Mutations in Target DNA Elements of Yeast HAP1 Modulate Its Transcriptional Activity without Affecting DNA Binding (original) (raw)

Abstract

The yeast zinc cluster protein HAP1, a member of the GAL4 family, is a transcriptional activator that binds as a homodimer to target DNA sequences. These targets include the upstream activating sequences of the CYC1 and CYC7 genes, which have no obvious sequence similarity. Even though both sites have the same affinity for HAP1, activation differs at these two sites, even when the sequences are placed in an identical promoter context. In addition, mutants of HAP1 that can bind to both sites but are specifically transcriptionally inactive at CYC7 have been previously isolated. In order to identify nucleotides that are responsible for this differential activity, we have performed random and site-directed mutagenesis of these target sites and assayed their binding to HAP1 in vitro and their activity in vivo in reporter plasmids. Our results show that HAP1 binding sites are degenerate forms of the direct repeat CGG N 3 TA N CGG N 3 TA. Moreover, we show that activity of HAP1 mutants defective for activation of the CYC7 gene is restored by specific mutations in the CYC7 binding site. Conversely, other mutations of the target sites prevent activation by HAP1, without interfering with DNA binding. The results suggest that the sequence of the target sites influences the conformation and, hence, the activity of DNA-bound HAP1.

Introduction

Many fungal transcriptional regulators contain a zinc finger known as a binuclear zinc cluster. The consensus sequence of the cluster is CX 2 CX 6 CX 5–9 CX 2 CX 6–8 C, where the cysteines bind two zinc atoms which coordinate folding of the domain. A well-characterized member of this family is GAL4, which binds as a homodimer to palindromic DNA sequences (CGG N 11 CCG), with each zinc cluster recognizing a CGG triplet, as shown by X-ray crystallography ( 1 and references therein). The zinc cluster domain is followed by a short linker region and a dimerization domain. While the spacing between the CGG triplets is 11 base pairs (bp) for GAL4, other zinc cluster proteins also recognize the same palindromic triplets, but with a different spacing. For example, the PUT3 and PPR1 activators bind CGG triplets spaced by 10 and 6 bp respectively ( 2–3 ). Moreover, the geometry of the zinc cluster controls binding specificity, as changing the spacing of the GAL4 site from 11 to 10 bp greatly reduces GAL4 binding in vitro ( 4 ). Construction of GAL4, PUT3 and PPR1 chimeric proteins has shown that the linker region is responsible for binding to a site of a given spacing ( 5–7 ).

HAP1, another member of this family of zinc cluster proteins ( 8–10 ), possesses an acidic activation domain at its C-terminus (residues 1307–1483) and an N-terminal DNA binding domain (residues 1–174) ( 10 , 11 ). The zinc cluster of HAP1 (residues 64–93) is homologous to that of GAL4 (43% identity) and the coiled coil sequence of the dimerization domain is similar to the domain identified in GAL4 ( 11 ). The middle region of HAP1 is required for regulation of DNA binding through interaction with heme. In the absence of heme HAP1 is present in a high molecular weight complex ( 12 , 13 ). Binding of heme to HAP1 apparently stimulates dissociation of HAP1 from the complex and subsequent homodimerization, which is a prerequisite for specific DNA binding.

Upstream activating sequences (UASs) for HAP1 (see Fig. 2 for DNA sequences) have been identified in the CYC1 and CYC7 genes, which encode isoforms of cytochrome c ( 14–16 ). In addition, other binding sites for HAP1 have been identified in the CTT1 (catalase T), the CYB2 (cytochrome b 2 ) and the CYT1 (cytochrome c 1 ) genes ( 17–19 ). These sites share little homology. This is particularly evident when comparing the UAS of CYC1 with that of CYC7 , which has led to the suggestion that HAP1 binds to unrelated DNA sequences ( 14 ).

Dissimilarity between these sites results in different levels of HAP1-dependent activation. Activity at the CYC1 promoter is higher than at CYC7 , even if the UASs are placed in a similar promoter context. This different transcriptional activity is not a result of different affinity of HAP1 for CYC1 and CYC7 ( 20 ). Furthermore, a single amino acid change in HAP1 of Ser63 to Arg, immediately N-terminal of the first cysteine of the zinc cluster, prevents binding to the UAS of CYC1 and results in greatly increased activation (10- to 100-fold) at CYC7 , even though the affinity of the mutant HAP1–18 for the UAS of CYC7 is similar to wild-type HAP1 ( 20 ). Other HAP1 mutants, termed positive control (PC) mutants, with mutations that flank the zinc cluster show wild-type binding and activation at CYC1 but are transcriptionally inactive at CYC7 , even though they show wild-type or increased binding at that site ( 21 ).

