.I. Mol. Biol. (1991) 218, 195-208

Comparison of the Crystal Structures of a Flavodoxin in its Three Oxidation States at Cryogenic Temperatures William Watt’, Alexander Tulinsky’, Richard P. Swenson3 and Keith D. Watenpaugh’ ‘Physical Upjohn

2Department

and Analytical Co., Kalamazoo,

Research 49007, U.S.A.

Chemistry

MI

of Chemistry, Michigan State University E. Lansing, MI 48824, l7.S.A.

3Department

of Biochermistry, Ohio State liniversity Col’umbus, OH 43270, U.S.A.

(Received 3 September 1990; accepted 5 November 1990) The focus of this study has been to determine t,he conformation of the holoprotein of recombinant flavodoxin from Desulfovihrio vulgaris with the FMN in each of its three oxidation states. The structure of the oxidized state of the wild-type flavodoxin at 20 A from LI. vulgar-is was used as a starting model for refinement. Diffraction experiments were conducted at low temperature (- 150°C) in order to maintain the oxidation state of int’erest throughout the intensity data collection. Crystals of flavodoxin, in the oxidized state. form yellow hipyramids by the standard hanging-drop method from 3.2 M-ammonium sulfate in 61 M-Tris. HCl buffer at pH 7.0 with protein concentrations ranging from 0.7% to (p9?/,. The reduced states of the crystals were achieved through the addition of sodium dithionite at pH 7.0 for the semiquinone (semi-reduced) and at pH 90 for the hydroquinone (fully reduced). Data sets consisting of one at room temperature (oxidized state) and three at low temperature (each oxidation state) were collected on a Nicolet P3F/Xentronics area at. 2.25 a detector X-ray diffractometer system. The four structures, hydroquinone resolution and all others at 1.9 A resolution, were refined by the restrained parameter leastR-values converged to 0.2 I squares progratn PROLSQ. The final crystallographic (hydroquinone), 0.20 (semiquinone). 0.20 (oxidized, low temperature), and 0.17 (oxidized, room temperature). The reduced st,ates of flavodoxin show a different conformation of the protein polypeptide chain (Asp61-Gly62) in the vicinity of NH(5) of the isoalloxazine group relative to the oxidized state. However, there are only slight conformational differences bet,wren the semiquinone and hydroquinone st)ates. In this report, structural comparisons of the three are made, with particular emphasis on the features that might be related to the difference in temperature of the diffraction data cxollections and differences in the oxidation stat,e of the FMN.

1. Introduction The electron-transfer mechanism associated with sulfa&reducing bacteria belonging to the genus I>esulfovl;brio has offered many opportunities to study structure/function relationships associated with biochemical redox chains. Although many redox carriers are known, the mechanism of electron transfer is still far from being completely clarified. This is, in part, due to a still incomplete understanding of the pathway for sulfate reduction. Flavodoxins contain one equivalent, of FMN. which is their only known prosthetic group, and t,hey lack transition metals such as iron commonly OOZd-2X36/91/050195

14 $03.OiI/O

found in ferredoxins. They are functionally equivalent, to the ferredoxins via their differences in redox pot*ent,ial (approx. -400 mV) between the semiquinone and hydroquinone states. It was first observed (Knight & Hardy, 1966) that clostIridial (nitrogenfixing) flavodoxin is synthesized only in iron-deficirnt media. Tn sulfate-reducing bacteria the pict,ure is somewhat more complex, since the relat,ive amount of ferredoxin or flavodoxin is dependent on the species and also according to growth conditions. The bacterium, Desu.lfovibrio vulgaris strain Hildenborough: synthesizes large amounts of flavodoxin but only small amounts of 195 8

1991 Acadrmic

l’rrus

I,imited

ferredoxin, even in high concentrations of iron (LeGall et al., 1979). Flavodoxins do not react dirert’ly with small molecules such as p.vridine nucleotides and their only known “sub&rates” are ot’her redox proteins. For example, they are reduced by NADPH and can couple NADPH oxidation to cytochrome c reduct)ion using NADP-ferredoxin reductase as a catalyst. In addition, fla.vodoxin also takes part in the phosphoroclastic cleavage of pyruvate,

