Refined mechanism of promoter nucleosome-depleted regions resetting after replication

Strains

All experiments presented in this study were performed using the budding yeast S. cerevisiae derived from the Anchor-Away genetic backgrounds [38]. A complete list of strains used in this study is provided in Supplementary Table S1. Strains were generated by genomic integration of tagging cassettes as described [39]. EdU-incorporating strains were generated by integrating a plasmid encoding Herpes simplex virus thymidine kinase (HSV-TK) and the human Equilibrative Nucleoside Transporter (hENT1), as previously described [40]. Yeast cells were grown in YEP medium (1% yeast extract, 2% peptone) supplemented with 2% glucose as carbon source at 30°C. Experiments were performed with exponential phase cells harvested between an OD₆₀₀ of 0.4–0.6, unless otherwise indicated. Anchor-away of targeted proteins was induced by the addition of 1 μg/ml of rapamycin (Rap) to the medium for 30 min.

Protein extraction by TCA and western blot

Protein extraction by TCA and western blot were largely performed as in [41]. Primary antibodies were diluted in TBST + 5% milk and incubated overnight at 4°C. Membranes were washed four times for 5 min each with fresh TBST, and horseradish peroxidase-conjugated or fluorescent secondary antibodies were applied for 1 h at room temperature. Chemiluminescence was detected using the Amersham ImageQuant 800 system. Trichostatin A (TSA) treatment was assayed using anibodies agaisnt H3K18ac (Abcam, ab1191) and H3 (Abcam, ab1791).

Cell cycle analysis by flow cytometry analysis

Flow cytometry was performed as described in [41] with unsynchronized cells treated or not with Rap for 30 min. Analyses were performed with Kaluza software.

Labeling of nascent DNA and crosslinking

Yeast cells from a fresh plate were grown overnight in YEPD medium at 30°C and diluted to OD₆₀₀ = 0.3 in YEPD medium next morning. Fifty mililiters of yeast cultures were used for each condition. After one cell cycle, Rap was added to corresponding samples at OD600 = 0.6 and incubated for 30 min at 30°C. For some experiments, TSA (Sigma–Aldrich, T8552) was also added for 30 min at a final concentration of 1 mM. Nascent DNA was labeled by addition of 5-ethynyl-20-deoxyuridine (EdU; Roth, 7845) at a final concentration of 10 μM for 2 min. After labeling, cells were immediately fixed with formaldehyde at a final concentration of 1% for 15 min followed by glycine addition at a final concentration of 125 mM for 5 min at room temperature. Cells were washed twice with HBS (50 mM HEPES–KOH, pH = 7.5, 140 mM NaCl), frozen in liquid nitrogen and kept at −80°C until further steps.

For verification of EdU pulse time, cells were arrested in G1-phase by addition of 200 ng/ml α-factor (PRIMM, 201307-00007) and incubated at 30°C for 2.5 h. EdU was added in the medium at the end of G1 arrest for 2 min and 50 ml of G1-arrested, EdU labeled cells were crosslinked with formaldehyde. The rest of the culture was washed with YEP (no sugar) medium for three times and cells were resuspended in prewarmed YEPD medium and incubated for 30 min at 30°C to release the cells into S-phase. At the end of S-phase release, EdU was added in the medium and cells were incubated at 30°C for an additional 2 min. S-phase released EdU labeled cells were then crosslinked with formaldehyde. Crosslinked cells were washed twice with HBS, frozen in liquid nitrogen and kept at −80°C until further steps.

MNase-seq

The MNase experiments were performed mainly as described in [42] with the following modifications. For chromatin extraction, frozen cell pellets were resuspended in 600 μl cell lysis buffer [50 mM HEPES–KOH, pH = 7.5, 140 mM NaCl, 1 mM ethylenediaminetetraacetic acid (EDTA), 1% Triton X-100, 0.1% sodium deoxycholate, supplemented with 1 mM PhenylMethylSulfonyl Fluoride (PMSF) and EDTA-free protease inhibitors] and 600 μl of glass beads (BioSpecProducts, 11079105) was added to the cells and mechanically lysed by bead beating with a Mini-Bead beater for 1 min 30 s. The cell extract was then centrifuged at 13 000 rpm for 30 min at 4°C to collect chromatin. The chromatin was resuspended in 600 μl of NP buffer (0.5 mM spermidine, 50 mM NaCl, 1 mM β-mercaptoethanol, 0.075% NP-40, 10 mM Tris–HCl, pH 7.4, 5 mM MgCl₂, 1 mM CaCl₂) per reaction. The DNA concentration was measured using Qubit double-stranded DNA (dsDNA) BR assay kit (Invitrogen, Q32850). Chromatin was diluted to ≈20–30 mg/ml using NP buffer to adjust chromatin density in each reaction. MNase treatment was performed using 0.5 μl of MNase (Thermo Scientific, 88216, 100 units/ml), incubated at 37°C and 100 μl of digestion product was aliquoted for different time points (3, 6, 9 min). The MNase digestion was stopped by adding a stop buffer [20 mM EDTA, 0.5% sodium dodecyl sulphate (SDS), 0.5 mg/ml proteinase K] in the reaction and incubated first at 42°C for 1 h and then at 65°C overnight. After reversal of crosslinking, DNA was extracted with phenol-chloroform-isoamylalcohol [PCI; (25:24:1 v/v), Roth, A156], treated with RNase A (Invitrogen, 12091021) and visualized on a 2.5% agarose gel. The nucleosomal DNA sample with the best MNase digestion time point (corresponding roughly to a ratio of 2 for mononucleosomes/dinucleosomes) was purified again with PCI extraction before proceeding further steps.

