J. Mol. Biol. (1992) 226, 227-238

Crystal Structure Solution and Refinement of the Semisynthetic Cobalt-substituted Bovine E$ythrocyte Superoxide Dismutase at 2-O A Resolution K. Djinovic ‘, A. Coda’, L. Antolini2, G. Pelosi3, A. Desideri4 M. Falcon?, G. Rotilio* and M. Bolognesi’t Studio

1Dipartimento di Genetica e Microbiologia and Centro Interuniversitario Macromolecole Informazionali, Universita di Pavia, Via Abbiategrasso 207, I-27100 Pavia, Italy

‘Dipartimento 31stituto

di Chimica, di Chimica

4Dipartimento

Universitd

Via Giuseppe Campi

Generale ed Inorganica, Universita Scienze, I-43100 Parma, Italy

di Chimica

5Dipartimento

di Modena, Modena, Italy

e Biologica, Messina, Italy

Organica

Universitd

di Biologia, Universitd di Roma “Vor Raimondo, 00173 Roma, Italy

(Received 14 November

di Parma,

Viale delle

di Mess&a,

Vergata”,

1991; accepted 3 March

183, I-41100

I-98166

Via Orazio

1992)

The semisynthetic Co-substituted bovine erythrocyte superoxide dismutasea (SOD) has been crystallized in a new crystalline form and the structure determined at 20 A (1 A = 01 nm) resolution. The crystals belong to space group P2,2,21 with cell constants: a = 51.0, b = 147.6, c = 47.5 A, and contain one dimeric molecule of 32,000 M, per asymmetric unit. The structure has been solved by molecular replacement techniques using the Cu,Zn bovine enzyme as a search model, and refined by molecular dynamics with the crystallographic pseudo-energy term, followed by conventional crystallographic refinement. The R-factor for the 18,964 unique reflections in the resolution range from 10-O to 2-O A is 0,176 for a model comprising 2188 protein atoms and 0200 solvent molecules; the root-mean-square deviation from the ideal bond lengths is 0010 A, and the average atomic temperature factor is 26.5 A2. The dimeric molecule of the enzyme is composed of two identical subunits related by a non-crystallographic S-fold axis. The subunit has as its structural scaffolding the conventional SOD-flattened antiparallel eight-stranded P-barrel, with three external loops. The co-ordination geometry of the metal center in the active site is fairly well preserved when compared with the native Cu,Zn bovine enzyme, Co2+ is in tetrahedral co-ordination, while the Cu2+ ligands show an uneven distortion from the square planar geometry. The least-squares superposition of the metals ligands and the catalytically important ArgJ41 of the native and Co-substituted enzyme yields a root-mean-square value of 0401 A, the largest deviation occurring at the Co2+ ligand As~81.~ An additional copper ligand, compatible with a water molecule, is observed at 2.38 A from Cu2+ in the active-site channel, at the supposed binding site of the 0; anion substrate. Several ordered water molecules have been observed on the protein surface and in the active-site channel; their structural locations coincide remarkably with those of related water molecules found in the crystal structure of the phylogenetically distant superoxide dismutase from yeast. Keywords: superoxide dismutase; copper; zinc; cobalt; X-ray

t Author to whom correspondence should be addressed. 227

0022-2836/92/130227-12

$03.00/O

0 1992 Academic Press Limited

228

K. Djinocic

1. Introduction Among the metalloproteins superoxide dismutases (SOD?) are almost unique enzymes, since they may contain either copper and zinc, or manganese or iron at the active site center. These proteins are considered to be widespread detoxifying enzymes, whose function is to disproportionate two superoxide radical anions into one molecule of oxygen and one of hydrogen peroxide (Fielden et al., 1974). While iron and manganese SOD can be traced to a common evolutionary origin, sharing well-defined molecular architecture, based on a common polypeptide chain of 20,000 M,, the Cu? Zn SOD is based on a polypeptide chain of 16,000 M,. thus totally unrelated to the other SODS (Stallings et al.. 1984). Cu,Zn SOD is active in solution as a homodimer, each chain hosting an independent active site containing the Cu2+,Zn2+ pair. The active-site crevice hosts positively charged residues that participate in the catalytic cycle of the enzyme, providing an electrostatic field that drives the substrate to the redox copper-zinc center for dismutation (Desideri et at., 1991). From a crystallographic viewpoint SODS from three different sources have been subjects of atomic resolution analyses. The three-dimensional structure of the enzymes from bovine erythrocytes (Tainer et al., 1982), spinach chloroplasts (Kitagawa et al.; 1991) and yeast (Djinovic ,et al., 1991) have been reported at 2.0, 2.0 and 2.5 A (1 A = @l nm) resolution, respec:tively; providing independent views of the same catalytic unit from widely different evolutionary sources. These investigations have shown that the enzyme tertiary structure is strongly conserved, and based on a Greek-key /?-barrel, topologically composed of two P-sheets of four b-strands each (Getzoff et al., 1989). Three of the major loops connecting P-strands in the barrel define a deep surface cleft, which accommodates the active-site ion pair. The co-ordination sphere of the two metal ions is strongly conserved in all Cu,Zn SODS and consists of four histidine residues in distorted square planar co-ordination around the Cu2 +, and of three histidines and one aspartate in distorted tetrahedral co-ordination around the Zn2+. The direct copper ligands are three NE2 atoms from residues His46, His61, and Hisll8, and one ND1 atom donated from the His44 side-chain; His61 is a bridging residue between Cu2+ and Zn2+, which is thus directly co-ordinated to ND1 atoms of His61, His69 and His78 and to ODl of Asp81. A fifth ligaad has been observed in the active-site channel at 2 A from the Cu2+, while the imidazole ring of His61 defines a plane containing both metal ions. t Abbreviations used: SOD. superoxide dismutase; n.m.r., nuclear magnetic resonance; b-SOD, bovine erythrocyte SOD; PEG, polyethylene glycol; r.m.s., root-mean-square; EMX, restrained crystallographic refinement with energy minimization; MDX, molecular dynamics refinement with crystallographic pseudoenergy; y-SOD, yeast SOD.