We wished to learn more about the nucleotides responsible for this differential activity of HAP1 (and its mutants) at CYC1 and CYC7 . We first performed saturation mutagenesis of the CYC1 and CYC7 UASs (as well as site-directed mutagenesis at other UASs of HAP1) to identify key nucleotides responsible for binding of HAP1. Our results show that all UASs of HAP1 are imperfect versions of the direct repeat CGG N 3 TA N CGG N 3 TA. Moreover, at a similar position the CYC7 site has two CGC triplets instead of the two CGGs. Changing these triplets to CGG restores the activity of HAP1 PC mutants, while the activity of HAP1–18 is greatly decreased.

Materials and Methods

Strain and media

Strain THl (Mat aura3–52 his4–519 ade1–100 leu2–2 Δ hap1∶∶hisG ) is a derivative of LPY22 ( 21 ). Most of the HAP1 coding region has been deleted in that strain. Rich (YPD) and synthetic (SD) media were prepared as described ( 25 ).

Plasmids

HAP1 expression vectors . A HIS4 version of SD5-HAP1 ( 20 ) was constructed by deleting the URA3 gene of SD5-HAP1 by cutting with Stu I, adding Not I linkers and ligating a Not I fragment (containing the HIS4 gene) from CYC7-5lacZ-HIS4 ( 21 ). SD5-HAP1-PC1-HIS, SD5-HAP1-PC2-HIS and SD5-HAP1-18- HIS were constructed by introducing Dra III fragments (containing the HAP1 mutations) from URA3 marked expression vectors ( 21 ) into Dra III-cut SD5-HAP1-HIS4. HAP1 expression vector deleted of its activation domain (SD5-HAP1ΔKp-HIS) was constructed by linearizing SD5-HAP1-HIS4 with Kpn I, treating with T4 DNA polymerase and inserting a Xba I linker (New England Biolabs) containing nonsense codons in the three reading frames.

Reporter plasmid . Plasmid pLG178-M was constructed by destroying the unique Mun I site in the 2µ origin of replication of pSLFΔ178K ( 26 ) by cutting with Mun I, filling in with Klenow fragment and ligating. pLG178-M was linearized with Xho I and an oligonucleotide (5′-TCGAGAGATCTAAAAAACAATTGC-3′) and its complement were inserted at that site to give p178MB. This plasmid has unique Mun I and Bgl II sites (flanked by Xho I sites) in front of a minimal CYC1 promoter driving lacZ transcription (see Fig. 1 ).

HAP1 UAS mutants . Spiked oligonucleotides (level of contamination 3% for each of the three other nucleotides) containing the UAS of CYC1 (5′-CGCGGATCCTCATCGTCCGTAAACCCCGGCCACTGTAGGAATTCGGA-3′) or the UAS of CYC7 (5′-CGCGGATCCGCTAATAGCGATAATAGCGAGGGCTGTAGGAATTCGGA-3′) were synthesized. They were then made double-stranded by hybridizing them with a second oligonucleotide (5′-TCCGAATTCCTACAG-3′) and filling in with Klenow fragment. The DNA was then cut with Eco RI and Bam HI, gel purified on a 4% polyacrylamide gel and subcloned into p178MB cut at the compatible sites with Mun I and Bgl II. A similar strategy, or PCR, was used to introduce specific nucleotide alterations, except that the oligonucleotides contained a Bgl II site instead of a Bam HI site. Mutations were identified by sequencing the double-stranded DNA using a kit from Pharmacia. UASs of CTT1 and CYB2 (and mutants) were generated as described above with oligonucleotides having the DNA sequences shown in Figure 2 .

Figure 1

Schematic representation of the plasmid reporter (p178MB) used in this study. Unique sites located upstream of a minimal (UAS-less) CYC1 promoter are shown. The arrow indicates the major transcriptional initiation site of the promoter. Single copies of UASs were inserted upstream of the minimal CYC1 promoter using the Bgl II and Mun I restriction sites (see Materials and Methods).

