Evidence That Both Arginine 102 and Arginine 747 A r e Involved in Substrate Binding to Neutral Endopeptidase (EC 3.4.24.11)"

Neutral endopeptidase (EC 3.4.24.11, NEP) is a Zn- metallopeptidase involved in the degradation of biolog-ically active peptides, notably the enkephalins and atrial natriuretic peptide. Recently, the structure of the active site of this enzyme has been probed by site- directed mutagenesis, and 4 amino acid residues have been identified, namely 2 histidines (His68a and Hises7), which act as zinc-binding ligands, a glutamate (Glds4) involved in catalysis, and an arginine residue (Arg"'), suggested to participate in substrate binding. Site-di-rected mutagenesis has now been used to investigate the role of 4 other arginine residues (Arg408, Arg409, Arg"', and Arg747) that have been proposed as possible active site residues and to further analyze the role of Arg"'. In each case, the arginine was replaced with a methionine, and both enzymatic activity and the IC60 values of several NEP inhibitors were measured for the mutated enzymes and compared to wild-type enzyme. The L-alanine- t-butyl CaCl, precipitation step and substitution of the phosphate buffer by 50 mM Tris-HC1, pH 7.4. Protein concentrations were determined by the method of Bradford (41). Computer Graphics Modelling of the Complex between Butanedione and in the Active Site of Thermolysin-Volume calculations and molecular structure display were carried out using Sybyl (Tripos), running on an Evans and Sutherland PS390, with a Microvax 3500 as host computer. Volume corresponded to the interactions between the van der Waals volume of the hypothetical modified arginine with the rest of the active site. To model the active site, the refined x-ray structure from the Protein Data Bank (PDB 3TLN entry) was used (42). The modified arginine was constructed using standard bond lengths and valency angles, then optimized using the Tripos Force Field.

system (2,3), the microvilli of the intestine (4)) the reticular cells of the lymph (5)) neutrophils (6), lymphocytes, lung (7), and the adrenal gland (8). It has also recently been shown to be the common acute lymphoblastic leukemia antigen (CALLA, CD10) (9,lO). Preferred substrates for the enzyme are short peptides which are cleaved on the NHz-terminal side of hydrophobic residues (reviewed in Ref. 11). Interest in the structure of the active site of NEP was initially stimulated by the observation that NEP inhibitors possess naloxonereversible antinociceptive properties, which can be linked to the enzyme's role in degrading the enkephalins in the central nervous system (reviewed in Refs. 12 and 13). Peripheral NEP also appears to be the enzyme mainly responsible for the degradation of atrial natriuretic peptide (14), and, in this context, NEP inhibitors may have therapeutic value as novel antihypertensives (15).
NEP is one example of an important group of Zn-metallopeptidases, of which only a few, including carboxypeptidase A (EC 3.4.17.1) and the bacterial endopeptidase thermolysin (EC 3.4.24.4, TLN), have been crystallized (16,17). Most studies so far have shown that the active site of NEP bears a closer resemblance to the active site of TLN than to that of carboxypeptidase, as expected for an endopeptidase. The two enzymes have similar specificities and stereochemical requirements (18)(19)(20)(21)(22), and both have a histidine residue in the active site (23,24) which may stabilize the transition state of the substrate-enzyme complex (25). In carboxypeptidase A, the equivalent residue is Tyr248 (16). In addition, NEP and TLN have limited sequence similarity, involving certain active site residues of the latter (26) and including the consensus sequence VXXHEXXH, which has been found in numerous other Zn-endopeptidases and in the Zn-exopeptidase aminopeptidase N, but not in the Zn-carboxypeptidases (27). In TLN, the 2 histidine residues in this sequence are zincbinding ligands, while the glutamate is involved in catalysis (17,18). The importance of the equivalent NEP residues (His583, His587, and Glu584) has recently been demonstrated by site-directed mutagenesis (28,29). A computer-assisted alignment of the likely secondary structures of NEP and TLN has also predicted a high degree of similarity at their active sites (30).
