doi:10.1016/j.jmb.2006.04.054 J. Mol. Biol. (2006) 360, 80–89

A Proximal Arginine R206 Participates in Switching ofthe Bradyrhizobium japonicum FixL Oxygen Sensor

Marie-Alda Gilles-Gonzalez1⁎, Ana Isabel Caceres2

Eduardo Henrique Silva Sousa1, Diana Rose Tomchick1

Chad Aaron Brautigam1, Constancio Gonzalez2 and Mischa Machius1

1Department of Biochemistry,University of TexasSouthwestern Medical Center,5323 Harry Hines Boulevard,Dallas, TX 75390-9038, USA2Department of Biochemistry,Molecular Biology, andPhysiology, Faculty of Medicine,University of Valladolid,47005 Valladolid, Spain

E-mail address of the correspondimagg@biochem.swmed.edu

0022-2836/$ - see front matter © 2006 E

In oxygen-sensing PAS domains, a conserved polar residue on the proximalside of the heme cofactor, usually arginine or histidine, interacts alternatelywith the protein in the ‘‘on-state’’ or the heme edge in the ‘‘off-state’’ butdoes not contact the bound ligand directly. We assessed the contributions ofthis residue in Bradyrhizobium japonicum FixL by determining the effects ofan R206A substitution on the heme-PAS structure, ligand affinity, andregulatory capacity. The crystal structures of the unliganded forms of theR206A and wild-type BjFixL heme-PAS domains were similar, except for amore ruffled porphyrin ring in R206A BjFixL and a relaxation of the H214residue and heme propionate 7 due to their lost interactions. The oxygenaffinity of R206A BjFixL (Kd∼350 μM) was 2.5 times lower than that ofBjFixL, and this was due to a higher off-rate constant for the R206Avariant.The enzymatic activities of the unliganded ‘‘on-state’’ forms, either deoxy ormet-R206A BjFixL, were comparable to each other and slightly lower(twofold less) than those of the corresponding BjFixL species. The moststriking difference between the two proteins was in the enzymatic activitiesof the liganded ‘‘off-state’’ forms. In particular, saturation with a regulatoryligand (the FeIII form with cyanide) caused a >2000-fold inhibition of theBjFixL phosphorylation of BjFixJ, but a 140-fold inhibition of this catalyticactivity in R206A BjFixL. Thus, in oxygen-sensing PAS domains, theinteractions of polar residues with the heme edge couple the heme-bindingdomain to a transmitter during signal transduction.

© 2006 Elsevier Ltd. All rights reserved.

*Corresponding author

Keywords: FixJ; PAS domain; response regulator; sensor kinase; E. coli Dos

Introduction

In the heme-based sensor proteins, a heme-binding domain controls a neighboring transmitterbased on the ligation state of the iron atom.1 TheFixL protein histidine kinases are the prototypesfor a large category of these signal transducers thatsense O2 with a heme cofactor inside a globularsensory module called a PAS domain.2,3 Prokary-otic heme-PAS domains are typically 30% identicalin amino-acid sequence and show their greatestconservation in the regions that contact the hemedirectly.4,5 These are: the Fα helix with thehistidine residue that coordinates to the heme

ng author:

lsevier Ltd. All rights reserve

iron on the proximal side, the FG loop that extendsfrom this helix along the heme edge containing thepropionate chains, and two antiparallel β-strands(Gβ, Hβ) that continue from the FG loop and linethe ligand-binding pocket on the distal side of theheme.A basic residue occupies the Fα9 position in all the

known FixLs and in the Acetobacter xylinum PDEA1and Escherichia coli Dos cyclic diguanylic acidphosphodiesterases (Figure 1). This residue is R206in the Bradyrhizobium japonicum FixL.3 The placementof R206 near the boundary of the Fα helix and FGloop is ideal for communicating to this loop thechanges triggered in the heme by binding of ligands.When the BjFixL heme is unliganded, i.e. the ‘‘onstate’’, the R206 side-chain forms hydrogen bonds tothe FG-loop (Figure 1).3,6–8 When the heme is boundto O2 in the ferrous form or CN– in the ferric form,i.e. the ‘‘off-state’’, the R206 guanido group forms a

d.

Figure 1. Structural elements implicated in BjFixL regulation and their occurrence in proven heme-PAS proteins. (a) Acontrast of the crystal structures of the wild-type BjFixLH ‘‘on-state’’ (left) and ‘‘off-state’’ (right) (PDB files 1XJ3 and 1LT0;unliganded versus liganded, respectively). Secondary structural elements are denoted A–I; the heme iron is shown as adark-red sphere; the heme carbon atoms are in green, oxygen atoms are in red, and nitrogen atoms are in blue; theimportant side-chains around R206 are shown explicitly, with the hydrogen bonding interactions as dotted lines andimportant water molecules as red spheres (b) An alignment of the Fα helix and FG-loop sequences for the Bradyrhizobiumjaponicum FixL, Sinorhizobium meliloti FixL, Methanobacterium thermoautotrophicum Dos, Acetobacter xylinum PDEA1, andEscherichia coli Dos sensor proteins. The absolutely conserved Fα3 residue (H200 in BjFixL), or proximal histidine,coordinates the protein to the heme iron (3, 32). The conserved Gβ-2 arginine residue (R220 in BjFixL) strongly influencesaffinity and regulation (6); it alternately interacts with the heme propionate 7 in the unliganded on-state or bound cyanidein the off-state. The usually basic Fα9 residue (R206 in BjFixL) interacts alternately with the FG loop in the BjFixL on-stateor the heme propionate 6 in the off-state.

