Lipogenesis and Redox Balance in Nitrogen-Fixing Pea Bacteroids

ABSTRACT

Within legume root nodules, rhizobia differentiate into bacteroids that oxidize host-derived dicarboxylic acids, which is assumed to occur via the tricarboxylic acid (TCA) cycle to generate NAD(P)H for reduction of N₂. Metabolic flux analysis of laboratory-grown Rhizobium leguminosarum showed that the flux from [¹³C]succinate was consistent with respiration of an obligate aerobe growing on a TCA cycle intermediate as the sole carbon source. However, the instability of fragile pea bacteroids prevented their steady-state labeling under N₂-fixing conditions. Therefore, comparative metabolomic profiling was used to compare free-living R. leguminosarum with pea bacteroids. While the TCA cycle was shown to be essential for maximal rates of N₂ fixation, levels of pyruvate (5.5-fold reduced), acetyl coenzyme A (acetyl-CoA; 50-fold reduced), free coenzyme A (33-fold reduced), and citrate (4.5-fold reduced) were much lower in bacteroids. Instead of completely oxidizing acetyl-CoA, pea bacteroids channel it into both lipid and the lipid-like polymer poly-β-hydroxybutyrate (PHB), the latter via a type III PHB synthase that is active only in bacteroids. Lipogenesis may be a fundamental requirement of the redox poise of electron donation to N₂ in all legume nodules. Direct reduction by NAD(P)H of the likely electron donors for nitrogenase, such as ferredoxin, is inconsistent with their redox potentials. Instead, bacteroids must balance the production of NAD(P)H from oxidation of acetyl-CoA in the TCA cycle with its storage in PHB and lipids.

IMPORTANCE Biological nitrogen fixation by symbiotic bacteria (rhizobia) in legume root nodules is an energy-expensive process. Within legume root nodules, rhizobia differentiate into bacteroids that oxidize host-derived dicarboxylic acids, which is assumed to occur via the TCA cycle to generate NAD(P)H for reduction of N₂. However, direct reduction of the likely electron donors for nitrogenase, such as ferredoxin, is inconsistent with their redox potentials. Instead, bacteroids must balance oxidation of plant-derived dicarboxylates in the TCA cycle with lipid synthesis. Pea bacteroids channel acetyl-CoA into both lipid and the lipid-like polymer poly-β-hydroxybutyrate, the latter via a type II PHB synthase. Lipogenesis is likely to be a fundamental requirement of the redox poise of electron donation to N₂ in all legume nodules.

INTRODUCTION

Biological reduction (or fixation) of atmospheric nitrogen (N₂) to ammonia (NH₃) provides up to 50% of the biosphere's available nitrogen, mostly through symbioses between soil bacteria (rhizobia) and legumes (1, 2). These symbioses are initiated by rhizobia infecting legume roots, resulting in the formation of nodules. Rhizobia differentiate into N₂-fixing bacteroids that express nitrogenase to reduce N₂ to NH₃ under microaerobic conditions (3). Bacteroids receive carbon from the legume while secreting NH₃ to the plant. The overall stoichiometry of N₂ fixation under ideal conditions is as follows:

$${\text{N}}_2+8{\text{e}}^{-}+8{\text{H}}^{+}+16\text{ATP}\to 2{\text{NH}}_3+{\text{H}}_2+16\text{ADP}+16{\text{P}}_{\text{i}}$$ (1)

Thus, 8 moles of electrons and protons and 16 moles of ATP reduce a single mole of N₂, making N₂ fixation energetically expensive.

Legumes energize bacteroid N₂ fixation by supplying dicarboxylates, principally malate (4), which must be oxidized to yield ATP and electrons to reduce N₂. Bacteroids metabolize malate by the use of NAD⁺-dependent malic enzyme (5–7) and pyruvate dehydrogenase to provide acetyl coenzyme A (acetyl-CoA), which can be completely oxidized in the tricarboxylic acid (TCA) cycle, yielding FADH₂ and NAD(P)H. The standard model is that NAD(P)H supplies electrons both to nitrogenase via ferredoxin, or via an equivalent low-potential electron donor, and to an electron transport chain for ATP synthesis (8, 9).

This model is supported by work in Rhizobium leguminosarum and Sinorhizobium meliloti, where TCA cycle mutants are unable to fix N₂ in symbiosis with pea (Pisum sativum) and alfalfa (Medicago sativa), respectively (10–13). However, the TCA cycle provides both reductant and biosynthetic precursors, so the abolition of N₂ fixation in these mutants could be due to insufficient NAD(P)H to directly power nitrogenase or, equally, could result from biosynthetic deficiencies. In contrast, in soybean (Glycine max) bacteroids, the TCA cycle either is dispensable for N₂ fixation or can be bypassed, with isocitrate dehydrogenase and 2-oxoglutarate dehydrogenase mutants of Bradyrhizobium japonicum able to fix N₂ at wild-type rates (14, 15). Moreover, standard midpoint potentials indicate that NAD(P)H is unlikely to donate electrons directly to ferredoxin (the E⁰′ for NAD⁺/NADH is −320 mV, for NADP⁺/NADPH is −324 mV, and for ferredoxin [Fe³⁺/Fe²⁺] is −484 mV) (16, 17). Thus, some other, as-yet-undefined mechanism must exist to transfer electrons to nitrogenase in root nodule bacteroids.

Finally, N₂-fixing bacteroids in nodules formed by soybean and common bean (Phaseolus vulgaris) accumulate large quantities of the lipid-like polymer poly-β-hydroxybutyrate (PHB), while bacteroids from pea, alfalfa, and clover (Trifolium spp.) apparently do not (18). While abolishing PHB synthesis does not adversely affect N₂ fixation rates in soybean and common bean (19–21), in Azorhizobium caulinodans, mutation of PHB synthase prevents N₂ fixation in both free-living and symbiotic forms (22), implying a fundamental role for PHB synthesis in at least some N₂-fixing rhizobia.

Determining how N₂ is fixed by bacteroids, in what is arguably the second most important nutrient assimilation cycle after photosynthesis, requires an understanding of bacteroid carbon metabolism. Metabolic profiling and flux analysis as well as mutational and N₂ fixation studies were used to investigate carbon flow in bacteroids. Remarkably, the results revealed that the TCA cycle is not the only sink for plant-derived carbon in symbiotic N₂ fixation; rather, pea bacteroids divert appreciable quantities of acetyl-CoA into the production of lipid or PHB. N₂-fixing bacteroids are therefore inherently lipogenic, and this is probably a metabolic requirement for N₂ fixation.

