PROTEINS: Structure, Function, and Genetics 8:352-364 ( 19901

The Structure of Rubredoxin From Desulfouibrio desulficricans Strain 27774 at 1.5 A Resolution Ronald E. Stenkamp, Larry C. Sieker, and Lyle H. Jensen Department of Biological Structure, SM-20, University of Washington, Seattle, Washington 98195 The structure of a small ruABSTRACT bredoxin from the bacterium Desulfovibrio desulfuricans has been determined and refined at 1.5 A resolution. The hairpin loop containing seven residues in other rubredoxins is missing in this 45 residue molecule, and once that fact was determined by amino acid sequencing studies, refinement progressed smoothly to an R value of 0.093 for all reflections from 5 to 1.5 A resolution. Nearly all of the water molecules in the well-ordered triclinic unit cell have been added to the crystallographic model. As in the other refined rubredoxin models, the F e S , complex is slightly distorted from ideal tetrahedral coordination.

rubredoxins using mutagenesis techniques would allow an assessment of the effects of individual amino acid substitutions on the iron complex, but it seems unlikely that a deletion of seven amino acids would be attempted in such a study. We felt this justified our examination of a functional, natural, altered protein. After considerable crystallographic effort, we found that the major structural change involves the deletion of seven residues near the middle of the polypeptide chain. We describe here the structure determination and refinement and present an analysis of the resultant model of this atypical rubredoxin. A preliminary account of work on this structure has appeared earlier.7

Key words: crystallography, refinement, structure comparison, molecular replacement, precision

Triclinic crystals of the protein were grown from a 1%(w/v) protein solution containing 0.1 M citrate (pH 4.0) and 25% saturated ammonium sulfate as described elsewhere.' Diffraction data to 1.5 A resolution were collected from a single crystal using 0-20 scans on a KRISEL Control updated Picker FACS-1 diffractometer. Details of the data collection are presented in Table I.

INTRODUCTION Rubredoxins are small, nonheme iron bacterial proteins containing one iron atom tetrahedrally coordinated by four cysteinyl sulfur atoms. Their presumed biological function is to serve as electron transport proteins. Specific receptors for some of them have been identified,' and in vitro studies have shown electron transfer interactions between rubredoxins and cjtochrome c3.' The redox potentials for rubredoxins vary from -60 mV to + 6 mV3 indicating that variations in structure and possibly activity exist in this class of proteins. Rubredoxins have been extensively studied crystallographically to provide a structural basis for understanding the molecular function, to determine how the small structure accommodates changes in the amino acid sequence, and to test and evaluate crystallographic methods. Rubredoxins from Clostridium pasteurianum (54 amino acids),4 Desulfovibrio vulgaris (52 amino acid^),^ and D.gigas (52 amino acids)6 have been refined at relatively high resolution to yield precise models of the polypeptides, Fe centers, and bound water molecules. Comparison of those refined models indicated only minor structural differences, so when a considerably smaller rubredoxin was isolated from D.desulfuricans (Strain 27774) (45 amino acids), we became interested in how its structure would differ from those of the other rubredoxins. Systematic alteration of C 1990 WILEY-LISS, INC.

EXPERIMENTAL

STRUCTURE DETERMINATION AND REFINEMENT The initial molecular model was generated by applying molecular replacement techniques"' to coordinates for the rubredoxin from D.vulgaris. Rotation functions'' were calculated for three resolution limits (26-5 A; 14-4.2 A, and 10-4 A) with a 15 A radius in the Patterson, and the correct orientation peak was present in all three. Because the position of the molecule is arbitrary in space group P1, there was no need for translation function calculations. F J l for the The crystallographicR ( = X ~ ~ F O [ - ~/CIF,/) initial model was 0.495 for the 5-2 resolution data. Modification of the model to one containing only glycine, alanine, or cysteine side chains reduced R to 0.492 and provided the starting point for refinement. The initial model development and refinement were

Received January 16, 1990; revision accepted May 11, 1990. Address reprint requests to Ronald Stenkamp, Department of Biological Structure, SM-20, University of Washington, Seattle, WA 98195.

353

RUBREDOXIN FROM D. DESULFURICANS

TABLE I. Crystal, Data Collection, and Data Reduction Information Crystal data Crystal size Space group Unit cell

0.12 x 0.24 x 0.56 mm P1 a = 24.916(6) A b = 17.779(4)8, c = 19.710(5)A a = 101.02(1)" p = 83.36(2)" y = 104.52(2)"

from least-squares fit to 22 centered reflections

One protein molecule per unit cell Data collection: Data collected on a Picker FACS-1 four-circle diffractometer CuK, radiation, A-1.5418 A w-20 scans 2"lmin scan rate Scan width = 0.64" + standard term for dispersion in 20 4 second backgrounds on either side of reflection maximum 20 = 60" (1.5 8, resolution) 10 standard reflections collected periodically throughout data collection 11,616 observations (includes Friedel mates) Data reduction: Empirical absorption corrections Deterioration correction obtained from fit to the function: c1 + c2x20xtime + c (2012xtime2 + c,xtime + c,xtime32 x20 + c,xtime2 10,127 observed reflections [ 2 a ( p ) cutoffl out of 10,835 reflections maximum absorption correction = 1.595 Maximum deterioration correction = 1.091 5,031 uni ue reflections R, = g)lI: F2)=0.030for 1,563 replicate reflections R , = X ( I p , - -I)/X p ) = 0 . 0 4 8 for 5,023 Friedel pairs

