Paf1 Restricts Gcn4 Occupancy and Antisense Transcription at the ARG1 Promoter

Abstract

The conserved Paf1 complex negatively regulates the expression of numerous genes, yet the mechanisms by which it represses gene expression are not well understood. In this study, we use the ARG1 gene as a model to investigate the repressive functions of the Paf1 complex in Saccharomyces cerevisiae. Our results indicate that Paf1 mediates repression of the ARG1 gene independently of the gene-specific repressor, ArgR/Mcm1. Rather, by promoting histone H2B lysine 123 ubiquitylation, Paf1 represses the ARG1 gene by negatively affecting Gcn4 occupancy at the promoter. Consistent with this observation, Gcn5 and its acetylation sites on histone H3 are required for full ARG1 derepression in paf1Δ cells, and the repressive effect of Paf1 is largely maintained when the ARG1 promoter directs transcription of a heterologous coding region. Derepression of the ARG1 gene in paf1Δ cells is accompanied by small changes in nucleosome occupancy, although these changes are subtle in comparison to those that accompany gene activation through amino acid starvation. Additionally, conditions that stimulate ARG1 transcription, including PAF1 deletion, lead to increased antisense transcription across the ARG1 promoter. This promoter-associated antisense transcription positively correlates with ARG1 sense transcription. Finally, our results indicate that Paf1 represses other genes through mechanisms similar to those used at the ARG1 gene.

INTRODUCTION

Eukaryotic organisms employ several mechanisms to repress gene expression. In some of the best studied cases, transcriptional repressors bind DNA and recruit corepressors that inhibit the basal transcriptional machinery, interfere with activator binding, or recruit histone-modifying proteins (reviewed in reference 55). In addition, nucleosomes within promoters and coding regions can inhibit RNA polymerase II (Pol II) recruitment and impede transcription elongation (reviewed in reference 3). More recently, the synthesis of noncoding RNAs (ncRNAs) has been implicated in transcriptional repression. For example, transcription across the Saccharomyces cerevisiae SER3 promoter inhibits SER3 expression by establishing a chromatin environment that obstructs activator binding (27, 42, 43).

A conserved, globally acting protein complex that has roles in gene repression and activation is the Paf1 complex (Paf1C), which consists of Paf1, Ctr9, Cdc73, Rtf1, and Leo1 in budding yeast (37, 46, 65, 68). Paf1C associates with RNA Pol II on open reading frames (ORFs) (37, 44, 58, 75), regulates the phosphorylation state of the RNA Pol II carboxy-terminal domain (CTD) (47, 54), and facilitates transcription elongation of chromatin templates in vitro (11, 35) and in vivo (71). Paf1C subunits are also required for the proper establishment of several histone modifications that mark active genes (12, 36, 52, 53, 78) and inhibition of H3 and H4 acetylation on the coding regions of active genes (12). In yeast, Paf1 and Rtf1 are required for monoubiquitylation of histone H2B lysine (K) 123 (40, 78, 79) and the subsequent methylation of histone H3 K4 and K79 (8, 21, 36, 52, 53, 69). In addition, Paf1 and Ctr9 are required for H3 K36 trimethylation (12). Human Paf1C is also required for these histone modifications (20, 34, 62, 82), which control the expression of many genes, including HOX genes (82) and genes that maintain embryonic stem cell identity (20). In addition to their functions during transcription elongation, the yeast and human Paf1 complexes are important for proper termination and RNA 3′ end formation of RNA Pol II transcripts (47, 50, 56, 62, 64).

Genome-wide transcriptional analyses indicate that Paf1C is required for the repression of many genes (56). The repressive functions of Paf1C are not fully understood but are of significant interest because of the many connections between this complex and misregulation of genes in human cancers, such as those of pancreatic, breast, and renal tissues (reviewed in references 10, 45, and 51). Here, we aim to elucidate the roles of yeast Paf1C in transcriptional repression. To this end, the ARG1 gene, which encodes arginosuccinate synthetase, an enzyme required for arginine biosynthesis, serves as a model locus of Paf1C-dependent transcriptional repression. The ARG1 gene is a valuable model gene because its transcription is modulated by well-characterized pathways. In the presence of arginine, the ArgR/Mcm1 complex, consisting of Arg80, Arg81, Arg82, and Mcm1, binds to arginine control elements in the ARG1 promoter and represses ARG1 transcription (1, 5, 13, 14, 19, 22, 23, 60). Under conditions of nutrient starvation, Gcn4 binds to sites within the ARG1 promoter and activates ARG1 transcription through the recruitment of multiple coactivators (19, 25, 29, 59, 70). The ARG1 gene was identified by microarray analysis to be one of several hundred genes negatively regulated by Paf1 in rich media (56). We have since found that in addition to Paf1, other members of Paf1C contribute to ARG1 repression (15). Furthermore, we have demonstrated that while Paf1 and Rtf1 mediate ARG1 repression by promoting H2B K123 ubiquitylation and H3 K4 methylation, Paf1 has additional repressive functions that remain to be elucidated (15).

In this study, we further explore the mechanism of repression by Paf1C and investigate the manner in which H2B K123 ubiquitylation represses transcription. Our results indicate that Paf1C-stimulated H2B K123 ubiquitylation inhibits association of Gcn4 with the ARG1 promoter, histone H3 acetylation by Gcn5, and ARG1 expression. Additionally, we have found that Paf1C impacts the pattern of ncRNA synthesis at the ARG1 gene by preventing antisense transcription from traversing the promoter. Our data suggest a model in which antisense transcription across the ARG1 promoter stimulates ARG1 sense transcription.

MATERIALS AND METHODS

Yeast strains and media.

Rich (yeast extract-peptone-dextrose [YPD]) and synthetic complete (SC) media were prepared as previously described (61). Where indicated, sulfometuron methyl (SM) was added to SC medium lacking isoleucine and valine (SC-IV) to a final concentration of 0.6 μg/ml. Yeast strains used in these studies are isogenic with FY2, a GAL2⁺ derivative of S288C, and listed in Table 1 (76). When possible, prototrophic strains were used to minimize variability in ARG1 repression due to auxotrophies (15). Histone H3 and H4 mutant strains were derived from matings between paf1Δ::KanMX cells and strains provided by Jef Boeke (16). Gene deletions and insertions were created by transforming diploid yeast strains with the appropriate PCR fragment (2, 61). Following sporulation and tetrad dissection, genotypes were confirmed by PCR analysis (2, 61). The NatMX cassette on pAG25 was amplified by PCR for NatMX gene replacement (24). To generate gcn4Δ::KanMX strains, the gcn4Δ::KanMX locus in the yeast deletion collection was amplified by PCR as described previously (66, 77). All other PCR fragments for KanMX gene replacement were made by PCR amplification of the KanMX cassette on pRS400 (6).

Table 1.