(Iryst,al

determined for (Ludwig et nl., 1976) in its three oxidation states, for the oxidized state from Anacystis nidulans (Smith et al., 1983: Laudenbach et al., 1987), and for the oxidized state from II. vulyaris (Watenpaugh et al.. 1973). However, they all show different protein structure in the immediate vicinity of the isoalloxazine ring upon reduction. Because the protein structure modifies the oxidation/reduction potentials differently in the different tlavodoxins, further knowledge of the of how structure can lead to a better understanding t’hese potentials are modified. tlavodoxin

structures from

have

C’lostridium

been

MP

2. Experimental Methods (a) Crystallization

and reduction

experiments

The recombinant flavodoxin (Krey et al., 1988) used in this study differs from the wild-type only by the replacement of the N terminus proline by an alanine. Yellow bipyramidal crystals grow from 3.2 M-ammonium sulfate. @I M-Tris. HCl (pH 7.0) with protein concentrations ranging from 0.7:/, to @9% in space group P4,2,2 (Watenpaugh et al.. 1972). Crystallizations were carried out by the hanging-drop technique at room temperature. The average size of the crystals was about I.0 mm x The 0.7 mm x @5 mm after about 1 month of growth. reduction to the semiquinone was carried out’ by transferring the oxidized crystals from a coverslip to its well solution and adding slightly more than 1 equivalent, of solid sodium dithionitr. The reduction time at) pH 7.0 to the semiquinone ranged from 20 to 30 min. The reduction to the hydroquinone state required the pH of the well solution to be adjusted slowly to 90 before the addition of sodium dithionite. The kinetics of reoxidation of the semiquinone to the oxidized state is moderately slow. whereas the reoxidation from the hydroquinone to the semiquinone state is more rapid. It was not necessary to carry out the reductions in an oxygen-free environment as was needed for a similar experiment with the C’lostridium. because the reduced MI’ (Smith et al.. 1977) flavodoxin states were maintained by flash cooling and then kept at low temperature (- 150°C) throughout, the data collection, (b) (Crystal mounting The methods that were used for collecting data at cryogenic temperatures (- 150°C) are similar to t,hosr discussed by Hope (1988). The csrystal, along with some mother liquor, was transferred by pipette to a depression spot’ plate containing approximately 2 ml of Type A Cargille lens immersion oil (R. I’. Cargille Laboratories. Inc. Cedar Grove, N,J 07009). For the reduced states, the lens immersion oil was purged with nitrogen gas for several hours before use. The excess mother liquor was

removed from the crystal Ir)y (*aret’uI suc+ioll through it smaller capillary (approx. 02 mm). The cryst,al was th(lll mounted on the end of a glass fiber and transferred to t ht. diffractometer in the manner described by Hop? (l!M).

data sets were area det,ector s-vstem using graphite monochromatized CuKa radiation from a sealed tube source (50 kV, 35 mA). The detector was mounted on the 20 arm at IX cm from t,hr c.ryst,ai with swing angle settings at 0’. 20” and 30” to the incident X-ray beam. The data’sets were collect,ed in 0.20” s(.ans/ frame around omega wit,h exposure times ranging from 60 to I20 s (depending on the posit’ion of the 28 arm). The 1st, data set was collected at room t,rmperature using a stalldard capillary mounted crystal and the remaining :I data set’s were collected at low t)etnperaturr with crystals mounted on glass capillaries. The data sets wer(s then eva,luated and reduced with t,he XEK’CEK software package (Howard et al., 1987). 4 summary of statistics for t,he different data sets is shown in Table 1. Table 2 indicates t’hr differences in diffraction intensities between the various oxidation states and temprrat,urrs. X-ray collected

