Structure of quinoprotein methylamine dehydrogenase at 0 2.25 A resolution

The EMBO Journal vol.8 no.8 pp.2171 -2178, 1989 Structure of quinoprotein methylamine dehydrogenase at 0 2.25 A resolution F.M.D.Vellieux, F.Huitema, H.Groendijk, K.H.Kalk, J.Frank Jzn.', J.A.Jongejan1, J.A.Duinel, K.Petratos2, J.Drenth and W.G.J.Hol Laboratory of Chemical Physics, University of Groningen, Nijenborgh 16, 9747 AG Groningen, 'Department of Microbiology and Enzymology, Delft University of Technology, Julianalaan 67, 2628 BC Delft, The Netherlands and 2EMBL Outstation, c/o DESY, Notkestrasse 85, 2000 Hamburg 52, FRG Communicated by W.G.J.Hol The three-dimensional structure of quinoprotein methyl- amine dehydrogenase from Thiobacillus versutus has been determined at 2.25 A resolution by a combination of multiple isomorphous replacement, phase extension by solvent flattening and partial structure phasing using molecular dynamics refinement. In the resulting map, the polypeptide chain for both subunits could be followed and an X-ray sequence was established. The tetrameric enzyme, made up of two heavy (H) and two light (L) subunits, is a flat parallelepiped with overall dimensions of -76 x 61 x 45 A. The H subunit, comprising 370 residues, is made up of two distinct segments: the first 31 residues form an extension which embraces one of the L subunits; the remaining residues are found in a disc-shaped domain. This domain is formed by a circular arrangement of seven topologically identical four- stranded antiparallel f-sheets, with - 7-fold symmetry. In spite of distinct differences, this arrangement is reminiscent of the structure found in influenza virus neuraminidase. The L subunit consists of 121 residues, out of which 53 form a fl-sheet scaffold of a central three-stranded antiparallel sheet flanked by two shorter two-stranded antiparallel sheets. The remaining residues are found in segments of irregular structure. This subunit is stabilized by six disulphide bridges, plus two covalent bridges involving the quinone co-factor and residues 57 and 107 of this subunit. The active site is located in a channel at the interface region between the H and L subunits, and the electron density in this part of the molecule suggests that the co-factor of this enzyme is not pyrrolo quinoline quinone (PQQ) itself, but might be instead a precursor of PQQ. Key words: methylamine dehydrogenase/quinoproteins/ protein crystallography/enzymes/pro-PQQ Introduction Oxidoreductases require the presence of co-factors to be enzymatically active. The structures of enzymes containing, e.g. FAD, FMN, NAD and NADP as co-enzymes (see, e.g. Rossman et al., 1975; Filman et al., 1982; Lindquist and Branden, 1985; Karplus and Schulz, 1987; Schreuder et al., 1988) are well characterized, but no structures have been reported so far of enzymes from which the recently discovered co-factor pyrrolo quinoline quinone (PQQ) has been obtained. In most of these PQQ-containing enzymes (which are collectively called 'quinoproteins'), the co-factor becomes detached from the protein on denaturation and apo-quinoproteins are effectively reconstituted with PQQ (Duine and Jongejan, 1989). Most interesting is the recent finding that PQQ is also the co-factor of several mammalian enzymes, including bovine plasma amine oxidase (Lobenstein-Verbeek et al., 1984), bovine adrenal dopamine f-hydroxylase (Van der Meer et al., 1988) and human placental lysyl oxidase (Van der Meer and Duine, 1986). In these cases, the co-factor is covalently bound to the protein and PQQ is detected in the form of a hydrazone after the enzyme has reacted with a hydrazine followed by pronase treatment. Although this indicates that the co-factor is attached via amide or ester bonds, no three-dimensional (3D) structure of a quinoprotein is available so that it is unknown how its co-factor is bound and which residues are involved. Methylamine dehydrogenase (EC 1.4.99.3) is an attractive model to study this aspect since it has similarity with the mammalian quinoprotein amine oxidases (EC 1.4.3.6) (Lobenstein-Verbeek et al., 1984). Methylamine dehydrogenase catalyses the following reaction: CH3NH3+ + H20 HCHO + NH4+ + 2H+ + 2e- A blue copper protein, amicyanin, acts as electron acceptor for the dehydrogenase. All methylamine dehydrogenases investigated so far have a