Cooperative Binding of Heat Shock Transcription Factor to the Hsp7O Promoter in Vivo and in Vitro*

The minimal promoter of the Drosophila hsp70 gene contains a TATA box and two nonidentical HSE se- quences, HSEI and HSEII, that synergistically activate the promoter. We have examined stereospecific align- ment and spatial constraints in this promoter. Similar to deletion of HSEII, insertion in the spacer between the HSEs of 1 to 5 or 11 to 14 nucleotides (nt) reduced pro- moter activity to about 10%. In contrast, HSEII was capable of contributing to promoter activity when the spacer was either shortened by 2 or 4 nt or extended by 6 to 10 or 16 or 18 nt. Hence, half of the possible helical arrangements of HSEs are compatible, whereas the other half are essentially incompatible with efficient promoter function. HSEII was ineffective when its distance to HSEI was increased by more than 18 nt. In vitro, HSEII is a weak and HSEI a strong binding site for heat shock transcription factor HSF, and HSF binds to HSEII cooperatively. To find out whether the above periodicity reflects cooperative binding of HSF in vivo or represents the need of stereoalignment for synergistic activation of transcription, the weak HSF binding site HSEII was replaced with the strong binding site HSEI. This substitu- tion greatly attenuated promoter periodicity, suggesting that the periodic effects are caused by cooperative bind- ing of HSF 4-nt deletions exhibit high relative activity. Insertion of 1 to 5 nt greatly reduces gene expression (note that an activity value of 0.1 reflects the level of activity of a mutant lacking HSEII entirely). Activity is largely restored when the insertion size is increased to 6-10 nt. The configuration produced by an 11-nt insertion is only weakly functional, whereas configurations resulting from 12 to 14-nt insertions are essentially nonfunctional. It appears that promoter function is rescued partially when the insertion size is increased by 16 or 18 nt. These results demonstrate periodic variation in promoter activity, suggesting the existence of pro-tein-protein interactions between HSF molecules binding to the two HSE sites andor between bound HSF molecules and other factors. Also compatible with this interpretation is the observation that the synergistic transcription effect of HSEII is strongly dependent on its distance from HSEI. Insertion of 10 nt reduced activity by at least 50%, and the synergistic effect was abolished by insertion of 20 nt.

The minimal promoter of the Drosophila hsp70 gene contains a TATA box and two nonidentical HSE sequences, HSEI and HSEII, that synergistically activate the promoter. We have examined stereospecific alignment and spatial constraints in this promoter. Similar to deletion of HSEII, insertion in the spacer between the HSEs of 1 to 5 or 11 to 14 nucleotides (nt) reduced promoter activity to about 10%. In contrast, HSEII was capable of contributing to promoter activity when the spacer was either shortened by 2 or 4 nt or extended by 6 to 10 or 16 or 18 nt. Hence, half of the possible helical arrangements of HSEs are compatible, whereas the other half are essentially incompatible with efficient promoter function. HSEII was ineffective when its distance to HSEI was increased by more than 18 nt. In vitro, HSEII is a weak and HSEI a strong binding site for heat shock transcription factor HSF, and HSF binds to HSEII cooperatively. To find out whether the above periodicity reflects cooperative binding of HSF in vivo or represents the need of stereoalignment for synergistic activation of transcription, the weak HSF binding site HSEII was replaced with the strong binding site HSEI. This substitution greatly attenuated promoter periodicity, suggesting that the periodic effects are caused by cooperative binding of HSF to HSEII, and that stereoalignment of HSEs is not required for transcription activation. In agreement, in vitro assays using spacer mutants revealed cooperative binding of purified, recombinant HSF to HSEII with a similar periodicity as observed in vivo. Changing the distance between TATA and the HSEs did not produce promoter periodicity, indicating that stereoalignment of these elements is not important.
Heat shock protein (hsp)' genes are a class of genes that are present in all cell types examined so far and are typically silent at temperatures at which normal growth occurs but are expressed at exceedingly high levels a t elevated temperatures or in cells suffering from other types of stress (1, 2). This regulation appears to include both transcriptional and post-transcriptional components (3). Analysis of hsp gene promoters initially focused on the Drosophila hsp7O gene ( 4 4 ) , but findings made with this promoter have been extended subsequently t o other * These work was supported by National Institutes of Health Grants GM31125 and GM25232. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "aduertzsement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. The abbreviations used are: hsp, heat shock protein; nt, nucleotide; hsp genes from Drosophila and other organisms (for a review, see Ref. 7). The unusually compact arrangement of sequence elements in the Drosophila hsp7O promoter has permitted the definition of a fully functional, minimal promoter segment (8-11) that comprises 88 nt of 5'-untranscribed and about 30 nt of transcribed sequence. The 5'-untranscribed region contains a consensus TATA box and two non-identical copies of a sequence element referred to as heat shock element (HSE). HSEs were shown to confer heat regulation on the hsp7O promoter (5, 6). Their action on transcription is synergistic; deletion of either HSE reduces transcription by 10-fold or more (12).
