed examination of the kinetics of the major late adenovirus transcription unit and the molecules produced from it also was greatly illuminating. For example, cleavage and polyA addition occurred within 45 sec of RNA polymerase passage of a chosen polyA site (Berget and Sharp 1978; Nevins and Darnell 1978a,b; LeMoullec et al. 1983). The briefly labeled, polyA-terminated, nuclear adenovirus-specific pre-mRNA appeared to stretch from (at least near) the transcription initiation site to one of the possible polyA sites (Fig. 2). These molecules had been cleaved and polyadenylated within seconds at one (and only one) of the five polyA⁺ sites used to process the primary transcript but had not yet been shortened to mRNA size by the splicing events that would produce one of the ∼15 much shorter mRNAs. Evidence was strong that each primary transcription event, nevertheless, continued to a position ∼28 kb from the initiation site, stopping at an apparent single transcription termination site or region near to, but not at, the end of the linear genome (Fraser et al. 1979).

However, it was soon convincingly shown with labeled late nuclear adenovirus transcripts that the major late transcript did, in fact, begin to be spliced at its 5′ end (Keohavong et al. 1982; Mariman et al. 1983) before transcription was complete. All the late adeno mRNAs contain three noncoding “leader” sequences: L1, L2, and L3 (Berget et al. 1977a,b; Broker et al. 1977; Chow and Broker 1978). L1 is ∼50 nt long and contains the 5′ cap structure; L2 is ∼80 nt long and maps ∼1100 nt downstream from L1; L3 is ∼100 nt long and maps ∼2500 nt downstream from L2. The L1to L2 splice occurs within seconds to minutes after transcription, but the L1-L2 splice to L3 requires over 10 min to occur, and additional time is required for final splicing to a downstream messenger RNA “body” that encodes protein (Fig. 3; Keohavong et al. 1982; Mariman et al. 1983). From these results, it was clear that some splices take longer than others and some can occur prior to completion of an entire pre-mRNA, a theme that has carried forward ever since.

Rate of RNAPII chain elongation

Knowing the speed of RNA chain elongation by RNAPII is crucial to the consideration of the timing and order of processing events and the nature and extent of the integration with transcription of transcriptionally important proteins (RNAPII and associated proteins) and processing factors.

The anti-viral compound DRB (5,6-dichloro-1-β-D-ribofuranosylbenzimidazole), which prevents chain elongation beyond promoter proximal pause sites by inhibiting the P-TEFb kinase (Core and Lis 2008; Core et al. 2008), has been extremely useful in determining chain elongation rate. For example, using DRB, Endre Egyhazi in 1974 (Egyhazi 1974) reported that the already started giant, 75S mRNA, encoding a secreted protein in the salivary gland of Chironomus tentans, continued to be elongated at ∼30 nt per sec (∼2 kb/min) at 25°, the temperature at which these larvae thrive.

The previously described older studies on labeled nascent RNA in cultured mammalian cells (Derman and Darnell 1974) afforded a chance using DRB to estimate chain elongation rate in mammalian cells. Already initiated hnRNA chains greater than a few hundred nucleotides continued to be labeled with ³H uridine in HeLa cells in the presence of DRB at an estimated ∼50 nt per sec (3 kb per min), slightly faster than the estimated rate in C. tentans (Sehgal et al. 1976, 1979).

Recently, Singh and Padgett (2009) used DRB to empty RNAPII from actively transcribed genes in cultured human cells. Upon removal of the drug, they extracted the nuclear RNA and scored transcriptional rates during resumption of transcription. The synthesis rate in pre-mRNA was scored by RT-PCR amplification of known sequences within many different very long primary transcripts. These extremely clear experiments showed a 3–4 kb per min rate of elongation throughout the primary transcripts, including very long introns (see Fig. 5). However, most important for later discussion, the transcription rate through long regions densely packed with exons also occurred at the same high rate. Multiple 5′ exon-intron and 3′ intron-exon junctions did not measurably impede the rate of RNAPII elongation (details discussed later) (see Fig. 5).

Singer and colleagues (Darzacq et al. 2007) installed in human cells specifically constructed, tandemly arranged genes, the inducible transcriptional product of which contained RNA sequences that bound a fluorescent protein probe so that the time required for transcription at the inserted site could be charted. They obtained a maximum rate of 4.3 kb per min, with some irregularity in polymerase movement at different sites of integration. A second report (Brody et al. 2011) using a very similar approach measured RNA polymerase elongation rate at 3.6 kb per min.

