Comparison of nucleotide diversity and symbiotic properties of Rhizobium meliloti populations from annual Medicago species

Forty-three isolates of Rhizobi~nz meliloti were trapped from Medicago (M. polymorpha, M. truncatula, M. rigidula, M. orbicularis and M. minima) and (M. satiua). were


Introduction
Annual Medicago species (or medics) are well agement and pasture reseeding in the French Mediterranean area [2]. For this region (Provence-C8te d'Azur, Languedoc-Roussillon and Gorse), five main and ubiquitous medic species (Me&cngc-, PO/V morpha, M. truncatula, M. rigidula. M. orbicularis M. minima) were recorded and selected for their suitability for pasture improvement [3]. Medics are more widely used in Australia where ley farming (a cereal pasture rotation) is practiced: they increase soil nitrogen. improve soil structure and control soil erosion [4,.5]. Since the use of nitrogenous fertilizers is not appropriate in these extensively planted areas, medics are enhanced by adapted Rhi:obiur?z-legume symbiosis and can thus supply nitrogen to other plants through their association with appropriate Rhkobium spp. To take advantage of this biological nitrogen input in semi-natural systems with economic and environmental benefits, suitable rhizobia must be present in the soil and must be able to establish an efficient symbiosis with medics. If not. efficient strains should be selected for introduction into soil as inoculant. But the establishment of inoculant strains can be hindered by the presence of ineffective populations of R. meliloti which can cause parasitic nodulation. Thus. to improve growth of medics, it is essential to characterize the native R. meliloti populations and the factors which influence their composition and dynamics.
Medics are associated with rhizobia belonging to the R. meliloti species. Their symbiotic interaction appears to be very specific: all R. meliloti strains are not able to nodulate all the Medicago species. and when they are, they do not always fix nitrogen. Eventually, when fixation does occur, the levels may vary. Brockwell and Hely [6.7] defined three R. meliloti groups based on their ability to elicit nodules on Medicago, and eight based on their similarities in fixing nitrogen on Medicago species. For instance, M. minima can form effective associations with a great number of R. meliloti strains whereas M. rugosa, M. rigidula or M. Iaciniata have very specific requirements [7-91. More recently, R. meliloti populations from annual and perennial Medicago were examined by multilocus enzyme electrophoresis (MLEE) and classical restriction fragment length polymorphism (RFLP) analyses on rRNA genes [IO]: R. meliluti originating from medics were divided into two highly diverging groups. which could coincide with different host-microsymbiont affinities.
To define the properties of indigenous R. meliloti populations occurring in southern France, we collected rhizobia by trapping with the most ubiquitous medics and then applied an approach in two steps for characterizing isolates. Screening all isolates by cross-inoculation methods and classifying rhizobia on the basis of their ability to nodulate and fix nitrogen is laborious and could be difficult to interpret properly, i.e. variations and inconsistencies were already noted by Brockwell and Hely [7]. Consequently, before examining isolate characters by plant tests, we first identified all isolates by the PCR-RFLP typing which provides an efficient and rapid method for classifying soil bacteria and rhizobia [II -131. Specific DNA regions are first amplified by polymerase chain reaction (PCR), then the PCR products digested by restriction endonucleases are separated by gel electrophoresis to reveal restriction fragment length polymorphism (RFLP). Universal primers were chosen in rRNA and rlif genes because prior studies had shown that different portions of the genes could produce genetic and phylogenetic polymorphism [ 1 I-151. The nodulation and nitrogen fixation specificity was then examined by cross-inoculation among each of the different PCR-RFLP groupings. We expected differences in genetic group ings to reveal consistent symbiotic traits. Moreover, one commercial perennial Medicago satila species was used since medics or luceme could preferentially trap some types of rhizobia. Comparisons were also made with reference R. meliloti strains.

