|Home | About | Journals | Submit | Contact Us | Français|
Across diverse eukaryotes, the Paf1 complex (Paf1C) plays critical roles in RNA polymerase II transcription elongation and regulation of histone modifications. Beyond these roles, the human and Saccharomyces cerevisiae Paf1 complexes also interact with RNA 3′-end processing components to affect transcript 3′-end formation. Specifically, the Saccharomyces cerevisiae Paf1C functions with the RNA binding proteins Nrd1 and Nab3 to regulate the termination of at least two small nucleolar RNAs (snoRNAs). To determine how Paf1C-dependent functions regulate snoRNA formation, we used high-density tiling arrays to analyze transcripts in paf1Δ cells and uncover new snoRNA targets of Paf1. Detailed examination of Paf1-regulated snoRNA genes revealed locus-specific requirements for Paf1-dependent posttranslational histone modifications. We also discovered roles for the transcriptional regulators Bur1-Bur2, Rad6, and Set2 in snoRNA 3′-end formation. Surprisingly, at some snoRNAs, this function of Rad6 appears to be primarily independent of its role in histone H2B monoubiquitylation. Cumulatively, our work reveals a broad requirement for the Paf1C in snoRNA 3′-end formation in S. cerevisiae, implicates the participation of transcriptional proteins and histone modifications in this process, and suggests that the Paf1C contributes to the fine tuning of nuanced levels of regulation that exist at individual loci.
Many proteins contribute to regulating RNA polymerase II (RNA Pol II) transcription to ensure accurate RNA synthesis. One such group of regulatory proteins is the conserved eukaryotic Paf1 (polymerase-associated factor 1) complex (Paf1C) (1, 2). The Paf1C is comprised of the Paf1, Cdc73, Ctr9, Rtf1, and Leo1 subunits in Saccharomyces cerevisiae (3–6). Originally discovered by virtue of its interaction with RNA Pol II, the yeast Paf1C associates with RNA Pol II at actively transcribed open reading frames (ORFs) from the start site of transcription to the poly(A) site (4, 7–9). The full recruitment of the yeast Paf1C to active genes requires the protein Spt5, which is phosphorylated by the Bur1-Bur2 cyclin-dependent kinase/cyclin (CDK-cyclin) complex (10–12). The human Paf1C also contains the Ski8 protein, which localizes to transcriptionally active genes in a manner dependent on the rest of the complex (13). The evolutionary conservation of the Paf1C and its important functions allow extrapolation of budding yeast studies to higher eukaryotes.
The best-characterized functions of the Paf1C are in regulating transcription elongation and promoting histone modifications. The yeast and human Paf1 complexes can stimulate efficient transcription elongation in vitro and in vivo (14–17). Transcription through a chromatin template is further regulated by the Paf1C-dependent posttranslational modification of histones. For example, the trimethylation of histone H3 at lysine 36 (K36) by the methyltransferase Set2 requires the Paf1C, mediated primarily through the Paf1 and Ctr9 subunits (18). The monoubiquitylation of yeast histone H2B K123, which is dependent on the ubiquitin-conjugating enzyme Rad6 and the ubiquitin protein ligase Bre1, also requires the Paf1C (19, 20). H2B K123 monoubiquitylation is required in turn for both Set1-mediated di- and trimethylation of histone H3 K4 and Dot1-catalyzed histone H3 K79 methylation (19–24). Analogously, the human Paf1C facilitates human Rad6 (hRad6) and human Bre1 (hBre1) recruitment leading to the monoubiquitylation of H2B K120, a mark that also engages in histone modification cross talk by facilitating hDot1-dependent methylation of H3 K79 and hSet1/MLL1-dependent methylation (MLL stands for mixed lineage, leukemia) of H3 K4 (13, 25). Found at sites of active transcription, these Paf1C-dependent histone modifications can affect recruitment of other proteins that participate in RNA Pol II transcription (23, 26–29). The function of hBre1/Rnf20 as a tumor suppressor highlights the clinical importance of these histone marks (30). The Paf1C in higher eukaryotes also has gene-specific functions, such as regulating the expression of genes dependent on the Wnt, Notch, and Hedgehog signaling pathways (31–34). These many roles may explain why perturbations of Paf1C in higher eukaryotes can alter stem cell pluripotency, development, antiviral responses, and cancer progression (31, 32, 34–40).
