.J. Mol. Biol. (1990) 211, 617-632

Crystal Structure of Plastocyanin from a Green Alga, Enteromorpha prolifera C. A. Collyer’“f, J. M. GUSS’, Y. Sugimura’, F. Yoshizaki’ and H. C. Freeman’x 1 Department of Inorganic Chemistry University of Sydney, Sydney 2006, Australia ‘Faculty

of Science, Toho University,

(Received 12 May

1989; accepted

Chiba

274, Japan

8 September

1989)

The crystal structure of the &containing protein plastocyanin (Nr 10,500) from the green alga Enteromorpha prolifera has been solved by molecular replacement. The structure was refined by constrained-restrained and restrained reciprocal space least-squares techniques. The refined model includes 111 solvent sites. There is evidence for alternate conformers at eight residues. The residual is 0.12 for a data set comprising 74% of all observations accessible at 1.85 A resolution. The P-sandwich structure of the algal plastocyanin is effectively the same as that of poplar leaf (Populus nigra var. itakica) plastocyanin determined at 1.6 A resolution. The sequence homology between the two proteins is 56%. Differences between the contacts in the hydrophobic core create some significant (0.5 to 1.2 A) movements of the polypeptide backbone, resulting in small differences between the orientations and separations of corresponding P-strands. These differences are most pronounced at the end of the molecule remote from the Cu site. The largest structural differences occur in the single non-p strand, which includes the sole turn of helix in the molecule: two of the residues in a prominent kink of the poplar plastocyanin backbone are missing from the algal plastocyanin sequence, and there is a significant change in the position of the helical segment in relation to the P-sandwich. Several other small but significant structural differences can be correlated with intermolecular contacts in the crystals. An intramolecular carboxyl-carboxylate hydrogen bond in the algal plastocyanin may be a,ssociated with an unusually high pK,. The dimensions of the Cu site in the two plastocyanins are, within the limits of precision, identical.

1. Introduction

288 (Colman et al., 1978). The Cu atom is not exposed to the solvent. Crystal structure analyses and refinements are available for the PC from poplar leaves (Populus nigra var. italica) in the oxidized [Cur’] state at pH 6.0 and pH 42 (Guss & Freeman, 1983), in the reduced [Cu’] state at six pH values between pH 3.8 and pH 7.8 (Guss et al., 1986), as a Hg”-substituted derivative (Church et al., 1986), and as the apoprotein (Garrett et al., 1984). In the crystalline state, the various chemical modifications cause few significant structural or conformational changes anywhere in the molecule, except in the vicinity of the Cu site. A knowledge of the structure of PC has been helpful in the interpretation of many aspects of the electron-transfer reactions of the protein with inorganic redox partners (Sykes, 1985; Brunschwig et al., 1985) and in viva (Haehnel, 1986). In most discussions of structure-function relationships involving PC it has been assumed that the structure of poplar PC in the crystalline state is a valid model

The “blue” or “type 1” copper-protein plastocyanin (Pc$) is an essential electron-carrier between photosystems II and I in the photosynthetic apparatus of all higher plants and some algae. The PC molecule consists of a single polypeptide of 97 to 105 amino acids (the number depending on the organism from which the protein is extracted), plus one Cu atom. The Cu site lies about 6 A (1 d = @l nm) below the surface at one end of the molecule, which is a b-sandwich with a slightly flattened cylindrical shape and approximate dimensions 40 A x 32 A x 7 Present address: Physics Department, Imperial College, Prince Consort Road, London SW7, U.K. $ Author for correspondence. $ Abbreviations used: PC, plastocyanin; Cu”Pc, plastocyanin in the oxidized state; Cu’Pc, plastocyanin in the reduced state; n.m.r., nuclear magnetic resonance; a.s.a., accessible surface area; e.s.d., estimated standard deviation; r.m.s., root mean square. 002%2836/90/030617-16 $03.00/O

617

0 1990 Academic Press Limited

pro1

100 *

90 +

DASKaSMSEEDLLNAKCETFEVAkSNMGEYSFYCSPHQGA~MV~KVTV~

Poplar E.

80 +

70 +

60 +

ifera

GVYCDPHSGAGMKMTITVQ

DADAISA-(E)-QYLNSKCQTVVRKLTTPGTY .

e

.

e

0

0

00

.aee

Figure 1. The amino acid sequences of the plastocyanins from E. prolifera (Simpson et al.; 1986) and Sopulus nigra var. italica (R. P. Ambler, unpublished work cited by Guss & Freeman (1983)). Residues printed in bold face are conserved in higher plant plastocyanins (Sykes, 1985). Filled circles indicate residues that are common to all known plastocyanin sequences. In the region of residues 57 to 61 the alignment is based on the superposition of the structures. Throughout this paper, residues are numbered according to their position in the poplar plastocyanin sequence. The exception is the single residue in E. prolifera plastocyanin that occurs insbead of residues 58-59-60 in poplar plastocyanin. This residue is labelled Glu59 * in the text and shown in parenthesis above to emphasize that it is structurally not equivalent to any residue in poplar plastocyanin (see Results).

for PCS from other sources and for PCS in solution. Evidence supporting this assumption has been provided by the close similarity between the ‘H n.m.r. spectra of Cu’Pcs from four higher plants (Freeman et al., 1978), by the high degree of sequence conservation among the proteins of the PC family (Guss & Freeman, 1983; Sykes, 1985; Simpson et al., 1986), and more recently by the successful use of models derived from the poplar PC structure to determine the structures of other PCS by the molecular replacement method from X-ray diffraction data (present work; Garrett, 1987; Tong, Guss & Freeman, unpublished results) and from two-dimensional n.m.r. data (Moore et al., 1988). There remains, however, a need for quantitative data that test the sensitivity of the molecular structure to changes in the amino acid sequence, intermolecular contacts and crystallization medium. In the present work we have been able to address two of these questions by studying the crystal structure of the PC from a green alga, Enteromorpha prolife’era. The polypeptide of this PC comprises 98 When aligned with the amino acid residues. sequence of poplar PC, the sequence of Enteromorpha prolifera PC has an additional residue at the NH, terminus but two residues fewer between residues 57 and 61 (Fig. 1). There is homology at 56 residues, leaving 40 amino acid differences. Comparisons with poplar PC thus afford ample opportunities for exploring the effects of amino acid structure. Further, changes on the tertiary E. prolifera Pc crystallizes in a space-group (24) that is different from that of poplar PC (P2,2,2,). The structural effects of intermolecular contacts, which must be different in the crystals of the two proteins, may accordingly also be assessed.

