Solution Structure of Porcine Delta Sleep-inducing Peptide Immunoreactive Peptide A Homolog of the ShortsightedGene Product*

The 77-residue delta sleep-inducing peptide immunoreactive peptide (DIP) is a close homolog of the Drosophila melanogaster shortsighted gene product. Porcine DIP (pDIP) and a peptide containing a leucine zipper-related partial sequence of pDIP, pDIP(9–46), was synthesized and studied by circular dichroism and nuclear magnetic resonance spectroscopy in combination with molecular dynamics calculations. Ultracentrifugation, size exclusion chromatography, and model calculations indicated that pDIP forms a dimer. This was confirmed by the observation of concentration-dependent thermal folding-unfolding transitions. From CD spectroscopy and thermal folding-unfolding transitions of pDIP(9–46), it was concluded that the dimerization of pDIP is a result of interaction between helical structures localized in the leucine zipper motif. The three-dimensional structure of the protein was determined with a modified simulated annealing protocol using experimental data derived from nuclear magnetic resonance spectra and a modeling approach based on an established strategy for coiled coil structures. The left-handed super helical structure of the leucine zipper type sequence resulting from the modeling approach is in agreement with known leucine zipper structures. In addition to the hydrophobic interactions between the amino acids at the heptade positions a and d, the structure of pDIP is stabilized by the formation of interhelical i to i′ + 5 salt bridges. This result was confirmed by the pH dependence of the thermal-folding transitions. In addition to the amphipatic helix of the leucine zipper, a second helix is formed in the NH2-terminal part of pDIP. This helix exhibits more 310-helix character and is less stable than the leucine zipper helix. For the COOH-terminal region of pDIP no elements of regular secondary structure were observed.

Delta sleep-inducing peptide immunoreactive peptide (DIP) 1 is a 77-residue NH 2 -terminally acetylated peptide that was originally isolated from porcine brain (pDIP) using polyclonal antibodies against the delta sleep-inducing peptide (DSIP) (1,2). Although DIP was detected via DSIP-specific antibodies that may recognize the sequence GGDA in DSIP and GGSA in pDIP, it is otherwise not sequence-related to this supposed sleep-inducing peptide. The pDIP sequence contains a putative leucine-zipper motif, a Pro/Glu rich domain, and three potential phosphorylation sites (2). DNA-binding capability of pDIP, however, is not obvious from its sequence, because it lacks the basic region found in the original basic region/leucine zipper (bZIP) DNA-binding domains. In the bZIP family of transcription factors, the leucine zipper acts as a dimerization domain and the upstream basic region as a DNA-binding domain (3,4). Recently, the human analog of pDIP was characterized by cDNA analysis, showing that human DIP (hDIP) differs from the porcine protein in only four residues. Notably, Arg 55 of pDIP is changed to a cysteine in hDIP. hDIP was shown to be present in a large number of tissues by reverse transcriptionpolymerase chain reaction/Southern hybridization (5).
pDIP shares significant homology with the murine TSC-22 protein and the product of the Drosophila melanogaster shortsighted (shs) gene, and the regions upstream of the leucine zipper are almost identical. TSC-22 was reported to be present in both the cytoplasmic and the nuclear fraction, and it has been discussed to function as a transcriptional regulator that is activated by transcription growth factor-␤1 and other growth factors of osteoblastic cells (6). The Drosophila shs gene product acts in the decapentaplegic pathway leading to photoreceptor differentiation (7). In contrast to TSC-22, shs gene product is a cytoplasmic protein present anterior to the furrow. The evolutionary conservation suggests the region upstream of the leucine zipper to be a distinct functional or structural domain.
The well known leucine zipper motif consists of two ␣-helices with the same sequential directionality forming a coiled coil. The coiled coil represents one of the most efficient packing modes of helices (8) and serves as a model for studies of protein stability and subunit interactions (9 -13). The two-stranded ␣-helical coiled coil is characterized by a heptade repeat denoted as "abcdefg" where positions a and d are usually occupied by large hydrophobic amino acids such as Leu, Ile, and Val (14,15), and positions e and g by oppositely charged amino acids (16,17). Residues at the a and d positions have profound effects on the oligomerization states of coiled coils (18 -20).