An additional straightforward biochemical technique recently measured chain elongation rate in cultured human cells (Wada et al. 2009) by following the synthesis of specific nascent pre-mRNA induced by the cytokine TNF α. The primary transcript elongation rate was measured in short intervals (min) by hybridization of induced nascent RNA to a tiling array of DNA. In addition, this group also used a fluorescent assay. A growth rate of 3.1 kb per min was obtained with both techniques (see Fig. 5C).

Satisfyingly, estimations of the synthetic rate of RNAPII in mammalian cells have been constant for 35 years; it is 3–4 kb per min. Thus, a 10-kb pre-mRNA transcript, for example, requires 3 min or less for complete synthesis. These time considerations are very relevant to the order of splicing and of polyA addition. The most direct measurement of RNAPII elongation rate in vivo in yeast, charted by the time of the accumulation of RNAPII on the 3′ ends of induced genes, was estimated to be ∼2 kb/min (Mason and Struhl 2005). Thus, in all these cases, the elongation rate was estimated to be ∼30–50 nt/sec, with all of the mammalian estimates falling in the faster range.

A recent review has professed doubt about whether these “average” elongation rates adequately allow for some “irregularity” in transcription rate (e.g., purposeful pausing at splice sites) (Oesterreich et al. 2011). We will return to this important unsolved question later.

Locating nascent RNA within the nucleus

Once proteins involved in splicing were identified, antibodies to these factors were promptly developed. Experiments in mammalian cells using fluorescent antibodies found processing proteins clustered in many punctate sites (“speckles”) scattered throughout the nucleus, as well as weaker staining on chromosomes. However, the speckles are not loci of processing (e.g., Huang and Spector 1991; for review, see Maniatis and Reed 2002). It seems quite likely the speckles are “storage bins” and not sites of pre-mRNA splicing. In experiments mentioned earlier in the section discussing chain elongation rates, Singer and colleagues (Darzacq et al. 2007) used constructed genes containing DNA sequences that, when inducibly transcribed, can be detected with the use of fluorescent proteins that bind the specific nascent RNA to act as a cytologic reporter. Similar reports (Brody et al. 2011) make clear that transcription and processing of the marked pre-mRNA transcripts occurs at (or very near) the chromosomal site of integrated genes. The nascent RNA and the integrated DNA can be separately detected by hybridization using differently marked fluorescent RNA and DNA probes. The two nucleic acid probes are in the same locale although the integration of the introduced and transcribed genes occurs at multiple sites. This transcription and beginning processing are not coincident with antibody-detected “processing protein clusters.” In view of the occurrence of frequent splicing before transcription is complete (discussed below), these experiments argue strongly that the bulk of splicing and polyadenylation appears to occur on nascent RNAs near the chromosomal site of synthesis.

Following selected nascent RNA from early transcription: Splicing of nascent RNA and time required

In experiments mentioned earlier on chain elongation rate, Singh and Padgett (2009) also provided some of the clearest evidence that splicing does occur on a variety of nascent, not yet completed pre-mRNAs in human cells. These experiments (see Fig. 5B) utilized DRB to empty genes of RNAPII transcripts, followed by removing the drug to allow transcription to resume. PCR primers that could detect unspliced exon–intron and intron–exon junctions, as well as primers to detect exon–exon unions, were used to score splicing at various times after DRB removal in multiple sites along the primary transcript. The unambiguous results showed frequent exon–exon union did occur to not yet fully transcribed molecules, with a higher frequency but not exclusively toward the 5′ end of the primary transcript, long before the RNAPII could possibly have reached all the way to the polyA site in these very long transcription units. However, even in these extremely long pre-mRNAs, only part of the necessary splicing had occurred in the 3′ half of the nascent pre-mRNA.

Singh and Padgett were also able to measure the time required for a splice to be completed as the RNAPII extended these nascent RNAs. Once again, the measured RNAPII rate of elongation of ∼3–4 kb per min allowed a measure of the time elapsed after an RNAPII complex passed both the 5′ exon–intron junction and the 3′ intron–exon junction before exon–exon union. The time for splicing after passage across to the 3′ intron–exon junction was ∼5 min for nine sites and 10 min for five other splice sites. (Also see Audibert et al. [2002], using a very different approach based on time for completion of an mRNA compared to time of synthesis.)

