*
To whom correspondence should be addressed
ABSTRACT
The heat shock transcription factor (HSF) is constitutively expressed in
Drosophila
cells as an inactive monomer. Upon heat shock HSF undergoes trimerization and
acquires high affinity DNA binding ability, leading to specific interaction with its cognate elements in heat shock promoters. Here we show that the transactivation function
of HSF is conferred by the extreme C-terminal region of the protein. Deletion analysis of HSF fragments fused
to the GAL4 DNA-binding domain demonstrates that transactivation is dependent on HSF
residues 610-691. This domain is located beyond the C-terminal heptad repeat (leucine zipper 4) whose presence or
integrity is dispensable for transactivation. The transactivation domain is
functional in the absence of heat shock and can be replaced by the extreme C-terminal region of human HSF1. The
Drosophila
and human HSF transactivation domains are both rich in hydrophobic and acidic
residues and may be structurally conserved, despite limited sequence identity.
In the cells of all organisms studied to date, a sudden increase in the ambient
temperature or treatment by a variety of other environmental stresses activates
a small set of so-called heat shock genes [for reviews see (
1
-
6
)]. In eukaryotes, this response is regulated by a transcriptional activator
called heat shock factor (HSF) [for recent reviews see (
7
-
10
)] which interacts with sets of at least three inverted AGAAN repeats (heat
shock elements or HSEs) present in heat shock promoters (
11
-
16
). With the exception of budding yeast, HSF exists in cells in a cryptic form as
an inactive monomer and acquires high DNA-binding affinity by the means of trimerization after stress induction (
17
-
23
). Such activated HSF gains also transcriptional competence and undergoes
hyperphosphorylation. In some instances uncoupling of DNA binding and
transcriptional activities was observed which suggested that these two events
could be regulated independently (
24
,
25
).
The DNA binding activity of metazoan HSF is subjected to negative control. When
expressed in bacteria, recombinant HSF constitutively forms trimers even at non-heat shock temperatures (
26
,
27
). Latent, heat-shock inducible HSF molecules accumulate however, when the protein is
expressed in frog oocytes (
28
), tissue culture cells (
18
,
29
,
30
) or in reticulocyte lysates (
31
). The temperature threshold of heat shock-dependent activation of HSF is affected by the intracellular environment
as shown by the behavior of human HSF1 protein expressed in
Drosophila
cells (
29
), tobacco protoplasts (
30
) and
Xenopus
oocytes (
28
,
32
) as well as by the effect of temperature adaptation on endogenous human HSF1 (
33
). These data suggest that factors present in eukaryotic cells are required for
suppressing the assembly of HSF trimers under normal conditions. One of
possible means of HSF suppression is through interactions with hsp70. An
association between HSF and hsp70 was observed both under normal conditions and
after heat stress and the overexpression of heat shock proteins facilitates the
kinetics of HSF trimer dissociation (
34
-
39
). The exact effect of heat shock proteins on HSF activity and conformation
awaits further characterization.
In the budding yeast
Saccharomyces cerevisiae
and
Kluyveromyces lactis
the regulation of HSF DNA binding activity is largely bypassed and HSF is
constitutively able to interact with heat shock promoters (
20
,
40
-
43
). In these cases, heat shock stimulates the transactivation function of
promoter-bound HSF trimers and hyperphosphorylation of serine and threonine
residues (
44
). The transactivation ability of budding yeast
HSF is localized in two separate domains broadly located near N-terminus and C-terminus, respectively (
45
,
46
). The C-terminal transactivation domain was shown to overlap a leucine zipper-like motif (
47
) and its activity is influenced by other parts of protein including a conserved
heptapeptide (CE2), the oligomerization domain as well as DNA binding domain (
45
,
46
,
48
-
50
). These regions suppress transactivation under non-stress conditions and specific mutations in those domains cause
constitutive activation of heat shock promoters. In addition, the effects of
mutations in two yeast hsp70 genes demonstrate that hsp70 may also be involved directly or indirectly in maintaining HSF in a transcriptionally suppressed state (
51
).
Studies of metazoan HSFs have concentrated mainly on sequences important for DNA
binding and trimer formation. Regions responsible for transactivation were
characterized only for tomato HSF proteins (
30
) and recently for human HSF1 (
52
). In this paper we identify a C-terminal fragment of
Drosophila
HSF necessary for efficient activation of a target promoter using chimeric
molecules comprised of HSF segments fused to the GAL4 DNA binding domain. We
also demonstrate that the corresponding region of human HSF1 harbors a potent
transactivation domain active in
Drosophila
cells despite lack of obvious homology at the protein sequence level.
