Identification and distribution of microsymbionts associated with soybean nodulation in Mozambican soils

Indigenous soybean rhizobial strains were isolated from root nodules sampled from farmers’ fields in Mozambique to determine their identity, distribution and symbiotic relationships. Plant infection assays revealed variable nodulation and symbiotic effectiveness among the 43 bacterial isolates tested. Strains from Ruace generally promoted greater whole-plant growth than the others. 16S rRNA-RFLP analysis of genomic DNA extracted from the rhizobial isolates produced different banding patterns, a clear indication of high bacterial diversity. However, the multilocus sequence analysis (MLSA) data showed alignment of the isolates with B. elkanii species. The 16S rRNA sequences of representative soybean isolates selected from each 16S rRNA-RFLP cluster showed their relatedness to B. elkanii, as well as to other Bradyrhizobium species. But a concatenated phylogeny of two housekeeping genes (glnII and gyrB) identified the soybean nodulating isolates as Bradyrhizobium, with very close relatedness to B. elkanii. The nifH and nodC sequences also showed that the majority of the test soybean isolates were closely related to B. elkanii, albeit the inconsistency with some isolates. Taken together, these findings suggest that the B. elkanii group are the preferred dominant microsymbiont of soybean grown in Mozambican soils. Furthermore, the distribution of soybean rhizobia in the agricultural soils of Mozambique was found to be markedly influenced by soil pH, followed by the concentrations of plant-available P and Mn. This study suggested that the identified isolates TUTMJM5, TUTMIITA5A and TUTLBC2B can be used as inoculants for increased soybean production in Mozambique.


Introduction
Interest in soybean cultivation in Africa is increasing due to its high protein content of grain for human consumption and for use as animal feed. In Mozambique, there is an increasing demand for soybean, driven largely by the poultry industry and for exportation. This has led to many farmers switching to soybean production [42], as it offers them an income opportunity for their livelihood [19]. Soybean cultivation in Mozambique is dominated by smallholder farmers who use little or no inputs. The inaccessibility of rhizobial inoculants is a major factor contributing to low soybean yields, which were about 450 kg/ha in 2004, and 1100 kg/ha in 2012 [2,12]. The increase was due to the application of inputs such as bacterial have been reported to effectively establish N 2 -fixing symbiosis with soybean [22,46,48]. Recently, Gyogluu et al. [14,15] found a marked effect of Bradyrhizobium inoculation on plant growth, N content, %Ndfa (nitrogen derived from the atmosphere), amount of N fixed and grain yield when compared to uninoculated plants. However, little is known about the diversity, distribution and symbiotic effectiveness of native soybean rhizobia of Mozambique. Also, little is known of the effect of environmental factors on the distribution of these indigenous soybean rhizobia. In this study, native soybean rhizobia were isolated from farms without any history of inoculation, and characterized in terms of phylogeny, distribution and symbiotic effectiveness with the ultimate aim of developing inoculants for increased soybean yields in Mozambique.

Soil analysis, bacterial isolation and characterization
Soybean nodules were sampled from TGx and non-TGx soybean varieties grown in smallholder farmers' fields in different provinces of Mozambique (Table 1). Healthy green soybean plants were dug out using a shovel and carried in paper bags to the laboratory. The rhizosphere soils were stored at 4 • C prior to nutrient analysis at the Department of Agriculture, Soil, Water and Plant Tissue Laboratory, Institute of Plant Production, Western Cape, South Africa. The climatic information and soil nutrient analysis of the sampling sites are shown in Table 1. Nodules harvested from each field, were surface-sterilized by rinsing in 95% ethanol, submerged in 3% NaOCl (commercial bleach) for 30 s, and rinsed six times with distilled water [47]. The nodules were aseptically dissected, the central portion squashed, and the nodule suspension streaked onto yeast mannitol agar (YMA) plates, and incubated at 28 • C. The plates were observed for bacterial colonies up to 20 d.