similar composition, consisting of two large, or heavy (H), and two small, or light (L), subunits, each small subunit containing a covalently bound co-factor. Sequence studies showed that this co-factor is attached at two sites to the protein chain (Ishii et al., 1983). Based on ESR and ENDOR spectroscopy the co-factor was proposed to be PQQ or a PQQ-like compound (De Beer et al., 1980). Recently, the latter suggestion received further attention when a PQQ-like structure was proposed, which lacked the carboxylic acid groups of PQQ, and was attached to the protein with Cys thio ether and Ser oxygen ether linkages (McIntire and Stults, 1986). However, application of the hydrazine method (Van der Meer et al., 1987) produced a hydrazone with the chromatographic and spectral properties of the hydrazone of authentic PQQ. Moreover, the hydrazone could be converted into PQQ, as judged from chromatographic and spectral data as well as activity in a bioassay. Therefore, it was concluded that the co-factor in methylamine dehydrogenase is PQQ, attached to the small subunit with pronase scissable bonds (Van der Meer et al., 1987). Methylamine dehydrogenase (or MADH) from the methylotrophic bacterium Thiobacillus versutus has a mol. 2171©cIRL Press F.M.D.Vellieux et al. Fig. 1. The structure of T. versutus methylamine dehydrogenase. Stereo view of the H2L2 tetramer along the crystallographic axis relating two HL dimers. The large, or H, subunits are shown in green; the small, or L, subunits in blue. Note how the N terminus extension of one large subunit 'embraces' the small subunit. Fig. 2. View of the small subunit in stereo with the ,B-strands in blue, the loops in green and the six disulphide bridges bridges in yellow. The model for the quinone co-factor is shown in red. wt of - 121 000, with the L and H subunits having a M, of 12 900 and 47 500 respectively (Vellieux et al., 1986). We describe here the X-ray structure of this quinoprotein. In the latest 2.25-A electron density map obtained, the polypeptide chains for both L and H subunits could be followed, an 'X-ray sequence' established, six disulphide bridges determined and the position of the co-factor elucidated. Results Molecular structure and subunit - subunit interactions The tetrameric enzyme is a flat parallelepiped with overall dimensions of - 76 x 61 x 45 A (Figure 1). The molecule can most easily be described as a dimer of HL dimers. The MADH tetramer is held together by extensive interactions taking place between subunits. It is interesting to note that no direct interactions occur between the two L subunits within the tetramer, as had earlier been predicted from cross-linking experiments (Matsumoto and Tobari, 1978b). The interactions which do occur can be divided into three classes. First, the less extensive interactions (in terms of the surface area involved) are found between the two large subunits, symmetrically across the molecular 2-fold axis (Figure 1). These H-H' interactions are quite limited in number since they only involve two residues from one loop in contact with two residues of the other, H', subunit. The second type of interaction is between the H subunit and the L' subunit belonging to a different H'L' dimer. These H - L' contacts mainly involve the interaction between the N-terminal arm of the H subunit, including the terminal helix, which 'embraces' the L' subunit (Figure 1). Finally, the most extensive interactions occur between H and L subunits within one HL dimer. These interactions involve many aromatic residues which might be important, since the active site cavity is found near this interface. Folding pattern of the small, co-factor binding, subunit In our present model the L subunit consists of 121 amino acid residues. This subunit contains a total of seven fl-strands connected by several stretches of rather irregular structure. The central element of this subunit is a long three-stranded antiparallel ,8-sheet with the usual left-handed twist formed by strands 4, 5 and 7 (Figures 2 and 5). This main structural element is flanked at the 'back' by two shorter two-stranded antiparallel sheets formed by strands 2 and 3, and by strands 1 and 6. Thus, the subunit is formed by a scaffold of 53 2172 3D structure of quinoprotein methylamine dehydrogenase Fig. 3. A stereo view of the main body of