The PAF Complex and Prf1/Rtf1 Delineate Distinct Cdk9-Dependent Pathways Regulating Transcription Elongation in Fission Yeast

Summary

Cyclin-dependent kinase 9 (Cdk9) promotes elongation by RNA polymerase II (RNAPII), mRNA processing, and co-transcriptional histone modification. Cdk9 phosphorylates multiple targets, including the conserved RNAPII elongation factor Spt5 and RNAPII itself, but how these different modifications mediate Cdk9 functions is not known. Here we describe two Cdk9-dependent pathways in the fission yeast Schizosaccharomyces pombe that involve distinct targets and elicit distinct biological outcomes. Phosphorylation of Spt5 by Cdk9 creates a direct binding site for Prf1/Rtf1, a transcription regulator with functional and physical links to the Polymerase Associated Factor (PAF) complex. PAF association with chromatin is also dependent on Cdk9 but involves alternate phosphoacceptor targets. Prf1 and PAF are biochemically separate in cell extracts, and genetic analyses show that Prf1 and PAF are functionally distinct and exert opposing effects on the RNAPII elongation complex. We propose that this opposition constitutes a Cdk9 auto-regulatory mechanism, such that a positive effect on elongation, driven by the PAF pathway, is kept in check by a negative effect of Prf1/Rtf1 and downstream mono-ubiquitylation of histone H2B. Thus, optimal RNAPII elongation may require balanced action of functionally distinct Cdk9 pathways.

Introduction

Regulation of RNA polymerase II (RNAPII) transcription elongation affects the expression of many eukaryotic genes [1]. A central regulator of metazoan elongation is positive transcription elongation factor b (P-TEFb), a complex comprising cyclin-dependent kinase 9 (CDK9) and a cyclin partner. P-TEFb was identified on the basis of its biochemical activity in overcoming promoter-proximal RNAPII pausing imposed by DRB-sensitivity inducing factor (DSIF; a complex of Spt4 and Spt5) and a negative elongation factor (NELF) [2]–[5]. Cdk9 orthologs are present and essential in Saccharomyces cerevisiae (Bur1) and Schizosaccharomyces pombe (Cdk9), where they have imputed positive roles in RNAPII elongation [6]–[12]. Cdk9 phosphorylates multiple proteins involved in elongation, including Spt5 and the Rpb1 subunit of RNAPII [13]. In addition to regulating elongation, these phosphorylations are implicated in the recruitment of chromatin-modifying enzymes and mRNA 3′-end processing factors [11], [14]–[16]. Cdk9 can also phosphorylate and activate the histone H2B ubiquitin-conjugating enzyme RAD6, indicating multiple links between Cdk9 and chromatin modification [17], [18].

Cdk9 phosphorylates both Rpb1 and Spt5 within repeated amino acid motifs at their C-termini. The Rpb1 C-terminal domain (CTD) repeat has consensus sequence Y¹SPTSPS⁷ and is involved in coordinating co-transcriptional events [14]. Cdk9 can phosphorylate multiple sites within the CTD heptad, but its roles in RNAPII elongation have been commonly attributed to Ser2 phosphorylation [13]. Spt5 contains a C-terminal repeat domain (which varies in size and sequence among taxa) that is phosphorylated by Cdk9 on a conserved threonine residue within the repeated motif [4], [11], [16], [19]. The relative contributions of specific Cdk9 targets to its various functions in transcription are not known.

The Polymerase-Associated Factor (PAF) complex is a key mediator of Cdk9 function during elongation. PAF is minimally composed of conserved subunits Paf1, Leo1, Cdc73, and Ctr9 [20], and its functions broadly overlap with those of Cdk9. In vitro, PAF cooperates with the phosphorylated Spt5 CTD to relieve promoter-proximal RNAPII pausing, and stimulates RNAPII elongation on DNA and chromatin templates [21]–[24]. Like Cdk9, PAF also functions in mRNA 3′ end processing and co-transcriptional chromatin modification [25]–[28]. In budding yeast, PAF recruitment to transcribed genes requires Bur1-dependent Spt5 phosphorylation, which promotes a direct interaction between the Spt5 CTD and the Rtf1 subunit of PAF [15], [16], [18], [29]–[31].

Rtf1 is a stable component of PAF in budding yeast, but is biochemically separate from PAF in metazoans [22], [26], [32]–[37]. Both PAF and Rtf1 are present at transcribed genes and are required for co-transcriptional formation of mono-ubiquitylated histone H2B (H2Bub1) and methylation of histone H3 lysine 4 (H3K4me) [25], [26]. Mutation or knockdown of PAF1 and RTF1 revealed shared functions in telomeric gene silencing in S. cerevisiae, somite development and hematopoiesis in zebrafish, regulation of flowering time in A. thaliana, and mouse embryonic stem cell pluripotency [25], [38]–[41]. In contrast, zebrafish Rtf1 and PAF were found to have distinct functions in cardiac development [34]. Characterization of PAF mutants in S. cerevisiae revealed pleiotropic deficiencies for paf1Δ and ctr9Δ mutants but few phenotypes overall in an rtf1Δ strain [42]. Indeed, an rtf1Δ mutation was found to suppress the phenotypes of paf1Δ, suggesting that the relationship between Rtf1 and PAF could, in some circumstances, be antagonistic [37]. Lethality of bur1Δ in budding yeast is also suppressed by rtf1Δ, as are phenotypes caused by mutation of the associated cyclin Bur2 [43], [44]. Thus, Rtf1 may counteract an important shared function of Cdk9 and PAF.

