Dominant associations of Ensifer medicae-Medicage polymorpha and Ensifer meliloti-Medicago lupulina in farmland and natural ecosystem

The nitrogen-xing rhizobia associated with Medicago polymorpha and M. lupulina in Yunnan, China have been poorly documented. This study aims to analyze the diversity of rhizobia isolated from these two Medicago species and investigate the impact of abiotic (soil properties) and biotic (plant hosts) factors on Medicago-associated rhizobia in this region. 91 rhizobial isolates were characterized by RFLP of 16S rDNA and 16S–23S IGS, BOX-PCR ngerprinting, nodulation assays and phylogeny analyses based on housekeeping and symbiosis genes. The genetic diversity of the rhizobial isolates was assessed by the BOX AIR pattern and Shannon index. Additionally, the correlation of soil properties and rhizobial distribution was determined by the constrained analysis of principle coordinates (CAP) based on Bray-Curtis distance of presence/absence (PA) transformed species data. All the tested contributed to the native than those of the same species from non-native hosts. The soil edaphic factor analysis elucidated that nitrogen, organic matter as well as Ca 2+ and Na + are the key factors to shape the biogeographical distribution of rhizobia. This study evidenced the microsymbiont preference of M. polymorpha to E. medicae and M. lupulina to E. meliloti, but also revealed the considerable impacts of both plant hosts and soil factors on the rhizobial diversity and biodistribution. rhizobia with and from and Kunming District of A total of 91 strains were isolated and characterized by 16S rDNA sequencing, 16S-23S rDNA IGS PCR-RFLP, BOX-PCR ngerprinting assay and nodulation tests. analyze correlation between samples also collected for physicochemical characterization. The results revealed M. polymorpha Ensifer and lupulina studied region. both host plants and soil nutrients rhizobia. the diversity of advantaged strains as contrast to the inecient Furthermore, in accordance with the another report (Bailly et al. 2007), our results indicated that the soil type (or ecological environment) have effect on rhizobial diversity. Rhizobial strains tested in this study (both Ensifer species strains) from natural ecosystem have higher nucleotide diversity than those from farmland. This suggested that Medicago symbionts might undergo more gene exchange events with other bacterial habitants in the natural ecosystem when compared to that in the farmland. Indeed, a greater value of genetic differentiation was observed in E. meliloti stains than in E. medicae strains that might be a result of ecological niche barrier. E. meliloti was more frequently recovered from nature ecosystem; while E. medicae strains were more readily isolated from M. polymorha in farmland. Selection force by environment and plant host may block rhizobial migration and prevent the subsequent gene exchange. This may also challenge the hypothesis that E. medicae is an evolutionary divergence from E. meliloti population (Biondi et al., 2003). Genome sequence analysis of more rhizobial species is needed to clarify the evolutionary relationship between E. medicae and E. meliloti. Instead, these two Ensifer species retained their own genetic characteristics. We found high levels of gene ow (Nm) within E. medicae or E. meliloti populations in the two habitats. Various subpopulations of the same Ensifer species could intermingle their genes through the horizontal gene transfer processes such as conjugation and transformation, which conferred sucient genetic exchanges among populations. This may be an important evolutionary force to contribute to the genetic diversity. Interestingly, phylogenies of some housekeeping and symbiotic genes suggested that gene transfer from E. kummerowiae, another Ensifer species that can nodulate M. marina and Kummerowia stipulacea (Wei Alías-Villegas., but not M. lupulina.


Introduction
Symbiotic nitrogen-xing system, root and/or stem nodules, established between legume plants and rhizobia is the most signi cant and e cient biological nitrogen xation system, which accounts for one quarter of the global nitrogen xation. Inside the nodules, rhizobia reduce the atmospheric nitrogen to ammonium and provide it to plants as nitrogen nutrient, while the plant partners supply carbohydrates to rhizobia as carbon and energy source. Therefore, both the plant and bacterial symbionts obtain bene ts from symbiosis. As a kind of extensively important forages and pasture plants, the Medicago species can form root nodules with their speci c rhizobia, and the nitrogen xation by the Medicago-associated rhizobia plays an important role in the world-wild pasture production.
