Inhibition of the FixL Sensor Kinase by the FixT Protein inSinorhizobium meliloti *

Nitrogen fixation in symbiotic rhizobia is subject to multiple levels of gene regulation. In Sinorhizobium meliloti, the alfalfa symbiont, the FixLJ two-component regulatory system plays a major role in inducing nitrogen fixation and respiration gene expression in response to the low ambient O2 concentration of the nodule. Here we report on the mode of action of the FixT protein, a recently identified repressor of nitrogen fixation gene expression in S. meliloti. First, we provide evidence that FixT prevents transcription of the intermediate key regulatory genes nifA and fixK by counteracting the activity of the FixLJ two-component system under otherwise inducing microoxic conditions. Second, we demonstrate that FixT acts as an inhibitor of the sensor hemoprotein kinase FixL, preventing the production or the accumulation of its phosphorylated form. FixT is thus a new example of a regulatory protein that blocks signal transduction in two-component systems at the level of the sensor kinase.

Nitrogen fixation in symbiotic rhizobia is subject to multiple levels of gene regulation. In Sinorhizobium meliloti, the alfalfa symbiont, the FixLJ two-component regulatory system plays a major role in inducing nitrogen fixation and respiration gene expression in response to the low ambient O 2 concentration of the nodule. Here we report on the mode of action of the FixT protein, a recently identified repressor of nitrogen fixation gene expression in S. meliloti. First, we provide evidence that FixT prevents transcription of the intermediate key regulatory genes nifA and fixK by counteracting the activity of the FixLJ two-component system under otherwise inducing microoxic conditions. Second, we demonstrate that FixT acts as an inhibitor of the sensor hemoprotein kinase FixL, preventing the production or the accumulation of its phosphorylated form.

FixT is thus a new example of a regulatory protein that blocks signal transduction in two-component systems at the level of the sensor kinase.
In legumes, nitrogen fixation takes place in a specialized organ, the nodule, that rhizobia elicit on the roots of their host plants. Within the nodule, rhizobia acquire the capacity to reduce N 2 to ammonia that is assimilated by the plant cells. In the Rhizobium-legume symbiosis, as in all nitrogen fixing systems that have been studied so far, nitrogen fixation is tightly regulated, primarily but not uniquely, at the transcriptional level. In Sinorhizobium meliloti (formerly Rhizobium meliloti), the alfalfa symbiont, both positive and negative control of nitrogen fixation gene expression have been described, although positive control is by far the best understood. Positive control is exerted by a two-component regulatory system, FixLJ, that senses the microoxic conditions that prevail inside the nodule (1,2). Indeed, the legume nodule is a microoxic environment whose dissolved O 2 concentration is in the 10 -30 nM range. Under microoxic (or anoxic) conditions, the sensor hemoprotein kinase FixL autophosphorylates from ATP on a conserved histidine residue (His 285 ) (3) and subsequently transfers its phosphate to one of the conserved aspartate residues (Asp 54 ) of the regulator protein FixJ (4). Phosphorylation of FixJ dramatically enhances its affinity for the promoters of two intermediate regulatory genes, nifA and fixK, that control expression of genes involved in the biosynthesis of nitrogenase and of an oxidase complex with high affinity for oxygen, respectively (5)(6)(7).
Early observations indicated that expression of nifA or fixK was enhanced in a fixK mutant (8). More recently, it was demonstrated that repression by fixK actually involved activation of a repressor gene, called fixT (9), located downstream of the fixLJ operon (Fig. 1A). We now provide evidence that the FixT protein prevents transcription of nifA and fixK by counteracting the activity of the master FixLJ two-component system under microoxic conditions. We further demonstrate that FixT behaves as an inhibitor of the FixL sensor kinase, preventing synthesis or accumulation of its phosphorylated form. Biological implications of this finding are discussed with respect to the regulation of symbiotic nitrogen fixation as well as in terms of signal processing by two-component regulatory systems in general.

EXPERIMENTAL PROCEDURES
Bacterial Strains-GMI939, -940, and -941 are derivatives of GMI211 (lac Sm r ), constructed as described in Ref. 9. Briefly, GMI939 was derived from GMI211 by site-directed inactivation of the Ϫ10 promoter region of the fixT 1 gene. GMI940 carries an ϳ5-kilobase pair deletion that removes the fixT 2 gene as well as the adjacent fixK and fixNOQP genes. GMI941 is a derivative of GMI940, in which the 5-kilobase pair deletion has been transduced into GMI939, using a phleomycin resistance marker associated with the deletion (9).
