Eur. J. Biochem. 248, (1997) 0 FEBS 1997

Transcription

1 Eur. J. Biochem. 248, (1997) 0 FEBS 1997 Mycothiol-dependent formaldehyde dehydrogenase, a prokaryotic medium-chain dehydrogenase/reductase, phylogenetically links different eukaroytic alcohol dehydrogenases Primary structure, conformational modelling and functional correlations Annika NORIN, Peter W. VAN OPHEM, Sander R. PIERSMA*, Bengt PERSSON, Johannis A. DUINE and Hans JORNVALL Department of Medical Biochemistry and Biophysics, Karolinska Institutet, Stockholm, Sweden Department of Microbiology and Enzymology, Delft University of Technology, Delft, The Netherlands (Received 17 February/l2 May 1997) - EJB /3 Prokaryotic mycothiol-dependent formaldehyde dehydrogenase has been structurally characterized by peptide analysis of the 360-residue protein chain and by molecular modelling and functional correlation with the conformational properties of zinc-containing alcohol dehydrogenases. The structure is found to be a divergent medium-chain dehydrogenase/reductase (MDR), at a phylogenetic position intermediate between the cluster of dimeric alcohol dehydrogenases of all classes (including the human foims), and several tetrameric reductaseddehydrogenases. Molecular modelling and functionally important residues suggest a fold of the mycothiol-dependent formaldehyde dehydrogenase related overall to that of MDR alcohol dehydrogenases, with the presence of the catalytic and structural zinc atoms, but otherwise much altered active-site relationships compatible with the different substrate specificity, and an altered loop structure compatible with differences in the quaternary structure. Residues typical of glutathione binding in class-i11 alcohol dehydrogenase are not present, consistent with that the mycothiol factor is not closely similar to glutathione. The molecular architecture is different from that of the constant alcohol dehydrogenases (of class-111 type) and the variable alcohol dehydrogenases (of class-i and class-i1 types), further supporting the unique structure of mycothiol-dependent formaldehyde dehydrogenase. Borders of internal chain-length differences between this and other MDR enzymes coincide in different combinations, supporting the concept of limited changes in loop regions within this whole family of proteins. Keywords: medium-chain dehydrogenase/reductase; formaldehyde/alcohol dehydrogenase; quaternary structure ; loop segment; molecular evolution. Alcohol dehydrogenase activity appears to be universal in eukaryotes, derived from enzymes of separate family assignments, and frequently of multiple occurrence in a complex fashion. Two such large protein families that have been much studied and of different derivation are the medium-chain dehydrogenasesheductases (MDR), and the short-chain dehydrogenased reductases (SDR). The MDR forms include the classical liver alcohol dehydrogenases with active-site zinc [l], and, for example, lens (-crystallin, which is apparently without zinc [2]. The SDR forms include the insect alcohol dehydrogenase, which has other structural features and an active-site Tyr-Xaa-Xaa-Xaa-Lys pattern [3] and enzymes that metabolize steroid hormones and Correspondence to H. Jornvall, Department of Medical Biochemistry and Biophysics, Karolinska Institutet, S Stockholm, Sweden Fax: hans.jornvall@mbb.ki.se Abbreviations. MDR, medium-chain dehydrogenase/reductase; SDR, short-chain dehydrogenaseheductase ; mycothiol, 1-0-(2 -[N-acetyl-~-cysteinyl]amido-2 -deoxy-a-d-glucopyranosyl-d-myoinositol. Enzymes. Alcohol dehydrogenase (EC 1.I.1.l) ; glutathione-dependent formaldehyde dehydrogenase (EC I) ; mycothiol-dependent formaldehyde dehydrogenase (EC ); chymotrypsin (EC I); trypsin (EC ); endoprotease Glu-C (EC ); endoprotease Lys-C (EC ); endoprotease Asp-N (EC ). prostaglandins [4]. ln prokaryotes, the pattern appears to be even more complex and several types have been detected [5]. Their structures and inter-relationships are little known. One such form is a factor-dependent formaldehyde dehydrogenase purified from the gram-positive methylotrophic bacterium Arnycolatopsis rnethanolicu [6] and other prokaryotes, such as Rhodococcus erythropolis [7]. It is known to be mycothiol-dependent formaldehyde dehydrogenase [8], mycothiol being the trivial name for 1-0-(2 - [N-acetyl-L-cyste~nyl]am~do-2 -deoxy-rx-~-glucopyranosyl)-d-niyoinositol. The enzyme has a broad substrate specificity, oxidizing higher