MATERIALS AND METHODS

Bacterial strains and culture conditions.

Bacterial strains and plasmids used in this study are detailed in Table 1. Rhizobium leguminosarum bv. viciae (Rlv3841) was grown at 28°C on tryptone yeast extract (TY) (28) or acid minimal salts (AMS) medium (29) with succinate (20 mM) and NH₄Cl (10 mM) as the sole carbon source and nitrogen source, respectively. Where appropriate, antibiotics were used at the following concentrations: streptomycin, 500 µg/ml; neomycin, 80 µg/ml; spectinomycin, 50 µg/ml; gentamicin, 20 µg/ml; and ampicillin, 50 µg/ml.

TABLE 1.

Strains, plasmids, and primers used in this study

Strain, plasmid, or primer	Genotype and/or description or sequencea	Reference or source
Strains
R. leguminosarum Rlv3841	Str derivative of R. leguminosarum bv. viciae strain 300	23
R. leguminosarum RU137	Rlv3841 phaC1::Tn5 Nmr	12
R. leguminosarum RU116	Rlv3841 sucD::Tn5 Nmr	12
R. leguminosarum RU156	Rlv3841 sucA::Tn5 Nmr	12
R. leguminosarum RU724	Rlv3841 sucA::Tn5-lacZ Nmr	12
R. leguminosarum RU725	Rlv3841 sucC::Tn5-lacZ Nmr	12
R. leguminosarum RU726	Rlv3841 sucB::Tn5-lacZ Nmr	12
R. leguminosarum RU733	Rlv3841 sucA::Tn5-lacZ Nmr	12
R. leguminosarum LMB814	Rlv3841 phaC2::Ω Str Spr	This work
R. leguminosarum LMB816	Rlv3841 phaC1::Tn5 phaC2::Ω Str Nmr Spr	This work
E. coli DH5α	Strain used for cloning; F− ϕ80lacZΔM15 Δ(lacZYA-argF)U169 recA1 endA1 hsdR17(rK− mK+) phoA supE44 thi-1 gyrA96 relA1	Invitrogen
Plasmids
pJET1.2/Blunt	PCR product cloning vector; Apr	Thermo-Fisher
pHP45-ΩSmSp	pHP derivative with ΩSmSp cassette; Smr Spr	24
pJQ200SK	pACYC derivative, P15A origin of replication insertional mutagenesis inactivation vector; Gmr Sucs	25
pRK2013	Helper plasmid used for mobilizing plasmids; ColE1 replicon with RK2 tra genes; Kmr	26
pLMB834	pr1645-1646 PCR product (2.8 kbp) from pRL100105 (phaC2) cloned into pJET1.2/Blunt; Apr	This work
pLMB835	pLMB834 with ΩSmSp cassette from pHP45-ΩSmSp cloned into unique EcoRI site; Apr Smr Spr	This work
pLMB838	pJQ200SK with BglII fragment from pLMB835 containing phaC2::Ω cloned into BamH1 site; Smr Spr Gmr Sucs	This work
Primers
pr1645	AACGCTACAGCGCAACGCTC	This work
pr1646	ACTTTCTTCGCTCCCGTCGG	This work
pr1647	ACCCCGAAGACGCTCGTCAT	This work
pr1648	ATGATCGTGACGGCATCGGC	This work
potfarforward	GACCTTTTGAATGACCTTTA	27
Tn5-1	ATAGCCTCTCCACCCAAGC	This work

^aAp, ampicillin; Gm, gentamicin; Km, kanamycin; Nm, neomycin; Sp, spectinomycin; St, streptomycin; Suc, succinate.

Metabolic flux analysis.

Rlv3841 cells grown in succinate-NH₄Cl AMS medium were harvested at mid-log phase (optical density at 600 nm [OD₆₀₀] of approximately 0.5) and subcultured into fresh AMS media to a reach a starting OD₆₀₀ of 0.02, with 20 mM [¹³C₄]succinate (20% fractional abundance). Cells were harvested at an OD₆₀₀ of 0.3 and centrifuged at 8,500 × g for 5 min. The resulting pellet was washed with fresh AMS medium and centrifuged, and the resulting cell pellet was extracted in 80% (vol/vol) ethanol at 80°C for 5 min prior to centrifugation at 12,000 × g for 5 min. The supernatant containing the soluble amino acids, organic acids, and sugars was dried by vacuum centrifugation. The insoluble pellet was rapidly frozen in liquid N₂ and freeze-dried. Protein in the insoluble fraction was hydrolyzed to its component amino acids by incubation with 6 M HCl for 24 h at 100°C.

Gas chromatography-mass spectrometry (GC-MS) analysis of derivatized amino acids, organic acids, and sugars was performed on an Agilent 7890A GC/5975C quadrupole MS system as described elsewhere (30). Amino acids and organic acids were analyzed after derivatization using N-tert-butyldimethylsilyl-N-methyltrifluoroacetamide (MTBSTFA) or N-methyl-N-(trimethylsilyl)-trifluroacetamide (MSTFA); sugars were treated with methoxyamine hydrochloride and then derivatized with MSTFA. Protein-derived and soluble amino acids were examined separately. Mass isotopomer abundances were quantified using Chemstation and corrected for the presence of naturally occurring heavy isotopes introduced during derivatization. The chemical fragments used for metabolic flux analysis are detailed in Table S1 in the supplemental material.

Metabolic modeling was performed with 13C-FLUX (version 20050329) using an iterative procedure described before (30, 31). A complete description of the model, which also defines the network carbon atom transitions, is provided in Table S2 in the supplemental material, and net flux data are provided in Table S3 in the supplemental material. During initial parameter fitting, fluxes corresponding to biomass outputs were allowed to vary, and the mean values from the 10 best-fit estimates were then used to constrain the network output flux values in subsequent simulations. Malate and oxaloacetate were combined into a single metabolite pool, as were phosphoenolpyruvate and pyruvate, to improve the determinability of fluxes between these intermediates. No adjustments were required to compensate for the contribution of preexisting unlabeled pools of metabolites. Molar fluxes are reported relative to a succinate uptake flux value of 1.

Material for metabolite profiling.