Northern Blot Analysis of fixK Expression-RNAs were extracted from 30-ml cultures (A 600 ϭ 0.4) in M9 medium of cells incubated under microaerobic conditions (2% oxygen) for 4 h (10). Bacterial cells were collected by centrifugation and incubated for 10 min at 65°C in 2 ml of prewarmed lysis solution (1.4% SDS, 4 mM EDTA, 75 g of proteinase K). Proteins were precipitated by adding 1 ml of NaCl 5 M at 4°C. Nucleic acids were precipitated from the supernatant by the addition of 1 volume of isopropyl alcohol, and the pellet was resuspended in nuclease-free water. DNA was eliminated by the addition of 7.5 units of fast protein liquid chromatography-pure RNase-free DNase I (Amersham Pharmacia Biotech). RNAs were further extracted with phenol/chloroform/isoamyl alcohol and chloroform/isoamyl alcohol and then precipitated with ethanol. The RNA pellet was washed with 70% ethanol and resuspended in nuclease-free water. RNAs were denaturated for 10 min at 75°C before loading on a 2.2 M formaldehyde, 1.5% agarose gel. RNAs were electrotransferred to a Hybond N nylon membrane (Amersham Pharmacia Biotech) and probed with a 32 P-labeled BglII-EcoRV fixK internal fragment. Equal loading of the different lanes in Fig. 1 was ensured by hybridization of the membrane with a hemA probe. For the purpose of quantification, a range of appropriate dilutions of the various RNA preparations was transferred to a Nylon membrane, and the hybridization signals were quantified on a phosphorus screen apparatus (Fuji).
Reverse Transcriptase-PCR 1 Analysis of nifA Expression-Microaerobic induction of nifA was achieved using the stoppered tube assay procedure (11). RNAs were purified as described above and quantitated by measuring the absorbance at 260 nm. The absence of contaminating DNA in the preparation was ensured by PCR amplification in the * This work was supported in part by European Union Biotech Program Grant BIOT4-CT97-2319 and by the Action Incitative Programmée-Institut National de la Recherche Agronomigue Microbiologie program. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18  absence of reverse transcriptase.
First-strand cDNA synthesis took place in a 17-l reaction volume. 1 l of total bacterial RNA (100 ng), 7.1 l of diethyl pyrocarbonatetreated H 2 O, and 1 l (50 ng) of the nifA or hemA reverse primer were heated at 70°C for 10 min and quickly chilled on ice. After a brief centrifugation, 3.4 l of 5ϫ buffer (250 mM Tris-HCl pH 8.3, 375 mM KCl, 15 mM MgCl 2 ), 2 l of 0.1 M dithiothreitol, and 1 l of a 25 mM dNTP solution were added and incubated at 42°C for 2 min. After the addition of 1 l (200 units) of Moloney murine leukemia virus reverse transcriptase (Life Technologies, Inc.) and thorough mixing by pipetting up and down, the reaction tube was incubated for 50 min at 42°C. Reverse transcriptase was inactivated by heating at 95°C for 5 min, since we have observed that heating at 70°C did not fully inactivate the enzyme.
For PCR amplification, 2.5 l of 10ϫ PCR buffer (200 mM Tris-HCl, pH 8.4, 500 mM KCl), 1 l of 50 mM MgCl 2, 1 l of dNTP mix (25 mM), 1 l of reverse primer (50 ng), 1 l of sense primer (50 ng), 0.5 l (2.5 units) of Taq DNA polymerase (Life Technologies, Inc.) were added to the first strand synthesis reaction mix. After gentle mixing, the reaction mixture was layered with one drop of silicone oil, heated to 94°C for 3 min, and submitted to 25 cycles of the following sequence (94°C for 30 s, 50°C for 30 s, 72°C for 1 min) and one cycle at 72°C for 10 min. Reverse transcriptase-PCR products were electrophoresed on a 2% agarose gel, blotted onto a nylon membrane (Biodyne A Transfer membrane; Pall), and probed with a 32 P-labeled probe internal to nifA or hemA genes.