2. Experimental (a)

Procedures data

Difraction

Crystals suitable for X-ray crystal structure analysis were grown in sodium phospha.te buffer (20 mar) with saturated ammonium sulphate solution at 5 “C (Yoshizaki et al., 1981). The crystals were transported and stored at room temperature in the mother liquor. Crystals for diffraction experiments were mourned in contact with the mother liquor. The measured pH of the mother liquor was 5%. The crystals belong to the tetragonal space group 14 with unit cell parameters a=53.98, e= 5948; Z= 8 (Freeman et al., 1983). There is 1 molecule per asymmet,ric unit. The solvent content, earlier reported as 60% (v/v), is 40 y0 (v/v).

Two crystals, A and B, were select,ed for data collection (Table 1). Counter measurements were made with Xi-

Table 3 Summary

of counter data sollection

Crystal A Crystal dimensions (mm) @4x@4x@2 d’nit cell parameters a (A) 53.7 59.3 c (4 Resolution (A) m-3.0 No. of measurements 2023 Final fractional decomposition (%) 10 Range of absorption corrections? 1.0-1.6 Condition for inclusion of reflections So. of unique observations &Ww) No. of measurements in R,(merge) t North et al. (1968).

Crystal

13

@5XO5XO~2 539 59.4 3.0-2.2 2908 16 1.1-17 I > 2cT(I) 3715 0.026 488

Crystal

I

Xtructure

I 100

300

500

IF01

Figure 2. Mean squared differences between structure amplitudes observed by counter and film methods, plotted as a function of the structure amplitude. The curve is a second-order polynomial fitted to the data.

filtered CuK, radiation on an Enraf-Nonius CAD-4/F diffractometer. The cited unit cell parameters were determined by counter measurements of 19 reflections from crystal B in the range 13.7” d f3< 148, and were later used in the refinement calculations. Intensities were measured only for reflections with h, L, 12 0. Lorentz, polarization, decomposition and absorption corrections were applied. The counter data with I> 20(I) from crystals A and B were placed on a common scale and merged. This data set represented S6o/o of the accessible observations to 2.2 A resolution (Table 1). High-resolution data were recorded on film using the X11 instrument at the European Molecular Biology Laboratory Outstation at the synchrotron DESY, Hamburg, using synchrotron X-rays with 1= 1.47 A. Crystal C (96 mm x 0.5 mm x 63 mm) was aligned by means of precession photographs before exposure to the synchrotron beam. The crystal was mounted with the 4-fold crystallographic axis along the rotation axis of the oscillation camera. The sum of all 2.8” rotations was 98.4”. The films were developed, scanned and processed on-site at DESY. Optical densities were recorded on an Optronics P-1000 film scanner, using a 100 pm raster and a maximum optical density photometer setting of 3. The resulting data to 185A resolution were processed using the program FILME, described by Nyborg & Wonacott (1977) as the Munich System. No absorption corrections were applied to the film measurements.

t R,(symmetry)

and R,(merge) are defined as follows:

619

of Plastocyanin

The software used for subsequent data-processing was the calculations and Fourier transformations program system PROTEIN (Steigemann, 1974). The observations from each film pack were treated as an independent set of measurements. The values of R,(symmetry)t for various film packs ranged from 0.050 to 0.120. Linear scale factors were determined by merging the film data with the scaled counter data. The components of partially recorded reflections on consecutive films were then summed. The final consolidated data set was produced by merging with unit weights the film observations (4 > d > 185 A) and the counter observations (a?> 2.2 A). The ith measurement Ii of a reflection having mean intensity I was rejected if JIi-Il+~I~ 2 025. The R,(merge) between the film and counter sets was 0.09 for 2537 common reflections. The final data set comprising 5415 observations represented 74% of the accessible observations to 1.85 A resolution. The weights to be attached to reflections during the refinement were estimated as follows. The reflections that occurred in both the counter and photographic data sets were sorted into groups according to the magnitude of F,. The averages of the structure amplitudes in ranges of IJ’,,l, and the average squares of the differences between corresponding structure amplitudes in the 2 sets of observations, were calculated. A second-order polynomial was fitted to the mean squared discrepancies as a function of the average structure amplitudes (Fig. 2). An empirical curve derived from 2 sets of independent observations in this way may be used to estimate the average systematic errors of structure amplitudes (Freeman & Guss, 1972). In the present work, the weight given to each F, in the least-squares refinement was the inverse of the corresponding mean squared discrepancy read from the curve in Fig. 2. (b) Molecular replacement calculations An atomic model based on the refined crystal structure of poplar PC (Guss & Freeman, 1983) was constructed for the Patterson searches. The model included (1) the polypeptide backbone with the exception of residues 58 to 61; (2) the side-chains of residues that are invariant in higher plant PCS (Fig. 1); (3) any atoms that occur in the sidechains of all known variants of variable residues; and (4) the Cu atom. The rotation function of Crowther (1972) was calculated. The molecular transform was computed from the atomic model placed in a box with 70 A edges. With data of 10 to 4 A resolution and a 23 A radius of integration, a single peak with an amplitude 1.6 times that of the next highest peak was found. A single 2-dimensional R-factor search using data of 9 to 6 A resolution produced a unique minimum at R = 941. This minimum defined the position of the molecule, since in space group 14 the z co-ordinate is fixed by the choice of origin. There were no steric inconsistencies in the resulting solution of the structure.

R, = (c) Rejinement

where N is the number of unique reflections (Friedel pairs being treated as equivalents), Ii,p is the ith estimate of the intensity of the rth reflection, and T, is the mean value of the intensity of the rth reflection. For R,(symmetry), n(r) is the number of measurements of the rth reflection in a particular film pack. For R,(merge), n(r) is the number of measurements of the rth reflection in the data set.

