The RNA Polymerase II CTD and Co-transcriptional Coupling
Imagine a factory floor. A single moving conveyor belt must trigger cutting, stitching, and packaging machines in order. It does this without a foreman standing over it. This is essentially the problem a eukaryotic cell faces every time it transcribes a gene. A single RNA polymerase II (Pol II) enzyme must synthesize a pre-mRNA. It must also ensure the transcript gets capped at its 5′ end. The transcript must be spliced to remove introns. It must also be cleaved and polyadenylated at its 3′ end. All of this happens while the RNA is still being made. The solution the cell evolved is elegant. Pol II carries its own mobile signaling platform. This is a long repetitive tail called the carboxy-terminal domain (CTD). Its chemical modifications change predictably as transcription proceeds. This matters because RNA processing is not a series of independent post-transcriptional cleanup steps. Instead, it is woven into transcription itself. This phenomenon is called co-transcriptional coupling. Coupling makes RNA processing dramatically more efficient and accurate. Without it, processing machines would have to find nascent RNA by chance diffusion. Coupling also creates a built-in quality-control logic. Capping enzymes only get recruited near the promoter. Splicing factors are handed off as the polymerase transcribes through exon-intron boundaries. Cleavage/polyadenylation factors are activated only once Pol II approaches the gene's end. The conceptual key is the "CTD code." This is a pattern of phosphorylation and other post-translational marks. These marks sit on the repeating heptapeptide units of the CTD. Different processing factors read them as they dock onto the transcribing polymerase. This is one of biology's clearest examples of a processive protein interpreting positional information along its own trajectory. It is a design principle that recurs from motor proteins to chromatin-modifying complexes. This article develops the mechanism of CTD-based coupling in molecular detail. It reviews the classical and modern experiments that established the code. It also works through examples of how the code is read, misread, and exploited. This lays the groundwork for the more detailed treatments of The 5' Cap and Its Functions, Splicing and the Spliceosome, and 3' Cleavage and Polyadenylation to follow.
- The Pol II CTD is a tandem repeat (Y1S2P3T4S5P6S7) of the heptapeptide, 52 repeats in humans, unstructured until modified.
- Cyclin-dependent kinases (CDK7, CDK9, CDK12/13) phosphorylate Ser5 and Ser2 at different transcription stages, creating a spatial/temporal phosphorylation gradient.
- Ser5-P (high near promoters) recruits the capping enzyme; Ser2-P (rising toward 3′ ends) recruits splicing and cleavage/polyadenylation factors.
- This "CTD code" is read by proteins with CTD-interaction domains (CIDs) that dock onto specific phospho-patterns, physically coupling transcription to RNA processing.
- Key evidence includes ChIP profiling of CTD phospho-marks across genes, CTD-truncation/mutation experiments, and structural work on CID-CTD complexes.
- The code is druggable and diagnostic: CDK9 inhibitors are anticancer leads, and CTD dysregulation is linked to neurodegeneration and cancer.
The Pol II CTD acts as a mobile, rewritable phosphorylation "barcode" that changes as the polymerase moves along a gene, recruiting capping, splicing, and 3′-processing machinery to the nascent RNA at exactly the right time and place.
The core idea
RNA polymerase II is built from twelve subunits, but only the largest subunit, RPB1, carries a unique structural feature absent from Pol I and Pol III: an extended, intrinsically unstructured tail projecting from its surface near the RNA exit channel, called the carboxy-terminal domain. The CTD is built from tandem repeats of a heptapeptide with the consensus sequence Tyr1-Ser2-Pro3-Thr4-Ser5-Pro6-Ser7. The repeat number scales roughly with genome complexity.
| Organism | CTD heptad repeats |
|---|---|
| Saccharomyces cerevisiae | 26 |
| Drosophila melanogaster | 42 |
| Homo sapiens | 52 |
Because the CTD has no fixed tertiary structure on its own, it behaves like a flexible scaffold that only acquires local, transient structure when bound by partner proteins or when heavily phosphorylated, which also drives liquid-liquid phase separation behavior important for forming transcriptional condensates. The core conceptual move is this: five of the seven residues in each heptad (Tyr1, Ser2, Thr4, Ser5, Ser7) can be reversibly phosphorylated, and Pro3/Pro6 can isomerize between cis and trans conformations. Different combinations of marks are installed and removed by specific kinases and phosphatases at specific stages of the transcription cycle—initiation, early elongation, productive elongation, and termination. Processing factors possess CTD-interacting domains (CIDs) or other phospho-CTD-binding modules that recognize particular phosphorylation patterns with modest but sufficient specificity. The result is that the identity of proteins riding on the polymerase changes as a function of where the polymerase is along the gene, not because of any external signal, but because of the CTD's own evolving chemical state. This is the essence of co-transcriptional coupling: RNA processing enzymes are recruited by reading the CTD, not by searching for RNA sequence motifs de novo in solution.
