Ribulose 1,5-Bisphosphate Carboxylase/Oxygenase Is Required in Bradyrhizobium diazoefficiens for Efficient Soybean Root Colonization and Competition for Nodulation

The Hyphomicrobiales (Rhizobiales) order contains soil bacteria with an irregular distribution of the Calvin–Benson–Bassham cycle (CBB). Key enzymes in the CBB cycle are ribulose 1,5-bisphosphate carboxylase/oxygenase (RuBisCO), whose large and small subunits are encoded in cbbL and cbbS, and phosphoribulokinase (PRK), encoded by cbbP. These genes are often found in cbb operons, regulated by the LysR-type regulator CbbR. In Bradyrhizobium, pertaining to this order and bearing photosynthetic and non-photosynthetic species, the number of cbbL and cbbS copies varies, for example: zero in B. manausense, one in B. diazoefficiens, two in B. japonicum, and three in Bradyrhizobium sp. BTAi. Few studies addressed the role of CBB in Bradyrhizobium spp. symbiosis with leguminous plants. To investigate the horizontal transfer of the cbb operon among Hyphomicrobiales, we compared phylogenetic trees for concatenated cbbL-cbbP-cbbR and housekeeping genes (atpD-gyrB-recA-rpoB-rpoD). The distribution was consistent, indicating no horizontal transfer of the cbb operon in Hyphomicrobiales. We constructed a ΔcbbLS mutant in B. diazoefficiens, which lost most of the coding sequence of cbbL and has a frameshift creating a stop codon at the N-terminus of cbbS. This mutant nodulated normally but had reduced competitiveness for nodulation and long-term adhesion to soybean (Glycine max (L.) Merr.) roots, indicating a CBB requirement for colonizing soybean rhizosphere.


Introduction
Certain soil bacteria possess the ability to fix CO 2 .Among them are several members of the Hyphomicrobiales order, which encompasses photosynthetic and non-photosynthetic bacterial species.Bradyrhizobium diazoefficiens [1] is a non-photosynthetic member of this order, which may thrive under different lifestyles in the soil, including free-living in planktonic or sessile states [2] or as symbionts, fixing atmospheric N 2 inside nodules of different tropical legumes such as soybean (Glycine max (L.) Merr.), cowpea (Vigna unguiculata (L.) Walp.), mung bean (V.radiata (L.) Wilczek), and siratro (Macroptilium atropurpureum (Moc.& Sessé ex DC.) Urb.) [3,4].To cope with such a diversity of lifestyles, B. diazoefficiens has a large genome (>9 Mbp) that encodes a wide menu of metabolic pathways [5].Among Plants 2024, 13, 2362 2 of 14 them is the Calvin-Benson-Bassham cycle (CBB), for which the cbb operon includes single copies of cbbL and cbbS encoding the large and small subunits of Type IC ribulose 1,5-bisphosphate carboxylase/oxygenase (RuBisCO) [6], cbbP for phosphoribulokinase (PRK) [7], cbbX for the RuBisCO activase [8], as well as other genes encoding enzymes of the pentose phosphate pathway (PPP) such as fructose 1,6-bisphosphatase (cbbF) [9], transketolase (cbbT) [10], fructose-bisphosphate aldolase (cbbA) [11], and ribulose-phosphate 3-epimerase (rpe) [12]-named here as cbbE as it is part of the cbb operon [13].In addition, the LysR-type transcriptional regulator cbbR [14] lies at the opposite strand near the 5' end of the cbb operon.Meanwhile, other species of the genus, such as B. japonicum, as well as Bradyrhizobium sp.CCGE-LA001, ORS 278, and B. oligotrophicum S58, possess an additional Type IC RuBisCO, while the photosynthetic strain Bradyrhizobium sp.BTAi also possesses a Type IAc RubisCO and an additional cbb operon [6].Surprisingly, there are two Bradyrhizobium spp.strains (B.manausense BR3351 and Bradyrhizobium sp.URHD0069) for which homologs to RuBisCO are absent in their genomic sequences.As in Bradyrhizobium, the distribution of these genes in Hyphomicrobiales is not uniform.For instance, members of Rhizobium, Sinorhizobium, and Ensifer possess them, while members of the closely related Agrobacterium do not, and in a nearby branch, members of Mesorhizobium also possess cbb genes while the closely related Aquamicrobium, Aminobacter, and Nitratireductor do not.Since there is evidence that, during evolution, RuBisCO genes were horizontally exchanged among different phyla including prokaryotes, archaea, and higher plants [15][16][17], this diversity in both cbb operon structures and photosynthetic capabilities suggests the possibility that, within Hyphomicrobiales, the cbb operon might also have been exchanged by horizontal gene transfer between photosynthetic and non-photosynthetic members of this order.
RuBisCO activity was detected in B. diazoefficiens cells growing chemoautotrophically in microaerobiosis using H 2 as electron donor and CO 2 as carbon source [18].Under these conditions, cbbP, cbbL, and cbbS are overexpressed in relation to growth in an ambient air atmosphere [19].In addition, B. diazoefficiens chemoautotrophic growth was observed in ambient air using thiosulfate as an electron donor in the absence of an externally added carbon source, and a cbbL-defective mutant is unable to grow in these conditions [20].For heterotrophic growth, B. diazoefficiens may use a wide range of carbon sources [1].Two important carbon sources are D-mannitol and L-arabinose, which follow two different pathways for their preparatory catabolic phases in aerobic conditions [21].Since B. diazoefficiens does not catabolize hexoses or pentoses through the Emden-Meyerhof-Parnas pathway ( [21] and references therein), in cultures grown with D-mannitol, the PPP is preferred; in addition, PRK and RuBisCO are synthesized, which suggests that CBB cycle is also at work, whereby this bacterium might add CO 2 to its total carbon assimilation [21].Differently, in aerobic cultures grown with L-arabinose, PRK and RuBisCO are not detected, and the first catabolic steps occur by the L-2-keto-3-deoxyarabonate (LKDA) pathway [21] because, with L-arabinose being the enantiomer of D-arabinose, the former cannot enter the PPP at the level of D-ribulose.These results suggest that, for B. diazoefficiens, the CBB cycle might play some role in (micro)aerobic environments even in the presence of certain carbon sources and, in particular, it may be important for some process in the soil atmosphere.The symbiotic association of B. diazoefficiens with soybean roots involves a preinfection processes characterized by rhizosphere colonization, bacterial adhesion to root surfaces, induction of root hairs deformation, and early root hair penetration.Although there are studies that demonstrated the role of RuBisCO in the symbiosis between Aeschynomene plants and photosynthetic bradyrhizobia [22], no studies addressed the possible role of the CBB cycle during the early symbiotic association of B. diazoefficiens with soybean roots.
To gain insight about the above-mentioned queries, here we compared the genomic regions containing the cbb operon in diverse Hyphomicrobiales and obtained a B. diazoefficiens mutant with a deletion in the catalytic region of cbbL and a stop codon in the N-terminus of cbbS to study its effects on early interaction with soybean plants.No conclusive evidence about horizontal gene transfer of the cbb operon in Hyphomicrobiales was obtained.
In turn, the mutant strain was impaired in root colonization as well as competition for nodulation, indicating that RuBisCO may be necessary for establishment of rhizobia in the rhizospheric environment.