Although both Rtf1 and PAF are required for formation of H2Bub1 and H3K4me, genetic analyses in S. cerevisiae indicate a particularly close functional connection between Rtf1 and these histone modifications [45], [46]. While the precise gene regulatory functions of H2Bub1 and H3K4me have not been elucidated, H2Bub1, like Rtf1, has been found to oppose Cdk9. In the fission yeast S. pombe, phenotypes associated with loss of H2Bub1 are suppressed by reduction of activity of Cdk9. These include a general 3′ shift of RNAPII density within genes, a pattern that could reflect enhanced elongation through coding regions caused by aberrant Cdk9 function [11]. H2Bub1 itself depends upon Cdk9 activity toward Spt5, and also contributes positively to Spt5 phosphorylation (Spt5-P) levels. Thus, H2Bub1 and Spt5-P may be part of an auto-regulatory mechanism that fine-tunes Cdk9 functions. The Cdk9 function antagonized by H2Bub1 involves a target other than Spt5, suggesting the involvement of a distinct Cdk9 pathway [11].

Using the fission yeast S. pombe as a model system, we report that PAF and Rtf1 (which we designate Prf1 in S. pombe) are involved in two distinct and opposing regulatory pathways, both of which depend on Cdk9 activity. We further show that these divergent functions involve different Cdk9 targets, and mirror the opposing functions of Cdk9 and H2Bub1 in elongation.

Results

PAF and Prf1/Rtf1 are biochemically separate and functionally distinct

To further explore the relationship between Cdk9 activity and H2Bub1 in S. pombe, we assessed the biochemical and functional properties of orthologs of known PAF complex subunits. We first tagged the genes encoding homologs of Paf1 (systematic ID # SPAC664.03), Ctr9 (SPAC27D7.14c; named tpr1⁺ in S. pombe), and Rtf1 [SPBC651.09c; which we henceforth refer to as Prf1, for PAF-related factor 1, to avoid confusion with the unrelated gene SPAC22F8.07c currently annotated as rtf1⁺ [47]] with the tandem affinity purification (TAP) tag and purified the factors from whole-cell extracts [48]. Purified material was analyzed by SDS-PAGE alongside that from an untagged strain. Polypeptides uniquely present in the TAP purifications were excised from the SDS gel and analyzed by tandem mass spectrometry. Material purified from paf1-TAP and tpr1-TAP strains contained orthologs of the previously defined PAF complex components Paf1, Leo1 (SPBC13E7.08c), Tpr1, and Cdc73 (SPBC17G9.02c) (Figure 1A). Prf1 was not visible by SDS-PAGE in the Tpr1-TAP or Paf1-TAP preparations, nor were PAF components visible in the Prf1-TAP preparation (Figure 1A). Thus, Prf1 and PAF are predominantly separate in S. pombe cell extracts.

To further confirm this finding we carried out in-solution mass spectrometry analysis on the Tpr1-TAP and Prf1-TAP preparations. Material derived from an untagged strain was analyzed in parallel as a control. This analysis identified the core PAF subunits Paf1, Leo1, Tpr1, and Cdc73 with high confidence in the Tpr1-TAP preparation, but not in the control or Prf1-TAP preparations (see Table S1). Conversely, Prf1 was identified with high confidence only in the Prf1-TAP purification.

We also investigated whether we could detect association between Prf1 and PAF in a single-step isolation procedure. We constructed a strain carrying both prf1-TAP and tpr1-13xmyc fusion genes. Whole-cell extracts from this strain, as well as from a tpr1-13xmyc strain, were incubated with IgG-sepharose beads to isolate Prf1-TAP and purified material was analyzed by western blotting. Whereas Prf1-TAP was efficiently purified, no Tpr1-13xmyc protein could be detected, further arguing that Prf1 and PAF do not stably associate in S. pombe cell extracts (Figure 1B).

We found that deletion of the S. pombe prf1⁺ gene, or of the tpr1⁺ and cdc73⁺ genes encoding PAF components, resulted in a total loss of H2Bub1 as gauged by immunoblotting, comparable to that observed in htb1-K119R (ubiquitin acceptor site) or brl2Δ (E3 ubiquitin ligase) mutants (Figure 2A). The presence of normal levels of H2Bub1 in the prf1-TAP and tpr1-TAP strains affirmed the functionality of the respective TAP-tagged proteins (Figure S1). As expected from the established dependence of H3K4me on H2Bub1 in yeast [49], [50], the prf1Δ, tpr1Δ, and cdc73Δ mutations also strongly diminished tri- and dimethyl forms of H3K4me, albeit to varying extents (Figure S2). Thus, Prf1 and PAF have a shared function in co-transcriptional histone modification. The paf1Δ and leo1Δ mutations caused only partial reduction in H2Bub1 and H3K4me levels, indicating divergent roles of PAF complex subunits in promoting these modifications (Figures 2A and S2). Because the H3K36me modification was maintained in these mutants, we infer a specific role for S. pombe PAF and Prf1 in the H2Bub1/H3K4me chromatin modification axis (Figure S2).

