The wild-type B. diazoefficiens strain employed in this work is a spontaneous Sm resistant derivative from the type-strain USDA 110. The lafR::Km mutant strain is a USDA 110 insertional mutant in which the Km resistance cassette from the plasmid pUC4K was inserted into the coding region of lafR (Mongiardini et al., 2017). We grew the bacteria for inoculum in liquid HM salts–yeast-extract medium (HMY) supplemented with either 0.5% (w/v) Mtl or 0.5% (w/v) Ara (Mongiardini et al., 2017) as the C sources (HMY-Mtl or HMY-Ara cultures, respectively), taking into account that the C-mole of each substrate is very similar (30.3 and 30.0 g.C-mole^-1 for Mtl and Ara, respectively). These cultures were grown at 30°C with rotary shaking at 180 rpm. We carried out the batch cultures of the wild-type and the mutant strain in an LH-210 Bioreactor (Inceltech, Toulouse, France) with a working volume of 4 l in HMY-Mtl or HMY-Ara at 30°C and 250 rpm with air flushed at 25–30 l.h^-1.

We studied the differential proteomes in B. diazoefficiens USDA 110 with protein samples obtained from bacteria in the exponential phase of growth (OD_{500 nm} = 1.0) in HMY-Mtl or HMY-Ara. We obtained the rhizobia samples from three independent cultures per condition by centrifugation at 7,150 × g, followed by two washes of the pellets with low-salt washing buffer (3 mM KCl, 1.5 mM KH₂PO₄, 68 mM NaCl, 9 mM NaH₂PO₄; Pérez-Giménez et al., 2012) and then resuspended the pellets in 10 mM Tris-HCl pH 8.5 up to an OD_{500 nm} = 10, adding 1 μl of phenyl methyl sulfonyl fluoride in Tris-HCl per ml. We then lysed the cells by ultrasound with a Sonifier 150 (Branson) sonicator employing six 3-min pulses at 38% power in an ice bath and centrifuged the lysate at 13,520 × g for 30 min to remove the unbroken cells and cell debris. We next centrifuged the resulting supernatant at 66,226 × g for 2 h to pellet the microsomal fraction, though also retaining that supernatant. To collect the membrane-enriched–protein fraction, we resuspended the pellets in 250 μl of rehydration solution (7 M urea, 2 M thiourea, 10% [v/v] isopropanol, 2% [v/v] triton X-100). In parallel, to each of those supernatants, we added 4 volumes of acetone for at least 12 h at -20°C to precipitate the cytoplasm-enriched proteins, before a final centrifugation at 13,520 × g for 20 min and resuspension in 400 μl of rehydration solution containing 7 M urea and 2 M thiourea. We quantified the proteins according to the method of Bradford (1976) and assessed the quality of the samples by one- and two-dimensional sodium-dodecyl-sulfate–polyacrylamide-gel electrophoresis (SDS-PAGE; Laemmli, 1970; Supplementary Figure S1).

We reduced the protein samples in rehydration solution with dithiothreitol in 50 mM ammonium bicarbonate at a final concentration of 10 mM (45 min, 56°C) followed by alkylation with iodoacetamide at a final concentration of 20 mM in the same solution (40 min, room temperature, in the dark). We precipitated the protein solution with 0.2 volumes of trichloroacetic acid at -20°C for at least 2 h followed by centrifugation at 14,000 × g for 10 min (4°C) and then washed the pellet twice with cold acetone before a drying at room temperature. In the CEQUIBIEM facility (Center of Chemical and Biochemical Studies by Mass Spectrometry, Universidad de Buenos Aires, Argentina), we resuspended the proteins to a final concentration of 2 mg ml^-1 in 50 mM ammonium bicarbonate, pH 8, followed by digestion with 15 μg ml^-1 trypsin (Promega V5111). We then purified and desalted the peptides with ZipTip C18 columns (Millipore) before performing a nano-HPLC separation coupled to MS on an EASY-nLC 1000 instrument (Proxeon, Thermo Scientific) using an Easy-Spray Column PepMap RSLC (P/N ES801) suitable for separating protein complexes with a high degree of resolution by reverse-phase chromatography (C18, 2 μm, 100 Å, 50 μm × 150 mm). The flow rate used for the nano column was 300 nl min^-1 with the solvents A (0.1% [v/v] formic acid in water) and B (0.1% [v/v] formic acid in acetonitrile), ranging from 93% A: 7% B (5 min) to 65% A: 35% B (120 min), and a 2-μl injection volume. The MS equipment contained a high-collision dissociation cell for fragmentation and an Orbitrap analyzer (Thermo Scientific, Q-Exactive).