Click reaction and streptavidin affinity capture

MNase digested and purified DNA fragments were subjected to a click reaction that adds biotin to the EdU-labeled newly synthesized DNA. Click reaction was performed in the following order to the indicated final concentrations and protected from light: 25 ng of DNA was mixed with 1 mM Biotin-PEG-Azide (Sigma–Aldrich, QBD10825). In a separate tube CuSO₄ (Sigma–Aldrich, C1297) and THPTA (Sigma–Aldrich, 762342) were added at 2 mM/4 mM ratio, respectively and mixed by vortexing (CuSO₄/THPTA mix, 1/2 ratio). Then DNA/Biotin PEG Azide mix was combined with CuSO4/THPTA and 10 mM ascorbic acid (Roth, 3525) was added to the mix. The reaction final volume was adjusted to 50 μl by using 10 mM Tris–HCl (pH = 7.5), mixed by vortexing and incubated at 37°C for ≈2 h with shaking at 800 rpm. The click reaction product was purified by ethanol precipitation. Biotin conjugated EdU-labeled DNA was then pulled down using 5 μl of Dynabeads M-280 streptavidin (Invitrogen, 11205D). Beads were pre-washed three times with 1× Tween buffer (5 mM Tris–HCl, pH = 8, 0.5 mM EDTA, 1 M NaCl, 0.05% tween) before being resuspended in 2× B&W Buffer (10 mM Tris–HCl, pH = 8, 1 mM EDTA, 2 M NaCl) and combined with an equal volume of biotinylated DNA. The reaction was incubated at room temperature for 30 min with gentle shaking. The beads were washed twice with 1× Tween buffer and once with 1× B&W Buffer (5 mM Tris–HCl, pH = 8, 0.5 mM EDTA, 1 M NaCl). Following washing steps, bead-bound DNA was eluted by adding sterile water and heating at 70°C for 5 min.

ChEC-seq

ChEC-seq experiments were performed mainly as described in [42] with the following modifications. The yeast strains were cultured the day before in YEPD medium and next morning they were diluted to OD₆₀₀ = 0.2 in YEPD. The 30 min Rap treatment (1 μg/ ml, final) was performed for anchoring away of targeted proteins and/or 30 min TSA treatment (1 mM final) at OD₆₀₀ = 0.5. For each condition, 50 ml cell cultures were harvested at room temperature. The cells were washed once in 1 ml Buffer A [15 mM Tris–HCl, pH 7.5, 80 mM KCl, 0.1 mM Ethylene glycol bis(2-aminoethyl)tetraacetic acid (EGTA), 0.2 mM spermine, 0.5 mM spermidine and EDTA-free protease inhibitor tablet (Roche, 04693159 001)]. Cells were resuspended in 693 μl of buffer A and 7 μl of 10% digitonin (0.1% final concentration; Sigma–Aldrich, D141) were added to facilitate permeabilization of the cells for 5 min at 30°C and mixed with the pipette several times during the incubation. CaCl₂ was added (final concentration 5 mM) to activate the MNase cleavage. For each sample 100 μl of cells were collected at 30 s after the induction of MNase digestion and were immediately mixed with 200 ml of 2× Stop solution (400 mM NaCl, 20 mM EDTA, 4 mM EGTA, 1% SDS). The cells were treated with Proteinase K (0.4 mg/ml final concentration) and incubated at 55°C for 30 min. The DNA was extracted using PCI extraction and precipitated with EtOH by adding 30 μg glycogen. Following RNase treatment, size selection of DNA fragments was performed as follows; 100 μl of AmpureXP beads (Beckman Coulter, A63881) were added in 40 μl of RNase-treated DNA and placed on a magnetic rack. The supernatant was then transferred to the new tube containing 10 mM Tris and 0.2 M NaCl for each reaction. The DNA was extracted with PCI and ethanol precipitated. The purified DNA samples were then used for library preparation using NEBNext UltraII DNA Library prep kit (NEB, E7370L) and were sequenced at the iGE3 platform of the University of Geneva.