Loss of H2Bub1 caused by htb1-K119R or brl2Δ mutations is associated with cell morphology and septation defects (including “twinned” septa and unseparated chains of cells) [11]. Microscopic examination of DAPI/calcofluor-stained cells revealed that prf1Δ caused qualitatively and quantitatively similar phenotypes (Figures 2B and 2C). To our surprise, the PAF complex mutations tpr1Δ and cdc73Δ, which eliminated H2Bub1 and reduced H3K4me levels, caused minimal cell morphology and septation defects. This indicates functional divergence between PAF and other factors promoting H2Bub1.

To further probe the functional relationships between Cdk9, PAF, Prf1, and H2Bub1, we assessed growth of the relevant mutant strains under conditions previously shown to make requirement for Cdk9 activity more stringent, including minimal medium (EMM) and medium containing the nucleoside biosynthesis inhibitor mycophenolic acid (MPA) [51]. We found that tpr1Δ and cdc73Δ phenocopied the hypomorphic cdk9-T212A mutant (which lacks the phosphorylation site for the CDK-activating enzyme Csk1; [7]) under these conditions, arguing for a functional link between these factors and Cdk9 activity (Figure 3). The paf1Δ and leo1Δ mutants shared the dramatic EMM sensitivity of tpr1Δ and cdc73Δ but were only mildly sensitive to MPA. Importantly, htb1-K119R and brl2Δ mutants lacked the phenotypic signature associated with reduced Cdk9 activity and loss of PAF, as did prf1Δ (Figure 3). The brl2Δ and prf1Δ mutations did cause moderate growth defects on EMM, perhaps due to H2Bub1-independent functions of Brl2 and Prf1 proteins. Thus, whereas PAF is functionally aligned with Cdk9, Prf1 is intimately connected to factors directly involved in H2Bub1 formation.

Recruitment of PAF and Prf1 to transcribed chromatin involves distinct Cdk9 targets

To further define the distinct functions of PAF and Prf1 we investigated mechanisms that target these factors to transcribed genes. Prompted by findings in budding yeast linking PAF chromatin association to Bur1 activity [15], [16], [30], we performed chromatin immunoprecipitation (ChIP) in prf1-TAP, tpr1-TAP, and paf1-TAP strains that also carried a Cdk9 variant (cdk9^as) engineered to be sensitive to inhibition by bulky ATP analogues [12]. We found that chromatin recruitment of all three proteins to the constitutive act1⁺ and adh1⁺ genes depended on Cdk9 activity (Figure 4A). Acute inhibition of Cdk9^as in these strains had no effect on overall levels of the respective TAP-tagged proteins (Figure S3A) and did not greatly impair RNAPII occupancy at act1⁺ or adh1⁺ (Figure S3B). This argues that chromatin association of both PAF and Prf1 depends on Cdk9 activity.

We reasoned that different Cdk9 targets might mediate the distinct functions of PAF and Prf1. The Spt5 CTD is a specific target of Cdk9 in S. pombe; the Spt5 CTD consists of tandem copies of a nonapeptide motif of consensus sequence T¹PAWNSGSK⁹. We reported previously that changing the Thr1 phosphoacceptor site to alanine reduced H2Bub1 levels [11]. We analyzed Prf1 and PAF recruitment to transcribed chromatin by ChIP in strains harboring Spt5 CTD mutations. The strains were constructed by replacing the endogenous spt5⁺ CTD domain (which contains 18 nonapeptide repeats) with a truncated (but fully functional) CTD domain containing eight consensus repeats or otherwise identical repeat arrays with T1A or T1E mutations [19]. The strains also carried the cdk9^as allele, which allowed assessment of the contribution of specific loss of Spt5-P to the effect of general Cdk9^as inhibition.

The spt5-T1A mutation reduced chromatin association of Prf1 at act1⁺ and adh1⁺ to a level close to that observed upon inhibition of Cdk9^as (5- to 10-fold; Figures 4B and S4A). Thus, Prf1 recruitment to transcribed chromatin strongly depends on phosphorylation of the Spt5 CTD by Cdk9. The phosphomimetic spt5-T1E mutation also impaired ChIP of Prf1, but rendered the residual crosslinking of Prf1 Cdk9-independent. Thus, although a negative charge at all Thr1 positions within a truncated CTD is not sufficient for normal Prf1 recruitment it does allow some Prf1 binding through an altered, Cdk9-independent mechanism.

In contrast, ChIP of Tpr1-TAP revealed Cdk9-dependent chromatin association in the presence of spt5-T1A and spt5-T1E mutations, as well as in a strain in which the Spt5 CTD is deleted altogether (spt5ΔC)(Figures 4B and S4A). This argues that PAF recruitment to chromatin involves a Cdk9 target other than the Spt5 CTD. The modest (∼2-fold) decrease in ChIP of Tpr1-TAP in the spt5-T1A and spt5ΔC mutants (that was not apparent in the phosphomimetic spt5-T1E mutant) suggests that Spt5-P does make some contribution to PAF recruitment. These data point to the existence of two functionally distinct Cdk9 pathways (which we will refer to as the PAF pathway and the Prf1 pathway) that involve distinct phosphoacceptor targets.

