Mutational Analysis of Active Site Residues of Human Adenosine Deaminase*

Adenosine deaminase was overexpressed in a baculovirus system. The pure recombinant and native en- zymes were identical in size, Zn2+ content, and activity. Five amino acids, in proximity to the active site, were replaced by mutagenesis. The altered enzymes were purified to homogeneity and compared to wild-type adenosine deaminase with respect to zinc content, enzymatic activity, and kinetic parameters. All but one of the alterations produced significant activity pertur- bations. Replacement of Cysz02 produced a protein that retained at least 30-40% of wild-type activity. In con- trast, replacements of His17, His214, Hiszs8, and Glu217 resulted in dramatic losses of enzyme activity. None of these mutants exhibited large variations in K,. The proteins produced from alterations of amino acids implicated in metal coordination were slightly activated by inclusion of Znz+ throughout purification. These experiments confirm that in the active enzyme Zn2+ plays a critical role in catalysis, that a histidine or glutamate residue plays a mechanistic role in the hy- drolytic deamination step, and that cysteine is not involved in the catalytic mechanism of adenosine deam- inase. These data support the roles for these amino acid residues suggested from the x-ray structure of murine adenosine deaminase (Wilson, D. K., Rudolf, F. B., and Quicho, F. A. (1991) Science 252, 1278-1284). Adenosine an

proteins produced from alterations of amino acids implicated in metal coordination were slightly activated by inclusion of Znz+ throughout purification. These experiments confirm that in the active enzyme Zn2+ plays a critical role in catalysis, that a histidine or glutamate residue plays a mechanistic role in the hydrolytic deamination step, and that cysteine is not involved in the catalytic mechanism of adenosine deaminase. These data support the roles for these amino acid residues suggested from the x-ray structure of murine adenosine deaminase (Wilson, D. K., Rudolf, F. B., and Quicho, F. A. (1991) Science 252, 1278-1284).
Adenosine deaminase (EC 3.5.4.4), an important enzyme of the purine salvage pathway, catalyzes the irreversible hydrolytic deamination of adenosine or deoxyadenosine to inosine or deoxyinosine. Adenosine deaminase is expressed at very high levels along the entire murine gastrointestinal tract, in thymic T cells and in decidual cells of the developing maternal-fetal interface (Lee, 1973;Knudsen et al., 1988 andWitte et al., 1991). In humans, the upper gastrointestinal tract is devoid of this enzyme activity, but high levels are expressed in the lower part of the tract.
The wide spectrum of adenosine deaminase activity in mammalian tissues portended an important role for purine metabolism in nutrition and reproduction. However, the entire purine salvage pathway, and adenosine deaminase in particular, became the focus of intense interest with the observation that hereditary deficiency of the enzyme in human infants is invariably associated with a form of severe * This work was supported in part by United States Public Health Service Grant CA26391 (to M. S. C.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "aduertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
$ Recipient of an Army Predoctoral Fellowship in Biotechnology.
Present address: Laboratory of Molecular Growth Regulation, § To whom correspondence should be addressed.
combined immunodeficiency (Giblett et al., 1972). These patients have no obvious gastrointestinal tract abnormalities, but they do exhibit a dramatic lymphopenia that seems to be a direct consequence of the absence of adenosine deaminase (Coleman et al., 1978;Donofrio et al., 1978). Potent inhibitors of adenosine deaminase are lympholytic in humans, and this property has been exploited in the treatment of certain leukemias, the hallmark of which is accumulation of differentiation-arrested lymphocytes (Coleman, 1983). Ground and transition state analog inhibitors have also proven useful in studies of the reaction mechanism of adenosine deaminase. With a rate enhancement of about lo'', this enzyme is among the most efficient that have been described (Frick et al., 1987). A hydrate tetrahedral intermediate has been postulated from a large number of chemical studies (Evans and Wolfenden, 1973;Wolfenden et al., 1969;Kurz and Frieden, 1983). The most convincing evidence for this intermediate reaction product is from 13C NMR studies of adenosine deaminase bound to purine riboside (1,6-dihydropurine riboside), in which a change of hybridization from sp2 to sp3 is detected . Subsequent UV and NMR studies confirmed that this inhibitor is bound as an oxygen adduct, presumably hydrated at the 1,6 position (Jones et al., 1989). This covalent hydrate with C6 in the adenosine deaminase-purine riboside complex has been confirmed recently by the determination of its structure by x-ray crystallography (Wilson et al., 1991). Unexpectedly, the crystal structure also revealed that adenosine deaminase is a metalloenzyme that complexes 1 mol of Zn2+ per mol of protein.
