Portal de Periódicos da CAPES

Tertiary Structure of Uracil-DNA Glycosylase Inhibitor Protein

1995; Elsevier BV; Volume: 270; Issue: 28 Linguagem: Inglês

10.1074/jbc.270.28.16840

1083-351X

Richard D. Beger, Suganthi Balasubramanian, Samuel Bennett, Dale W. Mosbaugh, Philip H. Bolton,

RNA and protein synthesis mechanisms

The Bacillus subtilis bacteriophage PBS2 uracil-DNA glycosylase inhibitor (Ugi) is an acidic protein of 84 amino acids that inactivates uracil-DNA glycosylase from diverse organisms. The secondary structure of Ugi consists of five anti-parallel β-strands and two α-helices (Balasubramanian, S., Beger, R. D., Bennett, S. E., Mosbaugh, D. W., and Bolton, P. H.(1995) J. Biol. Chem. 270, 296-303). The tertiary structure of Ugi has been determined by solution state multidimensional nuclear magnetic resonance. The Ugi structure contains an area of highly negative electrostatic potential produced by the close proximity of a number of acidic residues. The unfavorable interactions between these acidic residues are apparently accommodated by the stability of the β-strands. This negatively charged region is likely to play an important role in the binding of Ugi to uracil-DNA glycosylase. The Bacillus subtilis bacteriophage PBS2 uracil-DNA glycosylase inhibitor (Ugi) is an acidic protein of 84 amino acids that inactivates uracil-DNA glycosylase from diverse organisms. The secondary structure of Ugi consists of five anti-parallel β-strands and two α-helices (Balasubramanian, S., Beger, R. D., Bennett, S. E., Mosbaugh, D. W., and Bolton, P. H.(1995) J. Biol. Chem. 270, 296-303). The tertiary structure of Ugi has been determined by solution state multidimensional nuclear magnetic resonance. The Ugi structure contains an area of highly negative electrostatic potential produced by the close proximity of a number of acidic residues. The unfavorable interactions between these acidic residues are apparently accommodated by the stability of the β-strands. This negatively charged region is likely to play an important role in the binding of Ugi to uracil-DNA glycosylase. The Bacillus subtilis bacteriophage PBS2 uracil-DNA glycosylase inhibitor protein (Ugi)1 1The abbreviations used are: Ugiuracil-DNA glycosylase inhibitorUnguracil-DNA glycosylase. has been shown to inactivate Escherichia coli uracil-DNA glycosylase (Ung) by forming a stable Ung·Ugi protein complex in 1:1 stoichiometry(1Bennett S.E. Mosbaugh D.W. J. Biol. Chem. 1992; 267: 22512-22521Abstract Full Text PDF PubMed Google Scholar). Interaction with Ugi prevents the enzyme from binding to DNA and dissociates a preformed Ung·DNA complex(1Bennett S.E. Mosbaugh D.W. J. Biol. Chem. 1992; 267: 22512-22521Abstract Full Text PDF PubMed Google Scholar, 2Bennett S.E. Schimerlik M.I. Mosbaugh D.W. J. Biol. Chem. 1993; 268: 26879-26885Abstract Full Text PDF PubMed Google Scholar). Ugi bound in a preformed complex does not exchange with free Ugi(2Bennett S.E. Schimerlik M.I. Mosbaugh D.W. J. Biol. Chem. 1993; 268: 26879-26885Abstract Full Text PDF PubMed Google Scholar). Ugi also inactivates uracil-DNA glycosylase isolated from other sources, including Microccocus luteus, Mycoplasma lactucae, Saccharomyces cerevisiae, rat liver (nuclear and mitochondrial), herpes simplex virus types 1 and 2, human placenta, and KB cells(3Williams M.V. Pollack J.D. J. Bacteriol. 1990; 172: 2979-2985Crossref PubMed Google Scholar, 4Wang Z. Mosbaugh D.W. J. Biol. Chem. 1989; 264: 1163-1171Abstract Full Text PDF PubMed Google Scholar, 5Cone R. Bonura T. Freidberg E.C. J. Biol. Chem. 1980; 255: 10354-10358Abstract Full Text PDF PubMed Google Scholar, 6Karran P. Cone R. Freidberg E.C. Biochemistry. 1981; 20: 6092-6096Crossref PubMed Scopus (37) Google Scholar, 7Winters T.A. Williams M.V. Virology. 1993; 195: 315-326Crossref PubMed Scopus (14) Google Scholar). In vivo expression of the ugi gene in E. coli leads to inactivation of Ung and to bacteria with phenotypic properties consistent with other ung mutants(8Wang Z. Mosbaugh D.W. J. Bacteriol. 