X(lsz-

carried out in the absence of amino acid sequence information because we felt the high resolution data should suffice for determining the complete structure of the protein. Restrained least-squares refinement (PROLSQ)12at gradually increasing resolution limits was used extensively to improve the model. Many revisions were made based mainly on the interpretation of Fourier and difference Fourier maps with some reference to known amino acid sequences for other r ~ b r e d 0 x i n s . l ~ Early in the refinement process, the hairpin loop corresponding to residues 18-25 in the other rubredoxins did not behave well in refinement. Evidence for this can be seen in a plot of average main-chain temperature factors (Fig. 1)where the B values become quite large for this region in the incorrect models. The amino acid composition data indicated this rubredoxin contained between 45 and 48 residues, and comparison of the composition with the known sequences suggested that the deletions would most likely occur a t the C-terminus of the protein. This was incorrect, and, as it happens, the C-terminal portion of an adjoining molecule in the crystal extends into this "missing loop" region, complicating the map interpretation. Determined attempts to model this part of the

structure including (1) deletion of the loop and reintroduction of residues on the basis of subsequent difference maps, (2) addition of the hairpin loop from C. pasteurianum rubredoxin after superposing residues 17 and 26 (amino acid numbering system for the other rubredoxins), and (3) restriction of the resolution limits for the difference maps in case the refinement had introduced phase bias for the higher resolution reflections. None of these approaches yielded models which behaved well in refinement, but R was reduced to 0.208 for the 5-1.5 A data. At this point, the amino acid sequence became a ~ a i l a b l e and ' ~ showed that the entire hairpin loop, residues 19 through 25, was missing in this molecule. The correct polypeptide model was easily fit to the difference density (R=0.249 for the 5-1.5 A data), and refinement proceeded satisfactorily. Water oxygen atoms were added at occupancies proportional to the peak height of residual density in difference maps, hydrogen atoms were added a t calculated positions, van der Waals restraints were deleted from the refinement, and the structure factor contributions to the least-squares matrix were weighted more heavily. The structural model at this point contained only one disordered side chain,'" that of the cysteine a t

354

R.E. STENKAMP ET AL.

20.

w

3 _I

Q

> m 10.

+

Fig. 1. Main chain B values for models at various stages in the refinement. x , , and o denote coordinate sets which incorrectly model the region near the hairpin loop. 'denotes the final model.

Fig. 2. Stereoscopic drawing showing the two alternate side-chain conformations for Gln-25. The difference electron density generated by omitting the two water molecules from the model and fit with the second conformation is also included.

position 38 (C. pasteurianum numbering). Before the amino acid sequence was determined, the electron density for this residue was fit with a valine as found at this position in both C. pasteurianum and D. vulgaris rubredoxin. However, in this molecule, the residue is a cysteine. The density was fit with two Sy atoms of about half occupancy each. Refinement of this model yielded a n R of 0.087, but it was found that four water sites were too close to parts of the protein model. One of these was deleted, one was replaced by a disordered side chain for Glu-12, while two of them (see Fig. 2) were replaced by a n alternate conformation of the side chain of Gln-25. Correction of these side chains was followed by several refinement cycles. An anisotropic temperature factor for the Fe atom was introduced in the final cycles of least-squares refinement carried out by treating the calculated

structure factors for the rest of the structure as fixed contributions. The X-RAY System" was used for this refinement. This resulted in a n R of 0.093 for all 4867 reflections from 5 to 1.5 A resolution for the final model. The rms deviation between the bond lengths for the model and ideal values is 0.019 A. The parameters used for PROLSQ are listed in Table 11. The coordinates have been deposited in the Brookhaven Protein Data Bank. Table I11 contains the torsion angles for the final model while Figure 3 shows the average B values for the main-chain and side-chain atoms. The R value and scale for the model are shown in Figure 4 as functions of sin W h . Two additional points should be made concerning the refinement. First, the low R value for this structure merits comment. While an R of 0.093 is quite respectable for a protein structure determination, we believe that as data collection and refinement

355

RUBREDOXIN FROM D.DESULFURICANS

TABLE 11. PROLSQ Weighting Paramters (Final Cycle)* Parameter

Target

rms deviation of

SIGD3 SIGP SIGC

0.050 0.050

0.100

0.053 0.029 0.141

SIGB1 SIGB2 SIGB3 SIGB4 SIGT2 SIGT3 SIGT4

2.00 A2 2.00 3.00 3.00 3.0" 15.0 20.0

1.41A2 1.83 2.97 3.78 4.2" 11.9 16.0

Deviation from planar 1x4 distances Deviation from ideal planes Deviation from chiral volumes Deviation between temperature factors for Bonds with only main-chain atoms Angles with only main-chain atoms Bonds with at least one side-chain atom Angles with at least one side-chain atom Deviation for planar torsion angles Deviation for staggered torsion angles Deviation for orthonormal torsion angles

