ABSTRACT
Two transcription factors, the bHLH protein Pho4 and the homeodomain protein
Pho2, are required for transcriptional activation of the
PHO5
promoter in
Saccharomyces cerevisiae
. There are two essential Pho4 binding sites, corresponding to the regulatory
elements UASp1 and UASp2 at the
PHO5
promoter, but only a single, dispensable Pho2 binding site had previously been
identified. We have reinvestigated binding of Pho2 to the
PHO5
promoter using purified recombinant protein and have found multiple Pho2
binding sites of different affinities along the promoter. One of the high
affinity Pho2 sites largely overlaps the Pho4 binding site at UASp1.
Cooperative DNA binding of the two proteins to their overlapping sites,
resulting in a high-affinity ternary complex, was demonstrated. Pho2 and Pho4 also bind DNA
cooperatively at UASp2 where two Pho2 sites flank the Pho4 site. Finally, Pho2
facilitates binding of Pho4 to a third, cryptic Pho4 binding site which binds
Pho4 with lower affinity than UASp1 or UASp2. These results suggest that
cooperative DNA binding with Pho4 is integral to the mechanism by which Pho2
regulates transcription of the
PHO5
gene.
When the yeast
Saccharomyces cerevisiae
is grown under conditions of phosphate limitation, production of acid phosphatase is increased dramatically as a result of transcriptional activation of the
PHO5
gene (
1
). Genetic studies of the
PHO5
system revealed that two transcription factors, the basic-helix-loop-helix protein Pho4 and the homeodomain protein Pho2, are required
for the induction of the
PHO5
promoter (
2
). The activity of Pho4 is regulated in response to phosphate levels through phosphorylation by a cyclin-CDK complex, encoded by
PHO80
and
PHO85
, respectively. When phosphate is abundant, Pho4 is inactivated by phosphorylation, while upon phosphate starvation, the positive factor Pho81 prevents phosphorylation of Pho4 by inhibiting the Pho80-Pho85 complex (
3
). Recent data suggest that phosphorylation of Pho4 affects its subcellular localization with the phosphorylated form being localized in the cytoplasm and thereby unable to activate
PHO5
transcription (
4
).
Deletion analysis of the
PHO5
promoter demarcated two regulatory elements, UASp1 and UASp2 (
5
), to which Pho4 has been shown to bind
in vitro
(
6
).
In vivo
footprinting experiments revealed that Pho4 binds to both sites upon phosphate
starvation, but not at high phosphate conditions (
7
). In addition, Pho4 has been shown to bind to promoters of other genes which are regulated by phosphate, including
PHO8
(
8
),
PHO10
and
PHO11
(
9
),
PHO81
(
10
) and
PHO84
(
11
).
In contrast with the clear picture for Pho4, the role of Pho2, the other
activator of
PHO5
regulation, has proved more difficult to define. Pho2 is a homeodomain protein
which has been shown to regulate a diverse array of genes, including not only
PHO5
, but also the
HO
gene (
12
), the
HIS4
(
13
),
TRP4
(
14
) and the
ADE1
,
ADE2
,
ADE5
,
7
and
ADE8
genes (
15
). In each case, a Pho2 binding site has been demonstrated in the promoter region
in vitro
. In the
PHO5
promoter, a single Pho2 binding site was mapped between the two Pho4 binding
sites by
in vitro
footprinting (
6
). However, the role of Pho2 in
PHO5
regulation has remained enigmatic, because deletion of the Pho2 binding site
did not influence
PHO5
promoter activity significantly (
5
). In addition, activation of a heterologous promoter by a 31 bp sequence
containing UASp1 is fully Pho2-dependent, even though there was no evidence for Pho2 binding to this
element (
16
). Finally, although Pho2 is required for
PHO5
promoter activation, overexpression of Pho4 can give rise to a limited
activation of the
PHO5
promoter in the absence of Pho2 (
17
).
