ABSTRACT
We constructed plasmids encoding the sequences for the bZip modules of c-
Jun
and c-
Fos
which could then be expressed as soluble proteins in
Escherichia coli.
The purified bZip modules were tested for their binding capacities of synthetic
oligonucleotides containing either TRE or CRE recognition sites in
electrophoretic mobility shift assays and circular dichroism (CD).
Electrophoretic mobility shift assays showed that bZip
Jun
homodimers and bZip
Jun
/
Fos
heterodimers bind a collagenase-like TRE (CTGACTCAT) with dissociation constants of respectively 1.4*10-7 M and 5*10-8
M. As reported earlier [Patel
The AP-1 protein complex is a eukaryotic transcription factor that binds DNA in a
sequence specific manner. AP-1 was first identified by its role in the regulation of the human
metallothionein II
A
gene, and soon after it was shown that c-
Jun
and c-
Fos
are constituents of AP-1. Several related genes of
Jun
and
Fos
exist, giving rise to the families of either the
Jun
transcription factors or the
Fos
transcription factors. These members of the AP-1 family associate with each other to form a variety of homo- or heterodimeric complexes. A hallmark of AP-1 became its ability to mediate gene induction by the phorbol
ester tumor promoter 12-
O
-tetradecanoylphorbol-13-acetate (TPA) and led to the name TRE (TPA response element)
for its 7 bp recognition site (TGA(C/G)TCA), which is present in numerous
promoters. The same sequence is recognized by the transcriptional activator
GCN4, which is involved in the regulation of amino acid biosynthesis in yeast.
Binding of AP-1 to the TRE sequence can modulate transcription both positively and
negatively (for a review see
1
). The 8 bp recognition site CRE (cAMP responsive element) (TGACGTCA) is also a
target for
Jun
/
Jun
or
Jun
/
Fos
complexes (
2
,
3
).
Jun
and
Fos
belong to the bZip group of transcription factors, containing a DNA binding
domain with clustered basic amino acids (b) and an adjacent leucine zipper
motif (Zip) responsible for the dimerization of the transcription factors (for
a review on bZip proteins see 4). The leucine zipper mediated dimerization of
AP-1 proteins is concentration dependent and leads to increased [alpha]-helical secondary structure (
5
,
6
). Even as dimers, the basic domain of each monomer remains essentially
unstructured in solution. In the presence of DNA, the basic region adopts an [alpha]-helical conformation as a consequence of recognition and binding to
the consensus sequence (
5
). The crystal structure of the c-
Jun
/c-
Fos
bZip domain bound to a TRE sequence (
7
) showed that both subunits form continuous parallel [alpha]-helices. The C-terminal leucine zipper forms a coiled-coil structure. Both subunits make base-specific contacts with DNA in the major groove
via their basic N-terminal regions. These contacts are essentially the same as those
observed for the GCN4 bZip domain bound to a TRE (
8
) or a CRE sequence (
9
,
10
). In the latter case at least, the DNA undergoes a marked conformational change
including features reminiscent of A-DNA (C3'-endo conformation for certain deoxyriboses and an average
base-pair displacement of -1.4 Å). In the crystal structures of the GCN4-TRE and the
Jun
/
Fos
-TRE complexes, the DNA stays apparently closer to the B-conformation (
7
,
8
), although the average base pair displacement for the GCN4-TRE complex is far from neglectable (-0.9 Å) (
11
). In addition the unit vector normal to the base pairs also deviates
substantially from canonical B-DNA (
9
). Weiss
et al.
(
12
) have shown that the circular dichroism spectrum of both a TRE or a CRE
containing DNA is markedly increased upon addition of the GCN4 bZip domain. A
similar change has not been reported for the bZip domains of
Jun
and
Fos
(
5
). Here we show that the bZip domains of
Jun
and
Fos
also induce a marked increase in the CD spectra of a series of double-stranded oligonucleotides containing TRE and CRE sites. Within
experimental error the spectra of a bound TRE and a bound CRE are identical. On
the contrary, a
Jun
/
Jun
complex induces a somewhat more pronounced spectral change than a
Jun
/
Fos
complex for both TRE and CRE binding sites. These results were obtained with
recombinant bZip domains of c-
Jun
and c-
Fos
which are expressed as soluble proteins in
Escherichia coli,
and which do not contain any of the foreign helper sequences generally used for affinity purification.