1988; 170: 1082-1091Crossref PubMed Google Scholar). In vitro studies have indicated that Ugi is specific for uracil-DNA glycosylase, as it had no inhibitory effect on other DNA-metabolizing enzymes that were tested (4, 6). uracil-DNA glycosylase inhibitor uracil-DNA glycosylase. The ugi gene has been cloned, sequenced, and overexpressed in E. coli. It encodes a small (Mr = 9,477) heat-stable, non-globular polypeptide of 84 amino acid residues (1, 6). Ugi is an acidic protein (12 Glu, 6 Asp) with a pI of 4.2 that migrates anomalously during SDS-polyacrylamide electrophoresis(1Bennett S.E. Mosbaugh D.W. J. Biol. Chem. 1992; 267: 22512-22521Abstract Full Text PDF PubMed Google Scholar, 4Wang Z. Mosbaugh D.W. J. Biol. Chem. 1989; 264: 1163-1171Abstract Full Text PDF PubMed Google Scholar, 5Cone R. Bonura T. Freidberg E.C. J. Biol. Chem. 1980; 255: 10354-10358Abstract Full Text PDF PubMed Google Scholar). Whereas the native molecular weight as determined by gel filtration chromatography was ∼18,000, sedimentation equilibrium centrifugation established that Ugi existed as a monomer in solution(1Bennett S.E. Mosbaugh D.W. J. Biol. Chem. 1992; 267: 22512-22521Abstract Full Text PDF PubMed Google Scholar, 8Wang Z. Mosbaugh D.W. J. Bacteriol. 1988; 170: 1082-1091Crossref PubMed Google Scholar). Ugi inactivates E. coli uracil-DNA glycosylase noncompetitively; the Ki (0.14 μM) is more than 10-fold lower than the Km (1.7 μM) of the enzyme (9). Stopped-flow kinetic analysis suggests that the Ung/Ugi association is described by a two-step mechanism(2Bennett S.E. Schimerlik M.I. Mosbaugh D.W. J. Biol. Chem. 1993; 268: 26879-26885Abstract Full Text PDF PubMed Google Scholar). The first step involves a rapid pre-equilibrium or "docking," characterized by the dissociation constant K (1.3 μM), in which Ugi and Ung associate to form a pre-complex. The second step entails "locking," characterized by the rate constant k (195 s−1), in which optimal alignment is obtained. Two experiments indicate that the individual components of the Ung·Ugi complex are not apparently altered by the interaction. First, the dissociation of the complex by SDS-polyacrylamide gel electrophoresis revealed no detectable modification in the molecular weight of either protein subsequent to binding(1Bennett S.E. Mosbaugh D.W. J. Biol. Chem. 1992; 267: 22512-22521Abstract Full Text PDF PubMed Google Scholar). Second, when the Ung·Ugi complex was treated with urea and the individual proteins resolved by urea-polyacrylamide gel electrophoresis, both uracil-DNA glycosylase and inhibitor activities were recovered(1Bennett S.E. Mosbaugh D.W. J. Biol. Chem. 1992; 267: 22512-22521Abstract Full Text PDF PubMed Google Scholar). Several lines of evidence suggest that Ugi binds Ung at or near the DNA-binding site of the enzyme. Ung bound to Ugi no longer associates with DNA; Ung UV-cross-linked to single-stranded oligonucleotide fails to form complex with Ugi, and Ung complexed with Ugi does not UV-cross-link to dT20(10Bennett S.E. Jensen O.N. Barofsky D.F. Mosbaugh D.W. J. Biol. Chem. 1994; 269: 21870-21879Abstract Full Text PDF PubMed Google Scholar). Interestingly, the Ung peptides that photocross-link with dT20 are highly conserved in amino acid sequence when compared with nine other uracil-DNA glycosylases(11Upton C. Stuart D.T. McFadden G. Proc. Natl. Acad. Sci. U. S. A. 1993; 90: 4518-4522Crossref PubMed Scopus (75) Google Scholar). The fact that the Ugi protein from bacteriophage PBS2 is capable of inactivating uracil-DNA glycosylase from diverse biological systems, which are under no selective pressure to maintain an Ugi binding site, lends support to the notion that Ugi binds to the same site as DNA. The secondary structure of Ugi was recently determined by multidimensional NMR methods and found to consist of two helices and five anti-parallel β-strands(12Balasubramanian S. Beger R.D. Bennett S.E. Mosbaugh D.W. Bolton