Nucleosome Remodeling at the Yeast PHO8 and PHO84 Promoters without the Putatively Essential SWI/SNF Remodeler

Chromatin remodeling by ATP-dependent remodeling enzymes is crucial for all genomic processes, like transcription or replication. Eukaryotes harbor many remodeler types, and it is unclear why a given chromatin transition requires more or less stringently one or several remodelers. As a classical example, removal of budding yeast PHO8 and PHO84 promoter nucleosomes upon physiological gene induction by phosphate starvation essentially requires the SWI/SNF remodeling complex. This dependency on SWI/SNF may indicate specificity in remodeler recruitment, in recognition of nucleosomes as remodeling substrate or in remodeling outcome. By in vivo chromatin analyses of wild type and mutant yeast under various PHO regulon induction conditions, we found that overexpression of the remodeler-recruiting transactivator Pho4 allowed removal of PHO8 promoter nucleosomes without SWI/SNF. For PHO84 promoter nucleosome removal in the absence of SWI/SNF, an intranucleosomal Pho4 site, which likely altered the remodeling outcome via factor binding competition, was required in addition to such overexpression. Therefore, an essential remodeler requirement under physiological conditions need not reflect substrate specificity, but may reflect specific recruitment and/or remodeling outcomes.


Introduction
The packaging of eukaryotic genomes into chromatin, especially into nucleosomes consisting of 147 base pairs (bp) of DNA wrapped around a histone octamer [1,2], is primarily repressive to all processes that require access to DNA, like transcription or replication [3]. Therefore, in contrast to prokaryotes [4], the default state of eukaryotic genomes corresponds to an "off-state" and most genome activities necessitate opening ("remodeling") of chromatin structure and its nucleosome constituents [5]. This fits to the regulatory requirements of multicellular eukaryotes where only a small fraction of the genome is expressed in each cell. Unicellular eukaryotes, like the budding yeast Saccharomyces cerevisiae, are atypical in this regard, as they constitutively express most of their genes. Accordingly, the nucleosome organization at most yeast promoters corresponds to an "open door policy," [6] where a nuclease hypersensitive site (HSS) (for discussion of the alternative terms nucleosome-depleted region (NDR) versus nucleosome-free region (NFR) see [7,8]) of up to 150 bp length just upstream of the transcription start site allows access to the transcription machinery [5,8,9]. Nonetheless, some genes are repressed under standard growth conditions and become induced upon specific, often environmentally triggered, signaling. The promoters of these inducible genes are often kept in a repressed state by positioned nucleosomes over functional elements, like the TATA and UAS (upstream activating sequence, i.e., yeast enhancers) elements, and nucleosome remodeling is a crucial prerequisite [10] for their activation [11,12].
Promoters of the phosphate response (PHO) genes in S. cerevisiae are paradigmatic for this latter type (reviewed in [13,14]). PHO signaling is a nutrient-dependent regulatory pathway that responds to intracellular inorganic phosphate (P i ) levels [15,16]. The principal transactivator Pho4 activates numerous PHO genes constituting the PHO regulon. Of these, the PHO5, PHO8 and PHO84 genes serve since many years as models for nucleosome remodeling mechanisms in the course of promoter opening and transcriptional activation upon gene induction and are highly instructive for analogous mechanisms in multicellular eukaryotes. Under conditions of high phosphate supply (+P i conditions), Pho4 is phosphorylated by the Pho80/Pho85 cyclin/cyclin-dependent kinase complex. This phosphorylation prevents gene activation by Pho4 as it expedites nuclear export and prevents nuclear reimport of Pho4, as well as hampers the interaction with Pho4 s cooperative binding partner Pho2 [17][18][19][20][21][22]. Upon P i removal (−P i conditions) and concomitant decrease of intracellular P i levels, Pho80/Pho85 are inhibited by Pho81, and non-phosphorylated Pho4 accumulates in the nucleus, binds to its binding sites, the UASp elements (upstream activating sequence phosphate regulated), triggers promoter chromatin remodeling [23] and transactivates transcription [24]. Accordingly, deletion of PHO80 or PHO85 [25] or outtitration of the Pho80/Pho85 kinase activity by PHO4 overexpression [23] circumvents the actual PHO signalling pathway and leads to PHO gene induction even at +P i conditions.
A long series of studies in many labs, especially regarding the PHO5 promoter (reviewed in [13]), elucidated the promoter chromatin opening mechanism in exceptional detail. Pho4 recruits via its activation domain [26,27] chromatin modifying enzymes, like the SAGA complex [28,29], that lead to transient hyperacetylation of promoter nucleosome histones [30,31]. Pho4 also recruits, directly or indirectly, ATP-dependent chromatin remodeling enzymes, for example, the SWI/SNF complex [27,29,32]. Such "remodelers" translocate, assemble and disassemble nucleosomes, as well as change their composition with regard to histone variants [33][34][35][36][37], by using the energy of ATP hydrolysis. They are at the core of the promoter-opening mechanism.
One fundamental question regarding remodelers deals with their specificity and essential versus redundant roles. Many remodeling complexes in yeast, like ISW1a, ISW1b, ISW2, INO80, RSC and SWI/SNF, are able to slide nucleosomes along the DNA, but only two of these, RSC and SWI/SNF, are able to disassemble nucleosomes from the DNA [34,36,37]. The RSC complex is the only remodeler essential for viability in yeast [38], while all others can be deleted, even several of them at the same time [39], arguing that their functions are not essential or can be compensated by other remodelers. The SWR1 complex is an example of highly specialized, non-essential but non-compensated function as only this remodeler exchanges canonical histone H2A for the variant histone H2A.Z (Htz1 in yeast) in vivo [40]. Nucleosomes in an swr1 deletion strain hardly contain any Htz1 despite expression of Htz1 [41].