Figure 2

Alignment of the known UASs of HAP1 ( 14–19 ). Two possible orientations for the UAS of CYT1 are shown. Brackets show the nucleotides protected from DNase I by HAP1. Bold characters indicate nucleotides that match the CGG triplets of the ‘optimal’ sequence. Underlined characters indicate nucleotides that match the TA repeats of the ‘optimal’ sequence.

β-Galactosidase assays

Cells were grown to saturation in YEP containing 2% raffinose. They were then diluted into minimal medium containing 1% glucose and 1% galactose and grown for ∼12–18 h (OD 600 0.6–1.0) before assaying for β-galactosidase activity. β-Galactosidase assays were performed with permeabilized cells ( 27 ).

Preparation of extracts and DNA binding assays

Extracts were prepared and DNA binding assays were carried out as described ( 21 , 28 ) using a HAP1 expression vector under the control of UAS GAL (SD5-HAP1; 20). Probes were generated by PCR amplification using the reporter plasmids as templates and purified with G50 spin columns. Quantitative data were obtained by measuring the amount of radioactivity present in the retarded bands using a phosphorimager (Fuji).

Results

Strategy used to generate and characterize HAP1 UAS mutants

Reporter plasmids with mutated CYC1 and CYC7 UASs were generated using ‘spiked’ oligonucleotides (level of contamination 3% of each of the three other nucleotides), flanked by appropriate restriction sites (see Materials and Methods). Mutagenized UASs were then subcloned into a reporter plasmid containing a minimal CYC1 promoter driving lacZ transcription ( Fig 1 ). Mutations in the UASs were sequenced and the activity of the reporters was determined in vivo by transforming them, along with a HAP1 expression vector, into a hap1 − strain. In the presence of HAP1, activity at CYC1 and CYC7 was 6 and 2.7 U respectively ( Tables 1 and 2 ). Moreover, when a HAP1 mutant deleted of its activation domain was used (HAP1ΔKp) only background activity was measured (0.1–0.3 U β-galactosidase activity; data not shown), even though this HAP1 mutant showed wild-type DNA binding in vitro ( 21 ). This indicates that activity of these reporters was dependent on the activation domain of HAP1. Mutants were also tested for their in vitro DNA binding by the electrophoretic mobility shift assay (EMSA) using an extract prepared from a strain that contained a HAP1 overexpression vector under the control of UAS GAL4 .

Table 1

In vitro binding and in vivo activity of the UAS of CYC7 and mutants

Mutations that disrupt activity of HAP1 at CYC1

Many mutants generated by random mutagenesis of the CYC1 UAS displayed reduced binding and in vivo activity, like the single point mutants 1–9 and 1–14 ( Table 1 ). Activation from these sites is 10- to 20-fold lower than from the wild-type UAS of CYC1 . Strikingly, these mutations are located in CGG triplets, motifs recognized by the GAL4, PUT3 and PPR1 activators. Moreover, all other mutants with alterations in either triplet showed reduced in vitro binding and in vivo activation. For example, mutant 1–2, with the first CGG triplet changed to CAC, and mutant 1–28, with a change to GGG, showed reduced in vitro binding and in vivo activation. Similar results were observed with mutants 1–1, 1–2, 1–6, 1–7, 1–18, 1–22, 1–41 and 1–42, containing nucleotide alterations in either CGG triplet ( Table 1 ). These results suggest that two CGG triplets are important for binding of HAP1 to the UAS of CYC1 .

Interestingly, another class of mutants was isolated with mutations outside the CGG triplets. In the case of mutant 1–3, which carries five mutations, binding was not affected, but activation was 15 times lower than wild-type CYC1 . Similar results were seen with mutant 1–26. The properties observed at this UAS are analogous to those seen for HAP1 positive control mutants at the wild-type UAS of CYC7 , i.e. wild-type in vitro binding but no transcriptional activation ( 21 ).

Other mutants (1–10, 1–32, 1–40 and 1–43) with changes outside the triplets showed wild-type binding properties, as well as significant β-galactosidase activity, suggesting that many nucleotides outside the CGG triplets are not important for binding of HAP1.