aliphatic alcohols, as well as formaldehyde, the oxidation of formaldehyde requiring the mycothiol factor, while that of alcohols does not [6, 71. In this respect, the activity resembles that of one type of MDR, i.e. the class-111 alcohol dehydrogenase (91, which is identical to glutathione-dependent formaldehyde dehydrogenase [ 101 and has been characterized from archaeons [II, 121, human cell lines [13], prokaryotes [14, 151 and plants [16, 171. This class has constant structural properties 1181 and is the apparent ancestor of all the classes of the mammalian enzyme, through a series of gene duplications in the vertebrate line [16, 19, 201. Like the prokaryotic mycothioldependent formaldehyde dehydrogenase, the class-ill enzyme requires a factor for activity with formaldehyde, but that factor is glutathione [lo], while much lower activity with long-chain

2 Norin et al. (Eur: J. Biochem. 248) PQTVRGVIARSKGAPVELTDIVI PDPGPSEVTAL IATCAVCHTDLTYREGGINDEFPFLLGHEAAGTVESVGEGVDSVQPGD~~NWRA - I Intact t kc4-i P c t C 3 1 H OZ-----lt t; +-T7--It T16 T16 E26 +K3 + k K 6 ~ +op1- I OP2 b VCGQCRACKRGR?QYCFSTFNATQKMTLTDGTELTPALGIGAF~KTL~GQCTK~P~?AVAGLLGCG~GLG~~TGAVSR t c 5 - kc11 t c , ~ il 026 -k+ +T1 -i t- T9 T14 -T4 -I T15 IH I E25 b )-----K i~K7--Kl~I K10 t k O P 3 - l ~ O P 4 b Fig. 1. Primary structure of mycothiol-dependent formaldehyde dehydrogenase. The Val/Pro designation at position 85 indicates probable heterogeneity, where Pro was detected in one preparation and Val in another. Positions of all peptides analyzed are given by continuous lines for those positions that were sequenced. Peptide numbers refer to the elution order within each digest, while peptide letters indicate cleavage method: K (Lys-specific protease), D (Asp-specific protease), E (Glu-specific protease), T (trypsin), C (chymotrypsin), DP (acid cleavage of Asp-Pro bonds) SVAVIGCGAVGDAVIAGARLAG~KIIAVDRD~KLEWATELGATHT~ATETDWEAVQ~TGGFG~WIDAVGRPETWKQAFY~D * ~ ~ T1- +T6+1 T17 I +T~--IH t E25 + +El +-E2+ a El3 -E21+ ek10-b t K 2 + ~ K 1 2 +Kl1-0P AGTWLVGVPTPDMRLEMPLLDFFSRGGALKSSWYGDCLPERDFPVLIDLHLQGRLPLDKFVTERISLDDVE~FHTM~GEVLRS~ l-o23 +-Dl7 -+Dl --'LO ~~O T13-+-T16+ a TI5 -t - T12 ++T7+*T8+F T10 -i E I E27 It---E7 -+ El0 1 K11-1 K9 l -K5+-K8- alcohols is evident with the class-i11 alcohol dehydrogenase in the absence of glutathione [9] and the mycothiol-dependent prokaryotic enzyme in the absence of its factor [6]. Since glutathione is lacking in many gram-positive bacteria, the possibility exists that the mycothiol-dependent formaldehyde dehydrogenase is their class-i11 equivalent. Consistent with this possibility, the mycothiol-dependent formaldehyde dehydrogenase has a subunit molecular mass similar to those of MDR enzymes (40 kda), is zinc dependent, and has some alcohol dehydrogenase sequence similarity in a short N-terminal segment [6]. However, differences are also extensive, the mycothiol-dependent enzyme is apparently a trimer (MDR class-111 alcohol dehydrogenase is a dimer), and has about equal N-terminal sequence similarity with other MDR enzymes [6], which in turn have other inter-relationships [ 161. Hence, the structure and properties of the prokaryotic mycothiol-dependent formaldehyde dehydrogenase need clarification and are expected to yield information on basic relationships. In the present work, we have determined the primary structure of this dehydrogenase from A. methanolica, have found it to constitute a distant MDR analog, modelled it to that conformation [21-251, and found that it constitutes a distant relative, linking separate enzyme types. Furthermore, it exhibits an internal loop difference with special consequences that explain deviant properties. MATERIALS AND METHODS Protein analysis. Mycothiol-dependent alcohol dehydrogenase from Amycolatopsis methanolica, purified as described [6], was reduced with dithiothreitol and carboxymethylated by treatment with ''C-labelled iodoacetate in 6 M guanidine HCl [18]. Different samples of the carboxymethylated protein were treated with Achrornobacter lyticus Lys-specific protease (Waco), staphylococcal Glu-specific protease, Pseudomonas Asp-specific protease (both from Boehringer), trypsin (Worthington), and chymotrypsin (Sigma) in 0.1 M ammonium bicarbonate, ph 8.2. Peptides from these enzymatic digests were fractionated by reverse-phase HPLC (Waters 440 system) on Vydac C, (4.6 mmx250 mm), C, (2.1 mmx250 mm) and C,, (4.6 mm X250 mm) with gradients of