Analogs of Diaminopimelic Acid as Inhibitors of meso-Diaminopimelate Decarboxylase from Bacillus sphaericus and Wheat Germ*

Analogs (146) of diaminopimelic acid have been syn- thesized and tested for inhibition of meso-diaminopi-melate decarboxylases from Bacillus sphaericus IF0 3525 and from wheat germ (Triticum vulgaris). Di-fluoromethyldiaminopimelate I does not irreversibly inactivate or strongly competitively inhibit either enzyme. Lanthlonine sulfoxides (2ab, 2c, and 2d) are good competitive inhibitors (about 50% inhibition at 1 mM) of both decarboxylases. The meso and LL-isomers of Ianthionine sulfone (3ab and 3c) and lanthionine (6ab and 6c) are weaker competitive inhibitors (about 50% inhibition at 10-20 mM). The corresponding DD- isomers (3d and 6 4 are less effective. The N-modified analogs are the most potent competitive inhibitors. The inhibition constant (Ki) values for B. sphaericus and wheat germ decarboxylases with N-hydroxydiamino-pimelate 4 (mixture of isomers) are 0.91 and 0.71 mM, respectively; for the N-aminodiaminopimelate 5 (mix-ture of isomers) the Ki values are 0.10 and 0.084 mM, respectively. These N-modified analogs do not effectively inhibit L-lysine

Analogs (146) of diaminopimelic acid have been synthesized and tested for inhibition of meso-diaminopimelate decarboxylases from Bacillus sphaericus IF0 3525 and from wheat germ (Triticum vulgaris). Difluoromethyldiaminopimelate I does not irreversibly inactivate or strongly competitively inhibit either enzyme. Lanthlonine sulfoxides (2ab, 2c, and 2d) are good competitive inhibitors (about 50% inhibition at 1 mM) of both decarboxylases. The meso and LL-isomers of Ianthionine sulfone (3ab and 3c) and lanthionine (6ab and 6c) are weaker competitive inhibitors (about 50% inhibition at 10-20 mM). The corresponding DDisomers (3d and 6 4 are less effective. The N-modified analogs are the most potent competitive inhibitors. The inhibition constant (Ki) values for B. sphaericus and wheat germ decarboxylases with N-hydroxydiaminopimelate 4 (mixture of isomers) are 0.91 and 0.71 mM, respectively; for the N-aminodiaminopimelate 5 (mixture of isomers) the Ki values are 0.10 and 0.084 mM, respectively. These N-modified analogs do not effectively inhibit L-lysine decarboxylase. None of the compounds showed any time-dependent inactivation of the decarboxylases, in contrast to behavior of other pyridoxal phosphate-dependent enzymes with analogous substrate derivatives. Possible mechanisms of inhibition are discussed. In preliminary tests for antibiotic activity 4 and 5 both gave 75% growth inhibition of Bacillus megaterium at 20 pglml in defined media. Other analogs (I#) showed essentially no antibacterial activity.
However, mammals lack this metabolic pathway and require L-lysine as an essential dietary constituent (9). Mammals also rapidly excrete administered diaminopimelate and small peptides containing it (10, l l ) , some of which act as promoters of slow-wave sleep (11,12) or as immunoadjuvants (13). Since L-lysine is itself involved in peptidoglycan cross-linking in many Gram-positive bacteria (14) and is universally necessary for protein synthesis, analogs of meso-diaminopimelic acid which inhibit diaminopimelate decarboxylase could prove lethal to bacteria but should show little mammalian toxicity.