intensities of with a Nicolet

4 diffraction J’3F/Xentronics

(d) (‘rystallographic

r$nemvnt

et ~1. (1973) The structure described by Watenpaugh was used as a starting model for refinement of OXRTT with data from 8 t,o 2% A (1 a = 0.1 nm). Model rebuilding was carried out OII an Evans & Sutherland PS390 computer graphics system running FROM soft’ware (Jones, 1985: Pfliigrat’h rt al.. 1985) and using 2p0-r(‘, and rob-I/‘, electron density maps. All elec%ron refinement with the program density maps, after PROLSQ (Konnert & Hendrickson, 1980; Sheriff, l987), in order to reduce errant used Sim (1960) weighting calculated phases from Fourier syntheses. The refinement was completed after 123 cycles of restrained parameter least squares using the program PROLSQ with data from 8 to 1.9 A and 10 graphics rebuilding sessions. Bond lengths, bond angles and other parameters used to define standard geometry were based on a set compiled by Sielecki et al. (1979). The solvent atoms were added if they appeared in both the 2Pb- Fc and yO- ji’c maps and if there was at least 1 hydrogen bond to a protein donor or acceptor or another water molecule position previously determined. Occupancies and isotropic thermal parameters were refined in alternating cycles. The final model with 130 positions for solvent. has a crystallographic H-value of 017. The final roorrl~trrllprraturr model without the solvent) atoms was used as a starting model for rrtinrment with the low,-tenlI)rraturr data of the oxidized st,atr. The I st 4 chycles of rrfint:mrnt with data from 8 to 3.5 A were based on the model as a rigid body using the program CKYLSQ (Olthof-Hazekamp. 1987). Thr K-value dt~creasetl from (p56 t,o 049. The next 8 refinrment c.y~le:s with thr model partitionrd into rigid secondary structure groups (samtl range of data) decreased the m-value from 0.49 to @40. The data were then extended to 2.8 A. and 4 c*yclrs with (‘RYLSQ drc*reased the K-value to 0.36. An overall -__ 7 Abbreviations used: OXRT. oxidized state -room temperature; OXLT, oxidized state-low temperature; SQ, semiquinone state; HQ, hydroquinone state; n.m.r.. nuclear magnetic resonance: r.m.s.. root-mean-square.

Crystal

Structures

of a, Flavodoxin

197

Table 1 Crystal Space group Oxidation state c:rysta1 color Size (mm X mm x mm) llnit cell parameters

and intensity

data collection

P4,2,2 (tetragonal) OXRT OXLT Yellow Yellow 1~2Xo+Xo% 06x@6x~7 51.96

o = b (A)

13986 377,600 18,206

50.72 13904 357,683 17,705

a-190 15,055

cr-1.90 14,745

13,558

11.288 41,222 939 0.093

c (4

(Jrystal volume (A3) Molecular volume (A3)t Resolution range (A) No. possible refl. No. unique refl. collected No. refl. observed Z/u(Z) (final (11 A) &,!J

61,291 5.97 0066

t Molecule volume is calculated surface calculation.

summary

SQ Red/Purple l.OxO6x@7

HQ Pale yellow 06xti4xO.4

5144 139.62 369,445 17,731 00-190 15,439

51.36

139.38 367,663 17.748 co-225 13,516

13,507 60,332 316

0.100

from the molecular envelope obtained from a Connolly

9,313 .53,148 5-74

0080 (1983) dot

1 i I(z”‘)-zy) $ KS,, = !!!!i=’ N p,

zY

)

where N is the number of symmetry-related

reflections

thermal parameter of 20 8’ was held constant during the rigid-body refinements with CRYLSQ. This model was then used as a starting point for refinement with PROLSQ. The procedure for refinement is the same as that which was performed for the room-temperature model. The final model is based on 67 cycles of PROLSQ, data ranging from 8 to 1.9 8, 253 water molrcules. and gives an R-value of 0.199. The co-ordinates of the low-temperature oxidized model were employed to interpret the electron density for the semiquinone state and the co-ordinates of the semiquinone state were used as a starting point for refinement of the hydroquinone state. The initial models were refined

for 4 cycles with CRYLSQ as a rigid body in the same manner as previously mentioned for the low-temperature oxidized stat)e. Since a conformational change occurred along residues 61 t,o 63, reconstruction chain was carried out during a few

of the polypeptidr of the interactive

graphics sessions. The refinement of the semiquinone state with PROLSQ was completed in 128 cycles. data from 8 to 1.9 A. with 236 positions for water, and an R-value of 020. The refinement of the hydroquinone state was completed in a t,otal of 72 cycles. an R-value of @21 and

Table 2 Ayreement

statistics

between data sets R-factor1 o-35 0.39 0.58 042 w57 (k45

? Since the range of data for the hydroquinone form is only 2.25 A, t,he R-value was calculated only to that resolution.

f: R = C IIF,I-I~~ll/C(II~,l+I~,l~/~.