Authentication and measuring strain symbiotic efficiency
Seeds of soybean variety TGx1908-8F were surface-sterilized with 2% sodium hypochlorite (NaOCl) for 5 min, 70% alcohol for 1 min and thoroughly rinsed five times with sterile water. The seeds were aseptically planted in sterilised plastic pots containing sterile sand inside a laminar flow chamber. To prepare isolate cultures for inoculating the sterile-grown seedlings, one loopful of the bacterial culture was transferred to 6 ml yeast mannitol broth, incubated at 28 • C on a rotary shaker (65 cycles per minute) for seven days. Five days after planting (DAP), the seedlings were inoculated with 2 ml suspension of the bacterial culture (∼ 10 7 -10 8 cells/ml) [43]. The pots with seedlings were transferred to the glasshouse. Three replicate plants were grown per rhizobial isolate, and Dilworth's N-free nutrient solution was used to irrigate the plants [6]. Uninoculated and 5 mM NO 3 − -fed plants were included as controls. The plants were harvested at 45 DAP for assessing growth and nodulation. The authenticated (nodule-forming) isolates were tested for acid and alkali production using bromothymol blue (25 g/ml), as well as for bacterial colony size, shape, texture and number of days to grow and were further characterized genetically.
Genomic DNA extraction and PCR-amplification of 16S rRNA gene fragment Bacterial genomic DNA was extracted from authenticated rhizobial isolates using GenElut bacterial DNA isolation Kit (Sigma-Aldrich, USA) according to the manufacturer's instructions. PCR amplification of the 16S rRNA of bacterial genome was carried out with 1 l (40-60 ng/ul) DNA in a 25 l reaction volume with 3 l (5X) My Taq PCR buffer containing MgCl 2 and dNTPs, 5U Taq polymerase (Bioline, USA), 10 pM of each primer, and doubledistilled sterile water in a Thermal cycler (T100 Bio-Rad, USA). The primers and temperature profiles are described in Table S1. The size of the amplified products was estimated by gel electrophoresis on a 1.5% agarose gel stained with 1 g ml −1 ethidium bromide using a standard DNA marker (GeneDirex, 1kbp and 100 bp ladder), and photographed using a gel documentation system (GeldocTm XR+, Bio-RAD, USA).

Restriction Fragment Length Polymorphism (RFLP) analysis of the PCR-amplified 16S rRNA gene
The amplified 16S rRNA-PCR products were digested with three fast digest four base-cutting restriction endonucleases (HpaII, HindIII and MspI) separately following the instructions of the manufacturer (Thermo Scientific Lithuania, EU). The digested DNA fragments were separated by horizontal gel electrophoresis on a 3% agarose gel in Tris acetic acid EDTA (1X TAE) buffer at 85 V for 2.5 h, and photographed.
The RFLP groups were determined using the restriction fragment patterns generated by the different test restriction endonucleases. The relationship between isolates was established using data from the restriction enzymes that differentiated the iso-lates. A binary scoring system (1 for presence, and 0 for absence of a homologous band) was used to generate an input matrix for construction of a dendrogram using the unweighted pair-group method with the arithmetic mean algorithm (UPGMA) by means of NTSYS-Pc 2.1 software [39].
Multilocus (16S rRNA, glnII, gyrB, nifH, and nodC) sequence and phylogenetic analyses Representative isolates were selected from each Cluster based on the 16S rRNA PCR-RFLP analysis for subsequent sequence analysis of 16S rRNA, gyrB, glnII and symbiotic nifH and nodC genes. The PCR amplification of glnII, gyrB, nifH and nodC genes of the rhizobial genome was carried out as described above for 16S rRNA PCR amplification. The primers used and thermal cycle conditions are listed in Table S1. The PCR-amplified products of 16S rRNA, glnII, gyrB and symbiotic nifH and nodC genes were purified by Favour/Prep PCR purification kit (FAVORGEN, Sigma, USA), and sequenced (Macrogen, Netherlands). The ribosomal sequences were screened for chimeric sequences using UCHIME [11]. The quality of all nucleotide sequences was checked using BioEdit 7.0.0 software [16], and BLAST n programme was used in the GenBank database (http://www.ncbi.nlm.nih.gov/) to identify closely related species. Nucleotide sequences were submitted to NCBI GenBank to get accession numbers (Table S2). Reference type sequences were obtained from NCBI GenBank to align with sequences of the test isolates using MUSCLE [10] for the construction of phylogenetic tree by means of MEGA 6.0 program [45]. Phylogenetic trees were generated by Kimura 2-parameter method and evolutionary history was inferred using Neighbor-Joining method with 1000 bootstrap support [13,23,40]. Nucleotide information was recorded from conserved, variable, parsimony-informative, and singleton regions using consensus sequences.

