The chromatin structure at the PHO84 promoter undergoes extensive remodeling upon induction.
We characterized the PHO84
promoter chromatin structure under repressing conditions, i.e., in rich or synthetic medium with additional phosphate to ensure full repression, and under inducing conditions, i.e., synthetic phosphate-free medium. By DNase I indirect end-labeling analysis of the repressed state (+Pi
) we detected a short hypersensitive (sHS) region (about 150 bp), roughly between the MfeI and ApaI restriction sites, that was flanked by one positioned nucleosome upstream and one downstream (Fig. , upstream nucleosome and downstream nucleosome). This sHS region contained two closely positioned high-affinity Pho4 binding sites, UASpC and UASpD, whereas the two low-affinity sites, UASpB and UASpE, were occluded by the positioned upstream and downstream nucleosomes, respectively (Fig. ) (54
). In addition, we observed a broad hypersensitive region upstream of the BsrBI restriction site. Upon induction (−Pi
), the upstream nucleosome and at least one nucleosome downstream of the sHS region were remodeled, leading to an extended hypersensitive (eHS) region of about 500 bp. Its upstream border was almost fused to the broad hypersensitive region and the downstream border faded into the core promoter region around the TATA box and the transcriptional start site (Fig. ; see also Fig. , , and , below). This way UASpB and UASpE became accessible (Fig. ). Sometimes the eHS region appeared to contain a short region of lower DNase I accessibility between the MfeI and ApaI sites (see Fig. and ), which may reflect Pho4 and recruited factors bound to UASpC and UASpD. In Fig. the intensity of the broad hypersensitive region upstream of the BsrBI site appeared to change somewhat upon induction, which was probably attributable to an overall lower degree of digestion. However, in the majority of cases it did not undergo major changes upon induction (see Fig. , , and , −Pi
panels, below; also, data not shown). Therefore we refer to it as a constitutive hypersensitive region (cHS).
FIG. 1. The chromatin structure at the PHO84 promoter undergoes extensive remodeling upon induction. (A) DNase I indirect end-labeling analysis of the chromatin structure at the chromosomal PHO84 locus in wt (CY339 pCB84a) and pho4 (CY338) strains grown in phosphate-containing (more ...)
FIG. 4. Chromatin remodeling at the PHO84 promoter is incomplete and delayed in the absence of Snf2 and only delayed in the absence of Gcn5. (A) PHO84 promoter induction kinetics as in Fig. for wt (CY339 pCB84a), snf2 (CY409 pCB84a), and gcn5 (more ...)
FIG. 5. The Snf2 ATPase domain point mutant as well as a snf2 gcn5 double mutant show the same PHO84 promoter chromatin organization in the induced state as the snf2 deletion mutant. (A) DNase I indirect end labeling of the induced PHO84 promoter chromatin structure (more ...)
FIG. 8. Mutations that progressively destabilize the upstream nucleosome also progressively relieve its Snf2 dependency of remodeling, but destabilization or complete removal of the upstream nucleosome has only a small effect on the Snf2 dependency of overall (more ...)
The chromatin transition was fully dependent on the transactivator Pho4, as the PHO84 promoter chromatin pattern under inducing conditions in a pho4 deletion strain was virtually the same as the wild-type (wt) pattern of the repressed state (Fig. ). Interestingly, the unchanged nucleosome organization in a pho4 mutant suggested that the nucleosome positioning at the repressed promoter did not depend on binding of Pho4, e.g., to its linker binding sites UASpC and UASpD.
In addition to DNase I indirect end labeling, we mapped the PHO84 promoter chromatin structure of the repressed and the induced state more quantitatively by assaying the accessibility for several restriction enzymes along the promoter region that underwent the chromatin structure transition (Fig. ). Under +Pi conditions, the accessibilities for the various restriction enzymes were rather different, as would be expected for an organization into nucleosomes and nucleosome-free linker regions. The accessibilities at the HhaI and TaqI sites were the lowest, speaking for their protection by the upstream and downstream nucleosome, respectively. The BsrBI site was fully accessible under both repressing and inducing conditions, which was in agreement with its localization at the downstream start of the cHS region (Fig. ). The MfeI site was substantially but not fully accessible in the repressed state, indicating a location at the very border between the downstream nucleosome and the sHS region (Fig. ). Interestingly, a region of about 100 bp between the downstream nucleosome and the TATA box was only semiprotected in the repressed state, as the accessibilities for PacI, AgeI, and FokI were in the range of 43% (FokI) to 57% (AgeI) (Fig. ). This argued against a clearly positioned but rather for a less-organized nucleosome or for a chromatin structure with increased plasticity. Alternatively, some other DNA-protecting entity, e.g., an assembly of general transcription factors, could be responsible for this semiprotection.
In the induced state, all restriction enzyme sites tested in the promoter region of more than 500 bp upstream of the TATA box were highly accessible (Fig. ), confirming the presence of an extended hypersensitive region as observed by DNase I indirect end labeling (Fig. ) and suggesting that the whole region was mostly nucleosome free. Restriction enzyme accessibility assays also confirmed that the transition to this open chromatin state was dependent on Pho4 (Fig. ). For unknown reasons, the accessibilities at the HhaI, PacI, and AgeI sites, but not at the TaqI site, were even decreased under inducing compared to noninducing conditions in the pho4 strain.