Mutations that enhance activity of HAP1 at CYC1

Two mutants (1–8 and 1–16; Table 1 ) of the UAS of CYC1 showed increased transcriptional activity (3-fold). Mutant 1–8 carries a single nucleotide change 5 nt downstream of the second CGG triplet and mutant 1–16 has 4 nt alterations, including that found in 1–8. In order to rule out the possibility that these UASs could be bound by an activator other than HAP1 we measured their activity with a HAP1 mutant that is transcriptionally inactive (SD5-HAP1ΔKp-HIS; see Materials and Methods). Activity of the mutants was reduced to background levels (0.2 U β-galactosidase activity; data not shown), indicating that their activity was dependent on HAP1.

Mutations at the UAS of CYC7

The UAS of CYC7 has two CGC triplets instead of the two CGGs found at a similar position in CYC1 . Deletion of one G in one triplet (mutant 7–25) had drastic effects on HAP1 activity (8-fold reduction in activity; Table 2 ). Some other mutants with changes in these triplets also had decreased binding in vitro and activation in vivo . For example, mutant 7–20, with a change of the first CGC triplet to CTC, had 6-fold less β-galactosidase activity ( Table 2 ). Similarly, 7–30, which had reduced activity, carries a mutation in the second CGC triplet (along with two other changes). Interestingly, changing one CGC to CGG (7–24) increased the in vivo activity ∼4-fold. A single nucleotide change between the two CGC triplets ( 7–5 ) reduced in vitro binding ∼10-fold. Many mutations only minimally affected binding and activation at CYC7 . For instance, mutant 7–1 with 3 nt changes before the first CGC triplet, had wild-type activity. Similar results were obtained with mutants 7–2, 7–3, 7–8 and 7–34. All of these alterations fall outside the CGC repeats.

Affinity of mutants for HAP1

We then tested the affinity of various key mutants by competition assays, as shown in Figure 3 . As expected from previous results ( 20 ), the UASs of CYC1 and CYC7 had a similar affinity for HAP1. Higher in vivo activity of mutants 1–8 and 7–24 was correlated with an increased affinity of HAP1 for 7–24 and, to a lesser extent, for 1–8, as compared with the UAS of CYC1 . Conversely, 1–9, which had reduced in vitro binding, did not compete with the wild-type probe CYC1 , even at a 200 molar excess ( Fig. 3 ). Similarly, 7–5 showed reduced activity and affinity. On the other hand, mutants 1–3, 1–26 and 7I ( Table 3 ) showed increased affinity for HAP1 as compared with the wild-type UAS of CYC1 , but were transcriptionally inactive.

Figure 3

Affinity of HAP1 for selected UAS mutants. ( A ) Nucleotide sequence of the UASs used for competition assays. ( B ) Eletrophoretic mobility shift assay. HAP1 extract and the CYC1 probes were prepared as described in Materials and Methods. Lane — contains no competitor. Triangles represent increasing concentrations of cold competitors used in the assay: 4×, 20×, 50× and 200× molar excess over the labeled CYC1 probe. The arrow indicates the HAP1—DNA complex. Competition with 1–3 and 1–26 was performed on a separate gel.

Overall, data from random mutagenesis indicate that either CGG or CGC are important nucleotides for binding of HAP1. Changes at these triplets often resulted in lowered binding as exemplified by mutants 1–9, 1–14, 7–20 and 7–25. In addition, changing the second CGC triplet in CYC7 to CGG resulted in increased activity, suggesting that a CGG triplet is more favorable for binding of HAP1. Other key nucleotides are located 4 bp downstream of the first CGG/C) triplet and 5 bp downstream of the second CG(G/C) triplet, as shown by the single point mutants 7–5 and 1–8. Finally, many mutants (1–3, 1–10, 1–26, 1–32, 1–40, 1–43, 7–1, 7–2, 7–3 and 7–8) that carry mutations outside these above-mentioned nucleotides still showed strong binding to HAP1.