Much attention has been paid recently to the study of coiled * This work was supported by grants from the Fonds der Chemischen Industrie (FCI) (to P. R.) and a FCI Kekulé fellowship (to G. S.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
Sedimentation Analysis-Sedimentation experiments were performed at room temperature in a Beckman model E analytical ultracentrifuge equipped with a high-intensity light source, 10" recorder, and AnH-Ti rotor. Measurements under the conditions of the NMR experiments made use of cells with ultra-thin pathlength and schlierenoptics: sedimentation velocity at 68,000 rpm, high-speed sedimentation equilibrium at 30,000 and 24,000 rpm; UV scanning at 240 nm. Initial peptide concentrations were 0.2 and 1.1 mM in 50 mM potassium phosphate buffer, pH 7, and 50 mM potassium phosphate buffer, 100 mM Na 2 SO 4 , pH 7, respectively. The partial specific volume was calculated from the amino acid composition. The sedimentation coefficient was calculated from the slope in an ln(r) versus time plot, the molecular mass distribution was fitted assuming a monomer-dimer equilibrium using the ,,multeq 3b" program kindly provided by Dr. A. Minton. For the determination of the diffusion coefficient, areas and maximal gradients were obtained from 20-fold magnifications of schlieren photographies.
CD Spectroscopy-CD spectra were measured on a Jasco J600 CD spectropolarimeter. A water bath was used to control the cell temperature. Data were collected with a 0.2 nm step resolution, a time constant of 1 s, and a scan speed of 20 nm/min. Wavelength scans were performed at discrete temperatures from 25 to 90°C in a thermostatically controlled quartz cell of 0.5 or 0.05 cm pathlength, depending on peptide concentration. For each temperature point, spectra were obtained from 250 to 190 nm. Wavelength scans were processed by subtracting buffer scans taken at the same temperature and converting the data set to mean residual ellipticity prior to analysis. 10 mM phosphate buffer was used throughout.
Thermal transition curves were recorded from 5 to 90 or 95°C at a fixed wavelength of 222 nm. The samples were heated at intervals of 2.5-5°C. At each temperature the samples were allowed to equilibrate for 15-20 min. The reversibility of the denaturations was verified by slowly cooling the sample from final to initial temperature. Typically, more than 95% of the CD signal was recovered after cooling. The transition curves were normalized to the fraction of the unfolded peptide using the standard equation: f u ϭ ( 222 Ϫ n )/( u Ϫ n ), where n and u represent the ellipticity values for the fully folded and fully unfolded species, respectively, and 222 is the observed ellipticity at 222 nm. The transition temperature, T m , of each heat-induced reversible denaturation was determined by fitting the CD signal change at 222 nm ( 222 ) as a function of temperature to a two-state denaturation process using a least-squares fit program based on equations described elsewhere (29). The conformational stability ⌬G 0 was determined as ⌬G 0 ϭ ⌬G ϩ RT ln P, where P is the total protein concentration.
NMR Spectroscopy-NMR measurements were carried out on a standard Bruker AMX 600 spectrometer with 3.5 mM pDIP in 50 mM potassium phosphate buffer, pH 3.5, 298 K. Data from the following NMR spectra were employed for the sequence specific assignment of spin systems and the evaluation of nuclear Overhauser enhancement spectroscopy (NOESY) distance constraints: double quantum filtered correlated spectroscopy, total correlation spectroscopy with 40, 80, and 120 ms of mixing time, respectively, and a spin lock field of approximately 12 kHz and NOESY with mixing times of 100 and 200 ms, respectively. Solvent suppression was performed by continuous coherent irradiation prior to the first excitation pulse and during the mixing time in the NOESY experiment. The sweep width in 1 and 2 was 7246 Hz for all spectra. Quadrature detection was used in both dimensions with the time proportional phase incrementation technique in 1 (37). 4 K data points were collected in 2 and 512 data points in 1 . Zero filling of the time domain data resulted in a frequency-domain matrix with 1 and 2 K data points in 1 and 2 , respectively. All two-dimensional spectra were multiplied by a squared sineball function phase shifted by /4 for NOESY and for total correlation spectroscopy spectra and by /8 for COSY spectra. Base-line correction to the seventh order was used. Data were evaluated on X-window workstations with the NDee program package (Software Symbiose, Inc., Bayreuth, Germany). Chemical shift values are reported relative to external 2,2-dimethyl-2-silapentane sulfonate.