In the wt strain, the accessibility of the FokI cleavage site, which overlaps with the TATA box sequence (15
), also increased upon induction, but not to the same high level as for the other restriction enzyme sites. In addition, the accessibility of the FokI site in the induced state was quite variable. This altogether may be due to the poor performance of this restriction enzyme on chromatin templates or may indicate the presence of an unstable or partially remodeled nucleosome or of components of the general transcription machinery recruited to the TATA box under inducing conditions.
In summary, the restriction enzyme accessibility data in connection with the DNase I indirect end-labeling analysis led us to map the upstream and downstream nucleosome as shown in Fig. . The main guidelines were the location of the ApaI and MfeI sites just at the borders of the nucleosomes toward the sHS region. For the reasons stated above, we have not assigned clear nucleosomal positions to the region between the downstream nucleosome and the TATA box region but suggest a less-organized DNA protective structure there.
This less-organized structure together with the somewhat elevated accessibilities at the HhaI and TaqI sites suggested to us that there may be a low level of Pho4 present at the promoter even under repressive conditions. Under +Pi
conditions Pho4 is mostly phosphorylated at multiple sites and mainly located in the cytosol (37
), but some Pho4 may still be nuclear. For example, earlier we showed a Pho4 footprint at the repressed PHO8
) and sin
mutations in histone H4 showed significantly derepressed PHO5
activity in a UASp element-dependent, i.e., presumably Pho4-dependent, manner under otherwise-repressing conditions (81
). Such nuclear Pho4 may bind especially to the accessible high-affinity sites UASpC and UASpD in the sHS region. This could lead to some basal recruitment of chromatin remodeling activities and a partially remodeled chromatin structure. We tested this by restriction enzyme analysis of the PHO84
promoter region in a pho4
deletion strain under high-phosphate conditions (Fig. ). However, only the accessibility of the HhaI site was decreased significantly, arguing that there was some basal Pho4-dependent remodeling only of the upstream nucleosome in the repressed state. This may also be noticeable based on the slightly more spread out sHS region in the presence of Pho4 (Fig. , compare wt +Pi
). In contrast, the structure between the downstream nucleosome and the TATA box region was maintained semiopen also in the absence of Pho4.
Remodeling of PHO84 promoter chromatin upon induction results in histone depletion from the promoter.
The generation of an extended hypersensitive region at the induced PHO84
promoter was reminiscent of our previous findings for the PHO5
). Such hypersensitivity was found by ourselves and others to reflect not just altered nucleosomal structures but also nucleosome disassembly leading to histone eviction from the promoter regions (1
). We checked if histones were lost also from the induced PHO84
promoter. During PHO84
induction kinetics, the histone H3 occupancy was monitored by ChIP using an antibody directed against the C terminus of histone H3. The histone H3 occupancy dropped after 2 hours of induction to about 10% of the level under repressing conditions (Fig. ). At the same time there was no significant change of the histone H3 occupancy at a telomere control locus. Therefore, chromatin remodeling at the PHO84
promoter eventually led to histone eviction.
FIG. 2. Histones are depleted from the PHO84 promoter region upon induction. The induction kinetics after transfer of a wt strain (CY337) to phosphate-free medium was followed by ChIP using a histone H3 C-terminal antibody and amplicons at the PHO84 promoter, (more ...) The extent of chromatin remodeling critically depends on the intranucleosomal UASpE site.
A special feature of the PHO84
promoter is the presence of five Pho4 binding sites, UASpA to UASpE, which makes it one of the strongest PHO promoters (54
). Ogawa et al. (54
) showed previously by using a PPHO84
Z reporter construct and deleting an extensive upstream region that UASpA and UASpB were not required for full PHO84
activity. They further showed by site-directed mutagenesis that the low-affinity site UASpE in combination with either of the high-affinity sites UASpC or UASpD was necessary and sufficient for PHO84
regulation. We wished to check if any of these effects on promoter activity actually reflected effects on chromatin remodeling.
We set up an analogous reporter system by constructing plasmid pCB84a, for which the PHO84
promoter was coupled to the PHO5
coding gene. Thereby we avoided possible chromatin structure artifacts due to the close presence of the bacterial lac
Z DNA sequence (unpublished observations). The enzymatic activity of the secreted acid phosphatase Pho5 can be measured easily with intact cells and PHO5
transcriptional activity fully correlates with acid phosphatase activity, indicating no significant posttranscriptional regulation of PHO5
). Importantly, the endogenous copy of PHO5
was always deleted in strains where PHO84
reporter constructs were used.
Using the pCB84a construct we observed phosphate-regulated PHO84 promoter activity with a substantially higher basal and final level of Pho5 acid phosphatase activity than seen with the PHO5 promoter (Fig. ; see also Fig. , below). This was expected for the stronger PHO84 promoter.