The rotational and positional parameters found by molecular replacement were subjected to 10 cycles of constrained reciprocal space least-squares refinement using the program CORELS (Sussman et al., 1977; Sussman, 1983). The net change in molecular position was equivalent to a 4” rotation about an axis passing through the centre of gravity of the molecule, followed by a 0.5 A translation. The linear R-factor for the data in the range 9 to 6 A was reduced from 941 to 0.38. The agreement

C. A. Collyer

remained equally good t’o 3.3 ,%, the R-factor for data in the range 9 to 3.3 A being 639. with co-efficients A Fourier synthesis calculated 2F,- FC and phases CI, for 1119 reflections in the range 7 to 3.3 A was immediately interpretable. The atomic model and electron density were displayed on an Evans & Sutherland Picture System using the program PROD0 (Jones, 1978). Features of the map consistent with the primary st,ructure (Simpson et al.; 1986) included density along the ent#ire polypeptide backbone of the model with the exception of residues 8 to 11, 70 to 7 1 and 99; density corresponding to many of the O(peptide) atoms; density corresponding to the additional residue (“residue 0”) known to occur at the NH, terminus of E. prolifera PC; and continuous density bridging a kink at residues 57 to 61 in the poplar PC structure, as expected from the deletion of 2 residues in this region. A new residue 0 was added at the ?GH, terminus of the poplar PC model. Residues 58 to 60 were replaced by a new residue 59* having no direct equivalent in poplar PC. Stereochemical refinement produced shifts of 2 A in adjacent residues. The geometry of the modified model was then idealized using CORELS. Refinement against the X-ray data in the range 7 > d > 2.7 A led to marginal improvement in the residual from R = 0.31 to R = 029. The entire. polypeptide backbone except the region at residues 8 to 9 could now be traced in a 2F, - FC synthesis at 2.7 A resolution. Following PROLSQ refinement (Hendrickson & Konnert, 1980) at 2.1 A resolution, the missing region was finally traced as a bend similar to that found in poplar PC. The subsequent rounds of refinement consisted of: (1) an examination of the largest peaks in a difference map for indications of additional solvent molecules or uncorrected errors: (2) an examination of 2F0--F, maps for symptoms of errors in the model; (3) implementation of any indicated changes; (4) stereochemical refinement; (5) calculation of “removed” maps in regions where ambiguities persisted; (6) exclusion of atoms that occupied stereochemically implausible sites or did not’ lie in density; and (7) PROLSQ least-squares refinement to convergence. In step (2), all atoms with an isotropic temperature factor 245 8’ or with density 45 A’) was tested by a similar procedure. In the final 28 cycles the torsion angles of the 4 Cubinding side-chains were unrestrained. The torsion angle restraint was also lifted for the side-chain of Tyr62, which is involved in an intermolecular contact. The final 14 cycles were computed with hydrogen atoms included in the model at idealized positions. A number of hydrogen atom positions depended upon side-chain torsion angles that, were not fully determined by the known non-hydrogen atomic positions. An examination of local hydrogenbonding networks enabled some of the relevant torsion angles, and hence the positions of many of the missing hydrogen atoms, to be estimated. A total of 193 cycles of reciprocal space least-squares refinement were computed. They included 10 cycles of

et al.

rigid-body refinement of the search molecule, 36 cyclw of constrained-restrained refinement and 22 cycles of restrained refinement of a partial molecular model at medium resolution, and 126 cycles of restrained refinement of a substantially complete model against the entire data, set. When hydrogen atoms were included in the last rounds of refinement, the agreement with the observed data and with stereochemical criteria improved marginally, as also observed by Phillips (1980). The r.m.s. shifts in the positional and thermal parameters in the last refinement cycle were 0015 A and 0.18 A’, respectively. The final residual R was 0.117 for 5331 reflections in the range 8.4 > d > 1~9A. The st,atistics of the final model are given in Table 21. (d) Sites with partial

occupancy

The 8 residues that are modelled in 2 conformations account for 104 atomic sites with 0.5 occupanmy (i.e. 52 non-hydrogen atoms). In a,ddition, 14 of the 111 solvent sites are assigned 0.5 occupancy on the basis of alternative hydrogen-bonding patterns. The disordered protein groups and their mean temperature factors are listed in Table 3. (e) Atoms

with high temperature

factors

The average isotropic temperature factor for all atoms in the protein molecule is (B) = 16 A2. The highest individual atomic temperature factors are B= 44 and B = 54 A2 for protein and solvent atoms with occupancy 1.0; and B = 34 and 29 A2 for protein and solvent atoms with occupancy @5> respectively. The atoms with the highest values of R occur in the side-chains of Asp44 and Asp53, and high average values of B occur in several of the disordered side-chains (Table 3). The positions of all atoms and side-chains associated with high afomic thermal para,meters B > 30 A2 were carefully checked in F, - FC and 2F, - F, electron-density difference maps, Examples of 2F,-FF, densities at atomic sites with high atomic thermal parameters (Asp53) and at, atomic sites with 0.5 occupancy (Glu43) are given in Figs 3 and 4. (f) Accuracy and precision The identification of disordered backbone segments and side-chains, and t,he ability to ext,end parts of the solvent network beyond the first solvation shell, are indieators that the extensive refinement) has produced an accurate model of the E. prolifeera PC molecule. On the other hand. the low residual (R= 0.12, compared with R=0165 for poplar Cu”Pc and 0.15 to 0.17 for poplar Cu’Pc) does not necessarily imply that the structure has been determined more precisely than that of poplar PC. Although the data 7 The atomic positional parameters, atomic thermal parameters and structure factors have been deposited with the Protein Data Bank, Brookhaven National Laboratory, Upton NY 11973, U.S.A. Tables of: (1) water molecule environments; (2) intramolec-ular hydrogen bonds between atoms of the polypeptide backbone; and (3) intramolecular hydrogen bonds involving side-chain functional groups have been deposited as supplementary material with the British Library Lending Division as Supplementary Publication no. SCP40019. Copies may be obtained through Service Enquiries, British Library Lending Division, Boston Spa, Wetherby, West Yorkshire LS23 ?BQ, Engla,nd.

Crystal

Xtructure

621

of Plastocyanin

Figure 3. Electron density at a side-chain with high atomic thermal parameters (Asp53). The (ZF,--E”,) map is contoured at 0.6 eAe3 (i.e. 1 e.s.d.). The mean temperature factor B for the side-chain atoms is 31 8’.

Table 2 Xtatistics

of PROLXQ

least-squares

rejinement

Number of cycles Resolution (A) No. of reflections Threshold for inclusion No. of non-hydrogen atom sites No. of hydrogen atom sites Final residuai R

153 8.4-1.9 5331 I > 2a(l) 875 736 0.117

Target D

r.m.s. A

Max. A

Distances (8) Bonds Angles Dihedral X-H bonds X-H angles

0030 0040 0.050 0.100 @080

0.016 0034 0.033 0031 0023

0056 0.144 @163 @142 0092

Planes (d)

0.020

0010

0.03

0150

0132

0.44

Chiral

centres (A “)

contactst (A) Single Multiple H-bonded (X . Y) H-bonded (H . Y) Torsion angles (deg.) Omega Chi Aromatic

0.5 @5 0.5 0.5

0,155 0.209 @192 0.078

5.0 150 2@0

23 195 35.8

Thermal parameters B (A’) Main bonded 50 Main angle 7.5 Side bonded 10.0 Side angle 150

3.1 4.1 41 58

056 0.63 0.46 0.20 6 54 83

Restraints against excessive shift Positional parameters (A) 0.1 Thermal parameters (8’) 30 Reduction of van der Waals’ contact distances (d) Single @3 Multiple 0.0 H-bond 0.2 t Intermolecular contacts were unrestrained. such contacts were smaller than the maximum for intramolecular contacts.

