Genetically Engineered Zinc-chelating Adenylate Kinase fromEscherichia coli with Enhanced Thermal Stability*

In contrast with adenylate kinase from Gram-negative bacteria, the enzyme from Gram-positive organisms harbors a structural Zn2+ bound to 3 or 4 Cys residues in the structural motif Cys-X 2-Cys-X 16-Cys-X 2-Cys/Asp. Site-directed mutagenesis of His126, Ser129, Asp146, and Thr149 (corresponding to Cys130, Cys133, Cys150, and Cys153 in adenylate kinase from Bacillus stearothermophilus) in Escherichia coli adenylate kinase was undertaken for determining whether the presence of Cys residues is the only prerequisite to bind zinc or (possible) other cations. A number of variants of adenylate kinase from E. coli, containing 1–4 Cys residues were obtained, purified, and analyzed for metal content, structural integrity, activity, and thermodynamic stability. All mutants bearing 3 or 4 cysteine residues acquired zinc binding properties. Moreover, the quadruple mutant exhibited a remarkably high thermal stability as compared with the wild-type form with preservation of the kinetic parameters of the parent enzyme.

atom (5-7), a property which is due to the presence of 3 or 4 cysteine residues in the LID domain. Sequence alignment of AKs from Gram-positive and Gram-negative organisms, devoid of metal, showed that in the latter species the Cys residues are substituted with four other highly conserved amino acids, His, Ser, Asp, and Thr (Fig. 1B). This conservation suggests that these particular residues have some essential function, but different in the enzyme from the Gram-negative bacteria and eukaryotes. A noticeable exception is AK from the Gram-negative bacterium Paracoccus denitrificans. This enzyme not only conserves the Cys-containing sequence found in AK from Gram-positive species but binds zinc or iron (8).
In this study, we substituted His 126 , Ser 129 , Asp 146 , and Thr 149 in Escherichia coli adenylate kinase with cysteine residues. Our aim was to know whether a motif composed of 3 or 4 Cys residues generates a metal-binding site in AK or whether other structural factors contribute to the specificity (zinc versus iron or any other metal) or to the strength of the protein/metal interaction. On the other hand, we wanted to know the relevance of the metal binding for catalysis or stability of AK. A number of variants of AK e containing one to four cysteine residues were thus obtained. In agreement with previous studies on zinc-binding AKs, we found that the 3 and 4 cysteine modified forms of AK e acquired zinc binding properties. Moreover, the 4 cysteine-containing AK e exhibited an increased stability against thermal denaturation as compared with the wild-type form, with full conservation of its catalytic properties.
Site-directed Mutagenesis, DNA Sequence Analysis, and Growth Conditions-Site-directed mutagenesis was carried out according to Kunkel et al. (12). A 91 bases long primer (see Table I) containing 7 mismatched bases allowed several simultaneous substitutions in the adk gene. Mutant plamids from 48 randomly selected clones were further analyzed. A panel of single, double, triple, and quadruple mutants was obtained. Some additional variants, not resulting from this procedure, were created individually with appropriate primers (see Table I). Absence of any other mutation in the adk gene was checked on all plasmids. Overproduction of various AK forms was performed by growing strain BL21(DE3)/pDIA17 containing pVP1003 derivatives in LB medium (13) supplemented with 100 mg/liter ampicillin and 30 mg/liter chloramphenicol. Overproduction was carried out by adding 1 mM isopropyl-␤-D-thiogalactoside when the culture reached an absorbance at 600 nm of 1.0. Bacteria were harvested by centrifugation 3 h after induction.
Purification of AK e and Activity Assays-The adenylate kinase overproduced in E. coli was purified as described previously (14). When required, purified proteins were dialyzed against 50 mM ammonium bicarbonate, then lyophilized. Enzyme activity was determined at 30°C using the spectrophotometric assay (15). Measurements were made at 334 nm (0.5 ml final volume) using an Eppendorf ECOM 6122 photometer. One unit of enzyme activity corresponds to 1 mol of the product formed in 1 min at 30°C and pH 7.4 (in the direction of ATP formation). Protein concentration was determined according to Bradford (16), using a Bio-Rad kit. SDS-polyacrylamide gel electrophoresis was performed as described by Laemmli (17).
