Distinct functions of three chromatin remodelers in activator binding and preinitiation complex assembly

The nucleosome remodeling complexes (CRs) SWI/SNF, RSC, and Ino80C cooperate in evicting or repositioning nucleosomes to produce nucleosome depleted regions (NDRs) at the promoters of many yeast genes induced by amino acid starvation. We analyzed mutants depleted of the catalytic subunits of these CRs for binding of transcriptional activator Gcn4 and recruitment of TATA-binding protein (TBP) during preinitiation complex (PIC) assembly. RSC and Ino80 were found to enhance Gcn4 binding to both UAS elements in NDRs upstream of promoters and to unconventional binding sites within nucleosome-occupied coding sequences; and SWI/SNF contributes to UAS binding when RSC is depleted. All three CRs are actively recruited by Gcn4 to most UAS elements and appear to enhance Gcn4 binding by reducing nucleosome occupancies at the binding motifs, indicating a positive regulatory loop. SWI/SNF acts unexpectedly in WT cells to prevent excessive Gcn4 binding at many UAS elements, indicating a dual mode of action that is modulated by the presence of RSC. RSC and SWI/SNF collaborate to enhance TBP recruitment at Gcn4 target genes, together with Ino80C, in a manner associated with nucleosome eviction at the TBP binding sites. Cooperation among the CRs in TBP recruitment is also evident at the highly transcribed ribosomal protein genes, while RSC and Ino80C act more broadly than SWI/SNF at the majority of other constitutively expressed genes to stimulate this step in PIC assembly. Our findings indicate a complex interplay among the CRs in evicting promoter nucleosomes to regulate activator binding and stimulate PIC assembly.


Introduction
In the yeast Saccharomyces cerevisiae, most genes transcribed by RNA Polymerase II (Pol II) display a stereotypical pattern of nucleosome organization with a nucleosome-depleted region (NDR) of~120bp situated upstream of the coding sequences (CDS) and flanked by highly positioned "-1" and "+1" nucleosomes, with the transcription start site (TSS) generally located within the +1 nucleosome. Promoter elements and upstream activation sequences (UAS elements) reside within the NDR and may extend upstream into the -1 nucleosome (-1_Nuc) [1][2][3][4]. The exclusion of nucleosomes from the NDRs can facilitate efficient binding of transcriptional activator proteins at UAS elements [5][6][7]. The location of the TSS within the +1 nucleosome dictates that the latter is frequently evicted [8,9], or shifted in the 3' direction [10][11][12], during assembly of the Pol II transcription preinitiation complex (PIC).
Promoter nucleosome organization arises in part from cooperative or antagonistic actions of ATP-dependent chromatin remodeling (CR) complexes [13][14], which are recruited to NDRs or promoter-proximal nucleosomes of many yeast genes [14,15]. The CRs RSC and SWI/SNF generally move the -1 and +1 nucleosomes away from NDRs. RSC has been shown to function in this way at the majority of yeast genes [16][17][18] to maintain native NDR widths [19,20]. SWI/SNF, in contrast, acts in the promoters of a small fraction of genes that tend to have poor nucleosome phasing and wide NDRs [14,19] and to be highly expressed in WT cells [12]; and SWI/SNF and RSC exhibit functional redundancy at such genes [12,14]. The Ino80 complex (Ino80C) can have opposite effects on nucleosome positioning at different genes, and acts differently from RSC and SWI/SNF at certain genes to move the +1 nucleosome upstream to narrow, rather than widen, the NDR [14,21]. This Ino80C activity is much more pronounced on nuclear depletion of Isw2, catalytic subunit of the CR ISW2; and simultaneous depletion of both Ino80 and Isw2 leads to wider NDRs at a large number of genes and suppresses the widespread narrowing of NDRs conferred by depleting the RSC catalytic subunit Sth1 [14]. There is evidence that Ino80C also "edits" promoter nucleosomes to replace histone variant H2A.Z with conventional H2A [22]; however, there are conflicting findings regarding this activity [23]. Other evidence indicates that Ino80C can function in remodeling conventional nucleosomes containing H2A versus H2A.Z [24][25][26].
Transcriptional activation of yeast genes is mediated by binding of activator proteins to UAS elements that recruit an array of co-factors to facilitate binding of general transcription factors (GTFs) to the promoter, including TATA-binding protein (TBP), and in recruitment �2-fold induction of Pol II occupancies averaged across the CDS on SM treatment. Parallel ChIP-Seq analysis of histone H3 revealed a marked eviction of nucleosomes in the promoter intervals spanning the -1 and +1 nucleosomes and intervening NDRs at 70 of these SMinduced genes, and lesser increases in H3 occupancies at the remaining 134 induced genes [30]. As noted above, examining mutants lacking the catalytic subunits of SWI/SNF (snf2Δ) or Ino80C (ino80Δ), or conditionally depleted of the essential catalytic subunit of RSC by transcriptional shut-off (P TET -STH1), revealed overlapping roles for these three CRs in evicting promoter nucleosomes and inducing transcription of SM-induced genes [12,26,30]. Here, we set out to examine their roles in binding of Gcn4 to UAS elements, which increases sharply on translational induction of Gcn4 protein by SM [46].
Interestingly, our previous ChIP-seq analysis of Gcn4 in SM-treated WT cells revealed that only~40% of the 546 sites of induced binding detected throughout the genome are located upstream of genes (5' sites), with the majority occurring instead within the CDS (ORF sites). Roughly 70% of the genes with conventional 5' sites show evidence of transcriptional activation in amino acid starved cells and are likely direct targets of Gcn4, whereas only~20% of the genes with only ORF binding sites are transcriptionally activated. Mutation of the internal Gcn4 binding sites at a number of the latter genes was shown to impair their transcriptional activation by SM, demonstrating that Gcn4 can activate 5'-positioned promoters by binding within the downstream ORFs. Gcn4 binding to the majority of 5' sites occurs within NDRs; while the binding at ORF sites generally occurs within linkers between genic nucleosomes. Moreover, only~30% of all predicted Gcn4 binding motifs in the genome are occupied by Gcn4 in SM-induced cells, and the two most important features of bound motifs are (i) a strong match to the consensus Gcn4 motif and (ii) lower than average nucleosome occupancy [46]. These findings suggest that Gcn4 binding is inefficient when its binding motif maps within a nucleosome, in agreement with previous findings that Gcn4 binding to the HIS4 UAS is enhanced by Rap1 and its ability to exclude nucleosomes [5,6]. Nucleosome occupancy is also a key determinant of binding by activator Leu3 in the yeast genome [47]. Recent results indicating cooperation between different GRFs and RSC in downstream positioning of the +1 nucleosome for NDR formation [29], led us to wonder whether Gcn4 binding to 5' sites at many UASs might be stimulated by RSC.
The results of the current study indicate that Gcn4 binding at many 5' sites, as well as ORF sites, is stimulated by RSC, by SWI/SNF in cells depleted of RSC, and by Ino80C. Surprisingly, when RSC is present, SWI/SNF acts primarily to limit rather than enhance Gcn4 occupancies at 5' sites, suggesting a dual role for SWI/SNF in activator binding. We found that Gcn4 recruits all three CRs to 5' sites, consistent with their direct roles, and indicating feedback loops that regulate Gcn4 binding at UAS elements. Other evidence indicates that the CRs enhance Gcn4 binding by reducing nucleosome occupancies of the Gcn4 binding motifs, supporting the notion that Gcn4 binding is impeded by nucleosomes. Examining the effects of deleting/depleting the CRs on TBP occupancies showed that RSC and SWI/SNF collaborate to enhance TBP recruitment at Gcn4 target genes, as does Ino80C, in a manner associated with nucleosome eviction at the TBP binding sites. Cooperation among the CRs is also evident at the highly transcribed ribosomal protein genes (RPGs), while RSC and Ino80C act more broadly than SWI/SNF to stimulate this step in PIC assembly throughout the rest of the genome.

