Insights into the Phylogeny, Nodule Function, and Biogeographic Distribution of Microsymbionts Nodulating the Orphan Kersting’s Groundnut [Macrotyloma geocarpum (Harms) Marechal & Baudet] in African Soils

Rhizobia play important roles in agroecosystems, where they contribute to improving overall soil health through their symbiotic relationship with legumes. This study explored the microsymbionts nodulating Kersting’s groundnut, a neglected orphan legume. The results revealed the presence of different bradyrhizobial populations with high N2-fixing efficiencies as the dominant symbionts of this legume across diverse agroecologies in Africa. Our findings represent a useful contribution to the literature in terms of the community of microsymbionts nodulating a neglected cultivated legume and its potential for elevation as a major food crop. The presence of potentially novel bradyrhizobial symbionts of Kersting’s groundnut found in this study offers an opportunity for future studies to properly describe, characterize, and delineate these isolates functionally and phylogenetically for use in inoculant production to enhance food/nutritional security.


K ersting's groundnut [Macrotyloma geocarpum (Harms) Marechal & Baudet] is an
indigenous African legume which is native to the regions around Togo and central Benin in West Africa (1). This crop is currently cultivated at the subsistence level by older farmers in Ghana and neighboring West African countries (2,3). It has great potential as a food security crop because of the high protein level (23.1%) and the essential amino acid composition of the grain (4). Kersting's groundnut is a droughttolerant legume (5) and is also adapted to growth in N-deficient soils due to its ability to fix atmospheric N 2 when in symbiosis with soil bacteria called rhizobia. The symbiotic process leads to a high supply of N to the plant and the production of substantial grain yield without external chemical inputs (6,7). Despite its N 2 -fixing ability, as well as its nutritional and medicinal benefits (2,8,9), Kersting's groundnut is labeled as an orphan crop and is endangered due to neglect by farmers and researchers (10).
As a grain legume adapted to growth under drought and nutrient limitations in African soils (11), Kersting's groundnut probably harbors novel rhizobia with superior N 2 -fixing traits under those conditions. However, other than the crop being nodulated by Bradyrhizobium sp. strain CB756 under glasshouse conditions (8), there is currently no information on its specific microsymbionts in Africa, its continent of origin. This is despite the fact that studies of rhizobial symbionts of underutilized legumes, such as Kersting's groundnut, in previously unexplored environments could lead to the identification of compatible and effective rhizobial strains that can support the species' survival in such locations. For example, similar studies involving the underutilized Bambara groundnut (Vigna subterranea L. Verdc.) revealed the presence of potentially novel microsymbionts responsible for its nodulation in Ghanaian and South African soils (12).
Several studies have reported the presence of high diversity among rhizobial populations responsible for grain legume nodulation in African soils (12)(13)(14)(15)(16)(17)(18). Globally, the list of N 2 -fixing rhizobia responsible for legume nodulation consists of over 100 species from 14 bacterial genera originating from diverse legumes and contrasting environments across different continents (19). These sets of information have contributed significantly to the current progress in tapping the benefits of the legumerhizobium symbiosis for crop production, especially through the use of inoculant formulations with superior N 2 -fixing rhizobial strains (20,21). However, much remains to be done, as inoculation failures are often encountered when inoculant strains are introduced into new environments (7,20,22), thus stressing the need to bioprospect for effective and highly adapted indigenous rhizobial strains for grain legume nodulation in their native environments. Although rhizobial population studies have contributed significantly to our current understanding of the factors influencing their distribution and survival in diverse environments, it is their N 2 -fixing efficacy that is often considered a prerequisite for their selection as strains for inoculant production. Assessing the symbiotic efficiency of rhizobial strains alongside their diversity and phylogeny is therefore a useful approach to tap the agricultural and ecological benefits of the legume-rhizobium symbiosis (15,21). During the symbiotic process, which commences via intricate molecular dialogue between legumes and their compatible rhizobia, bacteroids in root nodules exchange fixed N for C compounds from photosynthesis (23). As a result, photosynthetic rates and shoot biomass accumulation are often used as measures of N 2 -fixing efficiency in purely symbiotic systems (14).