acetonitrile (from 0 to 60 %J) in 0.1 % aqueous trifluoroacetic acid. For chemical cleavage of Asp-Pro bonds, the protein was treated with 70% formic acid for 24 h at room temperature and additionally for the same period at 37 C. Sequence degradations were performed with ABI 477 instruments or a MilliGen Prosequencer 6600, all equipped with on-line phenylthiohydantoin analysis. Amino acid compositions were determined with an LKB-Pharmacia Alpha-Plus instrument after hydrolysis in evacuated tubes with 6 M HCV0.5 % phenol at 110 C for 24 h. Comparisons, molecular modelling, substrate docking and phylogenetic relationships. The primary structure determined was compared with the amino acid sequences of other alcohol dehydrogenases and related enzymes in the SwissProt database [26]. Molecular modelling was performed with ICM (version 2.6, Molsoft LLC, 1996) [27] with the known threedimensional structure of human class-id alcohol dehydrogenase as template. Tethers were imposed between residues of the class-i structure and those of the mycothiol-dependent alcohol dehydrogenase structure. After minimization of the tethers, all methyl groups were rotated to reduce clashes, followed by iterative combined geometry and energy optimization. After adjustments of polar hydrogen positions, the whole molecule was subjected to free energy minimization. The loop comprising residues was further modelled in ICM using a Monte Carlo minimization p!ocedure after giving rotational freedom to all bonds within 5.0 A of this loop. Docking of formaldehyde-conjugated mycothiol to the subunit model was performed using non-rigid docking based upon a Monte Carlo procedure [27], allowing free movement of the substrate, the rotatable bonds of the substrate, and the x angles of the substrate-binding residues at positions 43,47, 52-54, 88, 89, 105, , 127, 128, 276, 277, 279,281,283 and , and with an additional distance restraint of A between the hydroxyl oxygen and the catalytic zinc. After the ini-

3 284 Norin et al. (Em J. Biackern. 248) Fig. 2. Alignment of mycothiol-dependent formaldehyde dehydrogenase (MD-FDH) and human alcohol dehydrogenase of class I (ADH class I), yeast ethanol dehydrogenase (ADH yeast) and quinone oxidoreductase (QOR) of Escherichiu cozi. Strictly conserved residues in all these enzymes are marked against a gray background. Deviating loop structures are coloured green (ADH class I), red (MD-FDH), yellow (ADH yeast), and orange (QOK). Arrows indicate the positions of the ADH class I zinc ligands. tial docking, the distance restraint was removed and the substrate and substrate-binding residues were subjected to energy minimizations. Phylogenetic relationships were evaluated from tree constructions using the program CLUSTAL W [28] with bootstrap analysis for statistical evaluation [29]. RESULTS Primary structure. The primary structure of mycothiol-dependent formaldehyde dehydrogenase was determined by sequence analysis of the intact carboxymethylated protein and of constituent peptides obtained through five different proteolytic treatments (data not shown). The results show a 360-residue primary structure (Fig. l), corresponding to a subunit molecular mass of 37.7 kda. All regions were obtained in overlapping segments, but the segment corresponding to positions was frequently obtained in low yield, presumably because of oxidative modifications of Cys residues at positions 38 and 41. To secure this part of the structure, acid cleavage of the Asp-Pro bond between positions 25 and 26 (and of those between 148 and 149 and between 152 and 153) was utilized. Evidence for heterogeneity was obtained at position 85, where Pro was detected in one preparation (used for the digest with the Asp protease and detected in peptide D22; Fig. I), while Val was detected in another preparation (used for digestions with chymotrypsin and trypsin, and therefore detected in peptides C17 and T16; Fig. 1). The protein N-terminus was not blocked, in contrast to that of many MDR alcohol dehydrogenases [30], and direct sequencer analysis of the N-terminal segment was possible in agreement with earlier observations [6]. The structure obtained was identical to that of a previous report until residue 22 but the structures deviate after that, presumably because of earlier, out-ofphase interpretations. The entire structure now obtained is in good agreement with the total composition of the purified protein after hydrolysis (Table 1).