Site-directed mutagenesis of the UASs of CYC1 and CYC7

Based on the above results and comparison with other UASs of HAP1 (see Fig. 2 ), we concentrated on specific bases within the CGG/C motifs, as well as the downstream TA motifs, and made other alterations using site-directed mutagenesis, as shown in Table 3 . First, we changed either (or both) CGG triplet(s) in CYC1 to the sequence found at similar positions in CYC7 : CGC. This resulted in a reduction in binding and activation (10-fold) (mutants 1A, 1B and 1C; Table 3 ). Conversely, changing CGC to CGG in CYC7 resulted in increased activity (mutants 7D, 7E and 7F; Table 3 ). In addition, any alteration of the TA nucleotides in CYC7 reduced binding and activation. For example, changing 1 nt between the two CGC triplets (mutant 7K; Table 3 ) had drastic effects on binding of HAP1 in vitro and activation. Similar results were obtained with mutants 7H, 7I and 7J. At similar positions the CYC1 site has TT and AT sequences ( Table 3 ). Mutant 1G (TT→TA) showed increased activity, as did 1F (AT→AA). However, an AT→TT change had a negative effect (mutant 1E). Finally, mutating the CYC1 site to match the CGC triplets and the TA repeats found in CYC7 (mutant 1D; Table 3 ) gave levels of activation that were 30% of the UAS of CYC7 , indicating a secondary role for flanking sequences. From these data it appears that HAP1 can recognize a CGG repeat and also a CGC repeat, although less efficiently. In addition, TA repeats are also important for HAP1 binding. Their presence (and probably that of some other nucleotides, as suggested by mutant 1D) seems to be more important when HAP1 is bound to the weaker CGC motif of CYC7 . We propose that the ‘optimal’ UAS for HAP1 is the direct repeat CGG N 3 TA N CGG N 3 TA.

Table 2

In vitro binding and in vivo activity of the UAS of CYC7 and mutants

Activity of HAP1 mutants

Since a major difference between the CYC1 and CYC7 sites is the sequence of the two triplets, we tested the possibility that they are be responsible for the hyperactivity of HAP1–18 and the PC phenotype of HAP1 mutants, i.e. a defect in activation but not binding at CYC7 . As expected ( 21 ), HAP1-PC1 had wild-type activity at CYC1 and gave background activity at CYC7 ( Table 3 ), while HAP1–18 shows the opposite pattern of activity ( Table 3 ). Changing either CGC triplet (or both) of CYC7 to CGG resulted in increased activity of HAP1-PC1 and reduced activity of HAP1–18 (mutants 7D, 7E and 7F; Table 3 ). Similar results were observed with another PC mutant, HAP1–PC2 (data not shown). In addition, activation by HAP1–PC1 was greater when the second CGC triplet was changed to CGG as compared with the first one. The opposite pattern was seen with HAP1–18, where some activity was retained with a mutation in the first CGC triplet (mutant 7D), while HAP1–18 was transcriptionally inactive with a mutation in the second CGG triplet (7F). The increased activity (>50-fold) of HAP1–PC1 at 7E and 7F (relative to wild-type CYC7 ) cannot be explained by greater binding, since only a modest increase in binding (<2-fold) was observed with HAP1-PC1 tested with 7E and 7F probes in EMSA (data not shown). These results show that the effect of HAP1 PC mutations can be suppressed by mutating 1 or 2 nt of the CYC7 site. Conversely, activity of HAP1-PC1 was abolished when the CGG triplets of CYC1 were mutated to CGC (mutants 1B and 1D). It is possible that mutants of HAP1 adopt a different conformation according to the site they are bound to; this would result in different transcriptional activity (see Discussion).

Table 3

In vitro binding and in vivo activity of UASs of CYC1, CYC7 and mutants using expression vectors for HAP1 and mutants

Site-directed mutagenesis at other UASs of HAP1

Comparisons with other known UASs of HAP1 are shown in Figure 2 (see also Table 4 ). All these sites have, in the middle of the UAS, the sequence ‘TA N CGG’, which matches the ‘optimal’ sequence. However, instead of CGG as the first triplet, the sequences TGG and AGG are found in CTT1 and CYB2 respectively. In addition, the sequences TT and GC are found downstream of the second CGG triplet for CTT1 and CYB2 respectively, as compared with TA for the ‘optimal’ sequence. If HAP1 binds similarly to these sites, then changes that would allow a better fit with the ‘optimal’ sequence should result in higher activity. Mutating the first triplet of the UAS of CTT1 from TGG to CGG resulted in 7-fold higher activity ( Table 4 ). Similarly, changing AGG of CYB2 to CGG also increased the β-galactosidase activity >5-fold. Increases were also observed when nucleotides were mutated at a position equivalent to the second TA of CYC7 . However, these changes resulted in a more modest increase (2-fold) in activity ( Table 4 ). Comparison of the UAS of CYT1 with the optimal sequence shows that this site has the consensus sequence except for the second TA, which is CC. The second TA appears to be less important if other nucleotides match the ‘optimal sequence’. Taken together the data show that all HAP1 binding sites are related sequences, i.e. they are imperfect versions of an ‘optimal site’ that is a direct repeat.

