Eur. J. Biochem. 234’, 190-196 (1996) 0 FEBS 1996

Crystal structure of flavodoxin from Desulfovibrio desulfuricans ATCC 27774 in two oxidation states Antonio ROMERO’, Jorge CALDEIRA“, Jean LEGALL’. Isabel MOURA4, JosC J. G. MOURA“ and Maria J. ROMAO’ I Consejo Superior de Investigaciones Cientificas, Departamento de Cristalografia, Instituto de Quimica Fisica “Rocasolano”, Madrid, Spain

Instituto de Tecnologia Quimica e Biolbgica, Oeiras, Portugal Department of Biochemistry and Molecular Biology, University of Georgia, Athens GA, USA Departamento de Quimica, e C. Q. F. B. da Faculdade de Ciencias e Tecnologia, Universidade Nova de Lisboa, Portugal

(Received 16 November 1995/8 February 1996) - EJB 95 1905/3

The crystal structures of the flavodoxin from Desulfovibrio desulfikricans ATCC 27774 have been determined and refined for both oxidized and semi-reduced forms to final crystallographic R-factors of 17.9 % (0.8-0.205-nm resolution) and 19.4% (0.8-0.215-nm resolution) respectively. Native flavodoxin crystals were grown from ammonium sulfate with cell constants a = b = 9.59 nm, c = 3.37 nm (oxidized crystals) and they belong to space group P3,21. Semireduced crystals showed some changes in cell dimensions: a = b = 9.51 nm, c = 3.35 nm. The three-dimensional structures are similar to other known flavodoxins and deviations are found essentially in the isoalloxazine ring environment. Conforinational changes are observed between both redox states and a flip of the Gly61 -Met62 peptide bond occurs upon one-electron reduction of the FMN group. These changes influence the redox potential of the oxidized/ semiquinone couple.

Modulation of the redox potentials is known to be related to the association constant of the FMN group to the protein. The flavodoxin from D. desulfuricans now studied has a large span between E2 (oxidi:zed -+ semiquinone) and El (semiquinone - hydroquinone) redox potentials, both these values being substantially more positive within known flavodoxins. A comparison of their FMN environment was made in both oxidation states in order to correlate functional and structural differences.

Keywords: flavodoxin ; sulfate-reducing bacteria; X-ray ; three-dimensional structure ; crystallization.

Flavodoxins are a group of well known electron transfer pro- teins [ I ] that possess one FMN group buried in a single polypep- tide chain. Flavodoxins can be classified in two generic classes, according to their niolecular mass (15 or 23 kDa). These simple flavoproteins are isolated froin prokaryotes and algae, but have never been found in higher organisms. The FMN cofactor may occur in three different oxidation states (oxidized, semiquinone, and hydroquinone) when bound to the protein matrix.

In sulfate reducers, flavodoxin is a component of the sulfite reductase system [2,, 31 and of the pyruvate dehydrogenase sys- tem [4]. Ferredoxin can replace flavodoxin [5 , 61 and the role of the redox partners has been studied [7]. In contrast to Clos- tridium pasteurianum [8], in Desulfovibrio species both ilavo- doxin and ferredoxin are synthesized by the bacteria. In D. gi- gas, for example, it has been noted that the cells contain about 10 times more flavodoxin than ferredoxin even in the presence of high concentrations of iron [4]. The reason for these differ- ences with the clostridia is still unknown but could be an indica- tion that flavodoxin and ferredoxin play different and specific physiological roles in Desulfbvibrio.

The ilavodoxin isolated from D. desulfuricans ATCC 27774, a sulfate-reducing organism that can utilize nitrate as an alterna- tive electron acceptor, was previously purified, analyzed accord- ing to its redox properties, and the amino acid sequence deter- mined [9]. The mid-point redox potentials of this flavodoxin

Correspondmce ta A. Romero, Instituto de Quimica Fisica “Rocaso-

Fax: +34 1 5642431. lano”, 119 Serrano, E-28006 Madrid, Spain

were determined by ultraviolet/visible and EPR methods cou- pled to potentiometry measurements and their pH dependence studied in detail. The redox potential E,, for the couple oxidized/ semiquinone forms at pH 6.7 and 25 “C is -40 mV, while the value for the semiquinonehydroquinone forms ( E l ) , at the same pH is -387 mV. E2 varies linearly with pH, while E, is indepen- dent of pH at high values. However, at low pH (<7.0), this value is less negative, compatible with a redox-linked proton- ation of the flavodoxin hydroquinone. The modulation of redox potentials is due to different binding energies of the FMN in different oxidation states to the protein. This is discussed in terms of the three-dimensional structures.