RSC, SWI/SNF and Ino80C modulate Gcn4 binding within NDRs and coding regions
Our previous ChIP-seq analysis of Gcn4 in SM-induced WT cells identified 117 Gcn4 binding sites located 5' of an annotated TSS in the canonical location of yeast UAS elements, which appear to mediate transcriptional activation of the adjacent genes during amino acid starvation (5' Gcn4 peaks). An additional 62 Gcn4 peaks were identified within coding sequences that appear to activate transcription of the full-length transcript of the corresponding gene or an adjacent gene lacking a Gcn4 peak (ORF peaks) [46]. As reported previously [46], averaging the Gcn4 occupancies measured by ChIP-seq over the 5' or ORF peaks and aligning them to the Gcn4 binding motifs reveals that Gcn4 occupancies are highly induced by SM in WT cells, and that the summits of the occupancy peaks coincide precisely with the Gcn4 binding motifs (Fig 1A(i) and 1B(i), yellow vs. blue).
Our previous ChIP-seq analysis of histone H3 using chromatin fragmented by micrococcal nuclease digestion, which preferentially cuts within linkers, allowed us to estimate the positions of nucleosome dyads from the midpoints of immunoprecipitated DNA fragments. Averaging dyad densities over all 5' or ORF peaks reveals that Gcn4 binding motifs of 5' sites are generally located in the centers of NDRs, whereas the motifs for ORF peaks reside in the linkers between the genic nucleosomes observed in untreated WT cells (Fig 1E and 1F, red tracings). The dyad densities of the -1 and +1 nucleosomes and the intervening NDRs at the 5' Gcn4 peaks are diminished, and the NDRs become wider, in response to SM treatment ( Fig  1E, blue vs. red), reflecting eviction and repositioning of the -1 and +1 nucleosomes [12]. The dyad densities of the genic nucleosomes flanking ORF peaks are also diminished; however, the spacing between them appears unaltered by SM treatment (Fig 1F, blue vs. red). The fact that Gcn4 binds preferentially to motifs located outside of nucleosomes, and our previous findings that eviction or repositioning of promoter nucleosomes induced by SM treatment involves SWI/SNF, RSC and Ino80C [12,26], led us to examine whether Gcn4 binding is enhanced by one or more of these CRs.
To this end, we conducted Gcn4 ChIP-seq analysis on snf2Δ and ino80Δ mutants, lacking the catalytic subunits of SWI/SNF or Ino80C, respectively, and on a P TET -STH1 strain in which the RSC catalytic subunit was transcriptionally repressed by addition of doxycycline (Dox) for 8h, in cells treated with SM to induce Gcn4. We similarly examined a snf2Δ P TET -STH1 double mutant treated with Dox to examine the effects of depleting RSC in cells lacking SWI/SNF. We observed a slight reduction in the averaged Gcn4 occupancies for the group of 5' sites on depleting Sth1 in P TET -STH1 versus WT cells under inducing conditions (P TET -STH1_I vs. WT_I), but a more substantial reduction of~50% in the induced snf2Δ P TET -STH1 double mutant versus WT I cells (Fig 1A(i), red & cyan vs. yellow). Surprisingly, in the snf2Δ single mutant, the averaged Gcn4 occupancy was markedly increased compared to WT cells (Fig 1A(i), grey vs. yellow). Calculating the mean Gcn4 occupancy per nucleotide across each 5' site (with boundaries defined as described in Materials and Methods), revealed a significant reduction in median Gcn4 occupancy in snf2Δ P TET -STH1_I cells, but an increased median occupancy in snf2Δ_I cells, compared to WT_I cells (Fig 1C(i), cols. 5 & 3 vs. 2). These opposite effects on median Gcn4 occupancies conferred by snf2Δ (increased binding) versus the snf2Δ P TET -STH1 double mutant (decreased Gcn4 binding) are evident in comparing the results obtained from the individual biological replicates for these two mutants and WT_I cells (S1A Fig, grey or cyan vs. yellow), which were combined to generate the composite results in Fig 1C. High reproducibility of results from biological replicates is also illustrated for six representative genes with 5' Gcn4 sites in S2A-S2C Fig. The ORF Gcn4 peaks also showed reductions in P TET -STH1_I and snf2Δ P TET -STH1_I cells, but no significant increase in snf2Δ_I cells (Fig 1B(i) and 1D(i)). Similar analyses showed that the ino80Δ mutation conferred moderate reductions in both average and median Gcn4 occupancies at the 5' sites (Fig 1A(ii) and 1C(ii), gold vs. yellow), which is less severe than that just described for the snf2Δ P TET -STH1 double mutant. The ino80Δ mutation also reduced Gcn4 occupancies at ORF peaks (Fig 1B(ii) and 1D(ii)), to an extent comparable to that shown above for the P TET -STH1_I single mutation (Fig 1B(i) and 1D(i)).
To explore further the contributions of the CRs at individual Gcn4 motifs, we sorted the 5' Gcn4 peaks according to their reductions in Gcn4 occupancies in the snf2Δ P TET -STH1_I double mutant compared to WT_I cells. A heat map showing the differences in Gcn4 occupancies between the double mutant and WT for this ordering of 5' sites is shown in Fig 2A(i), which , and SM-induced CR mutants for (i) snf2Δ_I (gray), P TET -STH1_I (red), snf2Δ P TET -STH1_I (cyan); and (ii) WT_I (yellow) and ino80Δ_I (gold). In these and all similar plots below, yeast strain/condition labels adjacent to the tracings are listed in decreasing order of summit heights. (C-D) Notched box plots of log 2 Gcn4 occupancies per nucleotide averaged over the peak coordinates assigned by MACS2 analysis of (C) 5' and (D) ORF Gcn4 peaks for (i) WT_U, WT_I, and CR mutants snf2Δ_I, P TET -STH1_I and snf2Δ P TET -STH1_I; and (ii) WT_U, WT_I, and ino80Δ_I samples, color-coded as in (A-B). For these and all subsequent box plots, each box depicts the interquartile range containing 50% of the data, intersected by the median; the notch indicates a 95% confidence interval (CI) around the median, and p values for the significance of differences in median values calculated by the Mann-Whitney-Wilcoxon test are indicated as follows: ���� , � 0.0001; ��� , � 0.001; �� , � 0.01; � , � 0.05; ns, > 0.05. Unlogged median Gcn4 occupancies are indicated under the respective strain labels. Data analyzed in (A-D) were calculated from Gcn4 ChIP-seq analysis of sonicated chromatin from 2-4 biological replicates and normalized to the average occupancy per nucleotide on each chromosome for each data set. (E-F) Average dyad densities calculated from H3 MNase-ChIP-seq data aligned to the Gcn4 consensus motifs for (B) 5' and (D) ORF Gcn4 peaks in WT cells either untreated (_U) or SM-treated (_I). Midpoints (dyads) of nucleosome-size sequences between 125 and 175 bp were mapped with respect to Gcn4 consensus motifs. Average profiles were smoothed using a moving average filter with a span of 31 bp. The data were normalized to the average occupancy per nucleotide on each chromosome for each data set.
https://doi.org/10.1371/journal.pgen.1010277.g001 was calculated from the corresponding occupancies for these two strains displayed in the heat maps of Fig 2B(i)-(ii). Thus, the Gcn4 5' sites in the top quartile of the maps (designated Set_1 in Fig 2B) exhibit the greatest occupancy reductions in the double mutant versus WT cells, as indicated by the darkest blue hues in the difference map of Fig 2A(i). Similar difference maps were generated for the P TET -STH1 or snf2Δ single mutants, shown in Fig 2A(ii)-(iii), based on the Gcn4 occupancies measured in these strains displayed in Fig 2B(iii)-(iv). The Set_1 peaks display relatively smaller occupancy decreases in the P TET -STH1_I single mutant vs. WT_I than observed in the double mutant (lighter blue hues in the difference map of Fig 2A(ii) vs. Fig 2A(i) for the top quartile). In snf2Δ_I cells, the Set_1 peaks variously show small decreases, no change, or small increases in Gcn4 occupancy compared to WT_I cells (Fig 2A(iii), top quartile). The Gcn4 peaks in the middle two quartiles of the maps (dubbed Set_2 in Fig 2B) show strong to moderate reductions in the double mutant, lesser reductions in the P TET -STH1_I single mutant, but generally increased occupancies in snf2Δ_I cells versus WT_I cells (Fig 2A(i)-(iii), middle quartiles). The genes in the bottom quartile of the maps (Set_3) tend to show similar, small occupancy increases or decreases in both the double and single P TET -STH1_I mutants (Fig 2A(i)-(ii)), while showing marked increases in Gcn4 occupancies in snf2Δ_I cells (Fig 2A(iii)).
The trends depicted in the heat maps of Fig 2A and 2B were confirmed by quantifying the Gcn4_I occupancies in each peak and calculating the median values for the three sets of peaks. As expected, Set_1 displays a greater reduction in Gcn4 median occupancy in the double mutant versus the P TET -STH1_I single mutant, but no significant change in the snf2Δ single mutant (Fig 2C(i), cols. 3-5 vs. 2). Set_2 also shows a greater reduction in median occupancy in the double mutant versus the P TET -STH1_I single mutant, even though Gcn4 occupancy is significantly elevated in the snf2Δ single mutant containing WT STH1 (Fig 2C (ii), cols. 3-5 vs. 2). Set_3 shows no significant decrease in median Gcn4 occupancy in either of the P TET -STH1_I mutants, but increased occupancies in the snf2Δ single mutant (Fig 2C(iii), cols. 3-5 vs. 2). The differential effects of the snf2Δ and P TET -STH1 mutations on Gcn4 binding in the three sets of 5' sites summarized in Fig 2C are also evident for the biological replicates for each of the mutants (S1B Fig), including increased Gcn4 binding in both snf2Δ replicates for Set_3 but not for Set_1 sites (grey vs. yellow), and decreased Gcn4 binding in all three replicates of the snf2ΔP TET -STH1 double mutant for Set_1 but not for Set_3 sites (cyan vs. yellow).
The results in Fig Set_3). In contrast, the snf2Δ single mutation confers variable changes for Set_1 sites, but moderate or strong increases in Gcn4 occupancies for Set_2 and Set_3 sites, respectively (S3C Fig). Finally, we constructed a sectored scatterplot of the changes in Gcn4 binding conferred by snf2Δ versus the snf2Δ P TET -STH1 double mutation, color-coded according to Set_1, _2, or _3 sites. Although the changes are significantly correlated for all 5' sites (R = 0.64, P < 1.  Together, these findings indicate that RSC enhances Gcn4 binding for Set_1 and Set_2 sites, with a greater contribution for Set_1, but has weak and variable effects on Set_3 sites. In WT cells, SWI/SNF generally inhibits, rather than promotes, Gcn4 binding at the Set_2 and _3 sites, with the greatest inhibition for the Set_3 sites that are largely insensitive to RSC, while SWI/SNF has variable effects on the Set_1 sites most dependent on RSC. In cells depleted of RSC, however, SWI/SNF promotes rather than inhibits binding at 5' sites in Set_1 and _2, thus compensating for reduced RSC function at the peaks most dependent on RSC; and SWI/SNF no longer inhibits Gcn4 binding to most peaks in Set_3. One way to explain these last findings for Set_3 would be to propose that the inhibitory effect of SWI/SNF on Gcn4 binding is offset by the stimulatory role that SWI/SNF exerts upon depletion of Sth1, with little net change in Gcn4 occupancy found in the double mutant. Differential effects of the P TET -STH1_I and snf2Δ mutations on Gcn4 occupancies for the three Sets of 5' sites described above are illustrated for two archetypal members of each Set in S4A-S4C ((i)-(ii)) Fig.
It is worth noting that we previously obtained results fully consistent with those reported here by ChIP analysis of Gcn4 for particular genes, where occupancies were calculated as the ratios of input DNA recovered in the immunoprecipitates (determined by PCR analysis) and corrected for the corresponding ratios observed for a non-transcribed sequence on chromosome V [30]. Thus, as summarized in S3A Table, we reported that Gcn4 occupancy was elevated in snf2Δ versus WT cells by 2.1-and 1.6-fold in the UASs of Set_3 genes ARG1 and ARG4, respectively, elevated by 1.5-fold at the UAS of Set_ 2 gene CPA2, and did not increase at the Set_1 gene HIS4, in agreement with the ChIP-Seq analysis of ARG1 and HIS4 shown in S4C(i) and S4A(i) Fig. Extending the PCR-ChIP analysis further to include the P TET -STH1 and snf2Δ P TET -STH1 mutants, we observed a marked decrease in Gcn4 binding at the Set_1 gene HIS4 only in the double mutant, and no decrease in either mutant for the Set_3 gene ARG1 (S3B Table), again consistent with our Gcn4 ChIP-seq results in S4A(i) and S4C(i) Fig. The effects of the CR mutations on Gcn4 occupancies at binding sites in coding regions were less complex, as the majority of Gcn4 ORF peaks show relatively strong reductions in the snf2Δ P TET -STH1 double mutant, moderate reductions in the P TET -STH1_I single mutant, and little change in the snf2Δ_I single mutant (see difference maps in Fig 2D(i)-(iii), based on the heat maps in Fig 2E(i)-(iv)). Thus, as noted above for 5' Gcn4 peaks, RSC generally stimulates Gcn4 binding and SWI/SNF partially compensates for RSC function on depletion of RSC, but there is no evidence for an inhibitory effect of SWI/SNF, on Gcn4 binding at the ORF peaks. Interestingly, the magnitude of the reduction in Gcn4 occupancy at the ORF peaks in the snf2Δ P TET -STH1 double mutant versus WT cells is strongly correlated with Gcn4 occupancy at these peaks in WT_I cells (Fig 2F(ii)), suggesting an increasing requirement for RSC or SWI/SNF as the occupancy at ORF sites increases. A significant but considerably weaker correlation exists for 5' Gcn4 peaks (Fig 2F(i)), possibly reflecting the more complex interplay between RSC and SWI/SNF at the 5' sites.
Analyzing the effects of ino80Δ for the same ordering of 5' Gcn4 peaks described above reveals that the 5' sites in Set_1 show reductions of lesser magnitude compared to those given by snf2Δ P TET -STH1; whereas comparable, smaller reductions are conferred by ino80Δ and To identify all 5' sites with a heightened dependence on Ino80C, we ordered them differently based on their decreased Gcn4 occupancies in ino80Δ _I versus WT_I cells, generating heat maps with peaks exhibiting the largest occupancy reductions conferred by ino80Δ at the top of the maps (S6A (i)-(iii) Fig). As expected for this ordering, ino80Δ substantially decreases the median occupancy for the top quartile of peaks (designated Set_I), but not for the lower quartiles in Sets_II or _III (S6C(i)-(iii) Fig). The corresponding difference maps for the P TET -STH1 and snf2Δ P TET -STH1 mutations for this same ordering of 5' sites reveals more uniform reductions in Gcn4 occupancies across the sites compared to those given by ino80Δ (S6B(i)-(ii) vs. S6A(i) Fig); and the snf2Δ single mutation generally confers increased Gcn4 occupancies for sites in Sets_I, and _II that show decreased occupancies in ino80Δ _I versus WT_I cells (S6B(iii) vs. S6A(i) Fig). Thus, certain 5' sites have a particularly strong dependence on Ino80C, but little requirement for RSC or SWI/SNF in Gcn4 binding in WT cells. The importance of Ino80C for WT Gcn4 binding is illustrated for the YHR162W and HIS4 genes in S12E (i)-(ii) Fig (rows 2-3).
We considered the possibility that the changes in Gcn4 occupancies in the mutants described above involve changes in Gcn4 expression. Quantifying the occupancies of Pol II across the GCN4 coding sequences from our previous ChIP-seq analysis of Rpb3 indicated similar reductions in GCN4 transcription of~32-37% in the induced snf2Δ P TET -STH1 and P TET -STH1 mutants, but no change in snf2Δ_I or ino80Δ_I cells [12,26]  To examine this last issue further, we reasoned that if reduced Gcn4 occupancies arise primarily from reduced Gcn4 abundance, then the occupancy reductions should occur preferentially at binding sites of lower affinity or accessibility in chromatin. We showed previously that Gcn4 occupancies in WT cells are dictated primarily by their match to the consensus motif, with a secondary contribution from the distance of the motif from the dyad of the nearest nucleosome [46]. Consistent with this, Gcn4 occupancies in WT_I cells are positively correlated with the match of the binding motif to the consensus sequence (quantified as "Find Individual Motif Occurrences" (FIMO) scores) for both 5' and ORF peaks of Gcn4 occupancies (S7C and S7D Fig). Accordingly, we used Gcn4 occupancies in WT_I cells as a proxy for affinity/accessibility of the binding motifs in chromatin. As shown above in Fig [26], it appears that intrinsic properties of high-occupancy Gcn4 binding sites render them relatively more dependent on CR function for robust Gcn4 binding.
Evidence that SWI/SNF, RSC, and Ino80C are recruited by Gcn4 to NDRs and act directly to regulate Gcn4 binding To provide evidence that SWI/SNF, RSC and Ino80C function directly at Gcn4 motifs to regulate Gcn4 binding, we examined whether these CRs are recruited to Gcn4 binding sites. To this end, we conducted ChIP-seq analysis of the Myc epitope-tagged catalytic subunits, Snf2, Sth1, and Ino80, and corrected the occupancies for those obtained from ChIP-seq of an isogenic untagged strain. Heatmaps of the corrected CR occupancies sorted by the WT_I Gcn4 occupancies revealed that Snf2-myc occupancies are generally centered on the Gcn4 binding motifs, and that they tend to increase with increasing Gcn4 binding (Fig 3A(i)), yielding a direct correlation between Snf2-myc_I and Gcn4_I occupancies (Fig 3B(i)). A precise coincidence in the occupancies of Gcn4 and Snf2-myc is evident for several 5' Gcn4 peaks associated with particularly high induced Snf2-myc occupancies (S8A- S8E Fig). There is also evidence, however, that Snf2-myc occupancy is greatest upstream of the Gcn4 motif at a subset of 5' sites ( Fig 3A(i)), which might indicate interaction of SWI/SNF with another transcription factor bound at these promoters instead of Gcn4. The Sth1-myc occupancies are also generally centered on the Gcn4 motifs, but they are lower overall (Fig 3A(ii)), and the correlation between Gcn4 and Sth1-myc occupancies is less pronounced (Fig 3B(ii)); however, Snf2-myc and Sth1-myc occupancies at the 5' Gcn4 peaks are positively correlated (S9A(i) Fig). Ino80-myc occupancies also peak at, or somewhat upstream, of the Gcn4 motifs at most 5' sites, and show a positive correlation with WT_I Gcn4 occupancies (Fig 3B(iii)); although they are higher than expected at the 5' sites of lowest Gcn4 occupancies found at the bottom of the heat map in Fig  3A(iii). The Ino80-myc occupancies are additionally correlated with those for both Snf2-myc and Sth1-myc (S9A(ii)-(iii) Fig). Importantly, the median occupancies of Snf2-myc, Sth1-myc, and Ino80-myc for all 5' sites were significantly greater on induction of Gcn4 by SM compared to uninduced cells (Fig 3C(i)-(iii)). Together, these findings suggest that Gcn4 directly recruits SWI/SNF, Ino80C, and to a lesser degree RSC, to its 5' binding sites.
Supporting this last conclusion, we showed previously that Gcn4 directly interacts with SWI/SNF and RSC in cells extracts in a manner dependent on key hydrophobic residues in the Gcn4 activation domain crucial for activation of Gcn4 target genes [48,49]. Moreover, we showed that recruitment of both SWI/SNF and RSC to the ARG1 UAS in SM-treated cells, as judged by ChIP analysis of Snf2-myc and Rsc8-myc, was diminished by deletion of GCN4 [49]. Confirming our conclusion reached above that Gcn4 also recruits Ino80C, we found here that Ino80-myc occupancies at the 5' Together with the above findings that Snf2-myc, Sth1-myc, and Ino80-myc occupancies peak in the vicinity of the Gcn4 binding motifs at the majority of 5' sites (Figs 3A and S8), that occupancies of all three CRs are induced in parallel with Gcn4 induction by SM ( Fig 3C) and are positively correlated with Gcn4 occupancies for the group of 5' genes ( Fig 3B), we consider it very likely that SWI/SNF, RSC, and Ino80C are all recruited directly by Gcn4 to the majority of 5' genes.
We next sought evidence that SWI/SNF acts directly to dissociate Gcn4 from the 5' binding sites where it negatively regulates Gcn4 binding. Sorting the 5' motifs according to their increases in Gcn4 occupancy in snf2Δ_I versus WT_I cells (Fig 3D (i)-(iii)), we found that Snf2-myc occupancies are generally higher at the 5' motifs exhibiting the largest increases in Sth1-myc and (iii) Ino80-myc (corrected by subtracting the background myc-occupancies from untagged WT_I cells as described in Methods) at the 5' Gcn4 peaks, sorted by decreasing Gcn4 occupancies in WT_I. Occupancies of the myc-tagged CR subunits were calculated from ChIP-seq data of mildly sonicated chromatin from 2 biological replicates for Snf2-myc_I and Sth1-myc_I, and 3 biological replicates for Ino80-myc_I. (B) Scatterplots of log 2 Gcn4 occupancies in WT_I vs. corrected occupancies for (i) Snf2-myc_I, (ii) Sth1-myc_I, or (iii) Ino80-myc_I within ±100 bp windows surrounding the Gcn4 motifs of 5' sites. Pearson correlation coefficients (R) and associated p values are indicated. (C) Notched box plots of (i) Snf2-myc, (ii) Sth1-myc, or (iii) Ino80-myc corrected occupancies per nucleotide within ±100 bp windows surrounding the Gcn4 motifs of 5' sites in uninduced (_U) and SM induced (_I) conditions. (D) Heat maps depicting differences in Gcn4 occupancies between snf2Δ_I and WT_I cells (i); Gcn4 occupancies surrounding the motifs of 5' sites in (ii) WT_I or (iii) snf2Δ_I cells; and (iv) Snf2-myc_I corrected occupancies surrounding the Gcn4 motifs of 5' sites. Gcn4 5' sites were sorted by increasing order of the ratio of Gcn4 occupancies in snf2Δ_I vs. WT_I cells. (E) Scatterplot of corrected Snf2-myc occupancies within ±100 bp windows surrounding the Gcn4 motifs of 5' sites vs. the log 2 ratios of Gcn4 occupancies in snf2Δ_I vs. WT_I cells at 5' Gcn4 peaks. Pearson correlation coefficients (R) and associated p values are indicated. (F) Heat maps depicting differences in Gcn4 occupancies between ino80Δ_I and WT_I cells (i); Gcn4 occupancies surrounding the motifs of 5' sites in (ii) WT_I or (iii) ino80Δ_I cells; and (iv) corrected Ino80-myc_I occupancies surrounding the Gcn4 motifs of 5' sites. Gcn4 5' sites were sorted by increasing order of the ratio of Gcn4 occupancies in ino80Δ_I vs. WT_I cells, and the first (Set_I, n = 30), middle two (Set_II, n = 57) and fourth (Set_III, n = 30) quartiles of fold-changes are depicted. (G) Same analysis shown in (E) except for corrected Ino80-myc occupancies vs. the log 2 ratios of Gcn4 occupancies in ino80Δ_I vs. WT_I cells. https://doi.org/10.1371/journal.pgen.1010277.g003 Gcn4 occupancies conferred by the snf2Δ single mutation (and thus most negatively regulated by SWI/SNF), which are positioned at the bottom of the heat map in Fig 3D(iv)); and that there is a significant positive correlation between these two parameters for all 5' sites ( Fig 3E). Interestingly, a small enrichment of Snf2-myc is also evident for the 5' sites located at the top of the map in Fig 3D(iv), containing the subset of sites where SWI/SNF stimulates rather than impedes Gcn4 binding in WT cells. Overall, these findings support the idea that in WT cells containing RSC, SWI/SNF directly regulates Gcn4 binding at 5' sites, reducing binding at many sites but enhancing Gcn4 binding at a small subset of sites.
Carrying out the same analysis for Ino80C reveals that the 5' sites showing the strongest reductions in Gcn4 binding conferred by ino80Δ (at the top of the heat maps in Fig 3F(i)-(iii)) tend to have higher Ino80-myc occupancies (Fig 3F(iv)), in this case producing a negative correlation between the two parameters ( Fig 3G). The correlation is likely weakened by the moderate enrichment for Ino80-myc at the bottom of the heat map for the 5' sites that show moderately increased Gcn4 binding in ino80Δ cells (Fig 3F(iv)). The elevated Ino80-myc occupancies at these latter sites could be explained by an inhibitory effect of Ino80C on Gcn4 binding, akin to our findings for SWI/SNF. Together, these results suggest that Ino80C also directly regulates Gcn4 binding at 5' sites, functioning like RSC to enhance binding at most sites but possibly mimicking SWI/SNF in reducing Gcn4 occupancy at a small subset of binding sites. Our findings that RSC and Ino80C are recruited to 5' sites on SM induction, and generally enhance Gcn4 occupancies, suggest that they function in a positive feedback loop. The fact that SWI/SNF is also highly recruited to 5' sites where it generally reduces Gcn4 occupancies (at least in WT cells), suggests that SWI/SNF acts directly to dampen the positive feedback loop established by the other CRs.