Chromatin immunoprecipitation-quantitative PCR (qPCR)

Chromatin, prepared as described previously in [42] from 50 ml of Rpb1-AA strain (OD₆₀₀ ∼0.5) treated or not with Rap (1 μg/ml) for 30 min, was immunoprecipitated with an antibody against Pol II CTD (8WG16, BioLegend). Both IP and Input DNA fragments were amplified using a CFX96 Real-Time polymerase chain reaction (PCR) machine (Bio-Rad) and SYBR Green PCR Master Mix (Applied Biosystems, cat# 4472942).

Data analyses

For MNase-seq experiments, fastq files were first trimmed from their adapters using Trim Galore! (https://github.com/FelixKrueger/TrimGalore). Then, trimmed files were aligned using Bowtie2 on the sacCer3 genome in a paired-end manner with the following options: ‘-k 20–end-to-end–sensitive -X 800’ [43]. BigWig files were generated using bamCoverage from deepTools2 [44] (with the following options: normalization in counts per million (cpm), bin size of 1 bp, exclusion of the chrM for normalization, regions centered with respect to the fragment length, a minimum mapping quality of 20, duplicates removal, size selection of 120–200 bp. For ChEC-experiments, the same processing was followed with the exception of a size selection from 0 to 200 bp. Means of experiments were obtained with the bigwigCompare command from deepTools2 with a bin size of 1bp. Differences in signal between conditions were also calculated using bigwigCompare. Metagene plots and heatmaps were obtained using the plotProfile and plotHeatmap commands from deepTools2, respectively. Enrichment over mean plots were made as followed: average plots were calculated for the whole set of −1/+1 nucleosomes in the −100 to +100 bp window around the dyad, then every position relative to the feature was divided by the average number of reads over this −100/+100 bp window. It allows to normalize the profiles of nascent chromatin showing a reduced read depth due to purification. For most of the experiments, the means of two replicates are presented in main figures while separate replicates are depicted in supplementary figures. For Figs 1–4 and 6, the difference in +1 peak position per gene was calculated using a custom Python script deposited on Zenodo. For Fig. 5, the shifted and nonshifted populations were extracted using DANPOS3 (https://github.com/sklasfeld/DANPOS3) with the replicate 2 of the Reb1-AA strain in the absence of Rap.

−1/+1, +4 annotations and gene/replication origins lists

All the coordinates and gene/replication origins names are listed in Supplementary Table S2. −1/+1/+4 coordinates and NDR centers were obtained from [42] and [45], respectively. Replication origins coordinates were retrieved from [46]. Reb1 targets were extracted from the RNA-seq analysis in a Reb1-AA strain from [47]. Abf1 targets were retrieved from the Rpb1 chromatin immunoprecipitation (ChIP)-seq in an Abf1-AA strain from [48]. Rsc3 targets were defined from a k-means clustering of the difference in MNase-seq profile around +1 nucleosome in the Rsc3-AA (Fig. 3). Highly transcribed and lowly transcribed genes were retrieved from the RNA-seq in an Nrd1-AA strain in the absence of Rap [49].

Reanalysis of existing data

Data from [50] for MNase-seq in absence or presence of TSA or with the epl1 mutant were analyzed following the same procedure as the data analyses section. For MNase-(ChIP)-seq data from [51], single end reads were virtually extended to 150 bp and the centers of these fragments were extracted using deepTools2. For each of the histone modification presented, the corresponding total MNase-seq signal (Input) was subtracted.

Statistical analyses and model design

Plots and statistical analyses were performed using GraphPad Prism 10.0 and the SciPy library in Python. Statistical analyses for the boxplots comparing nucleosome positioning to steady-state were performed as follows. For nontarget genes, we randomly sampled n genes, where n corresponds to the number of target genes, and repeated this procedure 20 000 times; the mean P-value was calculated using the Mann–Whitney test. For strains in which the anchor-away system affects all genes, 500 genes were randomly selected, and the Mann–Whitney test was repeated 20 000 times, with the mean of the resulting P-values reported. When n < 1500, a regular Mann–Whitney test was performed. ns if P-value >.05, ^∗<.05, ^∗∗<.01, ^∗∗∗<.001, ^∗∗∗∗<.0001.

Purification of nascent chromatin after a short EdU pulse

In order to map the position of nucleosomes on the nascent DNA, we used an approach in which S. cerevisiae cells engineered for the uptake of 5-Ethynyl-2′-deoxyUridine (EdU), an analog of thymidine, are incubated in an EdU-containing medium for a short pulse [52]. Considering the rapid average speed of the replication fork in S. cerevisiae cells and the rapid maturation of yeast chromatin [34, 53], we opted for a 2-min DNA labeling. As a proof of principle, we assayed whether a 2 min pulse was enough to specifically detect newly replicated DNA around early replication origins as compared to the late ones in an S-phase synchronized population of cells (Supplementary Fig. S1A). To perform this, we synchronized S. cerevisiae cells in G1-phase before releasing them into S-phase for 30 min in order to enrich in early replicated domains initiated from efficient autonomously replicated sequences. EdU was added for 2 min after a 30 min release from G1-phase. To evaluate the background, G1-arrested cells were separately pulsed with EdU for 2 min during G1-arrest. The chromatin was then extracted, MNase digested to obtain the footprint of nucleosomes along the DNA, labeled with biotin via a Click-it reaction and purified by streptavidin beads. Finally, the eluted material was paired-end sequenced (Supplementary Fig. S1A). Notably, the purification of 2 min EdU-labeled DNA was specific enough to enrich in early replicated DNA (Supplementary Fig. S1B and C). Thus, despite a short pulse of EdU labeling, our experimental set-up allows the detection of newly replicated chromatin among a large majority of nonlabeled DNA.