81Importance of a Heme-PAS Proximal Arginine

hydrogen bond to the heme propionate group 6(Figure 1).3,8 An ensuing displacement of the FG loopby about 2 Å, due to the changed polar interactionson binding of regulatory ligand, is postulated to playa regulatory role.3What aspects of the heme pockets in heme-PAS

proteins contribute to O2 affinity? In BjFixL pro-teins, the distal Gβ-2 arginine (R220) is the onlypolar residue sufficiently near to bound O2 tointeract with this ligand directly. This residuealternately interacts with the heme propionategroup 7 or a regulatory ligand: bound O2 in theferrous state or cyanide in the ferric state.3,8Removal of this guanido group from the full-lengthprotein by R220A substitution led to an approxi-mately tenfold lower O2 affinity compared to BjFixL(Kd∼1500 μM instead of the wild-type 140 μM).6 Anidentical effect of this residue on the affinity hassince been reported for two different FixL variants:

the isolated heme-binding domain of BjFixL (i.e.R220A BjFixLH compared to BjFixLH), and anotherFixL (i.e. the R214 substitutions in Sinorhizobiummeliloti FixLT compared to the wild-type).9,10 An-other arginine residue of BjFixL, the proximal Fα9residue (R206), responds to the changed heme-ironcoordination by alternately forming polar bonds toa heme propionate group or the FG-loop region.3,8The R206 guanido group does not form hydrogenbonds to bound O2 and exerts its effects, if any, viaother electrostatic interactions, e.g. with the hemeedge. Might the R206 residue modulate the O2affinity by influencing the heme electronic structureand the ability of the iron atom to change itscoordination?What factors govern transduction of signal, in

this case the inhibition of the kinase activity ofFixL by its ligand-bound heme-PAS domain? Invivo, one important function of BjFixL and its

82 Importance of a Heme-PAS Proximal Arginine

response-regulator partner, the transcription factorBjFixJ, is an adaptive symbiotic induction ofseveral high-affinity terminal oxidases before ni-trogen fixation can occur.11,12 These oxidases allowB. japonicum to continue respiring aerobically,despite the nearly anoxic conditions needed forfixing nitrogen. In vitro, the unliganded form ofBjFixL catalyzes the phosphorylation of BjFixJ withATP.6 Saturation of the ferrous form with O2inhibits the kinase activity.For BjFixL substitutions that dramatically lower

the O2 affinity, cyanide saturation of the ferric statecan nevertheless provide an accurate measure of theinhibition supplied by strongly regulatory ligands.6The oxy-form of BjFixL and many other O2-bindingheme proteins resembles a superoxide-bound ferricform (i.e. FeIIO2↔FeIIIO2

–), and thus the O2-boundferrous state is isoelectronic with the cyanide-boundferric state (FeIIICN–). A functional and structuralcorrespondence of the unliganded FeII and FeIIIforms (active) and the liganded FeIIO2 and FeIIICN–

forms (inactive) is well established for BjFixL.3,8 Inparticular, cyanide is known to induce the sameconformational changes in the ferric state that O2induces in the ferrous state.The cyanide inhibition of BjFixL is reduced about

20-fold by a R220A substitution.6 Nevertheless, thepresence of this residue cannot account for all theligand-induced inhibition seen for the wild-type(>2000-fold). Thus, these results have suggested thatR220 operates as part of a regulatory ensemble.6Here, we consider the importance of the R206residue to heme-PAS domain structure, ligandbinding, and signal transduction in BjFixL.

Results

Ligand binding

The R206 residue makes a small stabilizingcontribution to O2 binding in BjFixL, despite the

Figure 2. Affinity of ferrous R206A BjFixLH for O2. Thatm=101,325 Pa) of O2 at 25 °C results in 78% saturation ofregion is in the inset (top right). A Hill plot is shown (topR206A BjFixLH (4 μM) with O2 (50–1000 μM) at 25 °C, in 10changes in fractional saturation were calculated from multiplwild-type deoxy and oxy-BjFixLH as the bases; these data aequilibrium constant of 350 μM for binding of O2, and a n v

inability of its guanido group to interact directlywithO2. This is evident from the failure to saturate theferrous form with 1 atm (1 atm=101,325 Pa) of pureO2 (Figure 2). From aHill plot of the data obtained bytitrating the R206A mutant directly with O2, the Hillcoefficient n was 1.0 and indicated no cooperativity;the dissociation equilibrium constant (Kd) was350 μM and thus 2.5 times higher than the value forthe wild-type (Figure 2; Table 1). This slightly lowerO2 affinity of the R206Avariant was manifested as afaster rate of O2 dissociation (koff ∼29 s−1) comparedto the wild-type (koff ∼20 s−1) (Figure 3; Table 1).13The parameters for binding of CO did not differ

significantly for the R206A mutant and wild-typeproteins. The CO association rate constant was0.016 μM−1s−1 (Table 1).13 The off-rate constant,measured by displacing CO with NO, was 0.063 s−1and in the same range as the wild-type (Table 1).From these kinetic parameters, the equilibriumdissociation constant for binding of CO to theR206A variant is estimated to be around 4 μM, avalue similar to that of the wild-type.Cyanide or imidazole readily saturated the met

form of the R206A mutant. From the measuredkinetics of cyanide binding, the dissociation equi-librium constant is estimated to be about 2 μM, avalue similar to that of the wild-type (Table 1).6 Inparticular, the on-rate constant for binding ofcyanide (7.5×10−5 μM−1s−1) was in the samerange as the value obtained for the wild-type(Table 1).6 The cyanide off-rate constant of1.3×10−4 s−1, measured by displacing the cyanidewith imidazole, was also similar to that of the wild-type (Table 1).6 For binding of imidazole to themet-forms, the R206A mutant on-rate constant(kon=8.5 mM−1s−1) was in the same range as thatof the wild-type (Table 1).6

Crystal structure

To investigate any structural consequences ofthe R206A substitution, we determined the crystalstructure of this mutated heme-PAS domain in

e exposure of deoxy-R206A BjFixLH (blue) to 1 atm (1the protein (red). A tenfold expansion of the 500–700 nmleft) for the saturation data obtained by titrating deoxy-0 mM Tris–HCl (pH 8.0), 4 mM β-mercaptoethanol; thee linear regression of whole spectra, taking the spectra ofre consistent with the absorption spectra, a dissociationalue=1.