The mouse c
-Jun
gene was first digested with
Sma
I and
Pst
I. The resulting 0.94 kb DNA fragment was purified and redigested with
Hin
fI and
Ban
I. The
Hin
fI/
Ban
I fragment codes for the bZip domain of mouse c-
Jun
spanning amino acids 250 to 327, a sequence that is identical to human c-
Jun
247 to 324. This
Hin
fI/
Ban
I fragment was cloned into the
Eco
RI/
Hin
dIII site of plasmid pKK223-3 (Amp
R
; Pharmacia, Uppsala, Sweden) giving rise to plasmid pKK-
Jun
. The synthetic oligonucleotide linkers (Table
1
) I (front end) and II (back end) were used to insert the
Jun
fragment into the vector. A 1.2 kb
Pvu
II fragment of the human c
-Fos
gene was recut with
Hin
fI and
Sau
3A. This fragment codes for the human c-
Fos
bZip domain, spanning amino acids 137 to 208. The
Hin
fI/
Sau
3A DNA was cloned into the same
Eco
RI/
Hin
dIII sites using linkers III (front end) and IV (back end), to give plasmid pKK-
Fos
.
Plasmids pKK-
Fos
and pKK-
Jun
were transformed into
E.coli
strain RB791 (
13
). The bacteria harboring the pKK-
Fos
plasmid were grown in LB medium (10 g/l Bacto-tryptone, 5 g/l yeast extract, 10 g/l NaCl, 100 mg/l ampicillin) at 37oC. Bacteria transformed with the corresponding pKK-
Jun
plasmid expressed higher amounts of
Jun
247-324
protein in minimal medium as compared with LB medium. These transformants were
grown at 37oC in M9 medium (8.5 g/l Na
2
HPO
4
.
2H
2
O, 3 g/l KH
2
PO
4
, 0.5 g/l NaCl, 1 g/l NH
4
Cl pH 7.4, 50 mg/l ampicillin) supplemented with 0.4% (w/v) glucose, 1 mM MgSO
4
, 50 [mu]M CaCl
2
, and 2 [mu]M thiamin. In both cases, an overnight starter culture was diluted 10-fold and incubated until A
550
= 0.8. Gene expression was induced with 1 mM ITPG, the cells grown for another 3
h, and harvested by centrifugation. About 12 g of wet bacterial paste was
resuspended in 48 ml CED-0.3 buffer (300 mM NaCl, 20 mM citric acid, 10 mM EDTA, 5 mM DTT, pH 5).
The lysis buffer was supplemented with 1 mM PMSF, 1 mM benzamidine, 1 [mu]M pepstatin, 5 [mu]M aprotinin; all further purification steps using CED buffer were done
in presence of 1 mM PMSF only. Cells were disrupted with a French press at 500 kg/cm
2
pressure. The cleared supernatant was batch-incubated overnight with 15 ml of CED-0.3 equilibrated Heparin Ultrogel (IBF, Villeneuve-la-Garonne, France) at 4oC. The suspension was transferred to a column and
washed with CED-0.3. A linear gradient using CED-0.3 and CED-2 (2 M NaCl, 20 mM citric acid, 10 mM EDTA, 5 mM DTT, pH 5; 1
mM PMSF) was applied to elute the proteins at a flow rate of 1 ml/min. The
collected fractions were analyzed using SDS-PAGE. Ammonium sulfate was gradually added to the pooled bZip fractions
under stirring up to 80% saturation. The precipitated proteins were dissolved
in HPLC buffer A (0.05% TFA/H
2
O) and injected on a semipreparative C-4 reversed phase coloumn (Macherey Nagel, Düren, Germany). A linear gradient between 32% and 40% HPLC buffer B
(0.05% TFA in acetonitrile) was applied at a flow rate of 2 ml/min. The
collected proteins were lyophilized and redissolved in deionized water. Protein
concentration was determined according to Scopes (
14
) and the proteins were stored at -20oC in the presence of 5 mM DTT to avoid cysteine oxidation.
