Structural and Sequence Comparisons of Quinone Oxidoreductase, ζ-Crystallin, and Glucose and Alcohol Dehydrogenases

ARCHIVES OF BIOCHEMISTRY AND BIOPHYSICS

Vol. 328, No. 1, April 1, pp. 173–183, 1996 Article No. 0158

Structural and Sequence Comparisons of Quinone Oxidoreductase, z-Crystallin, and Glucose and Alcohol Dehydrogenases Karen J. Edwards,*,1 John D. Barton,* Jamie Rossjohn,† Jennifer M. Thorn,* Garry L. Taylor,‡ and David L. Ollis* *Centre for Molecular Structure and Function, Research School of Chemistry, Australian National University, Canberra, Australian Capital Territory 0200, Australia; †St. Vincent’s Institute of Medical Research, Fitzroy, Victoria 3065, Australia; and ‡School of Biology and Biochemistry, Bath University, Claverton Down, Bath BA2 7AY, United Kingdom

Received September 8, 1995, and in revised form January 15, 1996

Quinone oxidoreductase, z-crystallin, glucose dehydrogenase, and alcohol dehydrogenase belong to a superfamily of medium-chain dehydrogenase/reductases. The crystal structures of Escherichia coli quinone oxidoreductase (QOR) and Thermoplasma acidophilum glucose dehydrogenase have recently been determined and are compared here with the well-known structure of horse liver alcohol dehydrogenase. A structurally based comparison of these three enzymes confirms that they possess extensive overall structural homology despite low sequence identity. The most significant difference is the absence of the catalytic and structural zinc ions in QOR. A multiple structurebased sequence alignment has been constructed for the three enzymes and extended to include z-crystallin, an eye lens structural protein with quinone oxidoreductase activity and high sequence identity to E. coli quinone oxidoreductase. Residues which are important for catalysis have been altered and the functions and activities of the enzymes have diverged, illustrating a classic example of divergent evolution among a superfamily of enzymes. q 1996 Academic Press, Inc. Key Words: alcohol dehydrogenase; glucose dehydrogenase; medium chain dehydrogenase; quinone oxidoreductase; z-crystallin.

Despite the rapid rate at which new protein structures are being determined, the appearance of a new protein fold remains a rarity. While it is clear that a 1 To whom correspondence should be addressed. Fax: 61-6-249 0750.

single protein fold can be used for a number of functions, the aspects of a structure that make it suitable for multiple functions are not obvious. In the case of enzymes the situation is more complex: they are capable of binding substrates and of catalyzing reactions between bound groups. In this paper, attention is focused on two new enzyme structures which are structurally homologous to horse liver alcohol dehydrogenase (LADH).2 All three are redox enzymes and share a common three-dimensional structural fold yet exhibit dissimilar cellular functions and substrate specificities. The most obvious differences are associated with the zinc ions that are bound to some of these proteins and with their catalytic residues. In LADH these bound metals are responsible for the catalytic activity and structural stability of the protein. Despite the importance of zinc ions in LADH, other members of the family function effectively in the absence of bound metal ions. Alcohol dehydrogenases occur in a wide range of organisms spanning the Eukarya, Bacteria, and Archaea (1–3). A medium-chain dehydrogenase/reductase (MDR) family has recently been defined, extended, and subdivided to include sequence-related dehydrogenases and reductases (4). Those classified as dehydrogenases require zinc ions (the Zn-ADHs) for activity, whereas the reductases do not necessarily require metal ions for activity. The Zn-ADHs catalyze the re2 Abbreviations used: QOR, quinone oxidoreductase; LADH, liver alcohol dehydrogenase; GDH, glucose dehydrogenase; Zn-ADH, zinc alcohol dehydrogenase; MDR, medium-chain dehydrogenase/reductase; NADH, nicotinamide adenine dinucleotide; NADPH, nicotinamide-adenine dinucleotide phosphate; DMSO, dimethyl sulfoxide; rms, root mean square.

173

0003-9861/96 $18.00 Copyright q 1996 by Academic Press, Inc. All rights of reproduction in any form reserved.

/ 6b15$$9366

02-27-96 12:06:43

arcas

AP: Archives

174

EDWARDS ET AL.