Efforts of decades have been focused on the diversity analysis of Medicago-associated rhizobia aiming at screening the rhizobial strains with high nitrogen-xing e ciency. Up to date, rhizobia isolated from the nodules of Medicago species (M. sativa, M. truncatula, M. lupulina, and M. orbicularis) were mainly classi ed into the genera Ensifer and Rhizobium (Hou et Villegas et al., 2006). E. medicae was reported as the microsymbiont for multiple Medicago species in Tunisia and France (Zribi et al., 2004). Although this species shared a common host M. sativa with its sister species E. meliloti, it could be distinguished from E. meliloti by its capacity to nodulate and x nitrogen with M. polymorpha L.  and E. medicae was reported as the unique symbiont of this plant . Although another report showed that E. meliloti could nodulate with M. polymorpha as well (Howieson and Ewing, 1986), only ine cient nodules or root swellings are induced (Alías- Villegas et al., 2015). In addition, some strains isolated from root nodules of Medicago species were identi ed as members of Rhizobium species, including R. mongolense, R.galegae, R. gallium, R. etli and R. leguminosarum (Hou et al., 2009;Sebbane et al., 2006;Van Berkum et al., 1998;Zakhia et al., 2004).
There are 6 annual and 12 perennial Medicago species in China (Fang and Li, 2019). Among these plants, species M. polymorpha and M. lupulina were naturalized in Yunnan Province located in the tropical region in China, which are annual broadleaf herb and annual or short-lived perennial herb, respectively, cultivated as green manure in rice eld or grown spontaneously. These two plants are important forage serving as green feed for livestock, due to their ability to form nitrogen-xing nodules. Both Medicago species play an integral role in sustainable agriculture, enriching soil with bioavailable nitrogen source in rice eld. In particular, M. polymorpha has been cultivated as green manure in some regions of Yunnan, where the soils are mainly acidic and contain relatively more aluminum than those in the northern regions of China. M. polymorpha is found adaptive to the humid and warm climate of Yunnan, turning out to be the only high-nitrogen forage in this region. However, e cient rhizobia associated with M. polymorpha might be lacking, this plant species has not spread widely in rice eld in Yunnan, and was only restricted to regions of Chuxiong Prefecture and Lufeng County. Therefore, selection and application of highly effective rhizobial strains is essential for M. polymorpha cultivation in these regions, where this plant is needed as green manure or forage for improving the extensive agriculture. However, up to date there has been no document about microsymbionts associated with M. polymorpha in China.
To better understand the speci c interactions and synchronal evolution between rhizobia and M. polymorpha, but also collect rhizobial strains for screening highly effective isolates as inoculant applied in rice green manure cultivating, it would be helpful to investigate the rhizobial diversity. In the present study, rhizobia associated with M. polymorpha and M. lupulina were collected from Chuxiong Prefecture, Dehong Prefecture, Yuxi County and Kunming District of Yunnan . A total of 91 strains were isolated and characterized by 16S rDNA sequencing, 16S-23S rDNA IGS PCR-RFLP, BOX-PCR ngerprinting assay and nodulation tests. In order to analyze the correlation between rhizobial distribution and the soil properties, soil samples were also collected for physicochemical characterization. The results revealed that M. polymorpha preferred Ensifer medicae and M. lupulina preferred Ensifer meliloti in the studied region. Additionally, both host plants and soil nutrients had strong impacts on the genetic diversity of Medicagoassociated rhizobia.

