Site-directed Mutagenesis and Characterization of Uracil-DNA Glycosylase Inhibitor Protein

Bacteriophage PBS2 uracil-DNA glycosylase inhibitor (Ugi) protein inactivates uracil-DNA glycosylase (Ung) by acting as a DNA mimic to bind Ung in an irreversible complex. Seven mutant Ugi proteins (E20I, E27A, E28L, E30L, E31L, D61G, and E78V) were created to assess the role of various negatively charged residues in the binding mechanism. Each mutant Ugi protein was purified and characterized with respect to inhibitor activity and Ung binding properties relative to the wild type Ugi. Analysis of the Ugi protein solution structures by nuclear magnetic resonance indicated that the mutant Ugi proteins were folded into the same general conformation as wild type Ugi. All seven of the Ugi proteins were capable of forming a Ung·Ugi complex but varied considerably in their individual ability to inhibit Ung activity. Like the wild type Ugi, five of the mutants formed an irreversible complex with Ung; however, the binding of Ugi E20I and E28L to Ung was shown to be reversible. The tertiary structure of [13C,15N]Ugi in complex with Ung was determined by solution state multi-dimensional nuclear magnetic resonance and compared with the unbound Ugi structure. Structural and functional analysis of these proteins have elucidated the two-step mechanism involved in Ung·Ugi association and irreversible complex formation.

The Bacillus subtilis bacteriophage PBS1 and -2 exhibit a unique genetic system that naturally contains uracil in place of thymine in a double-stranded DNA genome (1,2). Stable incorporation of uracil residues into the phage DNA is achieved by the substitution of dUTP for dTTP as precursor in DNA syn-thesis and the concomitant inactivation of the host uracilmediated base excision DNA repair pathway (2)(3)(4). To block uracil-DNA repair and protect the uracil-containing phage DNA from degradation, an early phage gene (ugi) 1 is expressed that inhibits the B. subtilis uracil-DNA glycosylase. The amino acid sequences of the PBS1 and -2 Ugi proteins appear to be identical (5)(6)(7).
The PBS2 ugi gene encodes a small (9,474 dalton), monomeric, heat stable protein of 84 amino acids that inactivates uracil-DNA glycosylases from diverse biological sources (5,8,9). The ugi gene product is an unusually acidic protein (12 Glu,6 Asp) with a pI ϭ 4.2 that migrates anomalously during SDS-polyacrylamide gel electrophoresis (5,10,11). Ugi inactivates Ung by forming a tightly bound noncovalent complex with 1:1 stoichiometry that is essentially irreversible under physiological conditions (10,12). Stopped-flow kinetic studies of the Ugi interaction with Escherichia coli Ung indicate that complex formation is accomplished through a two-step binding reaction (12). In the initial step, the association between free Ugi and Ung is characterized by a rapid pre-equilibrium reaction with a dissociation constant (K d ) of 1.3 M; the second step, the formation of an irreversible complex, is characterized by the rate constant k ϭ 195 s Ϫ1 . Thus, Ung⅐Ugi complex formation initiates with a "docking" interaction that facilitates optimal alignment between the two proteins. If correct alignment between Ung and Ugi does not occur, a reversible association will transpire. If, however, proper alignment is achieved, then a "locked" complex quickly follows.
The secondary and tertiary structures of free Ugi were recently determined by solution state multi-dimensional NMR techniques and found to include two ␣-helices and five antiparallel ␤-strands as illustrated in Fig. 1 (13,14). The five contiguous ␤-strands are connected by short loop regions to form an anti-parallel ␤-sheet. Analysis of the electrostatic potential of Ugi revealed several striking features (14). Seven of the 18 acidic amino acid residues (Glu-20, Asp-48, Glu-49, Asp-52, Glu-53, Asp-74, and Glu-78) come together to form a region of high negative potential on one face of the protein.
Each of the residues that form this electrostatic region or "knob" are located immediately adjacent to or terminate a ␤-strand. Two other acidic amino acid residues (Asp-40 and Asp-61) are also in juxtaposition to the end of ␤-strands; Glu-78 and Glu-64 reside in the loop regions (14). The remaining seven negatively charged residues are located in the ␣1-helix (Asp-6, Glu-9, and Glu-11) and ␣2-helix (Glu-27, Glu-28, Glu-30, and Glu-31). Both the ␣1and ␣2-helix elements project away from the ␤-sheet and are located on potentially flexible arms of the polypeptide (14). Furthermore, the ␣2-helix is longitudinally segmented into a hydrophobic face and a negatively charged face where the four glutamic acid residues protrude.
Several lines of evidence suggest that some of the negatively charged amino acid residues of Ugi may act as a DNA mimic and mediate the interaction with Ung. First, UV-catalyzed cross-linking of oligonucleotide (dT) 20 to the DNA-binding site of Ung blocked Ugi from forming a Ung⅐Ugi complex (15). Second, the x-ray crystallographic structure of Ugi in complex with human (16) and HSV-1 (17) uracil-DNA glycosylase reveals that the interfacing surface of Ugi shares shape and electrostatic complementarity to the DNA-binding groove of the enzyme (16,17). Third, the negative electrostatic knob of Ugi exhibits an electrostatic potential of Ͼ6.6 kcal, which is similar to that generated by the negatively charged phosphate backbone of DNA (14). Fourth, the recent x-ray structure of human uracil-DNA glycosylase complexed with a 10-bp oligonucleotide containing a target G⅐U mispair reveals the DNA complexed at the same site as Ugi (18). Fifth, charge neutralization by carbodiimide-mediated adduction of Ugi carboxylic acid residues caused a decrease in inhibitor protein activity (19). Finally, chemical adduction of specific glutamic acid residues (Glu-28 and Glu-31) of Ugi located in the ␣2-helix correlated with the formation of an unstable Ung⅐Ugi complex (19).