FIG. 1. Pairwise superposition of the Ca backbones for (a) QOR and GDH and (b) QOR and LADH monomers. QOR is shown in black in each figure; GDH and LADH are shown in gray.

duction of a wide variety of medium- to long-chain alcohols, with the horse LADH being the most extensively studied (5, 6). LADH is homodimeric, with each monomer consisting of a catalytic and nucleotide-binding domain. The catalytic domain comprises mainly b-sheet structures while the nucleotide-binding domain contains the characteristic babab motif known as the ‘‘Rossman fold’’ (7). LADH specifically requires the cofactor NAD/ for activity.

/ 6b15$$9366

02-27-96 12:06:43

arcas

The X-ray crystal structures of two further representative members of this MDR superfamily have recently been reported. These are an Escherichia coli quinone oxidoreductase (8, 9), which catalyzes the NAD(P)Hdependent reduction of quinone substrates, and a Thermoplasma acidophilum glucose dehydrogenase (10), which catalyzes the conversion of glucose to gluconate. These three enzymes have several functional and structural differences. GDH belongs to the zinc-containing

AP: Archives

175

COMPARISON OF FOUR MEDIUM-CHAIN DEHYDROGENASES TABLE I

Pairwise Ca rms Deviations between QOR:LADH and QOR:GDH. Monomer

Dimer

Domains aligned:

LADH

GDH

Catalytic domain Dinucleotide binding domain All common Ca

˚ (158) 1.9 A ˚ (139) 2.5 A ˚ (297) 2.7 A

˚ (149) 3.6 A ˚ (127) 3.5 A ˚ (261) 3.8 A

LADH 2.7 3.7 3.7

˚ A ˚ A ˚ A

GDHa 7.6 4.7 7.5

˚ A ˚ A ˚ A

Note. The number of atoms used in the calculation is given in parentheses. a One-half of the GDH tetramer, composed of two monomers equivalent to the QOR and LADH dimers.

dehydrogenase subfamily, as does LADH, whereas QOR is a member of the reductase subfamily and does not possess metal atoms. In addition, both QOR and LADH are dimeric, whereas GDH is tetrameric. Despite the low sequence identity between these three enzymes (LADH and GDH have 21 and 17% sequence identity with QOR, respectively), they exhibit extensive structural homology, not only in their nucleotidebinding domains but also in their catalytic domains. The MDR superfamily contains a number of other enzymes which have been identified on the basis of their sequence similarity. These have been detailed by Persson and coworkers (4). Of particular importance to the present study is the family of z-crystallins which, like QOR, belong to the reductase subfamily of MDRs. The z-crystallins also have quinone oxidoreductase activity and require NADPH as a cofactor (11). z-Crystallin is a major structural eye lens protein of camels, llamas, and some hystricomorphic rodents, and appears to belong to a group of crystallins which have been recruited from functional enzymes (12). Extensive sequence-based alignments and comparisons have been performed to analyze structure–function and evolutionary relationships between members of the MDR superfamily (4, 13, 14). It should be noted that these studies have been based exclusively on the structure of LADH. Here we compare the structures of QOR, GDH, and LADH to find the elements of their structure which are most important for their function and which have been conserved during the course of evolution. A structurebased sequence alignment has been constructed for the three enzymes and extended to include z-crystallin. QOR, GDH, and LADH have strong overall similarity with numerous structurally conserved regions. Functional considerations relating to both the nucleotidebinding and the substrate-binding sites are discussed. METHODS Structurally based sequence alignment. The structures of QOR, GDH, and LADH were superposed using the program SHP (D. Stu-

/ 6b15$$9366

02-27-96 12:06:43

arcas

art, unpublished results) and used to construct a structure-based sequence alignment within the GCG package (15). Since QOR is in its binary complex with the reduced form of NADP, comparisons were made to a similar LADH complex. Coordinates for the LADH/ NADH/DMSO crystal structure were obtained from the Brookhaven Databank (16) [Accession No. PDB2OHX.ENT (17)]. Coordinates for QOR are available from the Brookhaven Databank (Accession No. PDB1QOR.ENT) and those for GDH may be obtained from Garry Taylor. z-Crystallin (guinea pig) was included in the alignment due to its significant sequence identity and activity similar to that of QOR. On the basis of the structure-based sequence alignment, pairwise least-squares fitting of coordinates was carried out for common Ca atoms in QOR and LADH and QOR and GDH. Visual inspection of these alignments gave refined common structural regions which were used for final pairwise alignments and rms deviation calculations. Only those residues which were clearly homologous were used for the final alignments. Accessible and electrostatic surfaces. Accessible surface areas were calculated with X-PLOR (18) using the algorithm of Lee and ˚ . Solvent-accessible surfaces Richards (19) with a probe radius of 1.6 A were displayed with the Alberta/Caltech version of FRODO/TOM (20) used in conjection with Connolly’s ‘‘ms’’ program (21). Electrostatics for the substrate-binding site were determined using GRASP (22) (data not shown).