the large subunit, with the seven 'W' (3-sheet motifs in different colours. The direction of view is along the pseudo-7-fold axis. Fig. 4. Electron density of the quinone co-factor in methylamine dehydrogenase with the postulated 'pro-PQQ' shown. This is the density of a 2.25 A map with combined phases obtained from the refined partial model of both subunits using an 'X-ray sequence'. The atoms of the co-factor were not included in refinement or phase calculations so that an entirely unbiased view of the density is obtained. The two covalent links with the polypeptide chain are visible in the density, but have not been modelled. residues in a fl-sheet conformation, while the remaining residues are found in loops and segments of random coil. The structure is further maintained by numerous covalent linkages. The L subunit is expected to contain a total of 12 cysteine residues, as deduced from amino acid composition studies (F.Huitema, J.A.Duine and J.J.Beintema, unpub- lished results). In the final electron density map, six peaks of high density were found to interconnect the polypeptide chain. These were interpreted as the six disulphide bridges shown in Figure 5, with their location being compatible with that of the 12 cysteines found in the sequence of the L subunit of MADH from Pseudomonas AMI (Ishii et al., 1983). These disulphide bridges are probably important for stabil- izing the 3D structure of this subunit, as is observed from denaturation experiments (Matsumoto and Tobari, 1978a). Two other covalent bridges are found, involving the co-factor and residues 57 and 107 of this subunit, which will be discussed below in detail. Topology of the large subunit This subunit contains 370 amino acid residues in our current model. This is less than expected for a subunit of Mr 47 500 (Vellieux et al., 1986) which could be due to mobility or disorder of one or both chain termini leading to weak density in the map, or it could be the result of an overestimate of the mol. wt, or both. The large subunit consists of two distinct segments. The first 31 residues of the subunit form the extended arm which is part of the H - L' contact region mentioned above. This N-terminal extension comprises a 4.5 turn a-helix connected to a stretch of 15 residues in an extended conformation. This stretch is fully accessible to solvent but nevertheless very well defined in our electron density map, where it is seen to form an additional fl-strand, running alongside strand 3 of the L' subunit. The remaining 339 residues are found in a compact, disc-shaped domain. This main body of the subunit is formed by seven four- stranded antiparallel f-sheets, each with exactly the same 2173 F.M.D.Vellieux et al. VAL ASP PRO ARG ALAI| LYS TRP GLN PRO GLN ASP ASN ASP ILE bLN ALAI| LYS ASP IIKY II RP I ARL HIS 61Y JU ILE SCS l CYS 38 | ARSLY SER| VAL ASN [TYR LY YS61- S S-S\ GLY 38L3YS CYS S 86 LEU~~~~~~~~~~~~~~~~~~~~~~~~~~~ESSAR86 7a4 l ALA_CYSs--CCs 10877717 ASP THR GLYS ASP29 LEU ARS4b-S SER ASN I TYR CGYL 90 CYS 23 IGLU AL RA GLY CYSY297VAL GLU TYR ASNSNS-S LYSASPALAASN ~ ~ ~ ~ FROASPILIE 0 CY PE SLY SLY CYS78 SLYLALPSROSFSER 120~~~~~~~~~6 LSAPALA ASF8 APROL ILE CYSPH GLGY 127 OX 108 19 S - YS8 36 Fig. 5. The X-ray sequence of the small subunit depicted together with representations of the disuiphide bridges and the residues involved in covalent links to the quinone co-factor. The boxes arranged in zig-zag represent the residues found in 3-sheets regions. A total of seven [3-strands are found in the subunit, which comprise the following residues: 30-35 (strand 1); 40-42 (strand 2); 45-47 (strand 3); 50-64 (strand 4); 68-78 (strand 5): 86-90 (strand 6); 118- 127 (strand 7). These strands form three [3i-sheets in the following manner: sheet A by strands 1 and 6; sheet B by strands 2 and 3, and the central sheet C by strands 4, 5 and 7. Note that the X-ray sequence begins at residue 7, since we did not find any electron density corresponding to the first six residues in this N-terminal region, when compared with recent sequence data obtained by F.Huitema, J.A.Duine and J.J.Beintema (unpublished results). 