Skip Navigation

This Article
Right arrow Abstract Freely available
Right arrow Print PDF (146K) Freely available
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
Right arrow Email this article to a friend
Right arrow Similar articles in this journal
Right arrow Similar articles in ISI Web of Science
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Add to My Personal Archive
Right arrow Download to citation manager
Right arrow Search for citing articles in:
ISI Web of Science (2)
Right arrowRequest Permissions
Right arrow Commercial Re-use Guidelines
for Open Access NAR Content
Google Scholar
Right arrow Articles by Conte, M. R.
Right arrow Articles by Bloomberg, G.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Conte, M. R.
Right arrow Articles by Bloomberg, G.
Social Bookmarking
Add to CiteULike   Add to Connotea   Add to Del.icio.us  
What's this?

© 1997 Oxford University Press 3808-3815

Solution structure of the ATF-2 recognition site and its interaction with the ATF-2 peptide

Solution structure of the ATF-2 recognition site and its interaction with the ATF-2 peptide Maria Rosaria Conte*, Andrew N. Lane and Graham Bloomberg1

Division of Molecular Structure, National Institute for Medical Research, The Ridgeway, Mill Hill, London NW7 1AA, UK and 1Department of Biochemistry, School of Medicine, University Walk, University of Bristol, Bristol BS8 1TD, UK

Received July 7, 1997; Revised and Accepted August 12, 1997

PDB no. BNL-9505

ABSTRACT

The effect of leucine zipper proteins binding to the DNA recognition site is controversial. Results from crystallography, gel and solution methods have led to opposite conclusions about the conformation of the DNA in the complex. The role of the DNA binding site in the recognition process and in the gene induction mediated by transcription factors needs to be investigated further. In this article the self-complementary 16 bp oligodeoxynucleotide (CATGTGACGTCACATG)2, which contains the cAMP response element recognised by numerous transcription factors of the leucine zipper family, has been examined free from proteins and in its interaction with the mammalian activating transcription factor 2. The recognition process has been investigated by circular dichroism analysis, which has revealed conformational changes in both DNA and protein upon binding. The solution structure of the 16mer, important in order to define the effects induced by binding of leucine zipper proteins and the intrisic bending properties of DNA, has been determined from NMR data using direct refinement against NOE intensities, analysis of scalar coupling constants and restrained molecular dynamics calculations. Final structures starting from the A and B forms of DNA agreed to a pairwise root mean square deviation (r.m.s.d.) of 1.04 ± 0.3 Å (0.7 ± 0.2 Å to the average) for all atoms. The terminal base pairs were less well determined, and the pairwise deviation of the 12 core bp was 0.83 ± 0.27 Å (0.55 ± 0.19 Å to the average). The final structures are within the B-family with an average helical twist of 36 ± 2o. No significant intrinsic DNA bend is shown in the activating transcription factor regulatory site. However, there are substantial deviations from the canonical B-DNA (r.m.s.d. = 3.6 Å) in the core of the molecule, associated with relatively large base inclinations.

INTRODUCTION

The prokaryotic and eukaryotic gene induction by transcription factors is accomplished by the binding of specific transcription activators to a cis-regulatory element which is communicated to the basal transcription machinery at the start-site of transcription (1 ). In this picture the DNA is distorted via protein-induced bends, enabling the bringing together of the distantly positioned regulatory element and the basal transcription machinery. Testing of this general hypothesis has in some instances demonstrated that the specific function of protein-induced bends in DNA can be subserved by heterologous protein-induced bends or by intrinsic (sequence-directed) elements of DNA curvature (2 ).

Protein-induced bending of DNA has been investigated for numerous transcription factors including members of the leucine zipper protein family. Whilst it has been demonstrated that some transcriptional regulators, such as the Escherichia coli catabolite activator protein (CAP) (3 ,4 ) and high mobility group (HMG) box proteins (5 ,6 ), are able to bend their DNA targets, there exists some uncertainty whether the leucine zipper trancription factors induce different DNA bends and orientation (see below).

As far as leucine zipper/DNA complexes are concerned, X-ray structures of GCN4 bound to two different recognition sites [AP-1 and ATF (activating transcription factor)/CREB] (7 ,8 ) and the Fos-Jun heterodimer bound to the AP-1 site (9 ) indicate that the protein conformation changes on binding whereas the DNA conformation is essentially in an unkinked B conformation. This is in agreement with the data on the Fos-Jun heterodimer complexed by the AP-1 target site obtained by a combination of gel and solution methods (10 ) but appears to contradict some gel electrophoretic phasing analysis results which suggest that the DNA (AP-1 site) is bent in the complex with Fos-Jun heterodimer and Jun homodimer (11 ,12 ). On the other end, the anomalous electrophoretic mobility results observed with GCN4 (13 ) and with Myc/Max (14 ) were interpreted as an effect of the shape of the complex, a consequence of the leucine zipper protein motif rather than DNA bending, in agreement with the X-ray results. It appears that the source of disagreement involves the inconsistencies and disparity of the different approaches used, which include circular permutation assay, phasing analysis and DNA cyclization assays (15 ).

In order to understand the mechanism of leucine zipper/DNA recognition process, we have used different solution methods [gel electrophoresis, circular dichroism (CD), NMR and molecular modelling; see below], to analyse the interaction and the conformations of both the DNA and protein components so that a proper comparison of the effects of forming the specific complex can be determined.

The octanucleotide 5'-TGACGTCA-3', identified as a cAMP response element (CRE) is an enhancer that responds to increases in the intracellular cAMP concentration (16 ,17 ). The highly conserved CRE element has been found in the transcriptional regulatory regions of a large number of eukaryotic genes, such as those of somatostatin, fibronectin, human vasoactive intestinal polypeptide (VIP), human chorionic gonadotropin [alpha]-subunit ([alpha]-hCG) tyrosine hydroxylase, human proenkephalin, P-enolpyruvate carboxykinase and human growth hormone (HGh) (18 ) (for reviews see 19 ,20 ). It has also been shown to confer responsiveness to E1a in several adenovirus genes by interaction with ATF-2 (21 ,22 ).