Reduced Gcn4 occupancies are frequently associated with defective nucleosome eviction at Gcn4 binding motifs in mutants depleted of Sth1
The most likely mechanism for regulation of Gcn4 binding by CRs is the displacement of nucleosomes to either expose the motif and enhance binding or to occlude the motif and impede binding. ChIP-seq analysis of histone H3 in sonicated chromatin reveals reduced H3 occupancies surrounding the Gcn4 motifs within the NDRs of genes harboring 5' binding sites on SM induction of WT cells (Fig 4A, yellow vs. blue). These reductions in H3 seen in WT are diminished to similar extents in the snf2Δ and P TET -STH1 single mutants, but to a greater extent in the snf2Δ P TET -STH1 double mutant (Fig 4A, grey, red, & cyan vs. yellow), indicating cooperation between RSC and SWI/SNF in evicting nucleosomes at 5' Gcn4 sites. Similar conclusions were reached independently from H3 MNase ChIP-seq analysis of the same mutants (S10A Fig).
To examine the relationship between defects in H3 eviction and Gcn4 binding at particular 5' sites, we first plotted the changes in H3 occupancies in snf2Δ P TET -STH1_I versus WT_I cells in the regions surrounding the 5' Gcn4 motifs, sorted as above by their reductions in Gcn4 binding in this double mutant (Fig 4B(i)-(ii)). The resulting difference map reveals an obvious tendency for the 5' sites with the greatest reductions in Gcn4 binding, ie. in Sets_1 and _2 (dark blue hues in Fig 4B(i)), to exhibit sizable increases in H3 occupancies at the Gcn4 motifs in the double mutant (orange hues in Fig 4B(ii)), consistent with the idea that defective nucleosome eviction contributes to reduced Gcn4 binding in cells depleted of both RSC and SWI/SNF. Surprisingly, however, 5' sites at the bottom of the H3 difference map in Fig 4B(ii), belonging to Set_3, also show marked increases in H3 occupancies, but exhibit the smallest changes in Gcn4 binding observed among all 5' sites in the double mutant (Fig 4B(i))-a complexity we address further below.
The aforementioned trends in the heat maps were examined more closely by quantifying the changes in H3 occupancies for a 200bp window surrounding the Gcn4 motifs and comparing them to the corresponding changes in Gcn4 occupancies in the different mutants. In WT cells, SM-induction confers the expected reduced median H3 occupancies in parallel with