'W' topology (Figures 3 and 6). These seven fl-sheets are named WI -W7. The four individual strands of each sheet, or W, are named A- D; A being the strand nearest to the centre of the subunit, D running along the surface of the subunit (Figure 6). Connections between consecutive ,B-sheets are made up by long loops linking strand D of one f-sheet with strand A of the next fl-sheet. The last sheet, W7, con- tains both termini, with the N-terminal extension entering the main body of the subunit as strand B and strand A formed by the C-terminal segment of the subunit (Figure 6). This is the only deviation from the otherwise identical topology seen in all other Ws. Internal pseudo-symmetry of the H subunit The circularly symmetrical arrangement of seven nearly identical units, described above as Ws, is clearly one of the most striking features of the MADH structure. If the seven repeated units are viewed from the 'top' of the disc, it becomes immediately obvious that the structure contains a pseudo 7-fold axis of symmetry, whose direction is approximately parallel to that of the A strands of each W (Figure 3). This unexpected observation is confirmed by pair- wise superpositions which were carried out for these W motifs. The et carbons in the fl-strand regions of all Ws are found to superimpose very well with r.m.s. deviations rang- 2174 3D structure of quinoprotein methylamine dehydrogenase W2 W5 W7 WI Fig. 6. Schematic diagram of topology and numbering of sheets and strands in the large subunit of MADH. Table I. Comparison of 13-sheets (Ws) in the large subunit of methylamine dehydrogenase WI W2 W3 W4 W5 W6 W7 WI - 1.1 2.2 1.4 1.0 1.2 1.3 W2 47.5 - 2.0 2.3 1.1 1.0 1.6 W3 98.6 44.9 - 1.8 2.2 2.2 2.2 W4 150.3 103.2 48.2 - 1.5 1.4 1.4 W5 208.2 162.4 119.2 58.6 - 1.1 1.0 W6 269.3 224.8 178.4 120.2 59.9 - 1.0 W7 315.2 270.8 224.7 167.1 108.3 48.9 - 7-fold (308.6) (257.1) (205.7) (154.3) (102.9) (51.4) - Results of superimposing the seven 13-sheets of the large subunit of MADH onto each other. The upper half of the matrix contains the r.m.s. deviations in A. The left half gives the rotation angle required to obtain the best superposition of two sheets. From the variation of this rotation angle along the diagonals, an impression of the deviation from perfect 7-fold symmetry is obtained. The lower line gives the rotation angles corresponding with an ideal 7-fold axis, given in the order 6/7 x 3600, 5/7 x 360°, etc., for easy comparison with the numbers in the line immediately above. ing from 1.0 to 2.3 A (Table I). When the results of this superposition ofWs are analysed more carefully (Table II) a rather precise direction of the internal pseudo-symmetry axis can be found. However, the rotation angles thus obtained show considerable deviations from ideal 7-fold symmetry. Nevertheless, this remarkable structural repetition strongly suggests that the large subunit of MADH evolved from an ancestor made up of seven fully identical ,B-sheets. Relationship of the H subunit to neuraminidase MADH is the second example of a protein structure with a circular arrangement of topologically identical four- stranded antiparallel 3-sheets. This uncommon motif had earlier been recognized as the main structural element of the influenza virus neuraminidase (Varghese et al., 1983), which Table II. Angular relationship of Ws in H-MADH First Second xa Pb aC r.m.s. Number of 3-sheet $-sheet ( (0) (0) dev. equivalent (A) atoms WI W4 150.3 258.8 100.2 1.4 26 W2 W5 162.4 258.6 102.7 1.1 27 W3 W6 178.4 259.1 104.6 2.2 25 W4 W7 167.1 263.8 102.2 1.4 27 W5 WI 151.8 259.6 97.7 1.0 30 W6 W2 135.2 257.4 100.0 1.0 28 W7 W3 135.3 256.2 105.5 2.3 26 Average value (0) 154.4 259.1 101.8 SD (0) 16.0 2.4 2.7 aThe polar angles x, p and V describe a rotation as defined in Rossman and Blow (1963). Superposition of pairs of $-sheets yielding a surprisingly similar direction of the rotation axis as shown by the small standard deviations of p and 1. However the angle of rotation about this axis, x, varies considerably. A basic feature to note is the observation that W3 and W6 can be superimposed onto each other by a pseudo 2-fold rather than by a rotation of 4/7 x 3600, or 154.3°. Also note that the average rotation angle x in this table differs only 0.10 from that of an ideal 7-fold, but that the standard deviation is very large. catalyses the hydrolysis of the a-ketosidic bond between sialic acid and an adjacent sugar residue. When pair-wise superpositions are carried out after careful selection of equivalent residues in