Báo cáo Y học: Determination of the consensus binding sequence for the purified embryonic heat shock factor 2 potx

Determination of the consensus binding sequence for the purified
embryonic heat shock factor 2
Martine Manuel
1,
*
,
†, Murielle Rallu
1,
*
,
‡, Marie-The
´
re
`
se Loones
1
, Vincenzo Zimarino
2
, Vale
´
rie Mezger
1
and Michel Morange
1
Laboratoire de Biologie Mole
´
culaire du Stress, Unite
´
de Ge
´
ne

´
tique Mole
´
culaire UMR8541, Ecole Normale Supe
´
rieure, Paris, France;
2
DIBIT, San Raffaele Scientific Institute, Milan, Italy
Heat shock transcription factors (HSFs) are characterized
by their ability, upon activation, to bind to heat shock
response elements (HSE) present in the promoter of their
target genes. HSE are composed of inverted repeats of the
pentamer nGAAm. In this study, we compare the
embryonic HSF2 protein, purified from F9 embryonal
carcinoma cells tumor, and the in vitro synthesized HSF2.
We show that the context of HSF2 synthesis influences its
thermosensitivity and DNA-binding properties. Therefore,
we determined the consensus binding sequence for the
purified embryonic HSF2 by the technique of systematic
evolution of ligands by exponential enrichment (SELEX).
We show that embryonic HSF2 prefers sites containing
three or four nGAAm inverted pentamers and that its
optimal binding sequence contains the 8-mer palindromic
core 5¢-TTCTAGAA-3¢. The consensus binding sequence
for the embryonic HSF2 will be very helpful to identify
new targets for this factor, during developmental and
differentiation processes.
Keywords: heat shock transcription factor-2; protein purifi-
cation; cooperativity; SELEX; consensus binding sequence.
Heat shock factor 2 (HSF2) belongs to the vertebrate heat

shock factor family that also includes HSF1, HSF3 and
HSF4 [1–5]. The members of the HSF family are defined by
their ability to specifically bind the regulatory sequence heat
shock element (HSE) [6]. Located in the regulatory regions
of heat shock genes, HSE consists of the inverted repeat of a
basal element nGAAm [7]. Two inverted repeats are
sufficient for Drosophila HSF binding, but optimal binding
is obtained with three repeats [8]. In agreement with this
observation, the activated form of HSFs has been demon-
strated to be a trimer in yeast [9], in Drosophila [10], in
human [11,12] or in mouse [13]. The HSE-binding activity of
heat shock factors is not constitutive, but induced by
various stresses, by differentiation or developmental pro-
cesses. HSF1 and HSF3 are activated by stresses that elicit
the so-called Ôheat shock responseÕ and induce the tran-
scription of heat shock genes. HSF1 corresponds to the
paradigm member of the family and is the functional
homolog, for its function in the heat shock response, of the
unique HSF found in yeast and Drosophila. Avian HSF3 is
activated by more severe stresses than HSF1, but is also
required for an optimal response to stress [14,15]. Indeed,
avian cells expressing HSF1, but in which the HSF3 gene
has been disrupted, exhibit a diminished response to stress,
even at mild heat shock temperatures [14]. Athough
heterotrimers were never detected, HSFs may interact with
each other in a more complex way.
HSF4 is an exception and constitutively binds DNA as a
trimer in the absence of stress. Its expression is regulated in a
tissue-specific manner [5,16]. The Hsf4 gene generates both
an activator or a repressor of heat shock genes by alternative

splicing; the tissue-specificity of the two forms may create a
modulation of expression of hsps in the different tissues.
In contrast to HSF1 and HSF3, HSF2 is not activated in
response to heat shock or other cellular stresses. It is found
in a trimeric DNA-binding form during hemin-induced
differentiation of the human erythroleukemia cells K562, in
mouse embryonal carcinoma (EC) cells, and during mouse
embryogenesis and spermatogenesis. During the differenti-
ation of K562 cells, HSF2 is converted from an inert dimeric
form to a DNA-binding trimer that is able to induce the
transcription of Hsp70 gene [17–19]. In this system, it seems
that although HSF1 and HSF2 are activated by distinct
signals, they also induce a similar profile of heat shock gene
transcription [17,18]. It was therefore suggested that in
mammalian cells, HSF1 was responsible for heat shock gene
induction upon stress, while HSF2 was responsible for the
high spontaneous expression of heat shock genes, which is
observed in the absence of stress in EC cells, and during
mouse embryogenesis and spermatogenesis.
However, an accumulation of data shows that the
contribution of HSF2 to the transcriptional regulation of
heat shock genes remains unclear. Indeed, athough HSF2
Correspondence to M. Morange, Laboratoire de Biologie Mole
´
culaire
du Stress, Unite
´
de Ge
´
ne

´
tique Mole
´
culaire UMR8541, Ecole Normale
Supe
´
rieure, 46 rue d’Ulm, 75230 Paris cedex 05, France.
Fax: + 33 1 44 32 39 41, Tel.: + 33 1 44 32 39 46,
E-mail:
Abbreviations: HSF, heat shock transcription factor; HSE, heat shock
response elements; SELEX, systematic evolution of ligands by
exponential enrichment; EC, embryonal carcinoma; in vitro
synthesized, i.v.s.
*Note: these authors contributed equally to this work.
Present address: Department of Biomedical Sciences,
University of Edinburgh, UK.
àPresent address: Developmental Genetics Program, Skirball Institute
for Biomolecular Medicine, NYU Medical Center, New York, USA.
(Received 5 December 2001, revised 28 February 2002,
accepted 5 April 2002)
Eur. J. Biochem. 269, 2527–2537 (2002) Ó FEBS 2002 doi:10.1046/j.1432-1033.2002.02917.x
displays a strong DNA-binding activity in EC cells [20,21],
the HSE region of Hsp70 promoter was found unoccupied
by HSF2 [21]. Controversial data suggest that, in contrast to
what was observed in mouse, HSF2 does not display any
DNA-binding activity at any stage of the rat seminiferous
epithelial cycle and that HSF2 expression does not correlate
with any HSP expression pattern [22]. No correlation is
found during pre- or post-implantation embryogenesis
between the expression patterns of major HSPs and HSF2

profiles [23]. Even in the case of the K562 cell system, where
the HSE sites were found occupied in vivo by HSF2 during
hemin-induced differentiation [18], other data suggest a role
of HSF1 and not HSF2 in the hemin-induced transcription
of Hsp70 gene [24], re-addressing the respective role of the
two factors in Hsp70 expression during differentiation.
Therefore, the role of HSF2 during differentiation and
development is likely distinct from a simple inducer of heat
shock genes in nonstress conditions, in differentiation or
developmental situations. Its role is still not unravelled and
its targets as a transcription factor unknown.
Studies performed on recombinant HSF1 and HSF2,
produced in E. coli, using random oligonucleotide selection
have shown that they display slightly distinct preferences,
although both factors bind to the 5¢-nGAAm-3¢ basal motif
[25]. Recombinant HSF2, in contrast to HSF1, does not
bind to HSE in a cooperative manner. We purified HSF2
from F9 mouse embryonal carcinoma tumors and analyzed
its DNA-binding properties at various temperatures in
comparison with in vitro synthesized (i.v.s.) HSF2 protein,
produced in reticulocyte lysates. This study demonstrates
that the DNA-binding properties of the purified HSF2 are
different from those of the i.v.s. HSF2. This suggests that
HSF2 function is highly sensitive to the environment in
which it is synthesized. We therefore decided to determine
the consensus binding sequence for the purified embryonic
factor, by a SELEX assay using a semirandom oligonucle-
otides library. We found that the embryonic factor requires
at least three 5¢-nGAAm-3¢ motifs and that its optimal
binding sequence contains a palindromic 8-mer core