*All restraint scale weights were set to 1.0, except for the van der Waals restraints which were given zero weight. The structure factors were weighted by AFSIG x u(R where AFSIG was 12.0.

techniques improve, more precise (and hopefully accurate) protein models will become possible. The present R value is, however, somewhat seductive in that it encourages the thought that this structure is comparable to those normally found in small molecule crystal structure analyses. An R of less than 10% for a 500 atom, disordered, isotropic small molecule structure would be considered evidence for a reasonably well-determined structure. One should note though that most small molecule structure determinations are carried out a t significantly higher resolution. A data set collected to a 20 of 55" for MoK, radiation is a 0.77 A resolution data set. For this rubredoxin, a data set at that resolution would contain approximately 40,000 reflections. If an extrapolation of the average structure factor amplitudes and differences is made (based in part on data shown in Fig. 4), the R value for the current model with a 0.77 A data set would be approximately 0.16, not nearly as encouraging an R as 0.093. Our conclusion is that a t an R of 0.093 at 1.5 A resolution, significant errors may still remain in the data and model which are not properly accounted for. Second, the correct structure determination depended crucially on the determination of the amino acid sequence. In retrospect, several features of the crystal structure contributed to our difficulties with the "missing loop" region. At 5.0 A resolution, electron density consistent with the normal loop structure suggested that the deletions were at the end of the polypeptide, not in the middle. Furthermore, a t high resolution, the appearance of a five-membered ring a t the position of proline 20 in the other rubredoxins (at the top of their hairpin loops) again seemed to confirm the presence of the loop in this molecule. The C-terminus of a neighboring molecule, the side chain of histidine 18, and highly ordered water molecules account for the density in this region. Also, we would like to point out that even if

we had solved the loop problem it is doubtful that the disordered Cys-38 could ever have been identified by crystallographic means a t this resolution.

THE STRUCTURE As can be seen in Figure 5, the fold of the polypeptide chain of rubredoxin from D. desulfuricans is similar to that of the other rubredoxins except near the residues a t the ends of the missing seven residue loop. The original amino acid sequence alignment where this shortened rubredoxin was compared to other rubredoxin sequenced4 placed the deletion between residues 19 and 27 in order to superpose an aspartic acid with the Asp found in several other rubredoxins. Based on the structure (as can be seen in Fig. 5), the deletion is more correctly placed between residues 18 and 26, although residue 18 is very much in the center of the changes in the path of the polypeptide chain. The rubredoxins from C . pasteurianum, D. uulgaris, and D . gigas can be superposed17 on the D. desulfuricans molecule on the basis of residues 1-16 and 27-52. After doing so, the rms distances between the Ca atoms of each molecule and those of D. desulfuricans are 0.61, 0.51, and 0.66 A, respectively. These values are somewhat larger than expected from the estimated standard deviations and are indicative of structural differences among the molecules on a scale not easily noticed in a visual inspection of the superposed Ca atoms (see Fig. 5). Figure 6 is a plot of the distances between the C a in the D. desulfuricans rubredoxin and the other three molecules after they have been superposed. The plot shows the two regions where most of the structural differences occur; the loop region (residues 16 through 27) and the regions near residues 34-36 and 46-47. A detailed view of how the shortened polypeptide cuts across the base of the hairpin loop is shown in

R.E. STENKAMP ET AL.

TABLE 111. Main-Chain and Side-Chain Torsion Angles Residue 1 Met 2 Gln 3 Lys 4 Tyr 5 Val 6 Cys 7 Asn 8 Val 9 cys 10 Gly 11 Tyr 12 Glu 13 Tyr 14 Asp 15 Pro 16 Ala 17 Glu 18 His 26 Asp 27 Asn 28 Val 29 Pro 30 Phe 31 Asp 32 Gln 33 Leu 34 Pro 35 Asp 36 Asp 37 np 38 Cys 39 cys 40 Pro 41 Val 42 Cys 43 Gly 44 Val 45 Ser 46 Lys 47 Asp 48 Gln 49 Phe 50 Ser 51 Pro 52 Ala

*

4 -61 -91 - 129 -119 - 68 - 62 - 87 -118 81 - 58 -116 - 88 - 109 -49 - 60 - 88 - 132 59 66 - 82 -66 -57 -72 - 95 -68 -67 -61 - 92 - 63 -138 -49 - 79 - 98 -118 86 - 74 -79 - 59 -70 - 106 -111 - 127 -61 - 154 ~~

149 121 151 145 141 123 -30 -46 -1 0 137 123 109 112 -35 -26 0 47 30 27 114 155 - 36 -5 -11 146 162 -27 2 140 162 135 - 10 -44 -11 0 160 171 - 23 -6 26 154 147 142

U

0

-172 176 178 177 175 -179 -171 -171 177 176 177 173 177 -179 -178 178 - 177 176 175 173 178 178 -175 177 -177 - 172 174 - 175 178 172 178 - 177 179 - 170 176 178 177 -176 -179 179 176 176 175 - 176