FIG. 3. Effects of Pho4 binding site deletions on PHO84 promoter induction kinetics and chromatin remodeling. (A) The PHO84 promoter induction kinetics after shift to phosphate-free medium was monitored in a pho5 strain (CY339) bearing reporter plasmids where (more ...)
FIG. 9. Induction of the PHO84 and PHO5 promoters is delayed in the absence of Rtt109, but the effect is weaker for PHO84 induction and not as pronounced there in an asf1 strain. (A) PHO5 promoter induction kinetics as in Fig. but for wt (W303a), (more ...)
The PHO84 promoter chromatin structure on the plasmid underwent the same regulated transition as the endogenous chromosomal locus (compare Fig. and for DNase I mapping; data not shown for restriction enzyme accessibilities). It should be noted that the region far upstream of the PHO84 promoter, which is used for probing in indirect end-labeling techniques, was different between the plasmid and the chromosomal locus, thus allowing for a distinction of both loci within the same cell by differential probing and therefore excellent internal control. Due to the different relative position of the secondary cleavage site at the plasmid and chromosomal locus, the DNase I indirect end-labeling fragments at the plasmid locus were 214 bp smaller, leading to a more stretched out appearance of the plasmid chromatin patterns on the blot. Possible minor changes in nucleosome positions between the chromosomal and the plasmid locus could still be undetected by this low-resolution approach.
Using this reporter plasmid, a set of promoter variants similar to the ones of Ogawa et al. (54
) was constructed: a truncated version, plasmid pCB84b, in which effectively the upstream nucleosome and UASpA and UASpB were deleted (ΔΔUASpAB [schematic in Fig. ]), and point mutants for either one of the Pho4 binding sites, UASpC, UASpD, and UASpE, or for two sites together, i.e., UASpCEmut or UASpDEmut. For the truncated promoter the proper positioning of the downstream nucleosome in the repressed state and the generation of the corresponding extended hypersensitive region (truncated eHS type) upon induction were confirmed by DNase I indirect end labeling (data not shown).
Induction of the truncated promoter ΔΔUASpAB as monitored by acid phosphatase activity was very similar to the wt promoter (Fig. ). Mutation of the accessible high-affinity sites, UASpC or UASpD, affected the final promoter activity rather slightly, with the effect of the UASpD mutation being a bit more pronounced (Fig. ). In contrast, the absence of the intranucleosomal low-affinity site, UASpE, had a much stronger effect, reducing the final promoter strength by more than 50%. The combination of mutations in the UASpE and either UASpC or UASpD sites drastically reduced the final promoter activity to about 25% and 15% of the wt activity, respectively. We conclude, in agreement with Ogawa et al. (54
), that the contribution of UASpC and UASpD was redundant, whereas UASpE contributed about half the promoter activity by itself. Further, there was some cooperativity between the intranucleosomal UASpE site and the accessible site UASpD and maybe also UASpC, as the effects of the double mutants were larger than the sum of the effects of each single mutant.
Next we examined if the effects on promoter strength were a consequence of inefficient promoter chromatin remodeling or of an effect downstream of chromatin opening. The DNase I indirect end-labeling patterns under inducing conditions of the UASpCmut or UASpDmut promoter variants were the same as for the wt promoter (data not shown), which was in agreement with a rather slight effect of these mutations on promoter activity. The finding that one UASp element in the sHS linker was sufficient for full remodeling of the upstream and downstream nucleosome is similar to the PHO8
but different from the PHO5
promoter, where the linker site UASp1 alone was not sufficient for chromatin remodeling (25
). This may be because UASp1 at the PHO5
promoter is a low-affinity binding site, in contrast to the high-affinity linker sites at the PHO84
Any promoter variant lacking UASpE showed a hypersensitive region under inducing conditions that was less extensive in the downstream direction (eHS*) (Fig. , schematic). This was especially clear in the DNase I patterns of the induced UASpCEmut and UASpDEmut promoter variants (Fig. and data not shown), in which the extended hypersensitive region (eHS*) extended only up to about the SpeI marker band (−259 bp) (Fig. ), which was introduced with the UASpEmut point mutation and marked therefore the position of UASpE. In contrast, the eHS region of the induced wt promoter pattern reached further downstream beyond the AgeI marker (bp −172) (Fig. and ). This less-extensive eHS* region was less clearly visible in the DNase I pattern of the UASpEmut variant (Fig. ), but less extensive remodeling downstream of the SpeI site was confirmed also for this variant by a reduced final accessibility of the AgeI site (Fig. , table). We concluded that UASpE is essentially required for remodeling of the region between the downstream nucleosome and the TATA box.
Gcn5 is not essential for PHO84 promoter remodeling, but its absence causes a strong delay in histone eviction kinetics and concomitant promoter induction.
Previously, we found that remodeling of the chromatin structure at the weak PHO8
promoter was critically dependent on Gcn5 and Snf2 (28
). At the stronger PHO5
promoter only the rate of chromatin remodeling was strongly decreased in the absence of Gcn5 or Snf2 (6
), but eventually full remodeling was achieved. We wondered if remodeling at the even stronger PHO84
promoter would be mostly or even fully independent of the presence of these cofactors.
