Gds1 Interacts with NuA4 To Promote H4 Acetylation at Ribosomal Protein Genes

ABSTRACT

In our previously published studies, RNA polymerase II transcription initiation complexes were assembled from yeast nuclear extracts onto immobilized transcription templates and analyzed by quantitative mass spectrometry. In addition to the expected basal factors and coactivators, we discovered that the uncharacterized protein Gds1/YOR355W showed activator-stimulated association with promoter DNA. Gds1 coprecipitated with the histone H4 acetyltransferase NuA4, and its levels often tracked with NuA4 in immobilized-template experiments. GDS1 deletion led to a reduction in H4 acetylation in vivo and caused other phenotypes consistent with a partial loss of NuA4 activity. Genome-wide chromatin immunoprecipitation revealed that the reduction in H4 acetylation was strongest at ribosomal protein gene promoters and other genes with high NuA4 occupancy. Therefore, while Gds1 is not a stoichiometric subunit of NuA4, we propose that it interacts with and modulates NuA4 in specific promoter contexts. Gds1 has no obvious metazoan homolog, but the Alphafold2 algorithm predicts that a section of Gds1 resembles the winged-helix/forkhead domain found in DNA-binding proteins such as the FOX transcription factors and histone H1.

INTRODUCTION

Gene expression by RNA polymerase II (RNApII) involves multiple steps and many factors. Transcription is repressed by chromatin occlusion of promoter DNA sequences, so gene activation often requires nucleosome removal or sliding carried out by the Swi/Snf complex and other ATP-dependent chromatin remodelers (1). The subsequent exposure of the core promoter sequences allows RNApII and the basal transcription factors to assemble a preinitiation complex (PIC). This chromatin derepression is targeted to specific promoters by transcription activators that bind near regulatory promoter sequences known as enhancers or upstream activating sequences (UASs). In addition to Swi/Snf and the basal transcription machinery, transcription activators can also recruit the histone acetyltransferases (HATs) SAGA and NuA4 (1, 2). The resulting promoter-localized acetylations are recognized by complexes having one or more bromodomains (3). This group includes Swi/Snf, SAGA, and TFIID, thereby creating a positive-feedback mechanism to reinforce acetylation and displacement of the promoter-occluding nucleosomes.

Both histones H3 and H4 are highly acetylated in active promoter regions (4, 5). However, these two histones are modified by distinct HATs: SAGA and NuA3 are responsible for H3 acetylation (H3ac), while NuA4 targets H4 (6). Interestingly, both complexes contain the Tra1 protein, which directly interacts with some transcription activation domains, providing a simple mechanism for the coordinated targeting of H3 and H4 acetylation (7). Deletion or point mutations in the H3 and H4 N-terminal tails, where modification sites are concentrated, or their respective HATs suggest distinct functions for H3ac and H4ac. This difference is presumably because acetylated H3 and H4 tails are “read” by distinct complexes (8). For example, the bromodomain protein Bdf1, a component of both TFIID and the SWR-C complex, preferentially binds acetylated H4 (9, 10), while Swi/Snf recognizes acetylated H3 (11).

We have been studying transcription initiation and elongation using mass spectrometry analysis of transcription complexes assembled from nuclear extracts onto immobilized DNA templates (2, 12–14). In addition to the expected transcription factors, we found that the uncharacterized protein Gds1 was recruited to DNA in an activator-stimulated manner. GDS1 was first identified as a high-copy-number suppressor of slow growth caused by nam9-1, a mutant allele of a mitochondrial ribosomal protein gene (RPG) (15). While Gds1 may therefore have a mitochondrial function, other data suggest a nuclear function. Localization studies show the protein in both the nucleus and cytoplasm (16), and many of the reported physical and genetic interactions of Gds1 are with nuclear proteins (17–19). Particularly relevant to this study, Gds1 was identified as a possible substrate for the NuA4 HAT (20).

Following up on these results, we show here that Gds1 genetically and physically interacts with NuA4. Genome-wide chromatin immunoprecipitation sequencing (ChIP-seq) analysis reveals that Gds1 promotes promoter-proximal H4 acetylation at a subset of genes targeted by NuA4, in particular RPGs and others activated by Rap1. Although there is no obvious metazoan homolog of Gds1, structural predictions suggest that Gds1 contains a winged-helix/forkhead domain, similar to those found in many DNA-binding factors. While the molecular function of Gds1 remains unclear, our results will help focus future experiments toward solving this mystery.