For data acquisition, we used the XCalibur 3.0.63 (Thermo Scientific) software with an equipment configuration that performed a simultaneous peptide identification during the chromatographic separation. The equipment made a full MS (resolution: 70,000) and a tandem MS/MS with the 15 best signal-to-noise–ratio peaks in each cycle (resolution: 17,500), with a dynamic exclusion range to avoid fragmentation of the same peak more than once during the same elution profile of the chromatogram. This method facilitated the highest number of cycles per unit time. A voltage of 3.5 kV was used for electro-spray ionization (Thermo Scientific, EASY-SPRAY).

We calculated the areas under the peaks by means of the algorithms used by the Proteome Discoverer software (Thermo Scientific, 1.4 version). The program was designed to search within the B. diazoefficiens USDA 110 data base UP000002526 (Uniprot) after the execution of a virtual trypsin digestion. We performed Proteome Discoverer searches with a precursor-ion tolerance of 10 ppm and a fragment-ion-mass tolerance of 0.05 Da specifying the oxidation of methionine as the dynamic modification and carbamidomethylation as the static modification. We set the peptide level of confidence at “high” (at a 1% false-discovery rate) and finally processed the data of the areas obtained with the Perseus software (Max Planck Institute of Biochemistry, 1.5.5.3 version, freely available), which informatics executes an in-depth statistical analysis. We replaced the missing values with 0 when proteins were present in both instances (i.e., Mtl and Ara) and with 1 when proteins were absent in one (i.e., either Mtl or Ara). Through this algorithm, we were able to identify 3,334 proteins, which species are detailed in Supplementary Table S1.

We searched the protein lists obtained differentially with Mtl or Ara with the Rhizobase¹, Uniprot², and KEGG Pathways³ databases to look for each protein’s possible function. We then classified the proteins according to the pathways in which those species might take part and used the proteins’ gene tags from Rhizobase to locate them in the KEGG-Pathways maps. Then, we focused on those pathways for which groups of enzymes-instead of single enzymes-were differentially found either in HMY-Mtl or HMY-Ara-grown cultures, to raise hypotheses about the differential metabolic fluxes in each carbon source. Finally, to validate these hypotheses, we continued evaluating whether the other enzymes of each candidate pathway were actually detected by our proteomic experiments, even if these latter enzymes were not included in the differential pools (Supplementary Table S1). With this information, we modeled the stoichiometry of the pathways, considering only those proteins that we detected experimentally at least once in our biological samples (Supplementary Tables S6–S12).

To look for possible operon organization, we searched those proteins that had successive gene tags (Rhizobase), in either differential pool of proteins, with the MicrobesOnline⁴ to corroborate whether those proteins constituted parts of operons in the B. diazoefficiens USDA 110 genome (Supplementary Tables S2, S3).

We extracted the PHB from whole cells by two successive centrifugations at 12,000 × g during 40 min each. Then, we homogenized pellets with sodium hypochlorite overnight at room temperature, precipitated with 1:1 alcohol-acetone, and resuspended in chloroform. Next, we quantified the PHB by a previously described spectrophotometric method (Law and Slepecky, 1961).

We determined the O₂ and CO₂ concentrations in the supplied and emitted gasses using a paramagnetic oxygen analyzer (Servomex 1100 A; Norwood, MA, United States) and an infrared CO₂ analyzer (Horiba PIR 2000; Japan) and measured the gas-flow rates with a bubble flow meter. We determined biomass dry weight as described (Herbert et al., 1971) and measured the substrate (Mtl or Ara) concentration in culture supernatants (from either the used HMY-Mtl or the HMY-Ara) by HPLC. To this end, we filtered the samples through 0.22-μm nitrocellulose membranes and manually injected 10 μl of each sample with a rheodyne 7725i injector into a Carbopac PA1 column maintained at 35 °C. The chromatography involved a mobile phase of 0.1 M NaOH and a flow rate of 1.0 ml.min^-1, with peak heights measured by a Waters 2465 electrochemical detector and the substrate concentrations quantified by means of the software Empower Pro, 2002 Waters.

We calculated the μ_max from the linear regression of the Napierian logarithmic values (ln) of biomass dry weight during the exponential growth phase and determined the qO_2max plus the CO₂ production by a mass-balance method (Cooney et al., 1977). We calculated the biomass yield and specific rates of substrate consumption as described (Roels, 1983).