For some of the Pho2-dependent genes, there are indications of a role for Pho2 in interacting
with other transcription factors. At the
HO
promoter, a Pho2 binding site is located next to a Swi5 binding site, and it
was shown that the two proteins bind to their sites cooperatively (
12
). In the case of the
HIS4
promoter, a Pho2 (Bas2) protected region largely overlaps the Bas1 footprint.
Although Pho2 and Bas1 can bind to this region simultaneously, no cooperative
interactions between the two proteins were detected (
13
). In contrast, at the
TRP4
promoter, a Pho2 binding site completely overlaps one of the two Gcn4 binding
sites, and the two proteins were found to bind DNA in a mutually exclusive manner (
14
). Recent experiments have demonstrated an interaction between Pho2 and a specific domain of Pho4 using the yeast two hybrid system (
18
), raising the possibility of an additional role of Pho2 also at the
PHO5
promoter.
In an attempt to resolve some of the questions concerning the mechanism by which
Pho2 contributes to
PHO5
regulation, we have reinvestigated Pho2 binding at the
PHO5
promoter and examined the possibility of Pho2-Pho4 interactions. Our results show that there are multiple Pho2 binding
sites at the
PHO5
promoter. One of them significantly overlaps with the Pho4 binding site at
UASp1, while another two sites flank the second Pho4 binding site, UASp2. In
addition, a cryptic Pho4 binding site has been mapped 60 bp downstream of UASp2
with a Pho2 binding site at its 3' side. Binding studies reveal that Pho2 binds cooperatively with Pho4 at
each Pho4 binding site, providing a mechanism that explains the absolute
requirement of Pho2 for
PHO5
promoter activation.
The
PHO2-HIS
expression plasmid has been described previously (
19
).
The
PHO4-HIS
expression plasmid was created from a plasmid in which an
Nco
I site had been inserted at the
PHO4
start codon (provided by A. Hinnen). An
Nco
I-
Bbr
PI fragment of this plasmid was cloned into
Nco
I-
Eag
I-digested pET-21d (Novagen). This adds 10 amino acids to the C-terminus of Pho4, the last six of which are the histidine tag.
PHO5
DNA restriction fragments used for DNase I footprinting and gel shift analyses were derived from
PHO5
-
lacZ
fusions (
20
) containing the wild-type promoter as well as from subcloned derivatives providing restriction
sites not present in the wild-type promoter.
Purification of the Pho2-HIS fusion protein has been described previously (
19
,
21
). Pho4-HIS protein was similarly purified using Qiagen Ni
2+
-NTA
-
agarose except that the column was washed with buffer containing 60 mM imidazole
prior to eluting with buffer containing 1 M imidazole. Peak fractions
containing Pho4-HIS or Pho2-HIS were pooled and dialyzed against 20 mM Tris-HCl, pH 8.0, 100 mM NaCl, 0.5 mM DTT, 10% glycerol, 0.5 mM
EDTA, 0.2 mM PMSF, 1 mM benzamidine. As judged by SDS-gel electrophoresis, both proteins were highly purified. A monoclonal
antibody against the histidine tag was obtained from Dianova, Hamburg, Germany.
Purified Pho2-HIS and/or Pho4-HIS were incubated with labeled DNA fragments (~15 000 c.p.m.) for 30 min at room temperature in 25 [mu]l of 10 mM Tris-HCl, pH 7.4, 50 mM NaCl, 5 mM MgCl
2
, 1 mM DTT and 10 [mu]g/ml poly(dI-dC). After the incubation, 25 [mu]l of 1 U/ml DNase I, diluted in the same reaction buffer
containing in addition 1 mM CaCl
2
, was added and the reaction was incubated at room temperature for 1-2 min. The reaction was stopped by addition of 50 [mu]l of stop solution containing 0.4 M NaCl, 0.4% SDS and 25 [mu]g/ml of salmon sperm DNA. The DNA was precipitated with ethanol and analyzed on 6 or 8% polyacrylamide-8 M urea gels.