Purification of FixT-FixT was essentially purified by affinity chromatography of a fusion protein between FixT and MalE, the maltosebinding protein (Biolabs). DNA corresponding to the putative fixT open reading frame was generated by PCR (Expand TM Long Template PCR System; Roche Molecular Biochemicals) using the following two oligonucleotides: 5Ј-CCCAAGCTTGTTGGATGGGAAAACC and 5Ј-CCG-GAATTCACTAAGGGAAATTCA as primers. These primer oligonucleotides carried HindIII and EcoRI sites (underlined) that were used to clone the fragment into a pBluescript KSϩ vector (Stratagene) for verification of the sequence generated by PCR. The EcoRI-HindIII fragment was subsequently recloned between the EcoRI and XmnI sites of pMal-c2. The resulting plasmid (pMF116) was transformed into the TB1 strain of Escherichia coli.
MalE-FixT purification was done essentially as described by Biolabs Inc. 3 liters of complete LB medium supplemented with 0.2% glucose and ampicillin (50 g/ml) were inoculated with 30 ml of an overnight culture of TB1(pMF116). When the A 600 of the culture reached about 0.5, isopropyl-1-thio-␤-D-galactopyranoside (2 mM) was added, and the culture was incubated for an additional 2 h at 30°C. Cells were collected by centrifugation, washed, resuspended in C buffer (20 mM Tris-HCl, pH 7.4, 200 mM NaCl, 1 mM EDTA, 1 mM dithiothreitol), and finally disrupted by sonication. The crude cell extract (15 ml) was clarified by centrifugation, diluted to 200 ml with C buffer, and loaded on top of a 40-ml amylose-agarose column (Biolabs). After extensive washing (10 volumes) of the column with C buffer, bound MalE-FixT protein was eluted with C buffer supplemented with 10 mM maltose. Fractions containing MalE-FixT were screened by SDS-PAGE using an Amersham Pharmacia Biotech Phast Gel System apparatus. Fractions containing pure MalE-FixT were pooled and concentrated by centrifugation over Centripep 10 devices (Amicon). Fractions were stored at Ϫ80°C in the presence of 5% glycerol before use. Quantification of the protein was done by Coomassie Blue staining of protein separated on a SDS-acrylamide gel as compared with standards. The purity of the preparation was estimated to be 80%.
Part of the above MalE-FixT protein (20 mg) was incubated with Xa factor protease (200 g, Roche Molecular Biochemicals) for 2 h at room temperature. The protein mix was then sieved twice on a Sephacryl S200-HR column (Amersham Pharmacia Biotech). Fractions containing pure FixT were pooled, concentrated, and stored at Ϫ80°C in the presence of 5% glycerol. The purity of this FixT preparation was estimated to be 90% from inspection of a Coomassie Blue-stained SDSpolyacrylamide gel.
Phosphorylation Assays of FixL and FixJ and Transcription Assays-The FixL protein used in this work, FixL122, is a truncation of the full-length FixL protein that lacks the N-terminal part that normally anchors the protein to the membrane (12). FixL122 and FixJ were purified as described previously (5).
Phosphorylation assays were essentially as described (5) except that [␥-33 P]ATP (0.4 mM) was used as labeling source (final specific activity ϳ5000 dpm/pmol). All reactions were at 28°C. Phosphorylation reactions were loaded on SDS-10% polyacrylamide gels. When indicated, reactions were performed under an anaerobic hood facility. Dried gels were autoradiographed at room temperature. When required, quantification was obtained by liquid scintillation counting of excised band gels.
Transcription assays were essentially as described (5). The assay monitors the production of a 370-nucleotide-long RNA from supercoiled pJMR300 plasmid, a pTE103 derivative that contains the 5Ј-end of the fixK gene bracketed between two strong terminators of transcription. All reactions were performed at 28°C. Three tubes containing FixJ (6 M) and FixL122 (6 M) proteins were incubated under anoxic conditions in the presence of 0.4 mM ATP for 20 min. In two of the tubes, FixT (9 M) or buffer (control) was added prior to ATP addition. In the third tube, FixT was added only after 10 min. Formation of closed complexes was allowed for 10 min after the addition of E. coli RNA polymerase-70 holoenzyme (600 nM) and 12.5 nM supercoiled DNA template. This step was also performed under anoxic conditions, since closed complexes are unstable and dissociate in the absence of phosphorylated FixJ protein.
Isomerization of closed complexes into open complexes and elongation of RNAs was initiated by adding UTP and GTP (0.4 mM) and 0.1 mM [␣-32 P]CTP (2 Ci/mmol). Heparin was also added, so that the assay was a single-round transcription assay.