S. cerevisiae strains used in this study

Strain	Genotype
KA116	MATaARG80-13×MYC::HIS3 GCN4-3×HA::KanMX paf1Δ::KanMX his3Δ0
KA119	MATaARG80-13×MYC::HIS3 his3Δ0
KA120	MATα (hht2-hhf2)::HHTS-URA3/HHFS (hht1-hhf1)Δ::NatMX ura3-52
KA121	MATα (hht2-hhf2)::HHTS-URA3/HHFS (hht1-hhf1)Δ::NatMX paf1Δ::KanMX ura3Δ0
KA122	MATa (hht2-hhf2)::hhts-K9A-URA3/HHFS (hht1-hhf1)Δ::NatMX ura3Δ0
KA123	MATa (hht2-hhf2)::hhts-K9A-URA3/HHFS (hht1-hhf1)Δ::NatMX paf1Δ::KanMX ura3Δ0
KA124	MATα (hht2-hhf2)::hhts-K14A-URA3/HHFS (hht1-hhf1)Δ::NatMX ura3Δ0
KA125	MATα (hht2-hhf2)::hhts-K14A-URA3/HHFS (hht1-hhf1)Δ::NatMX paf1Δ::KanMX ura3-52
KA126	MATα (hht2-hhf2)::hhts-K4,9,14,18A-URA3/HHFS (hht1-hhf1)Δ::NatMX ura3-52
KA127	MATα (hht2-hhf2)::hhts-K4,9,14,18A-URA3/HHFS (hht1-hhf1)Δ::NatMX paf1Δ::KanMX ura3-52
KA175	MATa (hht2-hhf2)::hhts-K4,9,14,18A-URA3/HHFS (hht1-hhf1)Δ::NatMX paf1Δ::KanMX gcn4Δ::KanMX ura3-52
KA176	MATa (hht2-hhf2)::hhts-K4,9,14,18A-URA3/HHFS (hht1-hhf1)Δ::NatMX paf1Δ::KanMX gcn4Δ::KanMX ura3-52
KA177	MATa (hht2-hhf2)::hhts-K4,9,14,18A-URA3/HHFS (hht1-hhf1)Δ::NatMX paf1Δ::KanMX gcn4Δ::KanMX ura3-52
KA178	MATα (hht2-hhf2)::hhts-K4,9,14,18A-URA3/HHFS (hht1-hhf1)Δ::NatMX paf1Δ::KanMX gcn4Δ::KanMX ura3-52
KA179	MATα (hht2-hhf2)::hhts-K4,9,14,18A-URA3/HHFS (hht1-hhf1)Δ::NatMX paf1Δ::KanMX gcn4Δ::KanMX ura3-52
KY803	MATα his3Δ200 lys2-173R2 leu2Δ(0 or 1) ura3(Δ0 or 52)
KY1348	MATα gcn5Δ::HIS3 paf1Δ::URA3 his3Δ200 leu2Δ1 ura3-52
KY1699	MATα
KY1700	MATα paf1Δ::KanMX
KY1701	MATapaf1Δ::KanMX leu2Δ0
KY1702	MATapaf1Δ::KanMX leu2Δ0 ura3Δ0
KY1703	MATartf1Δ::KanMX
KY1704	MATα rtf1Δ::KanMX
KY1708	MATα gcn4Δ::KanMX
KY1709	MATα arg80Δ::KanMX
KY1719	MATα paf1Δ::KanMX gcn4Δ::KanMX
KY1720	MATapaf1Δ::KanMX arg80Δ::KanMX
KY1728	MATα GCN4-3×HA::KanMX
KY1729	MATaGCN4-3×HA::KanMX arg80Δ::KanMX
KY1730	MATaGCN4-3×HA::KanMX paf1Δ::KanMX arg80Δ::KanMX
KY1736	MATα ARG1p-HIS31-663 his3Δ::NatMX
KY1737	MATaARG1p-HIS31-663 his3Δ::NatMX paf1Δ::KanMX
KY2022	MATaGCN4-3×HA::KanMX HTA1-htb1-K123R (hta2-htb2)Δ::KanMX
KY2023	MATα GCN4-3×HA::KanMX HTA1-htb1-K123R (hta2-htb2)Δ::KanMX
KY2024	MATaGCN4-3×HA::KanMX HTA1-htb1-K123R (hta2-htb2)Δ::KanMX
KY2112	MATa (lyp1−290-1836)Δ::ARG1−497-1263 arg1Δ::NatMX
KY2115	MATa (lyp1−290-1836)Δ::ARG1−497-1263 arg1Δ::NatMX paf1Δ::KanMX
KY2118	MATα (lyp1−290-2059)Δ::ARG1−497-1553 arg1Δ::NatMX
KY2121	MATα (lyp1−290-2059)Δ::ARG1−497-1553 arg1Δ::NatMX paf1Δ::KanMX
KY2182	MATα ARG1p-HIS31-663 his3Δ::NatMX GCN4-3×HA::KanMX
KY2183	MATaARG1p-HIS31-663 his3Δ::NatMX GCN4-3×HA::KanMX
KY2184	MATα ARG1p-HIS31-663 his3Δ::NatMX GCN4-3×HA::KanMX
KY2185	MATα ARG1p-HIS31-663 his3Δ::NatMX GCN4-3×HA::KanMX paf1Δ::KanMX
KY2186	MATaARG1p-HIS31-663 his3Δ::NatMX GCN4-3×HA::KanMX paf1Δ::KanMX
KY2187	MATα ARG1p-HIS31-663 his3Δ::NatMX GCN4-3×HA::KanMX paf1Δ::KanMX
KY2188	MATα GCN4-3×HA::KanMX (hta2-htb2)Δ::KanMX
KY2189	MATα GCN4-3×HA::KanMX (hta2-htb2)Δ::KanMX
OKA178	MATaARG80-13×MYC::HIS3 GCN4-3×HA::KanMX his3Δ0

To create strains in which the ARG1 promoter and coding region were integrated at the LYP1 locus, an ARG1 fragment (−497 to +1263) was amplified by PCR with primers that permitted replacement of the LYP1 gene (−290 to +1836), such that the ARG1 promoter and coding region were adjacent to the LYP1 3′ untranslated region (UTR). For strains in which the LYP1 promoter, coding region, and 3′ UTR were replaced with those of the ARG1 gene, an ARG1 fragment (−497 to +1553) was amplified by PCR to allow replacement of the LYP1 gene (−290 to +2059). PCR products were transformed into arg1Δ::NatMX haploid strains. Transformants were selected on SC medium lacking arginine, and proper integration was confirmed by using PCR analysis and by measuring resistance to thialysine. The resulting strains were mated to paf1Δ::KanMX strains to obtain lyp1Δ::ARG1 arg1Δ::NatMX paf1Δ::KanMX strains.

To create strains containing the HIS3 coding region under the control of the ARG1 promoter and 3′ UTR, the HIS3 coding region (+1 to +663) was amplified from wild-type genomic DNA with primers that permitted replacement of the ARG1 coding region with that of the HIS3 gene. The purified PCR product was transformed into a his3Δ::NatMX haploid strain. Transformants were selected on SC medium lacking histidine. The resulting strain was mated to paf1Δ::KanMX strains to obtain ARG1p-HIS3_1-663 his3Δ::NatMX paf1Δ::KanMX strains.

Northern analysis.