Bennett et al. (12) suggested that after the Ung/Ugi association, the transition to the locked configuration may involve a conformational change in either one or both proteins. Subsequently, Sanderson and Mosbaugh (19) proposed that the locking reaction is caused predominantly by a change in Ugi structure. This position is supported by a comparison of the crystal structures of free human and HSV-1 uracil-DNA glycosylase with the structures of each enzyme in complex with Ugi (16,17,20,21). In both cases, the tertiary structure of the enzyme shows only minor structural changes. In contrast, a comparison of the heteronuclear multiple quantum correlation spectra of free and bound [ 15 N]Ugi indicates that many residues of Ugi undergo conformational change upon binding to Ung (14). At present, the tertiary structure of the unbound Ugi protein in solution was determined solely by solution state NMR techniques (14). Also, a comparison of the solution structure of free Ugi with the crystal structure of Ugi complexed with either the human or HSV-1 uracil-DNA glycosylase demonstrates significant structural changes occur in Ugi (14,16,17). A more complete understanding of the docking and locking reactions may well be gained by determining the solution state structure of Ugi in complex with Ung.
In the present report we (i) conduct site-directed mutagenesis of seven acidic residues of Ugi; (ii) purify each mutant Ugi protein to apparent homogeneity; (iii) characterize each mutant Ugi with regard to specific activity and Ung⅐Ugi complex stability and reversibility; (iv) determine the structural similarity between wild type Ugi and the Ugi mutant proteins using NMR methods; (v) compare the free and complexed Ugi solution structures; and (vi) model the interactions in the wild type and mutant Ung⅐Ugi complexes.  (22) and pSB1051 (12) were constructed as described previously. Oligonucleotides were synthesized using an Applied Biosystems 380B DNA Synthesizer by the Center for Gene Research and Biotechnology (Oregon State University).

Materials-Restriction
Site-directed Mutagenesis of the Uracil-DNA Glycosylase Inhibitor Gene-The first step in the site-directed mutagenesis procedure involved subcloning of the ugi gene into a pBluescript-based phagemid to produce single-stranded DNA. The pZWtac1 EcoRI-HindIII restriction fragment (726 bp) containing the ugi gene ( Fig. 2) was inserted into the corresponding EcoRI and HindIII sites of pBluescript II SK(Ϫ) using T4 DNA ligase. The resulting phagemid (pAL) was transformed into E. coli JM105, plated on LB plates containing 100 g/ml ampicillin, 40 mM isopropyl-1-thio-␤-D-galactopyranoside, and 40 g/ml 5-bromo-4-chloro-3-indolyl ␤-D-galactopyranoside, after which pAL DNA was purified from white colonies. E. coli CJ236 (dut, ung) was then transformed with phagemid pAL and grown at 37°C in 1.0 liter of 2 ϫ YT medium supplemented with 34 g/ml chloramphenicol and 100 g/ml ampicillin. Upon reaching a cell density of 10 8 cells/ml (1 A 600 nm ϭ 8 ϫ 10 8 cells/ml), uridine was added to a final concentration of 0.25 g/ml; VCS-M13 helper phage was added at a multiplicity of infection equal to 1.0, and incubation was continued at 37°C for 1.5 h. Kanamycin (26 g/ml final concentration) was added to select for infected E. coli cells and growth continued for an additional 5.5 h. The culture was centrifuged at 7000 rpm for 15 min at 4°C in a GSA (Sorvall) rotor, and the supernatant fraction was processed to precipitate pAL phage with the addition of 0.25 volume of a 15% PEG-8000 and 2.5 M NaCl solution. Phage DNA was isolated from the supernatant fraction following extractions with phenol and chloroform:isoamyl alcohol (24:1) and precipitation with ethanol (23). The precipitated DNA was centrifuged at 9500 FIG. 1. Tertiary structure of the uracil-DNA glycosylase inhibitor protein and location of Glu and Asp residues. The tertiary structure of Ugi determined by solution state NMR techniques (14) is shown with the 12 Glu residues in red and 6 Asp residues in yellow. Secondary structural elements include the ␣1-helix (Ser-5 to Lys-14), ␣2-helix (Glu-27 to Asn-35), ␤1-strand (Glu-20 to Met-24), ␤2-strand (Ile-41 to Asp-48), ␤3-strand (Glu-53 to Ser-60), ␤4-strand (Ala-69 to Asp-74), and ␤5-strand (Asn-79 to Leu-84). rpm for 20 min at 4°C in a SA600 (Sorvall) rotor, air dried, and resuspended in 500 l of TE buffer (10 mM Tris-HCl (pH 8.0), 1 mM EDTA). This single-stranded uracil-substituted DNA that contained the antisense ugi gene sequence was termed pALU(ss) DNA.
The second step involved in vitro primer extension of various oligonucleotides containing site-directed mutations in the ugi structural gene. Deblocked/deprotected oligonucleotides were purified by Sephadex G-25 chromatography and phosphorylated at the 5Ј end using ATP and T4 polynucleotide kinase as described previously (23). Defined DNA primer/templates were constructed by annealing various oligonucleotides (20 pmol) to pALU(ss) DNA at a 3:1 (primer:template) ratio in a 100-l volume as described (24,25). Primer extension reaction mixtures (100 l) contained 20 mM Hepes-KOH (pH 8.0), 2 mM dithiothreitol, 10 mM MgCl 2 , 2 mM ATP, 500 mM each of dATP, dCTP, dGTP, and dTTP, 0.5 mg/ml BSA, 6 units of T4 DNA polymerase, 200 units of T4 DNA ligase, and 1.8 pmol of the heteroduplex pALU DNA primer/ template. Following incubation for 5 min at 25°C and then 2 h at 37°C, a sample (10 l) terminated with the addition of 1.5 l of 0.1 M EDTA was analyzed by 1% agarose gel electrophoresis to determine the extent of primer extension. Transformation of E. coli JM105 CCMB80 competent cells was performed with 10 l of the primer extension reaction mixture (26). Transformed bacterial colonies were grown on LB plates containing 100 g/ml ampicillin, and isolated colonies were subsequently grown to saturation in 2 ml of 2 ϫ YT medium supplemented with ampicillin. Plasmid DNA (pSugi) was isolated using the Wizard Miniprep DNA purification technique (Promega). Isolated plasmids (pSugi) were analyzed by 1% agarose gel electrophoresis after three independent restriction endonuclease digestions as follows: (i) HindIII, (ii) HindIII plus EcoRI, and (iii) the appropriate novel restriction endonuclease for the site introduced into the pSugi DNA by site-directed mutagenesis (Fig. 2).