5¢-TTCTAGAA-3¢. This result is in contrast to what was
found for the recombinant HSF2.
MATERIALS AND METHODS
Oligonucleotides
The oligonucleotides used in this study are shown in Table 1.
Embryonal carcinoma (EC) cell culture, acquisition
of tumors in 129 mice and purification of HSF factors
F9 EC cells were grown and extracts were prepared as
previously described [20]. F9 tumor cells were obtained by
subcutaneous injection of 2 · 10
6
F9 cells in 5-week-old
syngenic mice (strain 129). Tumors were allowed to grow for
about 2 weeks. After cervical dislocation, the tumors were
rapidly dissected and immediately frozen in dry ice until use
for extraction. Appropriate measures were taken to minim-
ize animals pain or discomfort, in accordance with the
European Communities Council Directive of 24 November
1986 (86/609/EEC).
For HSF2 protein purification, 26 g of crude material
(18 tumors) were extracted with 300 mL of extraction
buffer (10 m
M
Hepes pH 7.9, 0.4
M
NaCl, 0.1
M
EGTA,
0.5 m
M

dithiothreitol, 5% glycerol, 0.5 m
M
phenyl-
methanesulfonyl fluoride supplemented with 1 lgÆmL
)1
pepstatin and 1 lgÆmL
)1
aprotinin). Whole-cell extracts
were clarified by centrifugation for 30 min at 100 000 g
and the supernatants were stored at )80 °C. The final
protein concentration of the extracts averaged
5.8 mgÆmL
)1
. The complete purification of HSF2 protein
was performed by adaptation of a three-step protocol
previously described by Wu et al.[26].
(a) Whole-cell F9 tumor extracts were applied on an
heparine-sepharose column (CL-6B, Pharmacia), washed
with 300 mL of equilibration buffer (0.15
M
NaCl, 20 m
M
Hepes pH 7.9, 0.1 m
M
EGTA, 10% glycerol, 0.5 m
M
PMSF, 1 lgÆmL
)1
pepstatin and 1 lgÆmL
)1

aprotinin).
Bound proteins were eluted with a linear salt gradient
from 0.15 to 1.5
M
NaCl. Fractions were analyzed by
electromobility shift assay (EMSA) and those containing
an HSE-binding activity (0.2 to 0.6
M
NaCl) were pooled.
The yield and purification factors were calculated for this
column and were found to be equal to 87% and 5.6,
respectively.
(b) A DNA-affinity resin was prepared by coupling
HSE sequences to a CNBr-activated sepharose (CL-4B;
Pharmacia Biotech), according to Kanodaga and Tjian
[27]. The synthetic HSE oligonucleotide CTAGAAGCTT,
similar to that of Sorger & Pelham [28], was annealed with
itself in order to form double-stranded molecules with
protruding ends, which were subsequently ligated. This
resulted in the formation of polymers of about
100–200 bp, as estimated by agarose gels, that were linked
to the resin. Fractions containing the HSE-binding activity
were pooled and adjusted, by dilution, to 0.35
M
NaCl,
26 m
M
Hepes pH 7.6, 20% glycerol, 0.3 m
M
dithiothrei-

tol, 0.5 m
M
phenylmethanesulfonyl fluoride. The diluted
fractions were incubated overnight at 4 °C, under gentle
agitation, in the presence of resin, protease inhibitors
(1 lgÆmL
)1
pepstatin and aprotinin) as well as
1.5 lgÆmL
)1
poly(dI-dC).poly(dI-dC) to avoid nonspecific
interactions. After extensive washing with 0.2
M
NaCl
equilibration buffer (26 m
M
Hepes pH 7.6, 20% glycerol,
0.1% NP40, 0.1 m
M
EGTA, 0.3 m
M
dithiothreitol,
Table 1. Oligonucleotide sequences used in this study.
HSE2 5¢-TCGACAGATCTCCTAGAACGTTCTAGA
AGCTTCGAGAGGATTC-3¢
2U519m 5¢-CAGAATCTTCTCGATAGTTAGG-3¢
SHVAL 5¢-CTAGAACGTTCTAGAAGCTTCGAGA-3¢
SHVAL-SPZ 5¢-CTAGAACGTTCTAGAGAGTTTCCAG-3¢
NOG 5¢-CTAGAACGTTCTAGGGGGGGGGG-3¢
NOA 5¢-CTAGAACGTTCTAAAAAAAAAAA-3¢

MTH 5¢-CTAGAACGTTCTAAAAATTTCCAG-3¢
MCL 5¢-CTAGAACGTTCTAAAAAATTTCCAG-3¢
SHC 5¢-CTAGAACGTTCTAGAGAGAGAGAGA-3¢
JUL 5¢-CTAGAACGTTCTAGAACGTTCTCA-3¢
Deg-sb 5¢-CACGTGCGCTGGTACN
3
GAANNTTC
N
14
GGCTATCGACTGGCG-3¢
CL39 5¢-ATGGAACATTCTAGAACCTTCTCTT-3¢
CL83 5¢-AGAGAACATTCTAGAACATGGGTAC-3¢
83woTA 5¢-AGAGAACATTCACGAACATGGGTAC-3¢
39woGAA 5¢-ATGCACCATTCTAGAACCTTCTCTT-3¢
83woGAA 5¢-AGACACCATTCTAGAACATGGGTAC-3¢
2528 M. Manuel et al. (Eur. J. Biochem. 269) Ó FEBS 2002
0.5 m
M
phenylmethanesulfonyl fluoride, 1 lgÆmL
)1
pepst-
atin and aprotinin), proteins specifically bound to the resin
were eluted by steps of increasing NaCl concentration
(0.3 to 2
M
). Fractions were analysed by EMSA and those
containing HSE-binding activity (0.6 to 1.6
M
NaCl) were
pooled, concentrated 20-fold and analyzed by SDS/