4 Table 1. Total composition of mycothiol-dependent formaldehyde dehydrogenase. Hydrolytic values are the ratios obtained after acid hydrolysis at 110 C for 24 h with 6 M HCU0.5 % phenol, without corrections for slow release, destruction or impurities. The low recovery of Val is consistent with nine Val-Val, Val-Val-Val, Val-Ile, or Val-Val-Ile sequences. Sequence values refer to the variant form with Va185. Amino acid Hydrolytic value Sequence value Norin et al. (EUK J. Biochenz. 248) 285 molar ratio CYS AspIAsn Thr Ser GlulGln Pro GlY Ala Val Met Ile Leu TYr Phe TrP His LYS Arg Sum 10 26r / Fig. 3. Molecular model of the mycothiol-dependent enzyme. Only one structural replacement shows extensive differences from that of the crystallographic structure, i.e. the loop coloured red in mycothiol-dependent formaldehyde dehydrogenase and green in human class-i alcohol dehydrogenase. The starts of the corresponding loop segments are indicated for mycothiol-dependent formaldehyde dehydrogenase (Nf, red), class-i alcohol dehydrogenase (Na, green), quinone oxidoreductase (Nq, orange) and yeast alcohol dehydrogenase (Ny, yellow) showing them to to differ substantially, while the C-terminal ends coincide. The red arrow indicates a deletion in mycothiol-dependent formaldehyde dehydrogenase compared with class-i alcohol dehydrogenase. Table 2. Overall residue identities of mycothiol-dependent formaldehyde dehydrogenase with human alcohol dehydrogenases classes I-V and microbial alcohol dehydrogenases. The yeast enzyme is the classical, ethanol-active yeast alcohol dehydrogenase [30, 311, which is named separately to distinguish it from the recently characterized yeast class-i11 alcohol dehydrogenase (32, 331 and the human classical liver alcohol dehydrogenase class I. Similarity of A. methnnolica mycothioldependent formaldehyde dehydrogenase to Human alcohol dehydrogenase class I Human alcohol dehydrogenase class I1 Human alcohol dehydrogenase class I11 Human alcohol dehydrogenase class IV Human alcohol dehydrogenase class V Yeast ethanol dehydrogenase Yeast alcohol dehydrogenase class 111 Pnrncoccus alcohol dehydrogenase Residue identities % Structural characteristics. The primary structure determined is unique but distantly related to those of medium-chain alcohol dehydrogenases and other MDR enzymes. Relationships are only at the 20-30% level, but cover the entire protein chains, and ascribe the protein to the MDR family. Alignments show that overall relationships are about equally distant with the separate classes of mammalian alcohol dehydrogenases, and with the prokaryotic and other microbial enzymes (Table 2) Nevertheless, throughout the system, and independent of species, relationships are somewhat closer with class-ill alcohol dehydrogenases than with the other classes, supporting the view that class 111 is an early ancestor, reflecting its link with this form of early separation. The alignments identify an internal segment with insertions and deletions compared with alcohol dehydrogenases and other MDR enzymes (Fig. 2). The position of these internal insertions and deletions correspond to a superficial loop in the tertiary structure of those MDR enzymes that have been crystallographically analyzed, the horse class-i [ 211, human class-i [22], and cod class-1411 hybrid [24] alcohol dehydrogenases, and prokaryotic [-crystallin [23 1. Furthermore, the absence of parts of this loop segment in other MDR enzymes, such as yeast alcohol dehydrogenase [31 J and sorbitol dehydrogenase [35], has been ascribed to a more complex quateriiary structure than that of the dimeric mammalian liver forms, which have the loop. Consequently, the present structure was submitted to molecular modelling starting from the human class-i alcohol dehydrogenase to evaluate the loop structure and other segments. The modelling confirms the overall relationships, with closely related tertiary structures, thus assigning the tertiary structure of the enzyme to the group of typical MDR enzymes, in spite of the low sequence similarity. The modelling further shows that the loop structure is unique to this mycothiol-dependent enzyme, which deviates from the fold of the other MDR enzymes in just the loop segment (Fig. 3). The loop differences are regular, ending at identical places in all the highly divergent forms analyzed, i.e. class-i alcohol dehydrogenase, sorbitol dehydrogenase and quinone oxidoreductase (Figs 2 and 3). The present enzyme has a loop size intermediate between that of the dimeric liver alcohol dehydrogenases and the tetrameric sorbitol dehydrogenases and [-crystallins (Fig. 3). Hence, the previous conclusion, that this loop structure is responsible for differences in quaternary structure [31, 351, is supported. The mycothioldependent enzyme has been reported to be a triiner [6), and it appears possible that the loop nature provides an explanation for this special quaternary structure. Regarding residues of functional importance, cysteine residues in the mycothiol-dependent enzyme are found at exactly those positions where cysteine residues ligand the two zinc atoms of the liver alcohol dehydrogenase subunits, i.e. at positions 41, 161,92,95, 98, 106 (Fig. 2), which correspond to positions 46, 174 (active-site zinc atom), 97, 100, 103 and 121 (structural zinc atom), respectively, in class-i human alcohol de-