Structure Calculations-Distance information was obtained from two-dimensional NOESY spectra in H 2 O/ 2 H 2 O (9:1) and in 99.998% 2 H 2 O. NOESY cross peaks were classified into three categories according to their volume intensity as estimated from the number of contours in NOESY spectra: strong 0.18 -0.27 nm; medium, 0.18 -0.4 nm; weak, 1.8 -5.5 nm. Pseudoatom corrections were used to adjust distances that involved nonstereospecifically assigned protons such as methyl groups or aromatic ring protons (38). Three-dimensional structures were calculated with the X-PLOR 3.1 package (39). The standard protocols for ab initio simulated annealing and simulated annealing refinement were applied with some modifications. The initial structure calculations started from an extended template with satisfactory local geometry. For the leucine zipper domain a modeling approach for coiled coil proteins (40) was used. The method draws upon knowledge of the oligomerization state, the helix directionality, and the properties of heptade repeat sequences. Unknown structural parameters are heavily sampled. The coiled coil twist angle, for example, is sampled with an initial range from Ϫ35°to 35°in one degree increments. The initial C ␣ positions were those of a regular ␣-helix, and the initial separation between the helices was set to 10 Å. For each initial structure, side chain and backbone atoms were grown from the C ␣ position by applying a protocol similar to those used for the generation of initial coordinates in NMR structure determination (41). Each structure was relaxed with the following simulated annealing protocol: (i) a 5 ps molecular dynamics slow-cooling stage from 500 to 300 K, (ii) a 20 ps constant temperature molecular dynamics (MD) simulation at 300 K, and (iii) 1000 steps of conjugate gradient minimization. During the slow-cooling stage, C ␣ atoms were held in place with harmonic point restraints that were slowly reduced. The helical hydrogen bond restraints were active during all stages, but no other restraints were applied during the constant temperature MD and energy minimization stages. A time step of 0.5 fs was used for temperatures above 350 K during the slow-cooling stage, otherwise a time step of 1 fs was applied. The coordinates for the residues in the leucine zipper domain thus obtained were used as a reference set in the ab initio simulated annealing and the simulated annealing refinement to restrain the main coordinate set. The restraints were incorporated as point restraints in the form of a harmonic potential.
Unrestrained MD-Unrestrained MD calculations were carried out using the parameters for a representative leucine zipper structure and the TIP3P water model (42) that was supplied with the standard X-PLOR force field (39). The overlay was achieved by placing the protein in the center of a cubic water box (6.33 nm) and deleting all solvent molecules closer than 0.26 nm to any heavy atom of the protein. Close nonbonded solute-solvent interactions were removed in two steps. First, 100 cycles of conjugate gradient energy minimization (43) were carried out, keeping the positions of all protein atoms fixed. Second, in 300 cycles of energy minimization, a harmonic potential was used to restrain the protein to its original conformation. During the first 15 ps of the MD calculations, the system was gradually heated to 300 K while coupled to an external water bath (44). The MD calculations were carried out using the Verlet algorithm (45) with a time step of 1 fs. The SHAKE algorithm (46) was used to constrain covalent bond lengths. A dielectric constant of 1.0 was applied with a scaling factor of 0.4 for 1-4 electrostatic interactions. All nonbonded interactions were cut off at a distance of 0.85 nm. During the whole simulation of 200 ps, minimum image periodic boundary conditions were used. Coordinates, energies, and velocities were saved every 0.5 ps for further analysis. Simulations and analyses were performed on Hewlett Packard HP 735 computers. A 1-ps simulation required about 3 h of CPU time.
A second unrestrained MD calculation was carried out with the leucine zipper of GCN4 using the crystal structure (47) (PDB coordinates 2ZTA) as a starting structure. Protons were added to this structure, and an energy minimization was carried out.

RESULTS AND DISCUSSION
Solid-phase Synthesis-In contrast to many other peptides with comparable lengths prepared by solid-phase peptide syn-thesis, application of well established Fmoc chemistry to the synthesis of pDIP resulted in a crude product containing the desired product as the major component. The analytical reversed-phase HPLC of the crude product shows pDIP as the last eluting compound of the highest absorbance (Fig. 1). Because most other components in the crude product are characterized by significantly shorter retention times, it is most probable that these are by-products were derived from truncations during solid-phase assembly of the peptide. The COOH-terminal Pro/Glu-rich domain, however, caused no difficulties. HPLC from the crude pDIP(9 -46) indicated a similar quality of the crude product. Both peptides were obtained in a purity of about 93% after only one preparative HPLC separation according to mass spectrometry and HPLC analysis. pDIP and the leucine zipper-containing fragment are eluted from the analytical HPLC at about 60% buffer B (Fig. 1), demonstrating that the peptides are retarded by significant interactions with the hydrophobic column.
