NMR and Computer Modeling Studies of the Conformations of

Vol . 259,No. 18, Issue of September 25, pp. 1143611447,1984 Printed in U.S. A.

T H E JOURNAL OF BIOLOGICAL CHEMISTRY 0 1984 by The American Society of Biological Chemists, Inc

NMR and Computer ModelingStudies of the Conformations of Glutathione Derivatives at the Active Site of Glyoxalase I* (Received for publication, February 21, 1984)

Paul R. RosevearSQ, Siv Sellinll, Bengt Mannervikll, Irwin D. KuntzII, and Albert S. Mildvan$ From the $Department of Physiological Chemistry, Johns Hopkins University School of Medicine, Baltimore, Maryland 21205, the VDepartment of Biochemistry, Arrhenius Laboratory, University of Stockholm, S-10691 Stockholm, Sweden, and the ((Departmentof Pharmucetltical Chemistry, University ofCalifornia, San Francisco, California 94143

The conformations of four derivativesof glutathione bound at the active site of the metalloenzyme glyoxalase I have been determined by NMR measurements and by computer model building using a distance geometry aDDroach. Paramagnetic effects of Mn2+-glyoxa1aseI on-the longitudinal relaxation rates of the carbonbound protons of the substrate analog S-(acetony1)glutathione at threefrequencies, the hydrophobic competitive inhibitor S-(propy1)glutathione at four frequencies, and thecharged competitive inhibitor &(carboxymethy1)glutathione at a single frequency were used to calculate Mn2+to proton distances in each complex. These and previously determined distances from Mn2+to the protons and 13C-enrichedcarbon atoms of the product S-(D-lactoyl)glutathionewere used in a distance geometry program to compute the conformations of each enzyme-bound derivative which best fit the measured distances and other known constraints such as bond lengths, van der Waals radii, planar and trans-peptide bonds, and thioester linkages. The distance geometry program also provided a measure of the uniqueness of the conformations consistent with the experimental data.Extended Y-shaped conformations were detected for each of the bound glutathione derivatives, similarto the x-ray structure and the theoretically calculated conformation of glutathione itself, suggesting this to be a low energy form. Acceptable conformations of each enzyme-bound derivative fell into two classes with the metal either above or below the mean plane through the glutathione compound. The conformational uncertainty within each class was relatively small, ranging from deviations of 0.9-1.9 A in the averagepositions of each of the atoms. A small but significant difference in the conformation of the substrateanalog as compared to the product was detected in the position of the reaction center carbon directly bonded to the glutathione sulfuratom. Unlike the second-sphere metal complexes formed by the bound substrate analog, the product, or thehydropho-

bic competitive inhibitor, the chargedcompetitive inhibitor S-(carboxymethy1)glutathione binds farther from the metal, in the thirdcoordination sphere.

Glyoxalase I, S-(D-lactoyl)glutathionemethylglyoxal-lyase (isomerizing) (EC 4.4.1.5) is the first of two enzymes in the glyoxalase system. The postulated metabolic role of this system is the detoxication of reactive a-ketoaldehydes such as methylglyoxal,formed from methylglyoxal synthase (1,2), amine oxidase (3), and side reactions of triosephosphate isomerase (4) and aldolase (5), toyield metabolically inert ahydroxycarboxylic acids such as D-laCtate. Glutathione functions as a coenzyme in this system. Glyoxalase I is a dimeric Zn’+-metalloenzyme (Mr= 46,000), which catalyzes the isomerization of the hemimercaptal adduct of methylglyoxal and glutathione to form thelactoyl-thioester: OH

0

OH

I1

I

CH-C-CH”S-G

0 Gly I

I

II

+CH,-CH-C--S-G

(1)

The mechanism of isomerization has been shown to involve an intramolecular proton transfer, presumably with a n enediol intermediate (6-8). The two Zn2+atoms per dimer canbe replaced by Mn2+ or Co2+ with specific activities of approximately 66% of the native Zn2+-enzyme(9). Optical and EPR spectral studies of the Co2+-substituted enzyme (10) and EXAFS studies of the Zn2+-enzyme(11) revealed distorted octahedral coordination of the metal. TheEXAFSstudiesalso gave evidencefor imidazole ligands donated by the protein to the Zn2+ (11). Nuclear relaxation studies demonstrated 2 metal-bound water molecules(12). Furthermore, measured distances from the enzyme-bound Mn2+and Co2+ to the carbon-bound protons of the product S-(D-lactoyl)glutathioneestablished a kinetically competent second sphere enzyme-metal(1igand)-product complex (13). Distances from the enzyme-bound Mn2+ to the 13C-enriched lactoyl carbonyl andhydroxymethylene carbons of the product constrained the lactoyl carbonyl group to point * This investigation was supported by National Institutes of Health toward the metal, with the carbonyl oxygen positioned to Grants AM 28616 and GM 19267 and National Science Foundation accept a hydrogen bond from an intervening ligand (14).Such Grant PCM 8121355, and by grants from the Swedish Natural Science a structure would facilitate polarizationof the carbonyl group Research Council. Support for the computation was provided by National Institutes of Health Grant CA-22780. The 360 MHz studies in the reverse glyoxalase reaction (13, 15). The intervening were done at the Middle Atlantic Regional NMR Facility which is metal ligand may be the water molecule that is immobilized supported by National Institutes of Health Grant RR-542. A prelim- when S-(D-lactoyl)glutathioneis bound to the active site of inary report of this work has been published (Rosevear, P. R., Sellin, the enzyme (12, 13). S., Mannervik, B., and Mildvan, A. S. (1982) Fed. Proc. 41, 1152). X-ray studies have determined the conformationof glutaThe costs of publication of this article were defrayed in part by the thione itself (16) and of glutathione bound to the active site payment of page charges. This article must therefore be hereby marked “aduertisement” in accordance with 18 U.S.C. Section 1734 of glutathione peroxidase (17) as well as of GSSG bound to glutathione reductase in the crystalline state (18,19). Howsolely to indicate this fact. ever, no information exists on the conformation of enzymePostdoctoral Fellow of the American Cancer Society (PF 2111).

11436

Conformation of Glutathione Derivatives bound glutathione on glyoxalase I or on any other enzyme in solution. Our previous nuclear relaxation studies hadfocussed on the structure of the reaction center, the lactoyl-thioester portion of the enzyme-product complex, and did notconsider the conformation of the glutathione moiety, until additional NMR data on several glutathione derivatives were obtained. Such data arenow available. The present NMR and model-building study was undertaken to determine and compare the overall conformations of four enzyme-bound glutathione derivatives on glyoxalase I: the product, S-(D-lactoyl)glutathione,a substrate analog, S(acetonyl)glutathione, a hydrophobic competitive inhibitor, S-(propyl)glutathione, and a charged competitive inhibitor, S-(carboxymethy1)glutathione.Distances from Mn2+ at the active site of the enzyme to selected nuclei of these glutathione derivatives were determined by the paramagnetic probe TI method (13, 14). These distances were then used to build models of these compounds by hand and by computer, using a distancegeometry program which calculates molecular conformations from known and measured interatomic distances (20, 21).