the fl-strands of both structures (with coordinates kindly made available by Dr P.Colman, CSIRO, Australia) the r.m.s. deviation between equivalent a carbon atoms is found to be quite small, i.e. between 0.6 and 2.8 A. It must be noted, however, that despite the great similarity of the W motifs, the two subunits are nevertheless distinctly different in several respects. First, whereas the H subunit of MADH has a quite regular packing of f-sheets, that of neuraminidase is not so regular, therefore giving a more irregularly shaped subunit. Thus exact superposition of one complete neuraminidase subunit on top of H-MADH is not feasible-at least not with r.m.s. deviations < 4 A. A second difference between the two structures is found in the sheet containing both chain termini, which is W7 in our notation. In H-MADH the A-strand of W7 is provided by the C terminus and strands B, C and D by the N terminus of the subunit. In neuraminidase, strands A, B and C are provided by the C terminus and only strand D by the N terminus. A third difference is the fifth folding unit, which is a 'perfect' W motif in H-MADH, but appears to be an irregularly shaped loop region in neuraminidase. In spite of these differences, the intriguing possibility exists that both enzymes evolved from a common ancestor, which consisted of seven identical W folding units. A complex series of gene duplication and fusion events, followed by divergence, then led to the observed differences in W5 and W7. Once sequence information becomes available, this postulated evolutionary relationship can be studied at the level of primary structure. The active site region and the structure of the co- factor The active site of MADH is situated in a channel, or cavity, found at the interface region between the large and 2175 F.M.D.Vellieux et al. a H02 CYS 108 \ PHE109 S-S I \ 5 CYS 78 l48S-~~AfAl - 1. I SY t ,9,. I Fig. 7. (a) Structure and atomic numbering of pyrroloquinoline quinone (PQQ). (b) Schematic drawing of a possible interpretation of the electron density shown in Figure 4 (see text also). It should be clear that several features of this model need independent confirmation. In particular the nature of the group providing two covalent links with the polypeptide chain can obviously not be derived with any certainty from an electron density map calculated with phases from a model where an 'X-ray sequence' had to be used. These links are therefore indicated with dotted lines. The density for the 2-carboxylate group is weak and, hence, this group is also shown with dotted lines. small subunit of each HL dimer. The presence of numerous hydrophobic side chains seen in this cavity suggests that the reaction will proceed in a very hydrophobic, and possibly water depleted environment. The density for the co-factor is connected at two positions to the polypeptide chain of the small subunit (Figure 4). The two attachment sites are at residues 57 and 107 in our X-ray sequence. These connections are very similar to those found for the co-factor of the Pseudomonas AM1 enzyme, as obtained from amino acid sequence studies (Ishii et al., 1983). Surprisingly, only two of the PQQ rings fit nicely into the density, whereas the third one, the pyridine ring, could not be inserted convincingly. The 3.0 A Fo-Fc difference Fourier map, calculated for the phenylhydrazine- derivatized enzyme, showed extra density at the presumed 5 carbonyl position [numbering of atoms is analogous to that of PQQ (Figure 7a)], thereby defining the location of this carbonyl group of the co-factor. Careful model building based on these findings suggested that the six-membered pyridine ring of PQQ is not present as such in the native MADH co-factor. In fact, both the 2.25-A native map, and the 3.0-A map for the hydrazine treated enzyme suggest that this third ring is not closed. Both covalent linkages to the polypeptide chain are seen to occur via this part of the co-factor, for which the 7-carboxylate group density is well defined. Taking into account that the bacterial biosynthetic pathway of free PQQ presumes the addition of a glutamic acid to an indole structure derived from tyrosine (Houck et al., 1988; Van Kleef and Duine, 1988), and application of the hydrazine method to the enzyme suggests that all structural elements from PQQ should be present (Van der