How Rhizobia Adapt to the Nodule Environment

ABSTRACT

Rhizobia are a phylogenetically diverse group of soil bacteria that engage in mutualistic interactions with legume plants. Although specifics of the symbioses differ between strains and plants, all symbioses ultimately result in the formation of specialized root nodule organs that host the nitrogen-fixing microsymbionts called bacteroids. Inside nodules, bacteroids encounter unique conditions that necessitate the global reprogramming of physiological processes and the rerouting of their metabolism. Decades of research have addressed these questions using genetics, omics approaches, and, more recently, computational modeling. Here, we discuss the common adaptations of rhizobia to the nodule environment that define the core principles of bacteroid functioning. All bacteroids are growth arrested and perform energy-intensive nitrogen fixation fueled by plant-provided C₄-dicarboxylates at nanomolar oxygen levels. At the same time, bacteroids are subject to host control and sanctioning that ultimately determine their fitness and have fundamental importance for the evolution of a stable mutualistic relationship.

INTRODUCTION

Plants are dependent on nutrients such as phosphorus, nitrogen, and sulfur to be present in a bioavailable form, which often limits plant growth. To facilitate nutrient uptake, mutualistic symbioses between plants and root-associated microorganisms have evolved. The symbiosis with arbuscular mycorrhiza (AM) originates back to the Devonian period more than 400 million years ago (1). The same signaling pathway that facilitated AM symbioses has since been adopted for other symbionts and is shared among all intracellular root symbionts (2, 3). This includes various root nodule symbioses (4, 5) that evolved around 110 million years ago (6), where the microsymbionts are either actinomycetes or rhizobia. These diazotrophic bacteria fix atmospheric N₂ gas into bioavailable ammonia. Biological nitrogen fixation accounts for approximately 50% of the bioavailable nitrogen on earth (7) and is of great importance for sustainable agriculture and the protection of susceptible ecosystems from leached fertilizers (8, 9).

The lack of genetic tools for the microsymbiont has hampered the investigation of actinorhizal symbioses (10), making rhizobium-legume symbioses the most researched and best-understood interactions so far (11). The symbioses between rhizobia (belonging to alpha- and betaproteobacteria [12]) and legumes (Fabaceae) exist with a variety of different infection mechanisms and nodule morphologies (13–15). Although bacterial entry via cracks in the root epidermis and Nod factor-independent infection processes exist (16–19), the best-studied mechanism is root invasion through root hairs via infection threads (ITs). Most major crop legumes and many model plants such as Medicago spp., Pisum sativum (pea), Lotus japonicus, Phaseolus vulgaris (bean), and Glycine max (soybean) follow this type of infection (20–26). Soil-dwelling rhizobia sense flavonoids exuded by the plant (27), and compatible flavonoids trigger the production of Nod factors (lipochitooligosaccharides) (for reviews, see references 28 and 29). Nod factor signaling leads to the induction of nodule primordia in the root cortex and causes the root hairs to curl their tips, resulting in the characteristic shape called “shepherd’s crook.” Bacteria attached to root hairs may be entrapped by this curling and form a microcolony from which they enter the root hair via an IT that is formed by the inverted growth of the plant cell wall and membrane. Since the microcolony is derived from one or very few founder cells, ITs are a major bottleneck for bacteria during invasion (30, 31).

Inside ITs, bacteria are thought to move mainly by cell division and potentially some gliding motility (32, 33). Eventually, the ITs reach the preformed nodule primordia, and rhizobia are taken up into individual plant cells by endocytosis, enclosed by a symbiosome membrane of plant origin. There, they multiply until the plant cell is densely packed with bacteria, which differentiate into their nitrogen-fixing form called a bacteroid. Nodules usually have one of two distinctive morphologies, depending on the host plant (34, 35): (i) determinate nodules, lacking a persistent meristem, therefore ceasing to grow at some point and forming spherical nodules, or (ii) indeterminate nodules with a persistent meristem, resulting in indefinite growth, branching, or lobe formation of the nodule, giving rise to elongated or irregular shapes. The detailed infection processes and differences between determinate and indeterminate nodules have been extensively discussed elsewhere (36–39).

COMPETITIVENESS

Host plants can influence the composition of the root microbiome (40, 41), but rhizobia must compete with various other bacteria for root colonization and with other compatible rhizobial strains for nodulation. Competitiveness is a complex trait: different abiotic factors such as soil pH (42–44) or nutrient availability (45) have an impact on the outcome of symbioses. Biotic factors also affect competitiveness, for example, the host-symbiont pairing (46–48), plant- or strain-intrinsic factors such as type VI secretion systems (49, 50), exopolysaccharide (EPS) production (51), and catabolic capacity in rhizobia for various substrates such as myo-inositol (52–54), glycerol (55), arabinose, protocatechuate (56), rhamnose (57), homoserine (58), or erythritol (59). During the transition from a soil-dwelling bacterium to a nitrogen-fixing bacteroid, rhizobia encounter drastically changing conditions, but it has not always been addressed at which stage of symbiosis formation a mutant strain is affected. Competitiveness in the rhizosphere is largely determined by metabolic functions and motility (48). Transcriptomics and mutant studies have shown that Rhizobium leguminosarum depends on a variety of substrates such as C₄-dicarboxylates, C₂-organic acids, and aromatic compounds for growth in the rhizosphere. However, sugar and sugar alcohol transport systems were similarly induced, showing that metabolism in the rhizosphere is highly complex (60). It is well documented that strains and mutants can be competitive in rhizosphere colonization but nonetheless lack competitiveness for nodulation (52, 61, 62), highlighting the need for changing adaptations throughout the infection process.