Exonuclease I11 protection (13,14) and footprinting assays (15) identified a protein, referred t o now as heat shock transcription factor (HSF), that binds to HSEs in a heat-induced fashion. Partially purified HSF was found to stimulate transcription in vitro, and, as for transcription in vivo, this effect required the presence of both HSEs in the minimal hsp70 promoter (15,16). HSF binds in vitro to the upstream HSE (HSEII) with a 12.5-fold higher affinity in the presence than in the absence of the downstream element (HSEI;Ref. 16). Recent experiments with highly purified HSF have revealed that active HSF exists as a homotrimer and that in vitro binding of HSF to extended synthetic HSE sites is highly cooperative (17)(18)(19).
HSF binding to the adjacent HSE sites in the minimal hsp70 promoter may participate in different kinds of protein-protein interactions: first, HSF bound t o the HSEI site may interact with a second HSF molecule to facilitate its binding to HSEII. Second, HSF molecules bound to the two HSE sites interact, either individually or as a complex, with the transcription machinery, causing the synergistic activation of the promoter (20). Whereas the latter type of interaction may or may not be dependent on the stereoalignment of HSE sites with respect to each other and/or to downstream sites, cooperative binding interactions could be expected to require alignment of the HSE sites. Since cooperative binding of HSF to HSEs had already been observed in in vitro assays, we wished to find out whether these interactions could also occur in vivo. To this end, the stereoalignment of the two HSE sequences in the minimal promoter was altered systematically by increasing or decreasing the size of the spacer separating them. Analysis of the transient activity of the mutant promoters in Drosophila cells revealed discrete, periodic changes in transcriptional activity, consistent with the occurrence of cooperative binding interactions in vivo. Parallel assays of cooperative binding of HSF in vitro using a similar set of mutant promoters revealed a comparable periodic pattern, supporting this interpretation. To confirm that cooperative binding interactions involving HSF occurred in vivo as well as to find out whether interactions of HSF with the transcription complex may be dependent on stereoalignment of one or both HSEs and the TATA box, similar transient expression assays were carried out with two additional series of mutant promoters, the first testing the effects on promoter periodicity of the replacement of the weak HSF binding site HSEII with the high affinity HSF binding site HSEI, and the second the effects of changes in the distance between the HSE sequences and the TATA box. The results support the conclusion that cooperative binding of HSF to HSE sequences occurs in vivo and may be an important aspect of the regulation of promoter activity in hsp genes. Furthermore, the data also suggest that stereoalignment of bound HSF molecules and the transcription complex is not required for the synergistic activation of the promoter. We have attempted to integrate these findings in a model of hsp70 transcription regulation that involves interactions between HSF molecules and a post-initiation transcription complex (Ref. 21, and references therein).

MATERIALS AND METHODS
Plasmid Constructions-All constructs were derived from pD88 (12). In this construct, a Drosophila melanogaster hsp70 gene segment that included, following an XhoI linker, 88 n t of 5'-nontranscribed sequence, the entire RNA leader region, and the first seven hsp70 codons is linked in-frame to a truncated P-galactosidase gene. A 2.4-kilobase long segment containing 3'-nontranslated and nontranscribed sequences from a D. melanogaster hsp70 gene was placed immediately downstream from the P-galactosidase gene. To prepare mutant promoters, use was made of the XhoI linker site a t -93, a BssHII site at -68, an NruI site at -53, and an MspI site a t -39. Mutants were constructed by replacement of segments between the above sites or by insertion at these sites of appropriate synthetic, double-stranded oligonucleotides. All mutant promoters were characterized by extensive restriction digestion and by nucleotide sequencing using the dideoxy method and the oligonucleotide primer 5'-CGCTCCGTAGACGAAGCGCC.
Cell Culture, Dansfection, and Measurements of 6-Galactosidase Activity-D. melanogaster S3 cells grown in Schneider D. melanogaster medium (Life Technologies Inc.) with 10% heat-inactivated fetal calf serum in roller bottles or in flasks were distributed into an appropriate number of 60-mm dishes. One hour later, after cells had attached and formed a nearly confluent layer, cells were carefully washed with Schneider D. melanogaster medium, and then transfected with 5 pg of DNA as described previously (22). Each experiment included quadruplicate cultures transfected with construct D88 and with up to 20 other constructs as well as mock-transfected cells. After overnight incubation at room temperature, half the cultures were sealed with parafilm and placed in a 37 "C water bath. Fifteen min later, actinomycin D was added at a final concentration of 1 pg/ml to all cultures which were then sealed as above and incubated for another 105 min at 37 "C. Cells were then collected, extracts prepared, and P-galactosidase activities measured by the colorimetric assay using o-nitrophenyl-o-6-galactopyranoside as the substrate (22). Absorbance of the reactions was measured in plastic cuvettes a t 420 and 550 nm, and P-galactosidase activities were calculated by the formulaAi,,, -1.1 x A,,,. Values were corrected for the background absorbance of mock-transfected cells. Activities of promoters are expressed as fractions of the activity in extracts from heattreated, D88-transfected cells. Mean values of relative activities and standard deviations calculated from 3-6 independent experiments are given in the Figures.