RESULTS

Activator-stimulated recruitment of Gds1 to promoters.

We have used an immobilized-template assay to isolate and analyze RNApII initiation (2, 12, 13) and elongation (14, 21) complexes in vitro. In these experiments, RNApII PICs were assembled from a yeast nuclear extract onto a template containing one Gal4-binding site upstream of the HIS4 core promoter region (Fig. 1A). Digestion with PstI at a unique site upstream of the promoter released the downstream DNA template and associated proteins (Fig. 1B). As previously reported (2), quantitative mass spectrometry shows Gal4-VP16 activator-stimulated binding of RNApII, the basal initiation factors, Mediator, and the coactivators Swi/Snf, SAGA, and NuA4 (Fig. 1C). The relative fold enrichment ratios of the coactivators were higher than that of RNApII and its initiation factors, consistent with a basal level of PIC assembly on the naked DNA template.

FIG 1.

Gds1 is recruited to promoters by the transcription activator Gal4-VP16. (A) Schematic of DNA templates used for immobilized-template analysis. Biotinylated fragments were PCR amplified from pSH515 (HIS4pro), pSH514 (HIS4pro.mTATA), or pUC18-G5CYC1 G− (CYC1pro) using oligonucleotides listed in Table S2 in the supplemental material. WT, wild type. (B) Workflow diagram for the isolation and characterization of RNApII PICs. (C) Volcano plot showing activator-dependent enrichment (x axis) (the vertical dashed line indicates the 95% confidence level using mixed-model analysis) of individual proteins on the HIS4pro template versus statistical significance (y axis) (the horizontal dashed line shows a P value of 0.05). Quantitative mass spectrometry samples from two biological replicates were analyzed with two technical replicates each. Each dot represents the averaged value for a single protein. Subunits within individual complexes are color-coded as indicated. (D) Immunoblots (IB) from immobilized-template experiments showing that the DNA template, transcription activator, and TATA box contribute to PIC and Gds1 binding. Note that SAGA (Ada3) and NuA4 (Esa1) are recruited by the activator but are not affected by the TATA box mutation. (E) Immunoblots from immobilized-template experiments showing that Gds1 is recruited to both the HIS4 and CYC1 core promoters.

In addition to the expected transcription factors, we also repeatedly observed activator-dependent enrichment of Gds1/YOR355W, a protein of unknown function. To validate the mass spectrometry results, a yeast nuclear extract containing tandem affinity purification (TAP)-tagged Gds1 was used for PIC formation and immunoblot assays. Gds1 associated with immobilized transcription template DNA but not beads alone (Fig. 1D). Gds1 binding to DNA was stimulated by Gal4-VP16 and decreased by mutation of the HIS4 TATA box, similar to the pattern seen for RNApII and several basal transcription factors. NuA4 (Esa1) and SAGA (Ada3) coactivators were also increased by the activator. Gds1 association was not specific for the HIS4 promoter, as similar activator-stimulated binding was also seen on the CYC1 promoter (Fig. 1E). The immobilized-template results suggest that Gds1 may somehow be linked to the RNApII transcription machinery.

Gds1 interacts with the NuA4 complex.

High-throughput synthetic-lethality studies reported genetic interactions between GDS1 deletion and those for factors related to RNApII transcription, RNA processing, DNA replication and repair, and ribosomal biogenesis (17–19) (https://www.yeastgenome.org/locus/S000005882/interaction). Interactors included genes for Eaf1, Eaf3, and Yng2, three nonessential subunits of the NuA4 HAT complex (19). Another study found that Gds1 not only physically interacts with NuA4 but also was a substrate for its lysine acetyltransferase activity (20). We confirmed the interaction between Gds1 and NuA4 by coprecipitation of the NuA4 catalytic subunit Esa1 with Gds1-TAP (Fig. 2A).

FIG 2.