Oligonucleotides were synthesized on an Applied Biosystems synthesizer and the
resulting single-stranded DNA purified by C-18 reversed phase HPLC. Equimolar amounts of complementary strands
were annealed in TE (10 mM Tris, 1 mM EDTA, pH 7.9). Samples were injected on
an analytical size exclusion HPLC column (SEC-2000, Beckman, Fullerton, CA, USA) equilibrated with phosphate buffered
saline (PBS; 150 mM NaCl, 10 mM Na-phosphate, pH 7.0). Those samples containing more than 5% of single
stranded DNA contaminants were purified on a SEC-2000 column, the collected fractions of dsDNA were ethanol precipitated,
dried and redissolved in TE buffer.
Proteins at varying concentrations were incubated with about 2500 c.p.m. of
32
P-labeled oligonucleotides in 10 [mu]l total sample volume. Equimolar amounts of separately purified
Jun
and
Fos
peptides were associated by an incubation of 30 min at 25oC prior to the formation of the protein-DNA complex. Under these conditions heterodimer formation and
subunit exchange of
Jun
and
Fos
bZip domains occurs with a half-life of less than 10 s (
15
). The DNA interaction buffer contained 50 mM Tris-HCl, pH 8.0, 5 mM MgCl
2
, 5 mM DTT, 1 mM EDTA, 10% glycerol, 0.1% Nonidet P-40, 2 [mu]g of bovine serum albumin, and 0.1 [mu]g of poly(dI-dC). The reaction mixtures were incubated for 30 min at room
temperature and subsequently run on a 5% polyacrylamide gel (with 1:19
bisacrylamide) at 4oC. The running buffer was 0.5* TBE (45 mM Tris, 45 mM Borate, 1 mM EDTA, pH 8.3). The data were
analyzed using either a Fuji 600 phosphoimager or a BioRad video scanning
system.
CD spectra were recorded on a Jobin Yvon dichrograph under constant nitrogen
flush at room temperature. The spectra were recorded between 200 and 320 nm
using 1 mm or 10 mm path-length quartz cells (Hellma, Müllheim, Germany). The proteins were dissolved in PBS-200 (200 mM NaCl, 10 mM Na-phosphate pH 7.5, 1 mM DTT) buffer. The spectra
presented here represent the baseline corrected mean of at least three scans of
the same sample. A starting concentration of 0.04 mg/ml of oligonucleotide was
used for the protein titration experiments. Equal volumes of protein stock
solutions were added until a maximum of DNA signal change was obtained. The
relative signal changes were corrected for dilution, which was kept beneath 3%.
The DNA CD spectra (Fig.
5
)
show the absolute signal change without correction for dilution.
The DNA binding capacity of the bZip modules were tested in electrophoretic
mobility shift assays (EMSA). For this purpose we used synthetic
oligonucleotides of different length (Fig.
1
B). One series of oligonucleotides contained a binding site derived from the
collagenase TRE, whereas the other series harbored the corresponding CRE site.
As a control, we used two oligonucleotides. TRE-21 mut. contained a mutated TRE site for which two base pairs in each TRE
half-site were simply swapped. These changes in the recognition site were not
sufficient to abolish protein binding activity completely, since some retarded
DNA could be detected in EMSA. The other control oligonucleotide, GEM-21 contained a scrambled, non-palindromic sequence with the same number of G/C and A/T base pairs
as the TRE consensus site. For this oligonucleotide no bound DNA was detectable
in EMSA. All specific oligonucleotides formed well-defined complexes with both the
Jun
/
Jun
and
Jun
/
Fos
bZip domains.
The TRE-21 collagenase-like oligonucleotide was used to determine the equilibrium
dissociation constant
K
d
of either the c-
Jun
247-324
homodimer or the c-
Jun
247-324
/c-
Fos
137-208
heterodimer DNA complex (Fig.
3
). The
K
d
value corresponds approximately to the protein concentration necessary to bind
50% of the DNA if the protein is in large excess over the DNA concentration.
For the c-
Jun
247-324
/c-
Fos
137-208
complex 50% of bound DNA is reached upon addition of about 5 * 10
-8
M of total protein concentration.
The bZip modules were further tested by circular dichroism spectroscopy.
Circular dichroism is a useful structural tool for the characterization of both
protein and nucleic acid conformation (for a recent review see ref.