EXPERIMENTAL METHODS

Crystallization. Crystallization experiments were performed with D. desulfuricans ATCC 27774 flavodoxin isolated and puri- fied as described [9]. Crystals were obtained at room temper- ature froin a protein solution of 8-10 mg/ml in 10 mM Tris pH = 7.6. Ammonium sulfate is known to be the precipitant used in the successful crystallization of flavodoxins in general. Also in this case we obtained clusters of very thin needles using 2.0-2.5 M ammonium sulfate as precipitant and pH = 5.0-5.5. Due to their dimensions (< 0.05 mm in thickness), these needles were of insufficient quality for X-ray analysis. A substantial im- provement could be achieved by adding 10% 2-methyl-2,4- pentanediol to the crystallization buffer, which contained 2.3 M ammonium sulfate, 0.1 M sodium acetate or 0.1 M sodium po-

Romero et al. (EUK J . Biochern. 239) 191

Table 1. Crystallographic data for flavodoxin. R,,,,,, (0 = ( l I - ( l ) l ) / Z .

Parameter Value for flavodoxin

oxidized (yellow) semireduced (purple)

Crystal size (mm) 0.2 x 0.2 x 1 .0 o.1.5x0.1sxo.8 Space group P3,21 P3,21 Unit cell parameters (nm)

Resolution range (nm)

u = b = 9.59 c = 3.37 m-0.205

u = h = 9.51 c = 3.3s z-0.215

Number of unique reflections 10 938 9173

R,,,,, (70) 8.2 1.5.5 Completeness (%) 94.2 87.6 90.2 93.3 (resolution shell, nm) (m-0.205) (0.21 -0.205) (m-0.217) (0.225-0.217)

tassium phosphate, pH = 5.0. Crystals were prepared by mixing equal amounts of the flavodoxin and crystallization solutions. After an incubation period, the initially formed precipitate was centrifuged off and drops were prepared. Long yellow crystals of 0.2X0.2X1.0 mm were obtained by the vapour diffusion method within a few days. These crystals diffract to beyond 0.21-nm resolution. The crystals were characterized by precession pho- tographs and the space group identified as P3,21 or P3,21. The correct enantiomorph was identified from the structure solution using Patterson search techniques (see below). The cell con- stants are a = b = 9.59 nm, c = 3.37 nm. The packing density is 0.0028 nm'/Da, corresponding to one flavodoxin molecule/ asymmetric unit.

Preparation of reduced crystals. Reduced flavodoxin crys- tals were obtained directly from the oxidized ones. Since the original conditions of growth were at pH = 5.0, the crystals were first carefully transferred to 2.5 M ammonium sulfate, pH = 7.0, the pH required for the reduction to the semiquinone state [lo]. Within 10-15 min of 100 mM sodium dithionite be- ing slowly added, the crystal colour gradually changed from yel- low to purple, corresponding to the stable semireduced state. The reoxidation process of these reduced crystals is slow and, after a 24-h period, the purple colour did not change signifi- cantly, considering that an excess of sodium dithionite is present. However, under exposure to X-rays, we found that the semire- duced crystals were more sensitive than the oxidized ones and special care was taken to minimize the presence of oxygen in the capillary. For data collection, crystals were mounted under a flush of argon in capillaries first purged with argon. Despite these precautions, the semireduced crystals suffered much more from radiation damage than the oxidized crystals.

