Topological mapping of the cysteine residues of N-carbamyl-D-amino-acid amidohydrolase and their role in enzymatic activity.

The N-carbamyl-D-amino-acid amidohydrolase from Agrobacterium radiobacter NRRL B11291, the enzyme used for the industrial production Of D-amino acids, was cloned, sequenced, and expressed in Escherichia coli. The protein, a dimer constituted by two identical subunits of 34,000 Da with five cysteines each, was susceptible to aggregation under oxidizing conditions and highly sensitive to hydrogen peroxide. To investigate the role of the cysteines in enzyme stability and activity, mutant proteins were constructed by site-directed mutagenesis in which the five residues were substituted by either Ala or Ser. Only the mutant carrying the Cys172 substitution was catalytically inactive, and the other mutants maintained the same specific activity as the wild type enzyme. The crucial role of Cys172 in enzymatic activity was also confirmed by chemical derivatization of the protein with iodoacetate. Furthermore, chemical derivatizations using both acrylamide and Ellman's reagent revealed that (i) none of the five cysteines is engaged in disulfide bridges, (ii) Cys172 is easily accessible to the solvent, (iii) Cys193 and Cys250 appear to be buried in the protein core, and (iv) Cys243 and Cys279 seem to be located within or in proximity of external loops and are derivatized under mild denaturing conditions. These data are discussed in light of the possible mechanisms of enzyme inactivation and catalytic reaction.

Optically active D-amino acids have attained a wide variety of commercial applications as intermediates for the production of fine chemicals, including ␤-lactam antibiotics, peptide hormones, and pesticides (1,2). In particular, D-phenylglycine and D-p-hydroxyphenylglycine are among the most important chiral building blocks for the production of semisynthetic penicillins and cephalosporins such as ampicillin and amoxicillin.
Several optically active D-amino acids are currently produced in a two-step reaction process starting from D,L-5 monosubstituted hydantoins that are inexpensively synthesized from the corresponding aldehydes (3). In the first step, the substrate is hydrolyzed by a D-specific hydantoinase to give a D-carbamyl derivative. Subsequently, the carbamyl derivative is converted to the corresponding D-amino acid either by chemical methods (4) or by a second enzymatic step catalyzed by an N-carbamyl-D-amino-acid amidohydrolase (hereinafter carbamylase) (5). Because chemical methods have high reaction temperatures, low yields, long reaction times, and generate large amounts of waste, the enzymatic hydrolysis of the N-carbamyl derivatives is highly preferred. Indeed, the use of the D-hydantoinase plus carbamylase two-enzyme system is considered one of the most successful industrial applications of enzyme technology.
Several microorganisms expressing both enzymatic activities have been isolated, and the optimal reaction conditions and the biochemical properties of the two enzymes have been studied in some detail (6,7). From what has been published so far, it appears that both the activity and stability of the carbamylase are negatively affected by oxidizing conditions, suggesting that one or more cysteine residues are present in the enzyme. Indeed, the sensitivity of the enzyme to oxidizing conditions is one of the most serious drawbacks of the enzymatic D-amino acid production process, and a strict anaerobic regime is required to allow the completion of the substrate to product conversion (8).
To shed light on the role of the cysteines in the activity and stability of this important industrial enzyme, we decided to study in detail the N-carbamyl-D-amino-acid amidohydrolase from Agrobacterium radiobacter NRRL B11291, a strain that is used industrially for D-amino acid production. In this paper we describe the characterization of the recombinant A. radiobacter N-carbamyl-D-amino-acid amidohydrolase expressed in Escherichia coli. In particular, the role of the five cysteines in the activity of the enzyme has been thoroughly investigated using both chemical and genetic methods. In addition, a topological mapping has been undertaken using a chemical approach in which the availability of the cysteine residues to derivatizing agents has been assessed in native and denaturated enzyme.

MATERIALS AND METHODS
Plasmids and Strains-E. coli 71/18(pSM214) was used for the expression of the recombinant N-carbamyl-D-amino-acid amidohydrolase. pSM214 is an E. coli-Bacillus subtilis shuttle expression vector constructed in our laboratory (9). Phage M13mp8 and E. coli TG1 were used for site-directed mutagenesis. E. coli was transformed using either the CaCl 2 method (10) or electroporation (11).