Zinc Content-The metal in various forms of AK was quantitated colorimetrically, using the metal-binding dye PAR as described previously (5, 7) and by atomic absorption spectrophotometry, using a graphite furnace instrument. The protein samples and the zinc standard solutions were diluted with water purified to 18.2 megohms/cm resistivity. In all cases, the background levels of zinc were insignificant.
Differential Scanning Calorimetry-The thermal stability of different proteins was studied by differential scanning calorimetry using an ultrasensitive Microcal MC-2D instrument at a scanning rate of approximately 50°C/h. Proteins in 50 mM Tris-HCl buffer (pH 7.4) were in the range of 1-1.5 mg/ml. Differential scanning calorimetry data were analyzed by the software provided by Microcal Inc., Northampton, MA.
Nomenclature-The mutants were named according to the position of key residues in the motif 126 His-X 2 -Ser-X 16 -Asp-X 2 -Thr 149 . Thus, the HSDT variant is the wild-type AK e . AK e C 4 corresponds to the 4 Cyssubstituted enzyme, AK e HC 3 to HCCC, AK e C 3 T to CCCT, AK e C 2 DT to CCDT, AK e HSC 2 to HSCC, and AK e HC 2 T to HCCT.

Overproduction and Purification of Cysteine-substituted
Variants of AK e -To create a 4-Cys-substituted mutant of AK e , a single nondegenerate 91-base oligonucleotide, was designed spanning the adk gene region corresponding to the LID domain in the protein (Table I). The 7 mismatched bases in the oligonucleotide allow simultaneous substitutions of His 126 , Ser 129 , Asp 146 , and Thr 149 codons with cysteines. Out of 48 randomly selected clones, two-thirds carried one or several substitutions with cysteine residue(s) and one-third harbored the expected quadruple modification. Considering the length of the oligonucleotide and the relatively low yield of its synthesis, the mutagenesis reaction was fairly effective and produced in one step a panel of single, double, and triple mutants displaying different positions of substitution. Missing species were constructed with appropriate primers (Table I).
The overproduced variants of AK e (about 20% of the soluble E. coli proteins) exhibited high specific activity in crude extracts (90 -120 units/mg of protein), close to that of the wildtype AK e , indicating that substitution with cysteine of any of the four targeted residues has no direct consequences on enzyme activity. Chromatography on blue Sepharose and Ultrogel AcA54 yielded homogeneous preparations of enzymes. The double Cys-substituted AK e species, HC 2 T, and CSDC were inactivated during purification in the absence of reducing agents. They conserved only 5% (23 units/mg of protein) and 37% (175 units/mg of protein) of wild-type activity.
Metal Binding-The zinc content of different variants of AK e , was quantified either by atomic absorption spectrophotometry or with the metal-binding dye PAR. The enzymes were first reacted with PMPS (18,19), the formation of the PMPSsulfhydryl chromophore being followed at 250 nm. Linear incorporation of PMPS into the proteins was observed up to 3.4 Ϯ 0.3 equiv./mole of C 4 mutant, 2.3 Ϯ 0.1 equiv./mole of HC 3 and C 3 T mutants ( Fig. 2A). The released Zn 2ϩ (0.73-0.82 mol of zinc/mol of protein) was determined spectrophotometrically with PAR (Fig. 2B). Atomic absorption spectrophotometry confirmed that the quadruple and the triple Cys mutations con- ferred to the protein the ability to bind the metal (0.8 Ϯ 0.1 mol of zinc/mol of protein). Less than 0.03 mol of zinc/mol of protein was found in the wild-type AK e . No iron was observed in AK e C 4 when E. coli was cultivated in minimal medium supplemented with this metal (data not shown).
Reaction of Wild-type AK e and of Cysteine-substituted Enzymes with DTNB-Wild-type AK e contains a buried cysteine residue in position 77. It reacted with DTNB only in the presence of urea over 2 M (15). The same was true for AK from Bacillus subtilis and AK from Bacillus stearothermophilus, although they contain, besides the conserved Cys 77 , 3 and, respectively, 4 other Cys residues in the LID domain (5,7). It was, therefore, surprising to find that Zn 2ϩ -chelating AK e variants reacted with DTNB under native conditions (Fig. 3). The kinetics of the reaction with DTNB of these mutants was fitted to a single exponential equation. Over 0.5 mM DTNB, the values of k obs (5.10 Ϫ3 s Ϫ1 for AK e C 4 , 7.10 Ϫ3 s Ϫ1 for AK e HC 3 and 14.10 Ϫ3 s Ϫ1 for AK e C 3 T) were practically independent on the concentration of thiol reagent. Thus, the first order process might reflect the dissociation rate of the AK e -Zn 2ϩ complex.