P.H. J. Biol. Chem. 1995; 270: 296-303Abstract Full Text Full Text PDF PubMed Scopus (13) Google Scholar). Six loop or turn regions were identified that connect the β-strands sequentially to one another to form a contiguous anti-parallel sheet. All five of the β-strands have terminal acidic amino acid residues on the same side of the anti-parallel sheet(12Balasubramanian S. Beger R.D. Bennett S.E. Mosbaugh D.W. Bolton P.H. J. Biol. Chem. 1995; 270: 296-303Abstract Full Text Full Text PDF PubMed Scopus (13) Google Scholar). In this paper, we describe the tertiary structure of Ugi as determined by solution-state multidimensional NMR and calculate the electrostatic potential of the protein. The tertiary structure brings together a large number of negatively charged residues to form an "electrostatic knob" that extends over about 18 Å on one face of the protein. The electrostatic potential of this region is similar to that generated by the phosphates in DNA. This large area of negative charge is likely to be the portion of Ugi that interacts with the DNA binding site of Ung. Preliminary results indicate that Ugi undergoes some conformational change upon binding to Ung. [15N]Ugi was purified as previously described(12Balasubramanian S. Beger R.D. Bennett S.E. Mosbaugh D.W. Bolton P.H. J. Biol. Chem. 1995; 270: 296-303Abstract Full Text Full Text PDF PubMed Scopus (13) Google Scholar). Briefly, E. coli JM105 transformed with pZWtac1 was grown to ∼7.5 × 108 cells/ml at 37°C in 10 liters of M9 medium, in which [15N]ammonium chloride (0.1%) provided the sole nitrogen source. Expression of the ugi gene was induced by the addition of isopropyl-1-thio-β-D-galactopyranoside to 1 mM, and after 3 h of additional growth, cells were harvested by centrifugation and stored at −80°C. Purification was carried out essentially as described by Wang et al.(9Wang Z. Smith D.G. Mosbaugh D.W. Gene (Amst.). 1991; 99: 31-37Crossref PubMed Scopus (30) Google Scholar). After purification to apparent homogeneity, Ugi was diafiltered against NMR buffer (25 mM deuterated Tris, 0.2 mM EDTA, 0.2 EGTA, and 100 mM NaCl, pH 7.0) and concentrated to ∼2.2 mM. E. coli uracil-DNA glycosylase was purified to apparent homogeneity as previously described (1Bennett S.E. Mosbaugh D.W. J. Biol. Chem. 1992; 267: 22512-22521Abstract Full Text PDF PubMed Google Scholar) and concentrated under 55 p.s.i. of prepurified N2 in an Amicon stirred cell equipped with a 62-mm diameter YM10 DIAFLO membrane (Amicon) to 24.5 μM. A portion (625 μl) of [15N]Ugi was combined with 54.3 ml of Ung and incubated at 4°C for 1 h. The protein mixture was then concentrated under prepurified N2 in an Amicon 10-ml stirred cell (25-mm diameter, YM10 membrane) and diafiltered against 125 ml of NMR buffer to a final volume of 1.8 ml. The concentration of the complex was determined by absorbance spectroscopy using the extinction coefficient ε280 = 5.4 × 104, which represents the sum of the extinction coefficients of the protein components (4.2 × 104 for Ung and 1.2 × 104 for Ugi) and found to be 739 μM. Samples (9 μg) of the complex reaction taken before and after concentration and diafiltration were analyzed by nondenaturing 12.5% polyacrylamide gel electrophoresis as previously described(1Bennett S.E. Mosbaugh D.W. J. Biol. Chem. 1992; 267: 22512-22521Abstract Full Text PDF PubMed Google Scholar); greater than 95% of the protein was in complex as judged by Coomassie Blue staining. All of the NMR spectra were obtained using a Varian Unityplus 400 spectrometer equipped with a Nalorac ID400 probe. Three-dimensional NOESY-HMQC and TOCSY-HSQC were obtained using uniformly 15N-labeled Ugi (13-16). Sensitivity enhanced TOCSY-HSQC data obtained using a 60-ms mixing time were applied to group the spin systems of the 15N amide-labeled residues. TOCSY-HSQC data were collected at 17, 30, and 37°C with eight transients per increment and obtained by the States-Haberkorn method along each indirectly detected