Several groups have used the arginine-specific reagents butanedione (23,(31)(32)(33) and phenylglyoxal (32,33) to demonstrate the presence of an arginine residue at the active site of NEP. In this respect, the enzyme appears to differ from TLN, which is unaffected by butanedione (31), even though is known to be in the active site and to participate in substrate binding (18). From the computer-assisted alignment (30)) 4 arginines Arg408, Arg409, Arga9, or ArgT4' have been suggested as possible active site residues of NEP. In addition, Arg"' has been identified as being irreversibly labeled after  (33). Site-directed mutagenesis has now been used here to assess the role of these 5 arginine residues, by individually replacing each of them with a methionine residue. The results indicate that two, Arg747 and Arg'02, may be involved in substrate binding. ArgI4" appears to interact with the carbonyl amide group of the Pi residue, and Arg'" could interact, as suggested (33), with the free carboxyl group of some substrates, e.g., enkephalins. In addition, computer modelling has been used to explain the differences between NEP and TLN in butanedione sensitivity.
Synthesis of the Neutral Endopeptidase-24.11 Inhibitor Phenylakmyl-$(CH2NH)-alunine-A solution of N-(t-butyloxycarbonyl) phenylalaninal (1 eq), obtained as previously described (36), in 1% acetic acid/methanol was successively heated with 1 eq of L-alaninet-butyl ester and 3 eq of sodium cyanoborohydride. The mixture was stirred for 1 h at room temperature, then evaporated to dryness, and the residue was dissolved in ethyl acetate. After washing (HzO, NaHC03, citric acid, NaCl), the organic layer was dried over Na2S04 and evaporated in uacuo. The residue was purified by chromatography on silica gel using CHZCl2/Et20, 8 2 as eluent. The protected pseudodipeptide was treated with 50% trifluoroacetic acid in CH,Cl, at 0 "C. After evaporation in uacuo, the residue was chromatographed on silica gel using iPrOH/NH40H, 9:l as eluent (RF 0.17). Site-directed Mutagenesis of Neutral Endopeptidase cDNA and Expression in COS-1 Cells-Conversion of the Arg codon to Met was accomplished by oligonucleotide-directed mutagenesis according to the method of Taylor et al. (37), with an M13 subclone containing the proper fragment of NEP cDNA. Mutant cDNAs were screened by sequencing, and a fragment containing the mutated region was isolated from the replicating form of the M13 recombinant phage and substituted in pSVENK19, an expression vector using the SV40 early promotor to express NEP in COS-1 cells, as previously described (28). Restriction mapping ensured that no rearrangement of DNA had occurred, and the mutated regions were sequenced to confirm that the mutation had been introduced in the expression vector.
Transfected COS-1 cell membranes were solubilized at a protein concentration of 1 mg/ml in 50 mM Tris-buffered saline, pH 7.4, containing 1% (w/v) n-octyl glucoside for 30 min at 4 "C. The preparation was centrifuged at 10,000 X g for 15 min, and the supernatant was taken for enzyme assays. Western blotting was used to show the presence of the 94-kilodalton protein and to determine the quantity of NEP present, as previously reported (38). Extracts were resolved on 8% sodium dodecyl sulfate-polyacrylamide gel electrophoresis and transferred to 0.45 pM nitrocellulose film.
Assay for Neutral Endopeptidase 24.1 I Actiuity-NEP activity and IC60 values were measured using 20 nM ~-Ala'[tyrosyl-3, 53H]-leucine enkephalin as substrate. Assays were carried out in 100 pl of 50 mM Tris-HC1, pH 7.4, at 25 "C, and the reaction was stopped by adding 10 pl of 0.5 M HCI. The metabolite [3H]tyrosyl-glycyl-glycine was separated using Porapak Q beads (Waters) as previously described (39). When COS-1 membrane preparations were used, the incubations also contained 1 p M captopril and 10 p~ bestatin (inhibitors of angiotensin-converting enzyme and aminopeptidases, respectively), and control tubes also contained 1 p~ concentration of the NEP inhibitor thiorphan. When K, and V,, values were determined, the substrate was D-AlaZ-leucine enkephalin used over a 70-fold concentration range, with 40 nM D-Ala2[tyrosyl-3, 53H]-leucine enkephalin included as tracer.
Enzyme activity was also determined by the continuous fluorometric assay described by Florentin et al. (35), using 30 p~ DAGNPG or DAGNPG-amide as substrates. Reactions were carried out at 37 "C in 50 mM Tris-HC1, pH 7.4 buffer and contained 0.4 pg of the enzyme. The reaction was monitored at 562 nm (ex = 342 nm).