PAGE. Silver staining of the gel revealed the presence
of several bands.
(c) Therefore, active fractions were subjected to a second
DNA-affinity chromatography. Fractions from the first
affinity chromatography were brought a second time to
0.35
M
NaCl, incubated with the HSE-affinity resin and
eluted exactly as before. Elution of HSE-binding activity
occured between 0.6 and 1.7
M
NaCl. The active fractions
were pooled and used for gel-shift assay.
During the two successive steps of DNA-affinity chro-
matography, fractions were collected in silanized tubes to
prevent sticking on plastic walls. Total yields in HSE-
binding activity and protein amount were estimated and
allowed to calculate a total purification factor equal to 3000.
SDS/PAGE and Western-blot analysis
Fractions containing HSF2 protein were pooled and loaded
on G25 sephadex (NAP columns; Pharmacia Biotech) in
order to discard most of the salts. Eluted material from the
Sephadex columns was then lyophilized and resuspended in
water so that the final volume was 100-fold less than at the
beginning. Three quarters of this concentrated material was
loaded on a 10% polyacrylamide gel and revealed by silver
staining in parallel with known concentrations of BSA to
estimate the amounts of purified protein. The last quarter
was used for Western blotting after transfer to a nitrocel-
lulose filter. HSF2 polyclonal antibodies were used at

1 : 2500 dilution as previously described [23]. Detection was
performed using the ECL peroxydase detection system
(Amersham).
Electromobility shift assays (EMSA)
Binding reactions were performed as described previously
[20]. Both strands of the DNA template were
32
P end-
labeled using T4 polynucleotide kinase and [c-
32
P]ATP.
Fourteen microliters of extracts containing 10–20 lgof
proteins from crude extracts, 0.7 ng of pure HSF2 protein
or 3 lLofin vitro translated proteins were mixed with 9 lL
of binding solution [0.2 ng of
32
P-labeled double-stranded
DNA template, 4 lg of double stranded polydI-dC, 9% (w/
v) Ficoll, 44 m
M
Hepes pH 7.6, 2.2 m
M
MgCl
2
and 88 m
M
KCl]. In competition experiments, 20 ng of unlabeled
double stranded DNA template were added to the binding
solution. The reaction mixtures were loaded on a 4%
acrylamide gel (acrylamide/bisacrylamide, 29 : 1, w/w) in

0.25 · Tris/borate/EDTA buffer.
Analysis of HSF2 thermosensitivity properties
HSF2 factors (i.v.s. or embryonic) were incubated at a
moderate (37 °C) or high temperature (44 or 45 °C) and
samples were taken at increasing periods of time, brought to
room temperature and subjected to the binding reaction in
presence of the oligonucleotide HSE2. Samples were then
immediately loaded on the migrating gel. Quantification of
the signal in the specific retarded complexes was performed
using a Bas1000 Imager (Fuji) after 1 h exposure. Arbitrary
values measured at distinct incubation times were standard-
ized to the initial value.
Multiple probes band shift assay
Synthetic oligonucleotides, containing an increasing num-
ber of the conserved 5 bp units nGAAm (organized in
contiguous arrays where each unit is inverted relative to
the immediately flanking one) and their complementary
strands were obtained from Genset (Paris, France). The
same oligonucleotides as those described by Xiao et al.
[29] were used, where n and m are A and T, respectively,
for GAA and TTC. Flanking sequences, added to this
core region in order to limit self-annealing, were identical
to those present at the ends of the oligonucleotide used
for affinity chromatography. According to the number of
repeats, oligonucleotides were named Rep2, Rep3, Rep4,
Rep5 and Rep6.
Binding reactions were performed as described above,
except that the binding solution contained a total amount
0.2 ng of
32

P end-labeled double-stranded oligonucleotides
corresponding to a mixture of Rep2, Rep3, Rep4, Rep5
and Rep6 at the same concentration. Protein extracts and
range of protein amounts used to perform these experi-
ments were as follows: 0.7 ng of HSF2 protein purified to
homogeneity (supplemented with 200 lgofBSA),and
3 lL of recombinant HSF2 protein expressed in reticulo-
cyte lysates. Binding reactions were performed at room
temperature during increasing periods of time ranging
from 0.5 min to 3 h and were followed by pore exclusion
limit electrophoresis. Samples were loaded on a 3–10%
gradient acrylamide gel (acrylamide/bisacrylamide, 29 : 1,
w/w) and migration was performed for 6 h at 350 V in
0.25 · Tris/borate/EDTA buffer, until the complexes
reached a position in the gel preventing their migration.
The position of specific complexes was detected by direct
autoradiography.
Bands containing the specific complexes as well as free
DNA were cut out of the gel and oligonucleotides present in
these gel slices were eluted overnight in distilled water at
37 °C. Samples were extracted once in phenol-chloroform
and once in chloroform, then concentrated in speed-vacuum
apparatus. They were then directly resuspended in the
sequencing loading buffer and analyzed on a denaturing
10% polyacrylamide gel in 1x TBE. The relative amounts of
the different oligonucleotides contained in each band were
quantified as previously described.
SELEX assay
The SELEX procedure was performed according to a
strategy described previously [30].

Preparation of a random sequence library
The 55-mer oligonucleotides Deg-sb (5¢-CACGTGCGC
TGGTACN
3
GAAN
2
TTCN
14
GGCTATCGACTGGCG-
3¢), containing two inverted trimers GAA and 19 random
nucleotides, and two PCR primers: P1, corresponding to the
first (top strand) 15 bases, and P2, complementary to the
last (bottom strand) 15 bases, were manufactured by
Ó FEBS 2002 Determining the optimal binding sequence for HSF2 (Eur. J. Biochem. 269) 2529
Eurobio (Les Ulis, France). A random sequence library,
Sel0, was generated by a primer extension reaction carried
out with Deg-sb as template and the (bottom) primer P2.
800 pmol of Deg-sb, annealed to a mix of 1600 pmol of cold
P2 and 80 pmol of radiolabeled P2, were extended with
100 U of Klenow fragment in a 200-lL Klenow reaction
mixture. The extended products were purified on a 12%
acrylamide gel.
Selection and amplification of sequences
that bind HSF2
Sel0 (450 ng in 90 lL of binding solution) was mixed with
30 mL of pooled elution fractions of purified embryonic
HSF2 ( 7 ng) and 1 mg of BSA. The reaction mixture
was incubated 15 min at room temperature and loaded on
a 4% acrylamide gel. After migration, the wet gel was
wrapped with Saran and exposed to X-ray film. The gel

region harboring HSF2-Sel0 complexes was localized by
comparison with the electrophoretic mobility of the
HSF2-radiolabeled Shvalspz complex that was loaded
on the adjacent control lane. An appropriate gel slice was
excised (large enough to take into account the smeary
binding pattern displayed by the purified HSF2 protein)
and soaked overnight at 37 °C in elution buffer (0.3
M
NaCl, 1 m
M
EDTA, 0.1% SDS). The eluted DNA was
purified on a Sephadex G-25 column (NAP-25 column,
Pharmacia Biotech) and concentrated to a volume of
50 lL in water. A 5-lL sample was added to a PCR
mixture together with 75 pmol of primer P1, 75 pmol of
primer P2 and 2.5 U of Tfl DNA polymerase (Promega)
in a final volume of 100 lL containing 20 m
M
Tris/acetate
(pH 9), 10 m
M
ammonium sulfate, 75 m
M
potassium
acetate, 0.05% Tween 20, 1.25 m
M
MgSO
4
,and75l
M