Table 1. Ligand-binding parameters of the R206A and wild-type BjFixL proteins at pH 8.0 and 25 °C

O2 CO

kon (μM−1s−1) koff (s−1) Kd (μM) kon (μM−1s−1) koff (s

−1) Kd (μM)

R206A BjFixLH 0.16 29 350 0.016 0.063 4.0 a

BjFixL (WT) b 0.14 20 142 0.008 0.04 4.8

CN– Imidazole

kon (μM−1s−1) koff (s−1) Kd (μM) kon (mM−1s−1)

R206A BjFixLH 7.5×10−5 1.3×10−4 1.7 a 8.5BjFixL (WT) c 1.1×10−4 1.2×10−4 0.94 4.9

Except where indicated, the parameters of R206A BjFixLH were measured directly. The average relative error is ±15%.a Calculated from kinetic constant.b Gilles-Gonzalez et al. (1994).13c Dunham et al. (2003).6

83Importance of a Heme-PAS Proximal Arginine

the unliganded form. Crystals were obtained forthe R206A BjFixLH under conditions similar tothose used for wild-type BjFixLH.3 A crystallo-graphic model was constructed and refined usingdata to a minimum Bragg spacing (dmin) of 2.30 Å(Table 2). Overall, the R206A BjFixLH structureresembled previously determined structures ofunliganded BjFixLH,3,7 with root-mean-squaredeviation (rmsd) for all Cα atoms of 0.28 Åcompared to unliganded BjFixLH (PDB entry1XJ3) and 0.55 Å compared to the cyanide-bound BjFixLH (PDB entry 1LT0). In both theunliganded and cyanomet forms of BjFixLH, R206interacts directly with the H214 residue of the FGloop as well as directly and indirectly (through a

Figure 3. Kinetic parameters for binding of O2 toferrous R206A BjFixLH at pH 8 and 25 °C. The mainFigure shows a kinetic trace of the dissociation of O2 at25 °C, as monitored after mixing, in a stopped-flowspectrometer, equal volumes of oxy-R206A BjFixLH and10 mM sodium dithionite in 100 mM Tris–HCl (pH 8.0).The dithionite instantaneously consumed any freed O2.The inset shows a determination of the association rateconstant kon=0.16 μM−1s−1 for binding of O2 from theslope of a linear plot of the dependence of kobs on ligandconcentration. Binding of O2 (100–600 μM) to deoxy-R206A BjFixLH (8 μM) was followed at 438 nm aftermixing, in a stopped-flow spectrometer, equal volumes ofligand and protein solutions in 100 mM Tris–HCl (pH 8.0).Each point represents three measurements of the apparentrate of association at 25 °C.

water molecule) with the heme propionate groups(Figure 1). The absence of these interactions fromR206A BjFixLH results in two immediate struc-tural consequences (Figure 4). First, the H214imidazole ring in R206A BjFixLH adopts aconformation that is rotated by about 90°compared to its position in unliganded BjFixLH.Thus, in the wild-type, the H214 side-chainconformation is presumably specifically held inits rotameric conformation through its interactionswith R206. Second, the heme propionate group 7rotates about 180° around the CAA-CBA bond(Figure 4). The porphyrin ring is also slightlymore curved than in wild-type BjFixLH. Specifi-cally, the carbon atoms of the A and C pyrrolerings move upwards and away from the proximalhistidine residue (H200) that is coordinated to theheme iron (Figure 4). This results in a maximumshift of about 0.35 Å for the C2A and C3C atoms.At the same time, the nitrogen atoms in rings B

Table 2. Data collection and refinement statistics

A. Data collectionSpace group R32Unit cell dimensions

a=b (Å) 127.12c (Å) 58.57

Resolution (Å) 27.07–2.30(2.34–2.30)

Completeness (%) 99.9 (100.0)Rmerge (%) 5.1 (62.4)I/σ(I) 36.2 (2.9)Multiplicity 7.4 (6.3)Wilson B-factor (Å2) 49.10

B. RefinementResolution (Å) 20.00–2.30No. reflections

Rwork 7565Rfree 611Rwork (%) 18.18Rfree (%) 21.85

Average B-factor (Å2) 43.52r.m.s. deviationsBond lengths (Å) 0.020Bond angles (deg.) 1.884

Numbers in parentheses refer to the highest-resolution shell.

Figure 4. Structure of R206A BjFixLH. (a) The overall structure, oriented as in Figure 1. The absence of the R206guanido group from unliganded BjFixLH causes H214 and the heme propionate 7 to adopt different rotamericconformations. Interactions are ruptured between the D212 carboxylate and the Fα helix, as well as between the H214imidazole and the heme edge. These structural changes appear not to affect the R220–propionate couple, which isconsidered important to the BjFixL on-state (6), and they do not lead to displacement of the FG loop from its position inwild-type BjFixLH. (b) A comparison of the heme structure in BjFixLH and R206A BjFixLH. The carbon atoms belongingto the heme in the wild-type enzyme are in grey, those in the R206A variant are in green; the central iron atom is in darkred; the pyrrole rings are labeled A–D, and propionate groups 6 and 7 are indicated. The absence of the R206 interactionsincreases the ruffling of the porphyrin ring.

84 Importance of a Heme-PAS Proximal Arginine

and D are shifted by about 0.2 Å in the samedirection.