All A values for

A values quoted

sets used for all the refinements have similar limits of resolution (1.85 d for E. prolifera Cu”Pc, 1.6 a for poplar Cu”Pc, 1.7 to 2.05 A for poplar Cu’Pc), the proportion of the accessible reflections included in the refinement is substantially lower in the present work (74%, compared with 80 to 93 %). Thus the lower residual R is a result of a lower ratio of observations to variable parameters, the number of observations having been reduced by the use of stringent criteria to reject deviant data while the number of variables is increased by the inclusion of alternate conformers and a large number of solvent sites. In the structure analysis of poplar Cu”Pc, a comparison between the results of refinements based on 2 independent data sets led to an estimate of about 013 a for the r.m.s. error in the C, N and 0 atomic positions (Guss & Freeman, 1983). It was emphasized that this estimate was an average that included contributions from more precisely determined atomic positions in the polypeptide

Table 3 Residues rejined in two conformations First disordered Residue

atom

Lys4 4’ Asnl8 18 Glu43 43 Glu59*t 59*’ Asp61 61’ Asp85 85’ Lys93 93 Gln99 99

CG

Mean temperature Main chain

CG N N N

16 14 7 8 3 6

CG CD N

30 29

factor (A’) Side chain 23 29 22 26 27 27 19 18 9 18 15 15 29 27 30 33

The second conformer of a disordered residue is identified the symbol ‘. t See legend to Fig. 1 for significance of the label “59*“.

by

C. A. Collyer

et al.

Figure 4. Electron density at a disordered residue (Glu43). The (SFO- F,) map is contoured at,.0.3 eam3 (i.e. 0.3 e.s.d.j, The 2 conformers are labelled 43 and 43’, respectively.

backbone and less precisely determined atomic positions in surface residues with high thermal parameters. The estimated standard deviation of the Cu-ligand bondlengths was estimated to be 005 8. Bearing in mind the reservations stated in the preceding paragraph, we estimate: (1) the r.m.s. error of the atomic positions in the polypeptide backbone of E. prol$era PC as @15 8, and (2) the standard deviation of the Cu-ligand bond-lengths as 007 A.

Table 4 Comparison

between the Cu sites of E. prolifera poplar Cu”Pc s

and

Protein E. prolifem Cu-liyand bond-lengths (A) Cu-N(His37) Cu-N(HisR7)

I.89

cu-S(Cys84)

217 212

CWS(Met92)

2.92

Bond-angles at Cu atom (deg.) S(His37)-Cu-N(His87) -S(Cys84)

-S(Met92) N(His87)-Cu-S(Cys84) -S(Met92) S(Cys84)-Cu-S(Met92)

104 125

90 120 102 108

Torsion angles in Cu-binding side-chains (deg.) His37 X1 -55 x2

His87

X,

Cys84

x2 x,

Met92

-65 -63

139 166

X1 x2

178 163

x3

173

Temperature factors B of Cu atom and G.-binding (averaged) (A “) 11 cu 5 His37 8 His87 Cys84

Met92

8 8

Poplar

2.04 2.10 2.13 290 97 132 85 123 103

108 -67 -62 -61 138 170

191 165 172 side-chains 13 9 8 9 7

a.

es&s

We shall describe and discuss t,he molecular structure of E. prolifera PC by comparing it with t,hat of poplar Cu”Pc at 1.6 ,& resolut8ion (GUS & Freeman, 1983). The poplar Cu”Pc structure is a useful benchmark because it has been refined to fairly high precision and is supported by independent structure refinements of the same protein under a variety of chemical conditions. It will occasionally be convenient to describe a difference between E. proliferera and poplar Pcs as a “movement” or “change” from the poplar PC structure. In such instances it should be remembered that in an evolutionary sense differences between the two PCS are more likely to represent changes from the algal to the higher piant protein than vice versa.

(a) The Cu site The Cu sites of E. prolifera Cu”Pc and poplar CuI’Pc have closely similar dimensions, side-chain torsion angles and thermal parameters (Table 4). The largest bond-length difference between the two proteins, 0.15 & occurs at Cu-S(His37). We attach no significance to this difference since, given estimates of 0.07 a and 0.05 a for the standard deviaLions of the Cu-ligand bond-lengths in E. prolifera and poplar PCS, respectively (see above), it is smaller than twice its own e.s.d., 0.09 8.

(b) Superposition poplar

of E. prolifera Cu”Pc 8

and

Figure 5 represents the results of a’ superposition of the polypeptide backbones of E. prolifera and poplar Cu”Pcs by the met,hod of Kabsch (1978). Residues 0, 51 to 61 and 99 were excluded from the calculations. The reasons for the exclusions will become clear in the following discussion. The r.m.s. difference between corresponding atomic positions in the two proteins, averaged over all the atoms

Crystal

Xtructure

of Plastocyanin

623

Figure 5. Superposition of C” backbones of E. prolifera Cu”Pc (heavier lines, arrowheads at residues 0, 10, ., 90) and poplar Cu”Pc. The small circle represents the position of the Cu atom. The large circle indicates the region where 2 residues that occur in poplar PC between 57 and 61 are absent in E. prolifera PC. (This and other molecular diagrams were produced using a program by Lesk & Hardman (1982).)

