INTRODUCTION

Lysyl oxidase (EC 1.4.3.13) is unique among the mammalian copper amine oxidases by catalyzing a critical post-translational modification essential to the biogenesis of connective tissue matrices. This enzyme initiates covalent cross-linking between and within the molecular units of elastin and of collagen by oxidizing peptidyl lysine in these proteins to peptidyl -aminoadipic--semialdehyde (1, 2). The peptidyl aldehyde can then condense with neighboring -amino groups or peptidyl aldehydes to form the covalent cross-linkages found in fibrillar collagen and elastin. Lysyl oxidase contains a tightly bound copper cofactor as well as a covalently bound carbonyl prosthetic group recently identified as lysine tyrosylquinone (3).

Lysyl oxidase catalyzes primary amine oxidation through a ping pong bi ter kinetic mechanism (4, 5). Following initial Schiff base formation with the LTQ¹ cofactor, the bound substrate undergoes rate-limiting, general base-facilitated -proton abstraction (6). Electrons migrating from the resulting carbanion reduce the carbonyl cofactor, followed by hydrolysis of the product imine intermediate to release the aldehyde product. The reduced enzyme, retaining the amino function of the substrate, is reoxidized by molecular oxygen to produce hydrogen peroxide and ammonia, regenerating the oxidized enzyme and completing the catalytic cycle.

The role of lysyl oxidase in the growth and repair of connective tissues has been well documented. Markedly increased levels of LO activity are observed in a variety of fibrotic diseases in which excess collagen is deposited in the affected tissues, as in models of atherosclerosis, hypertension, and liver and pulmonary fibrosis (2). The possibility that the development of fibrosis may be restricted by the specific suppression of lysyl oxidase activity has stimulated the search for selective and potent inhibitors of this enzyme. These efforts have identified mechanism-based and ground-state inhibitors, including -substituted haloethylamines (7), benzylamines substituted with electronegative para-substituents (8), and 1,2-diamines (9), each of which appear to inhibit as adducts of the carbonyl cofactor.

In the present report, we describe our observations that homocysteine thiolactone and the oxygen and selenium lactone analogues of this compound are active site-directed, irreversible inhibitors of lysyl oxidase. The selenium and sulfur lactones are the most potent of these and are selective for lysyl oxidase, among mechanistically similar copper-dependent mammalian amine oxidases tested in the present study. Notably, HCTL occurs in mammalian systems as a metabolic by-product of methyl transfer from S-adenosylhomocysteine. Moreover, the accumulation of HCTL has been suggested to be related to mechanisms of carcinogenesis and atherogenesis (10-12), whereas it has also been shown to thiolate proteins, including low density lipoproteins in vitro (13, 14). Elucidation of interactions of these compounds with lysyl oxidase should increase options for the design of antifibrotic agents and aid in the understanding of the biological effects of HCTL.

MATERIALS AND METHODS

Homocysteine thiolactone hydrochloride, homocysteine, methionine, seleno-DL-methionine, D-homoserine, homoserine lactone hydrochloride, homovanillic acid, horseradish peroxidase, diaminopentane dihydrochloride, BAPN, pyridoxal hydrochloride, [¹⁴C]NaCN, deuterated water, porcine kidney diamine oxidase, and bovine plasma amine oxidase were obtained from Sigma. Phenylhydrazine hydrochloride and Darco G-60 activated carbon were obtained from Fisher. Pyrroloquinoline quinone was a product of Fluka Chemical Corp. Hydrogen iodide, deuterium iodide, and -bromoethylamine hydrobromide were products of Aldrich. [U-¹⁴C]Phenylhydrazine was purchased from ICN Pharmaceuticals, Irvine, CA. [³⁵S]Methionine and [1-¹⁴C]iodoacetamide were purchased from Amersham Life Science. 4-Ethyl-5-(butylamino)-1,2-benzoquinone (Fig. 1) served as a model of the LTQ carbonyl cofactor of lysyl oxidase and was generously provided by Dr. Judith P. Klinman of the Departments of Chemistry and Molecular and Cell Biology, University of California, Berkeley.