Symbiotic data analysis
The symbiotic data (nodule number, nodule dry weight, plant dry weight) were subjected to a One-Way ANOVA using STATIS-TICA (Sta Soft Inc., Tulsa, OK, USA, 2012) [44] to test for significant differences, and means separated using Duncan's multiple range test (DMRT) at P ≤ 0.05 Table 2 Nodulation and symbiotic efficacy of forty three (43) native soybean bacteria isolates carried out under glasshouse conditions, and sampled at 45 days after planting.

Factors responsible for rhizobial distribution in Mozambican soils
Constrained correspondence analysis (CCA) was performed to determine the correlation between environmental factors and 16S rRNA-RFLP data of the rhizobial communities. The combination of explanatory variables was determined to describe the most influential variables by conducting an ANOVA permutation test with 999 permutations in a reduced model (P < 0.05). The analysis was done by means of R (version 2.15.3) using the package vegan [32].

Isolate characterization and symbiotic efficacy
Seventy pure bacterial isolates from soybean nodules were tested for their ability to form root nodules on soybean (the homologous host) in fulfilment of Koch's postulates. About 61% of the isolates (43 out of 70) induced effective root nodules on the host plant under microbiologically-strict glasshouse conditions. Their effectiveness was evidenced by the pink red internal colouration of the nodules, as well as the dark green colour of leaves when compared to the whitish internal colour of ineffective nodules and the yellow leaves of plants inoculated with ineffective isolates. Symbiotic response of the 43 tested isolates with soybean variety TGx1908-8F revealed significantly marked differences in nodule number (P ≤ 0.001), nodule dry weight (P ≤ 0.01) and plant dry weight (P ≤ 0.001). The number of nodules formed per plant ranged from 2 for isolates TUTMJM6A and TUTMCJ7B, to 52 for TUTMFJ4. Isolates TUTM19373A, TUTMJM5, TUTMI-ITA5A, TUTLBC2B, TUTRAH5B2, TUTRJN3A1, TUTMIITA5A1 and TUTMCJ5B1 were highly effective, and recorded greater nodule number and plant biomass relative to NO 3 -fed plants ( Table 2).
The 43 effective isolates obtained after authentication varied in shape, with 79% forming round colonies and 21% dome-shaped colonies. The majority of the isolates had a colony diameter of 2-3 mm. About 80% of the isolates were opaque, and 20% translucent. One isolate was gummy, while the remaining were mostly smooth and non-gummy. A bromothymol blue (BTB) test showed that 80% of the isolates tested were alkali-producing (turning BTB blue), while 20% were acid-producing (turning BTB yellow) (data not shown).
Restriction Fragment Length Polymorphism (RFLP) analysis of the PCR-amplified 16S rRNA gene PCR amplification using the 16S rRNA primers produced 1500 bp DNA fragments from the genome of the bacterial isolates. Digesting the 16S rRNA fragment with three different restriction enzymes (i.e. HaeIII, HpaII and MspI) resulted in clear polymorphic RFLP band patterns. The RFLP patterns produced by each of the restriction enzymes are shown in Table S3.
Digestion of 16S rRNA genomic region of the soybean isolate with restriction endonucleases showed MspI and HpaII had the most polymorphic (18 patterns) activity compared to HaeIII (14 patterns). The RFLP banding patterns generated from the three restriction endonucleases were used to construct a dendrogram by applying the UPGMA method. The analysis resulted in three major clusters (Fig. 1). Cluster I comprised 19 isolates, Cluster II 15 isolates, and Cluster III had 8 isolates. A total of 32 combined 16S rRNA-RFLP patterns were obtained. All the isolates grouped together with a 0.0-1.0 Jaccard's similarity co-efficient. One isolate (TUTMCJ7B) did not cluster with any of the other test isolates and showed 0.0 Jaccard's similarity co-efficient relative to the others. Cluster I, which comprised 19 isolates, grouped together with a similarity co-efficient of 0.10, while with Clusters II and III each was grouped with a similarity co-efficient of 0.01 and 0.02, respectively. Clustering of isolates was not based on location and soybean varieties (Fig. 1).