Data collection. The X-ray intensities were collected with a MAR imaging plate system (Hendrixaenfter), using CuK, radi- ation from a Rigaku rotating anode generator operating at 5.4 kW, and a crystal to detector distance of 90 mm. Crystals were rotated almost around the c axis and data sets were col- lected with 1 .5" scandframe and exposure time of 1000 s. The oxidized crystals were kept at 5 "C throughout data collection, while the semireduced crystals were cooled to -5°C with a stream of cold air. For both oxidized and semireduced crystals, the intensity data were processed and scaled with the program MOSFLM, version 5.2 [ I l l and CCP4 package [12]. Data col- lection statistics for both data sets are summarized in Table 1.

Structure solution. The structure was solved by Patterson search techniques using as a search model the coordinates of the homologous (48.6%) D. vulgaris tlavodoxin, refined to 0.19-nm resolution [I31 and obtained from the Brookhaven Protein Data Bank. A solution was found by either using XPLOR package

[14] with PC refinement of rotation function peaks, or Patterson search methods [15-171 and routines implemented in PRO- TEIN [18].

For the rotation function calculation with XPLOR, the search model was placed in a P1 cubic box of l0XlOXlO nm. Structure factor calculations for the model used data over 0.8-0.32 nm and a model Patterson map was calculated with a 0.025-nm sam- pling grid. The 7000 largest vectors of the Patterson map were selected. Product correlation function of crystal and model Pat- terson functions were calculated in the range 0- = 0-120", 0, = 0-90" and 0, = 0-360" [19] in 1.5" interval. The 6000 highest peaks of the rotation function were subsequently ana- lyzed by PC refinement method 1141. The refinement clearly showed one major peak in the rotation function occurring at 0, = 200.4", 0, = 45", 0, = 132.8", clearly above the next peak. The model was rotated according to the orientation found and the translation search performed with XPLOR, first two- dimensionally along x and y , and finally along z ; this gave a clear solution with a maximum T function of 0.298, which is more than 3a above the next solution. This result was obtained for the correct space group P3,21; calculations in P3,21 gave no solutions for the translation function.

In parallel, and using the orientation found with XPLOR, another translation function 1201 was calculated using programs written by E. G. Lattman. The resulting map was very weak and failed to give the correct position. Another approach was successfully attempted using the analytical packing function of Stubbs and Huber [21]. The highest peak was found at 5 . 4 ~ with the next peak below 4.0a. Both independently determined solutions are identical.

Refinement. The correctly oriented and positioned flavo- doxin molecule was refined first as a rigid body using 0.8-0.3- nm data and the program XPLOR, version 3.1 [14]. Besides standard XPLOR force field parameters (paraml Bx.pro), stan- dard group geometries of Engh and Huber [22] were employed for the refinement. Residue topology and parameter files for FMN have been taken from the XPLOR force field parameters which will allow a butterfly twist about N.5 -N10. The non-com- inon residues between D. vulgaris and D. desulfuricnns ATCC 27774 flavodoxin sequences were replaced by alanines and the oxidized flavodoxin data used for the refinement.

After 30 cycles of rigid body refinement and 200 cycles of positional refinement, the initial R factor of 49 % dropped to 34.3 %. At this stage, the correct amino acid sequence was grad- ually included throughout the refinement, accompanied by map inspection on an interactive graphics display (Evans and Suther- land). Manual rebuilding of residues not well defined was per- formed with program FRODO /23]. When all the 146 residues

192 Romero et al. (Eur: J. Biochem. 239)

Table 2. Quality of the final model.

Model parameters

toured at 4.0a showed high density in the loop region Gly61- Met62. This observation suggested a peptide flip around Gly61- Met62 peptide bond, which was accordingly modeled and subse- quently refined. The final R factor for the semiquinone form of D. desuljiuricans tlavodoxin was 19.4% for 8738 reflections (0.8-0.215 nm).