Solution of the crystal structure of a mammalian adenosine deaminase provided knowledge of the amino acids in the active site. However, at pH 4.2, where crystals were generated for x-ray analysis, adenosine deaminase is almost completely inactive, and at this pH the substrate analogue, purine riboside is only weakly bound (Wolfenden and Kati, 1991). Construction of mutations in active site residues coupled with determination of functional consequences of each mutation under conditions of optimal enzyme activity, will permit detailed characterization of the reaction pathway and description of enzyme intermediates.
In this study, guided in selection of targets by the crystal structure, we have altered key amino acid residues within the active site of human adenosine deaminase, an enzyme that is highly homologous to its murine counterpart. The recombinant enzymes were expressed in a baculovirus system and purified to homogeneity on a monoclonal antibody affinity column. Kinetic characteristics, stabilities, and metal binding capacities of the altered enzymes were assessed and correlated with mechanistic models.

EXPERIMENTAL PROCEDURES
Materials-Oligodeoxynucleotide primers used in constructing mutations and sequencing were synthesized at the University of Ken-

5464
This is an Open Access article under the CC BY license.  36S]dATP and [14C]adenosine were Du Pont-New England Nuclear products. The polyclonal antibody used in these experiments was raised in rabbits in our laboratory against homogeneous human adenosine deaminase. The anti-adenosine deaminase monoclonal antibody (NlD1) used in the study was also generated in our laboratory and propagated in ascites fluid (Philips et al., 1987). All other reagents were of the highest commercial grade available.
Bacterial Strains and Vectors-The Escherichia coli strains used for plasmid propagation were CJ236 and DH5a (Bethesda Research Laboratories). The plasmid vector M13 mp18 was used for sitedirected mutagenesis of the adenosine deaminase cDNA. The baculovirus transfer vector pAcC4, a generous gift from Cetus, was used in homologous recombination experiments to construct specific baculovirus variants.
Viruses, Cells, and Larvae-Autographa californica nuclear polyhedrosis virus (ACMNPV strain Li, Invitrogen Corp.) and Spodoptera frugiperda (Sf-9) insect cells (Invitrogen) used were propagated in the laboratory and used in the protein expression experiments. The second insect cell line, High-5 (Invitrogen), was derived from eggs of the cabbage looper and was an alternate host for recombinant baculovirus propagation. Both insect cell lines were cultured in Grace's media containing 10% fetal calf serum. Trichoplusia ni larvae were produced in this laboratory by methods previously described (Medin et al., 1990(Medin et al., , 1992. Site-directed Mutagenesis-To facilitate site-directed mutagenesis of selected regions of the enzyme, the full length adenosine deaminase cDNA was cloned into the vector M13 mp18. The cDNA was released from the vector pBR8 (Medin et al., 1990) with the restriction enzymes NcoI and HinfI. Phosphorylated linkers that convert the HinfI site to a NcoI site were ligated to the HinfI -end of the cDNA. This adenosine deaminase cDNA was isolated after subcloning into the NcoI site of pAcC4 by digestion with NcoI. Phosphorylated NcoI to EcoRI linkers were ligated to the adenosine deaminase cDNA, and this cDNA was digested with EcoRI and ligated into the dephosphorylated M13 mp18 at the EcoRI site. This construct (M13 adenosine deaminase) was used in the development of all site-specific mutant forms of adenosine deaminase. Site-directed mutagenesis was carried out using the method of Kunkel et al. (1987). M13 adenosine deaminase was transformed into E. coli strain CJ 236 (dut-, ung-). Singlestranded uracil containing M13 adenosine deaminase was isolated and annealed to synthetic phosphorylated mutagenic oligomers ( Table I). The second strand of DNA was synthesized using Sequenase version I1 and dNTPs. After ligation the reaction product containing the newly synthesized second strand was transformed into (dut', ung+) E. coli strain DH5a. The presence of the desired mutation was confirmed by DNA sequencing using dideoxynucleotides.
Expression of Recombinant Wild-type and Mutant Human Adenosine Deaminase in Baculovirus-infected Larvae and Insect Cells-The mutated cDNA constructs were released from M13 mp18 by digestion with NcoI and subcloned into pAcC4 at the NcoI site. In order to produce the desired baculovirus strains, Sf-9 insect cells were cotransfected with extracellular virus (ACMNPV) DNA and pAcC4 containing adenosine deaminase. Recombinant viruses were identified, isolated, and purified as in our previous study (Medin et al., 1990).