Materials And Methods
Nodule collection and soil physicochemical characterization Root nodules of M. polymorpha and M. lupulina, as well as soil samples in the root zone of these plants were collected from farmlands and natural ecosystem on 7 sampling locations (Dabanqiao, Jinshan, Yanzhan, Cangling, Jiangchuan, Yinjiang and Yiliang towns) with 39 sample sites in Yunnan (Fig. S1). In each sampling site, 5-10 plants were sampled for collecting the e cient root nodules. Soils were sampled compositely from the root zone of nodule sampled plants (5-20 cm in depth) and were mixed as a single sample for each sampling sites. Complete nodules were dissected immediately from rinsed roots and stored over dehydrated silica gel in closed drying tube until their use for rhizobial isolation in the laboratory. Soil samples were ground and passed through 2-mm mesh screens for determining the physicochemical properties. Soil available nitrogen (N), available phosphate (P) (using Bray's hydrochloric acid uoride ammonium by extraction method), and available potassium (K + ) (by ammonium acetate extraction plus ame photometry) were determined with the standard procedures (Du and Gao, 2006). Soil pH was measured using a pH meter (Mettler Toledo) by suspending 5 g soil in 5 mL of distilled water, and organic matter (OM) was measured using the potassium dichromate volumetric method (Du and Gao, 2006). Soil contents of Cland HCO 3 were determined by silver nitrate titration and potentiometer titration, respectively. Concentrations of Ca 2+ , Mg 2+ , Na + , Fe 2+ , K + and SO 4 2were measured using inductively coupled plasma atomic emission spectrometry.
Rhizobial isolation and ITS/BOX ngerprinting Nodules stored in closed drying tubes were immersed in sterile water for 1-2 hours. The rehydrated nodules were surface sterilized by immerging in 95% (v/v) ethanol for 30 s and in 0.2% mercuric chloride for 2 to 3 min (depending on the nodule diameter), following by rinsing six times in sterile water. Then, the nodules were crushed separately and the liquid from each nodule was spread on plates of yeast-mannitol agar (YMA) and incubated at 28°C for 48h (Vincent, 1970). The obtained bacterial colonies were isolated and puri ed by repeatedly streaking on the same medium plates. Cultures of pure isolates were stored in YM broth supplied with 30% of glycerol at -70 °C.
For each isolate, as well as the type strains E. meliloti USDA 1002 T and E. medicae A321 T , genomic DNA was extracted extracted by guanidine isothiocyanate method using FastPure Bacteria DNA isolation Mini kit (Nanjing Vazyme Biotech CO., Ltd) from 5 mL of YM culture agitated (120 rpm) overnight at 28°C. Using the DNA extract as template, the BOXAIR primer 5´-CTA CGG CAA GGC GAC GCT GAC G-3´ (Versalovic et al., 1991) was used for BOX-PCR in a total volume of 25 μL reaction mixture with the PCR procedure of Nick et al. (1999). The PCR products were separated by electrophoresis in 1.5% (w/v) agarose gels containing ethidium bromide (0.5 mg ml -1 ) and were photographed under UV light. The BOX pro les were distinguished by their different band patterns, e.g. the isolates sharing the same pattern were designed as the same BOX pattern and were treated as colones of the same strain. For analysis of restriction fragment length polymorphism (FRLP), 16S-23S rRNA intergenic spacer (IGS) was ampli ed in 25 ml volume with primers FGPS1490 (5′-TGC GGC TGG ATC ACC TCC TT-3′) and FGPS132 (5′-CCG GGT TTC CCC ATT CGG-3′), and the corresponding PCR protocol (Laguerre et al., 1996). Aliquot of 5 ml of the PCR products were used to veri ed the IGS ampli cation (about 900 bp) by electrophoresis in 1% (w/v) agarose gel. Aliquot of 5-10 µL, depending on the concentration, was digested separately with the restriction endonucleases Hae (GG|CC), RsaI (GT|AC), HifI (G|ANTC) and MspI (C|CGG) (Laguerre et al., 1994) at 37°C for 6h, as speci ed by the manufacturer with an excess of enzyme (5U per reaction). The restriction fragments were separated by horizontal electrophoresis in agarose (2%, w/v) gels (14 cm in length) at 80V for 3h and were visualized by staining with ethidium bromide. Strains or isolates with different RFLP patterns were designated into distinct IGS types.