Signal transduction

From the initial rates of the BjFixJ phosphorylationwith ATP, the key difference between the reactionscatalyzed by R206A BjFixL versus BjFixL was theirresponse to regulatory ligands (Figure 5; Table 3).The unliganded ferric R206A BjFixL protein hadabout one-half of the enzymatic turnover activity(43 h−1) of the wild-type BjFixL (81 h−1), suggestingthat the presence of the R206 residue normallyfavors the on-state. This is quite unlike the effect ofthe R220A substitution, which results in a proteinfive times more active than the wild-type in the on-state, suggesting that the R220 residue normallystabilizes the off-state.6 Saturation of the heme withcyanide caused a more than tenfold lower inhibitionfor R206A mutant compared to wild-type BjFixL(140-fold inhibition compared to >2000 fold, respec-

tively). These findings implicate the R206 residue inregulation.

Discussion

Ligand binding

Although it is the heme prosthetic group that bindsO2 in heme proteins, the surrounding proteinenvelope utterly governs the ability of the heme todo so. Indeed, heme and most of its derivatives willnot bind O2 reversibly at all, but instead becomeoxidized by O2 to the ferric state, in an essentiallyirreversible reaction.What are the processes bywhichproteins direct the reactivity of the heme? Hemeproteins arewell known to hinder or assistO2 bindingby controlling the formation and breakage of bondsbetween O2 and the protein.14 It is much lessrecognized and understood that heme proteins can

Figure 5. Influence of inhibitory ligand on the kinase activity of R206A BjFixL at pH 8.0 and 25 °C. Theautoradiograms show time-courses of FixL-catalyzed turnovers of BjFixJ to phospho-BjFixJ carried out for: 1.0 min,2.5 min, 5.0 min, 10 min, 20 min, and 30 min, in 20 mM Tris–HCl (pH 8) 50 mM KCl, 50 μM MnCl2, 5% (v/v)ethylene glycol, and 1 mM ATP/MgCl2, at 25 °C. Each reaction contained 7.8 pmol of FixL and 194 pmol of FixJ. Theleft side shows autoradiographs from reactions catalyzed by deoxy and aerobic ferrous wild-type BjFixL and R206ABjFixL (top), and reactions catalyzed by the met- and cyanomet-forms of wild-type BjFixL and R206A BjFixL(bottom). The activities of deoxy-WT and met-WT are identical; to show the activities of the inhibited species, theferric forms (met- and cyano-, below) were over-exposed compared to the ferrous forms (deoxy and aerobic, above).The right side shows a quantification of the phospho-BjFixJ produced by met-BjFixL (squares), met-R206A BjFixL(diamonds) and cyanomet-R206A BjFixL (inset: note the 100-fold smaller scale). The incomplete saturation of R206ABjFixL (58% deoxy:42% oxy) with O2 in air enhances the activity of the protein compared to the wild-type, but this ismainly because of the substantially higher deoxy fraction. Quantifications are therefore shown only for the ferricforms (Table 3).

85Importance of a Heme-PAS Proximal Arginine

achieve similar results by hindering or assisting thetransfer of electron density from O2 to the heme, aprocess that need not involve a direct interaction ofthe protein with O2. For example, compared withother heme proteins that reversibly bind O2, FixLproteins form a bond to O2 quite reluctantly. Anaverage deoxy-FixL molecule will wait 19 ms in air-saturated buffer before it finally combines with O2:about 50 times longer than a sperm-whale myoglobinmolecule would hesitate (Table 1).13 Once the Fe-O2bond is formed, however, it lasts almost as long as theanalogous bond in sperm-whalemyoglobin (35 versus50 ms) (Table 1).13 The polar interactions of bound O2with these two proteins fail to explain the vastlydifferent affinities of the very same heme embeddedin sperm-whale myoglobin versus that in FixL.

Our results show that the placement of a basicresidue at the Fα9 position permits BjFixL to achieve afine modulation of its affinity for O2 withoutmarshalling a direct interaction of the protein withthis ligand. The R206 guanido group at Fα9 is on theproximal side of the heme, well away from boundO2.This residue can reach the heme only at thepropionate 6 carboxylate group. Nevertheless, itsreplacement with alanine results in a protein with anO2 affinity measurably lower than that of the wild-type (Kd(O2)∼350 μM instead of 140 μM) (Table 1).Olson and colleagues have noted a similar drop inaffinity (three-to fivefold) on substituting the R45residue (CD3) of sperm-whale myoglobin with serineor glycine.15 Analogously to the FixL Fα9 residue, themyoglobin CD3 arginine forms a hydrogen bond

Table 3. Influence of cyanide on the kinase reactions ofR206A or wild-type BjFixL at pH 8.0 and 23 °C

Activity (h−1) Inhibition factor

BjFixL R206A BjFixL BjFixL R206A BjFixL

Met (FeIII) 81 43 1.0 1.0Cyanomet

(FeIIICN–)<0.03 0.30 >2700 140

The reaction measured was the catalytic turnover of the proteinsubstrate BjFixJ to the phospho-form in the presence of ATP.Inhibition factor is defined as the activity of the unliganded formdivided by the activity of a liganded form in the same oxidationstate. Quantification of ligand inhibition is most valid for theferric forms, where the activity of the unliganded state can becompared to the activity of the same oxidation state whileentirely saturated with cyanide.