In Fig. 2B, FixT was added after phosphorylation of FixJ by acetylphosphate for 1 h at 25°C (5). The transcription assay was then run in the presence of FixT as described above. Alternatively, the FixJ-P preparation was run over a Centricon TM 3000 microconcentrator (Amicon) in order to eliminate acetylphosphate before adding FixT.
Purification of FixL-Phosphate-600 l of FixL122 protein (6 M) were allowed to autophosphorylate for 60 min under standard conditions (see above). The reaction mix was loaded on top of a G25 Sephadex sieving column. Fractions containing phosphorylated (radioactive) FixL122 were individually screened by TLC (polyethyleneimine-cellulose, Schleicher & Schuell) for the presence of contaminating ATP, ADP, and P i using 0.75 M potassium dihydrogenophosphate (pH 3.5) as a buffer. Pure fractions were pooled, concentrated at 4°C, and used immediately.
Phosphotransfer Assays-Purified FixL122-phosphate (6 M; see above) and FixJ (6 M) were incubated at 28°C in standard phosphorylation buffer under aerobic conditions. FixT action was tested by preincubating FixT and FixL122-phosphate before the FixJ addition (t ϭ 0). Phosphotransfer was allowed to proceed for 5, 10, or 15 mn before the addition of SDS-PAGE loading buffer.
Design and Purification of the FixLC Protein-The C-terminal part of FixL (from amino acid position 266 to the end of the protein, position 505) was cloned into a pMalc2 vector to obtain a fusion protein between FixLC and MalE. DNA fragment corresponding to FixLC was purified from pDPD147 plasmid (12). pDPD147 was first restricted by NheI, treated with Klenow polymerase, and restricted by BamHI. The fragment of interest was purified on the gel and cloned into XmnI-and BamHI-restricted pMalc2 vector. The resulting plasmid (pAM266L) was transformed into the TB1 strain of E. coli, and the fixLC insert was checked by sequencing. MalE-FixLC purification was done as described above for FixT. Fractions containing the MalE-FixLC fusion protein were pooled and treated with Xa factor protease. The cleaved preparation was concentrated and stored at Ϫ80°C in the presence of 5% glycerol. Its purity was estimated to be ϳ50%.
Phosphorylation Assays for KinA, DegS, and NtrB Proteins-Purified proteins were kindly provided by Dr. Tarek Mzadek (Institut Pasteur, Paris, France). Standard phosphorylation assays of these proteins were as described by Dahl et al. (13). Briefly, phosphorylation took place at room temperature in a buffer (100 mM Tris-HCl, pH 8.0, 200 mM KCl, 4 mM CaCl 2 , 4 mM MgCl 2, 0.5 mM dithiothreitol, 0.1 mM EDTA, 5% (v/v) glycerol). Phosphorylation was initiated by adding 5 Ci of [␥-33 P]ATP and 2.5 M ATP. Reactions were stopped in SDS-loading buffer after 5 min of phosphorylation. FixT action was tested after preincubating FixT with KinA, NtrB, or DegS for 5 min prior to ATP addition. Alternatively, KinA, DegS, and NtrB were phosphorylated under the conditions described above for FixL.

RESULTS
fixT Lowers fixK and nifA Transcript Levels in Vivo-fixT is a duplicated gene on the pSym-a megaplasmid of S. meliloti. One copy of fixT, that we shall call fixT 1 , lies between the regulatory fixLJ and fixK genes (Fig. 1A) and is transcribed from a fixK-dependent promoter (9). The second copy of fixT, fixT 2 , is located ϳ260 kilobase pairs from fixT 1 in a reiterated region that includes a second copy of fixK as well as of the fixNOQP operon (14). nifA instead is unlinked to fixT genes.
Northern blot analysis of RNAs extracted from fixT S. meliloti mutant strains and congenic controls incubated under microoxic conditions showed that the levels of fixK transcript were enhanced in fixT mutants (Fig. 1B). Selective inactivation of fixT 1 (9) led to a 10-fold increase in the level of fixK RNA as compared with the wild-type GMI211 strain (lane 4). Similarly, a strain deleted for fixT 2 as well as for the adjacent fixK and fixN genes displayed an increase in the level of fixK RNA (lane 2) as compared with GMI211. For a reason that we ignore, this effect was, however, smaller than the one observed with the fixT 1 mutant. The combination of the two sets of mutations in the same strain (GMI941, lane 3) resulted in an ϳ50 -60-fold increase in the level of fixK transcript as compared with the wild-type control strain.