Phylogenetic Relationships of Housekeeping Genes and the Cbb Operon in Hyphomicrobiales
To compare the phylogenetic relationships of the CBB pathway with those of housekeeping genes, we concatenated cbbR, cbbL, and cbbP on the one hand, and atpD, gyrB, recA, rpoB, and rpoD on the other, and we compared their distributions among various Hyphomicrobiales.We observed a high degree of consistency between the trees obtained for each group of concatenates, except for Labrys sp.KNU-23, which appeared closely related to Microvirga massiliensis JC 119 and Methyloferula stellata AR4 in the housekeeping tree, while in the cbb tree it appeared in the same branch as B. diazoefficiens USDA 110, Rhodopseudomonas palustris CGA009, Tardiphaga sp.vice 352, and Nitrobacter winogradskyi Nb-255 (Figure 1).Since this change in the phylogenetic position of the Labrys sp.KNU-23 cbb operon suggested a horizontal gene transfer event, we looked for possible insertion sequences or tRNA sequences at the ends of this operon and evaluated possible changes in its GC contents with respect to the DNA flanking region.Although we observed a tRNA-Arg at 5 kb upstream ccbR, no other putative insertion sequences were found.Moreover, the GC content in this region was not different to that of the flanking regions or the general GC content of this species.Perhaps genomic signatures, if they existed, have been lost over time.
terminus of cbbS to study its effects on early interaction with soybean plants.No conclusive evidence about horizontal gene transfer of the cbb operon in Hyphomicrobiales was obtained.In turn, the mutant strain was impaired in root colonization as well as competition for nodulation, indicating that RuBisCO may be necessary for establishment of rhizobia in the rhizospheric environment.

Phylogenetic Relationships of Housekeeping Genes and the Cbb Operon in Hyphomicrobiales
To compare the phylogenetic relationships of the CBB pathway with those of housekeeping genes, we concatenated cbbR, cbbL, and cbbP on the one hand, and atpD, gyrB, recA, rpoB, and rpoD on the other, and we compared their distributions among various Hyphomicrobiales.We observed a high degree of consistency between the trees obtained for each group of concatenates, except for Labrys sp.KNU-23, which appeared closely related to Microvirga massiliensis JC 119 and Methyloferula stellata AR4 in the housekeeping tree, while in the cbb tree it appeared in the same branch as B. diazoefficiens USDA 110, Rhodopseudomonas palustris CGA009, Tardiphaga sp.vice 352, and Nitrobacter winogradskyi Nb-255 (Figure 1).Since this change in the phylogenetic position of the Labrys sp.KNU-23 cbb operon suggested a horizontal gene transfer event, we looked for possible insertion sequences or tRNA sequences at the ends of this operon and evaluated possible changes in its GC contents with respect to the DNA flanking region.Although we observed a tRNA-Arg at 5 kb upstream ccbR, no other putative insertion sequences were found.Moreover, the GC content in this region was not different to that of the flanking regions or the general GC content of this species.Perhaps genomic signatures, if they existed, have been lost over time.Plants 2024, 13, 2362 4 of 14 Some of the species analysed in Figure 1 possess a second copy of cbbL, although not in the context of an additional cbb operon.Therefore, we did not concatenate these extra cbbL copies with cbbP and cbbR because an artifact might be produced by mixing sequences that are in the cbb operon with the monocistronic cbbL copies.Instead, we performed a second phylogenetic analysis using only cbbL and observed that all the cbbL monocistronic paralogs grouped together in a separate clade (Supplementary Figure S1).This result indicates that there is considerable sequence divergence between cbbL paralogs in each of the species, suggesting that these extra copies could be originated in gene duplications rather than horizontal gene transfer, a hypothesis already considered for evolution of RuBisCO in a wider taxonomic context [15].In agreement with this hypothesis, in the known example of Bradyrhizobium sp.ORS278, the cbbL paralog seems inactive [22].

Different Effects of ∆cbbLS Mutation on Early and Late Root Adhesion and Colonization
We obtained a deletion mutant in B. diazoefficiens USDA 110 by removing 1092 bp encompassing the cbbL coding sequence (including the catalytic site) and the 5 ′ -UTR (containing the RBS) of the adjacent cbbS gene and introduced a stop codon in the Nterminus of cbbS (Figure 2 and Supplementary Materials).Since the coding sequences of the two RuBisCO subunits were affected, we named this mutant ∆cbbLS.
Plants 2024, 13, x FOR PEER REVIEW 4 of 15 Some of the species analysed in Figure 1 possess a second copy of cbbL, although not in the context of an additional cbb operon.Therefore, we did not concatenate these extra cbbL copies with cbbP and cbbR because an artifact might be produced by mixing sequences that are in the cbb operon with the monocistronic cbbL copies.Instead, we performed a second phylogenetic analysis using only cbbL and observed that all the cbbL monocistronic paralogs grouped together in a separate clade (Supplementary Figure S1).This result indicates that there is considerable sequence divergence between cbbL paralogs in each of the species, suggesting that these extra copies could be originated in gene duplications rather than horizontal gene transfer, a hypothesis already considered for evolution of RuBisCO in a wider taxonomic context [15].In agreement with this hypothesis, in the known example of Bradyrhizobium sp.ORS278, the cbbL paralog seems inactive [22].