While it seems clear-although mechanistically still unexplained-that the SWR1 complex exerts a very specific activity that no other remodeler can provide, it is much less clear why a certain remodeling process, like promoter chromatin opening upon induction, should essentially depend on one particular remodeler even though others exert the same or very similar activities, at least in vitro.
To address this question, we turned to the PHO8 and PHO84 promoters in budding yeast. At both promoters, there are nucleosomes that cannot be removed in the absence of SWI/SNF activity, even under full physiological induction conditions [42][43][44]. This encompasses all PHO8 promoter nucleosomes and therefore prevents PHO8 activation, while it relates to only the so-called "upstream nucleosome" (see below) at the PHO84 promoter without much effect on full PHO84 expression. In addition to SWI/SNF, the INO80 complex is also involved in opening both promoters, but not in such an essential way [42,45]. Curiously, even though both promoters depend stringently on SWI/SNF they do not require the RSC complex [46], which would seem an appropriate substitute for SWI/SNF, as both these remodeling complexes belong to the same remodeler family, show very similar activities in terms of nucleosome sliding and disassembly in vitro [34,36,37] and co-operate at other inducible promoters [47,48].
As all three PHO promoters are coactivated by the same transactivator Pho4, their different dependencies on chromatin cofactors for chromatin opening seemed unlikely due to different recruitment specificities, i.e., in principle, the same set of cofactors should be recruited to each promoter. Instead, it seems possible that there is something particular, maybe linked to DNA sequence or histone modifications, about some nucleosomes that caused the strict SWI/SNF dependency of PHO8 and PHO84 promoter nucleosome remodeling in vivo. Conversely, SWI/SNF may possess a mechanistically unique activity. Maybe only this remodeler was able to remodel these particular nucleosomes and/or only SWI/SNF could remove these nucleosomes as remodeling outcome, whereas others would (re-)generate the nucleosome pattern of the repressed state so that on time average, it seemed as if they did not remodel such nucleosomes. Indeed, the chromatin state of the PHO8 promoter and of the upstream nucleosome at the PHO84 promoter in snf2 mutants, where SWI/SNF activity is abolished, always displayed a pattern similar to the repressed state in wt cells [42,43], i.e., other non-SWI/SNF remodelers always regenerated this state after replication.
Mechanistically, the stringent SWI/SNF requirement for remodeling PHO8 and PHO84 promoter nucleosomes could reflect at least three different scenarios. First, such particular nucleosomes may need an especially high local remodeling activity and SWI/SNF may be the only remodeler recruited by Pho4 to sufficiently high local levels. Second, only the SWI/SNF complex may efficiently accept such nucleosomes as substrates for remodeling. Third, other remodelers may also remodel such nucleosomes, but only SWI/SNF removes them in the end while the others tend to put them back so that on time average it seems as if they do not remodel such nucleosomes. Thus, the problem could lie in either the local remodeler concentration, in the nucleosome as substrate, or in the remodelerspecific remodeling outcome. In metaphorical terms, the question is if SWI/SNF is the only screwdriver ready at hand to unscrew that screw (recruitment specificity), or if SWI/SNF is the only screwdriver that fits that screw (substrate specificity), or if SWI/SNF is the only screwdriver that will unscrew while other screwdrivers mostly tighten that screw (remodeling outcome specificity). The first scenario should allow SWI/SNF-independent remodeling upon enhanced recruitment of other remodelers, while the second and third scenarios should not allow it. Such strict mechanistic dependency would contrast with the finding that remodelers of different families were all able to remodel nucleosomes of widely different intrinsic stabilities in vitro [53]. Nonetheless, some aspects of nucleosome properties and remodeler functions in vivo may be missed by such in vitro studies. The third scenario would be in line with in vitro demonstrations that different remodelers lead to different remodeling outcomes, e.g., nucleosome sliding, eviction or compositional changes [34,36,37,54] and generate different steady state nucleosome positioning on the same DNA sequence [55][56][57]. Conversely, changing the DNA sequence in a nucleosome can alter remodeling outcomes and remodeler preferences. As a prominent example, the RSC complex prefers remodeling nucleosomes that contain poly(dA:dT) sequences [58] and removes them in a directional way, i.e., in the 5 direction from the poly(dA) sequence [56]. In retrospect, this explains our earlier observation that introduction of poly(dA:dT) sequences into the "upstream nucleosome" at the PHO84 promoter relaxed its SWI/SNF requirement or remodeling in vivo [42]. The poly(dA:dT)-containing nucleosome likely became a preferred substrate for RSC, which circumvented the SWI/SNF requirement.
To distinguish these scenarios, we asked if it was possible to circumvent the strict SWI/SNF requirement for nucleosome remodeling at the PHO8 and PHO84 promoters. We found two conditions, insertion of an intranucleosomal UASp element or overexpression of PHO4 that, each on its own, led to PHO8 promoter nucleosome remodeling in the absence of SWI/SNF activity, while PHO84 promoter nucleosome remodeling required the combination of both. This argues against substrate specificity but for recruitment strength and/or remodeling outcome specificity underlying the otherwise strict SWI/SNF requirement at both promoters.

An Additional Intranucleosomal High-Affinity UASp Element Allows PHO8 Promoter Nucleosome Remodeling without SWI/SNF Activity
The PHO8 gene encodes an vacuolar alkaline phosphatase [59]. The PHO8 promoter in its repressed state (+P i conditions) is organized into three positioned nucleosomes and three short HSSs ( Figure 1A, [60]). The HSS closest to the gene start is just upstream of the TATA box and the other two contain the UASp2 and UASp1 elements, respectively. The consensus sequence of UASp elements is an E-box motif CACGTG and the extent of matching this motif as well as surrounding bases modulates Pho4 binding strength [61][62][63][64]. UASp2 (CACGTGG) is a high-affinity binding site, while UASp1 (CACGCTT) has low affinity and little role in transactivation as its deletion reduces induced levels of Pho8 alkaline phosphatase activity relative to wild type (wt) by only 5% [65]. Upon PHO8 induction, the −3 nucleosome flanked by the UASp elements becomes remodeled into an extended HSS of about 300 bp, as seen by increased DNaseI or MNase accessibility in indirect end labeling analyses, as well as by increased accessibility of the intranucleosomal HpaI restriction site [60]. This extended HSS and increased HpaI accessibility is not seen in snf2∆ deletion mutants lacking the Snf2 ATPase subunit of the SWI/SNF remodeling complex or if Snf2 ATPase activity is abolished by a point mutation in the snf2K798A allele ( Figure 1B, [43]).