Reb1 and Abf1 TFs are dispensable for −1/+1 positioning shortly after replication

After assessing its validity, we implemented our experimental scheme in anchor-away strains in which either TFs, CRs or RNAPII can be conditionally depleted by Rap treatment (Fig. 1A). To assess the position of virtually all the nucleosomes on newly replicated chromatin and to not restrict our analyses to early replicated areas, we performed the experiments with nonsynchronized cells. We also paired-end sequenced the steady-state chromatin in parallel in order to compare the positioning of steady-state nucleosomes with their respective positioning in nascent chromatin. Importantly, none of the target depletions used in our study lead to an accumulation of cells in a specific phase of the cell cycle, suggesting that progression during the S phase remains comparable (Supplementary Fig. S1D).

We first tested the importance of Reb1 TF in resetting promoter NDR architecture after replication. Reb1 depletion only affects the expression of a minority of genes that was previously defined [47, 49]. Thus, we performed our analyses by making the distinction between Reb1 ‘target’ and ‘nontarget’ genes (Supplementary Fig. S2A). As already reported, Reb1 depletion leads to a shrinkage of the promoter NDR on the steady-state chromatin at the target genes (Fig. 1B) [14, 28]. We then focused on the positioning of the +1 nucleosome (Fig. 1C, right panel and Supplementary Fig. S2B). We also analyzed the data at the +4 nucleosome, located within the transcription unit. For most of the tested strains (unless otherwise specified), the observed phenotype was similar to that at the +1 nucleosome, highlighting the importance of NDR establishment in organizing nucleosomal arrays (Supplementary Fig. S2C).

Comparison of nascent and steady-state +1 positioning in the absence of Rap for the nontarget genes revealed that within 2 min of EdU, most of the NDR-flanking nucleosomes were correctly positioned in agreement with a rapid resetting of promoter NDRs in S. cerevisiae (Fig. 1C and D, bottom panels) [34]. However, the positioning of the +1 nucleosome was significantly different between nascent and steady-state chromatin for target genes in the absence of Reb1 (Fig. 1C and D, top panels and Supplementary Fig. S2B). Indeed, while steady-state +1 nucleosomes are shifted upstream, nascent +1 are positioned as in the −Rap condition. We made a similar observation for the −1 nucleosome, which is shifted downstream in the steady-state chromatin (Fig. 1C left).

We performed the same analysis by depleting another TF, Abf1. Genes were split into ‘targets’ and ‘nontargets’ (Supplementary Fig. S2D and E) [48]. +1 nucleosome positioning on nascent chromatin in the absence of Abf1 is not significantly different to its positioning in a −Rap condition while steady-state +1 is shifted upstream (Fig. 1E and F, upper panels). The −1 and +4 nucleosomes show the same trend (Supplementary Fig. S2F and G).

Taken together, these results indicate that Reb1 and Abf1 TFs are not required for establishing a proper positioning of the −1/+1 nucleosomes immediately after replication but participate in the long-term maintenance of their steady-state positioning.

Maintenance of histone acetylation rescues −1/+1 positioning defects caused by the absence of Reb1

Our previous work proposed that H3K18 acetylation (H3K18ac) on +1 nucleosomes is associated with a right shift while H3K36 trimethylation (H3K36me3) is associated with a left shift [42]. We decided to deepen this observation by systematically assaying the positioning of modified +1 nucleosomes. To perform this, we reanalyzed the dataset generated by the Friedman and Rando labs [51]. In this work, the MNase-ChIP-seq of the different histone modifications generated an unprecedented high resolution of modified nucleosomes. We reasoned that the sequencing of the input would indicate the sum of all the unmodified/modified nucleosomes while the mapping of a given modification would indicate its specific positioning relative to this sum (Fig. 2A). Thus, we subtracted the corresponding input from its associated modifications and plotted the signal as a Z-score centered on the +1 nucleosome dyad (Fig. 2B). All H3 acetylated nucleosomes are right shifted as compared to the position of the +1 nucleosome dyad (Fig. 2B). Moreover, the H3K18ac profile obtained by MNase-ChIP-seq from our previous work shows a similar behavior [42]. These analyses happen to be true for the −1 nucleosome but instead with an associated left shift implying that histone H3 acetylation increases promoter NDR length (Supplementary Fig. S3A). While it has been known that acetylation of histone H4 positively correlated with gene expression, we observed that the acetylation of different lysine residues on histone H4 does not induce a similar shift in −1/+1 positioning, with the exception of H4K16ac showing a H3K36me3/H3K79me3-like positioning (Supplementary Fig. S3B and C).