Insertion sequencing (INSeq) experiments using saturated random transposon mutant libraries identified essential genes in R. leguminosarum during four distinct stages of the symbiosis: rhizosphere growth, root attachment, nodule bacteria (viable bacteria within nodules), and differentiated bacteroids (63). In these experiments, bacteria were under selective pressure by pea plants, and mutants lacking competitiveness were lost at the affected and subsequent stages. Interestingly, while several genes were associated with competitiveness in the rhizosphere and root attachment, they were not necessarily involved in competitive nodulation, further indicating that persistence on the root surface is a distinct trait. In total, 390 genes were associated with competitive nodulation. Among these, 146 were already needed in the rhizosphere, 33 were needed in the root-attached stage, and 211 were needed in ITs and predifferentiated bacteroids. Functions related to competitive nodulation from the rhizosphere onward include lipopolysaccharide (LPS) and EPS biosynthesis, the NtrBC system and the σ⁵⁴ factor RpoN, and the stringent response. LPS and EPS modifications have a known role in evading the plant immune response during IT formation (64–70), and the stringent response has been shown to be involved at multiple stages during the Sinorhizobium-Medicago symbiosis (71–73). During later infection stages, allophanate, aldehyde, erythritol metabolism, and glycogen synthesis play important roles. Interestingly, in agreement with previous screens (48), almost all mutations causing auxotrophies lacked competitiveness, indicating that de novo synthesis of essential building blocks is crucial even when they are exogenously present in root exudates (74, 75).

The large number of genes associated with nodulation found outside the symbiotic plasmids in R. leguminosarum (63) is consistent with the findings for the chimeric Ralstonia solanacearum/Cupriavidus taiwanensis-Mimosa pudica experimental evolution of a symbiosis (76). Although the initial chimera was unable to form nodules, nodulation was achieved and competitiveness was gradually increased with repeated inoculation cycles on M. pudica seedlings. Strikingly, the ability to nodulate was directly linked with competitiveness (77). After initial infection was achieved by a loss of pathogenicity, a major improvement in competitiveness was seen when additional mutations in global transcriptional regulators enhanced intracellular colonization and persistence (78, 79), further highlighting that the ability to competitively nodulate legume plants heavily relies on the global rewiring of bacterial gene expression outside the acquired classical symbiosis genes (nod, nif, fix, etc.) found on mobile symbiotic plasmids or islands (80–84).

Notably, competitiveness for initial nodulation (but not persistence) is not affected by a strain’s efficiency in nitrogen fixation (85). Under field conditions, indigenous rhizobia with an inferior fixation capacity may outcompete the “elite” rhizobial strains inoculated on the legume seeds, resulting in suboptimal yields (86, 87). The dependence on environmental parameters and the overall complexity based on symbiotic and general-function genes (88) make competitive strains with high nitrogen fixation efficiencies difficult to identify. A recently reported barcoded and P_nifH-driven superfolder green fluorescent protein (sfGFP)-based toolkit allows the simultaneous high-throughput analysis of a strain’s competitiveness and nitrogen fixation rate (89). It enabled the identification of highly effective strains by measuring fixation rates of individual strains in single nodules. This approach facilitates the parallelized inoculation of multiple strains and the simultaneous assessment of their nodulation competitiveness and nitrogen fixation ability.

ADAPTATIONS TO NITROGEN FIXATION

Root nodules are de novo plant organs that share root- and stem-typical traits (90) and are formed only in the presence of compatible rhizobia. Their specialized function—to host intracellular bacteria for nitrogen fixation—makes the conditions inside nodules unique. Nodules are immunocompromised compartments with an immune response that differs from that of the rest of the plant, but they are also confined and insulated from the root, allowing them to host large numbers of intracellular bacteria (91). Despite some distinguishing properties specific to different rhizobia and their plant hosts, all rhizobium-legume symbioses share some universal characteristics. With the goal of highlighting these core changes defining the transition from free-living bacteria to nitrogen-fixing bacteroids, we performed a meta-analysis of 18 transcriptome and 1 proteome data sets published in 13 different studies on various symbioses (92–104) (Table 1). All genes/proteins were classified according to the Clusters of Orthologous Genes (COG) system (105) to enable interspecies comparisons.

TABLE 1.