To search for conditions that could allow PHO8 promoter nucleosome removal in the absence of SWI/SNF, we focused on nucleosome −3. "Nucleosome removal" means here any loss of canonical nucleosome structure as monitored by increased accessibility to DNaseI or restriction endonucleases. At the PHO5 promoter, nucleosome remodeling is aided by the endogenous intranucleosomal UASp2 site, as it contributes binding competition between the high-affinity DNA binder Pho4 and the histone octamer to the remodeling mechanism, like a pry bar [66]. Therefore, we introduced an additional highaffinity UASp element into the −3 nucleosome of the PHO8 promoter and tested if this allowed removal of this nucleosome in the absence of SWI/SNF. Indeed, the generation of a high-affinity CACGTGG element by three point mutations within the −3 nucleosome ( Figure 1A) led to increased HpaI accessibility (54 ± 3% for intranucleosomal high-affinity UASp vs. 9 ± 2% for wt promoter, Figure 1B) and increased DNaseI sensitivity ( Figure 1C, HSS between −388 and −874 marker bands, see  for examples of DNaseI patterns with not removed −3 nucleosome) in this region upon -P i induction in snf2∆ cells, which was similar to the chromatin opening extent of the wt PHO8 promoter in wt cells (61 ± 8%). This increase depended on the proper UASp sequence as a scrambled, still palindromic non-UASp element (CTCGAGG, Figure 2A, top panel) with the same GC content as the high-affinity UASp at the same position did not have this effect (17 ± 2% HpaI accessibility, Figure 2B).   Neither introduction of an intranucleosomal scrambled non-UASp site nor of an addition ternucleosomal high-affinity UASp site allowed PHO8 promoter opening without SWI/SNF activit Schematics analogous to Figure 1A showing the sequence changes for introduction of the scrambled UASp site in the −3 nucleosome (top) and the conversion of the low affinity UASp1 site into a high-af site (bottom) in the plasmid locus. (B) DNaseI indirect end labeling for the PHO8 plasmid locus analo to Figure 1C for the indicated genotypes and growth condition (-Pi). Corresponding HpaI accessi values analogous to Figure 1B are given below the blots. The condition -Pi snf2∆ intranucl. high aff. U is a biological replicate of the same condition in Figure 1C.
At the PHO5 promoter, enhanced remodeling by the intranucleosomal UASp ment was mainly due to its intranucleosomal location and not just due to increased P recruitment potential caused by the presence of an additional high-affinity UASp elem [66]. To test whether this was also true at the PHO8 promoter, we turned its low aff  Figure 1A showing the sequence changes for introduction of the scrambled non-UASp site in the −3 nucleosome (top) and the conversion of the low affinity UASp1 site into a high-affinity site (bottom) in the plasmid locus. (B) DNaseI indirect end labeling for the PHO8 plasmid locus analogous to Figure 1C for the indicated genotypes and growth condition (−P i ). Corresponding HpaI accessibility values analogous to Figure 1B are given below the blots. The condition −Pi snf2∆ intranucl. high aff. UASp is a biological replicate of the same condition in Figure 1C. PHO5 promoter chromatin opening triggered by Pho4 depends on its activation domain and the viral VP16 activation domain is a less potent substitute if fused to the Pho4 DNA binding domain (DBD) [26]. Similarly, we showed here that the PHO8 promoter nucleosome removal in the absence of SWI/SNF activity was also not possible if just the Pho4 DBD without the Pho4 activation domain was overexpressed (26 ± 1% HpaI accessibility, Figure 4) and was much less extensive upon overexpression of the VP16 activation domain fused to the Pho4 DBD (HpaI accessibility of 38 ± 2% for Pho4DBD-VP16 and less DNaseI hypersensitivity vs. 59 ± 6% HpaI accessibility and full DNaseI hypersensitivity for full length Pho4, Figure 4). As the Pho4 activation domain recruits chromatin cofactors [27], we concluded that increased nuclear Pho4 levels via PHO4 overexpression led to increased recruitment of other remodeling enzyme(s) that is/are able to remodel PHO8 promoter nucleosomes in the absence of SWI/SNF.

PHO80 Deletion Is Less Effective than PHO4 Overexpression Regarding PHO8 Promoter Nucleosome Removal without SWI/SNF
It seemed that PHO8 promoter nucleosome removal without SWI/SNF required elevated nuclear Pho4 levels. We therefore tested the pho80∆ deletion allele as another way

RSC Rather Hinders than Helps Circumventing the SWI/SNF Requirement for PHO8 Promoter Nucleosome Removal under Conditions of PHO4 Overexpression
RSC is the other member of the SWI/SNF remodeler family in yeast [49] that cooperates with SWI/SNF in remodeling at the PHO5 promoter [46] as well as at other highly expressed genes in yeast [47,48], and may provide the other remodeling activity in the absence of SWI/SNF for remodeling of the PHO8 promoter −3 nucleosome upon PHO4 overexpression. However, there were also observations that may argue the opposite way. We showed that RSC depletion via a temperature-sensitive sth1td degron allele at 37°C did not affect PHO8 promoter opening upon physiological induction [46]. Moreover, RSC ablation via a temperature-sensitive rsc3-ts allele at 37 °C led to some PHO8 promoter nucleosome removal already under +Pi conditions as seen by increased HpaI accessibility (47% in rsc3-ts [69]) and an altered DNaseI cleavage pattern similar to opened chromatin upon -Pi induction, even in the absence of Pho4 (rsc3-ts pho4 double mutant, [69]).