To confirm the concept that histone H3 acetylated +1 nucleosomes are right-shifted, we reanalyzed MNase-seq data of cells treated with TSA in order to inhibit histone deacetylases and to increase histone H3 acetylation levels [50]. Indeed, inhibition of deacetylation leads to a right shift of the +1 nucleosome and a left shift of the −1 nucleosome (Fig. 2C and Supplementary Fig. S3D).

Histone acetyltransferases directly interact with transcriptional activators to mediate −1/+1 histone acetylations (Fig. 2D). Epl1 is a component of two HATs: NuA4 and Piccolo [54] complexes both feature a HAT module comprising Esa1, Epl1, and Yng2. Notably, only NuA4 contains Tra1, mediating the interaction of this HAT with transcription activators [55]. When the 348 C-terminal amino acids of Epl1 (Epl11–485) are deleted, it disrupts the incorporation of the HAT module into NuA4. As a result, cells expressing Epl11–485 are believed to have Piccolo but lack NuA4 [54]. We took advantage of published MNase-seq data using the epl1-485 mutant [50]. This mutant reveals a left shift of most, if not all, of the +1 nucleosomes while the −1 nucleosomes are right shifted (Fig. 2E and Supplementary Fig. S3E), supporting that Epl1-dependent histone acetylation promotes NDR opening.

Finally, we tested the −1/+1 positioning in steady-state chromatin upon TSA treatment in the absence of Reb1 TF. Our analysis revealed that maintenance of histone acetylation is able to compensate for the lack of Reb1 (Fig. 2F–H and Supplementary Fig. S3F). Thus, Reb1 does not have a nucleosome displacing factor activity by itself but it is more certainly the Reb1-mediated acetylation that properly positions −1/+1 nucleosomes.

The RSC ATPase activity and INO80 are essential for −1/+1 nucleosome repositioning after replication

RSC and INO80 are the two main CRs involved in NDR setting in vitro [19]. Moreover, in vivo, RSC has been described as a pusher while INO80 is acting as a puller [23]. These two CRs are targeting most, if not all, promoter NDRs in S. cerevisiae. Indeed, RSC depletion via nuclear depletion of the ATPase catalytic subunit Sth1 leads to a shrinkage of promoter NDRs over the whole genome (Supplementary Fig. S4A). Contrarily, nuclear depletion of Ino80 carrying the ATPase activity of INO80 globally induces an expansion of the promoter NDRs (Supplementary Fig. S4B).

Analysis of the +1 position on nascent chromatin in the absence of Sth1 revealed that, contrary to GRFs, the left shift is already visible on the newly synthesized chromatin (Fig. 3A and B, and Supplementary Fig. S4E and G). Moreover, the right shift in the absence of Ino80 is also detected at the level of nascent chromatin (Fig. 3C and D, and Supplementary Fig. S4F and G). The phenotype of the depletion of these two CRs appears shortly after replication indicating that they act upstream of TFs for promoter NDR resetting. Symmetrical results were obtained with the −1 nucleosome (Supplementary Fig. S4C and D).

One proposed model for RSC recruitment to promoter NDRs is through the recognition of a CGSG motif on the DNA via the Rsc3/Rsc30 DNA interaction module [56, 57]. Surprisingly, depletion of the essential Rsc3 subunit of the RSC complex leads to the shrinkage of only a fraction of promoter NDRs as compared to the Sth1 nuclear depletion (Supplementary Fig. S5B), allowing us to define ‘target’ and ‘nontarget’ genes in the absence of Rsc3 (Supplementary Fig. S5A). With respect to the chromatin maturation after replication, depletion of Rsc3 does not change the +1 or −1 positioning as compared to the −Rap condition (Fig. 3E and F and Supplementary Fig. S5C–E). This conveys that the Rsc3 DNA binding module is not required for the reestablishment of the +1 nucleosome after replication but is essential for its long-term correct placement as observed in the case of TFs depletion. In other words, the RSC ATPase and DNA-binding modules clearly show different activities regarding NDR formation and maintenance, respectively.

Finally, TSA treatment did not change the effect of Ino80 and Sth1 depletion on steady-state chromatin implying that CRs are acting upstream of histone acetylation (Fig. 3G and H).