Despite the nutritional and medicinal value of Kersting's groundnut, as well as its potential as a food/nutritional security crop, there is currently little information on the microsymbionts associated with its nodulation in African soils. The aim of this study was to provide a detailed insight into the morphogenetic diversity, phylogenetic positions, biogeographic distribution, and symbiotic functioning of indigenous rhizobia associated with Kersting's groundnut nodulation in Ghana, South Africa, and Mozambique. Although the crop is native to West Africa, this study further explored the potential wider distribution of its microsymbionts outside the cultivation areas (South Africa and Mozambique). The possible effects of host plant and soil factors on the diversity and biogeography of rhizobial isolates were also explored using multivariate analysis. variables (pH and Ca, K, Mg, Mn, Na, and P concentrations) which showed significant influence on bradyrhizobial distribution (P Յ 0.05 at permutation 999) (Fig. 3). The results of the CCA revealed a total mean square contingency coefficient (inertia) of 22.34, of which 53% was explained by both CCA1 and CCA2 axes. Soil Na and P concentrations showed stronger correlations with the positive direction of the first canonical axis (CCA1), while Ca and Mg correlated with the same CCA1 axis but in the at a cutoff point of 70% similarity (indicated by the vertical red line) are numbered. Where consecutive isolates possess unique PCR profiles, the numbering is skipped and continued at the next group of isolates. Cluster analysis was done with the unweighted pair group method with arithmetic mean (UPGMA) algorithm using the software BioNumerics 7.6. Applied and Environmental Microbiology negative direction (Fig. 3). Similarly, soil pH and K concentration showed significant correlation in opposite directions with the second canonical axis (CCA2). Furthermore, the concentrations of soil P, Na, Mg, and Ca had a greater influence on the distribution of microsymbionts present on the CCA1 axis, just as pH and soil K concentration showed a greater influence on the isolates on the CCA2 axis. The isolates from Nyankpala were more influenced by soil pH than were isolates from other locations (Fig. 3). Phylogeny of Kersting's groundnut isolates based on 16S rRNA gene sequencing. PCR amplification of the 16S rRNA gene yielded bands of approximately 1,500 bp for the 41 representative isolates selected from the different BOX-PCR clusters. Phylogenetic analysis of the 16S rRNA gene assigned the 41 test isolates to seven groups (groups I to VII) within the genus Bradyrhizobium (Fig. 4). In group I, isolates TUTMGGH10, TUTMGGH12, TUTMGGH56, TUTMGGH77, and TUTMGGH86 from Ghana shared 98.3 to 100% sequence similarity with B. kavangense 14-3 T and B. subterraneum 58 2-1 T (Fig. 4). Group II comprised isolates from Mozambique and South Africa (sharing 99.7 to 100% sequence similarity with each other) which stood apart from any reference type strains but shared 99.7 to 99.8% sequence similarity with B. kavangense 14-3 T (Fig.  4). In group III, isolates TUTMGMZQ190, TUTMGMZQ195, and TUTMGMZQ206 (95.3 to 100% sequence similarity), each obtained from nodules of a different landrace in Mozambique (Table S1), stood separately from any reference type strains but shared 95.1 to 99.7% sequence similarity with Bradyrhizobium cajani AMBPC1010 T , Bradyrhizobium liaoningense 2281 (USDA 3622) T , Bradyrhizobium ottawaense OO99 T , and B. kavangense 14-3 T , the most closely related reference type strains. Interestingly, each of the isolates in group III occupied a different BOX-PCR cluster ( Fig. 1 and 4). In group IV, isolates TUTMGSA123, TUTMGSA137, and TUTMGSA145 (95.3% to 100% sequence similarity) from Nelspruit in South Africa shared close relatedness to Bradyrhizobium huanghuaihaiense, Bradyrhizobium ingae, and Bradyrhizobium iriomotense, with 99.5 to 100% sequence similarity (Fig. 4). Again, the three isolates in group IV each occupied FIG 2 Geographic origin of rhizobial isolates, as well as the distribution of isolates from study locations in different BOX-PCR clusters. For each location, the number of segments indicates the number of BOX-PCR clusters occupied by isolates from that location. Uppercase letters in each segment represent the labels of BOX-PCR clusters as indicated in Fig. 1. The area of each segment is proportional to the number of isolates from a given location occupying that cluster. N, number of different rhizobial isolates. a different BOX-PCR cluster and were also obtained from nodules of different Kersting's groundnut landraces ( Fig. 1 and 4). Although most clusters contained isolates from the same country, isolate TUTMGSA125 from South Africa grouped with other isolates from different locations in Ghana (group VI) (93.3 to 100% sequence similarity) and shared 93.6 to 99.7% sequence similarity with Bradyrhizobium ganzhouense RITF806 T , B. huanghuaihaiense CCBAU 23303 T , B. ingae, and B. iriomotense (Fig. 4). Group VII was the largest and similarly comprised isolates from Ghana and South Africa, which shared close relatedness and had 97.8 to 100% sequence similarity with B. elkanii and B. pachyrhizi PAC48 T (Fig. 4).
Phylogeny based on housekeeping genes atpD, glnII, gyrB, and rpoB. PCR amplification of the atpD, glnII, gyrB, and rpoB genes, which code for ATP synthase subunit beta, glutamine synthetase II, DNA gyrase subunit B, and RNA polymerase subunit B, respectively, produced 600-bp, 650-bp, 600-bp, and 550-bp PCR-amplified products. The sequences of atpD, glnII, gyrB, and rpoB were obtained for 34, 43, 40, and 43 isolates, respectively. The variations in the number of sequences obtained were due to the poor quality of some sequences. The individual gene-based phylogenies were congruent with each other and with the 16S rRNA-based phylogeny, with few excep-  tions, and thus placed the test isolates in distinct clusters within the genus Bradyrhizobium ( Fig. S2 to S5). For example, although isolates TUTMGSA169 and TUTMGSA170 were distantly related in the 16S rRNA phylogeny, the two isolates were tightly grouped in the atpD phylogeny (group VI) and shared a close relationship with isolates TUTMGSA181, TUTMGSA160, and TUTMGSA162 in groups V, V, and IV in the glnII, gyrB, and rpoB phylogenies, respectively ( Fig. 4 and S2 to S5). Similarly, although isolate TUTMGGH77 showed close relatedness to TUTMGSA160, TUTMGSA162, and TUT-MGSA181 in the atpD phylogeny, the isolate consistently shared a closer relationship with B. kavangense in the glnII, gyrB, and rpoB phylogenies, with 100%, 100%, and 99% sequence similarity, respectively, and 99% bootstrap support ( Fig. 4 and S2 to S5).