The third step of the procedure involved subcloning the ugi genes containing site-directed mutations from pSugi to the pKK223-3 derived overexpression vector pZWtac1, replacing the wild type ugi gene. Both pSugi and pZWtac1 DNA (1 g) were separately digested with excess HindIII and EcoRI, and the products were resolved by 0.8% agarose (low melting point) gel electrophoresis. Bands corresponding to the ugi gene containing DNA fragment (726 bp) from pSugi and the 4.6-kb fragment from pZWtac1 were excised, DNA extracted, and purified using glass milk (27). Samples containing the 726-bp and 4.6-kb EcoRI/ HindIII DNA fragments were mixed, and the complementary ends were joined using T4 DNA ligase. The ligation reaction mixture (10 l) contained 50 mM Tris-HCl (pH 7.5), 10 mM MgCl 2 , 10 mM dithiothreitol, 1 mM ATP, 25 g/ml BSA, 0.75 pmol of the 726-bp fragment, and 0.25 pmol of the 4.6-kb DNA fragment. Following incubation for ϳ16 h at 16°C, the ligation mixture was used to transform either E. coli JM105 or XL2-Blue competent cells (26). Transformed colonies were isolated after growth on LB plates supplemented with 100 g/ml ampicillin and grown overnight in 2 ml of 2 ϫ YT medium containing ampicillin. Plasmid DNA was then isolated using the Wizard Miniprep procedure and analyzed using restriction endonuclease digestion for HindIII, HindIII plus EcoRI, and the appropriate introduced restriction sites (Fig. 2) as described above. In the resulting plasmids (pKugi), the mutant ugi gene was expressed under the control of the isopropyl-1thio-␤-D-galactopyranoside-inducible tac promoter.
DNA Sequence Analysis-Double-stranded pKugi DNA containing site-directed mutations were used as templates for nucleotide sequencing using the dideoxynucleotide chain termination method originally described by Sanger et al. (28). Either primer FP/PKK that was complementary to the ugi sense strand at position Ϫ143 to Ϫ122 or primer IRPUGI that hybridized to the antisense strand at position ϩ379 to ϩ401 was used to initiate DNA polymerase-mediated synthesis. DNA sequencing was conducted using an Applied Biosystems model 373A sequencer by the Center for Gene Research and Biotechnology (Oregon State University).

Purification of [ 3 H]Ung⅐Ugi
Complexes-Wild type Ugi or various site-directed mutants of Ugi protein were mixed with [ 3 H]Ung in buffer A (30 mM Tris-HCl (pH 7.4), 1 mM EDTA, 1 mM dithiothreitol, 5% (w/v) glycerol) containing 50 mM NaCl and incubated at 25°C for 10 min and then at 4°C for 20 min. Following complex formation, each sample was applied to a DE52 cellulose column equilibrated in buffer A containing 50 mM NaCl, washed with equilibration buffer, and step-eluted, as described previously (19). Fractions (1 ml) were collected and samples were analyzed for 3 H radioactivity. The [ 3 H]Ung⅐Ugi complex was detected by 18% nondenaturing polyacrylamide gel electrophoresis, and fractions containing complex were pooled and concentrated using a Centriplus-10 (Amicon) concentrator.
Enzyme Assays-Uracil-DNA glycosylase inhibitor activity was measured under previously described conditions (10). When appropriate, Ugi was diluted with IDB buffer (50 mM Tris-HCl (pH 8.0), 1 mM EDTA, 1 mM dithiothreitol, 100 mM NaCl). One unit of uracil-DNA glycosylase inhibitor inactivates 1 unit of uracil-DNA glycosylase in the standard reaction. Uracil-DNA glycosylase activity was similarly measured except that Ugi was omitted (10). One unit of uracil-DNA glycosylase is defined as the amount that releases 1 nmol of uracil/h under standard reaction conditions.
Electrophoresis-Sodium dodecyl sulfate-polyacrylamide slab gel electrophoresis was performed essentially as described by Laemmli (29) and modified by Bennett et al. (12).
Nondenaturing polyacrylamide slab gel electrophoresis was performed at 4°C essentially as described by Sanderson and Mosbaugh (19) with resolving gels containing 18% acrylamide and 0.39% N,NЈmethylenebis (acrylamide). The gel was immediately stained using the rapid stain procedure described by Reisner (30) and modified by Sanderson and Mosbaugh (19).
Nondenaturing polyacrylamide tube gel electrophoresis was conducted using the same components in the resolving gel (9 ϫ 0.6 cm diameter) and stacking gel (1 cm) as described above. Following electrophoresis the resolving gel was either stained with Coomassie Brilliant Blue G-250 or sliced horizontally into 3.1-mm sections, placed into scintillation vials, dehydrated overnight, and solubilized in 500 l of H 2 O 2 at 55°C for 24 -36 h as described by Sanderson and Mosbaugh (19). After complete solubilization, 5 ml of Formula 989 fluor was added and 3 H and 35 S radioactivity was measured by scintillation spectrometry.