Different Effects of ΔcbbLS Mutation on Early and Late Root Adhesion and Colonization
We obtained a deletion mutant in B. diazoefficiens USDA 110 by removing 1092 bp encompassing the cbbL coding sequence (including the catalytic site) and the 5'-UTR (containing the RBS) of the adjacent cbbS gene and introduced a stop codon in the N-terminus of cbbS (Figure 2 and Supplementary Materials).Since the coding sequences of the two RuBisCO subunits were affected, we named this mutant ΔcbbLS.The TGA stop codon that is generated at the 5' of the cbbS ORF is underlined.The wild-type intergenic region between cbbL and cbbS contains 11 bp (not to scale in the scheme).The 1092 bp deletion eliminated most of the coding region from cbbL including its catalytic domain, the cbbL stop codon, as well as the cbbS RBS.The TGA stop codon generated at the N-terminus of cbbS is in-frame with the ATG start codon of cbbL, but not in-frame with the cbbS ATG start codon.In addition, five additional stop codons were generated downstream (not indicated in the scheme; for details, see supplementary Figure S6).The mutant ΔcbbLS gene fragment contains an additional 21 bp segment added by the mutagenesis strategy.
Then, we studied the early interaction of this mutant with soybean plants.For these experiments, we obtained B. diazoefficiens cells from liquid cultures in HMY-Mtl grown at the exponential phase, which are optimal conditions for cbb operon expression, the percentage of viable cells, adhesiveness to plant roots, infectiveness, and competitiveness for nodulation [21,23,24].In addition, we observed that ΔcbbLS and the wild-type USDA 110 grew equally well in liquid HMY-Mtl (Figure 3) whereby differences in adsorption or root The TGA stop codon that is generated at the 5' of the cbbS ORF is underlined.The wild-type intergenic region between cbbL and cbbS contains 11 bp (not to scale in the scheme).The 1092 bp deletion eliminated most of the coding region from cbbL including its catalytic domain, the cbbL stop codon, as well as the cbbS RBS.The TGA stop codon generated at the N-terminus of cbbS is in-frame with the ATG start codon of cbbL, but not in-frame with the cbbS ATG start codon.In addition, five additional stop codons were generated downstream (not indicated in the scheme; for details, see supplementary Figure S6).The mutant ∆cbbLS gene fragment contains an additional 21 bp segment added by the mutagenesis strategy.
Then, we studied the early interaction of this mutant with soybean plants.For these experiments, we obtained B. diazoefficiens cells from liquid cultures in HMY-Mtl grown at the exponential phase, which are optimal conditions for cbb operon expression, the percentage of viable cells, adhesiveness to plant roots, infectiveness, and competitiveness for nodulation [21,23,24].In addition, we observed that ∆cbbLS and the wild-type USDA 110 grew equally well in liquid HMY-Mtl (Figure 3) whereby differences in adsorp-tion or root colonization described below cannot be attributed to differences in previous inoculum growth.
Plants 2024, 13, x FOR PEER REVIEW 5 of 15 colonization described below cannot be attributed to differences in previous inoculum growth.In short-term experiments, we incubated soybean roots in liquid MFS for 1 h with USDA 110 and the ΔcbbLS mutant and observed that in these conditions early adsorption of the ΔcbbLS mutant was higher than that of the wild type (Figure 4a).This higher adsorption correlated with a higher production of extracellular polysaccharides by the Δcb-bLS mutant, which had 6.6 ± 0.6 µg per optical density unit at 500 nm (OD500), while the wild type had 4.4 ± 0.2 µg extracellular polysaccharide per OD500 unit.In another series of experiments, we incubated ΔcbbLS with the plants for 48 h in vermiculite pots at 28 °C to measure late adsorption and rhizosphere colonization.We measured the colonization of the root environment by detaching bacteria from rhizosphere and roots with a mild wash and firmly adsorbed bacteria by vortexing at maximum speed.In comparison to the wild type, a substantial reduction in the numbers of CFU released was noticed, especially in the fraction corresponding to firm adsorption to root In short-term experiments, we incubated soybean roots in liquid MFS for 1 h with USDA 110 and the ∆cbbLS mutant and observed that in these conditions early adsorption of the ∆cbbLS mutant was higher than that of the wild type (Figure 4a).This higher adsorption correlated with a higher production of extracellular polysaccharides by the ∆cbbLS mutant, which had 6.6 ± 0.6 µg per optical density unit at 500 nm (OD 500 ), while the wild type had 4.4 ± 0.2 µg extracellular polysaccharide per OD 500 unit.
Plants 2024, 13, x FOR PEER REVIEW 5 of 15 colonization described below cannot be attributed to differences in previous inoculum growth.In short-term experiments, we incubated soybean roots in liquid MFS for 1 h with USDA 110 and the ΔcbbLS mutant and observed that in these conditions early adsorption of the ΔcbbLS mutant was higher than that of the wild type (Figure 4a).This higher adsorption correlated with a higher production of extracellular polysaccharides by the Δcb-bLS mutant, which had 6.6 ± 0.6 µg per optical density unit at 500 nm (OD500), while the wild type had 4.4 ± 0.2 µg extracellular polysaccharide per OD500 unit.In another series of experiments, we incubated ΔcbbLS with the plants for 48 h in vermiculite pots at 28 °C to measure late adsorption and rhizosphere colonization.We measured the colonization of the root environment by detaching bacteria from rhizosphere and roots with a mild wash and firmly adsorbed bacteria by vortexing at maximum speed.In comparison to the wild type, a substantial reduction in the numbers of CFU released was noticed, especially in the fraction corresponding to firm adsorption to root In another series of experiments, we incubated ∆cbbLS with the plants for 48 h in vermiculite pots at 28 • C to measure late adsorption and rhizosphere colonization.We measured the colonization of the root environment by detaching bacteria from rhizosphere and roots with a mild wash and firmly adsorbed bacteria by vortexing at maximum speed.In comparison to the wild type, a substantial reduction in the numbers of CFU released was noticed, especially in the fraction corresponding to firm adsorption to root surfaces (Figure 4b,c).Therefore, it seems plausible that, despite a lower initial adsorption, the wild type either achieved higher firm adsorption [25,26] or grew better in the rhizospheric environment than the ∆cbbLS mutant leading to better colonization, although these two possibilities are not mutually exclusive.