on Glyoxalase I

11437

methylene and methyl protons, relaxation rates of the carbon-bound protons of S-(acetony1)glutathione were measured in 80% HzOwhich precluded measurements a t 60 MHz.Selective proton irradiationwas used to suppress the residual HzO signal at 100, 250, and 360 MHz. Relaxation rates of the protonsof S-(carboxymethy1)glutathionewere measured only at 250 MHz. Longitudinal relaxation rates were determined using the 90" (homogeneity spoil)-r-90' pulse sequence (25) at 60 and 100 MHz. The 180"-~-90"(25) and the saturation recovery (26) pulse sequences were used at the higher frequencies and gave indistinguishable results. Transverse relaxation rates were calculated from line width measurements at half-height ( A v ) using the relation 1/Tz = T A V . The relaxation rates of the carbon-bound protons of S-(D-lactoy1)glutathione (13), S-(acetonyl)glutathione,S-(propyl)glutathione, and S-(carboxymethy1)glutathionewere measured in the presence of Mn2+-substitutedglyoxalase I or with the native Zn2+-enzyme,which served as the diamagnetic control. The paramagnetic contributions to the longitudinal (l/Tlp) and transverse (l/Tu) relaxation rates were calculated as the difference in rates measured in the presence of equal concentrations of paramagnetic and diamagnetic enzymes. At the 1000-fold concentration ratios of glutathione derivatives to enzyme used in these studies, no significant diamagnetic effects of the Zn2+-enzymeon l/T1 or 1/T2 of the glutathione compounds were detected. Paramagnetic contributions to the longitudinal and transverse relaxation rates were used to obtain metal to proton distances and lower limit-exchange rates, respectively, using the general theory EXPERIMENTAL PROCEDURES reviewed elsewhere (25, 27). Distance Geometry Method-The distance geometry approach of Materials Kuntz and Crippen (20,21) was used to determine the conformations Glyoxalase I from human erythrocyteswas purified to homogeneity of the glutathione derivatives which best fit the measured MnZ+as previously described, and a mixture of the three isoenzymes was nucleus distances, and to evaluate the uniqueness of these conforused (22). Preparation of the metal-free apoenzyme and reconstitu- mations. The algorithm calculates the conformation of the molecule tion with either ZnZ+,Mn2+, or Co2+ were performed as described using known and experimentally determined distance constraints. previously (9). S-(D-1actoyl)glutathionewas prepared as described by The constraints are utilized to obtaina random trial set of all Ball and Vander Jagt (23), whereas S-(acetonyl)glutathione,S-(pro- intramolecular distances. These distances are then used to obtain a pyl)glutathione, S-(carboxymethyl)glutathione, and S-@-bromoben- set of atomic coordinates (X,Y, and 2) which are optimized to give zy1)glutathione were prepared by modifications of method A of Vince the best agreement possible with the initial distance constraints. The et al. (24). degree of agreement of a given structure with the initial constraints Immersible concentrators with a molecular weight cut-off of 10,OOO is expressed quantitatively by a score which measures the fourth daltonsobtained from Millipore Corp. (model CX) wereused to power of the deviation of the distances within the structure from concentrate and deuterate the enzyme solutions. 'H20, obtained from those given by the initial constraints. This process is repeated and a Stohler Isotope Chemicals, Inc., was distilled before use. Deuterated number of structural solutions are obtained. Acceptable solutions Tris base was also obtained from Stohler. All other chemicals were found by the distance geometry program are then compared by direct of the highest purity commercially available and were used without examination on a Vector General computer graphics system, and by further purification. computer, using the HOMOLOGY program (28) which optimally superimposes one rigid structure ontoanother by rotation and transMethods lation, minimizing the differences in atomic positions. Such comparEnzyme Preparation and Assay-The activity of glyoxalase I was isons permitted qualitative and quantitative evaluation of the uniquemeasured using a spectrophotometric assay with methylglyoxal as the ness of the conformation of a givenmolecule as defined by the 2-oxoaldehyde substrate (6, 9). Protein concentrations were deter- measured distances. The use of the HOMOLOGY program also mined by absorbance a t 280 nm (c = 0.80 mg/ml"). Before use in the permitted detailed comparisons of the optimal conformation of the NMR experiments, glyoxalase I was concentrated, using Millipore product S-(D-lactoyl)glutathionewith that of the substrate analog Simmersible CX concentrators, and diluted 10-fold with 10 mM Tris- (acetony1)glutathione and the conformation of hydrophobic analog S-(propy1)glutathione with that of the charged analog S-(carboxyC1, pH 7.6, twice to remove the 10% methanol contained in the enzyme storage buffer. Deuteration of glyoxalase I was achieved by methy1)glutathione. Since several modifications of the distance geometry approach four 10-fold dilutions and concentrations using 10 mM deuterated Tris-C1 in 'HzO, pH* 7.4.' All buffer and reagent solutions used in were required for the high resolution search of the conformations of the NMR studies were treated with Chelex 100 before use to remove the glutathione derivatives, this method willnowbe described in greater detail. In the conformational search procedure, each atom of trace metals. Magnetic Resonance Studies-Proton NMR spectra were obtained a given glutathione derivative is represented by a point at a known at 25 "C and at 60, 100, 250, and 360 MHz with the Varian NV-60, distance from other atoms, including Mn2+.If the distance between Varian XL-100, Bruker WM 250, and Bruker WH 360 FT spectrom- any two atoms not directly bonded to each other is unknown, its eters, respectively. Sample volumes of0.35 ml were used in 5-mm lower bound is set by the sum of the van der Waals radii. Its upper NMR tubes, which typically contained 10 mM deuterated Tris-C1, bound is set at 25 A to permit the conformation of the molecule to pH* 7.4, a derivative of glutathione (20 mM), glyoxalase I (0-70 p~ vary freely, but within the total size constraint of the molecule as defined by known bond lengths and other experimental measuresites), and 'H20. Spectra a t 250 MHz were obtained with 16 bit A/D conversion. A line broadening of 2 Hzwas applied to all spectra ments. obtained at 60 MHz. Such filtering was not used a t higher frequencies. To obtain correct bonding or 1-2 distances and bond angles, which Chemical shifts were calculated from the external reference sodium are expressed as 1-3 distances, as well as reasonable values for 4,4-dimethyl-4-silapentane-l-sulfonate. After the NMR experiments, dihedral angles, which are expressed as 1-4 distances, initial S t N C which occasionally lasted up to 12 h, glyoxalase I was found to retain tures of each of the glutathione derivatives were built from existing a t least 90% of its initial activity. x-ray data. This was accomplished by linking the x-ray structure of Longitudinal (1/T1) and transverse (l/Tz)relaxation rates of the glutathione itself (16) to that of appropriate molecular fragments carbon-bound protons of S-(propyl)glutathione, were obtained a t 60, from other x-ray structures using the program DOCK (29). The D100, 250, and 360 MHz. To prevent the exchange of the acetonyl lactoyl moiety was obtained from the atoms C(2), C(3), and C(6) of the x-ray structure of MnZ+-citratedecahydrate (30), the acetonyl from the atoms C(6), C(7), and C(8) of lactoylcholine iodide (31), the The abbreviation used is: pH*, pH measured in 2Hz0. propyl from the atoms C(4), C(5), and C(6) of potassium caproate

11438

Conformation Glutathione of Derivatives on

(32), and thecarboxymethyl from the atoms C(1) and C(2) of acetylcarnitine hydrochloride monohydrate (33). Additional hydrogen atoms were generated where necessary using the program CALCAT (34). The unusually short 1.19 A C(9)-C(lO)bond found in the x-ray structure of glutathione (16) was changed to themore appropriate CC single bond length of 1.54 8, (35). All other 1-2 bond lengths were then obtained from these structures and fixed at their respective values with errors of 0.05 A. Bond angles of methylene and methyl groups were defined by setting geminal interproton distances at 1.75 and 1.74 A, respectively, as determined by neutron diffraction (36). To insure tetrahedral sp3 methyl and methylene carbon atoms, it was found necessary to set the errors of the 1-3 distances between pairs of H atoms at 0.01 A. Other bond angles were fixed by using the 1-3 distances observed in the x-ray structures, usually with an error of 0.1 A. All dihedral angles were initially set by the 1-4 pistances in the x-ray structures and were allowed to vary by +1.3 A, to allow a complete conformational search. For the enzyme-bound glutathione derivatives, the two peptide bonds were fixed to be trans and planar as was found in the x-ray structure of glutathione itself (16) and as is generally found for peptide bonds inproteins (37). For trans peptide bonds, the 1-4 distances from the amide proton to thecarbonyl oxygen and between the two (Y carbons were fixed to thevalues found in the x-ray structure of glutathione with errors of +0.1 A. To ensure planarity of the peptide bonds and thecarboxylate groups, a subroutine was added to the algorithm which minimizes the weighted sum of the squares of the distances from the best mean plane through the atoms defined in that plane.* The thioester group of S-(D-lactoyl)glutathionewas also defined to be planar as was found by a comparison of the available x-ray data on thioesters (38). Although trans thioesters have been shown to be of lower energy than cis (38), the distance constraints imposed on the thioester group were left sufficiently free to allow eithera trans or cis planarthioester group in the final derived structure. The asymmetric a-carbons of the L-glutamyl and L-cysteinyl residues of glutathione and the hydroxymethylene carbon of the D-laCtOyl moiety of S-(D-lactoyl)ghtathionewere constrained to have the correct chirality by an additional penalty term based on quantitative deviations from the chiral volume. The chiral volume, Vch, for each asymmetric carbon atom was calculated using Equation 2.