Nuclear Magnetic Resonance Analysis-All of the wild type and mutant Ugi protein samples were concentrated to 7-18 mg/ml and dialyzed against NMR buffer containing 25 mM deuterated Tris, 0.2 mM EDTA, 0.2 mM EGTA, and 100 mM NaCl, at pH 7.0. NOESY experiments were carried out using a 150-ms mixing time at 25°C on a Varian INOVA 500 MHz NMR spectrometer. The spectral width in each dimension was 7500 Hz. The final pulse in the NOESY experiment was replaced with a watergate gradient pulse sequence and also a gradient pulse before each equilibration delay. The watergate sequence used a 1-ms z-direction gradient followed by a 2.2-ms selective shaped pulse for water and a 180°pulse followed by another 2.2-ms selective shaped pulse and another 1-ms z-direction gradient. A weak z-direction gradient was applied in the first half of t 1 and its negative applied in the second half of t 1 . Each NOESY experiment had 512 increments in t 1 with an acquisition time of 137 ms. Each NOESY spectrum was transferred into 1024 ϫ 4096 points and weighted using shifted Gaussians along each dimension.
Protein Structure Determination of [ 13 C, 15 N]Ugi Complexed to Ung in Solution-Samples of [ 13 C, 15 N]Ugi were complexed with unlabeled Ung as described previously (14). All of the NMR spectra were obtained with the sample at 30°C using a Varian Unityplus 400 spectrometer equipped with a Nalorac ID400 probe (31)(32)(33)(34). Data were acquired using States-Haberkorn for the indirectly detected dimension and using shifted Gaussians in the data processing along each dimension. The results of a 60-ms mixing time 15 N TOCSY-HSQC spectrum were used to identify spin systems. The experiment had an acquisition time of 0.108 s and there were 128 increments of t 1 and 24 increments of t 2 for each data set. The data were linearly predicted to 256 points in t 1 and 48 points in t 2 before being Fourier-transformed into 512 ϫ 128 ϫ 1024 points. The spectral widths were 5000 Hz for F 1 , 1500 Hz for F 2 , and 5000 Hz for F 3 .
A 15 N/ 1 H NOESY-HMQC spectrum was recorded with a mixing time of 200 ms and 16 transients per increment. There were 1024 points in F 3 and an acquisition time of 0.108 s was used. There were 128 increments in t 1 and 20 increments in t 2 . The data were linearly predicted to 256 points in t 1 and 40 points in t 2 before being Fourier-transformed into 512 ϫ 128 ϫ 1024 points. The 15 N NOESY-HMQC data was used along with the 15 N TOCSY-HSQC to make the chemical shift identification with the 15 N and HN chemical shifts of Ugi. These assignments led to 707 NOE constraints for structure determination. The spectral widths were 5000 Hz for F 1 , 1500 Hz for F 2 , and 5000 Hz for F 3 .
A 13 C TOCSY-HSQC-SE spectrum with a 40-ms mixing time was used to group the spin systems for 13 C-labeled atoms. The data were collected with 16 transients per increment. The acquisition time was 0.108 s. There were 256 increments of t 1 for each of the complex data sets. The data were linearly predicted to 512 points in t 1 before being Fourier-transformed into 512 ϫ 1 ϫ 1024 points. The spectral widths were 5000 Hz in F 1 and 12000 Hz in F 2 .
A 13 C NOESY-HMQC with a mixing time of 150 ms was collected with 16 transients per increment. The acquisition time was 0.0832 s.
There were 128 increments in t 1 and 48 in t 2 . The data were linearly predicted to 228 points in t 1 and 96 points in t 2 before being Fouriertransformed into 512 ϫ 256 ϫ 1024 points. The spectral widths were 5000 Hz in F 1 , 12000 Hz in F 2 , and 5000 Hz for F 3 . These assignments led to 325 NOE constraints. A two-dimensional 13 C NOESY-HMQC with a mixing time of 160 ms was collected with 16 transients per increment. There were 330 increments in t 1 with the offset set in the middle of the aromatic region. Analysis of this NOE spectrum gave 35 NOE constraints. The normalized Z4-score analysis of 1 HN, 1 H␣, 13 C␤, and 15 N chemical shifts for Ugi produced 106 and dihedral constraints (35).
The constraints were grouped into strong, medium, and weak. A strong NOE peak was constrained to 1.8 Ͻr Ͻ4.0 Å, a medium NOE peak was constrained to 2.1 Ͻ r Ͻ 4.5 Å, and a weak NOE peak was constrained to 2.4 Ͻ r Ͻ 5.0 Å during simulated annealing and refined simulated annealing protein structure determinations protocols. Once the secondary structure of Ugi in the complex was determined there were 24 sets of hydrogen bonds that were used for a total of 48 constraints. The hydrogen bond constrained the oxygen to amide proton to be 1.8 Ͻ r Ͻ 2.5 Å and the oxygen to nitrogen distance to be 2.5 Ͻ r Ͻ 3.3 Å. The normalized Z4 score analysis of chemical shifts for Ugi produced 106 and dihedral constraints (35).
The simulated annealing and refinements protocols followed the same procedures as described for the structure of the free uracil-DNA glycosylase inhibitor protein (14) as were previously reported (36). The simulated annealing and refinement protocols were run on an IBM 3CT running X-PLOR 3.1 (37).