Data sets used in the meta-analysis^a

Species (rhizobium strain-plant host)	Reference data set	Bacteroids terminally differentiated	Reference
Rhizobium leguminosarum bv. viciae 3841-Pisum sativum	Bacteria grown in MM (mid-exponential phase)	Yes	98
Rhizobium leguminosarum bv. viciae 3841-Vicia cracca	Bacteria grown in MM (mid-exponential phase)	Yes	98
Rhizobium leguminosarum bv. viciae A34-Pisum sativum	Bacteria grown in MM (mid- to late exponential phase)	Yes	99
Rhizobium leguminosarum bv. phaseoli 4292-Phaseolus vulgaris	Bacteria grown in MM (mid- to late exponential phase)	No	99
Rhizobium etli CFN42-Phaseolus vulgaris	Bacteria grown in MM (early exponential phase)	No	92
Rhizobium etli CFN42-Phaseolus vulgaris	Bacteria grown in MM (stationary phase)	No	92
Sinorhizobium meliloti 2011-Medicago truncatula	FIId: distal fraction of ZII; contains cells during infection and differentiation	Yes	102
Sinorhizobium meliloti 1021-Medicago truncatula	Bacteria grown in CM (mid-exponential phase)	Yes	95
Sinorhizobium meliloti 1021-Medicago sativa cv. Europe	Bacteria grown in MM (mid-exponential phase)	Yes	104
Sinorhizobium medicae-Medicago truncatula	FI: nonfixing zone of the nodule	Yes	97
Sinorhizobium sp. strain NGR234-Vigna unguiculata	Bacteria grown in CM (mid-exponential phase)	No	101
Sinorhizobium sp. NGR234-Leucaena leucocephala	Bacteria grown in CM (mid-exponential phase)	No	101
Bradyrhizobium sp. ORS285-Aeschynomene afraspera	Bacteria grown in CM (mid-exponential phase)	Yes	100
Bradyrhizobium sp. ORS285-Aeschynomene indica	Bacteria grown in CM (mid-exponential phase)	Yes	100
Bradyrhizobium japonicum USDA110-Glycine max	Bacteria grown in CM (mid-exponential phase)	No	96
Bradyrhizobium japonicum USDA110-Vigna unguiculata	Bacteria grown in CM (mid-exponential phase)	No	103
Bradyrhizobium japonicum USDA110-Macroptilium atropurpureum	Bacteria grown in CM (mid-exponential phase)	No	103
Azorhizobium caulinodans ORS571-Sesbania rostrata	Bacteria grown in MM (mid- to late exponential phase)	No	94
Mesorhizobium huakuii 7653R-Astragalus sinicus	Bacteria grown in CM (late exponential phase)	Yes	93

^aTranscriptome data were used in all cases apart from the Sinorhizobium medicae study, which is a proteome data set. All data sets compare bacteroids with the reference conditions described in the second column. MM, minimal medium; CM, complex medium.

Overall, most of the subsystems contained more downregulated than upregulated genes, indicating a reduced physiological complexity in bacteroids (Fig. 1). Notable exceptions to the general downregulation were “energy production and conversion” and “inorganic ion transport and metabolism” (see Table S1 in the supplemental material). Upregulated features in these subsystems comprised nitrogenase components (nif genes) and electron transfer proteins (including fix genes) such as ferredoxins, which act as electron donors for nitrogenase (106, 107) (Table S2). Free-living diazotrophs such as Klebsiella regulate nif gene expression in response to nitrogen starvation and low oxygen concentrations (108), whereas the nitrogen status in many bacteroids plays either no or only a minor role in nif gene activation, which is mainly induced by low oxygen concentrations (see Adaptations to Microaerobic Conditions, below). All rhizobia regulate nif and fix gene expression using similar sets of oxygen-responsive pathways (FixLJ or hFixL FxkR, FixK/Fnr, and NifA). The pathways differ in their sensitivities to oxygen (109), leading to staggered activation and thus a gradual adaptation of the invading rhizobia to the microoxic nodule environment (110). However, each species uses them in a slightly different way, and the interconnections between the pathways differ as well, a topic that has been discussed elsewhere (111–115). In addition to nif genes, the expression of iron and/or molybdenum transporters was increased across the various data sets, consistent with the requirement for these metals as cofactors of the nitrogenase enzyme (116).

FIG 1.

Central features of bacteroids determined from genome scale data sets. A meta-analysis of 18 transcriptome and 1 proteome data sets for bacteroids of various rhizobial strains was performed. The bar graph shows the number of data sets in which a COG category contained more upregulated than downregulated (black) or more downregulated than upregulated (white) genes/proteins. Where the total number is <19, some data sets contained either no differentially expressed genes/proteins or equal numbers of up- and downregulated genes/proteins in this category. Arrows indicate the 3 categories that were most often significantly enriched within upregulated (black) or downregulated (white) genes/proteins, with numbers in parentheses indicating the number of data sets for which significant up/downregulation was observed. P values were determined with a hypergeometric test followed by Bonferroni multiple-test correction, and a P value of <0.05 was considered significant. For the same analysis separated according to terminally differentiated and not terminally differentiated bacteroids, see Fig. S1 in the supplemental material.