Much less understood is the function of the Paf1C in RNA transcript termination and 3′-end formation. In both yeast and humans, the Paf1C has been shown to interact with RNA cleavage and polyadenylation factors (41, 42). In yeast, the Paf1C has also been shown to affect poly(A) site utilization and the proper 3′-end formation of small nucleolar RNAs (snoRNAs) (43, 44). snoRNAs constitute an important class of noncoding RNAs (ncRNAs), which function during rRNA processing reactions and have been implicated in tumorigenesis and diseases, such as Prader-Willi syndrome (45–47). In addition to the Paf1C, the 3′-end formation of certain yeast snoRNAs requires the RNA binding proteins Nrd1 and Nab3, the helicase Sen1, and the peptidyl prolyl-cis/trans-isomerase Ess1 (44, 48–52). Extended snoRNA transcripts are processed to their mature length by the exosome, a conserved nucleolytic complex (53, 54). Transcripts produced from two snoRNA genes, SNR13 and SNR47, have been shown to require Paf1 to promote efficient 3′-end formation and prevent the extension of RNAs into downstream genes (44). Previous work has implicated Paf1C-dependent histone modifications in this process, such as H2B K123 ubiquitylation affecting 3′-end formation of SNR47 and SNR13 and H3 K4 trimethylation contributing to SNR13 termination (55, 56). It remains an open question whether these modifications are universally required at all Paf1-dependent snoRNA targets.
To better understand snoRNA 3′-end formation and the involvement of the Paf1C in this process, we explored the roles of the Paf1C and histone modifications at all potential snoRNA targets. Examination of high-density tiling arrays to find RNA transcripts affected by the loss of PAF1 uncovered additional snoRNA transcripts that require the Paf1C for their proper 3′-end formation. Through further analysis of these snoRNAs, we demonstrate locus-specific levels of regulation by showing that the H2B ubiquitylation pathway is critical at only a subset of Paf1-targeted snoRNAs. Our studies on the involvement of Rad6 in snoRNA termination reveal that Rad6 can have roles at snoRNAs that are primarily independent of H2B K123 ubiquitylation. Furthermore, we have discovered functions for the Bur1-Bur2 complex and the Set2 methyltransferase in the formation of these ncRNA transcripts. Taken together, these results improve our understanding of the roles that the Paf1C and its dependent histone modifications play in gene expression and transcription termination.
S. cerevisiae strains used in this study are listed in Table 1 and, unless otherwise noted, are isogenic to strain FY2, a GAL2+ derivative of strain S288C (57). Yeast transformations, gene disruptions, and genetic crosses were performed as previously described (58, 59). All integrations and gene disruptions were confirmed by PCR. Strains with an integrated copy of the htb1-K123R (htb1 gene with K changed to R at position 123 of the encoded H2B protein) allele have been described (56). As was done previously (60), a truncated version of Set2 was generated (amino acids 1 to 261) using PCR amplification of pFA6a-13Myc-kanMX6 (61) integrated into the endogenous SET2 locus. Strains were confirmed by PCR and Western blot analysis, which demonstrated the presence of the epitope tag, the absence of histone H3 K36 trimethylation, and the presence of H3 K36 dimethylation. The LEU2-marked control plasmid (pADH1-HIS3-CYC1) and SNR47 termination reporter plasmid [pADH1-SNR47(70)-HIS3-CYC1] used in Fig. 7 have been described previously (62). The termination reporter plasmid contains 70 bp of the SNR47 3′-end formation element sufficient for termination just upstream of HIS3. Cells that cannot properly terminate transcription within these 70 bp will express a read-through transcript containing HIS3 that allows for strong growth on media lacking histidine. The control plasmid lacks these 70 bp, and all cells with this plasmid express HIS3 from the ADH1 promoter. Unless otherwise noted, cells were grown at 30°C in yeast extract-peptone-dextrose (YPD) medium (59).