18
). Conformational changes induced upon formation of a protein-DNA complex may be generally attributed to one or the other species,
since the composite spectrum is dominated by the DNA between 260-290 nm, and by the protein between 210-230 nm. In the following, protein CD spectra will be expressed as
usual by the molar ellipticity [Theta], whereas the DNA CD spectra are expressed as [Delta][epsilon] = [epsilon]
L
- [epsilon]
R
, where [epsilon]
L
and [epsilon]
R
represent the molar extinction coefficients of, respectively, left and right circularly polarized light. The numerical conversion between the two measures of CD is straightforward ([[Theta]] = 3298 [Delta][epsilon] ).
At a concentration of 20 [mu]M in PBS200, the c-
Jun
247-324
homodimer and the c-
Jun
247-324
/c-
Fos
137-208
heterodimer have [Theta] values at 222 nm of -20.000 degrees * cm
2
* dmol
-1
(Fig.
4
). Assuming a [Theta] value of -33.000 for 100% [alpha]-helical secondary structure (
19
,
20
), this indicates an [alpha]-helix content of about 60% for both the homo- and the heterodimer, corresponding quite closely to the
fraction of amino acids being part of the leucine zipper (Fig.
1
). The incomplete structure of the protein is thus most likely due to the
disordered N-terminal basic region and possibly also a disordered C-terminal tail of the bZip modules. c-
Fos
137-208
alone ([Theta] value of -11.000) appears to be only ~30% helical under these conditions. Furthermore, no DNA
binding activity could be detected for c-
Fos
137-208
neither in EMSA, nor in circular dichroism experiments (Fig.
4
).
Formation of the protein-DNA complex not only induces a conformational change of
the bZip modules, but also changes the DNA CD spectra, indicating changes in
the conformation of the DNA. Addition of either the c-
Jun
247-324
homodimer or the c-
Jun
247-324
/c-
Fos
137-208
heterodimer leads to a shift of the oligonucleotide spectra towards lower
wavelength and a marked increase in signal intensity of the positive CD band at
280 nm.
Figure
5
A and B show the CD spectra of TRE-21 in the absence and presence of the bZip domains. The c-
Jun
247-324
/c-
Fos
137-208
heterodimer (B) leads to a slightly less pronounced signal increase than the c-
Jun
247-324
homodimer (A), i.e. an increase of 1.48-fold at 280 nm instead of 1.65-fold, respectively.
Figure
5
C and D show the corresponding spectra of CRE-22. Again, the heterodimer induces a somewhat smaller change in the DNA
spectrum than the c-
Jun
247-324
homodimer (1.45-fold instead of 1.58-fold). The spectral changes for both complexes are even more
pronounced for CRE-14 (
Jun
/
Jun
1.75-fold;
Jun
/
Fos
1.6-fold) and TRE-13 (
Jun
/
Jun
1.75-fold;
Jun
/
Fos
1.6-fold). This may principally reflect the difference in length of the two
pairs of double-stranded oligonucleotides.
Assuming that the spectral changes are mostly due to specific protein binding to
the TRE and CRE elements within the oligonucleotides of length
l
= 13, 14, 21, 22 and 29 base pairs (Fig.
1
), one would indeed expect that only a segment of n base pairs would be pertubed
upon protein binding, whereas the remaining (
l -
n) base pairs would stay in a canonical, most likely B-conformation, i.e.:
[Delta][epsilon]
max
= n/
l
[middot] [Delta][epsilon]
n
+ (
l
- n)/l [middot] [Delta][epsilon]
0
1
with [Delta][epsilon]
max
being the observed signal upon complex formation under saturating conditions, [Delta][epsilon]
n
the signal of the perturbed base pairs, and [Delta][epsilon]
0
the signal of the unperturbed base pairs. This equation predicts a linear
relationship between [Delta][epsilon]
max
/[Delta][epsilon]
0
and 1/
l
, i.e:
[Delta][epsilon]
max
/[epsilon]
0
= (n [middot] [Delta][epsilon]
n
/[Delta][epsilon]
0
- n) [middot] (1/
l
) + 1
2
Figure
6
A shows that a plot of [Delta][epsilon]
max
/[Delta][epsilon]
0
versus 1/
l
does indeed exhibit a linear relationship for both the
Jun
247-324
homodimer and the
Jun
247-324
/
Fos
137-208
heterodimer complex, and that the homodimer induces consistently larger signal
changes than the heterodimer.