Protein Modeling-Starting with the HSV-1 uracil-DNA glyco-sylase⅐Ugi complex co-crystal coordinates described by Savva and Pearl (6,17), mutant Ugi forms in complex were generated by exchanging an individual wild type Ugi amino acid with a mutant residue using the residue replacement command in INSIGHT (BIOSYM). This is thermodynamically reasonable as all the mutant Ugi structures are similar to that of the free Ugi structure as evidenced by their NOESY spectra. Complete free energy analysis of the transition from the free to the bound form of Ugi is not computationally feasible. Therefore, rigid body energy minimizations were performed to determine a reasonable estimate of the ⌬⌬E involved between mutant forms of Ugi and wild type Ugi when bound to Ung (38). These calculations do not take into account the energy differences involved in the structural conformation changes that occur during binding to Ung. Rigid energy minimizations were then executed using an IBM 3CT running X-PLOR 3.1 (36). The rigid energy minimization procedure utilized all residues within 5 Å of the ␣2-helix and ␤1-strand of Ugi which included 40 residues of Ugi and 38 residues of uracil-DNA glycosylase. A dielectric constant of 6.0 was used to compensate for not using water with a cut-on distance of 6.0 Å and a cutoff distance of 6.5 Å. Energy minimizations were conducted using a two-step method. The first step involved 1000 iterations of rigid energy minimization with a large van der Waals radius but without considering electrostatic forces. In the second step, 2000 iterations were conducted with both electrostatic interactions and normal van der Waals radius influencing the structure. After the rigid energy minimizations converged, the minimized structure of the uracil-DNA glycosylase⅐Ugi complex emerged. Each individual unbound wild type and mutant Ugi structure was similarly generated. Interaction energies were calculated by combining the van der Waals, electrostatic, and hydrogen bond energies of the enzyme-inhibitor complex and unbound Ugi. Changes in interaction energies, ⌬E int , are defined as the difference in the interaction energies of the uracil-DNA glycosylase⅐Ugi wild type and mutant complex. The differences in the change in the interactive energies ⌬⌬E int are defined by subtracting the difference of the ⌬E int of the wild type and mutant Ugi from the ⌬E int of the mutant Ugi-containing complex.

Site-directed Mutagenesis of the Uracil-DNA Glycosylase
Inhibitor Gene-To investigate the role of specific negatively charged amino acid residues in the Ung/Ugi interaction, sitedirected mutagenesis producing single amino acid substitutions was performed on the ugi gene. The specific sites and substitutions selected were based on knowledge of the 1.9-Å crystal structure of Ugi complexed with human uracil-DNA glycosylase (16). Significant similarity exists between the human and E. coli enzyme around the proposed sites of Ung/Ugi interaction (Table I). Oligonucleotides were synthesized that introduced a codon change at seven Glu or Asp sites and a new restriction endonuclease cleavage site into the ugi gene as indicated in Fig. 2. To overproduce the mutant Ugi proteins, the EcoRI/HindIII DNA fragment containing the ugi structural gene was subcloned into the overexpression vector pKK223-3 producing a set of pKugi plasmids. Two methods were used to verify that the engineered mutations had been introduced into each pKugi DNA. First, restriction endonuclease digestions were conducted to establish the presence of the newly introduced recognition site within the EcoRI/HindIII DNA (726 bp) fragment. Second, dideoxynucleotide chain termination DNA sequencing of double-stranded pKugi DNAs was performed (data not shown). For all mutants, the entire ugi gene was bidirectionally sequenced and the results confirmed the designed nucleotide changes, exclusively.
Purification and Specific Activity of the Mutant Ugi Proteins-To facilitate characterization of the inhibitor activity exhibited by wild type Ugi and seven mutant Ugi proteins, each protein was overproduced using the appropriate pKugi vector and purified according to Sanderson and Mosbaugh (19). The purity of Ugi from the final purification step (fraction IV) was analyzed using 20% SDS-polyacrylamide gel electrophoresis (Fig. 3A). As previously observed the electrophoretic mobility of wild type Ugi was greater than that predicted for a 9474-dalton protein (10). Each mutant Ugi protein migrated with a unique slower mobility with respect to wild type Ugi, consistent with the elimination of a negatively charged residue by site-directed mutagenesis. However, these observations also imply that the mutant Ugi proteins exhibit different propensities to bind SDS or adopt to different protein conformations during electrophoresis, since each mutant protein carries the same charge. The specific activity of each purified Ugi protein was determined under standard conditions (Fig. 3B). Ugi E20I was essentially void of inhibitory activity, displaying ϳ1% of the wild type specific activity, whereas Ugi E78V was virtually unaffected, displaying ϳ105% activity. The four mutations in Glu residues located in the ␣2-helix (E27A, E28L, E30L, and E31L) showed progressively decreased levels of activity with 95, 88, 70, and 53% of control activity, respectively. Significant inactivation was also observed with the Ugi D61G protein which showed ϳ25% of wild type Ugi activity.
Ability of Mutant Ugi Proteins to Form a Complex with Ung-To determine whether the mutant Ugi proteins were able to form a Ung⅐Ugi complex, a 3-fold molar excess of Ung was incubated with each Ugi protein under standard binding conditions. The resultant Ung⅐Ugi complexes were resolved from the component proteins by nondenaturing polyacrylamide gel electrophoresis (Fig. 4). As controls, free Ung, wild type Ugi, and a 3:1 ratio of Ung⅐Ugi were analyzed for comparison with mutant forms of free Ugi and Ung⅐Ugi complexes. Each mutant Ugi protein migrated as a single band with a mobility similar to but slightly slower than that of wild type Ugi. In each case, the mutant Ugi proteins formed a Ugi⅐Ung complex that also migrated slightly slower than the wild type complex. With the exception of Ugi E20I, it appeared that each mutant Ugi protein formed a stable and complete complex with Ung since no free Ugi was detected. In contrast, the appearance of some free Ugi E20I, less Ung⅐Ugi E20I complex, and a smear of protein between the Ung and Ung⅐Ugi E20I complex bands suggested that Ugi E20I formed an unstable complex (Fig. 4, lane 5).