Znactiuation by Butanedione-Solubilized COS-1 membrane preparations, containing between 0.2 and 0.5 pg of NEP, were incubated in 100 mM sodium borate, pH 8.0, at 30 "C for 45 min with 10 mM butanedione. NEP activity was measured after diluting aliquots in 50 mM Tris-HC1, pH 7.4, to butanedione concentrations (50.2 mM), which were found, in control experiments, to have no effect on NEP activity over the time course of the assay (30 min).
Inactiuation by Phenylglyoxal-Rabbit kidney NEP or solubilized COS-1 membrane preparations, containing between 0.2 and 0.5 pg of NEP, were incubated, in the dark, in 100 mM ethylmorpholinylacetate, pH 7.4, at 30 "C for 45 min with 20 mM phenylglyoxal. NEP activity was measured after diluting aliquots in 50 mM Tris-HC1, pH 7.4, to phenylglyoxal concentrations (50.4 mM) which did not affect NEP activity over the time course of the assay. In protection experiments, NEP inhibitors were included in the incubations with rabbit kidney NEP, and enzyme activity was measured after dialysis of the preparation against 50 mM Tris-HC1, pH 7.4, at 4 "C.
Purification of Rabbit Kidney Neutral Endopeptidase-24.11-Rabbit kidney NEP was purified by monoclonal antibody affinity chromatography as previously described (40), with the omission of the CaCl, precipitation step and substitution of the phosphate buffer by 50 mM Tris-HC1, pH 7.4. Protein concentrations were determined by the method of Bradford (41).
Computer Graphics Modelling of the Complex between Butanedione and in the Active Site of Thermolysin-Volume calculations and molecular structure display were carried out using Sybyl (Tripos), running on an Evans and Sutherland PS390, with a Microvax 3500 as host computer. Volume corresponded to the interactions between the van der Waals volume of the hypothetical modified arginine with the rest of the active site. To model the active site, the refined x-ray structure from the Protein Data Bank (PDB 3TLN entry) was used (42). The modified arginine was constructed using standard bond lengths and valency angles, then optimized using the Tripos Force Field.

Mutagenesis of NEP cDNA and Expression of the Mutated
Enzymes-Protein modification studies (23,(31)(32)(33) and similarities in the specificity of NEP and TLN (19)(20)(21)(22) strongly suggest that at least 1 and maybe 2 arginine residues are present in the active site of NEP. Arg408, Arg409, Arg659, and Arg747 have been proposed as possible candidates by computerassisted alignment (30), and, in a recent study [14C]phenylglyoxal was shown to label Arg'" (33). To probe further the role of these arginine residues, site-directed mutagenesis has been used to substitute methionine for these residues in recombinant NEP. Several expression vectors carrying these mutations were derived from the already described plasmid pSVENK19-"659, pSVENK19-"747, and pSVENK19-M-102 had a methionine residue substituted for arginine 408, 409, 659, 747, and 102, respectively, and pSVENK19-M-102/ 747 had methionine residues substituted for both arginine 102 and arginine 747. These plasmid DNAs and the control nonmutated plasmid pSVENK19 were transiently expressed in COS-1 cells, and the presence of the mutated and nonmutated enzymes at the surface of the transfected cells was determined by binding of the lZ5I-2Bl2 monoclonal antibody (28). In each case, similar binding was observed, indicating that the mutations introduced did not interfere with the transport of NEP to the cell surface (results not shown). As the 2B12 antibody pSVENK19 (28). pSVENK19-"408, pSVENK19-"409, the Active Site of Neutral Endopeptidase-21.1 1 recognizes a conformational epitope, this also indicates that the overall tertiary structure of the mutant enzyme does not appear to be drastically altered. In addition, Western blot analysis, using a polyclonal antibody (38), showed the presence of the 94-kilodalton protein in all cases (Fig. 1).

Enzymatic Properties of Mutated Neutral Endopeptidase-
24.11-The K,,, and V,,, values for the degradation of D-Ala'leucine enkephalin were found to be very similar for the wildtype enzyme, Met4"-NEP, Met409-NEP, and Met6"-NEP. Furthermore, the ICso values of the NEP inhibitors thiorphan and phenylalanyl-alanine were also unchanged for these mutated enzymes (results not shown).