Mechanism
Step 1 — CTD is unphosphorylated at Pol II recruitment. When Pol II is assembled into the preinitiation complex (PIC) at a promoter, together with general transcription factors (TFIIA, B, D, E, F, H), its CTD is largely unphosphorylated (this hypophosphorylated form is historically called Pol IIA). TFIIB and Mediator make contacts that depend on this unmodified state; Mediator's interaction with the unphosphorylated CTD helps stabilize PIC assembly.
Step 2 — CDK7 (as part of TFIIH) phosphorylates Ser5 at initiation. TFIIH contains a kinase module (CDK7–cyclin H–MAT1 in humans, Kin28 in budding yeast) that phosphorylates Ser5 of the CTD heptad as Pol II clears the promoter. High Ser5-P marks the polymerase in the first ~50–100 nucleotides of transcription. This mark does two things simultaneously:
- It helps release Pol II from Mediator and the PIC, promoting promoter escape.
- It creates a docking surface for the mRNA capping enzyme.
Step 3 — Ser5-P recruits capping enzyme co-transcriptionally. The mRNA capping enzyme (in mammals, a bifunctional protein with RNA triphosphatase and guanylyltransferase activities, working with a separate methyltransferase) binds Ser5-phosphorylated CTD with high affinity, and low affinity for unphosphorylated or Ser2-phosphorylated CTD. Because the nascent transcript's 5′ end emerges from the exit channel right next to the CTD, capping enzyme recruited to Ser5-P CTD is positioned exactly where it needs to act—on an RNA of about 25–30 nucleotides, the point at which capping normally occurs.
Step 4 — Promoter-proximal pausing and the shift to Ser2 phosphorylation. In metazoans, Pol II frequently pauses 30–60 nucleotides downstream of the transcription start site, held by the negative elongation factors NELF and DSIF (Spt4/Spt5). Release into productive elongation requires P-TEFb (CDK9–cyclin T), which phosphorylates:
- Spt5 (converting DSIF from a pausing factor to an elongation factor)
- NELF (causing its dissociation)
- Ser2 of the CTD heptad
As elongation proceeds, CDK9 (and, in metazoans, the elongation-specific kinases CDK12 and CDK13) progressively increases Ser2 phosphorylation while CDK7-associated phosphatases (notably Rtr1/RPAP2 and Ssu72) begin removing Ser5-P. The result is a phosphorylation gradient along the length of a gene: ``` 5′ end: Ser5-P HIGH, Ser2-P LOW → capping machinery favored middle: Ser5-P falling, Ser2-P rising → splicing factors favored 3′ end: Ser5-P LOW, Ser2-P HIGH → cleavage/polyadenylation factors favored ```
Step 5 — Ser2-P and splicing factor recruitment. Components of the spliceosome and splicing regulators (including the essential factor Prp40 in yeast, and U2AF-associated factors and SR proteins in mammals) are recruited in part through interactions favored by rising Ser2-P and by the general architecture of the elongating polymerase, coupling intron removal to the pace of transcription. Splicing is also influenced by transcription elongation rate itself: slower Pol II elongation (e.g., caused by DNA sequences that intrinsically slow the enzyme, or by nucleosome barriers) favors inclusion of weak upstream exons because it gives spliceosomal components more time to assemble on the upstream splice site before a competing downstream site emerges—a phenomenon called kinetic coupling.
Step 6 — Ser2-P (and Tyr1 dephosphorylation) recruits 3′-end processing factors. As Pol II approaches the poly(A) signal (canonical hexamer AAUAAA) near a gene's 3′ end, high Ser2-P CTD, along with loss of Tyr1 phosphorylation, favors binding of the cleavage and polyadenylation specificity factor (CPSF) and cleavage stimulation factor (CstF). These assemble at the poly(A) site, cleave the transcript, and add a poly(A) tail; CTD-bound termination factors (including the Sen1/SETX helicase in the "torpedo" model) then help Pol II release from the DNA template, terminating transcription.