The wild-type adenosine deaminase and the mutants H214A, H214N, E217A, H238A, and C262A were overexpressed in the 2' . ni larvae (Medin et al., 1990). Briefly, 10 p1 of each viral stock (-lo9 plaque-forming units) was injected into the larvae at the fourth instar stage. The infection was allowed to continue for 4 days after which the larvae were collected and frozen immediately at -70 "C. The mutants H17A and H214L were overexpressed in Sf-9 and High-5 cells. For cellular infection, 2.5 X 107cells in T-175 flasks were infected at a multiplicity of infection of 10 with the appropriate virus stock. Cells were harvested 65-h postinfection, washed twice with cold phosphate-buffered saline, and frozen at -70 "C. Purification of Recombinant Proteins-Wild-type adenosine deaminase was purified from frozen larvae by adenosine-affinity chromatography (Medin et al., 1990). Briefly, frozen larvae (28 g) containing the recombinant protein were homogenized in a buffer (10 mM sodium acetate, pH 6.4, 2 mM EGTA, 5 mM benzamidine, 10 mM 6aminocaproic acid, 5 mM phenylmethylsulfonyl fluoride), and centrifuged at 30,000 X g for 30 min. Protamine sulfate was added to the crude extract and allowed to precipitate. The clarified supernatant obtained after centrifugation was loaded onto a DEAE-Sephadex column (25 X 5 cm). The protein was eluted with the same buffer containing 0.5 M sodium chloride and concentrated by ammonium sulfate precipitation. The precipitate, resuspended in a minimum volume (-8 ml) of phosphate-buffered saline, pH 7.4, was applied to an adenosine-Sepharose column (110 X 1.5 cm) (Schrader and Stacy, 1977), and the fractions containing the major adenosine deaminase activity were pooled and concentrated by ultrafiltration.
Mutant proteins were isolated by using a monoclonal antibody affinity column. The matrix was constructed by cross-linking the antibody (NlD1, the isotype with the highest affinity for human adenosine deaminase) to protein A-Sepharose (PAS-NlD1). The immunoaffinity column was made as described by Philips et al. (1987). The isolation of the proteins on the monoclonal antibody column was essentially the same as previously described (Philips et al., 1987) with modifications in the initial extraction procedure. Frozen larvae (3-10 g) were suspended in three volumes of cold extraction buffer containing several protease inhibitors (50 mM potassium phosphate, pH 6.8, 10 mM 6-aminocaproic acid, 5 mM benzamidine, 2 mM EGTA, 20 p~ leupeptin, 8 p~ pepstatin, and 1 mM phenylmethylsulfonyl fluoride) and homogenized on ice with a Tekmar tissue homogenizer. The homogenate was centrifuged at 30,000 X g for 30 min. The pellet was re-extracted with half the original volume of buffer, homogenized, and centrifuged as before. Supernatants from both extractions were mixed (crude extract) and brought to 30% ammonium sulfate saturation. The supernatant obtained after the first centrifugation was further subjected to another cycle of fractionation with ammonium sulfate (70%). The precipitate was resuspended in the original volume of buffer used and incubated with pre-equilibrated protein A-Sepharose-monoclonal antibody matrix overnight on a rotating platform. The elution and concentration of the proteins (in 4-6 M urea) were done essentially as described before. The entire operation was carried out at 4 "C.
For extraction of recombinant proteins from cells, frozen cells were suspended in four volumes of extraction buffer (see above, 6-aminocaproic acid was omitted and aprotinin was added to a final concentration of 5 pglml) and sonicated on ice (two 30-s bursts). The sonicate was centrifuged at 30,000 X g for 30 min. The pellet was resuspended in half the volume of original buffer, homogenized, and centrifuged again. Both the supernatants were mixed with the protein A-Sepharose antibody column matrix. The rest of the procedure was identical to that described above.
Protein Determination-Protein concentrations were determined by the Coomassie Blue dye binding method (Bradford, 1976) using bovine serum albumin as the standard and reagents from Bio-Rad. For zinc determination experiments, protein concentrations were measured by amino acid analysis at the University of Kentucky,

SDS-Polyacrylamide Gel Electrophoresis and Western
Blotting-SDS-polyacrylamide gel electrophoresis was performed according to the method of Laemmli (1970) in a mini-gel apparatus (Bio-Rad). To visualize proteins the gels were either stained with Coomassie Brilliant Blue or with silver staining procedures (Wray et al., 1981). Adenosine deaminase in the crude extracts was determined by Western blot analysis. Electrotransfer of proteins onto nitrocellulose sheets was performed according to Burnette et al. (1981). Blots of the protein were analyzed with anti-human adenosine deaminase polyclonal antibody (1:500, dilution) or anti-human adenosine deaminase monoclonal antibody (1:250, dilution). The second antibody used in the procedure was goat anti-rabbit or anti-mouse IgG alkaline phosphatase conjugate (Bio-Rad). The immunoreactive proteins were detected with an alkaline phosphatase conjugate substrate kit from Bio-Rad.