Glyoxalase I

nuclear relaxation measurements(13,14) and are summarized in Table IV. The 250 MHz 'H NMR spectra of the other three glutathione derivatives(20 mM) are given in Fig. 1. The spectrum of S-(acetony1)glutathione (Fig. l.4)was obtained in 80% H 2 0 t oprevent exchange of the acetonyl methylene protons. The intense water signalwas suppressed by continuous protonirradiation.Spectra of S-(propy1)glutathione (Fig. 1B) and S-(carboxymethy1)glutathione (Fig. IC) were obtained in 'H20. The resonances of the three glutathione derivatives were assigned (Table I) by selective homonuclear decoupling, by comparison withS-(D-1actoyl)glutathione(13), coupling data on amino + and with chemical shift and spin-spin acids and peptides (41). The Gly(C,H2) protons give a singletresonance due to deuteration of the Gly(NH) proton. Small differences in the chemical shifts of the two gly(C,Hn) protons were observed in S-(propy1)glutathione and S-(carboxymethy1)glutathione (Fig. 1, B and C; Table I). Broadening of the Gly(C,H2) resonance in the S-(acetony1)glutathioneanalog, presumably due tocoupling to the NH proton in 80% H20, prevented the possible detection of chemical shiftnonequivalence. Chemical shift differences in the Gly(C,H2) protons havepreviously been observed in glutathioneat p H values lower than 7.5 (42). In all glutathione derivatives the Glu(C,H) resonance, expected to be a triplet resulting from a 3J,p= 8.0 Hz (41, 43), was found by selective homonuclear decoupling to be under the Gly(C,H2) resonance. The individual Cys(C,H2) protons are resolved, and coupled to the C,H giving rise to an ABX pattern (Fig. 1; Table I) in all three glutathione derivatives as was previously observed for S-(D-1actoyl)glutathione (13). The downfield Cys(C&-proton is labeled A and the upfield proton is labeled B (Fig. 1). The stereochemical assignments of the Cys(COHP) protons in oxytocin have beenaccomplished using stereoselective deuteration (40). Using the sameassignV c h = (V,-V,).[(Vz-V4) x (V3-VdI (2) ments for the Cys(CpH2) protons in the glutathione derivaWhere V, through V, are the position vectors of the atoms directly tives, the A and B resonances correspond to the pro-R and bonded to the asymmetric carbon atom, labeled in ascending rank pro-S, protons, respectively. The methylene protons (C( l)Hz) carboxymethylene moiety of S-(carboxymethyl)according to theCahn-Ingold-Prelog system (21,39). Negative values of the of v , h indicate the R configuration and positive values denote the s glutathione are also individually resolved at 250 MHz (Fig. configuration. The chirality penalty term was also used to define the 1C; Table I). chirality of the pro-R and pro-S protons of the cysteinyl CBH2moiety, Paramagnetic Effects of Mn2+-GlyowalaseI on S-(Acetonylh, since distances from Mn2+to each of these protons could be individS-(Propyl)-, and S-(Carboxymethy1)glutathione-Paramagually measured. The pro-R and p r o 3 proton resonances of the CBH2 group of cysteine have been previously assigned by selective deuter- netic effects of Mn'+-glyoxalase I on the glutathione analogs ation (40). The chiral volumes used in the distance geometry algo- are exemplified by,the broadening of the proton resonances rithm for the a-carbons of L-glutamyl and L-cysteinyl, the CpHz of S-(propy1)glutathione (19.6 mM) upon the additionof 18.6 carbon of the cysteinyl moiety, and the hydroxymethylene carbon of ,.LM Mn2+-glyoxalase I (Fig. 2B) whencomparedwiththe the D-laCtOyl group were 6.9, 8.2,7.5, and -7.8, respectively. proton line widths in the absence of added Mn2+-glyoxalaseI All of the experimentally measured Mn2+to proton and Mn2+to carbon (14) distances with their respective errors were used to define (Fig. 2 A ) . This increase in line widths is dominated by the the position of the Mn2+and to determine the conformation of the paramagnetic effects on l/Tz. Paramagnetic effects on 1/T1, enzyme-bound glutathione derivative. In cases where only a lower of the glutathione derivatives (16.3-19.6 mM), were deterlimit Mn2+to proton distancewas obtained, the distance was entered mined by detailed titrations of each derivative with Mn2+as such into the program. In the S-(acetony1)glutathione structure an glyoxalase 1 (0-56.9 pM), using titrations with Zn2+-glyoxalase additional constraint was placed on the carbonyl group of the acetonyl I (0-70.0 p ~ under ) identical conditions as the diamagnetic moiety, oiz. that the oxygen point toward the metal atom as was found experimentally for the lactoyl carbonyl oxygen (14). This was controls (Fig. 3). From the inhibitor constants for S-(acetony1)glutathione accomplished by- setting an upper limit Mn2+ to carbonyl oxygen (KI= 1.5 2 0.3 mM), S-(propy1)glutathione (& = 0.32 f 0.04 distance of 5.1 A. The mechanistic reasons for this constraint are discussed. mM), and S-(carboxymethy1)glutathione (KI = 1.3 f 0.2 mM (44)) it could be shown that theactive site of the enzyme was RESULTS AND DISCUSSION 94-98% saturatedwiththeselinear competitive inhibitors normalized paramag'H N M R of S-(Acetonyl)glutathione, S-(Propyl)glutathion, under the experimental conditions. The (1/ and S-(Carboxymethy1)glutathione-Distances from Mn2+ at netic contributions to the longitudinal relaxation rates (l/fTzP)obtained the active siteof glyoxalase I to 7 protons and2 carbon atoms fTlp)(Fig. 3) and transverse relaxation rates of S-(D-lactoyl)glutathione were previously determined by from these titrations using thenegligibly small effects of the Zn2+-enzyme on the relaxation rates as the respective diamagnetic controls, are summarized in Table 11. When the excess overthe enzyme the largest We are grateful to Dr. L. Mario Amzel for adding this subroutine substrate analog is in great value of l/fT,, for each analog (Table11) may be used to set to theprogram.

Conformation of Glutathione Derivatives on Glyoxalase 1

I

A

4.0 FIG. 1. Proton NMR spectra of glutathione derivatives. A, S-(acetony1)glutathione (20 mM) in 20% 'H20 containing 10 mM deuterated Tris-C1, pH* 7.4, in a total volume of 0.35 ml. B, S-(propy1)glutathione (20 mM) in *H20. Other components are as described in A . C, S"(carboxymethy1)glutathione (20 mM) in 'H,O. Other components are as described in A . NMR spectra were obtained at 250 MHz hy using 32 transients, each of 8192 data points, an acquisition time of 1.4 s, a delay time of 15 s, a spectral width of 3000 Hz, quadrature phase detection, 16 bit A/D conversion, a 90" pulse, and saturation by irradiation of the residual HzO signal. T = 25 "C. Assignments are asgiven in Table I except for 1-1 which is an unidentified impurity.