Lysyl oxidase was isolated from calf aorta as a co-purified mixture of four individual variants (15) resolving as a single band at 32 kDa by SDS-PAGE (16). The individual ionic forms exhibit common structural and apparently identical mechanistic features (17, 18), further noted by the presence of a single dipeptide sequence containing the carbonyl cofactor, which was isolated from a proteolytic digest of the mixture of variants of bovine aorta lysyl oxidase (3).

Lysyl oxidase activity was assayed either against 125,000 cpm of a recombinant tropoelastin substrate expressed and labeled with L-[4,5-³H]lysine in a bacterial expression system (19) or against 500,000 cpm of L-[4,5-³H]lysine-labeled chick calvarial collagen substrate (20). Tritiated water formed by lysyl oxidase action was isolated by vacuum distillation and quantified by liquid scintillation spectrometry. Enzyme activity was also assayed against nonpeptidyl amine substrates by a horseradish peroxidase-coupled fluorescence method (21). Error ranges for all assay data, including those used to express kinetic isotope effects, were determined as the 90% confidence intervals from unweighted linear regression analyses according to Mendenhall (22).

Porcine kidney diamine oxidase and bovine plasma amine oxidase activities were determined at 37 °C at pH 7.2 or 7.4, respectively, by a horseradish peroxidase-coupled fluorescence assay modified from that used for lysyl oxidase (21). Reaction mixtures contained enzyme, 2.5 mM putrescine or benzylamine for assay of diamine oxidase or plasma amine oxidase, respectively, 40 µg of horseradish peroxidase, and 0.7 mM sodium homovanillate, in 16 mM potassium phosphate in a total volume of 2 ml. Background rates for each enzyme were determined in the presence of enzyme-saturating concentrations of inhibitors, i.e. 100 µM aminoguanidine for diamine oxidase (23) or 100 µM -bromoethylamine for plasma amine oxidase (24).

[1-¹⁴C]BAPN (55 mCi mmol¹) was synthesized from [¹⁴C]NaCN (55 mCi mmol¹) and -bromoethylamine hydrobromide. [¹⁴C]Sodium cyanide (1 mCi; 0.02 mmol) was completely dissolved in dimethyl sulfoxide, and then 1.3 µl (0.01 mmol) of triethylamine was added, followed by addition of 0.01 mmol of -bromoethylamine in dimethyl sulfoxide. The reaction mixture was stirred gently at 37 °C for 4 days to assure complete reaction. An identical reaction mixture containing [¹²C]NaCN was incubated in parallel as a control to monitor the status of the reaction by thin layer chromatography on silica gel plates using 1-butanol:acetic acid:water (4:1:1) as the mobile phase. At the completion of the reaction, concentrated HCl was added to bring the pH to 4-5 to quench traces of [¹⁴C]NaCN remaining in the solution. A Dynamax-C8 preparative scale HPLC column (Rainin Instruments, Woburn, MA) was employed for the purification of the synthesized compound. Elution gradients were generated between Buffer A (0.05% trifluoroacetic acid in water) and Buffer B (0.03% trifluoroacetic acid in acetonitrile) at 25 °C at a flow rate of 2 ml/min. Peak elution was monitored at 232 nm. Optimal separation was achieved by the addition of Buffer B to 85% from 0% within 1 min at 25 min after application of the sample to the HPLC column as indicated by TLC and UV absorption. Radioactive BAPN eluted between 45 and 48 min and was detected by monitoring fractions for radioactivity and for the appearance of BAPN by TLC (R_F value, 0.36), visualizing by reaction with ninhydrin. The final product was >95% pure, and the overall yield was 36%.

[³⁵S]HCTL was prepared from [³⁵S]methionine (82 mCi mmol¹) according to Baernstein (25) with a few modifications. Thus, 0.5 mCi of L-[³⁵S]methionine was added to 1.2 ml of 57% hydrogen iodide, and the mixture was refluxed at 137 °C for 3.5 h. The resulting solution was extracted several times with ether, the aqueous phase was diluted with water and then dried by lyophilization. The product was dissolved in hot absolute alcohol and then precipitated by addition of three volumes of pure ether. The precipitate was recovered by centrifugation and was further washed with ether. The purity of the product was confirmed by its absorption spectrum, by TLC, and by NMR spectroscopy. The overall yield was >90%. SeHCL was prepared by the same method, substituting selenomethionine for methionine.