Total RNA was isolated from cells grown to log phase and subjected to Northern blot analysis with random prime-labeled, PCR-amplified DNA probes as described previously (63). The SNR47-YDR042C, SNR48-ERG25, and SNR79-SEN2 probes were designed to detect transcription downstream of snoRNA genes, whereas the SNR47 probe used in Fig. 6C was designed against sequences internal to SNR47 (see Table 2 for a list of primers). These snoRNA probes were made using [α-32P]dATP and [α-32P]dTTP; SCR1 probes were made using [α-32P]dATP. Signals were quantified using ImageJ software and were made relative to the SCR1 loading control signal. The relative signal from the wild-type control strain was set equal to one within each Northern blot analysis. For quantification of all Northern blot analyses, signals were averaged for at least three independent sample preparations. Error bars represent plus and minus 1 standard error from the mean (SEM).
Total RNA was isolated as described above and then subjected to DNase treatment using Ambion Turbo DNA-free (catalog no. AM1907) and RNase inhibitor (catalog no. AM2682). cDNA was generated using Ambion RETROscript kit (catalog no. AM1710) with random hexamers and oligo(dT) primers. Real-time PCRs utilized SYBR green (Fermentas) and high-efficiency primers downstream of snoRNAs (see Table 2 for information on primers used). Reactions were run using an Applied Biosystems 7300 real-time PCR system. Signals were normalized to that of ACT1 and the relative transcript level of wild-type cells was set at one. For controls, reactions lacking reverse transcriptase or template were performed. Error bars represent plus and minus 1 SEM. P values were determined using Student's t test. At least three independent biological replicates were used to generate cDNA, and qPCRs with each replicate were performed in triplicate.
For each genotype, three independent biological replicates of total RNA were isolated and DNase treated as described above. Strand-specific cDNA synthesis reactions were performed on each sample using SuperScript II reverse transcriptase (Invitrogen) or with a reaction mixture with no reverse transcriptase as a control, as described before (64). These reaction mixtures contained both a primer designed to reverse transcribe transcripts that extended downstream of the snoRNA gene sequence and a primer to reverse transcribe the ACT1 mRNA. cDNA synthesis or control reaction mixtures with no reverse transcriptase were amplified by PCR. Two volumes of cDNA (1× and 6×) were used as the templates in PCRs to ensure signal linearity. Primers used for cDNA synthesis and PCRs can be found in Table 2.
Chromatin was isolated from cells grown to log phase, and chromatin immunoprecipitation (ChIP) was performed as previously described (65). Chromatin was isolated in triplicate for each genotype. Immunoprecipitation (IP) of sonicated chromatin was performed with an anti-Rpb3 antibody (catalog no. W0012; NeoClone Biotechnology) and protein G-coupled Sepharose beads (protein G Sepharose 4 fast flow; GE Healthcare). For each experiment, a no-antibody control IP reaction was done with wild-type chromatin using just protein G beads. IP and input DNA were used as the templates in quantitative real-time PCR (performed as described above). The primers used in this study are shown in Table 2. The ChIP assays shown in Fig. 4 and and55 were performed at the same time; therefore, the wild-type and no-antibody controls are the same in both figures.
RNA was isolated from isogenic wild-type (KY2276) and paf1Δ (KY1702) cells grown in triplicate, treated with DNase (GE Healthcare), and purified with RNeasy minikit (Qiagen) as previously described (66). cDNA was synthesized using random hexamers and oligo(dT) with SuperScript II reverse transcriptase (Invitrogen) and purified with MinElute columns (Qiagen) (66). Because the cDNA samples were prepared in the absence of actinomycin D, analysis focused on transcripts encoded in the sense direction (67). cDNA fragmentation, labeling, and array hybridization were done as previously described (67). Arrays were designed in collaboration with Affymetrix (catalog no. PN 520055) and contain 6.5 million oligonucleotide features, 25-nucleotide probes spaced every 8 bp covering one strand of the S. cerevisiae genome sequence, and a second set of probes offset 4 bp to cover the other strand. High-density tiling arrays were analyzed and compared using Affymetrix tiling analysis software (TAS). The results from triplicate samples were averaged, the coding sequence of each ORF was divided into 80 equal-sized bins, and probes hybridizing to those regions were used to assign a hybridization signal intensity for each bin. Bin values were used to determine the mean and median signal intensity of each ORF (excluding dubious ORFs and mitochondrial genes) for the wild-type and paf1Δ samples. The comparison of wild-type and paf1Δ signals within ORFs can be found in Data set S1 in the supplemental material. To examine potential read-through snoRNA transcripts, bins of 10 bp each from sequences downstream of each genomic snoRNA gene were assigned a hybridization signal, with coordinates defined as in the Saccharomyces Genome Database (SGD). We calculated the average distance from a genomic snoRNA gene to the next downstream sense ORF as ~380 bp and the average distance from a genomic snoRNA gene to a downstream antisense ORF as ~230 bp. Therefore, we examined 200 bp downstream to try to minimize the influence of transcription of downstream ORFs. Bin values were used to determine the mean and median signal intensity of 200 bp downstream of each snoRNA for wild-type and paf1Δ samples. For snoRNAs found within introns, 150 bp downstream was examined. The snoRNAs were ranked according to increased downstream transcription in the absence of PAF1 (Table 3; see Data set S1 in the supplemental material).