Table 4

In vivo activity of UASs of CTT1, CYB2 and mutants

Discussion

HAP1 DNA targets are related imperfect direct repeats

We have performed random mutagenesis of the HAP1 binding sites found in the CYC1 and CYC7 genes. The results show that HAP1 binds to a direct repeat with the ‘optimal’ sequence CGG N 3 TA N CGG N 3 TA. This is in contrast to other zinc binuclear cluster proteins, such as GAL4, PPR1 and PUT3, which recognize palindromic sequences containing inverted CGG triplets ( 2 , 3 , 29 ). HAP1 can accommodate some changes at these triplets. For instance, the UASs of CTT1 and that of CYB2 have the sequence TGG N 6 CGG and AGG N 6 CGG, respectively ( Fig. 2 ). However, activation at these sites is increased 5- to 7-fold when TGG or AGG is mutated to CGG. Variation of the CGG triplets is also seen for other zinc cluster proteins. For instance, one UAS GAL4 has an AGG triplet instead of CGG ( 30 ).

Divergence from the ‘optimal’ site is even greater in CYC7 , where two CGC triplets are found instead of the two CGGs. Again, mutating these triplets to CGG results in greater activation. This is in agreement with studies performed on the native CYC7 promoter ( 31 ). In addition, our mutational analysis shows that the 2 nt (TA) located 4 bp downstream of the two CGG/C triplets are important for binding of HAP1. Increased activity at the UAS of CYC1 was observed when some nucleotides were mutated at positions equivalent to the first or the second TA of CYC 7 ( Table 3 ). However, similar changes resulted in a more modest increase in activity of CYB2 and CTT1 ( Table 4 ). Comparison of the UAS of CYT1 ( Fig. 2 ) shows that this site has the consensus sequence except for the second TA, which is CC. The second TA appears to be less critical if other nucleotides match the ‘optimal’ sequence. More drastic effects are seen when mutating the second TA in CYC7 . Since two CGC triplets instead of CGG are present at that site, it is likely that the second TA sequence helps to stabilize interaction of HAP1 with that UAS. Moreover, the CYC7 site, as opposed to other HAP1 UASs, is an almost perfect direct repeat, but is the only site that does not have at least one CGG triplet. Other nucleotides must also be important for binding of HAP1 to that site, as suggested by a mutant of the UAS of CYC1 (mutant 1D, Table 3 ). This mutant has the repeat CGC N 3 TA N CGC N 3 TA found in CYC7 , but shows reduced binding and activation as compared with the wild-type UAS of CYC7 . Therefore, the presence of two unfavorable CGC triplets in CYC7 is compensated for by other nucleotides that form an almost perfect direct repeat. Moreover, mutant 7E, which matches the ‘optimal’ sequence, shows a lower activity than mutant CTT1A, in agreement with a secondary role for other nucleotides located outside the ‘optimal’ sequence.

Taken together these results show that HAP1 binds to related DNA sequences. Our mutational analysis is in agreement with recently published experiments ( 32 ) where random site selection was used to identify HAP1 binding sites. However, in that study the second TA repeat was not identified as being important for HAP1 binding. Our results show that this second repeat is important for increased activity at CYC1, CTT1 and CYB2 , and essential for activity at CYC7 .

Discrimination between palindromic and directly repeated sequences

All of the known target sites for GAL4 or the Kluyveromyces lactis homolog LAC9 are palindromic (see 4 , 29 , 33 for a compilation of the UASs). Conversely, all the known binding sites for HAP1 are imperfect direct repeats ( 14–19 ). In addition, no binding of HAP1 could be detected with various palindromic sequences derived from the consensus UAS GAL4 with spacing between the two triplets varying from 1 to 14 bp (unpublished results). This is in agreement with the random site selection, where only sequences with direct repeats were recovered ( 32 ). In addition, changing the orientation of the second CGG triplet to generate a palindromic sequence prevents binding of HAP1 ( 32 ). These observations suggest that there are constraints that prevent HAP1 from binding to a palindromic sequence and GAL4 from binding to a direct repeat. It has been suggested that HAP1 binds to a direct repeat through swiveling of one DNA binding domain relative to the dimerization domain ( 32 ). However, if the DNA binding domain of HAP1 shows such flexibility, HAP1 should also be able to bind to a palindromic sequence. One possibility is that the relatively long N-terminal segment of HAP1 that precedes its zinc finger (63 amino acids as compared with 10 for GAL4) prevents HAP1 from binding to a palindromic sequence. Orientation and spacing of the CGG triplets appear to be major determinants for the binding specificity of a given zinc cluster protein. This contrasts with nuclear receptors, which show more flexibility for their binding sites ( 34–38 ). Thus our results show that all HAP1 sites are related and are imperfect direct repeats with the optimal sequence CGG N 3 TA N CGG N 3 TA.