General upregulation was further found for chaperones, indicating a response to stress factors in nodules and restructuring of the bacteroid proteome. This agrees with the established importance of chaperones in bacteroids. Nodules formed by a clpB mutant of Mesorhizobium ciceri contained only a few bacteroids (117), and GroESL chaperones were found to be essential for symbiosis (118, 119) and to interact with nodule-specific cysteine-rich (NCR) peptides (120) (see Adaptations to a Nongrowing State, below). Further examples of stress responses include glutathione S-transferases, which were consistently upregulated in bacteroids. Mutants unable to produce glutathione, an important antioxidant molecule, showed reduced symbiotic efficiency (121, 122). Similarly, enzymes involved in the detoxification of reactive oxygen species, such as superoxide dismutases and catalases, were commonly upregulated, in agreement with the presence of reactive oxygen species due to auto-oxidation of leghemoglobin and ferredoxin (123).

ADAPTATIONS TO MICROAEROBIC CONDITIONS

The iron-sulfur clusters of the nitrogenase enzyme are highly susceptible to molecular oxygen (124), and the nodule cortex therefore contains an oxygen diffusion barrier. This limits the amount of free oxygen within the nodule, creating a microoxic environment of around 11 nM free oxygen (112, 125). For comparison, the freely diffused oxygen concentration in water at atmospheric oxygen levels is around 255 μM. Nitrogen fixation is an energy-intensive process, and energy in biological systems is most efficiently generated by respiratory chains with molecular oxygen as the final electron acceptor (126). To circumvent this apparent paradox, where oxygen is needed to energize nitrogen fixation and nitrogenase is sensitive to oxygen, rhizobia possess high-affinity terminal oxidases that allow respiration at low oxygen concentrations. Rhizobia can usually express a complex set of different respiratory oxygen reductases, reflecting their diverse lifestyles (127–130). Biochemical and genetic data show that a cytochrome c oxidase, encoded by fixNOQP, is a high-affinity cbb₃-type oxidase responsible for respiration under microaerobic conditions in rhizobia. Biochemical assays revealed a K_m value of 4 to 7 nM for oxygen bound to membranes of anaerobically grown bradyrhizobia, corresponding to the cbb₃-type cytochrome oxidase FixNOQP (131). These values are compatible with the low oxygen concentrations in nodules. Mutants of fixNOQP in Bradyrhizobium japonicum had only marginal nitrogenase activity (132, 133), whereas multiple copies of fixNOQP are present in the genomes of Mesorhizobium loti, Sinorhizobium meliloti, Rhizobium etli, and R. leguminosarum. In S. meliloti and R. leguminosarum, the two copies are functionally redundant, but only one copy is functional during symbiosis in R. etli (134–136). Mutation of the (single-copy) cbb₃-type oxidase in Azorhizobium caulinodans resulted in only a 50% reduction of nitrogenase activity in symbiosis (137). The remaining activity could be abolished by creating a double mutant with a bd-type quinol oxidase. This type of oxidase is crucial for nitrogen fixation in the free-living diazotrophs Klebsiella oxytoca (138) and Azotobacter vinelandii (139). Recently, genome sequences of betaproteobacterial rhizobia showed that Paraburkholderia strains lack fixNOQP but contain bd-type quinol oxidase genes. However, functional studies are missing (140). In accordance with the described adaptations to the microaerobic nodule environment, several cytochromes and high-affinity cytochrome oxidases showed upregulation in our meta-analysis. Interestingly, subunits of the ATP synthase were mostly downregulated, indicating that the overall energy demand in bacteroids is lower than that in exponentially growing cells.

Nodules have a characteristic red color due to the accumulation of leghemoglobins. These plant-derived proteins are found exclusively in the plant cytoplasm and provide buffering capacity for free oxygen as well as facilitating the diffusion of oxygen toward symbiosomes (141–144). Mutation or silencing of leghemoglobin biosynthesis resulted in inefficient symbioses (145, 146). The free oxygen concentration was, however, only marginally higher in mutant than in wild-type nodules. Although there was a slight increase in oxygen available for respiration, the ATP/ADP ratio dropped compared to the wild type (146). Thus, the main function of leghemoglobins is not to keep free oxygen concentrations low (which is achieved by the oxygen diffusion barrier) but rather to enhance the oxygen supply to symbiosomes, which is limiting because of the diffusion barrier. In this function, leghemoglobins are equivalent to myoglobins in animals.

ADAPTATIONS TO A NONGROWING STATE

In addition to changes induced by microoxic conditions, bacteroids are characterized by their nongrowing state. Growth arrest in free-living bacteria is usually induced by unfavorable conditions such as starvation, oxygen depletion, or antibiotic substances. Accordingly, the main adaptations to such conditions focus on persistence, i.e., maintaining low levels of proton motive force, synthesizing integral cell building blocks, slowing metabolism, and catabolizing nonessential endogenous compounds, which enables bacteria to stay viable for months or years (147).