Gds1 interacts with the NuA4 complex. (A) Extracts carrying wild-type or TAP-tagged Gds1 were incubated with IgG beads, and pellets were probed for the NuA4 subunit Esa1 by immunoblotting. (B) A gds1Δ strain was tested alongside NuA4 subunit deletion strains for sensitivity to methyl methanesulfonate (MMS) (0.02%, vol/vol) or rapamycin (25 mM) by serial dilution spotting. (C and D) Quantitative mass spectrometry showing that Gds1 resembles NuA4 subunits in differentially associating with VP16 and Gcn4 activation domains (C) and in response to NTPs (D).

To test if Gds1 might have a function related to NuA4, a gds1Δ strain was tested for phenotypes seen upon the deletion of nonessential NuA4 subunits, including sensitivity to rapamycin and the alkylating agent methyl methanesulfonate (MMS) (22). Like eaf3Δ, eaf5Δ, eaf6Δ, eaf7Δ, yng2Δ, and yaf9Δ strains, a gds1Δ strain had slowed growth in the presence of MMS (Fig. 2B). It also showed sensitivity to rapamycin similar to those of the eaf6Δ and eaf7Δ strains but less so than the other NuA4 subunit deletions.

Previous immobilized-template experiments (2) showed that some coactivators have differential responses to two different activation domains. SAGA and Swi/Snf were strongly recruited by Gal4 DNA-binding domain fusions to either the VP16 or Gcn4 activation domain. In contrast, NuA4 responded much more strongly to VP16 than Gcn4 (2). In a plot of the mass spectrometry enrichment ratios for these two activators (Fig. 2C), most activator-recruited factors lie below the diagonal, signifying a stronger response to Gcn4. In contrast, Gds1 is found with the NuA4 subunits (purple spots) above the diagonal, indicating a preference for the VP16 activation domain. Gds1 also behaves similarly to NuA4 subunits in its response to nucleotide triphosphates (NTPs) (Fig. 2D, +NTP). Based on all of these observations, we propose that Gds1 interacts with NuA4 in the context of transcription.

Gds1 affects H4 acetylation at NuA4 target genes.

To test for a functional linkage between Gds1 and the NuA4 complex, acetylations of histones H3 and H4 were examined in lysates from wild-type and gds1Δ cells. Normalized to total histones (H3 and H2B), H4ac was reduced by 62% (±9%) in gds1Δ cells (Fig. 3A). In contrast, H3ac was unaffected. This H4ac-specific effect is consistent with a change in NuA4 activity.

FIG 3.

Gds1 promotes histone H4 acetylation. (A) Immunoblotting of whole-cell extracts from GDS1 or gds1Δ cells was carried out using the indicated antibodies. (B) Promoter-proximal H4 acetylation was decreased by the loss of Gds1. Heat maps show data from ChIP-seq analysis for H4ac (orange) and H3 (gray) compared in GDS1 or gds1Δ cells. Each horizontal line represents an individual gene, and genes were ordered by the H4ac signal in the GDS1 strain. For the three panels shown for each immunoprecipitation (IP), the two left panels show the relative levels, while the right panel shows the difference (change = gds1Δ − GDS1, where blue color intensity represents negative values and orange represents positive values). (C) Metagene analysis of H4 acetylation and H3 data shown in panel B. TSS, transcription start site.

To further probe this effect, H4ac was mapped using genome-wide chromatin immunoprecipitation sequencing (ChIP-seq) in isogenic GDS1 and gds1Δ strains. Both heat maps (Fig. 3B) and metagene analyses (Fig. 3C) showed that the H4 acetylation peaks at the 5′ regions of most genes were somewhat decreased. In contrast, total nucleosome occupancy, monitored using anti-H3 antibody, showed no significant change by GDS1 deletion.

The magnitude of the H4ac change for each gene was plotted against its overall acetylation level (Fig. 4A). Although there was some reduction of H4ac at most genes (median gds1Δ/GDS1 log₂ ratio of 0.89), the contribution of Gds1 to acetylation appeared strongest at the most highly acetylated genes. In particular, we found that the ribosomal protein genes (RPGs) (red spots in Fig. 4A) made up many of the most severely affected genes. Inspection of individual RPG profiles generally showed a pronounced decrease of the H4ac peak overlapping the position of the +1 nucleosome (Fig. 4B). In many cases, this drop was accompanied by an increase in the total H3 signal at that location, consistent with the role of H4ac in targeting chromatin remodelers and other transcription factors to the 5′ ends of genes. The drop in H4ac at RPGs was verified using standard ChIP-PCR (Fig. 4C and D). Consistent with immunoblotting (Fig. 3A), H3ac was not similarly affected, with the possible exception of RPL16A (Fig. 4D).