A number of potentially more effective inhibitors, 1-5, can be envisioned (Fig. 2) based on substrate structure and consideration of the extensively studied mechanism of pyridoxal phosphate-dependent decarboxylases (23-25). The a-difluoromethyl analog 1 appeared especially promising because many a-difluoromethyl amino acids which have been examined are potent irreversible inhibitors of the corresponding amino acid decarboxylases (26-31). Although lanthionine 6 is only a weak competitive inhibitor of diaminopimelate decarboxylase (17,18,20), the sulfoxide 2 and sulfone 3 could be more likely to undergo pyridoxal phosphate-catalyzed @-elimination which may lead to enzyme inactivation (26,27,(32)(33)(34). The N-hydroxy analog 4 was chosen because N-hydroxyglutamic acid has been reported to be an irreversible inhibitor of various pyridoxal phosphate-dependent enzymes which metabolize glutamate, including glutamate decarboxylase (35). Finally, the N-amino (hydrazino) derivative 5 was expected to be a potent competitive inhibitor in analogy to the behavior of other a-hydrazino acids with enzymes using pyr- idoxal phosphate as a cofactor (30,36). In the present work we describe the synthesis of diaminopimelate analogs 1-5, their interaction with meso-diaminopimelate decarboxylase from Bacillus sphuericus (17) and from wheat germ (Triticum uulgaris) (18), and some preliminary in vitro tests for antibiotic activity.

RESULTS
Synthesis of Diaminopimelate Analogs-The difluoromethyl diaminopimelate I was prepared as a mixture of all possible isomers ( a d , Fig. 2) by the route illustrated in Scheme 1. The known (37) di-tert-butyl diaminopimelate 7 reacted with benzaldehyde to give the corresponding diimine 8, the lithiated anion of which was treated with chlorodifluoromethane by the procedure of Bey et al. (38). Deprotection under acidic conditions gave compound 1 in 15% overall yield as a mixture of stereoisomers which could not be separated using various chromatographic techniques.
The pure meso-, LL-, and DD-isomers of lanthionine (6ab, 6c, and 6d, respectively) were prepared by condensation of Dor L-cysteine with the appropriate D-or L-isomer of p-chloroalanine (39). Each lanthionine isomer was then individually oxidized with hydrogen peroxide to the corresponding sulfoxide Zab, 2c, or 2d by a modified literature procedure (40).
Since no epimerization occurs at C-2 or C-6, the sulfoxides are stereochemically pure except for 2ab which may be an unresolved mixture of two stereoisomers at the sulfur atom. The isomerically pure meso-, LL-, and DD-lanthionine sulfones (3ab, 3c, and 3d, respectively) were obtained by performic acid oxidation of the corresponding lanthionines 6ab, 6c, and 6d.
Synthesis of N-hydroxydiaminopimelate 4 was accomplished as shown in Scheme 2. The monocarbobenzoxy derivative 9 of di-tert-butyl diaminopimelate 7 was oxidized by the The N-aminodiaminopimelate (hydrazino analog) 5 was prepared as illustrated in Scheme 3. Racemic a-aminopimelic acid was N-protected by the method of Sheehan and Guziec (42) to give compound 11. This was converted to its bis-acid chloride and treated with the lithium derivative of the Roxazolidone 12 described by Evans et al. (43). An easily separable mixture of 13 and its diastereomer having opposite configuration at the a-carbon is produced. Reaction of the lithiated anion of 13 with dibenzyl azodicarboxylate afforded 14 with good stereoselectivity,' but epimerization occurred during basic cleavage of the chiral oxazolidone auxiliary to give 15. Hydrogenolysis of the protecting groups produced the N-aminodiaminopimelate 5 as an optically inactive mixture of stereoisomers.
Interaction of Substrate Analogs with meso-Diaminopimelate Decarboxylase-The diaminopimelate decarboxylases from B. sphaericus (17) and wheat germ (7'. uulgaris) (18) were tested for inhibition of release of 14C02 from [1,7-14C] diaminopimelate (16) by substrate analogs 1 4 (Fig. 2). Although the pure meso-, LL-, and DD-isomers (ab, c, and d, respectively) of the sulfur-containing amino acids 2, 3, and 6 were examined individually, the other analogs, 1, 4, and 5, were each used as statistical mixtures of all possible stereoisomers 0 4 . The isomeric forms of the latter compounds could not be separated by a variety of chromatographic methods. However, the preference of these decarboxylases for the meso-isomer and their absolute specificity for decarboxylation of a D-(R) center (17) suggested that la, 4a, and 5a would be the most potent inhibitors. The results are shown in Table I.