5 286 Norin et al. (Em J. Biochem. 248) Fig. 4. Stereo-drawing of the side-chains at the active site of mycothiol-dependent alcohol dehydrogenase model (red) and class-i alcohol dehydrogenase (blue). The catalytic zinc atom is shown as a blue sphere (CPK model). Fig. 5. Stereoview of formaldehyde-conjugated mycothiol (green) docked into the active site of the modelled mycothiol-dependent formaldehyde dehydrogenase subunit. Side-chains of residues within 3 A of the substrate are coloured red. The catalytic zinc atom is shown as a gray sphere (CPK model). The backbone is shown in black. The general orientation of the subunit is with the coenzyme-binding domain toward the bottom and the catalytic domain toward the top. hydrogenase. At the active site, the mycothiol-dependent ADH has two tryptophan residues. Trp304 substitutes Val318 in class I and is oriented similarly (Fig. 4), while Trp88 replaces Phe93 and points in a different direction because of space restrictions. Residues 43, 52 and 279 (corresponding to 48, 57 and 294 in class I) at the other side of the substrate-binding pocket occupy positions mainly similar to those in the class-i enzyme. However, Thr94 in class I is replaced by Arg89, the side-chain of which prolongs into the active site, indicating a possible function of this position in substrate binding. The docking calculations show that there is sufficient space for the formaldehyde-conjugated mycothiol close to the zinc atom at the active site (Fig. 5). Most of the surrounding residues are hydrophobic, but Arg89, Asnlll and Gln114 constitute the closest polar residues and may be of interest in binding. However, at the distant end of the mycothiol, the modelling need not be final and would be influenced by the subunit interactions, which are likely to be altered, but are unknown in the mycothiol-dependent enzyme. In conclusion, apart from the active-site and loop differences, and the possibly altered quaternary structure, the modelling is compatible with largely unaltered positions for major parts of the subunit, including the catalytic zinc and all zinc ligands. Although common in MDR enzymes, this pattern of functional conservation is not mandatory in the family, and several MDR members lack one or both of the zinc atoms and corresponding features [2, 351. Phylogenetic relationships. Calculation of an evolutionary tree shows that the mycothiol-dependent enzyme occupies a position intermediate between two clusters of other MDR enzymes [l], thus linking the cluster of vertebrate alcohol dehydrogenases with that of yeast ethanol dehydrogenaselvertebrate sorbitol dehydrogenaselother enzymes (Fig. 6). This intermediate position gives support to the characteristics of both groups, and will help to delineate more distantly related forms. It also establishes the importance of substrate-associated factors, such as glutathione in the class-i11 form and mycothiol in the present form, but it does not suggest a close relationship with the glutathione-dependent vertebrate alcohol dehydrogenases. Furthermore, as shown above, the intermediate phylogenetic position appears relevant in relation to the quaternary structure of the enzyme and its segment of internal loop differences. DISCUSSION Mycothiol-dependent formaldehyde dehydrogenase. The analysis establishes the nature of the prokaryotic mycothiol-dependent formaldehyde dehydrogenase. It is an MDR, and has a position phylogenetically intermediate between the two clusters of dimeric alcohol dehydrogenases and tetrameric reductased dehydrogenases (Fig. 6). Liganding residues at the catalytic and structural zinc atoms of alcohol dehydrogenases are present in the mycothiol-dependent enzyme (Fig. 2), and molecular modelling of this protein suggests that the conformation is related to that of the zinc-containing alcohol dehydrogenases and other MDR enzymes (Fig. 3). Only an internal loop structure shows pronounced differences (Fig. 3) that correlate with altered quaternary structures. Hence, we conclude that the mycothiol-dependent formaldehyde dehydrogenase is an MDR enzyme with active-site zinc and a binding site for coenzyme and that is related to those of other MDR enzymes, but with a unique factor binding.