13 0 -5 2 -1 1 -4 1 -1 -4 9 - 13 -11 -5 5 0 0 -11 -9 1 0 2 1 -4 -9 8 0 4 -8 -6 -4 18 0 9 -16 2 -3 5 6 0 -4 0 -3 10

P)

X1

-59 - 178 -57 - 75 69 - 175 - 63 170 79

X2

-

170

-

62

-

64

166 171 - 174 - 175 - 27 70 55 -61 -75 171 25 - 163 61 - 162 - 58 14 -67 69 -171 56 179 22 171 73 -

-

-61 - 65 - 69 72 - 70 - 82 - 60 3

-

59 64

~

Figure 7. The polypeptides have quite similar conformations up to residue 16 and beyond residue 27. The short-cut across the base of the loop appears to be a rather drastic structural change, but in fact, most of the hydrogen bond interactions between the main-chain atoms are unperturbed by the amino acid deletions. The carbonyl of residue 17 and the amide of residue 26 participate in the same hydrogen bonds in all four proteins, and the carbonyl oxygen of residue 18 participates in the same hydrogen bonds as does the carbonyl of residue 25 in the other rubredoxins. However, the amide of residue 18 interacts with the carbonyl oxygen of residue 14 rather than 15 as in the other proteins. Table IV lists distances between equivalent atoms in D. desulfuricans and D. uulgaris rubredoxins and shows that the major differences occur near the main-chain atoms of residue

18. The movement of the amide is sufficient to realign its hydrogen bond with the carbonyl oxygen one residue down the polypeptide chain. Deletion of the hairpin loop has no other effect on main-chain hydrogen bonding interactions in these molecules, mainly due to the fact that in the larger rubredoxins, there are no main-chain hydrogen bonds between residues in the loop and the rest of the protein. However, the removal of the loop causes a large change in the accessible surface area of the protein, the largest change occurring for tryptophan 37. In the larger rubredoxins, the tryptophan is almost completely buried in the protein while for D . desulfuricans rubredoxin, its accessible surface is much increased. In this crystal form two water molecules are close enough to hydrogen bond t o the nitrogen atom of the aromatic ring, N d .

RUBREDOXIN FROM D.DESULFURZCANS

357

0 0

0.

I

f

I

I

I

I

I

0

10

20

30

40

50

SEQUENCE NUMBER

t i .50

0.40

I

X

v

I

0

0.304

0

W

3

X

J

a

>

a

W

J 6

u

cn

0.50

0.20

ninon

0.101 EU01

0.00

0

v

1.00

0 0000000

f

0.00

I

I

I

0.20

0.10

SIN

I

I

I

-

-

0 . LO A

0.076 A

0.00 b

I

0.30

(THETA) /LAMBDA

Fig. 3. Average 8 values for the main-chain ( x ) atoms in the final model for D. desulforicans rubredoxin. Individual 8 values for side-chain atoms (0).

Fig. 4. R ( =ZllFol - ~ f c ~ / L Z ~( xf o) ~and ) scale ( =ZlfclnlFoI) ( 0 ) as functions of sin e/h. Coordinate enor estimates from reference 22 are superposed on the R value.

Another region showing relatively large variation among the rubredoxins is located between residues 34 and 36, as shown in Figure 6. We became aware of the special nature of this region by inspecting plots of thermal parameters as a function of sequence for the well-refined models (see Fig. 8). Our initial thought was that the higher thermal param-

eters for the main-chain atoms of residues 35 through 37 for D. desulfuricans rubredoxin were indicative of some required change of the structural model in this area. However, investigation of difference maps, comparison of the superposed structures, and a comparison of the packing environments in this region in the four rubredoxin crystals indicated

R.E.STENKAMP ET AL.

358

Fig. 5. Stereoscopic view of the (Y carbon atoms and the Fe-S complex in D. desulfuricans rubredoxin (open bonds) superposed on the a carbon plot from D. vulgaris rubredoxin (lines).

I ffl

2.0-

Z

0

m a 4

u 4

1.5-

I

a

J 4

z w

W

1.0-

3

+ W

m W

uZ

0.5-

4

+In H

a 0.0

f

I

I

I

I

I

0

10

20

30

40

50

SEQUENCE NUMBER

Fig. 6. Distances between the a carbon atoms of D. desulfuricans rubredoxin and D. vulgaris rubredoxin ( x ), D. gigas rubredoxin (0). and C. pasteurianurn rubredoxin (+) after superposition of the latter three on the first.

that the structural differences are real and are due to changes in the intermolecular contacts. In the other three rubredoxins, several intermolecular hydrogen bonds involve residues 35,36, and 37,and they make up part of the crystal packing interactions between the protein molecules in those structures. In the smaller D. desulfuricans rubredoxin, the hydrogen bonding interactions made by these residues all are to water molecules in the crystal. The higher B values are consistent with weaker intermolecular interactions in this region which would allow for greater variation in the structures.

rubredoxin differs from the others is near the Nterminus of the protein. The large differences in the Ca positions are consistent with the larger thermal parameters for the N termini in the three larger rubredoxins. The Fe-S, complex in this rubredoxin (see Fig. 9) is not significantly different from the metal centers in the larger rubredoxins. The Fe atom is tetrahedrally coordinated to four cysteinyl sulfur atoms, with no significant variation in the Fe-S bond lengths. These values agree with those found in several FeS, complexesls”o (see Table V).