Enzyme Assay and Kinetic Studies-Adenosine deaminase activity was detected by using the radiochemical assay described before (Coleman and Hutton, 1975) with ["Cladenosine as the substrate. One unit of adenosine deaminase activity is defined as the amount of enzyme required to produce 1 pmol of inosine/min at 37 "C. For each protein, kinetic measurements were done at a series of six concentrations of adenosine so as to bracket the expected K,,, value. The concentration of enzyme ranged from 0.75 nM for the wild type to 1 pM for the least active mutant proteins. Kinetic constants were determined by an enzyme kinetics program (Trinity software, Campton, NH). The nonlinear regression analyses from this program are reported in this paper.
Thermolysin Digestion of Wild Type and Mutant Adenosine Deaminase-The structural integrity of each mutant protein was studied by performing a thermolysin digestion at a variety of temperatures. The method used by Polesky et al. (1990) was followed with slight modifications. The proteins were stable to thermolysin at an enzyme:protein ratio of 1:50. Therefore an enzyme:protein ratio of 1:lO was used. Digestions were carried out for 45 min and were terminated by the addition of protein gel sample buffer. The digestion products were electrophoresed on a 12.5% polyacrylamide gel and visualized by staining with Coomassie Brilliant Blue.
Zinc Content of Wild-type and Mutant Adenosine Deaminase-T w o methods, flame atomic absorption spectrometry and graphite furnace atomic absorption spectrometry, were used for zinc analysis. Absorbance peaks by both methods were measured at 213.9 nm. All values shown are averages of duplicate determinations. The detection limits for zinc by the above two methods were 120 and 1.5 nM, respectively.
For zinc analysis the protein samples were extensively dialyzed against 5 mM HEPES buffer (passed through a Chelex-100 column) in metal-free dialysis tubing prepared as suggested by Auld (1988). The concentration of residual zinc in the buffer was subtracted from those in the unknown protein samples.

RESULTS
The goal of this study was to produce recombinant human adenosine deaminase proteins in which amino acid residues that have been implicated in or are close to the enzyme active site (Wilson et al., 1991) have been mutationally altered. 5 amino acid residues were targeted for substitution, Hisz3*, Glu217, His214, His17, and CysZ6' (Fig. 1). Each of these was converted to Ala except the residue which was changed to Asn, Leu, and Ala. The first 4 residues in this group are predicted to be involved in ligand binding or catalysis. The CysZ6' is close to the active site pocket, but is blocked by a wall of residues (Hisz3', SeP5 and Aspzg5).
We have previously demonstrated that wild-type human adenosine deaminase is produced in very high amounts in T. ni larvae and can be readily purified from larvae infected with recombinant baculovirus containing the adenosine deaminase open reading frame under the control of the polyhedrin promoter (Medin et al., 1990). Therefore, we selected the baculovirus system for production of the seven mutant proteins.
Expression and Immunodetection of Recombinant Proteins-Each adenosine deaminase mutant was generated by subcloning the cDNA into M13 and using the mutagenesis technique pioneered by Kunkel et al. (1987). Recombinant plasmids were isolated, identified by restriction analysis, confirmed by sequence analysis, and subcloned into the transfer vector.
Each of the transfer vector DNAs containing the adenosine deaminase open reading frame was then used in homologous recombination with wild-type baculovirus DNA to generate recombinant baculovirus for each mutation. The relative levels of wild-type and mutant adenosine deaminases produced in infected larvae were assessed by immunoblot analyses of crude extracts. No immunoreactive protein was detected in crude extract of insect larvae infected with nonrecombinant virus (Fig. 2, lane 9).
Immunoreactive protein with a mobility equal to that observed in a preparation of purified wild-type adenosine deaminase (lane 10) was observed in clarified extracts containing mutants H238A, E217A, C262A, H214N, H214A, H214L, and H17A and a small amount of degraded adenosine deaminase was observed in lanes 1-6 (lower band). However, the amounts of antigen present in equivalent aliquots of the extracts varied dramatically. The mutant protein that contained Ala in place of CysZ6* (lane 4 ) and Ala in 2 residues that are predicted to interact with the substrate, H238A and E217A, either a t position 1 of the heterocyclic ring (lane 3 ) or with the incoming hydroxyl residue (lane 2), were produced in quantities equivalent to the wild-type recombinant protein (lune 1).