We note that global levels of Spt5-P were reduced by prf1Δ, as was found previously for htb1-K119R and brl2Δ mutations (Figure S5A). This implicates Prf1 as part of the positive feedback loop linking Spt5-P and H2Bub1. Mutations in PAF components either did not affect (tpr1Δ and cdc73Δ) or increased (paf1Δ and leo1Δ) global levels of Spt5-P, further highlighting the specific functional connection between Spt5-P and the Prf1 pathway.

The Rpb1 CTD is a target of Cdk9. In S. pombe, Cdk9 contributes to global levels of Rpb1-Ser2 phosphorylation and promotes the formation of multiply phosphorylated Rpb1 CTD repeats [12], [51]. Budding yeast PAF components Cdc73 and Ctr9 both interact with phosphorylated Rpb1 CTD peptides in vitro and enhance global levels of Rpb1-Ser2 phosphorylation in vivo [30], [32], [52]. This latter function is shared by the corresponding PAF components in S. pombe (but not by Prf1; see Figure S5B). To investigate the role of Rpb1 CTD phosphorylation in Prf1 and PAF recruitment, we carried out ChIP assays in strains harboring phosphoacceptor site mutations in all the Rpb1 CTD repeats. The “wild-type” control in these experiments carried a truncated but fully functional Rpb1 CTD domain [53]. ChIP assays using an rpb1-S2AS7A double mutant (in which both serines 2 and 7 are changed to alanine in every CTD repeat) showed very little effect on chromatin association of either Prf1-TAP or Tpr1-TAP at the active genes analyzed (although a ∼2-fold, gene-specific reduction of Prf1-TAP association was observed at act1⁺) (Figures 4C and S4B).

We also tested the effect of an rpb1-S5A mutation, the lethality of which was bypassed by fusion of rpb1-S5A to the MCE1 gene encoding the mammalian mRNA capping enzyme [53]. As compared to the wild-type rpb1-MCE1 fusion, the rpb1-S5A-MCE1 fusion had little impact on crosslinking of either Prf1 or Tpr1 to the two loci tested (Figure S6). These data suggest that Rpb1 CTD phosphorylation on serines 2, 5, or 7 is not required for Prf1 or PAF association with transcribed chromatin in vivo.

The marked dependence of Prf1 chromatin association on Spt5-P led us to test whether this could reflect a direct interaction between the Spt5 CTD and Prf1. Biotinylated peptides corresponding to the unmodified or the Thr1-P form of the Spt5 CTD were coupled to streptavidin beads and incubated with TAP-purified, native Prf1 or PAF. We detected binding of Prf1 (∼10% bound) to the unmodified Spt5 CTD peptide that was enhanced by phosphorylation (∼35% bound) (Figures 5A and 5B). This interaction was specific, as we detected no binding of Prf1 to Rpb1 CTD phospho-peptides modified at Ser2, Ser5, Ser7, or Ser5 and Ser7. Given that TAP-purified Prf1 is not visibly associated with other proteins (Figure 1), this argues that there is a direct interaction between Prf1 and Spt5-P.

PAF (purified from the tpr1-TAP strain) also bound (∼5% of total) to the unmodified Spt5 CTD peptide but failed to bind detectably to the Spt5-P peptide, suggesting that, in vitro, phosphorylation inhibits PAF association with the Spt5 CTD (Figures 5A and 5B). Similar results were obtained with PAF purified from the paf1-TAP strain (Figure S7). Thus, we presume that the modest negative effect of the spt5-T1A mutation on Tpr1-TAP crosslinking to active genes was due to another Spt5 CTD-binding factor. Consistent with our ChIP data, we did not detect PAF association with unmodified or singly modified Rpb1 CTD peptides, and detected weak association with the peptide that was doubly phosphorylated at Ser5 and Ser7.

Evidence that PAF and Prf1 have opposing effects on RNAPII elongation

We previously argued that the relationship between Cdk9 and H2Bub1 is mediated by two Cdk9 pathways, one involving Spt5-P and another involving Spt5-P and an additional target [11]. Our data suggested the two pathways were functionally opposed. We wished to test whether the Prf1 and PAF pathways described here might mediate these opposing functions.

We first assessed genetic interactions between prf1Δ and reduction of Cdk9 activity. We introduced the cdk9^as allele into the prf1Δ background and analyzed double mutants by DAPI/calcofluor staining after inhibition of Cdk9^as with 3-MB-PP1. Cdk9^as inhibition strongly suppressed the cell morphology and septation phenotypes caused by prf1Δ (as it did for brl2Δ; Figure 6A). A hypomorphic cdk9-T212A mutation similarly rescued these phenotypes. These results indicate that the cell morphology and septation defects in prf1Δ, brl2Δ, and htb1-K119R strains depend upon Cdk9 activity, and suggest that Prf1, Brl2, and H2Bub1 itself act to limit some aspect of Cdk9 function.