Enzymes and Reagents-Restriction and modification enzymes were purchased from Boehringer Mannheim and from New England Biolabs and used according to the manufacturers' specifications. Oligodeoxynucleotides were synthesized using a Beckman P1000 DNA synthesizer. Sequencing was performed manually on both strands using the Sequenase sequencing kit (U. S. Biochemical Corp.) and [␣-35 S]dATP from DuPont NEN. Chromatographic material was from Pharmacia Biotech Inc. Guanidium hydrochloride, ␤-mercaptoethanol, dithiothreitol, acrylamide, iodoacetate, and 5,5Ј-dithiobis-2-nitrobenzoic acid (Ellman's reagent) were from Fluka, and N-carbamyl-D-p-hydroxyphenyl glycine was prepared from the corresponding amino acid and potassium cyanate according to the procedure described by Stark and Smyth (12).

TCGTTGGAGATGAACAT-3Ј.
For all the mutagenesis reactions, the wild type gene was isolated from plasmid pSM637 as an EcoRI-HindIII fragment and cloned in M13mp8. The mutated EcoRI-HindIII fragment was used to replace the wild type gene of pSM637, and the presence of the mutations on the derived plasmids was confirmed by sequence analysis. The mutated plasmids were used to transform E. coli 71/18, and the positive transformants were grown on LB medium supplemented with 20 g/ml chloramphenicol for the production of the mutant proteins.
Protein Analysis of Cellular Extracts-Cells were grown overnight at 37°C in LB medium supplemented with 20 g/ml chloramphenicol and collected by centrifugation. For analytical purposes, cells from 1-ml cultures were suspended in 300 l of 20 mM sodium phosphate, pH 7, 20% glycerol and sonicated to homogeneity. After sonication the suspension was centrifuged in an Eppendorf centrifuge for 10 min at 4°C, and 15-20 l of the supernatant were analyzed by SDS-polyacrylamide gel electrophoresis (PAGE) 1 (13). Total protein concentration in the crude extracts and in the purified enzyme preparations was evaluated using the Bradford procedure (14).
Enzyme Assay-N-Carbamyl-D-amino-acid amidohydrolase activity was measured using 500 l of 200 mM Na-PO 4 buffer, pH 7, containing 120 mM N-carbamyl-D-p-hydroxyphenyl glycine. The reaction was carried out at 40°C, and the NH 4 ϩ formed was determined using the Berthelot photocolorimetric method as modified by Weatherburn (15).
Purification of N-Carbamyl-D-amino-acid Amidohydrolase-E. coli cells from 1-liter cultures were resuspended in 20 ml of 25 mM Tris-HCl, pH 7.0, containing 20% (v/v) of glycerol (Buffer A) and lysed by two consecutive passages in a French Press apparatus set at 18,000 p.s.i. The soluble proteins were collected by centrifugation (30,000 ϫ g for 20 min) and loaded onto a Q-Sepharose FF column equilibrated with the same buffer. The column was first washed with Buffer A and then eluted with a linear gradient of 0 -0.4 M NaCl. The enzyme-containing fractions identified by SDS-PAGE analysis were pooled and loaded onto a nickel-activated chelating Sepharose FF column equilibrated in Buffer A. The N-carbamyl-D-amino-acid amidohydrolase activity was then eluted from the column with a linear gradient of 0 -0.2 M imidazole HCl, pH 7.0, in Buffer A. The fractions containing the enzyme that eluted at about 0.1 M imidazole were pooled and concentrated using an ultrafiltration device equipped with a YM-10 membrane (Amicon). The proteins were finally loaded onto a Superose 12 gel filtration column equilibrated in 25 mM sodium phosphate buffer, pH 7.0, containing 100 mM NaCl and 20% (v/v) glycerol. All the purification steps were carried out at 4°C, and the buffers were extensively degassed with helium before use.
Protein Derivatization with Acrylamide-Protein derivatization with acrylamide was performed according to Brune (16). The alkylated enzyme was purified on a Phenyl 5PW column (TOSOH Corp., 4 ϫ 75 mm) following the same HPLC method described for separation of fragments (see below). Under these conditions the alkylated protein has a slightly lower retention time than the underivatized enzyme.
Protein Derivatization with Disulfide Reagents-The purified protein (1.3 mg) dissolved in 0.1 M imidazole HCl buffer, pH 7.0, containing 20% glycerol was treated with either Ellman's reagent (1:1 w/w) or oxidized glutathione (1:1 w/w) and incubated at room temperature for 30 min. Excess reagent was removed using a PD-10 gel filtration column.