6 Norin et al. (ELK J. Biochem. 248) 287 Paracoccus Ill \ Human SDH / distinct enzyme, with different binding interactions, and a separate factor, mycothiol [8]. The evolution of a separate form of formaldehyde dehydrogenase in these bacteria is consistent with the concept that formaldehyde elimination is crucial. Not only is the formaldehyde-active class-i11 alcohol dehydrogenase the parent form of the animal alcohol dehydrogenases [18] and are the catalytic properties toward formaldehyde increased in microorganisms [34], but a separate form of formaldehyde dehydrogenase is present in microorganisms. It may be concluded that formaldehyde elimination is universally important and has evolved in different manners, similarly to the ethanol dehydrogenase, which is found in multiple forms of repeated origin in different organisms [ 381. MD-FDH Thermophilus GDH I Rat ER Fig. 6. Unrooted phylogenetic tree illustrating relationships between the structnre of rnycothiol-dependent formaldehyde dehydrogenase (MD-FDH) and other MDR proteins. Roman numbers indicate classes I-V of the human enzyme and class 111 of the enzyme from other sourceb. SDH, sorbitol dehydrogenase; TDH, threonine dehydrogenase; EtOH, ethanol-active alcohol dehydrogenase; jcr, [-crystallin; QOR, quinone oxidoreductase ; ER, enoyl reductase; GDH, glucose dehydrogenase. Mycothiol-dependent formaldehyde dehydrogenase is intermediate between the left-hand cluster of class-i11 alcohol dehydrogenase classes derived therefrom (as revealed by those human forms), and the righthand cluster of reductases and other non-alcohol dehydrogenases. Evidence for heterogeneity was detected at one position (position 85) where one preparation gave Pro and another Val. Although further positions with contradictory data were not obtained, additional microheterogeneities should not be excluded. Functional assignments. Since the overall assignment has been established, it is meaningful to consider differences that can explain the mycothiol-dependent nature of the enzyme. Several residue differences compared with alcohol dehydrogenases affect the substrate-binding site (Fig. 4). In particular, residues lining the inner part of the substrate pocket in alcohol dehydrogenase have been replaced, such as PheY3 (class I)/Tyr (class 111), which is changed to Trp88 in the present enzyme, Phel40 (class I)/Tyr (class HI), which is changed to Ala127, and Thr94 (classi)/ile (class HI), which is changed to Arg89. In total, 8 compared with class I11 and 10 compared with class I, of 12 lining residues have been changed. Thr46 of the present enzyme corresponds to His51 of alcohol dehydrogenase class I, which has been ascribed a role in the proton-release mechanism of the reaction. The His residue has been replaced by other residues in human class-ti and class-i11 alcohol dehydrogenases, and in several other MDR enzymes. Therefore, the release mechanism is different in the mycothiol-dependent and the other MDR enzynies. In addition, residues of particular importance in glutathione binding, Arg115 and Asp57 of class-i11 alcohol dehydrogenase [36, 371, are not conserved. However, there is an arginin? residue, Arg89, much closer to the active-site Zn atom (7.9 A between C[ and Zn) in the present enzyme (Fig.4) than the glutathione-)inding Arg11.5 is in class-i11 alcohol dehydrogenase (14.7 A; Brookhaven Data Bank, entry ITEH), suggesting that charge relationships differ in space between the two formaldehyde dehydrogenases. It may be concluded that the present enzyme is not just a prokaryotic class-i11 alcohol dehydrogenase, as perhaps anticipated from its dependence on a factor, but a Molecular architecture. Regarding the molecular architecture of MDR enzymes, two patterns are relevant, and illustrated by the present enzyme. One concerns the areas of above-average variability, reflecting emergence of activity. Three such segments have been defined in the class-i alcohol dehydrogenase family, and two in the class-111 family [I 81. The mycothiol-dependent enzyme has none of these extremes of segment variability, further supporting its unique role and class-independent origin outside the cluster of vertebrate alcohol dehydrogenases. Instead, the present enzyme has segment variability that overlaps in part with those of class I or class 111. Nevertheless, segments around the ligands to the active site zinc atom are well conserved, especially with those of class-i11 enzymes, supporting a conservation pattern in this active-site property, in particular toward class 111, and hence a functional equivalence of the two formaldehyde-metabolizing enzymes. The other architecture of special interest concerns the loop structures adjacent to the second zinc atom. In this segment, i- crystallin is lacking a long continuous part, compatible with the absence of two loop regions present in liver alcohol dehydrogenase [2]. Sorbitol dehydrogenase is lacking half of this segment, compatible with the absence of one of the loop regions [35], while the human class-i1 enzyme has an internally elongated protein chain in this segment [IY]. These properties coincide with border segments in the mycothiol-dependent enzyme (Fig. 2). Thus, a segment is missing in this region of this enzyme, and the C-terminal end of that segment in mycothiol-dependent formaldehyde dehydrogenase coincides closely with that of the missing segment in yeast alcohol dehydrogense, quinone oxidoreductase (Fig. 2) and sorbitol dehydrogenase Similarly, the start position of that missing segment in the present structure is adjacent to the start of the missing segment in the class-1 and class-111 enzymes compared with class-ll alcohol dehydrogenase Consequently, border regions are not random but occur in very similar positions when present. Hence, these missing or extra segments probably refer to single genetic events of insertions or deletions, reflecting the successive