We grew the wild-type and the lafR::Km mutant from starter cultures as described above, and when the bacteria reached the exponential growth phase, diluted each resulting suspension to (1.5 ± 0.2).10⁹ colony-forming units (CFUs).ml^-1 either in HMY-Mtl or HMY-Ara. We then inoculated 100 μl of each of these diluted suspensions into 10 ml of each medium followed by cultivation at 30°C with rotary shaking at 180 rpm. In parallel, we mixed 50 μl of each strain together and inoculated the resulting 100 μl of the 1:1 admixture into 10 ml of the corresponding culture medium, in order to ensure that at time zero of the experiment, all cultures started with the same total cell densities. Then, we continued cultivation under the same conditions until the cultures reached the stationary phase, which growth took 4.0 days in HMY-Mtl or 2.5 days in HMY-Ara. At these times we removed two 100-μl aliquots from each culture, used one for serial dilutions and CFU counts, and reinoculated the other into 10 ml fresh medium for further cultivation under the same conditions to strengthen the competition between the two strains. Once these subcultures reached the stationary phase, we repeated the process once more to obtain a third subculture, counting CFUs from each subculture after the appropriate dilution in yeast-extract mannitol agar (YMA) supplemented with either Sm (400 μg.ml^-1) for wild-type selection or Km (150 μg.ml^-1) for mutant selection in at least five replicas. The whole experiment was repeated in its entirety.

For proteomic comparisons, we employed duplicate or triplicate independent cultures (biologic replicates) from each culture medium and each subcellular fraction. With these replicates we constructed volcano plots to compare the samples using the logarithmic values of the ratio of the areas under the peaks along with the statistical significance of those values assessed by the Student-t-test (not shown). Hence, for each Mtl/Ara pair we plotted -log₁₀ Student-t-test p-value on the y axis versus log₂ fold change (Mtl area/Ara area) on the x axis. We considered differentially expressed proteins to be those with fold-changes greater than 2 (less than -1 or greater than 1 on the x axis of the graph) and with p-values lower than 0.05 (of values greater than 1.3 on the y axis of the graph). We also employed the same p-value threshold to determine the proteins present under only one condition (either growth in Mtl or Ara); then, for each C-source, combined the differentially expressed proteins—i.e., those with significantly greater fold-changes on that C-source and those exclusively present under that growth condition—in a single group to obtain the two lists comprising the Mtl-differential pool (MtlDP, Supplementary Table S2), and the Ara- differential pool (AraDP, Supplementary Table S3). Comparisons between these differential pools were carried out only for the groups of enzymes that belonged to the same pathway or part of a pathway, and that demonstrated consistent patterns in their variation. For all other experiments we used duplicate independent biologic replicas with at least three technical replicas each. The values reported are the means ± SD.

The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE (Vizcaíno et al., 2016) partner repository with the dataset identifier PXD008263 and 10.6019/PXD008263.

We obtained the total proteins from the cytoplasm- or membrane-enriched fractions of B. diazoefficiens USDA 110 grown in triplicate independent batch cultures in HMY medium, either with 0.5% (w/v) Mtl or 0.5% (w/v) Ara as the C-source. The bacteria were harvested in the exponential-growth phase at an OD_{500 nm} = 1.0. Upon assessing the quality of the samples by SDS-PAGE, we also corroborated that lateral flagellins were produced in only the HMY-Ara (Supplementary Figure S1). Next, the entire set of proteins obtained was analyzed by nano-HPLC coupled in tandem to mass spectrometry (nLC-MS/MS). Through an in-depth analysis of the proteomics data, we identified the 3,334 proteins listed in Supplementary Table S1, those representing 40% of the 8,317 predicted proteins of the whole B. diazoefficiens proteome (Kaneko et al., 2002), despite the differences in yield among the samples (Table 1). The membrane-enriched fractions obtained in HMY-Mtl and HMY-Ara contained 43 and 38% of the total proteins respectively (Table 1), which agrees with the 40% of proteins classified as cell wall (GO: 0005618), cell envelope (GO: 030313), and cell membrane (GO: 0044425) in the cell part fraction (GO: 0044464) in Microbes OnLine (consulted on March 28, 2018)⁵. We then looked for possible differences among the proteomes from HMY-Mtl or HMY-Ara cultures, including differential protein abundances along with the presence or absence of proteins in a given sample and averaged the results from at least two biologic replicas per condition and subcellular fraction. In this way, we identified 266 differential proteins pertaining to the HMY-Mtl cultures (Supplementary Table S2) and 237 to the HMY-Ara cultures (Supplementary Table S3)—that in the following will be referred to as the Mtl-differential pool (MtlDP) and the Ara-differential pool (AraDP), respectively. Although the proteins belonging to these pools were obtained from the cytoplasm- or membrane-enriched fractions, 115 proteins (43%) and 91 proteins (38%) appeared in both fractions in MtlDP and AraDP respectively. In addition, about one-third of the proteins that were observed as differentially expressed between Ara and Mtl in one fraction were not significantly so in the other, in at least one sample per fraction (Supplementary Tables S2, S3).