Description of the site
A 5 m X IO m soil plot, where five species of annual spontaneous medics naturally coexisted, was located in the Mediterranean garrigue (Combaillaux, HCrault. France) in May 1986. Medic species were identified as Medicago polymorpha CL.). M. truncat-L&I (Gaertn.), M. rigidula CL.). M. orbicularis (All.) and M. minima (Grufb.). Samples of surface soil from a part of the site where no legumes were growing, were collected aseptically from the top 10 Downloaded from https://academic.oup.com/femsec/article-abstract/19/2/71/791581 by guest on 15 March 2020 cm of the profile; they were then mixed and stored at 4°C less than 24 h before use. The physical and chemical properties of the soil were as follows: sand 486 mg . g-l; silt 259 mg.g-'; clay 227 mg'g'; total calcareous 476 mg . g-'; active calcareous 142 mg.g-'; pH (1: 1 soil -water) 8.6; organic matter 28 mg.g-'.

Isolation of rhizobia
Seeds from the five medics collected on the site and from a commercial perennial luceme (M. satica cv. magali) were used as hosts to trap rhizobia. They were surface sterilized in 1% calcium hypochlorite for 2 min and washed with seven changes of sterile water [16]. They were scarified by a spot hot-shock treatment with an iron microtip of a thermoregulated soldering station (JBC, Montpellier, France) to improve seed germination, and then incubated on a 0.7% agar medium. They were germinated for 72 h Table I Isolates of R. meliloti and reference strains in a growth chamber at 28°C. The pregerminated seeds were placed aseptically in Gibson tubes [17] containing vermiculite with a nitrogen-free nutrient plant solution [ 181. Soil was homogenized by shaking for 30 min and then 20 g of the soil were added to 190 ml of sterilized water to obtain a dilution of 10-l) which was vigorously shaken. One ml of the dilution was then used to inoculate eight plants grown axenically in Gibson tubes. Eight control tubes were not inoculated. The plants were grown for one month in a room with controlled environment (70% moisture), under a combination of metal halide and mercury vapor lamps that provided a photon flux density of 600 pmol . m-' . s-' (400 to 700 nm) was obtained from each nodule. Single-colony isolates were purified on yeast extract-mannitol (YEM) agar [ 161. A total of 43 isolates was obtained and stored on YEM slants at 4°C or in 50% glycerol at -80°C. These isolates and the reference R. meliloti strains are listed in Table I. Reference strains belonging to both divisions A and B [lo] were included for comparative analysis.

Genomic DNA extraction
Small-scale preparations of total bacterial DNA were obtained by growing each strain on a tryptoneyeast (TY) medium [20] agar plate for 4 days at 28°C. Cells were scraped off in TE8 buffer (Tris-HCl 50 mM, EDTA 50 mM, NaCl 100 mM, pH 8) and pelleted in a microcentrifuge tube for 2 min ( 10 000 X g, 15°C). The cell pellets were washed twice with TE8 buffer and finally resuspended in 0.5 ml TE8. The bacterial suspension was lysed by passing the microtubes through liquid nitrogen and boiling water twice for 1 min each time. The samples were first extracted with phenol equilibrated with TE8 buffer and then with chloroform. DNA was precipitated by adding 1 volume of isopropanol and 0.1 volume of NaCl (1 M) and collected by centrifugation for 10 min in a microcentrifuge.
The resulting DNA pellet was washed with 70% ethanol and 95% ethanol and then allowed to air-dry. The dried DNA was dissolved in 20 ~1 of pure water at 37°C for I h. This procedure routinely yielded 100 to 200 ng DNA PI-1 that was sufficiently pure to be used as a template for the polymerase chain reaction (PCR).

Data analysis
To allow a quantitative comparison of the different PCR-RFLP analyses, restriction fragment patterns from R. meliloti isolates and references were compared pairwise for the presence or absence of restriction fragments. The proportion of shared fragments (F), termed Dice's similarity coefficient, was calculated as F = 2.n,,/(n, + n,>, where n,, is the number of common fragments between isolate x and y, with the three restriction endonucleases used and n, and n, the numbers of fragments from isolates x and y, respectively [25]. Nucleotide diversity, n, the average number of nucleotide differences between isolates, was calculated as n= [(-In F/r], where r is the number of nucleotide base pairs in the restriction endonuclease recognition site [25]. All restriction endonucleases used in this study to digest amplified DNA recognize sites of r = 4. An unweighted pair-group method using arithmetic average (UPGMA) dendrograms of nucleotide diversities was constructed by the method of Sneath and Sokal [26]. A phylogenetic dendrogram was also constructed by the neighbor-joining method [27] with the Restsite software program developed by Miller [28]. The observed frequencies of PCR-RFLP patterns were compared with expected values using chi-square ( x2) tests (when expected numbers were at least five in each class).