Proteins containing mutations in the 2 residues that are thought to be coordinated to Zn2+ (H17A and H214A, H214N  and H214L), were present a t lower levels (lanes 5-8). In fact, immunodetection of antigen in the crude extracts containing H214L and H17A (lanes 7 and 8 ) was difficult, indicating that these proteins accumulated to much lower levels than wild-type adenosine deaminase in infected larvae. Therefore, expression of these two mutants in infected Sf9 and T. ni  insect cell cultures was compared. Accumulation of both mutant proteins in these cell lines was equally improved (Fig. 2, lunes 12 and 13), though it was still much lower than for wild-type or any of the other mutant constructs (data not shown). The reason for the low accumulation of these mutant proteins has not been resolved.
The recombinant adenosine deaminase produced in this system was not totally soluble. The pellet fractions obtained from the initial clarification of the crude extracts from wildtype and mutant proteins contained large quantities of immunoreactive antigen (data not shown). Attempts to solubilize recombinant adenosine deaminase from the pellets (using nonionic detergents, urea, or high salt) yielded only an additional 10-15% of the antigen (data not shown). The two mutants (H214L and H17A) that were present in the extracts in low quantities exhibited correspondingly smaller amounts of antigen in the pellet fractions, suggesting that solubility of these proteins was about the same as that of the wild-type adenosine deaminase.
Purification of Wild-type and Mutant Adenosine Deaminase-The wild-type adenosine deaminase was purified using a two-column procedure (including an adenosine-Sepharose column) that we have previously described (Medin et al., 1990). In the present studies the procedure was scaled to about 30 g of larvae from which 25-30 mg of recombinant protein was purified, corresponding to a yield of 25%. The purified enzyme is shown in Fig. 3 (lane 1 ). This purification procedure was next applied to mutants C262A, E217A, H214A, H214L, and H214A. The retardation of wild-type and all mutant proteins by the adenosine-Sepharose matrix was similar. Thus binding of substrate was not dramatically altered in any of the mutant proteins.
Unexpectedly, a low level of enzyme activity was detected in all mutant preparations except C262A, which exhibited high levels of adenosine deaminase activity. Therefore, we reexamined the uninfected larvae and larvae infected with baculovirus (that did not contain any adenosine deaminase coding sequences) for a deaminase-like activity. Both sets of larvae were subjected to the purification procedure. Indeed, a deaminating activity (1-5 units/mg) was detected in the eluant from the adenosine-Sepharose column from uninfected larval extract. However when infected larval extract was subjected to the purification protocol, the deaminating activity dropped to 0.1-0.5 units per mg. Undoubtedly, viral replication repressed host protein synthesis (O'Reilly et al., 1992). This remaining deaminase-like activity was insignificant when compared to wild-type adenosine deaminase (400 units/ mg), but might have interfered with detection of authentic activity in some of the mutant enzymes. In order to eliminate the insect deaminating activity from the mutant protein preparations, an alternate procedure was introduced for purification of all mutant recombinant proteins.
Monoclonal antibody (NlD1) directed against human adenosine deaminase (Philips et al., 1987) was used to construct an immunoaffinity column. This antibody, when tested in a Western blot against deaminating activity purified from uninfected larvae, showed no detectable antigen (data not shown). In addition, no adenosine deaminase antigen or activity was recovered from uninfected larval extracts when subjected to a purification procedure using the immunoaffinity column. As a further precaution against cross-contamination of the mutant proteins, individual monoclonal antibody affinity columns were constructed for each mutant and used only in the purification of that protein.
The results of the purification of the seven mutants are shown in Fig. 3. Each of these proteins has been purified 10 to 25 times. The proteins, detected by silver staining, were produced in excellent yield and purity, except for two constructs. H214L and H17A consistently gave low yields when purified from either infected larvae or cells. These two mutants have been purified 15 times from different cell preparations with similar results. The two preparations shown in the figure were from recombinant virus-infected cells and exhibited significant degradation of adenosine deaminase, as confirmed by Western blotting (data not shown). These data suggested that the very low antigen yields for these two mutants may reflect an altered protein stability.