FIG 4.

Gds1 stimulation of H4ac is particularly strong at ribosomal protein genes (RPGs). (A) Plot of H4ac levels (log₂ of the ChIP-seq read counts) versus the effect of GDS1 deletion on acetylation (gds1Δ/GDS1 ratio of read counts, expressed as log₂ values). Each spot represents one gene, and red spots designate RPGs. (B) Browser tracks of H3 (black, GDS1; gray, gds1Δ) and H4ac (red, GDS1; orange, gds1Δ) for RPL9B, RPS17B, RPS25B, RPS3, and RPL16A. Green lines bracket the position of the +1 nucleosomes. Axis scales for any given protein or modification are the same between conditions. (C, left) Browser tracks for RPS13, as described above for panel B. (Right) Standard ChIP-PCR, with a schematic of probes at the top. Note that the values for H3ac and H4ac have been normalized to total H3. Error bars designate standard deviations from three replicates. *** designates a P value of <0.01. (D) Validation of ChIP-seq data at the 8 indicated promoters, as for panel C.

RPGs are among the most highly expressed and most TFIID-dependent (23–26) genes in yeast. To determine whether either of these two properties correlated with Gds1 function, nonribosomal protein genes were sorted into quintiles based on their mRNA expression level (Fig. 5A) or decrease in a taf1-1 mutant (Fig. 5B). In both analyses, the change in H4ac levels upon GDS1 deletion was far less than that at RPGs, and there was no statistically significant difference between quintiles. Therefore, Gds1 dependence does not simply correlate with strong expression or a requirement for TBP-associated factor (TAF) function.

FIG 5.

The Gds1 effect on H4ac is strong at Rap1-binding sites but is not simply correlated with transcription frequency or TFIID dependence. (A) Box-and-whisker plots showing the gds1Δ/GDS1 ratio for RPGs or non-RPGs sorted into quintiles by transcript abundance. (B) Similar analysis, sorting non-RPGs by the magnitude of the transcription decrease upon TAF1 mutation (23). (C) Metagene analysis for Rap1-binding sites (left) or tRNAs (right), mapping ChIP-seq data from Fig. 3.

The transcription factor Rap1 binds and activates many RPG promoters. Moreover, Rap1 is believed to recruit the NuA4 complex such that high H4ac levels are observed around Rap1-binding sites (27, 28). A metagene analysis for H3 and H4ac centered around Rap1-binding sites showed a significant decrease of H4ac and an increase of H3 in gds1Δ cells (Fig. 5C, left). In contrast, no similar effect was seen at tRNA loci, which also have high H4ac levels (Fig. 5C, right).

We next asked whether the response to Gds1 correlated with NuA4 occupancy. ChIP-seq data for the Epl1 subunit of NuA4 (29) were plotted against the levels of H4ac (Fig. 6A). Surprisingly, a clear linear correlation was not seen. However, RPGs were clearly clustered among genes that had both the highest H4ac levels and NuA4 occupancy (Fig. 6A, left, red dots). When the response to gds1 deletion was mapped onto this plot (Fig. 6A, right, color gradient), it was clear that genes with the strongest Gds1 stimulatory effect on H4ac (green and blue dots) were highly enriched among those with the highest levels of NuA4. This effect is also apparent in heat maps of individual genes (Fig. 6B). Plotting the H4ac change for genes with the highest levels (the top 4%) of Epl1 shows that they respond to Gds1 nearly as strongly as RPGs (Fig. 6C).

FIG 6.