5.5

11439

1-1

3:O 2.5 2:o CHEMICAL SHIFT 8 (ppm)

1:5

Gly (CaHz)

1-1

2:o

C

2:5 4.0 3:O

GIY

3:5

CHEMICAL SHIFT 8 (ppm)

1.5

1 .o

methylene

Glu

Glu

a lower limit to its exchange rate ( ~ / T , , J into the paramagnetic tathione, respectively, by more than an order of magnitude environment (25). The l/fTzp value of the acetonyl(C(1)HJ (Table 11), indicating that the l/fTlpvalues of each of these protons (7.1 ? 0.6 X lo4s-', Table 11)exceeds all l/fT, values analogs are not exchange-limited. A similarly fast exchange of the S-(acetony1)glutathione protons by more than anorder rate at 25 "C of the product S-(D-1actoyl)glutathione(21.9 X of magnitude (Table 11),indicating that the l/fTlpvalues are lo4 s-' (13)) waspreviously estimated by the temperature not limited by the exchange rate of S-(acetony1)glutathione dependence of l/fTzp into the paramagnetic environment. Thus, the l/fTlp values Displacement Experimentswith S-(p-Bromobenzy1)glutacan be used to calculate Mn2+ to proton distances. Similarly,thione-To establish active site binding of the glutathione the l/fTzp values of the propyl(C(3)Hz)(1.9 +- 0.3 X lo4 s-', analogs under the conditions of the NMR experiments, and Table 11) and of theCys(CBHB) of S-(carboxymethy1)glu- to estimate the outer sphere contributions to l/fTlp and 1/ tathione (5.0 k 1.0 x lo3s-l, Table 11) exceed all the l/fTlp fTzp (13,14,27), displacement of the glutathioneanalogs were values of S-(propy1)glutathione and S-(carboxymethy1)glu- carried out using the tight-binding competitive inhibitor S-

11440

Conformation of Glutathione Derivatives onGlyoxalase I TABLEI Proton chemical shifts and spin-spin coupling constants forS-(acetonyl)glutathione, S-(propyl)glutathwn, and S-fcarboxvmethvlldutathione Glutathione derivatives S-(Acetony1)glutathione

Resonance

Chemical shift" (ppm from DSSc)

S4Propyl)glutathione

Spin-spin coupling constantsb 3Jm#

Chemical shift' (ppm from DSS')

Others

S-(Carboxymethy1)glutathione

Spin-spin coupling constantsb

Chemical shift" (ppm from

DSS')

Others

"@

-d

4.58

4.57

3.75

3.77' 3.79

3.72' 3.77 3.07 2.91 2.53 2.17 3.24h 3.25

5.2 3.01 7.9 2.84 2.52 2.13 3.60

'JAB

= 13.9 3.09

4.8 8.5

= 13.9 2.87 'Joy =:2.55 7.4 'Jp, =: 2.16 7.4

2~~

2.59

Spin-spin coupling constantab

3J.s

Others

'J,,(A)= 4.8 'Jep(~) = 8.5

' J m = 13.6

'5-

= 13.6 'Joy = 6.8 'Jp7 '16.8

1.59 0.95

2.29

Chemical shifts are accurate to within k0.02 ppm. Coupling constants are accurate to within k0.5 Hz. DSS, sodium 4,4 dimethyl-4-silapentane-1-sulfonate. Not resolved due to the large HZ0 resonance. e Protons of Gly(C,H,) were individually resolved (see text). 'Assigned to the individually resolved Cys(CBHz)methylene protons (see Fig. 1). Carbon atoms of the acetonyl, propyl, and carboxymethyl moieties are numbered as shown C(3)-C(Z)-C(l)-Scys. The methylene protons of the carboxymethyl moiety are individually resolved (see Fig. IC).

A GJUICTHZ) Propyl HJ

I,

1

CyS(CaH)

FIG. 2. Paramagnetic effects of Mn2+-glyoxalase I on the carbonbound protons of S-(propy1)glutathione. A, proton NMR spectrum at 250 MHz of S-(propy1)glutathione (19.6 mM). B, components as in A with the addition of Mn'+-glyoxalase I (18.6 pM sites), C, components as in B, with the addition of 529 pM S-@-bromobenzyl)glutathione. Conditions and components are otherwise as described in the legend to Fig. 1.

"---

I

0

1,

A

I

C

/J

45

40

3'5

30

25

20

I '5

10

05

Chemlcal S h l f t 6 (PPMI

(p-bromobenzy1)glutathione. Previous titrations of the product S-(D-lactoyl)glutathione complex of Mn2+-glyoxa1aseI with this inhibitor, measuring the decrease in 1/T* of the methyl protons of the product and assuming competition between the product and the inhibitor yielded a KO value for the inhibitor (1.8 +- 0.4 p~ (13))in good agreement with that found by direct titration of the Mn2+-enzymewith the inhibitor (1.1 k 0.6 &M (12, 45, 46)),and with the kinetically determined KI value (1.7 -+ 0.2 pM (46)). The addition of 529 HM S-(p-bromobenzy1)glutathioneto a

solution containing 18.6 PM Mn2+-glyoxaIaseI and 19.6 mM S-(propy1)glutathione decreased the paramagnetic effect on the 1/T2as measured by a decrease in the paramagnetic line broadening (Fig. 2, B and C ) . Decreases in the paramagnetic effects on the l/Tl of 88 k 9, 81 -+ 9, and 88 +- 15% were observed for S-(acetonyl)glutathione,S-(propyl)glutathione, and S-(carboxymethyI)glutathione,respectively, upon addition of 529 FM S-(p-bromobenzy1)glutathione.From the inhibitor constants of the glutathione analogs and the dissociation constant for S-(p-bromobenzy1)gIutathione(KD= 1.8

*

Glyoxalase I

Conformation of Glutathione Derivatives on

11441

by the expected amounts, upon saturation with the competi0.4 ~ L M(13)) reductions of the paramagnetic effects on the longitudinal relaxation rates of 94 f 4, 83 k 10, and 95 f 5% tiveinhibitor S-(p-bromobenzy1)glutathione indicate active for S-(acetonyl)glutathione, S-(propyl)glutathione,and S- site binding of the glutathione analogs under the conditions (carboxymethyl)glutathione,respectively, were expected. The of the NMR experiments, and negligibly small outer sphere profound reductions of the paramagnetic effects of Mn2+- contributions to the relaxation rates. Determination of CorrelationTimesandDistancesfrom of the glutathioneanalogs, glyoxalase I on the relaxation rates

0

5

10 [ZnZ+lb (pM)

-

15

0

10

20

30

40

50

60

70

[Znz+]b ( p M )

FIG. 3. Effects of Mn2+-and Zn2+-substitutedglyoxalase I on the longitudinal relaxation rates of the carbon-bound proton resonances of glutathione derivatives. A, titration of S(acetony1)glutathione (19.6 mM) with Mn2+-glyoxalase I (upper) or Zn-glyoxalase I (lower). B, titration of S-(propy1)glutathione (19.6 mM) with Mn2+-glyoxalase I (upper) or Zn2+-glyoxalaseI (lower).C, titration of S-(carboxymethy1)glutathione (16.3 mM) withMn2+glyoxalase I (upper)or Zn2+-glyoxalaseI (lower). Broken lines indicate the effects of addingthe competitive inhibitorS-(p-bromobenzy1)glutathione (529 PM). Concentrationsof [Mn2+Ib and [Zn*+]brefer to the enzyme site concentrations of glyoxalase I. Error bars on each point represent the average error obtained by computeranalysis. Proton NMR spectra were obtained a t 250 MHz in 10 rnM deuterated Tris-CI buffer, pH' 7.4, in 'HzO using 16 transients with 8192 data points, spectral width of 3000 Hz, an acquisition time of 1.4 s, 16 bit A/D conversion, quadrature detection, a saturation-recovery pulse sequence, and saturation of the residual Hz0 signal by continuous irradiation. T = 25 "C.

TABLEI1 Paramagnetic effects of Mn2+-glyoxalase Z on the protons of S-(acetonyOglutathione, S-(propyoglutathione, and S-(carboxymethyl)glutathione at 250 M H z These parameters are calculatedfrom the titrationsof Fig. 3 where the conditions aregiven. S"(Acetony1)glutathione

S-(Propy1)glutathione -___

S-(Carhoxymethy1)glutathione ~~

Resonance Resonance llfT1, llfTzP 1NTm 1lfTzP __ _ .1lfTlp 5260 80 f 60 3,000 f 1,000 5220 Gb(C,HJ G~Y(C,HJ NDb 920 f 80 Cys(C,H) 170 f 40 Cys(C,H) 1,300 f 200 15,000 f 5,000 Cys(CPHA) 660 f 80 240 5z 70 CYS(CBHA) 3,100 f 150 130 f 100 5,000 i 1,000 CY~CPHB) 3,000 f 200 14,000 f 5,000 Cys(C,H,) 5320 220 f 100 Glu(C&z) 5170 Glu(CoH4 Glu(C-,HJ 520 f 120 440 f 60 Glu(C,Hz) 100 & 40 3,200 i 200 71,000 +- 6,000 Propyl(C(l)Hz) 1,600 f 60 2,000 f 1,000 Carboxymethyl(CH2)190 2 40 660 f 50 36,000 -C 5,000 Propyl(C(2)H.J 1,900 f 100 Propyl(C(S)H,) 1,200 f 80 19,000 f 3,000 a

Line broadening could not be accurately determined due tocomplex spin coupling. Not determined due to overlap with large solvent signal.