We examined the effect of fixT on nifA expression by reverse transcriptase-PCR, since we could not obtain consistent Northern blot hybridization patterns with a nifA probe. Wild-type (GMI211) and fixT (GMI941) S. meliloti cells were incubated under nearly anoxic conditions in order to maximize nifA expression (11). Comparison of the nifA-specific amplification products from extracted RNAs showed higher levels of nifA transcript in the double fixT mutant as compared with the wild-type strain (Fig. 1C, lanes 1 and 2). Expression of the oxygen-insensitive S. meliloti hemA gene (8) was unaffected by mutations in fixT (Fig. 1C, lanes 3 and 4).
These data suggested that fixT affected the synthesis or the stability of fixK and nifA transcripts. We took advantage of the availability of an in vitro transcription assay for these genes to discriminate between these possibilities.
Purified FixT Protein Blocks fixK Transcription in Vitro-Previous genetic evidence indicated that fixT encoded a repressor protein (9). The FixT protein, encoded by fixT 1 , was overproduced in Escherichia coli as a fusion protein with MalE, the maltose-binding protein, and purified by affinity chromatogra-phy over an amylose column (see "Experimental Procedures").
In the experiments described below, either the MalE-FixT protein or the cleaved and purified FixT protein has been used and has led to similar results.
A transcription assay for fixK was previously set up (5) that relies on the ability of phosphorylated FixJ protein to drive fixK transcription in vitro, at the genuine promoter, in the presence of E. coli 70-RNA polymerase holoenzyme. The assay is designed in such a way that transcription strictly depends upon phosphorylation of FixJ. It consists of three steps (see "Experimental Procedures" for details). 1) Pure FixJ protein is phosphorylated by incubation with pure FixL protein and ATP for 20 min. The FixL protein that is used, FixL122, is a truncated, soluble, oxygen-responsive form of the full-length FixL protein (5,12). 2) Closed complexes are allowed to form for 10 min by the addition of E. coli RNA polymerase-70 holoenzyme and a cloned fixK DNA transcription template to phosphorylated FixJ. 3) Isomerization of closed complexes and elongation of transcription is started upon the addition of the three missing nucleotides. The first two steps are performed under anoxic conditions, since they are readily reversible in air.
The addition of highly purified MalE-FixT protein at the onset of FixJ phosphorylation (step 1 above) resulted in a dramatic decrease in the level of fixK mRNA synthesized ( Fig.  2A, lane 3). The effect of MalE-FixT on fixK transcription was specific, since all other transcripts, which originate from transcription of the vector template DNA genes by E. coli holoenzyme, were unaffected. Thus, FixT specifically blocked initiation of transcription of fixK or elongation of the nascent fixK RNA. The inhibitory effect of FixT was also observed when FixT was added 10 min after FixJ phosphorylation was started ( Fig. 2A, lane 4; see "Discussion").
Interestingly, no inhibition of fixK transcription was observed when FixJ was phosphorylated from acetylphosphate (Fig. 2B), a low molecular weight phosphodonor (5, 15), instead of being phosphorylated from FixL-phosphate. Specifically, FixT had no effect on FixJ-driven transcription if FixT was added to FixJ at the end (Fig. 2B) or at the beginning of FixJ incubation with acetylphosphate (data not shown). Furthermore, FixT did not inhibit the transcriptional activity of a FixJ-phosphate preparation from which acetylphosphate had been eliminated by molecular sieving (data not shown). This demonstrated first that FixT did not directly dephosphorylate FixJ-phosphate (since the transcription assay relies on phosphorylated FixJ) and, second, that FixT did not inhibit FixJ-  phosphate transcriptional activity per se. Rather, these data suggested an implication of FixL in inhibition by FixT.
FixT Inhibits FixL-Phosphate Synthesis-When purified FixL is incubated in the presence of ␥-labeled [ 33 P]ATP, the amount of (auto)phosphorylated FixL protein rises linearly over at least the first 15 min of the reaction (16) (Fig. 3B). Subsequently, the amount of FixL-phosphate levels off because of the accumulation of both FixL-phosphate and ADP that promotes the reverse reaction, i.e. FixL-P ϩ ADP 3 FixL ϩ ATP (16).