Thermolabilities of Wild-type and Mutant Proteins-We explored thermolabilities of the proteins by comparing the patterns of products obtained from thermolysin digestion generated over a range of temperatures of all seven mutant and wild-type adenosine deaminases. The procedure was designed to detect differences in thermolability of wild-type and mutationally altered proteins. The rationale for this approach is based on differential sensitivities of denatured and native proteins and differential thermolabilities of wild-type and mutationally altered proteins. Each of the purified proteins was subjected to digestion by thermolysin a t varying temperatures. The results of this analysis for two of the mutants is shown in Fig. 4. Across the entire temperature spectrum, the wild-type protein and the two mutants, H238A and C262A, exhibited similar digestion patterns. Other mutants, H214A, H214N, and E217A (not shown in this figure) showed identical digestion patterns. The two low yield mutants, H17A and H214L, were also subjected to thermolysin digestion. However, the limited quantities of these proteins made the analysis difficult to interpret. The overall digestion patterns and temperature dependence appeared similar, but with small amounts of these two proteins available, subtle changes in the  digestion patterns would not have been observed. The thermolysin digestion experiments were also carried out by incubating the samples a t 66 "C for varying times, from 10 to 60 min (data not shown). No dramatic differences in digestion patterns were observed during this time scale.
Kinetic Parameters of Wild-type and Mutant Deriuatiues-Kinetic properties of the purified wild-type and mutant adenosine deaminases were determined using adenosine as substrate (Table 11). The apparent K,,, values for the mutants were similar to the wild-type enzyme. Thus, alterations in neither the side chains of the two-Zn2+ coordinating residues (His17 and His2I4) nor in the N1 interacting residue (Glu2") interfered with productive substrate binding. The only significant variation in K,,, @-fold) was observed for the H238A variant.
The VmaX and kc,, values were dramatically altered in the mutant proteins (Table 11). The most active of the proteins was C262A. This residue is highly conserved from humans to bacteria in all deaminases (Chang et al., 1991), and it lies in the crystal structure in the entrance to the enzyme active site (Wilson et al., 1991). When this protein was purified using the standard two-column procedure, specific activities in homogeneous preparations ranged from 120 to 160 units/mg. In contrast, when this protein was purified by using the monoclonal antibody column procedure, specific activities of the purified preparations ranged from 24 to 137 units/mg. The most common activities in these preparations in 15 separately purified samples of C262A was 50-90 units/mg. The lower activity of the proteins purified by an immunoaffinity procedure probably reflected the conditions required (4-6 M urea) to elute the pure protein from the column. Alteration of this conserved residue may impede refolding of the protein during the procedure used for removal of urea and result in a lower enzyme activity than that observed in the native enzyme. In contrast, wild-type adenosine deaminase purified by both methods exhibits identical specific activities (Philips et al. 1987).' None of the other mutant enzymes exhibited differential specific activities as a result of conventional purification or the immunoaffinity column procedure (see Footnote a in Table 11).
The second mutant adenosine deaminase that appeared to exhibit activity was H17A. Enzymatic activity varied dramatically among the 15 preparations of this mutant purified on the immunoaffinity column. The lowest activity observed was 0.2 units/mg and the highest activity was 28 units/mg. This residue and H214 are postulated to be involved in coordinating the zinc atom. Therefore we examined the possibility that variations in Zn2+ content throughout the purification procedures were related to the retention of adenosine deaminase activity in these altered proteins.
Our standard purification scheme utilized throughout potassium phosphate buffer, pH 6.8. Since phosphate is known to chelate Zn2+, we substituted for this buffer, Tris (which does not chelate Zn2+ (Sellin and Manneevvik, 1984)) and purified wild-type, H214A, H214N, H238A, and E217A in the presence or absence of exogenously added Zn2+. The results are shown in Table 111. In the presence of Zn", either by exogenous addition throughout the purification or by use of a nonchelating buffer, enzyme activity was increased by 5-to 6-fold for the H214 mutants. We confirmed that this apparent stimulation of adenosine deaminase activity was not a general phenomenon since the specific activities of wild-type, H238A, ' D. Bhaumik, J. Medin, K. Gathy, and M. S. Coleman, published observations. and E217A enzymes were identical to those obtained by using phosphate buffer in the purification procedures (Table 111).