We wished to clarify RSC's role in PHO8 promoter nucleosome remodeling in the absence of SWI/SNF. RSC ablation on its own (sth1td allele) and in the presence of SWI/SNF increased HpaI accessibility at -Pi and 37°C conditions to a similar relative extent (86 ± 5% for sth1td vs. 71 ± 8% for wt ( Figure 6)) as did PHO4 overexpression in the wt background at -Pi and 30°C (82 ± 3% with ( Figure 3) vs. 61 ± 8% without PHO4 overexpression ( Figure 1B)). Therefore, removing RSC had a similar effect as elevated Pho4 levels, which suggested that RSC is not only not needed for chromatin opening at the PHO8 promoter, at least in the presence of SWI/SNF, but may even counteract opening. Its absence allowed partial remodeling under repressive [69] and an increased remodeling extent under inducing conditions ( Figure 6). At the PHO5 promoter, enhanced remodeling by the intranucleosomal UASp element was mainly due to its intranucleosomal location and not just due to increased Pho4 recruitment potential caused by the presence of an additional high-affinity UASp element [66]. To test whether this was also true at the PHO8 promoter, we turned its low affinity UASp1 element, which is not intranucleosomal but located in a constitutive HSS (Figure 1A), into a high-affinity (CACGTGG) site (Figure 2A, bottom panel). This also led to a slight increase in HpaI accessibility (29 ± 6%) but hardly any increase in DNaseI sensitivity in the region of the −3 nucleosome upon induction in snf2∆ cells ( Figure 2B). The effect was clearly much less pronounced than for the intranucleosomal high-affinity UASp element (54 ± 3% HpaI accessibility).
We concluded that an additional intranucleosomal high-affinity UASp element allowed circumventing the otherwise strict dependency of PHO8 promoter nucleosome removal on SWI/SNF activity probably by additional binding competition and less so by increased Pho4 recruitment potential.

PHO4 Overexpression Is Sufficient for PHO8 Promoter Chromatin Remodeling without SWI/SNF
It is well established that the stringency of chromatin cofactor dependency for PHO5 promoter opening scales reciprocally with induction strength as mediated by the extent of Pho4 recruitment and thereby Pho4 occupancy at the promoter [32]. Therefore, we asked if enhanced Pho4 recruitment alone, beyond physiological levels and without binding competition via an intranucleosomal Pho4 site, could also allow PHO8 promoter nucleosome removal without SWI/SNF. Elevated PHO4 expression leads to substantial nuclear levels of non-phosphorylated Pho4 already under +P i conditions and, thereby, to partial PHO induction, resulting in rather extensive chromatin opening at the PHO5, PHO8 and PHO84 promoters in wt, but not in snf2 cells (Figure 3, [23,42,45]). Combining PHO4 overexpression with -P i conditions corresponds to boosted induction beyond physiological "full induction" conditions. Indeed, this boosted induction led to increased HpaI accessibility at the PHO8 promoter for wt cells compared to mere -P i conditions (HpaI accessibility of 82 ± 3% with vs. 61 ± 8% without PHO4 overexpression, respectively, Figures 1B and 3). Remarkably, even snf2 cells showed substantially increased DNaseI sensitivity and HpaI accessibility, consistently somewhat more extensive in the snf2K798A than in the snf2∆ mutant (59 ± 6% vs. 43 ± 9% HpaI accessibility, respectively, Figure 3). This difference between snf2 alleles was also apparent for the very limited extent of remodeling upon PHO induction without PHO4 overexpression (HpaI accessibility of 20 ± 1% for snf2K798A vs. 9 ± 2% for snf2∆, Figure 1B). Chromatin opening upon PHO4 overexpression at -P i conditions was not significantly increased if it was combined with the intranucleosomal UASp element (HpaI accessibility of 64 ± 18% with vs. 43 ± 9% without intranucleosomal UASp, Figures 1B and 3).
PHO5 promoter chromatin opening triggered by Pho4 depends on its activation domain and the viral VP16 activation domain is a less potent substitute if fused to the Pho4 DNA binding domain (DBD) [26]. Similarly, we showed here that the PHO8 promoter nucleosome removal in the absence of SWI/SNF activity was also not possible if just the Pho4 DBD without the Pho4 activation domain was overexpressed (26 ± 1% HpaI accessibility, Figure 4) and was much less extensive upon overexpression of the VP16 activation domain fused to the Pho4 DBD (HpaI accessibility of 38 ± 2% for Pho4DBD-VP16 and less DNaseI hypersensitivity vs. 59 ± 6% HpaI accessibility and full DNaseI hypersensitivity for full length Pho4, Figure 4). As the Pho4 activation domain recruits chromatin cofactors [27], we concluded that increased nuclear Pho4 levels via PHO4 overexpression led to increased recruitment of other remodeling enzyme(s) that is/are able to remodel PHO8 promoter nucleosomes in the absence of SWI/SNF.

PHO80 Deletion Is Less Effective Than PHO4 Overexpression Regarding PHO8 Promoter Nucleosome Removal without SWI/SNF
It seemed that PHO8 promoter nucleosome removal without SWI/SNF required elevated nuclear Pho4 levels. We therefore tested the pho80∆ deletion allele as another way of modulating nuclear Pho4 levels in vivo. PHO80 deletion leads to constitutive nuclear accumulation of non-phosphorylated Pho4 [17] such that PHO5 and PHO8 promoter chromatin becomes remodeled already at +P i conditions [60,67]. We wondered if the combination of the pho80 allele with -Pi inducing conditions, leading to increased induction at the PHO5 promoter [68], would allow PHO8 promoter opening without SWI/SNF activity. However, this was hardly the case as judged by the DNaseI pattern ( Figure 5) and also HpaI accessibility did not increase significantly (20 ± 1% in the corresponding snf2K798A PHO80 strain ( Figure 1B) vs. 25 ± 7% in snf2K798A pho80∆ ( Figure 5)). In addition, combining the pho80∆ allele with PHO4 overexpression and -P i induction did not generate more extensive nucleosome removal than without the pho80∆ allele (HpaI accessibility of 50 ± 1% with pho80 allele (Figure 5) vs. 59 ± 6% without (Figure 3)).