RSC ATPase activity is required for INO80 recruitment and puller activity

Although our data indicate that RSC and INO80 CRs are necessary for promoter NDR formation after replication, while TFs participate in a second step to their maintenance, whether a temporal hierarchy exists between RSC and INO80 remains to be clarified. To address this question, we performed the analysis of the +1 positioning on newly replicated chromatin following nuclear depletion of both Sth1 and Ino80. In this case, simultaneous depletion of both factors led to the shrinkage of most of the promoter NDRs on steady-state chromatin as already observed for Sth1 depletion alone, indicating that Ino80 depletion is epistatic to Sth1 depletion (Supplementary Fig. S6A) [23]. Thus, at steady-state, RSC can be assigned as working upstream of INO80. Read-out of the +1 nucleosome positioning on newly-replicated chromatin in the absence of these two CRs unravels a left shift indicating a dominant effect of the RSC nuclear depletion just following the replication process (Fig. 4A and B, and Supplementary Fig. S6B–D). Consequently, it implies that after replication, RSC is acting first, pushing the −1/+1 nucleosomes, before that INO80 restricts the expansion of NDRs.

To confirm this result, we addressed the interplay between RSC and INO80 recruitment by ChEC-seq [58, 59]. We performed this experiment by assessing the binding of Sth1-ChEC in a Ino80-AA strain and vice versa. First, we considered the recruitment of Sth1 and Ino80 at steady state. As already described, Sth1-ChEC reveals a bipartite recruitment being directly in contact with the DNA at the level of the NDR center but also in contact with the neighboring −1/+1 nucleosomes (Fig. 4C). This is in good agreement with the structural data proposing a similar conformational arrangement [56]. The profile of Ino80-ChEC shows a slightly different recruitment pattern being enriched on a broader region of the NDR and surrounding the sharp RSC/DNA contact peak (Fig. 4C). We then tackled the interdependency of RSC and INO80 (Fig. 4D and Supplementary Fig. S6E). The ChEC-seq data revealed that Ino80 binding decreases for the whole set of promoter NDRs in the absence of Sth1 while the absence of Ino80 leads to a spreading of Sth1 binding around its initial peak, consistent with the view that RSC is acting upstream of INO80. Moreover, it also adds that INO80 restricts RSC positioning over NDRs. Thus, despite the need of RSC for initial recruitment of INO80, the two CRs are then entering into a “tug-of-war” as already proposed [60].

RSC is known to interact with acetylated nucleosomes in vitro through its multiple bromodomains and its in vivo recruitment at promoters is influenced by the levels of histone acetylation occurring at the surrounding −1/+1 nucleosomes [43, 58, 59]. Therefore, we tested the effect of TSA treatment on RSC and INO80 binding at promoter NDRs at steady state. Interestingly, maintenance of acetylation leads to the spreading of RSC binding being now enriched around the NDR center. This is accompanied by a shift of INO80, keeping its intercalated position between RSC contact on the DNA and the −1/+1 nucleosomes (Fig. 4E and Supplementary Fig. S6F). It acknowledges a scenario in which maintenance of histone acetylation via TSA treatment increases the interaction with the RSC complex, hence stimulating its displacement on NDRs and setting INO80 signal accordingly.

Steady-state maturation of chromatin recapitulates the interplay between CRs and histone H3 acetylation

Our data support a scenario in which H3 acetylated nucleosomes tend to recruit RSC leading to the expansion of the NDR. Conversely, nucleosomes that are deacetylated might be better substrates for INO80.

To further test this hypothesis, we isolated extreme cases of differences in positioning between +1 peaks in nascent condition as compared to the steady-state one. We clustered +1 nucleosomes in two groups: the highly shifted cluster, corresponding to a right-to-left shift during maturation (>30bp), and the nonshifted cluster (Fig. 5A). When treating the cells with TSA, the highly shifted +1 nucleosome is maintained at a position similar to the nascent chromatin condition (Fig. 5A). This strongly suggests that histone acetylation reduces the pulling activity of INO80. Thus, at steady state, the +1 nucleosome of the highly shifted cluster might interact more with INO80 and become relatively deacetylated while the nonshifted cluster might be a substrate of RSC, enriched in acetylated histone H3 nucleosomes (Fig. 5B). We tested this hypothesis using the published dataset from the Friedman lab [51]. Indeed, the highly shifted cluster shows significantly lower levels of H3K9ac, H3K14ac, and H3K18ac at the +1 nucleosome but not for H3K23ac or H3K27ac (Fig. 5C). Interestingly, the most significant lower levels are observed for the H3K14ac which has been shown to be the main target of Sth1 bromodomain [61]. Moreover, the shifted cluster is significantly enriched in INO80 binding (Fig. 5D and E). Altogether, these observations are consistent with our model.

Interestingly, the highly shifted cluster is enriched in genes coding for ribosomal proteins (Fig. 5F, 4.7-fold over-enrichment, hypergeometric P-value = 1.9 × 10⁻¹³) which are already known to be INO80-dependent in terms of expression [62]. This may be due to their unique dynamics, characterized by low histone turnover alongside an active acetylation–deacetylation cycle [63, 64].

RNAPII is not involved in promoter NDR resetting after replication

Recent results in mESCs revealed a contribution of RNAPII in resetting the accessibility of promoter NDRs after replication [36, 65]. In S. cerevisiae, a thermosensitive mutant of Rpb1, the catalytic subunit of RNAPII, affects the −1/+1 positioning while Rpb1 anchor-away does not reveal any striking phenotype [24, 66].