RNase Assay of Hsp70 a n d Hsp70-P-Galactosidase Danscripts-Cells transfected as above were either not heat-treated or heat-treated for 30 min at 37 "C. Standard procedures were employed for the preparation of cytoplasmic RNA and for RNase mapping (23). RNA (10 pg) was hybridized to probe RNA overnight, exposed to RNases A and T I at 33 "C, and analyzed on a 4% polyacrylamide-urea gel. Probes: construct XSL (22) that is identical to pD88 except that an EcoRI linker had been inserted at the NruI site in the hsp70 promoter a t position -50 was digested with FspI (cuts a t @-galactosidase codon 52) and either EcoRI or NruI (a recognition site for this enzyme is located about 50 n t upstream from the hsp7O promoter segment). Hsp70 promoter and P-galactosidase sequence-containing fragments of about 450 and 530 nt, respectively, were gel-isolated and inserted in the SmaI site of pSP72. To prepare RNA probes, DNA from the resulting plasmids was linearized and added to a transcription reaction containing Sp6 RNA polymerase and [32PlCTP (Stratagene). Probe RNA was hybridized to cytoplasmic RNA from transfected S3 cells. Gel analysis of RNasetreated samples was then carried out. The probes detected transcripts of both transfected hsp70-P-galactosidase genes (400-nt protected fragment) and endogenous hsp70 genes (270-nt protected fragment). Inten-sities of the 400-nt fragments were standardized relative to the 270-nt fragments.
Expression a n d Purification of Drosophila HSF-HSF-Bac was constructed by insertion of a ScaI site 2 nucleotides upstream of the first ATG codon of Drosophila HSF cDNA (a gift by Carl Wu;Ref. 24). A ScaI-EcoRI fragment of HSF was blunted and inserted into a filled-in NheI site of pJVPlOZ baculovims DNA (a gift from C. E. Richardson).
Recombinant baculovirus was propagated in Spodoptera frugiperda Sf21 insect cells (obtained from A. Wood). Sf21 cells were co-transfected with wild type AcNPV baculovirus and HSF-Bac DNA. Recombinant virus was identified visually as blue plaques in medium containing the colorimetric P-galactosidase substrate 5-bromo-4-chloro-3-indoyl P-Dgalactoside lacking the occlusion bodies characteristic of wild type virus. Recombinant virus was then purified by seven rounds of infection and plaque isolation. Identification of recombinant virus was confirmed by Western blot using an anti-HSF antibody raised against a GST-HSF fusion protein or gel-shift assay with lysates from infected cells. To purify recombinant HSF, Dichoplusia ni Tn5B1-4 cells (obtained from R. R. Granados) were plated at about 25% confluence in 100-mm dishes (3 x IO6 cells/plate) and were infected with recombinant virus at an multiplicity of infection of 10. Cells were heat-treated for 20 min at 37 "C 48 h after infection, and nuclear extract was prepared as described (25) except that 0.25 mM ~-l-tosylamido-2-phenylethyl chloromethyl ketone, 0.025 mM l-chloro-3-tosylamido-7-amino-2-heptanone, 2 pg/ml pepstatin A, and 1 mM benzamidine were present during the homogenization and nuclear lysis steps. After dialysis and centrifugation for 5 min a t 10,000 x g, the extract was applied to a 10-ml heparinagarose column equilibrated in bufferA (20 mM Hepes, pH 7.9, 0.1 mM EDTA, 0.5 mM dithiothreitol, 0.5 mM phenylmethylsulfonyl fluoride, and 10% glycerol) containing 0.2 M NaC1. The column was washed with the same buffer, and HSF was eluted with a 0.2-1.5 M NaCl gradient. HSF activity was followed by gel-shift assay using an end-labeled oligonucleotide HSE probe. Active fractions were pooled, diluted with buffer A to a NaCl concentration of 0.05 M, and applied to a 5-ml Mono Q column (Pharmacia LKB Biotechnology Inc.) equilibrated in buffer A and 0.05 M NaCl. HSF was eluted with a 0.05-1.0 M NaCl gradient. HSF activity eluted a t 0.2 M NaCI.
DNA Binding Assays-Labeled DNA probes were prepared as follows: DNA of pD88 and mutant derivatives containing insertions in the spacer region separating the two HSEs were digested with XhoI (-93) and AccI (-151, and promoter fragments were purified on 2% low melt agarose gels. Fragments were recovered from gel slices by heating a t 65 "C for 10 min, followed by phenol extraction and ethanol precipitation. Fragments were labeled at their XhoI ends by the Klenow DNA polymerase reaction that was carried out in the presence of ["2PldGTP, and unincorporated deoxynucleotides were removed by chromatography on spin columns. DNA binding reactions were performed for 16 h a t room temperature. Reactions (30 pl) contained 1 pl of an appropriate dilution of purified recombinant HSF, end-labeled DNA fragment (50,000 cpm corresponding to about 25 fmol of DNA), 1 pg of poly(dI-dC) x poly(dI-dC) and 20 pg of bovine serum albumin in 20 mM Hepes, pH 7.9, 4 mM dithiothreitol, 0.1% Nonidet P-40, 10% glycerol, and 0.1 M NaCl. Reactions were electrophoresed on 2% native agarose gels in 50 mM Tris, 32 mM boric acid, and 1 mM EDTA.