First, we examined induction kinetics of the PHO84 promoter in gcn5 cells and found a strong delay in comparison to wt cells, even though the final induction level was unaffected (Fig. ). In agreement with this, the DNase I pattern of the fully induced promoter in the gcn5 mutant was the same as observed in wt cells (Fig. ). Therefore, the Gcn5 activity had no essential role for the final opening of the PHO84 promoter chromatin. This was confirmed further by restriction enzyme analysis of DNA accessibility at the entire promoter region under fully inducing conditions (Fig. , −Pi).
In analogy to our earlier findings at the PHO5
), we assumed that the kinetic delay on the activity level in the gcn5
mutant (Fig. ) was caused by a delay in the chromatin remodeling step. We quantified chromatin opening for wt and gcn5
cells by restriction enzyme accessibility at 1.5 h after shift to phosphate-free medium and by histone H3 ChIP during an induction time course. To our surprise, we did not catch much of a delay in the increase of restriction enzyme accessibility at this time point of induction. There was only a slight delay compared to wt in opening at the AgeI site, i.e., in the region between the downstream nucleosome and the TATA box (Fig. , .5 h, −Pi
). For comparison, chromatin remodeling at the PHO5
promoter, as probed by ClaI accessibility, was still strongly delayed after 3 hours of induction in a gcn5
). Nonetheless, we did observe a strong delay in histone eviction kinetics as monitored by histone H3 ChIP (Fig. ). Even after 2 hours of induction, there was six to seven times more histone H3 still present at the promoter in the gcn5
mutant than in the wt cells. Therefore, we observed for the first time a large disparity between restriction enzyme accessibility and histone H3 eviction kinetics during induction of a PHO promoter. We conclude that histone eviction, rather than an initial increase of DNA accessibility, appeared to be the rate-limiting step in PHO84
promoter opening in a gcn5
In the absence of Snf2, remodeling of the PHO84 promoter chromatin structure is only partial: the downstream nucleosome is fully remodeled but the upstream one is not at all.
Second, we examined PHO84
promoter induction kinetics in a snf2
mutant and observed a similar delay as with the gcn5
mutant, again with hardly any effect on the final level of induction (Fig. ). In marked contrast and much to our surprise, this final activity of the snf2
strain corresponded to an only partially open DNase I pattern of the induced PHO84
promoter, both on the chromosomal and the plasmid locus (Fig. and data not shown). The downstream nucleosome was remodeled, but the upstream one was not at all. In addition, we noticed that the spreading of the eHS region was less extensive in the downstream direction than in the wt case (eHS**) (Fig. , schematic) and confirmed this by a reduced final accessibility of the AgeI and PacI sites (Fig. , −Pi
). This reduced downstream spreading of the eHS** region was similar to the reduced spreading of the eHS* region in the UASpEmut variant (Fig. ). It was even somewhat more severe, as also the PacI site accessibility was reduced in the eHS** but not in the eHS* region (Fig. and , tables). Even though the eHS** region in the snf2
mutant was less remodeled than the eHS* region in the UASpEmut variant, it was still compatible with full final activity levels (Fig. ). So, we concluded that the lower final activity in the UASpEmut, and even more so in the UASpCEmut and UASpDEmut variants (Fig. ), was less due to compromised chromatin remodeling but mainly due to the reduced number of UASp elements (see also reference 41
). As the transition from the semiopen to the fully open state in the region between the downstream nucleosome and the TATA box was compromised in both the snf2
mutant and the UASpEmut variant, we suggest that recruitment of the SWI/SNF complex by UASpE-bound Pho4 was essential for chromatin remodeling in this region.
Restriction enzyme probing of the induced state in the snf2 mutant also confirmed the lack of remodeling of the upstream nucleosome, i.e., persistently low HhaI accessibility, and full remodeling of the downstream nucleosome, i.e., high TaqI accessibility (Fig. , −Pi). Altogether, this chromatin pattern constituted a third type of extended hypersensitive region (eHS**) (Fig. , schematic), where the upstream nucleosome was still present, the downstream nucleosome fully remodeled, and the region between the downstream nucleosome and the TATA box not fully remodeled.
The same partially remodeled DNase I pattern was also observed in the snf2K798A strain, which bears a point mutation in the Snf2 ATPase domain (Fig. ), confirming that the ATPase activity of Snf2 rather than some other feature of the SWI/SNF complex was responsible for the observed effect.