et al. Since the initial investigation on this enzyme. a major problem has been that the Zn2 + ion is optically and magnetically silent. In this respect a derivativr has been prepared in which Co(H) has selectively replaced the native Zn(I1) ion (Calabrese ef al.. 1972). The introduction of such chromophoric probe has provided the opportunity to perform several spectroscopic studies in order to elucidate the str‘uczture-function relationship of the metal &ster. First of all, the Cu,Co derivative was found to be elrc%rolr paramagnetic resonance silent (Rotilio d al.. 1974), due to the magnetic coupling occurring between thfa two metals (Morgenst,ern-Radaran d al.. 1986) and this property was used to propose close proximit? between the two metals, before X-ray data were available. The presence of such a coupling allowt~d detection of the proton nuclear magnet’ic resonanct~ (n.m.r.) signals of the amino acid side-chains (Y)ordinated to the metal (aluster and in particular of the copper-bound histidines in cases in which t htl first co-ordination sphere of the copper is scblect ivel! perturbed (Bertini it al.. 1985). These propert,& have been recently used to investigate the p~~rturbxtion of the n.m .r. spectrum caused FJ: t,hr> additiotl of negatively charged anions (Kertlnl et ~1.. 1985: Paci e1 al.. 1988), which are supposed to by good substrate analogues. Furthermore. since the ('u.( "o derivative has a catalytic rficiency comparabk to that of t,he native enzyme. the chromophorr proptar ties of the cobalt site enabled the involvement of t,htb zinc in t,he ca,talytic mechanism to bc monitored. (McAdam et al., 1977). In view of this csonsideration. it is fundamental to assess the structural identity of t)he zinc and cobalt, enzymes by direct, analysis. with particular reference to the potentialities of tl. m .r. approaches, such as t,wo-dimensional n.m.r. and t,hr nuclear Overhauser effect,, which require unatnbguous assignment, of resonances on the basis of proper crystallographic co-ordinates. We have crystallized Cu,Co bovine SOD$ in a new crystalline form suitable for high-resolution crystal lographic investigation. Tn this paper we report the experimental details pertaining t,o solution of t)h (w/v) poly-

1 Residues of Cu,Co b-SOD have been numbered according to the scheme of h-SOT> (Tainer d nl.. 1982). The 2 Cu.Co b-SOD chains present in t,hr asymmetricunit have been identified by the letters A and R: their sequence numbers range from 1 to 151. 201 to 351. respectively.

Structure

of Cu,Co Superoxide

ethylene glycol (PEG, average M, = 4000) and @1 M-NaCl were equilibrated against a reservoir solution containing 16:/, (w/v) PEG at the same pH, at room temperature. The crystals grew as elongated prisms of about 1.2 mm x03 mm x@3 mm in 1 to 2 months. The preliminary characterization of the crystal form obtained under these conditions was performed by photographic methods from inspection of precession photographs. The crystals were found to be in the orthorhombic space group P2,212, with cell constants a = 51.0, b = 147.6, c = 475 A; assuming that the asymmetric unit contains one SOD dimer, the solvent content of this crystalline 1968). Under the same form is 56% (Matthews, experimental conditions isomorphous crystals of the native Cu,Zn b-SOD could be grown. Two crystals, mounted in a thin-walled glass capillary, were used for data collection using the synchrotron radiation source at the EMBL X11 beamline at the DORIS storage ring, DESY, Hamburg, employing a locally developed image plate system as detector and a wavelength of 1.009 A. The DORIS ring was operating in a single-bunch mode at 5.3 GeV and 40 to 100 mA, with approximately 3h between injections. The crystal was mounted on an Arndt-Wonacott oscillation camera (Nyborg & Wonacott. 1977) with the [ 1001 direction parallel to the rotation axis. and was constantly kept at 4°C by a stream of dry cooled air. The crystal-to-det)ector distances were 261 and 200 mm for low- and high-resolution data collection, respectively, and exposure times ranged between 7 and 1064 s: accordingly, the high-resolution data set (up to 2.0 A spacing). on the first crystal, were collected using 3”, 2” and 1.5” rotation angular ranges. A second crystal was used t.o record the blind region data as well as part of the high-resolution reflections using angular rotation ranges of 075”. 1.0”. 1.5” and 2.0”. Refinement of orientation and integration of the intensities were performed using the MOSFLM suite of programs (Leslie et al., 1986) modified for processing of the image plate data. Merging of the observed intensities into a unique data set of reflections was carried out using ROTAVATA/AGROVATA programs from the CCP4 suite. supplied by the SERC Daresbury Laboratory (U.K.). The intensities were converted to structure factor amplit,udes using the program TRUNCATE (French & Wilson. 1978). A total of 116,401 measurements were merged to yield 18,964 unique reflections (7>.2?; of the t)heoret,ical reflections in the 10.0 to 2.0 A resolution range). The merging R, factor (R, = ElIi- l/ClIJ, where Ii is the intensity of an observation, is the mean value of t,he reflection. and the summations are over all reflections) was 6-So/;. (h) Moleculnr

replacement

The starting model for the molecular replacement investigat.ion was the .‘orange-yellow” dimer of t,he bovine Cu,Zn superoxide dismutase. partially refined at, 2.0 A resolution, to an R-factor of 2%5 o/0 (Tainer et al., 1982). and deposited with the Protein Data Bank (Bernst.ein et nl.. 1977) as dataset 2SOD. The MERLOT (Fitzgerald. 1988) parkage of programs, which includes the fast-rotation function of Crowther (1972), the rotation function of Lattman & Love (1970) and a translation of Crowt,her & Blow (1967), was used function throughout. The searcxh model was placed in an-orthogqnal reference cell with dimensions of 100 A x 100 A x 100 A for calculation of the structure factors used to generate harmonic eoeficients for the fast rotation function. The structure factors were calculated up to 3.0 A spacings with an