Correlation between in vitro DNA binding and in vivo transactivation

Some mutants do not show a correlation between in vitro binding and in vivo activation. For instance, in the case of mutant 1–3 ( Table 1 ), which carries five mutations, binding is not affected, but activation is 15 times lower than wild-type CYC1 . Similarly, mutant 1–26, with one alteration 3 nt downstream of the first CGG triplet, shows greatly reduced activation. Affinity of these UASs for HAP1 was shown to be equivalent to that of CYC1 by competition assay ( Fig. 3 ). Similar results were observed for nuclear receptors, where a heterodimer formed of LXR/RXR binds to various sites but activates only a certain subset. ( 39 ).

In addition, the properties observed at these UASs are analogous to those seen for HAP1 PC mutants (which carry amino acid changes in the DNA binding domain), i.e. wild-type in vitro binding but no transcriptional activation (21, Table 3 ). For instance, HAP1-PC1 shows background activity at CYC7 , while binding of HAP1-PC1 to CYC7 was shown to be stronger than to wild-type CYC7 ( 21 ). However, when either CGC triplet is changed to CGG increased activity was observed ( Table 3 ). Binding of HAP1-PC1 to CYC7 was shown to be stronger than wild-type CYC7 ( 21 ). In contrast to PC mutants, HAP1–18 shows a dramatic increase in activation at the UAS of CYC7 (∼16 times more than wild-type HAP1), even though it has a similar affinity for the UAS of CYC7 as compared with wild-type HAP1 ( 20 ). Introduction of CGG triplets decreases the activity of HAP1–18 as opposed to the effect seen for PC mutants, as suggested from studies with the intact CYC7 promoter ( 31 ).

We proposed ( 21 ) that the phenotype of the PC mutants could be explained by the fact that these mutants would prevent interaction of the DNA binding domain of HAP1 with a cofactor protein that would normally act in synergy with the activation domain of HAP1. The model was based on genetic evidence that suggested that a mutant of GAL11, GAL11P, increases the activity of a GAL4 mutant by interacting with its DNA binding domain ( 40 ), which contains a zinc finger homologous to that of HAP1. However, more recent data showed that wild-type GAL11, unlike GAL11P, does not interact with GAL4 and is, rather, a component of the RNA polymerase II holoenzyme ( 41–42 ). Therefore, HAP1-PC mutants may simply have an altered conformation that may inhibit the activation domain. Changing the CGC triplets to CGG would alter the conformation of PC mutants, enabling them to be active, as seen when bound to the CYC1 site, which contains two CGG triplets. Conversely, HAP1–18 could have an alternate conformation that would allow very efficient interaction with some components of the basic transcriptional machinery. A similar model has been suggested to explain the phenotype of PC mutants of the glucocorticoid receptor ( 43 ), where it is proposed that DNA can act as an allosteric effector.

In conclusion, two different types of mutations modulate the activity of HAP1 without changing its affinity for target sites. Firstly, mutations in the DNA binding domain of HAP1 can lead to increased (HAP1–18) or decreased (PC mutants) activity at CYC7 . Secondly, mutations in the target sites for HAP1 also affect activation of HAP1 (or mutants). It will be interesting to determine if these two types of mutations have an allosteric effect on the activation domain of HAP1.

Acknowledgements

We are grateful to Andrew Nice for help in sequencing mutants. We thank Caterina Russo for synthesis of the spiked oligonucleotides. We are grateful to Drs John White and Hans Zingg for critical reading of the manuscript. We also thank members of the Laboratory of Molecular Endocrinology for helpful suggestions. This work was supported by grants from the Medical Research Council of Canada and the Fonds de la Recherche en Santé du Québec. NH is the recipient of a FCAR (Quebec government) studentship. BT is a scholar of the Medical Research Council of Canada.