The Gds1 effect on H4ac is strongest on promoters with high NuA4 occupancy. (A) Scatterplot of NuA4 occupancy (Epl1 ChIP signal reads [29]) versus H4ac. The left panel shows each promoter as a spot, with RPGs highlighted in red. The right panel shows the same data, with the effect of Gds1 on H4 acetylation (log₂ GDS1/gds1Δ ratio) color-coded. Note that the most Gds1-dependent genes (green-blue) are those with the highest Epl1 levels. (B) Heat maps showing H4ac (orange) on individual genes sorted by Epl1 occupancy (purple). The rightmost heat map shows the relative change upon GDS1 deletion, with dark blue indicating no change and white representing the strongest loss. (C) Box-and-whisker plots showing the gds1Δ/GDS1 ratios for all genes, RPGs, and Epl1 target genes (defined as the top 4% of genes sorted by the Epl1 ChIP signal at the promoter [29]). **, P < 0.05; ***, P < 0.01. (D) Standard ChIP-quantitative PCR (qPCR) was performed for the indicated proteins on the indicated RPG promoters, using chromatin from GDS1 and gds1Δ cells. The gds1Δ/GDS1 ratios are plotted, with error bars showing standard deviations from three replicates.

We next asked whether Gds1 and NuA4 affected each other’s binding to promoters. Unfortunately, efforts to chromatin immunoprecipitate Gds1-TAP were unsuccessful. However, GDS1 deletion did not affect the cross-linking of Rap1 or the NuA4 subunit Esa1 to several RPG promoters that had reduced H4ac (Fig. 6D). Interestingly, RNApII subunit Rpb3 ChIP was also reduced at several RPGs in the gds1Δ strain, suggesting that transcription was reduced. Based on these results, as well as the experiments described above, we propose that Gds1 promotes NuA4 activity, but not recruitment, at a subset of RNApII-transcribed genes with high NuA4 levels, particularly at RPGs.

Nuclear extracts were prepared from isogenic GDS1 and gds1Δ strains to test for effects on in vitro transcription. Both extracts were active and responsive to Gal4-VP16 (see Fig. S1A in the supplemental material). Linearized HIS4 promoter DNA was assembled into chromatin using recombinant histones and used for immobilized-template experiments that included acetyl coenzyme A (acetyl-CoA) (Fig. S1B and C). No differences were observed in activator-stimulated binding of RNApII or TFIIH (Kin28). Furthermore, Gal4-VP16 stimulated H4 acetylation, presumably through NuA4 recruitment, but this effect was independent of Gds1 (Fig. S1D). Therefore, our current in vitro system is unable to reproduce the in vivo effects of Gds1 on histone acetylation, precluding further biochemical analysis.

Gds1 is predicted to have a winged-helix/forkhead domain.

Although most NuA4 subunits are conserved over the eukaryotic lineage, a BLASTP search found no metazoan proteins related to Gds1. Homologs were detected among other fungal species, with conservation primarily mapping to a central region encompassing amino acids 90 to 230 (Fig. S2A). A subset of the aligned proteins also showed weaker sequence similarity near the C terminus of the protein.

Conserved regions often correspond to structural units. Three secondary structure prediction programs (Phyre2 [30], Robetta [31], and I-TASSER [32]) all predicted four or five alpha helices within the central conserved domain (Fig. S2B). Recently, the Alphafold2 algorithm (33) was shown to have a superior ability to predict three-dimensional (3D) structures. The regions of Gds1 predicted by this program to be structured (https://alphafold.ebi.ac.uk/entry/P41913) showed a striking correspondence to the BLAST alignments. The weakly conserved C-terminal region may have several helices, and an extensive fold is predicted in amino acids 90 to 230 (Fig. 7A). The predicted Gds1 central domain comprises five alpha helices followed by two antiparallel beta sheets and a final alpha helix (Fig. 7B).

FIG 7.

Gds1 is predicted to contain a forkhead/winged-helix domain. (A) Predicted aligned error plot from the Alphafold2 prediction for Gds1 (https://alphafold.ebi.ac.uk/entry/P41913). Green represents the expected error in the distance predicted between any two amino acid positions plotted on the x and y axes. Darker color signifies a higher confidence in the relative position. (B) The FOXP2 forkhead/winged-helix domain (magenta) (PDB accession number 2A07, chain J) and the predicted structure of Gds1 residues 80 to 230 (green) are shown separately and superimposed (center). See Fig. S3B in the supplemental material for a similar superposition of the forkhead/winged-helix domains from Gds1 and histone H1. aa, amino acids.