RESULTS AND DISCUSSION

Figure 1 clearly shows that the overall structures of the QOR, GDH, and LADH monomers are similar. Table I gives rms deviations between QOR, GDH, and LADH monomer subunits and dimers. Superposition of the structures of the three enzymes enabled a structurally based sequence alignment to be constructed and the relationship between sequence and structural conservation examined. Structure-Based Sequence Alignment A multiple structure-based sequence alignment has been constructed for the three MDR enzymes LADH, GDH, and QOR. The sequence of z-crystallin was added to this alignment due to its similarity to QOR. Figure 2 shows the alignment of QOR, GDH, LADH, and zcrystallin with boxed areas representing structurally conserved regions. These regions predominantly contain secondary structural elements. The largest differ-

AP: Archives

176

EDWARDS ET AL.

FIG. 2. Structure-based sequence alignment of QOR (Q), LADH (L), GDH (G), and z-crystallin (Z). Residue numbers for each protein are given in parentheses at the end of each line and secondary structure elements are labeled for QOR. The catalytic domain of QOR encompasses residues 2–124 and 266–327; the nucleotide binding domain contains residues 125–265. Residues identical in all four proteins are shaded in dark gray while similar residues are shaded in light gray. Residues involved in binding the catalytic zinc of GDH and LADH are marked (M). Residues which bind the structural zinc of both enzymes (F), GDH (FG), and LADH (FA) are also marked. Boxed areas represent structurally homologous regions of the three proteins.

ences are found mainly in the loop regions connecting secondary structure elements where numerous insertions and deletions are to be found. Despite the fact that the sequences of QOR, GDH, and LADH possess very few strictly conserved residues (21 in total, of which 9 are glycine) and low sequence similarity, it can be seen that extensive structural homology exists between these enzymes. This is evident not only in the nucleotide-binding domain but also in the catalytic domain.

/ 6b15$$9366

02-27-96 12:06:43

arcas

Although the relationship between these enzymes was predicted on the basis of their sequence alignments (4, 13, 14), the exact structural homology between these enzymes would have been difficult to predict based solely on sequence comparisons. In the light of the present three-dimensional structural information the multiple sequence alignment reported for these proteins and the other MDRs (4) could be significantly improved. One of the reasons for poor sequence alignment is that although most insertions and deletions occur in

AP: Archives

COMPARISON OF FOUR MEDIUM-CHAIN DEHYDROGENASES

177

FIG. 3. Sequence alignment for QOR (Q), GDH (G), and LADH (L) extracted from that reported by Persson et al. (4), showing the corrected alignment based on the structures of the enzymes. Secondary structure elements for QOR are labeled. The lines connecting residues represent the correct sequence alignment based on the structural alignment.

loop regions linking core secondary structures, errors are often attributable to misplacing of a gap in a core secondary structure region. Choice of scoring schemes and gap penalty functions for the alignment algorithm may also affect the accuracy of the alignment. Factors which may have affected the accuracy of the alignment of the MDR enzymes (4) are the low sequence identity between LADH and GDH and the use of a single structural model. Figure 3 shows the sequence alignment for QOR, GDH, and LADH extracted from Persson and coworkers (4). As can be seen there are numerous regions where there are discrepancies. The most significant is the poor alignment of the structural Zn loop region for GDH (residues 90–127). The region from aC to bF, which constitutes greater than half the Rossman fold and includes the nucleotide-binding motif, is also poorly aligned. This is of particular concern since the Rossman fold is one of the most highly conserved protein folds known and many predictions are based upon its structure. The second part of the catalytic domain (a4, b11, a5, and b12) is also misaligned. Such flaws in the sequence-based alignments have the potential to result in incorrect assumptions about the structure

/ 6b15$$9366

02-27-96 12:06:43

arcas

of these enzymes. In particular, specific functions cannot be assigned to individual residues with any certainty. Additionally, phylogenetic trees are constructed from such multiple sequence-based alignments and may thus be misleading. Since any inferences from a sequence alignment are dependent upon its accuracy, we would caution that care should be exercised when making structure–function predictions based solely on sequence comparisons of evolutionarily divergent proteins with low sequence identity. The accuracy of sequence alignments can be improved by the inclusion of additional structural information into the gap penalty function (23) of the alignment algorithm. This can effectively be achieved by using more accurate secondary structure information derived from a number of known structures within an enzyme family. Homology modeling can be a useful technique for predicting the structure of an amino acid sequence modeled from the known structure of a protein with a homologous sequence (24). An initial step in homology modeling involves the construction of a three-dimensional framework for the structurally conserved regions within a protein family (25). In many cases a better model is produced from a framework rather than

AP: Archives

178

EDWARDS ET AL.