Phylogenetic analysis of concatenated sequences. Due to variations in the number of sequences obtained for each gene, two concatenated phylogenetic analyses were carried out based on the sequences of atpD-glnII-gyrB-rpoB for 32 isolates (Fig. S6) and glnII-gyrB-rpoB for 40 isolates (Fig. 5). The phylogenies based on both concatenated sequences were congruent and successfully refined the discrepancies in the singlegene phylogenies. In both concatenated trees, the test isolates were placed in 10 distinct phylogenetic clusters (clusters I to X) belonging to the genus Bradyrhizobium ( Fig. 5 and S6). As in the individual gene phylogenies, most of the Kersting's groundnut isolates were highly divergent from the known Bradyrhizobium type strains in the concatenated phylograms. Except for isolates TUTMGSA108, TUTMGSA109, TUTMGSA121, TUTMGSA126, and TUTMGSA179 from South Africa, which grouped with B. pachyrhizi PAC48 T (98.1 to 99.0% sequence similarity), and isolate TUTMGGH77 from Ghana, which also showed close relatedness to B. kavangense 14-3 T (99.0% to 99.7% sequence similarity) in the atpD-glnII-gyrB-rpoB and glnII-gyrB-rpoB phylogenies, the remaining isolates formed distinct clusters with high divergence from the reference type strains ( Fig. 5 and S6). Although isolate TUTMGGH52 grouped with B. vignae 7-2 T in the 16S rRNA, glnII, and rpoB phylogenies, with 99.6%, 99.1%, and 100% sequence similarity, respectively, the isolate shared a closer relationship with B. subterraneum 54 1-1 T in the atpD and gyrB phylogenies, with 96.6% and 96.9% sequence identities, respectively, due to the absence of the B. vignae 7-2 T strain in the gyrB phylogram ( Fig. S2 to S5). Consequently, isolate TUTMGGH52 grouped with B. subterraneum 54 1-1 T in the atpD-glnII-gyrB-rpoB phylogeny (group II), with 96.6% sequence identity and 71% bootstrap support (Fig. S6), and together with isolates TUTMGGH14, TUTMGGH55, TUTMGGH59, and TUTMGGH80 formed group I in the glnII-gyrB-rpoB phylogeny and showed close relatedness to B. subterraneum 54 1-1 T (96.6% sequence identity and 99% bootstrap support) (Fig. 5). Isolates TUTMGGH9, TUTMGGH92, and TUTMGGH94 from Ghana (group IX of both concatenated phylograms) showed a high divergence from the reference type strains but shared sequence similarities of 95.7 to 95.8% with B. pachyrhizi PAC48 T and 96.5 to 96.6% with Bradyrhizobium viridifuturi SEMIA 690 T , the most closely related type strains ( Fig. 5 and S6). Although most isolates from the same country grouped together, they often originated from nodules of different hosts ( Fig. 5 and S6).
Phylogeny based on symbiosis-related (nifH and nodC) genes. PCR amplification of the nifH and nodC genes yielded 800-and 300-bp band sizes. A total of 35 and 28 sequences were obtained for the nifH and nodC genes of the test isolates, respectively. The nifH-and nodC-based phylogenies were congruent with the concatenated gene phylogenies, albeit with few discrepancies (Fig. 5, 6, S6, and S7). For example, although isolates TUTMGGH14, TUTMGGH52, TUTMGGH55, TUTMGGH59, and TUTMGGH80 from Ghana grouped together in all single-gene phylogenies (but were absent in the atpD phylogeny, except for TUTMGGH52) and in the concatenated glnII-gyrB-rpoB phylogeny, isolate TUTMGGH80 diverged from the other isolates in the nifH phylogeny ( Fig.  S7) and from isolate TUTMGGH14 in the nodC phylogeny (Fig. 6). As found with the individual housekeeping gene phylogenies, isolate TUTMGGH14 clustered with TUTMGGH52, TUTMGGH55, and TUTMGGH59 as B. vignae in group II of the nifH  (Fig. S7), but it also clustered with isolates TUTMGGH77 and TUTMGSA125 from Ghana and South Africa, respectively, in the nodC phylogeny (group III), with 81.7 to 96.7% sequence similarity and 58% bootstrap support (Fig. 6). Furthermore, though isolate TUTMGSA123 belonged to the B. japonicum group in all housekeeping gene and nifH gene phylogenies, it clustered with other isolates from Ghana and South Africa from the B. elkanii group in the nodC phylogeny (90.3 to 100% sequence similarity and 95% bootstrap support in group V) and together with those isolates shared 95.6 to 98.9% sequence similarity with Bradyrhizobium embrapense SEMIA 6208 T , Bradyrhizobium tropiciagri SEMIA 6148 T , and Bradyrhizobium viridifuturi SEMIA 690 T , members of B. viridifuturi sv. tropici in that phylogram (Fig. 6). The remaining isolates formed distinct clusters in the nodC phylogeny and showed high divergence from all the known reference type strains or symbiovars (Fig. 6). As with the concatenated trees, all the phylogenetic groups in the nodC and nifH phylograms contained isolates from different hosts ( Fig. 6 and S7).
