Cohesin facilitates nucleosome invasion by transcription factors

Abstract

Nucleosomes present a major barrier to transcription factor (TF) binding. However, a subgroup of TFs known as pioneer factors (PFs) can recognize motifs covered by nucleosomes and initiate chromatin opening. PFs also bind nucleosomal substrates with high affinity in vitro, which may facilitate their nucleosome invasion in vivo. Here, we show that LexA, a bacterial TF with poor nucleosome binding in vitro, can rapidly invade into a well-positioned nucleosome from a motif positioned at the dyad when expressed ectopically in yeast. This notable contrast between LexA binding in vitro and in vivo raises the possibility that TFs can exploit nucleosome dynamics in vivo to access occluded sites. Unexpectedly, we find that LexA-mediated chromatin opening can occur in the absence of DNA replication, chromatin remodeling, histone turnover, and a few histone chaperones. Instead, nucleosome invasion by LexA and a native PF, Cbf1, is promoted by the cohesin complex, illustrating an intriguing connection between cohesin and TF binding. Together, our results demonstrate that even nonpioneer TFs like LexA can bind and displace nucleosomes in vivo, through a process facilitated by cohesin.

Nonpioneer transcription factor can breach nucleosomes in vivo, and this process is facilitated by cohesin but not remodelers.

INTRODUCTION

Transcription factors (TFs) play a central role in orchestrating gene expression programs. Understanding the factors that promote or inhibit TF binding is therefore critical in elucidating the mechanisms of gene regulation. TFs typically function by recognizing specific DNA sequence motifs, which are often embedded within nucleosomes in eukaryotic genomes. Because histone proteins make extensive contacts with DNA, nucleosome-embedded motifs are less accessible, significantly restricting the ability of most TFs to engage with their target sites and regulate transcription (1–3). Over the past few decades, a specialized class of TFs called pioneer factors (PFs) has emerged. These PFs have the unique ability to engage with nucleosomal substrates, increase chromatin accessibility, and facilitate the binding of downstream TFs that would otherwise be excluded by the nucleosome barrier (4–6). In higher eukaryotes, PFs are often regarded as “master regulators” for differentiation and cell fate commitment. For example, OCT4 and SOX2 are essential in maintaining embryonic stem cell identity, and FOXA1/2 are required for proper endoderm and hepatic development and differentiation (7, 8).

Despite extensive research, identifying the biochemical properties of TFs that confer pioneering activity remains a topic of debate (9, 10). Many biochemical studies of purified PFs, such as Reb1, Cbf1 in yeast, FOXA1, OCT4, SOX2 in mammals, and Zelda in Drosophila, have shown that these factors exhibit similar binding affinities for naked and nucleosomal DNA (4–6, 11, 12). Some PFs achieve this using a “dissociation rate compensation” mechanism, in which they dissociate more slowly from nucleosomal DNA through interactions with histones to enhance their binding affinity toward nucleosomal substrates (11, 13, 14). Inside cells, PFs display slow nuclear mobility and the ability to scan and access closed chromatin (15). Together, these studies support a model where PFs have special intrinsic biochemical properties that allow them to effectively engage nucleosomes, which eventually leads to nucleosome displacement and/or histone eviction.

In contrast, other evidence suggests that these intrinsic nucleosome-binding properties are not strictly necessary for pioneering activity. Nucleosome displacement results from TFs and histones competing for the same genomic locus, and this competition can be modulated by the TF concentration. For example, a yeast TF Pho4 binds nucleosomes poorly and lacks dissociation rate compensation in vitro, yet it can displace nucleosomes when overexpressed in yeast (13). Rfx1, another yeast TF with weak pioneering activity, can be converted into a strong PF at elevated expression levels (16). In mammalian cells, HNF4A is typically classified as a non-PF because it binds downstream of FOXA1 during hepatocyte differentiation and preferentially targets open, deoxyribonuclease-hypersensitive sites (15, 17). However, when overexpressed in human cells, HNF4A can bind and open previously inaccessible chromatin regions (18, 19). These observations indicate that sequence-specific TFs with high concentration may be sufficient to confer pioneering activity, raising the possibility that any TF can function as a PF in certain conditions.

One caveat of these in vivo experiments is that TFs are assessed in their endogenous cellular context, where their pioneering activities can potentially be affected by interactions with other TFs and/or coregulators. Moreover, studies in mammalian cells typically rely on ATAC-seq to infer chromatin accessibility, a method that lacks single-nucleosome resolution. Even when nucleosome-level resolution is achieved with micrococcal nuclease (MNase) mapping, most nucleosomes in mammalian genomes adopt fuzzy positioning (20), making it difficult to evaluate the exact location of TF motifs within the nucleosome.

To address these limitations, we used a minimal system by expressing the Escherichia coli TF LexA in budding yeast to investigate its nucleosome-displacing activity. LexA is a small transcriptional repressor (∼23 kDa) that binds as a dimer to a palindromic consensus motif (LexO) with high affinity [dissociation constant (K_d) ∼ 1 nM] (3). It is composed of an N-terminal helix-turn-helix DNA binding domain along with a C-terminal dimerization domain. Structural studies of the LexA-DNA complex reveal that LexA binding induces DNA bending toward LexA, a configuration that is incompatible with nucleosomes (21). Consistent with this idea, LexA was shown to bind poorly to nucleosomes (22). When the LexO is placed at 8 bp from the entry side of a Widom 601 nucleosome, the LexA binding rate is reduced by two orders of magnitude, while its dissociation rate is increased up to three orders of magnitude, resulting in markedly lowered binding affinity compared to naked DNA (3). As a bacterial TF, it is unlikely for LexA to have specific coordinated interactions with euchromatic chromatin regulators. Therefore, using this artificial system, we can ask whether a TF with strong DNA binding but weak nucleosome binding can access its nucleosome-embedded motif in vivo, and if yes, which cellular processes promote this activity.

Here, we show that LexA can efficiently and rapidly invade into nucleosomes and generate a local nucleosome-depleted region (NDR) in yeast, even when its motif is positioned at the dyad of a well-positioned nucleosome. As this process likely relies on nucleosome dynamics in vivo, we systematically measured LexA-mediated nucleosome invasion in the absence of factors typically associated with nucleosome dynamics. Unexpectedly, we find that chromatin opening by LexA occurs independently of DNA replication, chromatin remodeling, and a few histone chaperones. Instead, this process is inhibited by histone turnover and facilitated by the cohesin complex. Cohesin also promotes nucleosome invasion and chromatin binding by endogenous yeast PF Cbf1. These findings reveal an unanticipated connection between cohesin, nucleosome dynamics, chromatin accessibility, and TF binding.

RESULTS

LexA binds weakly to nucleosomes in vitro but invades into nucleosomes rapidly in vivo

Previous studies of LexA-nucleosome interactions used the Widom 601 nucleosome positioning element (3). Nucleosomes assembled on 601 are highly stable and may pose an unusually strong barrier for LexA binding. To determine whether LexA’s nucleosome binding property also applies to naturally occurring nucleosomes, we examined LexA binding to a yeast native nucleosome, YNN1, which is significantly less stable than 601 (23). On the basis of previously determined nucleosome positioning over YNN1 (23), we placed LexO at the dyad (position 64 to 83 bp) and measured LexA binding using electrophoretic mobility shift assay (EMSA). Consistent with previous reports (3), LexA shows strong binding to LexO on naked DNA with a dissociation constant K_D < 5 nM (Fig. 1, A and B). However, when the same DNA is reconstituted into nucleosomes, no detectable band shift occurs until LexA concentration reaches 300 nM (Fig. 1, A and B). The same EMSA pattern occurs on nucleosomes lacking LexO sites, indicating that the band shift is due to nonspecific binding. These results confirm that LexA exhibits poor nucleosome binding in vitro, even when using a biologically relevant, less stable nucleosome substrate.

Fig. 1. LexA binds poorly to nucleosomes in vitro but invades rapidly into nucleosomes in vivo.

