Development of highly potent and selective phosphinic peptide inhibitors of zinc endopeptidase 24-15 using combinatorial chemistry.

Several hundred phosphinic peptides having the general formula Z-(L,D)Pheψ(PO2CH2)(L,D)Xaa‘-Yaa‘-Zaa‘, where Xaa‘ = Gly or Ala and Yaa‘ and Zaa‘ represent 20 different amino acids, have been synthesized by the combinatorial chemistry approach. Peptide mixtures or individual peptides were evaluated for their ability to inhibit the rat brain zinc endopeptidases 24-15 and 24-16. Numerous phosphinic peptides of this series act as potent (Ki in the nanomolar range) mixed inhibitors of these two peptidases. However, our systematic and comparative strategy led us to delineate the residues located in P2′ and P3′ positions of the inhibitors that are preferred by these two peptidases. Thus, endopeptidase 24-15 exhibits a marked preference for inhibitors containing a basic residue (Arg or Lys) in the P2′ position, while 24-16 prefers a proline in this position. The P3′ position has less influence on the inhibitory potency and selectivity, both peptidases preferring a hydrophobic residue at this position. On the basis of these observations, we have prepared highly potent and selective inhibitors of endopeptidase 24-15. The Z-(L,D)Pheψ(PO2CH2)(L,D)Ala-Arg-Met compound (mixture of the four diastereoisomers) displays a Ki value of 70 pM for endopeptidase 24-15. The most selective inhibitor of endopeptidase 24-15 in this series, Z-(L,D)Pheψ(PO2CH2)(L,D)Ala-Arg-Phe, exhibits a Ki value of 0.160 nM and is more than 3 orders of magnitude less potent toward endopeptidase 24-16 (Ki = 530 nM). Furthermore, at 1 μM this selective inhibitor is unable to affect the activity of several other zinc peptidases, namely endopeptidase 24-11, angiotensin-converting enzyme, aminopeptidase M, leucine aminopeptidase, and carboxypeptidases A and B. Therefore, Z-(L,D)Pheψ(PO2CH2)(L,D)Ala-Arg-Phe can be considered as the most potent and specific inhibitor of endopeptidase 24-15 developed to date. This new inhibitor should be useful in assessing the contribution of this proteolytic activity in the physiological inactivation of neuropeptides known to be hydrolyzed, at least in vitro, by endopeptidase 24-15. Our study also demonstrates that the combinatorial chemistry approach leading to the development of phosphinic peptide libraries is a powerful strategy for discovering highly potent and selective inhibitors of zinc metalloproteases and should find a broader application in studies of this important class of enzymes.

Several hundred phosphinic peptides having the general formula Z-(L,D) Phe(PO 2 CH 2 ) (L,D) Xaa-Yaa-Zaa, where Xaa ‫؍‬ Gly or Ala and Yaa and Zaa represent 20 different amino acids, have been synthesized by the combinatorial chemistry approach. Peptide mixtures or individual peptides were evaluated for their ability to inhibit the rat brain zinc endopeptidases 24-15 and 24-16. Numerous phosphinic peptides of this series act as potent (K i in the nanomolar range) mixed inhibitors of these two peptidases. However, our systematic and comparative strategy led us to delineate the residues located in P 2 and P 3 positions of the inhibitors that are preferred by these two peptidases. Thus, endopeptidase 24-15 exhibits a marked preference for inhibitors containing a basic residue (Arg or Lys) in the P 2  The endopeptidase 24-15 (EC 3.4.24.15) 1 belongs to the zinc metalloprotease family (1) and resembles a peptidase previously purified from rabbit brain by Camargo et al. (2). Later, endopeptidase 24-15 was named thimet oligopeptidase with respect to the thiol and metal dependence of its catalytic activity (3)(4)(5). Molecular cloning of the cDNA of endopeptidase 24-15 revealed a HEXXH motif, which characterizes peptidases belonging to this family, and a cysteine residue, lying on the C-terminal side of the second histidine of the zinc binding motif (6). This cysteine residue has been proposed to be responsible for activation of the enzyme by 2-mercaptoethanol or dithiothreitol, as well as for inhibition of the enzyme activity by thiol reactive reagents (6).