All effector plasmids were constructed in the pRmHa3 vector containing the
Drosophila
metallothionein promoter and the alcohol dehydrogenase 3' untranslated region (
29
). To generate control constructs pG and pGVP16, the region encoding the GAL4
DNA-binding domain (codons 1-147,
Eco
RI-
Bgl
II fragments) alone or together with the VP16 transactivation domain (codons 413-490,
Eco
RI-
Bam
HI fragment) was transferred into
Eco
RI-
Bam
HI site of the pRmHa3 vector from plasmid pSGGal4VP16 (
53
). GAL4-HSF chimeras were made by inserting various fragments of
Drosophila
HSF gene from plasmid pRmHSF20 downstream of the GAL4 DNA-binding domain coding sequence (see below and Fig.
1
for details). C-terminal and internal deletions were generated using convenient
restriction sites or by means of PCR, utilizing primers containing type II
restriction enzyme
Sap
I recognition sequences, thus avoiding generation of additional codons at the
junctions. PCR-derived regions and all subcloning junctions were verified by sequencing.
Details of subcloning are listed below (refer to Fig.
1
for the restriction map):
Construct I
. Blunted
Bst
BI-
Bam
HI fragment of the
Drosophila
HSF gene was inserted into blunted
Nhe
I (generated by filling original
Hin
dIII site)-
Bam
HI sites of plasmid pG.
Constructs II-IV and VI
.
Stu
I-
Bam
HI fragment of PCR products generated using a forward primer: 5'-GGGAATTCAAA
Construct V.
Bcl
I-
Bam
HI fragment of construct I was excised, ends filled and ligated.
Construct VII.
Blunted
Bst
BI-
Bam
HI fragment of the
Drosophila
HSF C-terminal deletion mutant 1-585 (Clos
et al
., 1993) was subcloned as in construct I.
Construct VIII.
Stu
I-
Bam
HI fragment was excised from construct I, ends filled and ligated.
Construct IX.
Blunted
Nco
I-
Bam
HI fragment of the
Drosophila
HSF gene was inserted into blunted
Hin
dIII-
Bam
HI sites of plasmid pG.
Construct X.
Blunted
Afl
III-
Bam
HI fragment of the
Drosophila
HSF gene was subcloned as in construct I.
Construct XI.
Blunted
Bsp
EI-
Bam
HI fragment of the
Drosophila
HSF gene was subcloned as in construct I.
Construct XII.
Stu
I-
Bam
HI fragment of the
Drosophila
HSF gene was subcloned as in construct I.
Constructs XIII-XVI.
Acc
I-
Bcl
I fragment was excised from constructs I-IV, respectively, ends filled and ligated.
Construct XVII.
PCR fragments of the human HSF1 gene coding residues 384-529 (forward primer 5'-TAT
Rabbit polyclonal antibody 943 against
Drosophila
HSF was described previously (
22
). Murine monoclonal antibody RK5C1 against GAL4 DNA-binding domain were obtained from Santa Cruz Biochemicals. Murine monoclonal antibody 3a3, cross-reacting with
Drosophila
hsc70 was purchased from Affinity Bioreagents. Secondary antibodies conjugated
with horseradish peroxidase (donkey anti-rabbit IgG-HRP and sheep anti-mouse IgG-HRP) were obtained from Amersham and Chemicon,
respectively. Fluorescein labeled anti-mouse IgG and rhodamine-labeled anti-rabbit IgG were from Jackson Laboratories.
Drosophila
Schneider line 2 cells were cultured in serum-free HyQ-CCM3 medium (Hyclone) in tightly closed flasks at 22oC. For transfection 3.5 ml of 1:5 dilution of a dense culture
were seeded into each 25 cm
2
flask and cells were allowed to grow for 24 h, medium was then replaced and DNA-lipofectin complexes were added [~5 [mu]g of total plasmid DNA diluted to 20 [mu]l with water and mixed with 30 [mu]l of lipofectin (Gibco/BRL)]. Expression of transfected
GAL4-HSF constructs was induced with 0.7 mM CuSO
4
as specified (
54
). Cells in flasks were washed once with PBS, then fresh medium was added and,
to induce the heat shock response, tightly closed flasks were submerged for 20 min in a water bath set to 37.5oC. When necessary, cells were allowed to recover at room temperature for 90
min. The flasks were chilled, cells were scraped, the cell suspension was
aliquoted into three Eppendorf tubes and centrifuged for 30 s at 12 000 r.p.m.
Cell pellets were immediately frozen on dry ice and stored at -80oC.