In this study we report the construction, overexpression and purification of the
soluble bZip modules
Jun
247-324
and c-
Fos
137-208
. Their secondary structure and their ability to bind oligonucleotides
containing TRE or CRE sites were tested by using CD and EMSA, respectively. The
CD measurements of protein dimers in the presence of stoichiometric amounts of
double stranded oligonucleotides revealed different impacts of
Jun
/
Jun
and
Jun
/
Fos
bZip dimers on the DNA spectra.
Jun
homodimers changed the DNA signal of all four specific oligonucleotides to a
somewhat stronger extent than
Jun
/
Fos
heterodimers suggesting that the homodimer might induce a more pronounced
conformational change than the heterodimer. So far only the X-ray structure of a
Jun
/
Fos
bZip-TRE complex has been described (
7
).
The influence of c-
Jun
and c-
Fos
bZip domains on oligonucleotide conformation has not been reported in an
earlier CD study (
5
). It is likely that a longer oligonucleotide was used in that study, which may
have masked the DNA signal changes. In our study, the TRE-29 oligonucleotide spectrum is considerably less sensitive to protein
binding than those of the shorter TRE-13 and CRE-14 oligonucleotides (Fig.
6
A).
Kerppola and Curran (
21
-
23
) reported that
Jun
homodimers and
Jun
/
Fos
heterodimers bend DNA in opposite directions. The difference in the CD spectra
induced by these two species is however unlikely to be linked to differential
bending, since circular dichroism seems to be fairly insensitive to DNA bending
(
24
). Using a DNA cyclization assay and EMSA experiments with different DNA
constructs, Sitlani and Crothers (
25
) concluded recently that neither
Jun
homodimers nor
Jun
/
Fos
heterodimers induce substantial DNA bending.
The observed changes in the DNA CD spectra upon complex formation with the
Jun
,
Fos
and GCN4 bZip domains would however be compatible with the induction of a DNA
structure related to the A-form. Upon addition of ethanol, DNA undergoes a B to A transition. Ivanov
and Krylov (
26
) show a series of spectra with partial A-like character. The spectral changes observed for a transition from B-DNA to a situation where about 30% of the DNA adopts an A-structure would nicely match the changes which we observe upon
complex formation with the
Jun
and
Fos
bZip domains, i.e. a ~2-fold increase of the CD intensity at 280 nm and a blue-shift of about 4nm.
For TRE-13 and CRE-14 this would correspond to ~4 bp in a pure A conformation, or more likely to an
intermediate conformation between A and B over a longer stretch of the binding site. This kind of
`intermediate' structure has been described by Nekludova and Pabo (
11
) and applies also to the GCN4-CRE complex (
9
,
10
) which reveals structural features reminiscent of A-DNA as outlined above. Throughout the entire fragment, the average
displacement of the base-pair centerline from the helical axis is -1.4 Å (
9
) as compared with -4 to -5 Å for pure A DNA and no displacement for B-DNA (
11
). These features would be compatible with a partial B -> A-like transition suggested by the CD spectral changes observed upon
addition of the
Jun
and
Fos
bZip domains (this work) and GCN4 bZip domains (
12
).
Surprisingly, we observe the same structural changes in solution for both the
CRE and the TRE complex, since the TRE structure in both the GCN4-TRE (
8
) and the
Jun
/
Fos
-TRE complex (
7
) has been described as being in the B-form. However, a reexamination of the GCN4-TRE structure by Nekludova and Pabo (
11
) revealed an overall displacement of -0.9 Å suggesting a significant deviation from canonical B-DNA. Nevertheless the CRE crystal structure within the GCN4
complex seems more A-like than the corresponding TRE structure (
9
). It might be that in solution the TRE structure within the complex is somewhat
different than in the crystal and that TRE and CRE adopt essentially the same
conformation in solution at least as far as base stacking is concerned.
We thank Paolo Sassone-Corsi for supplying the plasmids containing the c
-Jun
and c
-Fos
genes, Pierre Lepage for the mass spectral analyses and Annie Hoefft for
synthesizing the oligonucleotides. M.J. was supported by a fellowship of the
Gottlieb Daimler- and Karl Benz-foundation, Germany.
REFERENCES
Return