[-²H]HCTL was synthesized in two steps designed to minimize the opening of the thiolactone ring. Racemic [-²H]methionine was prepared from L-[-¹H]methionine according to Fujihara and Schowen (26). L-Methionine (10.7 mmol) was dissolved at room temperature in 10 ml of deuterium oxide containing 20 mmol of sodium hydroxide followed by the addition of 1.0 mmol of pyridoxal hydrochloride. The mixture was then autoclaved at 15 p.s.i. steam pressure for 30 min. The mixture was then cooled on ice and adjusted to pH 4 with HCl. Continued incubation at 0 °C resulted in the precipitation of crystalline material. The crystalline precipitate was washed with cold water and methanol by filtration, the product dissolved in water containing 50 mg of Darco G-60 activated carbon and filtered. Methanol was added, and the product was precipitated by incubation at 4 °C. The crystalline -deuterated methionine was isolated by filtration and dried under vacuum. Deuterium incorporation was determined by NMR spectroscopy to be >90% of theory. Deuterated HCTL was then synthesized from the deuterated methionine and deuterium iodide (57%) according to the procedure for synthesizing [³⁵S]HCTL (25). Deuterium incorporation into the final product was estimated as >83% by NMR, and overall yield was >80%. When hydrogen iodide was employed in the second step, the deuterium incorporation decreased for unknown reasons.

D-HSL was prepared by refluxing 250 mg of D-homoserine in 2 ml of 2 N HCl for 2.5 h at 120 °C. At the completion of the reaction, the mixture was cooled to room temperature, diluted with water, and freeze-dried. The resulting solid was redissolved in a minimum amount of hot absolute alcohol, reprecipitated, and washed with pure ether. The stereochemical purity of the product was confirmed by circular dichroism, and the overall yield was >90%.

SDS-PAGE was carried out in 0.1% SDS on slab gels (14 × 10 × 0.15 cm) of 12.5% acrylamide using the Tris/glycine buffer system of Laemmli (16). Gels were stained for 30 min in 0.1% Coomassie Blue (R-250) and then destained with 40% methanol, 10% acetic acid. Gels were also treated with EN³HANCE^TM (NEN Life Science Products) according to the protocol of the manufacturer before autoradiography.

Lysyl oxidase was reacted with [1-¹⁴C]iodoacetamide by a modification of a published procedure (27). Purified bovine aorta lysyl oxidase (15 µg) was incubated at 37 °C in the presence or absence (control) of 80 µM HCTL in 16 mM potassium phosphate, 1.2 M urea, pH 7.8, in a total volume of 35 µl for 2 h, fully inactivating the enzyme incubated in the presence of the inhibitor. Solid urea as well as 10 µl of 8 M urea, 0.22 M NH₄HCO₃, pH 8.5, was then added, bringing the urea concentration to 8 M. The mixture was further incubated at 50 °C for 0.5 h to denature the protein. The mixture was cooled to room temperature, [1-¹⁴C]iodoacetamide was added at a 100-fold molar excess relative to disulfide bond content of the native protein (27), and the mixture was incubated at room temperature in the dark for 1 h. The samples were initially and intermittently flushed with nitrogen to prevent oxidation of free sulfhydryls that may have formed during the reaction. At the end of the reaction, loading buffer containing SDS in the absence of a disulfide reductant was added and the samples were subjected to SDS-PAGE immediately.

LO (15 µg) was preincubated in the presence or absence of 100 µM BAPN, 80 µM phenylhydrazine, or 60 µM HCTL in 16 mM potassium phosphate buffer, 1.2 M urea, pH 7.8, at 37 °C for 1 h and then incubated for an additional 1 h with 60 µM [³⁵S]HCTL (82 mCi mmol¹), 100 µM [1-¹⁴C]BAPN (55 mCi mmol¹), or 80 µM [U-¹⁴C]phenylhydrazine (7.5 mCi mmol¹), in the presence or absence of nonisotopic inhibitors as specified. Unbound ligands were removed by exhaustive dialysis against 2 M urea, 16 mM potassium phosphate, pH 7.8, followed by dialysis against distilled water. The dialyzed samples were concentrated in vacuo and resolved by SDS-PAGE and analyzed by autoradiography.