For determination of relative binding constants the procedure described by Liu-Johnson et al. (26) was employed. Typically, five endlabeled fragments containing spacer sequences of different lengths were mixed together and used in a single DNA binding reaction. Following electrophoresis on a native gel and autoradiography, areas containing complexed (complex I contained HSF bound to HSEI, and complex I1 HSF bound to both HSEI and HSEII as verified by DNase footprinting) and unbound DNA were cut out and placed in Eppendorf tubes. DNA was recovered by freezing the gel slices, and the tubes were vortexed for 20 s and centrifuged for 20 min at 4 "C in an Eppendorf Microfuge. Supernatant solutions were phenol-chloroform-extracted, and DNAprecipitated by ethanol. Air-dried DNA was dissolved in formamide sample buffer and electrophoresed in a 5% acrylamide-urea gel (approximately equal amounts of cpm were loaded). A betascope was used to quantitate the intensities of the bands in the autoradiographs. Relative binding constants of different fragments n and m were calculated by using the equation KJK, = (C,,/Dn)/(Cm/Dm), where C is the amount of labeled DNA in complex I1 and D the amount of labeled DNA in the unbound fraction. Values are averages of at least three independent experiments.

RESULTS
The Minimal Hsp70 Promoter-The basic construct, D88, used in this study contains a 350-nucleotide-long hsp70 gene Values are relative to the activity of D88 and represent means from three to six independent experiments ("Materials and Methods"). The standard deviation is indicated for each measurement of mutant promoter activity. Non-heat shock activities of cells transfected with D88 or any of the mutants were 0.05 or less (not shown). B, relevant nucleotide sequences of the hsp7O promoter segments of wild type construct D88 and mutants. HSEII and HSEI sites are boxed, n t conforming to the consensus NGAAN motif are underlined, and spacer sequences separating these elements are shown below. Insertions are denoted by +, and deletions by -, followed by the number of n t inserted or deleted. C, RNase protection assay of S3 cells transfected with D88, +4 or +4(a) ( Table I). Transfected cells were heat-treated, and cytoplasmic RNA was prepared as described under "Materials and Methods." Results obtained with two different probes are shown (Pl, 450 nt; P2, 530 nt). Below the figure are relative amounts of hsp7O RNA. For comparison, P-galactosidase ex-fragment consisting of 88 nucleotides of 5'-nontranscribed sequence, the RNA leader region, and the first seven codons that are linked in-frame to a truncated bacterial 0-galactosidase gene (12). The 5'-nontranscribed segment includes both HSEII and HSEI. Previous studies have indicated that the most critical structural element of an HSE is a 5-nt motif, NGAAN, multiples of which must be arranged in alternating orientation (22,27). Competent HSEs minimally include three NGAAN motifs, but longer arrays, in which some of the motifs are lacking, are also functional provided that the overall spacing between the remaining NGAAN motifs is conserved (22). The nucleotide sequence of HSEI that contains an arrangement of three perfect and one imperfect NGAAN motifs and is located between positions -65 and -46 is CGCCTCGAATGTTCGC-GAAA (Fig. LB). HSEII that consists of one perfect and three imperfect NGAAN motifs and resides between -88 and -69 has the nucleotide sequence GCTCTCGTTGGTTCGAGAGA. The functional 5' boundary of HSEII lies at -87 (in the first imperfect NGAAN motif; Ref. 12). The sequence element GAGAGA that partially overlaps HSEII (-74 to -69) represents a binding site for a general Drosophila transcription factor that has been shown to be required for the normal expression of a number of genes (28,291. Although this factor binds the GAGAGAmotif at the boundary of HSEII in the hsp7O promoter (301, it is not known to play a role in the activity of this promoter.

A Insertion(+) or deletion(-) in base pairs
Measurements of Promoter Activity-Activities of mutant genes were compared with that of D88 (or construct OS in the experiment in Fig. 4) in a transfection assay. Briefly, D. melanogaster S3 cells were transfected with construct D88 or its derivatives, and after overnight incubation at room temperature, the cells were placed in a 37 "C water bath for heat treatment or incubated further at room temperature. Actinomycin D was added 15 min after the beginning of heat treatment to heat-treated and not heat-treated cultures to block mRNA synthesis, and the cultures were incubated at 37 "C for an additional 105 min prior to measurement of 0-galactosidase activity. A high level of 0-galactosidase activity was measured in DWtransfected heat-treated (exposed to heat prior to the addition of actinomycin D) cells, whereas not heat-treated cells had at least 20-fold lower activity (see caption to Fig. 1). As indicated by the results of multiple parallel assays, transfections were highly reproducible within experiments. None of the mutant promoters employed in this study were constitutively active (non-heat shock values omitted). To test whether the 0-galactosidase assays accurately reflected the level of transcription, expression from some of the mutant promoters was also analyzed by parallel RNase protection assays (see examples in Fig. 10. The intensity of the 400-nt protected fragment from transfected genes, standardized relative to the 270-nt protected fragment from endogenous genes, was found to closely parallel the 0-galactosidase activity measurements. Periodic Variation in Promoter Activity Caused by Systematic Changes in HSEIIIHSEI Spacer Length-To investigate the importance for promoter function of the relative positions of HSEII and HSEI that are 23 nt apart in the wild type hsp70 promoter (center-to-center distance), the spacer region between the HSEs was shortened by deletion or lengthened by insertion, at 1-or 