To address these discrepancies, we performed the nascent nucleosome positioning assay in the presence or absence of Rpb1 (Supplementary Fig. S7A). We made a distinction between the first and fifth quantiles of genes using their expression levels as a proxy to unveil their intrinsic dependence on RNAPII (Fig. 6A and B). Neither for highly nor lowly expressed genes, RNAPII appears to be important for −1/+1 positioning after replication (Fig. 6B). However, we could notice interesting differences between highly and lowly expressed genes when considering levels of +1 nucleosomes. The nascent +1 nucleosome levels of highly expressed genes are higher than those of steady-state chromatin in the absence of Rap. Thus, transcription restart might induce a partial depletion of +1 nucleosomes for highly expressed genes. If it is the case, +1 levels between nascent and steady-state chromatin in the absence of RNAPII for highly expressed genes should be equivalent, which appears to be the case. Such an effect is not observed for lowly expressed genes, pointing out the importance of transcription in +1 nucleosome depletion. Altogether, our data strongly suggest that the +1 nucleosome is first positioned after replication before being partially depleted by the transcription process implying that the transcription did not resume during the 2 min of EdU pulse. Therefore, while RNAPII may influence the steady-state levels of the +1 nucleosomes, the transcription process does not appear as being essential for NDR resetting post-replication in S. cerevisiae. However, data at the +4 nucleosome indicate a long-term role of transcription in maintaining proper nucleosome organization of transcription units (Supplementary Fig. S7B).

RSC ATPase activity as the initiator of promoter NDR resetting post-replication

How is RSC specifically targeted and/or acting at promoters? It is proposed that the sub-stochiometric subunits Rsc3 and Rsc30 constitute the DNA binding module recognizing a CGCG motif widely contained in yeast promoters [56, 57, 67]. However, our results show that Rsc3 depletion, contrary to Sth1 depletion, does not impact −1/+1 positioning immediately after replication and only affects a subset of gene promoters on steady-state chromatin (Fig. 3). Regarding its behavior in our assays, Rsc3 depletion phenotypes are more related to depletion of Reb1 and Abf1; Rsc3 could therefore be considered as a TF as already proposed [25, 57].

In another scenario, RSC might be recruited by newly-deposited acetylated histones. Indeed, RSC interacts with acetylated histones in vitro through its multiple bromodomains [68, 69]. Our results show that TSA treatment favors RSC pushing activity at −1/+1 nucleosomes, leading to its symmetric progression over the promoter from its nucleation site (Fig. 4). Thus, it could be appealing to propose that specific histone H3 acetylation at −1/+1 nucleosomes might trap the RSC complex post-replication. Indeed, results obtained in yeast cells indicate that freshly deposited nucleosomes tend to be acetylated [70]. However, they are not promoter specific. Thus, the acetylation status might not be enough to explain this specific action at promoters. One specific feature of promoter NDRs are the poly-dA/dT stretches that induce an intrinsic depletion of nucleosomes and that are known to stimulate and orient RSC activity [13, 14, 19, 21]. Thus, the combination of AT-rich-dependent depletion of nucleosomes together with histone H3 acetylation might preferentially activate RSC at promoters and not in genes bodies.

Finally, a recent report revealed a physical interaction between RSC and the Mediator complex, working as a co-activator at promoters [71]. Whether Mediator is involved in NDR resetting after replication remains to be investigated.

The pulling activity of INO80 follows RSC and restrains its impact

Our results clearly show that INO80 activity and recruitment are subsequent to RSC binding and pusher action (Fig. 4). Contrary to RSC, INO80 activity on chromatin is largely nonspecific and is linked with longer NDRs [26]. Thus, one tempting scenario is that RSC would first push the nucleosome leaving enough linker DNA for INO80 to bind and pull the nucleosome. In agreement with such a proposal, INO80 binding is different from the RSC profile, surrounding the nucleation site of RSC. Moreover, when treating with TSA, INO80 binding is displaced in accordance with RSC binding (Fig. 4).

Another question is whether RSC and INO80 bind at the same time? This seems quite unlikely. Indeed, our results indicate that RSC activity depends on acetylated −1/+1 nucleosomes while INO80 functions more with deacetylated nucleosomes. This goes well with the “tug-of-war” model in which CRs are competing constantly to maintain NDRs [60].

Recent studies report that hexasomes are the preferred substrate of INO80 in vitro [72, 73]. Further studies will be needed to integrate these observations in our model.

In this study, we did not cover other CRs such as SWI/SNF or ISW2 described as acting at promoters since, unlike RSC and INO80, they only target a subpopulation of genes [23, 26].