Another region where the structure of this smaller

There is an interesting distortion of the tetrahe-

359

RUBREDOXIN FROM D. DESULFURICANS

Fig. 7. Stereoscopic view of the polypeptide in D. desulfuricans rubredoxin near the seven amino acid deletion (open bonds) and the longer flap in D. vulgaris rubredoxin (lines).

TABLE IV. Distances for Individual Atoms at the Shortcut Across the Seven Residue Loop Between the D. desulfiricans Model and a Superposed D. vulgaris Model Atoms in D. vulgaris model

Atoms in D . desulfuricans Model

Atoms in D . vulgaris model

Ca(17) C(17) -

-

-

dral symmetry of the complex as shown by the bond angles about the iron atom. The deviations from 109.4’ are significant for this refined molecule. The pattern of distortions in the S-Fe-S bond angles holds when all four rubredoxin structures are compared (see Table V). The Sy(G)-Fe-Sy(S), Sy(6)-Fc+ Sy(39), Sy(9bF&y(42), and Sy(39)-Fe-Sy(42) angles are larger than the tetrahedral angle, and the Sy(6)-Fe-Sy(42) and Sy(9)-F+Sy(39) angles are smaller, consistent with an expansion of the complex along the axis passing through the Fe atom and between residues 6 and 42 (and 9 and 39). Interestingly, the expansion is seen in all of the proteins and five of the model complexes listed in Table V. Only the first inorganic compound is observed to be compressed.” The expansion observed in the model compounds in different crystal packing environ-

6.95 A 4.84 A 2.64 A 1.58 A 2.34 A 1.49 A 1.11 A 1.03 A

-

-

-

-

ments and with a number of different ligands has led to the belief that the distortion of the complex is “inherent to the structure” of these compounds.’s Why the one complex is compressed is not understood. Many of the small molecule model complexes possess D,, symmetry which in the case of the proteins is reduced to approximately C, symmetry. This can be seen in Figure 9 where residues 6 and 39 are related by a pseudo 2-fold axis as are cysteines 9 and 42. This pattern has been noted before4 and can be observed in the torsion angles about the C Q s y bonds and in the Fe-Sy-Cp bond angles. Dr. E.T. Adman (personal communication) has also pointed out that the sulfur atoms of residues 6 and 39 are involved in three N-H . . . S hydrogen bonds while those of residues 9 and 42 participate in only two.

360

R.E. STENKAMP ET AL.

i

X 20.W 3 _I


m 10.

-

0

10

I

1

I

I

20

30

40

50

SEQUENCE NUMBER

Fig. 8. Average B values for the main-chain atoms in all four rubredoxin models. D. desulfuncans rubredoxin ( x ) , D. vulgaris rubredoxin (+), D. gigas rubredoxin (o),and C. pasteurianurn rubredoxin (*).

-6

CYS

CYS

39

Fig. 9. The iron-sulfur complex in D. desulfuricans rubredoxin.

The additional atoms and close contacts with Sy(6) and Sy(39) would be consistent with the decrease in the CP-Sy-Fe bond angle. The observation of a distortion in the metal coordination sphere immediately raises the issue of the precision of the refined model. Due to the inclusion of stereochemical restraints in the block diagonal least-squares, the accuracy of standard deviations of the positional parameters derived from the leastsquares matrix is difficult to determine. The values obtained from the inverse matrix are on the order of 0.001,0.002, and 0.004 A for the positions of the iron, sulfur, and carbon atoms, respectively, and are

much too small for this structure. Presumably the sparseness of the least-squares matrix causes this underestimate of the standard deviations. We can also judge the standard deviation in the positions by examining the distribution of various bond lengths, although this will be affected to an unknown extent by the stereochemical bond length restraints. Figure 10 shows the distribution of CaCf3 bond distances. The average Ca-Cf3 distance is 1.542 A. The rms deviation and standard deviation of this sample are 0.017 and 0.017 A. The R values plotted in Figure 4, when compared to Luzatti’s plots for estimated standard deviations, 22 would correspond to standard deviations in positions of about 0.08 A. To test the effects of restraints, we gradually reduced their weights in successive cycles of refinement to yield a “freely refined’model. The rms shift for all atoms in the structure was 0.098 A. This probably reflects more on the convergence and restraints than on the precision of the model, but its is gratifying that the model did not diverge. We have also made estimates of the standard deviations using Cruickshank’s methodz3 as modified by Chambers and S t r o ~ d . ’The ~ resulting standard deviations in position are 0.107, 0.095, 0.094, 0.053 0.015, and 0.450 A for the C, N, 0, S, and Fe atoms and the water molecules, respectively. The estimates of the standard deviations presented here emphasize the difficulty in assessing the precision of molecular models obtained with data a t less than atomic resolution and refined using restraints. Given these problems, a realistic estimate