We previously reported that phosphodiepryl03, a phosphonamide peptide, acts as a potent mixed inhibitor of 24-15 and 24-16, with K i values in the nanomolar range (10). This inhibitor was shown to be unable to block the activity of several other zinc peptidases such as endopeptidase 24-11, angiotensin-converting enzyme, aminopeptidase M, leucine aminopeptidase, and carboxypeptidases A and B. We particularly studied the effect of this inhibitor in vivo on the neurotensin catabolism, since we previously established that 24-15 and 24-16 participated in vitro in the inactivation of this neuropeptide (11,12). We established that phosphodiepryl03 prevented neurotensin degradation in vivo in vascularly perfused dog ileum (13). More recently, several phosphodiepryl03 analogues were also proved to be potent but still acting as mixed inhibitors of * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
In this paper, we demonstrate that the synthesis, by combinatorial chemistry, of several hundred different phosphinic peptides having the general formula Z-(L,D) Phe(PO 2 -CH 2 ) (L,D) XaaЈ-YaaЈ-ZaaЈ has led to the discovery of both highly potent and selective inhibitors of 24-15.

MATERIALS AND METHODS
Diisopropyl fluorophosphate-treated carboxypeptidases A and B, leucine aminopeptidase, and angiotensin-converting enzyme were from Sigma. Endopeptidase 24.11 was purified and kindly provided by Drs. P. Crine and G. Boileau (Département de Biochimie, Université de Montréal, Canada). All the Fmoc-amino acid derivatives, the Mcc-Pro-Leu-Gly-Pro (D) Lys(Dnp) synthetic substrate and the 2-chlorotrityl resin were from Novabiochem.

24-15 and 24-16
Assays-Unless otherwise noted, all assays were performed at 25°C in 5 mM tricine/NaOH buffer, pH 7.5, containing 0.1 mg/ml bovine serum albumin, in the absence  or in the presence (24-15) of 0.1 mM dithiothreitol. Under these conditions, the enzymes retain their activity for 24 h. Endopeptidase activities were assayed with Mcc-Pro-Leu-Gly-Pro-(D)Lys(Dnp) substrate as previously described (17,18). Continuous assays were performed by recording the hydrolysis of this quenched substrate with a fluorimeter, setting excitation and emission wavelengths at 347 and 405 nm, respectively. In typical experiments, a cuvette containing 0.9 ml of buffer, 9 M of substrate was brought to thermal equilibrium in a jacketed holder in a compartment of a Perkin-Elmer LS 50 luminiscence spectrometer cell. The temperature was maintained by water circulating from a Haake F3 bath. Addition of 3-30 l of enzyme (final concentration, 0.1-1 nM) initiated the reaction.
Determination of K i Values-In most cases, the steady state velocities observed with different inhibitor concentrations can be determined from the progress curves before significant substrate depletion. Under these conditions, the inhibition constants K i can be directly determined from Dixon plots. However, for slow, tight binding inhibitors, equilibrium between the enzyme and inhibitor cannot be reached before complete substrate depletion. In these cases, 0.1-1 nM of enzyme was allowed to equilibrate overnight or longer with increasing concentrations of inhibitor; then, the reaction was initiated by addition of 10 l of substrate to determine the residual free enzyme concentration. Due to the slow dissociation rates for most of the inhibitors, it was possible to neglect the effect of the addition of a competitive substrate on the equilibrium position. Determination of the equilibrium position allows us to calculate the K i value using Equation 1, which takes into account mutual depletion of enzyme and inhibitor (19,20). All the values reported for K i were reproducible within Ϯ5%. Inhibitor Synthesis-Solid phase syntheses were performed manually or on a model 357 Advanced ChemTech multiple peptide synthesizer. The side chain protecting groups used were Asp(tBu), Glu(tBu), Ser(tBu), Thr(tBu), Tyr(tBu), Lys(Boc), His(Trt), Asn(Trt), Gln(Trt), and Arg(Pmc). The phosphinic pseudodipeptides Z-(L,D) Phe(PO 2 -CH 2 )GlyOH and Z-(L,D) Phe(PO 2 -CH 2 ) (L,D) AlaOH were prepared as previously described (15). The protection of the hydroxyphosphinyl group by the adamantyl group will be published elsewhere. 2 The protected phosphinic pseudodipeptides Z-(L,D) Phe(PO (OAd) -CH 2 )GlyOH and Z-(L,D) Phe(PO (OAd) -CH 2 ) (L,D) AlaOH were incorporated as a block during the solid phase synthesis.