DNA fragments were labeled with [[gamma]-
32
P]ATP (6000 Ci/mmol) using polynucleotide kinase, and ~5000 c.p.m. of labeled DNA was used per binding reaction. Protein-DNA binding reactions (10 [mu]l) were performed in 15 mM Tris-HCl, pH 8.0, 75 mM NaCl, 7.5% glycerol, 12.5 mM DTT, 0.375 mM EDTA, 750 [mu]g/ml bovine serum albumin and 25 [mu]g/ml poly(dI-dC). The amount of protein used in each
binding reaction is indicated in arbitrary units in the figure legends. One
unit of Pho4 and Pho2 corresponds to ~5 and 6 ng protein, respectively, as determined by SDS-gel electrophoresis. Reactions were incubated at room temperature
for 30-45 min before they were loaded onto a 5% polyacrylamide gel made in TBE
buffer (45 mM Tris, 45 mM boric acid, 1.25 mM EDTA, pH 8.3). Electrophoresis was in the same buffer for ~1.5 h at 250 V, then the gel was dried and autoradiographed with an intensifying
screen.
In order to look for additional Pho2 targets in the
PHO5
promoter, we have reinvestigated binding of Pho2 which was purified by the Hexa-HIS-tag method. DNA fragments encompassing the two Pho4 binding sites,
UASp1 and UASp2, were examined by
in vitro
DNase I footprinting. As shown in Figure
1
, several protected regions are seen on both the upper and the lower strand in the presence of added Pho2.
Two footprints, separated by a short unprotected region, are found between
UASp1 and UASp2 from -271 to -285 and from -291 to -320 (Fig.
1
A). This area contains the previously determined Pho2 binding site (-277 to -296) (
6
). In addition to these two regions, Pho2 protection is observed from -358 to -385 (Fig.
1
B), i.e. over a sequence partially overlapping UASp1. A stepwise decrease in the
Pho2 concentration results in the simultaneous loss of Pho2 protection at all
three regions, suggesting similar affinities of Pho2 for these binding sites (Fig.
1
A and B).
Pho2 protection (and enhancement) is also observed upstream of UASp1 as well as
downstream of UASp2. To map these sites more precisely, suitable DNA fragments
were examined. As shown in Figure
1
C, the region upstream of UASp1 contains multiple closely adjacent sites
protected by Pho2, in addition to the one partially overlapping UASp1. Two
sites are found downstream of UASp2 from -223 to -236 and -169 to -182 and an additional protected area around -110 (Fig.
1
D).
As summarized below in Figure
9
, Pho2 can bind with different affinities (see below) to multiple sites between
the TATA box and the upstream
Bam
HI site. One of the strong Pho2 binding sites partly overlaps the previously
determined Pho4 binding site at UASp1, which is consistent with a recent report
that Pho2 from a yeast extract can bind to an oligonucleotide containing the UASp1 sequence (
22
). In addition, there are two Pho2 binding sites that flank UASp2. It was of
interest, therefore, to determine if Pho4 and Pho2 can simultaneously bind to
UASp1 as well as to UASp2, and if binding of one protein influences binding of
the other.
We first examined binding of Pho2 and Pho4 to a DNA fragment containing UASp1
(Fig.
2
). The two proteins appear to be able to bind simultaneously to UASp1, even
though their individual footprints overlap significantly. Eleven out of 19
nucleotides protected by Pho4 on the bottom strand are also protected by Pho2.
The protection pattern found with both proteins is larger by a few nucleotides
in the downstream direction than the sum of the individual patterns.
The Pho4 binding site at UASp2 does not overlap with a Pho2 binding site, but
Pho2 sites are immediately adjacent (Fig.
4
). There are only 2-5 unprotected nucleotides between the Pho4 and downstream Pho2 protected
regions, while the upstream Pho2 footprint is at least 7 nucleotides removed from the Pho4 footprint. None of the footprints are qualitatively altered when both Pho2 and Pho4 are
present in the binding reaction.
Cooperative DNA binding of Pho4 and Pho2 to UASp2 was examined by gel shift
experiments using a fragment containing UASp2 and the two flanking Pho2 sites.