Sequential Protein Derivatization with Iodoacetate and Acrylamide-The purified protein (1.3 mg) dissolved in 0.1 M imidazole HCl buffer, pH 8.0, containing 20% glycerol was treated with 200 mM iodoacetate and incubated at room temperature for 20 min. Excess reagent was removed using a PD-10 gel filtration column, and the derivatized protein collected in 3 ml. The protein was then denatured by the addition of guanidinium HCl (3 g) and, after adjusting the pH to 8.0 with diisopropylethylamine, alkylated with a 6 M aqueous solution of acrylamide (0.6 ml). The reaction was carried out at room temperature for 14 h. Excess reagent was again removed using a PD-10 gel filtration column.
CNBr Cleavage-Cleavage of the enzyme was carried out either in 1 M HCl containing 6 M guanidinium hydrochloride or in 70% trifluoroacetic acid using 100 mg/ml of CNBr. The reaction mixtures were incubated for 16 h at room temperature in the dark under helium atmosphere and deluted with water, and the excess CNBr was removed by lyophylization. The peptide fragments were purified on reverse phase HPLC C8 columns (4 ϫ 125/250 mm, Lichrospher 100 RP-8, Hewlett Packard) using a Beckman System Gold equipped with a diode array detector. The separation was achieved at a flow rate of 1 ml/min, applying a linear gradient of 10 -100% Buffer B in Buffer A (Buffer A: water containing 0.1% trifluoroacetic acid; Buffer B: 90% acetonitrile/ 10% water/0.075% trifluoroacetic acid).
N-terminal Protein Sequencing-Automated Edman degradation of peptides from CNBr cleavage was performed on a Beckman LF 3000 Protein Sequencer by means of the standard Porton sequencing routine 40 and on-line reverse phase HPLC analysis of the PTH amino acids. PTH S-carboxymethylcysteine derived from the iodoacetate reaction eluted just before PTH serine, whereas PTH Cys-S-␤-propionamide (PTH Cys␤-Pam) obtained upon acrylamide treatment eluted between PTH Glu and PTH His. Finally, the dithiothreitol adduct of dehydroalanine that is obtained during the Edman degradation reaction from cysteine-Ellman's reagent mixed disulfides (17) eluted between PTH Ala and PTH Tyr.
H 2 O 2 Inactivation of N-Carbamyl-D-amino-acid Amidohydrolase and Catalase Protection Assay-N-Carbamyl-D-amino-acid amidohydrolase was exposed to increasing concentrations of H 2 O 2 (0 -0.2 mM) at room temperature. After a 15-min incubation, the residual activity was determined under the standard assay conditions. Catalase protection experiments were carried out by incubating the enzyme in the presence of increasing concentrations of catalase (0 -5 units) at a H 2 O 2 concentration of 0.1 mM. One catalase unit is defined as the amount of enzyme that decomposes 1 mol of H 2 O 2 in 1 min at 37°C.

Characterization of the N-Carbamyl-D-amino-acid Amidohydrolase Gene and Its Expression in E.
coli-The sequence of the N-carbamyl-D-amino-acid amidohydrolase gene, whose isolation will be described elsewhere, 2 is shown in Fig. 1. The gene encodes a protein of 304 amino acids with a calculated molecular weight of 34,000 and pI value of 5.99. From the sequence analysis, five cysteines are present at positions 172, 193, 243, 250, and 279.
The gene coding for the N-carbamyl-D-amino-acid amidohydrolase was inserted into plasmid pSM214 to give plasmid pSM637 (Fig. 2), which was used for its heterologous expression in E. coli. The enzyme was expressed under the control of a constitutive promoter at a level that on the basis of SDS-PAGE analysis was estimated to be more than 5% of the total soluble proteins (data not shown).
Properties of N-Carbamyl-D-amino-acid Amidohydrolase Purified from E. coli- Table I summarizes the results of the procedure developed to purify the recombinant N-carbamyl-Damino-acid amidohydrolase. The procedure, which allowed a 20-fold purification with a final yield of 34%, gave an enzyme more than 95% pure as judged by SDS-PAGE and HPLC analysis. Under the standard assay conditions, the specific activity of the purified enzyme was 10 mol min Ϫ1 mg Ϫ1 .