86 Importance of a Heme-PAS Proximal Arginine

with the heme propionate group 6 as one of its mainpolar interactions. In a complementary study, Haya-shi and colleagues replaced each heme propionategroup ofmyoglobin, in turn,with amethyl group andreported similar effects on binding of O2 to myoglo-bins reconstituted with these altered hemes.16

Regulation

In contrast to the relatively small effect of the R206Asubstitution on ligand binding by BjFixL, thismutation strongly and precisely influences signaltransduction, i.e. the regulated kinase activity. Duringsignaling by FixL proteins, the O2-regulated step is anintermediate phosphorylation of the FixL enzyme,while it catalyzes a phosphoryl transfer from ATP tothe FixJ transcription factor.17–21 Interestingly, FixJenhances this initial reaction of FixL with ATP,although it does not chemically participate in thisstep.20,21 Therefore, studies of regulation are bestdone as examinations of the turnover of FixJ and ATP,rather than the non-catalytic phosphorylation of FixLwithATP in the absence of FixJ. An earlier study of theeffect of the R220 residue on signal transduction byBjFixL, led us to conclude that this arginine normallystabilizes the off-state.6 The enzymatic activity of thisform (deoxy or met) was five times higher in R220ABjFixL than in BjFixL, suggesting that the loss of thisguanido group shifts a pre-existing equilibrium ofactive and inactive enzyme molecules toward theactive state. The near failure to inhibit this state and anincreased planarity of the heme were also consistentwith an increased stability of the deoxy form in theR220A mutant.6 By contrast, we now see that theR206A BjFixL substitution slightly lowers the activityof the unliganded form, suggesting that the R206guanido group normally stabilizes the on-state(Figure 5; Table 3). Notably, cyanide caused ameasurable, 140-fold, inhibition of the ferric R206ABjFixL kinase, although it completely shut down thewild-type BjFixL, i.e. >2000-fold inhibition (i.e.activity of the inhibited form below the limit ofdetection of our current experiments).The heme-binding domain of the unliganded form

of R206A BjFixL revealed several structural differ-ences compared to the wild-type. Most notably, the

absence of interactions with the R206 side-chainbroke the interactions between the H214 imidazoleand the heme propionate groups, and led the H214residue and the heme propionate groyup 7 to relaxinto new conformations (Figure 4). Consequently, theporphyrin ring adopted a more ‘‘ruffled’’ conforma-tion (Figure 4). Thus, the direct and indirect polarinteractions of R206 in the wild-type protein appearto contribute to a flattening of the porphyrin ring.Attempts to prepare crystals of liganded forms of

the mutant by diffusing cyanide into crystals of themet form, as has been done for met-BjFixLH, causedthe crystals to disintegrate. Similarly, co-crystalliza-tion of the cyanomet form with ligands under thesame conditions that produced diffracting crystalsof cyanomet-BjFixLH, did not lead to crystalgrowth. Apparently, the structural changes accom-panying ligand binding in the R206A mutant areincompatible with the crystal lattice. Yet, liganded,off-state structures have been observed for wild-type BjFixLH under crystallization conditions sim-ilar to those used in this study.3,8 The structuralchanges associated with ligand binding are there-fore likely more pronounced in the R206A mutantthan in the wild-type. It seems that R206 serves as aclamp that stabilizes the porphyrin ring anddampens structural changes upon ligand binding.Overall, our results are consistent with the Fα9 and

Gβ-2 arginine-propionate couples constituting keycomplementary elements of the FixL switch. The Fα9arginine–propionate 6 couple,which forms in theO2-bound state, controls the equilibrium distribution ofactive and inactive enzyme for the off-state. Like-wise, the Gβ-2 arginine–propionate 7 couple, whichexists in the unliganded state, controls the propor-tions of active and inactive enzyme for the on-state.As the electrostatic charges nearest to the heme

center, it is not surprising that the propionate groupsare increasingly implicated in the propagation ofregulatory conformational changes of heme proteins.Horiuchi and colleagues have noted a 400-folddeceleration of the reduction rate compared to thewild-type, when a conserved arginine–propionatecouple in the Pseudomonas putida cytochromeP450cam was ruptured by replacing the argininewith cysteine.22 The reason for the slowed reduction,a diminished affinity (20-fold) for an electron-transfermediator protein, suggested that the arginine–propionate interactions were essential for a confor-mational change preceding electron transfer. Garciaand colleagues have posited, on the basis of multipleavailable P450 structures, that the breakage orformation of an arginine–propionate couple canopen or close a channel that links the heme pocketto the protein surface.23 Along similar lines, Fergu-son-Miller, Cukier, and their colleagues showed, fortheRhodobacter sphaeroides cytochrome c oxidase, thatthe interactions of the R481 and R482 residues withthe heme a3 propionate groups underpin a conservedhydrogen bond network extending to the dinuclearCuA center, a magnesium ion, and several subunit IIamino acid residues.24,25 The disruption of thisnetwork by an R482P substitution slowed the rate

87Importance of a Heme-PAS Proximal Arginine

of electron transfer from the CuA to heme a 2000-fold,in this case because of a sub-optimal orientation forthe interaction between cytochrome c and cyto-chrome c oxidase.24 The pH-dependence of theslow step in the internal reductions confirmed aninvolvement of the R481–D-ring propionate couple(pKa∼6.3) in electron transfer and proton pumping.26An allosteric enhancement of FixL autophosphoryla-tion by FixJ, and the inhibition of this reaction bybinding of O2 to the heme-PAS in FixL, presume thatwell-oriented interactions of FixL with FixJ, and ofthe FixL heme-PAS with the kinase, are necessary forsignal transduction.21,27 It is reasonable to supposethat, as in the examples described above, thearginine–propionate couples in BjFixL trigger con-formational changes that propagate to the heme-PASsurface for detection by the kinase domain and theBjFixJ substrate. For the broader class of heme-basedsensors,1 such a mechanism has the appeal of beingreadily transposable to alternative regulatory part-ners and enzymatic transmitters.

Materials and Methods

Genetic manipulations

DNAs corresponding to full-length BjfixL (codons 1–505) or its heme-PAS domain-encoding region (codons141–270) served as the templates for preparing R206ABjfixL and R206A BjfixLH by the QuickChange site-directed mutagenesis protocol (Stratagene). The mutationwas confirmed by sequencing both strands of eachDNA. E.coli strain TG1 was transformed with each final constructcontaining the gene or gene segment under tac promotercontrol on a plasmid conferring ampicillin-resistance.

