X-ray Structure of Lipoamide Dehydrogenase from Azotobacter

J. Mol. Biol. (1989) 266, 365-379

X-ray Structure of Lipoamide Dehydrogenase from Azotobacter vinelandii Determined by a Combination

of Molecular and Isomorphous Replacement Techniques

A. J. Schierbeek, M. B. A. Swarte, B. W. Dijkstra, G. Vriend R. J. Read, W. G. J. Hol, J. Drenth

Laboratory of Chemical Physics, University of Groningen Nijenborgh 16, 9747 AG Groningen, The Netherlands

and C. Betzel

EMBL Outstation, Notkestrasse 85 2000 Hamburg 52, B.R.D.

(Received 25 July 1988, and in revised form 9 November 1988)

The crystal structure of lipoamide dehydrogenase from Azotobacter vinelandii has been determined by a combination of molecular replacement and isomorphous replacement techniques yielding eventually a good-quality 26 A electron density map. Initially, the structure determination was attempted by molecular replacement procedures alone using a model of human glutathione reductase, which has 26% sequence identity with this bacterial dehydrogenase. The rotation function yielded the correct orientation of the model structure both when the glutathione reductase dimer and monomer were used as starting model. The translation function could not be solved, however.

Consequently, data for two heavy-atom derivatives were collected using the Hamburg synchotron facilities. The derivatives had several sites in common, which was presumably a major reason why the electron density map obtained by isomorphous information alone was of poor quality. Application of solvent flattening procedures cleaned up the map considerably, however, showing clearly the outline of the lipoamide dehydrogenase dimer, which has a molecular weight of 100,000.

Application of the “phased translation function”, which combines the phase information of both isomorphous and molecular replacement, led to an unambiguous determination of the position of the model structure in the lipoamide dehydrogenase unit cell. The non- crystallographic 2-fold axis of the dimer was optimized by several cycles of constrained- restrained least-squares refinement and subsequently used for phase improvement by 2-fold density averaging. After ten cycles at 3.5 A, the resolution was gradually extended to 2.8 A in another 140 cycles. The 2.8 A electron density distribution obtained in this manner was of much improved quality and allowed building of an atomic model of A. vinelandii lipoamide dehydrogenase.

It appears that in the orthorhombic crystals used each dimer is involved in contacts with eight surrounding dimers, leaving unexplained why the crystals are rather fragile, Contacts between subunits within one dimer, which are quite extensive, can be divided into two regions separated by a cavity. In one of the contact regions, the level of sequence identity with glutathione reductase is very low but it is quite high in the other.

The folding of the polypeptide chain in each subunit is quite similar to that of glutathione reductase, as is the extended conformation of the co-enzyme FAD. The structure of A. vinelandii lipoamide dehydrogenase solved forms a starting point for investigating details of the catalytic mechanism as well as studying the interactions of this enzyme with its partners in the pyruvate dehydrogenase multi-enzyme complex.

0022-2836/59/06036s-15 $03.00/O 365

0 1989 Academic Press Limited

366 A. J. Schierbeek et al.

1. Introduction

Lipoamide deh ydrogenase (E.C. 181.4; LipDHaset) is a flavin-containing disulphide oxidoreductase, which is part of both the pyruvate dehydrogenase complex (PDC) and the 2-0x0- glutarate dehydrogenase complex (OGDC: Reed, 1974). It functions also in the glycine cleavage system (Kochi & Kikuchi, 1976). It functions as a dimer which, according to the known sequence of the Azotobacter vinelandii LipDHase, has an M, value of 103,966 (Westphal & de Kok, 1988). It catalyses the reoxidation of a dihydrolipoamide group, which is bound in amide linkage to a lysine of the core enzyme of these complexes:

Lip(SH), +NAD+ + Lip& +NADH +H+.

Lipoamide dehydrogenase shares both structural and mechanistical properties with the enzyme glutathione reductase (GRase: Williams, 1976) and with other members of the family of llavin- containing disulphide oxidoreductases, mercuric reductase (Fox & Walsh, 1983) and the recently discovered trypanothione reductase (Shames et al., 1986; Krauth-Siegel et al., 1987). A detailed analysis of the homology between lipoamide dehydrogenase from Escherichia coli (Stephens et al., 1983) and GRase from human erythrocytes (Krauth-Siegel et al., 1982; Thieme et al., 1981; Karplus & Schulz, 1987) showed that the homology between the two enzymes extends over the entire molecule (Rice et al., 1984). The gene sequences of five LipDHases have been determined (Otulakowski & Robinson, 1987; Westphal & de Kok, 1988; Benen et al., unpublished results; Browning et al., 1988; Pons et al., 1988). The alignment of LipDHase from A. vinelandii and GRase of human erythrocytes is shown in Figure 1.

The reaction catalysed by LipDHase shows great similarity to that of GRase (Williams, 1976). It is, however, NAD-dependent, whereas the reaction of GRase is NADP-dependent and physiologically the LipDHase reaction is directed in the reverse direction. In addition, the substrates of the two enzymes are obviously completely different. Another interesting feature of LipDHase is a functional association with the dihydrolipoyltrans- acetylase from the PDC and the dihydrolipoyltrans- succinylase from the OGDC. This implies func- tionally very important areas at the surface of LipDHase.

To obtain more insight into the reaction catalysed by LipDHase and into the role it has in the multienzyme systems in which it participates, we have solved the three-dimensional structure of

t Abbreviations used: LipDHase, lipoamide dehydrogenase; Lip, lipoemide; PDC, pyruvate dehydrogenase complex; OGDC, 2oxo-glutarate dehydrogenase complex; GRaae, glutathione reductasc; EMP, ethylmercuriphosphate; pCMBS, p-chloro-mercuri- benzene sulphonate; m.i.r., multiple isomorphous replacement; r.m.s., root-mean-square; s.i.r., single isomorphous replacement.

LipDHase from A. vinelandii by X-ray diffraction methods. At the same time, it will help to explain the functional differences and similarities between LipDHase and GRase.

2. Materials and Methods

(a) PuriJieation, crystallizatkm and preparation of heavy-atom derivatives

Purified LipDHase from A. vinelandii was a generous gift from Dr A. de Kok and co-workers from the Agricultural University Wageningen, Holland. It was purified from preparations of OGDC by a method described by Westphal & de Kok (1988). Bright yellow crystals were grown in melting-point capillaries from solutions containing 21 to 28% (w/v) PEG 4000 (Baker) in a 0.1 M-potassium phosphate buffer (pH 7.5) in the presence of 91 mM-EDTA as described (Schierbeek et al., 1983). The crystals grow as thin plates, 0.6 mm x 0.4 mm x 0.1 mm. They could be enlarged by seeding techniques (Thaller et al., 1981) to dimensions of 0.8 mm x 0.6 mm x 0.2 mm. The space group is P2,212,, with cell dimensions a = 64.1 A, b = 83.8 A, c = 1920 A and there is 1 dimer per asymmetric unit. The crystals diffract to at least 2.2 A resolution (1 A = 0.1 nm).