Symbiotic effectiveness
Isolates were tested by cross-inoculation for their ability to form nodules and to efficiently fix nitrogen on the six species of Medicago. Surface-sterilized seedlings of Medicago were grown axenically as described above. One ml of each rhizobial suspension from an early stationary-phase culture in YEM broth (containing 10~6-10-7 cfu . ml-') was inoculated in each seedling in Gibson tubes [17]. Eight replicates were done for each isolate and reference strain. Eight uninoculated plants and eight plants supplemented with nitrogen (10 mM of Ca(NO,),) were used as controls. The experiments were conducted in the same growth conditions as described above. The plants were harvested after 50 days, dried at 70°C for 48 h and then weighed to estimate the nitrogen fixation efficiencies. The experiments were repeated at least twice. Shoot dry matter data were subjected to an analysis of variance [29]. The data being numerous, ANOVA was done in two steps. First, variance analyses were performed within each group of Combaillaux isolates originated from one Medicago species in order to assay intra-species variations. Secondly, the dry matter weight averages of each isolate were compared between themselves in order to assay inter-species variations, and compared with the dry matter data of reference strains and controls.

Results
The isolates we had obtained at Combaillaux were first subjected to PCR-RFLP analysis on ribosomal and nif DNA regions. Then a subsample of 33 rhizobia was chosen among each of the different PCR-RFLP groupings and each of the six plant origins in order to determine their nodulation and fixation abilities.

I. PCR-RFLP analysis
Electrophoretic analysis of uncut PCR products indicated that all amplified fragments were similar in overall size: approximately 1.5 kb and 2.7 kb for the ribosomal 16s rDNA and 16SrDNA + IGS regions respectively, and 1.35 kb for the nifK-D region (data not shown). The amplified 16s rDNA region size was in agreement with those of 1478 bp estimated from the R. meliloti 16s rDNA sequences available in databases. Amplified DNA regions were then digested by three four-base cutting endonucleases to detect polymorphism by RFLP.
Among the 43 Combaillaux isolates, two profiles Rl and R2 were identified on the amplified DNA region nested between the 16s rDNA and 23s rDNA genes (Fig. 1). Two isolates (CP41 and CP4111) trapped from M. polymorpha (profile R2) were indeed distinguishable from the other 41 isolates (profile Rl) which showed the same patterns for each of the three restriction enzymes tested. The reference strains M29 and WSM533 presented two ribosomal DNA region patterns identical to the Combaillaux isolates (profiles Rl and R2 respectively).
Of the 20 possible combinations between the five ribosomal and the four nif profiles, three were recovered from Combaillaux bacterial populations and four additional ones were found among reference strains which originated from different geographic locations. Pattern RlNl representing 74% of all isolates, predominated in each sample from one plant species. Its frequency did not differ significantly (P < 0.05) between the six Medicago species ( x' = 4.49: 5 degrees of freedom). The two other patterns Rl N2 and R2N3 represented 21% and 5% of the total sampling respectively. The RlNl and RlN2 distributions according to the six trapping plants were statistically similar.
Dendrograms A and B were generated by UP-GMA clustering from the data sets obtained with ribosomal and rrif typings, respectively (Fig. 3). In ribosomal dendrogram A, rhizobia grouped at 0.024 of diversity and could be divided into two distinct clusters (AI and AII). Branch AI contains the Rl profile isolates and the strains, M29, L5-30 and 201 1 while branch AI1 is constituted of the R2 profile isolates and strains WSM533, M3 and M104. Dendrogram B, obtained by analysis of the four nif patterns, has three branches: branches BI and BII defined at 0.029 of diversity, and a more divergent one corresponding to the N4 pattern represented by the single strain M29. Except for M29, rhizobial populations from the AI cluster belonged to the BI cluster and those from the AI1 cluster recovered the BII cluster (Fig. 3). The tree inferred by the neighbor-joining method agreed with the UPGMA dendrogram topography (data not shown). With the same method, the strain M29, which has a highly divergent nif profile, clustered in branch I of the rDNA dendrogram A while it was excluded from branch I of the nif dendrogram B.