In analogy to the gcn5 mutant, we examined whether the kinetic delay of PHO84 promoter induction in the snf2 mutant (Fig. ) corresponded not only to the aforementioned reduction in the final extent of remodeling but also to a kinetic delay of chromatin opening, for example, at the TaqI site in the downstream nucleosome. After 1.5 h of induction there was not much delay in opening of the TaqI or any other site, based on the 1.5-h values for the snf2 strain compared to wt and normalized to their respective −Pi values (Fig. ). However, histone eviction kinetics measured by histone H3 ChIP in snf2 cells showed a strong delay (Fig. ). At present we are unsure why the final level of histone occupancy at the induced PHO84 promoter in snf2 cells as measured by histone H3 ChIP was not much higher than for the wt and gcn5 strains. This would be expected due to the continued presence of the upstream nucleosome in the snf2 strain. The resolution of our ChIP analysis (about 500 bp) cannot distinguish between the upstream and the downstream nucleosome, because the amplicon used (Fig. , schematic) will score fragments from both nucleosome regions. However, as the upstream nucleosome was not remodeled at all and as the downstream region close to the TATA box was remodeled to a lesser extent than in the wt (see above), we assume that histone H3 ChIP mainly monitored remodeling of the downstream nucleosome. Therefore, the delayed histone eviction in the snf2 mutant argues for a role of Snf2 in remodeling of the downstream nucleosome. Similar to the case of the gcn5 mutant, also here histone eviction seemed to be the rate-limiting step.
As remodeling of the downstream nucleosome was eventually complete but kinetically delayed at the histone eviction step in both the snf2 and gcn5 single mutants, we wondered if the downstream nucleosome may not open up at all in a snf2 gcn5 double mutant. This was not the case, as the DNase I pattern of the fully induced PHO84 promoter in the snf2 gcn5 double mutant was indistinguishable from that found in snf2 cells (Fig. ).
Previously, it was shown by us and others that submaximal induction conditions can exacerbate the dependency of PHO5
promoter chromatin remodeling on chromatin cofactors (19
). Such submaximal induction conditions may be achieved by using low-phosphate rather than phosphate-free medium (19
) or by overexpression of Pho4 in high-phosphate medium (25
). We tested under the latter conditions whether the differential requirement of Snf2 for remodeling of the downstream and the upstream nucleosome still persisted at submaximal induction. DNase I indirect end-labeling analysis under these submaximal induction conditions showed the same pattern as under fully inducing conditions, for both the wt as well as the snf2K798A
mutant (Fig. ). So, even at such low induction levels the downstream nucleosome could be remodeled without Snf2 activity, demonstrating further the different degree of Snf2 requirement for remodeling of the upstream and downstream nucleosome.
The semiopen chromatin structure close to the TATA box is not sufficient, and basal remodeling of the upstream nucleosome is not necessary for substantial basal PHO84 transcription.
, and gcn5
mutants all had a decreased basal level of transcription (Fig. and data not shown) (69
). In all these three mutants the semiopen less-organized chromatin structure between the downstream nucleosome and the TATA box was not affected in the repressed state. Therefore, this semiopen structure was not sufficient for sustaining substantial basal transcription under repressing conditions.
Nonetheless, in all three mutants the accessibility of the HhaI site under repressing conditions was reduced in comparison to wt, in snf2 and gcn5 cells even more so than in the pho4 mutant (Fig. , +Pi, and 1D, table). The reduced HhaI accessibility might have been responsible for the reduced basal transcription. In the wt, the targeted recruitment of Snf2 and Gcn5 by Pho4 could keep the upstream nucleosome in a partially remodeled state, which would allow partial access to UASpB and lead to even more remodeling of the upstream nucleosome and high basal transcription. To test this, we introduced a point mutation in UASpB and found indeed that the HhaI site accessibility under +Pi conditions (19 ± 2%) was significantly lower than at the wt promoter and similar to that of the wt promoter in the pho4 mutant (17 ± 2%) (Fig. ). However, despite this lower HhaI accessibility there was hardly any effect on the basal level of activity for the UASpBmut construct (data not shown), arguing that UASpB and basal remodeling of the upstream nucleosome were not necessary for the substantial basal transcription. In addition, mutation of the other intranucleosomal site, UASpE, which analogously may have been involved in basal remodeling of the downstream nucleosome, did not affect basal transcription either (Fig. ).
Ino80 is not essential for chromatin opening at the entire PHO84 promoter, neither in wt nor in snf2 cells, but its absence causes a strong delay in chromatin opening kinetics.
As we had already observed a cooperation between Snf2 and Ino80 for chromatin remodeling at the PHO5
), and as others have shown a recruitment of both Snf2 and Ino80 to the PHO84
promoter upon induction (23
), we investigated the role of Ino80 for PHO84
promoter opening. In particular, there was the possibility that Ino80 would be the alternative remodeler for remodeling of the downstream nucleosome in the absence of Snf2.
The absence of Ino80 by itself did not prevent full remodeling of the PHO84 promoter chromatin structure, i.e., the DNase I pattern of an ino80 mutant under fully inducing conditions corresponded to the eHS type of the wt (Fig. ) and the accessibility of restriction enzymes along the promoter region increased to almost-wt levels (Fig. ). Further, the DNase I pattern of the induced promoter in the snf2 ino80 double mutant was indistinguishable from the pattern of the snf2 single mutant (Fig. ). Together, these results argue that Ino80 was neither essentially required for remodeling under fully inducing conditions in the wt strain nor for remodeling of the downstream nucleosome in the absence of Snf2. Nonetheless, the chromatin opening kinetics in the ino80 strain was strongly delayed over the entire promoter region after 1.5 h of induction as examined by restriction enzyme accessibility (Fig. ). Therefore, Ino80 is clearly involved in the wt chromatin remodeling pathway at the PHO84 promoter.