In this article the hexadecamer d(CATGTGACGTCACATG)2, containing the consensus CRE core sequence and flanking regions for ATF-2 (23 ), has been investigated in the interaction with the basic region-leucine zipper ATF-2 homodimer by CD. Preliminary results (see below) showed that in addition to the expected increase in helical content of the peptide, there is also a change in the near UV CD of the DNA on forming a specific complex, as has been observed for the TPA Response Element (TRE) and CRE sequences interacting with the Jun homodimer and the Fos/Jun heterodimer (24 ). Since the DNA appears to undergo conformational changes upon the binding of ATF-2, it is essential to determine the structure and the conformation of the DNA in solution.

In the present article the solution structure of ATF binding site has been extensively investigated by NMR and molecular modelling. It is noteworthy to say that no high resolution structure of any leucine zipper protein DNA binding sites has been reported in literature. There are two reasons for examining the free DNA in detail. First, to ascertain whether the conformation changes significantly on forming the complex, and/or which distortions occur, since for a complete analysis, the conformational change of both protein and DNA are needed to assess the nature of the DNA/protein contacts and provide a free energy estimation of the binding process. The second involves the question that the free DNA could be intrinsically bent, as shown by DNA cyclization analysis of ATF binding site (10 ). In the latter case the leucine zipper transcription factors might require this structural specificity for DNA recognition and contact.

On the other hand, if the DNA binding site is not bent in solution as well as in the complex with leucine zipper proteins, it means that the leucine zipper transcription factors do not induce different DNA bend and orientation, but the intrinsic DNA properties, together with the contributions of other factors involved, may direct the assembly of the initiation complexes with distinct structural and functional properties.

MATERIALS AND METHODS

Materials

The HPLC-purified 16mer d(CATGTGACGTCACATG)2 was purchased from Oswel Research Products Ltd (Southampton, UK), and used without further purification. The DNA was dissolved in aqueous buffer (5 mM sodium phosphate, 100 mM KCl, 0.01 mM EDTA, pH 7) and annealed by slow cooling from 80oC. The DNA was dialysed against the same buffer for 2 days and lyophilised. The sample was redissolved in 0.6 ml of 2H2O or 0.6 ml 1H2O containing 10% 2H2O. The final concentration of DNA was 2.1 mM in strands.

The basic region-leucine zipper ATF-2 (55 amino acid peptide in the monomeric form) was synthesised by standard solid-phase methods and purified by HPLC on a reverse phase (C18) column, using a water/acetonitrile gradient (0.1% TFA). The peptide was then dialysed against buffer (200 mM KCl, 0.05 mM EDTA, pH 7) and lyophilised. The concentration of the peptide was determined from the UV absorbance at 279.6 nm, using the appropriate molar extinction coefficient for a single tryptophan residue per monomer.

Circular dichroism

The CD spectra were recorded on a JASCO Model J-600 spectropolarimeter at 25oC. All spectra were the average of five scans and were corrected for the baseline and normalised with respect to concentration (that is, for both peptide and DNA values are reported as molar ellipticities per residue). The buffer used was 5 mM sodium phosphate, 100 mM KCl, pH 7, and the cell path lengths were 1 and 10 mm.

The 16mer (2 * 10-6 M duplex) and the ATF-2 peptide (4 * 10-6 M homodimer) were scanned over the wavelength ranges of 220-330 nm (near UV) and 190-260 nm (far UV). The titration of the DNA with the peptide and that of the peptide with DNA were followed reading the spectra in the near and in the far UV, respectively.

Electrophoresis

Electrophoresis experiments were carried out using 20% polyacrylamide mini gels (Atto Corp) in Tris-Borate-EDTA pH 8 at room temperature (23oC). Gels were pre-electrophoresed for 20 min before loading. Duplex DNA (3.2 [mu]M) was dissolved in the run buffer supplemented with 25% glycerol, and mixed with increasing concentrations of peptide and incubated for 30 min before loading on to the gel. Tracking dye (bromophenol blue) was not mixed with the samples, but run separately in the outermost lanes. Gels were run at 120-200 V until the dye approached the end of the gel. Samples were stained with ethidium bromide, and visualised by fluorescence of the intercalated dye.

NMR spectroscopy

1H NMR spectra were recorded at 11.75 and 14.1 T on Varian UnityPlus and Varian Unity spectrometers, respectively. 2,2-Dimethyl-2-silapentane-5-sulphonate (DSS) was used for internal chemical shift referencing.

NOESY spectra in 2H2O were recorded at 25, 40 and 50oC using the method of States et al. (25 ), with mixing times of 30, 50, 100, 250 and 300 ms, and acquisition times of 0.06 and 0.7 s in t1 and t2 respectively. Prior to Fourier transformation, the free induction decays were zero-filled to 8192 or 16 384 points in t2 and 2048 points in t1, and multiplied by an unshifted Gaussian weighting function in both dimensions.

Driven truncated NOE experiments (26 ) in 2H2O were recorded with eight irradiation times from 30 to 500 ms at 25oC and from 50 to 600 ms at 40oC. The effective rotational correlation time of the duplex was determined from the cross-relaxation constant of each C(H6-H5) vector, at 25 and 40oC as previously described (27 -29 ).

Spectra in 1H2O were recorded at 4 and 10oC using a Watergate gradient pulse sequence (30 ) to suppress the solvent signal. NOESY spectra were acquired at 14.1 T with mixing times of 50 and 250 ms at 4oC and 75 and 300 ms at 10oC. Typical acquisition times were 0.05 and 0.4 s in t1 and t2 respectively. The data matrices were apodized using an unshifted Gaussian function and zero-filled to obtain a matrix of 16 384 or 8192 by 2048 complex points.

31P NMR spectra were recorded at 9.4 T on a Bruker AM400 spectrometer, using methylene diphosphonate as an external chemical shift reference. A proton-detected heteronuclear shift correlation spectrum was recorded at 40oC using the method described by Sklenar et al. (31 ). Limits on 1H-31P coupling constants were estimated using simulations and the observed antiphase splittings in F2.