The expression array data are available at ArrayExpress under accession no. E-MTAB-1342.
To define the scope of the involvement of the Paf1C in snoRNA 3′-end formation, we looked for snoRNA genes across the genome at which the absence of PAF1 leads to the synthesis of RNA transcripts that are extended at their 3′ ends. Analysis of high-density genome tiling arrays probed with labeled cDNA samples of transcripts prepared from wild-type and paf1Δ cells revealed many new examples of snoRNA genes affected by the deletion of PAF1 (see Materials and Methods). To identify Paf1-regulated snoRNAs, the ratio of transcript levels in paf1Δ cells relative to wild-type cells was calculated for sequences downstream of each snoRNA gene (see Materials and Methods; also see Data set S1 in the supplemental material). One-third of snoRNA loci were calculated to have at least a 1.7-fold increase in downstream transcription in the absence of PAF1 (Table 3; results for all snoRNAs are given in Data set S1). This list included the known Paf1-regulated gene SNR47 and many additional snoRNA genes, such as SNR48. These newly identified snoRNA targets of Paf1 represent both major functional classes of snoRNA genes (H/ACA box and C/D box). Using Northern blot analysis and RT-qPCR to measure the extended snoRNA transcripts in paf1Δ cells, we confirmed that the SNR47, SNR48, SNR79, SNR85, and SNR32 genes require Paf1 for proper RNA 3′-end formation (Fig. 1). Taken together, these findings demonstrate that Paf1 is required for proper 3′-end formation of many snoRNAs throughout the yeast genome.
Given that not all functions of the Paf1C are shared equally by all complex members, we wanted to examine each component of the Paf1C for its contributions to accurate snoRNA 3′-end formation (1, 2). We chose to analyze a known Paf1 target gene, SNR47, and SNR48, the snoRNA gene showing the strongest dependence on Paf1 as measured by the array analysis. The results of our Northern blot analyses showed that Paf1 and Ctr9 seem to be the most critical Paf1C components for proper SNR47 RNA 3′-end formation, and in agreement with our previous observations, Cdc73 and Rtf1 play lesser but still significant roles at SNR47 (Fig. 2A) (44, 56). Similar to the results observed at the SNR47 gene paf1Δ and ctr9Δ cells had the highest levels of extended transcripts at SNR48, while leo1Δ cells behaved similarly to wild-type cells at both loci (Fig. 2). Unexpectedly, we could not detect a significant role for RTF1 at SNR48, whereas deleting CDC73 caused a small but reproducible increase in 3′-end extended SNR48 transcripts (Fig. 2B). The finding that the requirement for Rtf1 at SNR47 and SNR48 is different was surprising because we had previously shown that Rtf1-mediated histone H2B K123 monoubiquitylation is necessary for proper 3′-end formation of the snoRNAs SNR13 and SNR47, which both require Paf1 (56). The result at SNR48 indicates that H2B K123 ubiquitylation may not always have a role in 3′-end formation of snoRNAs targeted by Paf1.