244 sites of water. A summary of the refinement statistics for each oxidation state is shown in Table 3. The atomic co-ordinates have been deposited with the Protein 1)ata Bank (Bernstein ct al., 1977).

3. Results and Discussion (a) Structurr i-\ ribbon representation (Priestle, 19X8) of the folding of flavodoxin is shown in Figure 1. The overall shape of the molecule can be described as an ablate spheroid with approximate dimensions of 25 Lh x 40 a x 40 A. The core of the molecule consists of a five-stranded parallel beta-sheet flanked on each side by a pa,ir of alpha-helices. There are 147 residues or 1133 non-hydrogen atoms in the structure with about one-third of the residues cont’ributing to the beta-sheet, one-third contributing to helices and the final third in turns with extended chain conformations. The regions of the molecule that are involved in the five beta-strands are: residues 2 to 10, 31 to 38, 51 to 60, 85 to 95, and 123 to 129. The four alpha-helical regions are: residues 13 to 28, 70 to 81: 103 to 115 and 132 to 148. The remainder of the molecule consists mostly of beta-turns: residues 10 to 13 and 27 to 30 are type 1. 37 to 40, 97 to 100 and 128 to 131 are t,ype ITT; 83 to 86 is type II; and 61 to 64 is type II’ (Table 4). The FMN lies mostly below the surface of the molecule. The structure of the hydrogen bonding wit)hin the alpha-helices and between t’he betastrands provides stability to the folding of the molecule. For example, a hydrogen bond exists between Asp95 and N97 in D. vulgaris as well as in A. nidulans, and Clostridium MP so it appears that Asp may be conserved at this site because it, stabi-

Table 3 Summary

of least-squares

Measured reflections Reflections used in refinement ( > 2~) is, Geometrical

re$nement

SQ

HQ

11,288 10,459

13,507 10,568

12,471 8,694

30.5 0.199

242 0203

0.213 I.82

OXK,T

OXLT

13,588 11,964 2.46 0170 conformityt

Target r~ Distances l-2 l-3 1-4 Planes Peptides Other Chiral volumes Non-bonded contacts Single torsion Possible H-bonds Other Thermal parameters l-2 (main-chain atoms) l-3 l-2 (side-chain atoms) I -3 Number of solvent atoms

parameters

Final model

0030 0.040 0050

0024 0.054 0050

0022 0.056 0057

0034 0060 0065

0.030 0.059 0.057

0040 0040 0300

0018 0.018 0.241

0.019 0.018 0.245

0.034 0.044 0337

0025 0025 0.341

0.500 omo 0500

0.188 0.320 @193

0.199 0.373 0.235

0.210 @353 0.219

0240 0356 0.304

I.500 3000 2QOO 4000

I.811 2.667 3638 5601 130

1.650 2.355 2.951 4.000 253

1.300 1.988 2.630 3981 236

1263 1.987 2.030 3059 244

t Final model values are the r.m.s. deviation

from the expected geometry

for the restraint class. factors between pairs of atoms. The notation 1-2, 1-3, l-4 refers to atoms related by a bond, bond angle, or dihedral angle, respectively. The target ~7is the inverse square root of the weight applied to this type of geometry restraint throughout the refinement. Reflection residuals are weighted by l/f(s)‘, where f(s) was the actual number used in the refinement and was obtained by approximation of 1/2(IF,,-B’J) (i.e. f(s) = 1/2W-~cI))

Thermal parameter values are the mean difference in isotropic temperature

lizes the backbone conformation by forming t’he hydrogen bond (Baker & Hubbard, 1984). The isoalloxazine ring appears to be planar and is buried between t’wo segments of polypeptide chain and the side groups Trp60 and Tyr98. The Tyr98 side-chain is almost coplanar with the flavin group (approx. 8’ between the plane normals) whereas Trp60 is about

Figure I). dyaris drawing.