of each dNTP. Eight such reaction mixtures were set up.
The samples were heated for 1 min at 94 °C (hot start).
For each 35 cycles of PCR, samples were denatured at
94 °C for 30 s, annealed at 46 °C for 30 s, and extended
at 75 °C for 15 s. All eight reaction mixtures were pooled
and the DNA was purified on a 12% acrylamide gel.
About 50–100 ng of DNA was used for the next cycle of
the SELEX procedure.
Cloning of the products of selection
DNA amplified from the last cycle of selection was rendered
blunt-ended using T4 DNA polymerase and inserted at the
EcoRV site of pBluescript (pKS+, Stratagene).
Sequencing of the products of selection
After each round of selection, the amplified selected
DNA was sequenced as follows using the T7-sequencing
kit from Pharmacia Biotech with the following modifi-
cations to take into account the short size of the
sequences. P2 (10 ng) was end-labeled with [c-
32
P]ATP
and annealed to 10 ng of selected DNA in a 14-lL
volume containing 0.15
M
Tris/HCl (pH 7.6), 15 m
M
MgCl
2
and 23 m
M
dithiothreitol. The mix was boiled

for 5 min and left on ice for 10 min 4 U of T7 DNA
polymerase in 2 lL of dilution buffer [20 m
M
Tris/HCl
(pH 7.5), 5 m
M
dithiothreitol, 100 lgÆmL
)1
BSA and 5%
glycerol], 4 lLof33m
M
NaCl and 1 lLof100m
M
MnCl
2
, 150 m
M
sodium isocitrate were added to the
annealing mix on ice. 4.5 lL of this mixture were added to
2.5 lL of each of the four ddNTP Mix-Short solutions
[840 l
M
each dN
1
TP, dN
2
TP, dN
3
TP; 93.5 l
M

dN
4
TP;
14 l
M
ddN
4
TP; 40 m
M
Tris/HCl (pH 7.6) and 50 m
M
NaCl]. The reaction mix was incubated at 37 °Cfor
20 min The sequences were analysed on a 10% denaturing
acrylamide gel.
Individual Sel6 clones were sequenced using the T7
sequencing kit from Pharmacia Biotech and the T7 primer
according to the manufacturer’s instructions.
RESULTS
Purification of mHSF2 protein from EC cells
Sufficient starting amounts for the purification of HSF2
protein were obtained from tumors of F9 embryonal
carcinoma cells. These tumors were produced by injection
of F9 cells, in which HSF2 is highly expressed, under the
skin of syngenic mice. We verified that extracts produced
from tumor cells displayed an HSE-binding activity similar
to that of extracts from in vitro cultivated F9 cells (data not
shown), showing that mouse or tissue manipulations did not
uncover any stress-inducible activity (due to HSF1 protein).
The complete procedure for HSF2 purification combined
heparin and DNA affinity chromatographies [26]. HSF2

protein elution profile was monitored by the presence of an
HSE-binding activity in gel-shift assay (at room tempera-
ture). The first step of this purification procedure (i.e. the
heparine–sepharose chromatography) led to the separation
of HSF2 protein from 80% of the proteins present in crude
extracts. The following steps consisted of two HSE-affinity
chromatographies (see Materials and methods). After the
first one, HSF2 protein was separated from most of the
remaining proteins but a few of them were still co-eluted
with it. Therefore, HSF2-containing fractions were reloaded
on the same column in order to obtain a pure protein.
Analysis on SDS/PAGE after silver-staining showed one
unique band of  70 kDa (Fig. 1A). This band was
recognized by HSF2 antibodies (Fig. 1B) and comigrated
with one protein product present in reticulocyte lysates
expressing HSF2 protein. Thus, it appeared that HSF2
protein from F9 embryonic cells was purified to near
homogeneity. The purification factor was estimated to be
equal to 3000.
The pure protein was stable at )70 °C and could sustain
more than two cycles of freeze-thawing. However, gel shift
assays with pure protein gave poor reproducible results, and
we considered that, at these low protein concentrations, the
rare molecules of HSF2 protein might stick on the tube
walls, even when silanized. Therefore, we added 200 lgof
BSA to each point of binding reaction and got a reprodu-
cible stabilization of the purified HSF2 protein. We called
HSF2 purified from F9 tumor cells Ôembryonic HSF2Õ.
Conditions of binding and elution of HSF2 protein, in the
affinity column, gave several informative results about its

properties. Indeed, whereas binding conditions of HSF2
protein to the heparine–sepharose resin were similar to that
of Drosophila HSF, conditions used for the HSE-DNA
affinity chromatography were quite different. HSF2-con-
taining fractions required a longer incubation time with the
resin in order to bring the reaction to completion and the
2530 M. Manuel et al. (Eur. J. Biochem. 269) Ó FEBS 2002
ionic strength had to be increased to 0.35
M
NaCl (in
comparison to 0.25
M
NaCl for Drosophila HSF). In fact,
we showed that optimal binding to HSE sequences occurred
at slightly higher NaCl concentrations for HSF2 protein
(present in extracts from F9 control cells) than for HSF1
(present in extracts from F9 heat-shocked cells), the
Drosophila HSF homolog, which could explain the discrep-
ancy observed between HSF2 and Drosophila HSF in
binding the resin (Fig. 2). Besides this differential sensitivity
to ionic strength conditions, other components of the
binding buffer did not differentially affect HSF2, except for
MgCl
2
(optimal concentrations: 0 m
M
for HSF2, 1 m
M
for
HSF1), which appeared slightly detrimental to HSF2

binding to DNA (data not shown).
The purified embryonic HSF2 protein displays
a different thermosensitivity than the
i
.
v
.s. factor
I.v.s. HSF1 and HSF2 proteins display very distinct
behaviors. HSF1 protein produced in reticulocyte lysates
is active for DNA binding, provided that the extracts have
first been heated. In contrast, HSF2 protein shows a
constitutive HSE-binding activity but loses this activity
upon heat treatment [3,32]. Therefore, it appeared that the
DNA-binding activity of HSF2 protein was much more
sensitive than that of HSF1 protein, at least when synthes-
ized in vitro.
Using electromobility shift assay (EMSA), we analyzed
the thermosensitivity properties of the purified embryonic
HSF2 in comparison with the i.v.s. factor, produced in
reticulocyte lysates.
I.v.s. or embryonic purified proteins were incubated at
various temperatures before being subjected to EMSA. This
experiment allowed to analyze the sensitivity properties of
soluble HSF2 proteins, by measuring their remaining
capacity to bind their target sequences after exposure to
denaturating temperatures. The remaining ability of the
factors to bind a consensus target was quantified and
plotted as a function of time.
The inactivation ratio of pure embryonic HSF2 protein
was estimated to be about 20% after 20 min at 37 °Cand