To prepare samples of free-living Rlv3841, six independent cultures of Rlv3841, derived from six isolated colonies of the strain, were grown in AMS medium on a gyratory shaker at 250 rpm to an OD₆₀₀ of 0.4. Cell pellets were collected by centrifugation (5,000 × g, 5 min), washed with isolation buffer (8 mM K₂HPO₄, 2 mM KH₂PO₄, 2 mM MgCl₂), and stored at −80°C for later use in metabolite profiling.

To prepare bacteroid and nodule cytosolic samples, seeds of P. sativum cv. Avola were surfaced sterilized with 70% (vol/vol) ethanol for 30 s, rinsed once in sterile water, and then immersed in a 2% (wt/vol) NaOCl solution for 2 min, prior to 10 rinses in sterile water. Seeds were sown into 2-liter beakers containing washed and autoclaved fine-grade vermiculite. Six independent cultures of test strain Rlv3841 or test strain RU116, derived from six isolated colonies of each strain, were prepared. A 1-ml aliquot of each culture was inoculated into a minimum of six pots, at cell densities of between 5 × 10⁷ and 9 × 10⁷ cells ml⁻¹. Seeds were initially sown in duplicate and thinned to one plant per pot after 7 days. Plants were watered once with 250 ml nitrogen-free nutrient solution as previously described (29) and were incubated in an illuminated environment-controlled growth room at 22°C on a 16-h day, 8-h night cycle.

Plants were harvested at 28 days postinoculation (dpi) for metabolomic profiling. Approximately 1.5 g of nodule tissue was excised from plants from each set of pots. Nodules were ground in isolation buffer (8 mM K₂HPO₄, 2 mM KH₂PO₄, 2 mM MgCl₂), and the homogenate was passed through muslin and centrifuged (250 × g for 5 min) to remove plant debris before a further round of centrifugation (5,000 × g, 10 min) was performed to pellet the bacteroids. The resulting supernatant, representing the nodule cytosol fraction, was freeze-dried; the pellet, representing the bacteroid fraction, was washed twice with isolation buffer and centrifuged (5,000 × g, 10 min); and the pellets were frozen at −80°C for later use in metabolite profiling.

Metabolite profiling platform.

Determinations of metabolomic profiles of free-living, bacteroid, and nodule cytosol samples were each performed using nonbiased, global metabolome profiling technology based on GC-MS and ultra-high-performance liquid chromatography/mass spectrometry/tandem mass spectrometry (UHLC/MS/MS²) platforms (32, 33) developed by Metabolon (Durham, NC). Six replicate samples from each treatment (free-living, bacteroid, and nodule cytosol samples) were extracted using an automated MicroLab Star system (Hamilton). Recovery standards were added prior to the first step in the extraction process for quality control purposes. To monitor total process variability, a series of technical replicates were taken from a pool made from small aliquots of all of the experimental samples. These were spaced evenly among the randomly ordered experimental samples, and all consistently detected metabolites were monitored for reproducibility. Sample preparation was conducted using methanol extraction to remove the protein fraction while allowing maximum recovery of small molecules. The resulting extract was divided into two fractions, one for analysis by LC and one for analysis by GC. Samples were placed briefly on a TurboVap evaporator (Zymark) to remove the organic solvent. Each sample was frozen and dried under a vacuum. Samples were then prepared for the appropriate instrument (either LC/MS or GC/MS).

The LC/MS portion of the platform was based on a Waters Acquity UHPLC and a Thermo-Finnigan LTQ mass spectrometer, which consisted of an electrospray ionization source and a linear ion trap mass analyzer. The sample extract was split into two aliquots, dried, and then reconstituted in acidic or basic LC-compatible solvents, each of which contained 11 or more injection standards at fixed concentrations. One aliquot was analyzed under acidic positive-ion optimized conditions and the other under basic negative-ion optimized conditions in two independent injections using separate dedicated columns. Extracts reconstituted in acidic conditions were subjected to gradient elution using water and methanol, with both containing 0.1% (vol/vol) formic acid, while the basic extracts, which also used water and methanol, contained 6.5 mM NH₄HCO₃. The MS analysis alternated between MS and data-dependent MS/MS scans using dynamic exclusion.

The samples destined for GC/MS analysis were redried using vacuum desiccation for a minimum of 24 h prior to being derivatized using dried N₂ and bistrimethyl-silyl-trifluoroacetamide (BSTFA). The GC column was 5% phenyl, and the temperature ramp was 40°C to 300°C over a 16-min period. Samples were analyzed on a Thermo-Finnigan Trace DSQ fast-scanning single-quadrupole gas chromatograph mass spectrometer using electron impact ionization.

Compound identification, data handling, and statistical analysis.

For metabolite profiling, identification of known chemical entities was based on comparisons to metabolomic library entries of purified standards as previously described (33, 34). Statistical analysis was performed using the software packages Array Studio (Omicsoft) and R (http://www.r-project.org/). Where a given metabolite was not detected in a particular sample, the observed minimum detected value for that metabolite from the analysis was assigned, under the assumption that the missing values were not random but resulted from the amount of the compound being below the limit of detection. Data for free-living and bacteroid samples were then normalized to protein content, as determined by Bradford assay (35). For the comparison of the bacteroids to the nodule cytosol, normalization was performed by extracting proportional amounts of bacteroid and cytosolic fractions of matched starting samples. That is, the total yield of bacteroid and cytosolic fractions for each sample was known, and a constant percentage of each fraction was analyzed in order to compare the relative amounts of metabolites in each fraction. The statistical model utilized the matched-pair nature of the samples to account for absolute differences between the samples. Welch's two-sample t test was used to identify metabolites that differed significantly between experimental groups (P < 0.05), and the false-discovery rate (FDR) was also calculated (36) to account for the multiple comparisons that normally occur in metabolomic-based studies (Q < 0.1). Thus, metabolites were considered to be significantly different if they met the statistical significance criteria (P < 0.05 and Q < 0.10).

Assessment of N₂ fixation.

Plants intended for assessment of N₂ fixation were grown as described above in “Material for metabolite profiling,” with the following exceptions. For measurement of N₂ fixation by acetylene reduction assay, plants were grown in 1-liter pots and harvested at the onset of flowering (21 dpi). Whole plants were removed from growth pots and transferred to 250-ml sealed bottles. After the rates of acetylene reduction of detached nodules were measured, nodules were excised and immediately transferred into a 25-ml bottle and assayed. Rates of N₂ fixation were determined by the amount of acetylene reduced after 1 h in an atmosphere consisting of 95% air–5% acetylene, as previously described (37). Following the acetylene reduction assay, bacteroid protein was quantified by excising nodules from roots and grinding in 40 mM HEPES (pH 7.0). The homogenate was passed through muslin and the eluate centrifuged (250 × g for 5 min) to remove plant debris. The supernatant was then centrifuged (5,000 × g, 10 min) to pellet the bacteroids. Bacteroids were lysed by two rounds of ribolysing on a FastPrep FP120 Ribolyser (BIO101/Savant) at a setting of 6.5 for 30 s, with samples kept on ice for 5 min between rounds. The protein content in the resulting supernatant was determined by Bradford assay (35) using a Pierce Coomassie assay kit (Pierce; catalog no. 23200) with bovine serum albumin (BSA) as the protein standard.