Value for flavodoxin

oxidized reduced

Rrns deviation from target values of: bond distances (nm) Bond angles (")

0.001 3 0.0014 2.93 2.98

dihedrals:(") 25.94 26.07 RESULTS AND DISCUSSION impropers (") 2.2s 2.38

Quality of the model. The final model consists of one flavo- doxin molecule and 92 water molecules for the oxidized form all atoms 0.2501

0.2354 and 83 for the semireduced form. Root-mean-square (rms) devi- protein atoms FMN group 0.1107 0,1217 ations from target values and average 5 factors are summarized solvent 0.4261 0.3601 on Table 2 for both structures and show that the final models

Average B frlctor (nm') for: o,2332 0.2236

Model coordinate error [31] (nm) 0.0201 0.0221

and the FMN proiithetic group were included, the R factor dropped to 24.7 %, for data between 0.8-0.205-nm resolution. At this stage we found out that, in the C-terminal region, one further residue could be fitted into the density, probably a valine. Water molecules were then included at stereochemically reason- able positions, where peaks were found in F,,-F, maps above 3.0a, using program MAIN [24]. With 93 water molecules in- cluded, individual restrained B factors were also refined and the R factor converged to 21.3% (data 0.8-0.205 nm).

From 2F,,-F, density maps calculated at this stage, and comparison with difference maps, the presence of a second addi- tional residue was evident and a leucine could be easily built at the C terminus. Both new added residues, Val147 and Leu148, refined very well.

In the final model refinement procedures, including indivi- dual 5 factors, the R factor converged to 17.9% for 10649 re- flections (0.8-0.205 nm for the oxidized flavodoxin structure).

Fur the refinement of the semireduced flavodoxin model, the refined model of the oxidized form was used. Initially, a differ- ence Fourier niap was calculated with coefficients F,,,, from the semireduced and F,.,,,c from the oxidized form. This map showed significant differences between both structures only in the vicin- ity of the FMN prosthetic group, as expected. The map con-

have good stereochemistry. With one exception, conformational 4,y angles for both structures lie within areas of low energy in a Ramachandran plot. In the oxidized form, Met62 lies in a for- bidden region with 4 = 53" v, = -101". This residue interacts with the isoalloxazine ring and is involved in conformational changes which occur upon reduction to the semiquinone form (as discussed below). A resulting peptide flip places Met62 close to the cn-region of a Ramachandran plot (4 = -108"; y = -83"), for the semireduced structure (see section below).

Structures of the oxidized and reduced flavodoxin. The rib- bon representation of Fig. 1 shows clearly the typical alp topol- ogy of this class of proteins: a central five-stranded parallel p- sheet flanked by a pair of parallel cx-helices on either side. The secondary structure as well as the polypeptide folding of the D. desulfuvicuns flavodoxin molecule, described in this work, is very similar to the homologous D. vulgaris protein [13]. All the backbone atoms of the two flavodoxins are superimposed with a rms deviation of 0.099 nm. A superposition of the a-carbon atoms of the two structures is shown in Fig. 2, which shows differences in loops close to the FMN group, mainly comprising two regions Ala59-Glu66 and Asp95 -Glu99 (discussed be- low). However, a sequence comparison of both flavodoxin mole- cules in the FMN binding region shows similarity in terms of primary sequence (Fig. 3).

The FMN prosthetic group is rather exposed to the solvent with the ribityl moiety pointing to the inside of the molecule where it establishes many hydrogen bonding contacts to the pro-

Fig. 1. Stereo ribbon diagram of D. desuZfuricans flavodoxin with the bound FMN group represented in ball-and-stick mode. Produced by MOLSCRIPT [30].

Roinero et al. (Eur J. Biochern. 239) 193

A

Fig. 2. Superposition of Cn plots of flavodoxin from D. desulfuricans (thicker line) and the one from D. vulgaris (thinner line).

9 10 11 12 13 14 15 16 17

D. desulfuricans S D. vulgaris Y

57 58 59 60 61 62 63 64 65 66 67 69 70

93 94 95 96 97 98 99 100 101 102 103 104 105

Fig. 3. Sequence comparison of D. desulfuricans and D. vulgaris fla- vodoxins in the FMN binding region. A shaded block indicates contact with FMN in both D. rlesulfiricntzs and 0. vdguris; a black block indi- cates contact with FMN in D. desulfuricans; a clear block indicates a conserved residue.

tein atoms (Fig. 4, Table 3). The phosphate group is anchored at the N terminus of helix 1, establishing several hydrogen bonds to free NH groups at the end of the helix. The isoalloxazine ring is practically planar and enclosed by the aromatic rings of Trp60 and Tyr98. The tyrosine phenyl ring is almost parallel to the flavin ring with an angle between plane normals of 9.5", while the tryptophan indole ring is tipped by 47". These values are i n agreement with other structurally resolved flavodoxins [13, 251.