80% after 20 min at 45 °C (Fig. 3). I.v.s. HSF2 protein was
also denatured by incubation at 37 °Cor45°C(Fig.3).At
high temperatures, the i.v.s. factor appeared to be signifi-
cantly more rapidly inactivated than the embryonic factor.
The inactivation of the i.v.s. protein observed at 37 °C
occurred in a limited manner and, unexpectedly, was
preceded by a transient phase of activation. Therefore,
incubation of the i.v.s. HSF2 at a moderate temperature
highly activated its DNA-binding abilities. This result was
uppermost striking as HSF2 appeared to be quite sensitive to
high temperature when synthesized in vitro [3]. Furthermore,
the pure embryonic factor did not behave in the same way.
Thus, HSF2 protein synthesized in the reticulocyte lysates
displayed a specific ability to become further activated
following a short exposure to a moderate temperature. This
seemed not to be characteristic of the factor itself but rather
of the conditions in which it had been produced.
The purified
i
.
v
.s. and embryonic HSF2 proteins exhibit
differences of cooperativity in DNA binding
In order to look for the cooperativity of HSF2 binding to
HSE sequences, we used the same methodology as that
Fig. 2. Effect of ionic (NaCl) strength on Heat-Shock Factors 1 and 2
DNA binding activities. Whole cell extracts from control unshocked
(F9C, corresponding to HSF2) or heat-shocked (F9HS, corresponding
to HSF1) F9 cells were incubated with labeled HSE oligonucleotide
under varying NaCl concentrations. After

PHOSPHORIMAGER
quanti-
fication, data were reported as fractions of the maximal value. Extracts
from heat-shocked cells (F9HS, HSF1) are plotted as circles; extracts
from control cells (F9C, HSF2) are plotted as triangles.
Fig. 1. Purification to homogeneity of HSF2 from F9 tumor extracts.
Elution fractions from the first and second cycle of HSE-affinity col-
umn (as well as HSF2 synthesized in reticulocyte lysates) were run on
SDS/PAGE after 100-fold concentration. (A) Silver staining. The
multiple bands observed above the 70 kDa i.v.s. HSF2 likely corres-
pond to additional proteins present in reticulocyte lysates. The smear
observed above the 70 kDa purified embryonic HSF2 is due to
remaining salts. (B) Western blot analysis using the HSF2 antiserum at
a 1 : 5000 concentration.
Ó FEBS 2002 Determining the optimal binding sequence for HSF2 (Eur. J. Biochem. 269) 2531
described by Liu-Johnson et al.[33],andappliedto
Drosophila HSF by Xiao et al. [29]. The strategy consists
in measuring the affinity of the factors for a series of
oligonucleotides containing an increasing number of the
binding motif. Immediately following the addition of the
different oligonucleotides, the factor recognizes equally well
all the oligonucleotides but, as time proceeds and if the
binding reaction is cooperative, the factor will bind more
and more preferentially to the sequences that contain a
higher number of motifs.
To be able to estimate quantitatively the cooperativity of
HSF2 binding to its sites, we measured, in the purified
complexes, the ratio of the oligonucleotides (rep4–6) as
compared to rep3. Both the i.v.s. and the purified embryonic
HSF2 display a higher affinity for sequences containing a

higher number of consensus trimers. In the case of the
purified embryonic HSF2, the ratio rep6/rep3 reaches the
plateau very quickly (15–20 min) (Fig. 4B). In contrast, in
thecaseofthei.v.s. HSF2, the ratio rep6/rep3 still increases
after 100 min (Fig. 4A). Therefore, the i.v.s. HSF2 displays
a higher cooperativity than the purified embryonic HSF2.
Embryonic HSF2 has a weak affinity for a good binding
sequence selected by the recombinant factor
Kroeger & Morimoto [25] had selected, from a random-
sequenced DNA library, sequences that could bind the
recombinant protein HSF2, synthesized in E. coli. Among
the selected sequences, the oligonucleotide 2U519 was
bound with a very good affinity.
We compared, by EMSA, the affinity of the purified
embryonic HSF2 for the oligonucleotide 2U519 and for the
oligonucleotides Shval and Shvalspz. Shval was commonly
used in the laboratory to detect HSE-binding activity in cell
extracts. It was designed according to the description, at
that time, of the basic heat shock response element; it
contains four inverted nGAAm pentamers. Shvalspz is a
modified version of Shval; it contains two inverted nGAAm
pentamers followed by the weak HSE sequence present in
the promoter of the putative target of HSF2 in spermato-
genesis, the testis specific gene Hsp70.2. While the i.v.s.
factor displayed, as expected, a better affinity for the
oligonucleotide 2U519 compared to Shval, in contrast, the
purified embryonic protein had a better affinity for Shval
(Fig. 5). Strikingly, the purified embryonic factor could bind
Shvalspz with a much higher affinity than 2U519 or Shval.
Fig. 4. Comparison of the cooperativity properties of in vivo and i.v.s.

HSF2. Proteins were incubated at room temperature with the mix of
rep oligonucleotides during increasing periods of time. The binding
mixture was then subjected to pore exclusion limit electrophoresis.
Bands containing the specific complexes as well as free DNA were cut
out of the gel and oligonucleotides present in these gel slices were
eluted. The relative amounts of the different oligonucleotides con-
tained in each band were quantified as previously described. The figure
shows the ratios rep6/rep3, rep5/rep3 and rep4/rep3 in the whole
complex, as a function of the incubation time. (A) I.v.s. HSF2. (B)
purified embryonic HSF2.
Fig. 3. Comparison of the thermosensitivity of in vivo and i.v.s. HSF2 at
37 °C and 45 °C. Proteins were incubated at 37 or 45 °Cduring
increasing times and thereafter submitted to gel-shift assay. Signal
amounts in the specific retarded complexes were quantified and then
compared to the initial value.
2532 M. Manuel et al. (Eur. J. Biochem. 269) Ó FEBS 2002
Determination of the consensus binding sequence
for the purified embryonic mHSF2 by SELEX
As we found that the recombinant and embryonic HSF2
proteins displayed significant differences in their DNA
binding properties, we decided to determine the consensus
binding sequence for the purified embryonic mHSF2 by the
technique of SELEX. We followed the procedure described
by Blackwell [30]. A semirandom-sequence oligonucleotide
library, Deg-sb, was ordered from Eurobio. These
55-nucleotide sequences contain a central semirandom
sequence, with 19 randomized nucleotides and two inverted
GAA trimers, flanked by two 15 fixed nucleotide sequences
for PCR amplification and sequencing. The two trimers in
the central sequence are separated by two randomized