We looked for the possible functions of these proteins in the Rhizobase, the Uniprot, and the KEGG databases and found that only 21.8% of the proteins in the MtlDP and 35.0% in the AraDP were annotated with well defined functions (Supplementary Table S4). This subset of proteins is listed in Supplementary Table S5, there classified by function according to the gene-ontology (GO)–biological-processes assignment in Uniprot. On the basis of this subset, we intended to evaluate the differential cellular status in the HMY-Mtl, and HMY-Ara cultures; focussing on the enzymes clearly identified as participating in, at once, Mtl and Ara catabolism, PHB synthesis, and biotin metabolism along with those proteins involved in cell motility and chemotaxis.

In our attempt to deduce the metabolic pathways operating with one or the other C-source, we considered only sets of differentially expressed enzymes acting in the same pathway of B. diazoefficiens according to the KEGG database, since metabolic fluxes are seldom influenced by changes in a single enzyme of one pathway (Fell, 2005). By this means, we were able to identify substantial differences between the two metabolic pathways that may be inferred from the protein profiles obtained (Table 2).

Mtl and Ara appeared to be catabolized by the two rather different pathways that are summarized in Figure 1. The proteome of the Mtl-grown cells contained two salient features: (i) the lack of phosphofructokinase (Bll2580 and Blr4659; EC 2.7.1.11; Supplementary Table S1), indicating, in agreement with previous reports, that the Emden–Meyerhof–Parnas pathway is inactive under this condition (Sosa-Saavedra et al., 2001), and ii) the strong induction of phosphoribulokinase (Blr2582; EC 2.7.1.19) along with both subunits of ribulose-bisphosphate carboxylase-oxygenase (RuBisCO; EC 4.1.1.39), the latter encoded by the cbbL (blr2585) and cbbS (blr2586) loci. Because these enzymes are key members of the CBB cycle, their presence in only the MtlDP suggested that the carboxylation and/or oxygenation of ribulose 1,5 bisphosphate might have been taking place in the HMY-Mtl cultures. Five other enzymes were present in only the MtlDP that catalyzed reactions shared by the pentose-phosphate pathway (PP) and the CBB cycle (Figure 1 and Table 2). In addition, the Mtl-dehydrogenase activity, which catalysis is required to oxidize Mtl to D-fructose, was measured experimentally (Kuykendall and Elkan, 1977) but the gene encoding that enzyme was not annotated in the B. diazoefficiens genome. We observed that Blr3225, which enzyme is similar to the glucose/ribitol oxidoreductase, is present in the MtlDP. The corresponding structural gene is a neighbor of three genes encoding proteins also found in MtlDP—namely the putative sorbitol/Mtl transport system (blr3224), a D-ribitol kinase (EC 2.7.1.47; blr3226), and the D-fructokinase (EC 2.7.1.4; blr3227) that may be required to phosphorylate the D-fructose produced after Mtl reduction (Table 2). Therefore, we would suggest that Blr3225 is the missing inducible Mtl-dehydrogenase EC 1.1.1.67 described by Kuykendall and Elkan (1977). In addition, two other enzymes in the MtlDP cytoplasm-enriched fraction are also present at similar levels in the membrane-enriched fractions from HMY-Mtl and HMY-Ara cultures. Those proteins, which are the (putative) fructose-1,6-bisphosphate aldolase (Bll1521; EC 4.1.2.13) and the phosphoglycolate phosphatase (Bll3754; EC 3.1.3.18), might also carry out functions in HMY-Ara metabolism.

Data from the literature indicate that B. diazoefficiens is able to fix CO₂ through the CBB pathway (Hanus et al., 1979; Masuda et al., 2010), and that the affinity ratio of the B. diazoefficiens RuBisCO for CO₂ and O₂ is in the order of that of higher plants (Horken and Tabita, 1999). To ascertain whether B. diazoefficiens possessed activity of the RuBisCO carboxylase when grown in HMY-Mtl, we compared the development of bacterial colonies in solid HMY-Mtl or HMY-Ara, either in a normal atmosphere, or in one enriched with 5% (v/v) CO₂ (Figure 2). In a normal atmosphere the HMY-Mtl colonies had a slightly higher degree of mucoidy than the HMY-Ara colonies; but the colonies developed in HMY-Mtl in the atmosphere with 5% CO₂ were substantially more mucoid, indicating that under this condition B. diazoefficiens USDA 110 produced more EPS. In contrast, the colonies formed in HMY-Ara at 5% CO₂ were indistinguishable from those developing in the normal atmosphere, thus further corroborating that the changes observed were restricted to the Mtl-growth condition.