(A) EMSA of purified LexA binding to naked or nucleosomal YNN1 DNA containing a LexO motif at the nucleosome dyad (left) or lacking LexO (right). Naked DNA (0.5 nM; top) or nucleosomes (0.5 nM; bottom) were incubated with increasing amounts of LexA. The asterisk marks LexA bound to free DNA. (B) Quantification of EMSA results in (A). Intensities of unbound DNA (top) and nucleosome (bottom) bands are shown. Error bars indicate the SE from three biological replicates. LexO did not alter the nucleosome gel shift pattern, indicating a lack of motif-specific binding. (C) Schematic of the construct used to induce LexA expression. (D) LexA invades nucleosome −4 at a modified HOpr through a LexO motif positioned at the dyad. Top: promoter with the engineered LexO site. Bottom: nucleosome occupancy measured by MNase-qPCR before (gray) and after (blue) LexA induction and at a control HOpr lacking LexO after induction (green). The orange dashed line marks the LexO site. Two biological replicates were analyzed. (E) Time course of LexA invasion into nucleosome −4 during induction. Darker colors indicate longer induction times. Three biological replicates were analyzed. (F) Quantification of the change in nucleosome occupancy at the LexO site from (E). (G) LexA ChIP-qPCR at nucleosome −4 during induction on LexO-containing (blue) and LexO-lacking (gray) templates. Two biological replicates were analyzed. (H) Quantification of LexA induction levels by anti-HA Western blot. LexA-HA abundance was estimated by comparison to HA-tagged endogenous Reb1, Cbf1, and Ser2 (see Materials and Methods). Two biological replicates were analyzed. ORF, open reading frame. bp, base pair.

Next, we assessed LexA binding and nucleosome invasion in vivo by inducing LexA expression with 17β-estradiol in a Saccharomyces cerevisiae strain (Fig. 1C) (24). This strain also harbors a single LexO site engineered into a well-positioned nucleosome −4 within a modified HO promoter (HOpr) (Fig. 1D). The Swi5 binding sites in this HOpr are mutated to maintain a constitutive nucleosome array across the cell cycle (16, 25). In the absence of LexA, MNase mapping confirms that LexO resides at the dyad of nucleosome −4 (Fig. 1D and fig. S1A). After LexA induction, nucleosome occupancy near LexO undergoes a marked change: nucleosome −4 dyad is converted into a short linker, flanked by two nucleosomes that become completely out of phase with the original array (Fig. 1D). The same LexA-induced nucleosome repositioning can be detected by MNase sequencing (MNase-seq; fig. S1B). This effect is dependent on LexO, as LexA induction does not alter nucleosome positioning on the HOpr background (Fig. 1D). These results indicate that LexA can invade into a nucleosome via a single motif at the dyad and displace adjacent nucleosomes in vivo.

To measure the dynamics of LexA invasion, we carried out MNase measurements over HOpr following LexA induction at different time points (Fig. 1E). Nucleosome −4 occupancy is altered within 15 min of LexA induction, and its depletion is nearly complete by 30 min (Fig. 1F). We further monitored LexA binding over time by chromatin immunoprecipitation (ChIP), which detected significant LexA binding at 15 min of LexA induction (Fig. 1G), consistent with the rapid depletion of nucleosome −4. To test whether this nucleosome invasion activity is a result of exceptionally high LexA concentration in cells, we used Western blotting to quantify the concentration of hemagglutinin (HA)–tagged LexA throughout induction. By comparing the band intensities of LexA-HA to three HA-tagged endogenous proteins of known concentrations, we estimated LexA level to be ∼30 nM at 15 min and ∼100 nM at 30 min (Materials and Methods; Fig. 1H and fig. S1, C and D). Notably, LexA does not show nucleosome binding in vitro at these concentrations (Fig. 1A).

Given this apparent contradiction, we asked whether the rotational setting of the LexO in Fig. 1D happens to promote LexA invasion into nucleosome −4. To test this, we repositioned LexO by shifting it 3 bp from its original location (p64 to p67), which will significantly alter its rotational orientation on the nucleosome −4. Similar nucleosome displacement was observed regardless of the LexO positioning (fig. S1, E and F), indicating that the rotational setting has minimal impact on LexA invasion. Collectively, these findings demonstrate that LexA binds poorly to nucleosomes in vitro but can efficiently invade nucleosomes in vivo.

Depletion of chromatin remodelers does not affect the rate of nucleosome invasion by LexA

We hypothesized that the discrepancy between LexA’s behavior in vitro and in vivo may be due to differences in nucleosome dynamics. Under physiologically relevant salt conditions, purified nucleosomes are relatively stable, exhibiting limited unwrapping or repositioning (26, 27). In contrast, nucleosomes in living cells are subject to constant perturbation by various regulatory factors. In particular, chromatin remodelers (CRs) use energy derived from adenosine triphosphate (ATP) hydrolysis to translocate nucleosomes or evict histones from DNA (28). This raises the possibility that CR-dependent nucleosome dynamics may transiently expose otherwise occluded motifs and facilitate TF binding (Fig. 2A).

Fig. 2. The rate of LexA invasion is not affected by depletion of CRs.

(A) Cartoon illustrating how the ATP-dependent remodeling activity of CRs could facilitate nucleosome displacement and promote LexA binding. (B) Sth1 anchor-away. Fluorescent images of yeast cells expressing Sth1-GFP before (left) and after (right) rapamycin treatment confirm nuclear depletion of Sth1. (C) Time course of LexA invasion into nucleosome −4 in the presence (top) or absence of Sth1 (bottom). (D) Quantification of nucleosome occupancy at the LexO site during LexA induction with or without Sth1. (E) Simultaneous depletion of multiple CRs. Fluorescence imaging confirms concurrent anchor-away of Sth1-GFP, Isw2-GFP, and Chd1-GFP, and a Western blot shows the auxin-induced degradation of Ino80 and Isw1. Actin serves as a loading control. LexA is induced to similar levels regardless of CR depletion. (F) Time course of LexA invasion into nucleosome −4 in the presence (top) or absence of multiple CRs (bottom). (G) Quantification of nucleosome occupancy at the LexO site during LexA induction with or without multiple CRs. bp, base pair.

To determine whether nucleosome invasion by LexA depends on CR activity, we first examined the role of RSC, the most abundant CR in budding yeast and the only CR essential for cell viability (29). Depletion of the adenosine triphosphatase (ATPase) subunit of RSC, Sth1, results in genome-wide shrinkage of NDRs, indicating that RSC broadly promotes chromatin accessibility (30–32) and could, in principle, facilitate LexA invasion. We generated an anchor-away strain to rapidly deplete green fluorescent protein (GFP)–tagged Sth1. Fluorescent imaging confirms efficient removal of Sth1 from the nucleus following rapamycin treatment (Fig. 2B). We then assessed nucleosome occupancy during LexA induction with and without Sth1. Sth1 depletion results in fuzzy nucleosome positioning over the HOpr (Fig. 2C, −Sth1, 0 min), consistent with altered remodeling activity. However, LexA-mediated nucleosome displacement remains unchanged, with comparable levels of nucleosome depletion at 15, 30, and 60 min of LexA induction in the presence or absence of Sth1 (Fig. 2, C and D). Similar reduction in nucleosome occupancy at the LexO site can be detected by MNase-seq (fig. S1G). These findings indicate that RSC is not required for LexA invasion into nucleosomes.

Since different CRs may function redundantly, we constructed another strain to simultaneously remove five major remodelers (Sth1, Isw1, Isw2, Chd1, and Ino80) through a combination of anchor-away and auxin-induced degradation (Fig. 2E) (31). Removal of these CRs has no impact on LexA induction levels (Fig. 2E). Similar to Sth1 depletion alone, the elimination of all five CRs leads to fuzzy nucleosome positioning over the HOpr (Fig. 2F). Despite the loss of most of the remodeling activities in the cell, LexA still efficiently displaces nucleosomes near the LexO site with no detectable changes in the rate of nucleosome invasion (Fig. 2, F and G). Together, these results illustrate that LexA-induced nucleosome displacement occurs independently of most ATP-dependent CRs.