2-nt intervals. Fig. LB shows the wild type construct D88 and a set of derivatives containing insertions or deletions within the spacer region. Care was taken to avoid the inclusion of any in-phase NGAAN or closely related motifs (GAG, CTC) in the altered spacer regions. Also, the GAGA sequence was pression levels of the same constructs, measured in parallel cultures, are given in the bottom line.  preserved. Results of activity assays of these mutants are shown in Fig. lA: promoters with 2-or 4-nt deletions exhibit high relative activity. Insertion of 1 to 5 nt greatly reduces gene expression (note that an activity value of 0.1 reflects the level of activity of a mutant lacking HSEII entirely). Activity is largely restored when the insertion size is increased to 6-10 nt. The configuration produced by an 11-nt insertion is only weakly functional, whereas configurations resulting from 12 to 14-nt insertions are essentially nonfunctional. It appears that promoter function is rescued partially when the insertion size is increased by 16 or 18 nt. These results demonstrate periodic variation in promoter activity, suggesting the existence of protein-protein interactions between HSF molecules binding to the two HSE sites andor between bound HSF molecules and other factors. Also compatible with this interpretation is the observation that the synergistic transcription effect of HSEII is strongly dependent on its distance from HSEI. Insertion of 10 nt reduced activity by at least 50%, and the synergistic effect was abolished by insertion of 20 nt. Insertions within the spacer region, while affecting its length, also cause sequence changes that may affect its flexibility. Since the HSEII and HSEI sites are only a short distance apart, these differences in the properties of the spacer regions may influence binding of and subsequent interactions between HSF molecules at the two sites. To find out whether the observed periodic changes in promoter activity occurred largely as a function of the number of nt inserted in or deleted from the spacer region, the effects of differences in the nucleotide sequence and composition of the spacer region on promoter function were tested. Promoters with spacers of different nucleotide sequence were prepared for a number of insertion sizes (see examples, in Table I). It was observed that differences in spacer sequence caused variation in promoter activity at most insertion sizes. The most drastic relative changes were seen with +1 and +5 insertions that apparently can produce configurations of HSEs intermediate in character between fully functional (insertion sizes 0 and +6) and nonfunctional (insertion sizes +2 and +4) arrangements. It may not be surprising that the formation of these intermediate configurations is critically dependent on the nucleotide sequence of the spacer. It was also observed that insertion mutant +4a was more than twice as active as other +4 insertions. This may be due to the unusual nature of the nucleotide sequence in the 4a spacer (a long stretch of alternating G and C residues) that may be capable of assuming a non-B DNA conformation. Since overall the differences in promoter activity were small, they do not affect the validity of our observation of spacer length-dependent periodic changes in promoter activity. In fact, the data obtained with mutants containing alternative spacer sequences, of which examples are shown in Table I, reveal a periodicity almost identical to that observed with the original set of mutants shown in Fig. lA.
The -1 deletion reducing promoter activity to about 10% of D88 represents an anomaly considering that larger deletions exhibit much higher activities, and that in the -1 deletion the HSEs are similarly aligned as in the much more active +9 insertions (Fig. lA). In the -1 deletion but not in D88, the NGAGN element (that may serve as a functional component of HSE in lieu of NGAAN; note that the corresponding HSEII inversion mutant in Fig. 5 has the element NGTGN) at -71 is in-phase with the NGAAN motifs of HSEI. As a consequence, the HSEI binding site for HSF may have been effectively extended preventing binding of HSF to HSEII. To test this possibility the -71 NGAGN element was changed to NGCCN in mutant -lGCC. Consistent with the hypothesis, this construct was considerably more active than the original -1 deletion ( Table I).
Alignment of HSE and TATA Sequences Is Not Important for Promoter Function-The results described above suggest that stereospecific alignment of HSEs may be required to allow either for cooperative binding of HSF to HSEII or for interactions between HSE-bound HSF molecules needed for transcriptional stimulation. Alternatively, they may reflect a need for alignment of HSEII with sequence elements downstream from HSEI. To distinguish between these possibilities, we constructed a series of promoter variants with deletions or insertions of different lengths between HSEI and the TATA box (Fig.  2B). The former interpretation predicts that only monotonous distance effects should be observed except where inhibition occurs due to the closeness of protein binding sites. If, however, the alternative interpretation is true, then periodic variation in promoter activity should be revealed by the experiment. Results showed that the different mutant promoters had similar activities with the exception of those with 4or 6-nt deletions (Fig. 2 A ) . Hence, there is no need for a specific alignment of HSEs and downstream sequences. The drastic reduction in activity caused by deletion of 4 or 6 nt suggests that these deletions may place HSEI and the TATA box in such proximity as to prevent the simultaneous binding of HSF and components of the transcription complex. Thus, the periodic effects caused by changing the length of the spacer between the HSEs reflects the need for stereoalignment of the two HSE sites and the HSF molecules binding to these sequences.