None of the analogs exhibited any irreversible inhibition of either decarboxylase. Surprisingly, the difluoromethyl diaminopimelate 1 was not even a good competitive inhibitor. Lanthionines 6 and the sulfones 3 (Fig. 2) are also weak competitive inhibitors, with the meso (ab) and LL ( c ) isomers being generally stronger than DD ( d ) isomers. The lanthionine sulfoxides 2 show much stronger inhibition. Interestingly, in the case of the B. sphuericus enzyme, the LL-lanthionine sulfoxide 2c appears to be slightly more inhibitory than the meso compound 2ab. Since the N-hydroxy and N-amino diaminopimelates 4 and 5 are the most potent inhibitors, inhibition constants (Ki) were determined for these two analogs by standard methods (44,45)  Mixture of stereoisomers.
zyme. The N-hydroxy compound 4 showed weak competitive inhibition (85% activity of control with 7 mM of this analog); the N-amino derivative 5 tested at 0.2 mM was noninhibitory. Antibacterial Actiuity-Preliminary in vitro tests for antibiotic properties of diaminopimelate analogs 1 4 were done in both complex and defined media. Since most complex media derived from plant sources may be expected to contain L-lysine and some diaminopimelate, defined media lacking these amino acids was used to ensure de novo synthesis of bacterial L-lysine. The difluoromethyl analog I caused no inhibition of growth when tested at 400 pg/ml against Escherichia coli, Bacillus subtilis, or Bacillus cereus in defined

DISCUSSION
It is helpful to consider the mechanism of meso-diaminopimelate decarboxylase in order to understand its behavior with substrate analogs. During decarboxylation of a-amino acids by pyridoxal phosphate-dependent decarboxylases the bond between the a-carbon and the carboxyl carbon of the substrate is expected to be nearly perpendicular to the plane of the cofactor's conjugated ?r system (Fig. 3) (23-25). The cofactor essentially stores the electrons of the cleaved bond until protonation from solvent can occur. This reaction proceeds with retention of configuration with all pyridoxal phosphate-dependent a-decarboxylases investigated to date except for meso-diaminopimelate decarboxylase which shows inversion (15, 16). If a substrate analog is used which has a leaving group (especially fluorine) attached to a 8-carbon of the aamino acid, the intermediate incipient anion (Br or Bz, Fig.  3) can in theory cause elimination to generate an electrondeficient conjugated system. Stereoelectronic considerations for elimination reactions (48) require that the bond between the @carbon and its attached leaving group also be aligned nearly perpendicular to the plane of the conjugated substrateinhibitor complex. The leaving group may be either above or below this plane, depending on the location of the binding site of the distal group (side chain) (49) which then determines the geometry of the coplanar C,-N and C,-Co bonds in the intermediate B , or B p The elimination reaction generates species (Dl and D A Fig. 3) which are highly reactive and prone to attack by nucleophiles at the 8-carbon or at C-4' of the cofactor (26,27). This may lead to direct covalent attachment of the inhibitor-cofactor complex to a group in the enzyme active site, or alternatively, it may cause cleavage at C-4' to release a reactive enamine which can be attacked by the enzyme-cofactor complex in an electrophilic fashion (26, 32-34).
Based on such considerations and on extensive precedent with other pyridoxal phosphate-dependent a-decarboxylases (26-31), the a-difluoromethyl diaminopimelate 1 was expected to be a potent inactivator of the meso-diaminopimelate decarboxylases. However, the total lack of irreversible or even strong competitive inhibition shows that this analog cannot bind effectively to the enzyme active site. Apparently both the plant and bacterial diaminopimelate decarboxylases enforce the stringent stereochemical requirement for the DLisomer of substrate by a "tight fit" in the region surrounding the a-carbon and do not permit replacement of the a-hydrogen by a larger group (e.g. difluoromethyl). This contrasts with the behavior of most other pyridoxal phosphate-dependent a-decarboxylases which easily accommodate an a-methyl or a-difluoromethyl group (30, 50). Elimination of a group " X at the 8-carbon of the side chain should circumvent this difficulty and lead to enzyme inhibition. The sulfoxides 2 and sulfones 3 were investigated because a number of pyridoxal phosphate-catalyzed P-eliminations of sulfur-containing groups (including thiols) are known (51, 52), and the parent meso-lanthionine 6ab is decarboxylated by diaminopimelate decarboxylase at 5% of the rate of the natural substrate (17). Although the LL-and mesoisomers of lanthionine sulfoxide 2 and sulfone 3 are competitive inhibitors (Table I), they do not show time-dependent inactivation of the decarboxylases. The failure of the desired elimination reaction may be due to poor leaving ability of the sulfur-containing group caused by lack of protonation. Alternatively, it may be due to misalignment of the bond between the @-carbon and sulfur, i.e. the distal binding site for the side chain places the sulfur in position R' or R" instead of at X in intermediate B2 (Fig. 3). If an unfavorable conformation is responsible for failure of elimination, replacement of the phydrogens of meso-diaminopimelate by halogen (e.g. fluorine) may afford a potent irreversible inhibitor of the decarboxylase. The reasons for stronger inhibitory activity of the sulfoxides 2 relative to the sulfones 3 or the parent lanthionines 6 are not clear. This effect may be due to some specific secondary binding of the sulfoxide functionality because the "wrong" DD-and LL-isomers also appear to compete effectively.