Relative Ability of Mutant and Wild Type Ugi Proteins to Complex with Ung-Competition experiments were conducted to determine the relative ability of each mutant Ugi protein to form a complex with Ung in the presence of wild type Ugi. Ung was incubated with a 2-fold molar excess of a mixture of Ugi and/or Ugi E27A at various ratios. The proteins were then resolved by nondenaturing polyacrylamide gel electrophoresis and detected by Coomassie Blue staining (Fig. 5A, lanes 1-6). Under these conditions, the Ung⅐Ugi E27A complexes were only partially resolved, whereas free Ugi and Ugi E27A were separated as independent bands. To quantitatively analyze the  (17), that involving PBS2 Ugi and human UDG1 were described from a 1.9-Å crystal structure (16). c E. coli Ung amino acid residues corresponding to the aligned HSV-1 and human uracil-DNA glycosylase polypeptides are indicated (9). d Chemical interactions listed include the following: H1, a hydrogen bond between carboxylate and Thr backbone amide or Arg side chains; H2, a pair of hydrogen bonds between the carboxylate and Ser backbone amide and side chain O␥; WH1, water-mediated hydrogen bond between the carboxylate and Ser-O␥; WH2, water-mediated hydrogen bonds with His backbone amide and Ser-O␥; WHN, hydrogen-bonded network with ordered solvent molecules and backbone atoms of UDG1 residues; and SB, salt bridge. ability of mutant Ugi proteins to compete with the wild type inhibitor protein, similar experiments were conducted after mixing each mutant Ugi with wild type [ 35 S]Ugi and incubating the mixtures with Ung. Following electrophoresis, 35 S radioactivity was detected in two bands that corresponded to [ 35 S]Ugi free and in complex. Thus, the amount of [ 35 S]Ugi detected in the complex band reflected the competitive ability of the mutant inhibitor protein to stably associate with Ung while in the presence of wild type Ugi. As a control, various ratios of [ 35 S]Ugi to Ugi (both wild type proteins) were mixed and analyzed by electrophoresis (Fig. 5B, black bars) Fig. 6A. Analysis of fractions across the peak by nondenaturing polyacrylamide gel electrophoresis verified that Ͼ95% of [ 3 H]Ung formed complex and that no detectable free [ 3 H]Ung or Ugi was observed in these fractions (Fig. 6A, inset). Stable preformed [ 3 H]Ung⅐Ugi complexes were isolated using this procedure for each mutant Ugi protein (Fig. 6B) (Fig. 6C). This value represents a background level when comparing results with the mutant preformed complexes. Of the six mutant Ugi contained in preformed complexes, only Ugi E28L was significantly displaced by wild type [ 35 S]Ugi (Fig. 6C) 5,7,9,11,13,15 (Fig. 7, closed circles) (Fig. 7, open circles). While the Ugi E28L mutant was capable of forming a stable complex with Ung, an irreversible complex was not achieved.

Solution State Structure of Mutant and Wild Type [ 15 N]Ugi
Proteins-NMR structural determinations were made to analyze and compare the polypeptide structures of the wild type and seven mutant Ugi proteins. Each Ugi protein showed onedimensional proton spectra consistent with a well ordered and folded structure (data not shown). The one-dimensional spectra obtained on the samples in 2 H 2 O also indicated that all eight Ugi proteins exhibited about the same number of slowly exchanging amide protons. In addition, the distribution of the amide proton chemical shifts was consistent with each mutant Ugi containing a high percentage of ␤-structure, as is the case for wild type Ugi (13,14). Structural determinations were also made by comparing secondary structural NOE peaks from amide to amide and amide to ␣-NOESY spectra. The amide to amide NOESY spectra for wild type and mutant Ugi forms are shown in Fig. 8. Each mutant Ugi protein was found to contain two ␣-helices and five ␤-strands identical to the secondary structural elements as exhibited by the unbound wild type Ugi protein. The NOESY spectra of each mutant Ugi was compared with that of the assigned wild type spectrum. Detailed examination showed that Ugi E20I, D61G, and E78V have structures that are very similar to wild type Ugi. However, the chemical shifts of many of the cross-peaks in the Ugi E20I spectrum are distinct from those of the wild type protein. Analysis of the NOESY spectra for Ugi E27A, E28L, E30L, and E31L likewise indicated close structural similarity to wild type Ugi, with the exception of the ␣2-helix length. Ugi E27A and E28L contained an ␣2-helix that was shorter at the N-terminal end, whereas Ugi E30L and E31L exhibited a shorter ␣2-helix at the Cterminal end.
Solution State Structure of [ 13 C, 15 N]Ugi Complexed to Ung-To determine the structure of Ugi bound to Ung, a sample of [ 13 C, 15 N]Ugi (820 nmol) was combined with an excess of unlabeled Ung (1220 nmol), and the Ung⅐[ 13 C, 15 N]Ugi complex (1.27 mM) was prepared as described previously (14). NMR structural determinations were made using 15 N-TOCSY-HSQC, 15 N/ 1 H-NOESY-HMQC, 13 C-TOCSY-HSQC, and 13 C-TOCSY-HSQC-SE spectra, and the solution state structure of [ 13 C, 15 N]Ugi complexed to Ung is shown in Fig. 9. A comparison of free [ 15 N]Ugi (Fig. 9A) to [ 13 C, 15 N]Ugi bound to Ung (Fig.  9B) indicates that significant structural change of Ugi occurred as a consequence of complex formation. The electrostatic surfaces of the free (Fig. 9, C and E) and complexed Ugi protein (Fig. 9, D and E) were evaluated using the GRASP program (38). DISCUSSION We have used site-directed mutagenesis to assess the role of specific negatively charged amino acids in Ugi activity. Three structural domains of Ugi were targeted that included the ␤l-strand (E20I), the ␣2-helix (E27A, E28L, E30L, and E31L), and the loop regions joining the anti-parallel ␤-strands (D61G, E78V). To gain information about the structural changes induced by the specific amino acid replacements, NMR spectral analysis was performed for each Ugi protein. The one-dimensional proton spectra of the wild type and mutant Ugi proteins appeared to be quite similar indicating that each Ugi protein folded in much the same manner. The results of binding experiments indicated that each mutant Ugi protein remained capable of associating with Ung and forming a Ung⅐Ugi complex. However, complex stability and reversibility was found to be altered by some amino acid substitutions. These results suggest that none of the individual Ugi amino acids examined play an essential role in mediating Ung/Ugi binding. Rather, the negatively charged residues act collectively to facilitate stable complex formation.