During the course of the structure refinement, we found out that two further residues could be added to the C-terminal resi- due Lys146 of the published primary sequence [9]. The good quality of the electron density maps (Fig. 5 ) allowed us to build the X-ray sequence Va1147-Leu148, which is favoured by the very hydrophobic environment in this region of the molecule : Val147 contacts to lle120 and Leu148 is surrounded by Va14, Va126, His31, Ala52 and Leu145. Furthermore, the first methio- nine residue is not seen in density. The published primary se- quence was accordingly modified, with SerZ as the first residue and ending with Va1147-Leu148, and the calculated molecular mass (15408.4 Da) now agrees, within experimental error, with the molecular mass of the apo-flavodoxin determined by mass spectrometry (1 5 420 ? 19 Da).

FMN environment in the two didation states. In Fig. 4 is shown a pictorial representation of the FMN group interactions to neighbouring protein atoms of tlavodoxin in the oxidized and semiquinone forms. Hydrogen bonds (Table 3) taken into ac- count are only those with favorable geometry. The whole pros-

B

Fig. 4. Stereo representation of the FMN group hydrogen bonding interactions to the protein environment for (A) oxidized and (B) semi-reduced forms fo flavodoxin.

Table 3. Hydrogen bonds between the FMN and the apoprotein or solvent in both oxidation states.

FMN Apoproteinlsolvent Distance in flavodoxin

oxidized reduced

nm

N1 N3 N5 0 2

0 4

02' 03' 04'

OP1

OP2

OP3

N 0 0 N N N 0 0 0

N62 0 N

N N

OY

01'

OY OY

Oyl N

Asp95 His100 Gly61 Asp95

Met62 Wat306 Ah59 Wat3 18 Ser93 Am14 Wat312 Trhl2 Thrl2 Am14 Serll Ser58 SerlO ThrlS Thrl5

Cysl02

0.354 0.289

0.325 0.300 0.330 0.210 0.296 0.315 0.263 0.296 0.346 0.278 0.254 0.296 0.287 0.259 0.277 0.277 0.281

-

0.366 0.211 0.269 0.340 0.295

0.258 0.293 0.337 0.210 0.306

0.277 0.248 0.302 0.287 0.259 0.211 0.282 0.261

-

~

194 Romero et al. (Eur: J . Bzochem. 239)

F i g 5 Stereo plot 0 1 the 2F,-F, electron density map (contouring level la) at the C-terminus region, around the two new residues (Va1147-Leu148) added to the published sequence [9].

thetic group is tightly bound to the protein and most of the in- teractions are similar to other flavodoxin structures 113, 25, 261. The phosphate group establishes O+HN hydrogen bonds to res- idues of the N terminus of helix 1, plus three further contacts to 01' of SerlO, ThrlS and Ser.58. The ribityl moiety is also stabi- lized by several hydrogen bonding contacts to backbone and side-chain atoms as well as to internal water molecules. The isoalloxazine ring is imbedded in the protein, stacked between the aromatic groups Trp60 and Tyr98. It is further stabilized by hydrogen bonds to backbone atoms of segments Gly61 -Met62, Asp95 -Cysl02, and two internal water molecules. This hy-

I drogen bonding network is essentially conserved in the known flavodoxin structures which is not surprising since, apart from two water molecult:s, only backbone atoms are involved and they belong to conserved secondary structure. From the two water molecules, water W306 (Fig. 4) is also present in the D. vulgaris structure in a similar environment. One significant dif- ference between D. clesuljiuricuns flavodoxin and the other two short chain flavodoxins (D. vulgaris and Clostridiurn M P ) is the hydrogen bond FMN, NI--+HN, Asp95. The distance observed in D. desu/juricuns flavodoxin is, within the experimental error (0.02 nm), too long to be considered as a favorable interaction (0.354 nm for the oxidized form and 0.366 nm for the semiqui- none). The corresponding values for the D. vulgaris flavodoxin are 0.323 and 0.3'37 nm respectively 1131 and 0.292 and 0.304 nm (FMN, N1-HN, Gly89) for Clostridiuin MP. [27]. In D. desulfuuicarzs flavodoxin, the FMN group may therefore be less tightly bound to explain the different redox properties of the different flavodoxins (as discussed below).