11442

Conformation of Glutathione Derivatives

on Glyoxalase I

Mn2+ to S-(Acetonyl)glutathione, S-(Propyl)glutathione, and S-(Carboxymethyl)glutathione on Glyoxalase I-As discussed elsewhere (27), when the rate of ligandexchange into the paramagnetic environment is rapid compared to l/fTlp and the outer sphere contribution to l/fT, is small, as shown abovefor the glyoxalase I-Mn2+-glutathioneanalog com( r )to the plexes, the equation relating Mn2+-proton distances normalized relaxation time UTl,) simplifies to the following form:

previously found tohave a correlation time ( T J of 0.74 f 0.17 ns and anf(7J value of 0.92 f 0.03 ns at 250 MHz (13). The narrow range of 7= values for the S-(D-laCtOyl),S-(acetonyl), and S-(propy1)glutathione complexes indicate that 7, is not very sensitive to the precise nature of the glutathione derivative. This mutual consistency is even greater for the f(r,) values of the three complexes (0.92 f 0.03, 0.85 f 0.08, and 0.70 f 0.24 ns, respectively) which agree within experimental error. Since the relaxation rates of the glyoxalase I-Mn2+-S(carboxymethy1)glutathione complex were measured only a t 250 MHz, the f ( 7 J of this complex was reasonably assumed to be the average of the f ( 7 , ) values of the three other complexes for which this parameter was measured (0.82 ? 0.30 In Equation3, C is a product of physical constants numerically ns). Using this average ~ ( T Jvalue and thel/fTlpvalues (Table equal to 812 .Idf3 for Mn2+-proton interactions, q is the II), distances from boundMn2' to the carbon-bound protons relative stoichiometry of Mn2+ and the glutathione derivative bound to glyoxalase I, 7 , is the correlation timefor the Mn2+proton dipolar interaction, andW Iand us are the nuclear and electron precession frequencies. The q values for the glutathione analogswere taken as 1.0 based on their linear competitive inhibition of the glyoxalase I reaction (44)3and by their essentiallycomplete displacement from Mn2+-glyoxalase I by the potent competitive inhibitor S-(p-bromobenzy1)glutathione which is known to bind with 1:l stoichiometry (12,45, 46). The correlation times(7J in Equation 3 were calculated for the Mn2+-glyoxalaseI-glutathione analogcomplexes from the frequency dependence of their l/fTl, valuesaccording to Equations 3 and 4.

TABLE111 Correlation timesof glyoxalase I-Mn2+-analogcomplexes The solutions contained 10 mM deuterated Tris-C1 buffer in'H20, pH* = 7.4. S-(Acetony1)glutathione (19.3 mM) with either 7.5 pM Zn2+-or Mnz+-substituted glyoxalase I sites. S-(Propy1)glutathione (19.3 mM) with either 12.2 p~ Zn2+-or Mn2+-glyoxalaseI sites. Resonances of glutathione analogs

A.-C. Aronsson, S. Sellin, and B. Mannervik, unpublished work.

j(7.)

B

100 250 360 100 250 360 100 250 360

0.90 0.94 0.89 0.54 1.30 2.19 0.81 0.89 0.98

2.05 0.89 0.49 1.46 0.76 0.26 1.93 0.90 0.50

1.8

1.3

0.60

0.89

1.4

1.8

60 100 250 360 60 100 250 360 100 250 360

2.60 1.75 2.52 6.29 0.65 0.73 1.06 0.41 1.04 0.84 0.67

3.98 0.61 0.45 0.09 1.835.3 0.95 0.84 0.87 0.85 0.92 0.61

0.46

4.4

MHz

S-(Acetony1)glutathione Glu(CpHz) Glu(C,Hz)

Equation 4 describes the frequencydependence of T ~ the , longitudinalelectronspinrelaxationtime of Mn2+, which generally dominates 7 , in macromolecular complexes of Mn2+ (25,27,47). The constantB depends on the zero-field splitting of the parameters, and 7,is the time constant for distortions symmetry at the metal ion. Values of B, 7,, and at each frequency, calculated by computer fitting of the frequency dependence of l/fT1, to Equations 3 and 4 (Table 111), are within the range expected for the enzyme-bound Mn2+ ion (25). The presence of 80% H,O in the S-(acetony1)glutathione complex precluded the measurement of relaxation rates at 60 MHz. The individual correlation times at 250 MHz for the Glu(CoHn), Glu(C,H2), and acetonyl (C(3)H3) protons (Table 111) were averaged to yield a 7, = 1.04 f 0.22 ns for the glyoxalase I-Mn2+-S-(acetony1)glutathionecomplex. This range of 7c values yielded a much narrower rangeof values of thecorrelationfunction ( f ( 7 , ) = 0.85 f 0.08 ns) which is defined as those terms in the inner parenthesis of Equation 3. From f ( 7 J and the l/fTlpvalues (Table 11), distances from enzyme-bound Mn2+ to the carbon-bound protons of S-(acetony1)glutathione were calculated using Equation 3 (Table IV). Errors in the distances reflect the measured errors in both l/fTlp and in f ( 7 J (Table IV). For the glyoxalase I-Mn2+-S-(propy1)glutathionecomplex, the individual correlation timesat 250 MHz of the Gly(C,Hz), propyl(C(2)Hz), and propyl(C(3)H2) resonances (Table 111) were averaged to yield a T = = 1.47 f 0.9 ns. This range of T~ values yields an ~ ( T J= 0.70 f 0.24 ns. From the l/fTlp values (Table 11) and the f ( 7 J value, distances from the enzymebound Mn2+ to the carbon-bound protons of S-(propy1)glutathione were calculated using Equation 3 (Table IV). The glyoxalase I-Mn2+-S-(D-lactoyl)glutathionecomplex was

T.

Frequency

Acetonyl(C(3)H3)

1.1

1.2

3.8

~

TABLEIV Distances on Mn2+-substitutedglyoxalase I from Mn2+ tonuclei of bound S-(D-lactoyl)glutathione,S-(acetonyl)glutathione, S-(propyl)glutathione, andS-(carboxymethy1)glutathione Glutathione analogs Glutathione and substituent nuclei

s-(D-LaCtoyl)'

S-(AceS-(Protonyl)

PYU

S-(Carboxymethyl)

r(A) r14 r(& 6'4 210.0 11.2 f 2.8 11.5 f 0.7 29.9 7.7 + 0.6 10.6 + 0.8 7.8 + 0.3 9.2 f 0.4 8.5 f 0.4 7.3 f 0.5 10.0 f 0.7 6.3 f 0.4 11.0 f 2.0 8.0 f 0.3 6.5 f 0.3 9.8 i 1.4 210.6 29.6 12.0 f 1.2 11.0 f 0.8 9.1 k 0.7 8.5 t 0.8 11.5 +- 1.0 6.5 f 0.3 7.1 f 0.5 10.4 f 1.0 6.9 f 0.4 8.5 f 0.2 8.4 f 0.5 7.4 f 0.5

5.7 + 0.3 6.1 f 0.5

"For S-(D-Lactoy1)glutathioneMn2+to proton distances are from Ref. 13 and Mn2+to carbon distances are from Ref. 14.