The first Fmoc amino acids were attached to the 2-chlorotrityl resin according to Barlos et al. (21). The degree of substitution of each resin sample was determined according to Meienhofer et al. (22), but using ⑀ ϭ 7040 M Ϫ1 ⅐cm Ϫ1 at 300 nm (23). The Fmoc groups were removed with 50% piperidine in dimethylformamide. Coupling of the next residue was achieved using the 2-(1H-benzotriazol-1-yl)1,1,3,3-tetramethyluronium hexafluorophosphate/diisopropylethylamine in situ strategy. Typically, 3 equivalents of Fmoc amino acid and 4 equivalents of diisopropylethylamine in dimethylformamide or N-methylpyrrolidone were added to the resin, and the reaction was allowed to proceed for 30 min. The coupling of Z-(L,D) Phe(PO (OAd) -CH 2 ) (L,D) XaaOH was achieved using 1.5 equivalents of this block. Fully protected peptides were cleaved from 2-chlorotrityl resin with a mixture of glacial acetic acid, trifluoroethanol, and dichloromethane (2/2/6) during 2 h. Solutions of protected peptides were dried in vacuo. Protective groups were removed by the action of trifluoroacetic acid containing 5% H 2 O, 5% thioanisol, 5% phenol, 2.5% ethanedithiol, and 22.5% of dichloromethane. Solutions of deprotected peptides were concentrated in vacuo, and peptides were treated with 2.5 equivalents of NaHCO 3 in water. Aqueous solutions were repeatedly extracted with cold diethylether and then lyophilized. Peptide mixture homogeneity was checked by amino acid analysis, allowing us also to determine concentrations of these solutions. Single peptides were purified by preparative HPLC (Vydac 218TP1022 column) performed on a Gilson system equipped with a variable wavelength detector. Peptide purities were checked by TLC, analytical HPLC (Vydac 218TP104 column), and mass spectrometry.
Construction of the Peptide Library by Combinatorial Chemistry-The protocol employed to synthesize the various mixtures of phosphinic peptides of formula Z-(L,D) Phe(PO 2 CH 2 )Gly-YaaЈ-ZaaЈ, where YaaЈ and ZaaЈ represent 20 different amino acids, is briefly described in Fig. 1. A first amino acids mixture is prepared by combining the 20 different amino acids linked to the 2-chlorotrityl resin (ZaaЈ-R samples) (Cys was omitted and replaced by Nle). This homogenized mixture is then split into 20 identical samples to which a known amino acid (YaaЈ) is coupled. This gives rise to 20 different mixtures, each of them containing 20 different dipeptides. The next step is the introduction of the phosphinic dipeptide moiety (Z-(L,D) Phe(PO (OAd) -CH 2 )Gly) in each sample. Side chain deprotection and cleavage of the peptides from the 20 different resins allowed the recovery of 20 different phosphinic tetrapeptide mixtures, each containing 40 different tetrapeptides. Amino acid analysis of these mixtures confirmed the expected amino acid composition and indicates that the ZaaЈ position has incorporated equimolar amounts of the 20 different amino acids.
The protocol used to synthesize the mixtures of phosphinic peptides of formula Z-(L,D) Phe(PO 2 CH 2 )-(L,D) Ala-YaaЈ-ZaaЈ was similar to that one described for the phosphinic peptide mixtures containing the Z-(L,D) Phe(PO 2 CH 2 )Gly moiety. However, in this case due to the presence of both (L,D) Phe and (L,D) Ala residues in these peptides, the final 20 different mixtures each contain 80 different phosphinic peptides.