AK-Zn 2ϩ ª Zn 2ϩ ϩ AK e O ¡ DTNB AK* (Eq. 1) As preincubation of AK e C 4 with ZnCl 2 did not affect the k obs (data not shown), one might assume that difference in affinity for metal is primarily due to the dissociation rate constant for protein-metal complex. AK e (in 10 mM Tris-HCl, pH 8.0) was treated with 20 l of 10 mM DTNB, then the absorbance increase was read at 412 nm. The ratio of thiols reacted to moles of AK e was calculated using a molecular mass of 23.5 kDa. Symbols are the same as those used in Fig. 2, to which wild-type AK e (छ) and AK sub (ƒ) were added. The reactivity toward DTNB of the double substituted species of AK e was relevant for the structural changes into the LID domain of the bacterial enzyme. Thus, the HSC 2 variant reacted with DTNB under denaturing conditions. This means that Cys 146 and Cys 149 in this mutant are not exposed to the solvent and do not form a disulfide bridge as might be expected. On the other hand, the C 2 DT variant easily forms an intramolecular disulfide bridge, as suggested by its lack of reactivity toward DTNB both under nondenaturing and denaturing conditions. An even more complicated behavior was found with CSDC and HC 2 T forms of AK e , both of which were inactivated during the purification. Under nondenaturing conditions, 1.3-1.6 mol of SH/mol of enzymes were titrated with DTNB, indicating that the extra thiol groups were free and accessible. The observed inactivation of these species of AK e is likely due to structural deformation of the LID domain, which propagates to the CORE of the molecule.
Thermal Stability and Proteolysis by Trypsin of Cys-modified Mutants of AK e -In preliminary experiments, different proteins were heated for 10 min at various temperatures between 40 and 80°C, after which the residual enzyme activity was determined. The wild-type AK e and the C 3 T, CSDT, HCDT, and HSDC mutants were half-inactivated at temperature between 51 and 54°C; the C 4 and HC 3 variants exhibited a higher thermal stability (half-inactivation at 65 and 58°C, respectively) than the wild-type AK e , whereas the C 2 DT mutant was less resistant (half-inactivation at 46°C).
The thermal stability of the C 4 and HC 3 variants was further examined by microcalorimetry. The excess heat capacity curve for the wild-type AK e , C 4 , and HC 3 mutants is shown in Fig. 4. The T m values (63 and 55.7°C, respectively, instead of 51.8°C for the wild-type enzyme) were reproducible within 0.1°C. Inspection of Fig. 4 suggests that, at least under the conditions of the calorimetric experiments, the cooperativity of the denaturation process decreases significantly in the case of C 4 mutant. A detailed analysis of structural and energetic properties of this variant is described in a companion study (20).
Limited proteolysis was used as a test of conformational changes in AK e induced by various amino acid substitutions. Inactivation of the bacterial enzyme by TPCK-trypsin followed first order kinetics (7). The triple and quadruple Cys mutants of AK e showed similar or slightly higher resistance against trypsin digestion (t1 ⁄2 between 26 and 40 min) as compared with the wild-type enzyme. The other modified variants of AK e , except the HCDT mutant, exhibited a much lower resistance to proteolysis (t1 ⁄2 Ͻ 3 min). Sequence analysis of the proteolytic fragments indicated that 131 R-V 132 and 141 K-F 142 bonds located into the LID domain became sensible to the attack by trypsin; the 14-kDa fragment accumulated upon proteolysis corresponds to the segment 1-131 of the molecule (Fig. 5).