Sequence Analysis-Residues 19 to 40 of pDIP have been proposed to form a leucine zipper (2). The arrangement of the sequence in an ␣-helical structure may be modeled schematically (Fig. 2). For clarity, the helical wheel representation for the maximal five heptades (Met 9 -Leu 43 ) is shown as a helix with a pitch of 3.5 residues/turn as found in coiled coil structures. The interface between the two helices consists of hydrophobic residues with the exception of Asn 31 and Asn 38 . Asparagine pairs are proposed to form an interhelical H-bond at the GCN4 homodimer interface (47). Recently, Lumb and Kim (48) showed that these Asn residues impart specificity for the formation of a two-stranded parallel coiled coil at the expense of stability. Electrostatic interactions between charged residues at positions e and g are sterically allowed in coiled coil structures (47) (Fig. 2, curved arrows), and formation of salt bridges between the two helices could further enhance dimer stability (49,50).
Protein secondary structure prediction suggests a high amount of helical secondary structure for pDIP, and a long helix starting from the NH 2 terminus and spanning the leucine zipper domain (with the exception of the residues around Asn 30 ) is predicted by all of the methods used (Fig. 2). The algorithm by Woolfson and Alber (51) distinguishes between amino acid sequences of dimeric and trimeric coiled coils. This method predicts a preference for dimer formation of the pDIP leucine zipper domain. The ⌬-score measuring the difference of dimer and trimer formation propensities of the test sequence was found to be 2.4, characteristic for dimer formation. Sequence analysis thus suggests that pDIP forms a two-stranded coiled coil. Dimerization may be achieved through a leucine zipper in which the hydrophobic interaction is complemented by the interaction of oppositely charged residues. Our subsequent spectroscopic studies were aimed at further defining structural characteristics of pDIP in solution.
Oligomerization State-pDIP fragment-(9 -46) eluted from a Superdex 75 gel-filtration column as a symmetric single peak with the molecular weight expected for a dimer (Fig. 3). In contrast, full-length pDIP eluted as a symmetric peak with an apparent molecular mass of 41.2 kDa, closer to a pentameric (43.8 kDa) than to the expected dimeric (17.5 kDa) form, indicating the presence of higher oligomers in solution (Fig. 3).
The state of association of pDIP was further analyzed by sedimentation equilibrium ultracentrifugation, because this technique does not suffer as much from peak shape dependence as does gel filtration. On the contrary, boundary analysis and curve fitting allow the minimal molecular mass and the association constant of assemblies to be quantified. Making use of an ultrathin cell (2 mm pathlength), combined with schlierenoptics, the sedimentation coefficient and diffusion constant were determined under the conditions of the NMR experiments. At c pDIP ϭ 2 mM and 68,000 rpm, no significant trailing of the schlieren peak was detectable; with s 20,w 0 ϭ 1.09 Ϯ 0.02 S and D 20,w ϭ (6.59 Ϯ 0.10) 10 Ϫ7 cm 2 s Ϫ1 , the maximum molecular mass is found to be M s,D ϭ 15,720 Ϯ 1300 Da.
Sedimentation equilibrium experiments made use of the meniscus depletion technique (52) applying A. Minton's ,,multeq 3b" program, the A 280 nm versus r profiles were fitted assuming a monomer-dimer equilibrium. Fig. 4 shows the result; the fit is perfect for both potassium phosphate buffer in the absence and presence of 100 mM Na 2 SO 4 . The residuals do not exceed Ϯ0.01. The association constant is K a ϭ 8.0 10 2 M Ϫ1 . The minimum molecular mass, extrapolating the ln(c) versus r 2 profile to zero concentration (meniscus depletion), yields weight average M r ϭ 8730 Ϯ 1295 in accordance with the calculated monomer.
CD Spectroscopy-To examine helical content and thermal stability of pDIP, CD studies were performed at different temperatures (Fig. 5B). The well defined isodichroic point around 202 nm suggests that the denaturation equilibrium can be described on the basis of a two-state structural transition. Minima at 207 and 221 nm and a maximum at 190 nm dem- onstrate that pDIP is largely ␣-helical as predicted. Spectra for pDIP (9 -46) also indicate predominantly ␣-helical structure (Fig. 5A). The helix content for pDIP and pDIP(9 -46) was determined according to Price (53) to be 57 and 61%, respectively. The lower relative helical content for pDIP(9 -46) (74% expected for the four heptades) could reflect the occurrence of two pairs of Asn at a positions. These polar residues are known to be less stabilizing compared with hydrophobic residues at this position in the heptade repeat (48). A similar result was obtained for CD studies of the Max homodimer, a member of the b-HLH-ZIP family (54).