Ara degradation, for its part, is carried out by the L-KDA pathway—similar to the Entner–Doudoroff scheme—coupled to oxalate oxidation (Watanabe et al., 2006; Koch et al., 2014). Accordingly, 11 enzymes of those pathways were found in the AraDP pool (Table 2). The L-KDA sequence bifurcates into the pyruvate and glycolaldehyde branches upon the formation of the L-KDA intermediate, the latter being further oxidized to glyoxylate (Figure 1). Whereas pyruvate is decarboxylated to acetyl-CoA, glyoxylate may be in part converted to oxalate and oxidized to CO₂ through the participation of formyl-CoA:oxalate CoA-transferase (Frc, Bll3156; EC 2.8.3.16) and oxalyl-CoA carboxy-lyase (Oxc, Bll3157; EC 4.1.1.8). Another portion of the glyoxylate pool may be converted to glycerate through reaction with the NADH formed by oxalate oxidation and then continue on the pathway to pyruvate (Figure 1). Among the differentially expressed proteins, the Oxc was present in the AraDP cytoplasm-enriched fraction and was also observed at similar levels in the Ara and Mtl membrane-enriched fractions (Table 2). Therefore, the activity of this enzyme might be required in concert with that of the Frc—which enzyme, though being more abundant in the AraDP cytoplasm-enriched fraction, was absent in the MtlDP membrane-enriched fraction—in order to detoxify intracellular oxalate, which detoxification might also be required in cells grown on HMY-Mtl.

We obtained the balances of potential ATP production and O₂ consumption per C-mole of substrate to estimate the efficiency of energy generation by each one of the pathways schematized in Figure 1 as well as the production of pyruvate before that intermediate enters the TCA-cycle in order to assess the anaplerotic potential of the metabolic flux in that direction. To compare the energetic potentials of these pathways, we assumed that all the C-source is fully oxidized with maximum energy recovery as ATP, and assumed maximal P/O values of 3 for NADH and 2 for FADH₂. In addition, we assumed that all the NADH and FADH₂ produced in these pathways transfer their electrons to O₂, which under the present circumstances, would be the final electrons acceptor.

The energetic balance of the deduced Mtl catabolic pathway was difficult because, in lacking direct enzyme-activity measurements, we could not estimate to what extent the PP and CBB pathways were operating and, if the CBB pathway was in fact active, what the proportion of carboxylase versus oxygenase activity was in the RuBisCO catalysis. Therefore, we obtained the pathways from KEGG with the enzymes that we actually detected in our samples—independently of whether they were differentially expressed or not—and calculated the stoichiometries taking into account all three possibilities, as summarized in Table 3. If, as one possibility, the RuBisCO were not functional, we would expect that Mtl was catabolized by the PP pathway, as shown in Supplementary Figure S2, in which circumstance the stoichiometry can be readily calculated (Supplementary Table S6). If, alternatively, the RuBisCO were functional, the question would remain regarding the relative participation of the carboxylase versus the oxygenase activity, which consideration we approached by establishing three different scenarios: oxygenase activity alone, 50% of each activity, or carboxylase activity alone. When we tried to calculate the stoichiometries assuming oxygenase as the sole activity in the RuBisCO catalysis, we found an imbalance of three C atoms, which discrepancy we could compensate for in two ways: either by delivering the glycolate produced in the ribulose-1,5-bisphosphate oxygenation to the oxalate-oxidation pathway (Supplementary Figure S3), or by recycling part of the C atoms through the THF pathway (Supplementary Figure S4). This latter possibility was suggested by the induction of Bll6549 (Table 2); which locus, according to the KEGG Pathways, encodes a D-3-phosphoglycerate: NAD⁺ oxidoreductase (EC 1.1.1.95) catalyzing the conversion of D-3-phosphoglycerate into 3-phosphonooxypyruvate, and that product may then enter the THF pathway after transaminations from serine and glycine (Supplementary Figure S4). Supplementary Tables S7, S8, respectively, summarize the resulting stoichiometries. In addition to the higher yield of ATP and pyruvate upon THF-recycling (Table 3), we wish to specify that the oxalate oxidation requires at least two enzymes that are in the AraDP (Supplementary Table S7). Although these two enzymes are also present in Mtl-grown cultures (Table 2), all these results indicate that C-recycling through the THF pathway should be more likely than the direct glycolate oxidation through the oxalate-oxidation pathway. Nevertheless, we maintained the same set of alternative possibilities when we considered that RuBisCO catalysis involved 50% carboxylase and 50% oxygenase activities, as exemplified in Supplementary Figures S5, S6, which postulation led to the stoichiometries of Supplementary Tables S9, S10. Finally, the assumption of a sole RuBisCO-carboxylase activity (Supplementary Figure S7) led to a pathway that did not produce glycolate, so that all the carbon atoms of ribulose 1,5-bisphosphate, together with the fixed CO₂, were converted to glycerate, whose further metabolism to pyruvate would enable the possibility of entering into the TCA-cycle (Supplementary Table S11).