included in the superposition, is 0.5 A. The largest difference is 1.4 A. We shall use the r.m.s. difference between the atomic positions as a convenient and plausible threshold of significance for “real” structural differences. By implication, differences below this threshold are ascribed to random errors rather than to changes in the primary structure or changes in intermolecular contacts. The value 0.5 A represents 2.5 times (0~15~+0~13~)1’2, where 615 A and 0.13 A are the r.m.s. errors in the atomic positions of E. prolifera and poplar PCS, respectively (see above). In antecedent comparisons between poplar Cu”Pc and Cu’Pc, and between pairs of Cu’Pc structures, somewhat lower thresholds of significance were derived from plots of the r.m.s. differences between corres-

ponding atomic positions as functions of the atomic thermal parameter B (Guss et al., 1986). For example, the r.m.s. differences were 615 A for polypeptide backbone atoms with B = 30 A2, and 0.25 A for side-chain atoms with B = 35 A2, respectively. In the polypeptide backbones of both E. prolifera and poplar Cu”Pcs, the highest atomic thermal parameters are about 30 A2. The use of 0.5 A as a threshold of significance is thus conservative. The close correspondence between the two PC structures is obvious from Figure 5. A number of differences can also be clearly seen. An additional residue 0 extends the NH,-terminal strand of the algal PC parallel to the neighbouring j-strands. The kink in the backbone at residues 43 to 44 and the short helical segment at residues 51 to 56 appear to

Figure 6. Stereo view comparing the polypeptide backbones of E. prolifera PC (filled black atoms) and poplar PC near residues 57 to 61. The side-chains of Ala57 in E. prolifera PC and Met57 in poplar PC are also shown. The molecules are superposed as described in the text. The residue numbers refer to the poplar PC sequence.

C. A. CO&W eL al.

60 Residue

70

80

90

number

Figure 7. Mean differences between positions of corresponding backbone atoms after superposition of the E. prvi+a and poplar PC structures. The values for residues that are part of the P-strands are shown as bold lines. For residues 43, 61 and 99 (which are disordered in the E. proZ$%ra PC structure) the differences are calculated from the means of the atomic positions in the 2 conformers. Residue 100 represents the Cu atom.

be significantly displaced. Finally, the P-bend at residues 58 to 60 in poplar PC is absent, in the algal PC as a result of the deletion of two residues (see Fig. 1). Details of the two structures in this region are compared in Figure 6. In E. prolifera PC a single residue replaces residues 58 to 60 in poplar PC. We label this residue “Glu59*” to emphasize that it is not structurally equivalent to any of the residues Ser58-Glu59-Glu60 in poplar PC. The sequence alignment in Figure 1 thus has a structural rationale. A more quantitative overview of the structural differences between the two PCS is given in Figure 7 where the mean differences in position are averaged over the backbone atoms of each residue. The r.m.s. value of these averaged differences is 96 A for the residues included in the superposition. Significant differences (defined above as being aO.5 A) are found at residues 2, 8, 12, 17-22, 26-27, 30, 35-36, 43-44, 50-57, 61-71, 88-89, 91 and 99; the largest discrepancies (2 l-2 A) occur at residues 52-54, 57, 61 and 99. Qualitative descriptions of the backbone changes that are associated with some of these differences have been given above. The rationalization of the differences in terms of intra- and interforces molecular requires a more detailed consideration of the molecular structure.

(c) Description

of P-sandwich

structure

The PC molecule is best described as a P-sandwich, i.e. two /?-sheets (I and II) separated by a hydrophobic core (Chothia & Lesk, 1982; Guss’ &

Freeman, 1983). Seven strands of the polypeptide backbone (strands 1 to 4 and 6 to 8) have substantial p character and contribute to the P-sheets (Fig. 8). An additional strand (strand 5) has no fi character and includes the sole segment of helix in the molecule. Most of the side-chains in the hydrophobic core belong to residues in the P-sheets. The residues are identified in Figure 8. Four other residues also have side-chains buried in the hydrophobic core: 46 (in strand 4, where it lies between the /l-turns at residues 42 to 45 and 47 to 50), 55, 57 and 63 (all in the non-b strand 5). In poplar PC these residues are identified as “buried” by having accessible surface areas 3.3& 580... Ser58 is absent), 65 0 . . .68 N ( > 3.3 A)? and 84 0 . . . 88 N (a, bad 0 . . H-N angle is noted in both PCS). Two hydrogen bonds in E. prolifera PC correspond to distances > 3.6 A in poplar PC, namely 39 3,. .57 0 and 85 0.. .88 X. Lines fitted to the six major segments of/3-strand in the P-sa,ndwich have closely similar positions and directions in the two PCS (Fig. 9). The largest displacement is observed for strand 6 in b-sheet 1; and

625

Crystal Structure of Plastocyanin

(a) S6

53

Sl

S2A

54

57

se

526

II

I

Figure 8. Residues forming P-sheets I and II in PCS. The /L&rands are numbered according to Guss &

Freeman (1983). Only those residues that have side-chains on the interior surfaces of the B-sheets are labelled. Single symbols indicate residues that are homologous in E. prolifeyferaand poplar PCS. Double symbols indicate nonhomologous residues (left-hand symbol, E. proZifera PC; right-hand symbol, poplar PC).

the smallest for strand 7 in P-sheet II. The divergences are most marked at the end of the molecule remote from the Cu atom. The calculations show that the relationship between the P-sheets is substantially the same in both PCS. On the other hand, the divergences and displacements are not pairs of negligible, the angles between corresponding lines differing by up to 5”. It therefore becomes appropriate to use the general trends in the P-sheet structure revealed by the least-squares averaging process as a guide in looking for local perturbations.

Figure region in (a) when (b) when

(b)

10. Projections

E. prolifera

of the C” atoms of the helical

PC (heavier lines) and poplar PC:

the molecules are superposed as in Fig. 5; and the helices alone are superposed.

two structures along the non-p strand 5 is poor from one end to the other (Fig. 5). The largest differences are centred on the two deletions/insertions between residues 58 and 60, and are presumably transmitted to the adjacent residues 57 and 61. The fit is, however, remarkably poor over the entire helical segment at residues 50 to 56 (Fig. 10(a)). If the helical segments in the two structures are treated as rigid bodies and they alone are superposed, the agreement between them is much improved (Fig. 10(b)). The transformation between the two helix positions requires a rotation of 20” about an axis through the helix centre of gravity, followed by a translation of 1.1 8. In other words, the structure of the helical segment is conserved in the two PCS; the positions of the helical segments with respect to the rest of the molecule are different. In E. prolifera PC the helical segment is closer to P-sheet II. (e) Sequence-dependent changes

(d) Shift in helical segment In contrast with the absence of pronounced differE. prolifera and poplar PCS in the p-street regions, the correspondence between the

ences between

in tertiary structure

As already mentioned, the movements of the P-strands in response to differences between the

primary

structures are small. We do not attempt to

53

St 0 cu

57

S2B 56

~~

SB

56 SB

Figure 9. Stereo diagram comparing the positions of a number of fi-strands in E. prolifera PC (heavier lines) and poplar PC after superposition of the molecules. The p-segments consisting of 5 or more residues are represented by straight lines fitted to the backbone atomic co-ordinates, and are labelled as in Fig. 8. The angles between corresponding lines in the 2 PCS differ by less than 2” for P-segments in the same P-sheet, and by less than 5” for P-segments in difj'erent P-sheets.