Molecular modelling and structure refinement

Restraints. NMR cross-peak volumes were determined as previously described (32 ) and normalised to the C(H6-H5) cross peak volume. Volumes were independently confirmed using the integration routines within FELIX 2.30 (Biosym, San Diego).

Nucleotide conformations were analysed according to a two-state model i.e. P(S), [chi](S) and P(N), [chi](N) using the program NUCFIT (33 ). The intranucleotide distances were then calculated for the major conformation obtained from the best fit solutions. The dihedral angles [chi] best defined by the data were restrained to ±10o, whereas those determined by using a smaller number of intranucleotide NOEs were restrained to ±20o. Sugar conformations were restrained by H1'-H4' and H2'-H4' distances, determined from the NOE intensities or calculated from the values of P(S) obtained from the coupling constants as previously reported (34 ). In some calculations, the sugar conformation was also restrained by the backbone angle [delta] (C5'-C4'-C3'-O3') which was calculated from P (pseudorotation phase angle) and [Phi]m (maximum amplitude) (35 ). The backbone angle [gamma] was determined from measurements of [Sigma]4' in DQF-COSY spectra and the width at half-height of H4' resonances taken from NOESY cross sections (from H1' and H2''), and comparing the NOESY intensities for the base to H5'/H5'' interactions. For all residues except the terminal ones, [gamma] was restrained to 0-100o as determined from [Sigma]4' (and see below).

Heavy atoms involved in Watson-Crick base pairing were restrained according to the standard distances (2.8-3.25 Å) (36 ), on the basis of the NMR spectra of the exchangeable protons. In addition, distance constraints for AC2H-TN3H, CN4H-GN1H were also supplied based on the observation of intense NOEs for these pairs of protons [2.2-3.2 Å for AC2H-TN3H, 2.5-4.0 Å for GN1H-CN4H (2) and 3.0-4.5 Å for GN1H-CN4H (1)]. Molecular dynamics (MD) calculations. All MD calculations were carried out on a Silicon Graphics Iris IndigoII workstation, using Discover 2.9 (Biosym, San Diego). Structure calculations were carried out using the Amber force field starting from different initial coordinates for the d(CATGTGACGTCACATG)2 duplex, including standard B-DNA, standard A-DNA and coordinates generated after a short (3 ps) free dynamics run starting from B-DNA. Solvent and counterions were not explicitly considered in the calculation, but their effects were simulated by using a distance-dependent dielectric constant with [epsilon] = rij for the molecular mechanics (MM) and MD steps. Additional calculations were performed without electrostatics to evaluate the relative importance of the experimental constraints and the force field.

The protocol employed for structure refinement consisted stepwise of: (i) 1000 steps of conjugate gradient minimization, (ii) MD with heating to 300 K during 30 ps (1 fs timestep), (iii) 200 ps MD (1 fs time step) with coupling to heat bath at 300 K, sampling at 20 fs intervals and (iv) 1000 steps of conjugate gradient minimization. After repetitions of steps (i), (ii), (iii) and (iv) the refinement concluded with an rms gradient of <0.05 kcal/mol/Å. Calculations starting at 1000 K converged to the same minimum. Soft constraints were applied throughout the refinement for all nucleotides, with force constants of (40 kcal/mol/Å2 for distances and 40 kcal/mol/rad2 for torsions). When the initial coordinates of standard A-DNA were employed, the structure refinement protocol used a further 1000 steps of conjugate gradient minimization before step (i), applying constraints but not the full force field (scaling the torsion, non-bond and coulombic interaction to 0.05).

Structures were analysed using InsightII (Biosym, San Diego).

RESULTS

Interaction of the ATF-2 peptide with the target site

Circular dichroism. The CD spectra of nucleic acids provide a global overview of the nature of the base-stacking within the helix. The CD spectra of d(CATGTGACGTCACATG)2 recorded over a wavelength range of 220-330 nm is characteristic of B-DNA (Fig. 1 A) (37 ). Figure 1 A shows two points of the titration of the 16mer with the ATF-2 peptide, corresponding to 1:0.5 and 1:1 DNA (duplex)/peptide (homodimer) concentrations. The effect of the specific binding of the peptide to the central 8 bp (CRE) of the DNA is an overall change of the base-stacking pattern in the CRE (Fig. 1 A), indicated by a small shift of the DNA spectra towards lower wavelength and by an increase of the positive signal. Since the peptide has no significant CD signal in the near UV, as shown in Figure 1 A, changes in the DNA CD profile implies conformational changes of CRE upon ATF-2 binding. This behaviour is qualitatively similar to those of TRE and CRE sequences bound to the Jun homodimer and Fos/Jun heterodimer (24 ).


Figure 1. (A)CD spectrum of ATF binding site and its titration with the ATF-2 peptide, as described in the text, cell path length 1 mm, 25oC. (B) CD spectrum of ATF-2 peptide and its titration with the DNA binding site, cell path length 10 mm, 25oC.

The far UV CD spectrum of ATF-2 peptide in the experimental conditions used (Materials and Methods) indicates that a region of the protein (~30% of the homodimer), probably the dimerization domain, is [alpha]-helical (minimum at ~220 and 208 nm, maximum at 193 nm), whereas the basic region appears as a mobile flexible segment folded into a loose helix in solution. This is in agreement with data acquired for other basic region-leucine zipper peptides in similar conditions (38 ,39 ). The molar ellipticity of ATF-2 peptide increases upon binding of DNA, as shown in Figure 1 B, indicating a substantial increase in the [alpha]-helix content in the protein. According to a computational fit blending ideal [alpha]-helix and random coil spectra of polylysine, the amount of [alpha]-helix in the complex is ~70%. The 16mer has little CD in this region (Fig. 1 B), so that the effect is due to a change in the peptide structure, in agreement with data obtained for other leucine zipper proteins (24 ,38 ,39 ).

Titration analysis carried out at micromolar concentrations are in agreement with the expected 1:1 DNA duplex/ATF-2 homodimer stoichiometry. This indicates that the Kd (constant of dissociation) of the complex is <<1 [mu]M. The CD signal changes were completely reversible.