Gene expression and protein purification

Procedures for overexpressing the R206A BjfixL andR206A BjfixLH were similar to those described for BjfixLand BjfixLH in E. coli strain TG1, except that the cells werecultured in a BIOFLO 3000 New Brunswick Scientificfermentor in 5 l of Terrific Broth.13 The cells wereharvested after 6 h of induction with 1 mM IPTG andlysed with a French pressure cell (Thermo Spectronic).Purifications were carried out as described.3,13,17

Preparation of the ferrous and ferric forms and theirliganded derivatives

The deoxy forms were obtained by reducing theproteins in an anaerobic glove box (Coy LaboratoryProducts Inc.) by either of two methods: (i) incubationfor more than 10 min with 10 mM dithiothreitol; (ii)treatment with an excess of sodium dithionite andimmediate removal of this reducing agent anaerobicallyby gel-filtration through a Sephadex-G25 column (Amer-sham Biosciences) equilibrated with buffer containing10 mM β-mercaptoethanol. The oxy forms were preparedby diluting a small aliquot of deoxy protein intophosphorylation buffer containing 4 mM dithiothreitoland letting the sample equilibrate with air for a fewminutes. The carbonmonoxy form was prepared bymixing the deoxy form with CO-saturated buffer. The

met forms were prepared by letting the proteins oxidize inair for >2 h or oxidizing them with equimolar levels ofpotassium ferricyanide. The cyanomet and imidazolemetforms were prepared by adding KCN (10 mM) orimidazole (200 mM) to the met forms.

Measurements of ligand binding

Static absorption spectra were measured with a Cary4000 UV-Vis spectrophotometer (Varian Analytical Instru-ments, Walnut Creek, CA) for the purified proteins in50 mM Tris–HCl (pH 8.0), 50 mMNaCl, 5% (v/v) glycerolat 23 °C in quartz cuvettes. Laser-flash photolysis andstopped-flow measurements were done with a LKS.60laser kinetic spectrometer fitted with a Pi-star stopped-flow drive unit (Applied Photophysics Ltd, Leatherhead,UK). For sample excitation, the LKS.60 spectrometer wascoupled to a Quantel Brilliant B Nd:YAG laser withsecond-harmonic generation. Data acquisition was pro-vided by an Agilent 54830B digital oscilloscope for fastmeasurements or a 12-bit ADC card within the instrumentworkstation for slow measurements.

Association rates

All protein or ligand solutions were in 100 mM Tris–HCl (pH 8.0), and amore than tenfold excess of ligandwasmixed with protein in all the experiments. Although theprotein concentrations are given as the total concentrationof monomer, FixL is entirely dimeric under those condi-tions (monomer-dimer Kd ∼0.1 μM).20 Ligand associationwas followed at 25 °C, at one or two wavelengths ofmaximum difference between the starting and the finalspecies. Specifically, ligand association was monitored at438 nm for O2, 425 nm and 441 nm for CO, 396 nm and427 nm for CN–, and 388 nm and 419 nm for imidazole.The deoxy forms (2–10 μM) were mixed rapidly in astopped-flow spectrometer with 20–400 μM O2 or 50–500 μM CO. Alternatively, the ferrous proteins wereequilibrated with 25–200 μM CO, and the rates ofrebinding were followed after flash photolysis. Ligandassociation to the ferric forms was followed after mixingeach protein (11 μM) with 1–10 mM KCN or 2.5–20 mMimidazole in the stopped-flow spectrometer. For eachchosen concentration of ligand, the apparent rate ofassociation was determined at least three times. Rateconstants were calculated by linear regression from plotsof kobs versus ligand concentration.

Dissociation rates

Rates of ligand dissociation were measured for proteinsin 100 mM Tris–HCl (pH 8.0) at 25 °C, as follows. Onevolume of oxy protein (14 μM ferrous protein equilibratedwith 500 μM O2) was mixed in a stopped-flow spectrom-eter with one volume of 10 mM sodium dithionite, so thatthe dithionite instantaneously consumed any free O2, andthe dissociation of O2 was followed. Dissociation of COwas monitored by mixing the carbonmonoxy protein(8 μM ferrous protein equilibrated with 50 μM CO) with alarge excess of NO (1.95 mM) in the stopped-flowspectrometer and following formation of the nitrosylform at 428 nm. Dissociation of cyanide was monitored bymixing the cyanomet protein (immediately after riddingthis protein of free CN– by gel-filtration) with 250 mMimidazole and following the formation of the imidazole-met form over 16 h with a Cary 4000 UV-Visiblespectrophotometer. The cyanide dissociation rates were

88 Importance of a Heme-PAS Proximal Arginine

calculated from the saturation changes determined frommultiple linear regression analysis of whole 350–700 nmabsorption spectra showing clear isosbestic points.

Equilibrium binding

The affinities of the proteins for O2 were measureddirectly by titrating the deoxy forms (4 μM) with 50–1000 μM O2, in buffer supplemented with 4 mM β-mercaptoethanol at 25 °C. For each protein, the dissoci-ation equilibrium constant for binding of O2 wasdetermined from saturation changes calculated by multi-ple linear regression analysis of whole spectra.

Phosphorylation assays

The catalytic phosphorylation of BjFixJ by BjFixL wasmeasured. These turnover assays were done by incubating1 μM full-length BjFixL kinase (either wild-type or R206A)with 25 μM BjFixJ in phosphorylation buffer (50 mM Tris–HCl (pH 8.0), 50 mM KCl, 50 μM MnCl2, 5.0% (v/v)ethylene glycol) and starting the reactions by introducing1 mMATP/MgCl2 (unlabeled ATP from Roche and [γ-32P]ATP fromAmersham Biosciences, specific activity 0.21 Ci/mmol) at 23 °C. After 1 min, 2.5 min, 5 min, 10 min,20 min, and 30 min, aliquots (10 μl) of the reaction weremixed with one-third volume of stop buffer (0.50 M Tris–HCl (pH 6.8), 40 mM EDTA, 4.0% (w/v) sodium dodecylsulfate, 0.20 M NaCl, 50% (v/v) glycerol, 2.0% (v/v) β-mercaptoethanol). For every reaction course, the status ofthe heme iron was verified from the 250–700 nmabsorption spectra before and after the assay. Theproducts were electrophoresed on 15% (w/v) polyacryl-amide gels.28 The phosphorylated protein in the dried gelswas quantified with a phosphorimager (Bio-Rad PersonalMolecular Imager FX).