C. A. @olEyeret al.

626

Strand

5

Strand P-Sheet

6

Strand

5

Strand

6

P-Sheet

I

1

(b)

(a)

Figure 11. Two sections, separated by 2.5 A, through corresponding regions of computer-generated space-filling models of: (a) E. prolijera PC; and (b) poplar PC, near residue 57. There are sequence differences between poplar and E. prolifera PCSat residue 57 (Met -+ Ala) in the non-p strand 5, and at residues 70 (Phe -+ Val) and 72 (Val -+ Arg) in /?-sheet I. _

I

account for all of them. In the following examples the causes and effects seem to be clear-cut. (i) Strands 5 (helical segment), 4 (P-sheet II) and 6 (P-sheet I) The difference between the positions of the helical segment in poplar PC and E. prolij’era PC is related to the following changes in primary structure: Poplar Ile146 Ser53 Met57 Phe70 Va172

-+ -+ -+ --j -+

E. prolifera Va146 Asp53 Ala57 Va170 Arg72

In poplar PC the position of the helix appears to be determined by hydrophobic contacts. In particular, Ala52 is sterically hindered by the bulky sidechain of Ile46, a “buried” residue (solvent-accessible area --)+z; h zl-y, 4+x, -++z; it-x, s-y, ++z; g -++y, f-x, --:+z; :+x, p+z. j -++y, +-x, *+z; k:-y,

Ala23

is formed

by both

conformers.

part among the 16 intermoIecular hydrogen bonds in poplar PC crystals (Table 9 of Guss & Freeman, 1983), and vice versa. Only seven residues occurring at the same positions in the sequence of poplar PC as the 13 hydrogen-bonded residues in E. prolifera PC form an intermolecular hydrogen-bond at all (residues 8, 61, 69, 71, 85, 89, 99). In three cases the residues are non-homologous (residues 71, 85, 99) and in all seven cases the contact in poplar PC is with a residue different from that in E. prolifera PC. A close non-bonded contact appears to play a significant role in E. prolifera PC (see below), but no contact of this type is observed in poplar PC. The structural consequences of these differences are as follows. (i) Non-P-strand 5 Two residues where significant polypeptide backbone shifts occur, Asp61-Tyr62, are involved in N(peptide)-H . . . O(carboxy1) bonds from Tyr62 to Asp61 between molecules related by the 4-fold axis at x=y=O. (ii) /l-Strand 6 Three residues in b-strand 6, Thr69. . . Va171Arg72, form five hydrogen-bonds to four residues on the northern double-loop between b-strands 7 and 8, Asp85Pro86-His87 *a. Glu89, in a molecule related to the origin molecule by a rotation of 7112about the 4-fold axis at x = y = 0. The residues involved in these hydrogen bonds comprise the two most extended regions of contact with neighbouring molecules. There are significant polypeptide backbone shifts at Thr69, Va171 and Ser88-Gly89. (iii) The northern loop8 A significant difference between the backbone conformations of E. prolifera and poplar PCS on the loop at the north end of strand 1 can be correlated with a number of differences between the crystal structures. In poplar PC, Asp8 is hydrogen-bonded to Serl7, and the peptide group of Asp9 makes a close contact with the side-chain of Glu18, in a symmetry-related molecule. In E. prolifera PC, Asp8 is hydrogen-bonded to Thr79 in a symmetry-related

C. A. Collyer et al.

630

molecule, there are no other close contacts, and the presence of Gly7 instead of Ala7 is likely to confer greater conformational freedom on the polypeptide backbone. Consistent with the latter hypothesis, the mean temperature parameters B of the backbone atoms at Asp%AspS-GlylO in E. prolifera PC are about 10 A2 higher than in poplar PC. On the northern loop between strands 3 and 4, the side-chain of Pro36 in E. prolifera PC is in van der Waals’ contact with Gln68 in a symmetryrelated molecule. No similar contact occurs in poplar PC. The conformation of the Gln68 side-chain (like that of the Glu68 side-chain in poplar PC) is stabilized by two intramolecular hydrogen-bonds (N Asn64 ’ * . O”l Gln68 and N Ser65. . . 0”’ Gln68), so that any adjustment that may be required to relieve an unfavourable contact is likely to occur in the origin molecule. This accounts nicely for a change in the Pro36 side-chain ring conformation, which is CY-endo in E. prolifera PC but CY-exo in poplar PC. Significant differences between the two PCS continue to the polypeptide backbone at Phe35-Pro36. It is to be noted that, despite the significant change in the backbone at Phe35 and a consequential change of the side-chain, the position of in the orientation the Phe35 phenyl ring remains the same. A CY-endo conformation for Pro36 has previously been observed in poplar Cu’Pc at low pH values (Guss et al., 1986) and in Hg”-substituted poplar PC (Church et aE., 1986). The present work supports the suggestion that this region of the PC molecule is exceptionally flexible (Church et al., 1986). Intermolecular hydrogen-bonding at four of the residues on the northern loop between strands 7 and 8, Asp85-Pro86-His87 . . * Gly89, has been mentioned above. These contacts help to rationalize the significant polypeptide backbone shifts at Ser88-Gly89. (iv) The COO terminus Although an intramolecular hydrogen bond with Ala23 accounm for the large polypeptide backbone shift at the COO-terminal residue Gln99 (see above), a hydrogen-bond between the Gln99 side-chain and Ala23 in a symmetry-related molecule must contribute to the observed conformation.

4. Discussion (a) Amino acid changes and P-sandwich structure In view of the importance that we attach to the comparison between E. prolifera and poplar PCS, we similarities emphasize a number of technical between the two structure analyses. The proteins were crystallized under similar conditions of high ammonium sulphate concentration and at comparable pH values. The structures were refined by similar protocols and to high resolution. The P-sheets and other components of the secondary structure are well defined in both structure analyses, and the thermal parameters for the polypeptide backbone atoms in the P-sheets are relatively low.