Step 7 — Phosphatases reset the cycle. Following termination, CTD phosphatases (Fcp1 removing Ser2-P, Ssu72 removing Ser5-P) strip the marks so that recycled Pol II can re-enter the unphosphorylated pool and be reused for another round of transcription initiation.
The evidence
The CTD-coupling model rests on several distinct, mutually reinforcing lines of experimental evidence, developed mostly between the late 1980s and the 2000s.
1. Discovery and essentiality of the CTD (Corden and colleagues, 1985). James Corden's laboratory cloned the gene for the largest subunit of mammalian Pol II and discovered it carried an unusual, highly conserved tandem repeat absent from Pol I and Pol III. Subsequent work showed that in yeast, truncating the CTD below roughly 8–10 repeats (out of 26) is lethal, and that viability requires a minimum repeat number, establishing that the CTD is not incidental but essential for normal Pol II function—this immediately raised the question of what essential job it performed.
2. Two Pol II forms differ by phosphorylation (IIA vs IIO). Early biochemical fractionation of Pol II identified two electrophoretically distinct forms: Pol IIA (hypophosphorylated, faster-migrating) and Pol IIO (hyperphosphorylated, slower-migrating due to the added mass and charge of many phosphates). Pol IIA is preferentially found in preinitiation complexes; Pol IIO is the elongating form found on gene bodies. This was an early clue that phosphorylation state changes with the polymerase's functional stage, not just its identity.
3. Kinase identification: CTD kinases map onto specific transcription stages. Genetic and biochemical work in yeast and human systems identified Kin28/CDK7 (part of TFIIH) as the kinase responsible for Ser5 phosphorylation at initiation, and Ctk1/CDK9 (P-TEFb) as responsible for Ser2 phosphorylation during elongation. Temperature-sensitive kin28 mutants in yeast lose Ser5-P and show defects specifically in capping enzyme recruitment and promoter escape—directly linking a specific kinase, a specific phospho-mark, and a specific downstream processing step.
4. ChIP mapping of CTD phospho-marks across gene bodies (Komarnitsky, Cho & Buratowicz, 2000; Ahn, Kim & Buratowski, and others in the early 2000s). Using phospho-specific antibodies (notably the H14 antibody for Ser5-P and H5 for Ser2-P) in chromatin immunoprecipitation (ChIP) experiments across yeast and human genes, researchers showed directly that Ser5-P is enriched at the 5′ ends of genes and declines toward the 3′ end, while Ser2-P is low at the 5′ end and rises steadily toward the 3′ end. This gave the phosphorylation-gradient model its quantitative, genome-mapped backbone: it is not a hypothesis but a measured ChIP profile reproduced across thousands of genes in later genome-wide (ChIP-seq) studies.
5. Direct biochemical recruitment of capping enzyme by Ser5-P CTD (Cho, Takagi, Moteki & Buratowski, 1997; McCracken et al., 1997; Ho and Shuman, 1999). In vitro binding assays showed that mammalian and yeast capping enzymes bind a Ser5-phosphorylated CTD peptide substantially more tightly than an unphosphorylated or Ser2-phosphorylated peptide, and that CTD phosphorylation stimulates capping enzyme's guanylyltransferase activity. Structural studies later showed the capping enzyme's CTD-interacting domain directly contacts phosphates on Ser5.
6. CTD truncation blocks 3′-end processing (McCracken et al., 1997, Nature). In a landmark paper, transcripts made by a truncated Pol II lacking a functional CTD were shown to be capped inefficiently, spliced poorly, and—critically—failed to be properly cleaved and polyadenylated at their 3′ ends, even though 3′-processing factors were present and active in the same nuclear extract. This experiment is often treated as the single clearest demonstration that the CTD itself, not merely Pol II catalytic activity, is required for coupling RNA processing to transcription: the same polymerase enzymatic activity minus its tail cannot support proper mRNA maturation.