Symbiotic efficiency and photosynthetic functioning induced by isolates. A one-way analysis of variance (ANOVA) revealed variable nodulation (nodule number and nodule dry matter [DM]), shoot dry matter accumulation (shoot DM), photosynthetic rates (A), stomatal conductance (gs), plant transpiration (E), leaf chlorophyll content, and relative symbiotic efficiency among representative isolates used in the concatenated glnII-gyrB-rpoB phylogenetic analysis ( Table 1). The isolates from Mozambique which formed two separate clusters (groups VI and VII) in the glnII-gyrB-rpoB phylogram, induced more nodules than did the commercial Bradyrhizobium sp. strain CB756 and most other isolates from Ghana and South Africa (Table 1). Isolate TUTMG-MZQ206 elicited the highest nodulation (348 Ϯ 23.2 nodules per plant) in Kersting's groundnut, followed by TUTMGMZQ190, TUTMGMZQ194, TUTMGMZQ198, and TUT-MGMZQ187 (303 Ϯ 35.4, 192 Ϯ 19.1, 185 Ϯ 3.3, and 168 Ϯ 15.5 nodules per plant, respectively). Other isolates which also induced higher nodulation than the commercial Bradyrhizobium sp. strain CB756 included TUTMGGH52 and TUTMGGH53 from Ghana, TUTMGSA109, TUTMGSA126, TUTMGSA137, TUTMGSA121, and TUTMGSA160 from South Africa, as well as isolate TUTMGMZQ195 from Mozambique (Table 1). In this study, 56% of the test isolates induced nodule numbers similar to that of the commercial strain Bradyrhizobium sp. CB756. Increased nodule numbers were generally accompanied by high nodule DM, albeit with few discrepancies due to observed variations in nodule size ( Table 1). As a result, there was a significant positive correlation between nodule number and nodule DM induced by the test isolates (r ϭ 0.67, P Ͻ 0.001, r 2 ϭ 0.45). In general, the isolates which elicited greater nodulation also caused an increase in leaf chlorophyll concentrations, stomatal conductance, transpiration rate, photosynthetic rate, and shoot DM (Table 1). These observations were supported by significantly positive correlations obtained when nodule number and nodule DM were each plotted against stomatal conductance, transpiration, photosynthesis, chlorophyll concentration, and shoot DM (Fig. 7). Moreover, there were significantly positive correlations between shoot DM and stomatal conductance (r ϭ 0.16, P Ͻ 0.001, r 2 ϭ 0.03), leaf chlorophyll concentration (r ϭ 0.33, P Ͻ 0.001, r 2 ϭ 0.11), and photosynthetic rates (r ϭ 0.33, P Ͻ 0.001, r 2 ϭ 0.11) induced by the test isolates.
Of the isolates tested, TUTMGGH50, TUTMGSA121, TUTMGSA126, TUTMGMZQ190, and TUTMGMZQ206 elicited significantly higher relative efficiency (202 to 268%) in terms of shoot biomass accumulation than did the commercial strain CB756 (relative efficiency, 140%). However, except for isolates TUTMGGH11, TUTMGGH59, TUTMGGH77, and TUTMGGH86, which elicited relatively lower nodulation leading to lower efficiency (83.1 to 95.1% relative efficiency), the remaining isolates together with the commercial Bradyrhizobium sp. strain CB756 showed high N 2 -fixing efficiency regardless of statis-  tical significance, with relative efficiencies ranging from 100% to 268% (Table 1). In this study, even the relatively lower-performing isolates induced much higher leaf chlorophyll levels than those fed with 5 mM nitrate (Table 1). Of the relatively low-performing isolates, TUTMGGH11 and TUTMGGH86 from Ghana shared close relatedness in the  glnII-gyrB-rpoB phylogeny, while the remaining isolates were separately grouped in the same phylogram (Fig. 5).

DISCUSSION
Kersting's groundnut is nodulated by genetically diverse bradyrhizobia in African soils. The identification of diverse rhizobial populations in soils offers an opportunity for the selection of efficient strains for inoculant formulation (13). Moreover, the potential of legume nitrogen fixation for increased crop yields can be enhanced by exploring new regions for novel rhizobia and legume germplasm in support of agriculture from the discovery of elite microsymbionts (24). This study therefore explored the colony characteristics, genetic diversity, and phylogenetic positions of Kersting's groundnut isolates obtained from diverse agroecologies in Ghana, South Africa, and Mozambique.
