In this work, we investigated the osmoadaptive mechanisms used by four native rhizobia isolated from root nodules of P. vulgaris cultivated in north Tunisia . Strains R. etli 12a3, R. gallicum bv. phaseoli 8a3 and R. leguminsarum 31c3 are potentially good inoculants as they were infective and showed efficient nitrogen fixation in symbiosis with P. vulgaris . In addition, Agrobacterium 10c2 was able to colonize preformed P. vulgaris nodules  and to specifically favour nodulation by some local strains , suggesting that it might be used as co-inoculant. Our results confirm the strain affiliations proposed by Mhandi et al. [24, 28]. In addition, on the basis of its phylogenetic relatedness to the A. tumefaciens type strain, Agrobacterium 10c2 is proposed in this work to be renamed as A. tumefaciens 10c2.
As shown by 13C- and 1H-NMR analyses, the long-term response of the four Rhizobium strains to NaCl involved the accumulation of trehalose, mannitol and glutamate; but the latter one was only observed in R. leguminsarum 31c3 and R. tropici CIAT 899. The reason why glutamate was not present in the extracts of R. gallicum bv. phaseoli 8a3 and R. etli 12a3 is unknown. It might be that glutamate is accumulated in these strains only during the early osmostress response, as the most commonly used charge counterbalance for K+ influx , but it is subsequently replaced by trehalose. This was previously demonstrated in S. meliloti by Gouffi et al . On the other hand, mannosucrose and glutamate were the main osmolytes in A. tumefaciens 10c2 grown at high salinity, whereas at low salt only mannitol was observed. Mannosucrose accumulation was found to be NaCl-dependent in A. tumefaciens 10c2 (this study), A tumefaciens strains C58 and NT1  and in rhizobial isolates from Acacia nodules , supporting the hypothesis that this compatible solute participates in alleviating osmotic stress. However, isolation and analysis of osmosensitive mutants would be necessary to prove the latter statement, and additional mechanisms involved in A. tumefaciens 10c2 osmoadaptation cannot be ruled out.
In the tested strains, mannitol was not accumulated when glucose was used as a carbon source (Figure 4, and data not shown). On the other hand, cells grown with [1/6-13C]mannitol as a carbon source accumulated [1/6-13C]mannitol, indicating that mannitol was not synthesized de novo but accumulated upon transport from the external medium. Bacteria rarely synthesize mannitol as a compatible solute, but it is frequent to find it as an external osmoprotectant . In general, uptake and accumulation of osmoprotectants is preferred over the synthesis of endogenous compatible solutes, as the latter is energetically more costly . However, R. tropici CIAT 899 and A. tumefaciens 10c2 used mannitol both as carbon source and as an osmoprotectant solute at low salinity, but mannitol was replaced by endogenous compatible solutes (i.e. trehalose or mannosucrose) when cells were exposed to hyperosmotic stress (see Figures 3 and 4). This finding may be explained by two, non-exclusive, reasons: (i) that trehalose and mannosucrose are better osmolytes than mannitol, and/or (ii) that energy-requiring systems, other than trehalose or mannosucrose synthesis, were operating at high salinity, and mannitol catabolism was enhanced in detriment of its accumulation.
The role of trehalose as a compatible solute involved in bacterial tolerance to osmotic stress has been widely demonstrated in the literature. Thus, E. coli , S. meliloti  and B. japonicum  mutants lacking the otsA gene for the synthesis of trehalose are osmosensitive. In another study, Alarico et al.  found a direct correlation between the presence of genes for trehalose synthesis (otsA/otsB) in Thermus thermophilus strains and their halotolerance. In this work, we found that trehalose synthesis in R. tropici CIAT 899 is osmoregulated (Figure 6), suggesting the involvement of trehalose in the osmotolerance of this strain. However, we could not find a direct correlation between the trehalose content of the rhizobial strains and their osmotolerance. On the contrary, trehalose levels in the less salt tolerant strains grown at 0.1 M NaCl were 10 fold-higher than those of the more salt-tolerant R. tropici CIAT 899 grown under the same conditions (Figure 6). Therefore, trehalose alone cannot account for the higher osmotolerance of R tropici CIAT 899. It is improbable that accumulation of mannitol by R tropici CIAT 899 conferred it a higher halotolerance, as mannitol was also accumulated by the less salt-tolerant strains. Other salt-induced responses, as modifications in the pattern of extracellular polysaccharides and lipopolysaccharides might be involved . Upon transposon mutagenesis, Nogales et al  identified eight gene loci required for adaptation of R tropici CIAT 899 to high salinity. These included genes involved in regulation of gene expression, genes related to synthesis, assembly, and maturation of proteins, and genes related with cellular buildup and maintenance.