The specific chemical interactions that stabilize the Ung⅐Ugi complex can be inferred from those in the x-ray crystallographic structures identified of HSV-1 (17) and human (16) uracil-DNA glycosylase⅐Ugi complexes. Such a comparison is justified since E. coli Ung shares extensive amino acid homology with its HSV-1 and human counterparts (39), and both co-crystal structures show significant structural similarity (16,17). The structure of Ugi in complex with Ung has been determined by conventional solution state methods and found to be essentially the same as the crystal structure (16,17). As indicated in Table I, the locations and types of interactions linking Ugi residues with either HSV-1 or human uracil-DNA glycosylase were found to be highly conserved. Amino acid sequence alignment of E. coli Ung to both the HSV-1 and human enzyme revealed identical or conservative substitutions at the sites of Ugi interaction. The ability of Ugi to perform DNA mimicry has apparently capitalized on the conservation of Ung residues located in the highly conserved DNA-binding pocket (16,18,21). This striking amino acid correspondence suggests that similar interactions most likely mediate the Ugi association with all three uracil-DNA glycosylases examined here and possibly others.
The Ugi E20I protein, although capable of forming a Ung⅐Ugi complex, did not completely block Ung activity, presumably due to an inability to form a stable complex with Ung. The instability of this association was evident from the dissociation detected during nondenaturing polyacrylamide gel electrophoresis, the inability to isolate a Ung⅐Ugi E20I complex by anion exchange chromatography, and the ineffectiveness of Ugi E20I to compete with wild type Ugi for Ung binding. The position of Glu-20 is apparently stabilized by a pair of hydrogen bonds between the carboxylate side chain of the conserved Ser-88 backbone amide and O␥ of E. coli Ung (Table I). In addition, a water-mediated hydrogen bond may also form between Ugi Glu-20 and Ung Ser-189, as has been described for the complex involving the HSV-1 enzyme (17). The loss of Ugi E20I activity may be explained by a weakening of these interactions due to charge neutralization or peptide conformational change surrounding this key residue. Protein modeling indicated that the van der Waals energy dropped considerably (Ϫ17 kcal/mol); in this case, more than enough to stop the interaction. Taken together the results suggest that Ugi E20I forms a frail unlocked complex that fails to prevent Ung association with uracil-DNA.
The four mutations (E27A, E28L, E30L, and E31L) created within the ␣2-helix had quite different effects on Ugi activity. While Ugi E27A retained near full inhibitor activity, the other three mutations caused a progressive reduction of Ugi-specific activity (E28L Ͼ E30L Ͼ E31L) with Ugi E31L maintaining ϳ50% wild type activity. The influence of these mutations was particularly interesting since a major structural difference between the free and complexed forms of Ugi involves the orientation of the ␣2-helix (Fig. 9, A and B). Furthermore, x-ray crystallographic studies have indicated that when in complex the ␣2-helix and ␤1-sheet resides over the DNA-binding groove and provides the majority of contacts between the enzyme and inhibitor (16,17). Therefore, it was not surprising that Ugi E27A activity was unaffected, since Glu-27 has not been impli-cated in complex interaction (Table I). The results obtained for Ugi E28L and E31L support the inference drawn from chemical modification that Glu-28 and/or Glu-31 play an important role  (14). The tertiary structure of [ 13 C, 15 N]Ugi bound to E. coli Ung was determined by solution state multidimensional NMR techniques as described under "Experimental Procedures." Several secondary structural elements are highlighted in both free Ugi (A) and in complex (B) structures as follows: ␣1-helix (light blue); ␣2-helix (dark blue); ␤l-strand (red); ␤2-␤5-strands (salmon); and the loop between ␤3and ␤4-strands (yellow). The location of the N-terminal (N), C-terminal (C), Glu-28 (28), and Glu-3l (31) residues are also indicated. The electrostatic surfaces of the free (C) and complexed (D) forms of Ugi were generated using the program GRASP (38) as described previously (14). Structures A and B of the free and bound Ugi correspond to the same view as indicated in C and D, respectively. The bottom panel depicts Ugi rotated 180°. The electrostatic potentials were calculated with a dielectic constant of 6.0 for the protein and 80.0 for the solvent. The ionic strength of the solution was set to 0. Only the charges of the side chains of Lys, Asp, Asn, Glu, and Gln residues were used. The electrostatic potential cutoff was set to 6.6 kcal/mol, and the regions with a negative potential of this magnitude are shown in red, and the regions with a positive potential of this magnitude are shown in blue.
in promoting stable Ung⅐Ugi complex formation (19). The unique ability of Ugi E28L to form a stable but reversible complex when challenged with wild type Ugi indicates that this residue plays a critical role in forming the locked complex. Like Glu-20, Glu-28 appears to form hydrogen bonds with a conserved Ser (Ser-166 of E. coli Ung) amide and the side chain O␥ (Table I). In addition, Glu-28 also forms water-mediated hydrogen bonds to a universally conserved active site His backbone amide and Ser O␥ atom (His-187 and Ser-192 of E. coli Ung). We speculate that these contacts are responsible, at least in part, for creating the irreversible nature of the Ung⅐Ugi complex. Additionally, the observation that Ugi E31L, like wild type Ugi, formed an essentially irreversible complex with Ung argues that Glu-31 does not play a major role in the locking reaction.
The two mutations in the loop regions connecting the consecutive ␤-strands of Ugi provided distinctly different results. Ugi D61G caused ϳ75% reduction of activity, whereas Ugi E78V showed a specific activity equivalent to wild type Ugi. The inability of the E78V mutation to affect activity may be explained since Glu-78 resides within the electrostatic knob region of Ugi that contains seven Glu or Asp residues (14). The results suggest that neutralizing the negative charge of Glu-78 may have little effect on the overall inhibitory action due to the relatively small individual contribution of Glu-78. The involvements of Glu-61 in Ugi/Ung binding remains to be determined; however, the reduced activity of Ugi D61G was not attributed to a defective locking reaction.