In this study, the majority of isolates induced crown nodulation on their homologous Kersting's groundnut host in the glasshouse, an observation that is usually attributed to low movement of rhizobia in the soil profile (25,26). The test isolates varied in the number of days to colony appearance and other morphological traits (Table S1), which suggests some diversity. The dendrogram constructed based on BOX-PCR profiles grouped the isolates into 23 major clusters, indicating the presence of genetically diverse rhizobial populations responsible for Kersting's groundnut nodulation in the studied environments. As with Kersting's groundnut isolates in this study, high genetic variability has been reported among rhizobial symbionts of several grain legumes, including soybean (27), cowpea (15,16,28), Bambara groundnut (12), groundnut (29), and the common bean (17) in African soils, which represents an important biological pool that can be exploited for enhanced symbiosis and increased grain yield through inoculant formulation.
The distribution of Kersting's groundnut isolates in this study was markedly influenced by the geographic origins of the isolates. The observed tendency for most isolates to show close relatedness in the dendrogram generated from BOX-PCR profiles ( Fig. 1) based on common geographic origin suggests the presence of country-specific Bradyrhizobium strains responsible for Kersting's groundnut nodulation in the wider geographic locations studied. The influence of geographic location on the distribution of rhizobia has been previously reported for cowpea isolates from Senegal and Ghana (15,18,28). In this study, the isolates from Nelspruit in South Africa were the most diverse, as shown by their presence in 13 out of the 23 major BOX-PCR clusters identified; this is in agreement with earlier studies which also revealed greater diversity among South African bradyrhizobial populations nodulating cowpea and Bambara groundnut than in isolates from Ghana or Botswana (12,16). Interestingly, although the isolates from Nyankpala sites 1 and 2 in Ghana each occupied seven BOX-PCR clusters, none of these isolates from the two sites occupied the same cluster despite the proximity of their geographic origins (Table S1 and Fig. 2), suggesting that rhizobial diversity can vary greatly between different sites and niches within a geographic location. In this study, the Kersting's groundnut landraces had no influence on the clustering of the isolates, as most BOX-PCR clusters comprised isolates obtained from nodules of different landraces (Table S1). However, whether the observed higher genetic diversity observed among the nodule occupants of the dark-seeded Kersting's groundnut landraces in this study was directly or indirectly linked to their greater phenolic compound compositions when compared to that of the white-seeded Boli landrace (30) remains to be explored. Earlier studies have, however, reported a greater abundance of phenolic compounds (involved in signal exchange during the legumerhizobium symbiosis) in legume seeds with dark seedcoat pigmentation than in those with lighter seedcoat color (31,32).
Phylogeny of Kersting's groundnut rhizobial symbionts native to African soils. Multilocus sequence analysis in this study revealed the presence of diverse Bradyrhizobium species responsible for Kersting's groundnut nodulation across the contrasting environments in Africa. Phylogenetic analysis assigned the test isolates proximally to B. vignae 7-2 T , B. subterraneum 58 2-1 T , B. kavangense 14-3 T , B. liaoningense 2281 (USDA 3622) T , Bradyrhizobium yuanmingense LMG 21827 T , B. huanghuaihaiense CCBAU 23303 T , B. pachyrhizi PAC48 T , and the type strain of B. elkanii. However, most of the test isolates were highly divergent from the reference type strains and could represent novel Bradyrhizobium species yet to be described. The observed incongruences between the phylogenies of 16S rRNA and protein-coding genes could be attributed to horizontal gene transfer events and/or genetic rearrangement in those genes during evolution (19). However, the fact that isolates TUTMGGG14, TUTMGGH52, TUTMGGH55, TUTMGGH59, and TUTMGGH80 shared a closer relationship with B. vignae 7-2 T in all the phylogenies, with 99.1 to 99.6% sequence similarity, suggests that B. vignae 7-2 T , a strain originally isolated from root nodules of cowpea and Bambara groundnut in African soils (12,13), has a broad host range in its nodulation ability. Furthermore, the observed divergence of isolates TUTMGSA160, TUTMGSA162, and TUTMGSA181 from B. kavangense 14-3 T in the concatenated phylograms (sharing only 94.5 to 95.5% sequence similarity) despite their closeness with the type strain in the 16S rRNA phylogeny (99.7 to 99.8% sequence similarity) is consistent with earlier reports which indicate that high 16S rRNA similarity alone may not yield decisive species delineation (33). According to Helene et al. (33), Ն97% sequence similarity in protein-coding genes can be used as a cutoff for species delineation in bacterial phylogeny. However, except for a few isolates which showed close relatedness to B. subterraneum and B. pachyrhizi in this study, the majority of the test isolates were distantly related to the reference type strains and could represent novel Bradyrhizobium symbionts of Kersting's groundnut in African soils (Fig. 5 and S6). The presence of high numbers of potentially novel Bradyrhizobium species responsible for Kersting's groundnut nodulation was expected since this study provides a detailed report of the crop's native symbionts in its cultivation areas (Ghana) and beyond (South Africa and Mozambique).