We next determined whether PAF mutations, which phenocopy cdk9-T212A, could also suppress the cell morphology and septation phenotypes of htb1-K119R, brl2Δ, and prf1Δ mutants. We found that cdc73Δ, when combined with htb1-K119R, brl2Δ, or prf1Δ, suppressed aberrant septation. Similarly, brl2Δ septation phenotypes were effectively suppressed by tpr1Δ (Figure 6A). The dependence of these phenotypes on both Cdk9 activity and PAF is consistent with the Prf1 pathway acting in opposition to the PAF pathway.

We also tested whether htb1-K119R, brl2Δ, and prf1Δ could reciprocally suppress cdk9 and PAF mutant phenotypes. None of these mutations affected the growth of cdk9 or PAF mutants on minimal media. Whereas htb1-K119R suppressed the MPA sensitivity of cdk9-T212A, prf1Δ did not, and brl2Δ enhanced this phenotype (Figure S8). These results indicate that the Prf1 pathway does not generally limit PAF pathway function.

In order to ascertain how these two pathways regulate RNAPII elongation, we carried out RNAPII ChIP experiments in strains harboring mutations in either the Prf1 or PAF pathways. We quantified the ChIP by qPCR with primers in nup189⁺ and SPBC354.10⁺, genes that exemplify the 3′ shift in RNAPII distribution that we documented previously [11]. Consistent with our previous data, the htb1-K119R mutation led to an increase in RNAPII occupancy near the 3′ end of the nup189⁺ gene relative to the 5′ end (Figures 6B and S9A). A similar effect was also observed for brl2Δ and prf1Δ mutations, implying that the 3′ RNAPII shift is characteristic of the Prf1 pathway. In contrast, cdk9-T212A, cdc73Δ, and tpr1Δ mutations led to a relative decrease in RNAPII density toward the 3′ end of the gene, consistent with defective elongation (Figure 6B). The effects on RNAPII distribution were generally less significant at SPBC354.10⁺, but we nonetheless observed a similar pattern in which RNAPII occupancy increased toward the 3′ end of the gene in Prf1 pathway mutants and decreased or was unaffected in PAF pathway mutants (Figure S9A).

We also found that the Prf1 and PAF pathway mutations differentially affected the steady-state levels of mRNA produced from these genes. PAF pathway mutants exhibited modest decreases in mRNA levels, whereas no change or modest increases were observed in the Prf1 pathway mutant strains (Figure S9B). These changes are likely to reflect altered RNAPII elongation, since the absolute levels of RNAPII occupancy near the 5′ ends of these genes did not differ between Prf1 and PAF pathway mutant strains (Figure S9C). Overall, our data suggest that the PAF and Prf1 pathways have differential effects on behavior of the RNAPII elongation complex (Figure 7).

Discussion

Cdk9 influences multiple aspects of RNAPII transcription. Our results here reveal that phosphorylation of the Spt5 CTD by Cdk9 directly and specifically recruits the conserved transcriptional regulator Prf1/Rtf1 to transcribed chromatin. We further define a second Cdk9 pathway, involving phosphorylation of at least one additional target, which recruits the PAF complex. We show that these two pathways have distinct and opposing functions, suggesting that balanced action of Cdk9 pathways is a feature of RNAPII elongation.

Prf1/Rtf1 and PAF are biochemically separate in S. pombe

Prf1/Rtf1 and PAF do not form a stable complex in S. pombe whole-cell extracts, a finding consistent with the properties of these proteins in metazoans. Biochemical studies of human Rtf1 have shown that it can be assembled into a complex with PAF in vitro, and that a C-terminal fragment of Rtf1 can be detected in association with PAF purified from nuclear extracts [22]. Although the physiological significance of these findings has not been explored, they suggest the possibility of a regulated interaction between full-length Rtf1 and PAF. Modulation of this interaction could underlie the shared and distinct functions of Rtf1 and PAF in different biological contexts.

The similarities we have found between the biochemical properties of Prf1 and PAF in fission yeast and those of their metazoan orthologs underscore the utility of the S. pombe system as a model for study of these factors. S. pombe Prf1 and PAF also show functional resemblance to their human counterparts in that they promote the H2Bub1/H3K4me histone modification axis but not H3K36me, a modification that is PAF-dependent in S. cerevisiae [26], [43].

We found that individual subunits within S. pombe PAF were functionally distinct: loss of either Tpr1 or Cdc73 was associated with more severe phenotypic consequences in our assays than loss of Paf1 or Leo1. This contrasts with PAF functional organization in S. cerevisiae, where paf1Δ causes pleiotropic phenotypes [42]. Further investigation into physical interactions between individual PAF subunits and other factors will help to clarify these functional differences.