In respect to differences between both oxidized and semire- duced structures, relevant changes occur at the peptide Gly61- Met62. The superposition of all backbone atoms of the two structures, gives an rms deviation 0.02 nm although, in the Gly61-Met62 region, both structures deviate more than 0.1 nm. These differences result from a peptide flip of the Gly61-Met62 peptide, upon the one-electron reduction of the FMN group (Fig. 6). This conformational change was deduced after inspec- tion of 2F,,-F, and F,, (semi-reduced)-F, (oxidized) electron density maps of the semireduced form (Fig. 7 B). In the oxidized structure (Fig. 7A), the carbonyl group of Gly61 points away from the NS atom of the isoalloxazine ring, while in the semiqui- none form, positive electron density points to the now proton- ated NS atom. The conformational change due to the peptide flip Gly61-Met62, which occurs in the semiquinone form, con- tributes to the additional stabilization of this redox state (Fig. 4).

Fig. 6. A comparison of the FMN region and residues 60-64 of flavo- doxin from D.desulfiricans for both oxidation states: oxidized (thin- ner line) and semi-reduced (thicker line).

A

\

B

Fig.7. Stereo view of the 2F,,-F, electron density map (contouring level la) around the FMN region and peptide Gly61-Met62 for (A) the oxidized form and (B) the semi-reduced form of flavodoxin (ori- ented as in A). Hydrogen bonding interaction of O(61) to NS of the FMN group is marked by a broken line.

The carbonyl oxygen of Gly61 flips over to make a favorable hydrogen bond to the protonated N5 atom of the isolalloxazine ring. The orientation of the Gly61-Met62 peptide also favors a hydrogen bonding contact of Gly61, NH with 0 4 of the FMN group, mediated by a water molecule in both oxidation states.

Romero et al. (Eul: J. Biochem. 239) 195

Table 4. Redox potential values of flavodoxins with known tbree- dimensional structure (pH = 7.0).

Type Source E2 E,

mV

Short-chain Clostridium beijerinckii 1271 - 92 -399 D. vulgaris [13] -103 -438 D. desulfuricans - 40" -387

Long-chain Anabaena [25] -196 -425 Anacystis nidulans [28] -220 -415 Chondrus crispus [26] -222 -370 Free FMN -238 -172

a Value measured at pH 6.7

Comparison of D. desulfuricans flavodoxin with literature data. Analysis of the redox potential properties. As men- tioned before, flavodoxins have been grouped into two classes according to the length of their chains. So far, three X-ray struc- tures have been published for the long-chain flavodoxins [25, 26, 281 and two crystal structures for the short-chain flavodoxins [13, 271, in which class D. desulfuricans flavodoxin is included. These classes are structurally different: in the long-chain struc- tures a 20-residue segment is inserted in the middle of a p-strand of the short-chain structure. Although this insertion does not in- terfere with the FMN binding region, significant differences are found between the two classes of flavodoxins, in their redox properties. Table 4 summarizes redox potentials for the several flavodoxins with known three-dimensional structure. It is evi- dent from this list that redox potential values E, (oxidized-semiquinone) are more negative for the long-chain flavodoxin family and comparable to the E2 of free FMN, in contrast to the corresponding more positive values for the other class. In structural terms, there are relevant differences in the interaction of the isoalloxazine ring with the surrounding pro- tein atoms: the most striking difference is the fact that, in all three long-chain flavodoxins in the oxidized form, the N5 atom of the FMN group is hydrogen-bonded to main-chain NH (Anab- aena and Anacystis nidulans) or side chain Oyl groups (Chon- drus crispus). In the short-chain flavodoxin family, the N5 atom is free in the oxidized state. There are also differences within this class of flavodoxin, which we attempt to correlate with the relative values of E2 potentials. For this analysis, we have com- pared D. desulfuricans flavodoxin structure with the homolo- gous D. vulgaris structure. The calculated rms deviation between both structures, after superposition of all backbone atoms, is 0.099 nm, which shows that the folding of both structures is well conserved. The larger deviations are found in the regions AlaS9-Glu66 (0.110-0.335 nm) and Asp95-Glu99 (0.110- 0.485 nm). The carbonyl oxygen atom of Trp60 deviates 0.335 nm since, in this loop region, a significant conformational change is present in the D. desulfuricans structure in comparison to the D. vulgaris flavodoxin. The largest difference between both structures is 0.485 nm and is due to a conformational change in the loop 97-100 region and caused by a flip of the Glu99-His100 peptide. In D. vulgaris flavodoxin, the carbonyl group of Glu99 points to the solvent region, while in D. desul- furicans flavodoxin it points towards the isoalloxazine ring with a 0.290-nm distance to the oxygen atom 0 4 of the FMN group (Fig. 8). The resulting peptide flip places Glu99 in the allowed P-region of a Ramachandran plot ( 4 = -65"; cp = 161") whereas in D. vulgaris flavodoxin it is in the a-region ( 4 = -56"; q = -47"). This new conformation is very clearly seen