In contrast, when methionine was substituted for Arg747, Arg"' , or both Arg"' and Arg747, a significant change in the K,,, values was observed ( Table I). The K,,, of D-Ala'"leucine enkephalin for Met747-NEP was approximately 3-fold higher than for the wild-type enzyme, and, for Metl0'-NEP and Met102/747-NEP, it was increased 6-and 16-fold respectively. Little change was seen in VmaX values ( Table I).
The initial rate of degradation of DAGNPG was also altered for the mutated enzymes (Fig. 2), being reduced by 40%) 83%, and 91% for Met747-NEP, Met'''-NEP, and Met747"02-NEP, respectively. The degradation of DAGNPG-amide by the wildtype enzyme was only 22% of that of DAGNPG, confirming a previous observation that a free COOH-terminal carboxyl group is important for the fluorescent substrate (35). It was also reduced for Met747-NEP, but was unchanged for Met'''-NEP and Met747"02-NEP. The rate of degradation of the amidated substrate was relatively similar for all four enzyme preparations.

The Role of Arg747 and Arg"' in Substrate and Inhibitor
A B Binding-To further characterize the role of Arg747 and Arg"' in substrate and inhibitor binding, the inhibitory potencies of several NEP inhibitors, whose structures are shown in Fig. 3, were determined for the three mutated enzymes and compared to those for the wild-type enzyme. Small but significant increases of 4-and 6.4-fold in the ICso of thiorphan for Met747-NEP and Met"'-NEP, respectively, and a 30-fold increase for Met'02/747-NEP were observed (Table I). However, thiorphan has a nanomolar affinity for NEP, and the strength of the coordination of the thiol group to the zinc atom mimimizes the loss of other stabilizing factors such as stereochemical requirements, hydrogen bonds, van der Waals, and ionic interactions (43). Phenylalanyl-alanine, which has a micromolar affinity for the enzyme, was therefore tested along with two analogs of this dipeptide, phenylalanyl-alanine-amide and phenylalanyl-$(CH2NH)-alanine (Tables I1 and 111).

27-
For the wild-type enzyme, the IC, of phenylalanyl-alanineamide was 112-fold higher than that of phenylalanyl-alanine confirming previous observations that a free COOH-terminal carboxyl group is important for the binding of NEP inhibitors (44). Reduction of the carbonyl-amide bond of phenylalanylalanine gave the inhibitor phenylalanyl-$(CH2NH)-alanine, whose reduced amide group would be positively charged and therefore likely to be repelled by positively charged residues in the active site. It is probably this factor, coupled with the loss of a potential hydrogen bond with a putative NEP arginine, by removal of the inhibitor's carbonyl group, which led to the large loss in affinity of this compound for the wild-type enzyme.
For Met747-NEP, there were similar increases in the ICs0 values of both phenylalanyl-alanine (6-fold) and phenylalanyl-alanine-amide (4.2-fold), as compared to the wild-type enzyme. There was, however, no change in the ICso of phenylalanyl-$(CH2NH)-alanine (Table 11), suggesting that, in the wild-type enzyme, Arg747 could interact with the carbonyl group of a P;-P; amide bond (Fig. 3).
For Met"'-NEP the ICso of phenylalanyl-alanine was increased 108-fold, that of phenylalanyl-alanine-alamide 2.4fold, and that of phenylalanyl-$(CHzNH)-alanine 3-fold, with respect to unmodified NEP (Table 111). It was particularly significant that the ICso of phenylalanyl-alanine (136 pM) was similar to that of phenylalanyl-alanine-amide for the wildtype enzyme (140.6 PM) implying that by mutating Arg"' an interaction has been lost with the free carboxyl group of the Pi residue. putative Arg residues which could interact with residues in the PI and Pi positions, respectively. Above are shown the NEP inhibitors used in this study. In each case, the benzyl ring of the Pi residue of the inhibitor is designed to fit in the Si subsite and the SH groups of thiorphan and PEM-1 to form a complex with the zinc atom.
for all three inhibitors (Table 111).