Total RNA was isolated from cells grown in YPD medium at 30°C to a density of 1 × 10⁷ to 2 × 10⁷ cells/ml. Ten micrograms of total RNA was subjected to Northern analysis with random prime-labeled DNA probes for the ARG1 (+34 to +1201), HIS3 (−27 to +376), and SCR1 (−242 to +283) genes as described previously (15). A phosphorimager and ImageQuant software were used to quantify signals. ARG1 or HIS3 signals were normalized to the SCR1 gene, which serves as a loading control. The relative ARG1 signal in arg80Δ samples (see Fig. 1A; also data not shown), which were processed in parallel with experimental samples, was arbitrarily set equal to one. The relative HIS3 signal in wild-type samples was set equal to one. Quantitations of Northern analyses represent the means from at least three independent experiments. Error bars represent one standard deviation of the mean in Fig. 1A and B, 2, and 4C to E.

Fig 1.

Paf1 mediates ARG1 repression independently of Arg80 but inhibits Gcn4 association with the ARG1 promoter. (A) Quantitation of Northern analysis of ARG1 transcript levels in wild-type (KY1699), paf1Δ (KY1702), paf1Δ arg80Δ (KY1720), and arg80Δ (KY1709) strains relative to a loading control, SCR1. Transcript levels were quantified and normalized to the levels detected in arg80Δ (KY1709) cells as described in Materials and Methods. (B) Quantitation of Northern analysis of relative ARG1 transcript levels in wild-type (KY1699), paf1Δ (KY1702), paf1Δ gcn4Δ (KY1719), and gcn4Δ (KY1708) strains. (C) ChIP analysis of Gcn4 occupancy at the ARG1 promoter in wild-type (KY1728), paf1Δ (KA116), paf1Δ arg80Δ (KY1730), and arg80Δ (KY1729) strains expressing HA-tagged Gcn4 and an untagged Gcn4 control strain (KA119). The y axis was truncated to facilitate comparison of lower values. The value for the paf1Δ arg80Δ strain is indicated above the bar. (D) ChIP analysis of Gcn4 occupancy at the HIS7, TRP2, and ARO4 promoters in wild-type (KY1728) and paf1Δ (KA116) strains expressing HA-tagged Gcn4 and an untagged Gcn4 control strain (KA119). (E) Strand-specific RT-PCR examining HIS7, TRP2, and ARO4 sense transcript levels in wild-type (KY1699) and paf1Δ (KY1700) cells. ACT1 is used as a positive control for RT-PCR experiments. A minus indicates a reaction mixture lacking reverse transcriptase. Results are representative of those obtained in two independent experiments. Relative to the ATG for each gene, the 5′ ends of cDNA synthesis primers map to HIS7 (+969), TRP2 (+1142), and ARO4 (+613). Primers used in PCR amplify HIS7 (+633 to +853), TRP2 (+745 to +967), or ARO4 (+327 to +573). WT, wild type.

Fig 2.

Mutation of H3 acetylation sites partially restores ARG1 repression in the absence of Paf1. (A) Quantitation of Northern analysis of relative ARG1 expression in wild-type (KY205), paf1Δ (KY803), and paf1Δ gcn5Δ (KY1348) strains. (B) Quantitation of Northern analysis of ARG1 transcript levels in wild-type (KY1699), HHTS (hht1-hhf1)Δ (KA120), hhts-K9A (hht1-hhf1)Δ (KA122), hhts-K14A (hht1-hhf1)Δ (KA124), hhts-K4, 9, 14, 18A (hht1-hhf1)Δ (KA126), paf1Δ (KY1700), HHTS paf1Δ (hht1-hhf1)Δ (KA121), paf1Δ hhts-K9A (hht1-hhf1)Δ (KA123), paf1Δ hhts-K14A (hht1-hhf1)Δ (KA125), paf1Δ hhts-K4, 9, 14, 18A (hht1-hhf1)Δ (KA127), and paf1Δ gcn4Δ hhts-K4,9,14,18A (hht1-hhf1)Δ (KA175-179) strains. WT, wild type.

Fig 4.

Paf1 mediates repression through the ARG1 promoter. (A) Representation of the ARG1p-HIS3 gene where the ARG1 coding region was replaced with that of the HIS3 gene. (B) Left, Northern analysis examining HIS3 mRNA levels when expressed from the endogenous HIS3 chromosomal location in wild-type (KY1699), paf1Δ (KY1700), and rtf1Δ (KY1703) cells and when the HIS3 coding region is fused to the ARG1 promoter (ARG1p-HIS3) in wild-type (KY1736), paf1Δ (KY1737), and rtf1Δ (KY2175) cells. Right, Northern analysis of ARG1 transcript levels in wild-type (KY1699), paf1Δ (KY1700), and rtf1Δ (KY1704) strains. The SCR1 gene serves as a loading control. Results shown are representative of at least three independent experiments. (C) Quantitation of HIS3 transcript levels in wild-type (KY1699), paf1Δ (KY1700), and rtf1Δ (KY1703) cells in which HIS3 is expressed from its normal chromosomal location. (D) Quantitation of HIS3 transcript levels in wild-type (KY1736), paf1Δ (KY1737), and rtf1Δ (KY2175) cells in which the HIS3 coding region is under the control of the ARG1 promoter (ARG1p-HIS3). (E) Quantitation of ARG1 transcript levels in wild-type (KY1699), paf1Δ (KY1700), and rtf1Δ (KY1704) strains. These quantified data were published previously (reprinted from Fig. 1B in reference 15) and are shown here for clarity. (F) ChIP analysis of Gcn4 occupancy at the ARG1 promoter in wild-type (OKA178), paf1Δ (KA116), ARG1p-HIS3 (KY2182-84), and ARG1p-HIS3 paf1Δ (KY2185-87) strains expressing HA-tagged Gcn4 and an untagged control strain (KY1699). WT, wild type.

Chromatin immunoprecipitation (ChIP) assays.

Cells were grown to a density of 1 × 10⁷ to 2 × 10⁷ cell/ml in YPD medium and harvested for the isolation of chromatin as previously described (67). Immunoprecipitation (IP) of sonicated chromatin was performed as previously described (67). Agarose-conjugated antihemagglutinin (anti-HA; Santa Cruz Biotechnology) was used to immunoprecipitate a C-terminally 3× HA-tagged form of Gcn4. Input and IP DNA were used as templates in PCR mixtures containing [α-³²P]dATP. PCR primers amplified the ARG1 promoter (−450 to −200 relative to the translation start site). PCR products were resolved on 6% native polyacrylamide gels, and signals were quantified using a phosphorimager and ImageQuant software (Fig. 1C). Alternatively, input and IP DNA were used as templates in quantitative real-time PCRs with SYBR green detection (Fermentas) using PCR primers that amplified the ARG1 promoter (−383 to −221), the HIS7 promoter (−231 to −118), the TRP2 promoter (−180 to −78), the ARO4 promoter (−391 to −302), or the ARG3 promoter (−371 to −242). IP/input signals for ARG1, HIS7, TRP2, ARO4, and ARG3 genes were normalized to a subtelomeric control region on chromosome VI. The means from three independent experiments were plotted with standard errors of the means in Fig. 1C and D, 3C, 4F, and 8G.

Fig 3.