361

RUBREDOXIN FROM D . DESULFURICANS

TABLE V. Selected Bond Lengths, Bond Angles, and Torsion Angles for the FeS, Complexes in Several Rubredoxins and Model Compounds (a) Bond lengths, A Compound D. desulfuricans rubredoxin D . gigas rubredoxin C. pasteurianum rubredoxin D. vulgaris rubredoxin

Reference

FeSy(6)

FeSY(39)

FeSy(42)

This work

2.282

2.262

2.303

2.246

6 4

2.308 2.333

2.295 2.288

2.271 2.309

2.273 2.235

21

2.330 Fe-S 2.284 Fe-S1 2.359 2.379 Fe-S1 2.265 2.272 2.378

tFe(SC,oH,,)41-

18

[Fe(SC,H,),12;[Fe(S2C4O2),1

19 19

Complex A Complex B Complex C

20 20 20

Sy(6)FeComDound D . desulfuricans rubredoxin D. gigas rubredoxin C. pasteurianum rubredoxin D . vulgaris rubredoxin

FeSyW

Reference

2.290 Fe-S 2.284 FeS3 2.338 2.396 Fe-S2 2.258 2.252 2.324 (b) Bond angles,' Sy(6)-

2.270 Fe-S 2.284 Fe-S4 2.355 2.387 Fe-S4 2.279 2.278 2.347

2.276 Fe-S 2.284 Fe-S2 2.360 2.394 Fe-S3 2.268 2.265 2.376

Sy(6)-

Fe-

Fe-

Sy(9)Fe-

Sr(39)

Sr(42)

Sy(9)FeSV(39)

Sy(39k FeSy(42)

This work

111.3

112.7

104.6

100.7

114.5

113.3

6

114.5

111.4

106.1

103.4

109.5

112.0

4

113.8

108.8

104.0

103.7

114.2

112.4

Complex B Comalex C

21 18 19 19 20 20 20

106.0 114.4 97.89 95.77 105.8 106.7 106.9 FeSr(39k CP(39)

104.1 114.4 101.34 95.83 108.5 107.0 103.5 FeSy(42)Cp(42)

113.1 107.08 115.27 114.23 109.1 109.8 113.8

Reference

110.4 107.08 112.67 115.11 109.3 109.2 108.5 FeSY(9)CP(9)

109.4 107.08 111.47 112.47 111.4 112.2 114.9

Compound

114.0 107.08 119.0 124.86 112.6 112.0 109.1 FeSy(6)Cp(6)

This work

100.0

107.6

103.0

111.0

rubredoxin

6

103.7

108.0

99.3

108.7

rubredoxin

4

100.0

107.8

99.3

109.5

rubredoxin

21

102.8

109.1

100.4

111.7

FeSr(39k CP(39)Ca(39)

FeSy(42)Cp(42)Ca(42)

D . desulfuricans rubredoxin

D. gigas C. pasteurianum D . vulgaris

(c) Torsion angles

Compound D. desulfuricans rubredoxin D. gigas rubredoxin C. pasteuriunum rubredoxin D. uulgaris rubredoxin

Reference This work

FeSy(6)Cp(6)Ca(6)

FeSy(9)cpc9r Ca(9)

-180.0

-98.4

-175.4

-86.9

6

-171.7

-91.1

-177.6

-94.4

4

-170.0

-91.4

- 177.1

-86.7

21

-171.0

-89.6

-174.3

-88.8

362

R.E. STENKAMP ET AL.

12

10

0

m

0

6

4 L 0

4

a m 3 2

2

0

1 .4s

1.so

1 Dl

5

1.60

1.65

TANCE

Fig. 10. Distribution of Cu-Cp bond lengths.

of the standard deviation in the carbon positions is about 0.10 A. This leads to an estimated standard deviation in a carbon-carbon bond of 0.14 A, and scaling by the atomic numbers, would give a standard deviation for an F e S bond length of 0.05 A. In any case, none of the individual Fe-S bond distances differs by more than 0.05 A from the average value of 2.273 A. Thus the F e S distances are not significantly different. Good small-molecule structure determinations often lead to standard deviations of 0.003 in bond lengths and 0.3"in angles, and comparable scaling to obtain a standard deviation for the S-Fe-S angle in this molecule gives a value of about 5".One of the S-FeS angles deviates from the mean by 8.2",but the significance of the distortion observed in the complex rests mainly on the pattern observed in all four rubredoxin molecules rather than the statistical significance of any one of them. Another assessment of the quality of the structure determination and the experimental data, albeit a qualitative one, can be made by looking a t a final difference electron density map where the structure factors are calculated for the nonhydrogen atoms. Roughly half of the calculated hydrogen positions lie in residual density. It is testimony to the sensitivity of the X-ray crystallography experiment that hydrogen atoms can be observed in such large structures. The high resolution and precision of this structure analysis are due to the highly ordered crystals and are the consequences of the low solvent content and the ordered nature of the water molecules in the spaces between the protein molecules. At a first approximation, the packing of protein molecules in this triclinic crystal is simlilar to the packing of spheres with columns of solvent space running par-

