Transcription by yeast Pol II terminates by at least two mechanisms: the Rat1-dependent ‘torpedo’ pathway and the Sen1–Nrd1 pathway13
. The Rat1 pathway works at mRNA genes, whereas the Sen1–Nrd1 pathway functions at snoRNAs, CUTs and some short mRNAs13,23,28,30,41
. Both pathways involve interactions between the Pol II CTD and CID proteins. Rtt103 and Pcf11 function in the mRNA pathway and preferentially bind CTD Ser2P10–12,14
. Pcf11, which also binds nonphosphorylated CTD, is required for both mRNA and snoRNA termination13,37,42
, suggesting a function that is common for both pathways.
Here we show that the Nrd1 CID resembles the Pcf11 CID structurally, but has a different phosphorylation preference for the CTD. In vitro, Nrd1 binds strongly to CTD-Ser5P and slightly better to CTD-Ser2P/Ser5P. However, several findings indicate that Ser2 phosphorylation is not crucial in vivo. Pol II associated with Nrd1 in vivo reacts with antibody H14 (recognizing Ser5P), but not H5 (primarily recognizing Ser2P). Furthermore, whereas the Nrd1 CID is required for recruitment of Nrd1 to the 5′ ends of genes, deletion of the Ser2 kinase Ctk1 has no effect. Therefore, we conclude that Ser5P is the primary determinant of CTD interaction for Nrd1 in vivo.
The unexpected specificity of Nrd1 for CTD-Ser5P, the promoter-proximal phosphorylation state4
, explains several observations. Whereas both Rtt103 and Pcf11 cross-link at 3′ ends of Pol II– transcribed genes, Nrd1 cross-links strongly at 5′ ends and, to some extent, at 3′ ends ()13,43
. Mutation of Sen1 causes termination defects at snoRNA genes and mRNA genes shorter than 600 nt23
. Both Sen1 and Nrd1 are necessary for suppression of CUTs, the short unstable transcripts produced by cryptic promoters throughout the yeast genome23,29–32
. When a Nrd1-dependent terminator sequence is moved further downstream, where Ser2P predominates and Ser5P levels are lower, it no longer functions properly44,45
The different CTD specificities of Nrd1, Pcf11 and Rtt103 are also consistent with genetic observations suggesting that the Ser2 kinase Ctk1 acts in opposition to the Nrd1–Sen1–Nab3 complex18
. A cold-sensitive allele of NAB3
is suppressed by deletion of CTK1
. Nab3 overexpression exacerbates cold sensitivity caused by CTK1
deletion, whereas the nrd1-102
allele weakly suppresses ctk1
. Finally, increasing CTD-Ser2P levels by mutating the CTD phosphatase Fcp1 increases levels of read-through transcripts at a Nrd1-dependent terminator45
. These observations suggest competition between the two termination pathways, with Ser5P early in elongation favoring the Sen1 pathway via Nrd1 and Ser2P at later times favoring the poly-adenylation/torpedo pathway via Rtt103 and Pcf11.
It is unclear what leads the CIDs of Pcf11 and Nrd1 to have different specificities. The Nrd1 CID is similar to Pcf11 in overall conformation, and a central CTD binding pocket seems to be conserved11
. A superposition of Nrd1 and Pcf11 was used to model a possible Nrd1-CTD interaction ( and Supplementary Fig. 1
). Mutagenesis studies indicate that conserved residues in the CID pocket are necessary for Nrd1 binding to the CTD ( and ). Binding of the CTD in this conserved pocket may be independent of the CTD-phosphorylation status. Although Pcf11 binding to the CTD is enhanced by Ser2P, Pcf11 also binds unphosphorylated and doubly phosphorylated CTD10
. In the Pcf11 structure, the single observed Ser2 phosphate does not contact the CID11
. The Ser2P preference may be in part due to a hydrogen bond between the CTD Ser2 phosphate and the CTD Thr4 side chain that stabilizes the β-turn11
, but the unphosphorylated CTD shows an intrinsic propensity to form β-turns within the Ser2-Pro3-Thr4-Ser5 motif 46
. Thus, the specificity of CIDs for different CTD-phosphorylation sites may be determined by additional contacts outside the central binding pocket. This idea is supported by recent cocrystal structures of phosphorylated CTD bound to the CID of the mammalian SCAF8 protein, where CID surface residues directly contact the CTD phosphates47
. A sulfate ion bound to Nrd1 may represent a Ser5P CTD interaction site (). Consistent with this idea, point mutations in this region reduce affinity for CTD-Ser5P peptides ().
CTD binding is only one of several mechanisms for recruiting the Nrd1–Nab3–Sen1 complex to RNAs (Supplementary Fig. 5
online).Neither the CID nor the Nab3 interaction domain is essential for viability, but both contribute to interaction with the polymerase and Nrd1 recruitment ( and ). There may be partial redundancy, because deletion of both domains is lethal18
(). Both Nrd1 and Nab3 are sequence-specific RNA binding proteins that can be targeted to specific transcripts carrying the appropriate recognition sequences13,16,20,21,26,28,30,40,41,44
. This may explain cross-linking of Nrd1 observed to regions downstream from the promoter, where Ser5P levels are likely to be lower.
Once targeted to the RNA, the Nrd1–Nab3–Sen1 complex terminates transcription by a mechanism that may involve the helicase activity of Sen1. Sen1-mediated termination is coupled to RNA 3′ processing and degradation events mediated by the TRAMP–exosome complex. We previously demonstrated a physical interaction between the exosome–TRAMP and Nrd1 complexes and showed that this interaction recruits the exosome to RNAs containing Nrd1 binding sites21
. At snoRNAs, the recruitment of exosome results in 3′ end trimming21
. For CUTs and certain mRNAs, this pathway results in complete degradation of the transcript (reviewed in ref. 29
It remains to be seen how the S. cerevisiae
Nrd1–Sen1–exosome pathway relates to gene expression in higher eukaryotes. Metazoan genomes have multiple CID proteins, several of which also carry RRMs. Furthermore, there is a mammalian Sen1-like protein called Senataxin that has been implicated in several ataxia syndromes. In mammals, most snoRNAs are processed from mRNA introns, so this pathway may be used primarily for termination and degradation of cryptic transcripts rather than for snoRNA biogenesis. Recent transcript mapping and Pol II cross-linking studies in higher eukaryotes suggest that transcription is surprisingly widespread throughout the genome and that most of these transcripts do not correspond to coding genes or stable noncoding transcripts48
. Therefore, suppression of cryptic transcription by co-transcriptional targeting of termination and degradation machineries may be even more important in higher eukaryotes.