Conformation of Glutathione Derivatives on Glyoxalase 1

11443

Q ACTOn

FIG. 4. Alternative conforma- @ tions of enzyme-bound S-(D-lactoy1)glutathione on glyoxalase I computed by the distance geometry program, using the experimentally determined distances from Mn2+ (Table IV). The best fit structure (Class 11) is on the right. I-GLU

I ~

of S-(carboxymethy1ene)glutathione were calculated (Table

IV). In the enzyme complexes of all of the derivatives, the errors LACTOYL IF!. IV) reflect the measured in theabsolutedistances(Table errors in both l/fT,, (Table 11) and f ( T e ) . The errors in the distances arerelatively small sincea sixth root is taken in the Q distance calculations (Equation 3). Distances from Mn2+ to the carbon-bound protons of S-(D-lactoyl)-, S-(acetonyl)-, and S-(propy1)glutathione analogs ranged from6.3 to 12.0 A (Tacomplexes rather ble IV),consistent with second-sphere metal than inner-spherecomplexes of the enzyme-bound Mn2+. The GLY distances from Mn2+ to the protons .Of S-(carboxymethyl)glutathione range from 10.0 to 11.5 A (Table IV) resulting from the much smaller l / f T , values in this complex. These distances suggest a third-sphere complex of this analog. Computer Model Building of the GlutathioneDerivatives Using the Measured Mn2+to Proton and Carbon DistancesThe experimentally measured distances from the bound Mn2+ on glyoxalase I totheprotonsandcarbonatoms of the IV) were utilized together with the glutathione analogs (Table previouslydiscussed constraintsinthedistance geometry FIG. 5. Comparison of the two most different computed algorithm to determine the conformationof each bound ana- Class I1 conformations of enzyme-bound S-(D-1actoyl)gluof acceptable solutions as defined log and its uniqueness. For the important product of the tathione, showing the range glyoxalase 1reaction, S-(D-lactoyl)glutathione,110 structures by the experimental distances. were obtained from the distance geometry algorithm using in Fig. 4 with the metal closer to the reader in bothcases, to previously determinedMn2+toproton (13) andMn2+to carbon distances (14) (Table IV). Of the 110 structures, only emphasize their difference. These two classes of solutions 13 were foundtohave a suitably low deviation from the were alsofound for the S-(acetony1)-, S-(propyl)-, and Sconstrained distancesat the end of the minimization routine (carboxymethy1)glutathionederivatives. Since this ambiguity (520 A‘). Structureshaving a largerscore (220 A‘) were cannot be resolved by the present data, for brevity, further found to contain unsuitablyhigh-energy interactions, such as discussion of the conformations of the bound glutathione inappropriate bond lengths or excessive atomic overlap well derivatives will arbitrarily consider the class of structures exemplified on the rightside of Fig. 4 (Class 11). However, all withinthesums of theirvanderWaalsradii,andthese structures were not further considered. Each of the 13 struc- conclusions apply equally to the Class I structures illustrated tures having a low score was carefully examined and found to on the leftof Fig. 4. While it mayreasonably be assumed that of the same contain no atypical bond lengths, bondangles, van der Waal all glutathione derivatives bind in conformations class, none of our conclusions depend critically on this asradii overlaps, or deviations from the experimental MnZ+ to proton and Mn2+ to carbon distances. As expected when only sumption. Conformation of theEnzyme-bound Product,S-(D-Laca single reference point is available (25), acceptable solutions for the S-(D-1actoyl)glutathione structure were found to fall toy1)glutathione-The best Class I1 solution to theconformainto two classes, distinguishable by the positionof the metal, tion of enzyme-bound S-(D-1actoyl)glutathionewith a total deviation of 4.0 A4 from the constrained distances is shown with the Mn2+ being either in front of or behind the mean of the 44 plane of theglutathione moiety. Both of these classes of in Fig. 4. Thisdeviationisequivalenttoeach solutions for the product S-(D-lactoyl)glutathioneare shown covalent bond lengths in S-(D-lactoyl)ghtathione having a n average error of only 0.03 A. Each of the acceptable S-(D-lactoyl)ghtathione structures These distances are expressed to the nearest 0.01 A since they are based on differences between well defined coordinates. The mag- was superimposed upon another of the same class using the nitude of the differences reflect the uncertainties in the data. HOMOLOGY program. The Mn2+ atoms from both structures

11444

Conformation of Glutathione Derivatives on

Glyoxalase I

were weighted to force them to superimpose as a point of Mn2+ (Mn2+ tocarbonyl oxygen distance 55.2 A). Although reference. The root-squared deviation between the positions acceptable solutions could be found withthe carbonyl oxygen of corresponding atoms of the compared structures was used pointing away from the Mn2+,this additional constraint was to evaluate the difference between any two solutions. Fig. 5 added to insure that thesolutions from the distance geometry compares the two most different Class I1 structural solutions algorithm were consistent with the carbonyl orientation found of S-(D-lactoyl)glutathione showing the maximum uncer- experimentally in the product S-(D-lactoyl)glutathione comtainty in the conformation as defined by the experimental plex (13, 14). Such an orientation of the carbonyl group of Mn2+ to proton and Mn2+ to carbon distances. The total thesubstrate would facilitate its polarization and enediol deviation in atomic positions for this 45-atom structure is formationin the reaction mechanism (12-15). Of the 20 84.84 or 1.88 A per atom.4 As may be seen in Fig. 5 , regions solutions, 11were of suitably low error to be used in defining of the S-(D-lactoyl)glutathionestructure which are well con- the conformational uncertainty of S-(acetony1)glutathione strained by many experimentally measured distances (the D- bound to Mn2+-substitutedglyoxalase I. The larger yield of lactoyl and cysteinyl moieties) have a small conformational acceptable solutions for the S-(acetony1)glutathione derivauncertainty. Thus, short distances are found between the tive (55%) compared to the product, S-(D-1actoyl)glutathione corresponding Cys-C, carbons (0.54 A), the Cys-sulfur (0.38 (12%),may reflect the greater number of distances measured A), and the lactoyl C(l), C(2), and C(3) carbons (0.46, 0.57, to S-(wlactoy1)glutathione which provided additional conand 0.57 A, respectively) of the two extreme ~ t r u c t u r e s . ~straints on the product complex (Table IV). These extra However, the C(1) carboxylate and the Cz carbons of gluta- constraints may have resulted in more local minima being mateare poorly constrained by the lack of experimental reached which could not be further improved by refinement. distances to them, anddiffer in position in the two structures The best Class I1 structure of S-(acetony1)glutathionewith by 4.67 and 3.00 A, respectively. a total deviation of 5.2 A4 is shown in Fig. 7A. The distance As seen in Figs. 4 and 5 , the conformation of S-(D-lac- from Mn2+ to the carbonyl oxygen was found to be 5.1 A, toy1)glutathione bound to Mn2+-glyoxalaseI is extended and consistent with the requirement that it be(5.2A. In all Y-shaped with no intramolecular hydrogen bonds. The lactoyl structural solutions, this distance was never found to be less carbonyl group in theS-(D-1actoyl)glutathionestructure (Fig. than 5.0 A which is well beyond the Mn2+ to oxygen distance 4) points toward the enzyme-bound metal with the carbonyl of 2.08-2.37 A expected for direct coordination (13).Comparoxygen4.8 A from the metal as previously determined by ison of the two most different acceptable solutions of Shand model-building studies (14). As previously discussed, (acetony1)glutathione (Fig. 7B) shows the relatively small this distance is beyond the range of Mn2+to oxygen distances conformational unFertaintyaveraging 1.30 A per atom with a (2.08-2.37 A) expected for direct carbonyl coordination, but range of 0.1-5.39 A, as defined by the experimental Mn2+to is appropriate for an intervening ligand such as water (14). proton distances and the reasonable additional constraints Distances from the carbonyl oxygen to the enzyme-bound introduced. Small distances were found between correspondMn2+were found to range from 4.6 to 5.5 A in the acceptable ing atoms of the superimposed structures in regions well S-(D-lactoyl)glutathionestructures generated by the distance defined by the NMR data such as theCys-Cocarbon (0.19 A), geometry algorithm. All of these distances are consistent with the Cys-sulfur (0.15 A ) and the C(l),C(2), and C(3) carbons an intervening ligand of the MnZ+ hydrogen bonded to the of the acetonyl side chain (0.84,0.43, and 0.96 A, respectively). carbonyl oxygen. Such a hydrogen bond would transmit the As found with the enzyme-bound product, all computed electrophilic effect of the Mn2+to the lactoyl carbonyl group, structures of the bound substrate analog S-(acetony1)glufacilitating its polarization in the reversal of the glyoxalase I tathione have an extended Y-shaped conformation (Fig. 7B). reaction (12-14). The distance from Mn2+ tothe lactoyl To test for possible conformational differences between the hydroxyl oxygen of6.2 A found by the distance geometry bound substrateand product, acceptable solutions of the algorithm is also well outside the range expected for direct Mn2+to oxygen coordination. In all acceptable structures, the thioester group was found to be tram, as has been found in a survey of the known crystallographic structures containing the thioester group (38). Several acceptable S-(D-lactoyl)glutathioneconformations were compared with the crystallographic structure of glutathione itself which is also extended (16). Comparison of the x-ray structure of glutathione with the corresponding 35 atoms of the best S-(D-lactoyl)glutathione conformation on glyoxalase I (Fig. 6) yields an average deviation of only 1.58 8, per atom, and a range of0.40-3.43 A. This similarity suggests that extended conformations of glutathione may be at or near an energy minimum and that theenzyme does not greatly alter theglutathione conformation. Conformation of the Enzyme-bound Substrate Analog S(Acet0nyl)glutathione-If one of the two acetonyl C(l)H2 methylene protons of S-(acetony1)glutathione were replaced with a hydroxyl group, it would be thesubstrate of the glyoxalase I reaction. The distance geometry algorithm was used with the experimentally determinedMn2+ to proton distances in the enzyme complex of the substrate analog (Table IV). Twenty solutionswere obtained from the distance FIG. 6. Comparison of the best fit conformation of S-(Dgeometry program using the experimental Mn2+ to proton 1actoyl)glutathione on glyoxalase I as determined by NMR distances (Table IV) together with the additional constraint and distance geometry (filled bonds),with the x-ray structure that the acetonyl carbonyl group should point toward the of glutathione (open b o d ) (16).