RSC Rather Hinders Than Helps Circumventing the SWI/SNF Requirement for PHO8 Promoter Nucleosome Removal under Conditions of PHO4 Overexpression
RSC is the other member of the SWI/SNF remodeler family in yeast [49] that cooperates with SWI/SNF in remodeling at the PHO5 promoter [46] as well as at other highly expressed genes in yeast [47,48], and may provide the other remodeling activity in the absence of SWI/SNF for remodeling of the PHO8 promoter −3 nucleosome upon PHO4 overexpression. However, there were also observations that may argue the opposite way. We showed that RSC depletion via a temperature-sensitive sth1td degron allele at 37 • C did not affect PHO8 promoter opening upon physiological induction [46]. Moreover, RSC ablation via a temperature-sensitive rsc3-ts allele at 37 • C led to some PHO8 promoter nucleosome removal already under +P i conditions as seen by increased HpaI accessibility (47% in rsc3-ts [69]) and an altered DNaseI cleavage pattern similar to opened chromatin upon -P i induction, even in the absence of Pho4 (rsc3-ts pho4 double mutant, [69]).
We wished to clarify RSC's role in PHO8 promoter nucleosome remodeling in the absence of SWI/SNF. RSC ablation on its own (sth1td allele) and in the presence of SWI/SNF increased HpaI accessibility at -P i and 37 • C conditions to a similar relative extent (86 ± 5% for sth1td vs. 71 ± 8% for wt ( Figure 6)) as did PHO4 overexpression in the wt background at -P i and 30 • C (82 ± 3% with (Figure 3) vs. 61 ± 8% without PHO4 overexpression ( Figure 1B)). Therefore, removing RSC had a similar effect as elevated Pho4 levels, which suggested that RSC is not only not needed for chromatin opening at the PHO8 promoter, at least in the presence of SWI/SNF, but may even counteract opening. Its absence allowed partial remodeling under repressive [69] and an increased remodeling extent under inducing conditions ( Figure 6). To follow up on this, we asked if nucleosome remodeling would proceed to a degree in the absence of SWI/SNF if we combined it with RSC removal. We tested promoter opening at the HpaI site in an snf2∆ sth1td double mutant with PHO4 o pression upon -Pi induction and at 37 °C. Under these conditions, the HpaI site snf2∆ sth1td strain was slightly but significantly more open than in the snf2∆ strai 2% vs. 52 ± 3% (Figure 6)). This argues that RSC is not one of the other remodele rather hindrance than help for remodeling of the PHO8 promoter −3 nucleosome absence of SWI/SNF.

Combining PHO4 Overexpression with an Additional High-Affinity Intranucleosoma UASp Element Is Necessary to Allow Partial PHO84 Promoter Opening without SWI/SN
The PHO84 gene encodes a high-affinity Pi transporter of the plasma membran Its promoter is among the strongest PHO promoters and harbors five UASp ele (UASpA to UASpE, [64]). Similar to the PHO5 and PHO8 promoters, it also und extensive chromatin remodeling upon induction by phosphate starvation [42,68]. conditions, two high-affinity Pho4 sites (UASpC/D) reside in a constitutive shor This HSS is flanked by two well-positioned nucleosomes (the "upstream" and " stream" nucleosomes, respectively ( Figure 7A)) that each harbor a low affinity Ph (UASpB and UASpE, respectively). The TATA box region has an ambiguous nucle To follow up on this, we asked if nucleosome remodeling would proceed to a higher degree in the absence of SWI/SNF if we combined it with RSC removal. We tested PHO8 promoter opening at the HpaI site in an snf2∆ sth1td double mutant with PHO4 overexpression upon -P i induction and at 37 • C. Under these conditions, the HpaI site in the snf2∆ sth1td strain was slightly but significantly more open than in the snf2∆ strain (58 ± 2% vs. 52 ± 3% (Figure 6)). This argues that RSC is not one of the other remodelers, but rather hindrance than help for remodeling of the PHO8 promoter −3 nucleosome in the absence of SWI/SNF.

Combining PHO4 Overexpression with an Additional High-Affinity Intranucleosomal UASp Element Is Necessary to Allow Partial PHO84 Promoter Opening without SWI/SNF
The PHO84 gene encodes a high-affinity P i transporter of the plasma membrane [70]. Its promoter is among the strongest PHO promoters and harbors five UASp elements (UASpA to UASpE, [64]). Similar to the PHO5 and PHO8 promoters, it also undergoes extensive chromatin remodeling upon induction by phosphate starvation [42,68]. At +P i conditions, two high-affinity Pho4 sites (UASpC/D) reside in a constitutive short HSS. This HSS is flanked by two well-positioned nucleosomes (the "upstream" and "downstream" nucleosomes, respectively ( Figure 7A)) that each harbor a low affinity Pho4 site (UASpB and UASpE, respectively). The TATA box region has an ambiguous nucleosome organization of intermediate accessibility. Upon -P i induction, the complete PHO84 promoter region, including the upstream and downstream nucleosomes, as well as most of the TATA box region, turns into an extensive HSS of~500 bp length of virtually full nuclease accessibility. This chromatin remodeling and gene induction is mainly driven by the UASpC/D/E elements with very little contribution of the UASpA/B elements.  As remodeling of the upstream nucleosome at the PHO84 promoter depends strictly on the SWI/SNF complex [42], similar to PHO8 promoter nucleosome remodeling [43], we asked if this dependency could be circumvented in a similar manner as we found here for the PHO8 promoter. We first tried the combination of -P i induction and PHO4 overexpression. In contrast to the PHO8 promoter, this did not allow remodeling of the upstream nucleosome in snf2∆ or snf2K798A cells as monitored by HhaI accessibility (4 ± 1% in snf2∆ and 11 ± 3% in snf2K798A cells ( Figure 7B)), which probes specifically the SWI/SNF-dependent upstream nucleosome [42], and by DNaseI indirect end labeling ( Figure 7B). Second, we turned the intranucleosomal low affinity UASpB element (CACGTTG), located in the "upstream" nucleosome, via a single point mutation into a high affinity site (CACGTGG, Figure 7A). This also did not allow remodeling of the upstream nucleosome in snf2∆ cells upon overnight phosphate starvation (HhaI accessibility of 8 ± 2 % (Figure 8)).