Crystallization, data collection, structuredetermination and refinement

Crystals of the met-R206A BjFixLH mutant wereobtained at 4 °C via the hanging-drop, vapor-diffusionmethod by mixing equal amounts of protein (0.6–1.2 mMin 5.0 mM Tris–HCl, pH 8.0) and well solution (50 mMHepes (pH 7.5), 4.5 M NaCl, 5% (v/v) 2-methyl-2,4-pentanediol) and incubating the drop over 0.7 ml of wellsolution. Attempts to prepare crystals of liganded forms ofthe R206A mutant, for example by soaking met-R206ABjFixLH crystals with cyanide or by co-crystallization,were unsuccessful. Crystals were cryo-protected bystepwise transfer into well solution supplemented withup to 20% (v/v) glycerol and then flash-cooled in liquidpropane. The crystals exhibited the symmetry of spacegroup R32 with cell parameters of 127 Å×127 Å×58 Å.They contained one molecule of R206A BjFixLH moleculeper asymmetric unit and diffracted X-rays to a minimumBragg spacing, dmin, of 2.3 Å. Diffraction experiments werecarried out at 100 K at beamline 19ID (Advanced PhotonSource (APS), ArgonneNational Laboratory, Argonne, IL).Data sets were processed with the HKL2000 package.29

The met-R206A BjFixLH structure was determined bydifference Fourier techniques using the previously deter-mined structure of ferrous BjFixLH (PDB code 1lsw) as theinitial model. Refinement of the model was carried outwith the program Refmac5,30 of the CCP4 package,31 witha random subset of all data set aside for the calculation offree R factors. Manual adjustments to the model were

carried out with the program O.32 Data collection andrefinement statistics are given in Table 2.

Protein Data Bank accession code

The coordinates of R206A FixLH from B. japonicumweredeposited at the RCSB Protein Data Bank under PDBaccession code 2cmn.

Acknowledgements

We thank Kelly Krabill-Gerber for preparingmutant genes and Sandra Hill for providingexcellent technical assistance. The project wassupported by a Welch Foundation grant no. I-1575,by the National Research Initiative of the USDACooperative State Research, Education and Exten-sion Service, grant number 2002-35318-14039, andby a Spanish DGICYT grants BFI 2001-1713 and BFU2004-06394. Use of the Argonne National Laborato-ry Structural Biology Center beamlines at theAdvanced Photon Source was supported by the U.S. Department of Energy, Office of Energy Research,under contract no. W-31-109-ENG-38.

References

1. Gilles-Gonzalez, M. A. & Gonzalez, G. (2005). Heme-based sensors: defining characteristics, recent devel-opments, and regulatory hypotheses. J. Inorg. Biochem.99, 1–22.

2. Gilles-Gonzalez, M. A. & Gonzalez, G. (2004). Signaltransduction by heme-containing PAS-domain pro-teins. J. Appl. Physiol. 96, 774–783.

3. Gong, W., Hao, B., Mansy, S. S., Gonzalez, G., Gilles-Gonzalez, M. A. & Chan, M. K. (1998). Structure of abiological oxygen sensor: a new mechanism for heme-driven signal transduction. Proc. Natl Acad. Sci. USA,95, 15177–15182.

4. Delgado-Nixon, V. M., Gonzalez, G. & Gilles-Gonzalez, M. A. (2000). Dos, a heme-binding PASprotein from Escherichia coli is a direct oxygensensor. Biochemistry, 39, 2685–2691.

5. Chang, A. L., Tuckerman, J. R., Gonzalez, G., Mayer,R., Weinhouse, H., Volman, G. et al. (2001). Phospho-diesterase A1, a regulator of cellulose synthesis inAcetobacter xylinum, is a heme-based sensor. Biochem-istry, 40, 3420–3426.

6. Dunham, C. M., Dioum, E. M., Tuckerman, J. R.,Gonzalez, G., Scott, W. G. & Gilles-Gonzalez, M. A.(2003). A distal arginine in oxygen-sensing heme-PASdomains is essential to ligand binding, signal trans-duction, and structure. Biochemistry, 42, 7701–7708.

7. Key, J. & Moffat, K. (2005). Crystal structures of deoxyand CO-bound bjFixLH reveal details of ligandrecognition and signaling. Biochemistry, 44, 4627–4635.

8. Hao, B., Isaza, C., Arndt, J., Soltis, M. & Chan, M.K. (2002). Structure-based mechanism of O2 sensingand ligand discrimination by the FixL hemedomain of Bradyrhizobium japonicum. Biochemistry,41, 12952–12958.

9. Balland, V., Bouzhir-Sima, L., Kiger, L., Marden, M.C., Vos, M. H., Liebl, U. & Mattioli, T. A. (2005).

89Importance of a Heme-PAS Proximal Arginine

Role of arginine 220 in the oxygen sensor FixLfrom Bradyrhizobium japonicum. J. Biol. Chem. 280,15279–15288.

10. Tanaka, A., Nakamura, H., Shiro, Y. & Fujii, H. (2006).Roles of the heme distal residues of FixL in O2 sensing:a single convergent structure of the heme moiety isrelevant to the downregulation of kinase activity.Biochemistry, 45, 2515–2523.