1. Ribbon flavodoxin.

representation of’ backbone of The FMN is represented as a sticak

46” out of the plane. The hydrogen bonds between main-chain at’oms are shown in Figure 2. Hydrogen bonds were assumed to be favorable if distances of less than 32 a occurred between peptide nitrogen atoms and carbonyl oxygen atoms. (b) Ik’NLV bindiny

site

i,n the reduced states

The most, obvious differences between the oxidized and reduced struct’ures occur at, the peptide of Gly61-Asp62. The orientation of this peptide, t)he central unit in a, beta-turn, is changed significant,ly as a (sonsequence of reduction of the FMN bv sodium dithionitr. The shape of the density of Trp(iO-(ily61 -Asp6%Asp63 in the oxidized and semiquinone states is shown in Figure 3: tht carbonyl group of Gly61 point,s away from N(5) in the oxidized flavodoxin whereas the carbonyl group appears to point towards the N(j) in the semiqunone sta.t,e. The geometry at O(61 ) and the isoa,lloxazine S(5) favors a hydrogen bond involving thfl proton at Y(5) of the semiquinone state; the C) to N distance is 2.99 A and the 0 H- N atoms form an angle of 166”. Thus, the hydrogen bonding network along this peptide in t)he two st,ates is different’. +Smith et aZ. (1977) have carried out a similar st,udy of the Plostridium MI’ flavodoxin and they have ohserwtl

a similar

conformational

change

in t’hta

Crystal

Structures

199

of a Flavodoxin

Table 4 Torsion Sequence firs. LO-13 Kes. 27-31 Kes.

3741 44-47 Rrs. 45-48 lies. 61-65 ltes. 83-86 F&s. 97-100 Kes. 128-131

Res.

STTG ADAGY DASV GGLF GLFE GDDSI AEGR SYEY GDPR

angles of hetcr-turns

4’

v

d3

-98

-42 - 30 -2x

- 105 -81

-60 -56 -54 -62 70 -53 -90 -69

120 38 -69

126 157 11“2

- 56 55 -86 -83 70 -68 -68

Type of turn? I0 -8 -28 28 -8

TIT 11 1 II’ II HI III

-16 4 -37 -21

7 The dihedral angle of the 2 central residues (Res.) are from a tetrapeptide beta-turn (Richardson 1981).

cshain near N(5), with the carbonyl group flipping over to make the hydrogen bond, although the orientation of the polypeptide chain is quite different through this region. The scheme for the hydrogen bonding to the isoalloxazine ring and ribityl and phosphate sections of the FMN for the room-temperature model are shown in Figure 4. The hydrogen bonding network is similar to that which was proposed in the unrefined model by Watenpaugh et al. (1973). Moreover, t,he same network appears to remain present’ in each of the the only difference low-temperature structures, being the presence of additional water molecules polypeptide

near t,he rihit!;l

oxygen

atoms.

This

could

Residue Helix P-sheet

be the

structure ordering as a result of scheme of the hydrogen bonding around the isoalloxazine ring in the semiquinone state is shown in Figure 5. Wat15.5 consistent81y appears in all three oxidation states bridging O(4) of the FMN to O(62) and N( 100) with strong hydrogen bonding interactions to both FMN and t’he protein (Table 5). In the oxidized structure, hydrogen bonds are formed between 0(61) to Oa2 of Asp63 and N(62) t,o O(4) of t,he flavin: the distances are 2.9 and 3.1 8, respectively. In the reduced structures. the above hydrogen bonds of O(61) and N(62) are broken and only t,he new hydrogen bond of O(61) to N(5) is formed result. of the water

cooling.

The

--)

H-band

-

Coval

dir. bond

type E

Symmetry-related Solvent

Figure 2. Schematic diagram of hydrogen bonding betwern main-chain atoms (2; and 0) anti elements of’ secondary structure. Hydrogen bonds within a secondary structure element are not indicated. Heavy lines denote (-ovalent bonds (i.e. chain continuation) between main-chain atoms. Arrows indicate direction of hydrogen bond. Water bridging between 2 main-chain atoms indicated as numbered circles.

Figure 3. Stereoview of the FMN region and residues 60 to 63 and 93 to 95 of the (a) oxidized state at low temperature with its corresponding 2Fo- Fc density. Hydrogen bonding between N( 1) of FMN to N(95) of the apoprotein and O(4) of the FMN to N(62) of the apoprotein are indicated by the thin line; and (b) semiquinone state with its corresponding 2J’,,-- F, density. Hydrogen bonding between N(1) of FMN to N(95) of the apoprotein and N(5) of the FMN to O(61) of the apoprotein are indicated by the thin line.