A search for heavy-atom derivatives was hampered by the tendency of the crystals to crack after transfer to mother liquor, either with or without protein added. Therefore, the crystals were first equilibrated by vapour diffusion to a standard mother liquor consisting of 22% PEG 4000 and O-1 mM-EDTA in 0.1 M-potassium phosphate buffer (pH 7.5). Hereafter, the crystals were stable enough to be transferred to solutions of standard mother liquor to which heavy-atom compounds had been added. An ethylmercuriphosphate (EMP) derivative was obtained by a 16 h soak in a solution containing 0.5 mM- EMP. A 2nd derivative was obtained by soaking a crystal for 2 h periods in a solution of increasing concentration of p-chloro-mercuri-benzene sulphonate (pCMBS), up to a concentration of 2 mnr, where the crystal was left to soak for another 8 h. The crystals were mounted in mother liquor containing heavy metals.

(b) Data collection and processing

Native data to 2.2 A were collected on 5 crystals by the oscillation method on the X11 beam line of the DESY synchrotron in Hamburg (FRG) at 0°C. The crystal to film distance was 75 mm and the waveIength was set at 1.488 A. An oscillation range of 2” was used. Average exposure times were 10 to 15 min per filmpack, which consisted of 2 films per pack. Because many of the strong reflections were overexposed, even on the B-films, a 2nd native data set was collected to 3.5” on the X31 beamline of the DESY synchrotron. Two crystals were used and the crystal to film distance was 100 mm. For this dataset, exposure times were kept short, about 3 min per oscillation of 2.5 A, to ensure proper recording of the strong reflections.

The EMP and the pCMBS derivative data sets were also collected on the X31 beam line of the Hamburg synchrotron. Again, the wavelength was set at 1.488 A. A crystal to film distance of 100 mm was used and, for each filmpack of 2 films, the crystal made 2.5“ of oscillation. For the EMP derivative, data were collected to 2.8 A resolution, and for the &MBS derivative data went out to 3.5 to 4 A resolution.

All the films were digitized on a Scandig microdensi-

A. vinelandii Lipmmide Dehydrogenase 367

LipDHase GRase

( I) ( 1) ACRGEPGPOGPPPAAGAVAS

LipDHase GRase

LipDHase GRase

( 60) TGEVAIDVPTHIARKDGIVRNLTGGVASLIKANGVTLFEGHGKLL-AGK- ( 691 SC*GKFNWRVIKEKR*AY*SR*NAIYONNLTKSHIEIIR**AAFTSOP*P

u-t?,

LipOHase GRase

(126) KvEVTAADGSSQVLDTENVILASGSKPVEI--PPAPVOQDVIVOSTGALD

LipDHase (1761 GRase (163)

LipDHose GRase

1226) (233)

LipDHase GRase

(270) (263)

LipDHose GRase

(320) (333)

LipDHase GRase

(368) 1363)

LlpDHose GRase

(416) (433)

LipDHase (466)

GKEGKTALGG -------***

”

FQNVPG#LGVIGAGVIGLELGSVWARLGAEVTVLEA&KFLPAVDEGVAK LEEL*~S~I~**+Y*A~I~MAGILSA**SKTSLMIR&~V~RSF+~SMIST

--- -

EAQKILTKQGLKILLGARVTGTEVKNKGVTVKFVDAE-GE--KSO-AF-- NCTEE*ENA*VEV*KFSGIKEVKKTLS<GLE*SM*TyP*RLP~DV --

DKLIVAVGRRPVTTOLLAADSGVTLOERGFIYVODYCATSVPGVYAIGDV *C*LW*I**V+N+K~*SLNKL*I~OK t--,-

VRGAMLAHKASEEGVVVAERIAGHK--AGMNYDLIPAVIYTHPEIAGVGK CGKwL*TPVxIAA*RKLwH*LFEYwEOSKLD*NN**T~VFS**P*GT**L

n ” tJw

TEQALKAE-GV-AINVGVFPFAASGRAMAANOTAGFVKVIADAKTDRVLG ?UDEAIHKY*IENVKTYSTS+TPMYH*VTKRK*K<CVM*MVCAtj --

VHVIGPSAAELVQOGAIAHEFGTSAEDLGMMVFAHPALSEALHEAALAVS ~LG~*HL*GF*VWVK~AT~T*AI**TS**E*VTLR

l-3

GHAIHVANRKK

(2251 (2321

(269) (282)

(319) (332)

(367) (3621

(4151 (432)

(465) (478)

(4771

Figure 1. Alignment of the sequences of LipDHase from A. vinelandii (Westphal & de Kok, 1988) and GRase from human erythrocytes (Krauth-Siegel et al., 1982). An asterisk denotes a residue identical with that in LipDHase. The active-site sequences and the nucleotide-binding fold sequences are boxed. a-Helices in LipDHase are denoted with a curved line, B-strands with an arrowed line. The level of sequence identity of the 2 enzymes is 26%. The alignment of A. vinelandii LipDHase and GRase was obtained by using the program ALIGN of the Protein Information Retrieval system (PIR; Barker et al., 1984). Then, this alignment was modified according to the alignment of E. co& LipDHase and GRase by Rice et al. (1984). Still, the alignment in regions of very low levels of homology (e.g. residues 132 to 144 and 155 to 175) is quite uncertain.

II271 (136)

tometer. For the 2.2 A native films, a raster of 50 pm x 50 pm was used. All other films were scanned on a 100 pm x 100 pm raster.

The 2.2 A native data were processed and merged into a single dataset by programs of the Groningen BIOMOL protein structure determination software package. Relative scale and temperature factors were calculated by an extension of the method of Hamilton et al. (1965). The 3.5 A native data and the 2 heavy-atom derivative datasets were processed using a program written by Rossmann (1979), running on a Cyber 205. As the 3.5 A native data set contained hardly any overloads, only the A-films of each filmpack were processed. Also for the $?MBS dataset, only the A-films were processed, because at the time of processing a new feature had become available in the Purdue program package; the intensity of overloaded reflections is estimated on the basis of the intensity of the non-overloaded pixels and the spot profile (Rossmann, 1984). The derivative datasets were scaled on the native dataset using scale factors and full anisotropic temperature factors. Relevant data collection statistics may be found in Table 1.

The 2 native datasets were merged together into 1 dataset. First, the 3.5 A set was scaled to the 2.2 A set using scale factors and full anisotropic B-factors computed with data between 8 and 4 A. For the merged dataset, all reflections of the low-resolution dataset were used and all data between 8 and 2.2 A for the high- resolution set. For 230 reflections having a difference greater than 50% in F between the 2 data&s, a weighted average was calculated with a rather arbitrary weighting scheme. A weight of 0.9 was used for the measurement from the low-resolution dataset, and a weight of 0.1 for the measurement from the high- resolution dataset. Usually, these were strong reflections that were partly overloaded in the high-resolution set and were more reliable in the low-resolution set. The overall R-factor computed from the merged reflections was 11%. Statistics on the merging are shown in Table 2.