Prf1/Rtf1 is a direct downstream effector of phosphorylated Spt5 CTD

Our data show that Prf1/Rtf1 directly and specifically associates with the phosphorylated Spt5 CTD, a finding that is consistent with recent reports on budding yeast and human orthologs of this protein [30], [31], [54]. In vitro, budding yeast Rtf1 was found to bind Spt5-P and the phosphorylated Rpb1 CTD with comparable affinities, and Rtf1 mutations that abrogate its interaction with Spt5-P prevented the association of multiple PAF subunits with transcribed chromatin in vivo [30], [54]. The budding yeast Ctr9 and Cdc73 proteins could also interact with phosphorylated peptides corresponding to either the Spt5 CTD or the Rpb1 CTD [30]. The apparently overlapping mechanisms involved in recruitment of Rtf1 and other PAF components to chromatin in budding yeast are in line with the robust association of Rtf1 with PAF found in this organism. Our findings in S. pombe point to a specific and primary role for the Prf1/Spt5-P interaction in Prf1 recruitment to transcribed chromatin, and suggest an alternate recruitment mechanism for PAF.

The direct interaction between Prf1/Rtf1 and Spt5-P helps explain the critical role of Spt5-P in establishment of H2Bub1 levels. This interaction also raises mechanistic questions regarding the reciprocal dependence of Spt5-P on H2Bub1. Loss of Spt5-P in the htb1-K119R mutant is accompanied by a reduction in crosslinking of Cdk9 to transcribed chromatin [11]. In the prf1Δ strain, it is possible that loss of Spt5-P is also attributable to the absence of the Prf1/Spt5-P interaction, which could provoke dephosphorylation of Spt5-P.

The spt5-T1A mutation did not result in cell morphology and septation defects observed in prf1Δ, brl2Δ, or htb1-K119R mutant strains, suggesting that some Prf1 function is maintained in the absence of Spt5 CTD phosphorylation [11]. This is consistent with our ChIP results, which demonstrated a substantial but incomplete loss of Prf1 chromatin recruitment in spt5-T1A mutant strains. Since residual Prf1 recruitment was further reduced by Cdk9^as inhibition, other Cdk9 targets are likely to play some role in this process, although the bypass of Cdk9 dependence in the spt5-T1E mutant suggests that the Spt5 CTD is the primary Cdk9 target involved. Multiple domains of S. cerevisiae Rtf1 can be independently recruited to chromatin, indicating that mechanisms other than Spt5 phosphorylation are also important for Rtf1 chromatin association [31], [46], [55]. It is also possible that Prf1 may retain some function even when it is not stably bound to chromatin, as has been observed for the S. cerevisiae PAF complex [56].

PAF association with transcribed chromatin requires phosphorylation of multiple targets by Cdk9

We found that PAF recruitment to chromatin in vivo was only modestly sensitive to Spt5 CTD phosphorylation at both transcribed loci tested. Since we found no interaction between the purified PAF complex and Spt5-P peptide in vitro, we presume that this reflects an indirect effect on PAF association with chromatin mediated by another factor that binds to Spt5-P. This is unlikely to be Prf1, as PAF chromatin association is not affected by the spt5-T1E mutation (unlike that of Prf1).

Because crosslinking of Tpr1-TAP and Paf1-TAP to transcribed loci was dramatically reduced by Cdk9^as inhibition even in an spt5-T1A mutant, PAF recruitment must require at least one other Cdk9 target in addition to Spt5. Given that the Rpb1 CTD is an established target of Cdk9 activity, we were surprised to find that simultaneous elimination of Ser2 and Ser7 in the Rpb1 CTD repeat, or removal of Ser5, had no substantial effect on PAF recruitment in vivo. This could reflect PAF association with forms of the Rpb1 CTD repeat phosphorylated on both Ser5 and another residue, as has been suggested by recent biochemical experiments with budding yeast PAF components [30]. Alternatively, the role of the Rpb1 CTD may be redundant with the Spt5 CTD [19], [57]. Pleiotropic phenotypes associated with mutant combinations that include Rpb1-Ser5 have precluded thorough genetic tests of these possibilities. Recruitment of the PAF complex to chromatin may also involve another, as yet unidentified, Cdk9 target.

Balanced action of Prf1/Rtf1 and PAF pathways in RNAPII elongation

Septation and cell morphology phenotypes of Prf1 pathway mutants were reduced by cdk9-T212A, cdc73Δ, and tpr1Δ mutations, indicating that these components act aberrantly in the absence of the Prf1 pathway. Prf1 pathway mutations did not generally suppress cdk9 or PAF mutant phenotypes, suggesting that the Prf1 pathway has a specialized role in limiting certain aspects of PAF pathway function.

ChIP of RNAPII reveals differential effects of the two pathways on RNAPII distribution within the nup189⁺ and SPBC354.10⁺ gene-coding regions, with the PAF pathway promoting RNAPII occupancy at promoter-distal sequences and the Prf1 pathway limiting occupancy at these locations. This observation supports our earlier studies on the opposing effects of Cdk9 activity and H2Bub1 on intragenic RNAPII distribution, and implies that these findings also characterize the two Cdk9 pathways we have described.