nucleotides and they are located at 18 nucleotides from the
5¢ end of the sequence. The design of the semi-random
sequence was based on the high affinity of the embryonic
HSF2 observed for the oligonucleotide Shvalspz. The use of
a semi-random DNA library, in which a skeleton of HSE
site has been conserved, instead of a totally random DNA
library, allows a faster enrichment in HSF2 binding
sequences and the analysis of the selected sites using the
pool sequencing assay [30]. A double-stranded DNA
library, Sel0, was obtained from Degsb by a primer
extension reaction. Sel0 sequences were incubated with the
purified embryonic mHSF2 and the bound molecules, Sel1,
were isolated by EMSA, amplified by PCR, sequenced as a
pool and subjected to the next round of EMSA. After the
second cycle of selection, the sequence of the pool of selected
molecules revealed a significant enrichment in sequences
containing a third GAA trimer separated from the two fixed
trimers by a TA dinucleotide. This result was confirmed by
the next rounds of selection (Fig. 6). The SELEX assay was
stopped after six cycles of selection/amplification as no
difference could be observed between the sequence of Sel4,
Sel5 and Sel6.
At this step the preferred binding sequence for embryonic
HSF2 was 5¢-(A/G)(G/T)(A/G)GAA(C/T)(A/G)TTCTA
GAAN (A/G)(A/T)-3¢ (top strand), as could be determined
from the sequence of Sel6. Sel6 sequences were subcloned in
pBluescript and 57 individual clones were sequenced
(Fig. 7). Strikingly, almost all of the sequences contained
a third GAA trimer, one of them containing a GAT instead
of GAA, and 46 sequences displayed a TA dinucleotide

between the second and the third GAA trimer (all of the 11
remaining sequences displaying either the T or the A). 15
sequences contained a fourth inverted GAA trimer at the
right position, i.e. two nucleotides from the third trimer. A
consensus binding sequence was determined by calculating,
for each position of the central semi-random sequence, the
percentage of selected molecules containing each of the four
nucleotides (Fig. 8). This consensus sequence was composed
of the first fixed GAA trimer, a 8-mer TTCTAGAA core,
which was present in almost 100% of the selected sequences,
and a fourth inverted GAA motif. It is noticeable that the
preferred dinucleotide (top strand) preceding the first and
third GAA trimers is TA in both cases, and the preferred
dinucleotide (top strand) preceding the second and fourth
trimers is C
A
/
G
.
The affinity of the embryonic HSF2, present in F9 crude
extracts, for each of the 57 selected sequences was tested by
EMSA, and no significant differences were found between
them. HSF2 could bind the sequences containing three
trimers and those containing four trimers with comparable
affinities, as determined by the quantification of the
complexes (data not shown).
Complementary results on the important features
of the HSF2 binding site
We had noticed the very good affinity of HSF2 for the
oligonucleotide Shvalspz. In parallel with the SELEX assay,

Fig. 6. Comparison of the sequences of Sel0 and Sel4. The semirandom
sequence oligonucleotides before selection (Sel0) or after the fourth
round of selection (Sel4) were sequenced as a pool.
Fig. 5. Comparison of the affinity of the in vivo and i.v.s. HSF2 for the
oligonucleotides 2U519, Shval and Shvalspz.
32
P-Radiolabeled oligo-
nucleotides were incubated with i.v.s. or in vivo synthesized (embry-
onic) HSF2 and the HSF2-DNA complexes (arrow) were visualized by
EMSA and autoradiography.
Ó FEBS 2002 Determining the optimal binding sequence for HSF2 (Eur. J. Biochem. 269) 2533
we studied, by EMSA experiments, which features of the
oligonucleotide Shvalspz were crucial for the binding of
HSF2. For that purpose, several double-stranded oligonu-
cleotides were designed. In the oligonucleotides NOA and
NOG, the third and fourth (top strand) imperfect GAA
trimers of Shvalspz were replaced by a repetition of A, for
NOA, or G, for NOG. In SHC, the third trimer was kept
but the fourth imperfect trimer was replaced by a repetition
of the dinucleotide GA. In MCL and MTH, the third trimer
was replaced by a repetition of A. In MTH, the fourth
trimer was placed at the right position according to the
standard HSE sequence, i.e. seven nucleotides from the
second trimer instead of eight nucleotides in MCL. In JUL,
the 5¢ half (top strand) of Shvalspz, containing two perfect
inverted GAA trimers, was repeated twice.
The embryonic HSF2, present in E9.5 embryos or F9
cells crude extracts, could bind to JUL with a very good
affinity (Fig. 9). The sequence of JUL, as well as the
sequence of Shvalspz, are very close to the consensus

binding sequence determined by the SELEX assay (they
both contain the 8-mer core), which explains the very good
affinity of HSF2 for those oligonucleotides. HSF2 could not
bind to NOG (Fig. 9), NOA (not shown), MCL (Fig. 9)
and MTH (not shown). HSF2 could bind to SHC but with a
reduced affinity compared to Shvalspz (Fig. 9). These
results suggest that three adjacent nGAAm pentamers are
required for the binding of HSF2.
Similar experiments were carried out with double-stran-
ded oligonucleotides derived from two sequences selected by
Fig. 7. Sequences of individual Sel6 clones. After the sixth round of
selection, the selected oligonucleotides were subcloned in E. coli and
individual clones were sequenced.
Fig. 8. Consensus binding sequences for the purified embryonic HSF2
and for the recombinant HSF2. The consensus binding sequence for the
purified embryonic mHSF2 was obtained from the sequences of
individual Sel6 clones.
2534 M. Manuel et al. (Eur. J. Biochem. 269) Ó FEBS 2002
the SELEX assay, CL39 and CL83. CL39 contains four
GAA trimers while CL83 contains only three trimers. In the
oligonucleotides 39woGAA and 83woGAA, the first GAA
trimer (top strand) was replaced by a CAC trimer. In the
oligonucleotide 83woTA, the TA dinucleotide between
the second and third trimers (top strand) was replaced by
the AC dinucleotide. HSF2 could bind to 39woGAA with a
comparable affinity than to CL39 (Fig. 10A), but it could
not bind to 83woGAA (Fig. 10B). This result confirms that
the binding of HSF2 requires at least three nGAAm
pentamers. HSF2 could bind to 83woTA but with a slightly
reduced affinity compared to CL83 (Fig. 10B), confirming