Samples of whole cell extracts or cells directly solubilized in SDS-sample buffer were subjected to SDS-PAGE and electroblotted onto nitrocellulose. Filters were blocked
with 5% powdered skim milk (Difco) in PBS/0.1% Tween 20. Primary antibodies
were diluted 1:5000-1:20 000 in PBS/0.1% Tween 20 containing 0.5% BSA and filters were
incubated for 1 h at room temperature. After rinsing twice with water and
washing 2 * 5 min with PBS/0.1% Tween 20, filters were incubated for 1 h with
secondary antibody diluted 1:20 000 in PBS/0.1% Tween 20 containing 0.5% BSA.
Blots were then rinsed twice with water and 4 * 3 min with PBS/0.1% Tween 20, and developed with ECL reagents (Amersham).
Cells were grown, transfected and heat-shocked in 9 cm
2
chamber slides (Nunc), washed with ice-cold PBS and fixed in 4% formaldehyde in PBS on ice for 10 min, washed
with PBS and incubated in methanol for 20 min. After washing four times in cold
PBS, slides were incubated in a mixture of primary antibodies (1:500 in PBS
containing 0.5% BSA) for 1 h, then washed again four times in PBS and incubated
for 1 h in a mixture of secondary antibodies (1:500 in PBS containing 0.5%
BSA). Then slides were washed with PBS, stained with Hoechst 33258, washed
again, dried in the darkness, mounted and analyzed under the fluorescence
microscope.
Cells cotransfected with GAL4-HSF constructs and pG
5
BCAT plasmid were disrupted in 200 [mu]l of 0.25 M Tris-HCl, pH 7.5 by three cycles of freezing and thawing (dry ice and 37oC, respectively). Samples were centrifuged at 12 000 r.p.m. at 4oC and supernatant was stored at -80oC until analyzed. For assay, 1 [mu]l of supernatant was mixed with 2.5 [mu]l of [
14
C]-chloramphenicol and 5 [mu]l of acetyl-coenzyme A in a total volume of 37.5 [mu]l (adjusted with 0.25 M Tris-HCl, pH 7.5). Samples were incubated at 37oC for 30 min, then chilled on ice, supplemented
with 70 [mu]l 0.25 M Tris-HCl pH 7.5 and extracted with 1 ml of ethyl acetate. Organic phase
(0.9 ml) was collected after a brief centrifugation, dried in a Speed-Vac, dissolved in 15 [mu]l of ethyl acetate and spotted onto a TLC silica gel plate. After
chromatography in chloroform:methanol (95:5) mixture, the plate was air-dried and analyzed by autoradiography.
Whole cell extracts were prepared and gel mobility shift analysis was carried
out as described previously (
55
) using [
32
P]-5'-end-labeled double-stranded oligonucleotides corresponding to heat
shock element (HSE):
5"-GGGCAGAATCTTCTAGAATCAGC-3"
3"- CGTCTTAGAAGATCTTAGTCGGG-5"
or the GAL4 binding site:
5"-AATTCGGTCGGAGTACTGTCCTCCGACTCT-3"
3"- GCCAGCCTCATGACAGGAGGCTGAGAGATC-5"
Reaction mixtures were incubated for 10 min on ice (HSE oligonucleotide) or at
room temperature (GAL4 binding site oligonucleotide) before loading onto 0.8%
agarose/0.5* TBE gel. After electrophoresis gels were dried onto DEAE paper (DE81,
Whatman) and autoradiographed.
In order to exclude the activity of the endogenous
Drosophila
HSF from interfering with analysis of mutant
Drosophila
HSF proteins expressed by DNA transfection, we have resorted to chimeric HSF
molecules whereby the DNA binding domain of HSF is substituted with the DNA binding domain of the yeast transcriptional
factor GAL4. The use of GAL4 ensures both constitutive DNA binding to target
DNA sequences as well as nuclear localization of the chimeric proteins (
56
-
58
). To further avoid the potential formation of mixed trimers between GAL4-HSF and endogenous HSF, the trimerization domain of HSF was also removed
in the GAL4 chimeras. Hence, the hybrid genes we constructed carry the region
encoding the C-terminal half of
Drosophila
HSF.
Figure
1
presents schematic structure of different GAL4-HSF constructs used in our study. Plasmids containing these inserts were
transfected into
Drosophila
Schneider line 2 (SL-2) cells and the expression of chimeric proteins was induced by adding 0.7 mM Cu
2+
for 16 h, conditions which do not induce the endogenous heat shock genes (
54
). The efficiency of transfection was ~25-30% as estimated by immunostaining with anti-GAL4 antibody (data not shown, see also Fig.
