PROTEINS Structure, Function, and Genetics 1488-101 (1992)
A Structural Model for Human Dihydrolipoamide Dehydrogenase Joyce E.Jentoft, Menachem Shoham, Darren Hurst, and Mulchand S. Patel Department of Biochemistry, Case Western Reserve University School of Medicine, Cleveland, Ohio 441 06
ABSTRACT The hypothesis that dihydrolipoamide dehydrogenases (E,s) have tertiary structures very similar to that of human glutathione reductase (GR) w a s tested in detail b y three separate criteria: (1) by analyzing each putative secondary structural element for conservation of appropriate polarhonpolar regions, (2) by detailed comparison of putative active site residues in E,s with their authentic counterparts in human GR, and (3)b y comparison of residues at the putative dimeric interface of the E,s with the authentic residues in GR. All three criteria are satisfied in a convincing way for the 7 E,s that were considered, s u p porting the conclusion that the structural scaffolding and the overall tertiary structure (which determines the location of functional sites and residues) are remarkably similar for the E,s and for GR. These analyses together with the crystal structures of human erythrocyte GR formed the basis for construction of a molecular model €or human E,. The cofactor FAD and the substrates NAD and lipoic acid were also included in the model. Unexpectedly, the surface residues in the cleft that holds the lipoamide were found t o be highly charged and predominantly acidic, allowing u s to predict that the region around the lipoamide in the subunit should be basic in nature. The molecular model can be tested b y site-directed mutagenesis of residues predicted to be in the dihydrolipoamide acetyltransferase subunit binding cleft. o 1992 Wiley-Liss, Inc. Key words: structural scaffolding, sequence similarity, sequence identity, flavoprotein, homology modeling, lipoic acid, mitochondria1 enzyme INTRODUCTION Mammalian dihydrolipoamidedehydrogenase (E,) (EC 1.8.1.4),a homodimer, is a common component of the three a-ketoacid dehydrogenase complexes (pyruvate dehydrogenase complex, a-ketoglutarate dehydrogenase complex and the branched-chain a-keto acid dehydrogenase complex), and of the glycine cleavage system (where it is referred to as L protein).lb4 Each polypeptide (M,51,000) contains a noncovalently bound FAD which participates in ca0 1992 WILEY-LISS, INC.
talysis by transferring electrons to NAD . E3 reoxidizes the dihydrolipoamide moiety covalently linked to the lysyl residue of the acyltransferase components of the three a-ketoacid dehydrogenase complexes and of the hydrogen-carrier protein of the glycine cleavage system. Although the mechanism of the reaction catalyzed by E, is well characterized?T6 the structurefunction relationships of the enzyme are not known. E, belongs to the family of flavoenzymes known as the pyridine nucleotide-disulfide oxidoreductases. This family includes several structurally well-characterized flavoproteins such as human erythrocyte glutathione reductase (GR),5 thioredoxin reductase: Pseudomonas aeruginosa mercury(I1) reductase,8 and trypanothione r e d u c t a ~ e .The ~ crystal structure of human GR at 1.54 resolution was used to propose a stereochemical description of the catalytic mechanism of this NADPH-dependent en~ y m e . " - ~In~ the past few years, cloning studies have resulted in the deduction of the primary amino acid sequences of several different and also of E. coli GR" (for review see references 4 and 22), so that at present the amino acid sequences are known for many flavoproteins belonging to this family. Considerable homology has been demonstrated in the primary amino acid sequences among flavoproteins. For instance, human liver E, shows the following homology with the other proteins: 96% pig heart E,, 44% E . coli E,, and 33% human erythrocyte GR.4,14,15Earlier, we compared the primary amino acid sequences of four different E,s, human GR, and Ps. aeruginosa mercury(I1)r e d u c t a ~ e .In ~,~~ this report, we present an extensive structural analysis of the E,s, based on (1)comparison of primary sequences for 7 E,s and human GR, (2) evaluation of putative secondary structural regions of E3s using the known three-dimensional structure of human GR, and (3) utilization of these analyses and the crystal structure of human GR in building a model +
The abbreviations used: E,, dihydrolipoamide dehydrogenase; GR, glutathione reductase. Received March 28, 1991; revision accepted November 10. 1991. Address reprint request to M.S. Patel, Department of Biochemistry, Case Western Reserve University School of Medicine, 10900 Euclid Avenue, Cleveland, OH 44106-4935.