I ,

202

Cl’. IITott et al.

PMS Asp62

N5 04

ox

W(.58)

Ml

r’.y)-”

x3

0% 0% NI O”% 0”:~ on:1 ( Y4

O”4 06 06 06 06 07 07 07

SC)

i-l(d

L!.!N

3.09

24x

25x

2.50

-‘a

:+oo 3.%” ?.YT

3. I x 4.11 “.%4 3. I !f 2.1”

3.1 ?I 315 %90 3% 263 f.82

MU 32 I %M 3.1 I ‘.gj(j 3.21 259 2.55

3.11

280

261 2.71 P%i

“W 2~99 “M

2.73 2424 2x9

%~5X %90 3 Id 273

2.55 “99 %9X 259

“.:l(j “%i :w!) PB I

247

%m

p;n

?4!) 24.x

34; 3.17

273

2.45

2.49 2.44

3~07 NX

3.37

2.51

3~%0 %%X 3..‘.) -A “.I% 294

regarding the hydroquinone structure is that. the dist’ance between Trp60 and Tyr98 is approximately 0.6 A further apart than in the semiquinonr and oxidized structures. This is probably due t)o changes in the isoalloxazine ring as a consequence ot reduction. (c) Fbavin planarity

The conformational change observed at Gly61 in the semiquinone state was also observed in the hydroquinone, hence the differences between t)he t,wo reduced states are small. If a hydrogen was present on N(1) in the hydroquinone state, then a conformational change would be expected to occur in residues 94 to 96. Vervoort et al. (1986) have observed, by nuclear magnetic resonance (n.m.r.), a similar hydrogen bond pattern between the oxidized state and the hydroquinone state; in addition, they have observed a strong hydrogen bond between N(5)-H and the apoprotein. From their studies, they believe the hydroquinone state is very similar to the semiquinone state (as based on the “first model” proposed by Watenpaugh et al. (1976)) and that FMN is ionized. Since there is a hydrogen bond between N( 1) and Asp95 in all three forms, it can be assumed that the N(1) is ionized in the hydroquinone state. A similar observation was made in the hydroquinone state of the Clostridium MI-’ &vodoxin (Ludwig et al., 1976). One final observation

OXL’I’

07 OX

04

Figure 5. Hydrogen bonding network in the vicinity of the FMN in the semiquinone state. Open circles indicate oxygen atoms, dots nitrogen atoms. Hydrogen bonds are indicated by broken lines.

OSHT

O((il) \vat I79 wat 155 N(W) O( 100) N(M): S( 102) N(95)$ O(59) \vat 181 \1’at209 wat 151 Ny2( 14) N(l.5) S(l4) OY(10) Ok’ X(1-k) N(I%) N(l I) OY(l%) N(H)

04

:‘\:

Apoprot.ein/solvent.

in the hydroquinone

stair

Several years ago, the regulation of redox potential was believed (Massey & Hemmerich. 1980) to be related to the bend along the N(5)-N( 19) axis of the flavin molecule. However, recent studies by n.m.r. 1984a.b; Franken et nl., 1984; (Moonen et al., Vervoort et al., 1986) have indicated that’ the FMN remains planar in all three oxidation states in solution and hence is not responsible for the very low redox potential of t’he semiquinone/hydroqmnone couple. They believe that t)he low potential is associated with the electrost’atic interaction between t’hr hydroquinone anion and the acidic residues in close proximity to the bound flavin. In the 1). wlgaris structure, the overall shape appears to be planar: a bending angle (angle between the plane normals) of 16.7” was obtained from the restrained least squares refinement in the hydroquinone form; however. some of the pyrimidine moiety atoms moved out ot the plane. A final rigid-group refinement rnaintaining planarity of each of the outer rings of the isoalloxazine group resulted in a better estimate of the non-planarity of the group. The resulting angle

Oystal

Structures of a Flavodoxin

203 ---~--

---7

---___ Main 2“I_ IO

chain

I

I

I

/

I

I

I

I

I

I

20

30

40

50

60

70

00

90

100

110

1

120

---i-I 30

---140

RKRLIVYGSllGNTEYTRETIRROL~~~GYEVDSRO~~SV~~GGLffGf~LVLLGCSlUGD~SIEL~~Df~PLfDSL~ETGR~GRKV~CfGCGDSSY~YfCG~VD~l~EKLKNLG~~lV~OGLRIDGDPR~~KOOlVGU~HNVRG~:

Residue

Figure 6. r.m.s. difference (A) in positions flavodoxin at room and low temperature.