Depletion of histone chaperones does not affect the rate of nucleosome invasion by LexA

Another group of proteins that is associated with nucleosome dynamics is histone chaperones, which bind histones and facilitate nucleosome assembly and disassembly independently of ATP hydrolysis (33). In particular, we focused on FACT, an essential complex known to bind nucleosomes and induce major conformational changes (34). More specifically, FACT is involved in histone “recycling,” where DNA is temporarily unwrapped from histones and then reassembled (Fig. 3A) (35). This process has the potential to expose underlying DNA and contribute to nucleosome displacement. FACT is also highly abundant in yeast, enabling its widespread function in transcription, replication, and DNA repair (36). On the basis of this prior knowledge, we reasoned that FACT could be involved in facilitating LexA-mediated nucleosome displacement.

Fig. 3. Nucleosome invasion by LexA is not affected by depletion of histone chaperones.

(A) Cartoon illustrating two nucleosome dynamic processes mediated by histone chaperones, histone recycling, and nucleosome assembly/disassembly, both of which could enhance the accessibility of nucleosomal DNA. (B) Time course of LexA invasion into nucleosome −4 in WT (blue), −pob3 (red), or −spt16 (yellow) strains. Pob3 and Spt16 were depleted through temperature-sensitive mutations by shifting cells to the nonpermissive temperature of 37°C for 45 min before LexA induction. (C) Quantification of nucleosome occupancy change over the LexO site for time course in (B). (D) Time course of LexA invasion into nucleosome −4 with simultaneous depletion of Pob3, Asf1, and Nap1. The three proteins were depleted through temperature-sensitive mutation (Pob3), anchor-away (Asf1), and auxin-induced degradation (Nap1). (E) Quantification of nucleosome occupancy change over the LexO site for time course in (D). bp, base pair.

The yeast FACT complex consists of two core subunits, Spt16 and Pob3. To probe the relationship between FACT and LexA invasion, we used two well-characterized temperature-sensitive mutants, spt16-11 and pob3L78R (37, 38). Shifting these mutant strains from the permissive (22°C) to the nonpermissive temperature (37°C) causes rapid FACT destabilization and degradation (37, 38), ultimately resulting in cell death (fig. S2A). We monitored HOpr nucleosome occupancy following LexA induction in wild-type (WT) and FACT mutant strains at 37°C. At this elevated temperature, LexA induction occurs faster than at 30°C, leading to more rapid LexA invasion in WT cells (nucleosome −4 is almost completely displaced by 15 min) (Fig. 3B and fig. S2, B and C). Both Spt16 and Pob3 mutant cells show comparable levels of LexA induction and similar rates of nucleosome −4 displacement (Fig. 3C and fig. S2, B and C). The LexA-mediated nucleosome displacement in the absence of Pob3 can be reproduced in MNase-seq data (fig. S2D).

Given the potential functional redundancy among histone chaperones, we next investigated whether other chaperones compensate for FACT depletion, thereby masking its effect on LexA invasion. To test this, we generated a compound mutant in which the Pob3 mutation was combined with anchor-away depletion of Asf1 (an H3/H4 chaperone) and auxin-induced degradation of Nap1 (an H2A/H2B chaperone) (fig. S2, E and F). Even in the absence of all three chaperones, LexA can still rapidly and efficiently displace nucleosome −4 (Fig. 3, D and E). This suggests that LexA invasion occurs independently of FACT and other major histone chaperones.

Rapid replication-independent histone turnover inhibits LexA invasion into nucleosomes

We next tested another dynamic process called histone turnover, where old histones are replaced with new ones. While major histone turnover occurs during DNA replication, it can also take place outside of the S phase through replication-independent mechanisms (39, 40). The impact of cell cycle and DNA replication on LexA invasion will be discussed in the next section; here, we focus on replication-independent histone turnover. Since LexO is engineered at the nucleosome dyad, which predominantly interacts with H3/H4 tetramer, we specifically examined how H3/H4 turnover affects LexA invasion. We hypothesized that higher histone turnover would enhance DNA accessibility at the dyad, thereby promoting LexA invasion (Fig. 4A).

Fig. 4. Nucleosome invasion by LexA is inhibited by fast histone turnover.

(A) Cartoon illustrating how fast histone turnover could promote TF binding and nucleosome invasion. (B and C) Time course of nucleosome occupancy near LexO sites engineered into two fast turnover nucleosomes (B) or two low turnover nucleosomes (C). The LexO motifs are positioned near the dyad of these nucleosomes and are marked by the dashed lines. (D) Same as the left in (B), except Cbf1 (full length or DBD alone) is induced in a CBF1 deletion background, with a Cbf1 motif engineered at the same site in the fast turnover nucleosome. Both the Cbf1 full-length protein and DBD are able to displace the fast turnover nucleosome. (E) Same as the left in (B), except LexA is fused at its C terminus to the VP16 activation domain. Fusion with VP16 results in nucleosome displacement not observed with LexA alone. bp, base pair.

To test this idea, we first examined previously published H3 turnover data in G₁-arrested cells (41). In this dataset, the HOpr exhibits a relatively low level of histone turnover. However, the turnover dynamics at our modified HOpr construct, which is inserted into a nonnative locus, remain unclear. We therefore selected four well-positioned nucleosomes, two with high turnover rates (YIL163C and HXK1) and two with low turnover rates (YEL073C and YGL159W) (fig. S3, A and B), inserted LexO into the dyads of these nucleosomes at their endogenous loci, and measured nucleosome occupancies following LexA induction. Unexpectedly, both high-turnover nucleosomes are less displaced by LexA, with their occupancies dropping by only ∼35% after 1 hour of LexA induction, compared to an over 60% reduction at the two low turnover nucleosomes (Fig. 4, B and C).

The unexpected repressive effect of high histone turnover on LexA invasion led us to question whether native yeast PFs like Abf1 or Cbf1 can displace these nucleosomes. In the case of Abf1, we inserted its consensus motif into the same high-turnover nucleosome at YIL163C and measured the resulting nucleosome occupancy. Endogenous Abf1 effectively displaces this nucleosome and generates a local NDR (fig. S3C), indicating that it has stronger nucleosome displacement activity than LexA. For Cbf1, given our previous finding that the DNA binding domain (DBD) of Cbf1 confers its high nucleosome affinity (13), we asked whether full-length Cbf1 or the DBD alone can invade into high-turnover nucleosomes. Accordingly, we deleted the endogenous Cbf1 and replaced it with an inducible version encoding either full-length Cbf1 or the DBD alone. Both inducible proteins can invade into the high-turnover nucleosome at YIL163C containing engineered Cbf1 motif, although the NDR formed by the DBD is shorter than the one generated by the full-length protein (Fig. 4D). A similar difference in NDR length was observed at HOpr with an engineered Cbf1 motif (13). These findings suggest that, at least in these two cases, strong nucleosome binding by specialized DBDs enhances TF’s capability to open chromatin, while regions outside of the DBD play a role in modulating NDR length, presumably through cofactor recruitment.

In yeast, the activation domain VP16 is known to recruit cofactors such as SAGA and SWI/SNF and facilitate chromatin destabilization and opening (42). We next asked whether fusing LexA to VP16 could enhance LexA’s ability to invade into high-turnover nucleosomes. The LexA-VP16 fusion protein successfully establishes an NDR at the region of high turnover (Fig. 4E). In summary, these results reveal that rapid histone turnover may present a higher, instead of lower, barrier for TF accessibility, but TF invasion into these regions can be facilitated by strong nucleosome binding via DBD and/or recruitment of coactivators through regulatory domains. Despite these insights, it is still not clear how LexA invades into low-turnover nucleosomes.

Nucleosome invasion is less efficient during G₁ and independent of DNA replication

DNA accessibility may vary across the cell cycle. In particular, DNA replication may transiently expose TF binding motifs behind the replication fork, providing an opportunity for TFs to bind. To investigate the role of cell cycle in LexA-mediated nucleosome displacement, we arrested the cells in G₁ with α-factor (fig. S4A), induced LexA, and monitored nucleosome occupancy. In G₁-arrested cells, nucleosome displacement is significantly slower compared to cycling cells, with the half-life of nucleosome −4 increasing from <30 to ∼60 min (Fig. 5, A and B). LexA expression levels are comparable in G₁ versus cycling cells, indicating that the slowed invasion is not due to less LexA induction (Fig. 5C and fig. S4B).