When purified FixT protein was preincubated with FixL prior to the addition of ␥-labeled ATP (400 M), no FixL-phosphate accumulation could be detected after 10 min (Fig. 3A) or after 60 min of incubation (data not shown). In a second set of experiments (Fig. 3B), FixL was incubated with ␥-labeled ATP for 5 min prior to the addition of FixT, thus allowing a detectable amount of FixL-phosphate to be synthesized. Upon the addition of FixT, the level of phosphorylated FixL stabilized but did not decrease, thus excluding a prominent phosphatase activity associated with FixT. This was confirmed by purifying FixL-phosphate from ATP (and ADP) nucleotides in order to prevent further FixL phosphorylation. It was then observed that the addition of FixT (Fig. 3C) or FixT plus ATP (data not shown) did not enhance the rate of FixL-P dephosphorylation. This is evidence that FixT is not a phosphatase.
We also ruled out the possibility of an ATPase activity associated with FixT, since no P i production or depletion of the ATP pool could be detected on TLC plates after extended incubation of FixT with ␥-labeled ATP, even in the presence of FixL (data not shown). We were also unable to detect any (reasonably stable) phosphorylation of FixT after it was incubated with either radiolabeled ATP or FixL-phosphate and separated by SDS-PAGE under neutral pH conditions (data not shown). Thus, the model we presently favor is that FixT either blocks autophosphorylation of FixL from ATP (FixL ϩ ATP 3 FixL-P ϩ ADP) or enhances the rate of the reverse reaction (i.e. FixL-P ϩ ADP3 FixL ϩ ATP).
Both explanations are consistent with the further observation that adding FixT to FixL 60 min after the addition of ATP (i.e. when the phosphorylation reaction FixL ϩ ATP º FixL-P ϩ ADP had reached equilibrium) resulted in the disappearance of most of FixL-phosphate within 15 min (Fig. 4).
FixT Promotes FixJ-Phosphate Dephosphorylation in the Presence of FixL-When FixL and FixJ are incubated together in the presence of ␥-labeled ATP under anoxic conditions, the phosphate first enters FixL and then readily transfers to FixJ so that only FixJ-phosphate accumulates (Fig. 5, lanes 1-3). The level of phosphorylated FixJ levels off after a 30 -40-min incubation in our assay conditions because of the phosphatase activity associated with FixL (3,17). When FixT was added after a 60-min incubation, the level of phosphorylated FixJ decreased dramatically within 15 min (Fig. 5). FixT thus promoted dephosphorylation of FixJ-phosphate. As pointed out above (Fig. 2B), this was not due to a direct dephosphorylation of FixJ-phosphate by FixT. Rather, FixJ-phosphate disappearance was a consequence of the block in FixL-phosphate synthesis that prevented replenishing of the FixJ-phosphate pool (see "Discussion"). In addition, we observed that FixT did not affect phosphotransfer between FixL-phosphate and FixJ as monitored with purified (sieved), nucleotide-free, FixL-phosphate (data not shown).
Evidence for a Stoichiometric Interaction between FixT and FixL-Early trials led us to use FixT at a concentration of 9 M and FixL at 6 M (monomers) in the experiments reported above. We subsequently investigated how the activity of FixT varied with its concentration. In a first set of experiments, FixL concentration was kept at 6 M, and FixT was titrated down from 9 to 0.1 M. Half-maximum inhibition of FixL autophosphorylation was observed when FixT was ϳ1.2 M, i.e. at a 1:5 ratio between FixT and FixL (Fig. 6). Complete inhibition was  ; lanes 3, 5, and 7) or MalE-FixT protein (ϩ; 9 M) was added (lanes 4, 6, and 8). Samples were withdrawn at 70 min (lanes 3 and 4), 75 min (lanes 5 and 6), and 80 min (lanes 7 and 8) for SDS-PAGE analysis. Lane 2, positive control. In this tube, MalE-FixT protein was present throughout the phosphorylation reaction of FixL122. observed at ϳ3 M FixT, i.e. at a 1:2 ratio. Similar data were obtained at another concentration of FixL (2 M monomer) (Fig. 6).
These data are consistent with the action of FixT on FixL being stoichiometric rather than catalytic. Based on previous observations that FixL is a dimer in solution (5,18) and assuming that FixT is a monomer, full inhibition of FixL autophosphorylation would occur upon mixing one monomer of FixT with one dimer of FixL.