Thermal Denaturation-The peak of negative ellipticity at 222 nm differs significantly in the spectrum of native and denatured pDIP providing a probe for the thermal unfolding reaction that was studied at acidic and neutral pH (Fig. 6A). The sigmoidal shape of the curves suggests a two-state transition in both pH regions, implying a cooperative transition. The melting temperature, defined as the temperature where 50% of the peptide is in the unfolded state, increases from 326 to 337 K when the pH is increased from 2.5 to 6.
The melting temperature is concentration dependent, since it decreases with decreasing protein concentration (Fig. 6B). This observation is consistent with the idea of unfolding connected to dissociation of the two pDIP subunits. If the folded monomer state is essentially unpopulated, unfolding can be described with the two-state model and the energetics can be readily determined in a manner similar to that for monomer denaturation: with P t being the total protein concentration and f u the fraction of the unfolded protein. The two states are the native dimer, N 2 , and the denaturated monomer, D. If this two-state model for the unfolding reaction of a dimer provides a reasonable description, then identical values of ⌬G 0 (value for a protein concentration of 1 M) should follow from experiments performed at different protein concentration. Fig. 7 shows unfolding free energies calculated for two different protein concentrations of pDIP as a function of temperature. The free energies calculated at different protein concentrations assuming the two-state model for the unfolding reaction of a dimeric protein are the same within experimental error, consistent with this model. The conformational stability ⌬G 0 of 35 kJ/mol is comparable to the free enthalpy determined for the Arc repressor (40 kJ/mol) (29) or the dimerization domain of transcription factor LFB1 (48 kJ/mol) (55). For physiological protein concentrations, the free enthalpy is lower than 10 kJ/mol; a reasonable value if pDIP acts as a regulatory protein stabilized by ligand interaction. The pH dependence of the thermal stability of pDIP could be explained with the formation of interhelical salt bridges between ionizable side chains at positions e and g in the heptade repeat. The interhelical salt bridges of pDIP would thus contribute to the stability of the coiled coil. Recently, Yu et al. (56) have also shown that at low ionic strength (10 mM) the coiled coil was much more stable against temperature at neutral pH than at acidic pH, indicating that the Lys-Glu ion pairs present in the coiled coil interface at neutral pH contribute significantly to the stability of the coiled coil, contrasting the results from Lumb and Kim (57).
The results of a study of the concentration dependence of the thermal unfolding transition for pDIP (9 -46) are in agreement with that found for the full-length peptide (Fig. 6C), evidencing the existence of a dimerization area in the leucine zipper domain. A comparison of the shape, however, shows that the thermal unfolding transition of pDIP (9 -46) is biphasic rather than monophasic, indicating a three state transition. This fact may be explained by the existence of native monomers during the denaturation process or by the formation of higher oligomers in the native state of the peptide. The differences in the unfolding reaction of pDIP and pDIP(9 -46) indicate an influence of the residues outside the leucine zipper domain on for-  mation and stability of the dimer.
NMR-The NH-NH region of the NOESY spectrum often can be used to define elements of secondary structure. From the large ␣-helix content observed in the CD spectra, an extensive array of interresidue connectivities is expected in the NH-NH region of the NOESY spectrum of pDIP. The sharp cross peaks observed in this region were found, however, to correspond only to the residues NH 2 -terminal of the leucine zipper domain. In addition to the absence of narrow interresidue NH-NH cross peaks between amino acids of the leucine zipper domain, the fingerprint region in the double quantum filtered COSY and the total coherence spectroscopy experiments show only 60% of the total number of cross peaks expected for the amino acid sequence of pDIP. Following standard methodology (38) it was possible to assign these resonances to the residues NH 2 -and COOH-terminal of the leucine zipper domain.