UV absorption spectra were recorded at 25 °C in a Hewlett Packard model 8452A diode array spectrophotometer. HCTL (95 µM) was incubated with 44 µM LO in 16 mM potassium phosphate, 6 M urea, 30 mM NaCl, pH 7.8, in a 1-cm cuvette. Scanning was initiated within 1 min of mixing of the enzyme and inhibitor and repeated every 90 s for 30 min. Difference spectra were obtained by subtracting the spectrum of the initial scan (equivalent to that of free enzyme) from those of the subsequent scans.

RESULTS

The structures of the aminolactones and of other compounds used in these studies are shown in Fig. 1. Each of the aminolactones proved to inhibit the activity of purified lysyl oxidase. As shown (Fig. 2), Lineweaver-Burk plots of initial rate assay data obtained in the presence and absence of varied concentrations of HCTL, SeHCL, or HSL yielded plots that intersect at the 1/v axis, indicative of competitive modes of inhibition in each case. The stereospecificity of the inhibition varied. Thus, the D- and L-isomers of HCTL were equally inhibitory, as indexed by the K_I values determined from Lineweaver-Burk plots (28), whereas only the L-isomer of HSL inhibited the enzyme. The individual isomers of SeHCL were not available, and all data obtained with this compound were determined with the racemic mixture.

[View Larger Version of this Image (18K GIF file)]

[View Larger Version of this Image (18K GIF file)]

The reversible or irreversible nature of the inhibition was determined by preincubating lysyl oxidase at 37 °C with varied concentrations of each lactone followed by assay for residual activity by the peroxidase-coupled assay method. Aliquots of the preincubated enzyme-inhibitor mixtures were sufficiently diluted to reduce lactone levels to noninhibitory concentrations in the assay. Each of the lactones caused the time-dependent loss of lysyl oxidase activity, which was not reversed by dilution. Graphing the data as the logarithm of the percent remaining activity against time of preincubation resulted in a series of linear plots with increasing negative slope at increasing inhibitor concentrations (data not shown), indicating that the development of irreversibility followed first order kinetics. Inhibition was also seen to be fully irreversible by exhaustive dialysis of enzyme preincubated with sufficient concentrations of these inhibitors to fully inhibit activity. Secondary plots of the t1/2 values derived from the first order plots against 1/[lactone] were also linear (Fig. 3), consistent with site-saturation interactions between the inhibitors and the enzyme. This kinetic behavior is consistent with the mechanism summarized in Scheme I (29), according to which irreversible inhibition develops by conversion of the lactone, initially bound in a reversible EI complex, to a covalently linked, inactivated enzyme-inhibitor complex ([EI]*, Scheme I). The inhibitor dissociation constant, K_I, is equal to k₁/k₁, whereas k₂ is the limiting first order rate constant for inactivation (29). The first order rate of inactivation caused by 25 µM HCTL decreased 10-fold from 0.1 min¹ to 0.01 min¹ when 5 mM n-hexylamine, a productive substrate for lysyl oxidase, was present in the enzyme-inhibitor preincubation mixture, consistent with the competitive kinetics obtained (see Fig. 2) and with interaction of lactone inhibitors at the active site.

[View Larger Version of this Image (16K GIF file)]

Table I summarizes the constants derived from these kinetic studies with the three lactones. As seen, SeHCL has the greatest affinity for the enzyme, with a K_I value 2.5- and 50-fold less than those of HCTL and HSL, respectively. In contrast, the k₂ values of the three lactones are similar, although there is a tendency for the value of k₂ to increase in the order SeHCL < HCTL < HSL. The second order rate constants (k₂/K_I) are predominantly influenced by the K_I values (Table I).