FixT Targets the Carboxyl-terminal Kinase Domain of FixL-FixL is a modular protein (12,19,20) made up of three domains: 1) a N-terminal membrane-anchoring domain, which is not essential for activity or for oxygen sensing in vitro (this domain is missing in the soluble FixL122 protein used in this work), 2) a central oxygen-sensing heme-containing domain whose structure has been solved recently (20); and 3) a Cterminal kinase domain homologous to the kinase domain of sensor regulatory proteins. This isolated domain, previously called FixLC, is capable of constitutive autophosphorylation independently of oxygen, although with a lower efficiency than the full-length protein (19). We examined whether this isolated kinase domain was prone to inhibition by FixT.
A FixLC protein was purified that consisted of the last 240 amino acids fused to MalE (see "Experimental Procedures"). FixT significantly inhibited phosphorylation of FixLC, although inhibition was not complete (Fig. 7A, lane 4), even at a molar ratio of 6 between FixT and FixLC (data not shown). We conclude that the C-terminal kinase domain of FixL is the primary target for FixT, although we do not exclude the possibility that the central oxygen-sensing domain of FixL may contribute to the interaction with FixT.
FixT Is Not Active on the Heterologous NtrB, DegS, and KinA Sensor Kinases-Since the C-terminal kinase domain of FixL is conserved between the sensor proteins of the two-component family, it was of interest to test whether other targets could be identified for FixT. Three purified kinases were kindly provided by Dr. T. Mzadek: the NtrB protein from Salmonella typhimurium and the DegS and KinA sensor proteins from Bacillus subtilis. Two different phosphorylation conditions were tested. In a first set of experiments, KinA, NtrB, and DegS were phosphorylated in the presence of 2.5 M [␥-33 P]ATP, as described before (13). No effect of FixT on the phosphorylation of the heterologous kinases was observed (Fig.  7B). In a second set of experiments, we assayed phosphorylation of NtrB, DegS, and KinA under the conditions used for FixL (i.e. 0.4 mM ATP) and, again, found no effect of FixT (data not shown). Similar data were obtained with higher concentrations of FixT (27 M) (data not shown).

DISCUSSION
Mode of Action of FixT-fixT was originally identified by genetic studies that demonstrated its ability to repress expression of the regulatory genes fixK and nifA in S. meliloti (9). However, the mode of action of fixT was completely unknown. Using in vitro methods, we have now demonstrated that FixT counteracts the activity of the FixLJ two-component system that mediates nifA and fixK expression in response to oxygen in S. meliloti. Specifically, we provide evidence that FixT acts by inhibiting primarily FixL-phosphate synthesis.
How FixT achieves its effect is not completely clear yet. When added prior to ATP, FixT completely prevented incorporation of phosphate into FixL (Fig. 3A). This is consistent with FixT blocking autophosphorylation of the FixL protein from ATP. For example, FixT may prevent access of ATP to the nucleotide binding pocket of FixL or slow down histidyl-phosphate formation. By contrast, FixT had no detectable ATPase activity (data not shown) and thus certainly does not act by depleting the pool of ATP. Alternatively, FixT may favor rapid dephosphorylation of neosynthesized FixL-phosphate. We have presented evidence that FixT is not a phosphatase for FixL-P (Fig. 3C) and does not phosphorylate itself from FixL-P (data not shown). By contrast, we do not exclude the possibility that FixT may enhance the rate of dephosphorylation of FixL-phosphate in the presence of ADP (FixL-P ϩ ADP 3 FixL ϩ ATP), i.e. may favor the reverse reaction of autophosphorylation.
FixT does not resemble any known protein, nor does it have any prominent motif that could give indications regarding its mode of action. However, we presently have three clues: 1) FixT acts directly on FixL; 2) FixT targets the C-terminal domain of FixL; 3) the interaction between the two proteins is likely to be stoichiometric. Our working model is that FixT makes a protein-protein interaction with the carboxyl-terminal kinase domain of FixL that may either inhibit autophosphorylation of FixL or favor the reverse dephosphorylation reaction.