Commonly downregulated traits in the analyzed data sets that are shared with growth-arrested cells (92) include translation (ribosomal proteins), cell envelope biogenesis (outer membrane proteins, peptidoglycan, and exopolysaccharide synthesis), intracellular trafficking, cell cycle control and DNA replication, signal transduction, motility, and the F_oF₁-type ATPase (see Table S3 in the supplemental material). The ribosomal silencing factor RsfS, which inhibits the assembly of the 30S and 50S ribosomal subunits (148, 149), was generally upregulated, indicating additional downregulation of protein biosynthesis. Furthermore, ClpA, a subunit of the ClpAP protease, known for its role in cell cycle control and stationary-phase adaptation (150–152) but also potentially part of FixK regulation (153), was commonly upregulated in bacteroids. These findings suggest that a major adaptation to the nodule environment for bacteroids is related to growth arrest. Strikingly, the experimental evolution experiments described above found mutations in efpR to be associated with adaptation to intracellular colonization (78). EfpR was found to act as a positive regulator of several 30S ribosomal protein subunits and as a negative regulator of the adapter protein ClpS (78), which redirects protein aggregates to ClpAP (154), mimicking the bacteroid growth arrest adaptation.

In contrast to free-living growth-arrested bacteria, however, bacteroids are not nutrient starved, are well adapted for respiration under microoxic conditions, and retain high levels of metabolic activity to sustain nitrogen fixation. It thus remains unknown how and why bacteroids stop dividing. Some legumes induce terminal differentiation and chromosomal endoreduplication in bacteroids, meaning that the bacteroid is unable to dedifferentiate into a free-living bacterium (155, 156). Generally, hosts with determinate nodules (e.g., Lotus, soybean, and bean) harbor bacteroids that are similar to free-living bacteria, and differentiation is not terminal (U morphotype) (157). Legumes of the inverted-repeat-lacking clade and some Aeschynomene spp. contain enlarged and elongated bacteroids (E morphotype) that sometimes branch and become Y-shaped or irregularly shaped. Many legumes of the Dalbergioid cluster (e.g., peanut and Aeschynomene spp.) cause bacteroids to become enlarged and spherical (S morphotype). Rhizobia capable of nodulating different host plants will adopt the bacteroid morphotype of the respective host (31, 158, 159). Enlargement of the bacteroids in the case of E and S morphotypes is accompanied by endoreduplication. This differentiation leads to up to 24C (chromosomes) in E-morphotype S. meliloti bacteroids in symbiosis with Medicago (155). Similarly, Bradyrhizobium sp. strain ORS285 bacteroids reach 7C in Aeschynomene afraspera (E morphotype) and 16C in Aeschynomene indica (S morphotype) (160).

The key factor for bacterial endoreduplication is NCR peptides. These consist of 60 to 90 amino acids and display similarities with plant defensins (161, 162). NCR peptides are abundant in hosts harboring swollen bacteroids (around 700 are known in Medicago truncatula) but evolved independently in different clades (155, 156, 163). They are exclusively expressed in bacteroid-containing cells and are targeted to the bacteroid (164). Plant mutants with defects in the NCR peptide secretory pathway form ineffective nodules (164–166). Some NCR peptides have antimicrobial activities (167, 168), and rhizobia with increased sensitivity to NCR peptides form ineffective symbioses because they are rapidly killed after release from ITs into plant cells (160, 169–172). At the right dosage, however, NCR peptides will promote endoreduplication, even in free-living cultures (173), although the mechanisms by which they interfere with the bacterial cell cycle are not yet fully understood (167). The mode of action of some NCR peptides has been identified, for example, inhibition of FtsZ (the structural component of the Z ring involved in cell division) and ribosomal proteins (120). Moreover, rhizobia are successively exposed to different NCR peptides, suggesting that they act in an orchestrated manner, controlling various steps of rhizobial physiology (173–175). NCR peptide-mediated endoreduplication and terminal differentiation might contribute to the growth-arrested phenotype of bacteroids. However, NCR peptides are absent in hosts harboring unswollen (U morphotype) bacteroids (176), and endoreduplication and swelling of bacteroids happen only after bacteroids stop dividing (159), indicating the presence of other mechanisms that are yet to be discovered.

METABOLIC ADAPTATIONS

As a result of the adaptations required for symbiotic nitrogen fixation, bacteroid metabolism is unique in several respects: there is a strong downregulation of most biosynthetic functions compared to growth in the free-living state, while specialized enzymes for nitrogen fixation are highly induced (98, 177). In contrast to free-living bacteria in stationary phase, bacteroid metabolism is highly active due to the energy requirement of nitrogen fixation despite the microoxic environment. In addition, bacteroid metabolism is closely interdependent with the plant host because of bidirectional nutrient exchanges between the symbiotic partners (Fig. 2) (116, 178).

FIG 2.