Table I. Lactone constants Constant HSL (O) HCTL (S) Se-HCL (Se) KI (L) 420  ± 56 µM 21  ± 3 µM 8.3  ± 2.2 µM k2 (L) 0.28  ± 0.01 min1 0.18  ± 0.01 min1 0.12  ± 0.03 min1 k2/KI 11 s1M1 143 s1M1 241 s1M1

As shown in the top panel of Fig. 3, substitution of the -proton of HCTL with a deuteron has a marked effect on the t1/2 for inactivation, determined at 37 °C, and, thus, on the k₂ value for this thiolactone. The k₂ at 37 °C for [-¹H]HCTL is 0.18 min¹, whereas that for [ -²H]HCTL, prepared and tested as the racemate, is 0.055 min¹, giving a kinetic isotope effect of 3.3-fold. The magnitude of this deuterium kinetic isotope effect is similar to the primary kinetic isotope effects found for k_cat with productive substrates undergoing enzyme-catalyzed -proton abstraction, including n-butylamine (^Dk_cat, 4.3 at 37 °C; Ref. 9) and tyramine (^Dk_cat, 2.6 at 55 °C; Ref. 30). While similar (31), smaller (32) and significantly larger (33) ^Dk_cat values have been found reflecting -proton abstraction steps in other copper-dependent amine oxidases, the values for lysyl oxidase suggest that additional rate contributing steps may exist that partially suppress the isotope effect of the isotope-sensitive proton abstraction step. In toto, these results point to the conclusion that the bound lactone undergoes -proton abstraction as a rate-limiting step in the development of irreversible inhibition. The substitution of the -proton with a deuteron did not change the apparent K_I, as determined from Lineweaver-Burk plots of assays performed at 37 °C, resulting in a ^DK_I of 1 and a ^Dk₂/K_I equal to that for ^Dk₂. Thus, the K_I for the protonated lactone was 21.6 ± 0.5 µM, whereas that for the deuterated lactone was 22 ± 0.7 µM. These K_I values are presumed to reflect the true dissociation constants predominantly while largely neglecting the contribution of k₂. Thus, k₂ for HCTL is relatively slow (0.18 min¹), so that only a minor fraction (10%) of the available enzyme became irreversibly inactivated under the initial rate assay conditions employed and in the less than saturating concentration of HCTL used in the determination of K_I .

There was no apparent enzyme turnover as indexed by the lack of H₂O₂ production in incubations of lysyl oxidase with HCTL, indicating that HCTL is not a productive substrate for lysyl oxidase. Indeed, simple alkylamines stemming from secondary carbon atoms, as in 1-ethylpropylamine, were also found not to be substrates for lysyl oxidase (data not shown). Moreover, neither N-acetyl-HCTL, -amino--butyrolactone (structures shown in Fig. 1), nor DL-homocysteine inhibited lysyl oxidase at concentrations 1 mM, indicating the essentiality of the -amino group and the five-membered lactone structure in the inhibition by these compounds.

The inhibitory effect of HCTL on LO was also confirmed by assays against protein substrates of lysyl oxidase, yielding IC₅₀ values of 30 ± 3 µM and 220 ± 20 µM, using tropoelastin and collagen, respectively, as substrates. In comparison to the value of 23.0 ± 3.6 µM obtained with 1,5-diaminopentane as substrate, spatially extended interactions between the enzyme and its collagen substrate previously described (34) may account for the significantly decreased sensitivity noted with this substrate.

The specificity of the inhibition of selected copper-dependent amine oxidases by the selenium and sulfur lactones was investigated in view of the apparent similarity among the mechanisms of action of these enzymes (35). As shown (Fig. 4), HCTL and SeHCL are each highly selective for lysyl oxidase among these catalysts. Although LO lost all activity at 100 µM HCTL and at 10 µM SeHCL, the activities of plasma amine oxidase and diamine oxidase were only slightly affected at these concentrations of these lactones.