- 4 1

5 5 Fig. 8. Stereo superposition of the region around the FMN cofactor and the residue Glu99, of flavodoxin from D. desu&uricans (thicker line) and D. vulgaris (thinner line). The unfavourable interaction of O(99) to carbonyl oxygen atom 0 4 of the isoalloxazine ring is caused by a flip of the Glu99-His100 peptide.

in the electron density maps, where a significant positive differ- ence density appears clearly pointing to the atom 0 4 of the FMN group. Therefore, in this new orientation the carbonyl oxygen of Glu99 is very well defined with a low temperature factor (0.15 nm'). Also, in the semiquinone structure, the peptide is in a similar orientation and the corresponding distance is larger (0.32 nm). This structural difference looks as if it should destabi- lize the FMN in D. desulfuricans flavoprotein, being more pro- nounced in the oxidized form relative to the semiquinone, where the contact is 0.03 nm longer.

In addition, the structural flip of Gly61-Met62, favouring hydrogen-bonding contact with the flavin moiety in the one- electron reduced state, also contributes to the stabilisation of this form. This conformational change, upon reduction, has also been reported for other known flavodoxin structures in different oxi- dation states [13, 271.

However, the dissociation constants for the oxidized forms of D. desulfuricans [9] and D. vulgaris [29] flavodoxins, (0.1 nM and 0.2 nM respectively, determined under identical ex- perimental conditions) do not agree with the proposed destabili- zation resulting from 0-0 repulsion and other differences be- tween both flavodoxins may add together with the 0-0 contact, to give the net energy differences indicated by the binding con- stants and potentials. The flipping of the Gly61-Met62 segment is observed upon the semiquinone formation in both Desulfovi- brio flavodoxins, which favors a stabilization of this redox state, but the 099 (Glu)/04 (FMN) interaction seems an unique fea- ture in D. desulfuricans flavodoxin.

The modulation of redox potentials is known to be a function of the strength of the association constant. The differences in E2 redox potentials (-40 mV for D. desulfuricans at pH 6.7 ver- sus - 103 mV for D. vulgaris at pH 7.0) should therefore reflect the mentioned structural changes.

This work was supported by a NATO Collaborative Research grant (931 648 to Insrituto de Tecnologia Quimica e BioldgicdConsejo Supe- rior de Investigaciones Cientificas) and by Dil: Gen. de Investigacibn Cientifica y Ecnica (PB93-0120). The authors thank R. Huber (Max- Plank Institut f i r Biochemie, Martinsried, Germany) for advice and the use of the imaging plate MAR system for collecting X-ray data, L. Sieker for discussions on crystallization experiments, the Fermentation Plant U. G . A. and M. Y. Liu for the growth of the cells.

196

REFERENCES

Romero et al. (Eur: J. Biochem. 239)

15. Hoppe, W. (1957) Acta Crystallogr: 10, 750-751

1.

2.

3 .

4.

5.