PCR-amplification of glnII and gyrB housekeeping genes resulted in 700 bp single band of each isolate. The glnII and gyrB housekeeping genes respectively encode glutamine synthetase II, and DNA gyrase B-subunit. The sequence analysis of these housekeeping genes showed that the highest conserved (62.70%) and parsimonyinformative (26.46%) was found in glnII consensus sequences while gyrB sequences had the highest variable (63.49%) and singleton (18.15%) positions (Table S4).
The phylogenetic analysis of both glnII and gyrB housekeeping genes grouped the test isolates into two clusters (Figs. S1 & S2). The isolates of Cluster I were closely related to B. elkanii with 100 and 99% bootstrap support in their respective phylogenies and high sequence identity of 98.9-100%. Isolate TUTMCJ5B formed an outgroup of Cluster I in the glnII phylogeny and showed 96.8% sequence identity with B. elkanii but grouped in Cluster I in the gyrB phylogeny with 100% sequence identity. Isolate TUTDAIAP3B formed Cluster II and showed proximal relation with B. arachidis with 96% sequence identity while it grouped with TUTDAIAP8B in Cluster II in the gyrB phylogeny and showed 98.5% sequence identity with B. elkanii. The aligned sequences of glnII +gyrB were combined and 918 analysed sites were used to construct a concatenated tree to refine the positions of the test isolates (Fig. 3). The concatenated tree was based on 544 conserved, 368 variable, and 227 parsimony -informative sites (Table S4).
The phylogenetic tree constructed from the concatenated genes grouped the isolates into two major clusters (I-II) with B. elkanii type strain (Fig. 3) with sequence identity of 99.5-99.6%. However, test isolates TUTDAIAP8B and TUTMCJ5B were out-grouped and showed 98.6% sequence identity with B. elkanii. Isolate TUTDA-IAP3B stood alone without any reference type strains and showed proximal relation with B. elkanii with 95.1% sequence identity (Fig. 3) Phylogenetic analysis based on nifH and nodC gene sequences To determine the presence and diversity of the N 2 -fixing genes (nifH and nodC) among the soybean rhizobia, test isolates were used to successfully amplify 600 bp and 300 bp of the nifH and nodC genome regions, respectively, and sequenced. From incompatibility of primer pairs, only 11 isolates were able to yield nodC amplifications. Due to the presence of small length of nifH reference sequences, 274 nucleotide length sequences were used for phylogenetic analysis. The analysed nifH 274 nucleotide sequences comprised 165 conserved, 109 variables and 86 parsimony informative sites (Table S4). All the test isolates in the nifH phylogram grouped into two clusters (Cluster I-II).
Fourteen isolates in Cluster I were closely related to B. elkanii with 99.2% sequence identity and 88% bootstrap support. But isolates TUTDAIAP8B and TUTMCJ5B in Cluster II were closely aligned with B. subterraneum with 99.6 and 98.1% sequence identity, respectively (Fig. 4). The test isolates in the nodC phylogeny were placed in two Clusters. In Cluster I, the isolates were grouped with B. elkanii with 99.4-100% sequence identity. Isolates TUTMI-ITA5A1 and TUTNFM2A1 showed proximal relation with 90.8 and 98.4% sequence identity, respectively, with B. elkanii while isolate TUTDAIAP3B stood alone in the phylogram (Fig. 5).

Factors responsible for rhizobial distribution in Mozambican soils
The physico-chemical properties of soil (Mn, Fe, C, K, P, N and pH) and rainfall data were combined with 16S rRNA-RFLP score to determine the distribution of soybean-nodulating isolates in Mozambique. The results of CCA analysis showed that 70% of the variation were explained, while 30% were unexplained. The CCA analysis also showed that soil pH was the most important environmental variable influencing the distribution of soybean rhizobia (Fig. 6), followed by plant-available P and Mn in soil. According to the angles and lengths of the arrows, it was observed that pH and available P were strongly positively correlated (90%) with the distribution of soybean rhizobia at Mutequelesse (TUTMCJ9A,  TUTMCJ5B, TUTMCJ4B1, TUTMCJ4B, TUTMCJ10B and TUTMCJ7B). However, pH and rainfall were negatively correlated with the distribution of the soybean isolates at Ruace (TUTRAB5B1, TUTNSN3B, TUTRLR3B, TUTRSRH9A, TUTRJN3A1 and TUTRJN5A) and Angonia (TUTDAIAP8B and TUTDAIAP3B, TUTDAIAP1A and TUTDAIAP2A1). The distribution of soybean isolates TUTMFJ3BG (from Mutequelesse), TUTNFM2A1, TUTNFM3B (from Ntengo), TUTM19373A, TUTMIITA5A2, TUTMIITA5A and TUTMIITA5A1 (from Muriase) were moderately influenced by Mn (52%). But soil Fe and C had a low influence on the distribution of the soybean isolates as shown by their short arrows. The large angle between Mn and pH, as well as Mn and P indicates that they had independent effects on the distribution of the soybean isolates.