Fig. 5. Nucleosome invasion is less efficient during G₁, independent of DNA replication.

(A) Time course of LexA invasion into nucleosome −4 in WT cells arrested in G₁. (B) Quantification of nucleosome occupancy at the LexO site during LexA induction in G₁-arrested versus cycling cells. (C) Quantification of LexA levels when induced in G₁-arrested (red) versus cycling (blue) cells. (D) Steady-state nucleosome occupancy near nucleosome −4 following 3 hours of LexA induction with variable β-estradiol concentrations in cycling (top) or G₁-arrested (bottom) cells. (E) Quantification of LexA levels in response to induction with variable β-estradiol concentrations for G₁-arrested (red) and cycling (blue) cells. (F) Relationship between nucleosome occupancy at the LexO site [from (D)] and LexA induction levels [from (E)]. (G) Time course of LexA invasion into nucleosome −4 in WT cells arrested in G₂/M by nocodazole. (H and I) Same as in (B) and (C) but in G₂/M-arrested versus cycling cells. bp, base pair.

Since LexA level is coupled to time during induction, we sought to determine whether the slower nucleosome invasion in G₁ was due to a concentration-dependent effect (i.e., a higher level of LexA is required to invade into nucleosomes in G₁ and thus takes longer to reach an effective concentration) or a time-dependent effect (i.e., LexA can invade at the same concentration, but the invasion process takes longer in G₁). To distinguish between these two possibilities, we induced LexA for 3 hours to different steady-state levels by titrating 17β-estradiol and measured nucleosome occupancy afterward (Fig. 5D). LexA levels were similar between G₁-arrested and cycling cells under identical 17β-estradiol concentrations (Fig. 5E and fig. S4C). Yet, the nucleosome displacement efficiency remains significantly lower in G₁ (Fig. 5D). Combining the data from Fig. 5 (D and E) shows that higher LexA concentration is required to achieve nucleosome invasion in G₁ compared to cycling cells (Fig. 5F).

To test whether the reduced nucleosome invasion in G₁ is due to the absence of DNA replication, we arrested the cells in G₂/M with nocodazole (fig. S4A) and repeated the nucleosome occupancy and LexA expression measurements (Fig. 5, G to I). LexA induction levels in G₂/M-arrested cells are again comparable to cycling cells (Fig. 5I and fig. S4D). However, unlike in G₁, LexA-mediated nucleosome displacement in G₂/M occurs at the same rate as in cycling cells (Fig. 5, G and H). Since DNA replication is absent in both G₁ and G₂/M, but only G₁ showed reduced LexA invasion, these results suggest that the effect is not simply due to the lack of DNA replication. Instead, we suspect that a key factor promoting nucleosome dynamics is down-regulated in G₁, leading to the observed reduction in LexA-mediated invasion.

Cohesin facilitates nucleosome invasion by LexA

The observations above prompted us to identify factors that are transcriptionally down-regulated in G₁ and are also implicated in regulating nucleosome dynamics or chromatin accessibility. Out of the candidates listed in fig. S5A, we first investigated the histone acetyltransferase Rtt109, which acetylates newly synthesized H3 at K56 (43). The H3K56ac modification enhances DNA breathing near the nucleosome entry/exit site (44) and could potentially promote nucleosome dynamics during and after S phase. However, auxin-induced degradation of Rtt109 does not affect nucleosome displacement by LexA in cycling cells (fig. S5, B and C). We also evaluated Top2, an essential topoisomerase known to relieve DNA torsional strains by resolving both positive and negative supercoils, a process that could influence TF invasion into nucleosomes by affecting nucleosome stability (45, 46). We anchored-away Top2 and confirmed that these cells are no longer viable (fig. S5D). Despite this severe phenotype, nucleosome −4 displacement still occurs in the Top2-depleted strain at a rate comparable to control cells (fig. S5E).

We next focused on cohesin (Fig. 6A), a SMC complex previously implicated in promoting TF binding and DNA accessibility in higher eukaryotes (47). All three core cohesin subunits, Smc1, Smc3, and Mcd1, are transcriptionally repressed during G₁ (48), and the complex is also actively removed from G₁ chromosomes by the cohesin-releasing factors (49). To test the cohesin’s role in nucleosome displacement, we depleted an ATPase subunit Smc3 via auxin-induced degradation (fig. S6A) and measured nucleosome occupancy over the HOpr following LexA induction in cycling cells. LexA’s invasion rate is already slower in the Smc3-AID strain even without auxin treatment compared to the WT control, consistent with the observed partial degradation of Smc3 in −auxin condition (fig. S6, A and B). This effect is further enhanced in the presence of auxin, producing a delay comparable to that observed during G₁ arrest (Fig. 6B and fig. S6B). LexA induction is not affected by Smc3 depletion, confirming that the observed effect is not due to differences in LexA expression levels (Fig. 6C and fig. S6C). To further validate these observations, we extended our analysis to other cohesin subunits. Specifically, we targeted Mcd1 (the α-kleisin subunit, known as Rad21 in higher eukaryotes) and Scc2 (the cohesin loader) and repeated the nucleosome occupancy assay. Mcd1 and Scc2 depletion also result in slower LexA-induced nucleosome displacement, although the effects are less pronounced compared to Smc3 depletion (Fig. 6D).

Fig. 6. Cohesin facilitates nucleosome invasion by LexA.

(A) Cartoon depiction of the cohesin complex. (B) Time course of LexA invasion into nucleosome −4 in cells depleted of Smc3, Mcd1, or Scc2. (C) Quantification of LexA expression levels following induction in WT versus Smc3-depleted cells. (D) Quantification of nucleosome occupancy at the LexO site in WT versus Smc3-, Mcd1-, or Scc2-depleted cells. (E) Same as in (D) but measured in G₂/M-arrested cells. (F) Published Mcd1-v5 ChIP-seq data at the endogenous HOpr (top) (50). Mcd1-v5 ChIP-qPCR over engineered HOpr before (gray) and after (yellow) LexA induction. Arrows denote positions tested in ChIP-qPCR. (G) Time course of LexA invasion into nucleosome −4 in cells depleted of Smc2, a subunit of the condensin complex. bp, base pair.

In human cells, cohesin was proposed to enhance TF binding through a “bookmarking” mechanism, where cohesin maintains chromatin accessibility during mitosis, facilitating TF rebinding in the subsequent G₁ phase (47). This mechanism requires cell cycle progression from M to G₁. To test if that is the case in yeast, we degraded Smc3 in G₂/M-arrested cells and measured nucleosome occupancy during LexA induction. Cohesin depletion again resulted in reduced nucleosome displacement (Fig. 6E), suggesting that the cohesin effect can be decoupled from cell cycle progression. The bookmarking model also requires cohesin to bind to the same site as the TF that it facilitates. To test this, we performed Mcd1-V5 ChIP–quantitative polymerase chain reaction (qPCR) near the LexO site ± LexA induction (Fig. 6F). In the absence of LexA, the LexO site exhibits detectable but low Mcd1 enrichment, consistent with published Mcd1 ChIP sequencing (ChIP-seq) signals over the endogenous HOpr (50). This enrichment is not increased with LexA induction. These results argue against the bookmarking model and indicate that LexA binding and the resulting nucleosome rearrangement do not recruit or trap cohesin locally.

Condensin is another SMC complex that shares structural and functional similarities to cohesin. Both complexes have loop extrusion activities, and in yeast, cohesin mainly loop extrudes and condenses chromosome during mitosis, while condensin can loop extrude in G₁ (51, 52). To test whether condensin also contributes to LexA invasion, and specifically, if it promotes LexA-mediated nucleosome displacement in G₁, we repeated the LexA induction experiment while depleting Smc2, an ATPase subunit of condensin, in both cycling and G₁-arrested cells (fig. S6, D and E). As observed previously, nucleosome displacement is slower in G₁ compared to cycling cells. Smc2 depletion had little effect on the nucleosome displacement rate in cycling and G₁ cells (Fig. 6G), suggesting that condensin does not play a significant role in this process.