Figure 7. A comparison of conformations indicated by the thin lines.

bet.ween main-chain

and side-chain

of Arg24 at room and low temprrat,urr.

atoms for thr oxidized

Appropriate

statta of

hJ.drogyn bonding is

87

148

148

1

GlU

Figure 8. A comparison of the conformations indicated by the thin lines

of Lys87 at room and low temperature.

(twist or bend) was 3.4”. Finally, the ring seems to remain nearly planar in all three oxidation states and the bending or twisting that occurs appears to he within the experimental error of the atomic positions. (d) Comparison

of the room-

oxidized

and

low-temperature

state

The principal reason for examining the oxidized state at low temperature were: (1) to determine if the protein crystals could survive the sudden

Appropriate

hydrogen bonding is

change in temperature required t’o maintain the reduced forms, (2) to ascertain the qualit,y of diffraction at low temperatures, and (3) to uncover benefits of data collection at cryogenic temperatures. To evaluate the merit of collecting data at low temperature, room- and low-temperature structures were inspected where low-temperature data collection had improved the quality or clarity of the electron density maps. These differences have shown improved ordering of some, but not all, of the side in the low-temperature model. Some chains examples are discussed below. A plot of the average

Crystal Structures of a Flavodoxin

Wat

406

205

Wat

Figure 9. Conformation of Glu42 at room and low temperature. ilt low temperature better ordered. The hydrogen bonds are denoted by the thin lines.

deviation in position between the two models at the two temperatures as a function of residue (Honzatko, 1986) is shown in Figure 6. The average r.m.s. differences for main-chain atoms is 026 A. while t’hat of side-chains is 043 A. The largest deviation between the two models occurs at Arg24. In the room-temperature model, the side-chain extends outward and then curls slightly back toward the main chain. The conformation of the side-chain is stabilized through a hydrogen bond network from NH1 and N” to 06’ of Asp28, and NH1 to Watl62. The low-temperature model has the side-chain extending straight out with hydrogen bonds stabilizing it through N” to

406

the carboxyl

group appears to be

06’ and 06’ of Asp28, NH1 to Wat,374. and NH2 Tao 06’ of Asp62 of a symmetry-related molecule. Another example of different conformation between the two models occurs at Lys87. The room-temperature model shows the side-chain extending straight out into the solvent channel with the formation of a salt bridge between N5 and gE’ and OE2 of Glul18; the low-temperature model shows the side-chain curling slightly back toward the main chain with the formation of a salt bridge between Nr and the C terminus. These conformational differences are shown in Figures 7 and 8. Some of the disordered side-chains in the room temperature model have become ordered as a result

206

Figure 10. Room and low-temperature t)emperature model.

II’. I17rrtt rt al.

conformations

of the cooling. In the room-temperature model, Glu42 extends out into the solvent’ channel (Fig. 9) wit)h no apparent density (at lo contour level, where 0 is estimated as the standard deviation from t’he mean density value of t’he map) beyond (1”. Tn the low-temperature model, the side-chain is more ordered, with density covering the carboxyl group with a hydrogen bond, between OE1 and Wat389 and Wat406, apparently stabilizing the conformation. Tn the case of Glu99 (Fig. lo), there is density in both models but the low-temperature model is in a slightly different posit,ion with a hydrogen bond

__-__

of Glu99 with the appropriate

between

hydrogen

--.----.---.--

bonds in t,he low-

OE2 and Wat221

and Wat508 and 0” with These are just two examples ot’ atI improvement in density fitking as a result of cooling

Wat279.

the crystals to cryogenic temperatures. Most of t’hr structure is well ordered at room temperature so the cooling did not produce a significant improvement. in fitting t’he model to density. A plot of AH as a function of residue between the two different models (Fig. 11) shows the expected trend of decreasing H-factor as a result of lowering of the temperat’ure. i.e. the average AB of the main chain between the low- and room-temperature models is 2.09 AZ; the

Crystal

Structures

207

of a Flavodoxin Main chain

-e

IO

20

I 30

I 40

I 50

I 60

/ 80

I 70 Residue

Figure oxidized

11. Difference state.

R-factor

in thermal differences

parameters are in A*.