Changes in nucleosome occupancy accompany ARG1 transcription, and the H2B-K123R substitution leads to increased Gcn4 levels at the ARG1 promoter. (A) Top left, nucleosome scanning assay examining nucleosome occupancy at the ARG1 promoter and 5′ region in wild-type strains in repressing and inducing conditions. Wild-type (KY1699) cells were grown to log phase in SC-IV medium and mock treated with DMSO (WT) or treated with SM for 2 h (WT + SM) prior to cross-linking. Mononucleosomal DNA was prepared and subjected to real-time PCR analysis using primers that tile a chromosomal region containing the ARG1 promoter. The midpoint position of each PCR fragment relative to the translation start site is plotted on the x axis. Relative MNase protection is plotted on the y axis. Bottom, schematic representation of predicted nucleosome occupancy. Shaded circles represent nucleosomes. Top right, representative Northern blot examining ARG1 transcript levels in strains harvested for RNA prior to cross-linking. (B) Nucleosome occupancy at the ARG1 promoter region in wild-type (KY1699) and paf1Δ (KY1700 and KY1701) strains grown in YPD. Samples were processed and relative MNase protection calculated as in (A). Right: representative Northern analysis of ARG1 mRNA levels immediately prior to cross-linking. (C) ChIP analysis of Gcn4 occupancy at the ARG1 promoter in wild-type (OKA178), paf1Δ (KA116), (hta2-htb2)Δ (KY2188-89), and (hta2-htb2)Δ htb1-K123R (KY2022-2024) strains expressing HA-tagged Gcn4 and an untagged control strain (KY1699).

Fig 8.

Paf1 restricts antisense transcription and Gcn4 occupancy at the promoters of other Paf1-repressed genes. (A) Diagram of the locations of primers used for cDNA synthesis and PCR for strand-specific RT-PCR analysis of sense and antisense transcription at the SNZ1 locus. (B) Strand-specific RT-PCR examining SNZ1 sense transcription in wild-type (KY1699) and paf1Δ (KY1700) strains. Results are representative of two independent experiments. (C) Strand-specific RT-PCR examining antisense transcription across the SNZ1 promoter in wild-type (KY1699) and paf1Δ (KY1700) strains. Results are representative of two independent experiments. (D) Diagram of the locations of primers used for cDNA synthesis and PCR for strand-specific RT-PCR analysis of sense and antisense transcription at the ARG3 locus. (E) Strand-specific RT-PCR examining ARG3 sense transcription in wild-type (KY1699) and paf1Δ (KY1700) strains. The data shown are representative of two independent experiments. (F) Strand-specific RT-PCR examining antisense transcription across the ARG3 promoter in wild-type (KY1699) and paf1Δ (KY1700) strains. Results are representative of two independent experiments. (G) ChIP analysis of HA-Gcn4 occupancy at the ARG3 promoter in wild-type (OKA178) and paf1Δ (KA116) cells expressing HA-Gcn4 and a strain expressing untagged Gcn4 (KA119). WT, wild type.

Nucleosome scanning assays.

Cells were grown to a density of 2 × 10⁷ cells/ml in YPD or SC-IV medium with 0.6 μg/ml SM or dimethyl sulfoxide (DMSO) as a control. Nucleosome scanning assays were performed as previously described (27, 63). Control genomic DNA and mononucleosomal DNA, which was generated by digestion with micrococcal nuclease (MNase), was used as a template in real-time PCR using SYBR green (Fermentas) detection. PCR primers were designed to tile a region containing the ARG1 promoter from the translation stop codon of the GPD2 gene to approximately 500 bases into the ARG1 coding region. PCR amplicons were approximately 100 bp in length with approximately 70 bp of overlap with neighboring amplicons. ARG1 signals were normalized to a well-positioned nucleosome in the GAL1-10 promoter (7, 27, 41). Relative MNase protection for each primer set obtained from at least three independent experiments was plotted at the midpoint of the amplicon. Error bars represent one standard deviation of the mean in Fig. 3A and B.

Strand-specific RT-PCR.

Strand-specific reverse transcription-PCR (RT-PCR) was performed as previously described (57). Briefly, total RNA was DNase treated with Turbo DNase I (Ambion) at 37°C for 30 min and purified with a Qiagen RNeasy cleanup kit. Two micrograms of RNA was used in a strand-specific cDNA synthesis reaction with SuperScript II RT (Invitrogen) or a no-RT control reaction; both of these contained a primer designed to reverse transcribe the RNA of interest and a primer designed to reverse transcribe the ACT1 mRNA. cDNA synthesis or no-RT control reaction mixtures were amplified by PCR with GoTaq polymerase (Promega). At least two volumes of cDNA were used as templates in PCRs to ensure signal linearity. For analysis of HIS7, TRP2, and ARO4 gene expression, 4 μl of no-RT reaction mixtures and 0.5-μl, 2-μl, and 4-μl volumes of undiluted cDNA were used in PCRs to amplify HIS7, TRP2, and ARO4 genes. The same volumes of 1:2 dilutions of cDNA were used in PCRs to amplify the ACT1 gene. For analysis of ACT1 and ARG1 sense transcription, 3 μl of no-RT reaction mixtures and 1-μl and 3-μl volumes of 1:2 dilutions of cDNA were used in PCRs to amplify the ACT1 and ARG1 genes. Unless otherwise stated in the figure legend, 3 μl of no-RT reaction mixtures and 1- and 3-μl volumes of undiluted cDNA were used in PCRs to detect antisense transcription at the native ARG1 or ARG1p-HIS3 genes. Four microliters of no-RT reaction mixtures and 0.5-μl, 2-μl, and 4-μl volumes of undiluted or 1:2 dilutions of cDNA were used in PCRs to examine SNZ1 sense and ACT1 transcript levels, respectively. Eight microliters of no-RT reaction mixtures and 2-μl, 4-μl, and 8-μl volumes of undiluted cDNA were used in PCRs to examine SNZ1 antisense transcript levels. Six microliters of no-RT reaction mixtures and 1-μl, 3-μl, and 6-μl volumes of 1:2 dilutions of cDNA were used in PCRs to examine ARG3 sense, ARG3 antisense, and ACT1 transcript levels. For quantitation of antisense transcription traversing the ARG1 promoter, strand-specific cDNA was used in real-time PCR using SYBR green (Fermentas) detection. ARG1 antisense transcription was normalized to the ACT1 gene. The relative transcript level in paf1Δ cells was set arbitrarily equal to one.

RESULTS

Paf1 inhibits association of Gcn4 with the ARG1 promoter.

We previously showed that members of Paf1C contribute to ARG1 repression to various degrees (15). For example, deletion of the PAF1 gene leads to a greater increase in ARG1 transcript levels than deletion of the RTF1 gene (15). To investigate further the repressive functions of Paf1C, we asked whether Paf1 mediates repression of the ARG1 gene by promoting the activity of the ArgR/Mcm1 repressor complex. To address this question, we measured ARG1 mRNA levels by Northern analysis of paf1Δ, arg80Δ, and paf1Δ arg80Δ strains grown in rich medium. Deletion of the PAF1 or ARG80 gene individually resulted in ARG1 derepression as expected (Fig. 1A). However, ARG1 transcript levels in paf1Δ arg80Δ strains were significantly higher than that in either single-deletion strain, suggesting that Paf1 and Arg80 mediate repression through distinct pathways (Fig. 1A).