allel to each crystallographic axis and intersecting

in a larger cavity. In fact, the nonspherical nature of the protein allows contacts betwen a molecule and more than its six nearest neighbors in the lattice. Table VI lists the amount of accessible surface occluded by each crystal packing contact. While the major contacts are between molecules related by the unit cell translation operations, several of the contacts along the diagonals of the unit cell are significant. Both polar and nonpolar atoms occur at the crystal packing interfaces as seen in Table VI. Comparison with the atoms contributing t o the accessible surface of the isolated molecule indicates no significant partitioning of polar atoms between the crystal packing surfaces and the surface exposed to solvent molecules in the crystal. Two other features of the crystal structure are apparent in Table VI. One is that the crystal packing interactions along the crystallographic a axis are not as extensive as along the b and c axes. The accessible surface involved in interactions in this direction (186or 198 A2)is smaller than those along the other axes. This is consistent with a widening of the crystal's solvent region to make a much more extensive layer of water in the b,c plane. The second feature of these crystals seen in Table VI is that much of the protein's surface is involved in crystal packing interactions. Normally, most of the surface of protein molecules in crystals is covered by solvent molecules with the crystal packing interfaces being much smaller in extent. In fact, the crystal packing surfaces for C .pasteurianum rubredoxin make up only 49% of the total protein surface in the rhombohedra1crystals, so the larger amount of crystal packing surface observed for D . desulfuricans rubredoxin is likely caused by the particular packing interactions in this crystal form and not just the small size of the protein. As in other protein crystals, the water molecules occupying the entire solvent space are in a state of static or dynamic disorder. All of the water molecules except two are within 2.5to 3.5 A of each other and can be thought of as one network of hydrogen bonded water molecules. Only two solvent molecules, water 59 and water 71,are not connected to this net. Nearly all of the water molecules, 107 out of 121,are hydrogen bonded to the protein molecule. Only 14 are not involved in direct hydrogen bonds to the protein, and none is further than 4.5 A from the protein. This is another indication of the small size of the solvent region in these crystals. In summary, several features of the protein structure are apparent in Figure 11, a stereoview of the entire molecule. Delection of the hairpin loop causes a significant exposure of the edge of Trp-37.We would expect that hydrogen bonds to solvent molecules and changes in the accessible surface would affect the spectroscopic properties of the tryptophan, although no details concerning the spectra are yet available. In addition, the side of the molecule where the loop is

363

RUBREDOXIN FROM D . DESULFURICMS

TABLE VI. Accessible Surface Area (A2)Occluded by Interactions Between a Rubredoxin Molecule and Neighboring Ones Related by Unit Cell Translations*

X -1 -1 -1 0 0 0

0 0 0 1 1 1

Translation in Y -1 0 0 -1 -1

0 0 1 1 0

0 1

z 0 0 1 -1 0 -1 1 0 1 -1 0 0

Nonpolar area - 104.8 -78.9

Polar area -42.9 -40.0

0.0 -70.6 -165.5 -181.9 -141.9 - 178.0 -83.2 - 16.8 -87.1 -91.3

0.0 -67.1 -58.5 -39.0 -155.2 -70.8 -58.5 0.0 -49.8 -37.0

Total areas For crystal interfaces fraction For solvent cavity fraction

1610.0 1199.8 0.75 410.2 0.26

885.0 618.9 0.70 266.1 0.30

Charged area -12.3 -67.7 -23.1 -37.0 -62.7 -90.4 -25.8

0.0

Total area -160.0 -186.7 -23.1 -174.6 -286.7 -311.4 -322.9 -251.9 -157.0 -26.0 - 198.5 - 128.3

562.0 408.3 0.73 153.7 0.27

3058.0 2227.0 0.73 831.0 0.27

-3.1 -15.4 -9.2 -61.6

*Areas calculated using a program written by D.C. Teller.26

Fig. 11. Stereoview of the entire molecule. The number scheme for the residues contains a break between His-18 and Asp-26 to retain the residue numbering found in the other rubredoxins. Multiple conformations are included for the side chains of Glu-12, Gln-32, Cys-38,and Lys-46.

missing is now a concave surface with four aromatic amino acid side chains (Tyr-11, Tyr-13, His-18, and Trp-37) and two prolines (Pro-34,Pro-40) nearby. It is possible that this change in the surface, compared to the other rubredoxins, is used to control recognition of the molecule by its redox partners. Also, the side chains of Tyr-4, Phe-30, and Pro-51pack next to each other and may play a role in electron transfer to or from the Fe center. Further experimentation will be necessary to address these possibilities.

ACKNOWLEDGMENTS We wish to thank B. Prickril and Dr. J. LeGall for samples of the protein and helpful discussions and

the referees for some helpful comments. This work has been supported by USPHS Grants GM-13366 and GM-32663.