(B) as a function

average AU of the side-chains (not shown) between the low- and room-temperature models is 2.02 A2. As the temperature was decreased, the cell parameter changes can produce slight repacking of the molecules. The cell parameter changes and related atomic shifts result in surprisingly large changes in diffraction intensities (Table 2) and a So/; decrease in cryst)al volume (Table 1). During the cooling process. the protein volume decreased by approximatelv 34,, (the molecular volume calculation was based’on t’he volume of the molecular envelope from a Connolly (1983) dot, surface calculation and a larger decrease was observed in the solvent volume i.e. since t)he crystals contain approximately 60 y/;, solvent. the change in solvent volume from cooling was t’hen about 7%. This difference in relative decrease of the volume of protein VerSUS solvent probably results in many protein crystals not being successfully cooled and in the increased mosaic> character in these experiment’s, The authors thank Dr Barry C. Finzel for the many useful discussions and software in generating the Figures in this manuscript,.

References Baker. E. h’. &. Hubbard, R. E. (1984). Hydrogen bonding in globular proteins. Progr. Biophys. Mol. Biol. 44. 97-I 79. Bernstein. F. C.. Koetzle, T. F., Williams. G. J. B., Meyer. E. F.. Brice. M. I>., Rogers, J. B., Kennard, 0.. Shimanouchi. T. & Tasumi, M. (1977). The protein data bank: a computer-based archival file for macromolecular structures. .I. Mol. Biol. 112. 5X-542.

I 90

, 100

I

I

I

HO

120

130

--Tm----

140

number

of residue

(room

t,emperature,

low temperature)

of the

C’onnolly, N. L. (1983). Solvent-accessible surfaces of proteins and nucleic acids. Sciencr, 221, 709- 713. Franken, H. I).. Riiterjans, H. & ?rZiiller. F. (1984). invesGgat*ion of Nuclear-magnetic-resonance ‘%-labeled flavins, free and bound to Megasphaera apoflavodoxin. Jk. .J. JSiochrm. 138. elsdmii 481-489. Honzatko. R. B. (1986). Automat,ed calculation of coordinate transformations for the superposition of protein structures. Acta Crystallogr. sect. A, 42, 172--l 78. Hope, H. (1988). Cryocrystallography of biological macaromolecules: a generally applicable mrt,hod. Ada Crystallogr. sect. B, 44. 22-26. Howard. A. J.. Gilliland, G. I,., Finzrl. K. (‘. 8r I’oulos, T. L. (1987). The use of an imaging proportional counter in macromolecular crystallography, -7. AJqd. Crystallogr. 20, 383-387. Jones. T. A. (1985). lnt’eractive wmputrr graphics: FR,ODO. Methods Enzymol. 115, 167-l 7 I. Knight, E.. ,Jr & Hardy, R. W. F. (1966). Tsolation and characteristics of flavodoxin from nitrogen-fixing Plostridium pasteurianum, -7. Bid. (~‘hrm 241, “732SP756. Konnert. J. H. & Hendrickson, \V. .\. (1980). Incorporation of stereochemical informat,ion into refinement. 111 ~‘omputing in cnrystallographic Crystallography (Diamond, R.. Ramaseshan, S. & K.. eds). pp. 13.01--13.25, Tndian Venkatesan, Academy of Sciences. Hangalore. Krry. (:. I).. T’anin, E. F. & Swenson. 11. P. (1988). Cloning, nwleotide sequence and rxprrssion of thtx flavodoxin gene from Draulfwibrio w,lgrtris (Hildmborough). J. Riol. Own. 263. 154X% 15443. Laudenbacah. 1). E.. Straw. 11‘. A.. I’attridgr. K. At. $ J,udwig. M. L. (1987). Sequence and &ructure of Aw,acystis nidulans flavodoxin: wmparisons with flavodoxins from other species. In Flavins CITL~ Fltwoprotrina (Edmondson, I). E & M!lc(‘ormi(