These results argue against long pausing of RNAPII at every splice site (hinted at by Oesterreich et al. 2011). In two thoroughly analyzed genes, Singh and Padgett detected no difference in chain elongation rate of nascent RNA from one gene of an intron of 110 kb (no splices) compared to a region of 133 kb with five exons. In other genes, the elongation rate of 3–4 kb/min was the same in two regions of 174 kb with 49 exons and another 105-kb region with 35 exons. In the case of these two primary transcripts, if there were a pause of, say, 1 min at each 5′ exon–intron junction and 2 min at each branch point/polypyrimidine stretch that marks the 3′ intron–exon junction, it would triple the time observed for RNAPII to transcribe these regions at 3 kb/min. If the pause were only, say, ∼10 sec, the Singh and Padgett experiments might not have detected such minor interruptions. So, it seems clear from the outset of the consideration of integration of splicing and transcription (see Figs. 3, 4) that “cotranscriptional” splicing cannot mean that the RNAPII complex only leaves a newly transcribed intron after it has been removed, but brief pausing at specific sites seems possible and very much needs to be settled (see later discussion).

Also, in yeast that generally have only short transcripts that are spliced at single sites, only partial splicing during transcription is the rule. Using a high-throughput sequencing analysis of splicing in 239 yeast pre-mRNAs, Tardiff et al. (2006) searched for completion during transcription of these transcripts. A great many yeast genes with introns encode ribosomal proteins, and the pre-mRNA for these is relatively short, <3 kb. Tardiff and colleagues detected four times as much polyA⁺ unspliced RNA as spliced mRNA from genes with one or two short exons, reminiscent of the early nucleoside labeling experiments with globin, immunoglobin, and animal virus transcripts described in an earlier section. Furthermore, analysis of total pre-mRNA processing by determining binding (ChIP) of spliceosomal components (U1 snRNPs and U2, 4, 5 snRNPs) showed that the great majority of yeast pre-mRNAs did not have time during synthesis to assemble a complete spliceosome and to finish splicing during synthesis. Thus invariant, completed cotranscriptional splicing seemed not to be the rule, although cotranscriptional splicing in yeast can certainly occur in genes with longer 3′ exons.

While extrapolations about times of complex interactions in vitro to in vivo actuality are unwarranted, it is noteworthy that a recent study measuring the time required to assemble individual yeast spliceosomes in vitro was at least 10–15 min (Hoskins et al. 2011).

Recent studies on nascent RNA in mammalian cells and in Drosophila

Perhaps the most convincing and informative approach to studying the mechanics of pre-mRNA processing involves nuclear fractionation (using 0.3 M NaCl and 1 M urea), leaving chromatin (at least, nucleosomes) intact with the great majority of nascent pre-mRNA still associated. This technique (Wuarin and Schibler 1994) was originally used with radioactively labeled DNA probes and showed that the nascent transcripts of two genes in the mouse liver nuclei underwent some but not complete splicing while the transcript was still attached to the DNA in chromatin.

With the availability of complete genome sequences, nascent RNA in chromatin prepared by this technique was examined for splicing using sequence specific approaches by Amy Pandya-Jones and Douglas Black (Pandya-Jones and Black 2009). Their initial experiments concentrated on two long transcription units in cultured human cells, cellular src (csrc) and fibronectin, the genes for both of which are actively transcribed and are over 50 kb long with multiple introns. Pandya-Jones and Black adopted the strategy of using pairs of primers that span 5′ exon-intron and 3′ intron-exon junctions as well as primer pairs spanning neighboring exons in mRNA. Use of the two sets of primers accurately detected unspliced primary transcripts as well as the union of two neighboring exons. About 50% of introns were removed from the nascent transcripts and was mainly (not completely) orderly from 5′ to 3′. Moreover, several of the exons in the two pre-mRNAs were known to be alternatively spliced, and this regulated splicing was also detected on the nascent pre-mRNA but, again, was not totally complete (see Fig. 5B).

These experiments were important for another reason. Both chromatin-associated nascent RNA and any nuclear RNA that did not fractionate with chromatin were examined. There was a great deal more pre-mRNA detected that was bound to chromatin than in the unbound nuclear RNA faction. Very little incompletely spliced pre-mRNA was released from the chromatin. Furthermore, the experiments implied completed mRNA must be transported rapidly compared to the synthesis and processing time of pre-mRNA.