7. Structural biology of CTD-reader proteins. Crystal structures of CTD-interacting domains (CIDs) bound to phosphopeptides—for example, the CID of the yeast termination factor Pcf11, and of capping enzyme guanylyltransferase domains bound to Ser5-P peptides—show a shallow, positively charged groove that specifically accommodates the phosphate on Ser5 or Ser2 through hydrogen bonds and electrostatic contacts, explaining at atomic resolution how different phospho-marks are read with the right specificity by different partners. Together, these lines of evidence—genetic essentiality, biochemical fractionation, kinase-specific mutants, genome-wide ChIP mapping, in vitro reconstitution, and CTD-reader structures—converge on the same model: Pol II's CTD carries a spatially and temporally regulated phosphorylation code, and different processing machines are recruited by reading that code as transcription proceeds.
Worked examples
Example 1 — Interpreting a ChIP-seq metagene profile. Suppose a ChIP-seq experiment measures Ser5-P and Ser2-P signal across a 3,000 bp gene (transcription start site, TSS, at position 0; poly(A) site at position 3,000), and gives the following (idealized) relative signal at four positions:
| Position (bp from TSS) | Ser5-P signal | Ser2-P signal |
|---|---|---|
| +100 | 100 | 8 |
| +800 | 55 | 40 |
| +1800 | 25 | 70 |
| +2900 | 10 | 95 |
Question: A student proposes that capping enzyme should be recruited most strongly near +1800, since that is roughly the gene's midpoint. Is this correct, and why or why not? Reasoning: Capping enzyme is recruited by Ser5-P, not by gene midpoint per se. The data show Ser5-P peaking near the TSS (+100: signal 100) and declining steeply by +800 (signal 55, roughly a 45% drop) and further by +1800 (signal 25, a 75% drop from the peak). Capping itself is a very early, essentially co-transcriptional event that occurs once the nascent transcript reaches ~25–30 nucleotides—well before +800, let alone +1800. So the student's proposal is incorrect on two counts: mechanistically, capping is triggered by Ser5-P and RNA length, not by gene midpoint, and quantitatively, Ser5-P is already sharply declining well before the gene's middle. The Ser2-P profile, rising from 8 at +100 to 95 at +2900 (roughly a 12-fold increase), instead predicts that cleavage/polyadenylation factor recruitment (which depends on Ser2-P) should be strongest near the 3′ end, consistent with the biology of 3′-end processing occurring at the poly(A) site.
Example 2 — Predicting the effect of a CDK9 inhibitor. A researcher treats cells with flavopiridol, a potent CDK9 (P-TEFb) inhibitor, and examines nascent transcription by nuclear run-on and RNA-seq of chromatin-associated RNA. Question: Predict the effects on (a) promoter-proximal Pol II density, (b) Ser2-P levels genome-wide, (c) capping efficiency, and (d) 3′-end processing, and explain the logic for each. Reasoning:
- (a) Promoter-proximal Pol II density should increase. CDK9 phosphorylates NELF and Spt5 (DSIF) to release paused Pol II into productive elongation. Blocking CDK9 traps Pol II in the paused state 30–60 nt downstream of the TSS, so ChIP or nuclear run-on signal for Pol II should pile up at promoter-proximal regions and decline in gene bodies—this is in fact the classical experimental signature used to identify pausing and P-TEFb dependence.
- (b) Ser2-P should drop sharply genome-wide, particularly in gene bodies and toward 3′ ends, because CDK9 (along with CDK12/13, which are less sensitive to flavopiridol at low doses but still partly affected) is the principal Ser2 kinase; CDK7-driven Ser5-P near promoters should be largely unaffected since it is installed by a different kinase.
- (c) Capping efficiency should be relatively unaffected in the short term, since capping enzyme recruitment depends on Ser5-P (CDK7-dependent) and occurs very early in transcription, upstream of where P-TEFb normally acts; however, because Pol II fails to escape the pause site efficiently, fewer transcripts overall reach productive elongation, indirectly reducing the number of capped, full-length mRNAs even though the capping mechanism itself is intact.
- (d) 3′-end cleavage/polyadenylation should be strongly impaired for genes that require productive elongation to reach their poly(A) sites, both because Ser2-P (needed for CPSF/CstF recruitment) is depleted and because Pol II rarely reaches the 3′ end at all if it is stalled near the promoter.
This exercise illustrates how the CTD code translates a single pharmacological perturbation (CDK9 inhibition) into a coherent, multi-step, predictable set of downstream molecular consequences—precisely the logic exploited by CDK9 inhibitors now in clinical development for cancer.