11. Sciotti, M. A., Chanfon, A., Hennecke, H. & Fischer,H. M. (2003). Disparate oxygen responsiveness of tworegulatory cascades that control expression of symbi-otic genes in Bradyrhizobium japonicum. J. Bacteriol. 185,5639–5642.

12. Preisig, O., Zufferey, R., Thony-Meyer, L., Appleby,C. A. & Hennecke, H. (1996). A high-affinity cbb3-type cytochrome oxidase terminates the symbiosis-specific respiratory chain of Bradyrhizobium japoni-cum. J. Bacteriol. 178, 1532–1538.

13. Gilles-Gonzalez, M. A., Gonzalez, G., Perutz, M. F.,Kiger, L., Marden, M. C. & Poyart, C. (1994). Heme-based sensors, exemplified by the kinase FixL, are anew class of heme protein with distinctive ligandbinding and autoxidation. Biochemistry, 33, 8067–8073.

14. Olson, J. S. & Phillips, G. N. (1997). Myoglobindiscriminates between O2, NO, and CO by electro-static interactions with the bound ligand. J. Biol. Inorg.Chem. 2, 544–552.

15. Carver, T. E., Olson, J. S., Smerdon, S. J., Krzywda,S., Wilkinson, A. J., Gibson, Q. H. et al. (1991).Contributions of residue 45(CD3) and heme-6-propionate to the biomolecular and geminate re-combination reactions of myoglobin. Biochemistry, 30,4697–4705.

16. Hayashi, T., Matsuo, T., Hitomi, Y., Okawa, K., Suzuki,A., Shiro, Y. et al. (2002). Contribution of heme-propionate side-chains to structure and function ofmyoglobin: chemical approach by artificially createdprosthetic groups. J. Inorg. Biochem. 91, 94–100.

17. Gilles-Gonzalez, M. A., Ditta, G. S. & Helinski, D. R.(1991). A haemoprotein with kinase activity encodedby the oxygen sensor of Rhizobium meliloti.Nature, 350,170–172.

18. Gilles-Gonzalez, M. A. & Gonzalez, G. (1993).Regulation of the kinase activity of heme proteinFixL from the two-component system FixL/FixJ ofRhizobium meliloti. J. Biol. Chem. 268, 16293–16297.

19. Lois, A. F., Weinstein, M., Ditta, G. S. & Helinski, D. R.(1993). Autophosphorylation and phosphatase activ-ities of the oxygen-sensing protein FixL of Rhizobiummeliloti are coordinately regulated by oxygen. J. Biol.Chem. 268, 4370–4375.

20. Tuckerman, J. R., Gonzalez, G., Dioum, E. M. & Gilles-Gonzalez, M. A. (2002). Ligand and oxidation-state

specific regulation of the heme-based oxygen sensorFixL from Sinorhizobium meliloti. Biochemistry, 41,6170–6177.

21. Sousa, E. H. S., Gonzalez, G. & Gilles-Gonzalez,M.-A. (2005). Oxygen blocks reaction of the FixL/FixJ complex with ATP, but influences neither FixJnor ATP binding to FixL. Biochemistry, 44,15359–15365.

22. Koga, H., Sagara, Y., Yaoi, T., Tsujimura, M., Naka-mura, K., Sekimizu, K. et al. (1993). Essential role of theArg 112 residue of cytochrome P450cam for electrontransfer from reduced putidaredoxin. Fed. Eur. Bio-chem. Soc. 331, 109–112.

23. Oprea, T. I., Hummer, G. & Garcia, A. E. (1997).Identification of a functional water channel incytochrome P450 enzymes. Proc. Natl Acad. Sci. USA,94, 2133–2138.

24. Qian, J., Mills, D. A., Geren, L., Wang, K., Hoganson,C. W., Schmidt, B. et al. (2004). Role of the conservedarginine pair in proton and electron transfer incytochrome c oxidase. Biochemistry, 43, 5748–5756.

25. Seibold, S. A., Mills, D. A., Ferguson-Miller, S. &Cukier, R. I. (2005). Water chain formation andpossible proton pumping routes in Rhodobacter sphaer-oides cytochrome c oxidase: a molecular dynamicscomparison of the wild-type and R481K mutant.Biochemistry, 44, 10475–10485.

26. Branden, G., Branden, M., Schmidt, B., Mills, D. A.,Ferguson-Miller, S. & Brzezinski, P. (2005). Theprotonation state of a heme propionate controlselectron transfer in cytochrome c oxidase. Biochemis-try, 44, 10466–10474.

27. Tuckerman, J.R.,Gonzalez,G.&Gilles-Gonzalez,M.A.(2001). Complexation precedes phosphorylation fortwo-component regulatory system FixL/FixJ of Sinor-hizobiummeliloti. J. Mol. Biol. 308, 449–455.

28. Laemmli, U. K. (1970). Cleavage of structural proteinsduring the assembly of the head of bacteriophage T4.Nature, 227, 680–685.

29. Otwinowski, Z. & Minor, W. (1997). Processing of X-ray diffraction data collected in oscillation mode.Methods Enzymol. 276, 307–326.

30. Murshudov, G. N. (1997). Refinement of macromo-lecular structures by the maximum-likelihood meth-od. Acta Crystallog. sect. D, 53, 240–255.

31. Collaborative Computational Project, Number 4.(1994). The CCP4 suite: programs for protein crystal-lography. Acta Crystallog. sect. D, 50, 760–763.

32. Jones, T. A., Zou, J. Y., Cowan, S. W. & Kjeldgaard, M.(1991). Improved methods for building proteinmodels in electron density maps and the location oferrors in these models. Acta Crystallog. sect. A, 47,110–119.

Edited by R. Huber

(Received 13 February 2006; received in revised form 20 April 2006; accepted 25 April 2006)Available online 11 May 2006