PLOS GENETICS
increased Gcn4 occupancies for all three Sets of 5' sites (cf. Fig 4C and 4D, (i)-(iii), cols. 1-2). In both the P TET -STH1 and snf2Δ P TET -STH1 mutants, the H3 occupancies are increased significantly for Set_1 and Set_2, and also for Set_3 in the snf2Δ P TET -STH1 double mutant, with larger increases in the double versus single mutant (Fig 4C(i)-(iii), cols. 4-5 vs. 2). Similar results were obtained by H3 MNase ChIP-seq analysis, except that no significant increase was observed for Set_3 sites in either P TET -STH1 or snf2Δ P TET -STH1 cells (S10B Fig). Importantly, the increased H3 occupancies at the Set_1 sites are associated with decreased median Gcn4 occupancies in the P TET -STH1 and snf2Δ P TET -STH1 cells, and the same applies to the Set_2 sites in the double mutant (Fig 4D(i)-(ii), cols. 4-5 vs. 2). (The opposing changes in median H3 and Gcn4 occupancies shown in Fig 4C and 4D for Sets _1 and_2 are also evident for the corresponding biological replicates examined for WT and snf2Δ P TET -STH1 cells, as shown in S1B and S1C Fig, cyan vs. yellow.) Similar insights emerged from a sectored scatterplot of changes in H3 versus changes in Gcn4 occupancy in the double mutant compared to WT cells (S10C Fig), where nearly all of the 5' sites in Sets_1 and _2 (red and green points) fall into the upper left-hand quadrant exhibiting increased H3 but decreased Gcn4 occupancies in the double mutant. The association between reduced Gcn4 binding and increased H3 occupancies conferred by P TET -STH1 in the double mutant is also evident at the archetypal Set_1 and Set_2 genes depicted in S4A and S4B Fig. Together, these results support the conclusion that RSC and SWI/SNF (upon Sth1 depletion) cooperate in evicting nucleosomes to facilitate Gcn4 binding at most 5' sites belonging to Sets_1 and _2.
For the Set_3 sites, by contrast, increased H3 occupancies are generally not associated with decreased Gcn4 binding in the snf2Δ P TET -STH1 double mutant (cf. Fig 4C(iii) and 4D(iii), col. 5 vs. 2). As noted above, this conclusion applies even to the subset of sites at the bottom of the heat map in Fig 4B(ii) that show strong increases in H3 occupancies but exhibit the smallest reductions in Gcn4 binding among all 5' sites in response to the snf2Δ P TET -STH1 double mutation. These latter sites exhibit high-level SWI/SNF recruitment (Fig 4B(iv), consistent with a direct role of SWI/SNF in nucleosome eviction, which has little impact on Gcn4 binding. The uncoupling of Gcn4 and H3 occupancies among Set_3 sites is also evident in the scatterplot of S10C Fig, as these sites (blue points) show increases in H3 occupancies comparable to those found for Set_1 and Set_2 sites (red and green points) but much smaller decreases in Gcn4 binding. The uncoupling is particularly noteworthy for the Set_3 archetype ARG1 in the snf2Δ P TET -STH1 double mutant (S4C(i) Fig). We concluded above that SWI/SNF exerts offsetting stimulatory and inhibitory effects on Gcn4 binding at the Set_3 sites in the double mutant. Hence, to account for the uncoupling of changes in H3 and Gcn4 occupancies at Set_3 sites, we suggest that elimination of SWI/SNF in the double mutant confers reduced nucleosome eviction at these sites, but the resulting increased H3 occupancies do not reduce Gcn4 binding owing to concurrent loss of a distinct function of SWI/SNF that inhibits Gcn4 binding.
Uncoupling of changes in H3 and Gcn4 occupancy is also widespread in the snf2Δ single mutant. First, examining the sectored scatterplot of H3 versus Gcn4 occupancy changes in S10D Fig shows that snf2Δ increases the H3 occupancies for a large proportion of all 5' sites, with the majority of points mapping above the x-axis, but most points fall into the upper righthand quadrant indicating increased Gcn4 binding conferred by snf2Δ. Consistent with this, the snf2Δ single mutation leads to increased median H3 occupancies at the Set_1 sites (Fig 4C  (i), col. 3 vs. 2), but has no effect on median Gcn4 binding at these sites (Fig 4D(i), col. 3 vs. 2). (See also S1B and S1C Fig for results from biological replicates for the Set_1 sites in snf2Δ_I vs. WT_I cells, Set_1, grey vs. yellow). The finding that increased H3 occupancies is generally not accompanied by decreased Gcn4 binding in snf2Δ cells can be explained, as suggested above, if the reduced nucleosome eviction conferred by snf2Δ is offset by loss of the proposed SWI/SNF function that impedes Gcn4 binding.
Exploring further the changes in Gcn4 binding and H3 occupancies conferred by the snf2Δ single mutation using the heat map analysis shown in S11A Fig revealed that a small subset of Set_1 sites, located at the top of the map in panel (i), show decreased, rather than increased, Gcn4 binding in snf2Δ cells. Interestingly, these sites also exhibit the largest increases in H3 occupancies among all 5' sites conferred by snf2Δ, as shown in panel (ii). To evaluate more completely the minority fraction of 5' sites where SWI/SNF stimulates rather than impedes Gcn4 binding in WT cells, we sorted all of the 5' sites by their reductions in Gcn4 occupancies in snf2Δ versus WT cells, placing those sites at the top of the heat map (in Quantile 1) of S11B (i) Fig. Again, we observed that these sites exhibit the largest increases in H3 occupancies at the Gcn4 motifs among all 5' sites (S11B(ii) vs. (i) Fig). Hence, at this small subset of 5' sites, SWI/SNF plays the conventional role observed for RSC of evicting nucleosomes to promote Gcn4 binding, replacing or outweighing the negative effect of SWI/SNF on Gcn4 binding occurring at most 5' sites in WT cells containing RSC.
As described fully in S12 Fig, analysis of the ino80Δ mutant provides evidence that Ino80C enhances Gcn4 binding at a subset of 5' sites in Set_I (as defined in S6A(ii) Fig) by evicting nucleosomes surrounding the relevant binding motifs, which is highly similar to our findings above for the P TET -STH1 single mutant (Fig 4C and 4D, Set_1 genes). The association between reduced Gcn4 binding and increased H3 occupancies at the Gcn4 binding sites conferred by ino80Δ is illustrated for YHR162W and HIS4 in S12E Fig.
The results described above indicate that reduced Gcn4 binding conferred by the mutations that impair RSC or Ino80C can be attributed at least in part to increased H3 occupancies found at the binding motifs in these mutants, whereas the effects of snf2Δ are more complex owing to the opposing roles that SWI/SNF plays in (i) stimulating Gcn4 binding via nucleosome eviction, and (ii) impeding Gcn4 binding by an unknown mechanism. An intriguing possibility for the latter inhibitory function would involve sliding of the remaining, nonevicted nucleosomes over the Gcn4 binding motifs, in a manner demonstrated previously for yeast SWI/SNF and the Gal4 DNA binding domain in vitro [41]. However, because we do not observe reduced H3 occupancies in snf2Δ cells over the 5' sites that exhibit the largest increases in Gcn4 binding in this mutant (S11B(ii) Fig, quartile 4), it would be necessary to stipulate that the proposed sliding is transitory and does not produce increased steady-state H3 occupancies over the 5' sites in WT versus snf2Δ cells. In this event, it might be expected that the nucleosome positions would be "fuzzier" in WT compared to the snf2Δ cells. Examining our H3 MNase ChIP-Seq data for snf2Δ and WT cells does not reveal a broader peak of nucleosome dyad positions encompassing the Gcn4 motifs for the Gcn4 sites with the largest increases in Gcn4 binding in snf2Δ cells (Fig 4F(iii), quartile 4 sites). A limitation of this last analysis is that there are very few nucleosomes that overlap the Gcn4 motifs at this group of sites, which are centered in the middle of NDRs, making it difficult to assess whether the nucleosome peaks exhibit the broadening expected for fuzzier positioning. However, there is also no obvious broadening of the adjoining -1 or +1 nucleosomes surrounding these sites (Fig 4F(iii)). In contrast, the 5' sites that exhibit decreased (versus increased) Gcn4 binding in snf2Δ cells exhibit increased nucleosome dyad densities at the binding motifs (Fig 4F(i), quartile 1 sites), consistent with the H3 ChIP-Seq data shown in S11B(ii) Fig for these sites (at the top of the map), and is to be expected if SWI/SNF stimulates Gcn4 binding by evicting nucleosomes at these motifs. In summary, we have no direct evidence that SWI/SNF impedes Gcn4 binding at certain 5' sites by sliding nucleosomes over the binding motifs; accordingly, the inhibitory effect could involve another biochemical activity of SWI/SNF, or it could arise as an indirect consequence of impaired nucleosome eviction at another genomic location.