the importance of the dinucleotide TA.
DISCUSSION
In this article, we have compared the active HSF2 protein
purified from EC cells (or present in E9.5 embryos or F9
cells crude extracts), called ÔembryonicÕ or Ôin vivo synthes-
izedÕ, with the active HSF2 protein synthesized in vitro in
reticulocyte lysates. The stability at various temperatures
and cooperativity of HSF2 synthesized in these different
contexts were analyzed and found different. Kroeger et al.
[34] mentioned that the environment in which the HSF is
synthesized determines its activation state. Indeed, in most
cell lines or adult tissues, HSF2 is not constitutively active,
while it is constitutively active when synthesized in E. coli or
in reticulocyte lysates. But in our case, differences relate to a
factor already active, either purified from F9 cells or
synthesized in reticulocytes lysates.
(a) HSE-binding properties at various temperatures are
highly dependent on the context in which HSF2 was
synthesized. The stability of the i.v.s. HSF2 and that of the
embryonic purified HSF2 are markedly different. Progres-
sive inactivation of HSE-binding activity, at 37 °C, is not
observed for the i.v.s. HSF2 as it was for the protein purified
from F9 cells. Instead, a significant stimulation of its DNA-
binding activity is observed, as if the activation of HSF2 in
reticulocyte lysates was incomplete and could be further
achieved in the reaction buffer. This abnormal behavior is
observed for short incubation times which explains that
former results reported that incubation at 37 °Cfor60min
had no effect on HSF2 DNA-binding activity [3].
(b) The i.v.s. HSF2 displays a higher cooperativity than

the purified embryonic HSF2. The cooperativity displayed,
in our hands, by the factor synthesized in vitro in reticulocyte
lysates is in contrast with what was observed for an HSF2
factor synthesized in E. coli and subsequently purified [25].
Because of the differences observed between the HSF2
protein synthesized in vivo and the protein synthesized in
reticulocyte lysates, we can conclude that the in vivo context
of synthesis is very important to give HSF2 its properties.
Therefore, we selected optimal binding sites for the
purified embryonic HSF2, from a semi-random DNA
library, and determined the consensus binding sequence for
this embryonic factor. We found that the embryonic HSF2
(purified from, or present in crude extracts of F9 cells)
preferred sites containing three or four nGAAm inverted
pentamers and that it was unable to bind to sites containing
only two adjacent pentamers. This is in contrast to what was
previously found for a recombinant HSF2 produced in
E. coli, which preferred sites containing two or three
pentamers [25]. The consensus binding site for the embry-
onic HSF2 is more stringent than the sites found for the
recombinant factors HSF2 and HSF1, as it contains a 8-mer
palindromic core TTCTAGAA that was very strongly
selected, and as the other ÔNÕ positions of each pentamer are
not random. The consensus binding site for the recombinant
proteins HSF2 and HSF1 was determined by selecting
sequences that could bind efficiently the factors from a
totally random DNA library. In contrast, we used a semi-
random DNA library for the SELEX assay to study the
purified embryonic factor. This could be an explanation for
the higher stringency found for the embryonic HSF2

consensus binding sequence, compared to what was
described for the recombinant proteins, as the use of a
semi-random DNA library, in which two inverted GAA
trimers were fixed, allowed a much faster enrichment in
sequences binding the transcription factor with a high
affinity. However, the inability of the embryonic HSF2 to
bind sequences containing only two nGAAm pentamers
shows that this factor has a higher requirement for its
binding site than the recombinant factor.
Using the program
TARGETFINDER
[35], we identified
several genes containing the consensus binding site for the
embryonic HSF2 in their promoter. Those genes are
Fig. 10. Comparison of the affinity of embryonic HSF2, present in F9
cells extracts, for the oligonucleotides CL39 and 39woGAA and for the
oligonucleotides CL83, 83woGAA and 83woTA.
32
P-Radiolabeled
oligonucleotides were incubated with F9 cells extracts and the HSF2–
DNA complexes were visualized by EMSA and autoradiography.
HSF2 binds CL83 and CL39 with a comparable affinity, as determined
by the quantification of the complexes (not shown). The binding data
presented in (A) and (B) are from separate gels.
Fig. 9. Comparison of the affinity of embryonic HSF2, present in E9.5
embryos extracts, for the oligonucleotides Shvalspz, JUL, SHC, MCL
and NOG.
32
P-Radiolabeled oligonucleotides were incubated with
E9.5 embryo extracts and the HSF2–DNA complexes (arrow) were

visualized by EMSA and autoradiography.
Ó FEBS 2002 Determining the optimal binding sequence for HSF2 (Eur. J. Biochem. 269) 2535
currently being studied as genes potentially regulated by the
transcription factor HSF2, taking advantage of our Hsf2
)/)
mice [36].
We conclude from this study that the intrinsic properties
of HSF2, illustrated by those of the purified embryonic
factor, can be deeply modified by the cellular context in
which it is synthesized. Therefore, HSF2 which is active for
DNA-binding until midgestation (for a longer period in the
brain) could have targets and transcriptional capacity
different from one organ to another or at various times
during the mouse development.
The consensus binding site for the embryonic HSF2
will help us to identify new targets for this factor, that
might be involved in developmental and differentiation
processes.
ACKNOWLEDGEMENTS
We are grateful to Dr Agne
`
s Delahodde and Dr Olivier Jean-Jean for
the helpful advice about the SELEX assay. This work and M. Manuel
were supported by a grant (9293) and a fellowship from the Association
pour la Recherche sur le Cancer.
REFERENCES
1. Rabindran, S.K., Giorgi, G., Clos, J. & Wu, C. (1991) Molecular
cloning and expression of a human heat shock factor, HSF1. Proc.
Natl Acad. Sci. USA 88, 6906–6910.
2. Schuetz, T.J., Gallo, G.J., Sheldon, L., Tempst, P. & Kingston,

R.E. (1991) Isolation of a cDNA for HSF2: evidence for two heat
shock factor genes in humans. Proc. Natl Acad. Sci. USA 88,
6911–6915.
3. Sarge, K.D., Zimarino, V., Holm, K., Wu, C. & Morimoto, R.I.
(1991) Cloning and characterization of two mouse heat shock
factors with distinct inducible and constitutive DNA-binding
ability. Genes Dev. 5, 1902–1911.
4. Nakai, A. & Morimoto, R.I. (1993) Characterization of a novel
chicken heat shock transcription factor, heat shock factor 3, sug-
gests a new regulatory pathway. Mol. Cell Biol. 13, 1983–1997.
5. Nakai, A., Tanabe, M., Kawazoe, Y., Inazawa, J., Morimoto, R.I.
& Nagata, K. (1997) HSF4, a new member of the human heat
shock factor family which lacks properties of a transcriptional
activator. Mol. Cell Biol. 17, 469–481.
6. Wu, C. (1995) Heat shock transcription factors: structure and
regulation. Annu.Rev.CellDevBiol.11, 441–469.
7. Amin, J., Ananthan, J. & Voellmy, R. (1988) Key features of heat
shock regulatory elements. Mol. Cell. Biol. 8, 3761–3769.
8. Perisic, O., Xiao, H. & Lis, J.T. (1989) Stable binding of Droso-
phila heat shock factor to head-to-head and tail-to-tail repeats of a
conserved 5 bp recognition unit. Cell 59, 797–806.
9. Sorger, P.K. & Nelson, H.C. (1989) Trimerization of a yeast
transcriptional activator via a coiled-coil motif. Cell 59, 807–813.
10. Westwood, J.T. & Wu, C. (1993) Activation of Drosophila heat
shock factor: conformational change associated with a monomer-
to-trimer transition. Mol. Cell Biol. 13, 3481–3486.
11. Baler, R., Dahl, G. & Voellmy, R. (1993) Activation of human
heat shock genes is accompanied by oligomerization, modifica-
tion, and rapid translocation of heat shock transcription factor
HSF1. Mol. Cell Biol. 13, 2486–2496.