Regarding the L-KDA pathway in the Ara cultures, we confirmed the existence of the pathway deduced by Koch et al. (2014) and propose that the glycerate originating in one of the metabolic branches from glyoxylate is converted to pyruvate, which intermediate would then enter the TCA-cycle (Supplementary Figure S8). This proposal results in the construction of the stoichiometry of Supplementary Table S12.

According to these stoichiometric calculations, the potential yield of ATP was, at the most, 1.3 times higher in the Mtl- than in the Ara-catabolism calculations, whereas the potential O₂ consumption per C-mole of substrate was, at the most, 1.1 times higher in the Mtl than in the Ara calculations (Table 3). Furthermore, a high pyruvate production per C-mole of substrate was calculated in the Ara-catabolism calculations, which C yield might indicate a certain anaplerotic potential.

Four proteins related to the synthesis of nicotinamide-adenine dinucleotide (NAD) from L-aspartate are present in AraDP, although all of those gene products are simply upregulated; indicating that, as may be expected, none is completely absent from the HMY-Mtl cultures. (Figure 1 and Table 2). In addition, the 6,7-dimethyl-8-ribityllumazine synthase 2 (EC 2.5.1.78), which enzyme is involved in the synthesis of flavin-adenine dinucleotide (FAD), was observed in only the AraDP cytoplasmic-enriched fraction (Table 2). In parallel, one paralog of the A-chain of the ATP synthase (Bll1188) was observed in only the membrane-enriched fraction of the AraDP (Table 2). These results indicate that a higher production of electron acceptors might occur under these conditions, suggesting the possibility of a higher electron flux to the ETC in the HMY-Ara cultures than in the HMY-Mtl.

In contrast to these consistent trends, the changes in the ETC proteins were scattered (Supplementary Table S5), indicating either that we could not adequately separate these membrane-embedded proteins for MS analysis, or that a possible higher electron-flux through the ETC might be attained without major changes in the abundance of the ETC proteins.

Four proteins of the PHB pathway were found as differentially expressed in the cytoplasmic-enriched fractions between the two culture conditions. The regulator PhaR (Blr0227; Quelas et al., 2016b), the acetoacetyl-CoA reductase PhaB (Blr0225; EC 1.1.1.36), and the PHB synthase PhaC1 (Bll4360; EC 2.3.1.-)—all of which are involved in PHB synthesis—were found upregulated in MtlDP; whereas the inactive PhaC2 paralog of PHB synthase (Blr6073; Quelas et al., 2013) was present in only the cytoplasmic-enriched fraction of the AraDP (Table 2). In addition, the hypothetical protein Blr0908, which predicted gene product has more than a 65% identity to rhizobial PHB depolymerases, was likewise found in the AraDP in both the cytoplasmic- and the membrane-enriched fractions (Table 2). To our surprise, we also observed the truncated PHB synthase PhaC4 (Blr2885)—which protein appears to be inactive in PHB metabolism (Quelas et al., 2013)—in only the membrane-enriched fraction of the MtlDP (Supplementary Table S2).

If the PHB granules had coprecipitated with the membranes (Jendrossek and Pfeiffer, 2014), the presence of the proteins in the membrane-enriched fractions may actually be reflecting the fraction bound to PHB granules. Hence, we measured the PHB contents of HMY-Mtl- or HMY-Ara–grown cells in the stationary phase (OD_{500 nm} = 6.7 in the HMY-Mtl and 4.8 in the HMY-Ara cultures) and observed that HMY-Mtl-grown cells had 20.7 ± 0.6 fg of PHB.CFU^-1, whereas the HMY-Ara-grown cells had 13.7 ± 1.2 fg.CFU^-1, thus corroborating that the PHB synthesis and accumulation was higher under growth in HMY-Mtl.