Phylogenetic analyses of housekeeping genes and symbiotic genes
Based on the results of IGS-RFLP analyses and BOX-PCR patterns, isolates representing different clusters and sampling sites were chosen for the analysis of multiple gene sequencing. The 16S rRNA gene ampli ed by PCR as described previously with the primers fD1 (5´-AGA GTT TGA TCC TGG CTC AGA-3´) and rD1 (5´-AAG GAG GTG ATC CAG CC-3´) (Weisburg et al., 1991). Multilocus sequence analysis (MLSA) based on the ve housekeeping genes atpD (encoding for the ATP synthase beta-chain), recA (recombinase A), dnaK (DnaK chaperone), gyrB (DNA gyrase, beta-subunit) and glnA (glutamine synthetase I) was also performed in the present study, which has been widely used to differentiate rhizobial species ( Martens et al. (2007), atpD352F/atpD871R and gyrB343F/gyrB1043R described by Martens et al. (2008) were used to amplify the corresponding genes in 25 ml volume by PCR. The PCR products were checked by electrophoresis in 1% (w/v) agarose gel. After puri ed with the Solarbio DNA puri cation kit (Beijing Solarbio Science & Technology Co., Ltd.), the amplicons were sequenced commercially using the same primers in Beijing Genomics Institute (BGI). The sequences acquired in this study were aligned with those from type strains of the de ned bacterial species (obtained from the NCBI database by blasting), using Clustal W ( Thompson et al., 1997). Maximum likelihood phylogenetic trees were constructed and were bootstrapped with 1000 pesudo-replicates using Mega 6.1 (Tamura et al., 2013). Phylogenies were also constructed using the concatenated sequences of 16S rRNA and the ve housekeeping genes by Maximum likelihood method.
Fragments of the nifH gene (about 800 bp) and nodC gene (about 700 bp) were ampli ed with primer pairs nifHF/nifHR and nodCF540/nodCR1160, respectively, using the protocols of Laguerre et al. (2001). The visualization, puri cation and sequencing of the nifH and nodC amplicons were performed same as that mentioned for the housekeeping genes. All of the acquired nucleotide sequences were used for alignment with related genes extracted from GenBank database by Blast, and construction of the phylogenies was performed using the same methods described above for the keeping housekeeping genes.
All the obtained nucleic acid sequences were submitted in GenBank database under the accession numbers MT863814-863838 for nodC gene, MT863789-863813 for nifH, and others listed in Table S1.
Nodulation assays A total of 35 representative strains were used in the nodulation tests that were selected according to their different16S-23S IGS genotypes and BOX patterns, as well as phylogenies of 16S rRNA and ve keep housing gene sequencing. M. polymorpha and M. lupulina seeds were scari ed using concentrated sulfuric acid for 10 min, rinsed several times with sterile water, and then surface-sterilized in 3.2% (w/v) sodium hypochlorite followed by several rinses with sterile water. They were then placed on 0.8% water-agar at 4℃ for 3 days, and then germinated at 28℃ until the seedlings developed roots of 0.5-1 cm in length. Two seedlings were transplanted into a sterile glass tubes (30 × 200 cm) with nitrogen-free plant nutrient solution (Vincent, 1970) (Table 1). All of these strains were divided into ve rDNA types by IGS-RFLP analysis and displayed 39 BOX-AIR ngerprinting patterns (

Phylogenies based on housekeeping and symbiotic genes
Based on the BOX-PCR patterns as well as the IGS-RFLP pro les of all the bacterial isolates, 25 representatives were selected for sequencing the 16S rDNA and ve housekeeping genes (recA, glnII, atpD, dnaK and glnA) to determine their species a liation (Fig. S2-S7).