For assessment of N₂ fixation by plant biomass accumulation, plants were grown in 2-liter pots and were supplied with 200 ml of additional sterile water at 28 dpi. Plants were then harvested at 47 dpi by cutting shoots below the hypocotyl and drying at 60°C for 48 h prior to weighing.

Lipid analysis.

Bacteroids were collected from nodules harvested from plants grown for analysis of lipids as described in “Material for metabolite profiling” and harvested at 28 dpi. Nodules were ground in 20 mM HEPES buffer (pH 7.0) and purified by Percoll gradient (38). Cells of free-living Rlv3841 were grown in AMS medium with succinate and NH₄Cl and harvested at an OD₆₀₀ of 0.4 to 0.6 by centrifugation (5,000 × g for 10 min). The resultant bacteroid and cell pellets were stored at −80°C for later use. Bacteroid and cell pellets were lysed by the use of a Ribolyser as described above and centrifuged (10,000 × g for 10 min). The supernatant was then centrifuged (20,000 × g for 20 min) prior to further ultracentrifugation (60,000 × g for 60 min) to remove cell membranes. The supernatant was concentrated by vacuum centrifugation prior to lipid quantification performed using a triglyceride determination kit (Sigma; catalog no. TR0100). Protein determination was performed using the Bradford assay as described above.

Mutant construction and phenotyping.

To construct the phaC2 (pRL100105) mutant of Rlv3841, primers pr1645 and pr1646 (see Table S1 in the supplemental material) were used to amplify 2.8 kb of the region containing the gene and the PCR product was cloned into pJET1.2/blunt, giving plasmid pLMB834. The Ω-streptomycin/spectinomycin cassette from pHP45-ΩSmSp was cloned into the unique EcoRI site of pLMB834 to produce pLMB835. The BglII fragment from pLMB835 was cloned into pJQ200SK to produce pLMB839. Plasmid pLMB839 was then conjugated into strain Rlv3841, using pRK2013 as a helper plasmid, to produce phaC2 mutants as previously described (5), resulting in LMB814. The mutation was confirmed by PCR mapping using primer pairs pr1648-potfarforward and pr1657-potfarforward. Strain LMB816, the phaC1 (RL2098) phaC2 (pRL100105) double mutant, was made by using the general transducing phage RL38 to lyse strain RU137. The kanamycin-marked phaC1::Tn5 mutation was then back-transduced into LMB814 to generate LMB816, as previously described (39), and the mutation was confirmed by PCR mapping with pr1647-potfarforward, pr1648-potfarforward, and pr1647-Tn5-1 primer pairs. Assessment of N₂ fixation of the resulting mutants was performed as described above. Transmission electron microscopy was performed on nodules harvested from plants at 28 dpi, and the methods for nodule sectioning, staining, and microscopy were performed as detailed previously (20).

RESULTS

Metabolic flux analysis of free-living rhizobia.

Dicarboxylates are provided to bacteroids by plants to support N₂ fixation (3, 4), so the pathways operating in free-living Rhizobium leguminosarum bv. viciae 3841 (Rlv3841) growing on [¹³C₄]succinate were quantified. The major flux of succinate metabolism in Rlv3841 was via fumarate to malate (Fig. 1) and subsequently from malate to pyruvate and oxaloacetate to phosphoenolpyruvate. These fluxes would support the major metabolic requirements of cells growing on a TCA cycle intermediate for synthesis of acetyl-CoA to supply the TCA cycle and phosphoenolpyruvate for biosynthesis of sugars. Large fluxes were also detected in gluconeogenesis converting phosphoenolpyruvate to triose phosphates, in the oxidative decarboxylation of pyruvate to acetyl-CoA, and in the TCA cycle from oxaloacetate to 2-oxoglutarate. Overall, these fluxes are consistent with respiration of an obligate aerobe growing on a TCA cycle intermediate as the sole carbon source.

FIG 1.

Flux map of central carbon metabolism for free-living R. leguminosarum Rlv3841, grown on succinate and NH₄Cl. Net fluxes are expressed on a molar basis relative to succinate uptake. The thickness of each arrow is proportional to the net flux with the exception that fluxes representing <1% of succinate uptake are indicated by broken arrows. Biosynthetic outputs are shown in solid boxes, and metabolites treated as a single pool in the model are shown in dashed boxes. Flux identifiers, defined in Table S2 in the supplemental material, are shown in italics. The precise values for the deduced fluxes are presented in Table S4 in the supplemental material. Standard abbreviations are used for amino acids and metabolic intermediates, and PPP data represent the reversible nonoxidative steps of the pentose phosphate pathway.

Currently, metabolic flux analysis cannot be conducted on notoriously fragile isolated pea bacteroids (40). Nitrogenase activity, as measured by acetylene reduction, in isolated pea nodules collapsed 90 min after excision to less than 2% of that in nodules on roots (0.25 ± 0.03 versus 18.3 ± 2.5 nmol acetylene reduced · mg nodule⁻¹ · h⁻¹). This precludes labeling of nodule metabolites at the isotopic steady state under physiologically relevant conditions in an isolated system. Moreover, the likely low rate of protein turnover in nondividing bacteroids compromises the use of the labeling patterns of protein-derived amino acids to reflect those of their metabolic precursors. We therefore used metabolite profiling to examine the differences in levels of metabolic intermediates between cultured cells and bacteroids.

Bacteroid central metabolism.