As shown in Figure
5
A, binding of Pho4 to UASp2 was enhanced in the presence of Pho2. Furthermore,
the presence of Pho4 increases the apparent affinity of Pho2 for its binding
sites (not shown). To determine if binding of Pho2 to either of the single
sites is sufficient for cooperativity, two additional restriction fragments were analyzed, each containing only one of the two Pho2 sites in addition to UASp2. The upstream Pho2 site
appears to bind Pho2 somewhat more strongly than the downstream (compare the
ratios of bound versus free DNA in lanes 5 of Fig.
5
B and C). However, either site alone can give rise to cooperative DNA binding of
Pho4 and Pho2.
We wished to determine whether protein interactions between Pho2 and Pho4 are
sufficient to generate ternary complexes and/or enhance the DNA binding
activity of Pho4, and therefore, we examined binding of Pho4 in the absence or
presence of Pho2 to a 41 bp restriction fragment from UASp2 containing only the
Pho4 site (Fig.
6
). This is possible, because, unlike with UASp1, the Pho2 and Pho4 sites do not
overlap at UASp2. Pho4 binding to this fragment was identical in the presence
and absence of Pho2. No evidence of Pho2 binding was observed, nor was any
ternary complex detected (Fig.
6
), although even higher Pho2 concentrations were used than in the experiments of
Figure
5
. These results show that DNA binding by Pho2 is required for the formation of a stable ternary complex and cooperative interactions with Pho4.
As shown in Figure
1
D, one region protected by Pho2 is present 70 bp downstream of UASp2 (at
position -169 to -182). Adjacent to the Pho2 protected region, there is a sequence -CACATG- which corresponds to the Pho4 consensus binding site
(CACGTG) with just one mismatch. We therefore examined binding of Pho2 as well as Pho4
to this region. At lower Pho4 concentrations, which are still sufficient to
give clear footprints at UASp1 and UASp2, Pho4 binding is not detectable (not
shown). However, at higher Pho4 concentrations there is a footprint from
position -190 to -200, which is separated from the adjacent Pho2 footprint by only a
few nucleotides (Fig.
7
).
As outlined in the Introduction, earlier data had suggested that Pho2 might be
required for
PHO5
activity, without however actually binding directly to any UAS elements in the
PHO5
promoter. Such a model was based on the strict Pho2 dependence of the
PHO5
promoter, as well as the lack of demonstrable Pho2 binding sites in the
PHO5
promoter, other than a single site which is dispensable for
PHO5
promoter activity. We have now found that there are multiple Pho2 binding sites
at the
PHO5
promoter. It is difficult to purify full length Pho2 from
Escherichia
coli
extracts, and only after using the HIS-tag technology have we been able to obtain highly purified native Pho2.
This may account for our previous difficulties in obtaining clear Pho2
footprints and might explain why Pho2 binding was previously detected only at a
region of the
PHO5
promoter where we have now identified a cluster of strong binding sites.
Importantly, in the present study we have demonstrated that there is mutual
binding cooperativity between Pho2 and Pho4 at each Pho4 binding site, which
may resolve the contradiction regarding the Pho2 dependence of the
PHO5
promoter.
A recent paper by Hirst
et al
. (
18
) demonstrated that a Pho2-VP16 fusion could activate the
PHO5
promoter when it was coexpressed with a Pho4 derivative lacking an activation
domain. However from these results, it was not clear whether the Pho2-VP16 fusion was directly contacting DNA or rather was being targeted to
the
PHO5
promoter through an interaction with Pho4. Our data indicate that DNA binding
of Pho2 is indeed required for its recruitment to the promoter, and
furthermore, that Pho2 binding directly reinforces the affinity of Pho4 for
each of its binding sites.
Of the different mechanisms invoked to explain the role of Pho2 in the
activation of other promoters the one proposed by D. Stillman
et al
. in their studies of the
HO
promoter, where they demonstrated cooperative DNA binding of Pho2 with the zinc finger DNA binding protein Swi5 (
12
), appears the most relevant for the
PHO5
promoter. In addition, they have shown that the two proteins are important for
activity of the
HO
promoter
in vivo
. In contrast, attempts to demonstrate cooperative DNA binding of Pho2 and Bas1
at
the
HIS4
promoter
in vitro
were unsuccessful (
13
). In the
TRP4
promoter, Pho2 has been shown to bind to a site that overlaps a Gcn4 binding
site, yet the two proteins appear to bind in a mutually exclusive manner (
14
).