A combination of impaired Gcn4 binding and nucleosome eviction is associated with reduced transcription in the CR mutants
We asked next whether the reductions in Gcn4 binding at 5' sites in the P TET -STH1 and snf2Δ P TET -STH1 mutants confer reduced transcription of the associated genes. Indeed, for genes in Set_1 and Set_2, decreased Gcn4 binding is generally associated with reduced Rpb3 occupancies averaged over the downstream CDS in snf2Δ P TET -STH1_I versus WT_I cells. This trend is evident in both the heat-maps of Fig 4B (panel (iii) vs. (i), Sets_1-2) and in box-plot comparisons of median occupancies of Rpb3 (Fig 4E(i)-(ii), col. 5 vs. 2) versus Gcn4 (Fig 4D(i) Note however that these genes also exhibit increased promoter H3 occupancies in the snf2Δ P TET -STH1 and P TET -STH1 mutants (Fig 4C(i)-(ii) (cols. 4-5 vs. 2), which might contribute to their reduced transcription owing to impaired PIC assembly.
The genes in Set_3 exhibit increased promoter H3 levels in the snf2Δ P TET -STH1 double mutant (Fig 4C(iii), col. 5 vs. 2), which appear to be insufficient to impair transcription (Fig 4E  (iii), col. 5 vs. 2). Given the nearly WT Gcn4 occupancies of the Set_3 genes in this mutant ( Fig  4D(iii), col. 5 vs. 2), one possibility is that a reduction in Gcn4 binding is required in addition to defective nucleosome eviction for impaired transcription, as observed above for genes in Sets_1 and 2 in the double mutant, perhaps owing to diminished recruitment of other coactivators by Gcn4 at its reduced occupancies. This inference can also account for the finding that the snf2Δ single mutation reduces transcription of only the Set_1 genes at the top of the difference heat maps in S11A(i)-(iii) Fig, which exhibit both reduced Gcn4 binding and marked increases in promoter H3 occupancies. Analysis of the ino80Δ mutant by the same approaches, described fully in S12 Fig, likewise provides evidence that a combination of defects in Gcn4 binding and impaired promoter nucleosome eviction conferred by ino80Δ reduces the transcription of a subset of genes with 5' Gcn4 binding sites.

Evidence that the CRs stimulate TBP recruitment at SM-induced genes by evicting or displacing the +1 nucleosome
We examined next the contributions of SWI/SNF and RSC to PIC assembly at genes with 5' Gcn4 binding sites by conducting ChIP-seq of native TBP in the same CR mutants described above. The averaged occupancy plots revealed strong induction of TBP binding on SM treatment of WT cells, peaking~130 bp downstream of the 5' motifs ( Fig 5A(i), yellow vs. blue). Lower-level TBP binding also was induced~200 bp upstream of the motifs, consistent with activation of bidirectional promoters by Gcn4. Essentially identical results were obtained in WT cells treated with Dox prior to SM treatment (Fig 5A(ii), black vs. blue), which served as the WT control for the P TET -STH1, and snf2Δ P TET -STH1 mutants. The averaged TBP occupancies at both locations were substantially reduced in the SM-treated snf2Δ P TET -STH1 double mutant and ino80Δ strain (Fig 5A(ii), cyan vs. black; Fig 5A(i), gold vs. yellow), but were either unaffected or slightly increased, respectively, in the P TET -STH1 and snf2Δ single mutants (Fig 5A(ii), red vs. black; Fig 5A(i), grey vs. yellow). Interestingly, while not reducing TBP occupancies, the snf2Δ mutation shifted the position of TBP binding upstream towards the Gcn4 binding site (Fig 5A(i) and 5C(ii)). The latter might indicate that TBP is recruited by Gcn4 to the UAS but is not efficiently delivered to the core promoter downstream, but additional work is required to understand the underlying mechanism. For Gcn4 motifs located within ORFs, induction of TBP occupancy peaks in WT cells was again observed both upstream and downstream of the motifs (S13A and S13B Fig), consistent with our

Fig 5. Reduced TBP recruitment in SWI/SNF and RSC mutants is frequently associated with increased +1 nucleosome occupancies in promoters of genes with 5' Gcn4 binding sites. (A) (i)-(ii)
Averaged TBP occupancies surrounding Gcn4 5' motifs from ChIP-seq data of sonicated chromatin from at least 2 biological replicates of the indicated strains, with labels along the plots are arranged in decreasing order of summit heights. (B) (i)-(ii) Notched box plots of log 2 TBP occupancies measured within ±100 bp of TBP peak summits identified by MACS2 peak analysis in the promoters of 83 of the 5' Gcn4 target genes in the indicated yeast strains, color-coded as in (A). (C) Heat map depictions of differences in TBP occupancies surrounding TBP peak summits in the 83 5' genes from (B) sorted in decreasing order of WT_I TBP levels in (i) WT_I vs. WT_U, (ii) snf2Δ_I vs. WT_I, (iii) P TET -STH1_I vs. WT_I, (iv) snf2Δ P TET -STH1_I vs. WT_I, and (v) ino80Δ_I vs. WT_I, color-coded as indicated to the right. Two equal bins based on WT_I TBP of High_TBP or Low_TBP occupancies are indicated. (D) Notched box plots of factor occupancies in the 83 5' gene promoters for the two bins defined in (C) for (i) log 2 TBP within ±100 bp of TBP peak summits, (ii) log 2 Gcn4 at the respective Gcn4 peaks, (iii) H3 within ±100 bp of the TBP peak summits, and (iv) log 2 Rpb3 within the CDS of target genes. (E) Average dyad density from H3 MNase-ChIP-seq data aligned to the TBP summits of the 83 5' genes binned as in (C). Midpoints (dyads) of nucleosomal size sequences between 125 and 175 bp were averaged and plotted relative to the TBP identification of bidirectional antisense and subgenic sense transcripts induced by internal Gcn4 binding [46]. As above, the averaged TBP occupancies within ORFs was markedly reduced only in the snf2Δ P TET -STH1 double mutant and the ino80Δ strain (S13A and S13B Fig).
To quantify the effects of the CR mutations on TBP recruitment at individual genes with 5' Gcn4 peaks, we determined the TBP occupancies in a window of ±100 bp surrounding the summits of each TBP peak that we identified at 83 of the 117 genes with 5' Gcn4 sites. Consistent with the averaged TBP data in Fig 5A, the results revealed significant reductions in median TBP occupancies only in the snf2Δ P TET -STH1 and ino80Δ mutants compared to WT_I cells (Fig 5B(i)-(ii)). We also displayed the TBP occupancy changes gene-by-gene using TBP difference heat maps in which the 5' genes were sorted by their TBP occupancies in WT_I cells. As might be expected, the genes with highest TBP occupancies in WT_I cells also display the largest inductions of TBP binding on SM-treatment of WT cells (Fig 5C(i), High vs. Low group). Most of the genes in the High_TBP group (upper half of each map) show reductions in TBP binding in both the snf2Δ P TET -STH1 and ino80Δ mutants compared to WT induced cells ( Fig  5C(iv)-(v)); whereas only about one-half of these genes exhibit even moderately reduced TBP binding in the P TET -STH1 single mutant (Fig 5C(iii)), and none show reduced TBP binding in snf2Δ_I vs. WT_I cells (Fig 5C(ii)). Together, the results in Fig 5A-5C indicate overlapping functions of RSC and SWI/SNF in promoting TBP recruitment on Gcn4 binding at 5' sites, wherein a strong recruitment defect occurs only when both Sth1 and Snf2 are removed simultaneously in the snf2Δ P TET -STH1 double mutant. In contrast, eliminating Ino80 confers a marked reduction in TBP recruitment in otherwise WT cells at most of the genes exhibiting the highest levels of TBP recruitment by Gcn4 in WT cells.
Recently, we provided evidence that reductions in Pol II occupancies in ino80Δ cells are associated with both decreased TBP occupancies and increased promoter H3 occupancies at the SM-induced genes, consistent with the idea that Ino80C promotes Gcn4-activated PIC assembly and transcription by evicting promoter nucleosomes [26]. To extend this model to include RSC and SWI/SNF, we quantified the H3 occupancies surrounding the summits of the TBP peaks for the High_TBP and Low_TBP groups of 5' genes defined above. As expected, these H3 occupancies are reduced on SM-treatment of WT cells (Fig 5D(iii), High_ & Low_TBP, cols 1-2), consistent with eviction of promoter nucleosomes to facilitate TBP binding. Importantly, we observed increased H3 occupancies in the snf2Δ P TET -STH1 double mutant that are greater in magnitude than observed in the snf2Δ or P TET -STH1 single mutants under inducing conditions, particularly for the High_TBP group (Fig 5D(iii), High_ & Low_TBP, cols 3-5 vs. 2). We also examined the positions of nucleosome dyads measured by H3 MNase ChIP-Seq analysis, which revealed a shift upstream in the averaged dyad densities of +1 nucleosomes that results in greater dyad densities immediately 5' of the TBP summits in the mutants, which was most pronounced in the double mutant for the High_TBP group ( Fig  5E(i)-5(ii), cyan vs. grey & red vs. yellow). The fact that combining the snf2Δ and P TET -STH1 mutations in the double mutant confers additive increases in nucleosome occupancies at TBP binding sites that are coupled with additive reductions in TBP recruitment at many of these sites (noted above in Fig 5D(i), cols. 6 vs. 4-5) supports the model that RSC and SWI/SNF functionally cooperate to enhance Gcn4-activated TBP binding by eviction and repositioning of +1 nucleosomes. The coupling of reduced TBP binding and elevated H3 occupancies in the snf2Δ P TET -STH1 double mutant is illustrated for six archetypal genes with 5' Gcn4 sites in summits. Profiles were smoothed using a moving average filter with a span of 31 bp. Yeast strain labels are arranged in decreasing order of dyad densities at the +1 nucleosome spanning the TBP summits.
Consistent with our previous findings [26], reduced TBP recruitment conferred by ino80Δ at genes with 5' Gcn4 sites also is frequently associated with elevated H3 occupancies at the TBP binding sites (S13C (i)-(ii) Fig), as illustrated for four archetypal genes in S13D and S13E Fig. Finally, as might be expected, the Pol II (Rpb3) occupancies in WT_I cells are greater for the High_TBP versus Low_TBP group of genes containing 5' Gcn4 sites (Fig 5D(iv), High_ vs. Low_, col. 2). Importantly, we observed marked reductions in Rpb3 occupancies at these genes only in the snf2Δ P TET -STH1 double mutant, with smaller reductions in the P TET -STH1 single mutant and no significant changes in the snf2Δ single mutant (Fig 5D(iv), cols. 3-5 vs. 2). The fact that the changes in Poll II occupancies generally parallel the changes in TBP occupancies in these mutants (noted above in Fig 5D(i) and Fig 5C(ii)-(iv)) supports the model that RSC and SWI/SNF functionally cooperate to enhance Gcn4-activated transcription of these genes by stimulating PIC assembly.