On SDS-PAGE the molecular mass of the enzyme was estimated to be about 32,000 Da, in good agreement with the theoretical calculation from the sequence analysis. However, on Superose 12 HR 10/30 the protein eluted with an apparent molecular mass of about 67,000 Da, suggesting that the active form of the enzyme is a homodimer. The recombinant enzyme co-eluted with bovine serum albumine (67,000 Da) even in the presence of 10 mM ␤-mercaptoethanol, indicating that the two subunits of the homodimeric structure are not covalently associated (data not shown).
Under prolonged storage at 4°C in phosphate buffers, the enzyme progressively lost activity that could be recovered only partially by treatment with reducing agents such as ␤-mercaptoethanol (data not shown). The loss of activity paralleled the appearance of insoluble inactive aggregates in the enzyme preparation, suggesting that covalent intermolecular reactions took place under the storage conditions used. Nevertheless the purified enzyme was more stable in the absence rather than in the presence of ␤-mercaptoethanol. This can be explained by the fact that in the presence of air thiol groups have a propensity to generate hydrogen peroxide and superoxide anions, which in turn can lead to irreversible enzyme damage (18, 19).
To test the sensitivity of the enzyme to hydrogen peroxide, the activity was measured after incubation either in the presence or in the absence of H 2 O 2 . As shown in Fig. 3A, the enzyme was fully inactivated after 15 min of incubation with 0.1 mM H 2 O 2 . The enzyme was protected from H 2 O 2 inactivation when a sufficient amount of catalase was included in the reaction mixture before H 2 O 2 addition (Fig. 3B). The sensitivity of the enzyme to H 2 O 2 suggested that at least one cysteine residue takes part in the catalytic reaction.
To experimentally test whether one of the five cysteines present in the carbamylase from A. radiobacter NRRL B11291 is involved in catalysis, we expressed in E. coli five mutants of the enzyme in which each of the cysteine residues was replaced with alanine. Substitution of Cys 172 by alanine resulted in a completely inactive enzyme under the standard assay conditions, whereas substitution of any of the other cysteines had no effect on enzyme activity. Substitution of Cys 172 with serine also resulted in complete inactivation of the enzyme. A number of double mutants were also generated (Cys 243 /Cys 250 , Cys 250 /Cys 279 , Cys 193 /Cys 279 , Cys 172 /Cys 43 , and Cys 172 /Cys 250 ). Taken together, these data clearly indicated that only Cys 172 is strictly required for enzymatic activity.
Identification by Chemical Methods of the Cysteine Residue Involved in Catalysis-When the carbamylase from A. radiobacter NRRL B11291 is hydrolyzed by CNBr, a number of fragments are generated that can be resolved on reverse phase HPLC and identified by sequence analysis (Fig. 4). If this procedure is carried out after full denaturation of the enzyme in 6 M guanidinium hydrochloride and treatment with 1 M acrylamide, it is possible to establish whether any of the cysteines of the molecule is engaged in a disulfide bond. Free cysteines are alkylated by acrylamide, and on Edman degradation the compound PTH Cys␤-Pam is easily recognized in HPLC by standard PTH separation (16). Using the approach described above, we found that all five cysteine residues were alkylated upon acrylamide treatment (Table II), indicating that their thiol groups are in the reduced state.
Treatment of the enzyme with 0.2 M iodoacetate before the denaturation and the alkylation step resulted in modification of available cysteine thiols, and the Edman degradation product of the modified cysteines (PTH S-carboxymethylcysteine) can be distinguished from PTH Cys␤-Pam. Only Cys 172 was identified as PTH S-carboxymethylcysteine (Table II), whereas all other cysteines were recovered as PTH Cys␤-Pam, indicating that only Cys 172 is exposed to the aqueous solvent. Because the iodoacetate-treated enzyme was fully inactive (data not shown), we concluded that in agreement with the site-directed mutagenesis experiments, Cys 172 must be involved in the catalytic reaction.
Topological Mapping of the Cysteines in the N-Carbamyl-Damino-acid Amidohydrolase-From the data described above, it appears that Cys 172 is exposed to the solvent and therefore easily accessible to chemical reagents able to react with thiol groups. In fact Cys 172 was the only cysteine derivatized with iodoacetate, the others being chemically modifiable only after full denaturation of the enzyme with 6 M guanidinium hydrochloride.