In another recent contribution to the “efficiency of splicing” literature on chromatin-associated nascent RNA, Rosbash and colleagues (Khodor et al. 2011) reported on Drosophila S2 cultured cells and on total fly “head” RNA (using the technique of Wuarin and Schibler [1994] to prepare chromatin). In a very exhaustive analysis with high-throughput total RNA sequencing (RNA-seq), they found many introns were spliced in this nascent RNA at ∼80% efficiency. But, to quote, “1689 genes had 2793 introns with >50% cotranscriptional intron retention.”

Among the most recent experiments on nascent pre-mRNAs are the collaborative efforts in the laboratories of Douglas Black and Stephen Smale (Bhatt et al. 2012; A Pandya-Jones, DM Bhatt, C-H Lin, ST Smale, and DL Black, in prep.). They describe the RNA-seq examination of the chromosomally associated nascent pre-mRNAs that add considerably to the proof of retention on chromatin of not yet completely spliced pre-mRNAs, most likely even after polyA addition. In addition, the experiments suggest that transcriptionally complete but not completely spliced transcripts remain associated until processing is completed.

A distinct advantage of these latest experiments is that the authors stimulated the otherwise untreated cells under study (macrophages) with lipid A, an agent that induces the inflammatory response and the prompt induced increase of transcription in macrophage-like cells of many known genes, followed by a transcriptional decrease. The experiments reported on three classes of RNA: nascent chromosomal RNA, nucleoplasmic RNA (the nuclear RNA that did not sediment with chromatin), and cytoplasmic polyA+ RNA. The RNA_seq technique affords a quantitative assessment of the amount of transcribed nascent RNA (“RNA reads”) across the entire length of an induced gene. Thus, exon and intron content of pre-mRNA associated with any given gene was measured. Genes that exhibited at least a fivefold induction of chromatin-bound RNA (nascent RNA) were examined after 15, 30, 60, and 120 min. which afforded examination throughout a rise and fall of transcription reflected by a large increase and then a decrease in chromatin-associated nascent RNA for individual genes. There was a time delay in the appearance of any RNA in the nucleoplasm and a slightly longer time for an increase in the polyA-containing cytoplasmic fraction. An important point was re-emphasized—much more induced RNA existed on the chromatin than in the nucleoplasm from all of these acutely induced genes.

Many points pertinent to a consideration of processing were further illuminated by the details of the experiments. Virtually all of the induced nascent RNAs showed some splicing, but at least half of the final splices were not completed (especially those near the 3′ end of the transcript) by the time the full-length nascent RNA had reached the polyA site. An abrupt, steep fall-off of transcripts—“reads”—(not to zero) occurred at the polyA site of these pre-mRNAs, although it is firmly established that RNAPII most often continues transcription beyond the polyA site (see Fraser et al. 1978; Hofer and Darnell 1981 for old evidence examining pulse-labeled RNA). Recent experiments (for review, see Richard and Manley 2009) indicate that “torpedo” destruction of post-polyA site cleaved transcripts by a 5′ to 3′ exonuclease following closely behind RNAPII might not leave much evidence that these regions are, in fact, transcribed.

The results from the Black and Smale labs clearly indicate that the chromosomal pre-mRNA contained nonspliced transcripts, and most likely, some are already cleaved at polyA sites. The chromatin-associated transcripts were not proven to be polyadenylated, but reflecting on the old results showing, at most, 30 sec is necessary for polyA addition (Derman and Darnell 1974; Salditt-Georgieff et al. 1980), it seems likely that at least some of this incompletely spliced RNA has already been polyadenylated and is “waiting” both to complete splicing and to associate with exon junction complexes (Tange et al. 2004) or transport proteins (Stewart 2007) needed to safely enter the cytoplasm. Whether the unspliced sites are random or represent particularly “difficult” or “slow” splice junctions is not yet clear. In contrast, virtually all of the molecules in the nucleoplasm were completely spliced and polyadenylated, as were those in the cytoplasm.

A final note on the frequency of cotranscriptional removal of introns (and the attitude of authors toward their data) comes from high-throughput sequencing of nascent RNA. The impressive international effort of the ENCODE project in which presumed nascent RNA (not prepared by the Wuarin and Schibler technique) was subjected to “deep sequencing” and reported to be >50% and thus “predominantly cotranscriptional” (Tilgner et al. 2012). The Rosbash lab, using the Wuarin-Schibler technique, compared total Drosophila head and cell culture nascent RNA and mouse liver nascent RNA for splicing during transcription (Khodor et al. 2012). As noted earlier, ∼75% of total Drosophila nascent RNA was spliced, but only 45% of mouse liver nascent RNA was spliced (Khodor et al. 2011).