[View Larger Version of this Image (23K GIF file)]

HCTL can covalently derivatize proteins via aminolysis of the thiol ester by protein - and/or -amino groups (13, 14, 36), resulting in the gain of free sulfhydryl groups due to the opening of the thiolactone ring. To clarify whether this is relevant to the inhibition of LO by HCTL, the native enzyme was modified by HCTL as described and then reacted with [1-¹⁴C]iodoacetamide, using the isotopically labeled alkylating reagent to detect sulfhydryl groups newly introduced by reaction of the enzyme with the thiolactone. Autoradiography of SDS-PAGE analyses of the modified enzyme (Fig. 5) indicated that the low background level of [1-¹⁴C]iodoacetamide associated with the native enzyme was not significantly changed by the modification with HCTL. Prior reduction of the enzyme disulfide bonds by treatment of the native enzyme with 2-mercaptoethanol or dithiothreitol markedly increased incorporation of the isotope, consistent with the alkylation of the newly available cysteine residues. These data indicate that HCTL does not introduce free SH functions and are consistent with previous observations that each of the cysteine residues of native lysyl oxidase exists in disulfide linkage (15). The preparation of lysyl oxidase used in this experiment included a co-purified band at ~24 kDa, previously identified as a tyrosine-rich acidic matrix protein (37) and which was removed from LO preparations used in all other experiments of this study (27). As shown, HCTL also does not significantly alter the available sulfhydryl content of this protein.

1

2

3

[View Larger Version of this Image (46K GIF file)]

SDS-PAGE of lysyl oxidase previously incubated with [³⁵S]HCTL (60 µM; 82 mCi mmol¹) revealed that the band of lysyl oxidase resolving on the denaturing gel at 32 kDa retained radioactivity, thus providing direct evidence that HCTL is covalently linked to the inhibited enzyme. The incorporation of [³⁵S]HCTL into the enzyme correlating with complete inactivation was determined to be 0.45 ± 0.05 mol/mol of enzyme after removing unbound reagent by dialysis against 8 M urea. Notably, a maximum of 0.4-0.5 mol of [¹⁴C]phenylhydrazine/mol of LO are incorporated, and this correlates with full inactivation of the enzyme (3). These molar labeling values suggest that, at maximum, 50% of the purified enzyme molecules are catalytically functional, possibly reflecting the net oxidized carbonyl cofactor content. Competitive labeling experiments were carried out with HCTL and other chemical inactivators known to derivatize the active site (Fig. 6). As shown (lanes 1-4), prior denaturation of native LO or preincubation of native LO with unlabeled BAPN or PH prevented the subsequent incorporation of radioactive [³⁵S]HCTL. Similarly, LO preincubated with nonisotopic HCTL failed to incorporate radioactivity derived from [¹⁴C]phenylhydrazine (lanes 5 and 6) or [¹⁴C]BAPN (lanes 7 and 8). Nonisotopic PH also prevented the incorporation of [¹⁴C]BAPN. Apparently, HCTL, BAPN, and PH covalently modify the same or overlapping site(s) in the enzyme. In the case of PH, this site is known to be the LTQ cofactor (3). Moreover, the covalent modification of the active site by [³⁵S]HCTL requires the native structure of the enzyme.

1

2

[View Larger Version of this Image (31K GIF file)]

The possibility that the competing active site labels might have hydrolyzed HCTL by aminolysis prior to its reaction with lysyl oxidase was negated by spectroscopic assessment for the presence of the intact thiolactone (_max, 238 nm) following incubation with these competing agents in parallel in the absence of lysyl oxidase.

The perturbation of the chromophoric properties of LO and of the LTQ model by HCTL is shown as difference spectra in Fig. 7. The addition of HCTL to LO caused the time-dependent bleaching of the broad absorbance band measured at 520 nm characteristic of the oxidized state of the LTQ cofactor (3). Increases in the absorption bands at 290 nm and 390 nm developed simultaneously with the decrease at 520 nm. Closely similar spectral changes occurred upon addition of 5 mM HCTL to free LTQ under the same conditions, in that the absorption band at 520 nm was gradually bleached and the bands at 280 and 390 nm developed with time (Fig. 7).