We compared this central Alphafold2-predicted domain of Gds1 to known protein structures using PDBeFold (https://www.ebi.ac.uk/msd-srv/ssm/). The default settings produced multiple strong matches (root mean square deviation [RMSD] of the backbone of <2.3 Å) to the forkhead/winged-helix domain of FOX family transcription factors, specifically FOXP2, FOXC2, and FOXN3 (Fig. S3A). A superposition of the predicted Gds1 structure and the known FOXP2 structure is shown in Fig. 7B. By slightly relaxing the search parameters to require alignment over 50% of the protein length rather than the default 70%, similarity to many additional winged-helix proteins was found, with a particularly strong match (RMSD of 1.726 Å) to histone H1/H5 (Fig. S3B). In the context of both the FOX/forkhead factors and histone H1/H5, this winged-helix domain binds DNA, suggesting that Gds1 may directly contact DNA.

Gds1 regions outside the structured domains are made up of low-complexity sequences highly enriched for charged and polar residues and are predicted to be largely disordered. Interestingly, lysine residues within a disordered region of Gds1 are acetylated by NuA4 in vitro, although the functional relevance of these modifications is unclear (20). In our immobilized-template experiments, the absence or presence of acetyl-CoA did not change the amounts of Gds1 bound to promoters, arguing that acetylation is not required for Gds1 binding (Fig. S1E).

DISCUSSION

Genetic and biochemical studies have revealed a surprisingly large number of basal factors, coactivators, and chromatin-related proteins needed for activator-stimulated transcription. To study these factors as an integrated system, we have been using yeast nuclear extracts to assemble complexes on both naked DNA and chromatin templates (2, 12–14). Our proteomic analyses of initiation and elongation complexes suggest that nuclear extracts effectively reproduce many aspects of in vivo transcription. In addition to all of the expected coactivators and basal transcription factors, transcription on immobilized templates also recapitulates the cycle of RNApII phosphorylations driving the exchange of elongation and mRNA processing factors (14). Few unexpected or unidentified proteins were similarly enriched, suggesting that most or all of the key factors have been discovered.

One exception is Gds1, a poorly characterized protein that reproducibly showed activator-stimulated association with transcription templates (Fig. 1). GDS1 was first identified as a high-copy-number suppressor of nam9-1, a mutation in a mitochondrial ribosomal protein that cannot grow on media containing glycerol as a carbon source (15). Although Gds1 protein can be detected in mitochondria (34), it is located predominantly in the cytoplasm and nucleus (16). A GDS1 deletion shows synthetic sick and lethal phenotypes, with many nuclear proteins linked to transcription and ribosome biogenesis, including components of the NuA4 HAT complex (17–19).

The results presented here further connect Gds1 with NuA4. We confirm (Fig. 2A) a previous report that Gds1 coprecipitates with NuA4 (20) and show that the loss of Gds1 reduces H4 acetylation in vivo (Fig. 3 and 4). Although Gds1 is not an integral subunit of NuA4 (35), we find that it often behaved similarly to NuA4 subunits during quantitative mass spectrometry of transcription complexes (Fig. 2B and C). Strong drops in H4ac upon GDS1 deletion correlate with high NuA4 occupancy, but not all promoters with high H4ac levels depend on Gds1 (Fig. 3B and Fig. 4A). Furthermore, NuA4 recruitment does not seem to be reduced upon GDS1 deletion (Fig. 6D). Therefore, Gds1 may regulate NuA4 activity in a manner sensitive to promoter context. Consistent with this possibility, Gds1 binding to immobilized templates was increased by the presence of the TATA-containing core promoter, in this respect resembling PIC components more than NuA4. Unfortunately, while our in vitro transcription system exhibits activator-stimulated H4 acetylation, it apparently does not reproduce the stimulatory effect of Gds1 (see Fig. S1 in the supplemental material).