In an attempt to further investigate the spatial organization of the cysteine residues, native N-carbamyl-D-amino-acid amidohydrolase was treated with 1 M acrylamide, fully dena- FIG. 2. Restriction map of plasmid pSM637 used for the expression of the N-carbamyl-D-amino-acid amidohydrolase. The carbamylase gene is co-transcribed with the chloramphenicol acetyl transferase gene from pC194 using a B.subtilis-E. coli constitutive promoter (9). tured with 5 M guanidinium hydrochloride, and finally derivatized with the thiol-reacting Ellman's reagent. When the chemical modification of the five cysteines was analyzed by peptide sequencing, we found that cysteines 172, 243, and 279 had been alkylated by acrylamide (identified as PTH Cys␤-Pam), whereas Cys 193 and Cys 250 were identified as the dithiothreitol adducts of dehydroalanine (Table II), indicating that they had been modified by Ellman's reagent after full denaturation of the enzyme.
Considering that a 1 M solution of acrylamide is a weak protein denaturant, the experiments described above can be interpreted as indicating that Cys 243 and Cys 279 are exposed to the solvent by partial unfolding of the protein and are thus located within or in proximity of external loops. On the other hand, Cys 193 and Cys 250 are expected to be buried in the protein core and become accessible to the thiol-reacting compounds only after full denaturation.

DISCUSSION
The carbamyl-D-amino-acid amidohydrolases are important enzymes for the industrial production of D-amino acids, some of which are key intermediates for the synthesis of the ␤-lactam antibiotics. The relevant role of these enzymes in the pharmaceutical industry prompted us to study in more detail the biochemical properties of the carbamyl-D-amino-acid amidohydrolase from A.radiobacter, a bacterial strain that, due to its ability to synthesize large quantities of this enzyme together with hydantoinase, is currently used for the production of p-OH-phenylglycine and phenylglycine starting from the corresponding hydantoines (5,20). In particular, we focused our attention on the role of cysteine residues on the activity of the enzyme.
Sequence analysis of the carbamylase gene revealed that it encodes a protein of 32,000 Da containing five cysteines. On the basis of gel filtration and SDS-PAGE analyses, the enzyme was found to be organized in a nondisulfide bonded homodimeric structure.
The organization in more than one subunit appears to be a typical feature of amidohydrolases. For example, the N-carbamyl-D-amino-acid amidohydrolase from Comamonas sp. E222c, which has been recently characterized and whose N-terminal sequence is highly homologous to the corresponding region of the A. radiobacter enzyme, was proposed to have a trimeric structure with three identical subunits (7). Furthermore, the bacterial amidohydrolases that catalyze the hydrolysis of asparagine and glutamine are active as tetramers of identical

FIG. 3. Effect of H 2 O 2 on N-carbamyl-D-amino-acid amidohydrolase in the presence and absence of catalase.
A, the enzyme was incubated with different concentrations of hydrogen peroxide for 15 min (0 -0.2 mM) at room temperature, and the residual activity, expressed as percentage of initial activity, was determined using the standard assay conditions. B, the enzyme was incubated at room temperature for 15 min in the presence of both H 2 O 2 (0.1 mM) and catalase (0 -5 units), and the residual activity was determined.

FIG. 4. Reverse phase HPLC separation of the CNBr fragments obtained after N-carbamyl-D-amino-acid
amidohydrolase derivatization with acrylamide in denaturing conditions. The CNBr reaction was performed in 70% trifluoroacetic acid. The arrows indicate the Cys-containing fragments. In particular, the 13.5-min peak corresponds to fragment 170 -184, which contains Cys 172 , the peak that elutes at 14.3 min corresponds to fragment 185-220 containing Cys 193 , and the 25.8-min peak is the 240 -304 fragment containing Cys 243 , Cys 250 , and Cys 279 5.
protein chains with molecular weights in the range of 34,000 -36,000 per monomer (21,22). Finally, the mature form of the penicillin acylase from E. coli is a periplasmic 80,000-Da heterodimer having the A and B chains of 209 and 566 amino acids, respectively (23).