Modification of the substrate nitrogen presented an attractive alternative approach to inhibition of the decarboxylase. Cooper and Griffith (35) have observed that N-hydroxyglutamate irreversibly inhibits pyridoxal phosphate-dependent glutamate decarboxylase. These workers suggest formation of a very stable nitrone A3 (Fig. 3) in the active site as the cause of inhibition. Although N-hydroxydiaminopimelate 4 is a good competitive inhibitor of diaminopimelate decarboxylase from both wheat germ and B. sphericus, no irreversible inactivation was seen. The compound is specific for this enzyme and does not inhibit L-lysine decarboxylase. Although enzymecatalyzed decarboxylation of 4 appears unlikely, the possible occurrence of this process has not been rigorously disproven. Reversibility of the inhibition may result from binding of the N-hydroxy analog 4 in the active site by initiation of transimination with the cofactor-enzyme complex without completion of nitrone formation. Alternatively, the nitrone AS may form but may undergo rapid and reversible enzyme-catalyzed cleavage or diffusion out of the active site. In either case, the behavior of plant and bacterial diaminopimelate decarboxylases is similar and again in strong contrast to precedent with other pyridoxal phosphate-dependent enzymes.
The N-aminodiaminopimelate 5 was expected to be a very potent competitive inhibitor because hydrazino substrate analogs have inhibition constants in the range of 1-2 FM with a pyridoxal phosphate-dependent histidine decarboxylase (30) and with aspartate aminotransferase (36). Presumably a hydrazone results from interaction of such analogs with the cofactor in the active site (A4, Fig. 3). Because of the intervening nitrogen, such species are incapable of the normal reactions (e.g. decarboxylation) and are also much more stable to hydrolysis than normal substrate imines. Taking into consideration the strict stereochemical requirements of diaminopimelate decarboxylases and the fact that analog 5 is probably a statistical mixture of all stereoisomers (25% would be the correct 5a isomer), the observed inhibition constants (K; = 100 /IM for B. sphericus enzyme; Ki = 84 pM for wheat germ enzyme) indicate quite potent inhibition.
The lack of antibacterial activity of the difluoromethyl diaminopimelate 1 and the sulfur-containing analogs 2 and 3 is in accord with their inability to irreversibly inhibit mesodiaminopimelate decarboxylase. Although the N-modified an-alogs 4 and 5 are better competitive inhibitors of this enzyme, their antibiotic action is very limited in preliminary microbiological tests. Since B. megaterium is not affected by 4 and 5 in complex media but is strongly inhibited in defined media lacking L-lysine, one major difficulty appears to be rapid transport of exogenous L-lysine which circumvents the metabolic block. This could potentially be avoided by co-administration of a compound which interferes with transport of this amino acid. Another problem may be ineffective transport of the N-modified analogs 4 and 5 themselves into bacterial cells; this may account for the very limited antibiotic spectrum in defined media. It may be possible to overcome this problem by incorporation of the N-hydroxy-and N-aminodiaminopimelates 4 and 5 into dipeptides. Payne and co-workers (53-56) have shown that related peptides can enter bacterial cells more effectively, will release the toxic amino acid by hydrolysis, and will greatly enhance antibacterial activity. Studies on such dipeptides and on other inhibitors of enzymes in the diaminopimelate pathway are in progress.