Meer et al., 1987), the following hypothesis on the structure of the co-factor in MADH (Figure 7b) seems plausible: the co-factor has a quinone indole structure, derived from tyrosine, to which a glutamic acid moiety is attached by a C9a-&Cy linkage. The latter moiety is attached to the protein chain at two positions: (i) to residue 57 of the small subunit via the -y-carboxyl group of the glutamate, and (ii) to residue 107 of the subunit via the a-amino group of the glutamate. The PQQ-phenylhydrazone which is identified during characterization of the co-factor of MADH (Van der Meer et al., 1987) is then generated by ring closure to form the third, missing ring of PQQ during this procedure. Closure of the third ring will proceed rather easily, as found in routes of chemical synthesis of PQQ (Buchi et al., 1985) where this reaction occurs under mild conditions. At present the electron density map does not reveal the precise nature of the residues binding the co-factor, but it is clear that they are rather small. In view of the fact that at least one of the bonds connecting the co-factor to the protein appears to be pronase-sensitive (Van der Meer et al., 1987), peptide, ester or thioester bonds are good candidates. This implies that residue 107 might be an aspartic acid forming a peptide bond with the a-amino group of the glutamate moiety. Residue 57 could be a serine, threonine or cysteine involved in an ester of thioester bond with the ,y-carboxylate group. However, modelling these covalent linkages in the current electron density map did not prove to be straightforward. Clearly, further crystallographic, sequencing and chemical studies are required to fully reveal the secrets of this co-factor. Since the co-factor as found in MADH could be similar to one of the last intermediates in the biosynthesis route of free PQQ, we propose that it should be named 'pro-PQQ'. Conclusions The X-ray structure of T.versutus MADH gives, for the first time, a view on the active site of a quinoprotein, with its co-factor covalently bound to two protein side chains. Available evidence indicates that this co-factor is not PQQ itself, but rather pro-PQQ, containing both the quinone indole moiety, plus a protein bound glutamate residue. This observation agrees with the reported differences observed in the spectral properties of MADH and of other quinoprotein dehydrogenases where PQQ becomes detached as such upon denaturation (Kenney and McIntire, 1983; Husain et al., 2176 3D structure of quinoprotein methylamine dehydrogenase Table III. Data statistics, results of heavy atom parameter refinement: data collection and processing Data set No. of No. of Rmergea observations unique reflections Native 1 (film) 125 245 32 211 0.058 (on F) Native 2 (film) 317 421 49 688 0.103 (on I) K2PtL6 (film) 126 660 33 427 0.048 (on F) Pt(en)C12 (film) 91 914 24 533 0.086 (on F) U02(C2H302)2 (film) 127 066 27 988 0.098 (on I) phenylhydrazine-inhibited (FAST) 44 065 18 819 0.079 (on F) N aRmr _E lFi -Flmerge= E Fi Table IV. Data statistics, results of heavy atom parameter refinement: MIRAS phase calculation to 2.5 A Resolution (A) 13.33 8.24 5.96 4.76 3.84 3.26 2.86 2.50 Total 0.81 0.78 0.77 0.70 0.62 0.60 0.59 0.50 0.75 K2PtI6 F,/Eja 2.01 2.17 2.35 1.68 1.19 0.78 0.63 1.00 1.23 Rc,11lisa 0.58 0.47 0.51 0.64 0.79 0.92 0.96 0.74 0.71 Pt(en)CI2 Fc/Eja 2.51 2.36 2.14 1.38 1.02 0.83 0.75 - 1.22 Rcullis 0.45 0.53 0.54 0.61 0.88 0.93 1.29 - 0.69 U02(C2H302)2 Fc/Eja 2.55 2.21 2.39 1.83 1.38 1.54 1.73 1.42 1.68 Rcullisa 0.30 0.48 0.52 0.58 0.80 0.73 0.69 0.54 0.63 aR = E FPH FP + FH(calo for centric reflections only. ElIFPHI -IFPI 1987; Frank et al., 1988). This point has also been discussed by Hartmann and Klinman (1988). The mechanism by which the co-factor is incorporated into the enzyme remains most intriguing. One might speculate that MADH itself is involved in several of the steps leading to the formation of its functional co-factor. In this context it is interesting to refer to enzymes containing a side chain of the protein's constituent amino acids as a co-factor, in particular those using the free radical of tyrosine for catalytic conversions (Prince, 1988). If these residues are in fact primitive co-factors, pro-PQQ could be considered as an intermediate in the evolution of tyrosine to PQQ. Another startling