Based on the 16S rDNA phylogeny, the representatives were divided into two clades, with 10 isolates closely related to E. meliloti LMG6133 T (>99% similarity) and the remaining 15 clustered with S. medicae WSM A321 T (>99% similarity) (Fig. S2). The phylogenetic tree based on MLSA of the 5 concatenated housekeeping genes (Fig. 1) showed a similar topology to that of the 16S rDNA tree, and similarities greater than 97% were observed among the strains within each of the E. meliloti and S. medicae clades (including the type strain in each Intriguingly, some E. meliloti strains exhibited traces of genetic incongruence, which were evidenced by comparison of the housekeeping gene phylogenies of MLSA. In the phylogenic tree of aptD (Fig. S3), the 15 E. medicae representative strains together with the type strain clustered in one clade that was consistent to the result of MLSA. Nevertheless, the other 10 isolates formed 3 lineages within the clade intermingling with E. meliloti and E. kummerowiae type strains. Similar relationships were also observed in the recA phylogeny (Fig. S5), and the genetic incongruence between aptD/recA phylogeny and MLSA phylogeny suggested that the representative strains SWF67487 and SWF67523 might acquire genetic materials via horizontal gene transfer from S. kummerowiae strains during the natural evolution process. The gene transfer event of E. meliloti isolates from E. kummerowiae can be further supported by the phylogeny of symbiotic genes, since the 25 representatives were divided into 3 groups in the nodC phylogeny (Fig. 2), which were different from the results of MLSA. Particularly, 8 E. meliloti strains formed a monoclade together with S.
Taken together, the phylogenetic analyses based on multiple concatenated housekeeping gene sequences (MLSA) divided the 25 representatives into two groups, corresponding to E. meliloti and E. medicae. Traces of HGT events in some E. meliloti strains were evidenced by the differences between phylogenies of single housekeeping genes, and between the phylogenies of MSLA and symbiotic genes.
Symbiotic performance of representative strains on Medicago polymorpha Given the colonial morphology, hosts, and the genomic/phylogenetic analyses, 32 representative strains belonged to E. meliloti and E. medicae from both Medicago species were chosen for nodulation assays ( Biogeographical patterns and genetic differentiation of Medicago rhizobia The aforementioned phylogenic analysis indicated genetic incongruence in E. meliloti strains but not in E. medicae representatives. Then, we assessed the gene diversity of the rhizobial isolates by assessing the BOX AIR pattern and Shannon index. E. meliloti populations showed signi cantly higher gene diversity index than the E. medicae counterparts. Moreover, a high level of genetic differentiation (Fst=0.92054) was observed between E. meliloti and E. medicae, based on the sequences of 5 housekeeping genes ( Table 2). To know whether host speci city would affect genetic diversity of rhizobial isolates, we rstly compared the genetic diversity of symbiotic strains within the Ensifer species. The total E. medicae strains from M. polymorpha in both farmland and nature ecosystem showed higher genetic diversity to those from M. lupulina nodules according to the BOX-PCR pattern and Shannon index. Similarly, E. meliloti isolates from M. lupulina showed signi cantly higher genetic diversity than those from M. polymorpha nodules (Fig. 3). Thus, these data suggested that rhizobial strains from their native host tend to be more genetically diverse.
Then, we evaluated whether soil type would affect rhizobial diversity by performing the PCoA based on Bray-Curtis distance (Table S2, Fig. 4). The amount of multiple nutrient factors was extensively various in natural ecosystem while it tended to be more similar in the farmland sites, as the observation that the component 1 represented 53.4% of relative eigenvalues, and component 2 represented 20.04% of relative eigenvalues, after Cailliez correction. Furthermore, multivariate analysis of variance (MANOVA) revealed that the E. medicae distribution signi cantly differed between the two habitats (P=0.047), while no overt difference was observed for E. meliloti species between farmland and nature ecosystem. In addition, permutational multivariate analysis of variance (PERMANOVA) corroborated that soil type (farmland/natural ecosystem) accounted for 1.61% (P = 0.15) of variation in the observed beta-diversity of rhizobia (Bray-Curtis distance metric).
