A first step in gene expression is the recruitment of the DNA-transcribing enzyme RNA polymerase II (Pol II) to a gene, and the assembly of transcriptional machinery around it. Pol II can then initiate RNA synthesis. However, during transcription of most mammalian genes, Pol II does something peculiar — after synthesizing a short RNA molecule usually no longer than 60 nucleotides, it stops, awaiting further instructions before transcribing the remainder of the gene1. Such pausing and subsequent RNA elongation is central to gene regulation in animals, yet the mechanisms underlying this process have not been clear. In two papers in Nature, Vos et al.2,3 describe landmark structures that shed new light on Pol II pausing and release.
A heterodimer comprising the proteins SPT4 and SPT5 is crucial for the pausing of Pol II4. During transcription initiation, general transcription factors bind and occlude the regions of Pol II recognized by SPT5 — these factors must be released before SPT5 can associate. Thus, SPT5 binding occurs after transcription proper begins, and stable interactions between SPT5 and Pol II require a nascent RNA about 20 nucleotides in length to have formed5. Interactions with transcribing Pol II then enable SPT5 to recruit additional factors that govern Pol II activity and RNA processing4,5. One such factor is the negative elongation factor (NELF) protein complex, which comprises four subunits (NELF-A, -B, -C and -E)4.
In contrast to SPT5, which is evolutionarily conserved from bacteria all the way through to humans, no equivalents to the mammalian NELF proteins have been identified in bacteria, yeast, worms or plants4. The organisms that do contain a NELF complex are those that exhibit stable pausing of Pol II, implying a role for NELF in this process. Indeed, the release of NELF from Pol II is concomitant with escape from pausing into elongation1, and acute depletion of NELF both prevents normal pausing6 and increases premature termination7 (the process whereby Pol II inadvertently releases DNA, ceasing transcription). But the molecular basis of NELF activity has remained obscure. In particular, it has been unclear how NELF interacts with Pol II and how it might stabilize the paused state in a manner that prevents both continued RNA synthesis and transcription termination.
In the first of their papers, Vos et al.2 used cryo-electron microscopy to resolve the structure of a paused transcription complex at 3.2-ångström resolution. The authors assembled a highly purified structure on an artificial DNA–RNA scaffold that contains sequences known8 to strongly promote Pol II pausing, using pig Pol II along with human SPT5 and NELF complexes. The Pol II–SPT5–NELF complexes formed on this scaffold showed clear differences compared with previously published Pol II–SPT5 complexes in an actively transcribing conformation9. Whereas the DNA–RNA hybrid held within active Pol II has an unpaired DNA base that can be used as a template to direct addition of the next RNA nucleotide, the DNA–RNA hybrid in the paused complex is ‘tilted’ and lacks unpaired template DNA. Without a free DNA base in its active site, Pol II is unable to carry out RNA elongation.
This non-productive DNA–RNA hybrid conformation alone explains why Pol II pauses. But more importantly, the structure also reveals the role of NELF in this process. The researchers found that a protein lobe comprising the NELF-A and NELF-C subunits binds near a funnel region in Pol II through which nucleotides normally access the active site. The NELF lobe protrudes into the funnel, potentially restricting the entry of nucleotides needed for transcription. In addition, NELF restrains mobile loop domains in Pol II, such as the trigger loop, near the active site. This restraint locks the enzyme in the inactive conformation while simultaneously discouraging Pol II from sliding along the DNA, which can lead to transcription termination.
The NELF binding pocket near the Pol II funnel overlaps with a region that, when not occluded, can be bound by the factor TFIIS to stimulate elongation. Intriguingly, TFIIS has been shown to reactivate Pol II that adopts a non-productive, tilted DNA–RNA hybrid conformation10. Thus, Vos et al. propose that NELF also prevents Pol II reactivation by blocking TFIIS binding (Fig. 1).
The release of paused Pol II into elongation is triggered by the recruitment of the kinase enzyme P-TEFb, which phosphorylates Pol II and pause-inducing factors, triggering dissociation of NELF1. P-TEFb activity is accompanied by the recruitment to Pol II of the SPT6 protein and the polymerase-associated factor (PAF) protein complex. However, whether these elongation-associated factors directly affect Pol II pause release has been unclear. In the second of the papers, Vos et al.3 examined this possibility by assembling a structure that included a modified, elongation-permissive nucleic-acid scaffold and these activating proteins.
As anticipated, the DNA–RNA hybrid in the activated elongation complex is no longer tilted and adopts a conformation compatible with RNA synthesis. The authors found multiple sites phosphorylated by P-TEFb in both SPT5 and NELF. Phosphorylation at these sites might aid the opening of the interface between Pol II and SPT5, and lead to dissociation of NELF. Furthermore, the group showed that phosphorylation of SPT6 and a linker region in the carboxy-terminal domain of Pol II aided docking of SPT6 on the enzyme. Most strikingly, the structure revealed that the binding of NELF and PAF to Pol II is mutually exclusive. Thus, dissociation of NELF during pause release enables the binding of PAF as well as TFIIS, allowing transcription to proceed.
Taking these results together, a detailed molecular model of Pol II pausing and release begins to emerge. We note a recurring theme wherein mutually exclusive, overlapping binding sites for a succession of Pol II-associated factors enable an orderly exchange during the transcription cycle. Furthermore, the specificity of each protein’s interaction with the Pol II complex is ensured by multiple interaction interfaces, often with scaffold proteins such as SPT5 and the nucleic acids.
Of course, questions remain about the transition from pausing to productive elongation. For example, this work calls into question the roles of RNA-binding domains found in NELF subunits4. Surprisingly, Vos et al. showed that disruption of one such domain in NELF-E had no effect on pausing. It also remains to be seen whether the tilted DNA–RNA conformation observed by the authors is prevalent in vivo, and how the phosphorylation of pause-inducing factors drives pause release.
This work represents a fundamental jump in our understanding of pausing. The structures point to several appealing models for regulated pause release that can be tested in future work.
Nature 560, 560-561 (2018)