229

Dismutase

Table 1. fast rotation function (l), and Lattman-Love cross-rotation

A. Results of the Crowther

after origin

removal, function

(2) in Euler angles

a(“) (1)

175.0 177.0

(2)

R. Results

of Crowther-Blow after self-vector

PC”)

IT)

84.0 85.0

32.0 32.0

translation

function,

removal

xt

Yt

zt

r.m.s.J

50 94 44

29

27 77 50

9.71 8.44 8.63

50 79

t Fractions of’ the unit cell. $ r.m.s. is the root-mean-square

deviation

from t,he mean.

overall temperature factor (B) at @O A” in order to sharpen the peaks. The harmonic coefficients were calculated with upper and lower limit,s of 35 and 8.0 8, respectively; the integration radius being set to 2@5 A. The Eulerian angle increments between successive calculations of the rotation function were 2.5” for CI, 1’ for p and 5” for y. The origin-removed cross-rotation-function analysis gave a rotation peak with a r.m.s. value of 654 (r.m.s. is t,he root-mean-square deviation from the mean). As the success in determining the correct t’ranslation of the oriented model is highly dependent on how accurate the rotation angles are, the rotation function of Lattman & Love (1970) was applied in order to “refine” the determined rotation angles on a finer grid (11 steps of 1” around each angle) using the diffraction data between 100 and 40 A. In order to position the dimers with respect to the crystallographic symmetry elements. the refined values of the rotation search angles in Eulerian space were used. The correctness of the solution was assessed on the basis of an internally consistent set of vectors from the 3 Harker sections in the translation function. and of packing considerations. Calculations of a translation function map on a 1.0 A grid, for the data between l@O and 4 8, for the 3 Harker sections (X = l/2: r = l/2: 2 = l/2) gave a self-consistent set of vectors. The initial and with the refined values of the eulerian angles. along results of the translation search calculations are reported in Table 1. The solution was subsequently submitted to R-value minimization dWard et al., 1975) using the data between 10.0 and 40 A, with a decrease in the R-factor value from 046 to 0.43. The initial R-value for the correctly oriented and positioned model was 0.385. for the intensity data between 7.0 and 40 A resolution. A rigid body refinement process, allowing 3 positional and 3 rotational paramet’ers to vary independently for each of the 2 chains in the asymmetric unit, was st.arted in the 7-O to 4.0 A resolution range with an overall B factor of 2@0 A’, using the program TNT (Tronrud et al., 1987). The rigid body refinement did not yield any significant decrease in the conventional R-fartor after 3 cycles. This can possibly be explained by the very tight contact between the 2 subunits forming a functional dimer. which cannot be affected by t.he different crystalline forms. and. on the other hand. by a fairly precise molecular replacement solution. The refinement of the model proceeded with 8 cycles of conventional crystallo-

230

et al

K. Djinovic

Table 2

Table 3

Summary of the rigid body and molecular dynamics rejknement with crystallographic restraints using TNT (Tronrud et al., 1987) and GROMOS (van Gunsteren & Berendsen, 1987) program suites Resolution (4

&it

7.W.O 7.W3.5 60-30

0.385 0385 0.393 @392 0.411 0.247

6.0-2.3

0.338

0.265

@60 2.15

6G2.1 6G2.0

0.287 0.280

0.260 0.270

3.75 490

6C2.0

0.253

0.243

7.6

In (PS)

3 8

Comment Rigid body refinement Conventional refinement EMX/MDX EMX/MDX, followed by R-factor ref. EMX/MDX EMX/MDX, followed by B-factor ref. EMX/MDX

graphic refinement of the co-ordinates, with stereochemical restraints, for the intensity data in the resolution range from 6.0 to 35 A. As shown in Table 2, this procedure also did not bring about rapid decrease in the R-factor. Therefore, after inspection of the electron density map, the refinement was continued using molecular dynamics simulation with a crystallographic pseudo-energy term, because of its larger radius of convergence, owing to the presence of the kinetic energy that allows the conformational barriers to be overcome. (c) Molecular dynamics and conventional rejnement The GROMOS suite of programs (van Gunsteren & Berendsen, 1987) was used for molecular dynamics refinement of the starting positioned and oriented model, which was not manually corrected for bad stereochemistry and/ or contacts in any way. Pseudo-energy minimization was first’performed (EMX, a combination of energy minimization with X-ray restraints) to remove the initial strain in the structure due to bad contacts and poor chain stereochemistry. Molecular dynamics refinement including crystallographic potential (MDX) was typically performed in series of 605 to 0.6 ps per cycle, consisting of steps of At = 2 fs. The balance between the crystallographic Rfactor and the energy term was controlled applying a weight cr, to the X-ray potential term. EMX refinement in the resolution range from 6.0 to 30 A (100 cycles) was followed by the MDX re$nement at T = 300 K for the data between 6.0 and 30 A, and an isothermal algorithm (Berendsen et al., 1984) was applied to keep the overall temperature constant, using the temperature relaxation time z = 61 ps. No restraints on bond lengths and on the non-crystallographic P-fold axis relating the 2 subunits in the functional dimer of the asymmetric unit were used during EMX and MDX refinement procedures, R-factors of 290 A’ being assigned to each atom. The ox value for this stage of refinement was chosen to be approximately 60% of the observed ur value {CQ= [ < (IF,]--]F,])*> I*}. and was gently increased during the refinement procedure (low value of 8, means high weight for the X-ray energy terms). Subsequently the resolution was increased to 2.3 A spacing, limiting the low-resolution data to 60 A in order to avoid the contribution of the solvent structure at lower-angle data. After 2.75 ps simulation time (R = 0.276) the model was submitted to EMX refinement, followed by the refinement of isotropic temperature factors using the TNT suite of programs (Tronrud et aE., 1987). The refinement of the model with individual thermal parameters proceeded with the data in the 60 to

of

stereochemical

Stage

No. cyc./ &in

Summary

the crystallographic rejinement restraints using the TNT program (Tronrud et al., 1987)