[View Larger Version of this Image (33K GIF file)]

DISCUSSION

The present study introduces five-membered -aminolactones as a new class of irreversible inhibitor of lysyl oxidase. In toto, the protection against inactivation of the enzyme by a productive substrate, the competitive kinetics obtained, the mutually exclusive labeling of the enzyme by isotopic HCTL, PH, and BAPN, and the perturbation of the near UV-visible spectrum of lysyl oxidase by HCTL indicate that HCTL is an active site-directed covalent inhibitor of lysyl oxidase. The spectral perturbations and the competitive relationship between PH and HCTL specifically invoke the LTQ carbonyl cofactor as a site of interaction of HCTL with the enzyme, whereas the first order kinetics of enzyme inactivation are consistent with rate-limiting, enzyme-mediated processing of an initially reversible enzyme-HCTL complex to the inactivated, covalent complex. It is likely that rate-limiting -proton abstraction is involved in this process, in view of the significant primary deuterium kinetic isotope effect on the rate of inactivation. Thus, these kinetic and chemical characteristics of the inhibition categorize HCTL as a mechanism-based inactivator of lysyl oxidase. Although not all of these experimental criteria were applied to the oxygen or selenolactones, it seems highly likely that these are also mechanism-based inactivators and that all three of these competitive, irreversibly inhibitory lactones inactivate the enzyme by the same mechanism, particularly in view of the similar values of the first order rates of inactivation and their nearly identical chemistries.

The amino nitrogen, the carbonyl oxygen, and the internally esterified sulfur atoms of HCTL collectively might ligate the copper atom of the enzyme. However, such coordinate interactions would not account for the covalent, irreversible inhibition of the enzyme. Moreover, the lack of alkylatable sulfhydryl content in the enzyme inactivated by HCTL argues against the possibility that HCTL acylates a critical residue through its carbonyl function.

In view of the evidence that HCTL interacts with the carbonyl cofactor and appears to undergo -proton abstraction, a mechanistic pathway of inhibition is indicated the initial steps of which parallel those in the mechanism accounting for the oxidation of productive substrates by lysyl oxidase (5). As proposed in Scheme II, initial Schiff base formation with the cofactor (I) is followed by -proton abstraction and tautomerization of the resulting carbanion (II), thus following stages of processing of productive substrates (4, 5). The resulting conjugation between the reduced cofactor and the carbonyl oxygen of the bound lactone should then favor the stabilization of the complex by delocalization of electrons, illustrated here as the resonance pair, III and IV, irreversibly linking the lactone to the enzyme. Although an oxyanion is represented at carbon 4 of the cofactor-inhibitor adduct (III), the pK of this functional group at the active site is not known, and this oxygen may be at least partially protonated.

Scheme II.

As noted, the apparent affinities of the three lactone inhibitors for LO differ significantly, as reflected by the K_I values, with that for SeHCL approximating 1/3 and 1/50 of those for HCTL and L-HSL, respectively. Since the k₂ values span the comparatively small range of 0.12 (SeHCL) to 0.28 min¹ (HSL), the differences in the apparent second order rate constants for inactivation (k₂/K_I) predominantly reflect differences in binding affinities. Examination of the computer optimized three-dimensional structures of these lactones indicates that the configuration of the noninhibitory D-isomer of HSL differs significantly from those of SeHCL and HCTL. The D-isomers of the three lactones are arranged in the top row of Fig. 8 in overlapping fashion, aligned so that the -carbons, -protons, and the amino functions approximate the same positions in space as do the bonds between the -carbon and the carbonyl carbon for each. The S, Se, and O atoms of the different lactones project away from the viewer and the carbonyl oxygens project toward the viewer in this representation. These spatial restrictions reflect the strong likelihood that the binding and processing of each lactone requires precise orientations between the amino group of the lactone and the susceptible carbonyl of the LTQ cofactor and between the -proton of the lactone and the general base previously implicated in -proton abstraction in LO catalysis (6). It is apparent that the lactone ring of D-HSL is more "kinked" in the upward direction than those of HCTL and SeHCL. Inspection of SeHCL indicates that the five atoms of its lactone ring most closely approximate a planar relationship with each other, among these compounds. These configurational differences are consistent with the decreasing covalent radius of the Se (1.17 Å), S (1.03 Å), and O atoms (0.7 Å) increasing the strain on the five-membered lactone ring in the order SeHCL < HCTL < HSL. Since SeHCL has the highest affinity for the enzyme, this suggests that binding is inhibited by unfavorable interactions between the enzyme and those edges of the lactones that deviate most from planarity. Notably, the inhibitory L-isomer of HSL, shown in bold at the bottom of Fig. 8, deviates from planarity in the downward direction when the -amino and -proton are restricted as defined above, whereas the noninhibitory D-isomer deviates upward. These relationships between inhibitor potency and configuration of the lactone rings suggest that the microenvironment of the binding site for these inhibitors in LO conflicts with the edge of the lactone ring most distal to the -carbon, particularly when that ring projects in a cis relationship to the -amino group. The differences in electronegativity values for the O, S, and Se atoms (3.5, 2.44, and 2.48, respectively) likely do not contribute much to the inhibitory effects since the k₂ values are nearly independent of the ring heteroatom, as noted.