The phylogenies of nifH and nodC genes were congruent with each other, as well as with the phylogenies based on individual and concatenated gene sequences, albeit with a few inconsistencies. The incongruences between phylogenies based on symbiotic genes and housekeeping gene loci for some isolates (e.g., TUTMGGH59 and TUTMGGH80) indicate possible horizontal gene transfer and/or genetic recombination events, thus agreeing with earlier reports that symbiosis-related genes such as nifH and nodC are prone to transfers between bacterial strains (19). However, the observed congruency between phylogenies of symbiotic and housekeeping genes for some isolates in this study could be attributed to common evolutionary histories and/or vertical transfer of those genes to the isolates. Except for isolates in group V of the nodC phylogeny, which shared a closer relationship with B. tropiciagri, B. embrapense, and B. viridifuturi sv. tropici, most isolates in the phylogram (groups I to III) were highly divergent from all reference type strains, indicating that they could belong to novel symbiovars within the Bradyrhizobium genus. Although members of B. viridifuturi sv. tropici were earlier isolated from nodules of forage legumes in South America (34,35), a recent study also reported the presence of the symbiovar in root nodules of cowpea grown in African soils (15). Moreover, although isolates in group IV clustered separately in the nodC phylogeny, they shared 93.5 to 95.1% sequence similarity with Bradyrhizobium shewense, a strain isolated from Erythrina brucei but with a nodC gene similar to that of Bradyrhizobium arachidis isolated from Arachis hypogea (36). It would therefore seem that, like cowpea (another legume of African origin), the Kersting's groundnut is a promiscuous host which is capable of nodulating with diverse symbiovars within the Bradyrhizobium genus. The occurrence of bradyrhizobial isolates from different host landraces in the same phylogenetic clusters of the nodC and nifH genes suggests that host-symbiont compatibility may be conserved within the species.
Influence of soil properties on bradyrhizobial distribution. The distribution of the bradyrhizobial isolates in this study was markedly influenced by the soil chemical properties of the test locations. The South African isolates, which shared close relationships with B. elkanii in the phylogenetic analysis, were more influenced by the concentrations of P and Na in the soil. The effect of soil P concentration on the distribution of B. elkanii was previously reported by Zhang et al. (37). In another report, the concentration of soil P was also found to alter the diversity of rhizobia in soils (38). In this study, soil K content had a marginal effect on the distribution of isolates from Nyankpala site 1, which were closely related to B. yuanmingense, a finding consistent with earlier reports on the effect of soil K on the distribution of B. yuanmingense in Chinese soils (39). Recently, Puozaa et al. (12) also found a marked influence of soil mineral nutrients on bradyrhizobial distribution in African soils.
Symbiotic efficiency and photosynthetic functionality induced by native rhizobial isolates. Although knowledge of rhizobial biodiversity has contributed to our current understanding of the factors responsible for their survival and distribution, it is their N 2 -fixing efficiency that determines their usefulness for inoculant production aimed at improving the legume-rhizobium symbiosis for increased crop yields (21). In this study, phylogenetically distinct bradyrhizobia responsible for Kersting's groundnut nodulation were found to show markedly higher symbiotic efficiency than the commercial strain Bradyrhizobium sp. CB756 (Table 1). The presence of these potentially novel and high-N 2 -fixing Kersting's groundnut bradyrhizobia could explain why the crop has survived cultivation in nutrient-depleted soils across West Africa, where landraces were found to exhibit varied responses to Bradyrhizobium inoculation (3, 7). As the symbiotic process is sustained by a mutual exchange of N compounds from the rhizobial microsymbiont and photosynthate from the legume host, increased nodulation by test isolates was found to markedly stimulate higher photosynthesis, stomatal conductance, and leaf chlorophyll concentration; consequently, there was increased accumulation of shoot biomass over nitrate feeding. As a result, there were significant positive correlations when nodule number and nodule DM were each plotted against leaf photosynthetic rates, stomatal conductance, plant transpiration, leaf chlorophyll concentration, and shoot biomass (Fig. 7). Thus, increased N 2 fixation from effective nodulation enhanced RuBisCO and chlorophyll biosynthesis, stimulated greater stomatal conductance, and increased photosynthetic rates, which resulted in better plant growth and biomass accumulation than 5 mM nitrate feeding. Even the relatively low-performing rhizobial isolates induced greater photosynthetic rates and leaf chlorophyll levels, which resulted in 83% to 95% relative efficiency. This finding is consistent with other reports on the superiority of symbiotic N over nitrate feeding in supporting the growth of soybeans (40). Despite the relatively profuse nodulation elicited by isolates from Mozambique which formed distinct clusters in the concatenated trees (Fig. 5), their relative efficiencies were matched by several isolates from Ghana and South Africa which induced moderate nodulation on the homologous host (Table 2), thus agreeing with previous reports that the extent of nodulation is not the sole a Nyankpala site 1, UDS fields; Nyankpala site 2, CSIR-SARI research field. F, rhizobia were trapped by direct planting of Kersting's bean in the fields; G, rhizobia were trapped in the glasshouse using soils collected from the field. For each study location, soils were randomly sampled from several points on the field and bulked, and subsamples were used for single determination of soil chemical properties. b ND, not determined.
determinant of symbiotic efficiency (41). Moreover, the presence of highly efficient native rhizobial symbionts of Kersting's groundnut in the studied countries suggests their wider distribution beyond its current cultivation areas.