SWI/SNF functions with RSC and Ino80C in TBP recruitment only at highly expressed genes in non-stressed cells
Having found that SWI/SNF, RSC, and Ino80C all participate in TBP recruitment at genes with 5' Gcn4 binding sites in starved cells (Fig 5A and 5B), we next examined their relative importance for this step of PIC assembly at genes expressed at high levels in non-starved cells (ie. not treated with SM). In the absence of stress,~50% of Pol II and the transcriptional machinery is devoted to transcription of ribosomal protein genes (RPGs) [50]. Accordingly, we examined the RPGs as a group and divided the remaining constitutively expressed genes into ten deciles according to their Rpb3 occupancies. Interestingly, all four of the CR mutations conferred significantly decreased TBP occupancies at the RPGs, with the greatest reductions in the snf2Δ P TET -STH1 double mutant (Fig 6A), comparable marked reductions in the P TET -STH1 and ino80Δ single mutants, and somewhat smaller reductions in the snf2Δ single mutant (Fig 6A). These findings indicate functional cooperation between RSC and SWI/SNF, and a substantial contribution from Ino80C, in TBP recruitment at the RPGs. All of the mutants also displayed reduced Rpb3 occupancies at the RPGs, but here snf2Δ conferred reductions comparable to those observed in the snf2Δ P TET -STH1 double mutant and the ino80Δ mutant, with lesser reductions in the P TET -STH1 single mutant (Fig 6B). These last results might indicate that SWI/SNF makes an additional contribution to transcription beyond TBP recruitment at these genes.
Examining changes in TBP occupancies at the remaining genes reveals reduced TBP recruitment for the~40% most highly expressed genes in Deciles 1-4 in all four CR mutants, albeit of smaller magnitude compared to the RPGs (Fig 6C). By comparison to the RPGs, SWI/ SNF makes a smaller contribution than RSC for the genes in Deciles 1-4, such that reductions in TBP occupancies are smaller in the snf2Δ versus P TET -STH1 single mutant, and similar in magnitude between the P TET -STH1 single mutant and snf2Δ P TET -STH1 double mutant ( Fig  6C, cols. 1-3 for Deciles 1-4). The effect of ino80Δ is relatively greater in comparison to the other mutants for Deciles 1-4 compared to that observed for the RPGs, being comparable to the snf2Δ P TET -STH1 double mutant in reducing TBP recruitment for the former genes ( Fig  6C, Deciles 1-4, cols. 3-4). For most of the remaining genes in Deciles 5-9, it appears that

PLOS GENETICS
RSC and Ino80C make comparable, modest contributions to TBP recruitment with little involvement of SWI/SNF, as we observe similar reductions in TBP occupancies in the P TET -STH1, snf2Δ P TET -STH1, and ino80Δ mutants, but no significant change in snf2Δ cells (Fig 6C,  Deciles 5-9, cols. 1-4). All of the mutants show somewhat elevated TBP occupancies for Decile 10 ( Fig 6C, Decile 10, cols. 1-4), which might result from reduced competition with the more highly expressed genes whose ability to recruit TBP is impaired in the CR mutants. Overall, the results indicate that RSC and Ino80C function broadly to enhance TBP recruitment at constitutive genes regardless of gene expression levels, whereas SWI/SNF acts primarily at highly expressed genes, and particularly at the RPGs, to help stimulate this step of PIC assembly.
Examining averaged occupancies of Snf2-myc, Sth1-myc, and Ino80-myc at all genes in uninduced cells reveals enrichment of all three factors upstream of the TSS, with relatively higher averaged occupancies for Ino80-myc and Sth1-myc versus Snf2-myc (Fig 6D(i)). Interrogating individual genes using heat maps, sorting genes by Rpb3_I occupancies, shows that Sth1-myc and Ino80-myc are moderately enriched upstream of the TSS at nearly all genes, whereas Snf2-myc is more restricted to highly expressed genes near the top of the map (Fig 6D  (ii)-6(iv)). Thus, the relative occupancies of RSC, SWI/SNF, and Ino80C generally parallel their relative importance for TBP recruitment at these gene groups, which is consistent with direct contributions of the three CRs to this step of PIC assembly throughout the genome.
Finally, we compared the dependence on the three remodelers for TBP recruitment between genes that rely primarily on TFIID or utilize SAGA in addition to TFIID for this step of PIC assembly. The 655 genes designated as coactivator-redundant, which exhibit overlapping functions for SAGA or TFIID [51], show greater reductions in TBP in the P TET -STH1, snf2Δ P TET -STH1, and ino80Δ mutants than do the 4245 genes that rely primarily on TFIID (S15 Fig). Consistent with this, the most highly transcribed decile of the 4680 non-RPG genes (Decile 1), shown above to exhibit the greatest TBP reductions in all four remodeler mutants among all deciles (Fig 6C), are highly enriched for coactivator-redundant genes (53%) compared to Deciles 2-10 (5-19% coactivator-redundant genes) and the non-RPG genes as a group (~14%). Moreover, the genes with 5' Gcn4 sites, which show redundant contributions by SWI/SNF and RSC and a marked requirement for Ino80C for TBP recruitment (Fig 5B(i)-(ii)), are comprised of 45% coactivator-redundant genes. These findings suggest a tendency for coactivator-redundant genes to show a greater requirement than TFIID-dependent genes for RSC and Ino80C, and also for SWI/SNF among the most highly expressed members, for robust TBP recruitment. Interestingly, the RPGs are dramatic exceptions to this tendency, as they exhibit the strongest reductions in TBP recruitment in all four CR mutants of any gene group we analyzed but are almost exclusively (95%) TFIID-dependent genes. Thus, while showing minimal dependence on SAGA, the high-level transcription achieved by the RPGs involves the concerted functions of all three CRs for efficient TBP recruitment.