FIG. 4. MOLSCRIPT (32) diagram of the structure of QOR with bound NADPH cofactor. The shaded areas represent the structurally conserved regions found in QOR, GDH, and LADH and correspond to the boxed regions in Fig. 2. These are shaded as follows to represent conservation in each of the domains: light gray (catalytic domain), dark gray (interconnecting a-helices), and midgray (nucleotide-binding domain). The area colored black corresponds to residues 87–94 (QOR) and represents the region of the zinc-loop deletion in QOR. Secondary structure elements are labeled.

from an individual structure. On the basis of the structure-based sequence alignment of these enzymes, a structural framework for the MDR enzymes is proposed. QOR was chosen as the basis for this framework as it is the simplest of the three representative enzymes. This framework is depicted in Fig. 4 where the structurally conserved regions are proposed to define a structural framework for the MDR enzymes. Overall Structure Comparison The quaternary structures of the enzymes are somewhat different. LADH and QOR are dimeric, whereas

/ 6b15$$9366

02-27-96 12:06:43

arcas

GDH and z-crystallin are tetrameric. GDH tetramer contacts are predominantly via an additional loop found between a2 and bA and the structural lobe. As this loop is absent in z-crystallin the tetrameric association of z-crystallin must be different from that of GDH. The only residue involved in GDH tetramer formation which is conserved with z-crystallin (Gly 208:GDH) is strictly conserved across all four enzymes. The major site of dimer–dimer interactions is strand bF. Although this region is structurally conserved, there is no corresponding sequence identity. The only conserved residue involved in dimer formation (Gly 239:QOR) is in a region of minor interaction. As a result, the detailed subunit interactions between these enzymes differ markedly leading to significantly larger rms deviations for the dimer compared to the monomer (Table I). The nucleotide-binding fold and mode of cofactor binding appear to be structurally conserved throughout the dehydrogenase family (5, 26). The arrangement of secondary structural elements is highly conserved in the nucleotide-binding domains of QOR, GDH, and LADH. The main difference between GDH, QOR, and LADH is that GDH is missing helix aD. This additional helix, found between the two babab motifs of QOR and LADH, is not part of the classic Rossman fold (27), although it appears to be common in the MDRs (4). Secondary structure elements in the catalytic domain are also well conserved although the loop between helix a1 and strand b5 is quite different in all three enzymes. Strand b12 is absent in GDH, resulting in the formation of a four stranded b-sheet compared to the sixstranded sheet present in QOR and LADH (Fig. 2). The most significant difference between QOR and the other two enzymes is the absence of the catalytic and structural zinc ions. Although the overall structures of the substrate-binding sites are similar, the equivalent catalytic zinc-binding residues in QOR have been replaced by residues which are incapable of coordinating metal ions. The large structural zinc-binding loop (Fig. 5) found in LADH and GDH is completely absent in QOR. The absence of this loop in QOR increases the accessibility of the substrate-binding cleft to solvent. Table II lists the solvent accessibility of the active site for all three enzymes and also gives the area covered by the structural zinc-binding loop in GDH and LADH. The diminished (in GDH) or absent (in QOR) structural zinc loop also exposes a large cleft which runs through the catalytic domains of GDH and QOR. Previous workers (28) have pointed out that z-crystallins do not have this structural zinc-binding loop nor do they have residues which are appropriate for binding the catalytic zinc. The presence or absence of the zinc-binding loop also affects the arrangement of the secondary structure at