(c) Molecular replacement: rotation functions

The self-rotation function was calculated with Crowther’s fast rotation program (Crowther, 1972).


Table 1 Data collection statistics for the native and

derivative crystals

3.5 A 2.2 A Native Native EMP pCMBS

Number of crystals 2 5 3 10 Number of filmpacks 58 42 32 29 Average L,( %)t 4.7 4.2 3.9 48 R msrgc (%I$ 8.5 5.7 8.6 8.9 Total reflections measured 42,364 183,126 54,298 16,909 Unique reflections 11,630 44,990 22,259 7462 Max. resolution (A) 3.5 2.2 2.8 4.0 Average isomorphous

differences (%)I ~ 17.3 20.7 Completeness of data set (%) 86.1 81.0 87 86.0

tR,,, = Z/l--fl/Zfx 100.

S41,r,e = 2 ji, II~~v,~rtI - IE;I~&(N C IFpv.wl) X 100, where

N is the number of measurements of reflyction hkl and Fav,Yl is the average of the N structure amplitudes of this reflection.

4 The average isomorphous difference between the EMP and pCMBS data was 19.2%.

A peak with an arbitrary height of 9.7 relative to an origin peak of 50 was found at CI = 37.5”, /3 = 44” and y = 37*5”, using 2184 reflections between 8 and 4 A resolution and an integration radius of 24 A. This peak position corresponds in a polar axis system (Rossmann & Blow, 1962) to a 2-fold axis, K = 180”, with rp = -27” and $ = 56”. The 3rd highest peak had a peak height of 6.5.

LipDHase and GRase have a sequence identity of 26%. The active sites of the 2 enzymes are highly conserved and sequence similarities extend throughout the molecule (see Fig. 1). Therefore, GRase might serve as a reasonable molecular replacement model. The co- ordinates used were those of the 2 A model of GRase (Thieme et al., 1981) as present in the Brookhaven Protein Data Bank (Bernstein et aE., 1977). LipDHase crystallizes as a dimer per asymmetric unit, whereas GRase crystallizes as a monomer, which becomes a dimer through a crystallographic 2-fold rotation operation. The model used for the cross-rotation function was a complete GRase dimer, which was generated by applying the 2-fold rotation to the GRase monomer. The 2-fold axis of the dimer was set parallel to the z-axis as suggested by Moss (1985). In this orientation, rotation peaks belonging to

Table 2 Statistics of merging the high and low-resolution

native data sets

Resolution range (A) loo-8 83.5 3.5-2.2

Total no. of reflections in 3.5 A dataset

Total no. of reflections in 2.2 A dataset

No. of reflections unique for 3.5 A dataset

No. of reflections unique for 2.2 A dataset

R-factort Number of common reflections

1130 10,627 -

660 11,371 31,062

509 731

39 1475 31,062 11.1 11.3 621 9896 -

t R = Wd.,,aSCt 1 I - IFa.u.ct 2111W’dnf~.c, 1 I x 100 % for reflections common to datasets 1 and 2.

the same solution, but differing in a 180” rotation about the dimer 2-fold, will show up in the same fi section, with t,he same value of a, but with a 180” difference in y. The model was placed in an orthogonal cell with PI symmetry, which was made sufficiently large (a = 128 A, b = 128 A, c = 88 A) to avoid cross-vectors shorter than 20 A in the search model. The cross-rotation function (Crowther, 1972) was calculated with 3602 observed and 3868 calculated reflections between 8 and 4 A resolution, using an inner and outer integration radius of 4 A and 24 A, respectively. An overall B-factor of 15 A2 was used for the model. The 2 highest peaks were found in the section /? = 70” at a = 142.5” and y = 30” and 210”, with a relative peak height of 162, which was 5.4 times the standard deviation of the map. The 2nd highest pair of peaks was found in the section B = 90”, at a = 0” and y = 80” and 260”. These 2 peaks had a peak height of 155, which is 5.1 6. After interpolation, the highest peak position was at u = 142.4”, /l = 68” and y = 32.5”.

To confirm the 2-fold relation between the 2 highest peaks, another cross-rotation was calculated with a monomer of GRase as a model. The molecule was placed in a cell with a = 100 A, b = 84 A and c = 88 A. A cross- rotation function was calculated with the same parameters as before; i.e. data between 8 and 4 A and inner and outer radii in Patterson space of 4 and 24 A. In this case as well, where the 2-fold symmetry is not imposed, 2 peaks with a peak height of 4.9 cr were found at the same positions in the same section of/l = 70”, where the dimer model had given its maxima. The peaks in the section B = 90” had a height of 3.8 u. In Figure 2, the section /? = 70” is shown for the rotation functions with a monomer and a dimer as model.

The 2 peaks in the cross-rotation are related to each other by the 2-fold rotation observed in the self-rotation function, which is a 2nd confirmation of the correctness of the cross-rotation function result. A GRase dimer to which the rotation angles of the cross-rotation functions are applied now gives a model that has the same orientation as LipDHase in its unit cell. The positioning of this model in the cell is the next problem to be solved.

(d) Translation function studies

With the correctly oriented GRase dimer and/or monomer, we tried to find the position of the LipDHase molecule using the translation function of Crowther & Blow (1967). In order to obtain experience with the limitations of the translation function, a series of test calculations was done in which the influence of the various imput parameters was investigated, showing that the starting model can be very incomplete, but that the error in orientation has to be small (Schierbeek et al., 1985). However, whichever resolution range we tried in the case of LipDHase, the sections of the translation function were very noisy and no 3 mutually compatible peaks could be found. Changing the model by leaving out side-chains or regions that were thought to be very different also had no positive effect. Small changes in the orientation also gave no improvement. Therefore, it was decided to use the heavy-atom phase information we had obtained via the isomorphous replacement techniques to solve the structure of LipDHase.

(e) Multiple i8morphm~8 rephcem,ent

A difference Patterson from the EMP derivative could be solved easily using vector search options of the program VSFUN from the BIOMOL software package.

A. vinelandii Lipoamide Dehydrogenuse 369

360 -

a (deg.)

180 0

(a) 360-

180

A (b) Figure 2. Section p = 70” of the cross-rotation function (Crowther, 1972) showing 2 peaks at CI = 1425”, fi = 70”,

y = 30” and CL = 1425”, /? = 70” and y = 210” related by local 2-fold symmetry. (a) Fobs, LipDHase; Fc8,c, GRase monomer. (b) Fobs, LipDHase; Fcalc, GRase dimer.

Two major heavy-atom sites were found. “Lack-of- closure error” refinement of the heavy-atom parameters with the program PHARE (Bricogne, 1976) coupled with difference and residual Fouriers revealed 6 more heavy- atom sites. The heavy-atom sites could be grouped in pairs with similar occupancy, consistent with the presumed local 2-fold symmetry. A difference Fourier of the @MBS derivative, calculated with the single isomorphous replacement phases from the EMP derivative, revealed four major sites, two of which were common with the EMP derivative. All these sites could be confirmed in the difference Patterson. Here, also the X-foldedness was apparent from the occupancy of the heavy-atom sites. For the pCMBS derivative, 4 more sites could be found after heavy-atom refinement and Fourier techniques. The final parameters and statistics are shown in Tables 3 and 4.