Conformation Glutathione of

ACETONYL A

0 %

FIG. 7. Computed conformations of enzyme-bound S-(acetony1)glutathione based on the experimental distances of Table IV. A, best fit Class I1 conformation; B , comparison of the two most different Class I1 conformations showing the range of acceptable solutions.

Derivatives on Glyoxalase I

11445

substrate andproduct bindto differentprotein conformations (12). Conformation of Enzyme-bound S-fPropyl)glutathionDistances from Mn2+ to the protons of S-(propy1)glutathione were utilized as constraints in the distance geometry algorithm to evaluate the conformation of this hydrophobic glutathione derivative on glyoxalase I. From the 30 solutions obtained with the distance geometry algorithm, 13 had acceptably low deviations from the imposed constraints to be usable in defining the uniqueness of the conformation of S(propy1)glutathione. The best Class I1 solution of S-(propy1)glutathione with a deviation of 1.79 A" is shown in Fig. 9A. Again, the glutathione analog is shown to bind to Mn2+substituted glyoxalase I with an extended Y-conformation. The range of possible conformations of enzyme bound S (propy1)glutathione consistent with theMn2+-proton distances (Table IV) was determined by comparing the two most different but acceptable computed solutions (Fig. 9B). The average deviation of the two structures is 1.50 A per atom with distances from corresponding atoms ranging from 0.07 to 5.51 A. Sizable differences in the positions of the Cys-C, carbon (1.90 A), the Cys-sulfur (2.56 A), and the C(l), C(2), and C(3) carbons of the propyl substituent (2.35, 2.28, and 2.76 A, respectively) in the two structures indicate that the conformations of the cysteinyl and S-propyl moieties of the bound analog are not precisely defined. This ambiguity, which is surprising in view of the accurately determined distances from Mn2+ to thecarbon-bound protons of the cysteine and propyl groups (Table IV), was not initially detected by manual model-building but could easily be reproduced by stick models when the computed conformations were available. The conformational uncertainty of bound S-(propy1)glutathione may reflect the numerous allowed orientations of a flexible side chain containing no rigid sp2 carbon atoms or substituents larger than hydrogen. It is emphasized that this conformational ambiguity does not necessarily reflect mobility of the cysteinyl and propyl groups of the enzyme-bound inhibitor, but merely indicates that their positions are not precisely determined by the present data. Conformation of Enzyme-bound S-(Carb0xymethyl)glutathione-In contrast with hydrophobic derivatives of glutathione, charged species, such as S-(carboxymethy1)glutathione are weaker competitive inhibitors (44). The distances from Mn2+-substitutedglyoxalase I to the protons of S-(carboxy-

substrate analog and theproduct were compared by computer as described above. A comparison of the best fit conformation of S-(D-1actoyl)glutathionewith that conformation of S-(acetony1)glutathione most like it (Fig. 8) revfals differences in the positions of the Cys-C, carbon (0.92 A), the Cys-sulfur (1.46 A), the C(l), C(2), and C(3) carbons of the substituents at sulfur (1.64, 0.48, and 0.55A respectively), and of the GluC, (1.92 A). Of these differences in position, only that of the C( 1)of the substituentexceeded the uncertaintyin its location (k1.03 A) as measured by the range of distances of this atom from Mn2+in all acceptable solutions of both acetonyl and S(D-1actoyl)glutathione.This change in position of C(1) of the substituent suggests a conformational difference between the substrate and product at the reaction center. The ability of glyoxalase I to accommodate ligands with differing conformations at thisatomic position is relevant to the observation that thisenzyme apparently can bind and utilize both configurations of the thiohemiacetal as substrates (48). A conformational change of the enzyme during catalysis has been FIG. 8. Comparison of the best fit Class I1 conformation of suggested on the basis of observed protein fluorescence S-(n-1actoyl)glutathionewith that of the substrate analog Schanges (12, 46). An alternative explanation is therefore that (acetony1)glutathionemost like it.

Conformation of Glutathione Derivatives

11446

on Glyoxalase I

Y-GLU

A

+

A

P Mn

FIG.9. Computed conformations of enzyme-bound 5'-(propy1)glutathionebased on the experimental distances of Table IV. A, best fit Class I1 conformation; B, comparison of the two most different Class I1 conformations, showing the range of acceptable solutions. methy1)glutathione (Table IV) were used inthe distance geometry program to obtain 30 structures of which only four were acceptable. The best Class I1 solution with a deviation from the constraintsof 1.09 A4 is shown in Fig. 1OA. Although all of the protons of the cysteinyl moiety and its S-carboxymethyl substituent are farther from the Mn2+than the corresponding protons of the otherglutathione derivatives (Table IV), the conformation of bound S-(carboxymethy1)glutathione is also extended and Y-shaped (Fig. 10A). A comparison of the two most different Class I1 structures (Fig. 10B) shows, the conformation to be reasonably well defined by the experimentally measured Mn2+to proton distances (Fig. IOB), with an average deviation of 0.93 A per atom and a range of 0.322.35 A. The hydrophobic analog, S-(propyl)glutathione, is compared with the charged analog, S-(carboxymethy1)glutathione in Fig. 11. The larger Mn2+ to proton distances in the S(carboxymethy1)cysteine moiety demonstrate that this group is positioned farther away from the Mn2+ than is the S (propy1)cysteine (Fig. l l ) , as well as the S-(D-laCtOyl),and

FIG.10. Computed conformations of enzyme-bound S-(carboxymethy1)glutathionebased on the experimental distances of Table IV.A, best fit Class I1 conformation; B, comparison of the two most different Class I1 conformations, showing the range of acceptable solutions. Note that a different view is shown.

'*

*

I-GLU

FIG.11. Comparison of the best fit conformations of S-(carboxymethy1)glutathione and S-(propyl)glutathione, and of their positions with respect to the enzyme-bound Mn2+ on glyoxalase I. The Mn2+is closer to the viewer than are the glutathione derivatives.