On the one hand, all the Ensifer strains originated from natural ecosystem had a higher level of nucleotide diversity than those from farmland ( Table  2), suggesting that the soil type might contribute to rhizobial genetic diversity. The greater variation in the content of several nutrients in natural ecosystem might explain the higher genetic diversity of rhizobia. Consistently, moderate level of genetic differentiation was observed within E. medicae (Fst=0.05251) or E. meliloti (Fst=0.0501) between natural ecosystem and farmland. Furthermore, gene ow was detected for both E. meliloti and E. medicae strains between farmland and natural ecosystem. Altogether, these data suggested that Ensifer strains from natural ecosystem were more genetically diverse compared to those from farmland.

Deterministic factors for diversity of rhizobia from two Medicago plants
To identify which speci c soil factor could determine the genetic diversity of rhizobial populations, analysis of variance (ANOVA) based on a set of environmental factors was performed. Compared to natural ecosystem, the alfalfa farmland had a signi cantly higher content of multiple nutrient factors, including organic matter (OM), sodium (Na + ), magnesium (Mg 2+ ) and bicarbonate (HCO 3 -) (Table S3, S4). As shown in the CAP analysis based on Bray-Curtis distance of PA (presence/absence) transformed species data (Fig. 5A), the soil physiochemical data could explain 53.4% of variation in the diversity of ve IGS type species. Soil Ca 2+ , Na + , HCO 3 -, Cl -, SO 4 2and pH contributed to a signi cant partition in the rst component (63.51% of total variance). Meanwhile, nitrogen (N), OM, potassium (K), and phosphorus (P) played a minor role in the second component (14.31 % of total variance), showing negative effects on rhizobial distribution. Variation partition analysis revealed that 11.7%, 5.8%, 2.2% and 0.5% could be signi cantly explained by 6 ions, NPK, pH and OM variables respectively. These results indicated that Ca 2+ and Na + were major factors in shaping the genetic diversity of ve IGS types (Fig. 5B).
Rhizobial species correlated with different sets of explanatory variables were further identi ed. IGS type A was more likely found in the farmland sites with lower HCO 3-. IGS type C was more likely found in the nature ecosystem with high soil Ca 2+ , Na + , HCO 3 -, Cl -, lower nitrogen, K and OM. The IGS types B, D and E more likely appeared in the nature ecosystem sites with high P and low HCO 3 -. In summary, IGS types A and C have similar nutrition utilization, and both preferred soils with low potassium, OM, phosphorate and nitrogen contents, while IGS type C preferred soil with high contents of Ca 2+ and Na + (Fig. 5A).
To evaluate the correlation of rhizobial abundance with environment factors, CAP analysis based on Hellinger transformed species data was conducted (Fig. 5C). IGS type A had a signi cantly higher abundance in farmlands that were positively associated with pH. While more stringent than type A, type C was also negatively correlated with soil N, k, P and OM. IGS types B, D and E were not considered due to their extremely low abundance (Fig. 5C). Variation partition analysis of Hellinger transformed species data (Fig. 5D) exhibited almost the same pattern as the PA transformed species data (Fig. 5B). The range of soil physiochemical content for two rhizobial species identi ed in this study was given in Table S4.
Taken together, distribution of E. medicae strains was positively correlated with Na + in soil, and their abundance was also associated with low level of Nit and OM. E. meliloti isolates appeared to prefer good-quality soil conditions with high Ca 2+ and Na + contents, and negatively correlated with soil N and OM. To conclude, the soil contents of N, OM as well as Ca 2+ and Na + were the major soil factors to shape the distribution of rhizobial strains belonging to the two Ensifer species detected in the present study.