Discussion
The genetic diversity of native rhizobia nodulating soybean in Mozambique has little been studied. However, the renewed gov- ernment interest in promoting soybean production in Mozambique necessitates the creation of new knowledge for increasing the grain yield of this legume. Understanding the genetic diversity of soybean rhizobia is important for discovering new genera, species and strains for developing new technologies, including the production of inoculants, for increased soybean yield. To explore the genetic diversity of native soybean rhizobia in Mozambican soils, nodules were sampled from farmers' fields that had no previous history of rhizobia inoculation, and analysed genetically.
Although analysis of genomic DNA revealed a high diversity in the 16S rRNA-RFLP results as evidenced by the different banding patterns following digestion with the three enzymes, the MLSA data showed alignment of the isolates with B. elkanii species. These results (RFLP and phylogenies) suggest that, there is an intra-species diversity within the B. elkanii group of native rhizobia nodulating soybean in Mozambique. However, the 16S rRNA sequences grouped the isolates with B. elkanii, B. pachyrhizi, B. tropiciagri and B. ferriligni. This suggests that sequencing the 16S rRNA  gene alone is insufficient for differentiating Bradyrhizobium species [3,9,38], and indicates that rhizobial 16S rRNA gene phylogeny may not always accurately reflect prokaryotic phylogeny. Housekeeping genes are highly conserved among bacteria of the order Rhizobiales, as they encode for important proteins [52] and have therefore been widely used in studies of phylogenetic reconstruction of the order Rhizobiales [27,37]. As a result, two housekeeping genes (gyrB, glnII) were sequenced and analysed to support the 16S rRNA data. The phylogenetic analysis of individual housekeeping genes also suggested that most of the isolates were highly aligned with B. elkanii. However, the phylogenetic positions of isolates TUT-DAIAP3B, TUTDAIAP8B and TUTMCJ5B were not stable in single gene phylogenies, and this could suggest recombination, migration or horizontal gene transfer [8,36].
To fine-tune the individual housekeeping gene analysis, a concatenated phylogeny was done, and the results further confirmed that the isolates nodulating soybean in Mozambique were species of Bradyrhizobium, closely related to B. elkanii. Except isolates TUTMCJ5B and TUTDAIAP8B, there was phylogenetic congruence between the individual and concatenated housekeeping genes for all the isolates. The high similarity of the isolates (99.5-99.6%) with species belonging to the B. elkanii lineage indicates their close identity to B. elkanii [21], which suggests that the B. elkanii group may be the dominant microsymbiont of soybean in Mozambican soils.
Some studies have shown a relationship between soil environment and nodule occupancy. For example, B japonicum and B. elkanii were commonly isolated from acidic and neutral soils, while high pH decreased or completely eliminated their occupancy of root nodules in such soils [26,53]. In this study, the soybean strains obtained were mostly from acidic to neutral soils with pH ranging from 4.4 to 7.0. Previous studies have found B. elkanii to be the dominant microsymbiont nodulating soybean varieties in Kenya [18], Myanmar [41], as well as other parts of Africa [1,20], tropical and subtropical Asia [26], North America [5] and South America [4]. While B. japonicum and B. elkanii species are the known bacterial symbionts of soybean found across many geographic and climatic regions, worldwide B. liaoningense, B. yuanmingense and Bradyrhizobium canariense have been isolated from root nodules under limited climatic conditions [50,51]. Recently, Naamala et al. [29] showed that Bradyrhizobium diazoefficiens and an unknown group of novel Bradyrhizobium species are the dominant bacteria nodulating soybean in South African soils.