FIG. 6. Chromatin remodeling at the PHO84 promoter is delayed in the absence of Ino80. (A) DNase I indirect end-labeling analysis of the induced PHO84 promoter chromatin structure for an ino80 strain (BY4741-1). Labeling is as for Fig. and (more ...)
In contrast to Snf2 and Gcn5, Ino80 was not involved in keeping the upstream nucleosome in a partially remodeled state under repressing conditions (+Pi), as the HhaI accessibility was not affected in the ino80 mutant (Fig. , table, +Pi). A slight decrease in PacI accessibility may indicate that Ino80 has a minor role in positioning the downstream nucleosome under repressing conditions.
As presented above for the case of Snf2, we checked if PHO84 promoter opening became more dependent on Ino80 under submaximal conditions. Strikingly, the DNase I patterns of the snf2K798A and the ino80 mutants at submaximal induction were indistinguishable, i.e., under these conditions the upstream nucleosome became strictly dependent also on Ino80 (Fig. ).
The stricter cofactor requirements for remodeling of the upstream nucleosome correlates with higher intrinsic stability as measured in vitro and predicted in silico.
As shown above, remodeling of the upstream nucleosome was strictly dependent on Snf2, whereas remodeling of the downstream nucleosome was not (Fig. and ). In addition, remodeling of the upstream nucleosome was more dependent on Ino80 than remodeling of the downstream nucleosome (Fig. ). This constitutes a case of differential cofactor requirements for nucleosome remodeling within one and the same promoter.
We found earlier that the differential cofactor requirements for chromatin remodeling at the PHO5
promoters correlated with differential intrinsic stabilities of the positioned nucleosomes (31
). These stabilities were measured using our yeast extract chromatin assembly system that is able to generate the proper in vivo nucleosome positioning de novo in vitro (31
). In this system, plasmids bearing the yeast locus of interest are assembled by salt gradient dialysis into a chromatin structure with a specific but usually not proper, i.e., not in vivo-like, nucleosome positioning pattern. The in vivo-like pattern is induced in the next step by the addition of yeast whole-cell extract in the presence of energy. A so-far-unidentified energy-dependent activity in the yeast extract apparently constitutes the thermodynamic conditions for in vivo-like nucleosome positioning. In a next step, it is possible to compare the intrinsic stability of properly positioned nucleosomes by titrating the histone concentration. Under conditions of limiting histones (underassembled chromatin) there are fewer nucleosomes deposited onto the DNA than there are nucleosome positions available. Therefore, the multitude of alternative and mostly overlapping nucleosome positions will compete for nucleosome occupancy. Positions that are already occupied in equilibrium in underassembled chromatin are more stable than those that are occupied only in fully assembled chromatin (for a full discussion of this methodology see reference 31
). Using this approach, we observed previously that the proper positioning over the PHO5
promoter region could only be generated in fully assembled chromatin, whereas the proper PHO8
promoter pattern was also achieved in underassembled chromatin. Therefore, the intrinsic stability of the PHO8
promoter nucleosomes was higher than the stability of the PHO5
With the same methodology we compared the intrinsic stability of the upstream and downstream nucleosome at the PHO84
promoter (Fig. ). First, we prepared fully assembled salt gradient dialysis chromatin (histone octamer:DNA mass ratio set as 100%) using a plasmid with a 3.5-kb PHO84
insert as template and tested if the yeast extract would generate the in vivo pattern. Much to our surprise, we observed that the DNase I pattern of the salt gradient dialysis chromatin was already very similar to the in vivo pattern (Fig. , compare SGD and in vivo). This pattern was clearly different from a digest of free DNA and did not change much, as expected (31
), with the addition of yeast extract in the absence of energy. This was the first case out of 14 tested yeast loci (C. Wippo and P. Korber, unpublished results) where salt gradient dialysis by itself was already able to generate a very in vivo-like chromatin structure. This suggests that rather strong nucleosome positioning sequence elements in the PHO84
promoter lead to in vivo-like nucleosome positioning already under pure salt gradient dialysis conditions. Nonetheless, incubation with yeast extract and energy did make the pattern more similar to the in vivo pattern, especially regarding the relative band intensities and the upper part of the lane, i.e., the coding region (Fig. , compare SGD +Yex/ATP with in vivo). Therefore, the PHO84
promoter is one more example where our yeast extract in vitro assembly system constitutes conditions more similar to in vivo conditions for nucleosome positioning than salt gradient dialysis alone.