In conclusion, this study provides a detailed insight into the microsymbiont diversity, biogeographic distribution, and phylogenetic position of rhizobia associated with Kersting's groundnut nodulation in different African countries. The study further explored the symbiotic efficiency and photosynthetic function elicited by the test bradyrhizobial isolates on homologous Kersting's groundnut. The results revealed a marked influence of geographic origin and soil mineral nutrient concentrations on the distribution and diversity of potentially novel bradyrhizobial isolates. The observed genetic divergence of most Kersting's groundnut isolates from Bradyrhizobium reference strains in this study suggests that Africa could be a hot spot of bradyrhizobial biodiversity, a finding that compels further studies to be conducted to unravel other novel symbionts responsible for legume nodulation across the continent. Due to their high symbiotic efficiency, the indigenous Kersting's groundnut isolates in this study represent a useful biological resource that could be tapped for bioinoculant formulations for increased crop yields.

MATERIALS AND METHODS
Origin of Kersting's groundnut nodules used. The Kersting's groundnut nodules used in this study were harvested from plants grown at six sites in the Northern Region of Ghana (Nyankpala site 1, Nyankpala site 2, Savelugu, Gbalahi, Sognaayili, and Damongo in the Guinea savanna agroecology), two locations in South Africa (Nelspruit and Klipplaatdrift in the lowveld and middleveld areas of Mpumalanga Province, respectively), and one location (Muriaze) in Mozambique (Table 2 and Fig. 2). The bulk soils from each field were collected, pooled, and analyzed to obtain information on their chemical properties ( Table 2). The nodules were either collected from field-grown plants (F) or trapped in the glasshouse (G) using soils collected from the field ( ]) were used to trap rhizobia under field conditions. For this, the seeds of each landrace were surface sterilized in 95% ethanol (3 to 5 minutes), followed by immersion in NaOCl (commercial bleach) for 3 minutes. The seeds were then rinsed 5 times with sterile distilled water by changing the water each time. The sterilized seeds were sown in 3-m by 2-m plots with three replicate plots for each Kersting's groundnut landrace at Nyankpala 1, Sognaayili, Damongo, Klipplaatdrift, Nelspruit, and Muriaze. Weeds were controlled using a hand hoe. At flowering (55 days after planting), 5 plants were dug out from each plot, separated into shoots and nodulated roots, placed in prelabeled brown paper bags, and transported to the laboratory, where the roots were gently washed in running tap water to remove soil and adhering debris. Nodules were carefully observed and detached with small root segments, dehydrated on silica gel, and stored for bacterial isolation.
Trapping rhizobia in the glasshouse using soils from the field. Soils were collected from locations (namely, Nyankpala site 1, Savelugu, and Gbalahi) where field planting was not undertaken and used to trap rhizobia under glasshouse conditions using three phenotypically diverse Kersting's groundnut landraces (Puffeun, Boli, and Funsi) as hosts. For each landrace, two surface-sterilized seeds were planted in triplicate pots containing sterile (autoclaved) sand. Soil inocula were prepared by suspending 20 g of each soil sample in 1,000 ml of sterile distilled water. The suspension from each location was used to inoculate pots containing seedlings of the different landraces (15). The plants were fed with sterile (autoclaved) N-free nutrient solution (42) and, when necessary, irrigated with sterile distilled water. At flowering, the plants were uprooted and the roots washed with tap water to remove sand and adhering debris. Nodulation was observed, after which the nodules were removed from roots and dehydrated on silica gel for later isolation of root nodule bacteria.
Isolation of bacteria from root nodules. Bacterial isolation from root nodules of Kersting's groundnut was done using the procedure described by Somasegaran and Hoben (43). Healthy and functional nodules (10 to 15 nodules per landrace) were selected from each source for bacterial isolation using yeast mannitol agar (YMA) medium (43). Isolated single colonies were selected from each plate and restreaked onto YMA plates to obtain pure colonies for further characterization. Morphological characteristics, such as colony color, shape, size/diameter, and opacity, were recorded for each isolate. The isolates were named using the prefix for the Tshwane University of Technology ( Authentication of bacterial isolates in the glasshouse. To fulfill Koch's postulates, single-colony cultures were assessed for their ability to induce root nodules on their homologous Kersting's groundnut host (using the landrace Funsi) (44). Two surface-sterilized seeds (44) of Kersting's groundnut were planted in sterile (autoclaved) sand contained in sterile pots. The seedlings were thinned to one plant per pot after germination and grown in a naturally lit glasshouse at an average temperature of 28°C. There were four replicate pots for each isolate. Seven-day-old seedlings were inoculated with 1 ml YMA broth suspension of bacterial culture grown to exponential phase (10 6 to 10 7 cells/ml). Uninoculated plants, plants fed 5 mM NO 3 Ϫ1 , and plants inoculated with the commercial strain Bradyrhizobium sp. CB756 were used as controls. The inoculated seedlings were supplied with sterile N-free nutrient solution (42) and sterile distilled water in alternation. The plants were harvested at 60 days after planting (DAP) and assessed for nodulation.