In an alternative mechanism, Paf1 could promote ARG1 repression by inhibiting the activator Gcn4. To test this possibility, ARG1 transcript levels were measured in strains lacking the PAF1 gene, the GCN4 gene, or both genes. We found that deletion of the GCN4 gene suppressed the effect of deleting the PAF1 gene. ARG1 mRNA levels in the paf1Δ gcn4Δ double mutant were approximately 70% lower than those in the paf1Δ single mutant, indicating that the full level of ARG1 derepression that occurs in the absence of Paf1 requires Gcn4 (Fig. 1B).

To test if Paf1 influences Gcn4 occupancy at the ARG1 promoter, ChIP analysis was performed using strains expressing HA-tagged Gcn4 or an untagged control strain. Gcn4 levels are low in cells grown in rich medium, but ChIP analysis was sufficiently sensitive to detect HA-tagged Gcn4 at the ARG1 promoter under these conditions (Fig. 1C). Interestingly, Gcn4 occupancy was increased in both paf1Δ and arg80Δ strains compared to that in a wild-type strain, and paf1Δ arg80Δ strains exhibited higher Gcn4 occupancy than that of either single mutant (Fig. 1C). These results show that in paf1Δ strains ARG1 derepression is associated with increased promoter association of Gcn4 and suggest that Paf1 and Arg80 independently modulate Gcn4 levels at the ARG1 promoter.

We also analyzed Gcn4 association and transcript levels at several other Gcn4-regulated genes to address the possibility that the elevated promoter occupancy of Gcn4 at the ARG1 gene was due to increased GCN4 expression in paf1Δ strains. We found that deletion of the PAF1 gene did not increase Gcn4 occupancy at the promoters of HIS7, TRP2, and ARO4 genes, nor did it lead to increased expression of these genes as measured by RT-PCR (Fig. 1D and E). In addition, deletion of the PAF1 gene had little, if any, effect on GCN4 transcript levels (data not shown). Therefore, our current data do not support the idea that the paf1Δ mutation causes increased GCN4 expression and, as a consequence, increased Gcn4 occupancy at the ARG1 promoter.

Histone H3 acetylation sites are required for ARG1 derepression in the absence of Paf1.

Gcn4 stimulates ARG1 transcription by recruiting the Gcn5-containing histone acetyltransferase (HAT) complex SAGA, thereby promoting histone acetylation and recruitment of SWI/SNF, TBP, and RNA Pol II (25, 26, 38, 39, 59, 70). Therefore, Paf1 might repress ARG1 transcription by restricting Gcn4 occupancy and subsequent histone acetylation by Gcn5. Consistent with this hypothesis, we found that the full level of ARG1 derepression that occurs in paf1Δ cells requires Gcn5 (Fig. 2A).

Using strains in which specific histone H3 lysines were changed to alanines, we asked if Gcn5-targeted acetylation sites on H3 are required for ARG1 derepression in paf1Δ cells. In these strains, the HHT1-HHF1 genes have been deleted and HHT2-HHF2 genes have been replaced with a synthetic version, the HHTS-HHFS genes (16). Substitution of the HHT2-HHF2 genes with the HHTS-HHFS genes had no effect on ARG1 expression in the presence of Paf1 and did not significantly reduce ARG1 expression in paf1Δ cells (Fig. 2B). ARG1 transcript levels in hhts-K9A, hhts-K14A, and hhts-K4,9,14,18A strains were similar to those in HHTS strains, indicating that the acetylation of these H3 residues is not required for ARG1 repression in the presence of Paf1 (Fig. 2B). Interestingly, these H3 substitutions significantly restored repression in paf1Δ strains (Fig. 2B). Moreover, the reduction in ARG1 transcript levels caused by the hhts-K4,9,14,18A mutation was not further decreased by the deletion of the GCN4 gene (Fig. 2B). Together, these results suggest that Paf1 represses ARG1 transcription, at least in part, by inhibiting Gcn4 association at the ARG1 promoter and Gcn4-dependent modifications on histone H3. In contrast, mutation of histone H4 acetylation sites (hhfs-K12A, hhfs-K16A, and hhfs-K5,8,12,16R) did not restore repression in paf1Δ strains (data not shown).

Changes in nucleosome occupancy accompany ARG1 expression.

To investigate the possibility that Paf1 restricts promoter accessibility at the ARG1 gene by regulating nucleosome occupancy or positioning, we performed nucleosome scanning assays (63) using PCR primers that tile across the ARG1 promoter and 5′ coding region. We first examined changes in nucleosome occupancy that occur upon transcriptional activation, using wild-type strains grown to log phase in SC-IV medium and treated with SM, which increases cellular levels of Gcn4 (31), or mock treated with DMSO. In mock-treated wild-type strains, the nucleosome scanning assay revealed a region with relatively low MNase protection encompassing the ARG1 regulatory elements. This nucleosome-deficient region was flanked by peaks of high MNase protection (Fig. 3A). These results match well with genome-wide analyses of nucleosome occupancy in wild-type cells grown in rich media (32). Upon treatment of cells with SM, ARG1 transcript levels were increased (Fig. 3A, right), and the peaks of MNase protection were reduced (Fig. 3A), revealing that transcriptional induction in wild-type cells disrupts nucleosome occupancy and positioning at the ARG1 promoter and 5′ region.

To determine whether Paf1 influences nucleosome occupancy at the ARG1 gene in repressing conditions, wild-type and paf1Δ cells were grown to log phase in YPD medium and harvested for the nucleosome scanning assay. The relative MNase protection profile for wild-type cells grown in YPD medium (Fig. 3B) was nearly identical to that of wild-type cells grown in SC-IV medium (Fig. 3A). Interestingly, although the loss of Paf1 results in derepression of the ARG1 gene (Fig. 3A, right), the relative MNase protection profile in paf1Δ strains mostly resembled that of wild-type cells grown in rich medium. However, a slight reduction in occupancy and forward shifting of the nucleosome positioned over the ARG1 TATA box was observed (Fig. 3B). Therefore, as measured by the nucleosome-scanning assay, deletion of the PAF1 gene causes small changes in the nucleosome occupancy pattern at the ARG1 promoter.

H2B K123 ubiquitylation restricts Gcn4 occupancy at the ARG1 promoter.