The metabolic profiles of free-living and bacteroid forms of Rlv3841 were analyzed using nonbiased, untargeted metabolome analysis (32, 33). The metabolites whose levels were most highly elevated in bacteroids relative to free-living Rlv3841 were homoserine and asparagine (105- and 58-fold increases, respectively; Fig. 2). Levels of both were also high in the nodule cytosolic fraction relative to bacteroids (33- and 11-fold increases; Table S4 in the supplemental material), in accordance with previous observations (41, 42) and consistent with their known plant origin. Asparagine is made in the plant cytosol as the primary nitrogen export product from nodules (40). Furthermore, free asparagine is not made by Rlv3841, which, from analysis of its genome, uses the GatCAB pathway to insert asparagine into proteins by charging asparaginyl-tRNA with aspartate and then transamidating aspartate to asparagine (43). In addition, catabolism of asparagine and homoserine is not upregulated in bacteroids (44), nor do catabolic mutants show reduced N₂ fixation rates (45, 46), consistent with minor roles in symbiosis.

FIG 2.

Metabolite profile of Rlv3841 bacteroids versus Rlv3841 free-living cells showing fold change in metabolite abundance relative to Rlv3841 bacteroids. Bacteroids were isolated from nodules from pea plants 28 days postinoculation (dpi). Cells were harvested from log-phase cultures grown in AMS broth with 20 mM succinate and 10 mM NH₄Cl as the carbon and nitrogen sources, respectively. Bolded intermediates were detected by metabolite profiling, with a statistically significant fold difference (P < 0.05 by Welch's t test and Q < 0.1 for the false-discovery rate) denoted with a red (increase) or green (decrease) arrow. A > sign indicates that the metabolite was undetectable in either the free-living or the bacteroid sample, so the difference reported is therefore a lower-limit estimate of the fold change. Intact arrows indicate single-step enzyme-catalyzed reactions. Broken arrows indicate instances in which two or more enzyme-catalyzed steps are involved in a series of reactions. Abbreviations: UD, undetectable; BT, bacteroid; FL, free-living; GABA, γ-amino butyric acid; GSH, glutathione (reduced); GSSG, glutathione (oxidized); 2OG, 2-oxoglutarate; OAA, oxaloacetate; PEP, phosphoenolpyruvate.

Our fundamental issue of investigation is whether the TCA cycle is altered during symbiotic N₂ fixation. The dicarboxylates malate, fumarate, and succinate are the carbon sources for bacteroids in planta, and levels of all three were increased in bacteroids relative to free-living cells (Fig. 2). Moreover, levels of these metabolites were also much higher in the plant nodule cytosol fraction relative to bacteroids (malate 14-, fumarate 20-, and succinate 2.5-fold increased [see Table S4 in the supplemental material]), consistent with active plant dicarboxylate synthesis.

Metabolism of dicarboxylates by bacteroids is performed via malic enzyme and phosphoenolpyruvate carboxykinase to pyruvate and phosphoenolpyruvate, respectively, with pyruvate subsequently oxidatively decarboxylated to acetyl-CoA (5–7). The levels of intermediates of sugar metabolism such as 3-phosphoglycerate, fructose-6-phosphate, and glucose-6-phosphate and the pentose phosphate pathway (ribulose-5-phosphate and xylulose-5-phosphate) were greatly reduced (Fig. 2), suggesting that little sugar synthesis occurs in bacteroids. Remarkably, levels of pyruvate (5.5-fold reduced), acetyl-CoA (50-fold reduced), free coenzyme A (33-fold reduced), and citrate (4.5-fold reduced) were much lower in bacteroids (Fig. 2). In sharp contrast, the transcription and enzymatic activity of citrate synthase (RL2234; icdB) were increased 3.2- and 12-fold, respectively, and increases in the activity and transcription of other enzymes of the decarboxylating arm of the TCA cycle have been noted (44, 47). While such increased enzyme biosynthesis might indicate increased flux into the TCA cycle, it is equally consistent with lower feedback inhibition of the synthesis and activity of enzymes displayed by key intermediates such as acetyl-CoA and citrate (48, 49).

Carbon in the TCA cycle could also be channeled to glutamate, which is synthesized from 2-oxoglutarate by glutamine synthetase/glutamate synthase (GS/GOGAT) (50). However, glutamate levels were 20-fold lower in bacteroids relative to free-living cells (Fig. 2), consistent with GS/GOGAT activity being both low and not essential in mature bacteroids (51). Levels of metabolites derived from glutamate, including glutathione and N-acetylglutamate, were also reduced, while levels of many other amino acids were either altered only slightly or unchanged in bacteroids (Fig. 2).

However, steady-state metabolite levels do not represent flux. Low levels of pyruvate, acetyl-CoA, coenzyme A, and citrate in bacteroids may indicate a low rate of synthesis but can equally result from rapid turnover. Furthermore, metabolites may dramatically change concentrations during isolation of bacteroids from nodules. We addressed this by comparing wild-type bacteroids with mutant bacteroids defective in the TCA cycle, which should lead to different metabolite profiles. If low acetyl-CoA levels in wild-type bacteroids relative to free-living cells result from increased flux through the TCA cycle, then TCA cycle mutants should have elevated acetyl-CoA levels.

Metabolite profile of a TCA cycle mutant.

We previously isolated several Tn5 insertions in Rlv3841 genes encoding TCA cycle enzymes (12). Malate dehydrogenase, succinyl-CoA synthetase, and the E1 and E2 components of the 2-oxoglutarate dehydrogenase complex are transcribed from the mdh-sucCDAB operon (52). Mutations in sucA (RU156, RU724, and RU733) and sucB (RU726), encoding the E1 and E2 components of the 2-oxoglutarate dehydrogenase complex, respectively, abolished 2-oxoglutarate dehydrogenase activity (12), resulting in plants that failed to reduce acetylene (Fix⁻). Therefore, blocking the TCA cycle in Rlv3841 prevents N₂ fixation. However, we now show that strain RU116, mutated in sucD (encoding the β-subunit of succinyl-CoA synthetase) and originally scored as Fix⁻ based on yellowing of plants and small nodules but retaining low levels of succinyl-CoA synthetase activity (12), is able to reduce acetylene at 35% of the wild-type rate (Fig. 3). This mutation may affect the number of bacteroids in nodules or total nodule mass or reduce nitrogenase activity. However, acetylene reduction per unit bacteroid protein and levels of shoot dry matter of plants grown under nitrogen-free conditions inoculated with the sucD mutant were 45% and 51% of the wild-type values, respectively (Fig. 3). Therefore, sucD bacteroids have lowered N₂ fixation, presumably due to attenuation, but not complete blockage, of the TCA cycle.

FIG 3.