dimension. Each of the complex data sets was analyzed using 128 increments of t1 and 24 increments of t2. The data were linearly predicted to 256 points in t1 and 48 points in t2 prior to Fourier transformation into 512×128×1024 points using shifted Gaussians along each dimension. NOESY-HMQC data were obtained using a mixing time of 200 ms at 17, 30, and 37°C with 16 transients per increment and obtained by the Stakes-Haberkorn method along each indirectly detected dimension. There were 128 increments in t1 and 20 increments in t2. The data were linearly predicted to 256 points in t1 and 40 points in t2 before Fourier transformation into 512×128×1024 points using shifted Gaussians along each dimension. Representative planes from the NOESY-HMQC data have been previously presented(6Karran P. Cone R. Freidberg E.C. Biochemistry. 1981; 20: 6092-6096Crossref PubMed Scopus (37) Google Scholar). A series of eight J-modulated HSQC experiments were obtained at 25°C to determine the 3JHN-αH dihedral angle-coupling constants(13Kessler H. Seip S. Croasman W.R. Carlson R.M.K. Two-Dimensional NMR Spectroscopy. VCH, New York1994Google Scholar, 14LeMaster D.M. Prog. NMR Spectroscopy. 1994; 26: 371-419Abstract Full Text PDF Scopus (97) Google Scholar, 15Hull W.E. Croasman W.R. Carlson R.M.K. Two-Dimensional NMR Spectroscopy. VCH, New York1994: 67-456Google Scholar, 16Dyson H.J. Wright P.E. Croasman W.R. Carlson R.M.K. Two-Dimensional NMR Spectroscopy. VCH, New York1994: 655-698Google Scholar). Mixing times of 10, 30, 50, 60, 70, 80, 100, and 140 ms were used. Data were collected with 256 points along t1 and 2048 points during the acquisition time. The data were zero filled to 512 points along t1 and Gaussian shifted before Fourier transformation into 512×2048 points. The modulation of the volume, V, of each HSQC signal was linearly approximated to V(τ2) = constant*cos(π3JHN-αHτ2) with respect to the mixing time, τ2, to produce 67 NH-CαH dihedral angle-coupling constants. The linearly predicted time of the change in sign of a peaks amplitude was within 2% error of the exact non-linear expression for the J-modulated evolution of a HSQC signals. A series of seven XT1 HSQC (17Balasubramanian S. Nirmala R. Beveridge D.L. Bolton P.H. J. Magn. Reson. 1994; 104: 240-249Crossref Scopus (16) Google Scholar) experiments was obtained at 25°C to determine the 15N T1 times. Delays of 0, 35, 80, 150, 300, 600, and 1200 ms were used with 256 points along t1 and 2048 points during the acquisition time. The data were zero filled and Gaussian shifted along both dimensions prior to Fourier transformation into 1024×4096 points. The volume of each HSQC peak was fitted with respect to mixing time and least square fitted to V(t) = V(t = 0)*exp(-t/T1), with t the delay time, to produce a best fit T1 time for each 15N-labeled atom that could be accurately found in at least four of the seven XT1 HSQC experiments. A total of 968 interproton restraints were acquired from 200-ms three-dimensional HMQC NOESY data obtained at 17, 30, and 37°C experiments. Of these NOESY constraints, 270 were intraresidue and 636 were interresidue. Of the 636 interresidue NOESY constraints, 329 were sequential. The 908 NOESY restraints were divided into three categories corresponding to strong, medium, and weak. The respective ranges of the categories were 1.8-5.0, 2.1-5.0, and 2.4-5.0 Å. A total of 67 Φ dihedral constraints was obtained from a series of six J-modulated HMQC experiments. The 67 dihedral angle constraints were divided into three categories. When 3JNH-αH had a coupling constant greater than 8 Hz, Φ was constrained to −140 ± 40°. When the coupling constant was below 6 Hz and the residue was part of a helical secondary structure element, Φ was constrained to −80 ± 40°. Coupling constants below 6 Hz of residues not in a helical secondary structure element were not given any dihedral constraints. The rest of the coupling constants were constrained to −120 ± 40°. An initial linear Ugi structure was built using Biosym Insight II by piecing together the primary structure sequence with the residue append command. This structure was used as a template structure for the X-PLOR simulated annealing(16Dyson H.J. Wright P.E. Croasman W.R. Carlson R.M.K. Two-Dimensional NMR Spectroscopy. VCH, New York1994: 655-698Google Scholar, 18Brunger A.T. Clore G.M. Gronenborn A.M. Saffrich R. Nigles M. Science. 