∆cbbLS Mutation Affects Competition for Nodulation
To observe if this mutation produced an effect on nodulation, the ∆cbbLS strain and the wild-type USDA 110, grown in the same conditions as above, were inoculated each in five soybean plants, which were cultured under hydroponic conditions in pots irrigated with N-free MFS plant nutrient solution for 30 days after inoculation (dai).At this time, plants were removed, roots were washed, and they were observed for nodulation.Both the ∆cbbLS mutant and the wild type produced around 20 morphologically normal nodules per plant, without differences in nodules' positions or sizes (Supplementary Figure S2).There were also no differences in shoot dry weight (around 470 mg plant −1 ) or chlorophyll contents (around 31 SPAD units plant −1 ) between USDA 110 and the ∆cbbLS mutant.However, when each of these strains was coinoculated in a 1:1 proportion with LP 3004the nearly isogenic Sm-derivative from USDA 110 [27,28]-the wild type showed 10-fold higher nodule occupancy than that of the mutant strain at 30 dai, despite LP 3004 being consistently more competitive (Figure 5a).
Plants 2024, 13, x FOR PEER REVIEW 6 of 15 surfaces (Figure 4b,c).Therefore, it seems plausible that, despite a lower initial adsorption, the wild type either achieved higher firm adsorption [25,26] or grew better in the rhizospheric environment than the ΔcbbLS mutant leading to better colonization, although these two possibilities are not mutually exclusive.

ΔcbbLS Mutation Affects Competition for Nodulation
To observe if this mutation produced an effect on nodulation, the ΔcbbLS strain and the wild-type USDA 110, grown in the same conditions as above, were inoculated each in five soybean plants, which were cultured under hydroponic conditions in pots irrigated with N-free MFS plant nutrient solution for 30 days after inoculation (dai).At this time, plants were removed, roots were washed, and they were observed for nodulation.Both the ΔcbbLS mutant and the wild type produced around 20 morphologically normal nodules per plant, without differences in nodules' positions or sizes (Supplementary Figure S2).There were also no differences in shoot dry weight (around 470 mg plant −1 ) or chlorophyll contents (around 31 SPAD units plant −1 ) between USDA 110 and the ΔcbbLS mutant.However, when each of these strains was coinoculated in a 1:1 proportion with LP 3004the nearly isogenic Sm-derivative from USDA 110 [27,28]-the wild type showed 10-fold higher nodule occupancy than that of the mutant strain at 30 dai, despite LP 3004 being consistently more competitive (Figure 5a).,c).In (a), nodules occupied by either USDA 110 or the ΔcbbLS mutant were recorded as Sm-sensitive, whereby nodules occupied by LP 3004 alone could not be distinguished from those occupied by both competitor strains because all these nodules were Sm-resistant.Therefore, the sum of percentages in this competition is <100%, because it represents the Sm-sensitive nodules only.In (b,c), USDA 110 and the ΔcbbLS mutant were distinguished by PCR, whereby double occupation was recorded, and therefore the sum of percentages is >100% because nodules containing both strains were counted twice, i.e. as occupied by USDA 110 and by the ΔcbbLS mutant.The percentage of double occupation was 8.20 ± 6.44% (b) or 7.60 ± 5. In these experimental conditions, we were unable to distinguish nodules occupied by only LP 3004 from those occupied by both competitor strains together since both kinds of bacterial samples contained LP 3004 and therefore both grew in the presence of Sm.To assess double occupation, in separate experiments we measured direct competition  ,c).In (a), nodules occupied by either USDA 110 or the ∆cbbLS mutant were recorded as Sm-sensitive, whereby nodules occupied by LP 3004 alone could not be distinguished from those occupied by both competitor strains because all these nodules were Sm-resistant.Therefore, the sum of percentages in this competition is <100%, because it represents the Sm-sensitive nodules only.In (b,c), USDA 110 and the ∆cbbLS mutant were distinguished by PCR, whereby double occupation was recorded, and therefore the sum of percentages is >100% because nodules containing both strains were counted twice, i.e., as occupied by USDA 110 and by the ∆cbbLS mutant.The percentage of double occupation was 8.20 ± 6.44% (b) or 7.60 ± 5.28 (c) (average ± confidence interval with p < 0.05).In all graphs, bars represent the average from 10 (a) or 8 (b,c) plants, error bars represent confidence intervals with p < 0.05, and asterisks indicate significant differences with p < 0.005 obtained by one-way ANOVA.All results are representative from two independent experiments.In these experimental conditions, we were unable to distinguish nodules occupied by only LP 3004 from those occupied by both competitor strains together since both kinds of bacterial samples contained LP 3004 and therefore both grew in the presence of Sm.To assess double occupation, in separate experiments we measured direct competition between B. diazoefficiens USDA 110 and the ∆cbbLS mutant, by coinoculating both strains in a 1:1 proportion and carrying out the experiment as described above, and after obtaining the bacteria from the nodules we determined nodule occupation by PCR using the FwUp_RBC and RvDw_RBC primers (Table 1).In this way, the nodules occupied by USDA 110 rendered a single 1692 bp band, those occupied by the ∆cbbLS mutant rendered a single 621 bp band, and those occupied simultaneously by both strains (double occupation) rendered both bands.We observed that double nodule occupancy was around 8%, which is similar to values observed previously [27] and corroborated that USDA 110 occupied almost 10 times more nodules than each of the ∆cbbLS mutant clones (Figure 5b,c).In all these experiments, control plants inoculated with each strain alone formed nodules occupied by only the inoculated strain, while no nodules were observed in the uninoculated controls.Together, the similar nodulation achieved with each strain inoculated alone, and the 10-fold reduction in nodules occupation by the mutant strain under competition indicated that, although the ∆cbbLS mutant is able to carry out the complete infection and nodulation process, it is less efficient at some early step of the interaction.Our results above indicated that this step may be related to colonization of soybean roots.