To test the idea that H2B K123 ubiquitylation may not always function in 3′-end formation of snoRNAs targeted by Paf1, we examined RNA 3′-end formation at other Paf1-targeted snoRNA genes for dependence on Rtf1 and H2B K123 ubiquitylation. Using RT-qPCR, we found that cells lacking RTF1 or the gene encoding the H2B K123 ubiquitin ligase, BRE1, have significant levels of read-through transcripts at SNR47, consistent with previous Northern blot analyses (Fig. 3A) (56). To determine whether histone H2B ubiquitylation also plays an important role in 3′-end formation of other Paf1-regulated snoRNAs, we tested for read-through transcripts in rtf1Δ and bre1Δ cells at the SNR85, SNR32, and SNR48 genes, all of which are Paf1 dependent. Unlike the significant levels of read-through transcripts observed at SNR47 in rtf1Δ and bre1Δ cells, we did not detect the same levels of read-through transcripts at these other snoRNA loci in cells lacking RTF1 or BRE1 (Fig. 3A to toC).C). These results indicate a gene-dependent requirement among Paf1-regulated snoRNAs for the involvement of Rtf1-mediated H2B K123 ubiquitylation in RNA 3′-end formation. To better understand the different requirements for proper snoRNA 3′-end formation, we focused our studies on SNR47 and SNR48 as examples of Paf1-regulated snoRNA loci that are either dependent on Rtf1 or primarily Rtf1 independent.
Because the ubiquitin-conjugating enzyme Rad6 is required for Rtf1-mediated H2B K123 ubiquitylation, we examined whether the deletion of RAD6 affects 3′-end formation of SNR47 and SNR48 transcripts (19, 20). As with Rtf1 and Bre1, Rad6 was also shown to be necessary for proper 3′-end formation of SNR47 (56). Surprisingly, unlike RTF1 and BRE1, we found that RAD6 was strongly required for SNR48 3′-end formation, as rad6Δ cells have high levels of extended SNR48 transcripts (Fig. 3B and andC).C). These results suggest that Rad6 has a role in RNA 3′-end formation at SNR48 that is independent of its role in H2B K123 ubiquitylation. Consistent with this idea, the htb1-K123R substitution caused only a slight increase in the levels of 3′-extended SNR48 transcripts (Fig. 3B and andC).C). If the role of Rad6 in promoting RNA 3′-end formation at SNR48 is mainly independent of its H2B ubiquitylation function, we hypothesized that this role of Rad6 may also be independent of Paf1. To test this idea, we measured extended SNR48 transcripts in paf1Δ, rad6Δ, and paf1Δ rad6Δ cells, setting the extended transcript levels in paf1Δ cells to one for ease of comparison. Double mutant cells showed an additive defect in RNA 3′-end formation, suggesting that the function of Rad6 in SNR48 3′-end formation is also likely to be independent of Paf1 (Fig. 3D). Collectively, our results suggest that Rad6 can promote snoRNA formation through different mechanisms, which are either dependent or independent of Paf1C-regulated H2B K123 ubiquitylation.
The extended snoRNA transcripts could represent an RNA Pol II transcription termination defect in which the polymerase fails to terminate properly and continues transcription or could also reflect an RNA processing defect. Therefore, we analyzed the occupancy of Rpb3, an RNA Pol II subunit, across SNR47 and SNR48 in wild-type and paf1Δ cells. We found that cells lacking PAF1 had a significant increase in RNA Pol II levels downstream of SNR47 and SNR48 compared to wild-type cells (Fig. 4A and andB).B). These results indicate that in cells lacking PAF1, inefficient transcription termination by RNA Pol II contributes, at least in part, to the formation of 3′-extended snoRNA transcripts. The extended snoRNA transcripts are aberrant RNAs and possible substrates of the nuclear exosome, of which Rrp6 is an exonuclease component (54). We observed higher levels of SNR47 and SNR48 read-through transcripts in paf1Δ rrp6Δ double mutant cells than in paf1Δ or rrp6Δ single mutant cells (Fig. 4C and andD),D), suggesting that Paf1 likely functions independently of the exosome in preventing the accumulation of these extended transcripts. Taken together, our results show a requirement for Paf1 in RNA Pol II transcription termination at snoRNA genes.