Effect of Butanedwne Treatment on Recombinant Neutral Endopeptidase-24.1 1 -Treating NEP with the arginine-specific reagent butanedione results in rapid inhibition of the enzyme (23,31,32), thought to be due to the modification of a single arginine residue located in the active site (32). To identify this residue, wild-type, MET747, Met"', and Met'0z/747-NEP were treated with 10 mM butanedione, as described under "Materials and Methods." The incubation time (45 min) was chosen as this gave a maximum inhibition of the wild-type enzyme, with similar 3-fold increases in the K,,, of D-Alaz-leucine enkephalin and the ICs0 of thiorphan (Table I). This treatment did not however significantly change either the K,,, of D-Alaz-leucine enkephalin or the ICs0 values of any of the inhibitors for Met747-NEP (Tables I and 111). In agreement with this, the K,,, of the substrate and the IC60 of thiorphan were unchanged for butanedione-treated Arg'0z/747-NEP (Table I).
In contrast, butanedione treatment of Metlo2-NEP raised the K,,, of D-Alaz-leucine enkephalin 2.6-fold, and the ICs0 of thiorphan 4-fold (Table I). The ICs0 of phenylalanyl-alanine was increased by a factor of 7 and that of phenylalanylalanine-amide by a factor of 3. For phenylalanyl-$(CH2NH)alanine, however, there was only a slight increase in IC50 (1.5fold) (Table 111). These increases were generally similar to those observed when Met747-NEP was compared to wild-type NEP (Tables I and 111). A slight decrease (20-30%) in Vmax was observed for the wild-type and the three mutated enzymes after butanedione treatment.

Effect of Phenylglyoxal Treatment on Met747, Met"', Wild-type, and Rabbit Kidney Neutral
Endopeptidase-24.11-Phenylglyoxal has also been reported to inactivate NEP, but, unlike butanedione, this seems to be due to an interaction with more than 1 residue (32). In agreement with this, 20 mM phenylglyoxal was found to inactivate wildtype, Met747-NEP, MetLo2-NEP, and Met'0z/747-NEP (Fig. 4). The rate of inactivation was similar for wild-type, Met747-NEP, and Metlo2-NEP, and, after a 1-h incubation, the activity remaining was 10,12, and 15% of the controls, respectively. In contrast, phenylglyoxal-induced inactivation of NEP was much slower, with 45% activity remaining after 1 h.
Rabbit kidney NEP was also inactivated by 20 mM phenylglyoxal (90% after a 1-h incubation), and this could only be partially prevented by including the NEP inhibitors, thiorphan and PEM-1 (IC50 2 nM and 50 nM, respectively), at concentrations which would saturate the active site of the enzyme (Fig. 5). In the presence of 1 PM thiorphan, there was a 60% loss of activity after a 1-h incubation with the reagent. 25 PM PEM-1 afforded better protection, but there was still a 40% loss of activity after 1 h.
Computer Graphics Modelling of the Interaction of Butanedwne with Argo3 in the Active Site of Thermolysin-NEP is rapidly inactivated by butanedione while TLN is unaffected (31) even though the latter has an arginine (Argo3) in its active site shown to be involved in substrate binding (18). This has been interpreted as meaning that there is no equivalently positioned arginine residue in NEP (31). In TLN, Argo3 is located inside a deep cleft, formed on one side by HisZ3l, thought to stabilize the transition state of the enzymesubstrate complex, and on the other side by LeuZoz, which is partially involved in the definition of the S: hydrophobic subsite of the enzyme. Butanedione reacts with arginine, in the presence of borate ions, to form a cyclic moiety in which both methyls are on the same side (45). Starting from the xray structure of TLN, the two possible models of the modified arginine were constructed, the first (Fig. 6 A ) with the two methyl groups pointing in the direction of HisZ3l, the other (Fig. 623) with the methyl groups pointing in the direction of Leuzo2. Negative van der Waals forces were then defined by calculating the volume common to the methyls and the atoms of the enzyme. In the first model, both methyls were found to be involved in unfavorable interactions, primarily with HisZ3l and Leuzoz, but also with Hid4', one of the zinc atom ligands. In the second model, there was an unfavorable interaction primarily with the side chain of Leuzo2. These results could

IC50 values of NEP inhibitors for wild-type, Met747, and butanedione-treated Met747-NEP
The chemical formulas of the inhibitors used and their proposed mode of interaction with the NEP subsites are given in Fig. 1 and mutant preparations were incubated with the reagent as described under "Materials and Methods." Aliquots were removed at various times and diluted in 50 mM Tris-HC1 to measure activity. Enzyme activity is expressed as a percentage of controls treated in the same manner but without the addition of phenylglyoxal and is plotted against the incubation time with the reagent. therefore explain why butanedione has no effect on TLN activity (31). If in NEP an arginine residue is present in a position similar to Arg203 of TLN, then butanedione would probably not bind as shown in Fig. 6A as the NEP equivalents of Hisz3* and Hid4' are thought to be well conserved (20, 21). However, the second mode of binding (Fig. 6 B ) could still occur if there is no equivalent of Leuzo2 positioned to cause unfavorable interactions. This is likely since structure-activity studies have shown that the Si hydrophobic pocket of NEP is much larger than its equivalent in TLN (19,20). The immediate environment of an equivalent arginine in NEP could therefore differ sufficiently to allow butanedione modification.  -1 (0). Enzyme activity is expressed as a percentage of a control, treated in the same manner but without the addition of phenylglyoxal, and is plotted against the time of incubation with the reagent.