Symbiotic phenotype of sucD mutant (RU116) compared to wild-type Rlv3841. (a and b) N₂ fixation as measured by acetylene reduction on whole plants at dpi 28 expressed on (a) a per-plant basis (n = 6 per treatment) and (b) a per-unit bacteroid protein basis (n = 6), where the significance value (*, P < 0.05) was determined by Welch's t test. (c) A photograph of pea plants taken at dpi 47 with uninoculated water control (WC), Rlv3841, and sucD (RU116) treatments. (d) Mean shoot dry weights of dpi 42 peas (n = 12 per treatment), where treatments not sharing a letter differ significantly (P < 0.05; analysis of variance [ANOVA] and Tukey's honest significant difference [HSD]). In all cases, error bars represent standard errors of the means.

Metabolite profiles of the sucD mutant and wild-type bacteroids (Fig. 4) show that while succinate levels were similar in RU116 and wild-type bacteroids, levels of fumarate and malate were considerably lower in the mutant bacteroids, indicating reduced flux of carbon. Our key investigation concerns the decarboxylating arm of the TCA cycle. Predictably for a mutant strain blocked in the TCA cycle at succinyl-CoA synthetase, citrate levels were 11-fold higher in the sucD mutant than in the wild-type strain and levels of intermediates derived from 2-oxoglutarate, such as glutamate, glutathione, and 2-hydroxyglutarate, were all increased markedly (Fig. 4). Therefore, attenuation of succinyl-CoA synthetase activity caused an accumulation of metabolites prior to the activity of the decarboxylating arm of the TCA cycle. Thus, the TCA cycle operates in bacteroids and reducing its activity also reduced N₂ fixation. Crucially, though, while the levels of pyruvate of the two bacteroid types were similar, no acetyl-CoA or free coenzyme A was detected in the sucD mutant. If the only major route for acetyl-CoA metabolism is the TCA cycle, its levels should rise dramatically in strain RU116 (sucD). This suggests that acetyl-CoA has other large sinks that are independent of the TCA cycle. The presence of alternative sinks for acetyl-CoA would explain its very low level in bacteroids compared to free-living bacteria. It would also have profound implications for our understanding of Rhizobium-legume symbioses, as it suggests a major rerouting of central metabolism during N₂ fixation in pea bacteroids.

FIG 4.

Metabolite profile of sucD (RU116) bacteroids versus Rlv3841 bacteroids showing fold change in metabolite abundance relative to sucD bacteroids. Bacteroids were isolated from nodules from pea plants 28 dpi. Bolded intermediates were detected by metabolite profiling, with a statistically significant fold difference (P < 0.05 by Welch's t test and Q < 0.1 for the false-discovery rate) denoted with a red (increase) or green (decrease) arrow. Intact arrows indicate single-step enzyme-catalyzed reactions. Broken arrows indicate instances in which two or more enzyme-catalyzed steps are involved in a series of reactions. sucD bacteroids are attenuated in TCA cycle enzymes post-2-oxoglutarate (2-OG). Abbreviations: UD, undetectable; GABA, γ-amino butyric acid; GSH, glutathione (reduced); GSSG, glutathione (oxidized); OAA, oxaloacetate; PEP, phosphoenolpyruvate.

Lipids are a sink for acetyl-CoA in bacteroids.

Apart from its complete oxidation in the TCA cycle, the other major metabolic fate of acetyl-CoA is in lipogenesis. Two possible products of lipogenesis are poly-β-hydroxybutyrate (PHB) and fatty acids. Considerable attention has focused on PHB because it is abundant in soybean and common bean bacteroids, although it is thought to be absent in mature N₂-fixing bacteroids from indeterminate nodulating plants, including pea, alfalfa, and clover. In contrast, there has been relatively little quantification of bacteroid lipids, which we sought to address.

There was a range of chain lengths and degrees of unsaturation in the free fatty acids in both bacteroids and free-living succinate-grown cells (Table 2). Levels of long-chain free fatty acids (C₁₆ to C₂₀) were higher in bacteroids than in either free-living bacteria or nodule cytosolic fractions. There were also significantly higher levels of monoacylglycerols, with bacteroids containing highly elevated levels of 1-linoleoylglycerol (>57-fold), 1-palmitoylglycerol (16-fold), and 2-linoleoylglycerol (>13-fold) as well as 1-stearoylglycerol (3.9-fold) and 2-oleoylglycerol (5.8-fold). Moreover, the less efficient N₂-fixing sucD mutant strain showed levels of these lipid species that were significantly lower than those seen with wild-type Rlv3841 bacteroids. The presence of these molecules at high levels in wild-type Rlv3841 suggests that bacteroids use fatty acids as a sink for acetyl-CoA.

TABLE 2.

Comparison of fatty acids and monoacylglycerols detected in metabolite profiles

Lipid species	Fold changea in metabolite abundance relative to that in:
	Rlv3841 free-living samples	Nodule cytosol samples	sucD bacteroid samples
Free fatty acids
cis-Vaccenate (18:1n7)	1.99	5.91	4.00
Palmitoleate (16:1n7)	8.20	4.87	2.94
Linolenate (α or γ [18:3n3 or 6])	23.0	4.09	1.32
Linoleate (18:2n6)	18.7	3.62	2.13
Eicosenoate (20:1n9 or 11)	8.39	2.91	2.56
10-Heptadecenoate (17:1n7)	8.22	2.54	1.19
Dihomo-linoleate (20:2n6)	3.81	2.50	1.28
Stearate (18:0)	2.16	1.90	2.44
Palmitate (16:0)	3.72	1.88	2.94
Margarate (17:0)	3.94	1.15	1.89
Pelargonate (9:0)	0.75	0.48	2.17
Heptanoate (7:0)	0.19	0.19	0.46
Caproate (6:0)	19.7	0.16	0.71
Caprylate (8:0)	1.45	0.16	0.75
Isovalerate	1.30	0.05	0.36
Glycerolipids
1-Linoleoylglycerol (18:2)	57.2	8.04	9.09
2-Linoleoylglycerol (18:2)	13.2	5.38	6.25
2-Oleoylglycerol (18:1)	5.83	3.16	2.78
1-Stearoylglycerol (18:0)	3.86	0.33	3.85
1-Palmitoylglycerol (16:0)	15.6	0.27	4.35

^aValues are fold changes in metabolite abundance in Rlv3841 bacteroids relative to that in the indicated samples. Values highlighted with boldface indicate significant increases (P < 0.05 and Q < 0.1), and values highlighted with italics indicate significant decreases (P < 0.05 and Q < 0.1); data without highlighting showed no significant differences.