AP: Archives

COMPARISON OF FOUR MEDIUM-CHAIN DEHYDROGENASES

179

FIG. 5. Zinc-binding loops and nearby regions. Ca traces are shown for QOR (thick gray), GDH (black), and LADH (thin gray). Structural Zn atoms for GDH (black) and LADH (gray) are also shown. The N- and C-terminal residues of the zinc-binding loops of GDH (residues 90–127) and LADH (91–145), as well as the equivalent peptide of QOR (87–94), are marked (Zn N and Zn C). The ends of helix a3 of QOR (266–276) and LADH (319–325) and the equivalent peptide of GDH (305–308) are labeled. Helix a2 is labeled (QOR, 116–136; GDH, 149– 167; LADH, 167–187).

some distance from the loop itself. From the superimposed structures (Fig. 5), it can be seen that the zinc loops of LADH and GDH pass close to helix a3 of QOR. LADH and GDH have taken different strategies to avoid this collision. GDH does not have an equivalent to helix a3 (QOR), having instead a short, direct peptide and a shorter helix a4 (Fig. 5). LADH has an equivalent to helix a3 and a shortened helix a4. However, the LADH a3 turns in the direction opposite to that of QOR, resulting in a significant displacement of the helix relative to a3 in QOR. Further, the presence or absence of helix a3 affects the position of the C-terminal end of helix a2. Helix a2 is kinked in all three

TABLE II

Solvent-Accessible Surface Areas for the QOR, LADH, and GDH Monomers

LADH GDH QOR

Total accessible surface area ˚ 2) (A

Substrate-binding site area ˚ 2) (A

Area covered by Zn binding loop ˚ 2) (A

14,300 15,300 12,800

119 139 244

1600 920 —

/ 6b15$$9366

02-27-96 12:06:43

arcas

proteins. The degree and direction of this kink varies, resulting in the N-termini of helix a2 aligning well in all three proteins, whereas the C-termini do not. A ‘‘helix breaking’’ residue is present at the kink (Gly 175:LADH, Pro 156:GDH, Gly 123:QOR, Pro 132:zcrystallin). There seems to be no preference for proline or glycine in this position, suggesting that there is structural conservation without sequence conservation. Functional Considerations: Nucleotide Binding The bab nucleotide-binding regions for QOR and LADH are depicted in Fig. 6. Structurally based studies on sequence patterns found in the fingerprint region of the nucleotide-binding domain of a number of NAD(P)binding proteins (29, 30) have identified different requirements for nucleotide specificity. NAD requires a GXGXXG sequence motif plus a negatively charged residue at the end of strand bB, whereas specificity for NADP is achieved by a GXGXXA sequence motif and a positively charged residue near the end of strand bB. It appears that the nucleotide-binding motifs for QOR and z-crystallin are somewhat unusual since both enzymes have single residue insertions in the region between the first and the second conserved glycines. QOR

AP: Archives

180

EDWARDS ET AL.

FIG. 6. The bab (bA, aB, bB) nucleotide-binding region showing the nucleotide-binding motif (labeled) and bound cofactor for (a) QOR and (b) LADH. LADH has the classic GXGXXG nucleotide-binding motif, whereas QOR has a single residue insertion (Ala 149) to give an AXXGXXG motif. The different conformations for NADPH (QOR) and NADH (LADH) are clearly seen. The important residues, Val 172 (QOR) and Asp 223 (LADH), at the end of strand bB, are also labeled.

has an AXXGXXG motif (Fig. 6a); z-crystallin has GXXGXXG. Neither enzyme has the last alanine residue in the motif predicted for NADPH binding, confirming the proposal by Baker and coworkers (29) that this alanine residue is not a prerequisite for NADPH binding. Various LADH/NAD complexes have been reported (5). Table III lists the residues for QOR and LADH which are involved in cofactor binding. Although structurally equivalent residues are present, they are not necessarily involved in cofactor binding. The structures of QOR/NADPH and LADH/NADH show some notable differences (Fig. 6). NADPH when bound to QOR is bent and more compressed than NADH when bound to LADH. The region of adenosine binding in QOR is different from the corresponding region in LADH with the adenine ring being syn for QOR:NADPH and anti for LADH:NADH (5). In QOR the loop between bE and aF has moved closer to the cofactor to interact with the adenine ring. QOR is also alone among these three enzymes in possessing two serine residues (Ser 241 and Ser 242) inserted into the sequence immediately after the conserved region which includes strand bE (Fig.