An electron density map was calculated with data between 25 and 5 A using the multiple isomorphous replacement (m.i.r.) phases. The sections were plotted and copied on transparencies and inspected on a lightbox. Some proteinaceous density could be seen in a very noisy map, but it was nearly impossible to distinguish separate molecules or to see any 2-fold symmetry in the density. Therefore, we decided to carry out solvent flattening in order to improve the density and to obtain a better idea of where the protein was located in our cell.

(f) Solvent jhttening

From a m.i.r. map calculated with data between 25 and 4A, an envelope was obtained using the algorithm of Leslie (1987), 1st with a radius of 12 A and in later stages with a radius of 10 and 8 A. The solvent level, which is 52.3 y. assuming 1 LipDHase dimer/asymmetric unit, was set to a conservative 40% for the calculation of the

envelope. Fifteen cycles of solvent flattening at a constant resolution of 4 A were performed using the programs developed by Wang (1985). After every 5th cycle, a new mask was calculated. The R-factor (XjlF,,,l - IFobsll/ZIFobslr where lFinvl indicates the Fourier amplitude of the inverted electron density map), dropped from 386% after the 1st cycle to 22.9% after the 15th cycle for 8230 reflections between 25 and 4 A. The resultant mean figure of merit was 0.79, and the

Table 3 Rejined heavy-atom parameters for the EMP and

pCMBS derivatives of LipDHase

Relative B Derivative Site r Y 2 occupancy (A*)

EMP 1At 0.2759 0.5646 0.4599 0.384 12.29 1st 0.0478 0.2358 0.1846 0.376 20.39 2A 0.4403 0.5004 0.0344 0.162 7.58 2B 0.2668 0.4188 0.1898 0.147 10.11 4 0.4724 0.7419 0.2760 0.045 31.61 5 0.0251 0.4088 0.2814 0.052 30.41 6 0.1317 0.5621 0.4671 0.037 24.19 7 0.1535 0.0444 09428 0.047 12.19

pCMBS 1A 0.2828 0.5554 0.4606 0.304 17.46 18 0.0524 0.2300 0.1845 0.291 15.79 2A 0.4268 0.5032 0.0308 0.110 11.38 2B 0.2456 0.3930 0.1888 0.123 8.48 3A 0.3695 0.3743 0.0324 0.372 4.96 3B 0.1300 0.4230 0.1527 0.324 11.29 8 0.3936 0.5845 0.1794 0.052 5.41 9 0.4507 0.4162 0.0338 0.135 44.58

t Positions labelled with A and B arc related by non- crystallographic 2-fold symmetry. z, y, z are fractional co-ordinates.


Table 4 Phase calculation statistics for 8 EMP sites and 8 pCMBS sites of LipDHase

Resolution (A) No. of reflections Mean fig. of merit R (%)t EMP cusia Phasing power1 EMP Rcums (%H pCMBS Phasing power: pCMBS

12.6 8.4 6.3 5.0 4.2 3.6 3.2 2.8 Total 253 661 1211 2009 3054 4034 4635 4835 20691

o-51 0‘50 0.62 0.61 057 0.44 0.38 0.37 0.46 58.7 56.7 57.1 53.6 63.9 62.3 71.7 65.9 61.1 1.25 1.08 1.94 1.94 1.57 1.43 1.38 1.35 1.47 59.7 65.5 63.2 58.2 77.2 59.6 - - 66.1 1.47 1.40 1.75 1.88 1.50 1.50 - - 1.60

t Rmi. = WmI - I~P+fH(s.leJl

W~Fwl- IFPll for centric reflections.

$ Phasing power is the ratio of the root-mean-square heavy-atom scattering factor amplitude to the root-mean-square lack of closure error.

accumulated phase change with respect to the m.i.r. phases was 37-4”.

The resulting map was clearly an improvement compared to the original m.i.r. map. The boundaries of the molecule were very clear and there was no doubt where protein molecules could be found in the cell. In the protein region, however, it was hard to trace a contiguous protein chain and the density was often “chopped” in fragments. However, it was possible to find out which heavy-atom position belonged to which molecule, and which crystallographic symmetry operations had to be applied to place all heavy-atom positions in a region belonging to 1 dimeric LipDHase molecule. Assuming that the pairs of heavy-atom sites in both heavy-atom derivatives belong to subunits that are related by local 2-fold symmetry, a line could be drawn through the centres of the lines that connect these pairs of positions. The direction vector of this line could be expressed in a polar axis system with rp = -28” and $ = 58.5”. Looking along this line in the density, it was clear that the molecule is built out of 2 lobes with similar size and shape. The direction of the line also fits in very nicely with the result of the self-rotation function, where the maximum was found at rp = -27” and $ = 56”. At this stage, part of the translation problem was evidently solved, only the translation along the pseudo S-fold remaining unknown. This could have been solved in various ways, but the “phased translation function” looked to be the most elegant procedure to obtain all 3 translation components.

(g) Phased translation function

At this stage, we had an oriented model of GRase and an electron density, which clearly showed which regions belonged to protein and which to solvent. The best superposition of the model on the electron density is found at the position at which a calculated electron density overlaps best with the experimental electron density. This position can be found by calculating an overlap function (Rossmann & Blow, 1962; Colman et aE., 1976):

WI = S~~,s(~).~mode,(x--t) dx,

which, because it is a convolution of pMiz(x) with pmodel (-x), can be expressed as a Fourier transform. Since we have to integrate over the whole cell in real space, we have to expand to Pl for the reciprocal space analogue. So, for this “phased translation function”, the following Fourier summation has to be calculated:

T(t) = $ & %,lFo, WI IF,,,, (WI exp[i(at,mt - aoalc )I exp( - 2xih.tL

where IFcalcl and a,.ls refer to the structure factor amplitude and phase a, calculated for 1 oriented model placed in the unit cell of the unknown structure.

Since we do not know the hand of the heavy-atom derivatives in advance, we also have to calculate the following Fourier summation for the opposite hand:

WI = + 2 %mJFob. WI IFcslc (WI exp[i( - amR - a,,lc)] exp( - 2xih.t).

Thus, a structure factor calculation in Pl was carried out for a GRase dimer, rotated over a = 37%‘, /? = 112”, y = 32.5” and positioned at the origin of a triclinic LipDHase unit cell. The native LipDHase data with the phases and figure of merit after solvent flattening were expanded to Pl, and the phased translation functions were calculated. The highest peaks of these functions are shown in Table 5.

The highest peak in the 1st summation showed that the heavy-atom configuration chosen was correct. It also gave directly the translation to be applied to obtain the best superposition of the model and the electron density. So, in spite of the rather poor quality of the map, the phase information was sufficient to reveal the correct hand for our structure and the correct position of the oriented molecule. A more detailed derivation of a normalized phased translation function and some tests of its power are to be published elsewhere (Read & Schierbeek, 1989).