Using chemical methods we proved that the five cysteines of the enzyme are not engaged in intramolecular disulfide bonds. Sequence analysis of the CNBr-generated fragments containing the cysteine residues after treatment of the denatured enzyme with thiol-reacting acrylamide revealed that all the five cysteines were alkylated by the acrylamide. These data indicating the absence of disulfides in the carbamylase are consistent with the cytoplasmic nature of the enzyme. As a consequence of the presence of free thiols groups, the purified enzyme progressively lost activity that could be partially recovered when reducing agents such as dithiothreitol or ␤-mercaptoethanol were added to the enzyme preparation. However, somewhat surprisingly, we found that the shelf life of the enzyme was prolonged when the reducing agents were not included in the storage buffers. To reconcile these two apparently contradicting results, we hypothesized that the main enzyme-inactivating agent is hydrogen peroxide. It is known that in the presence of air the thiol groups have a propensity to generate H 2 O 2 and superoxide anions, which in turn can lead to irreversible enzyme damage (18,19).
Our experiments demonstrating the high sensitivity of the N-carbamyl-D-amino-acid amidohydrolase to H 2 O 2 supports this hypothesis. We suggest that the pathway for enzyme inactivation proceeds through various degrees of thiol oxidation, from inter-or intramolecular disulfide bonds to more oxidized products such as cysteic acid (24). The reducing agents exert two opposite effects. On one side they are beneficial, protecting the enzyme against reversible inactivation probably caused by disulfide bond formation. On the other hand, however, they are detrimental to enzyme activity because of their propensity to generate highly toxic H 2 O 2 . A direct consequence of this hypothesized mechanism for enzyme inactivation is that at least one of the cysteine residues of the enzyme takes part in the catalytic reaction.
The experiments described in this work clearly demonstrated that this is indeed the case. The Cys 172 3 Ala substitution produced a fully inactive enzyme, at least as judged by our standard assay, which can detect enzymes with specific activity 10,000-fold lower than the wild type. Similar conclusions were obtained following a chemical approach. Cys 172 was the only cysteine that under nondenaturing conditions could be chemically modified by either iodoacetate or Ellman's reagent. After modification, the enzyme was totally inactive.
The involvement of a cysteine in the enzymatic activity fits with the type of chemical reaction catalyzed by carbamylase. Among the amide bond-hydrolyzing enzymes, many utilize either the hydroxyl group of serine and threonine side chains or the thiol group of cysteine as a nucleophile to attack the scissile bond. Typical examples are the serine and cysteine proteases (25,26), the amydohydrolases, which catalyze the hydrolysis of asparagine and glutamine to their acidic forms (22,27,28), and the E. coli penicillin acylase responsible for the conversion of penicillin G to 6-aminopenicillanic acid (23). Interestingly, the N-carbamyl-D-amino-acid amidohydrolase from A. radiobacter shares about 25% identity with the aliphatic amidases of both Pseudomonas aeruginosa and Rhodococcus erythropolis (29,30) with the highest degree of homology being found in the region surrounding and including Cys 172 .
Grouping the A. radiobacter carbamylase in the "cysteine and serine proteases family" implies that (i) the reaction proceeds through the formation of an acyl-enzyme intermediate and (ii) a neighboring residue such as histidine or lysine should serve as a base to enhance the nucleophilicity of Cys 172 . Both properties can be experimentally tested. In particular, we recently developed a rapid screening procedure able to identify on agar plates carbamylase-deficient mutants. We are utilizing such screening protocol to shed light on the amino acid residues crucial for the enzymatic activity.
It has been shown that in some cysteine-dependent enzymes the active site cysteine can be substituted for serine without completely destroying the enzyme activity. For example, the substituted thymidylate synthases of both E. coli and bacteriophage T4 retain 0.02 and 0.07% activity of the wild type enzymes, respectively (31,32).
Our Cys 172 3 Ser mutant was at least 4 orders of magnitute less active than the wild type, indicating that the carbamylase has more structural and catalytic constrains than other cysteine-dependent enzymes. The resolution of the three-dimensional structure of the enzyme will be of great help to shed light on the details of the catalytic reaction. In this context, the availability of large quantities of pure enzyme will facilitate future protein crystallization experiments.
An interesting aspect of the work presented here is that by using different protocols for the derivatization of the thiol groups, including treatment of the enzyme under native, mild denaturing, and strong denaturing conditions, it has been possible to have some hints as to the topological position of the five cysteines present in the molecule. The data clearly show that the active site Cys 172 is readily accessible to the solvent, Cys 193 and Cys 250 are buried in the protein core, and Cys 243 and Cys 279 are probably located near external loops. These last two cysteines may be the ones involved in the formation of intermolecular disulfide bridges that we found to occur under oxidizing conditions. If this is the case, the replacement of Cys 243 and Cys 279 with other amino acids should be beneficial for the enzyme stability.