Central metabolic and transport reactions in bacteroids. Shown is a schematic map of major metabolic pathways and transport reactions in bacteroids as well as important flows of electrons and ATP. This figure was created with BioRender.

An important implication of nutrient provision by the plant is that downregulation in biosynthetic pathways can occur either because a compound is required only in small quantities in bacteroids or because it is provided by the plant. Multiple auxotrophic rhizobial mutants with a Fix⁺ phenotype are known, indicating that plants provide the respective compounds to their bacteroids (179). A striking illustration of this concept is symbiotic auxotrophy: mutants lacking the general amino acid permeases Aap and Bra in R. leguminosarum were severely impaired in their symbiotic efficiency, indicating a requirement for plant supply of amino acids (180). While free-living rhizobia are capable of synthesizing all proteinogenic amino acids, downregulation in branched-chain amino acid synthesis occurs in bacteroids, making them dependent on the supply of branched-chain amino acids by their host plant (181, 182). A similar downregulation was observed in our meta-analysis for serine biosynthesis. Indeed, a serine auxotrophic mutant of R. etli was fully proficient in symbiosis (183). Yet cysteine biosynthesis, which requires serine as a precursor, was one of the few upregulated pathways for amino acid biosynthesis. This upregulation is likely related to the production of iron-sulfur clusters for the nitrogenase enzyme, which starts with cysteine as a sulfur donor (184). It is therefore possible that plant hosts supply serine to bacteroids to assist with cysteine biosynthesis for nitrogenase assembly. Experimental evidence also supports the provision of amino acids by the plant host that are nonessential for bacteroid functioning. In pea, enzymes catabolizing γ-aminobutyric acid (GABA) are highly induced in bacteroids, and labeling studies indicated that GABA is supplied by the plant, but mutants of R. leguminosarum unable to catabolize GABA differentiated into functional bacteroids (185).

Apart from amino acids, a different example of plant-derived metabolites essential for nitrogenase activity is homocitrate, a component of the iron-molybdenum cofactor. In contrast to free-living diazotrophs, most symbiotic rhizobia are unable to synthesize homocitrate, and a homocitrate synthase mutant (Fen1) of L. japonicus was incapable of forming effective nodules (186). These observations illustrate the intricate control that plants exercise over their bacterial symbiont.

Legumes provide bacteroids with C₄-dicarboxylates as the main carbon source, primarily malate and succinate (187–189). Catabolism of C₄-dicarboxylates requires their conversion to both oxaloacetate and acetyl coenzyme A (acetyl-CoA), which are condensed to form citrate in the first step of the tricarboxylic acid (TCA) cycle. There are two pathways for generating acetyl-CoA from TCA cycle intermediates: (i) malate is oxidized to pyruvate by malic enzyme (ME), or (ii) phosphoenolpyruvate carboxykinase (PCK) converts oxaloacetate into phosphoenolpyruvate, which is converted into pyruvate by pyruvate kinase (PK). In both cases, acetyl-CoA is obtained from pyruvate via the pyruvate dehydrogenase complex. While all bacteroids are required to convert TCA cycle intermediates into acetyl-CoA, the synthesis routes differ. Bacteroids of S. meliloti and A. caulinodans exclusively use ME (190, 191), whereas Sinorhizobium fredii and R. leguminosarum utilize both ME and PCK/PK (191, 192). In addition, the activity of PCK is important for bacteroids to sustain gluconeogenesis, which is essential for the production of precursors for various metabolites (63, 193). The TCA cycle is the main pathway for dicarboxylate metabolism in bacteroids, and at least some TCA cycle enzymes were upregulated in the transcriptome/proteome data sets across different rhizobial species. However, studies with individual mutants have produced contradictory evidence on which parts of the TCA cycle are required for nitrogen fixation. Mutants in different TCA cycle enzymes were Fix⁻ in S. meliloti, S. fredii, Rhizobium tropici, and R. leguminosarum (194–198), but several TCA cycle mutants in B. japonicum still formed an efficient symbiosis (199–202). This may be explained by flux through anaplerotic reactions or the existence of uncharacterized isozymes (203). For example, an aconitase mutant of B. japonicum retained almost 30% of the aconitase activity (201). Additional mutant studies with careful evaluation of both enzyme activities and redirection of metabolic fluxes are therefore required to elucidate the essentiality of flux through different parts of the TCA cycle.

Dicarboxylate catabolism in the TCA cycle generates large amounts of reduced electron carriers (NADH and reduced flavin adenine dinucleotide [FADH₂]) required for ATP synthesis and nitrogen fixation (116, 178). Due to the low oxygen levels in nodules, flux through the TCA cycle may become disadvantageous because the oxidation of electron carriers in the respiratory chain is restricted, and the accumulation of reducing equivalents inhibits pyruvate dehydrogenase and α-ketoglutarate dehydrogenase (204–206). Bacteroids could therefore channel carbon into polymers such as polyhydroxybutyrate (PHB), glycogen, and lipids (106, 207–209). Accumulation of these polymers is observed in bacteroids, although the nature of the compounds differs between symbioses (106, 210). In B. japonicum, the regulator of PHB synthesis, PhaR, has been shown to repress various gene targets, notably FixK₂, a transcriptional regulator for adaptation to microoxic conditions (211, 212). This suggests a certain degree of coordination between PHB metabolism and symbiotic adaptation, at least in some rhizobial strains.