finding in MADH was the discovery of a 7-fold circular arrangement of fl-sheets, which constitute the main body of its large subunit. This, together with the obvious structural similarity to influenza virus neuraminidase, suggests that both structures might have evolved from the same ancestor. It is quite remarkable to observe such a close topological relationship between enzymes with entirely different functions present in methylotrophic bacteria and in the influenza virus. Materials and methods Crystal structure determination Crystals of MADH from T. versutus have been obtained as described previously (Vellieux et al., 1986), except that crystals were grown and stored at 4°C. A single crystal of excellent quality was obtained from virtually each hanging drop experiment. The space group is P3121, with a = b = 129.8 A, c = 104.3 A. The asymmetric unit contains one HL dimer with aMr of 60 400. The Vm is exceptionally high, 4.2 A3/dalton, which means that the fraction of the volume of the unit cell occupied by solvent is - 70% (Matthews, 1968). The enzyme as it is isolated contains the co-factor in its semiquinone form (De Beer et al., 1980). X-ray diffraction data for the native enzyme and three heavy atom derivatives were collected on 1-1.5° oscillation photographs using the X-31 beam line of the DESY-EMBL outstation in Hamburg, except for the low resolution native data which was collected using Cu Kcx radiation from an Elliot GX6 rotating anode generator. Data processing was carried out using established film processing programs (Schwager et al., 1975; Rossmann, 1979). Derivative data sets were scaled to native data using a local scaling procedure (Matthews and Czerwinski, 1975). Initial heavy atom substitution models were obtained from difference Patterson maps using vector search programs (Argos and Rossman, 1974), and were completed by difference Fourier methods. The final heavy atom model is composed of three sites for the uranyl acetate derivative, three sites for the platinium iodate derivative and five sites for the second platinium derivative. Phase refinement and phase calculation were carried out using the program PHARE (written by Dr G.Bricogne), with anomalous scattering information used for each derivative. From these data, first a multiple isomorphous replacement (Green et al., 1954) electron density distribution at 4.5 A resolution was calculated. This clearly showed the outline of the molecule and the large solvent regions in the crystals. The resolution was then gradually extended to 3.5 A in 45 solvent flattening cycles. From this point onwards, the crystal structure determination went as follows. (1) An improved high resolution native dataset became available and the MIRAS phasing procedure was repeated. Phases could be calculated to 2.5 A resolution with much improved statistics (Tables III and IV). (2) A 2.5 A resolution density map was obtained, and subjected to 60 cycles of solvent flattening and phase extension from 3.5 to 2.5 A resolution by a modification of the method by Frederick et al. (1984). (3) In the resulting map, the polypeptide chain in the large subunit region could be traced. A model for this subunit, where side chains were assigned solely on the basis of the electron density, was refined using molecular dynamics procedures (Bringer et al., 1987; Fujinaga et al., 1989). The R factor at this stage was 36%. The resulting partial model phases (Rice, 1981) were combined with 'solvent flattening phases' to produce a new 2.25 A electron density map. (4) The polypeptide chain in the small subunit of MADH was partially traced in this map and an incomplete model, made up of the large subunit plus segments of the small subunit, was refined. The R factor decreased to 30%. Again, phase combination was carried out to produce an improved electron density map. (5) In this density a revised model of the entire small subunit of MADH could now be built, and six disulphide bridges plus the co-factor unambiguously located. In this step an 'X-ray sequence' for the small subunit was deduced. The resulting model was subjected to a few cycles of molecular dynamics refinement. After phase combination a final 2.25-A resolution electron density map was obtained, and used to study the novel type of co-factor. Crystals of phenylhydrazine-derivatized enzyme