To date, three different enzymatic pathways have been described for trehalose synthesis in rhizobia (OtsAB, TreS and TreYZ; ). The most common two-step OtsAB pathway catalyzes the synthesis of trehalose from UDP-glucose and glucose 6-phosphate. Trehalose synthase (TreS) catalyzes the reversible conversion of maltose and trehalose. Finally, the two-step TreYZ pathway acts in the production of trehalose from a linear maltodextrin (e.g., glycogen) . In this work, we showed the presence of otsA within the genome of the four Rhizobium analyzed strains, suggesting that trehalose synthesis in these strains occurs at least via OtsAB. Synthesis of trehalose from maltooligosaccharides in R. tropici CIAT 899 was earlier reported , although TreY activity could not be detected . Interestingly, the phylogenetic position of OtsA from R. gallicum bv phaseoli 8a3 and R. etli 12a3 was not consistent with the 16S rDNA-based tree, suggesting the existence of lateral transfer events. Avonce et al.  also found inconsistencies in the topology of a proteobacterial OtsA-based tree, and suggested to be caused by either lateral gene transfer or differential loss of paralogs.
Cyclic (1→2)-β-glucans have a role in hyposmotic adaptation of the legume symbiont rhizobiaceae . In R. tropici CIAT 899 (and probably R. gallicum bv. phaseoli 8a3) cells grown at low salinity, the cyclic β-glucan was co-extracted with the cytoplasmic compatible solute pool, suggesting that high amounts of beta glucan were present in the periplasm.. As trehalose, cyclic (1→2)-β-glucans are synthesized from UDP-glucose . We found that mannitol and galactose were substrates for both trehalose and the β-glucan of R. tropici CIAT 899. In contrast, mannose was a substrate for the β-glucan but not for trehalose.. From the above data, we conclude that R. tropici CIAT 899 can convert mannitol and galactose into UDP-glucose and glucose-6-phosphate, the two trehalose precursors, but it cannot transform mannose into glucose-6-phosphate. In E. coli and other bacteria, galactose degradation pathway I (Leloir pathway) can yield both UDP-glucose and glucose-6-phosphate . Thus, a similar route might be operating in R. tropici CIAT 899. By using [1/6-13C]mannitol as a carbon source, we showed that both trehalose moieties, as well as the β-glucan units, where derived directly from mannitol. In E. coli and other bacteria, mannitol and mannose enter the cell via specific phosphotransferase systems so the first intracellular species are mannitol-1-phosphate and mannose-6-phosphate, respectively. In a second step, these phosphoderivatives are converted by a single dehydrogenase or isomerase reaction, respectively, into the glycolytic intermediate fructose-6-phosphate, which in turn is converted to glucose-6-phosphate by the action of a phosphoglucose isomerase [43, 44]. A search in the KEGG specialized pathway database  showed that the genomes of R. etli CFN 42, R. leguminosarum bv. viciae 3841, S. meliloti 1021, A. tumefaciens C58, Mesorhizobium loti MAFF303099, B. japonicum USDA 110 and Rhizobium sp. NGR 234, among others, do not carry the mtlA gene encoding the specific mannitol phosphotransferase, suggesting that in the Rhizobiaceae mannitol do not use a phosphotransferase system to enter the cell. Instead, we found the smoEFGK genes encoding a sorbitol/mannitol ABC transporter, mtlK (encoding a mannitol 2-dehydrogenase that converts mannitol to fructose), and xylA (encoding a xylose isomerase that converts fructose to glucose). By analogy with these phylogenetic relatives, we suggest that in R. tropici mannitol could be converted into glucose via fructose. In the case of mannose, we found that the above genomes carried manX, encoding the phosphohistidine-sugar phosphotransferase protein, suggesting that the first intracellular species is mannose-6-phosphate. The gene manA, encoding the mannose-6-phosphate isomerase (isomerizing mannose-6-phosphate into fructose-6-phosphate) is present in S. meliloti, Rhizobium sp. NGR 234, A. tumefaciens and B. japonicum, but not in R. etli, R. leguminosarum, or M. loti. This finding suggests that the latter microorganisms, and most probably R. tropici CIAT 899, cannot convert mannose-6-phosphate into fructose-6-phosphate, and consequently it cannot yield glucose-6-phosphate. R. etli, R. leguminosarum and M. loti carried noeK, encoding a phosphomannomutase that converts mannose-6-phosphate to mannose-1-phosphate, and noeJ, encoding a mannose-1-phosphate guanylyltransferase that converts mannose-1-phosphate to GDP-mannose, a precursor for glucan biosynthesis. In addition, R. tropici CIAT899 carries a noeJ-like gene, as described by Nogales et al . Again by analogy with its close relatives, we suggest that a similar pathway might be operating in R. tropici, explaining why this microorganism can synthesize the cyclic β-glucan from mannose, but cannot convert mannose into trehalose.