Zn2+ Content of Wild-type and Mutant Adenosine Deaminases-The purified recombinant proteins were analyzed for Zn2+ content using two methods: graphite furnace atomic absorption spectrometry and flame atomic absorption spectrometry (Table IV). The graphite furnace method is more sensitive, but since small quantities of protein are required (nanomolar range), the signal-to-noise ratio is smaller than with the atomic absorption method. Previous analyses, including the x-ray crystallography studies, revealed a single Zn2+ residue at the active site. In these experiments, wildtype enzyme, E217A, H238A, and C262A were demonstrated to contain a single Zn2+ when the atomic absorption spectrometry method was used. The graphite furnace method indicated 2 mol of Zn2+ per mol of protein for these enzyme preparations. We suspect that this higher ratio is simply due to the inherent problem in accurately measuring Zn2+ from small quantities of protein. Both types of H17A preparations (high and low activity) contained Zn2+, though the amounts of these proteins were so limited that accurate quantitation of molar ratios was not feasible. The His214 mutants also contained Zn2+ (with the exception of H214L) even though these preparations exhibited no enzyme activity in the absence of added Zn2+, but were slightly activated (H214A and H214N) when Zn2+ was added throughout the enzyme purification.

DISCUSSION
Adenosine deaminase is an enzyme of enormous interest. It plays important roles in nutrition and reproduction, and it serves a protective role against accumulation of deoxyadenosine in cells of the immune system. The catalytic power of this enzyme is among the highest known and extensive chemical studies with inhibitors have generated testable hypotheses about the active site. As expected for a protein with a central role in metabolism, the primary amino acid sequence of adenosine deaminase is highly conserved across species (Chang et al., 1991). The crystal structure of murine adenosine deaminase, complexed to the inhibitor purine riboside at pH 4.2, has recently been elucidated (Wilson et al., 1991), and active site residues have been identified. This information, coupled with access to the coding sequences of adenosine deaminases from other species, has made the testing of amino acid side chains in the active site feasible.
Human adenosine deaminase is slightly longer than the FAAS, flame atomic absorption spectrometry. Standards containing zinc ranged from 0.385 to 6.16 p~. Measured values ranged from 1.16 to 3.08 pM.
H17A, Zn2+ was detected in this preparation (designated *), but the small quantity of protein did not permit accurate quantitation. ND, not determined. murine enzyme (by 11 residues at the carboxyl end of the molecule) but shares greater than 80% identical side chains (Wiginton et al., 1984;Ingolia et al., 1985). Of the 59 amino acid differences between the two proteins, 26 are conservative substitutions. All of the residues implicated in binding substrate or catalyzing the hydrolytic reaction are identical in the human and mouse proteins (Chang et al., 1991). Therefore, we anticipated that residues identified in the mouse enzyme would serve identical functions in both enzymes. The major goal of this study was to test selected amino acid residues in the active site at a pH of optimal enzyme activity (6.8). Of particular interest were a Cys residue close to the active site, 3 His, and a Glu residue in the active site.
Chemical studies with adenosine deaminase from a number of sources have previously implicated a sulfhydryl group as a catalytically important residue, perhaps as the source for protonation of N1 of adenosine (Orsi et al., 1972;Wolfenden et al., 1967;Weiss et al., 1987). Furthermore, comparison of adenosine deaminase protein sequences from Escherichia coli to humans has demonstrated the presence of a conserved Cys (Chang et al., 1991) which is also positionally conserved in all known AMP deaminases. Thus, it seemed highly likely that this residue would be essential for enzyme catalysis or conformation.
In the mouse and human enzymes this conserved residue is at position 262. The interpretation of the crystal structure places this Cys out of the range of the active site and suggests that access by this sulfhydryl group to the active site is blocked by 3 residues involved in substrate binding or catalysis (Wilson et al., 1991). When this amino acid was converted to Ala as described in this study, there was only a modest effect on enzyme activity or kinetic parameters. The mutant protein was apparently correctly folded in uiuo, since enzyme which retained 30-40% of wild-type activity was obtained following conventional purification. When the protein was exposed to a denaturant during purification on an immunoaffinity column, about 20% of native enzyme activity was normally recovered under urea removal conditions in which 100% of wild-type adenosine deaminase activity and other mutant enzyme activities are recovered. The thermolability studies of the C262A mutant showed that there was no evidence of a drastically altered proteolytic digestion pattern upon heating. Thus, while this CysZ6' residue may play an important, but as yet poorly understood, structural role in the enzyme structure, it is clearly not essential for catalysis.
The most surprising revelation about adenosine deaminase which occurred with the solution of the crystal structure was the role of Zn2+ in the enzyme active site (Wilson et al., 1991). This feature of the enzyme was unanticipated since standard studies with metal chelators failed to detect any enzyme sensitivity to such agents (Zielke and Suelter, 1971). The central role of Zn2+ in the functioning of adenosine deaminase helps to explain in part earlier observations of the effects on the immune system produced by Zn2+ deficiency in humans and animals. Zinc deficiency causes atrophy of lymphoid tissues and abnormalities in T and B cell function (Cossack and Prasad, 1991) reminiscent of the severe combined immunodeficiency observed in children who lack adenosine deaminase (Coleman, et al., 1978).