No other effect of FixT was observed besides its effect on FixL phosphorylation. First, FixT did not detectably impair phosphotransfer between FixL-phosphate and FixJ (data not shown). Second, FixT had no effect on the stability of FixJphosphate (in the absence of FixL), as assayed by transcriptional activity (Fig. 2B). Note that direct assessment of FixJphosphate dephosphorylation by FixT was not appropriate because of the very low specific activity of acetyl-phosphate that can be achieved (5). Instead, transcription is an indirect but reliable assay for FixJ-phosphate dephosphorylation. Thus, the primary, if not unique, role of FixT is to inhibit FixL phosphorylation.
The reported effect of FixT on FixL-phosphate synthesis fully accounts for both the FixT-induced FixJ-phosphate dephosphorylation in the presence of FixL (Fig. 5) and the inhibition of fixK transcription ( Fig. 2A). Even under low oxygen conditions, phospho-FixJ is continuously and rapidly hydrolyzed because of the phosphatase activity associated with unphosphorylated FixL (16). In the presence of FixT, the FixL-phosphate pool drops with the consequence that phospho-FixJ cannot be replenished and thus disappears rapidly. FixJ-phosphate dephosphorylation in turn results in a dramatic (more than 100-fold) decrease of its capacity to bind and activate the nifA and fixK promoters (7). The comparison of the kinetics of disappearance of FixJ-phosphate and of inhibition of fixK transcription is consistent with this analysis (compare, for example, Fig. 5, lane 8, and Fig. 2A, lane 4).
Modulators of Two-component Systems-Two-component systems are of paramount importance in bacteria, where they probably constitute the most widespread device for signal transduction. E. coli and B. subtilis each use up to 30 different two-component systems to cope with environmental changes (21,22). Some two-component systems look very simple; in the FixLJ system, for example, a single protein, FixL, is able to sense its cognate signal (molecular oxygen) and to transduce it to the regulator protein, FixJ, by phosphotransfer. Moreover, both FixL and FixJ proteins have a modular structure (12,19,23). As a result, a combination of four domains suffices for oxygen signal transduction. Many two-component systems are more complex, involving more proteins than a simple sensor and regulator pair. Many of these proteins are phosphatases. For example, in B. subtilis, RAP phosphatases ensure dephosphorylation of the SpoOF-phosphate regulator protein (24). The SixA protein from E. coli is a phosphohistidine phosphatase that promotes dephosphorylation of the sensor ArcB protein (25). The PII protein in enteric ntr circuits is not a phosphatase per se but stimulates the phosphatase activity associated with the NtrB sensor kinase (26).
A more relevant comparison for FixT is KipI, for which the name antikinase was coined in bacteria. kipI was originally identified during the course of the B. subtilis genome sequencing project (27). Subsequently, KipI was shown to block sporulation of B. subtilis when overproduced in the absence of another protein, KipA, with which KipI normally forms a complex (28). KipI and FixT resemble each other by their inhibitory effect on the (auto)phosphorylation of their cognate sensor ki-nases, KinA and FixL, respectively. However, FixT and KipI display no amino acid sequence similarity to each other, and, thus, whether they are related in structure or mode of action is not known.
Biological Role of fixT in S. meliloti-The primary role of FixLJ in S. meliloti is to control, via nifA and fixK, nif and fix gene expression in response to oxygen availability. The present data reinforce the pivotal role of the FixLJ system in S. meliloti, as an integration site for both positive regulation by oxygen and negative regulation by FixT. Typically, molecular devices such as the KipI, FixT, or the above mentioned phosphatases may allow two-component systems to integrate multiple environmental regulatory signals as demonstrated for the RAP phosphatases of B. subtilis (24,28).
Since FixT interacts with FixL, one obvious possibility would be that FixT contributes to oxygen control. Available data indicate that FixL is sufficient for oxygen regulation of nifA and fixK expression both ex planta and in vitro and that fixT does not grossly affect oxygen regulation of the system (9). Furthermore, the central oxygen-sensing domain of FixL does not seem to be the primary target for FixT, and no difference in FixT activity was observed in vitro under anoxic versus oxic conditions (data not shown). Still, FixT may contribute to a fine tuning of the FixLJ response to oxygen, ensuring, for instance, that the FixL protein, which has a very low affinity for oxygen (29), does not become prematurely activated in response to a moderate or transient oxygen deprivation. Alternatively, FixT may prevent activation of the FixLJ system by low oxygen under conditions that would be otherwise antithetical to symbiotic nitrogen fixation such as, possibly, low energy charge, low carbon, or high nitrogen.