MODEL FOR DIHYDROLIF’OAMIDE DEHYDROGENASE
of human E, with bound cofactors and substrate. Preliminary accounts of the crystal structures of E, from yeastz3 at 4.5 A resolution and from A. uinelandiiZ4a t 2.8 A resolution have been reported. Since no coordinates of these structures are available to us, we have based our model-building studies on the crystal structure of human GR. Two important differences exist between human GR and human E,. Whereas E, catalyzes the oxidation of dihydrolipoamide, GR catalyzes the reverse reaction, namely the reduction of glutathione. Second, E, utilizes NAD+ and GR utilizes NADPH as cofactor. Nevertheless, the two enzymes share significant sequence homology, in particular in the regions of cofactor and substrate binding. Even for the case of E . coli E,, which has less sequence homology to human GR than human E, does, it was possible to deduce a very similar structural scaffolding between the two enzymes based on the sequence ofE. coli E , and the crystal structure of human GR.25
RESULTS AND DISCUSSION Comparison of the Sequence and Secondary Structure of E,s With GR, a Protein of Known Three-Dimensional Structure Three aspects of similarity between E,s and GR are compared using the set of aligned sequences: (1) the putative secondary structural elements of E,s are compared on an element-by-element basis with the structures and sequence in comparable regions of GR, (2) the putative amino acids in the substrate and coenzyme binding sites of E,s are compared with the amino acids at the binding site of GR, and (3) the putative amino acid residues in the subunitsubunit interface for forming dimeric E,s are compared with the amino acid residues in the dimer interface region of GR. As fully discussed below, this analysis indicates that all E,s possess the same core secondary structural elements as GR and there is strong conservation of active site residues, and residues a t the dimer interface between E,s and GR. Secondary structural elements The working hypothesis of this section is as follows: if the structures are similar, then the structural elements that form the hydrophobic core of the molecule should be conserved, and the nature of the amino acids within these elements should be conserved. The structural comparison is then based on the known structure of GR” and the primary structures of E,s from human (H),14,15porcine Yeast (y),17 E. coli (E),16A. vinelandii (A),18Pseudomoms putida (pp)19and P. fluorescens (Pf).” The sequence alignment of these E,s with human GR is shown in Figure 1. The sequences were initially aligned using a computer program “Micro Genie” (Beckman), and then revised manually by maximizing alignment within secondary structural elements and by Put-
89
ting most insertions and deletions into regions outside the putative secondary structural elements. The resultant alignment strongly conserves the pattern of polar and nonpolar residues within the secondary structural regions. This is shown by the comparison of human E, and GR, where identical patterns are indicated by the letters P (polar) and N (nonpolar) under the aligned sequences. The percentage identity plus percentage of conserved polarity (the combined percentage shown in parentheses) for the sequence alignment of each protein relative to human E,, as shown in Figure 1, was Pf-E,, 37.8% -t 29.5% (67.3%);Pp-E,, 36.9% + 31.2% (68.8%); A-E3,48.9%+ 27.4%(76.4%);E-E,,42.5% + 31.9% (74.4%);Y-E3, 55.5% 24.9%(80.4%);P-E,, 96.2% + 3.2%(99.4%);and H-GR, 30.0%+ 36.5%(66.5%). Additionally, comparison of identity and similarity within secondary structural elements for H-GR compared with H-E, was 30.3% + 42.9%(73.2%).The 30%identity of amino acid residues between human E, and human GR is not much less than that for the distantly related Pseudomonas E,s and human E,, and the overall conservation of polarity for all comparisons with human E, was over 65%.DoolittleZ6 has noted that a sequence identity of 25%is significant, and one of the earliest examples of modeling, of lactalbumin on the X-ray structure of hen egg white lysozyme, was based on 35%identity of amino acid residue^.'^ Doolittle also notedz6that modeling is most likely to give useful results if the two proteins share functional properties, a criterion clearly met by human E, and human GR, as well as the other members of the pyridine nucleotide disulfide oxidoreductase family. The tertiary structural elements of GR, starting from the amino terminus, are the flexible Nterminal segment (GR residues 1 to 181, the FAD domain (GR residues 19 to 157),the NAPDH domain (GR residues 158 to 293), the central domain (GR residues 294 to 364), and the interface domain (GR residues 365 to 478). The domain structure of E,, based on GR, is illustrated in Figure 2. The flexible N-terminal segment. This region is truncated from 18 residues in GR to 4 residues in human and porcine E,, 3 residues in yeast, E. coli E, and P. putida, and only 1 residue in A. vinelandii and P. fluorescens. The N-terminal region of the molecule is therefore likely to be in differing environments for E,s relative to GR. In addition, a more polar character would be expected in the E, molecules in the regions that are no longer shielded from solvent by a long N-terminal chain. Unfortunately we do not know what regions will be affected by this change because the N-terminal segment of GR is disordered in the crystal. The FAD domain. This domain (shown in Fig. 2, upper) provides one side of the FAD binding site. Its central feature is a large p-sheet shared with the central domain (shown edge-on in the lower left
+
90
J.E. JENTOFT ET AL.