Recruitment of the Paf1C to chromatin involves the Bur1-Bur2 complex (10–12, 18, 68). Consequently, we asked whether Bur1-Bur2 participates in snoRNA 3′-end formation at SNR47 and SNR48. Since BUR1 is essential for viability, we used bur2Δ cells to determine whether Bur1-Bur2 affects the synthesis of snoRNA transcripts. Interestingly, we found very high levels of 3′-extended transcripts at the Paf1-targeted snoRNAs SNR47 and SNR48 in the absence of BUR2 (Fig. 5A and andB).B). In fact, bur2Δ cells had much higher levels of extended snoRNA transcripts than paf1Δ cells, indicating that the loss of BUR2 is more detrimental to snoRNA 3′-end formation than loss of PAF1. In agreement with this, we also observed a strong requirement for Bur2 in promoting RNA Pol II transcription termination at SNR47 and SNR48 (Fig. 5C and andD).D). When we assessed the epistatic relationship between bur2Δ and paf1Δ, we found that cells lacking both BUR2 and PAF1 do not have increased read-through transcripts at SNR47 and SNR48 relative to bur2Δ cells (Fig. 5E and andF).F). The lack of an additive defect in the paf1Δ bur2Δ double mutant indicates that Paf1 and Bur2 likely function in the same pathway. However, Bur1-Bur2 may have additional functions outside this pathway, as suggested by the robust snoRNA termination defects observed in bur2Δ strains. Alternatively, there may be a threshold effect on snoRNA termination defects so that a defect much greater than that caused by loss of BUR2 may cause inviability. Confirming that the extended SNR47 and SNR48 transcripts in paf1Δ and bur2Δ strains are indeed read-through snoRNA transcripts, we demonstrated that these transcripts contain both the snoRNA sequence and sequences downstream of the snoRNA, using strand-specific RT-PCR or Northern blot analysis (Fig. 6). Together, these results suggest a greater requirement for Bur1-Bur2 than Paf1C in proper snoRNA 3′-end formation, likely reflecting the multiple roles of Bur1-Bur2 in the phosphorylation and recruitment of proteins integral to transcription and histone modifications (11, 68–70).
In addition to facilitating snoRNA transcription termination, Bur1-Bur2 and Paf1 are also required for full levels of histone H3 K36 trimethylation. Therefore, we hypothesized that there might be a role for H3 K36 trimethylation in snoRNA 3′-end formation (18). Because H3 K36 methylation is catalyzed by the histone methyltransferase Set2, we used set2Δ cells to examine whether a lack of H3 K36 methylation adversely affects 3′-end formation of SNR47 transcripts. Using both Northern blot analyses and RT-qPCR, we uncovered a role for Set2 in SNR47 RNA 3′-end formation (Fig. 7A and andB).B). Whereas cells lacking SET2 lose all H3 K36 methylation states (mono-, di-, and trimethylation), paf1Δ, ctr9Δ, and bur2Δ cells primarily lose H3 K36 trimethylation (18). Therefore, to mimic the H3 K36 methylation defect in these cells, we utilized a truncation allele of SET2, set2(1-261) (set2 gene that codes for amino acids 1 to 261 of the Set2 protein), which specifically abrogates H3 K36 trimethylation (60). We found that set2(1-261) cells show a defect in SNR47 3′-end formation similar to what we observed in set2Δ cells, suggesting that Set2-mediated trimethylation of H3 K36 is specifically important for snoRNA 3′-end formation (Fig. 7B). As independent confirmation of this result, cells lacking H3 K36 trimethylation were also impaired in termination within SNR47 sequences on an SNR47 termination reporter plasmid (Fig. 7C and andD)D) (62).
Given that Paf1 is needed to promote full levels of H2B monoubiquitylation, which in turn is needed for efficient 3′-end formation of SNR47 RNA, we asked whether H3 K36 trimethylation and H2B K123 ubiquitylation are part of the same pathway that promotes snoRNA 3′-end formation by measuring the levels of SNR47 read-through transcripts in set2Δ bre1Δ cells. RT-qPCR showed that set2Δ bre1Δ cells have increased levels of extended SNR47 transcripts relative to single mutant set2Δ or bre1Δ cells (Fig. 7E). This finding suggests that Paf1 can promote two pathways of snoRNA 3′-end formation, one through histone H2B ubiquitylation and one through histone H3 K36 trimethylation. While these two pathways are likely to be independent on the basis of our results showing an additive increase in read-through transcripts in set2Δ bre1Δ cells, we cannot exclude the possibility of some cross talk in light of previous findings that H2B K123 ubiquitylation can impact H3 K36 methylation at certain genes (71). Cumulatively, our results indicate that Paf1C-dependent histone modifications can make independent contributions to snoRNA 3′-end formation and underlie locus-specific regulation.
