We previously observed that in a S. meliloti wild-type strain, constitutive expression of fixT, driven by the strong neomycin promoter of plasmid pMF10 (Table 1), resulted in strong inhibition of the expression of a fixK-lacZ reporter fusion (pMF457 plasmid; Table 1), thus leading to white colonies on X-gal containing plates. We used this observation to screen for S. meliloti mutants that would escape repression by FixT. After random Tn5 mutagenesis of a strain overexpressing fixT, we isolated blue colonies on X-gal containing plates. These putative mutant colonies were subsequently assayed for β-galactosidase activity in liquid cultures under microoxic conditions. Four independent mutants were isolated, that were characterized by an unrepressed level of expression of the fixK gene despite the constitutive expression of fixT (Figure 1A; compare lanes 2 and 3). Two of the mutants possessed a Tn5 insertion in the same gene. One of these two mutants, GMI401, was further characterized. Southern-Blot analysis of genomic DNA digested by different restriction enzymes, revealed a single Tn5 insertion in GMI401 (data not shown). Transduction experiments using the N3 phage, showed genetic linkage between the mutant phenotype and the Tn5 insertion since reintroduction of the pMF10 plasmid in the transduced strain confirmed the lack of repression of fixK by overexpressed fixT (Figure 1A; lanes 5 and 6). This excluded the possibility that a mutation on one of the plasmids, pMF457 or pMF10, could have been responsible for the phenotype observed.

In the GMI401 mutant strain, the level of fixK gene expression in the absence of pMF10 was the same as in the wild-type strain (Figure 1A; compare lane 1 to lane 4 and lane 5). This result demonstrated that activation of fixK gene expression by the FixLJ two-component system was not affected in the mutant strain, but rather that the phenotype was genuinely due to decreased repression by fixT. Western-Blot analysis using an antibody directed against the FixT protein indicated that the level of FixT protein was the same in the GMI211(pMF10) wild-type strain and in the GMI401(pMF10) mutant strain (data not shown).

Similar results have been obtained using a nifA-lacZ fusion (pCHK57 plasmid). The inhibition of nifA-lacZ expression by overexpressed fixT that was observed in a GMI211 wild-type strain was not observed anymore in the GMI401 mutant strain (data not shown).

Altogether, these results tend to indicate that, in the GMI401 mutant strain, the absence of repression of fixK and nifA by fixT was due to a reduced inhibition of the FixLJ system by the FixT protein.

The GMI401 mutant strain produced nitrogen fixing nodules on M. sativa, as did the GMI211 wild-type strain (data not shown). Whereas the GMI211 parent strain expressing fixT constitutively produced non N₂-fixing nodules (Figure 1B), the GMI401 mutant strain was able to induce effective N₂-fixing nodules on alfalfa (Medicago sativa cv. gemini) thus enabling the plants to grow in the absence of combined nitrogen. Hence, the gene that modulates repression by FixT activity ex planta is also active in planta.

The Tn5 insertion in GMI401 was positioned on the pSymB megaplasmid of S. meliloti (see Materials and Methods), whereas the fixLJ, fixT and fixK genes are located on pSymA megaplasmid. The genomic DNA flanking the Tn5 insertion was sequenced (Figure 2 and Materials and Methods). Prediction of coding regions around the region of interest was performed using the FrameD program [7]. As shown in Figure 2, this analysis revealed a putative open reading frame encompassing the Tn5 insertion site and another putative orf located just dowstream. On the basis of the sequence analysis, these two genes may belong to the same operon, and the Tn5 insertion may thus affect expression of both genes. During the annotation phase of the S. meliloti genome sequencing project (http://sequence.toulouse.inra.fr/meliloti.html) that was run concomitantly to this work, the upstream gene was named asnO, for its homology to the Bacillus subtilis asnO gene. We have thus adopted this annotation in this paper. The deduced amino sequence of the second orf, that was named SMb20482, is similar to an acetyl-transferase in the amino-terminal part of the protein and to a cyanophycin synthetase gene in the carboxy-terminal part of SMb20482.

The deduced amino acid sequence of the gene carrying the Tn5 insertion is similar to that of glutamine-dependent asparagine synthetases of various bacteria including B. subtilis and Escherichia coli (see Figure 3). These proteins are members of the PurF family of glutamine-dependent amidotransferases. All PurF enzymes possess a conserved amino-terminal cysteine, which is essential for glutamine-dependent amidotransferase activity [8, 9]. This Cys2 residue is also conserved in the S. meliloti asnO product. The PurF-type amidotransferases possess 14 additional conserved residues in the amino-terminal glutamase domain [8, 9]. As shown by the alignment of Figure 3, all these residues are present in the asnO product of S. meliloti. We conclude that the AsnO protein belongs to the PurF family of glutamine amidotransferases. The carboxy-terminal domian of the proteins, carrying the synthetase activity, is less conserved.