Because of the observed changes in conformation of both DNA and peptide on forming the specific complex, it is necessary to determine the structure of both component in solution as well as in the complex.Electrophoresis. DNA band-shifts were performed as described in Materials and Methods. On increasing the concentration of ATF-2, the intensity of the free DNA band decreased, and concomitantly a new band appeared with a relative mobility of 1:10 (data not shown). The free DNA band disappeared at a peptide (monomer):DNA (duplex) ratio of ~2:1, indicating that the retarded band corresponds to 1 ATF-2 dimer to 1 DNA duplex. Even at concentration of peptide greatly in excess of the DNA, no additional bands were observed. Furthermore, the retarded band was sharp, and appeared to form complexes with the DNA essentially stoichiometrically, indicating that under these conditions, the dissociation constant is <<1 [mu]M, which is in agreement with the CD data. Experiments performed with DNA mixed sequences (in which local sequence variants of the 16mer were used) showed no evidence of binding to the ATF-2 peptide (data not shown).

NMR spectroscopy

Assignment of non-exchangeable and exchangeable protons. All the non-exchangeable protons of d(CATGTGACGTCACATG)2 except H5'/H5'' were assigned at 25, 40 and 50oC following the well-determined connectivities (40 ,41 ) in NOESY (nuclear Overhauser enhancement spectroscopy), TOCSY (total correlation spectroscopy) and DQF-COSY (double quantum filter correlation spectroscopy) spectra. Figure 2 A shows a section of a typical NOESY spectrum showing the base-H1', -H5 and -H3' regions and the sequential connectivities between base protons and H1' and H3'. H2' and H2'' were distinguished both by the relative intensities of the H1'-H2' and H1'-H2'' cross peaks in short mixing time NOESY spectra (H2'' giving the more intense peak), and from the analysis of the fine structure of these peaks in DQF-COSY, as described in detail elsewhere (34 ).


Figure 2. (A) NOESY spectrum of d(CATGTGACGTCACATG)2 in 2H2O. The spectrum was recorded at 14.1 T and 25oC with a mixing time of 300 ms. Acquisition times were 0.06 and 0.7 s in t1 and t2 respectively. The data matrix was apodized using a shifted Gaussian weighting function and zero-filled to 8192 by 2048 complex points. The H8/H6 interproton connectivities to H1' and H3' are shown. (B) Structures of the ATF-2 hexadecamer. Ten superimposed structures are shown in stereo.

Adenine C2 protons were assigned from the very weak intraresidue cross-peak to H1' and weak sequential and cross-strand NOEs to the H1' of the 3' neighbouring nucleotides in the 300 ms NOESY spectra (where resolved). The assignment of the AC2H resonances were confirmed from the strong NOE cross peaks to the N3H of the base-paired T in 1H2O NOESY spectra (see below).

Imino and amino protons were assigned from NOESY spectra recorded in 1H2O. At 4 and 10oC, all of the expected eight iminoprotons were observed in the spectral region between 13.73 and 12.41 p.p.m., which is consistent with the 2-fold symmetry of the ATF-2 recognition site. The terminal G16N1H, and all the TN3H and GN1H imino protons could be unambiguously assigned based on the characteristic chemical shifts, behaviour and the NOE connectivities (41 ).

The amino protons were assigned (where possible) from the NOE to the imino proton of the paired base (41 ). The assignment was checked by observing the very intense intra amino NOE. The amino protons of all the cytosine residues could be unambiguously assigned. The broad terminal C1-G16 imino proton did not show NOEs to the amino protons of C1, but it was possible to assign the latter protons from the presence of cross-peaks between them and the H5 of the same residue in the NOESY spectra. Assignment of the amino protons of A and G is more difficult because of exchange broadening of the resonances via rotation about the C-N bond, and only the amino protons of A14 could be identified confidently. Amino protons corresponding to those of A2 and A12 could be tentatively assigned, but some ambiguity remained from the lack of NOE between the A12 amino protons and spectral overlap. Neither the amino protons of A7 nor those of any of the G residues could be assigned.Rotational correlation time. The effective rotational correlation time of the duplex was determined from NOE build-up curves for the cytosine H6-H5 protons as described in Materials and Methods; the mean correlation times were 6.1 ± 0.2 ns at 25oC and 3.65 ± 0.2 ns at 40oC. This correlation time refers to a vector perpendicular to the helix axis, and contains contributions from the correlation times for rotation about both the long and short axes. For a 16 bp DNA duplex in the B conformation (see below), the axial ratio is ~2.6. The correlation time for end-over-end tumbling can then be calculated as ~1.6 times that of the measured value of the cytosine vectors (28 ). Hence, the correlation times for motion of the long axis are ~9.8 ns at 25oC and 5.8 ns at 40oC. The effect of this anisotropy on NOEs is taken into account in NUCFIT (33 ).Solution conformation of d(CATGTGACGTCACATG)2. The 1H NMR spectra of d(CATGTGACGTCACATG)2 showed only a single set of resonances indicating that the 16mer is symmetric in solution, on average. The conformations of the sugars have been previously described in terms of a two-state equilibrium between S and N states, using a combination of scalar couplings, NOE intensities and a spectral simulation method. All of the sugars were at least 65% in the S conformation, and on average 88% for non-terminal residues, with values of P(S) in the range 138-166o (34 ) (Table 1 ). The glycosidic torsion angles were determined from time-dependent NOEs assuming a two-state equilibrium in which the NOEs are expressed as a linear combination of two states, namely [chi](S)P(S) and [chi](N)P(N). As the mole fraction of the N state was low, the conformation of this state was kept fixed at [chi](N)P(N) = -160o, 9o. The parameters that describe the S state were then found by non-linear regression against the experimental data (33 ,42 ) using several different starting values for P(S), and fs chosen near to the values calculated from coupling constants and different values of [chi](S). This procedure is analogous to the so-called direct refinement (43 ,44 ). The glycosidic torsion angles [[chi](S)] for all non-terminal residues were found in the range of -110 to -129o (±10o/20o) (Table 1 ) at both 25 and 40oC.