/ 6b15$$9366

02-27-96 12:06:43

arcas

2). This insertion maximizes contact between QOR and the adenosine moiety of the cofactor. The insertion of an extra residue (Ala 150) in the QOR sequence appears to accommodate the adenine ribose phosphate group by widening the cleft in this area. The widened cleft in QOR leads to less direct contact between the cofactor and QOR with a chain of water molecules being used to mediate interactions between enzyme and cofactor. QOR has a small hydrophobic residue, Val 172, instead of the larger negatively charged residue (Asp 223) required for NADH specificity in LADH (31). This mutation is essential in providing sufficient space for the additional adenine ribose phosphate group of the NADPH cofactor. QOR has two positively charged residues, Lys 177 and Arg 317, which surround the phosphate group and stabilize the negative charge on the phosphate by electrostatic interactions. The interaction with Lys 177 was easily predicted based on the requirement for a positively charged residue at the end of strand bB (30). The refined structure of QOR has shown, however, that Arg 317 is also of importance in

AP: Archives

181

COMPARISON OF FOUR MEDIUM-CHAIN DEHYDROGENASES TABLE III

Comparison of Cofactor-Binding Residues for LADH and QOR QOR/NADPH

LADH/NADH

Conserved region?

Gly 173 (Val 217) (Arg 219) Ala 148* Gly 151* Gly 152 (Val 172) Lys 177 Phe 42 Gly 321* (Ile 43) Tyr 46 (Thr 127) Val 153 (Asn 240) Asn 41* (Leu 123) Leu 266* Phe 238 Pro 264 Ser 216 Ser 241a Tyr 130 Tyr 192 Arg 317

Ile 224 Ile 269 Arg 271 Gly 199 Gly 210 Gly 202 Asp 223 Lys 228 Arg 47 Arg 369 Ser 48 His 51 Thr 178 Val 203 Val 294 Cys 46 Cys 174 Phe 319 Val 292 Ala 317 (Val 268) — (Gly 180) (Pro 244) (Gly 365)

N Y Y Y Y Y Y N Y N Y Y Y Y Y Y Y N Y Y Y N Y Y N

His 317, which may interact with the phosphate group. This residue represents an insertion between the QOR residues Glu 315 and Ser 316. These histidine residues will be less flexible than the lysine and arginine QOR counterparts. The greater adaptability in this region may allow QOR to bind both NADPH and NADH while the relative inflexibility of z-crystallin limits it to NADPH specificity. Functional Consideration: The Substrate Binding Site Despite the high structural conservation in the region of the active site (helices a1 and a2 and strand b1), there are major functional differences between these enzymes. The most significant is that QOR does not possess a metal equivalent to the catalytic zinc of LADH and GDH and must therefore have a different mechanism for catalysis. LADH has a hydrophobic binding pocket (7), whereas GDH possesses a charged binding site (Rossjohn et al., unpublished results). The substrate-binding site for QOR [identified from the crystal structure (9) and based on analogy to LADH] has charged and neutral regions. Table IV details the residues that line the substratebinding site in LADH and GDH and their equivalent residues in QOR and z-crystallin. A comparison of

Note. Residues marked (*) do not directly interact with the cofactor but make water-mediated contacts. Residues in parentheses do not interact with the cofactor. a QOR:Ser 241 is in an insertion region and has no equivalent in LADH.

providing electrostatic interactions to the phosphate group. Inspection of the structure strongly suggests that QOR should show NADPH specificity, but it appears to be active with both NADPH and NADH (Lilley et al., unpublished results). The above observations may provide an explanation for the NADPH specificity shown by z-crystallin. zCrystallin has a GXXGXXG binding motif. The first glycine (Gly 156) in the motif may be less versatile than the alanine of QOR. Inspection of the sequence alignment suggests that Lys 187 of z-crystallin may be equivalent to Lys 177 of QOR. Examination of the sequence alignment with concomitant visualization of the structure of QOR suggests that this residue is not in a suitable position for interaction with the phosphate group. z-Crystallin:His 200, equivalent to QOR:Tyr 192, may provide electrostatic interactions with the phosphate group. Unlike QOR, which has a flexible arginine (Arg 317) in the tight turn between b12 and a5, z-crystallin has a serine residue at this position. zCrystallin does, however, have a possible alternative,

/ 6b15$$9366

02-27-96 12:06:43

arcas

TABLE IV

Residues Lining the Substrate-Binding Sites for QORa z-Crystallin, LADH, and GDH QOR

z-Crystallin

LADH

GDH

— Asn 41 Thr 63 Leu 123 — — — — Ile 43 Asn 240 Tyr 52 Ser 265 Leu 266 Arg 263

— Asn 48 Thr 71 Ile 131 — — — — Val 50 Cys 248 — Ser 270 Leu 271 Gly 268

Zn Cys 46b His 67b Cys 174b Leu 116c Phe 93 Leu 141c Phe 140c Ser 48d Val 294 Leu 57 Ile 318 Phe 319e Gly 316e