Beyond Nrd1, Nab3, and Sen1, the proper 3′-end formation of snoRNAs requires the functions of additional factors such as the exosome, the TRAMP complex (Trf4-Air2-Mtr4p polyadenylation complex), and proteins that interact with RNA Pol II, including Pcf11, Ess1, and the Paf1C (44, 48, 51, 72–74). Here we have used high-resolution tiling arrays to investigate the requirement for Paf1C in snoRNA 3′-end formation on a genome-wide scale. Our list of snoRNA genes most strongly affected by the deletion of PAF1 encompasses one-third of the genomic snoRNA genes (Table 3) and likely underestimates the scope of the Paf1C effect, as a known Paf1C-regulated snoRNA gene, SNR13, fell below a 1.7-fold increase in calculated downstream transcription (see Data set S1 in the supplemental material). We verified the Paf1 dependence of a subset of these newly identified snoRNA gene targets of Paf1 and showed that the extended snoRNA transcripts in paf1Δ cells arise, at least in part, through defective transcription termination. Although we cannot exclude the possibility that Paf1 also contributes to the processing of snoRNAs, the results of our double mutant studies indicate that Paf1 and the nuclear exosome have nonoverlapping functions.
Our identification of new snoRNA targets of the Paf1C allowed us to define roles for additional factors in snoRNA 3′-end formation, such as Rad6, Bur1-Bur2, and Set2. Interestingly, we also found evidence indicating that regulation can exist in a locus-specific manner. For example, Rtf1, Bre1, and histone H2B ubiquitylation are important at SNR47 and SNR13 but are not as critical at SNR48, SNR32, and SNR85, yet all these genes require Paf1. Supporting our results, functions of the Paf1C in transcription elongation and nucleosome occupancy that are independent of histone modifications have been reported (15, 75). Furthermore, RNA transcripts do not always show the same extent of defects in NAB3, NRD1, and SEN1 mutants and strains with mutations in other known regulators, suggesting that more complex levels of regulation may exist at individual loci (48).
To identify snoRNA genes with increased levels of read-through transcription in paf1Δ cells and minimize any interfering effects of signals from downstream genes, we compared expression levels within 200 bp downstream of genomic snoRNA genes for the isogenic paf1Δ and wild-type strains (200 bp is lower than the average distance we calculated between genomic snoRNAs and downstream sense and antisense ORFs). However, it remained possible that signals from certain downstream genes were impacting our analysis. To address this concern, we analyzed two groups of snoRNA loci, which had either the highest or lowest levels of snoRNA read-through transcription in paf1Δ cells relative to wild-type cells, but we failed to find a correlation between the calculated level of transcription downstream of the snoRNA gene and the distance to the nearest ORF. In addition, we did not find a correlation between our calculated snoRNA read-through levels and the expression or stability of the closest downstream mRNA (data not shown). Therefore, while the predicted levels of read-through transcription at some individual snoRNA loci may be influenced by a downstream gene, we do not feel these effects impact the majority of the snoRNAs we have classified as Paf1 dependent.
Importantly, many of the newly identified snoRNA targets of Paf1 are regulated by the Nrd1-Nab3 pathway or are bound by those proteins (51, 62, 76, 77). We confirmed directly that the Nrd1 RNA binding protein is needed at SNR47 and SNR48 for RNA 3'-end formation (data not shown). Furthermore, only five snoRNAs on our list of Paf1-regulated snoRNA genes (Table 3) were not among the top 100 cross-linked RNAs for Nrd1 or Nab3 in transcriptome-wide binding analyses (76). Interestingly, over 60% of the snoRNAs that are in the bottom third when ranked by Paf1 dependence are not found among the top cross-linked RNAs for Nrd1 and Nab3. These observations indicate an enrichment of the most strongly Paf1-regulated snoRNAs among the collection of RNAs highly bound by Nrd1 and Nab3 and suggest that Paf1 is important for the termination of snoRNAs that are targeted by Nrd1 and Nab3. Further work will be required to mechanistically define how Paf1 functionally cooperates with these RNA binding proteins.