Attribution of the L (R configuration) or D (S configuration) stereochemistry of the pseudophenylalanine residue was determined according to the procedure previously described (24). RESULTS We previously reported the ability of a phosphonamide peptide, phenylethyl-(PO 2 -NH)Gly-Pro-Nle (phosphodiepryl03), to act as a potent mixed inhibitor of both 24-16 and 24-15 (10). The structure of this inhibitor, which is thought to encompass the S 1 to S 3Ј subsites of the catalytic site of these enzymes, was therefore used as the starting point of a strategy aimed at developing fully specific blockers of 24-15. Preliminary experiments examined the putative influence of 1) substitution of the phosphonamide surrogate by a phosphinic group; 2) stereochemical modifications introduced at the P 1 , P 2Ј , and P 3Ј posi-tions; and 3) esterification of the C terminus of the inhibitor. Table I indicates that the phosphinic peptide appears slightly less potent than the corresponding phosphonamide peptide (compounds 2 and 1). Furthermore, introduction of D-amino acids at the P 1 , P 2Ј , and P 3Ј positions drastically reduces the potency of the inhibitors toward both peptidases (see compounds 3, 4, and 5). Finally, esterification of the C-terminal carboxyl group decreases the inhibitor affinity, indicating that the binding of these phosphorus peptides requires a free Cterminal carboxylate (compounds 2 and 6). Based on these preliminary observations, the search for new compounds was undertaken by checking systematically the role played by the P 2Ј and P 3Ј positions in the potency and selectivity of the inhibitors.
Preference of 24-15 and 24-16 for the P 2Ј and P 3Ј Positions of the Phosphinic Peptide Inhibitors-20 different peptide mixtures having the general formula Z-Phe(PO 2 CH 2 )Gly-YaaЈ-ZaaЈ, each containing a single amino acid in the YaaЈ position, while a mixture of 20 different amino acids is present in the ZaaЈ position, were prepared as described under "Materials and Methods." The determination of the relative potency displayed by each mixture makes it possible to determine the preference of both peptidases for the inhibitor P 2Ј position. The data in Fig.  2 indicate that the preferred amino acid residues at the P 2Ј position of the inhibitor are arginine and lysine for 24-15 and proline for 24-16. Based on these results, the synthesis of 20 different phosphinic peptides of the general formula Z-(L,D) Phe(PO 2 CH 2 )Gly-Arg-ZaaЈ, with ZaaЈ representing 20 different amino acids, was performed. Each peptide was purified and its inhibitory potency determined on 24-15 and 24-16. The influence of the P 3Ј position on the selectivity of these inhibitors toward these peptidases is shown in Fig. 3. Two inhibitors, containing a phenylalanine or a methionine in the P 3Ј position, are identified as rather selective compounds for 24-15, with selectivity factors of 410 and 380, respectively. From this diagram, it appears that aromatic, hydrophobic, and basic side chains in the P 3Ј positions also confer some selectivity for 24-15 to the phosphinic peptide, while the presence of a small, acidic amino acid or proline residue reduces the inhibitor selectivity. Furthermore, from Table II, it can be seen that the two selective phosphinic peptides, with a Phe or Met residue in the P 3Ј position, are also potent inhibitors of 24-15. From the potency point of view, 24-15 seems to prefer a linear hydrophobic side chain, with a clear preference for the Met residue in the P 3Ј position of the inhibitor. (25), as well as recent results in our laboratory (14), have shown that an alanine residue in the P 1Ј position of the inhibitor slightly increases the potency of the inhibitors, as compared to the compounds containing a glycine in this position. Accordingly, phosphinic peptide mixtures containing an alanine in the P 1Ј position were synthesized to investigate the effect of this residue both on inhibitor potency and selectivity. Based on the results observed in this study, four peptide mixtures containing, respectively, a Pro, Lys, Arg, or Nle residue in the P 2Ј position were prepared. The IC 50 of these four mixtures toward endopeptidases 24-15 and 24-16, as well as their selectivity factors for these peptidases, are reported in Table III. The presence of either a Pro or Nle residue in the P 2Ј position does not lead to selective inhibitors, while a basic residue in the same position provides potent and selective inhibitors. Remarkably, as compared to the results reported in Fig. 2, with an Ala in the P 1Ј position, the inhibitor mixture containing an Arg in the P 2Ј position appears more selective than that one with a lysine in this position. This striking observation led us to synthesize several phosphinic peptides containing an alanine in the P 1Ј position and a Lys or Arg residue in the P 2Ј position. The data in Table IV confirm that the presence of arginine, as compared with lysine, influences the selectivity of the inhibitors and provides with Z-(L,D) Phe(PO 2 CH 2 ) (L,D) Ala-Arg-Phe a compound that is 3300 times more potent toward 24-15 than toward 24-16. As reported for the phosphonamide peptides (10,13), it is worth noting that this phosphinic peptide inhibitor at 1 M is unable to block the activity of several zinc peptidases, like endopeptidase 24-11, angiotensin-converting enzyme, aminopeptidase M, leucine aminopeptidase, and carboxypeptidases A and B (data not shown). Comparison of the results in Fig. 2 and Table  IV highlights the role played by the residue in the P 1Ј position in the inhibitor selectivity. However, either with a Gly or Ala   24 -15 and 24 -16 Values for K i were determined as described in the text. Assays were carried out in 5 mM tricine/NaOH, pH 7.5, 0.1 mg/ml bovine serum albumin in presence (endopeptidase 24 -15) or in absence (endopeptidase 24 -16) of 0.1 mM dithiothreitol. Selectivity corresponds to the ratio K i 24 -16 /K i 24 -15 .