We propose that Cdk9-dependent recruitment of PAF, through phosphorylation of the Spt5 CTD and other targets, promotes RNAPII elongation through gene coding regions, as supported by our RNAPII ChIP and mRNA expression data. Spt5 CTD phosphorylation also recruits Prf1/Rtf1 and promotes formation of H2Bub1, which may act as an auto-regulatory mechanism to balance the positive effects of the PAF pathway. Previous studies documenting a positive role for PAF in elongation have shown Rtf1 to be dispensable for this function, consistent with the distinctive behavior of Prf1 in S. pombe [21]–[23]. Although we did not observe significant effects of the Prf1 pathway mutations on expression of nup189⁺ and SPBC354.10⁺ (with the exception of a slight increase for SPBC354.10⁺ in the htb1-K119R mutant), it is possible that loss of the putative Cdk9 auto-regulatory function in these mutants would be associated with defective mRNA processing or export, rather than a change in mRNA levels per se. Further elucidation of the mechanisms through which these pathways affect gene expression will be an important goal of future research.

A negative role for the Prf1 pathway in elongation could be related to the functions of H2Bub1 in directing nucleosome assembly and histone deacetylation within gene coding regions [58]–[61]. Indeed, H2Bub1 and the Hos2 deacetylase complex have been linked to reduced RNAPII processivity in budding yeast [61].

It is notable that an interaction between PAF and the elongation factor TFIIS that stimulates elongation in mammalian cells is antagonized by the H2Bub1 ubiquitin ligase RNF20 [62]. Although we find no role for TFIIS in opposing H2Bub1 functions in S. pombe (data not shown), a role for the Prf1 pathway and H2Bub1 in altering PAF function in elongation may be conserved.

In metazoans, Cdk9 activity has an essential role in release of RNAPII from promoter-proximal pause sites. Phosphorylation of multiple targets, including the Rpb1 CTD, Spt5, and NELF, is thought to drive RNAPII from the pause, implying coordinate action of Cdk9 phosphorylation events in pause release [4], [63]. Interestingly, the experiments defining the role of PAF in P-TEFb-dependent pause release in vitro used PAF complex devoid of Rtf1 [21]. We suggest that the P-TEFb pause release function in metazoans, which involves multiple Cdk9 targets and PAF, is analogous to the PAF pathway we have documented in S. pombe. The metazoan analog of the Prf1 pathway might involve Rtf1 recruitment via Spt5-P subsequent to pause release, as part of a mechanism to keep the stimulatory effects of PAF on elongation in check. Thus, the roles of Cdk9 targets in regulating the RNAPII elongation complex may diverge during the course of elongation.

Materials and Methods

Strains and media

S. pombe strains used in this study are listed in Table S2. Liquid cultures used standard YES media (yeast extract 5 g/L, D-glucose 30 g/L, supplemented with 250 mg/L each of histidine, leucine, adenine and uracil). Minimal media (EMM) was purchased from MP Biomedicals. EMM complete was supplemented with 250 mg/L each of histidine, leucine, adenine, and uracil. Mycophenolic acid (MPA; Bioshop) was dissolved in dimethyl sulfoxide (DMSO) and used at a final concentration of 25 µg/mL. The TAP and 13xmyc tags were introduced at the relevant chromosomal loci as described previously [64]. All other genetic manipulations were carried out using standard methods [65].

Tandem affinity purification and mass spectrometry

Paf1-TAP, Prf1-TAP and Tpr1-TAP were purified from whole cell extracts as described [48]. For mass spectrometry analysis of polypeptides excised from polyacrylamide gels, purified material was concentrated by addition of 0.02% sodium deoxycholate, incubation on ice for 30 min, and precipitation with trichloroacetic acid (15% final concentration) overnight. The precipitate was collected by centrifugation, washed twice with ice-cold ethanol and then resuspended with 10 µl of SDS-PAGE sample buffer. Proteins were resolved by SDS-PAGE on a 4–20% gradient acrylamide gel and visualized by Coomassie (Bio-rad) or silver staining before mass spectrometry analysis.

To analyze the purified material in solution, the TAP purification procedure was modified as follows: Glycerol and detergent were omitted from the final wash and elution steps, and EGTA was replaced with EDTA in the elution buffer. The eluted material was subjected to mass spectrometry analysis, details of which are provided in the Supplemental Material (see Text S1).

For single-step TAP purification, whole cell extracts were incubated with IgG-sepharose beads (GE Healthcare) and the beads were collected and washed as described [48]. Beads were suspended in SDS sample buffer and bound material was analyzed by SDS-PAGE and western blotting.

Immunological methods

Immunoblotting was performed as previously described [11]. The following commercial antibodies were used: H2Bub1 (Active Motif #39623), H3K4me1 (Abcam #ab8895), H3K4me2 (Abcam #ab32356), H3K4me3 (Abcam #ab8580), H3K36me3 (Abcam #ab9050), histone H3 (Abcam #ab1791), histone H2B (Millipore #07-371), TAP tag (Fisher Scientific #PICAB1001), Rpb1 (8WG16; Covance #MMS-126R-200), Rpb1-Ser2-P (H5; Covance #MMS-129R), myc (9E10; Covance #MMS-150P). Antibodies against S. pombe Spt5 and Spt5-P were described previously [11].