Cohesin facilitates the binding and nucleosome invasion by a native PF, Cbf1

We next examined whether the effect of cohesin described above extends to a native PF, Cbf1. For this experiment, endogenous Cbf1 was depleted alone or in combination with Smc3 using auxin-induced degradation (fig. S7A), and exogenous Cbf1 was reintroduced through 17β-estradiol induction. We then performed MNase and ChIP assays at multiple time points to monitor Cbf1 binding and nucleosome invasion in the presence or absence of cohesin (Fig. 7A). Western blot analysis confirmed that Smc3 depletion does not affect Cbf1 induction levels (fig. S7, B and C).

Fig. 7. Cohesin facilitates the binding and nucleosome invasion by a native PF, Cbf1.

(A) Experimental workflow for Cbf1 induction ± Smc3 depletion, followed by genome-wide measurement of Cbf1 binding and nucleosome positioning. (B) Low throughput analysis of Cbf1 binding and nucleosome invasion at an engineered motif positioned near the dyad of HOpr nucleosome −4. Left: quantification of nucleosome occupancy at nucleosome −4 following Cbf1 induction in WT versus Smc3-depleted cells. Right: Cbf1 ChIP-qPCR signals at the same site. (C) Effect of Smc3 depletion on nucleosome occupancy near genome-wide NDRs. Average profile (top) and heatmap (bottom) of nucleosome occupancy aligned at NDR centers. The y axis of the heatmap represents MNase-seq fragment sizes. (D and E) MNase-seq and Cbf1 ChIP-seq signals over time at two representative loci where Cbf1 binds to open chromatin (D) or within nucleosomes (E) in the presence (left) or absence (right) of Smc3. (F and G) Average nucleosome occupancy and Cbf1 ChIP-seq signals following Cbf1 induction over all Cbf1 binding sites in open chromatin (F) or embedded within nucleosomes (G). At nucleosome-embedded sites, Cbf1 induction triggers local nucleosome depletion, which is slowed in the absence of Smc3. Two biological replicates were analyzed for each condition. (H) Nucleosome depletion at nucleosome-embedded Cbf1 sites during Cbf1 induction, grouped by local cohesin density (based on Mcd1 ChIP-seq signals). Two biological replicates were analyzed for each condition. (I) Same as in (H) but grouped by local H3 turnover signals. Two biological replicates were analyzed for each condition. bp, base pair.

We first carried out a low-throughput analysis using a single engineered Cbf1 motif at the HOpr nucleosome −4 dyad. Following Cbf1 induction, nucleosome −4 is displaced, and an NDR formed near the Cbf1 motif (fig. S7D). Both Cbf1 binding and nucleosome displacement at this site are slowed by Smc3 depletion (Fig. 7B and fig. S7D), demonstrating that cohesin also facilitates nucleosome invasion by a native PF, at least at this engineered site.

To evaluate this effect at a genome-wide scale, we performed MNase-seq and Cbf1 ChIP-seq during Cbf1 induction. Across all time points (t = 0, 30, 60, and 90 min), MNase-seq profiles exhibit consistent differences ± Smc3 when aligned at the centers of genome-wide NDRs. Compared to WT, Smc3-depleted cells exhibit higher average occupancy within NDRs and greater fuzziness of flanking nucleosomes (Fig. 7C and fig. S7E). Since PFs are known to maintain nucleosome depletion within NDRs and form barriers to phase neighboring nucleosomes, these observations suggest that PF activity is compromised in the absence of cohesin.

We next focused on Cbf1 activity at its target sites. ChIP-seq reveals a progressive increase in Cbf1 binding over the induction time course, with 148 ChIP-seq peaks detected at 90 min. Most of these peaks (137 of 148) contain Cbf1 motifs. These motif-containing sites closely match previously mapped endogenous Cbf1 targets, both by ChIP-seq (53) and ChIP-exo (54), although the latter identified more sites, likely due to differences in signal-to-noise ratios and thresholding (fig. S7F). Among these 137 sites, MNase-seq data at t = 0 indicate that 75 reside within broad NDRs (example in Fig. 7D), 29 are embedded within well-positioned nucleosomes (example in Fig. 7E), and the rest are in short linkers, at nucleosome edges, or ambiguous (fig. S7G). Upon Cbf1 induction, nucleosome-embedded sites underwent progressive nucleosome loss and transitioned into NDRs, consistent with the pioneering activity of Cbf1 (Fig. 7E and fig. S7G). We also identified 107 additional Cbf1 motifs that show significant nucleosome depletion over time despite lacking strong Cbf1 ChIP-seq signals (fig. S7G). These are likely bona fide Cbf1 targets missed by ChIP-seq due to low signal-to-noise or biases against nucleosome-bound events, and most of them overlap with ChIP-exo sites (fig. S7F). We therefore combined these 107 sites with the 29 ChIP-seq positive sites as the nucleosome-embedded Cbf1 targets (n = 136).

Aggregate analyses of MNase-seq signals centered around nucleosome-embedded Cbf1 sites confirm that nucleosome displacement is slowed down by Smc3 depletion (Fig. 7, F and G). Cbf1 binding is also reduced in the absence of Smc3 by 15 to 20% for both NDR- and nucleosome-localized sites after 60-min induction (Fig. 7, F and G). Therefore, cohesin facilitates Cbf1 binding to its target sites across the genome, regardless of their locations relative to nucleosomes.

We next asked whether this effect depends on cohesin’s local enrichment. Nucleosome-embedded Cbf1 sites were stratified by cohesin density based on the Mcd1 ChIP-seq signals (fig. S7H) (50). Unexpectedly, sites with high cohesin level do not show faster nucleosome depletion; if anything, depletion was slightly slower. In addition, Smc3 depletion slows down Cbf1 nucleosome invasion similarly for both high- and low-density sites (Fig. 7H). Thus, the effect of cohesin is not limited to sites with high local enrichment and is likely due to broader chromatin changes mediated by cohesin.

Next, we examined the relationship between Cbf1 invasion and histone H3 turnover. Our data in Fig. 4 show that Cbf1 can invade into a nucleosome with a high H3 turnover level, but we cannot exclude the possibility that high turnover may affect the dynamics of this invasion quantitatively. To test this, Cbf1 nucleosomal sites were separated into high- and low-turnover groups based on published H3 turnover signals (fig. S7H) (41). Consistent with our findings for LexA (Fig. 4, B and C), Cbf1 displaces nucleosomes with low H3 turnover more efficiently, and both categories show slower nucleosome invasion upon Smc3 depletion (Fig. 7I). These results indicate that high H3 turnover also has inhibitory effects on Cbf1 binding, and cohesin promotes its nucleosome invasion at sites with both high and low H3 turnover.

Last, Cbf1 induction also allows us to assess whether local RSC density influences nucleosome invasion. We used previously published RSC ChEC-seq data (55) to categorize Cbf1 nucleosome invasion sites into two groups, RSC enriched (RSC+) and RSC depleted (RSC−) (fig. S8A). The initial average nucleosome occupancy is lower at RSC+ sites than at RSC− sites. Nevertheless, Cbf1-mediated nucleosome depletion was highly similar between the two groups (fig. S8B). We next performed MNase-seq during Cbf1 induction in the absence of RSC (Sth1 anchor-away). Western blot analysis shows that Cbf1 induction driven by the MET3pr is slower when Sth1 is anchored away (fig. S8C), which explains why nucleosomes are displaced to a less extent (fig. S8D). However, the decrease in nucleosome occupancy is still very similar between RSC+ and RSC− sites. Together, these results support the conclusion that RSC is not required for the nucleosome invasion by Cbf1.

DISCUSSION

By establishing a nucleosome invasion model in WT and mutant budding yeast using the bacterial TF LexA, this study uncovered several unexpected findings. First, despite its poor nucleosome-binding ability in vitro, LexA can rapidly invade into a well-positioned nucleosome in vivo with a binding motif placed at the dyad, an inaccessible region on the nucleosome. Second, the nucleosome invasion capability of LexA does not depend on factors typically associated with nucleosome dynamics, including CRs and histone chaperones. Third, contrary to expectations, rapid histone turnover inhibits rather than facilitates nucleosome displacement mediated by LexA and Cbf1. Last, the cohesin complex, best known for its roles in chromosome compaction and segregation, emerges as a key factor that promotes nucleosome invasion by LexA as well as by endogenous PFs. Although the precise mechanisms underlying these observations remain unclear, we explore some possible explanations below.