Influence of the P 2Ј Positions in Phosphinic Peptide Mixtures Having the General Formula Z-(L,D) Phe(PO 2 CH 2 ) (L,D) Ala-YaaЈ-ZaaЈ-Previous inhibitor studies
Inhibitors -P 1 ----P 1Ј -P 2Ј -P 3Ј - residue in the P 1Ј position, it can be seen that the preference of 24-15 in the P 3Ј position is still for methionine.

DISCUSSION
The development of phosphinic peptide chemistry by solid phase synthesis, in conjunction with the combinatorial chemistry approach (26,27), makes it possible to prepare rapidly a huge number of different phosphinic peptides. All these phosphinic peptides, as good analogues of the substrates of zinc metalloproteases in the transition state, are expected to be potent inhibitors of this enzyme family. In fact, as demonstrated in this study, but also in previous work (15), phosphinic peptides provided that they contain the right amino acid sequence, which ensures an optimal recognition of these inhibitors by the target zinc metalloprotease, are highly potent inhibitors of this class of proteases. In this work, we have identified a phosphinic sample (Z-(L,D) Phe(PO 2 CH 2 ) (L,D) Ala-Arg-Met, a mixture of four distereoisomers) displaying a K i value of 70 pM. Thus, the phosphinic peptides of this sample are by far much more potent inhibitors of 24-15 than the previously reported carboxyalkyl (25,28) or hydroxamate (29) peptide inhibitors, all exhibiting K i values in the range of 10 -100 nM.
The other important aspect of the approach is the possibility of optimizing the inhibitor selectivity by screening the peptide mixtures with different proteases. The power of the approach is well illustrated in this work by the recognition of the clear preference of 24-15 for an arginine or lysine residue in the P 2Ј position of the inhibitor, two residues that are much less tolerated at the same position by 24-16. 24-16 prefers a proline residue in this position than a basic one. Interestingly, the proline residue is also well accommodated by 24-15. This result may explain why, for a long time, it has been stated that 24-15 has a preference for proline in the P 2Ј position (28). This lack of selectivity, displayed by the two peptidases toward the recognition of peptide inhibitor with proline in the P 2Ј position, accounts for the fact that all the phosphinic peptides having the general formula Z-Phe-(PO 2 CH 2 )Gly-Pro-ZaaЈ behave as mixed inhibitors of these two peptidases (data not shown). Such results are also consistent with the view that these two endopeptidases have closely related active sites, and that even with a systematic approach, the development of highly selective inhibitors of these enzymes still remains a challenge. Thus, identification of selective inhibitors of 24-16 will require investigating the influence of the P 1 and P 1Ј positions in these inhibitors on the selectivity. Nevertheless, Z-Phe-(PO 2 CH 2 )Ala-Arg-Phe, which is 3 orders of magnitude more potent for 24-15 than for 24-16 and unable to block several other zinc endopeptidases, is to date the most specific inhibitor of 24-15. The marked preference of 24-15 for inhibitors having an Arg residue in the P 2Ј position probably explains the efficiency of this enzyme in cleaving bioactive peptides displaying such a structural feature. In this respect, it can be mentioned that the best specificity constant (k cat /K m ratio) for rat 24-15 has been observed for the hydrolysis of dynorphin A 1-8 (3,9). In this peptide (Tyr 1 -Gly-Gly-Phe-Leu-Arg-Arg-Ile 8 ), the cleavage occurs between the Leu and Arg residues. Thus, in the bound state, the Arg 7 of this peptide probably interacts with the S 2Ј subsite of 24-15. Hydrolysis of several bioactive peptides pos-sessing a C-terminal sequence like Xaa-Arg-XaaOH are now under investigation to confirm the effect of the P 2Ј arginine residue on the 24-15 specificity.