Structures of -aminolactones.

[View Larger Version of this Image (38K GIF file)]

The metabolic relationship between homocysteine and HCTL, as well as the possible impact of each on human disease, has been a subject of considerable investigative effort. Congenital homocystinuria, in which plasma homocysteine can increase from the normal level of 10 µM to as high as 400 µM, can derive from homozygous defects in the gene for cystathionine--synthase, a pyridoxal 5-phosphate-dependent enzyme catalyzing transulfuration between homocysteine and serine to yield cysteine, ammonia, and -ketobutyrate. Homocystinemia can also arise from defects in cobalamin-dependent metabolism, as seen in homozygous deficiency of the gene coding for methylenetetrahydrofolate reductase, the enzyme participating in the synthesis of methylated vitamin B₁₂. Transfer of the methyl group from S-adenosylmethionine converts this methyl donor to S-adenosylhomocysteine, from which free homocysteine is then released. HCTL can arise by a metabolic error-correcting process in which homocysteine is mis-activated by methionyl tRNA synthetase to form enzyme-bound homocysteinyl adenylate, from which homocysteine is released as free HCTL (38). Increased levels of homocysteine in homocystinuria have been correlated with premature atherosclerosis and thrombosis (39-41), and this has been suggested to reflect the thiolation of amino groups of low density lipoprotein (12). An initial finding of high levels of HCTL in human serum (42) contrasts with later reports noting its apparent absence in human serum (43, 44), possibly due to the presence of a thiolactonase enzyme activity identified by Dudman et al. (43). Nevertheless, in view of the reasonably high affinity of HCTL for LO (K_I, 21 µM), it remains possible that even modestly low equilibrium levels of HCTL might occur in congenital homocystinuria that could diminish levels of functional LO. In that regard, homocystinuria is accompanied by connective tissue defects. Indeed, lysine-derived cross-link content was significantly reduced (~3-fold) in type I collagen of homocystinuric patients (45). By the same token, vitamin B₆ (pyridoxine) deficiency results in decreased levels of cross-links in elastin and collagen. Although this was initially thought to be due to the presence of a pyridoxal or pyridoxal 5-phosphate cofactor in LO (46), more recent studies have identified the LO cofactor as peptidyl LTQ, as noted (3). Myers et al. (47) made the reasonable suggestion that free homocysteine, accumulating due to the resulting deficiency in vitamin B₆-dependent cystathionine -synthase activity in B₆-deficient states, may react to form thiazolidine derivatives of aldehydes stemming from LO action on peptidyl lysine, thus inhibiting condensation of peptidyl aldehydes to form mature, stable cross-links. The present results now raise the additional possibility that the connective tissue defects in homocystinuria and in vitamin B₆ deficiency may stem directly from the inhibition of LO by HCTL derived from homocysteine.

In view of the central importance of lysyl oxidase in connective tissue development and repair, and noting evidence that it may have additional functions on biology, including the suppression of ras-induced tumorigenesis (48) and as a chemotactic agent (49), the possibility that HCTL, a naturally occurring molecule, may suppress the activity of this catalyst in vivo has important biological implications.