Intrinsic versus context-dependent pioneering activity

PFs are often defined by their ability to bind nucleosomes in vitro with similar or only modestly reduced affinities compared to naked DNA. This property is “intrinsic” to these factors, as it arises from direct histone contacts and/or structural features that are compatible with nucleosomal DNA (5, 13). PFs are also functionally defined on the basis of their ability to invade into nucleosomes and open chromatin in vivo. The two definitions are generally believed to be linked, i.e., the propensity to bind nucleosomes in vitro should facilitate, or even enable, chromatin opening inside cells. However, LexA challenges this view by showing both poor nucleosome binding in vitro and efficient nucleosome invasion in vivo, clearly demonstrating that the former is not required for the latter. In previous work, we also observed that another bacterial TF, TetR, can bind and displace nucleosomes in yeast, although in that case the motif was placed closer to the nucleosome entry/exit region (16). These findings suggest that, at least in budding yeast, the ability to invade into nucleosomes may apply to any TFs with sufficiently high DNA binding affinity and or concentration.

Nonetheless, our data also reveal important distinctions between LexA and native PFs. LexA invasion results in the formation of a very narrow, linker-like NDR, whereas Cbf1 binding at the same position produces a broader NDR (Fig. 1E versus fig. S7D). This difference may be functionally significant, as wider NDRs expose more DNA, potentially allowing additional regulatory factors to be recruited. Furthermore, LexA has difficulty invading into nucleosomes with high histone turnover, while Cbf1, and even its DBD alone, can displace these nucleosomes. These observations suggest that native PFs still have stronger “pioneering activity” than LexA in vivo. Collectively, these observations suggest the need to reconsider how PFs are defined. Rather than solely relying on their nucleosome-binding affinity in vitro, defining PFs based on their ability to create broad NDRs or to displace highly dynamic nucleosomes in vivo may better reflect their pioneering functions.

Remodelers function downstream of TF invasion into nucleosomes

By eliminating most remodeling activities in the cells, we demonstrate that CRs are not required for LexA-mediated nucleosome displacement. This observation is consistent with the behavior of endogenous PFs, like Cbf1. This conclusion is supported by several prior studies. First, major CRs including RSC, ISWI, INO80, and SWI/SNF have been shown to be dispensable for the initial chromatin opening by endogenous PFs, although they play a role in modulating the width of the resulting NDRs (24). Second, simultaneous depletion of RSC, ISWI, INO80, and CHD1 in yeast has minimal impact on NDRs across the native yeast genome (31). Third, in live Drosophila hemocytes, the chromatin-binding behavior of the PF GAGA factor is largely independent of the recruited CRs NURF and PBAP, which selectively facilitate the binding of downstream factors (56).

Together, these results suggest that the primary function of CRs is to shape the size of NDRs following PF binding, rather than to initiate chromatin opening. This idea is supported by recent structural evidence showing that many CRs, including RSC and INO80, bind to extended linker DNA to engage nucleosomes, suggesting that preexisting NDRs are required for CRs to carry out remodeling activity (57, 58). Thus, a likely sequence of events during NDR formation is that PFs first bind to their target motifs within nucleosomes and generate localized NDRs, which serve as an entry point for CRs to bind and reposition adjacent nucleosomes to further adjust chromatin accessibility.

Notably, SWI/SNF was not depleted in our experiment because its loss severely impairs 17β-estradiol induction (24), thereby preventing LexA overexpression and subsequent analysis of nucleosome invasion. However, we do not believe that SWI/SNF accounts for the residual remodeling activity that enables LexA access. First, SWI/SNF is not very abundant (∼2000 complexes per cell) and thus unlikely to globally remodel the ∼60,000 nucleosomes in the yeast genome to allow rapid LexA invasion into a randomly selected loci. Second, SWI/SNF depletion affects only ∼250 NDRs genome-wide, without changing nucleosome occupancy at the NDR center. It also does not impair chromatin opening by endogenous PFs recruited to synthetic HOpr (24, 59). Together, these observations suggest that SWI/SNF acts primarily through locus-specific recruitment, modulating nucleosome dynamics at targeted sites rather than providing a global remodeling activity required for LexA-mediated invasion.

Two strategies to overcome rapid histone exchange

In both yeast and higher eukaryotes, rapid turnover of H3/H4 tetramers predominantly occurs at promoters, enhancers, and within actively transcribed gene bodies (39–41, 60). This dynamic histone exchange is usually thought to be associated with transcriptional activation, as it disrupts the canonical nucleosome structure and transiently exposes otherwise occluded DNA. Contrary to this view, we found that LexA is less efficient at invading into nucleosomes with rapid histone turnover, suggesting that at least some turnover sites may inhibit, rather than enhance, DNA accessibility.

The underlying mechanism for this inhibitory effect remains to be elucidated. However, the observation that the same high-turnover nucleosome can be displaced by endogenous PFs, including the Cbf1 DBD, and LexA fused with VP16 raises the following possibilities. Regions with rapid histone turnover are sites where nucleosomes are continuously disassembled and reassembled, and newly deposited histones may form structures that disfavor LexA binding, given its inherently low affinity and fast dissociation from nucleosomal DNA. Cbf1 or its DBD may remain stably bound during this disruptive process due to its ability to engage nucleosomal DNA. Alternatively, LexA binding may be sterically occluded by histone turnover machinery. In contrast, LexA-VP16 may persist at these sites by recruiting activating cofactors, such as SWI/SNF and SAGA, which can destabilize and disrupt nucleosomes and clear the roadblock for LexA to stably bind.

These findings support the existence of two mechanistically distinct strategies by which TFs can achieve stronger pioneering activity in vivo. The first is to develop specialized DBDs that can better dock onto nucleosomes, like what was observed for yeast PFs Reb1 and Cbf1 (11, 13) and several PFs in higher eukaryotes (5, 6, 61). Alternatively, TFs can use regions outside of DBD to recruit cofactors to enhance chromatin interaction and overcome the nucleosome barrier. Note that these two mechanisms are not mutually exclusive. Some PFs like FOXA1, SOX2, and Zelda may use both strategies for their chromatin opening and stable occupancy (62, 63).

Promotion of PF activity by the cohesin complex

By depleting cohesin subunits, we demonstrate that the cohesin complex enhances chromatin accessibility at genome-wide NDRs and accelerates nucleosome invasion by both the artificial TF LexA and the endogenous PF Cbf1. These results are consistent with a previous study in mammalian cells showing that cohesin is broadly associated with TF binding clusters and that cohesin loss leads to reduced DNA accessibility and TF occupancy (47, 64). Together, these observations suggest that cohesin may play a conserved role in promoting PF activity.

A previous mammalian study attributed this effect to a “mitotic bookmarking” mechanism, whereby cohesin remains bound during early mitosis to maintain open cluster sites, enabling TFs to re-engage in late M/next G₁ phase (47). However, a recent study indicates that cohesin is largely removed from chromosomes during mitosis in mammalian cells (65, 66). Furthermore, this model does not apply to budding yeast, where TFs are not broadly displaced during M phase due to modest chromatin compaction (67). Consistently, LexA induced at G₂/M can engage chromatin and mediate nucleosome displacement (Fig. 5G). The bookmarking mechanism also requires M-G₁ transition, yet the effect of cohesin on LexA invasion can be detected in cells arrested in G₂/M without cell cycle progression. Therefore, the effect of cohesin on chromatin accessibility and TF binding is unlikely due to bookmarking in yeast.

Although the precise mechanism remains unclear, two observations provide important clues. First, cohesin promotes Cbf1 binding at both NDR-localized and nucleosome-embedded sites. Second, this effect is not restricted to loci with strong cohesin enrichment. These results argue against a simple “site holding” model in which cohesin directly maintains open chromatin at its binding sites. Instead, cohesin appears to exert a global influence on chromatin accessibility. One possibility is that the passage of cohesin during loop extrusion physically reorganizes or untangles chromatin fibers, thereby enhancing accessibility (68). In addition, cohesin activity has been shown to generate torsional stress on chromatin (69), which can modulate the binding specificity and stability of nucleosomes and TFs (45). Further experiments will be required to test these hypotheses.