Among other factors that may influence the specificity of 24-15, the presence of a free C-terminal carboxylate group has been previously pointed out (28,30). Our data confirm the importance of a free C-terminal carboxylate group in the P 3Ј position of the inhibitor (Table I). It should be noticed that the carboxyalkyl peptide inhibitors developed by Orlowski et al. (25) do not contain a free carboxylate group at the P 3Ј position since this group is protected by a p-aminobenzoate moiety. Therefore, the free carboxylate group of these inhibitors is borne by the para-amino-benzoate group, which likely interacts with the S 4Ј or S 5Ј subsites of 24-15 (30). Altogether, these data suggested that there are several cationic loci in the active site of 24-15 and explain the ability of this endopeptidase to cleave peptidyl bonds located three, four, or five residues from the C-terminal free carboxylate group of a peptide.
Inhibition of 24-15 by the CPP-AAF-pAB inhibitor triggers an increased recovery of luteinizing hormone-releasing hormone in vivo (31) and potentiated the luteinizing hormonereleasing hormone-induced release of plasmatic luteinizing hormone and follicle-stimulating hormone (32). Subsequent studies reported the increased antinociceptive properties of dynorphin A 1-8 and leucine enkephalin-Arg-Gly-Leu after administration of this inhibitor (33). Finally, intravenous infusion of CPP-AAF-pAB rapidly slowed down the arterial pressure of normotensive rats (34), indicating that 24-15 could play a role in the control of the pressor response in mammals. However, these data are still controversial since it was recently reported that CPP-AAF-pAB could undergo proteolytic cleavage (35)(36)(37). This produces a catabolite that potently inhibits angiotensinconverting enzyme, whose role in blood pressure response has been well documented. Therefore, the potent and selective 24-15 inhibitors reported here represent novel tools for reexamining the reel contribution of this enzyme in the control of the above physiological processes. These molecules should also help us to establish the relative contribution of 24-15 and 24-16 in vivo in the neurotensin degradation (13,14).
This new series of inhibitors might also constitute valuable probes for evaluating the degree of similarity between 24-15 and the family of proteins that exhibit a high percentage of sequence identity with the 24-15 (38,39). In this connection, it will be of interest to determine the ability of the present inhibitors to block the peptidase activity of porcine-soluble angioten-  Table I. Selectivity corresponds to the ratio K i 24 -16 /K i 24 -15 .
Inhibitors -P 1 ------P 1Ј -P 2Ј -P 3Ј -   Table I. Inhibitors -P 1 ----P 1Ј -P 2Ј -P 3Ј - sin-binding protein, a protein which was recently shown to share a 65% sequence identity with porcine 24-15 (40). To our knowledge, our study is the first example of the development of phosphinic peptide libraries and their successful use for discovering potent and selective inhibitors of zinc metalloproteases. Recently, a synthetic procedure for developing peptide libraries containing a phosphonate group (PO 2 O) was published (41), but no results were given for the potency of these phosphonate peptides as zinc metalloprotease inhibitors. However, for bacterial collagenase and mammalian 24-15 and 24-16, it should be mentioned that phosphonate peptides have been found to be extremely poor inhibitors of these proteases (14,42). A similar result was reported for thermolysin (43). More recently, phosphinic peptides were also proved to be potent inhibitors of the astacin protease, another zinc metalloprotease (44). Thus, in our opinion, development of phosphinic peptide libraries should be an important future approach for discovering potent and selective inhibitors of other zinc metalloproteases (45), a family of proteases that has rapidly grown in the last few years and is of great interest.