A Structural Model for Human Dihydrolipoamide Dehydrogenase Joyce E.Jentoft, Menachem Shoham, Darren Hurst, and Mulchand S. Patel Department of Biochemistry, Case Western Reserve University School of Medicine, Cleveland, Ohio 441 06
ABSTRACT The hypothesis that dihydrolipoamide dehydrogenases (E,s) have tertiary structures very similar to that of human glutathione reductase (GR) w a s tested in detail b y three separate criteria: (1) by analyzing each putative secondary structural element for conservation of appropriate polarhonpolar regions, (2) by detailed comparison of putative active site residues in E,s with their authentic counterparts in human GR, and (3)b y comparison of residues at the putative dimeric interface of the E,s with the authentic residues in GR. All three criteria are satisfied in a convincing way for the 7 E,s that were considered, s u p porting the conclusion that the structural scaffolding and the overall tertiary structure (which determines the location of functional sites and residues) are remarkably similar for the E,s and for GR. These analyses together with the crystal structures of human erythrocyte GR formed the basis for construction of a molecular model €or human E,. The cofactor FAD and the substrates NAD and lipoic acid were also included in the model. Unexpectedly, the surface residues in the cleft that holds the lipoamide were found t o be highly charged and predominantly acidic, allowing u s to predict that the region around the lipoamide in the subunit should be basic in nature. The molecular model can be tested b y site-directed mutagenesis of residues predicted to be in the dihydrolipoamide acetyltransferase subunit binding cleft. o 1992 Wiley-Liss, Inc. Key words: structural scaffolding, sequence similarity, sequence identity, flavoprotein, homology modeling, lipoic acid, mitochondria1 enzyme INTRODUCTION Mammalian dihydrolipoamidedehydrogenase (E,) (EC 1.8.1.4),a homodimer, is a common component of the three a-ketoacid dehydrogenase complexes (pyruvate dehydrogenase complex, a-ketoglutarate dehydrogenase complex and the branched-chain a-keto acid dehydrogenase complex), and of the glycine cleavage system (where it is referred to as L protein).lb4 Each polypeptide (M,51,000) contains a noncovalently bound FAD which participates in ca0 1992 WILEY-LISS, INC.
talysis by transferring electrons to NAD . E3 reoxidizes the dihydrolipoamide moiety covalently linked to the lysyl residue of the acyltransferase components of the three a-ketoacid dehydrogenase complexes and of the hydrogen-carrier protein of the glycine cleavage system. Although the mechanism of the reaction catalyzed by E, is well characterized?T6 the structurefunction relationships of the enzyme are not known. E, belongs to the family of flavoenzymes known as the pyridine nucleotide-disulfide oxidoreductases. This family includes several structurally well-characterized flavoproteins such as human erythrocyte glutathione reductase (GR),5 thioredoxin reductase: Pseudomonas aeruginosa mercury(I1) reductase,8 and trypanothione r e d u c t a ~ e .The ~ crystal structure of human GR at 1.54 resolution was used to propose a stereochemical description of the catalytic mechanism of this NADPH-dependent en~ y m e . " - ~In~ the past few years, cloning studies have resulted in the deduction of the primary amino acid sequences of several different and also of E. coli GR" (for review see references 4 and 22), so that at present the amino acid sequences are known for many flavoproteins belonging to this family. Considerable homology has been demonstrated in the primary amino acid sequences among flavoproteins. For instance, human liver E, shows the following homology with the other proteins: 96% pig heart E,, 44% E . coli E,, and 33% human erythrocyte GR.4,14,15Earlier, we compared the primary amino acid sequences of four different E,s, human GR, and Ps. aeruginosa mercury(I1)r e d u c t a ~ e .In ~,~~ this report, we present an extensive structural analysis of the E,s, based on (1)comparison of primary sequences for 7 E,s and human GR, (2) evaluation of putative secondary structural regions of E3s using the known three-dimensional structure of human GR, and (3) utilization of these analyses and the crystal structure of human GR in building a model +
The abbreviations used: E,, dihydrolipoamide dehydrogenase; GR, glutathione reductase. Received March 28, 1991; revision accepted November 10. 1991. Address reprint request to M.S. Patel, Department of Biochemistry, Case Western Reserve University School of Medicine, 10900 Euclid Avenue, Cleveland, OH 44106-4935.