Resolution (4

4”i,(% ) &“(% 1

A

6.0-2.0

Ii C

10.0 2.0 I@&24

27.0 28-2 244

D

10.2~2~0

2 14

No. cycles

25.1

68

21.8

28

19.2 17.6

24 26

with’ suite

r.m.s. coord. shift 0009 0058 Om7 0009

2.1 A resolution range; after an additional 1 ps of simulationO time the crystallographic data were increased to 2.0 A resolution. and, after a total of 4.9 ps. the refinement of the thermal parameters was Forformed again for the intensity data between 60 and 2.0 A. The model with updated individual isotropic B-factors was submitted to an additional 2.7 ps of MDX simulation, followed by EMX refinement in order to release t’he system’s kinetic energy. This procedure led to an R-factor of6243 for the crystallographic data between 6.0 and 2.0 A. lip to this stage of the refinement no manual intervention was applied to the Cu,Co b-SOD model. The course of the rigid body and molecular dynamics refinement is presented in Table 2. The refinement of this molecular model was then con tinued by conventional restrained crystallographic refinement of the co-ordinates and temperature factors using the TNT suite of programs (Tronrud et al., 1987). Electron density maps with coefficients (2]1p,]- ]E’,]) and (IF,] - IF,]) and calculated phases were computed at regular intervals. whenever convergence of the refinement to a local minimum was detected by a limited decrease in the Kfactor and by small r.m.s. shifts in the stereochemical parameters, indicating the need for manual intervention. Table 3 gives a summary of the crystallographic refinement in terms of R-values and r.m.s. co-ordinates shifts. Four stages of restrained crystallographic refinement of co-ordinates and temperature factors, alternated wit,h model-building sessions on a graphics station, using the FRODO modelling program (Jones, 1978)> led to a final R-value of 0176 fo: the 18,876 observed structure factors in the 160 to 20 A resolution range. The metal-ligand bond-length restraints applied were taken from Orpen et al. (1989), as the average value of the respective bonds observed in low M, co-ordination compounds. After step B of the conventional crystallographic refinement the contribution of the ordered water molecules located in the difference electron density maps was included in t,he model. The solvent molecules were critically re-examined during each subsequent manual refitting of the model on the basis of hydrogen bonding criteria and of their temperature factors. Systematic inspections of difference electron density maps led to inclusion of 200 ordered solvent molecules in the present, model. During t,he last stage of the least-squares refinement the occupancies of the water molecules were refined as well. Table 4 presents the parameters of the final cycle of the refinement for t,he Cu,Co b-SOD model. Data processing and reduction were performed on the Ethernet-based MicroVAX cluster at EMBL, Hamburg (FRG), molecular replacement and molecular dynamics calculations on a Convex C220 8 computer at CILEA. Milan0 (Italy), while conventional refinement, and molecular graphics on VAX8530, MicroVax 3800 systems and on a PS390 Evans-Sutherland graphics computer

Structure of Cu,Co Superoxide Dismutase

Table 4 Refinement

parameters

and

rejbement

231

Table 5 results for

the final

cycle

Chain A

Parameter X0. of cycles R (%) No. of reflections (res. range A) No. of protein atoms No. of solvent atoms r.m.s. co-ord. shift, in the final cycle (A) r.m.s. deviatiqns from ideal values: Bond length (8) Bond angle (“) Torsion angle (“)Trigonal plane (A) Planar groups (A) Non-bonded contacts (A)

Residues with poorly dejked side-chains

26 17.6 18,876 (loo-20) 2188 (+4 metal ions) 200 0.009 0.010 1.748 2676 0008 0014 0.066

were carried out at the University of Pavia. The Figures were prepared using the molecular graphics package WHATIF (Vriend, 1990). The structure factors and the co-ordinate data sets of Cu,Co b-SOD have been deposited with the Protein Data Bank (Bernstein et al., 1977), from which copies may be obtained.

3. Results and Discussion (a) Quality of the model The quality of the refined Cu,Co b-SOD model has been assessed on the basis of the observed structure factors, using the (TV method proposed by Read (1986). From the analysis of the data repprted in Figure 1, a r.m.s. co-ordinate error of O-20 A can be derived. The overall stereochemistry of the model is close to ideality; a summary of the main stereochemical parameters describing the structure, as calculated by the TNT program package (Tronrud et al., 1987) is presented in Table 4. The electron density is well defined throughout the two polypeptide chains contained in the asymmetric unit, except for some of the larger side-chains (see Table 5), on the surface of the model, that show weak electron density and cannot be fitted properly to it. The calculation of the so-called “real space R-factor” (Jones et al., 1991) on a per-residue basis is given in Figure 2(a). Correspondingly, a profile

Lys3 Lyz9 Gln15 Lys23 Lys73 Lys89 Asn90 GlulO7 TyrlOS Lysl20 LyslBl

Chain B CE, NZ CD, CE, NZ CD, OEl, NE2 CE, NZ CE, NZ CG, CD, CE, NZ ND2 CG, CD, OEl, OE2 CD2, CE2 NZ CE, NZ

Lys3 Asp1 1 Lys23 Am51 Lys73 Lys89 Glu107 Lys151

CD, oD2 CE, liD2 NZ CD, CG, CE,

CE NZ

CE, NZ CD. OEl, OE2 NZ

of the refined main chain B-factors per residue is shown in Figure 2(b). The average temperature factor for the whole model, including 200 water molecules, is 265 8”. The average temperature factor of 2127 protein non-hydrogen atoms considered in the structure factor calculations is 24.4 A’. The inspection of the torsional angles +,t/~ defining the polypeptide chain conformation does not highlight evident deviations from the allowed regions of the Ramachandran plot (Fig. 3), with the exception of residues Ser66, Leu124 and Asn137, in both chains. All other residues displaying a positive value for the 4 conformational angle (see Fig. 3) are glycine. It is remarkable that, for Leu124, the same left-handed helical conformation is found in b-SOD and in y-SOD (Djinovic et al., 1991, 1992). The general model derived for Cu,Co b-SOD, presented in Figure 4, has been compared with the starting “orange-yellow” dimer of the Cu,Zn b-SOD