Because Paf1C-dependent H2B K123 ubiquitylation is required for ARG1 repression (15, 72), we asked whether the loss of this modification alters Gcn4 occupancy at the ARG1 promoter. ChIP analysis of HA-tagged Gcn4 was performed on strains that are deleted of the HTA2-HTB2 genes and contain the HTA1-HTB1 genes or HTA1-htb1-K123R genes. The htb1-K123R gene encodes a derivative of H2B that cannot be ubiquitylated on K123. Deletion of HTA2-HTB2 genes alone did not influence Gcn4 occupancy at the ARG1 gene (Fig. 3C, compare columns 1 and 3). However, similar to deletion of the PAF1 gene, the H2B-K123R substitution resulted in increased Gcn4 levels at the ARG1 promoter (Fig. 3C). Unlike results with the paf1Δ strain, Gcn4 occupancy was also higher at the ARO4 promoter in the htb1-K123R strain than in the control strain (data not shown), suggesting that, with respect to H2B K123 ubiquitylation, the effects of deleting the PAF1 gene are less severe than the effects of eliminating the ubiquitylation site. However, arguing against a general effect of the H2B-K123R substitution on Gcn4-regulated genes or Gcn4 protein levels, Gcn4 occupancy was not significantly elevated at the HIS7 or TRP2 gene in the htb1-K123R strain, and levels of expression of a GCN4-lacZ reporter construct (30, 48) were similar in the htb1-K123R and wild-type control strains (data not shown). Taken together, our results indicate that H2B K123 ubiquitylation restricts Gcn4 occupancy at the ARG1 promoter and provide a molecular explanation for the repressive functions of this modification and Paf1C at the ARG1 gene.

Paf1 mediates ARG1 repression through the promoter region.

If its impact on Gcn4 occupancy at the ARG1 gene constitutes an important repressive role for Paf1C, then the sequences required for Paf1-mediated repression should reside within the promoter. To test this idea, we substituted the ARG1 coding region with the HIS3 coding region at the endogenous ARG1 locus (Fig. 4A). HIS3 transcript levels were examined by Northern analysis in strains containing the ARG1p-HIS3 gene or the HIS3 gene at its normal location. The native HIS3 gene was expressed in wild-type cells, with a slight reduction or increase in transcript levels observed in the absence of Paf1 or Rtf1, respectively (Fig. 4B and C). However, when the HIS3 coding region was under the control of the ARG1 promoter, the HIS3 gene was strongly repressed in wild-type cells and derepressed in the absence of Paf1 or Rtf1 (Fig. 4B and D). Interestingly, paf1Δ and rtf1Δ cells exhibited very similar increases in ARG1p-HIS3 expression (Fig. 4B and D), whereas transcript levels from the native ARG1 gene are consistently higher in paf1Δ cells than in rtf1Δ cells (Fig. 4B and E) (15). At the promoters of both the ARG1 and ARG1p-HIS3 genes, deletion of the PAF1 gene resulted in increased Gcn4 occupancy (Fig. 4F). Taken together, these findings confirm the importance of the ARG1 promoter in mediating Paf1C-dependent repression of the ARG1 gene but also suggest that the ARG1 coding region is required for the full level of derepression observed in paf1Δ cells.

Paf1 prevents antisense transcription from traversing the ARG1 promoter.

Our previous analyses of the histone modification pattern at the ARG1 locus in repressing conditions suggested the possibility of antisense transcription. Specifically, we found histone H3 K4 trimethylation levels were highest at the 3′ end of the ARG1 gene, and H3 K36 trimethylation levels were highest at the 5′ end, a pattern that is reversed from that typically observed at active genes (15). Furthermore, previous genome-wide studies reported antisense transcripts traversing the ARG1 coding region in wild-type cells (18, 74, 80, 81).

To both confirm the presence of antisense transcription at the ARG1 locus and determine whether it is influenced by Paf1, we performed strand-specific RT-PCR assays to examine sense and antisense transcription at the ARG1 locus. As expected, a cDNA synthesis primer complementary to the ARG1 mRNA detected the derepression of ARG1 sense transcription that occurs in paf1Δ strains (Fig. 5A and B). To detect antisense transcription at the ARG1 locus, we used several different cDNA synthesis primers within the promoter and coding region (Fig. 5A). Using cDNA synthesis primers downstream of the ARG1 transcription start site (primers E and F), we detected antisense transcription that traverses the ARG1 coding region in wild-type and paf1Δ strains at similar levels (Fig. 5C). Using cDNA synthesis primers upstream of the ARG1 transcription start site (primers A to D), we found that antisense transcription within the ARG1 promoter occurs only at very low levels in wild-type cells (Fig. 5D). Interestingly, antisense transcription within the ARG1 promoter was significantly increased in paf1Δ cells compared to that in wild-type cells (Fig. 5D). These results observed with ethidium bromide-stained gels were confirmed using quantitative real-time PCR analysis of strand-specific cDNA generated using primer B (Fig. 5E). Together, these results confirm the presence of antisense transcription at the ARG1 locus and suggest that Paf1 prevents antisense transcription across the ARG1 promoter.

Fig 5.

Paf1 inhibits antisense transcription across the ARG1 promoter. (A) Diagram showing the locations of primers used in cDNA synthesis (arrows) and the PCR amplicon (bar) used for strand-specific RT-PCR experiments. (B) Strand-specific RT-PCR examining ARG1 sense transcription in wild-type (KY1699) and paf1Δ (KY1700) strains. Results shown are representative of more than three independent experiments. (C) Strand-specific RT-PCR examining antisense transcription within the ARG1 transcribed region (primers E and F) in wild-type (KY1699) and paf1Δ (KY1700) strains. Results shown are representative of two independent experiments. (D) Strand-specific RT-PCR examining antisense transcription within the ARG1 promoter (primers A to D) in wild-type (KY1699) and paf1Δ (KY1700) strains. Results shown are representative of at least two independent experiments. A minus indicates control reaction mixtures lacking reverse transcriptase. (E) Real-time PCR analysis of strand-specific cDNA generated using primer B in wild-type (KY1699) and paf1Δ (KY1700) strains. Results shown are the averages from two independent experiments. Error bars represent the ranges in the data. WT, wild type.

Antisense transcription at the ARG1 promoter requires the ARG1 coding region but not the 3′ UTR.

To begin to address the importance of antisense transcription at the ARG1 locus, we attempted to block antisense transcription by eliminating possible start sites for the antisense transcript(s). Previous genome-wide studies identified antisense transcripts arising from the ARG1 3′ UTR in wild-type cells (18, 80, 81). Therefore, we decided to integrate the ARG1 promoter and coding region at the LYP1 locus, such that the ARG1 promoter and coding region were fused to the LYP1 3′ UTR, which does not contain known start sites for antisense transcripts (18, 80) (Fig. 6A). As a control, we created a yeast strain in which the ARG1 promoter, coding region, and 3′ UTR were integrated at the LYP1 locus (Fig. 6A). The LYP1 3′ UTR did not affect the requirement for Paf1 in repression of ARG1 sense transcription (Fig. 6B). Surprisingly, however, strand-specific RT-PCR using a primer within the coding region detected antisense transcription in both wild-type and paf1Δ cells when the ARG1 promoter and coding region were adjacent to either the ARG1 or LYP1 3′ UTR (Fig. 6C). Furthermore, increased antisense transcription still occurred within the ARG1 promoter in paf1Δ cells, regardless of which 3′ UTR was adjacent to the coding region (Fig. 6D). Because replacement of the ARG1 3′ UTR with that of LYP1 did not eliminate antisense transcription across the ARG1 promoter and coding region, some antisense transcripts likely arise from start sites within the ARG1 coding region.

Fig 6.