3
). Figure
2
shows a Western blot analysis of transfected cells directly lysed in SDS-sample buffer. A monoclonal antibody specific for the DNA binding domain
of yeast GAL4 protein shows the expression of a number of GAL4-HSF chimeras (Fig.
2
A, lanes 4-9, arrowheads). Analysis of the same blot with polyclonal anti-
Drosophila
HSF antibody (Fig.
2
B) reveals that the indicated proteins contain HSF fragments and thus truly
represent chimeric GAL4-HSF. In relation to the expression of the endogenous
Drosophila
HSF, the overall expression levels of the GAL4-HSFs are greater by several-fold, although there are fluctuations between different GAL4-HSF constructs. Analysis of the same Western blot with anti-hsc70 antibody in Figure
2
C confirms equal protein loading in every lane.
To determine the subcellular distribution of chimeric GAL4-HSF proteins, transfected SL-2 cells were analyzed by indirect immunofluorescence. Figure
3
shows representative examples of such staining of unshocked cells. The GAL4 DNA-binding domain was always localized in interphase cell nuclei of cells
transfected with plasmid pG carrying the GAL4 DNA binding domain only, as
confirmed by staining for DNA. That distribution pattern is identical to the
nuclear localization of the endogenous HSF in both transfected (GAL4 positive)
and non-transfected (GAL4 negative) cells. Nuclear localization in the absence of
heat stress was observed for representative GAL4-HSF chimeras (constructs I and VI; Fig.
3
). In these cases the transfected cells also show increased staining with anti-HSF antibody. All other GAL4-HSF constructs were analyzed, and the subcellular distribution of
the GAL4 chimeric proteins was found to be nuclear in the absence of heat shock
and that distribution was not changed by heat stress (data not shown).
To analyze the DNA-binding activity of GAL4-HSF chimeras, whole cell extracts were prepared from transfected
cells. Electrophoretic mobility shift analysis indicated that these proteins
recognize target DNA constitutively and representative examples are shown in
Figure
4
A. As Western blot analysis showed that chimeric proteins are equally well
extracted from cells before and after heat shock (data not shown), the specific
DNA binding activity of GAL4-HSF chimeras is not affected by heat stress. This contrasts with heat
inducible DNA-binding activity displayed by endogenous HSF in the same extracts (Fig.
4
B). A similar constitutive DNA binding activity was found for the other GAL4-HSF constructs (data not shown).
To identify regions of
Drosophila
HSF responsible for transactivation, GAL4-HSF constructs were cotransfected into SL-2 cells with a CAT reporter gene controlled by a GAL4-responsive promoter (
53
). The expression of chimeric GAL4-HSF factors was induced with 0.7 mM Cu
2+
for 6 h, cells were then heat shocked for 20 min at 37.5oC and allowed to recover for 1.5 h at room temperature prior to analysis of CAT activity. Figure
5
shows that the C-terminal half of
Drosophila
HSF comprising residues 321-691 (construct I) was capable of robust transactivation. The level of CAT activity, normalized to the GAL4 DNA binding activity,
was 28.6% for this construct, about one-third the level that was conferred by the viral transactivator VP16
(pGVP16), set at 100%. The level of CAT activity was progressively reduced as
the C-terminal end was truncated to residue 674 (construct II; 16%), to residue
655 (construct III, 12.7%), and to residue 629 (construct IV; 9.5%). Further
deletions to residues 609, 602, 585 and 536 abolished transactivation by the
chimeric factors to undetectable levels. The absence of CAT activity for these
constructs did not result from a lack of GAL4-HSF expression as the expressed proteins could be visualized by Western
blot analysis using polyclonal antibody against
Drosophila
HSF (Fig.
5
B), nor from the lack of DNA binding activity as determined by the electric
mobility shift assay (Fig.
5
C). Hence, the results define a transactivation domain for
Drosophila
HSF that lies from residue 610, downstream of the C-terminal leucine zipper, to the end of the protein at residue 691.
N-terminal deletions of
Drosophila
HSF from residues 321-610 (constructs IX-XIII) did not result in a reduction of the ability of the GAL4-HSF chimeras to activate CAT expression (Fig.
5
). Instead, when internal sequences between 405 and 610 were removed (constructs
XI-XIII), the overall level of CAT activity was increased by up to several-fold of the undeleted construct, suggesting that parts of this
internal region may participate in negative regulation of the transactivation
domain. We have also analyzed the function of the transactivation domain of HSF
in the absence of the negative regulatory region by combining the progressive C-terminal deletions from constructs II, III and IV with the deletion of
residues 349-609 in construct XIII. In general agreement with the results shown above,
the deletions from the C-terminus of HSF in this context also revealed progressive reduction of CAT
activity (constructs XIV, XV and XVI) (Fig.