The human recombinant adenosine deaminase also contained Zn2+ as shown by analytical techniques. 2 residues were altered that have been implicated in Zn2+ binding, Hid7 and His214. The H17A was an unstable recombinant protein that exhibited an unaltered K , for substrate binding. The His214 mutants varied in stability. Substitution of His with Ala or Asn produced mutants that could be obtained in reasonable yield. The Leu substitution, however, produced a protein that was as unstable as the H17A. The concentration of Zn2+ available at all stages of the expression and purification of the recombinant proteins appeared to be important for recovery of enzyme activity. Use of a non-Zn2+-chelating buffer and addition of Zn2+ throughout the purification usually resulted in recovery of a small percentage of wild-type enzyme activity (0.05-0.12%). However, the addition of Zn2+ during prolonged dialysis of purified enzyme preparations exhibiting less than 0.02% activity was not effective in restoring any additional activity. These mutants were not devoid of Zn2+ as shown by the atomic absorption analyses. If Zn2+ was dislodged from the active site during purification by the use of phosphate buffer subsequent recovery of activity was never observed. These studies support the notion that Hid7 and His214 coordinate ZnZ+.
GluZi7 is proposed, on the basis of its proximity to the N1 of purine riboside in the crystal structure, to be involved in protonation of N1 of the adenosine purine ring (Wilson et al., 1991). The pK. of the side chain carboxylate of glutamate is 4.2. At the pH of maximal enzyme activity, 7, this residue would be negatively charged, unless its pK, has been altered by the environment of the active site. Removal of the carboxylate residue at this position caused a dramatic reduction in catalytic activity (10,000-fold) but no significant change in the K,,, for adenosine. Thus, the putative hydrogen bond of the potentially dissociable proton of G1u217 to the lone electron pair of N1, as proposed by Wilson et al. (1991), is probably not important in substrate binding but is important for catalysis as indicated by the large decrease in kcht. The function of G~u~~~ can be compensated by water in the absence of the carboxylate side chain, since activity is not completely eliminated in the altered protein.
Hisz3' and Aspzg5 have been proposed as potential proton donors for the ammonia leaving group at C6 of adenosine, based on distance calculations (Kati and Wolfenden, 1989;Wilson et al., 1991). From analysis of residue distances in the crystal structure (crystals were grown at pH 4.2), AspZg5 is proposed to be protonated during the course of the reaction via the abstraction of a proton from water. In an alternate proposal, Wolfenden and Kati (1991) have suggested that Hisz3' abstracts a proton from an attacking water molecule and thus is the source of the proton for the ammonia. Only one of these residues was altered in this study. When Hisz3' was changed to Alaz3', a protein was produced that exhibited drastically reduced enzyme activity (3000-fold decline) and a decreased K,,, for the adenosine substrate. These data indicated that Hisz3* did potentiate binding of adenosine. When the side chain was more hydrophobic, as in Alaz3', the substrate was more tightly bound than in the wild-type enzyme. Likewise, the decrease in kcat from 320 to 0.1 s-' indicated that Hisz3' may indeed be the source of the proton in the reaction, a fact that can be established only when we have a mutation of the Aspzg5 since activity was not completely abolished in this H238A mutant.
The decrease in both K,,, and kc,, observed in H238A at first suggested that this residue may interact with the substrate transition state more strongly than with the ground state. However, inspection of k-, and free energy change (6G k l ) values for all the mutant enzymes revealed a range from 13.7 kcal for H17A to 14.7 kcal for H238A. These relatively small differences among all of the mutants do not support a greater interaction for H238A in the transition state than any of the other residues that were altered. To achieve a significant free energy change in this reaction, the observed K,,, would be expected to change by at least 10-fold with a concurrent decrease in kc,, compared to the wild-type enzyme. Studies are currently underway to analyze the contribution of Hisz3' to binding the 60H group in the transition state.
The results presented herein from site-directed mutagenesis, when combined with previous information derived from NMR and x-ray crystallography, illustrate that in adenosine deaminase an SH group (CysZ6') is important for protein folding, but not enzyme catalysis; that GluZl7 and Hisz3' are both essential for catalysis; that correct orientation of Zn2+ is crucial for activity. The availability of specifically targeted active site mutants of adenosine deaminase will now permit the determination of detailed interactions of each of these amino acid residues with the substrate and with potent inhibitors of this important enzyme.