Discussion
The cbb operon is irregularly distributed among Hyphomicrobiales, making it difficult to appreciate the importance of the CBB cycle for adaptation of these bacteria to soil conditions.The genus Bradyrhizobium, within Hyphomicrobiales, is no exception, since some members such as B. diazoefficiens possess only one copy of the cbb operon, others such as the photosynthetic Bradyrhizobium sp.BTAi possess two, and others still such as B. manausense do not have any.This diversity suggests the existence of horizontal gene transfer events among these bacteria, but our comparison of the phylogenetic relationships between three CBB key genes of cbb operon with that of housekeeping genes indicated a strong correlation between both phylogenies.The exception to this rule was Labrys sp.KNU-23.This genus was first isolated from silt of Lake Mustijärv in Estonia by Vasil'eva and Semenov in 1984 [29], and since then, different Labrys species have been isolated and described from different parts of the world; notably, several strains were isolated from legume nodules [30,31] or plant tissues [32], sharing the environment with members of Bradyrhizobium and relatives.Therefore, these close contacts might favour the exchange of genetic material among bacterial cells.In addition, we found a sequence signature upstream of Labrys sp.KNU-23 cbb operon that might indicate horizontal gene transfer, although evidence is tenuous.
The cbb operon in B. diazoefficiens USDA 110 seems regulated by cbbR, as in the close relative Rhodopseudomonas palustris [14,33], although the inducer of CbbR is still unknown (Supplementary Figure S3).We observed that the cbb operon is induced in liquid aerobic cultures with D-mannitol as a carbon source, but not with L-arabinose as a carbon source, indicating some metabolic specificity in the conditions under which the CBB cycle is required in an ambient air atmosphere [21].Since the early preinfection events in the B. diazoefficiens-soybean symbiosis elapses in the soil atmosphere, we wondered if the CBB cycle might play some role at this stage of the symbiotic interaction by characterizing root colonization and nodulation of the B. diazoefficiens ∆cbbLS mutant.In short-time incubations, the ∆cbbLS mutant had a higher level of early adsorption than that of the wild type, but in longer term incubation, the ∆cbbLS mutant was impaired in root colonization.Adsorption of rhizobia to plant roots was characterized as a two-step process: an early and reversible adsorption occurs in the first few hours of contact between roots and bacteria, followed by a late and irreversible adsorption past 12 h or more [25].These different adsorption steps seem mediated by different adhesin molecules [26,34].Therefore, our results indicate that the ∆cbbLS mutant was differently affected for these kinds of adsorption.In this regard, the behaviour at early adsorption correlated with extracellular polysaccharide levels, which is known to affect this early preinfection step [24].As a non-exclusive alternative, we may consider that growth in the rhizosphere may have been hampered in the ∆cbbLS mutant.
In another series of experiments, the ∆cbbLS mutant strain nodulated soybean plants normally and produced the same numbers of nodules as the wild type at vegetative stage.The nodules produced by the mutant displayed an external morphology alike functional nodules, with similar distribution along the roots as wild-type nodules, indicating that the infection and nodulation processes were not altered by the mutation.In agreement with these observations, cbbL or cbbS transcripts or gene products were not detected in wild-type B. diazoefficiens bacteroids obtained from active soybean nodules [35][36][37] and cbbS transcripts were found only in bacteroids from senescent soybean nodules [37], indicating that RuBisCO is not required for the symbiotic lifestyle in B. diazoefficiens.
Differently from nodulation by each strain inoculated alone, when the ∆cbbLS mutant was coinoculated with the wild-type strain, the former occupied very few nodules although all nodules were normal again, indicating that some preinfection process was affected by the mutation, in agreement with our observation that ∆cbbLS was affected in its ability to colonize the soybean roots.Taken together, all these results indicate that, in B. diazoefficiens, RuBisCO should be necessary only in the free-living state before infection, as well as after abandoning the senescent nodule to reinitiate a new cycle of soil and rhizosphere colonization.
In general, bacteria thriving in the rhizosphere benefit from the nutrient richness found there, which comes from passive and active processes of exudation from the plant roots.The composition of root exudates is rather complex and includes aminoacids, organic acids, carbohydrates, secondary metabolites, and larger molecules such as polysaccharides, proteins, and glycoproteins, some of which are involved in molecular signal exchanges between rhizobia and the plant symbiont [38][39][40][41][42][43][44].In these studies, the presence of arabinose was detected in soybean root exudates, although it is not specified whether it is L-arabinose or D-arabinose, which is an important distinction because, while L-arabinose is catabolized through the LKDA pathway and induces the synthesis of lateral flagella in B. diazoefficiens [21], D-arabinose does not induce the synthesis of these flagella [45], and, according to KEGG metabolic charts for B. diazoefficiens, this isomer is probably catabolized by the PPP-CBB cycle.In addition, the presence of D-mannitol in soybean root exudates was never reported-which is intriguing because this polyol is one of the best C-sources for Bradyrhizobium spp.-although other sugars that may be catabolized similarly by this bacterium, such as D-glucose or D-fructose [1], are commonly observed in root exudates.Nevertheless, chemical detection and identification of all the molecular components of root exudates is difficult, and in addition, root exudate composition varies along root position, time, and according to plant status [46][47][48][49].
In contrast to B. diazoefficiens, it was shown that the photosynthetic Bradyrhizobium sp.ORS278 requires RuBisCO for efficient N 2 fixation in symbiosis with Aeschynomene indica L. [22].This bacterium possesses two copies of cbbL and cbbS, one set in the canonical cbb operon and the others at a different location in the genome, which were named 1 and 2, respectively.However, only cbbLS1 is expressed in A. indica nodules.A ∆cbbL1 mutant is severely affected by N 2 fixation in symbiosis with these plants but is not affected Plants 2024, 13, 2362 9 of 14 at all in N 2 fixation in vitro, which indicates that the effect of RuBisCO inactivation is not directly related to nitrogenase activity.Based on previous studies on photosynthetic bacteria, the authors proposed that during a precise time interval between decrease in oxygen concentration in the nodule and the onset of nitrogenase activity, an excess of reducing power might build up from electrons produced at the final steps of the oxidation of plant-supplied organic carbon, thus leading to imbalance of redox potential.To overcome this situation, the CBB cycle might serve as sink of reducing power excess, and therefore the absence of RuBisCO activity might indirectly preclude N 2 fixation by producing a metabolic disturbance in the nodule [22].In the case of the non-photosynthetic B. diazoefficiens, such disturbance might take place outside the nodule.However, it should be noted that both the wild type and the ∆cbbLS mutant grew equally well in HME-Mtl.Because, in our experiments, soybean plantlets were inoculated with HME-Mtl-grown rhizobia, it seems that, if an excess of reducing power is built up in the rhizosphere, root exudates should provide some electron donors in addition to the carbon sources.However, the possible presence of such additional electron donors was not reported at that time in the soybean root exudates.
In the rhizosphere environment, CO 2 concentration is high, especially around actively growing roots, due to the combined effects of root respiration, hampered diffusion of CO 2 in water, and air confinement in the soil pores [50,51].Therefore, CO 2 fixation might be an advantageous trait for bacteria that colonize and associate to plant roots.Hence, the role of RuBisCO (and PRK) in the rhizosphere might be more related to CO 2 fixation to harness the high CO 2 concentration around actively growing roots, where the points of entry by infection threads are located.However, more experimental evidence is required to proof this hypothesis, such as direct measurements of CO 2 consumption by the wild type and the mutant in the rhizosphere.If this is the case, an attractive possibility is that inoculants based on B. diazoefficiens grown with D-mannitol might sequestrate CO 2 in the rhizosphere, thus making an additional contribution to sustainable agriculture, beyond its largely known role as N 2 -fixer.Field experiments with measurements of CO 2 evolution in the soil surface are required to assess this possibility.