A notable result of the present work is that all significant differences between the polypeptide backbone conformations of E. prolifera and poplar PCS can be correlated either with differences between the amino acid sequences or with differences between the intermolecular contacts in the crystal. We have defined the r.m.s. difference between the positions of corresponding backbone atoms in the optimally superposed molecules, 65 A, to be the threshold of significance. In calculating this r.m.s. difference we have included only the contributions of residues that are clearly equivalent in the two molecules (see Results, section (b): above). When all residues are included, the r.m.s. difference is 96 A. These differences are somewhat higher than the value ~0.4 A found by Chothia & Lesk (1986) from a survey of protein structures that have been determined to high resolution in different crystal environments, eonfirming that the crystal structures of E. prolifera and poplar PCS do not merely represent the same molecule in different environments. While some of the observed differences are correlated with differences between intermolecular cont,acts (see Results, section (f), above), others must be attributed to changes and deletions in the amino acid sequence. The most dramatic difference between the two amino acid sequences is the replacement of residues 58-59-60 in poplar PC by a single residue 59* in E. prolifera PC. As a result of this difference, PC lacks a prominent kink in the polyE. prolifera peptide backbone, which in poplar PC causes the side-chains of the acidic residues Glu59-Glu60 to protrude from the surface of the molecule. In plant PCS, residues 59 to 61 and 42 to 45 have been described as an extended “acidic patch”, which ma,y be significant in biological electron-transfer as a means of facilitating electrostatic recognition of a redox partner (Colman et al., 1978; Davis et al., 1980; Simpson et al.; 1986; Moore et al., 1988). The loss of one acidic residue and the absence of a kink imply that the contribution of residues 59 to 61 to the acidic patch in E. prolifera PC is less effective than in poplar PC. Similar observations apply to the structure of PC from another green alga, Scenedesmus obliquus, which has recently been determined from twodimensional n.m.r. data (Moore et al., 1988). (The accuracy of the elegant n.m.r. structure analysis is confirmed by the present work.) Like E. prolifera PC, the protein from S. obliquus has two residues fewer than poplar PC at 58-59-60. The effects on the conformation of the polypeptide backbone in this region are similar. Inspection of the published struct’ural diagrams shows tha,t (at the 6” atom level) His57 in S. obliquus PC is structurally equivalent to Ala57 in E. prolifera PC and Met57 in poplar PC, and that the residues labelled Asp58 and Asp59 by Moore et al., correspond to Asp59* and Asp61 in the present structure. The two algal PC structures therefore bear the same relationship to poplar PC in this region. The disorder that we have noted at Asp59* and Asp61 is also found at the corresponding

Crystal

Structure

residues in S. obliquus PC; a third residue, Tyr68, is reported to be disordered in S. obliquus PC but we find no evidence for disorder at the corresponding residue, Va170, in E. prolifera PC. This difference may be real, in view of the differences between the shapes and volumes of the two types of side-chain. Surveys of the structures of several families of proteins have shown that the differences within a group can be correlated with the packing arrangements of internal side-chains (Lesk & Chothia, 1980, 1982; Chothia & Lesk, 1982, 1985). When amino acid substitutions are associated with changes in the volume of buried residues, the interactions between parts of the secondary structure may be affected. For example, a superposition of poplar PC with a structurally related “blue” Cu-protein, Pseudomonas aeruginosa azurin, shows that differences between the shapes and volumes of buried residues lead to the observed -4 A displacement of P-sheet I with respect to /?-sheet II and the Cu binding site (Chothia & Lesk, 1982). In contrast to this large displacement, the relative positions of p-sheets I and II in E. prolifera and poplar PCS are effectively identical, indicating that the 43 amino acid differences have been accommodated in other ways. We have already discussed the details of this accommodation and restrict our further comments to observations that appear to illustrate two general principles. Lesk & Chothia (1982) have identified a number of mechanisms by which mutations at the interface between the P-sheet domains in immunoglobulins are accommodated. The present work provides further examples of two of these mechanisms: (1) the changes at residues 70 to 72 (Fig. 11) and at residues 82-94-96 (Fig. 14) appear to represent local complementarity in mutations, believed to be a relatively rare phenomenon (Lesk & Chothia, 1982); (2) the side-chain of Met57 in poplar PC (and presumably in other higher-plant PCS where it occurs) is part of the hydrophobic core between the two P-sheets (Fig. ll), but Met57 belongs to strand 5, which is not a P-strand. The effect is closely similar to the lateral insertion of a side-chain from a loop into the region between two P-sheets. According to (Lesk & Chothia (1982), such an insertion can contribute to good packing in cases where the volumes of the side-chains between two b-sheets have been reduced by mutations. (b) Amino There

acid changes and charge distribution

are 18 amino acid differences between PC and poplar PC that involve a charged residue in one or other of the two molecules (Fig. 1). Nearly all these differences occur at exposed residues, i.e. at residues that in poplar PC have an accessible surface area 231 A2 (Guss & Freeman, 1983): there are differences involving charged sidechains at 16 among the 62 exposed residues, but only at one among 32 buried residues (a.s.a. 620 A2) and at one among five intermediate residues, respectively. (The total numbers of sideE. prolifera

631

of Plastocyanin

chain differences in the three groups are 31162, 215 and 9132.) The side-chain differences among the exposed residues comprise: hydrophobic (in poplar PC) wersus basic (in E. prolifera PC), 3; uncharged polar versus acidic, 2; basic versus hydrophobic, 3; basic versus polar uncharged, 2; acidic versus hydrophobic, 2; acidic versus uncharged polar, 3; and acidic deleted, 1. While these differences reflect the well-known fact that exposed residues have a higher probability of being charged and/or variable than do buried residues, they also give a qualitative and poplar PC indication that the E. prokifera molecules have significantly different charge distriThe overall formal charges on the butions. E. prolifera and poplar Cu”Pc molecules at pH 7 are - 6 and - 8, respectively. (c) Possible