It was not possible to detect diacyclglycerols or triacylglycerols in these samples as they fall outside the polarity range and upper size limit of the GC- and LC-MS techniques used. Therefore, membrane-free extracts were isolated by ultracentrifugation and their glycerolipid level quantified by enzyme assay. Levels of glycerolipids were 22-fold higher in bacteroids than in free-living cells (62 ± 2.66 ng/mg protein versus 2.8 ± 1.26 ng/mg protein, respectively). Bacteroids channel a large proportion of acetyl-CoA away from the TCA cycle and into lipids, suggesting that related storage mechanisms may be utilized under N₂-fixing conditions.

Pea bacteroids of Rlv3841 accumulate PHB.

PHB accumulation occurs in undifferentiated rhizobia in infection threads of pea nodules but is thought to be absent in bacteroids (20). When R. leguminosarum strain A34 was mutated in phaC, encoding a type I PHB synthase, it lacked detectable PHB both in infection thread bacteria and in bacteroids. This is consistent with the paradigm that bacteroids from indeterminate nodules such as pea and alfalfa do not make PHB in bacteroids. However, the genome of R. leguminosarum strain Rlv3841 has two PHB synthases: a type I PHB synthase on the chromosome (phaC1, RL2094) and a phaE (pRL100104) phaC2 (pRL100105) type III PHB synthase on the pRL10 symbiotic plasmid. The putative operon containing phaE and phaC2 is preceded by a consensus nifA promoter and was induced 7- to 40-fold in bacteroids, while phaC1 was not upregulated (44). As PHB is another lipogenic end product of acetyl-CoA metabolism, we investigated the symbiotic roles of these two PHB synthases in Rlv3841.

Previous work demonstrated that phaC1 was active in free-living Rlv3841, as mutation of this gene reduced PHB accumulation in the RU137 mutant by 93% relative to wild-type results (12), although the symbiotic performance of this phaC1 mutant was not determined. Therefore, we isolated a phaC2 single mutant (LMB814) and a phaC1 phaC2 double mutant (LMB816) in Rlv3841 and assessed their symbiotic phenotype, along with that of the original phaC1 mutant. While rates of N₂ fixation in phaC1 and phaC2 single mutants and phaC1 phaC2 double mutants were not significantly different from the rate seen with wild-type Rlv3841 (Fig. S1 in the supplemental material), examination of nodule sections by transmission electron microscopy (TEM) showed that PHB accumulation was altered. Pea nodules containing wild-type Rlv3841 exhibited large PHB droplets in bacteria in infection threads and smaller bodies in mature bacteroids (Fig. 5). Previously, when small PHB droplets were observed in bacteroids, it was assumed they were synthesized by bacteria in infection threads. However, while the phaC1 mutant harbored small PHB droplets in bacteroids, they were absent in the undifferentiated bacteria in infection threads. In contrast, PHB was largely absent in phaC2 mutant bacteroids but was abundant in bacteria occupying infection threads. Finally, PHB was absent from both bacteroids and bacteria in infection threads in the phaC1 phaC2 mutant. Therefore, Rlv3841 has two functional PHB synthases: one active in free-living and undifferentiated bacteria (type I, PhaC1) and the other active in bacteroids (type III, PhaE PhaC2). Although most sequenced rhizobia carry a type I PHB synthase, analysis of genome sequences shows that other rhizobia contain phaE and phaC2 genes, including strains forming symbiotic interactions not usually thought to make PHB, such as R. leguminosarum bv. viciae VF39 (pea) and R. leguminosarum bv. trifolii TA1 (clover) (Integrated Microbial Genomes [https://img.jgi.doe.gov/cgi-bin/w/main.cgi]). It is therefore likely that these other type III-harboring bacteroids also accumulate PHB, as has been demonstrated for Rlv3841.

FIG 5.

Transmission electron micrographs of pea nodules at 28 dpi. (a and b) Wild-type Rlv3841 bacteroids (a) and bacteroids in an infection thread (b), both showing PHB droplets. (c and d) Mutant phaC1 (RU137) bacteroids showing PHB accumulation (c); PHB accumulation was absent from infection threads (d). (e and f) Mutant phaC2 (LMB814) bacteroids where PHB droplets were largely absent (e); the bacteroids were abundant in infection threads (f). (g and h) Double mutant phaC1 phaC2 (LMB816) bacteroids (g) and bacteroids in infection threads (h), with PHB absent from both. Scale bars are 2 μm in panels a, c, e, and g and 1 μm in panels b, d, f, and h. Arrows point to PHB droplets which appear white.

DISCUSSION

The metabolism of free-living Rlv3841 growing on succinate as the sole carbon source is dominated by flux through the TCA cycle as well as anaplerotic and biosynthetic reactions. However, while the TCA cycle is essential for fully effective N₂ fixation in pea bacteroids, the accumulation of lipid shows a significant alternative fate for acetyl-CoA. Importantly, this observation is supported by the work of Miller and Tremblay (53), who showed that S. meliloti bacteroids from alfalfa nodules contain 34% of the total neutral lipid fraction as di- and triglycerides whereas these lipids were undetected in free-living S. meliloti. Moreover, the extraordinary deposition of PHB in bacteroids from common bean and soybean is an extreme example of carbon storage and redox balancing that has hitherto lacked a coherent explanation, particularly since preventing synthesis in these symbioses does not prevent N₂ fixation (19–21). Here we show that bacteroids of some strains of R. leguminosarum, such as Rlv3841, make PHB via a putative nifA-dependent type III PHB synthase. Therefore, the paradigm that mature bacteroids from indeterminate nodules (such as those formed on pea, alfalfa, and clover) do not synthesize PHB is incorrect. Most importantly, Rlv3841 bacteroids accumulate both PHB and lipid, showing that, even with acetyl-CoA incorporated into lipids, more acetyl-CoA accumulates in PHB. Thus, entry of acetyl-CoA into the TCA cycle must be limited, which implies that symbiotic N₂ fixation should be thought of as a fundamentally lipogenic process.