Int. J. Mol. Sci. 2023, 24,4949 As remodeling of the upstream nucleosome at the PHO84 promot on the SWI/SNF complex [42], similar to PHO8 promoter nucleosome re asked if this dependency could be circumvented in a similar manner as the PHO8 promoter. We first tried the combination of -Pi induction a pression. In contrast to the PHO8 promoter, this did not allow remodeli nucleosome in snf2∆ or snf2K798A cells as monitored by HhaI accessibil and 11 ± 3% in snf2K798A cells ( Figure 7B)), which probes specifically pendent upstream nucleosome [42], and by DNaseI indirect end labelin ond, we turned the intranucleosomal low affinity UASpB element (CA in the "upstream" nucleosome, via a single point mutation into a high GTGG, Figure 7A). This also did not allow remodeling of the upstre snf2∆ cells upon overnight phosphate starvation (HhaI accessibility of Figure 8. The combination of PHO4 overexpression with introduction of an ad somal high-affinity Pho4 binding site (high-affinity UASpB) allows partial re stream nucleosome at the PHO84 promoter in the absence of SWI/SNF activity. labeling for the plasmid PHO84 locus in snf2∆ cells (CY407) as in Figure 7B aft tion in phosphate-free medium (-Pi). "+ high aff. UASpB" corresponds to p where UASpB is mutated to a high-affinity UASp element according to Figur the plasmid locus was included in the gel on the right hand side, but the posi was estimated according to similar published gels ( Figure 3B,C in ref. [42]). Ev ibration of band gel positions it is clear that the protected region correspond nucleosome (see left hand side of figure) is not detected in the gel on the right Figure 8. The combination of PHO4 overexpression with introduction of an additional intranucleosomal high-affinity Pho4 binding site (high-affinity UASpB) allows partial remodeling of the upstream nucleosome at the PHO84 promoter in the absence of SWI/SNF activity. DNaseI indirect end labeling for the plasmid PHO84 locus in snf2∆ cells (CY407) as in Figure 7B after overnight incubation in phosphate-free medium (−P i ). "+ high aff. UASpB" corresponds to plasmid pCB84a-Bhi, where UASpB is mutated to a high-affinity UASp element according to Figure 7A. No marker for the plasmid locus was included in the gel on the right hand side, but the position of the HhaI site was estimated according to similar published gels ( Figure 3B,C in ref. [42]). Even without exact calibration of band gel positions it is clear that the protected region corresponding to the upstream nucleosome (see left hand side of figure) is not detected in the gel on the right hand side.
Only the combination of the high-affinity UASpB site with PHO4 overexpression and -P i induction led to partial remodeling of the upstream nucleosome in the absence of SWI/SNF, as seen by 31 ± 1% HhaI accessibility and a DNaseI pattern in the region of the upstream nucleosome that resembled the open state ( Figure 8, compare with the DNaseI pattern of the open state in wt under -P i conditions in Figure 7B).
We concluded that PHO84 promoter upstream nucleosome remodeling depended more strictly on SWI/SNF than remodeling at the PHO8 promoter. Circumventing the SWI/SNF dependency at the PHO84 promoter required the combination of enhanced Pho4 recruitment and an intranucleosomal high-affinity Pho4 site and even then remained only partial.

Discussion
Our study contributes to one of the central questions regarding chromatin remodeling mechanisms: why do certain chromatin transitions depend on specific chromatin cofactors in a more or less stringent way? The differential SWI/SNF requirement for the PHO5, PHO8 and PHO84 promoters served since many years as an example for the difference between relaxed versus essential requirement although these promoters depend on the same transcriptional transactivator Pho4 [13]. Part of this discussion was the assumption that there may be something special about PHO8 and PHO84 promoter nucleosomes so that only SWI/SNF and not even the closely related [49] and more abundant [71] RSC remodeling complex, which is still active in snf2 mutants, could remodel such nucleosomes. This aspect of nucleosome substrate and remodeling specificity in PHO promoter chromatin opening mechanisms has to be re-considered, as we now demonstrate conditions for PHO8 and PHO84 promoter nucleosome removal without SWI/SNF in vivo. We summarize these conditions in Table 1. Abbreviations: "nucl.": nucleosome, "PHO4-DBD": DNA binding domain of Pho4, "n.d.": not determined; "n.a.": not applicable.