Common misconceptions & traps
| Misconception | Why it's wrong |
|---|---|
| "The CTD code is a fixed, permanent modification set on one Pol II molecule." | It is dynamic: kinases and phosphatases continuously add and remove marks as the same polymerase molecule moves along the gene; the pattern is positional, not static. |
| "Capping, splicing, and 3′ processing are separate steps that happen after transcription is finished." | All three are substantially co-transcriptional; they occur on the nascent RNA while Pol II is still elongating, coordinated in real time via the CTD. |
| "Ser2-P and Ser5-P are the only important CTD marks." | Tyr1-P, Thr4-P, and Ser7-P are also biologically important (e.g., Tyr1-P suppresses premature termination factor binding; Ser7-P is required for snRNA gene 3′-end processing); the code is richer than a simple two-mark binary. |
| "All 52 human heptad repeats are identical and interchangeable." | The repeats are only close to consensus in the N-terminal half; the C-terminal repeats are progressively degenerate and have distinct, non-redundant functions, including some roles independent of phosphorylation. |
| "CTD phosphorylation state is set once at initiation and just persists." | Removal is just as important as addition: phosphatases like Ssu72, Fcp1, and RPAP2 actively reset the CTD, and their mistiming causes processing defects just as kinase mistiming does. |
The scientist behind it
Origins · 1 minJames L. Corden and how this was discovered
James Corden's laboratory in the mid-1980s cloned the gene encoding the largest subunit of eukaryotic RNA polymerase II and discovered, buried at its C-terminus, an extraordinary tandemly repeated heptapeptide sequence with no precedent in any known protein domain at the time. Corden recognized immediately that a domain this repetitive and this conserved across species as distant as yeast and mammals was unlikely to be accidental, and his subsequent genetic work in yeast—showing that truncating the repeat number below a threshold was lethal—established the CTD as an essential functional module rather than an evolutionary vestige. This opened three decades of work, contributed to substantially by Stephen Buratowski, David Bentley, Danny Reinberg, Jesper Svejstrup, Roger Kornberg's structural biology of Pol II, and many others, that gradually decoded what the repeats actually do: serve as a phosphorylation-encoded, positionally informative scaffold for RNA processing machinery.
Where this matters today
Cancer therapy. CDK9 (P-TEFb) and its associated cyclin T are hyperactivated or required in several cancers, particularly those driven by MYC, which depends on efficient Pol II pause release. CDK9 inhibitors (including flavopiridol/alvocidib and newer selective compounds) are in clinical trials, exploiting the fact that oncogene-driven transcription is unusually dependent on productive Ser2 phosphorylation and pause release. CDK12/13 inhibitors are being explored particularly in cancers with defective DNA damage response, since CDK12 also promotes transcription of long DNA-repair genes.
Antiviral and antiparasitic drug design. Several viruses (notably HIV, via the Tat protein, and various herpesviruses) hijack P-TEFb to drive their own transcription; understanding CTD-kinase coupling has informed strategies to block viral gene expression by targeting this hijacked machinery.
Neurodegenerative disease. Mutations and aggregation of RNA-binding and CTD-interacting proteins (including in ALS/FTD, where the CTD-associated splicing and processing machinery is disrupted) implicate defective co-transcriptional coupling in neurodegeneration; the CTD's phase-separation behavior, driven partly by its repetitive, low-complexity phosphorylation-dependent sequence, is an active area of research into transcriptional condensates and their dysregulation in disease.
Biotechnology and synthetic biology. Engineering CTD phosphorylation states or CTD-reader domains is used to build synthetic transcription-coupled reporters and to design orthogonal RNA processing circuits, exploiting the same code logic the cell uses natively.
Diagnostic and mechanistic research tools. Phospho-CTD-specific antibodies (Ser2-P, Ser5-P) remain a standard tool in ChIP-seq/PRO-seq experiments to distinguish paused versus elongating polymerase genome-wide, informing studies of gene regulation across nearly all of modern molecular biology, from developmental gene regulatory networks to cancer transcriptomics.
Ask the tutor: How does promoter-proximal pausing interact with the Ser5-P-to-Ser2-P transition to control which genes are rapidly inducible?
Why does Pol II need a CTD at all—why can't processing factors just find nascent RNA on their own?