By sequence comparisons of the Pho2 protected regions at the
PHO5
promoter, as well as those at the
HO,
HIS4
and
TRP4
promoters, the following consensus sequence for Pho2 binding emerges: 5'-(T/C)TAA(T/A)T(T/G)AAT-3'. At the
PHO5
promoter, the Pho2 protected region that overlaps with UASp1 and the two regions of
protection located between UASp1 and UASp2 contain sequences which fully match
the proposed consensus, while the Pho2 sites adjacent to the 3' side of UASp2 and at the 3' side of the newly mapped Pho4 site show 1 and 2 mismatches,
respectively. The apparent relative affinity of Pho2 for binding sites at the
PHO5
promoter, estimated from gel shift experiments, correlates well with the extent
their sequences match the proposed consensus.
A common motif found in the binding sites for many homeodomain proteins is the
TAAT sequence (
24
). Nonetheless, it is clear that sequences outside this core contribute to the
binding specificity of particular homeodomain proteins (
25
-
27
). The OCT homeodomain proteins, as well as some other homeotic proteins, bind
specifically to the sequence TNATTTGCAT (
25
,
28
), while the
Drosophila
homeodomain proteins
eve
,
zen, en, prd
and
ftz
all bind to the consensus sequence TCAATTAAAT (
26
,
29
). Both consensus sequences are similar (with two mismatches) to the one
proposed here for Pho2 binding. The similar binding specificity of Pho2 and the
Drosophila
homeodomain proteins is in agreement with the high degree of homology in the third recognition helix of Pho2,
eve
and
prd
(
30
).
Many homeodomain proteins are promiscuous in their capacity to bind DNA cooperatively with either homologous or heterologous homeodomain proteins and/or with non-homeodomain proteins. Cooperative binding of Pho2 with the zinc finger
protein Swi5 was previously demonstrated (
12
), and here we are reporting cooperative binding of Pho2 with the bHLH protein
Pho4, showing that the Pho2 protein can make cooperative interactions with
different classes of non-homeodomain proteins. One surprising aspect of the Pho4-Pho2 binding cooperativity is that it appears to be relatively
insensitive to the spacing between the Pho4 and Pho2 binding sites. Pho2
facilitated DNA binding by Pho4 regardless of whether the sites overlapped to a
significant extent (UASp1), or whether they were separated by up to 12 bp
(UASp2). In the
HO
promoter, increasing the space between the Pho2 and the Swi5 binding sites by
10 bp did not influence cooperative binding of these two proteins
in vitro
, nor promoter activity
in vivo
(
12
). This was interpreted to reflect substantial flexibility of the interaction
domains of Pho2 and Swi5, and the same may well apply to Pho2 and Pho4.
The alternative explanation that Pho2-Pho4 interactions alone are sufficient for the formation of a stable
ternary complex is ruled out by the experiment showing that binding sites for
both proteins are required to observe ternary complexes and cooperative binding
(Fig.
6
). Interactions between Pho2 and Pho4 have been demonstrated
in vivo
by the two hybrid assay (
18
), suggesting that Pho2 and Pho4 might bind each other even in the absence of
DNA-binding by both proteins. However, the yeast two hybrid system is a very
sensitive assay for protein interaction, which can even detect an interaction
between a kinase and its substrate (
31
). Therefore, the yeast two hybrid data probably reflect an interaction that
normally occurs when both proteins are bound to DNA.