Second, we repeated the salt gradient dialysis chromatin assembly with limiting histones (histone octamer:DNA mass ratio of 60%) and still obtained a rather in vivo-like pattern (Fig. ). This in vivo-like pattern again did not change upon the addition of yeast extract without energy. However, incubation with yeast extract in the presence of energy, i.e., conditions that should be closer to the in vivo conditions, had a differential effect on the regions upstream and downstream of the sHS region. The upstream nucleosome and the cHS region again became even more like the in vivo pattern, but the sHS region was so much extended further downstream that the position of the downstream nucleosome was compromised. The sHS region was always somewhat sharper in the pure salt gradient dialysis chromatin pattern and became fuzzier upon addition of yeast extract and energy, also with fully assembled chromatin templates (Fig. ; compare widths of brackets). But whereas the more fuzzy sHS region in the fully assembled chromatin (100%) resembled more the in vivo case, it was stretched too far downstream to be compatible with a proper positioning of the downstream nucleosome for the underassembled chromatin templates (60%) (Fig. ; compare widths of brackets). We stress that the more extensive sHS region under underassembled conditions compared to fully assembled conditions (Fig. ) was not due to the use of different degrees of DNase I digestion, as we saw such a difference significantly and repeatedly over a wider range of DNase I digestions (data not shown).
This differential effect on the upstream and downstream nucleosome was confirmed by restriction enzyme accessibility assays. The accessibility of the TaqI site in the downstream nucleosome increased much more (from 15% to 69%) (Fig. ) upon addition of extract and energy to underassembled chromatin than the accessibility of the HhaI site in the upstream nucleosome (from 10% to 36%). The overall lower accessibilities in the fully assembled chromatin compared to the in vivo situation probably reflected here a subpopulation of aggregated, i.e., indigestible, templates in vitro, which may form especially at such high histone concentration. Altogether, these results suggested that the downstream nucleosome was intrinsically less stably positioned in vivo than the upstream nucleosome. This correlated with its more relaxed cofactor requirements.
The finding of higher intrinsic stability of the upstream nucleosome also correlated strikingly with the prediction of the N
-score algorithm (84
) (Fig. ). The N
-score algorithm was trained on in vivo yeast nucleosome positioning data and used to predict the probability for nucleosome occupancy (positive values) or depletion (negative values) rather than exact positions. It showed a positive peak right in the middle of the upstream nucleosome, maybe suggesting an especially stable nucleosome here in vivo. In contrast, the DNA sequence underlying the downstream nucleosome was rather neutral, or even negative at its 3′ end, with regard to the propensity for nucleosome occupancy.
Introduction of destabilizing mutations into the DNA sequence of the upstream nucleosome relieves the Snf2 dependency for its remodeling in vivo.
So far, we correlated, in this and our previous study (31
), intrinsic nucleosome stability and the cofactor requirement. Next we wished to test directly if stability was causative for requirement. Extended stretches of poly(dA-dT) are known to be unfavorable for nucleosome formation in vivo and in vitro (4
). So we replaced a stretch of 10 or 19 consecutive bases with adenine deoxynucleotides (plasmids pCB84a-10A and −19A, respectively) in the middle of the upstream nucleosome region (Fig. ). As expected, such replacements led to increasingly more negative N
-scores for the region that was occupied by the upstream nucleosome in the wt promoter (Fig. ).
We needed to check if the upstream nucleosome would still form in vivo on these mutated DNA templates. DNase I mapping confirmed the presence of the upstream nucleosome for both variants in the wt and snf2 backgrounds (Fig. and data not shown). Restriction enzyme accessibility assays showed that there was no increase in HhaI site accessibility for the 10A replacement compared to the wt promoter (data not shown) but an increase for the 19A variant (from 25 to 40% in wt and from 15 to 48% in the snf2 background) was observed (Fig. ). This suggested a destabilized upstream nucleosome for the 19A variant already under repressive conditions. There was also a subtle shift in positioning as the sHS region extended more upstream beyond the ApaI marker (compare Fig. and ). This region of additional hypersensitivity at the 3′ border of the upstream nucleosome correlated with the region of the most negative N-score at about −550 (Fig. ).
The reduced stability of the 19A variant was directly assessed in our in vitro chromatin assembly assay (Fig. ). First, the upstream nucleosome formed neither with a limiting (60%) nor with the full (100%) complement of histones during salt gradient dialysis, but the DNase I pattern in this region was similar to that of the free DNA digest. This speaks for the lower nucleosome positioning power of the mutated DNA sequence under these conditions. Second, the addition of yeast extract and energy induced a more in vivo-like chromatin structure in the fully assembled (100%) chromatin template, with accessibilities for the HhaI and TaqI sites that were very similar to the in vivo values (Fig. ; compare 19A in the wt background [D] and 100% with yeast extract and energy [E]). This confirmed again that the unidentified energy-dependent activity in the yeast extract constitutes conditions for more in vivo-like nucleosome positioning. Third, addition of yeast extract and energy to the underassembled (60%) chromatin templates increased not only the TaqI site accessibility (from 22 to 66%) (Fig. ), similar as seen before for the wt promoter (from 15 to 69%) (Fig. ) but now also the HhaI site accessibility (from 47 to 73%). This argued for a low stability of both the upstream and downstream nucleosome.