References

1

,

Nature

,

1992

, vol.

356

(pg.

408

-

414

)

2

,

Mol. Cell. Biol.

,

1989

, vol.

9

(pg.

4706

-

4712

)

3

,

Mol. Cell. Biol.

,

1990

, vol.

10

(pg.

5257

-

5270

)

4

,

J. Biol. Chem.

,

1993

, vol.

268

(pg.

24699

-

24706

)

5

,

Science

,

1993

, vol.

261

(pg.

909

-

911

)

6

,

Nature

,

1989

, vol.

340

(pg.

724

-

727

)

7

,

Mol. Cell. Biol.

,

1990

, vol.

10

(pg.

5128

-

5137

)

8

,

Eur. J. Biochem.

,

1994

, vol.

225

(pg.

593

-

599

)

9

,

J. Mol. Biol.

,

1988

, vol.

204

(pg.

263

-

276

)

10

,

Cell

,

1989

, vol.

56

(pg.

291

-

301

)

11

,

Proc. Natl. Acad. Sci. USA

,

1993

, vol.

90

(pg.

2851

-

2855

)

12

,

EMBO J.

,

1993

, vol.

12

(pg.

1209

-

1218

)

13

,

J. Biol. Chem.

,

1994

, vol.

269

(pg.

14643

-

14647

)

14

,

Cell

,

1987

, vol.

49

(pg.

19

-

27

)

15

,

Mol. Cell. Biol.

,

1987

, vol.

7

(pg.

3252

-

3259

)

16

,

Mol. Cell. Biol.

,

1987

, vol.

7

(pg.

2212

-

2220

)

17

,

EMBO J.

,

1988

, vol.

7

(pg.

1799

-

1804

)

18

,

Mol. Cell. Biol.

,

1991

, vol.

11

(pg.

3762

-

3772

)

19

,

Mol. Cell. Biol.

,

1991

, vol.

11

(pg.

4934

-

4942

)

20

,

Nature

,

1989

, vol.

342

(pg.

200

-

203

)

21

,

Genes Dev.

,

1992

, vol.

6

(pg.

2001

-

2009

)

22

,

Proc. Natl. Acad. Sci. USA

,

1982

, vol.

79

(pg.

2236

-

2239

)

23

,

Cell

,

1983

, vol.

32

(pg.

319

-

325

)

24

,

Genes Dev.

,

1989

, vol.

3

(pg.

1590

-

1601

)

25

,

Methods in Yeast Genetics

,

1986

Cold Spring Harbor, NY

Cold Spring Harbor Laboratory Press

26

,

Mol. Cell. Biol.

,

1988

, vol.

8

(pg.

647

-

654

)

27

,

Cell

,

1983

, vol.

32

(pg.

1279

-

1286

)

28

,

Cell

,

1987

, vol.

49

(pg.

9

-

18

)

29

,

EMBO J.

,

1986

, vol.

5

(pg.

603

-

608

)

30

,

Mol. Cell. Biol.

,

1984

, vol.

4

(pg.

1440

-

1448

)

31

,

Mol. Cell. Biol.

,

1988

, vol.

8

(pg.

2275

-

2279

)

32

,

Genes Dev.

,

1994

, vol.

8

(pg.

2110

-

2119

)

33

,

Mol. Cell. Biol.

,

1991

, vol.

11

(pg.

1777

-

1784

)

34

,

Cell

,

1990

, vol.

61

(pg.

505

-

514

)

35

,

Cell

,

1992

, vol.

68

(pg.

731

-

742

)

36

,

Mol. Endocrinol.

,

1993

, vol.

7

(pg.

319

-

330

)

37

,

Genes Dev.

,

1993

, vol.

7

(pg.

295

-

307

)

38

,

Mol. Endocrinol.

,

1994

, vol.

8

(pg.

1193

-

1207

)

39

,

Genes Dev.

,

1995

, vol.

9

(pg.

1033

-

1045

)

40

,

Cell

,

1990

, vol.

63

(pg.

1299

-

1309

)

41

,

Cell

,

1994

, vol.

77

(pg.

599

-

608

)

42

,

Cell

,

1995

, vol.

81

(pg.

359

-

368

)

43

,

Genes Dev.

,

1994

, vol.

8

(pg.

2842

-

2856

)

© 1996 Oxford University Press