Phenotypic analyses

DAPI/calcofluor staining was carried out as described previously [11]. For spot tests, 4×10⁵ cells were harvested from exponentially growing cultures, and resuspended in YES media. Ten-fold serial dilutions of the culture were spotted onto the appropriate plates and incubated at 30°C for 3–6 days.

Chromatin immunoprecipitation

ChIP was carried out as previously described [48] with minor modifications. Briefly, 1.5×10⁷ cells grown in YES media were fixed with 1% formaldehyde for 30 min. Glycine (pH 2.5) was added at a final concentration of 125 mM to stop the crosslinking. Cells were washed twice with cold 1XTBS (20 mM Tris-HCl, pH 7.5, 150 mM NaCl, 1 mM Tween-20) and snap frozen in liquid nitrogen. Cell pellets were resuspended in 0.5 mL of lysis buffer [50 mM HEPES, pH 7.6, 150 mM NaCl, 1 mM EDTA, 1% Triton X-100, 0.1% sodium deoxycholate, 1 mM PMSF, protease inhibitor cocktail (Roche)], 200 µl of sterile glass beads were added, and cells were lysed at 4°C using a mini-beat beater (4×30 s with 1 min rests on ice). The lysates were transferred to 1.5 mL tubes and centrifuged at top speed for 15 min to collect the chromatin. Chromatin pellets were resuspended in 1 mL lysis buffer, sonicated for 20 minutes using a Bioruptor waterbath sonicator (Diagenode) (30 s ON/OFF with 2 min rest on ice after 10 min) and centrifuged at top speed for 5 minutes. The supernatant (700–800 µl) was incubated for 2–3 h with either 50 µl of IgG sepharose beads (GE Healthcare)(for TAP immunoprecipitation) or with 15 µl of protein G Dynabeads (Invitrogen) coupled to 8WG16 antibody (for RNAPII immunoprecipitation) at 4°C with rocking. The beads were washed successively with 0.5 mL of each of the following buffers: lysis buffer +0.1% SDS, lysis buffer +0.1% SDS+500 mM NaCl, LiCl buffer (10 mM Tris-HCl, pH 7.5, 1 mM EDTA, 250 mM LiCl, 0.5% sodium deoxycholate, 0.5% NP-40) and TE (10 mM Tris-HCl, pH 7.5, 1 mM EDTA). Chromatin was eluted with 100 µl of elution buffer (50 mM Tris-HCl, pH 7.5, 10 mM EDTA, 1% SDS) at 65°C for 30 min. Reversal of crosslinks, DNA purification, and quantitative PCR were as described [48]. Primers specific for the act1⁺, adh1⁺, nup189⁺, and SPBC354.10⁺ genes have been described previously [11], [66].

Peptide-binding assays

15 µl of streptavidin Dynabeads (Invitrogen) were washed three times with 400 µl of 1xPBS/0.1% Triton-X100. Beads were coupled to 50 µg biotinylated peptide in 380 µl of 1xPBS. Incubation was at room temperature for 3 h with rocking. Spt5 CTD peptides were synthesized at the Keck Foundation Research Laboratory (Yale University) and had the sequence [biotin]-GSKTPAWNSGSKTPAWNS. In the phospho-peptide both Thr residues were phosphorylated. Rpb1 CTD peptides were a generous gift from Francois Robert (IRCM, Montreal) and had the following sequences: [biotin]-YSPTSPSYSPTSPSYSPTSPS, [biotin]-YSPTSPSY(pSer)PTSPSYSPTSPS, [biotin]-SPTSPSYSPT(pSer)PSYSPTSPSY, [biotin]-TSPSYSPTSP(pSer)YSPTSPSYSP, [biotin]-PTSPSYSPT(pSer)P(pSer)YSPTSPSYS. Beads were washed twice with washing buffer (20 mM HEPES pH 7.6, 20% v/v glycerol, 1 mM EDTA, 0.1% Triton X-100, 250 mM KOAc, 10 mM β-glycero-3-phosphate, 1 mM PMSF) and resuspended with 150 µl of the binding buffer (20 mM HEPES pH 7.6, 0.1% Triton X-100, 1 mM PMSF, 10 mM β-glycero-3-phosphate). Binding reactions contained 20 mM Hepes pH 7.6, 0.075% Triton X-100, 0.75 mM PMSF, 7.5 mM β-glycero-3-phosphate, 5% glycerol, 37.5 mM KOAc, 2.5 mM MgOAc, 0.25 mM EDTA, 0.25 mM DTT and 50–200 ng of purified factors (∼50 ng for Prf1-TAP or 200 ng for Paf1-TAP and Tpr1-TAP) in a volume of 200 µl. The mixtures were incubated at 4°C for 1 h with shaking. Beads were washed 4 times with 1 mL of washing buffer, resuspended with 20 µl of SDS-PAGE sample buffer, boiled for 5 minutes at 95°C, and subjected to SDS-PAGE and immunoblotting with a TAP antibody. Signal intensities were quantified using ImageJ software and used to calculate percent bound.

Supporting Information

Attachment 1

Attachment 2

Attachment 3

Attachment 4

Attachment 5

Attachment 6

Attachment 7

Attachment 8

Attachment 9

Attachment 10

Attachment 11

Attachment 12