1993; 261: 328-331Crossref PubMed Scopus (147) Google Scholar). The energy function used during all of the simulated annealing and energy minimization routines was confined to covalent geometry terms (bonds, angles, torsion, and chirality), distance and dihedral angle constraints, and a non-bonded, van der Waals term. A soft square potential was used during the initial simulated annealing protocol for all NOE distance constraints within a specified "switching" region, and outside the switching region a soft asymptote was used: E=min(ceil(S·Kb·T/2cij2)*{a+b/Δsoftexp+cΔ}(Eq. 1) for R > d + dplus+ rsw or E=min(ceil(S·Kb· T/2cij2)*{Δexp}(Eq. 2) for R < d - dminus - rsw, where ceil is 1000 kcal/mol, S is a scale factor of 100 kcal/mol, T is the temperature 300 K, and Kb is the Boltzmann constant. Δ is the distance from the edge of the square well NOE constraints, and a and b are determined by X-PLOR such that the function is a smooth function at the point R = d + dplus+ rsw. The slope from the asymptote is given by c, which was set to 0.1 during the initial high temperature molecular dynamics; otherwise, it was set to 1.0. The switching distance was set to 0.5 Å. The exponent softexp was set to 1, and exp was set to 2. A square biharmonic potential of 100 was used for all dihedral constraints. The simulated annealing protocol started with 500 steps of energy minimization, followed by 6000 steps of a 4-fs timestep (24 ps total) molecular dynamics at 1000 K with a scaled energy function such that angular energies were multiplied by 0.4, torsions were multiplied by 0.1, and van der Waals terms were multiplied by 0.002. Another 3000 steps of 4-fs timestep (12 ps total) molecular dynamics at 1000 K were carried out with an intact energy function except for the van der Waals, which was scaled to 0.002 its normal value. The system was cooled by dropping the temperature at 50-K intervals to a final temperature of 200 K. At each temperature, 1000 steps of a 4-fs timestep (4 ps total) molecular dynamics were done while slowly scaling the van der Waals radius up to a final value of 0.75 at 200 K. This was followed by 2000 steps of square well conjugate energy minimization with the van der Waals energy function multiplied by 4.0. The 40 calculated Ugi structures with the lowest total energy and smallest number of violations were saved for 40 additional, separate iterations of refined simulated annealing. Refined simulated annealing consisted of molecular dynamics at 1000 K, and the temperature was lowered to 100 K in 50-K increments. At each temperature, 667 steps of a 2-fs timestep (1.34 ps total) molecular dynamics were done. A hard square potential of 400 kcal/mol was used for the NOE constraints: E-min(ceil(S·Kb·T/2cij2)*{a+b/Δsoft+cΔ}(Eq. 1) for R > d + dplus+rsw or E=min(ceil(S·Kb·T/2cij2)*{Δexp}(Eq. 2) for R > d + dminus+rsw where ceil is 1000 kcal/mol, S is 400 kcal/mol, T is 300 K, Kb is the Boltzmann constant, and Δ is the distance from the edge of the square well NOE constraints. The NOE distance constraints were 1.8-4.0, 2.1-4.5, and 2.4-5.0 Å for strong, medium, and weak NOE constraints, respectively. An additional 19 hydrogen bonds were put in as NOE distance constraints; only those hydrogen bonds with long amide exchange times that fit within the anti-parallel β-sheet secondary and tertiary structure scheme were used. Hydrogen bonds were constrained with a N-O distance of 2.5-3.3 Å and H-O distance of 1.8-2.5 Å. After the simulated annealing to 100 K, 500 steps of energy minimization with a van der Waals energy