In order to perform complementation trials, we cloned the asnO gene in the pBBRI-MCS3 plasmid [10], a shuttle vector that replicates in both E. coli and S. meliloti. A DNA fragment containing the entire asnO coding region and its 5' flanking promoter region, was cloned downstream of the lac promoter of the plasmid pBBRI-MCS3. We introduced the corresponding pBasn2 plasmid in a strain expressing fixT constitutively and monitored expression of fixK under microoxic conditions by RT-PCR experiments. Data indicated complementation of the GMI401 mutant strain by the asnO transgene (Figure 4). The restoration of the wild-type phenotype (ie inhibition of fixK expression in the presence of pMF10), demonstrated the implication of the asnO gene in the characteristic phenotype observed in the GMI401 mutant strain. However, restoration of the wild-type phenotype was not complete, for a reason that we do not understand.

Because of the homology of the S. meliloti asnO gene product with known asparagine synthetases, we tested whether the asnO gene was involved in asparagine biosynthesis, by three complementary approaches (data not shown).

First, we observed that the mutant strain GMI401 and the isogenic GMI211 wild-type strain, grew at equal and similar rates in minimal medium with or without asparagine.

Second, we tested the ability of pBasn2 to restore prototrophy of a E. coli asparagine auxotroph, a double asnB/asnA mutant strain (Table 1). In asparagine-supplemented minimal medium, all strains grew at equal rates. However, in the absence of added asparagine, the S. meliloti asnO gene under the control of the plac promoter was unable to complement the asparagine auxotrophy of the asnA/asnB E. coli mutant strain.

Third, addition of asparagine at different concentrations in minimal medium, did not lead to complementation of the S. meliloti GMI401 mutant phenotype, i.e did not restore repression of fixK gene expression by overexpressed fixT. This observation suggested that the mutant phenotype is not dependent on the level of asparagine, and hence, that the asnO gene does not mediate asparagine synthesis.

The reactions catalyzed by asparagine synthetases involve two different family of proteins depending on whether glutamine or ammonia is used as a nitrogen source. Members of the AsnA family, that are found in procaryotes, only use ammonia as the amino group donor [11, 12]. The AsnB family, which is found in both procaryotes and eucaryotes, preferentially uses glutamine as nitrogen source [13, 14]. E. coli possesses an asnA and an asnB gene [15] whereas three asparagine synthetase genes, asnB, asnH and asnO, which all belong to the AsnB family, have been characterized in B. subtilis. No member of the AsnA family has been found in B. subtilis [16]. Complete genome analysis indicated that S. meliloti carries no asnA-like gene and two asnB-type genes (see http://sequence.toulouse.inra.fr/meliloti.html) one of which is asnO and a second gene that was named asnB. Sequence analysis is consistent with the possibility that the asnB gene of S. meliloti might be the ortholog of the biosynthetic asnB gene of E. coli and B. subtilis whereas the asnO gene mutated in GMI401 may play a regulatory function, in relationship with nitrogen fixation and microoxic respiration (Figure 3). Similarly, the asnO gene of B. subtilis has a regulatory role in sporulation [16], whereas the ltsA gene of Corynebacterium glutamicum is involved in cell wall formation [17].

We have monitored by RT-PCR experiments expression of the asnO gene in a wild-type strain grown in minimal medium, in either oxic or microoxic conditions. The results showed that asnO gene expression was slightly enhanced in microoxic conditions as compared to oxic conditions (Figure 5). No effect of a fixJ mutation was observed on asnO expression using a GMI5704 fixJ mutant strain [18].

The bacterial strains and plasmids used in this study are listed in Table 1. S. meliloti GMI211, the symbiotically effective parent strain used in this study is a lac Sm^r, Nod+ Fix+ derivative of S meliloti RCR2011 [20]. Tn5 transposon mutagenesis of GMI211 was performed using the pRK602 plasmid (Table 1, [21]). In order to isolate random Tn5 insertions in the S. meliloti genome, we conjugated E. coli MM294(pRK602) donor strain with a S. meliloti GMI211(pMF10) recipient strains, and screened for streptomycin (100 μg/ml) and neomycin (100 μg/ml) resistant transconjugants.