Regarding the PHO8 promoter, the so far seemingly essential SWI/SNF requirement for nucleosome removal does not reflect a mechanistic specialty of this particular remodeling enzyme that would exclusively be able to remodel these special nucleosomes in vivo. Instead, we showed via boosted induction strength upon PHO4 overexpression that the SWI/SNF requirement at this promoter has to join the ranks of the many chromatin cofactors, e.g., Ino80, Gcn5, Asf1, that are involved in PHO promoter chromatin opening to a degree that reciprocally scales with induction strength [32,42,45,72,73]. Remodelers other than SWI/SNF were able to remodel PHO8 promoter nucleosomes, but seem to be recruited less efficiently or remodel less effectively so that their activity became apparent only upon increased Pho4 levels. At present, we do not know for sure if the other remodeler(s) is/are directly recruited by Pho4, as would be similar to remodeler recruitment by Gcn4 [47], but the strong effects of PHO4 overexpression and the dependency on the Pho4 activation domain argue in this direction. We also cannot distinguish if this forced recruitment leads to the same amount of local activity of other remodeler(s) that then remodel(s) with the same effectiveness as otherwise SWI/SNF does. We also do not know if the forced recruitment has to bring in (much) more of the other remodeler(s) as its/their specific remodeling activity per remodeler molecule may be lower than that of SWI/SNF. This will be difficult to distinguish, as it depends on accurate determination of local remodeler occupancy in vivo. Monitoring remodeler occupancy at PHO promoters by chromatin immunoprecipitation (ChIP) or other techniques is notoriously unreliable. For example, RSC binding to the PHO5, PHO8 and PHO84 promoters was detected by anti-Rsc9-ChIP-chip [74] and anti-Sth1-CUT & RUN [75], but not by native anti-Sth1-ChIP-seq [76] or anti-Rsc8-ChEC-seq [77]. Further, one would have to compare occupancies among remodelers, but all these techniques have unknown and difficult-to-calibrate remodeler-specific efficiencies.
Nonetheless, our data clearly show that forced induction strength allows circumventing the so far seemingly essential SWI/SNF requirement for PHO8 promoter opening. In addition, we also found that the remodeling outcome could be changed for the non-SWI/SNF remodeler(s) by introduction of an intranucleosomal high-affinity UASp, even at physiological induction strength. As only the intranucleosomal UASp and not just an additional UASp in the neighboring linker region showed this effect, this approach unlikely amounts to just another way of increased recruitment of Pho4 and consequently of other remodelers at the promoter. Instead, this likely is another case where binding competition between the specific DNA binding factor Pho4 and the histone octamer potentiates or alters chromatin remodeling outcome, as we showed previously at the PHO5 promoter [66]. The effect here is even more pronounced as binding competition not only makes a difference between weak versus strong remodeling at the PHO5, but between no and almost full opening at the PHO8 promoter.
At the PHO84 promoter, the SWI/SNF-dependency for remodeling of the upstream nucleosome probably corresponds to a combination of recruitment and remodeling outcome specificity as only the combination of forced induction and introduction of intranucleosomal high-affinity UASp element led to partial remodeling without SWI/SNF. It may seem surprising that remodeling at the PHO84 promoter looked rather extensive by DNaseI indirect end labeling but rather partial by restriction enzyme accessibility (Figure 8). This is due to the very limited digestion degrees employed for DNaseI indirect end labeling so that the patterns reflect the most nuclease-sensitive chromatin states while the majority of templates are still undigested in the region of interest (see strong bands in upper part of lanes for all DNaseI patterns). Therefore, DNaseI patterns are good for demonstrating if hypersensitive regions are generated and for showing how they look. However, this technique has to be complemented by a more quantitative assay, like restriction enzyme accessibility, to monitor which fraction of chromatin templates was actually remodeled. Accordingly, this "partially open" PHO84 promoter upstream nucleosome upon phosphate starvation and PHO4 overexpression in snf2∆ cells and in the presence of the intranucleosomal high-affinity UASp likely means that this nucleosome is fully remodeled for some templates, but only in a smaller fraction than in wt cells under full induction conditions. For sure, remodeling of PHO8 and PHO84 promoter nucleosomes without SWI/SNF calls for one or several other remodeling enzyme(s). At the PHO8 promoter, the RSC complex is unlikely to contribute to the other remodeler(s), but rather, seems to counteract them. RSC opposes PHO8 promoter opening with or without SWI/SNF, as HpaI accessibility was always higher upon RSC depletion. The effect was not large, but was significant, and may amount to another example of remodeler-specific remodeling outcome. In this sense, and in line with our earlier conclusions [69], we suggest that the outcome of PHO8 promoter chromatin remodeling by RSC may correspond to the nucleosome positioning pattern of the repressed, whereas the outcome of remodeling by SWI/SNF corresponds to the pattern of the induced state. RSC activity would be dominant at the promoter under +P i conditions, as it is globally much more abundant than SWI/SNF [71]. Upon PHO induction, Pho4 binds at the promoter and recruits SWI/SNF [27,32] and thereby increases local SWI/SNF levels so that SWI/SNF would overcome/outcompete the RSC activity and remodel this nucleosome in its SWI/SNF-specific way [78], leading to the extended hypersensitive site. Accordingly, if RSC is ablated, the remodeling outcome by SWI/SNF is increased under inducing conditions, as shown here (Figure 6), and already, the global SWI/SNF levels without Pho4-mediated local recruitment under +P i conditions led to partial nucleosome remodeling [69]. Conversely, other remodelers in the absence of SWI/SNF would also have to overcome RSC. Indeed, we noticed ( Figure 3) that the ATPase-dead version of the SWI/SNF complex (snf2K798A) allowed a bit more opening by other remodelers than the complete absence of the SWI/SNF complex (snf2∆). This tendency was the other way around at the PHO5 promoter where chromatin opening was more impaired in the snf2K798A than in the snf2∆ mutant [45]. These slight differences between both SWI/SNF-inactivating alleles at these two promoters may reflect that both SWI/SNF and RSC contain bromodomains [35] and compete for binding to acetylated histones that are present at both promoters during induction [30,31]. Therefore, we suggest that a nonfunctional SWI/SNF complex that still binds to acetylated histones hinders RSC's positive role during PHO5 promoter, but hinders RSC's negative role during PHO8 promoter opening.
While RSC is no help for PHO8 promoter opening, we cannot exclude that RSC may have a role at the PHO84 promoter in the presence of the intranucleosomal high-affinity UASp, as argued above regarding a possible change of remodeling outcome due to the addition of binding competition.