Construction of the ∆cbbLS Mutant
We carried out general cloning procedures essentially as described [54].For mutagenesis, we proceeded essentially as described [55].To delete a 1092 bp fragment containing the cbbL coding sequence and the 5 ′ -UTR of the adjacent cbbS from the B. diazoefficiens USDA 110 genome, we proceeded as follows: We amplified the fragments Up_RBC (encompassed by primers FwUp_RBC and RvUp_RBC) and Dw_RBC (encompassed by primers FwDw_RBC and RvDw_RBC) by PCR (Table 1).We designed the primers in order to possess 21 nt complementary sequences at the Up_RBC-3 ′ and Dw_RBC-5 ′ ends, in such a way that they may form a single fragment to be amplified with the FwUp_RBC and RvDw_RBC primers, giving rise to the 621 bp UpDw_RBC fragment (Supplementary Figure S4).In addition, we designed a BamHI site at the 21 bp joint between Up_RBC and Dw_RBC.We purified this fragment and inserted it in the pK18mobsacB suicide vector with EcoRI and HindIII, whose restriction sequences were incorporated in the 5 ′ ends of FwUp_RBC and RvDw_RBC primers (Table 1).In this way, we obtained the pCC1 plasmid, which was electroporated into E. coli DH5α.After selection of transformants carrying pCC1 in Km, we purified the plasmid and checked the correct insertion of the UpDw_RBC fragment by PCR with M13 primers and plasmid digestion (Supplementary Figure S5).Then, we electroporated the pCC1 plasmid into E. coli S17-1 and transferred it to B. diazoefficiens USDA 110 by biparental mating as described [55].Next, we selected single crossing-over B. diazoefficiens recombinants (Supplementary Figure S6) by plating in the presence of Cm and Km, and in a second step, we selected double crossing-over recombinants (Supplementary Figure S7) by plating in the presence of 10% (w/v) sucrose and confirming that the sucrose-resistant selected clones were Km-sensitive.Since we observed that the first four nucleotides in the cbbS coding sequence are ATGA, we planned the FwUp_RBC and RvUp_RBC primers so as to amplify the first 380 bp of cbbL and link it to the first bp of cbbS in order to produce a frameshift that replaces the ATG start codon of cbbS by a TGA stop codon (Supplementary Figure S8).In doing so, we observed that five additional stop codons were introduced in the frame-shifted cbbS sequence.We confirmed these mutations in the B. diazoefficiens derivative strain by sequencing the entire 818 bp region using the FwExt_RBC and RvExt_RBC primers (Table 1 and Supplementary Figure S6).Sequencing was carried out at Macrogen Inc. (Seoul, Republic of Korea).This mutant is referred as ∆cbbLS.

Sequence Comparisons and Construction of Phylogenetic Trees
We compared the amino acid sequence of CbbL from B. diazoefficiens USDA 110 to the BLASTP database [56], excluding the Bradyrhizobium genus and including only Hyphomicrobiales members.We set alignment parameters to e-value = 0.0005, BLOSUM62 matrix, and a maximum of 250 targets.From the sequences obtained, we chose those included in cbb operons and with higher identity to B. diazoefficiens USDA 110, then filtered them in such a way to avoid repeated species as well as those labelled as "multispecies".When a species was repeated in the list of sequences, we picked those sequences with the best match to B. diazoefficiens USDA 110 CbbL.In addition, we chose Rhodobacter sphaeroides as outgroup.Then, we included CbbP and CbbR sequences from the strains chosen in the first step in a concatenated CbbR-CbbP-CbbL sequence, which we submitted to a MEGA X analysis [57] to obtain the multiple alignment and to construct the phylogenetic tree using a neighbour-joining statistical analysis [58].In parallel, we carried out this procedure with these same strains using a concatenated sequence of the housekeeping proteins AtpD, GyrB, RecA, RpoB, and RpoD.In a separate analysis, we compared only CbbL, but while including all paralogs in the same set of species.

Plant Assays
Soybean seeds SYN4x1RR were kindly provided by Syngenta Argentina, Vicente López, Argentina.We surface-sterilized and germinated the seeds in water agar 1.5% (w/v) as described [23,24].For inoculation, we cultured B. diazoefficiens in liquid HMY-Mtl until exponential phase, evaluated by optical density at 500 nm (OD 500 ) and CFU counts.
To evaluate early adsorption, we incubated the bacteria with 10 roots for 1 h as described [23].After washing the soybean roots, we quantified the adsorbed bacteria by releasing them from the roots by shaking them separately in N-free modified Fåhraeus plant nutrient solution (MFS) [59] and counting the remaining bacteria adsorbed to each root by counting the developed microcolonies as described [23,33].From these counts, we obtained the adsorption index (%A) by relating the total number of bacteria adsorbed, expressed as percentage of the total number of bacteria inoculated [23].
To evaluate rhizosphere colonization, we surface-sterilized seeds and germinated them in water agar as above, and when rootlets reached 2-3 cm in length, we planted them in 5 sterile vermiculite pots per condition at 1 plant per pot, which were irrigated with MFS to field capacity.We inoculated each plant with 1 mL of each B. diazoefficiens strain separately, which were grown in the same conditions already described, capped the pots with sterile film, and left the pots at 28 • C in the dark during 48 h.We then removed the plants from each pot, cut the roots, and immersed each one in separate 50 mL Falcon tubes containing 10 mL of sterile MFS.We agitated the tubes in a rotary shaker at 180 rpm for 5 min and took samples from the MFS supernatants, diluted them, and plated them in YMA.Afterward, we carefully removed roots, blotted each one for a few seconds on sterile paper towels, and suspended each root in a new Falcon tube containing 10 mL fresh sterile MFS.We vortexed this second suspension at maximum speed for 2 min and, again, obtained samples from the MFS supernatant of each tube, diluted them, and plated them in YMA.We counted CFU in 10 replicas from each sample to evaluate the mild and tight adsorption to roots, respectively.In all these experiments, we inoculated the same number of plants with MFS without bacteria as negative controls.
For nodulation assays, we cultivated one plant per 500 mL pot containing sterile vermiculite watered with MFS at planting.Each pot received one pregerminated seedling, the number of pots per strain being indicated for each experiment in the Results section.We inoculated at planting with B. diazoefficiens USDA 110 or ∆cbbLS cultured in the same conditions as above.We grew the plants in the greenhouse at 30 • C/20 • C day/night temperature and provided watering as required, in sequences comprising two irrigations with sterile distilled water followed by one with sterile MFS.At thirty dai, we removed the plants, observed the nodules external morphology, and counted the nodules.
We measured competition for nodulation as described [57] using B. diazoefficiens LP 3004 as reference strain, since it is a Sm-resistant almost-isogenic derivative from USDA 110 [27,50].Each strain, namely USDA 110 or ∆cbbLS, were competed against LP 3004 in 1:1 mixed inocula, prepared as described above.Before mixing, we diluted culture samples so as to set their OD 500 at the same value for all strains, and then we carried out appropriate dilutions in MFS to set the initial CFU of each sample to approximately 10 6 •mL −1 .We corroborated the actual numbers of CFU•mL −1 by colony counting 10 replicas per sample in YMA.As controls, we inoculated sets of pots with each strain separately, or with sterile MFS without bacteria.We performed inoculation by irrigating the pots with the bacterial suspensions as described [60].Plants were cultivated as described above and collected at 30 dai.At this time, we removed nodules, surface-sterilized them, extracted the bacteria from the interior of each nodule, and cultured the bacterial extracts in YMA replica plates with or without 400 mg•L −1 Sm.Finally, we calculated the proportion of nodules containing antibiotic-sensitive bacteria for each plant as a measure of competitiveness.In these experimental conditions, we were unable to distinguish nodules occupied by only LP 3004 from those occupied by both competitor strains since both kinds of bacterial samples grew in the presence of Sm.To distinguish nodules occupied by both strains simultaneously, we coinoculated USDA 110 with each of two different clones of the ∆cbbLS mutant and determined nodules occupancy by PCR using the FwUp_RBC and RvDw_RBC primers as described above.
We measured chlorophyll contents in leaves and shoot dry weights as described [61].All the above-described experiments were carried out twice in their entirety, and results from one representative independent experiment are shown.