Diffusion-limited search for a specific RNA sequence or structure amid the crowded nucleus would be slow and error-prone, especially for events (like capping) that must happen almost immediately after transcription begins, within the first ~25–30 nucleotides. By tethering processing factors directly to the transcribing enzyme via the CTD, the cell effectively pre-positions each machine exactly where and when it is needed, converting a search problem into a delivery problem. The CTD also enforces order: capping factors are favored early (Ser5-P high), 3′-processing factors late (Ser2-P high), preventing, for example, premature polyadenylation before the transcript is complete. This co-transcriptional strategy is faster, more accurate, and inherently coordinates multiple processing events on the same molecule.
Is the CTD phosphorylation pattern the same on every gene, or does it vary?
The general shape of the gradient—high Ser5-P near the promoter, rising Ser2-P toward the 3′ end—is broadly conserved across most protein-coding genes, but the fine details vary substantially. Gene length, the strength and duration of promoter-proximal pausing, intron number and position, and local chromatin context all modulate exactly how sharply and quickly the transition occurs. Some genes show unusually prolonged pausing (common at genes regulated for rapid inducible responses, such as heat-shock genes), which changes the local CTD phospho-profile. Because of this variability, ChIP-seq or PRO-seq experiments typically report metagene-averaged profiles across thousands of genes to reveal the general trend, while acknowledging gene-to-gene heterogeneity.
What are Tyr1 and Thr4 phosphorylation used for, if Ser2 and Ser5 do most of the work?
Tyr1 phosphorylation, installed by non-canonical CTD kinases, is enriched broadly across gene bodies and appears to suppress premature binding of termination factors (such as Pcf11), helping keep Pol II in productive elongation mode until it reaches the correct 3′ end. Thr4 phosphorylation is comparatively less abundant but has been specifically implicated in histone gene 3′-end processing (a specialized non-polyadenylated pathway) in some systems. These marks show that the "two-letter" Ser2/Ser5 model, while a good first approximation, is a simplification; the fuller CTD code involves at least five modifiable residues combined combinatorially, and research is still refining what each additional mark contributes.
How does the CTD code connect to alternative splicing decisions?
Splicing factor recruitment via Ser2-P-rich CTD is necessary but not sufficient to explain alternative splicing outcomes; the rate at which Pol II elongates through a gene is an equally important variable. If Pol II pauses or elongates slowly over a weak, upstream splice site, spliceosomal components have more time to assemble there before a stronger downstream splice site becomes transcribed and available, favoring inclusion of that upstream exon. Because CDK9/CDK12 activity, chromatin structure, and DNA sequence-encoded pause sites all influence elongation rate, and because CTD phosphorylation state is intertwined with elongation rate, the CTD code and transcription kinetics act together—not the CTD phospho-mark alone—to shape which exons are ultimately included.
What would happen to a cell if Ssu72 (a Ser5-P phosphatase) were completely inactivated?
Loss of Ssu72 activity would be expected to cause persistent, inappropriately high Ser5-P levels extending further into gene bodies and toward 3′ ends than normal, since the mark would not be efficiently removed. Because Ssu72 also participates directly in 3′-end processing complexes and in gene looping (bringing the promoter and terminator into proximity), its loss disrupts proper termination and reduces the efficiency of Pol II recycling back to a hypophosphorylated (Pol IIA) state needed for new rounds of initiation. In practice, Ssu72 is essential in yeast, and its loss causes widespread defects in transcription termination and 3′-end formation, consistent with the CTD code depending on regulated removal of marks just as much as their addition.
Does the CTD do anything besides recruit RNA-processing enzymes?
Yes. The CTD also helps recruit chromatin-modifying enzymes co-transcriptionally—for example, Ser2-P CTD helps recruit the Set2 histone methyltransferase in yeast, which deposits H3K36 methylation in transcribed gene bodies, a mark that in turn influences splicing fidelity and suppresses spurious intragenic transcription initiation. The CTD's low-complexity, highly repetitive, phosphorylation-tunable sequence also promotes liquid-liquid phase separation, contributing to the formation of transcriptional condensates that concentrate Pol II, Mediator, and processing factors together at active genes. So the CTD functions as a general-purpose, dynamically modified communication hub linking transcription not just to RNA maturation but to chromatin state and nuclear organization.