Spacer Length-dependent Variation of Promoter Activity in Vivo and Cooperative Binding of HSF in Vitro Show Similar Periodicity-Does the observed periodicity in promoter activity reflect cooperative binding of HSF to HSEII? The extent of the role played by cooperative binding interactions between transcription factors is determined by a number of parameters including the affinities of the factors for their respective binding sites. Previous work has shown that in vitro HSF binds HSEII only weakly, but that the af'finity of HSF for this element is increased 12.5-fold when HSEI is present (16). To confirm that binding of HSF to HSEII is cooperative and to determine whether this interaction is influenced similarly as promoter activity by changes in the length of the spacer region between the HSEs, we decided to perform in vitro assays of HSF DNA binding. For in vitro DNA binding, purified recombinant Dro-  (+) and deletions (-) in the =&on separating HSEs from the TATA box. A, P-galactosidase activity assays in heat-treated transfected S3 cells. E , relevant nucleotide sequences from D88 and mutants. For other details, see Fig. 1. sophila HSF made in a baculovims expression system was incubated a t limiting concentration with sets of end-labeled fragments that included HSEII, HSEI, and spacer segments of different lengths. HSF.DNA complexes were separated by native gel electrophoresis, and DNA was extracted from complex I that contained HSF bound to HSEI and complex I1 in which HSF occupied both HSE sites (the nature of these complexes was verified by footprinting; data not shown). Recovered DNA was electrophoresed on a sequencing gel (Fig. 3), and relative constants for HSF binding to both HSEII and HSEI (complex 11) were calculated (Table 11). Since binding to the HSEI site (complex I) was unaffected by spacer length, differences in the constants reflected differences in HSF binding to the HSEII site. While HSF bound strongly to HSEII on a fragment containing the wild type spacer, moderately strong binding occurred to HSEII present in a probe fragment with a 1-nt deletion (-1GCC), and weak binding in fragments with 2-or 3-nt insertions. Binding was restored partially (+9 and +loa) or nearly completely (+6 and +8) when the insertion size was increased to 6-10 nt. Binding to HSEII was reduced somewhat FIG. 3. Arepresentative in nitro DNA binding assay. The binding of HSF to four different insertion mutants +2, +8, +12, and +18 was compared relative to D88. The diagram on the right represents a native agarose gel from which HSF.DNAcomplexes were isolated. C I , complex I containing HSF bound to HSEI; C 11, complex I1 containing HSF bound to both HSEI and HSEII; D, unbound DNA. The autoradiogram shows the different DNA species separated on a sequencing gel. See "Materials and Methods" for details.  Data from Fig. 1 and Table I. at insertion sizes of +12 and +14, but increased again a t larger insertion sizes. These results confirmed the existence of cooperative binding of HSF to HSEII since the relative binding constant changed as a function of the stereoalignment of HSE sequences. There appears to be a close correlation between changes in binding constants and promoter activity for insertion sizes of up to 16 nt. At an insertion size of 20 nt, however, distance appears to have a stronger negative effect on promoter activity than on HSEII binding in vitro, presumably reflecting differences in the conditions of DNA binding in vitro and in vivo. We conclude from these results that cooperative binding of HSF to HSEII adequately explains the periodic variations in promoter activity observed in vivo. Swap of HSEII for HSEI Provides Evidence for in Vivo Cooperative Binding of HSF-If cooperative binding interactions are critical for hsp7O promoter function in vivo, substitution of the weak HSF binding site HSEII with the strong site HSEI should reduce the need for cooperative binding of HSF. In this case, the substitution promoter should be less dependent on the stereoalignment of HSE sites than the HSEII-containing promoter. On the other hand, interactions between bound HSF molecules that may be required for transcriptional enhancement should not be affected by this swap of binding sites. Hence, the extent by which the substitution of the HSEII site reduces the amplitude of the periodic effects should provide a measure of the contribution of cooperative HSF binding to in vivo hsp70 promoter function. We employed transfection assays to test whether substitution of HSEII with HSEI reduced the need for stereoalignment of HSEs. HSEII was replaced with an HSEI sequence in such a way that the TTC element of the HSEI sequence occupied the same position as that in HSEII (i.e. -77 to -75; see Fig. 4B). The substitution promoter, OS, containing a spacer region of the same length as D88, was similar in activity to D88. To avoid confounding spacer sequence effects, many of the substitution promoters received the same spacer sequences that had been used in the previous sets of constructs (-lS, OS, 3s to 8S, and 14s to 18s; compare sequences in Fig. lB, Table I and Fig. 4B). Comparison of the activities of insertion promoters containing HSEII or the HSEI substitution revealed that the periodic effects were greatly attenuated in the substitution promoters: a 1-nt deletion reduced activity only by about 55% in the substitution promoter ( Fig.  4A) but by 90% in the promoter containing HSEII (Fig. lA).
Insertions of 1, 3, or 4 n t decreased expression from the substitution promoter by only 3040% (Fig. 4A) but from the HSEII-containing promoter by about 90% (75-90% for +1 insertions depending on the spacer sequence; Fig. lA and Table   I). Furthermore, a 14-nt insertion did not affect the activity of the substitution promoter but reduced that of the HSEII-containing promoter by 90%. Note that, within the range of insertion sizes examined (insertions of 22 nt or less), only modest distance effects were observed with the substitution promoter, although in some additional mutants with larger insertions (86-180 nt) promoter activity was greatly reduced (data not shown). This contrasts results obtained with the HSEII-containing promoter whose activity already decreased to 50% following insertion of 10 nt, and to less than 10% following insertion of 20-24 nt.