12. Zuo, J., Baler, R., Dahl, G. & Voellmy, R. (1994) Activation of the
DNA-binding ability of human heat shock transcription factor 1
may involve the transition from an intramolecular to an inter-
molecular triple-stranded coiled-coil structure. Mol. Cell Biol. 14,
7557–7568.
13. Sarge, K.D., Murphy, S.P. & Morimoto, R.I. (1993) Activation of
heat shock gene transcription by heat shock factor 1 involves
oligomerization, acquisition of DNA-binding activity, and nuclear
localization and can occur in the absence of stress. Mol. Cell Biol.
13, 1392–1407. [published errata appear in Mol. Cell Biol. (1993)
13, 3122–3123 and 13, 3838–3839]
14. Tanabe, M., Kawazoe, Y., Takeda, S., Morimoto, R.I., Nagata,
K. & Nakai, A. (1998) Disruption of the HSF3 gene results in the
severe reduction of heat shock gene expression and loss of ther-
motolerance. EMBO J. 17, 1750–1758.
15. Kawazoe, Y., Tanabe, M., Sasai, N., Nagata, K. & Nakai, A.
(1999) HSF3 is a major heat shock responsive factor during
chicken embryonic development. Eur. J. Biochem. 265, 688–697.
16. Tanabe, M., Sasai, N., Nagata, K., Liu, X.D., Liu, P.C., Thiele,
D.J. & Nakai, A. (1999) The mammalian HSF4 gene generates
both an activator and a repressor of heat shock genes by alter-
native splicing. J. Biol. Chem. 274, 27845–27856.
17. Sistonen, L., Sarge, K.D. & Morimoto, R.I. (1994) Human heat
shock factors 1 and 2 are differentially activated and can
synergistically induce hsp70 gene transcription. Mol. Cell Biol. 14,
2087–2099.
18. Sistonen, L., Sarge, K.D., Phillips, B., Abravaya, K. & Morimoto,
R.I. (1992) Activation of heat shock factor 2 during hemin-
induced differentiation of human erythroleukemia cells. Mol. Cell
Biol. 12, 4104–4111.

19. Theodorakis, N.G., Zand, D.J., Kotzbauer, P.T., Williams, G.T.
& Morimoto, R.I. (1989) Hemin-induced transcriptional activa-
tion of the HSP70 gene during erythroid maturation in k562 cells
is due to a heat shock factor-mediated stress response. Mol. Cell.
Biol. 9, 3166–3173.
20. Mezger, V., Bensaude, O. & Morange, M. (1989) Unusual levels of
HSE-binding activity in embryonal carcinoma cells. Mol. Cell.
Biol. 9, 3888–3896.
21. Murphy, S.P., Gorzowski, J.J., Sarge, K.D. & Phillips, B.
(1994) Characterization of constitutive HSF2 DNA-binding
activity in mouse embryonal carcinoma cells. Mol. Cell Biol. 14,
5309–5317.
22. Alastalo, T.P., Lonnstrom, M., Leppa, S., Kaarniranta, K., Pelto-
Huikko, M., Sistonen, L. & Parvinen, M. (1998) Stage-specific
expression and cellular localization of the heat shock factor 2
isoforms in the rat seminiferous epithelium. Exp.CellRes.240,
16–27.
23. Rallu, M., Loones, M., Lallemand, Y., Morimoto, R., Morange,
M. & Mezger, V. (1997) Function and regulation of heat shock
factor 2 during mouse embryogenesis. Proc. Natl Acad. Sci. USA
94, 2392–2397.
24. Yoshima, T., Yura, T. & Yanagi, H. (1998) Heat shock factor 1
mediates hemin-induced hsp70 gene transcription in K562 ery-
throleukemia cells. J. Biol. Chem. 273, 25466–25471.
25. Kroeger, P.E. & Morimoto, R.I. (1994) Selection of new HSF1
and HSF2 DNA-binding sites reveals difference in trimer
cooperativity. Mol. Cell Biol. 14, 7592–7603.
26. Wu, C., Wilson, S., Walker, B., Dawid, I., Paisley, T., Zimarino,
V. & Ueda, H. (1987) Purification and properties of Drosophila
heat shock activator protein. Science 238, 1247–1253.

27. Kadonaga, J.T. & Tjian, R. (1986) Affinity purification of
sequence-specific DNA binding proteins. Proc. Natl Acad. Sci.
USA 83, 5889–5893.
28. Sorger, P.K. & Pelham, H.R.B. (1987) Purification and char-
acterization of a heat-shock element binding protein from yeast.
EMBO J. 6, 3035–3041.
29. Xiao, H., Perisic, O. & Lis, J.T. (1991) Cooperative binding of
Drosophila heat shock factor to arrays of a conserved 5 bp unit.
Cell 64, 585–593.
30. Blackwell, T.K. (1995) Selection of protein binding sites from
random nucleic acid sequences. Methods Enzymol. 254, 604–618.
31. Sanger, F., Nicklen, S. & Coulson, A.R. (1977) DNA sequencing
with chain-terminating inhibitors. Proc. Natl Acad. Sci. USA 74,
5463–5467.
2536 M. Manuel et al. (Eur. J. Biochem. 269) Ó FEBS 2002
32. Mathew, A., Mathur, S.K., Jolly, C., Fox, S.G., Kim, S. &
Morimoto, R.I. (2001) Stress-specific activation and repression of
heat shock factors 1 and 2. Mol. Cell Biol. 21, 7163–7171.
33. Liu-Johnson, H.N., Gartenberg, M.R. & Crothers, D.M. (1986)
The DNA binding domain and bending angle of E. coli CAP
protein. Cell 47, 995–1005.
34. Kroeger, P.E., Sarge, K.D. & Morimoto, R.I. (1993) Mouse heat
shock transcription factors 1 and 2 prefer a trimeric binding site
but interact differently with the HSP70 heat shock element. Mol.
Cell Biol. 13, 3370–3383.
35. Lavorgna, G., Guffanti, A., Borsani, G., Ballabio, A. & Bonci-
nelli, E. (1999) TargetFinder: searching annotated sequence
databases for target genes of transcription factors. Bioinformatics
15, 172–173.
36. Kallio, M., Chang, Y., Manuel, M., Alastalo, T P., Rallu, M.,

Gitton, Y., Pirkalla, L., Loones, M T., Paslaru, L., Larney, S.,
Hiard, S., Morange, M., Sistonen, L. & Mezger, V. (2002) Brain
abnormalities, defective meiotic chromosome synapsis and female
subfertility in HSF2 null mice. EMBO J., in press.
Ó FEBS 2002 Determining the optimal binding sequence for HSF2 (Eur. J. Biochem. 269) 2537