The early stages of RNA Pol II transcription elongation are marked by several events that are linked to CTD Ser5 phosphorylation (8
). These events include mRNA capping, higher levels of histone H3K4 methylation, and an early termination pathway in yeast that is mediated by Nrd1, Nab3, and Sen1. The Nrd1 complex mediates both transcription termination and the subsequent recruitment of the nuclear exosome for snoRNA 3′-end trimming (51
). This complex can be recruited to genes via multiple mechanisms, including the recognition of specific RNA sequences by Nrd1 and Nab3, or by the interaction of the Nrd1 CID with CTD S5P. It is likely that individual genes will vary in being more or less dependent on each of these recruitment pathways.
Here we demonstrate that H3K4 trimethylation, a mark associated with active transcription, also contributes to efficient termination by the Nrd1/Sen1 pathway. The deletion of SET1 either by itself or in an nrd1 mutant background enhances the termination read-through of several snoRNAs and CUTs tested (A and D, F, , and ). Although these effects are much less than those seen for mutants of the actual termination factors themselves, our results indicate that chromatin modifications near promoters can influence the choice of whether to terminate soon after initiation or to continue elongating.
Interestingly, mutations in the Paf1 complex have also been reported to cause snoRNA termination defects by reducing the recruitment of the Nrd1 complex (48
). Since the Paf1 complex is required for proper H3K4 methylation by Set1, it is logical to ask whether the Paf1 complex effect might be mediated via Set1. Sheldon et al. (48
) concluded that this was unlikely because snR13 read-through transcripts were not seen in a SET1
deletion. While this clearly suggests that the Paf1 complex has additional functions related to snoRNA termination, our observations of read-through transcripts in set1
Δ cells make it worth revisiting the idea that some of the Paf1 complex occurs via a disruption of H3K4 methylation. We note that while this paper was under review, a follow-up study by the Arndt laboratory showed that the termination defect observed for Paf1 complex mutants is likely mediated via its effect on H2B ubiquitylation (57
), a prerequisite for H3K4 trimethylation.
The contribution to Nrd1-dependent termination by Set1 is clearly related to its histone methyltransferase activity. Histone H3 mutated at lysine 4 (F) or a catalytic-site mutant of Set1 (A to D) produced effects similar to those seen in set1
Δ cells. It appears that it is specifically H3K4 trimethylation that promotes Nrd1-dependent termination. The monoubiquitylation of H2BK123 by the Rad6/Bre1 complex, which is required for H3K4 di- and trimethylation by Set1 (29
), also promotes efficient Nrd1-dependent transcription termination (E) (57
). Similarly, the set1
Δ) allele, which abolishes H3K4 trimethylation but not H3K4 dimethylation (see Fig. S2 in the supplemental material) (18
), also exhibits a reduction of Nrd1 recruitment with subsequent termination defects (C and D and B). Given that snoRNAs and CUTs are terminated relatively close to the promoter, it makes sense that promoter-proximal H3K4 trimethylation would be linked to the Nrd1-Nab3-Sen1 termination pathway.
Set1 may promote SNR13
termination by affecting histone acetylation levels. Because the deletion of PHO23
also leads to read-through, we suspect that this effect occurs via the Rpd3L HDAC. Although sequence-specific DNA binding proteins such as Ume6 can recruit Rpd3L to repress specific promoters (12
), we speculate that the PHD fingers of the Rpd3L subunits Pho23 and Cti6 can bind H3K4me3 to moderate acetylation levels at active promoters. This model is supported by several findings. First, cells lacking Pho23 show a small but significant increase in acetylation at the SNR13
(B) and YEF3 (18
) promoters. Additionally, Rpd3 recruitment was observed in SNR13
regions (C and see Fig. S3A in the supplemental material) (data from reference 7
) where pho23
Δ results in termination defects (A and A). Another genome-wide ChIP analysis of Rpd3 showed that, in addition to being recruited to a small group of repressed promoters, the deacetylase was unexpectedly found to cross-link to active promoters proportional to their transcription rate (27
). At the time, those authors speculated that Rpd3 might be recruited to active promoters to dampen high levels of acetylation. A similar story emerged from genome-wide mapping of HATs and HDACs in mammalian cells. HDACs are recruited to active genes, and levels correlated positively with H3K4 methylation (63
). Therefore, both HATs and HDACs may functionally associate with active promoters. Supporting this view, we show that the blocking of NuA3 HAT recruitment by the deletion of YNG1
reverses the enhancement of the termination defect by pho23
Δ (D). Although it may appear paradoxical for H3K4me3 to recruit the antagonistic activities of both NuA3 (via Yng1) and Rpd3L (via Pho23 and Cti6) to the promoter, this mechanism may maintain balanced levels of histone acetylation or perhaps allow cycles of acetylation and deacetylation that are important for proper transcription.
In yeast, the choice between early Nrd1-dependent termination or continued elongation to downstream polyadenylation sites must be made for both snoRNAs and mRNA genes (8
). Set1 may influence this decision point by affecting the kinetics of early transcription (). This idea is supported by the sensitivity of set1
Δ or pho23
Δ cells to 6-AU as well as the genetic interaction with a RPB2
allele that slows elongation. The loss of H3K4 trimethylation may make it slightly harder for Pol II to travel through the promoter-proximal chromatin environment, perhaps due to downstream effects on nucleosome acetylation or remodeling. A slower transit time through this window might affect the modifications of Pol II or the recruitment of necessary elongation factors, resulting in termination defects for snoRNAs and CUTs. The recent observation that another slow Pol II allele, rpb1
), shows reduced capping efficiency (21
) supports this concept. Although recent reports failed to detect widespread mRNA changes in the absence of Set1 (18
), it is possible that a subset of genes could be affected by this mechanism.
A choice between early or late termination pathways might also underlie the regulation of genes in higher eukaryotes, where higher levels of Pol II accumulate near the 5′ ends of genes relative to further downstream (34
). While some of these promoter-proximal polymerase molecules may be paused, others may be destined for early termination (8
). It was recently found that bidirectional transcription is an inherent feature of many actively transcribed genes (reviewed in reference 9
). In yeast, the reverse-direction promoter-associated transcripts are terminated by the Nrd1-Nab3-Sen1 complex (40
). It will be interesting to see whether H3K4 methylation also plays a role in mediating early pausing and termination in metazoans.