were obtained by, first, oxidizing the free radical form of the co-factor by soaking in a solution of Wurster's blue. Afterwards, these oxidized crystals were transferred to a solution of the phenylhydrazine. The 3.0 A inhibited enzyme dataset was collected in 10' frames on a FAST area detector system (Arndt, 1982) using Cu Ka radiation from an Elliot GX2 1 rotating anode X-ray generator and processed using the MADNES program package (Pflugrath and Messerschmidt, 1986). Acknowledgements We thank Dr P.Colman (CSIRO, Australia) for kindly providing Cca coordinates for neuraminidase. Professor B.C.Wang (Pittsburgh, USA) provided an initial version of the programs used for solvent flattening. The help of Dr G.Vriend and Dr B.W.Dijkstra with graphics and computational work and of Dr K.S.Wilson for assistance in the use of the Hamburg EMBL synchrotron radiation facility is gratefully acknowledged. This work was supported by the Dutch Foundation for Chemical Research (SON), the BIOSON Research Institute and the Dutch Organization for Scientific Research (NWO). A grant from the'Werkgroep Gebruik Supercomputers' for use of the Cyber 205 in Amsterdam is gratefully acknowledged. 2177 F.M.D.Vellieux et al. References Argos,P. and Rossmann,M.G. (1974) Acta Crystallogr., A30, 672 -676. Arndt,U.W. (1982) Nucl. Instrum. Meth., 201, 21-25. Beer,R.de, Duine,J.A., Frank,J. and Large,P.J. (1980) Biochim. Biophys. Acta, 662, 370-374. Brunger,A.T., Kuriyan,J. and Karplus,M. (1987) Science, 235, 458-460. Buchi,G., Botkin,J.H., Lee,G.C.M. and Yakushijin,K. (1985) J. Am. Chem. Soc., 107, 5555-5556. Duine,J.A. and Jongejan,J.A. (1989) Annu. Rev. Biochem., 58, in press. Duine,J.A., Frank,J. and Jongejan,J.A. (1986) FEMS Microbiol. Rev., 32, 165-178. Filman,D.J., Bolin,J.T., Matthews,D.A. and Kraut,J. (1982) J. Bio. Chem., 257, 13663-13672. Frank,J., Dijkstra,M., Duine,J.A. and Balny,C. (1988) Eur. J. Biochem., 174, 331-338. Frederick,C.A., Grable,J., Melia,M., Samudzi,C., Jen-Jacobson,L., Wang,B.C., Greene,P., Boyer,H.W. and Rosenberg,J.M. (1984) Nature, 309, 327-331. Fujinaga,M., Gros,P. and van Gunsteren,W.F. (1989) J. Appl. Cryst., 22, 1-8. Green,D.W., Ingram,V.M. and Perutz,M.F. (1954) Proc. Roy. Soc., 225, 287-307. Hartmann,C. and Klinman,J.P. (1988) BioFactors, 1, 41-49. Houck,D.R., Hanners,J.L. and Unkefer,C.J. (1988) J. Am. Chem. Soc., 110, 6920-6921. Husain,M., Davidson,V.L., Gray,K.A. and Knaff,D.B. (1987) Biochemistry, 26, 4139-4143. Ishii,Y., Hase,T., Fukumori,Y., Matsubara,H. and Tobari,J. (1983) J. Biochem. (Tokyo), 93, 107-119. Karplus,P.A. and Schultz,G.E. (1987) J. Mol. Biol., 195, 701-729. Kenney,W.C. and McIntire,W. (1983) Biochemistry, 22, 3858-3868. Kleef,M.A.G. van and Duine,J.A. (1988) FEBS Lett., 237, 91-97. Lindqvist,Y. and Branden,C.-I. (1985) Proc. Natl. Acad. Sci. USA, 82, 6855-6859. Lobenstein-Verbeek,C.L., Jongejan,J.A., Frank,J. and Duine,J.A. (1984) FEBS Lett., 170, 305-309. Matsumoto,T. and Tobari,J. (1978a) J. Biochem. (Tokyo), 83, 1591-1597. Matsumoto,T. and Tobari,J. (1978b) J. Biochem. (Tokyo), 84, 461-465. Matthews,B.W. (1968) J. Mol. Biol., 33, 491-497. Matthews,B.W. and Czerwinski,E.W. (1975) Acta Crystallogr., A31, 480-487. McIntire,W.S. and Stults,J.T. (1986) Biochem. Biophys. Res. Commun., 141, 562-568. Meer,R.A.van der and Duine,J.A. (1986) Biochem. J., 239, 789-791. Meer,R.A.van der, Jongejan,J.A. and Duine,J.A. (1987) FEBS Lett., 221, 299-304. Meer,R.A.van der, Jongegan,J.A. and Duine,J.A. (1988) FEBS Lett., 231, 303 -307. Pflugrath,J.W. and Messerschmidt,A. (1986) MADNES Users' Guide. Prince,R.C. (1988) Trends Biochem. Sci., 13, 286-288. Rice,D.W. (1981) Acta Crystallogr., A37, 491-500. Rossman,M.G. (1979) J. Appi. Crystallogr., 12, 225-238. Rossmann,M.G. and Blow,D.M. (1963) Acta Crystallogr., 16, 39-45. Rossman,M.G., Liljas,A., Branden,C.-I. and Banaszak,L.J. (1975) In Boyer,P.D. (ed.), The Enzymes. Academic Press, New York, Vol.11, pp. 61 -102. Schreuder,H.A., van der Laan,J.M., Hol,W.G.J. and Drenth,J. (1988) J. Mol. Biol., 199, 637-648. Schwager,P., Bartels,K. and Jones,A. (1975) J. Appi. Crystallogr., 8, 275-280. Varghese,J.N., Laver,W.G. and Colman,P.W. (1983) Nature, 303, 35-40. Vellieux,F.M.D., Frank,J., Swarte,M.B.A., Groendijk,H., Duine,J.A., Drenth,J. and Hol,W.G.J. (1986) Eur. J. Biochem., 154, 383-386. Received on March 20, 1989 2178