d 0.1

0 20406080,0~20,4~60 Restdue

200 240 280 320 3 220 260 xx) 340 number

,.,

g g

60 I50 40 30 20 IO 0

I

204060800~2014~60

I

I

200 240 280 320 360 220 260 300 340

Residue number (b) (sin B/M2

Figure 1. Plot of lna, method @20 A.

x IO3 (II-‘)

versus ((sin0)/1)’ based on the of Read (1986), giving a mean co-ordinate error of

Figure 2. (a) Plot sequence average sequence

numbers B-factors numbers

of the real-space R-factor z)erfor the Cu,Co b-SOD model. (b) Plot of for the main chain atoms ver.sus for Cu,Co b-SOD model.

232

K. Djinovic -180

180

1 18C

+

+

-18(

.--

+ -h. +=?A

Figure 3. The $-+ plot for the refined model of Cu,Co-SOD. The inner lines define areas of fully allowed conformations with ‘t (NMYC) = 110” for the non-glycine residues. The outer lines show the regions obtained by relaxing the van der Waals’ contact constants as well as by allowing T to increase to 115” (Ramakrishnan & Ramachandran, 1965). All residues with positive 4 values are glycine. with the exception of Ser66. Leu124 and Asn137 (in both chains), which adolIt conformations close to left-handed

helix.

used for the molecular replacement searches. After least-squares superposition of the two Ca backbones, a r.m.x. deviation of 0%17 A is observed. The P backbones of the two overlayed st)ructures arc> displayed in Figure 5. As can be easily recognized from inspection of the two chains in the Figure, the largest’ deviations occur in equivalent regions of the two independent, A and B (*hains, building up the most exposed polypeptide loop of each subunit.

The Cu,Zn SODS are found in solution as fumtional dimers, whose stabilit,y is established by means of polar and apolar interactions across the dimer interface. A number of evolutionarily invariant or highly conserved residues (Get,zoff et al.. to very specific interact,ions 1989) contribute allowing the formation of a dimeric species with an internal Y-fold axis relating t,he two polypeptide chains. All four main-chain-to-main-chain hydrogen

Figure 4. Stereoscopic view of the pseudo-crystallographic dyad axis.

C” atom

tracing

et al. bonds between the two subunits observed in the Cu,Zn b-SOD dimer have been preserved in the crystal form of the Co-substituted enzyme. Table 6 presents a comparison of hydrogen bond lengths between (‘u.C’o and Cu,Zn h-SOD. and y-601). in which 11~149 is substituted by Leu. It can readily br ohserved that thca inter subunit interactions art’ strongly preserved in the two different, cryst,alline forms of the b-SOl)s as well as a,mong the phylogenetically distant yeast and hovine enzymes. Tk very specific interaction between the two polypeptide chains can also be appreciated from the formation of the heterodimrric forms. that, have been obtained in r*ifro by subunit excshange between (‘u,Zn SOl)s isolated from difierent organisms. in particular upon hrhat t,reatment (Tegelstrom. 1975). Moreover. a natural hetrrodimeric form has been found in t)issurs of t,hr amphibian Xrnopus lawis. which expresses two significantly different gSr,nesfor (!u,Zn SOI) ((lap0 rt 01.. 1990).

The crystal contacts that occur betwet~n symmrtrically equivalent molecules in the 1’2,2,2, c*rystalline form of t htx (‘LI.(‘o I)-SOI) arc’ reported in TabIt 7. They involve tnaitl-~hairl-to-main-(~~lain. maitlside-chain-t o-sidr-(*haill chain-to-sitle-chain. and hydrogen bonds. It can be seen that the inter molecular caontacts between different asytnmt~t tic. units units:

are formed Mween the same crizyrrw namely, between ;\ and A. or I< and

sub

l< polj

peptide czhains, except for the polar int~eract~ionh between thr C-terminal I,ys151 NZ atom of the IZ monomer, and the Glu75 OE2 atom of the A sub unit. Thcl same residue (Lys151) of t.hr A chain hydrogen bonds lo the solvent molecule OH 1 IciO (3.34A). whicah it) turn fhrms a c*ontac+ to (iIn NE:! (3.13 ioz) of the Es subunit, thus stabilizing the n]onOmer-to-moIlonier int,eractions iti t hr itS>'tW met,ric unit, The polar interaction betwren .A sut)the 1001 / units forms thr connections along the c~trntac~ting in t,hrl ~t.~st~aI, while direction surfaces solvent

ktwtac,n

formation

of

13 subunits

d&nr

t he

(.ryht al

channels along the [ lOO] rlir.rc%ion ilr t ht. crystal. The residues in the t,u’o polypept’idr c+hains that par%icipatcB 10 the intermolcc*ular c*ontac+s belong to the solvt~llt-exposed D-1 urn or loott regions. TLcsiduc* .Asp25 of thrh 23 to 26 b--t urr1. c.onnrc%ing %h antI 3c /I-strands. partic,if)atrb in t III3

of the (Ju,(‘o

main-cahain

b-SOI)

dimrr

to-main-cahain

hydrol-((Bn

as seen ~)~~r~~~~l(tic~tllar~to thf~

Structure

of Cu,Co Superoxide

Dismutase

233

Figure 5. Stereoscopic view of the C” atoms of Cu.Co b-SOD A subunit (continuous line) superimposed on the “orange” b-SOD subunit (broken line). Labels have been placed on every 15th c” atom.