In addition to the narrow cross peaks, broad resonances of generally small intensity were found in the NOESY spectrum. These resonances correspond to residues from the leucine zipper domain. The lineshape of these cross peaks could result from an exchange between monomeric and dimeric or oligomeric states of pDIP, the life time of these states being intermediate on the NMR time scale. Additionally, the leucine zipper domain may indeed be a rigid region, and thus protons therein would be subject to strong cross-relaxation and poor J-coupling (58). A similar behavior of the two subdomains in BR-LZ containing the DNA-binding domain and the leucine zipper domain of GCN4 was observed earlier (27).
Despite these difficulties, it was possible to assign a large number of backbone resonances starting from some well separated NH resonances in the downfield region. For this assignment, NH-NH NOEs and both sequential C ␣ H-NH(i ϩ 1) NOEs and medium range C ␣ H-NH(i ϩ 3) NOEs from the better resolved C ␣ H resonances were used (59). The strong sequential overlap in the side chain region does not allow tracing of the complete spin system for each backbone assignment. Using the two different assignment procedures for the residues from the leucine zipper domain and the remaining residues, the unambiguous sequence-specific assignment of 72 out of 77 amino acids was possible.
Comparing the chemical shifts of the Asn side chain protons, the differences between Asn 6 in the NH 2 -terminal region, and Asn 30 and Asn 37 in the leucine zipper (Table I)  The deviation of C ␣ proton chemical shifts from their random coil positions are sensitive indicators of secondary structure. Therefore, a secondary structure estimate according to the chemical shift index strategy (60) was used (Fig. 8A). C ␣ proton resonances shifted to high field, relative to the corresponding random coil C ␣ proton resonances, indicate local ␣-helical structure. High field shifts more than 0.1 ppm are marked by Ϫ1. For interpretation of the chemical shift index only resonances should be taken into account with the same sense chemical shift deviation for a stretch of more than three sequential residues (60). This procedure indicates for pDIP two helical regions, extending from Leu 3 to Arg 13 and from Gln 48 to Ser 54 . Assuming the chemical shift indices of Glu 21 , Glu 25 , Gln 33 , and Leu 40 to be Ϫ1 a third helical region from Val 16 to Lys 41 can be predicted. The stretch of negative chemical shift indices is interrupted around Asn 37 , one of the two polar residues known to be less stabilizing as compared with hydrophobic residues at this position in the heptade repeat (47,48,61).
A more quantitative description can be made (Fig. 8B) by using the absolute values of the chemical shift differences. The high field shifts of the C ␣ -proton resonances from residues in the leucine zipper are more pronounced than those of the other regions. This fact can be explained with the formation of a dimerization-stabilized helix. Particularly remarkable are the strong highfield shifts of the C ␣ -proton resonances from some hydrophobic residues at positions a and d.
No NOEs (i, i ϩ j, j Ͼ 8) were observed, especially no NOEs between residues NH 2 -and COOH-terminal to the leucine zipper domain, indicating that the leucine zipper domain forms a parallel coiled coil. The pattern and the intensities of sequential and medium range NOEs (i, i ϩ j, j Յ 4) (Fig. 9) indicate ␣-helical structure of the peptide between residues Met 1 and Val 12 . In the leucine zipper domain only a very small number of unambiguous NOEs were found.
Calculation and Description of the Structure-The structure was determined by a combinatorial strategy using NMR data supplemented by modeling approaches for coiled coil proteins (40). The assumptions used in the modeling strategy were all confirmed by our previous studies. The sequence of pDIP contains the typical leucine heptade repeats, CD spectra were characteristic of ␣-helical structure, the C ␣ H resonances show high field shifts compared with the random coil values, and various results indicate pDIP to form dimers. Modeling of the leucine zipper domain was performed as described under "Materials and Methods." The correlation of structures with a positive starting twist angle is superior to that of structures with a negative starting twist angle (Fig. 10). No transitions between structures with positive starting twist angle and negative starting twist angle were found during the calculations, indicating that the energy barrier between these two types of structure is too high to be overcome by this modeling approach. Positive twist angles correspond to a left-handed conformation of the super helix, which is in agreement with proposals by Crick (8) and with known leucine zipper structures (47). Further structural analyses show that the hydrophobic and electrostatic interactions are stronger in the case of left-handed conformations (Fig. 11). Most of the leucine-leucine distances in these structures are in the range of 5-6 Å when measured as average distances of the C ␥ . Identical values are found in the x-ray structure of GCN4 (47). Formation of salt bridges as defined by the distance between ionizable side chains being less than 4 Å (62) is indicated between Glu 34 and Lys 29 and between Glu 36 and Lys 41 (Fig. 2). In the modeled coiled coil structures, the leucine zippers corresponding to a left-handed conformation are stabilized by such intermonomer electrostatic interac- tions (Fig. 11B). The existence of salt bridges stabilizing the ␣-helical coiled coil is in agreement with the pH dependence of the thermal unfolding process.