Zn Cys 40b His 67b Glu 155b — Val 92 — — Thr 42d Thr 276 — Ser 303 Val 304e Ala 301e

Note. A dash denotes no equivalent residue. a The potential substrate-binding site for QOR has been identified based on the X-ray structure (9), analogy to LADH, and biochemical considerations. b Residues which bind the catalytic zinc in LADH and GDH. c Residues which are part of the structural zinc-binding loop in LADH. d Catalytic residues in LADH and GDH. e Equivalent residue in LADH and GDH but not part of substratebinding site.

AP: Archives

182

EDWARDS ET AL.

structurally equivalent residues to QOR shows marked differences in the nature of the residues surrounding the pocket. Residues Cys 46, His 67, and Cys 174 which bind the catalytic zinc in LADH have been replaced by Asn 41, Thr 63, and Leu 123 in QOR. A similar set of residues is found for z-crystallin. Further, the catalytic residue which binds the hydroxyl group of the alcohol or glucose substrate, Ser 48:LADH and Thr 42:GDH, has been replaced by a hydrophobic residue in both QOR and z-crystallin (Ile 43 and Val 150, respectively). There is only one other member of the MDR superfamily [the predicted protein product of the Trichoderma harzianu indc11 gene (4)] which shares this replacement. In LADH the basic residue, His 51, is involved in a proton shuttle with Ser 48 and O2*A of the bound NADP. The equivalent residue in QOR, Tyr 46, is also bound to the nicotinamide ribose sugar, but there is no possibility of a similar proton shuttle mechanism operating in QOR, as Ile 43:QOR is unable to make hydrogen bond contact with the nicotinamide ribose. In GDH, the equivalent residue to His 51:LADH is Gly 45:GDH. The glycine at this position allows the oxygen atom of Thr 42:GDH to transfer a proton directly to bulk solvent. The substrate-binding site of z-crystallin is similar to that of QOR, having four strictly conserved residues and two conservative changes. The most notable differences between QOR and z-crystallin are found at Arg 263, where the equivalent residue in z-crystallin is a glycine (Gly 268), and at Tyr 52 where there is no equivalent z-crystallin residue. No residues in the substrate binding site are conserved between QOR and LADH, and only one residue, Ser 265:QOR, is conserved between QOR and GDH. This gives the strongest indication of the divergent nature of these enzymes. Concluding Remarks These structurally based comparisons of three MDR enzymes, QOR, GDH, and LADH, confirm that despite low sequence identity they possess a high degree of homology in their overall fold. This is true not only for the nucleotide-binding domain but also for the catalytic domain. This study presents a classic example of divergent evolution of enzymes from a common ancestor and demonstrates that despite a high degree of divergence within the amino acid sequences a stable enzyme fold has been retained. The catalytic zinc ion has been lost in QOR and z-crystallin, and those residues important for catalysis have been significantly altered to accommodate the new activity of the enzyme. We propose that the structure of z-crystallin is comparable to that of QOR and that z-crystallin has a similar active site. A structural framework for the MDR superfamily has been proposed based on the structurally conserved regions common to QOR, GDH, and

/ 6b15$$9366

02-27-96 12:06:43

arcas

LADH and we anticipate that such a framework may be used to accurately model the structures of other MDR enzymes. Predictions based solely on multiple sequence alignments may be misleading. Comparisons of the crystal structures of several members of an enzyme family may therefore provide greater insight and understanding of the structure–function relationships within an enzyme family and result in improved multiple sequence alignments. ACKNOWLEDGMENTS J.R. is supported by a Royal Society Fellowship (1995) and thanks M. J. Danson and D. W. Hough for the GDH project. We also thank N. E. Dixon for his contribution to the QOR project.