MATERIALS AND METHODS

Plasmid and yeast strains

Standard methods were used for strain construction and plasmid cloning. All yeast strains were derived from the w303 background. The endogenous HOpr (−1298 to −272 relative to HO open reading frame) was deleted, and a modified HOpr lacking Swi5 binding sites was inserted into the CLN2 locus (16). LexA or LexA-HA, driven by a core GAL1 promoter, was integrated at the HIS3 locus. The same strain also expressed Gal4-ER-VP16, allowing LexA induction in the presence of 17β-estradiol. Engineering LexO/Cbf1/Abf1 motifs into loci with various levels of H3 turnover was done using an FOA pop-out scheme with seamless editing.

For auxin-induced degradation strains, a 3x-v5-AID DNA cassette was amplified with primers containing 70 bp of DNA homology arms and integrated by homologous recombination to generate in-frame gene fusions. Protein degradation was verified by treating the samples with 500 μM 3-indoleacetic acid (Sigma-Aldrich, I2886), typically for 30 to 60 min, followed by Western blotting. Anchor-away strains were generated by tagging the target with FRB-TAP-GFP. Nuclear protein depletion was induced by treatment with rapamycin (1 μg/ml) for 45 min and confirmed by spotting assays or fluorescence microscopy. Temperature-sensitive FACT mutants were grown at 20°C for all experiments unless otherwise indicated. For inactivation, cells were shifted to 37°C for 45 min.

For the inducible Cbf1-TAP strain in Fig. 3, we replaced the endogenous CBF1 promoter with the GAL1 promoter. An inducible Cbf1-TAP driven by MET3 promoter was integrated at the HIS3 locus. Cells were first grown in SCD + 20× Met to repress the endogenous Cbf1 expression and shifted to SCD-Met to induce Cbf1-TAP. The Cbf1-TAP strain in Fig. 7 was generated by tagging the C terminus of endogenous Cbf1 with an AID tag. An inducible Cbf1-TAP driven by a GAL1pr was inserted in the HIS3 locus which can be induced with 17β-estradiol.

For cell cycle arrest experiments, G₁ arrest was achieved with 5 μM α-factor (Zymo Y1001) in SCD-Met medium for 3 hours (equivalent to the doubling time of the anchor-away strain in synthetic media) (70). G₂/M arrest was achieved with nocodazole (15 μg/ml; Sigma-Aldrich, M1404) for 2 hours (one doubling time in yeast extract, peptone, and dextrose).

Cell cycle analysis

Cycling or arrested cells were grown to an optical density (OD) of 0.4, and 1 ml of culture from each condition was collected for fluorescence-activated cell sorting (FACS) analysis (71). Cells were fixed with ethanol overnight and stained with 2.5 μM Sytox Green in 50 mM sodium citrate buffer containing ribonuclease A (RNase A; 20 μg/ml). Following overnight proteinase K treatment (10 μl, 20 mg/ml), the cells were pulse-sonicated to reduce aggregates. FACS analysis was carried out using a Cytek Aurora (Cytek Biosciences) analyzer at the Huck Flow Cytometry Facility. The instrument was configured using conventional mode, and gain settings were optimized for FSC, SSC, and Sytox Green signals. Single cells were gated using FSC-A/H and Sytox Green-A/H bivariant plots. The DNA intensity (Sytox Green) was displayed in a histogram using linear scale with the G₁ peak near the first quartile and the G₂ peak near the second quartile of the histogram axis. For each sample, 50,000 cells were recorded, and the data files were analyzed using Kaluza software (Beckman Coulter).

In vitro nucleosome assembly

The YNN1 sequence was cloned in the pUC19 plasmid. A LexA-binding site (TACTGTATGAGCATACAGTA) was introduced at the p65 position (65 bp from the entry/exit site centered at the dyad) by site-directed mutagenesis (QuikChange kit, Agilent). DNA templates were labeled at the 5′ end with Cy3 using labeled oligonucleotides from Integrated DNA Technologies. All DNA constructs were amplified by PCR and purified using an anion-exchange MonoQ column (Cytiva) on a high-performance liquid chromatography system (Agilent). Purified DNA was stored in 0.5× TE buffer (5 mM tris base and 0.5 mM EDTA).

Recombinant histones (human H2A with K119C, human H2B, Xenopus laevis H3 with C110A, and human H4) were purchased from “The Histone Source” (Colorado State University; https://histonesource-colostate.nbsstore.net/). Histone octamers were refolded under a 1:1.3 molar ratio of H3 and H4 to H2A and H2B. Site-specific labeling was performed by conjugating Cy5 maleimide to H2A K119C. Labeled octamers were purified on a Superdex 200 column (Cytiva) using an ÄKTA Pure fast protein liquid chromatography system (Cytiva). Nucleosomes were reconstituted by salt dialysis using purified DNA and histone octamers mixed at a 1.5:1 molar ratio. The reconstituted nucleosomes were purified using sucrose gradient centrifugation on an Optima L-90K Ultracentrifuge (Beckman) at 41,000 rpm under 4°C for 22 hours. The 5 to 30% sucrose gradient was prepared with 0.5× TE buffer using Gradient Master (Biocomp). Purified nucleosomes were stored in 0.5× TE with 20% glycerol and aliquoted in −80°C.

Electrophoretic mobility shift assay

DNA (0.5 nM) or nucleosomes were incubated with variable concentrations of LexA in T130 buffer [10 mM tris-HCl (pH 8.0), 130 mM NaCl, 10% glycerol, and 0.0075% Tween 20] under room temperature for 20 min. The mixtures were run on a 5% acrylamide native gel with 0.3× TBE buffer (30 mM tris base, 27 mM boric acid, and 0.3 mM EDTA) for 90 min at 300 V.

MNase assay and MNase-seq

A total of 1.5 OD₆₆₀ units of cells were harvested and washed sequentially with 1 ml of water and 1 ml of 1 M sorbitol. The pellet was resuspended in 0.5 ml of spheroplasting solution [1 M sorbitol, 0.5 mM 2-mercaptoethanol, and zymolyase (0.18 mg/ml)] and incubated at room temperature for 7 min with gentle inversion. Spheroplasts were collected by centrifugation at 500g for 3 min, washed twice with 1 ml of 1 M sorbitol, and resuspended in 200 μl of digestion buffer [1 M sorbitol, 50 mM NaCl, 100 mM tris-Cl (pH 7.4), 5 mM MgCl₂, 1 mM CaCl₂, 1 mM 2-mercaptoethanol, 0.5 mM spermidine, and 0.075% NP-40]. Samples were incubated at 37°C for 8 min with 1 U of MNase (Sigma-Aldrich, N3755). Reactions were terminated by adding 20 μl of quench buffer (250 mM EDTA and 5% SDS). DNA was purified by phenol-chloroform extraction, and mononucleosome-sized DNA was gel-extracted before proceeding to stacking qPCR (72). Nucleosome occupancy was normalized to a reference nucleosome in the EXO84 terminator (24).

For MNase-seq, 1.5 OD₆₆₀ units of cells were digested with 1 U (Sigma-Aldrich, N3755) of MNase for 8 min that resulted in ∼80% genome converted into mononucleosomes. One hundred nanograms of MNase-digested DNA was used to generate sequencing libraries using the NEBNext Ultra II DNA Library Prep Kit for Illumina (catalog no. E7645L), following the manufacturer’s protocol. Libraries were sequenced on an Illumina NextSeq 2000 in 2 × 50 bp paired-end mode. About 20 million reads per sample were generated and aligned to the sacCer3 genome assembly or to our synthetic HOpr construct (figs. S1, B and G, and S2D) using Bowtie2 with presets “--very-sensitive.” Bigwig files were generated using deepTools bamcoverage (version 3.5.6) using a 130- to 170-bp size selection, and alignments were trimmed by 15 bp on either side using offset 15 −15. Two biological replica were performed at each time point.