Catalytic Properties of Cys-substituted AK e - Table II shows the kinetic parameters of wild-type AK e , compared with two zinc-containing variants. The K m for nucleotide substrates was similar for the three variants of bacterial enzyme, and excess of   4 , and AK e HC 3 K m (ADP) and V max (ADP) were determined from plots of 1/v versus 1/ADP 2 , which assumes that the two molecules of ADP bind to the enzyme with the same affinity. The apparent K m for AMP and for ATP was determined at a single fixed concentration of cosubstrates (1 mM ATP and 0.2 mM AMP). The V max (ATP, AMP) was obtained by extrapolating the reaction rates for infinite concentrations of ATP and AMP and assuming that the concentration of one nucleotide substrate does not affect the apparent K m for the second nucleotide substrate. AMP (above 0.3 mM) inhibited the activity of all forms at a similar extent. It should be mentioned that removal of metal ion did not affect the phosphorylating activity of apoAK e C 4 or apoAK e HC 3 , confirming that zinc does not participate in the kinase activity. DISCUSSION Zinc in proteins is responsible for a wide range of functions (21)(22)(23)(24). The design of zinc binding sites generates proteins with new interesting properties (25)(26)(27)(28). The fact that zinc is a structural component of AKs from Gram-positive bacteria (6) prompted us to create a similar metal site in the enzyme from Gram-negative species. As shown here, the presence of three or four cysteine residues in the consensus sequence 126 Cys-X 2 -Cys-X 16 -Cys-X 2 -Cys 149 led to a zinc binding site in E. coli AK. Moreover, a significant increase in thermostability of the C 4 variant as compared with the wild-type AK e was observed.
The crystal structure of AK e shows that the LID domain forms a single distorted antiparallel ␤-sheet, two turns and one loop structure (29). The ␤-sheet is stabilized by hydrogen backbone interactions and attractive forces between few side chains inside the ␤-sheet. The four amino acids (His 126 , Ser 129 , Asp 146 , and Thr 149 ) replaced by cysteine residues in the mutagenized protein belong to this hydrogen binding network (Fig. 6). Three additional amino acid side chains (Arg 131 , Glu 151 , and Tyr 133 ) stabilize the network of hydrogen bonds by connecting the ␤-sheet segments in a sandwich-like structure. In the LID domain of adenylate kinase from Gram-positive bacteria, Zn 2ϩ which is held by cysteine residues seems to substitute efficiently the hydrogen bond network (30). The crystal structure of adenylate kinase from B. stearothermophilus (entry 1.Zip in the Protein Data Bank) confirms this observation. Metal chelation not only preserves the above mentioned network but also enhanced the thermal stability of the protein. As the catalytic properties of the C 4 and C 3 containing variants of AK e are conserved, the zinc-chelating LID domain of the protein conserves also intact the ability to rotate and to move like a solid block on the ATP-binding pocket. In other words, the overall conformation of this domain remains intact, in agreement with circular dichroism and NMR structural analysis (20).
The biochemical characteristics of the others variants of AK e might be also viewed in the light of hydrogen bond network located into the LID domain. The double mutants with vicinal thiols (C 2 DT and HSC 2 ) conserve over 65% of the activity of the wild-type enzyme. On the contrary the two mutants, where each Cys residue is located on one side of the sandwich-like structure (HC 2 T and CSDC), are greatly affected in their activity. The loss of activity was independent on disulfide bridge formation as in the latter cases the SH groups are free.
Among the Cys-monosubstituted variants, the most conservative substitution concerns Ser 129 . The S129C mutant exhibited similar structural and catalytic properties with the wildtype enzyme and with another AK e mutant (S129F) previously described by Haase et al. (31). This last mutation, however, is conditionally lethal, and the bacteria do not survive at 42°C. It was concluded that AK e might be involved in other essential cellular functions, independent of phosphotranferase activity, such as phospholipid synthesis. This attractive hypothesis still awaits for experimental proofs. All other single Cys variants (except HCDT form) of the AK e , although active and with similar thermal stability as the wild-type enzyme, exhibited a considerably lower resistance against trypsin digestion. In other words, despite the fact that the single amino acid substitutions were "conservative" in terms of hydrogen bond formation, some subtle conformational changes into the LID domain occur, yielding proteins with higher susceptibility to proteolytic digestion.
In conclusion, this study highlighted the importance of some key residues into the LID domain of the AK e . Quadruple and triple Cys mutations stabilized the protein by chelation with zinc. Double mutants are the most exposed to conformational changes leading to inactivation, irrespective of the presence or absence of disulfide bridges. All single Cys mutants are active but only one (HCDT) conserves the stability of the wild-type protein.