REFERENCES 1. Persson, B., Krook, M., and Jo¨rnvall, H. (1991) Eur. J. Biochem. 200, 537–543. 2. Scopes, R. K. (1983) FEBS Lett. 156, 303–306. 3. Jo¨rnvall, H., Persson, B., and Jeffrey, J. (1987) Eur. J. Biochem. 167, 195–201. 4. Persson, B., Zigler, J. S., and Jo¨rnvall, H. (1994) Eur. J. Biochem. 226, 15–22. 5. Eklund, H., Samama, J-P., and Jones, T. A. (1984) Biochemistry 23, 5982–5996. 6. Ramaswamy, S., Eklund, H., and Plapp, B. V. (1994) Biochemistry 33, 5230–5237. 7. Eklund, H., Nordstro¨m, B., Zeppezauer, E., So¨derlund, G., Ohlsson, I., Boiwe, T., So¨derbury, B-O., Tapla, O., Bra¨nde´n, C-I., and ˚ keson, A ˚ . (1976) J. Mol. Biol. 102, 27–59. A 8. Edwards, K. J., Thorn, J. M., Daniher, J. A., Dixon, N. E., and Ollis, D. L. (1994) J. Mol. Biol. 240, 501–503. 9. Thorn, J. M., Barton, J. D., Daniher, J. A., Dixon, N. E., Ollis, D. L., and Edwards, K. J. (1995) J. Mol. Biol. 249, 785–799. 10. John, J., Crennel, S. J., Hough, D. W., Danson, M. J., and Taylor, G. L. (1994) Structure 2, 385–393. 11. Rao, P. V., Krishna, C. M., and Zigler, J. S. (1992) J. Biol. Chem. 267, 96–102. 12. Rao, P. V., and Zigler, J. S. (1992) Biochim. Biophys. Acta 1117, 315–320. 13. Sun, H-W., and Plapp, B. V. (1992) J. Mol. Evol. 34, 522–535. 14. Jo¨rnvall, H., Persson, B., Du Bois, G. C., Lavers, G. C., Chen, J. H., Gonzalez, P., Rao, P. V., and Zigler, J. S., Jr. (1993) FEBS Lett. 322, 240–244. 15. Devereux, J., Haeberli, P., and Smithies, O. (1984) Nucleic Acids Res. 12, 387–395. 16. Bernstein, C. F., Koetzle, T. F., Williams, G. J. B., Meyer, E. F., Jr., Brice, M. D., Rodgers, J. R., Kennard, O., Shimanouchi, T., and Tasumi, M. (1977) J. Mol. Biol. 112, 535–542. 17. Al-Karadaghi, S., Cedergren-Zepperzauer, E. S., and Ho¨vmoller, S. (1994) Acta Crystallogr. D50, 793–807. 18. Bru¨nger, A. T. (1992) X-PLOR Version 3.1 Manual: A System for X-Ray Crystallography and NMR, Yale Univ. Press, New Haven, CT. 19. Lee, B., and Richards, F. M. (1972) J. Mol. Biol. 55, 379–400. 20. Jones, T. A. (1985) Methods Enzymol. 115, 157–171. 21. Connolly, M. L. (1983) Science 221, 709–713.

AP: Archives

COMPARISON OF FOUR MEDIUM-CHAIN DEHYDROGENASES 22. Nichols, A., Sharp, K. A., and Honig, B. (1991) Proteins: Struct. Funct. Genet. 11, 281–296. 23. Barton, G. J. (1990) Methods Enzymol. 183, 403–428. 24. Osguthorpe, D. (1989) Biochemist 11, 4–9. 25. Blundell, T., Carney, D., Gardner, S., Hayes, F., Howlin, B., Hubbard, T., Overington, J., Singh, D. A., Sibanda, B. L., and Sutcliffe, M. (1988) Eur. J. Biochem. 172, 513–520. 26. Ohlsson, I., Nordstro¨m, B., and Bra¨nde´n, C-I. (1974) J. Mol. Biol. 89, 339–354.

/ 6b15$$9366

02-27-96 12:06:43

arcas

183

27. Rao, S. T., and Rossman, M. G. (1973) J. Mol. Biol. 76, 241– 256. 28. Borra´s, T., Persson, B., and Jo¨rnvall, H. (1989) Biochemistry 28, 6133–6139. 29. Baker, P. J., Britton, K. L., Rice, D. W., Rob, A., and Stillman, T. J. (1992) J. Mol. Biol. 228, 662–671. 30. Wierenga, R. K., Terpstra, P., and Hol, W. G. J. (1986) J. Mol. Biol. 187, 101–107. 31. Fan, F., Lorenzen, J. A., and Plapp, B. V. (1991) Biochemistry 30, 6397–6401. 32. Kraulis, P. J. (1991) J. Appl. Crystallogr. 24, 946–950.

AP: Archives