The ARG1 3′ UTR is not required for antisense transcription. (A) Diagram of the ARG1 regions integrated at the LYP1 locus in the same orientation as the LYP1 gene. Strand-specific RT-PCR examining ARG1 sense transcription (B) and antisense transcription within the ARG1 open reading frame (C) and promoter (D) in wild-type (KY2112) and paf1Δ (KY2115) cells in which the ARG1 promoter and coding region are fused to the LYP1 3′ UTR and in wild-type (KY2118) and paf1Δ (KY2121) cells in which the ARG1 promoter, coding region, and 3′ UTR were integrated at the LYP1 locus. cDNA synthesis primers used to detect antisense transcription are noted and correspond to those shown in Fig. 5A. For analysis of antisense transcription with primer E, 6 μl of no-RT reaction mixtures (−) and 2-μl and 6-μl volumes of cDNA were used in PCRs to amplify the ARG1 gene. Results are representative of two independent experiments. (E) Locations of cDNA synthesis and PCR primers used to detect antisense transcription at the ARG1p-HIS3 promoter. (F) Strand-specific RT-PCR analysis to detect antisense transcription traversing the ARG1p-HIS3 promoter in wild-type (KY1736) and paf1Δ (KY1737) cells (left). ACT1 is used as a positive control for RT-PCR experiments (right). A minus indicates a reaction mixture lacking reverse transcriptase. WT, wild type.

Replacement of the ARG1 coding region with that of the HIS3 gene largely maintained Paf1-dependent repression (Fig. 4). However, as revealed by the similar levels of ARG1p-HIS3 expression in paf1Δ and rtf1Δ strains (Fig. 4B and D), the ARG1 coding region plays some role in supporting the high levels of ARG1 derepression typically observed for paf1Δ strains (Fig. 4B and E). To determine whether a loss in antisense transcription across the ARG1 promoter correlated with the reduction in ARG1p-HIS3 derepression in paf1Δ strains, we performed a strand-specific RT-PCR analysis of this hybrid gene (Fig. 6E and F). We found that reaction mixtures containing the same amount of cDNA used to detect antisense transcription at the native ARG1 locus did not detect antisense transcription traversing the promoter in ARG1p-HIS3 strains, even in paf1Δ cells (Fig. 6F). We confirmed that our PCR primers amplify genomic DNA and found that a very faint signal corresponding to antisense transcription at the ARG1p-HIS3 locus could be detected only if the amount of cDNA used in PCRs was doubled (data not shown). Our results demonstrate that replacement of the ARG1 coding region with the HIS3 coding region significantly reduces antisense transcription across the ARG1 promoter and indicate that the ARG1 coding region, possibly through supporting antisense transcription, is required for the full level of ARG1 derepression characteristic of paf1Δ strains. Further, the increase in Gcn4 occupancy observed at the ARG1p-HIS3 locus in paf1Δ cells (Fig. 4F) suggests that Paf1 restricts promoter association of Gcn4 and antisense transcription at the ARG1 promoter through different mechanisms.

ARG1 sense transcription positively correlates with antisense transcription across the ARG1 promoter.

Our data suggest a model in which antisense transcripts arising from start sites within the ARG1 coding region traverse the ARG1 promoter and promote ARG1 expression in paf1Δ cells. If antisense transcription over the ARG1 promoter and ARG1 sense transcription are positively correlated, then ARG1 sense transcription could also influence antisense transcription. To examine this possibility, we lowered the level of ARG1 sense transcription by deleting the GCN4 gene and examined the effect on antisense transcription. Consistent with our Northern analysis, strand-specific RT-PCR detected a reduced level of ARG1 sense transcription in paf1Δ cells when the GCN4 gene was deleted (Fig. 7A). As detected by conventional PCR (Fig. 7B) and quantitative real-time PCR (Fig. 7C), deletion of the GCN4 gene also reduced the level of antisense transcription traversing the ARG1 promoter in paf1Δ cells, but it did not affect antisense transcript levels within the ARG1 coding region (data not shown). These results suggest that ARG1 sense transcription positively influences antisense transcription specifically across the ARG1 promoter.

Fig 7.

ARG1 sense transcription positively correlates with antisense transcription across the ARG1 promoter. (A) Strand-specific RT-PCR examining ARG1 sense transcription in wild-type (KY1699), paf1Δ (KY1702), paf1Δ gcn4Δ (KY1719), and gcn4Δ (KY1708) cells. The cDNA synthesis primer used to detect sense transcription is depicted in Fig. 5A. (B) Strand-specific RT-PCR examining antisense transcription across the ARG1 promoter in wild-type (KY1699), paf1Δ (KY1702), paf1Δ gcn4Δ (KY1719), and gcn4Δ (KY1708) cells. The cDNA synthesis primer (primer B) used to detect antisense transcription is depicted in Fig. 5A. Results shown are representative of two independent experiments. A minus indicates reaction mixtures lacking reverse transcriptase. (C) Real-time PCR analysis of strand-specific cDNA generated using primer B in wild-type (KY1699), paf1Δ (KY1702), paf1Δ gcn4Δ (KY1719), and gcn4Δ (KY1708) cells. Results shown are the averages from two independent experiments. Error bars represent the range in the data. Strand-specific RT-PCR analysis of ARG1 sense (D) and antisense (E) transcription in wild-type (KY1699) cells grown in SC-IV medium and mock treated with DMSO (WT) or treated with SM for 2 h (WT + SM). Results are representative of two independent experiments. WT, wild type.

To determine whether increased antisense transcription across the ARG1 promoter generally accompanies an increase in ARG1 sense transcription, we examined ARG1 antisense transcription in wild-type cells treated with SM. Strand-specific RT-PCR analysis detected the increased ARG1 sense transcription that occurs upon SM treatment (Fig. 7D). Interestingly, increased ARG1 sense transcription was also associated with increased antisense transcription traversing the ARG1 promoter in induced wild-type cells (Fig. 7E). These results indicate that increased ARG1 sense transcription, whether from deletion of the PAF1 gene or SM treatment, is accompanied by an increase in antisense transcription across the ARG1 promoter.

Paf1 restricts antisense transcription and Gcn4 association at the promoters of other Paf1-repressed genes.

As shown above, ARG1 derepression in paf1Δ cells is associated with increased antisense transcription and Gcn4 occupancy at the promoter. To determine whether similar events occur at other Paf1-repressed genes, we examined antisense transcription and activator binding at the promoters of two other Gcn4-regulated genes, the ARG3 and SNZ1 genes, which have been shown through microarray analyses to be repressed by Paf1 (56). Strand-specific RT-PCR analysis confirmed increased SNZ1 and ARG3 sense transcription in paf1Δ cells (Fig. 8A, B, D and E). Interestingly, similar to results with the ARG1 gene, we observed increased antisense transcription across the SNZ1 and ARG3 promoters in paf1Δ cells (Fig. 8C and F). To determine if the derepression of the SNZ1 and ARG3 genes is associated with increased Gcn4 occupancy at their promoters, we performed ChIP analyses using strains expressing HA-tagged Gcn4 or an untagged control strain. At the SNZ1 promoter, Gcn4 was not detected in either wild-type or paf1Δ strains at levels above that for the untagged control strain (data not shown). However, deletion of the PAF1 gene resulted in increased Gcn4 occupancy at the ARG3 promoter relative to that for wild-type cells (Fig. 8G). These results indicate that Paf1 represses multiple genes through similar mechanisms.