Discussion
In the present study, we systematically investigated, for the rst time, Our results demonstrated that the speci city for the two Medicago species and their microsymbionts appeared not so stringent, although M. polymorha was recognized as the sole host of E. medicae (Biondi et al., 2003;Brunel et al., 1996;Rome et al., 1996). While, it has been found that E. . Therefore, it could be estimated that the associations between the hosts M. polymorpha/M. lupulina and the microsymbionts E. medicae/E. meliloti were not only determined by the preference (or speci city) between the hosts and the rhizobia, but also might be regulated by the soil abiotic and biotic conditions. Furthermore, soil edaphic could explain the distinct geographical distribution for these two rhizobial species. From CAP analysis, E. meliloti were displayed strong negative correlation to soil contents of nitrogen and organic matter in nature ecosystem, while the negative correlation of E. medicae with nitrogen and organic matter contents was even more stringent. Garau et al. (2005) reported that E. medicae was mostly associated with medics that well adapted to moderately acid soils, such as M. polymorpha, M. arabica and M. murex; whereas E. meliloti was predominantly isolated from M. littoralis and M. tornata that naturally grow in soils with alkaline or neutral pH, and also from M. sativa in acidic soils as well (Ramirez-Bahena et al., 2015). In our studies, both E. medicae and E. meliloti strains were mostly isolated from alkaline soils, but the nutrient contents and iron patterns were different among the sampling sites, which formed the determinants to regulate the nodule occupancy of E. medicae and E. meliloti on M. polymorpha and M. lupulina.
By analyzing the BOX-AIR ngerprinting pro les of all the 91 isolates, we found that E. meliloti populations showed signi cantly higher gene diversity index than the E. medicae counterparts in two plant hosts and two habitats in Yunnan, China. Analysis on the nucleotide diversity of 25 selected representatives also supported this observation. This difference may be due to the strain number of the species in BOX pro le analysis. A high level of stringency in Medicago-rhizobia speci city might constrain the genetic diversity of strains with poor symbiosis e ciency. Since the more e cient rhizobial strains would outcompete in the nodulation process, thus being thrived and accumulated since they can favor plant growth by xing nitrogen more e ciently. Such selection of symbionts by plant host may not only lead to the formation of the dominant rhizobial populations in the sampling sites, but also could enhance the diversity of advantaged strains as contrast to the ine cient ones.
Furthermore, in accordance with the another report (Bailly et al. 2007), our results indicated that the soil type (or ecological environment) have effect on rhizobial diversity. Rhizobial strains tested in this study (both Ensifer species strains) from natural ecosystem have higher nucleotide diversity than those from farmland. This suggested that Medicago symbionts might undergo more gene exchange events with other bacterial habitants in the natural ecosystem when compared to that in the farmland. Indeed, a greater value of genetic differentiation was observed in E. meliloti stains than in E. medicae strains that might be a result of ecological niche barrier. E. meliloti was more frequently recovered from nature ecosystem; while E. medicae strains were more readily isolated from M. polymorha in farmland. Selection force by environment and plant host may block rhizobial migration and prevent the subsequent gene exchange. This may also challenge the hypothesis that E. medicae is an evolutionary divergence from E. meliloti population (Biondi et al., 2003). Genome sequence analysis of more rhizobial species is needed to clarify the evolutionary relationship between E. medicae and E. meliloti.
Instead, these two Ensifer species retained their own genetic characteristics. We found high levels of gene ow (Nm) within E. medicae or E. meliloti populations in the two habitats. Various subpopulations of the same Ensifer species could intermingle their genes through the horizontal gene transfer processes such as conjugation and transformation, which conferred su cient genetic exchanges among populations. This may be an important evolutionary force to contribute to the genetic diversity. Interestingly, phylogenies of some housekeeping and symbiotic genes suggested that gene There is no con ict of interest among authors. And all the authors listed have read the manuscript and approved the submission.
All data generated or analyzed during this study are included in this published article and its supplementary information les.