Discussion
In this report, we have provided evidence for distinct contributions of three CRs in regulating the binding of transcriptional activator Gcn4 at its numerous target genes, and illuminated functions of these CRs in controlling TBP recruitment for PIC assembly and transcription, both at Gcn4 target genes and more generally throughout the genome. Gcn4 binds in the NDRs of most of its target genes (5' sites), but within the CDS at a subset of activated genes (ORF sites) where it stimulates bidirectional transcription of antisense and subgenic sense transcripts as well as activating the 5'-positioned promoters [46]. We found that depleting or eliminating the catalytic subunits of RSC (Sth1) or Ino80C (Ino80) reduces Gcn4 occupancies at both 5' and ORF binding sites; with differential contributions of these two CRs at different 5' sites. Surprisingly, SWI/SNF functions oppositely at 5' sites in WT cells, generally reducing their Gcn4 occupancies, while having little effect at the ORF sites. In cells depleted of Sth1, by contrast, SWI/SNF functionally substitutes for RSC in promoting Gcn4 binding at both 5' and ORF sites, as eliminating Snf2 exacerbates the effect of depleting Sth1 in reducing Gcn4 occupancies at the sites most affected by depleting Sth1 alone from otherwise WT cells. These opposing effects of SWI/SNF appear to cancel out at the subset of 5' sites where SWI/SNF most strongly impedes Gcn4 binding in WT cells, such that the double mutation snf2Δ P TET -STH1 has little effect on their Gcn4 occupancies. Ino80C also stimulates Gcn4 binding at 5' sites that are highly dependent on RSC, but it stimulates binding at other 5' sites that are largely independent of RSC, and where SWI/SNF impedes rather than enhances Gcn4 binding. Thus, our results have uncovered distinct contributions of these three CRs to Gcn4 binding at different 5' sites.
The Gcn4 binding motifs of 5' sites tend to reside in the center of NDRs, while the ORF motifs are generally located in linkers connecting genic nucleosomes, suggesting that Gcn4 binding is impeded by inclusion of its recognition sequence within nucleosomes [46]. Consistent with this, our results suggest that RSC and Ino80C enhance Gcn4 binding by removing nucleosomes from the binding motifs in NDRs. Thus, the subset of 5' sites showing decreased Gcn4 occupancies in the P TET -STH1 single mutant or in the ino80Δ strain also tend to exhibit defective H3 eviction at these 5' sites in those mutants (Figs 4C and 4D(i) & S12B and S12C (i)). We observed similar inverse relationships between effects of the snf2Δ P TET -STH1 double mutation on H3 and Gcn4 occupancies for the subsets of 5' sites showing the largest reductions in Gcn4 occupancies in this double mutant (Sets_1 and _2). This correlation was not evident however for the 5' sites that are least impaired for Gcn4 binding in the double mutant (Set_3), where increased H3 occupancies were not accompanied by appreciable reductions in Gcn4 binding (Fig 4C and 4D, (iii) vs. (i)-(ii)), which we attributed to eliminating an inhibitory effect of SWI/SNF on Gcn4 binding at these sites by the snf2Δ mutation in the double mutant. The dual opposing effects of SWI/SNF on Gcn4 binding can also account for the fact that the 5' sites in Sets_1 and _2 show increased H3 occupancies in the snf2Δ single mutant but generally no reduction in Gcn4 binding (Fig 4C and 4D(i)-(ii)); and also that the sites in Set_3 exhibiting the least effect of snf2Δ on H3 eviction show increased Gcn4 binding in this mutant (Fig 4C  and 4D(iii)). Thus, the effect of snf2Δ on Gcn4 binding at any given site appears to reflect the relative importance of SWI/SNF in evicting nucleosomes (to enhance binding) versus its opposing role in diminishing Gcn4 binding.
The negative effect of SWI/SNF on Gcn4 binding in WT cells is not unprecedented. In vitro studies indicate that SWI/SNF can displace the Gal4 DNA binding domain from its DNA binding site by sliding a nucleosome across the binding site [41]; and human SWI/SNF was found to displace glucocorticoid receptor from its binding sequence in a reconstituted nucleosome array, contributing to the transient nature of GR interactions with the promoter in chromatin [42]. We considered the intriguing possibility that SWI/SNF functions similarly by sliding nucleosomes that have not been evicted from the NDR across Gcn4 binding motifs to displace bound Gcn4, offsetting the stimulatory effects of nucleosome eviction at these binding sites exerted by other CRs, and that the outcome of these opposing effects differs at different 5' sites. However, we failed to observe higher occupancies of H3 or greater densities of nucleosome dyads in WT cells at the motifs of 5' sites where snf2Δ increases Gcn4 binding, nor could we detect greater fuzziness of such nucleosomes or the adjoining -1 and +1 nucleosomes in WT cells that might reflect transient nucleosome sliding back and forth across the motifs. It is possible therefore that deleting SNF2 increases Gcn4 binding indirectly, eg. altering the expression or activity of an unknown factor that either reduces Gcn4 affinity or competes with Gcn4 for binding motifs. One finding disfavoring such an indirect mechanism, however, is that Snf2 occupancy at the Gcn4 sites is highly correlated with Snf2 function in suppressing Gcn4 binding (Fig 3D(i) vs. (iv)). Another possibility compatible with the latter is that SWI/SNF's direct interactions with Gcn4 itself [48,52] might dislodge Gcn4 from its binding motifs. Since the 5' motifs where SWI/SNF acts negatively are nearly devoid of nucleosomes, there would be no offsetting stimulatory effect of SWI/SNF through nucleosome eviction. Even in the case of RSC and Ino80C we cannot exclude the possibility that interactions of the CR with Gcn4 or other coactivators/PIC components influence Gcn4 binding independently of the established functions of the CRs in evicting or sliding nucleosomes.
By ChIP-seq analysis of myc-tagged versions of Sth1, Snf2, and Ino80, we obtained strong evidence that the CRs act directly to modulate Gcn4 binding at 5' sites, finding that their peak occupancies are centered around the Gcn4 binding motifs. These occupancies increased on induction of Gcn4 by SM treatment, consistent with widespread recruitment of the CRs by Gcn4 to UAS elements. Further evidence for the latter assertion was reported previously by demonstrating reduced SWI/SNF or RSC recruitment at the Gcn4 target gene ARG1 in a gcn4Δ mutant [49,53], and similar evidence was obtained here for Gcn4-dependent Ino80-myc recruitment for the entire cohort of 5' Gcn4 binding sites. We envision that the basal levels of RSC and Ino80C stimulate Gcn4 binding to the UAS elements as Gcn4 abundance begins to rise in response to starvation, and that the bound Gcn4 then recruits higher levels of these two CRs to create a positive feedback loop of mutually stimulatory Gcn4 and CR binding/recruitment as induction proceeds (Fig 7). Even though it is not the most parsimonious explanation, we cannot rule out the alternative possibility that an unknown factor is responsible for recruiting Gcn4, RSC and Ino80C to higher levels in starved versus unstarved cells. Basal levels of SWI/SNF might similarly enhance Gcn4 binding during the early stages of induction, but the increasingly higher levels of SWI/SNF recruited as Gcn4 occupancy rises would frequently act (2) Induction of Gcn4 expression by amino acid starvation (dark purple) leads to increased Gcn4 binding by mass action. The increased Gcn4 occupancy leads to additional recruitment of the CRs, which in turn increases eviction and displacement of nucleosomes from the Gcn4 motif to favor additional Gcn4 binding, thus establishing a positive feedback loop that promotes Gcn4 binding (3). At a subset of binding sites, increased SWI/SNF recruitment enhances an inhibitory effect on Gcn4 binding by an unknown mechanism, leading to dissociation of Gcn4 from the NDR (4) and reversing the stimulatory effects of other CRs on Gcn4 binding (5).
https://doi.org/10.1371/journal.pgen.1010277.g007 to impede Gcn4 binding in a manner that counteracts the enhanced Gcn4 binding stimulated by the other CRs. The basal levels of the CRs posited in this model could arise from stochastic interaction with NDRs, or from recruitment by basal levels of Gcn4 or other transcription factors bound to the UASs in non-starved cells. Presumably, the elevated occupancies of the CRs achieved at high-level Gcn4 binding enable them to diffuse laterally from the NDR to mediate the eviction and repositioning of the -1 and +1 nucleosomes we observed in starved WT cells [12,26,30].
Our model that Gcn4 stimulates its own binding by recruitment of CRs that evict nucleosomes from its binding sites (Fig 7) provides a molecular mechanism that may help to explain previous observations suggesting that an activation domain can facilitate DNA binding by the activator protein. Bunker and Kingston reported that tethering polycomb group proteins to a test promoter differentially interferes with promoter activation by various activation domains that were fused to the Gal4 DNA binding domain [54]. It is possible that the ability of an activation domain to recruit one or more CRs was instrumental in resisting the repressive effects of polycomb group proteins. Vashee and Kodadek found that the Gal4 activation domain could outperform the VP16 activation domain when each was linked to the Gal4 DNA binding domain for trans-activation specifically from weak Gal4 binding sites [55]. It is possible that the Gal4 AD was more efficient than the VP16 AD in recruiting CRs that eliminated nucleosomes surrounding the Gal4 binding sites in a manner that preferentially facilitated Gal4 binding to non-consensus motifs.
It seems likely that the increased nucleosome occupancies in the NDRs and core promoters of Gcn4 activated genes conferred by the CR mutations observed here and elsewhere [12,30] lead to reduced transcription initiation by occluding binding sites for TBP or other general transcription factors involved in PIC assembly. Indeed, we showed recently that reductions in Pol II occupancies in ino80Δ cells are associated with both decreased TBP occupancies and increased promoter H3 occupancies at the SM-induced genes [26]. Here, we observed additive reductions in TBP binding at genes with 5' Gcn4 sites on combining the snf2Δ and P TET -STH1 mutations in the double mutant, which was associated with the additive effects of these mutations in elevating nucleosome occupancies at the TBP binding sites. The relatively smaller increases in H3 occupancies observed in the P TET -STH1 or snf2Δ single mutants were associated with lesser reductions in TBP binding in the P TET -STH1 cells than found in the double mutant, or even increased TBP binding in snf2Δ cells. The latter may reflect the elevated Gcn4 occupancies found at most 5' sites in snf2Δ cells, which might enable robust TBP recruitment despite increased nucleosome occlusion of the core promoters.
The functional cooperation between SWI/SNF and RSC in TBP recruitment observed at Gcn4 target genes in SM-treated cells was also observed in non-starved cells for the group of highly expressed RP genes, for which snf2Δ impaired TBP recruitment in otherwise WT cells and exacerbated the recruitment defect conferred by P TET -STH1 in the double mutant. At the large majority of other genes, however, RSC is much more important than SWI/SNF in promoting TBP recruitment in non-starved cells, where it functions on par with Ino80C. The greater importance of RSC and Ino80C in TBP recruitment is mirrored by their greater occupancies compared to SWI/SNF at the majority of constitutively expressed genes. It is interesting that genes capable of utilizing SAGA or TFIID as a coactivator have a greater requirement for RSC and Ino80C compared to the genes that rely primarily on TFIID. Our findings support the idea that the three CRs act directly to enhance PIC assembly throughout the genome, with RSC and Ino80C acting broadly and with particular importance at coactivator-redundant genes, whereas SWI/SNF function is concentrated at certain highly expressed genes, including RP genes in non-starved cells and Gcn4-induced genes in amino acid-starved cells.

ChIP-seq and PCR ChIP analysis of Gcn4 occupancy
WT strain BY4741 and snf2Δ or ino80Δ mutant strains were cultured for at least 3 doublings in synthetic complete medium lacking isoleucine and valine (SC-Ilv) to log-phase (OD 600 = 0.6-0.8) and SM was added at 1 μg/ml for 25 min to induce Gcn4 synthesis. To transcriptionally deplete Sth1 from P TET -STH1 and snf2Δ P TET -STH1 strains, cells were cultured in SC-Ilv medium with 10 μg/μl doxycycline for 8-9h to log-phase (OD 600 = 0.4-0.6), allowing 2-3 doublings, and SM was added at 1 μg/ml for 25 min as above. ChIP-seq was conducted and DNA libraries for Illumina paired-end sequencing were prepared as described previously [12] except that chromatin samples containing 5 μg DNA were immunoprecipitated using Gcn4 antibodies [59] for 3h in experiments comparing WT with snf2Δ and P TET -STH1 mutants, and overnight in experiments comparing WT and ino80Δ strains. Biological replicates of Gcn4 ChIPseq data (and all other ChIP-seq data) are described in S2 Table. PCR-based ChIP analysis of Gcn4 occupancy at four target genes was performed as previously described [30].

ChIP-seq analysis of TBP
For analyzing TBP occupancies, strains were cultured in the presence or absence of SM treatment and subjected to ChIP-seq analysis using antibodies against native TBP, as previously described [26].
ChIP-seq analysis of Snf2-myc 13 , Sth1-myc 13 and Ino80-myc 13 SNF2-myc and STH1-myc strains were cultured in the presence or absence of SM treatment, and untagged WT strain BY4741 cultured with SM, were subjected to ChIP-seq analysis as described previously [30] with a few modifications. Yeast cells (100 ml) were cross-linked for 15 min with 10 ml formaldehyde solution (50 mM HEPES KOH, pH 7.5, 1 mM EDTA, 100 mM NaCl and 11% formaldehyde) and quenched with 15 ml of 2.5 M glycine. WCEs were prepared by glass-beads lysis in 400 μl FA lysis buffer (50 mM HEPES KOH, pH 7.5, 1 mM EDTA, 150 mM NaCl, 1% TritonX-100 and 0.1% Na-deoxycholate) with protease inhibitors for 45 min at 4˚C and the supernatant collected after removing the beads was pooled with 600 ml FA lysis buffer used for washing the beads. The resulting lysate was sonicated to yield DNA fragments of 300-500 bp and cleared by centrifugation. Chromatin samples containing 5.0 μg DNA were immunoprecipitated with anti-myc antibodies (Roche) for 3h. Paired-end sequencing libraries were prepared from immunoprecipitated DNA using Illumina paired-end kits from New England Biolabs (cat. #E7370 and #E7335). ChIP-seq analysis of Ino80-myc was previously described [26]. The myc occupancies at each nucleotide, normalized to the average myc occupancy per nucleotide on the chromosome, were calculated for each myc-tagged strain and for the untagged WT and gcn4Δ strains, and the latter values were subtracted from the former values to yield the corrected occupancies of the myc-tagged CR subunits in the tagged strains.