Extracellular Polysaccharide Determinations
We measured exopolysaccharide and capsular polysaccharide contents from bacterial cultures with anthrone as described [24], using glucose as standard.

Figure 1 .
Figure 1.Comparison of the phylogenetic trees obtained with concatenated atpD, gyrB, recA, rpoB, and rpoD housekeeping genes (left) or concatenated cbbR, cbbP, and cbbL as representative of the cbb operon (right).The position of Labrys sp.KNU-23, which differs between both trees, is indicated by dotted red lines.Rhodobacter sphaeroides 2.4.1 was used as outgroup.Numbers indicate the percentage of replicate trees in which the associated taxa clustered together in the bootstrap test (1000 replicates).

Figure 1 .
Figure 1.Comparison of the phylogenetic trees obtained with concatenated atpD, gyrB, recA, rpoB, and rpoD housekeeping genes (left) or concatenated cbbR, cbbP, and cbbL as representative of the cbb operon (right).The position of Labrys sp.KNU-23, which differs between both trees, is indicated by dotted red lines.Rhodobacter sphaeroides 2.4.1 was used as outgroup.Numbers indicate the percentage of replicate trees in which the associated taxa clustered together in the bootstrap test (1000 replicates).

Figure 2 .
Figure 2. Scheme of the USDA 110 wild-type and ΔcbbLS mutant DNA regions.The TGA stop codon that is generated at the 5' of the cbbS ORF is underlined.The wild-type intergenic region between cbbL and cbbS contains 11 bp (not to scale in the scheme).The 1092 bp deletion eliminated most of the coding region from cbbL including its catalytic domain, the cbbL stop codon, as well as the cbbS RBS.The TGA stop codon generated at the N-terminus of cbbS is in-frame with the ATG start codon of cbbL, but not in-frame with the cbbS ATG start codon.In addition, five additional stop codons were generated downstream (not indicated in the scheme; for details, see supplementary FigureS6).The mutant ΔcbbLS gene fragment contains an additional 21 bp segment added by the mutagenesis strategy.

Figure 2 .
Figure 2. Scheme of the USDA 110 wild-type and ∆cbbLS mutant DNA regions.The TGA stop codon that is generated at the 5' of the cbbS ORF is underlined.The wild-type intergenic region between cbbL and cbbS contains 11 bp (not to scale in the scheme).The 1092 bp deletion eliminated most of the coding region from cbbL including its catalytic domain, the cbbL stop codon, as well as the cbbS RBS.The TGA stop codon generated at the N-terminus of cbbS is in-frame with the ATG start codon of cbbL, but not in-frame with the cbbS ATG start codon.In addition, five additional stop codons were generated downstream (not indicated in the scheme; for details, see supplementary FigureS6).The mutant ∆cbbLS gene fragment contains an additional 21 bp segment added by the mutagenesis strategy.

Figure 3 .
Figure 3. Growth of USDA 110 and ΔcbbLS in HMY-Mtl medium.Data are averages ± SD from three independent cultures.When error bars are not shown, they are smaller than the symbol.

Figure 4 .
Figure 4. Early interactions of B. diazoefficiens USDA 110 and ΔcbbLS with soybean (Glycine max (L.) Merr.) roots.(a) Initial reversible adsorption measured after 1 h incubation; (b,c) late firm adsorption and rhizosphere colonization, measured after 48 h incubation.In (b), bacterial cells were recovered from roots and rhizosphere by shaking at 180 rpm during 5 min.In (c), the remaining bacterial cells after the first wash were recovered by vortexing at maximum speed during 2 min.In all graphs, bars represent average from 10 plants (a) or 5 plants (b,c), error bars represent confidence intervals with p < 0.05, and asterisks indicate significant differences with p < 0.05 (one asterisk) or 0.005 (three asterisks) obtained by one-way ANOVA.All results are representative from at least two independent experiments.

Figure 3 .
Figure 3. Growth of USDA 110 and ∆cbbLS in HMY-Mtl medium.Data are averages ± SD from three independent cultures.When error bars are not shown, they are smaller than the symbol.

Figure 3 .
Figure 3. Growth of USDA 110 and ΔcbbLS in HMY-Mtl medium.Data are averages ± SD from three independent cultures.When error bars are not shown, they are smaller than the symbol.

Figure 4 .
Figure 4. Early interactions of B. diazoefficiens USDA 110 and ΔcbbLS with soybean (Glycine max (L.) Merr.) roots.(a) Initial reversible adsorption measured after 1 h incubation; (b,c) late firm adsorption and rhizosphere colonization, measured after 48 h incubation.In (b), bacterial cells were recovered from roots and rhizosphere by shaking at 180 rpm during 5 min.In (c), the remaining bacterial cells after the first wash were recovered by vortexing at maximum speed during 2 min.In all graphs, bars represent average from 10 plants (a) or 5 plants (b,c), error bars represent confidence intervals with p < 0.05, and asterisks indicate significant differences with p < 0.05 (one asterisk) or 0.005 (three asterisks) obtained by one-way ANOVA.All results are representative from at least two independent experiments.