The important nodulation and nitrogen fixation genes nodC and nifH, respectively, are situated on interchangeable elements such as symbiotic islands and plasmids, and play an important role in the outcome of the legume-Rhizobium symbiosis. From the results of nifH gene sequence analysis, the majority of soybean isolates were closely related to B. elkanii. However, some of the isolates also had nifH genes that were closely related to B. subterraneum and B. arachidis, thus suggesting that the nifH gene of these isolates (TUT-DAIAP8B and TUTMCJ5B) probably have different origin compared to known bradyrhizobial strains. Even in 16SrRNA and concatenated gene sequence analyses, isolates TUTDAIAP8B and TUTMCJ5B showed low sequence identity with known B. elkanii type strains. The inconsistency of nodC with core gene phylogenies for isolates TUTMIITA5A1, TUTRAH5B2 and TUTNFM2A1 suggested differences in evolutionary history of both chromosomal and symbiotic genes [8].
Glasshouse nodulation assays revealed variable induction of root nodules and symbiotic effectiveness among the bacterial isolates. The nodules formed differed in shape and size, with smaller nodules ranging from 0.1 to 0.2 mm, medium-sized nodules ranging from 0.3 to 0.4 mm in diameter relative to 0.5-0.6 mm in diameter for bigger nodules. Isolates that induced bigger nodules generally produced fewer nodule numbers, while those that induce smaller nodules produce greater nodule numbers. This could explain why some isolates such as TUTM19043A which recorded only 7 nodules, showed nodule dry weight similar to TUTRSRH8B that produced more nodules. Some isolates which recorded higher nodule number (e.g. isolate TUTM19373A), also recorded a higher biomass (see Table 2). By contrast, isolate TUTMJM5 recorded lower nodule number but showed high symbiotic effectiveness comparable to isolate TUTM19373A. This finding indicates that nodule number alone is not a good measure of symbiotic effectiveness of a strain. Isolates TUTM19373A, TUTMJM5, TUTRJN3A1, TUTLBC2B,  TUTMCJ5B1 TUTMIITA5A1, TUTMIITA5A, TUTRAH5B2, TUTMFJ4,  TUTMFJ2B TUTDAIAP2A1, TUTRLR4B2 and TUTNFM3B were highly effective and promoted plant growth that was greater than that of 5 mM NO 3− -fed plants. These isolates occurred in Clusters I and II in the RFLP dendrogram, and representative isolates used for the 16S and nifH analysis clustered together with B. elkanii in both analyses. However, isolates TUTDAIAP8B (which showed poor symbiotic ability) and TUTRAB5B1 (which exhibited moderate effectiveness), occurred in Clusters I and II of the RFLP dendrogram but separated from the remaining isolates that aligned with B.elkanii and showed distant relation with B. ferriligni. However, TUTRAB5B1 had nitrogen-fixing genes similar to B. elkanii, while TUTDAIAP8B had nifH genes closely related to B. subterraneum. Isolates TUTNFM2A1 (moderately effective) and TUTMCJ4B (poorly effective) clustered together (Cluster III) in the 16S analysis, and showed close relatedness with B. elkanii in terms of their nitrogen-fixing genes.
Strains isolated from Ruace soils were particularly very interesting because they generally promoted greater whole-plant biomass. This observation was not surprising as earlier studies on quantification of N 2 fixation at three locations in Mozambique found that soybean plants from Ruace were symbiotically superior to their counterparts from the other locations irrespective of inoculation [14].
In this study, constrained correspondence analysis (CCA) was done in order to better understand the influence of rainfall and soil characteristics on the biogeographic distribution of native rhizobia nodulating soybean in Mozambique. As found in other studies [25,49,53], the distribution of soybean rhizobia was markedly influenced by soil pH. In fact, soil pH has been reported to be a major determinant of the biogeography of Bradyrhizobium species [25,26], followed by plant-available P and Mn in the soil. Elsewhere, soil pH, salinity and phosphate levels have also been identified as important parameters determining the biogeographic distribution of soybean rhizobia [17]. In conclusion, this study has identified B. elkanii as the major native rhizobial species nodulating soybean in Mozambican soils. Strain symbiotic effectiveness studies also suggest that some native rhizobia such as strains TUTM19373A, TUTMJM5, TUT-MIITA5A and TUTLBC2B present in Mozambican soils have the potential for use as inoculants to increase soybean production. The study further revealed that soil pH, P and Mn were the most important factors determining the distribution of soybean rhizobia in Mozambique.