As we showed that also the INO80 complex [42,45] has a role in physiological opening of the PHO8 and PHO84 promoters, we speculate that INO80 contributes to the other remodeling activity in the absence of SWI/SNF. Nonetheless, also other remodelers like Isw1 or Chd1 may contribute, as we showed for the PHO5 promoter [46]. Our finding that SWI/SNF-independent PHO8 promoter chromatin opening could be achieved either by increased recruitment or by additional binding competition may suggest that the other remodeler(s) differ(s) in the former vs. the latter case. These questions will be addressed in a future study.
In summary, this and future studies that investigate differential chromatin cofactor requirements across genomic loci help to understand the diversification of specificity in recruitment, substrate and remodeling outcome in the evolution of remodeler ATPases and other chromatin cofactors and their more or less essential roles in physiological remodeling pathways.

Strains, Media, Plasmids and Strain Construction
Saccharomyces cerevisiae strains used in this study are listed in Table 2. Strains CY397 ura3 and CY408 ura3 were derived from CY397 and CY408 after selection on 5-fluoroorotic acid (5-FOA)-containing medium and confirmed for uracil auxotrophy. Strain CY39780 was generated by transformation of CY397 with a linear DNA fragment of the pho8 locus where a LEU2 marker gene cassette was inserted into the PHO80 coding region. Strain CY407 pho8::KAN and CY397 pho8::KAN were generated by transformation of CY407 or CY397, respectively, with a linear DNA fragment generated by PCR using genomic DNA of strain Y04315 (EUROSCARF strain) as template and the primers 5 -CTTGCTAGCAACAATAGGCG-3 and 5 -AGGAAGAAGTTGGCTGGTAG-3 . Successful disruption of the genomic PHO8 ORF was confirmed by Southern blotting as hybridization with a probe hybridizing in the mid-coding region of the PHO8 gene no longer gave a signal. For repressive conditions (high phosphate, +P i ), yeast strains were grown at 30 • C in YPDA medium (YPD with 0.1 g/liter adenine plus 1 g/l KH 2 PO 4 ) for strains without and in YNB selection medium supplemented with the required amino acids for strains with plasmids. Phosphate starvation conditions always corresponded to incubation in phosphate-free synthetic medium with or without lack of amino acids for plasmid selection [45,80]. For transfer to phosphate-free medium, cells were grown and washed in water and resuspended in the phosphate-free medium. If not indicated otherwise, phosphate starvation was overnight.
For results at 37 • C in Figure 6, all strains were grown in YNB medium supplemented with 1 g/l KH 2 PO 4 (YNBP) either lacking uracil for strains with the sth1 td allele or without selection for all other strains. All strains were first grown to logarithmic phase at 24 • C and then shifted to 37 • C and either YNBP or phosphate-free medium prewarmed to 37 • C (without uracil for sth1 td containing strains) over night. For each experiment involving the sth1 td allele, the respective temperature-sensitivity phenotype was confirmed by growth arrest at 37 • C for more than 12 h after shifting back from phosphate-free to +P i medium, i.e., YNBP minus uracil medium.

Chromatin Analysis
The preparation of yeast nuclei and chromatin analysis of nuclei by DNaseI indirect end labeling and restriction endonuclease accessibility assays were done as previously described [10,46,80,82,83]. In brief, chromatin digested with DNaseI and deproteinized was secondarily cleaved with BglII or SspI for PHO8 and PHO84, respectively. DNA fragments separated on a 1.5% agarose gel (Loening buffer: 40 mM Tris, 12 mM NaOAc, 36.4 mM acetic acid, 1 mM EDTA) were transferred onto a Nylon membrane (Biodyne ® , PALL Corporation) by Southern blotting and specifically visualized with a probe abutting the respective restriction site. The probe for the chromosomal PHO84 locus was a PCR product corresponding to bases −1083 to −1428 from the ATG of the PHO84 ORF, the probe for the PHO8 locus was a PCR product corresponding to bases +78 to +568 from the ATG of the PHO8 ORF. To monitor the PHO84 plasmid locus, secondary cleavage for DNase I indirect end labeling was done with HindIII and the probe for the plasmid locus corresponded to the HindIII-BamHI fragment of pBR322. Probes were labeled with [α-32 P]dCTP using the kit PrimeIt II (Stratagene). The four marker bands (lane M) were generated by double digests with HindIII and either ClaI, AgeI, ApaI, or BsBI (from top to bottom in the lanes) for PHO84, and BglII and either EcoRV, NdeI, HindIII, or SacI (from top to bottom in the lanes) for PHO8. Hybridized membranes were exposed to X-ray films (Fuji Super RX). Films were scanned with an Epson Perfection V700 scanner, either in color or gray scale mode. Scans were imported into Adobe Photoshop CS6 and processed by conversion into grayscale format. Sometimes linear level adjustment was applied to the entire image.  Figures S1-S7). For restriction nuclease accessibility assays, chromatin was incubated with HpaI or HhaI and deproteinized DNA was cleaved with restriction enzymes that frame the HpaI or HhaI cleavage site: BglII/EcoRVfor HpaI at PHO8 or HindIII for HhaI at PHO84. We controlled that the restriction enzyme concentration was not limiting by using two enzyme concentrations that differed at least 2-fold (75 and 150 U for HpaI, 60 and 240 U for HhaI). The resulting DNA fragments were resolved in agarose gels and Southern blotted as for DNaseI mapping. Probes for the PHO8 and PHO8 chromosomal locus were the same as for DNase I mapping. To monitor HhaI accessibility at the PHO84 plasmid locus, BamHI and EcoRV were used for secondary cleavage and a PCR product from −557 to −310 from the ATG of the PHO84 ORF on the plasmid locus was used as probe. Quantification of the percentage of cleaved DNA was done by PhosphorImager analysis (Fuji FLA3000, Fujifilm imaging plate, BAS-MP) with Aida Image analyzer software v.4.27 (Raytest). The restriction enzyme accessibility [%] was calculated as the quotient [cut/ (cut + uncut)]. Error bars show either the range of two or the standard deviation of more than two biological replicates.