We found that ribosomal protein gene promoters, which typically have some of the highest histone acetylation levels, are highly enriched among the promoters most strongly affected by the loss of Gds1 (Fig. 5 and 6). RPGs comprise a unique regulon designed to tune mRNA and protein expression to the available amounts of the RNA polymerase I-transcribed rRNA (36). This balance is sensitive to the availability of nutrients and cellular growth and division rates. This signaling may even underlie the original genetic interaction between gds1Δ and a mitochondrial ribosomal protein mutation. However, the Gds1 effect on RPG H4ac is not simply an indirect effect of slower growth, as gds1Δ cells grow at normal rates on rich media (Fig. 2B).

Searches for homologs that might supply further insight into Gds1 function were mostly uninformative. However, structural prediction programs suggest that Gds1 contains a forkhead/winged-helix domain, which may suggest a direct interaction with DNA. Depending on the parameters used, Gds1 showed the closest structural similarity to the FOX family transcription factors or the histone H1/H5 family. The FOX proteins recognize specific DNA sequences, and many even bind in the context of a nucleosome to function as “pioneer” transcription factors (37). It will be interesting to see if Gds1 recognizes particular sequences, either at the UAS or near the core promoter, that might explain why some promoters are more strongly affected by its removal. Alternatively, Gds1 may resemble histone H1/H5 by binding DNA in a non-sequence-specific manner in the context of a NuA4-bound nucleosome, perhaps even occupying a space similar to that of histone H1 at the nucleosome dyad. While these are purely speculative ideas at this point, the information uncovered in this study should provoke future experiments that reveal more about how Gds1 fits into gene regulation.

MATERIALS AND METHODS

Strains, plasmids, and oligonucleotides.

Saccharomyces cerevisiae strains, oligonucleotides, and plasmids used in this study are listed in Tables S1 to S3 in the supplemental material.

Yeast techniques.

For phenotypic analyses, liquid cell cultures of the indicated yeast strains grown overnight were normalized for cell density, serially diluted (3-fold in each step), and spotted onto the indicated media.

Whole-cell extracts were made by glass bead lysis as previously described (38). For testing protein interactions, IgG-agarose beads were added to bind Gds1-TAP, and after washing, associated proteins were analyzed by elution in loading buffer. For all immunoblot assays, proteins were separated by SDS–10% polyacrylamide gel electrophoresis, transferred to an Immobilon P membrane (Millipore Sigma), and immunoblotted with the designated primary antibodies (listed in Table S4) and the appropriate secondary antibody conjugated with horseradish peroxidase. Detection was performed by chemiluminescence using the Pierce SuperSignal West detection reagent.

Immobilized-template binding assay.

Immobilized-template experiments were carried out using yeast nuclear extracts as previously described (2, 13, 14, 21). Briefly, the indicated DNA templates were prepared by PCR, with one primer containing a biotin residue for linkage to magnetic beads (MyOne Strep Dynabeads; Invitrogen). Immobilized templates underwent incubation with yeast nuclear extracts, followed by washes to remove unbound proteins and isolation of template-bound proteins using magnetic concentration. To analyze bound proteins by standard immunoblotting, the beads were resuspended in loading buffer and run on SDS-polyacrylamide gels. The antibodies used are listed in Table S4. For quantitative mass spectrometry, proteins were eluted from beads using PstI and subjected to multiplexed iTRAQ or tandem mass tag (TMT) labeling, followed by 3D liquid chromatography-tandem mass spectrometry (LC-MS/MS) analysis. Note that the quantitative mass spectrometry figures in this paper show reanalyses of data originally generated by Sikorski et al. (2) (Fig. 2B) and Joo et al. (14, 21) (Fig. 2C).

Chromatin immunoprecipitation and ChIP-seq.

Chromatin immunoprecipitations were performed using antibodies against Rpb3 (monoclonal, clone 1Y26, catalog number W0012; Neoclone), Rap1 (gift of Tony Weil, Vanderbilt University), histone H3 (catalog number ab1791; Abcam), acetylated H3 (catalog number 06-599; Millipore), or acetylated histone H4 (catalog number 06-598; Millipore). NuA4 ChIP was done using TAP-tagged Esa1 strains (Table S1) and IgG-Sepharose (catalog number 95017-050; Cytiva via VWR). Genome-wide data were generated and analyzed as previously described (13, 39). For gene-specific ChIPs, standard PCR was performed using primers listed in Table S2.

Data availability.

Genome-wide data have been deposited in the GEO database under accession number GSE185227.