REFERENCES 1. Petitdemange, H., Marczak, R., Blusson, H., Gay, R. Isolation and properties of reduced nicotinamide adenine dinucleotide rubredoxin oxidoreductase of Clostridium acetobutylicum. Biochem. Biophys. .~ Res. Commun. 91:125812651 1979. 2. Bell, G.R., Lee, J.-P., Peck, H.D., Jr., LeGall, J. Reactivity of Desulfovibrio gigas hydrogenase toward artificial and natural electron donors or acceptors. Biochimie 60:315320,1978. 3. Moura, I., Moura, J.J.G.,, Santos, M.H., Xavier, A.V., LeGall, J. Redox studies on rubredoxins from sulphate and sulphur reducing bacteria. FEBS Lett. 107:419-421,1979. 4. Watenpaugh, K.D., Sieker, L.C., Jensen, L.H. The struc-

364

R.E. STEN‘KAMPE:T AL.

ture of rubredoxin at 1.2 A resolution. J. Mol. Biol. 131: 509-522,1979. 5. Adman, E.T., Sieker, L.C., Jensen, L.H., Bruschi, M., LeGall, J. A Structural model of rubredoxin from Desulfovibrio vulgaris at 2 A resolution. J. Mol. Biol. 112: 113-120,1977. Frey, M., Sieker, L., Payan, F., Haser, R., Bruschi, M., 6. Pepe, G., LeGall, J . Rubredoxin from Desulfovibrio gigas. A molecular model of the oxidized form at 1.4 A resolution. J. Mol. Biol. 197525-541, 1987. 7. Sieker, L.C., Stenkamp, R.E., Jensen, L.H., Prickril, B., LeGall, J. Structure of rubredoxin from the bacterium Desulfovibrio desulfuricans. FEBS Lett. 208:73-76, 1986. 8. Sieker, L.C., Jensen, L.H., Prickril, B.C., LeGall, J. Crystallographic study of rubredoxin from the bacterium Desulfovibrio desulfuricans strain 27774. J. Mol. Biol. 171: 101-103, 1983. 9. North, A.C.T., Phillips, D.C., Mathews, F.S. A semi-empirical method of absorption correction. Acta Crystallogr. A24351-357,1968. 10. Rossmann. M.G. “The Molecular ReDlacement Method.” New York Gordon and Breach, 1972.11. Crowther, R.A. The fast rotation function. In: “The Molecular Replacement Method (Rossmann, M.G., ed.). New York Gordon and Breach, 1972: 173-185. 12. Hendrickson, W.A., Konnert, J.H. Incorporation of stereochemical information into crystallographic refinement. In: “Computing in Crystallography” (Diamond, R., Ramaseshan, S., Venkatesan, K., eds.). Bangalore: Indian Institute of Science, 1980: 13.01-13.23. 13. Yasanobu, K.T., Tanaka, M. The types, distribution in nature, structure-function, and evolutionary data of the ironsulfur proteins. In: “Iron-Sulfur Proteins,” Vol. I1 (Lovenberg, W., ed.). New York: Academic Press, 1973: 27-130. 14. Hormel, S., Walsh, K.A., Prickril, B.C., Titani, K., LeGall, J., Sieker, L.C. Amino acid sequence of rubredoxin from Desulfovibrio desulfuricans strain 27774. FEBS Lett. 201: 147-150,1986, 15., Smith, J.L., Hendrickson,W.A., Honzatko, R.B., Sheriff, S.

16.

17. 18.

19.

20.

21. 22. 23. 24.

25.

Structural heterogeneity in protein crystals. Biochemistry 25:5018-5027,1986. Stewart, J.M., Machin, P.A., Dickinson, C.W., Ammon, H.L., Heck, H., Flack, H. The X-RAY76system technical report. TR-446. Computer Science Center, Univ. of Maryland, College Park, Maryland, 1976. Ferro, D.R., Hermans, J. A different best rigid body molecular fit routine. Acta Crystallogr. A33: 345-347, 1977. Millar, M., Lee, J.F., Koch, S.A., Fikar, R. Synthetic models for the iron-sulfur protein rubredoxin: synthesis, structure, and properties of a highly symmetric iron(II1) tetrathiolate anion. Inorg. Chem. 21:4105-4106, 1982. Coucouvanis, D., Swensen, D., Baenziger, N.C., Murphy, C., Holah, D.G., Sfarnas, N., Simopoulos,A., Kostikas, A. Tetrahedral complexes containing the Fe”S, core. The syntheses, ground-state electronic structures, and crystal and molecular structures of the [P(C,H,),l,Fe(SC,H,), and [P(C,H,),l,Fe(S,C,O,), complexes. An analogue for the acJ . Am. Chem. SOC. tive site in reduced rubredoxins (WJ. 103:3350-3362, 1981. Lane, R.W., Ibers, J.A., Frankel, R.B., Papaefthymiou, G.D., Holm, R.H. Synthetic analogues of the active sites of iron-sulfur proteins. 14. Synthesis, properties and structures of bis(o-xylyl- a,a-dithiolato)ferrate(II,III) anions, analogues of oxidized and reduced rubredoxin sites. J. Am. Chem. SOC.99:84-98, 1977. Adman, E.T., personal communication. Luzzatti, V. Traitement statistique des erreurs dans la determination des structures cristallines. Acta Crystallogr. 5:802-810, 1952. Cruickshank, D.W.J. The accuracy of electron-density maps in X-ray analysis with special reference to dibenzyl. Acta Crystallogr. 255-82, 1949. Chambers, J.L., Stroud, R.M. The accuracy of refined protein structures: Comparison of two independently refined models of bovine trypsin. Acta Crystallogr. B35: 1861-1874,1979, Teller, D.C., personal communication.