S. meliloti strains were grown at 30°C in TY complex medium or in defined M9 medium [22] supplemented with 0.3 mM CaCl₂, 1 mM MgSO₄ and 2 μM biotin. Sterilised carbon sources were added at 20 mM final concentrations. Microoxic conditions were achieved as described by de Philip et al [23] (2% oxygen for 4 h). For testing induction of fixK-LacZ expression by microoxic conditions, S. meliloti strains were previously grown to OD₆₀₀ = 0.3 in M9 medium. β-Galactosidase assays were performed as described by de Philip et al. [23].

DNA techniques were carried out as described by Sambrook et al. [24]. S. meliloti genomic DNA was prepared as described by Chen and Kuo [25]. For Southern-Blot analysis, 1 μg of genomic DNA was digested at 37°C for 4 hours with the appropriate enzymes. DNA fragments were separated on a 0.8% agarose gel, transferred onto a nylon membrane (BiodyneA transfer membrane, Pall, East Hills, NY), and hybridized with a Tn5 ³²P-labelled probe.

A genomic fragment of S. meliloti GMI401 carrying the Tn5 insertion was cloned from a EcoRI-digested genome fragment into a pUC18 plasmid, resulting in the pUC23 plasmid (see Table 1). However, the presence of a EcoRI restriction site in the asnO gene (see Figure 2) did not permit to obtain the entire sequence of the gene from this construct. Therefore, a library of 96 BACs covering the S. meliloti genome [26], was screened to isolate the full-length asnO gene. PCR screening of the BAC library was performed using the oligonucleotide primers ASN+54 and ASN-9 shown in Figure 2. By this method, we could detect the presence of the asnO gene on BAC37, that was assigned to the pSymB megaplasmid [26]. Complete sequence of the asnO region was determined by a primer-walking approach on the BAC37 and pBH1 plasmid (see Table 1), which was obtained by subcloning a HindIII fragment of BAC37 in pBBR1-MCS3 [10]. Sequencing was performed on a ABI373 automated sequencer (ABI, Columbia, MD), using the ABI PRISM Dye terminator cycle sequencing ready kit (Perkin-Elmer, Oak, Brook, IL).

Putative open reading frame were predicted using the FrameD program [7] used for S. meliloti whole genome analysis (http://sequence.toulouse.inra.fr/meliloti.html). Protein analysis was done with the NCBI web page using the BLAST2 package program [27] against the NCBI-nr and Swiss-Prot databases. Homologous domains searches were driven using the ProDom database [28].

Medicago sativa cv. Gemini seedlings were aseptically grown on agar slants made up with nitrogen-free Fahraeus medium. Three-day-old plants were inoculated with the different S. meliloti strains and grown for 3 weeks before observation.

The bacteria from a 25 ml culture at 0.4 OD_600nm were harvested and RNA prepared with the Qiagen RNeasy kit as described by the manufacturer. DNA was eliminated by addition of 7.5 Units of FPLC-Pure RNase-free DNaseI (Amersham-Pharmacia Biotech). RNA was further extracted with phenol-chloroform and then precipitated with ethanol. After washing with 70% ethanol, the pellet was resuspended in nuclease-free water. RNA was quantified by absorbance measurements at 260 nm. Absence of DNA contamination was verified by PCR amplification.

RT-PCR reactions were performed according to the manufacturer, using the SuperScript™ One-Step RT-PCR System (GibcoBRL).

The following oligonucleotides were used for reverse transcription of the messenger RNA and amplification of the product to evaluate gene expression :

For the hemA gene: hemAr (reverse primer) : 5'-GTCGATCGCGTTCTT-3'; hemAf (foward primer) : 5'-TGGATGGGCTGCATCA-3'

For the asnO gene: RTA1 (reverse primer) : 5'-TGCGTATTCTCGACCTG-3'; RTA2 (foward primer) : 5'-TCGCGAAAATTGTAGATG-3'

For the fixK gene: KPR (reverse primer) : 5'- CCGATTACCAGAAGATGC-3'; KPF (foward primer) : 5'-TATCTACCGCCTCCTTTC-3'

RT-PCR products were electrophoresed on a 2% agarose gel, blotted onto a nylon membrane and probed with a ³²P-labeled DNA probe prepared from the gene of interest. Washing was done with 0.1XSSC, 0.1%SDS at 42°C during 30 min.