The nucleotide conformation is also partly described by the dihedral angle [gamma] (O5'-C5'-C4'-C3') (45 ). The g+ rotamer could readily be distinguished from the g- or t rotamers because in the g+ state, the J4'5' and J4'5'' couplings are both small (~3 Hz), whereas one coupling is large (~12 Hz) in the other two rotamers. The value of the sum of coupling constants [Sigma]4' (= 3J3'4' + 3J4'5' + 3J4'5'' + 4J4'P) taken from a cross-section through the H3'-H4' cross-peak of DQF-COSY spectra provides information about the size of 3J4'5' and 3J4'5'' couplings, since the value of 3J3'4' is known from the analysis of sugar conformation (34 ), and an upper limit to 4J4'P can be obtained from the P-H heteronuclear correlation experiment. We have found that this coupling is generally small (see below). For d(CATGTGACGTCACATG)2 all the values of [Sigma]4' taken from DQF-COSY at 50oC were found in the range of 6.5-7.5 Hz (data not shown) except for C1 where [Sigma]4' = 12 Hz. Even given the imprecision of the estimates of 4J(PH), this small value is consistent only with the g+ rotamer (0-100o). Obviously this method does not allow a precise determination of the angle [gamma], because the [Sigma]4' taken from H3'-H4' cross-section of DQF-COSY is affected by errors depending on the linewidth. Simulation of DQF-COSY spectra using the program Gamma (46 ) indicated that for the range of linewidths found for the hexadecamer (34 ), [Sigma]4' is underestimated for g+ rotamer, but it is overestimated for the rotamers g- and t, so that a value ~7.0 Hz unambiguously defines the g+ rotamer. [Sigma]4' was also determined indirectly by estimating the width at half-height of H4' resonances, which contains the sum of the coupling constants and the linewidth of H4'. Spectral overlap precluded measurements of the resonance width from 1D spectra; the measurements were made from cross-sections through H1' and H2' in NOESY spectra recorded with a mixing time of 300 ms at 25, 40 and 50oC (not shown). These estimates were in good agreement with the values estimated from DQF-COSY. Further, the intranucleotide NOEs H5'/H5'' were consistent also with the g+ rotamer (47 ). This information allowed us to restrain all [gamma] within the range 0-100o (except for C1 and G16). The values of P(S), [chi](S) and [gamma] are typical of nucleotides in the B conformation, which is supported by the observed internucleotide NOE intensities, in particular the sequential H2'(i) - H8/H6(i + 1) NOEs. The parameters that describe the S state were therefore determined, from which precise constraints for the nucleotide conformations could be derived.

Table 1 . Conformational analysis for d(CATGTGACGTCACATG)2
Base[chi](S) (o)P(S) (±15o) (o)fs (± 0.05)
C1-136 (±20o)1560.70
A2-110 (±10o)1560.85
T3-115 (±10o)1440.85
G4-125 (±20o)1550.85
T5-116 (±20o)1380.90
G6nd1500.93
A7-125 (±10o)1600.90
C8-118 (±10o)1440.82
G9-129 (±20o)1450.87
T10nd1380.94
C11-121 (±10o)1500.85
A12-120 (±10o)1600.93
C13-121 (±10o)1500.85
A14-123 (±20o)1580.85
T15-118 (±20o)1400.89
G16-129 (±10o)1660.78
Glycosidic torsion angles [chi](S) were determined as described in the text, and the sugar conformations were obtained as previously reported (34). nd, not determined.

The 31P NMR spectrum (not shown) had a dispersion of 0.52 p.p.m., which is typical of B-DNA in which all the backbone angles are in the usual rotameric states (48 ). The PH heteronuclear correlation experiment was too crowded to provide complete assignments of the phosphate resonances. However, there was no evidence in the H3'-P cross-peaks for coupling constants >5-6 Hz, consistent with [epsilon] in the usual range for the BI conformation.

Distance restraints for internucleotide and cross-strand distances were derived from NOE build-up curves. Because of the possible effects of conformational averaging at the nucleotide level, these distances were classified relatively loosely thus: (0-2.6, 2.6-3.5, 3.0-5.0 Å). 430 restraints were used in the calculations, including 164 intranucleotide distances, 92 torsion angles (including [chi], [delta] and [gamma]), 172 internucleotide sequential distances, 48 cross-strand distances (involving AC2H and CH5) and 40 H-bond distance restraints. Only conformationally-sensitive distance restraints have been used. In some calculations, we also provided loose constraints for the principal rotamers about the [epsilon] (-187o ± 50o) and [beta] (180o ± 50o) dihedral angles based on the 31P NMR spectra (see above).

Calculations were started from different conformations within the B family, and also from the A conformation. Apart from the terminal base pairs, which are relatively poorly constrained by the data, the final structures all are quite similar, as shown by the statistics in Table 2 . Thus, structures obtained from different starting A and B conformations refined to a pairwise root mean square deviation (r.m.s.d.) of 1.04 ± 0.3 Å for all atoms (0.7 ± 0.2 Å to the average structure). If the terminal 2 bp are ignored, the pairwise r.m.s.d. between the core 12 bp drops to 0.83 ± 0.27 Å (0.55 ± 0.19 Å to the average). This indicates that the central part of the duplex, including all of the consensus recognition site, is relatively well determined. The terminal base pairs are subject to fraying (see above), and also have a lower density of constraints than the core nucleotides. The quality of the structures depends on the nature as well as the number of constraints. Thus, if only the 338 distances are used (no torsions), the pairwise r.m.s.d. values are ~1.9 Å for all atoms. Considerable variation was found about the backbone angle [gamma] for these structures. Constraining [gamma] had a substantial effect on the overall convergence, leading to a typical pairwise r.m.s.d. value of ~1.2 Å. Adding further torsion constraints for [delta] and [chi] had a small effect, mainly because the nucleotide conformations are specified by the tight intraresidue distance constraints derived from detailed analysis of the NOE time-courses and coupling constants. These torsion constraints therefore provide redundant information. Finally, adding loose constraints for the [beta] and [epsilon] torsions had very little effect on the structure determination, suggesting that for B-like duplexes, there is no significant force to change these torsions from the values present in the starting structures (which are in a local energy minimum). There were no violations of torsion angles >1.5o, of which there were 2, and no distance violations >0.1 Å. The 10 best structures had total energies of within 10 kcal/mol, and the energy of the structures was less than -987 kcal/mol.