To ensure that the reduction in the amplitude of the periodic effects observed with the substitution mutants was due to the increased affinity of the substituted HSE site for HSF and not to sequence changes incidental to mutant construction, derivatives of D88 were also prepared in which the NCTCN element at the 5' end of HSEII (-88) was changed to the standard Nl"CN motif in an attempt to increase the affinity of the resident HSEII site. This sequence change caused a small increase in promoter activity. Insertion of 3 nt into the spacer region of this mutant promoter only reduced its activity by 55% (not shown) rather than by more than 90% as in the promoter containing the wild type HSEII sequence. This result is analogous to results obtained with the above substitution promoters and supports the conclusion that substitution of HSEII with a strong HSF binding site diminishes the magnitude of promoter periodicity, providing evidence for cooperative binding in vivo of HSF to HSEII. It is interesting to note that replacement of HSEII with HSEI does not increase promoter activity, underscoring the high efficiency of cooperative binding of HSF in vivo.
The Influence of the Orientation of HSEII on Promoter Periodicity-HSEII contains three imperfect and one perfect NGAAN motif. The latter motif lies in the promoter-proximal half of the HSEII element and may serve as the initial contact site of a cooperatively binding HSF molecule. In the inverted orientation of HSEII, the perfect motif is 5 nt further away from HSEI than in the original orientation. We therefore asked whether inversion of HSEII would change promoter periodicity by 5 nt or place a greater distance constraint on promoter activity than the original orientation.
Results of transfection experiments with an appropriate set of inversion mutants (Fig. 5B) indicated that HSEII inversion did not alter promoter periodicity (compare results in Figs. lA and 5 A ) . The main difference between promoters with inverted and un-inverted HSEII was that the activities of +8 to +10 insertions were considerably lower in the former than in the latter promoters. These results suggest that the initial contact in the cooperative binding reaction may not occur at the perfect motif but perhaps at the proximal end of HSEII. Completion of the binding reaction may become more difficult as the distance between the perfect motif and the contact site is increased.

DISCUSSION
Using sets of mutant promoters with altered spacing between HSEs and between HSEs and the TATA box region, we analyzed in detail how minimal hsp7O promoter function depends on the relative helix positions of the two regulatory sequence elements HSEI and HSEII and obtained evidence that cooperative binding of HSF to HSEII occurs in vivo and plays an important role in regulating the activity of this promoter. Depending on the relative helix positions of the HSE elements, cooperative binding of HSF to HSEII can or cannot occur, but once bound to the two elements, HSF molecules appear to be capable of interacting with the transcription machinery independent of the precise location of their binding sites in the promoter.
Three separate lines of experiments provide evidence for cooperative binding of HSF to HSEII in vivo: first, promoter activity changes (with a defined periodicity) as a function of the relative helix positions of the HSE sequences. This suggests the existence of protein-protein interactions between HSF molecules binding to the HSE sites. Second, in vitro DNA binding assays using purified HSF revealed that binding to HSEII is similarly dependent on the helix positions of the HSE sequences as promoter activity in vivo. These results not only confirm earlier studies demonstrating the existence of cooperative binding interactions in vitro but also suggest that the in vivo variation of promoter activity reflects the need for coop- erative binding of HSF in vivo. Third, in vitro binding studies indicated that HSEII binds HSF only weakly: removal of the adjacent HSEI site resulted in a 12.5-fold decrease in the affinity of HSEII for HSF (16). If the periodic variation in promoter activity reflected cooperative binding of HSF to HSEII rather than interactions between HSE-bound HSF molecules essential for synergistic activation of the promoter, replacement of the HSEII site with a second copy of an HSEI site should decrease the dependence of promoter function on the relative positions of the HSEs. On the other hand, if the positional requirements reflected the need for interactions between bound factors, replacement of the HSEII site should not alter the observed periodic effects on promoter function. Our observation that replacement of HSEII with HSEI greatly attenuates the positional effects strongly suggests that they reflect constraints imposed on cooperative binding of HSF in vivo. This interpretation is further supported by the findings that HSEII no longer contributes to promoter activity when its distance to HSEI is increased by more than 18 nt, and that promoter activity is reduced 24-fold when HSEII is removed by only 10 nt. In contrast, this distance dependence is greatly diminished in the constructs containing two HSEI sequences. Although cooperative binding interactions occurring over considerable distance have been described (31), such interactions are frequently highly sensitive to distance (32-34). One would therefore have predicted, as was indeed observed, that the HSEII t o HSEI substitution that reduces the dependence of the promoter on cooperative binding to HSE sequences would also mitigate the requirement for close proximity of the HSE sequences.