Reb1 and Abf1 are not acting as PFs after replication

TFs/NDFs are usually considered as equivalent to PFs in multicellular eukaryotes [25, 26, 74]. Indeed, Reb1 and Abf1 are involved in the maintenance of promoter NDRs in vivo and Reb1 interacts in vitro with its specific binding motif located within a nucleosome [14, 28, 29, 32]. The PF model fitted to re-organization of the chromatin after replication would place Reb1 and Abf1 as the first players in NDR resetting, which is not supported by our data thereby questioning the in vivo concept of PFs. Steady-state NDRs represent the average state of a balance between opening and closing [75, 76]. Thus, from a steady-state observation, it is impossible to extract the first occurring step. Our assay allows to follow the sequence of events and to clearly state that Reb1 and Abf1 are not acting as PFs for NDR resetting after replication.

Interestingly, TSA treatment rescues a proper positioning of −1/+1 nucleosomes in the absence of Reb1 implying that it is not the TF per se that maintains NDRs open, but rather their −1/+1 acetylation states (Fig. 2). Therefore, in this case, promoter NDR architecture appears to be independent of the presence of TFs. It might explain the apparent high percentage of genes in which TFs are not stably detected at yeast promoters [77]. However, our experiments did not address the transcription status in this context. Although numerous TFs show direct interactions with the transcription machinery [78], it is tempting to propose that a subset of TFs might set promoter architecture by promoting histone acetylation, which represents the keystone.

Based on this model, why would co-depletion of TFs and CRs be additive [14]? We hypothesize that the absence of TF tethering to promoters might leave enough linker DNA for INO80 to pull −1/+1 nucleosomes. Consistently, expression of the DNA-binding domain of Rap1 TF partially rescues the −1/+1 positioning [47]. Therefore, we propose that TFs have a dual function acting both as barriers limiting INO80 action and as activators of RSC pushing function by promoting histone acetylation of the flanking nucleosomes.

Although we did not detect PF activity, we only tested a limited set of TFs. Therefore, we cannot exclude the possibility that such activity may be TF-specific. Interestingly, physical interactions between RSC or its mammalian counterpart and factors directly involved in transcription, such as Hmo1 and β-catenin, have already been reported [79, 80].

Histone H3 acetylation helps the opening of promoter NDRs via −1/+1 nucleosomes shifts

Histones are heavily post-translationally modified leading to the hypothesis of the “histone code” [3]. While some modifications are associated with open or closed chromatin, their impact on nucleosome positioning has been poorly assessed. By reanalyzing high-resolution maps of histone modifications [51], we clearly establish that H3 acetylated nucleosomes correspond to a subpopulation of −1/+1 nucleosomes that is shifted relative to the average and associated with an expansion of the promoter NDRs (Fig. 2). In agreement with this concept, maintenance or decrease of histone acetylation shift the whole population of −1/+1 nucleosomes (Fig. 2).

Activation of genes during the metabolic cycle is associated with an expansion of promoter NDRs [81]. Strikingly, maximal expression correlates with a peak of H3K9ac and H3K18ac on +1 nucleosomes [82]. Moreover, exit from quiescence is associated with an opening of the promoter NDR as transcription resumes [83]. Altogether, it supports well our data proposing a link between H3 acetylation, chromatin opening and gene expression.

Surprisingly, acetylation of different lysine residues on histone H4 does not induce a similar shift in −1/+1 positioning despite their positive correlation with gene expression (Supplementary Fig. S3C) [51]. Thus, histone H4 acetylations stimulate gene expression independent of NDR expansion, probably by weakening the interaction between DNA and histones as already proposed [84].

This study, as well as our previous work, establish that H3K36me3, H3K79me3 and H4K16ac correlate with a decreased NDR width [42] (Supplementary Fig. S3B and C). Moreover, induction of an increase of H3K36me3 at −1/+1 nucleosomes reduces NDR length [42]. Altogether, these observations stress the importance of histone modifications in setting proper nucleosome positioning at NDRs required for optimal gene expression.

Extension of the model to multicellular eukaryotes

Within the SWI/SNF family, yeast RSC is related to the BAF complexes in mammals with PBAF being the closest one [85]. All of the BAF complexes are assembled around an ATPase, being BRG1 or BRM. Very recently, a study in mESCs has shown that nearly all of the promoter NDRs, as well as enhancers, rapidly lose their accessibility when treated with an inhibitor of BRG1/BRM [86]. Importantly, this accessibility loss can be compensated with the time by the EP400/TIP60 HAT, pointing out, as in our yeast model, a probable central role of histone acetylation in promoter opening. On the other hand, in mESCs, the INO80 complex colocalizes with the TATA-binding protein contained in the transcription pre-initiation complex marking most of the active promoters [87]. Thus, the crucial CRs of our model also vastly participate in promoter NDRs regulation in multicellular eukaryotes. However, their interplay with TFs and especially with PFs at promoters and enhancers remains obscure due to a lack in temporal resolution. Indeed, the two-step mechanism we propose, NDR formation and maintenance, which is involving the same players, cannot be inferred from steady state analyses. Short pulses of chromatin labeling coupled with acute depletion of TFs/CRs might reveal if this model can be extended to mammalian cells.