(h) Rigid body refinement

The position and orientation of the model was refined with a g-dimensional search using the program BRUTE (Fujinaga & Read, 1987) running on a Cyber 265. Around the starting positions obtained from the rotation function and the phased translation function, 8 new orientations that differed in 1” were checked at 13 x 13 x 13 positions differing by an increment of 0.5 A, which was later

Table 5 Highest peaks in the phased translation function

Peak no. peak height (;i, (1, &

Correct hand: 1 119 (11.00) 10.61 46.39 117.09 2 48 (5.7 u) 28.63 6943.5 99.97

Inverse hand: 1 37 (au) 24.83 33.88 4.80

2 36 (3~) 51.04 78.89 95.56

A. vinelandii Lipoamide Dehydrogenase

Figure 3. Packing of 4 “multi-Ala” GRase dimers in the unit cell of LipDHese after the rigid body refinement with BRUTE and some further refinement steps with CORELS (Sussmann, 1985). In the latter step, the 2 subunits were allowed to move independently; i.e. the local 2-fold was not enforced. The R-factor for this model was 47.0% for data between 10 and 5 A resolution.

reduced to an increment of 0.25 A. As a model, we used a “multi-alanine” model of GRase, which is a model in which all side-chains that are different in LipDHase were changed to alanine and all insertions were left out. Also the co-ordinates for the FAD coenzyme were omitted. The R-factor for data between 10 and 6 A dropped from 50.5% at the initial position (a = 37.5”, /l = 112.0”, y = 32.5”; x = 10.62 A, y = 46.42 A, z = 117.15 A) to 45.1 o/c for the optimized orientation and position (0: = 36.6”, /l = ill”, y = 30%‘; z = 11.75 A, y = 4790 A, .z = 116.25 A). In this position and orientation, the dimers pack nicely in the unit cell (Figure 3).

(i) Two-fold averaging of tb electron density map

To improve the phases resulting from the m.i.r. and the molecular replacement procedures, we applied 2-fold averaging of the electron density map according to methods described by Bricogne (1976).

The envelope and the 2-fold axis were obtained as follows. A GRase dimer was set in the orientation and position obtained with the program BRUTE. All side- chains that were different in GRase and LipDHase were changed with the MUTATE option (written by R. J. Read) of the program WHATIF (G. Vriend, unpublished results). This model was subjected to a series of rigid body refinements with the program CORELS (Sussmann, 1985) in which the number of rigid groups was increased from 1 (the complete dimer), via 2 (the 2 monomers) and 8 (2 times 4 domains) to 22 (helices and sheets). From this model, which had a final R-factor of 46% for data between 10 and 5 A, the orientation and position of the 2-fold axis was calculated by superimposing all backbone atoms of subunit 1 to those of subunit 2. The r.m.s. difference between the 2 subunits was 0.762 A for 1776 atoms.

An envelope was calculated from the co-ordinates with the program CRDENV (written by B. W. Dijkstra) by taking a sphere of 5 A around each atom and regarding all points within the spheres as points of the envelope. In cases where 2 dimer molecules made close contacts, points assigned to the envelope of the closest molecule were considered as part of the envelope. So, care was taken to ensure that the envelope did not overlap with those of symmetry-related molecules in the cell.

Phases were calculated for the model after the CORELS refinement to 3.5 A and combined with the m.i.r. phases. Weights were calculated with the program SIGMAA (Read, 1986). A map was calculated with the combined phases, which was then subjected to local S-fold averaging according to the method of Bricogne (1976). Ten cycles of e-fold averaging and solvent flattening were applied at constant 3.5 A resolution. No phase combination was applied so that, after every cycle

of averaging, a new map was calculated with the phases from map inversion and a structure factor weighting scheme according to Sim (1960). After 10 cycles of averaging at 3.5 A resolution the R-factor, which is defined as Z(l(F,b,I - lFi,,,ll)/E(lF,s&, with (F,,I being the observed structure factor amplitude and (PinvJ the structure factor amplitude after map inversion, was 16.6% for 7500 reflections. The drop in R-factor after the first 10 cycles is shown in Figure 4.

Phase improvement by 2-fold density averaging from 3.5 A to 2.8 A resolution was carried out in steps of 0.01 A. The averaged electron density map was inverted to give phases to 3.5 A. To these phases, the m.i.r. phases between 3.50 and 3.49 A were added. Because the resolution of the pCMBS data was only 4 A, these phases were essentially EMP s.i.r. phases. A map to 3.49 A was calculated, of which the coefficients used were: msimJ-Fo,,s(expicp,,, to 3.5 A and msimlF,bl expicp,,, between 3.50 and 3.49 A. On this map, 2 cycles of 2-fold averaging were applied. The s.i.r. phases were not used during the 2 density averaging steps. The resultant map was inverted to give phases to 3.49A resolution, to which the s.i.r. phases between 3.49 and 3.48 A were added and the averaging and solvent flattening was repeated for 2 cycles. This whole process was repeated, gradually increasing the resolution to 2.8OA. When this limit was reached, another 10 cycles were carried out at constant 2.8 A resolution. In total, 150 cycles of 2-fold averaging were performed. The final overall R-factor after averaging was 11.8% for 24,206 reflections. The R-factor as a function of resolution after every 20th cycle is shown in Fig. 5.

(j) Model building

Subunit 1 of the LipDHase dimer and the FAD coenzyme were fitted as well as possible in the electron density map obtained after 2-fold averaging using the interactive computer graphics program FRODO (Jones, 1985) on an Evans and Sutherland PS350 graphics system. Regions where insertions and deletions occur when aligning GRase and LipDHase were accounted for at this stage. The C-terminal 10 residues could not be seen in the density, so the model of subunit 1 contained only residues 1 to 466. After the model building of subunit 1, the 2nd subunit was generated by applying the 2-fold operation that was obtained after the CORELS refinement. In this way a complete model of the LipDHase dimer was obtained.

3. Results and Discussion Using a combination of molecular and

isomorphous replacement, we have determined the


0.60 z 2 0.50 Y fk

0.40

0 8-16 6.03 5.00 4.36 3.92 3.59 3*33 3.12 2-95 2.80

Resolution (a) Figure 4. Statistics of 10 cycles of phase improvement at, 3.5 A resolution. R-factors after cycles 1 to 10, with

R = W’o~sl - I~invll/W’o~l~ The UPP~ r curve is the R-factor after the 1st cycle of 2-fold averaging, the lower curve shows the R-factor after the 10th cycle.

three-dimensional structure of lipoamide dehydrogenase (LipDHase) from A. vinelandii. The first step of the molecular replacement problem, the determination of the orientation, gave quite clear results using a dimer of glutathione reductase (GRase) as a model. In the second step, the determination of the position of the molecule, no

solution could be found using the translation function of Crowther & Blow (1967). This indicates that, in this case, the limits of the power of molecular replacement have been reached.