Pho4 by itself has been reported to bind with lower affinity to a UASp1
oligonucleotide as compared with one containing UASp2 (
32
). However, as shown here, when larger promoter fragments containing UASp1 or
UASp2 were used instead, Pho4 binds to both sites with similar affinity. In
comparison, binding to a fragment containing the newly mapped Pho4 site is
about three to four times weaker. Except for the core consensus sequence, this
site does not show any homology to either UASp1 or UASp2. Deleting (
5
) or mutating this site (unpublished) does not lead to a significant drop in
promoter activity in an otherwise wild-type promoter. This is clearly different from the situation encountered
with UASp1 and UASp2. Mutating either of the two elements leads to a 90%
reduction in promoter strength (unpublished). In view of the strong cooperative
binding of Pho2 and Pho4 to the newly mapped Pho4 site it is still conceivable
though that this element is functional
in vivo
, and we are currently testing this possibility.
A role of Pho2 in increasing the affinity of Pho4 for its target sites is also
consistent with previous
in vivo
results. Disruption of
PHO2
renders chromatin at the
PHO5
promoter permanently closed, even under phosphate starvation conditions (
17
). Furthermore, binding of Pho4 to the
PHO5
promoter as assayed by DMS
in vivo
footprinting is lost in a
pho2
strain (
9
). Overexpression of Pho4, however, restores its ability to bind to its sites
and disrupt nucleosomal structure in the
PHO5
promoter even in the absence of Pho2, indicating binding of Pho4 to the UAS
elements. Nonetheless, expression of the
PHO5
gene is only ~25% of wild-type level (
17
). This suggests a role of Pho2 not only in helping Pho4 bind to its target
sites but also in enhancing transcriptional activation. A recent report has
proposed that interaction between Pho2 and Pho4 might increase the accessibility of the activation domain of Pho4, which might account for this second role (
33
).
Homeodomain proteins often bind to multiple sites at their target promoters.
Five binding sites for
the
bcd
protein were found upstream of the hunchback gene (
34
). Similarly, the Ubx protein binds cooperatively to clusters of its binding
sites found in several promoters (
35
). In many such cases, the binding of the homeodomain protein to multiple sites
has been shown to be cooperative. Our binding assays revealed that Pho2 binds
cooperatively by itself to a
PHO5
promoter fragment containing two or more Pho2 binding sites, even when those
Pho2 binding sites were separated by as much as 50 bp. Cooperative binding to
distal sites by other homeobox proteins has been postulated to occur through a
DNA looping mechanism (
35
).
To fully understand the activation process at the
PHO5
promoter, the chromatin structure of the promoter has to be taken into account.
The repressed
PHO5
promoter is covered with four positioned nucleosomes, which are disrupted upon
promoter activation in a Pho4-dependent process (
36
). Although binding of Pho4 to both UASp1 and UASp2 is required for chromatin
transition and transcriptional activation to occur (
7
), it is reasonable to assume that the activation process at the
PHO5
promoter is initiated through UASp1, since it is located in a short nucleosome-free region in the repressed promoter and therefore is available for
protein binding under repressing conditions (
7
). In contrast, UASp2 is located in the middle of nucleosome -2, which in the repressed state prevents binding of Pho4 to this site (
7
). On the basis of the finding that homeodomains make contacts in the major as
well as the minor groove of DNA, it was suggested that nucleosomes could
exclude homeodomain proteins from binding (
37
). Therefore, only the site overlapping UASp1 is accessible to Pho2 in the
repressed
PHO5
promoter and might therefore be of particular importance in
PHO5
regulation. Cooperative binding of Pho2 and Pho4, together with cooperative
binding of Pho2 by itself, could make the promoter exquisitely sensitive to small changes in the concentrations of the two regulatory proteins.
We thank D. Stillman and A. Hinnen for kind gifts of plasmids. This work was supported by the Deutsche Forschungsgemeinschaft (SFB 190) and Fonds der Chemischen Industrie. J.S. was supported by a National Science Foundation NATO postdoctoral fellowship.
*To whom correspondence should be addressed. Tel: +49 89 5996 420; Fax: +49 89
5996 440; Email: hoerz@bio.med.uni-muenchen.de
+
Present address: Department of Pathology, Washington University School of
Medicine, St Louis, MO 63110-1093, USA
REFERENCES
Return