In calculations in which the electrostatic contribution to the forcefield was omitted, globally similar structures were obtained, but with a greater spread, especially for the terminal 2 bp, and for the positions of the phosphate. The pairwise r.m.s.d. value for the core 12 bp was 1.6 ± 0.1 Å (1.1 ± 0.1 Å to the average). The pairwise r.m.s.d. for structures with and without electrostatics was 3.35 ± 0.2 Å. Thus, given the uncertainties in the proper treatment of the electrostatics of DNA, the positions of the phosphates cannot be considered well determined by NMR data in the absence of relatively tight, explicit experimental restraints, in agreement with the findings of Gonzalez et al. (49 ). The loose constraints on [beta] and [epsilon] as applied here have relatively weak restraining power. Considerably more accurate values of these torsion angles, for example from heteronuclear experiments with labelled DNA (50 ), would be needed to specify the backbone conformation more precisely.

Table 2 Statistics of rMD structure calculations
Constraint setNo. restraints/r.m.s.d./Å
 residueBf/<Bf><Bf>/BAf/<Af><Af/Bf>
1. 338 distances10.61.0 ± 0.83.2 ± 0.51.1 ± 1.01.9 ± 0.6
2. 338 distances + 28 [gamma]11.40.6 ± 0.43.0 ± 0.30.8 ± 0.41.06 ± 0.3
3. 338 distances + 28 [gamma] +32 [delta] +32 [chi]13.40.6 ± 0.33.1 ± 0.20.7 ± 0.41.06 ± 0.2
4. 338 distances + 28 [gamma] + 32 [delta] +32 [chi] + 30 [beta] + 30 [epsilon]15.30.6 ± 0.23.2 ± 0.20.7 ± 0.6 1.04 ± 0.3
Converged structures obtained using different sets of constraints were superimposed and the r.m.s.d. values calculated as described in the text. The distances restraints used included 82 * 2 intraresidue, 86 * 2 sequential, 48 * 2 cross-strand, and 40 H-bond distances. Averages were calculated from 10 structures. Bf denotes final structures starting from standard B-DNA (B) and Af denotes structures starting from standard A-DNA. The r.m.s.d. for standard B-DNA versus minimized B-DNA was 1.15 Å and for minimised B-DNA versus minimised A-DNA was 6.21 Å.

DISCUSSION

The interaction of d(CATGTGACGTCACATG)2 with the leucine zipper-basic region ATF-2 homodimer has been investigated in solution by CD analysis which has shown that the recognition process implies conformational changes for both protein and DNA. As reported for other leucine zipper proteins, the basic region of the peptide folds into an [alpha]-helix upon DNA binding. The near UV CD data on DNA show a conformational change upon protein binding, revealed by an overall change of the base-stacking pattern in the CRE sequence. The identification and the analysis of the DNA conformational changes require the structural analysis of the DNA recognition site free from protein, so that a proper comparison with the DNA conformation in the complex can be made.

The final structures shown in Figure 2 B clearly demonstrate that the ATF binding site is within the B family of conformations, but differs significantly from the canonical B structure (r.m.s.d. = 3.6 Å), and also from B-DNA energy minimised without experimental constraints (r.m.s.d. = 2.9 Å). This indicates that the experimental restraints determine the structure and implies that, in order to have a detailed and accurate picture of the recognition process, as well as to ascertain the effect of the binding of leucine zipper proteins, it would not be correct to assume that the conformation of the DNA recognition site is simply standard B (canonical structure and/or energy minimized without experimental constraints).

The r.m.s.d. values of the final structures, especially when considering just the core of the molecule, indicate a reasonably well determined structure. Similar results have been obtained for smaller fragments of DNA (51 ) and it turns out that well determined structures (except for the backbone) can also be obtained for long duplexes, where a biologically interesting sequence can be studied without the influence of end-effects.

The number of independent constraints was relatively small. However, by appropriate model fitting, at least substantial parts of the molecule can be defined to high precision (i.e. nucleotide conformations) whereas, because there are relatively fewer internucleotide and cross-strand NOEs, and more parameters to determine per dinucleotide, it is hardly surprising that the backbone parameters are less precisely defined with the data that can realistically be obtained by homonuclear methods.

In these structures, we have accounted for conformational averaging in the nucleotides within the context of a simple two-state conformational equilibrium, which is supported by extensive NOE and scalar coupling data. This model is sufficient to account for all of these data, but may be a simplification (52 ). Because of the much sparser experimental data available for the internucleotide interactions, and the greater number of degrees of freedom, we have chosen to use fairly loose internucleotide constraints. This results in a relatively poor determination of the helical parameters based on experimental data alone; the nature of the forcefield parameterisation becomes relatively more important, and therefore subject to greater uncertainty. The influence of conformational averaging in the sugars on the internucleotide NOEs, independent of any averaging in the backbone conformation, has not been assessed, though a possible method for doing this has been described by James' group (49 ,52 ).

The observation that all of the sugars are predominantly `S' is different from other versions of the ATF recognition site from the E2A promoter of adenovirus (54 ,55 ). It indicates that a characteristic backbone surface different from that of a standard B-DNA is not strictly required in the protein-DNA recognition process.

Helical twists, rise, propeller twists and base inclinations were calculated for the different structures using Curves version 5.1 (56 ) (Table 3 ). On average, the helical twists were 34.5 ± 3.0o (33.4 ± 2.3o for the 12 core bp) and the rise 3.4 ± 0.4 Å, with little difference between structures starting from B or A conformations, and are typical of B-DNA. The parameters for the terminal base steps were quite variable, and reflect the lack of constraints at the ends of the duplex.Propeller twists were small and showed no particular pattern along the sequence.