importance acidic

of a hydrogen patch

bond at the

In one of the conformers of the disordered residue Glu43, the side-chain carboxyl group is within hydrogen-bonding distance of the carboxyl group of Asp53 (01~43’ 0” . . . Asp53 O”l, 2.8 A; see footnote on p. 000 concerning supplementary material). Evidence that the positions of both side-chains have been determined accurately has been given in Figures 3 and 4. Since the crystals were maintained at PHnl,,, 5.8 during the diffraction measurements, both carboxyl groups should be in the deprotonated (carboxylate, -COO-) state. The short O... 0 contact of 2.8 A renders this unlikely and indicates the presence of a carboxyl . . . carboxylate hydrogen bond, -COOH . . . - OOC-. Interactions of this type are a feature of aspartic proteinases (James & Sielecki, 1983; Pearl & Blundell, 1984; Suguna et al., 1987). They also occur, albeit between adjacent molecules, in other proteins where they are believed to be associated with anomalously high pK, values (Sawyer & James, 1982). The prediction that a pair of hydrogen-bonded carboxylate groups should have an unusually high pK, is made by analogy with the high pK,, of maleic acid (6.1) compared with that of fumaric acid (44); an intramolecular carboxyl-carboxylate hydrogen bond is present in the former but impossible in the latter (cf. Brown et al., 1962). It is consistent with a pK, of -6 that about half of the molecules in E. prolifera PC at pH 5.8 are observed to have a hydrogen bond at Glu43’.. . Asp53. The implications of this hydrogen bond and the possibility of an unusually high pK, will be discussed in a subsequent paper (Collyer & Freeman, unpublished results). This research was supported by grants (X0/15377 and A2860032P from the Australian Research Grants Scheme Postgraduate (to H.C.F.), and by a Commonwealth Research Award (to C.A.C.). Access to the synchrotron radiation source at DESY (Hamburg, F.R.G.), and the assistance of Drs T. P. J. Garrett (Sydney), K. Bartels and H. Bartunik (Hamburg) during data collection, are gratefully acknowledged.

C. A. Collyer et al.

632

Brown, H. C., McDaniel, D. H. 85 Hafliger, 0. (1962). In Determination of Organic Structures by Physical Methods (Nachod, F. C. & Phillips, W. D., eds); vol. 1, pp. 567-662, Academic Press, New York. Brunsehwig, B. S., DeLaive, P. J., English, A. M., Goldberg, M., Gray, H. B., Mayo, S. L. & Sutin, N. (1985). Inorg. Chem. 24; 3743-3149. Chothia, C. & Lesk, A. M. (1982). J. Mol. Biol. 160, 309-323. Chothia, C. & Lesk, A. M. (1985). J. Mol. Biol. 182, 151-158. Chothia, C. & Lcsk, A. M. (1986). EMBO J. 5, 823-826. Church, W. B., Guss, J. M., Potter, J. J. & Freeman, H. C. (1986). J. Biol. Chem. 261, 234-237. Colman, P. M., Freeman, H. C., Guss; J. M., Murata, M., Norris, V. A., Ramshaw, J. A. M. & Venkatappa, M. P. (1978). Nature (London), 272, 319-324. Crowther, R. A. (1972). In The Molecular Replacement Method (Rossmann, M. G., ed.), pp. 174-178, Gordon & Breach, New York. Davis, D. J., Krogmann, D. W. & San Pietro, A. (1980). Plant

Physiol.

in GrystaUography (Diamond, R.; Ramaseshan, S. $ Venkatesan, K.? eds), pp. 13.01-13.23, Indian Academy of Scienoes? Bangalore. James, M. N. G. & Sielecki, A. R. (1983). J. Xoi. B?:oi. 163, 299-361. Jones, T. A, (1978). J. Appl. Crystallogr. 11; 268-272. Kabsch, W. (1978). Acta Crystallogr. sect. A; 34, 827-828. Lesk, A. M. & Chothia, C. (1980). J. Mol. Biol. 136, 225-270. Lesk, A. M. & Chothia, 6. (1982). J. Hoi. BioE. 160; 325-342. Lesk, A. M. & Hardman, K. D. (1982). Science, 216, 539-540. ,Moore, J. M.,,Case, D. A., Chazin, W. J., Gippert, G. P.: Havel., T. F.; Powls, R. & Wright, P. E. (1988). Science, 240, 314-317. North, A. C. T.: Phillips, D. C. & Matthews, F. S. (1968). Computing

References

65, 697-702.

Freeman, H. C. & Guss, J. M. (1972). Acta CrystaZZogr. sect. B, 28; 2090-2096.

Freeman, H. C., Norris, V. A., Ramshaw, J. A. M. & Wright, P. E. (1978). BEBS Letters, 86, 131-135. Freeman, H. C., Garrett, T. P. J., Guss, J. M., Murata, M., Yoshizaki, F., Sugimura, Y. & Shimokoriyama, M. (1983). J. Mol. Biol. 164, 351-353. Garrett, T. P. J. (1987). Ph.D. thesis, University of Sydney. Garrett, T. P. J., Clingeleffer, D. J., Guss: J. M., Rogers, S. J. & Freeman, H. C. (1984). J. Biol. Chem. 259, 2822-2825. Guss, J. M. & Freeman, H. C. (1983). J. Mol. Biol. 169, 521-563. Guss, J. M.; Harrowell, P. R., Murata, M., Norris, V. A. 8: Freeman, H. C. (1986). J. Mol. Biol. 192, 361-387. Haehnel, W. (1986). In Photosynthesis III. Encyclopaedia of Plant Physiology (Staehlin, L. A. & Arntzen, C. J., eds), vol. 19, pp. 547-559, Springer Verlag, BerIin. Hendrickson W. A. & Konnert, J. H. (1980). In

Acta Crystallogr.

Nyborg,

24,

351-359.

J. & Wonacott,

A. J. (1977). In The Rotation (Arndt, U. W. $ Wonacott, A. J., eds), pp. 139-151, North-Holland Publishing Company, Amsterdam. Pearl, L. & Blundell, T. (1984). FEBS Letters, 174, 96-101. Phillips, S. E. V. (1980). J. Mol. Biol. 142, 531-554. Sawyer, L. & James, M. N. 6;. (1982). Nature (London), 295, 79-80. Simpson, R. J., Moritz, R. L.: Nice, E. C., Grego, B., Yoshizaki, F., Sugimura, Y., Freeman, H. C. & Murata, M. (1986). Eur. J. Biochem. 157, 497-506. Steigemann, W. (1974) Ph.D. thesis, Technisehe Univ., Miinchen. Suguna, K., Bott, R. R., Padlan, E. A., Subramanian, E.. Sherriff, S., Cohen, G. H. & Davies, D. R. (1987). J. Mol. Biol. 196, 877-900. Sussman, J. L. (1983). In Methods and Applications in Crystallographic Computing (Hall, S. R. & Ashida, T.: eds), pp. 206-237, Clarendon Press, Oxford. Sussman, J. L., Holbrook, S. R., Church, G. M. & Kim, sect. A, 33, 800-804. S.-H. (1977). Acta Crystallogr. Sykes, A. G. (1985). Chem. Sot. Rev. 14, 283-315. Yoshizaki, F., Sugimura, Y. & Shimokoriyama, M. (1981). J. Biochem. 89, 1533-1539. Method

Edited by W. A. Hendrickson

in

Crystallography