A fundamental requirement for rhizobium-legume symbioses is the secretion of fixed nitrogen to the plant instead of its assimilation by the bacteroid. Ammonia is the main secretion product, but the secretion of alanine and aspartate has also been reported (213–216). The glutamine synthetase-glutamate oxoglutarate aminotransferase (GS-GOGAT) system, which assimilates ammonia into glutamate in free-living rhizobia, appears to be largely downregulated in bacteroids (116, 217). Rhizobia usually possess two glutamine synthetases (GSI/GlnA and GSII/GlnII), with some strains encoding a third enzyme (GSIII/GlnT). In general, only GSI activity is present at low levels in bacteroids, and the deletion of any of the GS genes does not impair the symbiosis (217, 218). Mutant studies have further shown GOGAT to be dispensable in S. meliloti (219, 220), R. etli (221), and R. leguminosarum (222). Overall, these results indicate no need for ammonia assimilation by the bacteroid, which may be due to amino acid provision by the plant or reduced protein biosynthesis as a result of growth arrest.

A recent study showed that the nitrogen-related phosphotransferase system in R. leguminosarum controls central carbon metabolism in response to carbon and nitrogen availability, and mutation of this system caused an ineffective symbiosis (223). This highlights the complex interconnection between carbon and nitrogen metabolism in bacteroids, which must be finely tuned to enable a functional symbiosis. The complex regulatory systems underlying rhizobial nitrogen metabolism have been comprehensively discussed elsewhere (179, 217, 224).

MODELING BACTEROID METABOLISM

Bacteroid metabolism is difficult to investigate experimentally due to the challenge of measuring nutrient exchange with the plant host, the multiple lifestyle adaptations required during symbiosis formation (63), and the fragile nature of isolated bacteroids (178). In addition to various omics techniques, constraint-based metabolic models have therefore become a popular tool for investigating rhizobium-legume symbioses (225). Metabolic models are knowledgebases of the reactions catalyzed by enzymes encoded in an organism’s genome. By imposing an objective function and assuming a steady state of all metabolic reactions, flux distributions are calculated using flux balance analysis (FBA) (226, 227). Models with different scopes have been reconstructed for various rhizobial species (see Table S4 in the supplemental material), and studies have progressed from describing bacteroid metabolism (228–231) to integrating bacterial and plant models as well as models of nodule zone-specific metabolism (232, 233). Rhizobial genomes are large and may include secondary replicons encoding niche-specific functions (60, 234). Metabolic models specific to the nitrogen-fixing state must therefore be constrained using transcriptome and proteome data sets (228) and/or appropriate objective functions (231, 234). Since direct measurement of reaction rates and exchange fluxes with the plant is infeasible, FBA studies have mostly used objective functions composed of generally accepted features of bacteroid metabolism, such as nitrogen fixation, amino acid secretion, and the production of carbon polymers (229, 230, 235).

All models published to date are consistent with experimental evidence for the use of the TCA cycle and oxidative phosphorylation as the main pathways fueling bacteroid metabolism. Similar to the disparities in experimental results, the operation of both a full and a partial TCA cycle during nitrogen fixation has been predicted. This may be due to strain-specific differences as well as the choice of constraints, in particular carbon and oxygen uptake rates. Synthesis of carbon polymers as well as amino acid exchange with the plant have also been investigated in silico. An increase in symbiotic efficiency was predicted for the deletion of glycogen or PHB synthesis in R. etli (229), which agrees with the enhanced symbiotic properties of a glycogen synthase mutant of R. tropici (236) but not with the phenotypes of mutants for PHB and glycogen synthases in R. leguminosarum and S. meliloti (207, 209). Most, but not all, models assume extensive amino acid cycling between the plant and bacteroid. This is based on early experimental studies suggesting glutamate supply by the plant (180), which, however, was later shown to be nonessential for bacteroids (181). With regard to amino acids secreted by the bacteroid, alanine has been predicted to be the main nitrogenous compound produced at low oxygen levels since it serves as a sink for both carbon and electrons (232).

A recent study of an integrated model for plant and bacteroid metabolism further addressed the question of why C₄-dicarboxylates are supplied to bacteroids instead of sucrose, which is abundant in nodules. Rhizobia differentiating into bacteroids were predicted to catabolize sugars (233), which is supported by studies of mutants in dicarboxylate transport, which are able to nodulate and differentiate into bacteroids but fail to fix nitrogen (187, 188). Modeling results reported previously (233) indicated that sucrose catabolism in fully differentiated bacteroids would achieve higher yields of fixed nitrogen and that proton supply by the plant is required to support the catabolism of C₄-dicarboxylates. The provision of C₄-dicarboxylates was further predicted to be driven by oxygen limitation of the plant mitochondria (233). Experimental evidence has been found for oxygen limitation of both plant (106) and bacteroid (237, 238) fractions, highlighting this as an outstanding question for future studies.