The reasonable conclusion from all of these extensive recent reports plus older work is that “cotranscriptional” splicing occurs, but for at least 50% of mammalian genes, and probably for all relative short (<10 kb or less?) genes, polymerization is complete and polyA addition has occurred before splicing is complete. The two major processing events are not obligatorily connected mechanistically.

Focus on the CTD

Strong evidence began building in the 1990s, with close to a consensus belief at present, that the –COOH terminal domain, the CTD, of the largest subunit of RNAPII plays a necessary and decisive role in ensuring efficiency of processing, with the inherent implication that proteins (complexes) are deposited on growing RNAs very quickly after the nascent RNA has emerged from the RNAPII exit channel.

The CTD consists of heptad repeats (Y₁S₂P₃T₄S₅P₆S₇; 52 in mammals and 26 in yeast) in which all three serines (S) 2, 5, and 7, threonine (T₄), and tyrosine (Y₁) are subject to phosphorylation at different stages of transcription, providing changing affinities for different processing factors (for review, see Buratowski 2009; Munoz et al. 2010; Hsin and Manley 2012). Moreover, it is proven that various proteins important in elongating and processing the pre-mRNA can bind to the phosphorylated CTD (for review, see Reed 2003; Gornemann et al. 2005; Listerman et al. 2006; Core and Lis 2008; Chen and Manley 2009; Richard and Manley 2009; Hsin and Manley 2012). These proteins include factors implicated in capping the pre-mRNA, splice site recognition, and splicing execution (e.g., SR proteins and polypyrimidine binding proteins, CDK kinases and phosphatases). Particularly well-studied are several different proteins involved in polyA cleavage and polyA addition. More recently, individual proteins and complexes capable of modifying chromatin (methylases, acetylases, etc.) have also been found capable of binding the CTD.

Meinhart et al. (2005) provide an excellent review of the physical structure of proteins bound to the CTD segments and discuss the probable dimensions of the CTD domain. No structure of the entire CTD domain is available, but they describe several known structures with phosphorylated heptad repeats. Crystal structures show binding domains of processing proteins encompassing 1, 2, or 3 phosphorylated heptad repeats. These findings suggest that as many as 8–10 proteins in yeast and twice that many in mammalian cells might bind simultaneously to a properly phosphorylated longitudinally structured CTD, not to mention additional associated factors not directly bound to the CTD. Moreover, the structure of the yeast RNAPII shows the CTD to be poised just at the exit channel of the whole enzyme. However, even with this enormous concentrating power of the CTD in bringing RNA processing proteins near to nascent RNA, it should be noted that there are many more different factors (e.g., >15 SR proteins alone) than CTD binding sites, and the stoichiometry of bound factors at any instant during transcription is unknown.

Choice between alternative polyA sites

At least 50% of primary transcription units in mammals have two or more polyA sites, and about one-quarter of these (10%–15% of total pre-mRNAs) result in mRNAs with different coding capacity of their 3′ terminal exons, as illustrated originally in the late adenovirus transcript (Edwalds-Gilbert et al. 1997; Tian et al. 2005; Di Giammartino et al. 2011).

In addition, ∼40% of human genes exhibit differential choices of polyA sites within UTRs that very likely result in removing sequences to which microRNAs bind and lead to mRNA turnover. These alternative UTR sites often are “weaker” polyA sites (for review, see Mayr and Bartel 2009; Hornyik et al. 2010; Rosonina and Manley 2010; Di Giammartino et al. 2011). Of considerable interest is the common use of the alternative upstream UTR sites in cancer cells. In several cases, this choice depends on an increase in one or more of the well-identified core polyA factors (for review, see Di Giammartino et al. 2011).

Finally, an important role of suppressing unwanted intronic polyA sites was recently elaborated by experiments from the Gideon Dreyfuss group. U1snRNPs bind to many such sites, inhibiting their use and thereby preserving the use of “proper” downstream polyA sites (Berg et al. 2012).

As the number of known processing factors has greatly increased, particularly in the last decade, many molecules including those known to be important for splicing have now also been implicated in 3′ end processing. Several examples of differential polyA choices illustrate regulated events in 3′ end formation involving 3′ coding exons.