DISCUSSION
Active site models of TLN and carboxypeptidase A are widely used as starting points for designing inhibitors for other Zn-metallopeptidases whose active site structures are unknown (12,13). The active sites of these two enzymes are similar in many aspects, although there are important differences which contribute to their different specificities. The positioning of an arginine residue which participates in substrate binding (18,22) is an example. In carboxypeptidase A, the position of the guanidinium group of Arg145 is such that it interacts with the carboxyl-terminal group of the substrate, thus preventing the binding of an extended polypeptide chain A -c g. tanedione and the guanidinium groups of in the active site of thermolysin, obtained by computer graphics modelling as described under "Materials and Methods." The amino acids belonging to the active site of TLN, His"', Hisz3', and LeuzM, are indicated while ArgZo3 is located behind Leu"' . The two possible orientations of the cyclic complexes resulting from the reaction between butanedione and ArgZo3 of TLN are shown, with the methyl groups of the heterocycle formed pointing either in the direction of His'31 (A) or in the direction of LeuzM ( B ) . Calculated negative van der Waals forces are shown as caged areas. It can be observed that in both cases (especially in A ) unfavorable interactions occur. (16). In TLN, is placed to the side of the substrate and forms hydrogen bonds with the amide carbonyl group of the Pi residue, allowing the binding of an extended polypeptide (17,18).
NEP, although originally classified as an endopeptidase (1) and having many features in common with TLN, can also act as a dipeptidylcarboxypeptidase, and, with some substrates, e.g. enkephalins, a free COOH-terminal carboxyl group in the Pi position is important for substrate binding (11,31). It has therefore been proposed that the active site of NEP contains an arginine residue which interacts with this group, making the enzyme, in this aspect, more "carboxypeptidase A-like'' However, structure-activity studies as well as site-directed mutagenesis have provided evidence both for a "TLN-like" and a carboxypeptidase A-like arginine in NEP (19)(20)(21)33), and the present results would seem to support this hypothesis. The simplest interpretation is that Arg747 acts as the TLNlike arginine and is modified by butanedione and phenylglyoxal and that, as previously suggested (33), Arg"' acts as the carboxypeptidase A-like arginine and is modified by phenylglyoxal but not by butanedione.
leucine enkephalin and in the IC, values of several NEP inhibitors were of an order expected for the loss of hydrogen bonds (46). The only inhibitor tested whose IC, was not modified for the mutated enzyme was phenylalanyl-+(CH,NH)-alanine, in which the carbonyl group of the S;-Si amide bond has been replaced by a CH, group, suggesting that it is the interaction with the carbonyl group in the other inhibitors that has been lost by mutating Arg747. In addition, neither the K, of D-Ala2-leucine enkephalin nor the ICso values of the inhibitors tested were affected by butanedione treatment of Met747-NEP.
The changes observed in the activity of Met'"-NEP were generally similar to those previously reported when Arg'" was mutated to glutamine (33). The most marked difference in ICs0 (Table 111) occurred with phenylalanyl-alanine (100-fold increase), whose ICso for Met'"-NEP was similar to that of phenylalanyl-alanine-amide for the wild-type enzyme, suggesting that the interaction lost by mutating Arglo2 is with the carboxyl group of the Pi residue. For Met'02/747-NEP, the increases in K, and IC,, values were roughly the equivalent of the sum of those obtained with Met747 and Met'"-NEP.