Western blotting and quantification

Yeast cultures (10 ml) were grown to ∼0.15 OD₆₆₀, harvested, washed with 0.5 ml of H₂O, and resuspended in 200 μl of 1 M NaOH. After incubation at room temperature for 10 min, cells were pelleted and resuspended in 50 μl of SDS–polyacrylamide gel electrophoresis (SDS-PAGE) sample buffer. Proteins were resolved by SDS-PAGE and transferred for Western blotting using standard protocols. Primary antibodies were used at the following concentrations: anti-LexA (Thermo Fisher Scientific, PA1-4966) 1:3000; anti-HA (Abcam, ab91100) 1:5000; anti-v5 (Abcam, ab27671) 1:2000; anti–α-tubulin (Abcam, ab184970) 1:15000; anti-actin (Abcam, ab230169) 1:2000; anti-H3 (Abcam, ab1791) 1:2000; anti–glyceraldehyde phosphate dehydrogenase (GAPDH; Abcam ab125247) 1:5000; anti-TAP (Thermo Fisher Scientific, CAB1001) 1:1000; and anti-myc (Abcam, ab32) 1:1000. Secondary antibodies were used at the following concentrations: goat anti-rabbit immunoglobulin G (IgG; Cell Signaling Technology, 7074s) 1:3000; goat anti-mouse IgG (Invitrogen, 62-6520) 1:2000; goat anti-rabbit (Bio-Rad, 12004162 and 12005869) 1:3000; and goat anti-mouse (Bio-Rad, 12004158 and 12005866) 1:3000.

To enable direct comparisons, inducible LexA and endogenous Cbf1, Reb1, and Ser2 were tagged at their C terminus with an HA tag. Reb1 and Cbf1 are known budding yeast PFs and serve as a good internal comparison for monitoring the expression of LexA. Protein copy numbers were obtained from SGD (Cbf1, 5757 ± 2714; Reb1, 5582 ± 1829; and Ser2, 11714 ± 2168) and converted to concentrations using the average volume of 42 μm³ and the nuclear volume of 3 μm³ in haploid yeast (73). The Western blot intensities of these proteins agreed well with their known copy numbers within the cell (fig. S1C).

For quantification, four concentrations of lysate were analyzed per sample to ensure that signals were within the linear detection range (only data within the linear range were used). A loading control confirmed equal protein loading, and an internal reference standard was included for normalization across blots. For Cbf1-TAP quantifications in fig. S7B, normalization was performed against total protein content combined with an internal standard. This is because the usual loading control, GAPDH, is regulated by Cbf1 [based on previously published RNA-seq measurements (74)], and its expression varied in Cbf1-AID strains ± auxin.

ChIP sequencing

Yeast cultures (50 ml) were grown in SCD-Met medium to an OD₆₆₀ of 0.4. Crosslinking was performed by adding 1.39 ml of 37% formaldehyde directly to each culture and incubating at room temperature for 20 min. Crosslinking was quenched with 2.7 ml of 2.5 M glycine for 5 min. Cells were pelleted at 3000 rpm for 3 min at 4°C, resuspended in 1 ml of tris-buffered saline (TBS), and washed twice with 1 ml of TBS. Cells were lysed by resuspension in 250 μl of FSPP buffer [50 mM Hepes-KOH (pH 7.5), 140 mM NaCl, 1 mM EDTA, 1% Triton X-100, 0.1% sodium deoxycholate, 1% protease inhibitor cocktail, and 1 mM phenylmethylsulfonyl fluoride] with 300-μl glass beads. Samples were vortexed at 4°C twice for 20 min each, separated by a 10-min break. Lysates were released by piercing the tube ends with a hot needle and centrifuging at 3000 rpm for 5 min at 4°C into 15-ml tubes. Lysates were recovered, supplemented with 1 ml of FSPP buffer, and sonicated (30 s on/off cycles for 7 min). After centrifugation at 14,000 rpm for 20 min at 4°C, the 200-μl supernatant was saved as input control.

For immunoprecipitation, 200 μl of lysate was diluted with 800 μl of FSPP buffer and incubated with 20 μl of preblocked magnetic IgG beads and 1 to 2 μg of antibody (TAP: Thermo Fisher Scientific, CAB1001; LexA: Thermo Fisher Scientific, PA1-4966; V5: Abcam, ab27671) overnight at 4°C with rotation. Beads were sequentially washed with (i) FA lysis buffer [50 mM Hepes-KOH (pH 7.5), 140 mM NaCl, 1 mM EDTA, 1% Triton X-100, and 0.1% sodium deoxycholate], (ii) FA lysis + 150 mM NaCl, (iii) FA lysis + 500 mM NaCl, (iv) LiCl buffer [0.25 M LiCl, 1% NP-40, 1% sodium deoxycholate, 1 mM EDTA, and 10 mM tris-HCl (pH 8.0)], and (v) TE buffer (pH 8.0). ChIP sample was then eluted with 400 μl of elution buffer [50 mM NaCl, 50 mM tris-HCl (pH 8.0), 10 mM EDTA, and 1% SDS] by rotating at 30°C for 30 min. Eluates were treated with 5 μl of proteinase K (20 mg/ml) overnight at 65°C. Input controls were processed in parallel by adding 160 μl of FA lysis buffer, 40 μl of 10% SDS, and 5 μl of proteinase K, followed by overnight incubation at 65°C.

DNA was extracted with phenol:chloroform:isoamyl alcohol. ChIP samples were resuspended in 40 μl of H₂O. Input samples were resuspended in 380 μl of TE buffer with 10 μl of RNase A (10 mg/ml) and incubated at 37°C for 30 min. DNA was washed with ethanol and resuspended in 40 μl of H₂O. Factor enrichment was verified through qPCR. For ChIP-seq, ∼5 million paired-end reads were generated and aligned to S. cerevisiae genome sacCer3 with Bowtie2 (version 2.5.4) using default parameters. Peak calling was performed with MACS2 (version 2.2.7.1). Bigwig files were generated using deepTools bamcoverage (version 3.5.6). Two biological replica were performed at each time point.

Statistical analysis

For all ChIP-seq and MNase-seq experiments, two biological replicates were performed for each sample. After confirming the reproducibility between replicates, replicates were merged for further analysis unless stated otherwise. For paired-end MNase-seq data, dyad and nucleosome positions were inferred using DANPOS3, which calculates fragment midpoints and genome-wide dyad density (75). In addition to average nucleosome coverage from all mapped reads (fig. S7E), we generated V-plots (76) to provide more information about nucleosome distributions by sorting the sequencing reads according to their fragment length (Fig. 7C). Nucleosome occupancy over Cbf1 motifs identified from ChIP-seq or ChIP-exo was quantified from MNase-seq data. For each paired-end fragment, centers were calculated from the fragment length and used for generating aggregate fragment midpoints centered around Cbf1 motifs. Midpoints falling within 10-bp bins were used for generating the occupancy plots. For Cbf1 ChIP-seq, binding was quantified over Cbf1 motifs using deepTools, and ChIP-seq signal was extracted. For each point in the time course, values were normalized to the 0-min time point to calculate the fold change (Fig. 7, F to I).

For Cbf1 analysis, we first scanned the yeast genome for potential consensus binding sites using published position weight matrix and the recommended cutoff (77). For each consensus site, we quantified both the MNase-seq signal at the motif center and the Cbf1 ChIP-seq signal within ±100 bp of the motif. Across the Cbf1 induction time course, nucleosome occupancy and Cbf1 enrichment at each site were normalized to the t = 0 condition, such that these values represent a fold change relative to baseline. On the basis of the combined ChIP-seq and MNase-seq profiles, we classified Cbf1 binding sites into four categories: (i) ChIP+ and NDR-localized, (ii) ChIP+ and nucleosome-embedded, (iii) ChIP− and nucleosome-embedded, and (iv) nucleosome edge or ambiguous (fig. S7G). Similarly, ChIP-seq densities for cohesin and H3 turnover were determined for each Cbf1 binding site using previously published datasets (41, 50) and grouped according to their density.

Supplementary Materials

This PDF file includes:

Figs. S1 to S8

Tables S1 and S2