Gas-exchange studies and symbiotic efficiency/effectiveness. The 40 bacterial isolates selected for multilocus sequence analysis (MLSA) were further evaluated for their N 2 -fixing efficiency. For this, gas-exchange studies were carried out to measure photosynthetic rates (A), stomatal conductance (gs), and leaf transpiration rates (E) on young and fully expanded trifoliate leaves of each replicate plant at 60 DAP using a portable infrared red gas analyzer, version 6.2 (LI-6400XT; LI-COR, Lincoln, NE, USA). The chamber conditions used included photosynthetic flux density of 1,000 mol m Ϫ2 s Ϫ1 , reference CO 2 concentration of 400 mol mol Ϫ1 , and flow rate of 500 mol s Ϫ1 . Gas-exchange measurements were carried out between 8:30 a.m. and 12:30 p.m. Leaf chlorophyll concentrations were determined using the same leaves used for gas-exchange measurements. Briefly, chlorophyll was extracted from 6 leaf discs (each with an area of 0.786 cm 2 and weighing~18.7 mg) using preheated (65°C) dimethyl sulfoxide for 30 min (45). The absorbance of the extracts at 645 nm and 663 nm was then measured on a Jenway 7300 spectrophotometer. The total chlorophyll concentration was calculated using the equations described by Richardson et al. (45). The plants were then uprooted and assessed for nodulation (nodule number and nodule dry matter) and shoot dry matter (shoot DM) after oven-drying at 60°C for 72 h. The relative efficiency (RE) of isolates was calculated as the shoot biomass of inoculated plants expressed as a percentage of the shoot biomass of the 5 mM KNO 3 -fed Kersting's groundnut plants (21).
Extraction of bacterial genomic DNA and BOX-PCR fingerprinting. The genomic DNA of authenticated rhizobial isolates was extracted using GenElute bacterial genomic DNA kits (Sigma-Aldrich, USA), following the manufacturer's instructions. DNA integrity was checked on a 0.8% agarose gel stained with ethidium bromide. The bacterial genomic DNA was subjected to repetitive extragenic palindromic-PCR (REP-PCR [BOX-PCR]) using the BOX-A1R primer (46). The final volume of the PCR mixture was 25 l, and it contained 1 l (50 to 70 ng/l) genomic DNA, 12.5 l MyTaq PCR mastermix (2ϫ) (Bioline USA), 1 l (10 M) BOX-A1R primer, and 10.5 ml sterile ultrapure H 2 O. PCR amplification was carried out in a thermal cycler (T100; Bio-Rad, USA) using standard temperature profiles, and the PCR products were subjected to gel electrophoresis (15). Cluster analysis was carried out with the unweighted pair group method with arithmetic mean (UPGMA) algorithm using the trial version of the software BioNumerics 7.6, with kind permission from BioNumerics.
PCR amplification of the 16S rRNA, protein-coding, and symbiotic genes. The PCR amplifications of 16S rRNA, protein-coding (atpD, glnII, gyrB, and rpoB), and symbiotic (nifH and nodC) genes were each carried out in a 25-l reaction volume containing 1 l DNA (50 to 70 ng/l), 3 l of MyTaq buffer (5ϫ), 1 l (10 M) each of forward and reverse primers of the gene of interest, 0.1 l Taq polymerase (5 U; Bioline, USA), and 18.9 l double-distilled ultrapure water using standard temperature profiles (15). The amplified gene products were confirmed by gel electrophoresis on a 1.2% agarose gel.
Sequencing and phylogenetic analysis. The amplified PCR products were purified using a PCR Cleanup kit (NEB, USA), following the manufacturer's instructions, and sent to Macrogen (The Netherlands) for sequencing. The quality of sequences was verified using the software BioEdit 7.0.9.0 (47). The BLASTn program was used to identify closely related species in the NCBI database. Pairwise and multiple-sequence alignments were carried out with CLUSTALW, and phylogenetic trees were constructed by means of the maximum likelihood statistical method using MEGA7 software (48). The evolutionary history was inferred by means of the maximum likelihood method based on the Kimura 2-parameter model (49). The robustness of branching was estimated using 1,000 bootstrap replicates (50).
Statistical analysis. The influence of soil chemical properties on the distribution of Kersting's groundnut rhizobial isolates was explored using canonical correspondence analysis (CCA) with vegan (version 2.4-2) (51) in the R software (52). We explored the soil chemical properties of the test locations which influenced the distribution of the Kersting's groundnut isolates. The general permutation test was used to assess the statistical significance of the ordination axes. The graph analysis is presented only for the soil factors that showed significant contributions. Quantitative data, including nodule number, nodule dry matter, shoot dry matter, photosynthetic rate (A), stomatal conductance (gs), leaf transpiration (E), and total chlorophyll concentration, were subjected to a 1-way ANOVA using the STATISTICA software version 10.0 (StatSoft). The Duncan multiple-range test was employed to separate the means that showed significant differences at a P value of Յ0.05.
Data availability. The sequences obtained were deposited in the NCBI GenBank database with the accession numbers MK183839 to MK183879 for the 16S rRNA gene, MK183880 to MK183913 for atpD, MK183914 to MK183953 for gyrB, MK183954 to MK183996 for rpoB, MK184040 to MK184078 for nifH, MK184079 to MK184106 for nodC, and MK184107 to MK184149 for glnII.

SUPPLEMENTAL MATERIAL
Supplemental material for this article may be found at https://doi.org/10.1128/AEM .00342-19.