6.

7.

8.

9.

10.

11.

12.

13.

14.

Mayhew, S. G. & Ludwig. M. L. (1975) in The enzymes (Boyer, P.

LeGall, J. & Hatchikian, E. C. (1967) C.R. Acad. Sci. Paris 264,

Peck, H. D. Jr & LeGall, J. (1982) Phil. Trans. R. Soc. Lorzd. B 298,

Hatchikian, E. C. & LeGall, J. (1970) Ann. Znst. Pasteur 118, 288- 301.

LeGall, J. & Dragani, N. (1966) Biochem. Biophys. Res. Commun. 23, 145.

Moura, J. J. G., Macedo, A. L. & Palma, P. N. (1994) Methods Enzymol. 243, 165-188.

Moura, J. J. G., Moura, I., Xavier, A. V., Hatchikian, E. C. & Le Gall, J. (1978) FEES Letr. 89, 177-179.

Knight, E. Jr & Hardy, R. W. E (1966) J. Bid. Chem. 241, 2752- 2756.

Caldeira, J., Palma, P. N., Regalla, M., Lampreia, J., Calvete, J., Schafer, W., LeGall, J., Moura, I. & Moura, J. J. G. (1994) Eur: J . Biochem. 220, 987-995.

Watenpaugh, K. I)., Sieker, L. C. & Jensen, L. H. (1976) in Flavins andjZuvoprotei.rzs (Singer, T. P., ed.) pp. 405-410, Elsevier, Am- sterdam.

Leslie, A. G. W. (1990) MOSFLM program, Abstract of the Crystal- lographic Computing School, Bischenberg.

Evans, P. R. (1990) The CCP4 program package, pp. 139-146, Ab- stract of the Crystallographic Computing School, Bischenberg.

Watt, W., Tulinsky, A., Swenson, R. P. & Watenpaugh, K. D. (1991) J. Mol. Biol. 218, 195-208.

Briinger, A. T. (1992) X-PLOR (version 3.1) manual, Yale Univer- sity, New Haven.

D., ed.) vol. 12B, pp. 57-109, Academic Press, New York.

2580-2583.

845 - 850.

16. Rossmann, M. G. & Blow, D. M. (1962) Acta Crystallogl: 15, 24-

17. Huber, R. (1965) Acra Ctystallogr: 19, 353-356. 18. Steigemann, W. (1974) PhD thesis, Technische Universitat

19. Rao, S. N., Jih, J. W. & Hartsuck, J. A. (1980) Acta Crystallogr:

20. Crowther, R. A. & Blow, D. M. (1967) Acta Ctystallogl: 23, 544-

21. Stubbs, M. T. & Huber, R. (1991) Acra Crystallogl: A47, 521 -526. 22. Engh, R. & Huber, R. (1991) Acta Crystallogr: A47, 392-400. 23. Jones, T. A. (1978) J. Appl. Crysrallogl: 15, 23-31. 24. Turk, D. (1992) PhD thesis, Technische Universitat Munchen. 25. Rao, S. T., Shaffie, F., Yu, C., Satyshur, K. A,, Stockman, B. J.,

Markley, J. L. & Sundaralingam, M. (1992) Prorein Sci. I , 3413- 1427.

26. Fukuyama, K., Malsubara, H. & Rogers, L. J. (1992) J . Mol. Bid.

27. Smith, W. W., Burnett, R. M., Darling, G. D. & Ludwig, M. L. (1977) J. Mol. Biol. 177, 195-225.

28. Smith, W. W., Pattridge, K. A,, Ludwig, M. L., Petsko, G. A,, Tser- noglou, D.. Tanaka, M. & Yasunobu, K. T. (1983) J. Mol. Bid.

29. Curley, G. P., Carr, M. C., Mayhew, S. G. & Voordouw, G. (1991)

30. Kraulis, P. J. (1991) J. Appl. Crystallogl: 24, 946-950. 31. Luzzati, V. (1952) Acta Crystulbogr. 5, 802-810.

31.

Munchen.

A36, 878-884.

548.

225, 775-789.

165, 737-755.

Eur: J. Biochem. 202, 1091 -1 100.