The complete oxidation of 1 mole of acetyl-CoA in the TCA cycle yields 4 moles of reducing equivalents [i.e., NAD(P)H or FADH₂]. In free-living rhizobia, this reductant can be channeled to the aerobic respiratory chain, driving oxidative phosphorylation, or used as a reductant in biosynthesis to fuel cell growth and division. However, mature pea bacteroids are in a metabolically active but nondividing state. In addition, N₂ fixation in legume root nodules occurs at microaerobic O₂ concentrations, estimated at 3 to 57 nM (54, 55). This low O₂ level is likely to restrict bacteroid respiration and hence TCA cycle activity, thereby forcing acetyl-CoA into lipids. While it is theoretically possible to have large rates of electron flux to a high-affinity terminal oxidase such as cbb₃ in bacteroids if O₂ flux is also high, the large-scale production of lipids and PHB suggests that this route is restricted. Instead, by channeling acetyl-CoA into lipid and PHB synthesis, bacteroids could overcome this metabolic constraint by consuming both carbon and reductant as NAD(P)H. Lipogenesis is a classic response of all domains of life to an excess of carbon and reductant that cannot be reoxidized by respiration or fermentation. Thus, free-living bacteria synthesize lipid when growth is nutritionally unbalanced, such as under O₂- or N₂-limited conditions (56, 57). During free-living N₂ fixation, both Azotobacter beijerinckii and A. caulinodans accumulate PHB, and in A. caulinodans, PHB synthesis is essential to both free-living and symbiotic N₂ fixation (22, 58). Thus, bacteroids may become lipogenic as a physiological response to the microaerobic environment inside legume root nodules.

Although aerobically growing free-living rhizobia and bacteroids differ with respect to O₂ supply and ability to divide, the other obvious metabolic difference is the supply of ATP and reductant for bacteroid nitrogenase. Bacteroids must supply reductant and ATP to nitrogenase, requiring 8 moles of electrons and 16 moles of ATP to reduce 1 mole of N₂ (equation 1). Although the electron source for nitrogenase in the free-living N₂-fixing bacterium Klebsiella pneumoniae, where electrons are transferred by NifJ (pyruvate:flavin oxidoreductase) and NifF (flavodoxin) complex from pyruvate to nitrogenase (59, 60), is well understood, it is unknown for rhizobia. In the classical model in rhizobia, all reductants generated by metabolism, primarily as NAD(P)H, can be allocated to all processes, including N₂ reduction or biosynthesis with excess reductant and ATP consumed by lipogenesis (Fig. 6). However, the standard redox potentials of NADH and ferredoxin (the E⁰′ value for NAD⁺/NADH is −320 mV and the ferredoxin Fe³⁺/Fe²⁺ value is −484 mV [16, 17]) suggest that it is unlikely that NADH donates electrons directly to ferredoxin and then to nitrogenase. An alternative explanation would be that a specific molecule acts as the low-potential electron donor to nitrogenase, an example being pyruvate oxidation by the NifJ-NifF complex in K. pneumoniae (59, 60). This process consumes four pyruvate molecules and produces four acetyl-CoA molecules to generate the eight electrons needed by nitrogenase. Since this complex is not present in rhizobia, an alternative pathway is required. One possibility is that the electron-transferring flavoprotein (ETF) complex, FixABCX, interacts with pyruvate dehydrogenase, as shown by genetic suppressor analysis in A. caulinodans (61). ETF complexes use electron bifurcation in anaerobic bacteria (62, 63), which might enable FixABCX to generate low-potential electrons for reduction of ferredoxin and then N₂ (Fig. 6). While unproven, such a mechanism would require 8 moles of pyruvate to reduce 1 mole of N₂ and would exacerbate the reductant problem because acetyl-CoA oxidized by the TCA cycle would generate excess NAD(P)H. In the absence of convincing experimental evidence for the pathway of electron donation to nitrogenase, we cannot complete a formal model to represent the electron and reductant balance. However, Fig. 6 illustrates how dramatically redox balance in bacteroids can be altered by the need for low-potential electrons for N₂ reduction.

FIG 6.

Two possible pathways of electron allocation in N₂-fixing bacteroids. (a) In the first scenario, NADH supplies electrons directly to nitrogenase as well as providing ATP from oxidative phosphorylation. A minimum of 2 moles of malate is required to be oxidized to acetyl-CoA to yield sufficient ATP and electrons to reduce 1 mole of N₂. (b) In the second scenario, electrons are supplied to nitrogenase via tight coupling with PDH and electron bifurcation through FixABCX, requiring 8 moles of malate to reduce 1 mole of N₂. The 16 moles of electrons liberated from the oxidation of 8 moles of pyruvate could undergo electron bifurcation at FixABCX, resulting in 8 (high-potential) electrons reducing CoQ via the Fix complex, while 8 (low-potential) electrons are channeled to nitrogenase and N₂ fixation. The 16 moles of ATP for N₂ fixation could be supplied from oxidative phosphorylation, for example, using the 8 (high-potential) electrons from FixABCX that enter the electron transport chain (ETC) at the level of CoQ (producing 6 moles of ATP), plus the reoxidation of 4 moles of NADH (of a total 8 moles of NADH) produced by malic enzyme (producing 10 moles of ATP). In this scheme there would be no need for all 8 moles of acetyl-CoA produced to enter the TCA cycle and undergo further oxidation to yield 24 moles of NADH and 8 moles of FADH₂, as that would be far in excess of the requirement to make the additional ATP required for N₂ fixation, leading to overreduction of the electron carrier pool. Instead, only a fraction of the acetyl-CoA might be further oxidized in the TCA cycle and the NADH produced used to support lipogenesis. The two models are not mutually exclusive, as, in panel a, free NADH might also interact with FixABCX, enabling low-potential electrons to be generated by bifurcation for reduction of ferredoxin. Note that a P:2e⁻ ratio of 2.5 is assumed for NADH and a ratio of 1.5 is assumed for electrons entering the ETC at the level of CoQ. For simplicity, we have not distinguished between NAD⁺ and NADP⁺ in this model. Abbreviations: CoQ, coenzyme Q; ME, malic enzyme; N2ase, nitrogenase; PDH, pyruvate dehydrogenase.

While A. caulinodans must synthesize PHB during N₂ fixation (22), synthesis can be blocked in bacteroids of peas, alfalfa, common bean, and soybean (19–21, 64, 65). The ability to prevent PHB synthesis and still have a functioning bacteroid may be explained by the existence of multiple storage sinks for acetyl-CoA, including PHB, free fatty acids, glycerolipids, and membrane phospholipids, with PHB itself being less important in these symbioses. Overall, bacteroids are highly lipogenic, with multiple lipid sinks for excess reductant. This applies to both determinate and indeterminate nodules and is likely to be an essential part of the energization of nitrogenase and associated redox balance in all N₂-fixing symbioses.

Funding Statement

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.