Promoter periodicity, reflecting the existence of requirements for stereoalignment of factor binding sites and the factors binding to these sites, has been first recognized in studies with an SV40 promoter (341, and similar observations were made with a number of other promoters (32,(34)(35)(36)(37)(38). Most pertinent to our experiments is a previous study describing periodicity in the Drosophila hsp7O promoter (39). In these studies concerned mainly with the demonstration of the existence of positional requirements, relative positions of factor binding sites were changed in 5-nt steps. While such extreme changes in relative positions of binding sites are appropriate to uncover evidence for protein-protein interactions, they do not provide any insight into the degree of flexibility inherent in such interactions. There are two components providing flexibility, the DNA underlying and between the binding sites and the binding proteins themselves. We have been able to study this aspect on the example of the adjacent HSE sites in the minimal hsp7O promoter by changing the relative helix positions of the sites by deletiodinsertion of 1 nt at the time. We observed a high degree of flexibility as half of all possible relative helical arrangements of HSE sites are conducive to cooperative binding interactions. Furthermore, subtracting the effect of distance per se, interactions appear to occur with similar but not identical efficiencies in all of these relative positions. The remaining arrangements (with the exception of intermediate configurations produced by certain specific spacer sequences in +1 and +5 insertions) are essentially equally nonfunctional, i.e. there is a sharp transition between functional and nonfunctional arrangements of HSE sequences. Hence, at least in the case of cooperative binding interactions involving HSF, these interactions approach the quality of an o d o f f switch: there essentially appear to exist only two types of positional arrangements of binding sites, one type permitting cooperative binding and resulting in nearly complete occupancy of both sites, and the other prohibitive of cooperative binding. It is tempting to speculate that positional requirements such as the ones observed for cooperative binding to HSE sequences could be exploited to provide an additional mechanism of gene regulation: with the proper helical arrangement of two sites required for synergistic activation a promoter may be regulatable by small changes in regional chromosomal superhelicity that may either permit or prohibit cooperative binding of regulatory proteins to the sites. Although there is no evidence that the hsp70 promoter may be regulated by such a mechanism, it is interesting to note that HSEII and HSEI are normally in positions in which rotation of the sites relative to each other by 35" in either direction (partly due to an anomaly discussed under "Results") greatly reduces promoter activity.
We report herein that hsp7O promoter activity is very sensitive to increases in distance between the HSE sequences. Conflicting results have been reported previously. An earlier study by this group suggested that hsp70 promoter variants with spacers of up to 370 nt in length are active (12). We have subsequently realized that all insertion mutants used in this study included in their spacers the sequence element CGAGAT located 5 n t downstream from HSEII that together with nearby sequences may constitute an additional imperfect HSF binding site with the nucleotide sequence ZGAGAGCGAZTC (underlined are incomplete GAA motifs). To test the hypothesis that the addition of this site artifactually strengthened the promoter, we replaced part of the spacer region in one of the mutants such that the sequence in question was changed to CGAGGG. Consistent with the hypothesis, the latter construct was considerably less active than any other mutant in the series (not shown). In a second study reporting similar findings (39), the activity of the promoter may have been influenced by multiple enhancers present in the vector employed that in-cluded several functional promoters in addition to that of the hsp70 gene. As discussed above, once the need for cooperative HSF binding interactions is removed, which to a large measure can be achieved by replacing HSEII with HSEI, promoter activity becomes relatively insensitive to changes in the alignment of the two required HSE sequences. Furthermore, there appear to be no stringent requirements concerning the relative positioning of HSE sequences and downstream sequence elements. Binding of regulatory transcription factors to upstream sites, in this case HSF binding to HSE sites, is thought to stabilize the initiation complex or to enhance its assembly via protein-protein interactions (40). Consequently, the stereoalignment of upstream transcription factor binding sites and downstream sequence elements should be critically important, especially when the sites are close (for examples, see Refs. 32,34,36, and 37). Nevertheless, a number of cases have been described where, as in the hsp7O promoter, the stereoalignment of sites appears to be of little importance for promoter function (33, 41-44). In several of these promoters, however, distances between relevant factor binding sites can only be increased within narrow limits, supporting the view that the factors binding to these sites may need to contact each other. Several different explanations have been proposed to account for the absence of a need for alignment of factor binding sites that may also be applied to the hsp7O promoter (42, 43, 45-49). However, an alternative and more specific explanation is suggested by several unusual properties of the hsp70 promoter: the TATA box region of the hsp70 gene is constitutively occupied by a protein, presumably a TFIID-like molecule (13, 14, 50, 51), even in the absence of heat shock. In contrast, HSF is not present in the promoter of the inactive gene but binds only following heat shock (13, 14). Cross-linking experiments have revealed that RNA polymerase I1 is present at the beginning of the hsp7O gene in the absence of stress. Following heat shock, the distribution of RNA polymerase I1 changes, and it becomes abundantly present in the entire transcribed region (52, 53). More recently it was found that the inactive promoter contains, on average, one molecule of RNA polymerase I1 that is capable of initiating transcription, but pauses at position +25 (21, 54). This suggests a model of transcription regulation at a postinitiation level that might obviate the need for exact positioning of HSE sites: transcription initiation complex may form on the unstressed hsp70 promoter in the absence of bound HSF, and transcription begins but, in the absence of stress, pauses after a short distance. When the cell is stressed, HSF is activated and binds t o HSE sites. As RNA polymerase I1 moves along its helical path at the beginning of the gene, the stereoalignment of the HSE-bound HSF molecules and RNApolymerase I1 is changing continually. At some point in time, the relevant domains of one of the bound HSF molecules and of RNA polymerase I1 or an associated factor become exactly juxtaposed, and a productive interaction between the molecules may occur, increasing the probability of RNA polymerase I1 to proceed past the pause site. The synergistic effect of the second, bound HSF molecule may be produced by a second, similar interaction with RNA polymerase 11, that may potentiate the likelihood for transcription of the entire hsp70 gene.