Figure 4 .
Figure 4. Early interactions of B. diazoefficiens USDA 110 and ∆cbbLS with soybean (Glycine max (L.) Merr.) roots.(a) Initial reversible adsorption measured after 1 h incubation; (b,c) late firm adsorption and rhizosphere colonization, measured after 48 h incubation.In (b), bacterial cells were recovered from roots and rhizosphere by shaking at 180 rpm during 5 min.In (c), the remaining bacterial cells after the first wash were recovered by vortexing at maximum speed during 2 min.In all graphs, bars represent average from 10 plants (a) or 5 plants (b,c), error bars represent confidence intervals with p < 0.05, and asterisks indicate significant differences with p < 0.05 (one asterisk) or 0.005 (three asterisks) obtained by one-way ANOVA.All results are representative from at least two independent experiments.

Figure 5 .
Figure 5. Competition for soybean (Glycine max (L.) Merr.) nodulation between B. diazoefficiens USDA 110 and ΔcbbLS mutants.(a) Percentage of nodules occupied by USDA 110 or ΔcbbLS in competition against the Sm-resistant LP 3004, a near-isogenic spontaneous derivative from USDA 110; (b,c) direct competition between USDA 110 and the ΔcbbLS mutant clone 1 (b) or clone 2 (c).Each bar represents the percentage of nodules occupied by the referred strain either as single occupant (a) or in single occupation + double occupation together with the other strain (b,c).In (a), nodules occupied by either USDA 110 or the ΔcbbLS mutant were recorded as Sm-sensitive, whereby nodules occupied by LP 3004 alone could not be distinguished from those occupied by both competitor strains because all these nodules were Sm-resistant.Therefore, the sum of percentages in this competition is <100%, because it represents the Sm-sensitive nodules only.In (b,c), USDA 110 and the ΔcbbLS mutant were distinguished by PCR, whereby double occupation was recorded, and therefore the sum of percentages is >100% because nodules containing both strains were counted twice, i.e. as occupied by USDA 110 and by the ΔcbbLS mutant.The percentage of double occupation was 8.20 ± 6.44% (b) or 7.60 ± 5.28 (c) (average ± confidence interval with p < 0.05).In all graphs, bars represent the average from 10 (a) or 8 (b,c) plants, error bars represent confidence intervals with p < 0.05, and asterisks indicate significant differences with p < 0.005 obtained by one-way ANOVA.All results are representative from two independent experiments.
Figure 5. Competition for soybean (Glycine max (L.) Merr.) nodulation between B. diazoefficiens USDA 110 and ΔcbbLS mutants.(a) Percentage of nodules occupied by USDA 110 or ΔcbbLS in competition against the Sm-resistant LP 3004, a near-isogenic spontaneous derivative from USDA 110; (b,c) direct competition between USDA 110 and the ΔcbbLS mutant clone 1 (b) or clone 2 (c).Each bar represents the percentage of nodules occupied by the referred strain either as single occupant (a) or in single occupation + double occupation together with the other strain (b,c).In (a), nodules occupied by either USDA 110 or the ΔcbbLS mutant were recorded as Sm-sensitive, whereby nodules occupied by LP 3004 alone could not be distinguished from those occupied by both competitor strains because all these nodules were Sm-resistant.Therefore, the sum of percentages in this competition is <100%, because it represents the Sm-sensitive nodules only.In (b,c), USDA 110 and the ΔcbbLS mutant were distinguished by PCR, whereby double occupation was recorded, and therefore the sum of percentages is >100% because nodules containing both strains were counted twice, i.e. as occupied by USDA 110 and by the ΔcbbLS mutant.The percentage of double occupation was 8.20 ± 6.44% (b) or 7.60 ± 5.28 (c) (average ± confidence interval with p < 0.05).In all graphs, bars represent the average from 10 (a) or 8 (b,c) plants, error bars represent confidence intervals with p < 0.05, and asterisks indicate significant differences with p < 0.005 obtained by one-way ANOVA.All results are representative from two independent experiments.

Figure 5 .
Figure 5. Competition for soybean (Glycine max (L.) Merr.) nodulation between B. diazoefficiens USDA 110 and ∆cbbLS mutants.(a) Percentage of nodules occupied by USDA 110 or ∆cbbLS in competition against the Sm-resistant LP 3004, a near-isogenic spontaneous derivative from USDA 110; (b,c) direct competition between USDA 110 and the ∆cbbLS mutant clone 1 (b) or clone 2 (c).Each bar represents the percentage of nodules occupied by the referred strain either as single occupant (a) or in single occupation + double occupation together with the other strain (b,c).In (a), nodules occupied by either USDA 110 or the ∆cbbLS mutant were recorded as Sm-sensitive, whereby nodules occupied by LP 3004 alone could not be distinguished from those occupied by both competitor strains because all these nodules were Sm-resistant.Therefore, the sum of percentages in this competition is <100%, because it represents the Sm-sensitive nodules only.In (b,c), USDA 110 and the ∆cbbLS mutant were distinguished by PCR, whereby double occupation was recorded, and therefore the sum of percentages is >100% because nodules containing both strains were counted twice, i.e., as occupied by USDA 110 and by the ∆cbbLS mutant.The percentage of double occupation was 8.20 ± 6.44% (b) or 7.60 ± 5.28 (c) (average ± confidence interval with p < 0.05).In all graphs, bars represent the average from 10 (a) or 8 (b,c) plants, error bars represent confidence intervals with p < 0.05, and asterisks indicate significant differences with p < 0.005 obtained by one-way ANOVA.All results are representative from two independent experiments.
: Phylogenetic analysis of CbbL sequences, Figure S2: nodules and symbiotic parameters from plants inoculated with the wild type or the ∆cbbLS separately, Figure S3: Structure and regulation of the cbb operon, Figure S4: Schematic diagram of the strategy used to clone the UpDw_RBC fragment into pK18mobsacB, Figure S5: Plasmid pCC1 containing the UpDw_RBC fragment, Figure S6: Single crossing-over insertion of pCC1 into

Table 1 .
Primers used in this study.The restriction sites introduced in the primers are indicated in bold, and the 21 nt fragments introduced to generate complementary sequences between RvUp_RBC and FwDw_RBC are underlined.