The phase information obtained with the multiple isomorphous replacement (m.i.r.) method using two heavy-atom derivatives was, in itself, far

1.00

O-90.

O-80.

0*70-

,I.....,... 8.16 6.03 5.00 4.36 3.92 3-59 3-33 3-12 2.95 2

Resolution (8)

Figure 5. Statistics of phase improvement from 3.5 A to 2.8 A resolution. R-factors are shown as function of resolution after every 20th cycle. R-factors are defined as R = Z(\Fob,l - ~F,J/Z(F,,(.

A. vinelandii Lipoamide Dehydrogenuse 373

from sufficient to allow tracing of the polypeptide chain. However, the m.i.r. phases gave a very clear solution to the translation problem when they were used in the phased translation function, together with the information from the model oriented according to the rotation function.

The power of phase improvement by the method of S-fold averaging has been shown in previous cases with rubisco (Schneider et al., 1986), insecti- cyanin (Holden et al., 1987) and monellin (Ogata et al., 1987). In the case of LipDHase, it also made a tremendous improvement to the electron density map, allowing an easy fitting of the polypeptide chain. Thus, a preliminary model was obtained for this 103,000 &Zr protein. In describing the structure, we have refrained from giving too much detail, in view of the fact that the model is not yet refined.

(a) The electron density map

The electron density map after averaging was a great improvement with respect to the m.i.r. map or the map calculated with combined phases. In Figures 6 and 7, some parts of the density are shown with the different phases just mentioned; m.i.r. phases, combined phases and phases after 2-fold averaging. In the averaged electron density map, the polypeptide chain could be traced easily and, apart from some side-chains at the surface of the molecule, the density for the side-chains correlated well with the size of those predicted by the sequence determination. At some places, however, the density was too weak to position the polypeptide chain with certainty; so, in these regions the model is not well-defined. These sites of low density were, in most cases, at positions where insertions or deletions with respect to GRase had occurred, i.e. at regions 36 to 44, 131 to 137 and 376 to 379 (LipDHase numbering), but also in two other regions connecting two pieces of secondary structure at’ residues 250 to 252 and 388 to 397. All these regions are on the surface of the molecule. Also, the last ten residues could not be seen in the density, so the last residue in our current model is residue 466.

(b) The overall structure

A stereo view of the C” tracing of the dimer is shown in Figure 8. The overall folding of the polypeptide chain is very similar to that of GRase. The four domain structure of GRase can be found also in LipDHase, with the FAD-binding domain comprising residues 1 to 151, the NAD-binding domain residues 152 to 280, the central domain residues 281 to 349, and the interface domain residues 350 to 476. The secondary structure elements are indicated in Figure I.

(c) Heavy-atom binding sites

As shown in Table 3, 16 heavy-atom positions were determined for the two derivatives in the m.i.r. procedure. Ten of the 16 positions are

arranged in pairs by the dimer S-fold (labelled A and B in Table 3). Four of these positions (sites IA, IB, 2A and 2B) are shared by both heavy-atom compounds. The sites arranged in pairs have high occupancy (sites IA, IB, 2A, 2B, 3A and 3B). Site 1 is located in a pocket near the only free thiol group of the protein at residue Cys306 and is surrounded by Ile316, Met347, Glu332 and Tyr349. So, although this free thiol cannot be labelled by N-ethyl-maleimide in the oxidized enzyme (de Kok et al., 1982), it is accessible to compounds like EMP and pCMBS. The second site is located near the active site disulphide Cys48Cys53, interacting on the other side with His450 and Pro451 from the other subunit. Site 3, which is present only in the pCMBS derivative, is located between two histidine residues that come from two different subunits, His458 and His327’, in the vicinity of Phe448. The fact that this heavy-atom compound squeezes between two subunits might explain the tendency of the crystals to crack in pCMBS-containing solutions. Using a “gradient” of pCMBS probably enables the side-chains to move away to allow the pCMBS to find its position without disrupting the crystal packing. The remaining heavy-atom sites have a much lower occupancy than the first, and are all located on sites for which it is not really obvious why a heavy-atom compound would bind at all.

(d) Crystal contacts

Each dimer of LipDHase in the crystal is surrounded by ten other dimers. Four “surrounding” dimers are related by a simple translation along r or y to the “central” dimer. The other six surrounding dimers are arranged as a kind of trigonal prism and are related by 2-fold screw axes to the central dimer. The residues t*hat contact a symmetry-related molecule within a radius of 5 A are listed in Table 6. Fourteen different regions are involved in crystal contacts. All these contacts are of a polar nature.

(e) The dimer interface

The dimer of LipDHase consists of two identical monomers that have intimate contacts with each other. The enzyme is assumed to be active only as a dimer, since the active site is composed of residues coming from two subunits. The active site disulphide bridge Cys48Cys53 is close to His450 and Glu455 from the other subunit. In GRase the same constellation is found with Cys58-Cys63 and His467’ and Glu473’ (primed numbers refer to residues in the “second” subunit).

The contact areas of the two subunits can be divided into two parts; a “lower” part and an “upper” part, separated by a cavity. In the lower part, which consists of a residue 65 to 101 and 389 to 393, the long arm that extrudes from the FAD- binding domain wraps around the local S-fold axis. This region has hardly any sequence identity with


ICI

Figure 6. Electron density around Phel17 with: (a) m.i.r. phases to 2.8 A; (b) m.i.r. phases to 2-8 A combined with model phases to 3.5 A; (c) phases after averaging to 2.8 A.

A. vinelandii Lipoamide Dehydrogenase 375

(bl

Figure 7. Electron density around the isoalloxazine ring of the FAD coenzyme with: (a) m.i.r. phases to 2.8 A; (b) m.i.r. phases to 2-8 A combined with model phases to 3-5 A; (c) phases after averaging to 2.8 A.


Figure 8. A stereo view of the C” tracing of the A. vinelandii LipDHase dimer.

the corresponding region in GRase. Of the 26 residues involved in dimer contacts, only three are conserved with respect to GRase. Still this long arm has a conformation that is similar to that in the GRase. In the upper part, which consists of residues 48 to 57, 325 to 357 and 422 to 458, many more residues are conserved. Of the 39 residues that are involved in interface contacts in this area, 16 are conserved between LipDHase and GRase. This ratio is significantly higher than the average percentage identity, indicating the importance of these residues for the proper functioning of the enzyme. The residues involved in the dimer interface are shown in Table 7.

(f) FAD binding

The electron density corresponding to the isoalloxazine part of FAD is shown in Figure 7. The

binding of FAD in the protein is shown in Figure 9. As can be seen, the FAD binds in an extended conformation, with the isoalloxazine buried deeply within the protein and the adenine part more exposed to solvent. The binding of FAD to GRase has been discussed by Schulz et al. (1982) and, on the basis of the sequence identity between LipDHase from E. coli and GRase from human erythrocytes, it was concluded that the binding of FAD to LipDHase was similar to that in GRase (Rice et al., 1984). This is confirmed by our electron density map. A list of amino acids contacting FAD in LipDHase is shown in Table 8.