Three proteins related to biotin biosynthesis were observed in only the MtlDP: 8-amino-7-oxononanoate synthase (Blr2097; EC 2.3.1.47) in the membrane-enriched fraction plus an ATP-dependent dethiobiotin synthetase (Blr2098; EC 6.3.3.3) and adenosylmethionine-8-amino-7-oxononanoate aminotransferase (Blr2099 and Blr2221; EC 2.6.1.62) in the cytoplasmic-enriched fraction—with the last of these (i.e., Blr2221) upregulated in that fraction. By contrast, in the AraDP a probable biotin-sulfoxide reductase encoded in bll6273 could be observed in only the cytoplasmic-enriched fraction (Table 2). These results suggested that a more active synthesis of biotin occurred in the HMY-Mtl-grown cells; whereas in the HMY-Ara cultures an accumulation of oxidized biotin would seem prevented (Figure 1). Biotin is a cofactor of various carboxylases, among which are the EC 6.4.1 ligases that catalyze the formation of carbon-carbon bonds (Erb, 2011). We observed two such carboxylases in the MtlDP: a putative propionyl-CoA carboxylase beta chain (Blr3940; EC 6.4.1.3) was present in only the membrane-enriched fraction, and a 3-methylcrotonyl-CoA carboxylase (alpha subunit Blr 4421 along with a beta subunit Blr0986; EC 6.4.1.4) was upregulated in both the cytoplasmic- and membrane-enriched fractions (Supplementary Table S2). In contrast, no carboxylases were detected in the AraDP.

The presence of the probable biotin-sulfoxide reductase in the AraDP cytoplasmic-enriched fraction suggested a higher oxidation rate of biotin. Therefore, upon searching for possible oxidative-damage–protecting proteins in the MtlDP and AraDP, we found two related to oxidative-damage mitigation in the MtlDP, and six in the AraDP (Supplementary Table S5). Their functions are somewhat superimposed, however; and in two instances the proteins that appeared as differentially expressed in one subcellular fraction were at similar levels in the other. In addition, since many paralogs exist for these functions, whether or not those differences are reflected in a true differential metabolism directed toward oxidative-damage mitigation remains unclear.

In agreement with our previous observations (Covelli et al., 2013; Mongiardini et al., 2017) and the results presented in Supplementary Figure S1B, we found the lateral-flagellar–filament structural flagellins Bll6865 and Bll6866, as well as the lateral-flagellar–hook protein Bll6858 in the AraDP (Table 2), but we failed to detect the majority of the other lateral-flagellar components. Nevertheless, we recently found the transcripts for these proteins produced in the HMY-Ara but not in the HMY-Mtl, under the control of the lafR transcriptional regulator (Mongiardini et al., 2017). Because flagellins and the flagellar-hook protein are the most abundant flagellar gene products, the lack of detection of the other proteins might have resulted from a low abundance or from a tight association with the flagellar structure that perhaps we did not manage to disassemble during the extraction. Indeed, we detected the rotor protein FliN_L (Bll6879) and the stator protein MotA (Bll6882)—but in only one sample each—in the HMY-Ara cultures (Supplementary Table S1). Moreover, the secreted flagellins observed in Supplementary Figure S1B require that at least the flagellar-export apparatus and the hook be assembled (Evans et al., 2014).

In addition to the lateral-flagellar structural proteins, we detected most of the proteins of one of the chemotaxis clusters in the AraDP. These proteins comprise those encoded in blr2343 to blr2347 plus blr2349—which loci are part of a functional chemotaxis cluster (Donati et al., 2011)—along with the putative methyl-accepting chemotaxis protein encoded in blr2931 (Table 2). This finding was somewhat surprising because the lateral flagella seemed not to respond to known chemotaxis signals, and Ara is not a strong chemoattractant for this strain (Quelas et al., 2016a). Moreover, although these proteins were detected in the cytoplasmic-enriched fraction, the majority were differential in the membrane-enriched fraction (Table 2). In particular, the functioning of the methyl-accepting chemotaxis proteins (Blr2345 and Blr2347) should occur in the membrane compartment.