It is important to note that carbon polymer synthesis and amino acid secretion are commonly included in the bacteroid objective function in addition to ammonia export. Since flux distributions calculated by FBA must produce all compounds that are part of the objective function, carbon and nitrogen allocations in metabolic models are constrained by forcing the stoichiometric production of these compounds. This was circumvented by an integrated model of plant and bacteroid metabolism and optimization for plant biomass production (233). In future studies, the application of comprehensive and unbiased modeling approaches, such as elementary flux mode enumeration (239), may be insightful to improve and expand our current understanding of bacteroid metabolism.

Apart from the direct characterization of bacteroid metabolism, the underlying gene-protein-reaction relationships in metabolic models are useful for guiding targeted investigation of gene essentiality. Seeing as symbiosis formation is a multistep process requiring various metabolic capabilities at different stages (63), it is difficult to distinguish between genes essential for infection and those for essential nitrogen fixation itself. Discrepancies between experimental and in silico gene essentiality can thus unravel metabolic requirements at different stages of symbiosis formation.

PLANT CONTROL AND METABOLIC SANCTIONING

Both computational and experimental studies have demonstrated how bacteroid metabolism can be modulated in response to nutrient supply by the plant, in particular carbon and oxygen (240, 241). Plant control over bacteroid metabolism is essential due to the high fitness cost of nodule formation and the maintenance of bacteroids. The plant host must therefore be capable of monitoring symbiotic performance and responding accordingly, which becomes even more important when considering that the efficiency of the symbiont cannot be evaluated prior to nodulation (85, 242). Mathematical models demonstrated that a substantial investment in nitrogen fixation will be advantageous for rhizobia only if the plant adjusts its resource allocation according to symbiotic performance (243). To maximize their fitness, reproductive bacteroids can divert some of the plant-supplied carbon into polymers and catabolize those after dedifferentiation into the free-living state. For nonreproductive bacteroids, the undifferentiated population in the nodule benefits from the carbon supply and will be released into the soil after nodule senescence (244). Thus, the rhizobium-legume symbioses seem prone to selecting for cheaters that fix no or little nitrogen. In fact, field isolates vary greatly in their nitrogen fixation efficiencies (89). Consequently, legumes must have the means to identify and punish inefficient strains to maximize the nitrogen return for their carbon investment. It is well established that Fix⁻ strains do not persist within mature nodules (245), and when nodules are coinhabited with a fixing strain, the Fix⁻ strain is sanctioned rapidly in a cell-autonomous way (246). Sanctioning results in a fitness increase of good symbionts since higher numbers will be released from nodules. Under experimental conditions, efficient symbionts will outcompete Fix⁻ strains in a few plant generations (247, 248) since plants readily select the most efficient strain (249), even when the initial cell numbers are lower than those of the less efficient strain. Experimental data on carbon and oxygen supply to cheating strains (85, 240, 241) show that plants sense and integrate the nitrogen output from bacteroids and readjust carbon investment (250) or apply oxygen sanctioning in response to poorly performing strains.

FINAL CONCLUSIONS AND FUTURE PERSPECTIVES

The bacteroid stage in the rhizobial life cycle features a unique combination of traits. It is characterized by a nongrowing state in an oxygen-depleted environment. Despite the downregulation of many biosynthetic pathways, bacteroids maintain a highly active and specialized metabolism driven by the catabolism of C₄-dicarboxylates. Rhizobia are equipped with specialized functions to cope with this environment, and global adjustments of cellular and metabolic processes must take place to ensure a successful transition to nitrogen-fixing bacteroids. Failure to do so is punished by the host plant and results in the reduced fitness of a rhizobial strain. Host plants impose control over their symbionts at multiple levels. Nutrient supply, mostly in the form of malate and succinate, can be reduced; the oxygen supply can be limited; or, in some cases, plants can target bacteroids directly with antimicrobial peptides to force them into terminal differentiation.

Despite major advances in our knowledge, the complex interactions in rhizobium-legume symbioses retain many unanswered questions. Modulation of the plant immune response by both the host and the symbiont to accommodate thousands of intracellular prokaryotes per cell, what it shares with the better-understood immune response during infection, and its interplay with host sanctioning and pathogens remain exciting research areas despite substantial published work (245, 251–253). Similarly, our understanding of nutrient supply to bacteroids beyond malate and branched-chain amino acids is fragmented, obscuring the stoichiometry of metabolic fluxes that drive nitrogen fixation. The rhizobium-legume symbioses are prime examples of biological systems whose complexity necessitates the combination of experimental and computational approaches. In silico modeling is a rapidly advancing field that will greatly contribute to our understanding of the symbioses and aid in advancing toward engineering synthetic symbioses.