Conformation of Glutathione Derivatives on the S-(acetony1)cysteine moieties (Table IV). Thus, if the other derivatives are considered to form second-sphere complexes of the enzyme-bound Mn2+,the S-substituted cysteine of S-(carboxymethy1)glutathioneis located in the third coordination sphere of Mn2+-glyoxalase I. Despite the fact that S(carboxymethy1)glutathione bindsto glyoxalase I differently from the product and from the other derivatives, it retains the extended Y-shaped conformation. The similarity in conformation of the product, S-(D-lactoy1)glutathione andthe other glutathione derivatives bound at the active site of glyoxalase I demonstrates the stability of the extendedY-shapedconformation of glutathionecompounds. A similar conformation of unbound glutathione has beendetermined by x-raydiffraction analysis (Fig. 6) (16) and by theoretical quantum mechanical calculations (49). In combination with the previous results from x-ray studies of glutathione peroxidase (17) and glutathione reductase (18), the present data suggestthat enzyme binding sites for glutathione are designed to accommodate a low-energy conformation of the peptide.

Glyoxalase I

11447

1757 20. Kuntz, I. D., Crippen, G . M., and Kollman, P. A. (1979) BWpolymers 18,939-957 21. Havel, T.F., Kuntz, I. D., and Crippen, G . M. (1983) Bull. Math. Biol. 45.665-720 22. Aronsson, A.-C., Tibbelin, G . , and Mannervik, B. (1979) Anal. Biochem. 9 2 , 390-393 23. Ball, J. C., and Vander Jagt, D. L. (1979) Anal. Biochem. 98, 472-477 24. Vince,R., Daluge, S., and Wadd, W. B. (1971) J . Med. Chem. 14,402-404 25. Mildvan, A. S., and Gupta, R. K. (1978) Methods Enzymol. 49, 322-359 26. Marklev. J. L., Horsley, - . W. J., and Klein, M. P. (1971) J . Chem. Phys: 55, 3604 27. Mildvan. A. S.. Granot. J.. Smith. G. M.. and Liebman. M.N. (1980)’Adu.inorg. Bidchem. 2, 211-236’ 28. Rao, S. T., and Rossmann, M. G . (1973) J . Mol. Biol. 76, 241256 29. Wood, W., Stodola, R. K., and Badler, N. (1982)DOCK Computer Program from The Institutefor Cancer Research, Philadelphia, PA 30. Carrell, H. L., and Glusker, J. P. (1973) Acta. Crystallogr. Sect. B Struct. Crystallogr. Cryst. Chem. 29, 638-640 Acknowledgments-We are grateful to Helen Berman, Jenny Glus- 31. Chothia, C., and Pauling, P. (1977) Acta. Crystallogr. Sect.B Struct. Crystallogr. Cryst. Chem. 33, 1851-1854 ker, William Wood, H. L. Carrell, and Michael Liebman of the T. R. (1952) Acta. Cryst. 5 , 14-17 32. Comer, Institute for Cancer Research, to E. Loren Buhle, Jr., Eric Suchanek, 33. Destro, R., and Heyda, A. (1977) Acta Crystallogr. Sect. B Struct. and L. Mario Amzel of the Johns Hopkins Medical School for help Crystallogr. Cryst. Chern. 33, 504-509 in the computations, and to Gudrun Tibbelin for excellent technical 34. Carrell, H. L. (1976) CALCAT Program from the Institute for assistance in the preparation of the enzyme. Cancer Research, Philadelphia, PA 35. MacGillavry, C.H., and Rieck, G . D. eds. (1962) International REFERENCES Tables for X-ray Crystallography, Vol. 111, p. 276,Kynoch Press, Birmingham, England 1. Ray, S., and Ray, M. (1981) J . Biol. Chem. 256,6230-6233 36. Fung, C. H., Feldmann, R. J., andMildvan, A. S. (1976) Biochem2. Hopper, D. J., and Cooper, R. A. (1972) Biochem. J . 128, 321istry 15,75-84 329 37. Schulz, G. E., and Schirmer, R. H. (1979) Principles of Protein 3. Ray, S., and Ray, M. (1983) J . Biol. Chem. 258,3461-3462 Structure pp. 17-26, Springer-Verlag, New York 4. Iyengar, R., and Rose, I. A. (1981) Biochemistry 20, 1223-1229 38. Zacharias, D. E., Murray-Rust, P., Preston, R. M., and Glusker, 5. Grazi, E., and Trombetta, G . (1978) Biochem. J . 175, 361-365 J. P. (1983) Arch. Biochem. Biophys. 222, 22-34 6. Racker, E. (1951) J . Biol. Chem. 190, 685-696 7. Hall, S. S., Doweyko, A. M., and Jordan, F. (1976) J. Am.Chem. 39. Cahn, R. S., Ingold, C., and Prelog, V. (1966) Angew. Chem. Znt. Ed. Engl. 5, 385-415 SOC.98,7460-7461 8. Chari, R. V. J., and Kozarich, J. W. (1981) J . Biol. Chem. 2 5 6 , 40. Fischman, A. J., Live, D. H., Wyssbrod, H.R., Agosta, W.C., and Cowburn, D. (1980) J. Am. Chem. SOC.102,2533-2539 9785-9788 9. Sellin, S., Aronsson, A.-C., and Mannervik, B. (1980) Acta. Chem. 41. Wuthrich, K. (1976) NMR in Biological Research: Peptides and Proteins pp. 42-55, Elsevier/North-Holland, New York S c a d . Ser. B Org. Chem. Biochem. 34,541-543 10. Sellin, S., Eriksson, L. E. G., Aronsson, A.-C., and Mannervik, 42. Fujiwara, S., Formicka-Kozlowska, G., and Kozlowski, H. (1977) Bull. Chem. SOC. JPN 50,3131-3135 B. (1983) J . Biol. Chem. 258,2091-2097 11. Garcia-Iniquez, L, Powers, L., Chance, B., Sellin, S., Mannervik, 43. Eriksson, B., and Eriksson, S. A. (1967) Acta. Chem. Scand. 21, B., and Mildvan, A. S. (1984) Biochemistry 23, 685-689 1304-1312 12. Sellin, S., Eriksson, L. E. G., and Mannervik, B. (1982) Biochem- 44. Sellin, S., Aronsson, A.-C., Eriksson, L. E. G., Larsen, K., Tibistry 2 1,4850-4857 belin, G., and Mannervik, B. (1983) in FunctionsofGlutathWne: 13. Sellin, S., Rosevear, P. R., Mannervik, B., and Mildvan, A. S. Biochemical, Physiological, Toxicological, and Clinical Aspects (1982) J . Biol. Chem. 257, 10023-10029 (Larsson, A., Orrenius, S., Holmgren, A., and Mannervik, B., 14. Rosevear, P. R., Chari, R. V. J., Kozarich, J . W., Sellin, S., eds) pp. 187-197, Raven Press, New York Mannervik, B., and Mildvan, A. S. (1983) J . Biol. Chem. 258, 45. Sellin, S., Aronsson, A.-C., Mannervik, B., and Eriksson, L. E. 6823-6826 G. (1981) Acta. Chem. Scand. Ser. E Org. Chem. Biochem. 35, 15. Sellin, S., and Mannervik, B. (1983) J . Biol. Chem. 258, 8872229-231 8875 46. Aronsson, A.-C., Sellin, S., Tibbelin, G . , and Mannervik, B. 16. Wright, W. B. (1958) Acta. Cryst. 11,632-642 (1981) Biochem. J . 197,67-75 17. Epp, O., Ladenstein, R., and Wendel, A. (1983) Eur. J . Biochem. 47. Reuben, J., and Cohn, M. (1970) J. Biot. Chem. 2 4 5 , 6539-6546 133,51-69 48. Grittis, C. E. F., Ong, L. H., Buettner, L., and Creighton, D. J. 18. Schulz, G. E., Schirmer, R. H., Sachsenheimer, W., and Pai, E. (1983) Biochemistry 22, 2945-2951 F. (1978) Nature (Lod.) 273,120-124 49. Laurence, P. R., and Thomson, C. (1980) Theoret. Chin. Acta 57,25-41 19. Pai, E.F., and Schulz, G. E. (1983) J . Biol. Chem. 258, 1752-