bonds between symmetry-related molecules in both subunits of the functional dimer; Asp25 N of the A chain bonds to Ser66 0 in the Zn ligand region of the 6,5 loop, while in the case of the B subunit the contacts are established between the carbonyl oxygen of Asp25 and the polypeptide nitrogen of Glu130 (a-helix region). On the other hand, the sidechains of Asp25 of bot’h subunits form side-chain-toside-chain hydrogen bonds to Asn63 and Lys68, respectively. Another main chain-to-main chain hydrogen bond between symmetry-related B subunits, stabilizing the crystal structure of Cu,Co b-SOD, is formed between the Pro100 0 (of the 4,7 loop) and the Glu131 N atom of the a-helical region. Residue Glu130 of the A subunit is involved in the intermolecular contacts to the main chain atoms of the 4 monomer Tyr108 N, (2.49 A), Ser105 0 (3.11 A) and Glu107 N (2.89 A) mediated by water molecule OH 1139, as shown in Figure 6, and also forms a direct contact to the0 nitrogen atom of symmetry-related Glu107 (3.12 A).

joining the two charged residues on the active-site rim is not preserved in the Cu,Co SOD structure. Glu130 of the A subunit is involved in the intermolecular crystal contacts, bridged by means of a solvent molecule, as discussed above, to the symmetry-related A subunit, as well as in a weak polar interaction with Lys134 of the same subunit. The carboxylate group of Glu130 in the B polypeptide chain does not participate in intermolecular or intramolecular hydrogen-bon+ng interactions. An alternative salt bridge of 3.33 A between Lys134 and Glu131 is formed in the B subunit, neutralizing the opposite charges of the two side-chains.

Table 7 Intermolecular

as observed

in thu Cu,Co

b-SOD IIistance

bridges

(d) Salt

contacts

There are three salt bridges identified within the Cu,Zn SOD subunit’; two of them are preserved in the Cu.Co enzyme crystalline form: Arg77. . .Asp99 linking the fi-barrel t’o loop 6,5, which contains all four Zn ligands. and Asp74. .Arg126, connecting 6.5 and 7,X loops, which form the active-site channel. The salt link bet’ween Glul30 and Lysl34.

Residue

Atom

(Chain

Residue

Atom

(‘ha.in

Asp25 ASJl2B ASP25 Thr56 Glu7.i Glu107 4sp1.5 Asp25 LNAf, Pro100 LeulOl SW103

s ODl OD2 OGl OE2 s 0 ODP 0 0 0 (I(:

A A A I4 A A 1% u 1% I3 I3 13

Ser66 Ser66 Asn63 As1190 Lys1.51 Glul30 Glu130 Lys68 Lys 120 Glu131 Am129 GlylZX

0 0 SD2 ODl NZ OEl s NZ NZ s ND:! 0

;I A A A I3 A 13 1% lj 13 13 13

Table 6 backbone hydrogen bonds as observed in the Cu,Co b-SOD orange-yellow dimer of the Cu,Zn b-SOD (B), and CD dimer of the y-SOD enzyme (C)

Subunit-to-subunit

(il),

Distance (A) R&due

Atom

Chain

Residue

Atom

(:lu49 (:I~112 lIPl49 Ile149

s 0 0 s

A A A $

Ilr149 Ilel49 Glg49 Glgl12

0 N N 0

(‘hain I3 u 8 l3

A

u

(‘

2.76 1-92 2.68 286

2.94 3.19 3.35 3.06

2.7 I 3.64 2.60 I%2

4)

3.39 2.93 3.13 3.21 3.08 3.12 3.27 3.17 3.45 2.97 3.18 2.96

234

K. Djinovic

et al.

J

Figure 6. Stereoscopic view of the crystal the main chain atoms of a symmetry-related

J contact between the side-chain of the Glu130 residue of the A subunit A monomer, mediated by the OH 1139 solvent molecule.

In the Cu,Zn b-SOD crystalline form two additional salt links occur between subunits across the dimer interface involving Argll3. .Glul07, and Lysl51. . .Asp50 ion pairs. In the Cu,Co b-SOD structure the Argll3 NH2 atom is connected to Gln47 OEl atom of the same subunit by means of the hydrogen bond interaction mediated by water molecules OH 1063 in the A, and OH 1054 in the B subunit, respectively. Gln47 NE2 ato? of the B chain hydrogen bonds to OH 1160 (3.12 A), which in turn is involved in a polar interact@ with Cterminal Lysl51 of the A subunit (3*35A), compensating in this way for the loss of the Lysl51. . .Asp50 interaction (see Fig. 7). On the other hand, Lys151 of the B subunit participates in a polar interaction with Glu75 of t’he symmetryrelated A subunit (see Table 7), forming in this way an ion pair. (e) Active

Figure 7. Stereoscopic Argll3

b-SOD, there were no significant differences considering the positional error associated with the atomic co-ordinates. On the other hand, from consideration of the published crystallographic data on Co2 + co-ordination model compounds, average Co2+-N and Co’+-0 distances of 2.01 and 2.08 A (Orpen et aE., 1989), respectively. are found. Inspection of the co-ordination geometry of the two metal ions indicates that Co’+ , as expected, is in tetrahedral co-ordination, whereas the four protein ligands of copper form a distorted square planar coordination sphere. An additional ligand for copper whose density is coqpatible with a water molecule is observed at 2,38 A from Cu’+ in the active-site Table 8 Comparison of the bond distances (A) sites of Cu,Co b-SOD (A) and Cu,Zn

site

The overall three-dimensional structure of the co-ordinating residues is active-site metal thoroughly conserved despite the chemical substitution of Zn2+ by Co’+. The co-ordination geometry of the metal center in the active site is described by the co-ordination bond distances listed in Table 8. From a comparison of the same bond distances as observed in Cu,Zn