The structure calculation of the full-length peptide was performed as described above under "Materials and Methods." The coordinates of a representative structure of the modeling approach were used as reference coordinates to restrain this part of the peptide harmonically to the known structure of the leucine zipper domain. For the residues NH 2 -and COOHterminal to the leucine zipper domain, only the NOEs determined from the 100 ms NOESY spectrum were used as experimental restraints, and no symmetry was imposed on the structure.
A schematic picture of the pDIP structure is shown in Fig.  12, and the energy and root mean square deviation (RMSD) values for 10 structures that resulted from this molecular dynamics calculation procedure are listed in Table II. The positive term for the NOE effective energy is caused by no more than two NOE violations larger than 0.05 nm per structure.
The leucine zipper region, Val 16 -Ala 44 , shows the lowest RMSD values due to the harmonic fitting. A second region with low flexibility was found in the NH 2 -terminal part of the peptide. Structural analysis with the PROMOTIF program (63) shows this part to form a helical secondary structure, as was already indicated from analysis of the NOE pattern ( Fig. 9) and the chemical shift index (Fig. 8). The helix is extending from Val 4 to Tyr 10 and shows 3 10 -character in the region from Leu 8 to Tyr 10 to some extent. The existence of a second helix, aminoterminal to the leucine zipper, is somewhat reminiscent of the arrangement of helices in bZIP proteins. In pDIP, however, this first helix is not basic and does not contain the consensus sequence found in most bZIP proteins, that is NXX(A)(A)XX(C/ S)R (64). Within the bZIP protein family members with a dimerization domain but no intact basic region responsible for the DNA-binding such as CHOP (65) act by dimerization with proteins containing the basic region and inhibiting their DNAbinding activity. The formation of such heterodimers would be a possible mode of action of pDIP.
The COOH-terminal region of the peptide is very flexible and does not show stable elements of regular secondary structure (Fig. 12). Structured parts of pDIP show higher homology within the sequences of DIP, TSC-22 proteins from different mammals, and two shs variants (Fig. 13): The NH 2 terminus is highly conserved (80% identity). Residues responsible for the hydrophobic and electrostatic interactions between the two ␣-helices in the leucine zipper domain are 88% identical. Lys 29 , Glu 34 , Glu 36 , and Lys 41 (sequence position in pDIP), all involved in salt bridge formation, are conserved in all sequences and the hydrophobic amino acids at positions a and d show high homology. The evolutionary conservation of these residues underscores their functionality. It is known that the residues at positions a and d influence the oligomerization state of the  coiled coil protein and the distribution of charged residues at positions e and g control the topology of the mutual helix arrangement (18,66). Therefore, the leucine zipper domains of the other proteins in the sequence alignment ( Fig. 13) also have an intrinsic disposition to form parallel two-stranded coiled coils.
The unstructured COOH-terminal region of pDIP, in contrast, shows only 20% identity within the aligned sequences. Although the proline-rich sequence of pDIP does not show homology to SH 3 -binding domains (67) it is possible that this region is responsible to protein-protein interaction. Prolinerich polypeptides have highly restricted mobility even before binding. Thus, binding leads to a smaller drop in entropy than it would for a normal more flexible peptide, and hence a larger overall binding energy is achieved (68).
Unrestrained Molecular Dynamics-Unrestrained MD calculations of the leucine zipper of pDIP were performed to compare its structural stability to that of the well known GCN4 leucine zipper structure from the bZIP family. The RMSD values of the backbone and side chain heavy atoms of the structures in the unrestrained molecular dynamics calculation as compared with the starting structure were determined for both structures (Fig. 14). For both proteins, the RMSD values increase early in the calculation as expected due to an equilibration with the solvent. For the remaining time the RMSD values of the backbone atoms remain relatively stable, whereas the RMSD values of the side chain heavy atoms still increase. The RMSD values for the backbone atoms of pDIP and GCN4 leucine zippers are of the same order after the simulation time of 200 ps (0.96 and 0.98 Å, respectively), indicating virtually identical conformational stability of these leucine zippers. This indicates the modeled leucine zipper of pDIP to exhibit the same degree of flexibility as the leucine zippers of the bZIP family.