evolution of the entire family, and fundamental additions or removals of entire loops, without affecting remaining parts of the tertiary structure. Differences in these segments have been suggested to correlate with the quaternary structures. [-Crystallin, sorbitol dehydrogenase and yeast alcohol dehydrogenase, lacking large parts of the internal segment, are tetrameric, while liver alcohol dehydrogenases, having the entire segment, are dimeric 12, 31, 351. Tt therefore appears significant that the present mycothioldependent formaldehyde dehydrogenase, with its intermediate position (Fig. 6) and intermediate relationship in the internal segment (Fig. 2), has an intermediate, trimeric quaternary structure 161. In conclusion, the structure shows a unique protein form, intermediate between those of previously characterized MDR

7 288 Norin et al. (Eul: J. Biochem. 248) enzymes, linking separate functions by residue replacements at functionally important positions, and showing considerable loop variations in otherwise conserved structures. This work was supported by the Swedish Medical Research Council (projects 13X-3532 and 03P-11312), and the Dutch National Science Foundation (NWO). REFERENCES 1. Persson, B., Zigler, J. S. Jr & Jornvall, H. (1994) A super-family of medium-chain dehydrogenasesheductases (MDR). Sub-lines including [-crystallin, alcohol and polyol dehydrogenases, quinone oxidoreductases, enoyl reductases, VAT-1 and other proteins, Eur: J. Biochem. 226, Borris, T., Persson, B. & Jornvall, H. (1989) Eye lens c-crystallin relationships to the family of long-chain alcohol/polyol dehydrogenases. Protein trimming and conservation of stable parts, Biochemistry 28, Jiirnvall, H., Persson, B., Krook, M., Atrian, S., GonzBlez-Duarte, R., Jeffery, J. & Ghosh, D. (1995) Short-chain dehydrogenasesl reductases (SDR), Biochemistry 34, Krook, M., Marekov, L. & J8rnval1, H. (1990) Purification and structural characterization of placental NAD -linked 15- hydroxyprostaglandin dehydrogenase. The primary structure reveals the enzyme to belong to the short-chain alcohol dehydrogenase family, Biochemistry 29, van Ophem, P. W. & Duine, J. A. (1993) Microbial alcohol, aldehyde and formate ester oxidoreductases, in Enzymology and molecular biology qf carbonyl metabolism 4 (Weiner, H., Crabb, D. W. & Flynn, T. G., eds) pp , Plenum, New York. 6. van Ophem, P. W., Van Beeumen, J. & Duine, J. A. (1992) NADlinked, factor-dependent formaldehyde dehydrogenase or trimeric, zinc-containing, long-chain alcohol dehydrogenase from Amycolatopsis Nzethanolica, Eur J. Biochem. 206, Eggeling, L. & Sahm, H. (1985) The formaldehyde deydrogenase of Rhodococcus erythropolis, a trimeric enzyme requiring a cofactor and active with alcohols, Eul: J. Biochem. 150, Misset-Smits, M., van Ophem, P. W., Sakuda, S. & Duine, J. A. (1997) Mycothiol, 1-0-(2 -[N-acetyl-L-cysteiny1]amido-2 -deoxy- ~-D-glUCOpyranoSyl)-D-myo-inOSitOl, is the factor of NADlfactordependent formaldehyde dehydrogenase. FEBS Lett. 409, Vallee, B. L. & Bazzone, T. J. (1983) Isozymes of human liver alcohol dehydrogenase, Curr: Top. Biol. Med. Res. 8, Koivusalo, M., Baumann, M. & Uotila, L. (1989) Evidence for the identity of glutathione-dependent formaldehyde dehydrogenase and class I11 alcohol dehydrogenase, FEBS Lett. 257, Ammendola, S., Raia, C. A,, Caruso, C., Camardella, L., D Auria, S., De Rosa, M. & Rossi, M. (1992) Thermostable NAD -dependent alcohol dehydrogenase from Sulfolobus solfataricus: gene and protein sequence determination and relationship to other alcohol dehydrogenases, Biochemistry 31, Cannio, R., Fiorentino, G., Carpinelli, P, Rossi, M. & Bartolucci, S. (1996) Cloning and overexpression in Escherichia coli of the genes encoding NAD-dependent alcohol dehydrogenase from two Sulfolobus species, J. Bucteriol. I78, Kaiser, R., Holmquist, B., Hempel, J., Vallee, B. L. & Jornvall, H. (1988) Class 111 human liver alcohol dehydrogenase: a novel structural type equidistantly related to the class I and class I1 enzymes, Biochemistry 27, Gutheil, W. G., Holmquist, B. & Vallee, B. L. (1992) Purification, characterization, and partial sequence of the glutathione-dependent formaldehyde dehydrogenase from Escherichia Cali: a class I11 alcohol dehydrogenase, Biochemistry 31, IS. Ras, J., van Ophem, P. W., Reijnders, W. N. M., van Spanning, R. J. M., Duine, J. A., Stouthamer, A. H. & Harms, N. (1995) Isolation, sequencing, and mutagenesis of the gene encoding NAD- and glutathione-dependent formaldehyde dehydrogenase (GD-FALDH) from Paracoccus denitrifcuns, in which GD-FALDH is essential for methylotropic growth, J. Bucteriol. 177, Shafqat, J., El-Ahmad, M., Danielsson, O., Martinez, M. C., Persson, B., Parts, X. & Jornvall, H. (1996) Pea formaldehydeactive class 111 alcohol dehydrogenase: common derivation of the plant and animal forms but not of the corresponding ethanolactive forms (classes I and P), Proc. Nut1 Acad. Sci. USA 93, Martinez, M. C., Achkor, H., Persson, B., Fernindez, M. R., Shafqat, J., Farrts, J., Jornvall, H. & Parks, X. (1996) Arabidopsis formaldehyde dehydrogenase. Molecular properties of plant class 111 alcohol dehydrogenase provide further insights into the origins, structure and function of plant class P and liver class I alcohol dehydrogenases, Eul: J. Biochem. 