The K,,, of ~-[~H]Ala~-leucine enkephalin and the IC,, of thiorphan were also unaffected by butanedione treatment of this doubly mutuated enzyme. In particular, it should be noted that the IC5, values of phenylalanyl-alanine, phenylalanylalanine-amide, and phenylalanyl-$(CH,NH)-alanine for Met'02/747-NEP were all in the same range, indicating that, with the two arginine interactions now cancelled, the enzyme has no preference between the three analogs. The results obtained using DAGNPG and DAGNPG-amide as substrates also support this hypothesis, with the rate of degradation of the amidated substrate being relatively similar for all four enzyme preparations.
In contrast to Met747 and Met'02/747-NEP, Met'"-NEP was still affected by butanedione, both the K,,, of the substrate and the ICso values of the inhibitors being increased between 3-and 6-fold. The smallest increase was for phenylalanyl-+(CH2NH)-alanine (1.5-fold). In addition, the increases in K, and ICs,, after treating Metlo2-NEP with butanedione, were of the same order of magnitude as those obtained by mutating Arg747, and the absolute values obtained were close to those for Met'02/747-NEP.
Butanedione has previously been shown to inactivate NEP by interacting with a single residue, thought to be in the active site, as NEP inhibitors completely protect the enzyme from inactivation (24,32). The fact that Metlo2-NEP can still be modified by this reagent therefore argues strongly in favor of a 2nd arginine residue at the active site of NEP, which the present findings suggest to be Arg747. The results also suggest that Arg747 interacts with the carbonyl group of the P; residue and therefore probably has a position in the active site of NEP similar to that of Arg203 in TLN. The previously reported differences in the sensitivity of the two enzymes to butanedione (31) can probably be explained in the light of the computer modelling results, showing that butanedione modification of Arg203 of TLN would cause unfavorable interactions with other groups in the active site (Fig. 6, A and B ) .
It is not clear why the reagent does not also interact with Arglo2 of NEP. However, from this point of view, it is interesting to note that Arg'02 is not located in the COOH-terminal region of NEP where the other active site residues so far identified are found (26,28,29). This suggests a folding of the protein which brings Arglo2 to the entrance of the active site cleft. As the complex formed between the guanidinium group of arginine and butanedione is relatively unstable, it may be more readily reversible with a group in a hydrophilic environment than with a group in the more protected environment of the interior of the active site. Arg747, as previously suggested (30), could be brought into the active site by a disulfide bridge involving Both Met747-NEP and Met"'-NEP were inactivated by phenylglyoxal. In the former case, the reagent could presumably interact with Arg"' and in the latter with Arg747. However, Met'02/747-NEP was also inactivated, albeit at a slower rate than either the wild-type or the singly mutated enzymes, suggesting the occurrence of one or more phenylglyoxal-sensitive arginine residues which are not in the active site itself. This is in agreement with the results of previous studies (32 , 33) showing that ['4C]phenylglyoxal is still incorporated into NEP in the presence of excess substrate or inhibitor and the present findings that NEP inhibitors cannot completely protect the enzyme from phenylglyoxal inactivation (Fig. 5). In addition, the phenylglyoxal inactivation curve for rabbit kidney NEP in the presence of thiorphan closely resembled that of phenylglyoxal inactivation of Met'02/747-NEP. The greater protection afforded by PEM-1 could be due to its long COOHterminal extension able to protect arginine residue(s) situated in the vicinity of the active site but not critically involved in stabilizing inhibitor or substrate binding as shown by the almost identical ICso values of phenylalanyl-alanine and its 2 analoes for Met'02/747-NEP and butanedione-treated Met'''-NEP.-In conclusion, 2 arginine residues seem to be located at the active site of NEP and to participate in substrate binding: Arg747, which binds to the carbonyl amide group of the P; residue, and Arg"' , which can bind to the free carboxyl group of a P6 residue. The presence of these 2 residues in the active site could explain why the enzyme has both endopeptidase and dipeptidylcarboxypeptidase activities. Furthermore, the assignment of these additional amino acids, critically involved in substrate and inhibitor binding, could aid a model of the active site of NEP to be constructed for docking experiments to improve the design of potent and selective inhibitors (47).