As can be seen from Table 8, most residues involved in contacts between FAD and the polypeptide chain are conserved in A. vinelandii LipDHase and GRase. A peculiar difference between the two FAD binding pockets is the substitution of Lys66 (GRase) by Ser56

Table 6 Contacts in LipDHase crystals

Residues “central molecule”~~$

Residues other moleculef

Symmetry operation@

Unit translation

(2% Y. 4

l-5 252, 253, 267-269 1 - 1 0 0

252, 253, 267-269 1-5 1 1 0 0

304, 305, 34&348, 4oS410 I’-3’ 2 -1 0 0

I’-3’ 304, 305, 346348, 409410 2 0 0 0

245’-252’, 159-165’ 320’-322’, 368’-375’ 2 0 1 0 320’-322’, 368’-375’ 245’-252’, 159’-165’ 2 -1 1 0

86 3 89 *, 7%80 157-165 3 l-l 0

157-165 86, 89, 73’-80 3 1 0 0

t All contacts within a 5 A radius of the “central” molecule, centered at r = 0.331, y = O-443 and z = 0.107, with the surrounding molecules are given.

$ Residues that have a primed number belong to the second subunit. $ The numbers refer to the following symmetry operations: 1, z, y, z; 2, x+ l/2. -y+ l/2, --z: 3, -x,

y+ l/2, --z+1/2.

A, vinelandii Lipoamide Dehydrogenase

Table 7

377

Dimer interface residues in LipDHase from A. vinelandii

Residue in Residue LipDHasc in GRase

Residue in LipDHase

Residue in GRase

Residue in LipDHase

Residue in GRase

Ser62 Lvs65 Gill68 Ala69 Phe73 Lys74 Leu75 His76 Gly77 lie78 Ser79 ThrXO Va183 Ala84 Ile85 Asp86 Thr89 Met90 Arg93 Gln96 Ile97 Am100 LeulOl

Thr72 His75 Phe78 Met79 Ala83 Asp84

Tyr85 Gly86 Phe87 Pro88 Ser89 Gly92 Lys93 Phe94 Asn95 Va198 1le99 LyslO2 Ala105 Tyr106 ArglO9 LeullO

Ser389 Arg391 Ala392 Met393

Cys48 cys53 Ile54 Lys57 Leu325 Ala326 His327 Lys328 Glu331 Glu338 Ile352 Pro353 Ala354 Va1355 Tyr357 Ser422 Glu425 Gln428

Met405 His408 Ala409 Val410

Cys58 Cys63 Va164 Lys67 Leu338 Thr339 Pro340 Va1341 Ala344 His351 Ile367 Pro368 Thr369 Va1370 Phe372 Gly439 Glu442 Gln445

Gln429 Ile432 Ala.433 Glu435 Phe436 Thr438 Ser439 Asp442 Leu443 Met445 Met446 Va1447 Phe448 Ala449 His450 Pro45 1 Ala452 I&u453 Ser454 Glu455 His458

Gly446 Val449 Ala450 Lys452 Met453 Ala455 Thr456 Asp459 Phe460 Asn462 Thr463 Va1464 Ala465 He466 His467 Pro468 Thr469 Ser470 Ser47 1 Glu472 Va1475

All residues that have a solvent-accessible surface area difference of more than 5 AZ when comparing monomer and dimer of LipDHase are shown. The ares were calculated with the program DSSP, written by Kabsch & Sander (1983). The first 27 residues, from Ser62 to Met393 in LipDHase, belong to the “lower” part. the other belong to the “higher” part of the interface as discussed in the text.

(LipDHase). In GRase, Lys66 forms an internal salt-bridge with Glu201, which is conserved in LipDHase as Glu194. No salt-bridge can be made with a serine residue, so Rice et al. (1984) speculated upon the possibility of another salt-bridge that could be made with the neighbouring lysine residue in LipDHase (Lys57). According to our electron density map, such a salt-bridge is indeed formed (Fig. 10). The density indicates that the side-chain of Lys57 curls back and points in the direction of Glu194. The speculation of Rice et al. (1984) was correct and the same polar environment that is present in GRase “below” the FAD can be found also in LipDHasr.

4. Conclusion

Using a combination of molecular and multiple isomorphous replacement, we have determined the three-dimensional structure of LipDHase from A. vinelandii. Refinement of the structure against the 2.2 A data we have collected will give a more accurate picture of the protein, allowing a detailed comparison of the structure of LipDHase with that of the refined structure of GRase. The enzyme is active in the crystal (T. Smale, personal com- munication), so the binding of a variety of compounds, like substrate-analogue and cofactors, in the crystal will give more information about the

Figure 9. Residues surrounding the FAD coenzyme. The following residues are drawn: 9 to 16, 32 to 36,44 to 57, 119 to 121, 149 to 153. 316 to 319, 324 to 329, 353 to 355, and the FAD in thick lines.

378 A. J. Sehierbeek et al.

Table 8 Residues involved in FAD binding in LipDHase jrorn A. vinelandii

Residue in LipDHase

Residue in GRase

Residue in LipDHase

Residue in GRase

Residue in LipDHase

Residue in GRase

Ile9 Ile16 GlylO Gly17 Ala11 Gly18 Gly12 Gly19 Pro13 Ser20 Gly14 Gly21 Gly15 Gly22 Ile32 Va149 Glu33 Glu50 Lys34 Ser51 Tyr35 His52 Gly46 Gly56 Thr47 Thr57

Cys48 Va151 Gly52 cys53 Ser56 Lys57 His120 Gly121 Ala149 serl50 Serl70 Leu174 Ilel91

Cys58 Va161 Gly62 Cys63 Lys66 Lys67 His129 Ala130 Ala155 Thrl56 Ser177 Phel81 lle198

Leu195 Arg278 Leu285 Glu317 Asp318 Met324 Leu32.5 Ala326 His327 Tvr357 I&450’ Pro451’

Met202 Arg291 Leu298 Gly330 Asp331 Leu337 Leu338 Thr339 Pro340 Phe372 His476 Pro468’

All residues within a distance of 5 A of the FAD coenzyme are listed, together with the homologous residues in GRase as aligned in Fig. 1. Residue numbers with a prime indicate amino acids of the “second” subunit.

Figure 10. Electron density indicating the presence of a salt-bridge between Lys57 and Glu194 in Azotobacter vinelundii LipDHase.

reaction mechanism of LipDHase. Also, the inter- action with other compounds of the pyruvate dehydrogenase can be studied by crystallizing complexes of LipDHase and the lipoyltrans- acetylase from A. vinelandii, with the ultimate goal being to elucidate completely the structure of the entire pyruvate dehydrogenase complex.

We are indebted to Dr A. de Kok and his colleagues at the Agricultural University in Wageningen. the Netherlands, for a generous supply of lipoamide dehydrogenase. It is a pleasure to thank Drs M. G. Rossmann, B. C. Wang and C. Sander for providing computer programs.