function multiplied by 4.0 were executed. This was followed by 100 steps of energy minimization with a hard square potential of 200 kcal/mol for both NOE and dihedral angle constraints. The 12 lowest energy structures with no NOE violations greater than 0.4 Å and no dihedral angle violations greater then 5° were saved. These structures have a 1.0 Å root mean square deviation for the heavy backbone atoms and 1.3 Å root mean square deviation for all heavy atoms. Fig. 4 contains the overlay of these 12 structures. The electrostatic surfaces were generated using the program GRASP(19Nicholls A. Sharp K.A. Honig B. Proteins. 1991; 11: 281-296Crossref PubMed Scopus (5315) Google Scholar). The electrostatic potentials of Ugi were calculated with a dielectric constant of 4.0 for the protein and 80.0 for the solvent. The ionic strength was set to zero. The charges of the side chains of the Lys, Asp, and Glu residues were used. The electrostatic potential of the duplex DNA was generated in the same manner with charges on the phosphate backbone. The DNA duplex used is the symmetric dimer of d(CGCGAATTCGCG) with an idealized B-70 structure. The secondary structure of Ugi was recently determined by multidimensional NMR methods and is shown in Fig. 1. The secondary structure has two helices and five β-strands that are arranged into a contiguous antiparallel sheet. Not present in secondary structural features are seven acidic residues: Glu-38, Asp-40, Glu-49, Asp-52, Asp-61, Glu-64, and Glu-78. In addition, all five of the β-strands have terminal acidic residues on the same side of the antiparallel sheet as shown in Fig. 1. The tertiary structure of Ugi was determined using the combination of the NMR results and simulated annealing. The relevant NMR information is summarized in Figure 2:, Figure 3:. The assignments of the resonances have been previously presented(12Balasubramanian S. Beger R.D. Bennett S.E. Mosbaugh D.W. Bolton P.H. J. Biol. Chem. 1995; 270: 296-303Abstract Full Text Full Text PDF PubMed Scopus (13) Google Scholar). The 15N T1 values have been determined, and the relaxation data are consistent with a well defined structure.Figure 3:The number and types of NOEs for each residue are indicated as are the 15N T1 values of the amide nitrogens. Also shown are the root mean square deviation for all heavy atoms of each residue as well as for the heavy backbone atoms of each residue.View Large Image Figure ViewerDownload Hi-res image Download (PPT) The overlay of the structure is shown in Fig. 4, and the average tertiary structure is shown in two perspectives in Fig. 5. In this structure, the two helices present in the protein are situated on the same face of the β-sheet. Neither of the helices appears to be in close proximity to any other structural element. The loop formed by residues 61-68 folds over the β-sheets. The β-strand and helix regions of the Ugi structure are all quite regular. The tertiary structure of Ugi was used to calculate the electrostatic potential that is shown in Fig. 5. The electrostatic potential of Ugi is quite striking. The tertiary structure brings together Glu-20, Asp-48, Glu-49, Asp-52, Glu-53, Asp-74, and Glu-78 to form a surface of highly negative potential, indicated by the red surface, on one face of the structure. The electrostatic potential of this region is greater than 6.6 kcal and is thus similar to that generated by the phosphate backbone of DNA. This large area of negative potential is likely to be a region of Ugi that interacts with the residues of Ung that naturally interact with the phosphates of DNA. The other side of Ugi is mostly neutral as shown in Fig. 5. The electrostatic potential of Ugi is compared with that of B-form DNA in Fig. 6. It is seen that both the areas and magnitudes of the electostatic potentials in the two instances are quite similar. Other regions of Ugi with negative potentials are found adjacent to the two helices. As the potentials in these regions are greater