Although the loss of histone H2B K123 ubiquitylation caused by deletion of BRE1 or RTF1 or the H2B-K123R substitution does not strongly impact transcript synthesis at SNR48, rad6Δ cells show high levels of SNR48 read-through transcripts. Therefore, our data suggest that the Rad6 E2 ubiquitin conjugase may be working at some snoRNA genes in a manner independent of Paf1 and histone modifications, possibly implicating the other E3 ubiquitin ligase partners of Rad6 in snoRNA 3′-end formation. Specifically, Rad6 has been shown to work with Rad18 in DNA repair pathways and Ubr1 in gene regulation (78, 79). However, we found that cells lacking either UBR1 or RAD18 did not exhibit SNR48 read-through transcripts like rad6Δ cells (data not shown). It is currently unknown whether the primary contribution of Rad6 in SNR48 3′-end formation involves combinatorial effects of its various ubiquitin ligase partners and their cellular roles or other as yet undefined functions of Rad6.
We have demonstrated a previously undescribed role for Set2-mediated histone H3 K36 trimethylation in snoRNA 3′-end formation. This result may explain why Paf1 and Ctr9 are the most critical Paf1C subunits to snoRNA 3′-end formation through their actions promoting both histone H3 K36 and H2B K123 modifications. We expect that the histone modifications at these snoRNA genes are important to impact chromatin structure and recruit downstream effectors that may act directly on transcription termination or elongation (55). At SNR13 for example, the Buratowski lab found that Set1-mediated H3 K4 trimethylation is important for proper SNR13 transcript termination in cells bearing a nrd1 mutation, most likely through the recruitment of histone deacetylase complexes (HDAC) to SNR13 (55). Set2 can work in a similar manner at certain genes by affecting the recruitment of the HDAC Rpd3S (60, 80, 81). Given that Rpd3S recruitment is thought to be dependent on H3 K36 dimethylation, which is present in the paf1Δ, bur2Δ, and set2(1-261) cells that show a termination defect at SNR47, the role of Set2 in snoRNA termination likely involves a different function, which is more dependent on H3 K36 trimethylation. One possibility is that recruitment of chromatin remodeling complexes such as Isw1b by H3 K36 trimethylation impacts snoRNA termination (82). We also established an important role for the Bur1-Bur2 CDK-cyclin complex in transcription termination at snoRNA genes by showing that bur2Δ cells have much higher levels of 3′-end extended snoRNAs than paf1Δ cells. We expect that Bur1-Bur2 is contributing to snoRNA termination through multiple functions, such as the phosphorylation of Rad6, Spt5, and the Rpb1 subunit of RNA Pol II, and also by promoting the recruitment of the Paf1C to chromatin for full levels of histone modifications (10–12, 68–70). In metazoans, the Paf1C also regulates histone modifications and transcription elongation and interacts with RNA 3′-end formation components. Therefore, enhancing our understanding of the functions and targets of the Paf1C in yeast may ultimately explain the critical importance of the Paf1C to development and disease progression in higher eukaryotes (2, 35).
We thank members of the Nislow lab for their technical advice and assistance and Jeff Corden and David Brow for sharing reagents and information. We also thank Allen Ho, Sarah Hainer, Kristin Klucevsek, Peggy Shirra, and Joe Martens for comments on the manuscript.
This research was supported by National Institutes of Health grant R01-GM52593 to K.M.A. and by award number F32GM093383 to B.N.T. from the National Institute of General Medical Sciences. L.E.H., M.G., and C.N. are supported by grants from the Canadian Institutes of Health Research (MOP-84305).
The content of this article is solely the responsibility of the authors and does not represent the views of the funding institutions.
Supplemental material for this article may be found at http://dx.doi.org/10.1128/MCB.01233-12.