Table 3 . Helical parameters for d(CATGTGACGTCACATG)2 calculated as described in the text using Curves version 5.1 (56)
StepRise (Å)Twist (o)Incl (o)Prop. twist (o)
T3  -6.9 15.5
 3.1131.4  
G4   0.65-11.4
 3.2936.9  
T5   4.7 6.9
 3.6032.4  
G6  10.2 -6.9
 3.7331.8  
A7  13.1 -2.6
 3.8936.1  
C8  17.1 0.2
 3.8730.2  
G9  17.4 -1.1
 3.8935.9  
T10  14.0 -2.4
 3.7332.0  
C11  10.9 -7.3
 3.5932.1  
A12   5.1 8.2
 3.3436.2  
C13   0-12.9
 3.1733.3  
A14  -8.2 15.1
mean3.5733.4 6.5 0.12
sd0.027 2.3 6.5 9
Helical parameters are averages over 10 structures and were calculated for the core 12 bp.

Helical parameters, shown in Table 3 , indicate that the ATF regulatory site is not significantly kinked in solution, is free from protein, andsuggest that leucine zipper do not require this structural specificy for DNA recognition and binding. Once the structure of the free DNA in solution is assessed, it will be possible to determine in detail the DNA confomational changes induced upon the binding of leucine zipper proteins. The results with CD showed that in addition to the expected increase in helical content of the peptide, there is also a change in the near UV CD of the DNA on forming a specific complex (see above). As observed for Jun homodimer and Fos/Jun heterodimer (24 ), the difference in the DNA CD spectra induced upon protein binding is unlikely to be linked to different bending of the DNA binding site, since CD seems to be fairly insensitive to DNA bending. Thermodynamics and structural studies on the ATF-2 and 16mer/ATF-2 complex are currently in progress (M.R.Conte, G.Bloomberg and A.N.Lane, unpublished results), in order to investigate in detail the nature of the recognition process and to define the DNA conformation in the complex. DNA bends are often manifest in the 31P NMR spectrum (57 ,58 ) or of the exchangeable protons (6 ). Preliminary 31P and 1H NMR results on the 16mer/ATF-2 complex (data not shown, M.R.Conte, G.Bloomberg and A.N.Lane, data to be published elsewhere) suggests that the DNA is not greatly distorted in the complex with ATF-2. These data would suggest that the leucine zipper proteins bind to a non-intrinsically-bent DNA binding site in a B-like conformation and induce DNA conformational changes, which appear not to be DNA bending.

In conclusion, preliminary speculations would indicate that the leucine zipper proteins are not responsible for bending the DNA in the assembly of the initiation complex.

The proton NMR assignment is available from M.R.C. or as

Supplementary Material via NAR Online.

ACKNOWLEDGEMENTS

This work was supported by the Medical Research Council of the UK, and a Wellcome Travelling Research Fellowship to M.R.C. We thank Dr S.Martin for assistance with CD studies and Dr J.Gyi for comments on the manuscript. NMR spectra were recorded at the MRC Biomedical NMR Facility.

REFERENCES

1 Nordheim, A. (1994) Nature 370, 177-178. MEDLINE Abstract

2 Goodman, S.D. and Nash, H.A. (1989) Nature (London), 341, 251-254.

3 Zinkel, S.S. and Crothers, D.M. (1990) Biopolymers 29, 29-38. MEDLINE Abstract

4 Schultz, S.C., Shields, G.C. and Steitz, T.A. (1991) Science 253, 1001-1007. MEDLINE Abstract

5 Pil, P.M., Chow, C.S. and Lippard, S.J. (1993) Proc. Natl. Acad. Sci. USA 90, 9465-9469. MEDLINE Abstract

6 Werner, M.M., Huth, J.R., Gronenborn, A.M. and Clore, G.M. (1995) Cell 81, 705-714.

7 Ellenberger, T.E., Brandl, C.J., Struhl, K. and Harrison, S.C. (1992) Cell 71, 1223-1237. MEDLINE Abstract

8 Konig, P. and Richmond, T.J. (1993) J. Mol. Biol. 233, 139-154. MEDLINE Abstract

9 Glover, J.N.M. and Harrison, S.C. (1995) Nature 373, 257-261.

10 Sitlani, A. and Crothers, D.M. (1996) Proc. Natl. Acad. Sci USA 93, 3248-3252. MEDLINE Abstract

11 Kerppola, T.K. and Curran, T. (1991) Curr. Opin. Struct. Biol. 1, 71-79.

12 Kerppola, T.K. (1996) Proc. Natl. Acad. Sci. USA 93, 10117-10122. MEDLINE Abstract

13 Gartenberg, M.R., Ampe, C., Steitz, T.A. and Crothers, D.M. (1990) Proc. Natl. Acad. Sci. USA 87, 6034-6038. MEDLINE Abstract

14 McCormick, R.J., Badalian T. and Fisher, D.E. (1996) Proc. Natl. Acad. Sci. USA 93, 14434-14439. MEDLINE Abstract

15 Hagerman, P.J. (1996) Proc. Natl. Acad. Sci. USA 93, 9993-9996. MEDLINE Abstract

16 Comb, M., Birnberg, N.C., Seasholtz, A., Herbert, E. and Goodman, H.M. (1986) Nature 323, 353-356. MEDLINE Abstract

17 Montminy, M.R., Sevarino, K.A., Wagner, J.A., Mandel, G. and Goodman, R.H. (1986) Proc. Natl. Acad. Sci. USA 83, 6682-6686. MEDLINE Abstract

18 Shepard, A.R., Zhang, W. and Eberhardt, N.L. (1994) J. Biol. Chem. 269, 1804-1814. MEDLINE Abstract

19 Meyer, T.E. and Habener, J.F. (1993) Endocrine Rev. 14, 269-290.

20 Montminy, M.R., Gonzalez, G.A. and Yamamoto, K.K. (1990) Trends Neurosci. 13, 184-188. MEDLINE Abstract

21 Tassios, P.T. and La Thangue, N.B. (1990) New Biol. 2, 1123-1134. MEDLINE Abstract

22 Liu, F. and Green, M.R. (1994) Nature 368, 520-525. MEDLINE Abstract

23 Jones, C. and Lee, K.A.W. (1991) Mol. Cell. Biol. 11, 4297-4305. MEDLINE Abstract

24 John, M., Leppik, R., Busch, S.J., Granger-Schnarr, M. and Schnarr, M. (1996) Nucleic Acids Res. 24, 4487-4494. MEDLINE Abstract