Calculations were performed on the Cyber 170/760 of the Groningen University Computer Centre, the PVAX-II of the BIOSON research institute and the Cyber 205 in Amsterdam. Access to the latter computer was made possible by a grant from the “Werkgroep Super- computers”, which is gratefully acknowledged.

References Barker. W. C.. Hunt. L. T., Orcutt. B. C.? George. D. G..

Yeh, L. S.. Chen, H. R., Blomquist, M. C., Johnson, G. C.. Seibel-Ross, E. I. & Ledkey. R. S. (1984). Report of the National Biomedical Research Foundation, Georgetown University. Washington. DC.

Bernstein, F. C., Koetzle, T. F., Williams, G. *J. B.. Meyer, E. F., Jr, Brice. M. I)., Rodgers, ,J. R., Kennard. O., Shimanouchi, T. & Tasumi, M. (1977). J. Mol. Biol. 112, 535-542.

Bricogne, G. (1976). Acta Crystallogr. sect. A. 32, 832-847. Browning, K. S., Uhlinger, D. ,J. & Reed. L. J. (1988).

Proc. Nat. dead. Sci., U.9.A. 85, 1831-1834. (Tolman, P. M., Fehlhammer, H. & Bartels. K. (1976). In

Crystallographic Computing Techniques (Ahmet. F. R., Humland, K. & Sedlacek, B., eds). pp. 248- 258, Munksgaard, Copenhagen.

Crowther. R. A. (1972). In The Molecular Replacement Method (Rossmann, M. G., ed.), pp. 173-178, Gordon & Breach, New York.

A. vinelandii Lipoamide Dehydrogenase 379

Crowther, R. A. & Blow. D. M. (1967). Acta CrystaZEogr. 23, 544-548.

de Kok, A., Visser, A. J. W. G. & de Graaff-Hess, A. C. (1982). In Flavins and Flavoproteins (Massey, V. & Williams. C. H., Jr, eds), pp. 61-65, Elsevier, North Holland, Inc.. Amsterdam.

Fox. B. & Walsh, C. T. (1983). Biochemistry, 22, 4082- 4088.

Fujinaga, M. & Read, R. J. (1987). J. AppZ. Crystallogr. 20. 517-521.

Hamilton, W. C., Rollett. J. S. & Sparks, R. A. (1965). Acta CrystaZZogr. 18, 129130.

Holden, H. M., Rypniewski, W. R., Law, J. H. & Rayment, I. (1987). EMBO J. 6, 15651570.

Jones, T. A. (1985). In Methods in Enzymology (Wyckoff, H. W., Hirs, C. H. W. & Timasheff, S. N., eds), vol. 115. pp. 157-171, Academic Press, New York.

Kabsch, W. & Sander, C. (1983). Biopolymers, 22, 2577- 2637.

Rossmann, M. G. & Blow, D. M. (1962). Acta, Crystallogr. 15, 24-31.

Schierbeek, A. J., van der Laan, J. M., Groendijk, H., Wierenga, R. K. & Drenth, J. (1983). J. Mol. BioZ. 165, 563-564.

Karplus, P. A. & Schulz, G. E. (1987). J. Mol. BioZ. 195, 701-729.

Kochi, H. & Kikuchi. G. (1976). Arch. Biochem. Biophys. 173, 71-81.

Krauth-Siegel, R. L., Blatterspiel, R., Saleh, M., Schiltz, E., Schirmer, R. H. & Untucht-Grau, R. (1982). Eur. J. Biochem. 12, 259267.

Krauth-Siegel, R. L., Enders, B., Henderson, G. B., Fairlamb, A. H. & Schirmer, R. H. (1987). Eur. J. B&hem. 164, 123-128.

Leslie, A. G. W. (1987). Acta CrystaZZogr. sect. A. 43, 134- 136.

Schierbeek, A. J., Renetseder, R., Dijkstra, B. W. & HOI, W. G. J. (1985). In lMoZecuZur Replacement (Machin, P. A., ed.), pp. 16-21, Science and Engineering Research Council, Daresbury.

Schneider, G., Lindqvist, Y., Brand&n, C. 1. & Lorimer, G. (1986). EMBO J. 5, 3409-3415.

Schulz, G. E., Schirmer, R. H. & Pai, E. F. (1982). .J. Mol. BioZ. 160, 287-308.

Shames, S. L., Fairlamb, A. H., Cerami, A. & Walsh, C. (1986). Biochemistry, 25, 35193526.

Sim, G. A. (1960). Acta Crystullogr. 13, 511-512: Stephens, P. E., Darlison, M. G., Lewis, H. M. & Guest,

J. R. (1983). Eur. ,I. B&hem. 133, 155-162. Sussmann, J. L. (1985). In Methods in Enzymology

(Wyckoff, H. W., Hirs, C. H. W. & Timasheff, S. N., eds), vol. 115, pp. 271-303, Academic Press, New York.

Moss, D. S. (1985). Acta CrystaZZogr. sect. A, 41, 470-475. Thieme, R., Pai, E. F., Schirmer, R. H. & Schulz, G. E. Ogata, C., Hatada, M., Tomlinson, G., Shin, W. C. & (1981). J. Mol. BioZ. 152, 763-782.

Kim, S. H. (1987). LVature (Lomlonf, 328, 73+742. Thaller, C., Weaver, L. H., Eichele, G., Wilson, E., Otulakowski. G. & Robinson, B. H. (1987). J. BioZ. Chem. Karlsson, R. & Jansonius, J. N. (1981). J. Mol. BioZ.

262, 17313-17318. 147, 465469. Pons, G., Raefsky-Estrin, C., Carothers, D., Pepin, R. A.,

Javed, A. A., Jesse, B. W., Ganapathi, M. K., Hamols, D. & Patel, M. S. (1988). Proc. Nat. Acad. Sci., U.S.A. 85, 1422-1426.

Wang, B. C. (1985). Methods EnzymoZ. 115, 90-112. Westphal, A. H. & de Kok, A. (1988). Eur. J. B&hem.

172, 294-305.

Read, R. J. (1986). Acta Crystallogr. sect. A, 42, 140-149. Read, R. J. & Schierbeek, A. J. (1989). J. AppZ.

CrystaZZogr. In the press.

Williams, C. H., Jr (1976). In The Enzymes (Boyer, P. D., ed.), vol. 13, pp. 89-173, Academic Press, New York.

Reed, L. J. (1974). Act. Chem. Res. 7, 4046. Rice, D. W., Schulz, G. E. & Guest, J. R. (1984). J. Mol.

BioZ. 174, 483496. Rossmann, M. G. (1979). J. AppZ. Crystallogr. 12, 226

238. Rossmann, M. G. (1984). In Biological Systems: Structure

and Analysis (Diakun, G. P. C Garner, C. D., eds), pp. 2840, Science and Engineering Research Council, Daresbury.

Edited by R. Huber

Documents

X-ray Structure of Lipoamide Dehydrogenase from Azotobacter