MODEL FOR DIHYDROLIF’OAMIDE DEHYDROGENASE
of human E, with bound cofactors and substrate. Preliminary accounts of the crystal structures of E, from yeastz3 at 4.5 A resolution and from A. uinelandiiZ4a t 2.8 A resolution have been reported. Since no coordinates of these structures are available to us, we have based our model-building studies on the crystal structure of human GR. Two important differences exist between human GR and human E,. Whereas E, catalyzes the oxidation of dihydrolipoamide, GR catalyzes the reverse reaction, namely the reduction of glutathione. Second, E, utilizes NAD+ and GR utilizes NADPH as cofactor. Nevertheless, the two enzymes share significant sequence homology, in particular in the regions of cofactor and substrate binding. Even for the case of E . coli E,, which has less sequence homology to human GR than human E, does, it was possible to deduce a very similar structural scaffolding between the two enzymes based on the sequence ofE. coli E , and the crystal structure of human GR.25
RESULTS AND DISCUSSION Comparison of the Sequence and Secondary Structure of E,s With GR, a Protein of Known Three-Dimensional Structure Three aspects of similarity between E,s and GR are compared using the set of aligned sequences: (1) the putative secondary structural elements of E,s are compared on an element-by-element basis with the structures and sequence in comparable regions of GR, (2) the putative amino acids in the substrate and coenzyme binding sites of E,s are compared with the amino acids at the binding site of GR, and (3) the putative amino acid residues in the subunitsubunit interface for forming dimeric E,s are compared with the amino acid residues in the dimer interface region of GR. As fully discussed below, this analysis indicates that all E,s possess the same core secondary structural elements as GR and there is strong conservation of active site residues, and residues a t the dimer interface between E,s and GR. Secondary structural elements The working hypothesis of this section is as follows: if the structures are similar, then the structural elements that form the hydrophobic core of the molecule should be conserved, and the nature of the amino acids within these elements should be conserved. The structural comparison is then based on the known structure of GR” and the primary structures of E,s from human (H),14,15porcine Yeast (y),17 E. coli (E),16A. vinelandii (A),18Pseudomoms putida (pp)19and P. fluorescens (Pf).” The sequence alignment of these E,s with human GR is shown in Figure 1. The sequences were initially aligned using a computer program “Micro Genie” (Beckman), and then revised manually by maximizing alignment within secondary structural elements and by Put-
89
ting most insertions and deletions into regions outside the putative secondary structural elements. The resultant alignment strongly conserves the pattern of polar and nonpolar residues within the secondary structural regions. This is shown by the comparison of human E, and GR, where identical patterns are indicated by the letters P (polar) and N (nonpolar) under the aligned sequences. The percentage identity plus percentage of conserved polarity (the combined percentage shown in parentheses) for the sequence alignment of each protein relative to human E,, as shown in Figure 1, was Pf-E,, 37.8% -t 29.5% (67.3%);Pp-E,, 36.9% + 31.2% (68.8%); A-E3,48.9%+ 27.4%(76.4%);E-E,,42.5% + 31.9% (74.4%);Y-E3, 55.5% 24.9%(80.4%);P-E,, 96.2% + 3.2%(99.4%);and H-GR, 30.0%+ 36.5%(66.5%). Additionally, comparison of identity and similarity within secondary structural elements for H-GR compared with H-E, was 30.3% + 42.9%(73.2%).The 30%identity of amino acid residues between human E, and human GR is not much less than that for the distantly related Pseudomonas E,s and human E,, and the overall conservation of polarity for all comparisons with human E, was over 65%.DoolittleZ6 has noted that a sequence identity of 25%is significant, and one of the earliest examples of modeling, of lactalbumin on the X-ray structure of hen egg white lysozyme, was based on 35%identity of amino acid residue^.'^ Doolittle also notedz6that modeling is most likely to give useful results if the two proteins share functional properties, a criterion clearly met by human E, and human GR, as well as the other members of the pyridine nucleotide disulfide oxidoreductase family. The tertiary structural elements of GR, starting from the amino terminus, are the flexible Nterminal segment (GR residues 1 to 181, the FAD domain (GR residues 19 to 157),the NAPDH domain (GR residues 158 to 293), the central domain (GR residues 294 to 364), and the interface domain (GR residues 365 to 478). The domain structure of E,, based on GR, is illustrated in Figure 2. The flexible N-terminal segment. This region is truncated from 18 residues in GR to 4 residues in human and porcine E,, 3 residues in yeast, E. coli E, and P. putida, and only 1 residue in A. vinelandii and P. fluorescens. The N-terminal region of the molecule is therefore likely to be in differing environments for E,s relative to GR. In addition, a more polar character would be expected in the E, molecules in the regions that are no longer shielded from solvent by a long N-terminal chain. Unfortunately we do not know what regions will be affected by this change because the N-terminal segment of GR is disordered in the crystal. The FAD domain. This domain (shown in Fig. 2, upper) provides one side of the FAD binding site. Its central feature is a large p-sheet shared with the central domain (shown edge-on in the lower left
+
90
J.E. JENTOFT ET AL.