Several lines of evidence have led to a proposal that free Ugi undergoes a conformational change during formation of the Ung⅐Ugi complex (14, 16, 17, 19 -21). A direct demonstration of this change is evident by comparing the NMR solution structure of free [ 15 N]Ugi with that of [ 13 C, 15 N]Ugi complexed to E. coli Ung (Fig. 9). Clearly, the tertiary structure of the free and bound Ugi are quite different; however, both forms of the protein contain similar secondary structural elements (i.e. two ␣-helices and five ␤-strands). The ␤2-␤3-␤4-␤5 portion of the anti-parallel ␤-sheet remains generally unchanged in the two structures and provides a focal point for comparison. The major transition between these structures involves a collapse of the polypeptide segments containing the ␣1and ␣2-helix. In the unbound state, both helices extend away from the core of Ugi (14). We speculate that this Y-shaped structure may arise from the negative charge repulsion between the negative electro-static knob and both the negatively charged ␣1and ␣2-helices. Upon binding to Ung, the flexible arms containing the ␣1and ␣2-helix reorient to allow the positioning of ␤1 and ␣2 over the positively charged DNA-binding pocket of uracil-DNA glycosylase (16,17). As a consequence, several other structural changes occur as follows: (i) the ␤1-strand becomes twisted; (ii) the ␣1and ␣2-helix move toward the core of Ugi; and (iii) the loop between the ␤3and ␤4-strands becomes slightly reoriented. The orientation and negative charge of Glu-28 in the DNA-binding pocket mediates the formation of the locked complex and excludes DNA. The involvement of a Ugi structural change may explain the specificity exhibited by this inhibitor protein toward uracil-DNA glycosylases acting through a mechanism involving DNA mimicry.
The structural changes that occur during complex formation have a pronounced effect on the electrostatics of Ugi (Fig. 9). The primary changes appear to result from the positions of the ␣1and ␣2-helices relative to the rest of the protein. The ␣1-helix is positioned behind the ␤-sheet of the complex structure shown, and the ␣2-helix is positioned in front. There are smaller changes of the ␤-strands. The helices appear to have relatively few interactions with the rest of the protein in the free form, and there may be no particularly large barriers between the free and bound conformations. The positioning in the bound state of the ␣1-helix effectively covers the electrostatic potential of the knob region that is exposed in the free protein. The position of the ␣2-helix in combination with the modest rearrangements of the ␤-strands gives rise to a large negative electrostatic potential on the face that forms most of the contacts in the Ung complex. This suggests that electrostatic interactions will play a considerable role in the complex and that the ␣2-helix appears to be involved in these interactions.
Molecular modeling studies were conducted using the cocrystal coordinates of the HSV-1 uracil-DNA glycosylase⅐Ugi complex and variations of the free and bound Ugi structure. The models only allowed differences in the position of the amino acid side chains corresponding to the mutant Ugi proteins. Information concerning the contributions to the energies of complex stability was assessed for the mutations in the ␤1-strand and ␣2-helix. The modeling indicated that the complex containing Ugi E20I has by far the highest energy, consistent with the low activity of this protein. The modeling suggested that Ugi E20I forms the same unbound protein FIG. 10. Molecular modeling of the Ugi E28L and E31L mutant proteins complexed with uracil-DNA glycosylase. The Ugi protein on the left represents a partial polypeptide structure of free Ugi as previously determined by Beger et al. (14). The ␣2-helix and loop region joining the ␤3and ␤4-strands are shown with the location of Glu-28 and Glu-31 indicated. A portion of the HSV-1 uracil-DNA glycosylase⅐Ugi complex derived from the co-crystal coordinates described by Savva and Pearl (6,17) is illustrated on the right. Modeled structures of Ugi E28L and E31L (yellow) complexed with HSV-1 uracil-DNA glycosylase (light blue) were obtained as described under "Experimental Procedures" and are shown in the top and bottom middle, respectively. The changes in the location of the enzyme and inhibitor side chains in the complexes containing Ugi E28L and E31L are depicted (red). structure as wild type Ugi but that there are very unfavorable (ϳ15 kcal) van der Waals interactions in the complex. In contrast, Ugi E27A and E30L were found to have energies that are essentially identical to that of wild type Ugi in the complex. This is consistent with neither Glu-27 nor Glu-30 residues participating in an interaction with the enzyme (Table I) and both Ugi E27A and E30L showing only partially reduced activity. Models of Ugi E28L and E31L were examined in an attempt to explain the reason that Ugi E28L was the only mutant protein capable of forming a stable but reversible complex. As shown in Fig. 10, the positions of the Leu side chains in both mutant Ugi polypeptides were quite similar to the wild type Glu, although they are shorter in length. Since both mutant proteins are structurally very similar to wild type, we infer that the absence of the carboxyl group precipitates the change in properties of each mutant. The Leu side chain should not be capable of mediating the hydrogen bond interactions that stabilize the enzyme-inhibitor complex (Table I). Under this condition Ugi E28L appears capable of conducting the docking reaction but not the locking reaction. Analysis of the energy terms showed that the electrostatics of the complexes with Ugi E28L and E31L are ϳ5 and ϳ2 kcal, respectively, less favorable than that of complex containing wild type Ugi. Both mutant proteins showed ϳ3 kcal less stability than wild type Ugi in complex based on van der Waals forces. Hence, Ugi E28L differed from the other mutations in the ␣2-helix in that it not only had the highest energy but was unfavorable in both electrostatic and van der Waals energy relative to the wild type Ugi in complex. This suggests that the locking reaction may involve both the electrostatic potential and the hydrogen bond interactions of Glu-28.
This study has demonstrated that Ugi exists in three different conformational states (free, unlocked, and locked Ugi) during the binding reaction with Ung. The involvement of a significant structural transformation and role of Glu-20 and Glu-28 in mediating the locking reaction has been demonstrated. However, several issues remain to be investigated concerning the structure and function of Ugi during Ung complex formation. First, do individual amino acid residues play an essential role in the docking reaction? Second, what is the effect of various mutations on the kinetics of the Ung-Ugi interaction? Third, what is the structure of E. coli Ung when complexed with Ugi? Additional protein structural and biochemical analysis will be required to elucidate these important issues.