Finally, both variants showed remodeling of the upstream nucleosome upon induction in a snf2 strain. The extent of remodeling as judged by DNase I indirect end labeling was substantial for both variants in comparison to the internal control of the wt promoter at the chromosome locus (Fig. ) and to the plasmid locus (data not shown). HhaI site accessibility assays confirmed a partial remodeling for the 10A variant and almost full remodeling for the 19A replacement variant (Fig. ). Altogether, these results argue strongly that the intrinsic stability of the upstream nucleosome in the wt promoter caused its strict Snf2 requirement for remodeling.
The destabilization or complete absence of the upstream nucleosome relieves the Snf2 dependency of promoter induction only partially.
In addition to the mechanistically interesting relationship between intrinsic stability and Snf2 dependency of remodeling of the upstream nucleosome, we asked further if the critical Snf2 dependency of remodeling the upstream nucleosome was the main cause for the Snf2 effect on overall PHO84 promoter induction kinetics (Fig. ). If so, the kinetic delay in a snf2 background should be reduced if the upstream nucleosome is destabilized (19A variant, plasmid pCB84a-19A) or absent (ΔΔUASpAB variant, plasmid pCB84b). We followed induction kinetics for both variants in the wt and snf2 backgrounds by acid phosphatase assay and compared them to the kinetics of the wt promoter in both backgrounds (Fig. ). For both variants the delay of induction in the snf2 mutant compared to the wt background was somewhat diminished, more so in the case of the truncated promoter and only very slightly in the case of the mutated promoter. This was more apparent after normalization of the phosphatase activity in the snf2 strains to the respective activity in the wt background at the same time points (Fig. ). Nonetheless, as the delay in the snf2 strains was still substantial in both cases, we reasoned that there was a significant Snf2 dependency of other parts of the PHO84 promoter besides the upstream nucleosome. For example, we showed specifically that the kinetics of remodeling the downstream nucleosome was dependent on Snf2, as histone eviction of the wt promoter was delayed in the snf2 mutant (Fig. ) (see above).
Since the HhaI accessibility of the PHO84 promoter variant in pCB84a-19A was considerably increased under repressive conditions in a snf2 strain (Fig. ) but did not result in a higher basal level of transcription (data not shown), it seemed again (see above) that Snf2 had an effect on basal transcription that was not necessarily linked to basal remodeling of the upstream nucleosome.
The histone acetyltransferase Rtt109 has a role for induction of both the PHO84 and the PHO5 promoters.
We and others found that the histone chaperone Asf1 is involved in the induction of the coregulated PHO5
). Recently, several groups reported the critical requirement of Asf1 for the activity of the histone acetyltransferase Rtt109, which acetylates histone H3 at lysine 56 (18
). This finding raised the question of whether an involvement of Asf1 reflects its role solely as histone chaperone or rather a role of Rtt109. We checked this for induction of the PHO5
promoter and observed that the delay in induction was virtually the same in the asf1
mutants and that there was no further delay in an asf1 rtt109
double mutant (Fig. ). This argued strongly that Asf1 and Rtt109 function together in the same pathway during PHO5
induction. We also noted that for both the asf1
mutant as well as the rtt109
mutant the basal PHO5
activity levels were slightly but significantly elevated.
In contrast, induction of PHO84 was significantly delayed only in the rtt109 but hardly at all in the asf1 mutant (Fig. ). The induction delay in the rtt109 mutant was due to a delay on the level of chromatin remodeling as monitored by restriction enzyme accessibility and histone ChIP assays (Fig. , D, and E). However, the effects were much less severe than those in the snf2, gcn5, or ino80 mutants (compare to Fig. and ), especially as they were rather limited to an early time of induction (45 min). There was hardly any effect on the level of restriction enzyme accessibilities for the asf1 mutant, and only at 45 min of induction was there a slight delay in histone eviction. This may constitute a weaker pendant to the effects in the gcn5 and snf2 strains, i.e., histone eviction being the rate-limiting step.
There was no differential Rtt109 requirement of the upstream and downstream nucleosome discernible, as the kinetics of restriction enzyme site accessibility were similarly delayed for the HhaI and the TaqI sites in the rtt109 mutant (Fig. ). We also checked the effects of the asf1 and rtt109 deletions on induction of the truncated pCB84b construct and got similar results as with the full-length pCB84a plasmid (Fig. ), speaking for a role of Rtt109 in remodeling of the downstream nucleosome but not excluding a role for remodeling of the upstream nucleosome as well.
The effects of the asf1 and rtt109 deletions on PHO5 and PHO84 induction showed some dependency on the strain background. In the BY4741 background, the rtt109 mutant showed a weaker delay for PHO5 induction than the asf1 mutant (data not shown). In the W303 background, the rtt109 mutant had a similar effect on PHO84 induction as in the BY4741 background, but here also the asf1 mutant had an appreciable effect, similar to that of the rtt109 mutant (data not shown).
It was shown that Rtt109 exists in a complex with another histone chaperone, Vps75 (78
); however, the absence of Vps75 caused hardly any effect on PHO5
induction (data not shown).