than 6.6 kcal, they could contribute to the binding to Ung; however, the areas of these potentials are not nearly as large as that of the main negative potential patch. Moreover, the negative potentials of these regions are not as strong as that of the main negative potential region, due in part to charge neutralization by adjacent Lys residues. Of the three aromatic residues, the two not present in secondary structural elements are Tyr-65 and Trp-68. Both Tyr-65 and Trp-68 are close to the large negative potential surface and could also be involved in the interaction of Ugi with Ung. Tyr-65 is close to His-44 in the tertiary structure of free Ugi, but this may not be the case when Ugi is bound to Ung. Tyr-47 is also in this region and may also be involved in the Ugi-Ung interactions. Thus, the tertiary structure of Ugi brings together a large number of negatively charged residues to form an electrostatic knob. This unusual feature is the result not only of the negatively charged residues at the ends of all five of the β-strands but of those in adjacent loop regions as well. The size of this knob as well as the magnitude of its electrostatic potential strongly suggest that this region is important in the interaction of Ugi with Ung. Since the electrostatic knob is confined to one face of the Ugi structure, it is likely that most of the interactions of Ugi with Ung will occur on this face. The charge-charge repulsion generated by bringing together so many negative charges is apparently offset by the stability of the β-sheet. The Ung·Ugi complex was prepared with 15N-labeled Ugi and the HSQC spectrum of the complex obtained. This spectrum is compared with that of free Ugi in Fig. 7. The results indicate that many residues of Ugi undergo considerable conformational change upon binding to Ung. Preliminary assignments of the resonances and preliminary analysis of NOE connectivities suggests that the secondary structure of Ugi does not change upon complex formation. It appears, rather, that the conformational changes are primarily confined to the loop regions. The presence of a conformational change is also consistent with prior kinetic studies of the Ung-Ugi interaction, which indicated that Ugi underwent a slow, irreversible isomerization or "locking" step upon binding to Ung(2Bennett S.E. Schimerlik M.I. Mosbaugh D.W. J. Biol. Chem. 1993; 268: 26879-26885Abstract Full Text PDF PubMed Google Scholar). This conformational change is most likely associated with this locking step. The shape of the electostatic knob in free Ugi is determined in part by the folding of the loop formed by residues 61-68 over the β-sheet. There is an apparent hydrogen bond between Tyr-65 and His-44 that constitutes a part of the interaction between the loop and the β-strands. Preliminary studies of the Ugi·Ung complex suggest that this interaction is disrupted in the complex; hence, the orientation of the 61-68 loop relative to the β-strands may be different in the complex. These results indicate that the Ugi protein folds in such a manner that a large region of highly negative electrostatic potential is generated. The acidic residues that comprise this region are the terminal residues of all five β-strands and those found in the loops. All four aromatic amino acid residues are located next to this electrostatic knob, and we speculate at least one of them may interact with the uracil binding site of Ung. Ugi undergoes a considerable conformational change upon binding to Ung, which is consistent with the kinetics of complex formation(2Bennett S.E. Schimerlik M.I. Mosbaugh D.W. J. Biol. Chem. 1993; 268: 26879-26885Abstract Full Text PDF PubMed Google Scholar). That this conformational change occurs primarily in the loop regions of Ugi may suggest that the electrostatic knob is present in the complex, although its size and shape may change upon complex formation. The assistance of Dr. Igor Goljer in performing some of the NMR experiments is appreciated.