and

A

Ii

in the active b-SOD (R) A

13

CU’+~NDI Cu’+-NE2 Cu’+-NE2 CU’+~NE~

His44 His46 His61 His118

2.09 2.29 2.16 2.13

2.01 2.11 2.21 2.09

(:o’+--ND1 &*+-ND1 / Co’+-ND1 CoZ+-ODl

His61 His69 His78 Asp81

2.1 1 2.01 2.02 1.99

209 2.14 2.04 I.91

CU’+~NDI (‘II*+-YE2 / Cu’+-NE2 Cu*+-NE2

His44 His46 His61 His118

202 “~18 I.99 2.14

1.96 2.1% I.98 1.99

Co’+-ND1 (‘o* +-ND1 Co*+--ND1 Co2’--0DI

His61 His69 His78 Asp81

I.99 2.32 1.97 I.96

2.07 “a-3 I.16 2.07

view of the Glu47 residue of the B subunit interacting across the OH 1054 molecule of the same subunit, and across OH 1160 with the Lys151 C-terminal residue of the A subunit.

with

Structure of Cu,Co Superoxide Dismutase

Figure 8. Comparison of the conformations of the residues important for the electrostatic the active site region (Cu,Co b-SOD, continuous line; Cu,Zn b-SOD, broken line).

channel leading to the protein surface. This is the supposed substrate binding site that is occupied by 0; during catalysis. The catalytic cycle of Cu,Zn SOD is under the direct control of a number of charged residues that build up the electrostatic field in the active site, and participate in successfully guiding and orienting the superoxide anion (Desideri et al., 1989; Getzoff et al., 1989; Sines et al., 1990; Desideri et al., 1991). It is therefore relevant to compare the conformations of such residues in the active sites of the two enzymes, which differ by the Zn*++Co*+ “point substitution”. Figure 8 shows the residues that are believed to be relevant for the electrostatics of the active site in the two enzymes, after least-squares superposition of the two C” chains. It can be readily appreciated that residue Argl41 maintains essentially an identical conformation in the two proteins, in relation to its substrate binding and orienting function. On the other hand, the other charged residues examined do not show a comparable conservation of their orientation. Residue Lys134 is also believed to participate in the electrostatic guidance of the substrate in the active site, although its contribution is probably mediated by the presence of the Glu130 negative charge. In both chains of Cu,Co b-SOD the electron density of the side-chain of Lys134 is well defined, and indicates that the Lys134 NZ atom of the A tubunit interacts with the Glu130 OE2 atom (372 A); in the-B subunit an alternative polar interaction of 3.33 A with the sidechain of Glu131 is observed, which however shows a slightly different conformation from that observed in the Cu,Zn protein. In the context of the electrostatic control of the b-SOD activity, the role of residues Glu119 and Lysl20 has recently been questioned, on the basis of the computer-simulated enzyme/substrate encounter through the use of Brownian dynamics, and considering the observed substitution of both these residues with neutral amino acids in the primary structure of SODS displaying almost identical catalytic parameters

235

guidance of the substrate to

(Sines et al., 1999). Also the conformation of residue Asp122, playing an important role in the structure of the active site cavity, because it bridges through hydrogen bonds His44 (co-ordinated to copper) and His69 (co-ordinated to zinc), is highly preserved in the two proteins. In the active-site region an almost ideal network of hydrogen-bonded water molecules stretches from the copper ion’s fifth ligand along the molecular space between the disulfide region of the 6,5 loop and the a-helix and the 138 to 141 segment of the polypeptide chain. Comparing the structure of the solvent in this region with that observed in y-SOD, striking similarities can be noticed despite only 55% sequence identity between the two phylogenetically distant proteins, implying a structural and functional role for the solvent itself. Indeed, a number of ordered water molecules are found in almost identical positions, participating in the scaffolding of the protein structure and in the orientation of the side-chains, by means of a complex hydrogen-bonding scheme. In particular, the positions of OH 1028, 1041, 1045, 1046, 1047, 1070, 1126, 1131 are strongly conserved in the two proteins, Among these, the four solvent peaks numbered OH 1046, OH 1126, OH 1047 and OH 1131, once overlayed on the y-SOD structure (on the basis of the proteins C” backbones only) are found to fall within the r.m.s. distance (as calculated from the comparison of the two enzymes) from equivalent solvent molecules in y-SOD. Inspection of the two protein structures shows that this coincidence is not fortuitous and reflects a structural role for these water molecules with respect to the activesite residues. Water molecules OH 1047 and OH 1131 are mostly responsible for the inaccessibility of the active-site center, together with the side-chain of Arg141. On the other hand, this side-chain is hydrogen bonded to OH 1947 and through this to OH 1131, which in turn is hydrogen bonded to Gly139 0. Moreover, the copper ion’s fifth ligand (OH 1991) is hydrogen bonded to OH 1131. Another

K. Djinovic

236

et al.

Figure 9. Stereoscopic drawing showing the geometry of the active-site metals and their ligands in Cu.C’o b-SOD (continuous line), superimposed on the same acbive-site residues as observed in the (‘u,Zn b-SOD enzymcl (broken line).