Bioinformatics analysis of ChIP-seq data
Paired-end sequencing (50 nt from each end) was conducted by the DNA Sequencing and Genomics core facility of the NHLBI, NIH. Sequence data were aligned to the SacCer3 version of the genome sequence using Bowtie2 [60] with parameters -X 1000-very-sensitive, to map sequences up to 1 kb with maximum accuracy. PCR duplicates from ChIP-seq data were removed using the samtools rmdup package. Numbers of aligned paired reads from each ChIP-seq experiment and correlation coefficients for genome-wide occupancy profiles of the different replicates are summarized in S2 Table. Raw genome-wide occupancy profiles were obtained from the alignment (.bam) files by counting the number of DNA fragments that overlapped with every bp, using the coverage methods for GRanges objects from the Genomi-cRanges package in R. To allow the comparison between different samples, each profile was normalized such that the average occupancy for each chromosome was equal to one. Heat maps showing alignments of multiple loci were generated in R using custom scripts (https:// github.com/rchereji/bamR). To visualize specific loci, igvtools was used to create tracks (tdf files) that were loaded in the Integrative Genomics Viewer (IGV, Broad Institute) [61]. Transcript end coordinates (TSS and TTS) were obtained from [62].
For calculation of per base pair occupancies in the given intervals, generation of averaged occupancies profiles and heat maps of occupancies or occupancy differences surrounding the given genomic features, a custom R scripts package available at github.com/rchereji/bamR was employed. These R scripts, originally designed to plot ChIP-seq occupancies surrounding genic features like TSS and TTS or within coding sequences, were adapted to generate heat maps of Gcn4 occupancy differences within peak coordinates, TBP occupancy differences surrounding the peak summits, and histone H3 occupancies and nucleosome dyad densities surrounding Gcn4 motifs or TBP summits. To do so, alternative genome annotation files were generated with either Gcn4 peak start and end coordinates or 200 bp intervals surrounding the TBP peak summits replacing ORF start and end coordinates; and with Gcn4 motif centers or TBP peak summits replacing TSS coordinates. The annotation information employed for these analyses are provided in S1 File. Box plots, scatter plots and line plots were generated using R package "ggpubr".

Identification of TBP peaks
To identify TBP peaks in the gene promoters of Gcn4 5' target genes, MACS2 (http://liulab. dfci.harvard.edu/MACS/) analysis was employed to TBP ChIP-seq data from two replicates of WT_I cultures (S2 Table), using a threshold for the p-value of 10 −3 (S1 File, sheet "MACS2_TBP_Ind"). Among 117 5' target genes, we identified TBP peaks in the promoters of 83 5' target genes (S1 File, sheet "TBP_peak_annotations"). MACS2 analysis of WT_U TBP ChIP-seq data was conducted as above using a p-value of 10 −3 for identifying the TBP peak within the promoter of 134 RP genes (S1 File, sheet "MACS2_TBP_Unind"). TBP peaks were assigned to specific genes by using the bedtools utility "closest" to identify the TBP peak summit closest to the TSS of each gene, followed by assessment of the TBP summit by examining the TBP data in the IGV browser. As tDNA genes often reside in the 5' non-coding regions of Pol II-transcribed genes and display large TBP peaks in both induced and uninduced conditions, the TBP occupancies within 250 bp surrounding tDNA genes (S1 File, sheet "tDNA_annotations") were subtracted from the TBP coverage in the bed file of merged replicate data and the corrected bed file was used to generate averaged profiles and calculate TBP occupancies within 200 bp intervals surrounding the TBP summits. The computation of TBP occupancies in the intervals comprised of 100 bp on either side of the TBP summits, as done in Fig 5B, was found to be unaffected if the TBP occupancies at neighboring tDNA genes were not subtracted in the manner just described.

Data and software availability
File S1 contains (i) annotations of Gcn4 and TBP occupancy peaks listing the coordinates of the Gcn4 occupancy peaks or the windows of ±100 bp surrounding the summits of each TBP peak used in plotting heat maps of Gcn4 or TBP occupancies, and (ii) results of MACS2 analysis of uninduced and SM induced TBP ChIP-seq data. A file listing the genomic coordinates of the 250 bp surrounding each tDNA gene is also provided in File S1. File S2 contains the data used to construct all plots. Unprocessed and analyzed ChIP-seq data have been deposited in the NCBI GEO database under accession number GSE192592 Supporting information S1 Table. Yeast strains used in this study. Names, genotypes and sources of each strain are given. (DOCX) S2 Table. Compilation of ChIP-seq replicate experiments. The strain and growth condition (SM-induced (I), or uninduced (U)), immunoprecipitating antiserum (IP), sample identification number, total number of sequencing reads (All PE reads), number of reads after removing duplicate reads (PE rmdup), correlation coefficient between indicated replicates, and source of data, are given for each ChIP-seq experiment. (DOCX) S3 Table. PCR ChIP data for Gcn4 binding at four exemplar target genes. Mean occupancies (+/-SEM or +/-SD) of Gcn4 in the UAS regions of the indicated genes is expressed as the ratio of input DNA recovered in the immunoprecipitates corrected for the same ratio measured for non-transcribed sequences from chromosome V (analyzed as a control for non-specific immunoprecipitation) as described previously [30]. (DOCX) S1 Fig. Supporting evidence that SWI/SNF and RSC have differential effects on Gcn4 binding at 5' sites. (A-B) Box plots of log 2 Gcn4 occupancies in biological replicates of WT_U, WT_I, or snf2Δ_I, P TET -STH1_I and snf2Δ P TET -STH1_I cells for all 5' sites (A), or in the three sets of Gcn4 5' sites defined in Fig 2B (B). (C) Box plots of H3 occupancies in biological replicates of WT_U, WT_I, or snf2Δ_I, P TET -STH1_I and snf2Δ P TET -STH1_I cells for the three sets of 5' Gcn4 sites defined in Fig 2B. (DOCX)

S2 Fig. Supporting evidence that SWI/SNF and RSC have differential effects on Gcn4 binding at particular 5' sites.
Gene browser profiles of Gcn4 occupancies from biological replicates for the indicated strains/conditions. The Gcn4 peak numbering assigned previously [46] is given at the top of each profile, and the Gcn4 occupancies per nucleotide averaged over the peaks determined in this study are listed next to each peak.  [12] by Western blot analysis in the indicated strains using Gcd6 signals analyzed in parallel as loading control. Band intensities for Gcn4 were normalized to those for Gcd6 in the same samples and the mean Gcn4/Gcd6 ratios determined from 3 biological replicates were plotted. Significance of differences in mean values was calculated with the student's t test.  [46] is given at the top of each profile, and corrected Snf2-myc occupancies per nucleotide within ±100 bp windows surrounding the Gcn4 motifs are listed next to each peak. (DOCX)  Fig 2A and 2B depicting H3 occupancies per base pair in the ±100 bp windows surrounding the Gcn4 motifs. H3 occupancies were calculated from H3 MNase-ChIP-seq data from at least 3 biological replicates of WT_U, WT_I, or snf2Δ_I, P TET -STH1_I and snf2Δ P TET -STH1_I cells. (C-D) Sectored scatterplots of the log 2 ratios of Gcn4 occupancies vs log 2 ratios of H3 occupancies per base pair in the ±100 bp windows surrounding the Gcn4 motifs in snf2Δ P TET -STH1_I vs. WT_I cells (C) or snf2Δ_I vs. WT_I cells (D). The 3 sets of Gcn4 5' sites defined in Fig 2B(ii) are color-coded as: Set_1, red rectangles; Set_2, green pluses; and Set_3, blue stars. (DOCX) S11 Fig. Defective eviction of nucleosomes associated with reduced Gcn4 occupancies at a subset of 5' Gcn4 peaks in snf2Δ_I cells. (A) (i)-(iii) Heat maps depicting differences between snf2Δ_I and WT_I cells for (i) Gcn4 occupancies measured as in Fig 2A(iii), (ii) H3 occupancies surrounding the Gcn4 motifs of 5' sites from H3 ChIP-seq data, and (iii) Rpb3 occupancies averaged over the CDS of 5' genes, for the Gcn4 5' sites sorted by increasing order of foldchanges in Gcn4 occupancies in snf2Δ P TET -STH1_I vs. WT_I cells. occupancies averaged over the CDS of genes with 5' sites. H3 and Rpb3 occupancies were calculated from ChIP-seq data of sonicated chromatin from at least 3 biological replicates of WT_U, WT_I and ino80Δ_I cells. P values from Mann-Whitney-Wilcoxon tests are indicated. The heat map of H3 occupancy changes conferred by ino80Δ around the 5' motifs ordered by the Gcn4 occupancy reductions in this mutant (panel A(i)) reveals that the 5' sites with the strongest reductions in Gcn4 binding in ino80Δ cells located at the top of the map (Set_I) show the strongest increases in H3 occupancies centered around the Gcn4 motifs (panel A(ii)). Moreover, the decreases in median Gcn4 occupancy for individual 5' sites conferred by ino80Δ are paralleled by increased median H3 occupancies for the Set_I group of 5' sites; whereas the sites in Set_II and III show no significant changes in median H3 or Gcn4 occupancies in ino80Δ_I versus WT_I cells (panels 12B-C, Sets_I-III, col. 3 vs. 2). The subset of genes with 5' sites that are most dependent on Ino80C for Gcn4 binding generally show the greatest reductions in Rpb3 occupancies (Set_I sites in panel A (iii) vs. (i)). Moreover, the Set_I genes, but not genes in Sets_II-III, show reduced median occupancies of both Gcn4 and Rpb3 in ino80Δ_I versus WT_I cells (panels C-D(i)-(iii), col. 3 vs. 2). As only the Set_I genes also exhibit increased median H3 occupancies in ino80Δ_I cells (panel B(i)-(iii), col. 3 vs. 2), it seems likely that a defect in Gcn4 binding (and subsequent impaired recruitment of other coactivators) in combination with loss of Ino80C-mediated promoter nucleosome eviction produces the reduced transcription of Set_I genes conferred by ino80Δ. (E) Gene browser profiles of Gcn4 and H3 occupancies from ChIP-seq analyses of sonicated chromatin for the indicated strains, as described in S4