241, Danielsson, O., Atrian, S., Luque, T., Hjelmqvist, L., Gonzilez-Duarte, R. & Jornvall, H. (1994) Fundamental molecular differences between alcohol dehydrogenase classes, Proc. Natl Acad. Sci. USA 91, Jornvall, H., Hoog, J.-O., von Bahr-Lindstrom, H. & Vallee, B. L. (1987) Mammalian alcohol dehydrogenases of separate classes : intermediates between different enzymes and intraclass isozymes, Proc. Natl Acad. Sci. USA 84, Parts, X., Cederlund, E., Moreno, A., Hjelmqvist, L., Farrks, J. & Jornvall, H. (1994) Mammalian class IV alcohol dehydrogenase (stomach ADH): structure, origin and correlation with enzymology, Proc. Natl Acad. Sci. USA 91, Eklund, H., Nordstrom, B., Zeppezauer, E., Soderlund, G., Ohls?on, I., Byiwe, T., Soderberg, B.-O., Tapia, O., Branden, C.-I. & Akeson, A. (1976) Three-{imensional structure of horse liver alcohol dehydrogenase at 2.4 A resolution, J. Mol. Biol. 102, Hurley, T. D., Bosron, W. F., Hamilton, J. A. & Amzel, L. M. (1991) Structure of human plpl alcohol dehydrogenase : catalytic effects of non-active-site substitutions. Proc. Nut1 Acad. Sci. USA 88, Thorn, J. M., Barton, J. D., Dixon, N. E., Ollis, D. L. & Edwards, K. J. (1995) Crystal structure of Escherichia coli QOR quinone oxidoreductase completed with NADPH, J. Mol. Biol. 249, Ramaswamy, S., El-Ahmad, M., Danielsson, O., Jornvall, H. & Eklund, H. (1996) Crystal structure of cod liver class I alcohol dehydrogenase : Substrate pocket and structurally variable segments, Protein Sci. 5, Moreno, A,, Farrks, J., Parks, X., Jornvall, H. & Persson, B. (1996) Molecular modelling of human gastric alcohol dehydrogenase (class IV) and substrate docking: differences towards the classical liver enzyme (class I), FEBS Lett. 395, Bairoch, A. & Apweiler, R. (1996) The SWISS-PROT protein sequence data bank and its new supplement TREMBL, Nucleic Acids Res. 24, Totrov, M. & Abagyan, R. (1994) Detailed ab initio prediction of lysozyme-antibody complex with 1.6 A accuracy, Nut. Struct. Biol. 1, Thompson, J. D., Higgins, D. G. & Gibson, T. J. (1994) CLUSTAL W: improving the sensitivity of progressive multiple sequence alignment through sequence weighting, position-specific gap penalties and weight matrix choice, Nucleic Acid.y Res. 22, Felsenstein, J. (1985) Confidence limits on phylogenies: an approach using the bootstrap, Evolution 39, Hjelmqvist, L., Hackett, M., Shafqat, J., Danielsson, O., Iida, J., Hendrickson, R. C., Michel, H., Shabanowitz, J., Hunt, D. F. & Jornvall, H. (1995) Multiplicity of N-terminal structures of medium-chain alcohol dehydrogenases. Mass-spectrometric analysis of plant, lower vertebrate and higher vertebrate class I, 11, and III forms of the enzyme, FEBS Lett. 367, Jornvall, H., Eklund, H. & Brandkn, C.-I. (1978) Subunit conformation of yeast alcohol dehydrogenase, J. Biol. Chem. 253, Bennetzen, J. L. & Hall, B. D. (1982) The primary structure of the Saccharomyces cerevisiae gene for alcohol dehydrogenase I, J. Bid. Chem. 257, Wehner, E. P., Rao, E. & Brendel, M. (1993) Molecular structure and genetic regulation of SFA, a gene responsible for resistance to formaldehyde in Saccharomyces cerevisiae, and characterization of its protein product, Mol. Gen. Genet. 237,

8 Norin et al. (Eur. J. Biochem FernBndez, M. R., Biosca, J. A,, Norin, A., Jornvall, H. & Paris, X. (1995) Class 111 alcohol dehydrogenase from Saccharomyces cerevisiae: Structural and enzymatic features differ toward the human/mammalian forms in a manner consistent with functional needs in formaldehyde detoxication, FEBS Left. 370, Eklund, H., Horjales, E., Jornvall, H., BrandCn, GI. & Jeffery, J. (1985) Molecular aspects of functional differences between alcohol and sorbitol dehydrogenases, Biochemistry 24, Engeland, K., Hoog, J.-O., Holmquist, B., Estonius, M., Jornvall, H. & Vallee, B. L. (1993) Mutation of Arg-115 of human class 111 alcohol dehydrogenase: a binding site required for formaldehyde dehydrogenase activity and fatty acid activation, Proc. Nafl Acud. Sci. USA YO, Estonius, M., Hoog, J.-O., Danielsson, 0. & Jomvall, H. (1994) Residues specific for class 111 alcohol dehydrogenase. Site-directed mutagenesis of the human enzyme. Biochemistry 33, Jornvall, H. (1994) The alcohol dehydrogenase system, in Toward a molecular basis of alcohol use und abuse (Jansson, B., Jornvall, H., Rydberg, U., Terenius, L. & Vallee, B. L., eds) pp , Birkhauser, Basel. 39. Eklund, H., Miiller-Wille, P., Horjales, E., Futer, O., Holmquist, B., Vallee, B. L., Hoog, J.-O., Kaiser, R. & Jornvall, H. (1990) Comparison of three classes of human liver alcohol dehydrogenase. Emphasis on different substrate-binding pockets, Eur: J. Biochevn. 193, Supplementary material. Mycothiol-dependent formaldehyde dehydrogenase, a prokaryotic medium-chain dehydrogenasel reductase, phylogenetically links different eukaroytic alcohol dehydrogenases. Fig. S1. Purification of the major digests analyzed, accounting for most of the peptides investigated. This information is available on request from the Editorial Office. One page is available.