6
).
To assess the ability of GAL4-HSF chimeras to respond to heat stress, the activity of GAL4-HSF constructs was determined in both control and heat shocked
cells. Analysis of CAT levels accumulating with or without heat shock treatment
revealed that all GAL-HSF constructs were able to constitutively activate the target promoter.
No significant increase of CAT activity could be quantified when the
transfected cells were subjected to heat shock. Representative examples
(constructs I and XII) are shown in Figure
7
A.
It is known that the yeast HSFs as well as the tomato HSFs contain potent C-terminal transactivation domains (
30
,
45
-
47
). Recently a transactivation function was also ascribed to the C-terminus of human HSF1 (
52
). To investigate if the human HSF1 C-terminal region can function in
Drosophila
cells, we substituted the 3'-end region of
Drosophila
HSF with the equivalent region of the human protein (construct XVII). Figure
7
B shows that this chimeric factor could function efficiently as a transactivator
in
Drosophila
cells. Hence, the C-terminal transactivation domain is functionally conserved between human
and insect HSFs.
To analyze regions of
Drosophila
HSF involved in transactivation of target genes we constructed chimeric genes
composed of the DNA binding domain of the yeast transcription factor GAL4 fused
to various segments derived from the C-terminal half of HSF. After transfection, the constructs were shown to
yield chimeric proteins which were constitutively localized to the cell nucleus
and exhibited DNA binding activity independent of heat shock. Analysis of the
ability of the GAL4-HSF proteins to activate transcription of a CAT reporter gene revealed
that transactivation required HSF residues 610-691. This domain of 80 amino
acids is located immediately C-terminal to leucine zipper 4, previously implicated in controlling the
inactive, monomeric conformation of HSF (
18
); hence, leucine zipper 4 is dispensable for transactivation.
Our results also revealed stress-independent negative regulation of the transactivation domain imposed by an internal region located
between residues 405-610 of
Drosophila
HSF. It is of interest that a study of GAL4-human HSF1 chimeras in human cells (
52
) also showed that a corresponding region of HSF1 conserved between human and
chicken factors could negatively regulate the C-terminal transactivation domain; in this case however, the negative
regulation was relieved upon heat shock. The disparity in the response to heat
stress could be due to the cell type-specific nature of such heat-inducible transactivation, which was observed in HeLa cells but not
in COS cells (
52
).
Although we were unable to clearly identify domains important for the heat shock
regulation of transactivation in the C-terminal half of
Drosophila
HSF, it is probable that such regions may exist in the N-terminal parts of HSF absent from the chimeras analyzed in this study.
Regions important for the suppression of transactivation under normal
conditions have been mapped to the N-terminal half of yeast HSF including leucine zippers 1-3 (
45
,
46
,
48
). Indeed, after submission of this manuscript, two studies (
59
,
60
) have reported contributions of the N-terminal regions of human HSF1 towards regulating the transactivation
domain in a heat shock-dependent manner. It will be of interest to confirm such regions in
Drosophila
HSF by additional studies.
Potent C-terminal transactivation domains encompassing leucine zipper motifs were described in HSFs from yeast
Saccharomyces cerevisiae
and
Kluyveromyces lactis
(
45
-
47
). In the case of tomato HSFs, the transactivation domain was localized near the C-terminus and a structural repeat containing conserved tryptophan residues was suggested to be involved in transactivation (
30
). Recently, a transactivation domain was mapped to the C-terminus of human HSF1 (
52
,
59
,
60
). The location of the transactivation domain near the C-terminus of
Drosophila
HSF demonstrates a similarity in the overall organization of HSF between
insects, plants and vertebrates.
Comparison of
Drosophila
HSF and human HSF1 protein C-terminal sequences shown in Figure
8
reveals the expected similarities in the conserved leucine zipper 4 region
important for regulating trimer formation. In the adjacent C-terminal region, there are many conservative exchanges between
Drosophila
HSF and human HSF1. The whole region contains abundant hydrophobic and acidic
residues in contrast to very few basic amino acids. The importance of these
residues in the function of the
Drosophila
HSF will require more detailed mutagenesis studies. It is of interest that the
C-terminal region of human HSF1 is able to functionally substitute for the
corresponding region of
Drosophila
HSF, suggesting the possibility of a structural conservation, despite limited
sequence identity.
We would like to thank all members of our laboratory for critical comments
during preparation of this manuscript. This work was supported by the
intramural research program of the National Cancer Institute. R.A. was a Howard
Hughes Medical Institute-National Institutes of Health Research Scholar.
REFERENCES