DISCUSSION

Here, we describe a detailed investigation of transcriptional repression by Paf1C, focusing on the Paf1 subunit and using the ARG1 model gene. Our results are consistent with the involvement of Paf1C in two modes of repression. First, by promoting H2B K123 ubiquitylation, Paf1 inhibits the association of Gcn4 with the ARG1 promoter. Second, Paf1 directly or indirectly restricts antisense transcription across the ARG1 promoter, and this antisense transcription is positively correlated with ARG1 sense transcription. Conditions that stimulate ARG1 sense transcription also lead to an increase in promoter-associated antisense transcription, suggesting the existence of a positive regulatory loop. These mechanisms of repression are not limited to the ARG1 gene, as we observed increased Gcn4 association and antisense transcription at the promoters of other genes in paf1 mutant strains.

Northern analysis of single and double mutant strains demonstrated that derepression of ARG1 transcription in paf1Δ cells was strongly dependent on Gcn4, and by ChIP analysis, we observed increased promoter occupancy of Gcn4 in these cells. Furthermore, deletion of GCN5 or mutation of histone H3 acetylation sites significantly restored ARG1 repression in paf1Δ strains. Therefore, by limiting Gcn4 occupancy at the ARG1 promoter, Paf1 likely inhibits a previously defined pathway in which Gcn4 recruits Gcn5, resulting in H3 acetylation and ARG1 expression. Based on previous studies, H3 acetylation may be required for ARG1 derepression in paf1Δ cells by stimulating recruitment of the RNA Pol II general transcription machinery to the ARG1 promoter and/or facilitating transcription elongation (25, 39, 70).

Apart from a slight reduction in occupancy and forward movement of the nucleosome positioned over the ARG1 TATA box, we observed little change in chromatin architecture within the ARG1 promoter or 5′ coding region in paf1Δ cells. In contrast, ARG1 expression was accompanied by dramatic alterations in nucleosome occupancy within the promoter and 5′ coding region in wild-type cells grown in inducing conditions. Therefore, our nucleosome scanning studies have revealed a mechanistic distinction between the ARG1 derepression that occurs in paf1Δ cells and the ARG1 induction occurring in wild-type cells. In addition, our analysis provides important new information on the chromatin structure at the ARG1 gene, an extensively employed model gene about which little is known with respect to nucleosome positioning under activating conditions.

We and others have previously shown that Paf1C-dependent histone modifications, specifically H2B K123 ubiquitylation and H3 K4 methylation, are important for ARG1 repression (15, 72). Similar to what we observed with paf1Δ cells, the loss of H2B K123 ubiquitylation resulted in increased Gcn4 occupancy at the ARG1 promoter. Genome-wide transcriptional analyses have demonstrated that H2B K123 ubiquitylation represses many genes in yeast (4, 49), and reversal of H2B K123 ubiquitylation by the deubiquitylating enzyme Ubp8 is required for full expression of inducible genes (17, 28, 33). These observations clearly indicate a repressive role for H2B K123 ubiquitylation, but the molecular basis for this mode of repression had not been defined. Our data suggest that H2B K123 ubiquitylation, mediated by Paf1C, can inhibit the promoter association of trans-acting factors. This effect is unlikely to be limited to Gcn4. We detected increased occupancy of Arg80, but not a complete ArgR/Mcm1 complex, at the ARG1 promoter in paf1Δ cells despite derepression of the gene (data not shown), and TBP levels are elevated at the ARG1 promoter in a rad6Δ strain (72). Although the Gcn4 and Arg80 binding sites are located in a region of relatively low nucleosome occupancy within the ARG1 promoter, the dynamic stability of nucleosomes within this region could influence protein binding. Indeed, H2B K123 ubiquitylation has been shown to increase nucleosome stability at the promoters of repressed genes (9), and Batta et al. very recently described a genome-wide role for H2B K123 ubiquitylation in promoting nucleosome occupancy (4). Batta et al. also reported an inhibitory effect of this modification on RNA Pol II recruitment at genes expressed at low levels (4). Interestingly, the slight changes in nucleosome occupancy detected at the ARG1 promoter in paf1Δ strains by our nucleosome scanning assays are consistent with the weak effects of paf1Δ on nucleosome patterns genome-wide compared to the effects of an H2B-K123A mutant (4).

The reversed patterns of histone modifications previously observed at the ARG1 gene led us to investigate the possibility of antisense transcription (15). In accordance with genome-wide studies reporting antisense transcripts traversing the ARG1 coding region in wild-type cells (18, 74, 80, 81), we detected antisense transcription at the ARG1 gene in both wild-type and paf1Δ cells. Therefore, the overall histone modification pattern detected at the ARG1 gene in repressing conditions likely reflects both antisense transcription and low levels of sense transcription (Fig. 5). Although genome-wide analyses identified antisense transcripts arising from the ARG1 3′ UTR (18, 80, 81), we found that replacement of the ARG1 3′ UTR did not eliminate antisense transcription. Therefore, antisense transcription apparently arises from multiple start sites located within both the ARG1 3′ UTR and coding region. Consistent with this idea, replacement of the ARG1 coding region with that of the HIS3 gene severely reduced antisense transcription across the coding region.

Our data demonstrate that the ARG1 coding region is required for the full level of derepression observed for paf1Δ cells. Although we cannot rule out the possibility that another feature of the coding region, such as sequence or nucleosome positioning, confers the high level of derepression observed in the absence of Paf1, our data are consistent with the idea that antisense transcription across the ARG1 promoter bolsters ARG1 sense transcription. Antisense transcription has been shown to stimulate sense transcription at other genes, such as the yeast PHO5 gene (73). Reciprocally, conditions that stimulate ARG1 sense transcription (i.e., in paf1Δ cells or in wild-type cells treated with SM) lead to an increase in antisense transcription across the ARG1 promoter. These observations demonstrate that increased antisense transcription across the promoter is not uniquely due to the absence of Paf1. Instead, this appears to be a more general consequence of increased sense transcription. Taken together, our results raise the possibility that sense and antisense transcription at the ARG1 gene engage in a self-reinforcing feedback loop to promote full expression of the gene.

The mechanisms that prevent antisense transcription from traversing the ARG1 promoter in repressing conditions remain to be elucidated. In the case of Paf1C, it is possible that previously described functions in histone modification and transcription termination play a role. For example, the loss of histone modifications in paf1Δ cells could lead to increased levels of cryptic transcripts oriented in the antisense direction. Alternatively, defects in transcription termination and RNA 3′ end formation previously observed with paf1Δ strains could lead to the extension of antisense transcripts into the promoter region. The generation of extended antisense transcripts in wild-type cells grown in inducing conditions is less easily explained, but it seems reasonable to speculate that the gross changes in nucleosome architecture occurring upon gene induction influence the initiation and/or termination of these transcripts. An important goal of future studies will be to determine the mechanisms that regulate the patterns of antisense transcription at ARG1 and other genes.