25 States, D.J., Haberkorn, R.A. and Ruben, D.J. (1982) J. Magn. Reson. 48, 286-292.

26 Wagner, G. and Wüthrich, K. (1979) J. Magn. Res. 33, 675-680.

27 Lane, A.N., Lefèvre, J.-F. and Jardetzky, O. (1986) J. Magn. Reson. 66, 201-218.

28 Birchall, A.J. and Lane, A.N. (1990) Eur. Biophys. J. 19, 73-78. MEDLINE Abstract

29 Conte, M.R., Jenkins, T.C. and Lane, A.N. (1995) Eur. J. Biochem. 229, 433-444. MEDLINE Abstract

30 Piotto, M., Saudek, V. and Sklenar, V. (1992) J. Biomol. Str. 2, 661-665.

31 Sklenar, V. Miyashiro, H., Zon, G. and Bax, A. (1986) FEBS Lett. 208, 94-96.

32 Lane, A.N., Jenkins, T.C., Brown, T. and Neidle, S. (1991) Biochemistry 30, 1372-1385. MEDLINE Abstract

33 Lane, A.N. (1990) Biochim. Biophys. Acta. 1049, 189-204. MEDLINE Abstract

34 Conte, M.R., Bauer, C.J. and Lane, A.N. (1996) J. Biomol. NMR 7, 190-206. MEDLINE Abstract

35 Rinkel, L.J., van der Marel, G.A., van Boom, J.H. and Altona, C. (1987) Eur. J. Biochem. 166, 87-101. MEDLINE Abstract

36 Saenger, W. (1984) Principles of Nucleic Acid Structure. Springer Verlag, New York.

37 Ivanov, V.I., Minchenkova, L.E., Minyat, E.E., Frank-Kanenetskii, M.D. and Scholkina, A.K. (1974) J. Mol. Biol. 87, 817-833. MEDLINE Abstract

38 Saudek, V., Pasley, H.S., Gibson, T., Gausepohl, H., Frank, R. and Pastore, A. (1991) Biochemistry 30, 1310-1317. MEDLINE Abstract

39 Stanojevic, D. and Verdine, L. (1995) Nature Struct. Biol. 2, 450-457.

40 Reid, B.R. (1987) Q. Rev. Biophys. 20, 1-34. MEDLINE Abstract

41 Wijmenga, S.S., Mooren, M.M.W. and Hilbers, C.W. (1993) in Roberts,G.C.K. (ed.), NMR of Macromolecules. A Practical Approach. IRL Press, Oxford, Ch. 8, pp. 217-288.

42 Lefèvre, J.-F., Lane, A.N. and Jardetzky, O. (1987) Biochemistry 26, 5076-5090. MEDLINE Abstract

43 Yip, P. and Case, D.A. (1989) J. Magn. Reson. 83, 643-648.

44 Robinson, H. and Wang, A.H.-J. (1992) Biochemistry 31, 3524-3533. MEDLINE Abstract

45 Wijmenga, S.S., Heus, H.A., Werten, B., van der Marel, G.A., van Boom, J.H. and Hilbers, C.W. (1994) J. Magn. Res. 103B, 134-141.

46 Smith, S.A., Levante T.O., Meier, B.H. and Ernst, R.R. (1994) J. Magn. Res. 106A, 75-105.

47 Kim, S.-G., Lin, L.J. and Reid, B.R. (1992) Biochemistry 31, 3564-3574. MEDLINE Abstract

48 Gorenstein, D.G. (1992) Methods Enzymol. 211, 254-286. MEDLINE Abstract

49 Gonzalez, C., Stec, W., Reynolds, M. and James, T.L. (1995) Biochemistry 34, 4969-4982. MEDLINE Abstract

50 Tate, S.-i, Kubo, Y., Ono, A. and Kainosho, M. (1995) J. Am. Chem. Soc. 117, 7227-7278.

51 Weisz, K., Shafer, R.H., Egan, W. and James, T.L. (1994) Biochemistry 33, 354-366. MEDLINE Abstract

52 Gochin, M. and James, T.L. (1990) Biochemistry 29, 11172-11180. MEDLINE Abstract

53 Lane, A.N. (1996) Magn. Res. Chem. 34, S3-S10.

54 Borden, K.L.B. (1993) Biochemistry, 32, 6506-6514.

55 Borden, K.L.B. (1994) Biochem. Biophys. Acta 1219, 505-514.

56 Lavery, L. and Sklenar, H (1988) J. Biomol. Struct. Dyn., 6, 63-91.

57 Beckmann, P., Martin, S.R. and Lane, A.N. (1993) Eur. Biophys. J. 21, 417-424. MEDLINE Abstract

58 Haqq, C.M., King, C.-Y., Ukiyama, E., Falsafi, S., Haqq, T.N., Donahoe, P.K. and Weiss, M.A. (1994) Science 266, 1494-1500. MEDLINE Abstract

* To whom correspondence should be addressed at present address: Department of Biochemistry, Imperial College of Science, Technology and Medicine, Exibition Road, South Kensington, London SW7 2AY, UK. Tel: +44 171 594 5315; Fax: +44 171 225 0960; Email: s.conte@ic.ac.uk
Add to CiteULike CiteULike   Add to Connotea Connotea   Add to Del.icio.us Del.icio.us    What's this?



This Article
Right arrow Abstract Freely available
Right arrow Print PDF (146K) Freely available
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
Right arrow Email this article to a friend
Right arrow Similar articles in this journal
Right arrow Similar articles in ISI Web of Science
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Add to My Personal Archive
Right arrow Download to citation manager
Right arrow Search for citing articles in:
ISI Web of Science (2)
Right arrowRequest Permissions
Right arrow Commercial Re-use Guidelines
for Open Access NAR Content
Google Scholar
Right arrow Articles by Conte, M. R.
Right arrow Articles by Bloomberg, G.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Conte, M. R.
Right arrow Articles by Bloomberg, G.
Social Bookmarking
Add to CiteULike   Add to Connotea   Add to Del.icio.us  
What's this?