water molecule (OH 1046), playing a structural role in the protein, hydrogen bonds to the His61 ND1 (the copper-cobalt bridging residue) and Lys134 0 atoms in the a-helical region. The remarkable coincidence of the location of several water molecules in two enzymes that are phylogenetically distant, but which nevertheless display quite comparable catalytic properties (O’Neill et al., 1988), stresses the relevance of the solvent contribution in stabilizing the protein’s tertiary structure and possibly in supporting its catalytic activity. Bearing in mind that the substrate of SOD is a small charged dioxygen anion whose dimensions compare favorably with those of a water molecule, it is possible that the solvent molecules observed in the strict neighborhood of the active site of the two different SODS indicate successive steps in the diffusion process of the substrate to the redox center. In this respect it is unfortunate that no solvent structure is yet available for the monoclinic crystal form of the Cu,Zn b-SOD. (f) Comparison

of the active sites

Concerning the perturbation of the enzyme active site structure in the presence of the foreign Co2+, Figure 9 shows an overlay of Cu,Co b-SOD active site with the native enzyme. The overlay has been performed taking into account only the active-site atoms shown: namely, those belonging to the amino acids directly co-ordinated to the metal ions, and Arg141. After least-squares superposit$on of the t’wo atom sets an r.m.s. value of 0.401 A is observed. Tnspection of the overlaid structures shows that the largest deviation present in Cu,Co b-SOD active site as compared to the Cu,Zn b-SOD occurs at residue atoms are Asp81, whose carboxylat? group displaced from 0.49 to 1.34 A. Despite this, the coordination bond is maintained within a distance of 1.99 and 1.96 B in the two subunits, which is in a good agreement with values in the literature. On the other hand, it is fairly difficult to explain the substantially different backbone conformation observed for Asp81 in the two proteins. As a general conclusion, it can be stated that t,he substitution of Co(H) for the native Zn(I1) does not bring about strong perturbations in the active-site

structure or ill the general conformation of the protein. The fairly high level of conservation of the active-site geometry in the &Co b-SOD, wit’h respect to the Cu,Zn b-SOD, is in line with t#he essentially identical catalytical properties of the native and semisynthetic enzymes (Calabrese el al., 1972; McAdam et al., 1977). The most important consequence of this work is, however, the availability of the crystallographic co-ordinates at 2.0 A for the Cu,Co b-SOD derivative, which will permit’ the definitive assignation of the n.m.r. paramagnetic shifts in order to compare structures determined by crystallographic and solution methods (Banci et al., 1989; Paci et al.. 1988, 1990).

Area per la Ricerca di Triestr (Italy) is t)hanked for financial support to Kristina Djinoric. This project was supported by grants from the Italian Eational Research Council target,-oriented Project “Hiotecnologie e Biostr~ispecial project .‘l’eptidi Bioattivi”. and mentazione”. from the Ministr,v of the University and Scientific Research

References Banci. L.. Bertini. I.. I,uchinat, C’.. Piccoli. M.. Scozzafava. A. & Turano. P. (1989). ‘H ?jOE studies on dic~opper(II)~dicobaIt(II) superoxidr dismutase. Inorg. (‘hem. 28. 465&4656. Berendsen. H ,J, (‘.. f’ostma. .J. I’. M.. Gunstrren. IV. F. van. IXNola. I\. & Haak, .J. 1~. (1984). Molecular dynamics with coupling to an external bath. ,I. (‘hw,. Pkys. M(8). 3684-3690. Bernstein. F. C.. Koetzle, T. F.. Williams, G. .I B.. Meyer. E. F., . & Taylor, R. (1989). Tables of bond lengths determined by X-ray and neutron diffraction. Part 2. Organometallic rompounds and co-ordination complexes of the d- and f-block metals. J. Chrm. Sot. Dalton, SlS83. Paci. M.. Desideri, A. & Rotilio. G. (1988). Cyanide binding to Cu,Zn superoxide dismutase. An ,ljMR study of the Co(I1) derivative. .I. Bial. (‘hem. 263 (I), 162-166. Pari. M.. Desideri. A., Sete. M., Falconi. )I. & Rotilio, G. (1990). Mapping the copper ligands of Cu,Zn superoxide dismutase by nuclear Overhauser enhancement, of the isotropically shifted ‘H-NMR lines of the Cu.Co derivative. FEBS Letters, 261, 23 l-236. Ramakrishnan. C. & Ramachandran. (:. X. (1965). Stereochemical criteria for polypeptide and protein cahain conformation. Biophys. J. 5. 909-933. Read. R. .J. (1986). Improved Fourier coefficients for maps using phases from partial st’rurtures with errors. Acta Crystallogr. sect. A. 42. 14&149. Rot’ilio, G., C’alabrese, L., Yondovi, B. & Blumberg. W.E. (1974). Electron paramagnetic resonance studies of cobalt-copper bovine superoxide dismutasr. J. Riol. (‘hem. 249. 3157-3160. Sines, ,J. ,J.. Allison. S. 8. & McCammon, ,J. A. (1990). Point charge distributions and elect,rostatir steering in enzyme/substrate encounter: Brownian dynamics of modified copper/zinc superoxide dismutases. Biochemistry. 29. 9403-9412. Stallings. W. (‘.. Pattridge, K. A.. Strong, R. K. &, Ludwig, M. I,. (1984). Manganese and iron superoxide dismutases are structural homologs. .J. Biol. Chem. 259. 10695-10699. Tainer, ,J. 9., Getzoff. E. D., Beem, K. M.; Richardson, tJ. S. & Richardson, D, C. (1982). Det’erminat,ion and analysis of the 2 A structure of copper. zinc superoxide dismutase. J. Mol. Biol. 160, 181-217. Tegelstrom. H. (1975). Interspecific hybridisation in vitro of superoxide dismutasr from various species. Hrrrditas. 81. 185-198. Tronrud. I). E., TenEyck. L. F. & Mat,thews. B. W. (1987). An rfhcient general-purpose least-squares

238

K. Diinovic

refinement program for macromolecular structures. Acta Crystallop. sect A, 43, 484-501. Van Gunsteren, W. F. & Berendsen, H. ,J. C. (1987). BIOMOS. Biomolecular Software. Laboratory of Physical Chemistry, University of Groningen, The Netherlands. Edited

et al. Vriend, G. (1990). WHAT IF: A molecular modelling and design program. J. Mol. Graph. 8, 52-56. Ward, K. B., Wishner, B. r., Lattman, E. E. t Love. W. E. (1975). Structure of deoxyhemoglobin A crystals grown from polyethylene glycol solutions. J. Mol. Riol. 98, 161-177.

by R. Huber