According to Table 3, the efficiency of ATP production per C-mole of substrate (y_ATP/S), as well as the balance of O₂ consumption per C-mole of substrate would have been similar in the HMY-Mtl or HMY-Ara cultures. Nonetheless, earlier studies on the respiration of soybean-nodulating rhizobia had demonstrated that the O₂-consumption rate of the strain 505 was higher in Ara- than in Mtl-grown cells (Thorne and Burris, 1940). In agreement with that report, in the present work, we detected that in HMY-Ara cultures an activation of NAD- and FAD-biosynthetic enzymes, a reduction of PHB accumulation, and an increase in the biotin-reducing activity occurred, all of which observations would indicate that HMY-Ara–grown cells might harbor a shortage of reducing power and a surplus of oxidative species relative to HMY-Mtl–grown rhizobia. We therefore calculated the qO_2max and other growth parameters of B. diazoefficiens USDA 110 growing in HMY-Mtl or HMY-Ara in independent duplicate bioreactor batch cultures during the exponential-growth phase. The specific growth rate in the exponential phase (μ_max) proved to be not significantly different between those two carbon sources, but the qO_2max of the wild-type strain was almost three times higher in the HMY-Ara than in the HMY-Mtl (Table 4), thus corroborating the results obtained by Thorne and Burris (1940). In addition, the y_x/s was two times higher in the HMY-Mtl (Table 4). In agreement with these observations, the cultures in HMY-Ara entered the stationary phase of growth at around 40–50 h of culture, while those in HMY-Mtl were continuing in the exponential phase even after 70 h, thus producing a substantially higher biomass by this time (data not shown).

In accordance with the qO_2max values, the specific rates of substrate consumption (qS_max) and CO₂ production (qCO_2max) also were around three times higher in the cultures of B. diazoefficiens in the HMY-Ara than in those in the HMY-Mtl (Table 4). Since the amount of biomass produced per mole of ATP (y_x/ATP) was the same for both culture conditions and the stoichiometric balances failed to reveal any substantial differences in the efficiency of ATP production per C-mole of substrate, the results would indicate that the cultures in the HMY-Ara consumed a larger proportion of the catabolic energy produced for simple maintenance—encompassing processes different from the production of biomass or C-products—than did those in the HMY-Mtl. One of these maintenance processes requiring energy may be the generation and use of the lateral flagella, which appendages are present in the HMY-Ara cultures but are scanty in the HMY-Mtl cultures (Mongiardini et al., 2017; see also Supplementary Figure S1B). To evaluate the contribution of the lateral flagella to qO_2max in HMY-Ara, we repeated the above measurements in both culture media with a B. diazoefficiens USDA 110 lafR::Km insertional mutant lacking the class IB regulator of lateral-flagellum synthesis and therefore unable to produce any lateral flagellar structures, including the flagellar motor (Mongiardini et al., 2017). We observed that μ_max remained within the same values; but in contrast to the considerable elevation in O₂-consumption on the part of the wild strain in HMY-Ara, the qO_2max of the lafR::Km mutant was similar in both the HMY-Ara and the HMY-Mtl (Table 4), thus indicating that the absence of lateral flagella substantially reduced the enhancement in O₂-consumption rate with Ara as the sole C-source. In agreement with these results, the values of y_x/s, qS_max, and qCO_2max in the lafR::Km mutant were similar upon growth in the HMY-Mtl and the HMY-Ara (Table 4); but entry into the stationary phase remained earlier in the HMY-Ara cultures than in the HMY-Mtl.

The differences in qO_2max and y_x/s among the wild-type and the lafR::Km mutant in HMY-Ara (Table 4) suggested that this mutant might outcompete the wild-type if both strains were to grow together. To investigate this possibility, we coinoculated the strains either in HMY-Ara or HMY-Mtl, initiating the cultures with 1:1 admixtures containing ca. (7.5 ± 0.2).10⁶ CFUs.ml^-1 of each strain at time zero. We left the cultures until they reached the stationary phase and then counted the numbers of CFUs.ml^-1 using different antibiotic resistances to distinguish the wild-type and the mutant. As controls, we maintained parallel cultures of each strain alone without competition [initial CFU ml^-1 of each strain alone at time zero: (1.5 ± 0.2).10⁷] and observed that 100% of the CFUs obtained from those populations carried the antibiotic-resistance mark of the corresponding strain, indicating that no antibiotic-resistant spontaneous mutants arose in significant numbers during the course of the experiment.

At the time of sampling for the CFU counting of the cocultures, we also diluted these stationary-phase cells to initiate new cultures, repeating the initial process twice in order to strengthen the competition among the strains in the second and third subcultures. Thus, we observed that after replicating the bacterial admixtures in successive cultures, the lafR::Km mutant clearly outcompeted the wild-type strain in HMY-Ara, reaching more than 95% of the cell population by the third subculture (Figure 3). Higher numbers of the mutant cells, however, also developed in the HMY-Mtl, but in this instance the competition was not so pronounced as in the HMY-Ara. These results were not attributable to intrinsic differences in survival since in the controls where the strains were cultured axenically the total number of CFUs.ml^-1 remained constant throughout the successive subcultures (not shown).