Symbiotic properties
A sample of 33 R. meliloti isolates and five references were assayed for both nodulation and nitrogen fixation efficiency in symbiosis by cross-inoculation with five annual species and one perennial species of Medicago ( Table 2).
The isolates nodulated the five other Medicago species from which they were not issued. However   Downloaded from https://academic.oup.com/femsec/article-abstract/19/2/71/791581 by guest on 15 March 2020 6.7 nodules per plant). In contrast, the four other M. po/vmorpha isolates could elicit, with high regularity a higher number of nodules (i = 23.2 rf: 38.3 nodules per plant). On M. polymorpha, differences in the aspect of nodules were observed: all the isolates, except CP41 and CP4111, formed rudimentary and white (Fig. 4A) nodules which we counted positively as they provide evidence of infection. In contrast, CP41 and CP4111 isolates elicited cylindrical and typical efficient nodules (Fig. 4C). This striking phenotypic trait was also shared by reference strains: strains M3, M 104 (Fig. 4B) and WSM533 triggered typical nodules while M29 and 2011 elicited rudimentary ones.
The shoot dry matter data are given in Table 2. Among the 33 Combaillaux isolates, two significant differences were detected: (i) on M. rigidula, dry matter was lower (P < 0.001) for two isolates (CP41 and CP4111) than for all other R. meliloti isolates and the two reference strains (2011 and M29): and (ii) on M. polymorpha, two isolates (CP41 and CP4111) trapped from M. polymorpha and two reference strains (WSM533 and M3) strains were able to develop an efficient nitrogen fixation (P < 0.001). The 29 other Combaillaux isolates and two other reference strains (2011 and M29) elicited non-effective and rudimentary nodules on M. polymorpha, with no significant difference in biomass over the uninoculated control. The fifth reference strain Ml04 tested on M. polymorpha showed plant biomass data slightly higher than the uninoculated control but not significantly different. However, plants in symbiosis with Ml04 were obviously greener and more vigorous (Fig. 4B) than uninoculated plants or plants associated to non-efficient Combaillaux isolates like CP61 (Fig. 4A). Probably, Ml04 is a strain with a low N,-fixation ability that we cannot reveal with the dry plant biomass method.

Comparison between molecular groupings and symbiotic properties
The most relevant symbiotic properties of Combaillaux isolates were noted on M. polymorpha plants: isolates with a RI ribosomal profile elicited rudimentary and non-efficient nodules on M. polymorpha whereas the two isolates with a R2 profile (CP4I and CP4111) triggered typical and efficient ones (Fig. 4A,C). Such properties were extended to five R. meliloti reference strains: strains M29 and 2011 close to Rl type, belonged to branch I and elicited rudimentary and non-efficient nodules on M. polymorpha while strains WSM533, M3 and M104, genetically close to R2 type, belonged to branch II and triggered cylindrical and mature ones. These typical nodules formed on M. polymorpha were efficient except for Ml04 which appears to be a low symbiotic N,-fixer (Fig. 4B). Moreover, rhizobia isolates with a Rl profile seem to nodulate more poorly M. rigid&z plants than isolates identified by a R2 profile. The correlation with the same symbiotic properties can be recovered from groups I and II depicted by nif pattern analysis except for the strain M29. In contrast, the subdivision within branch I (i.e. Nl and N2 groupings) could not be related to the cross-nodulation and -fixation results.

Discussion
Two markedly divergent R. meliloti groups, which we designated groups I and II, were found in a restricted soil area: a dominant one (I) was trapped by the six Medicago species and a minor one (II) by

M. polymorpha.
Our classification is based on the analysis of two chromosomal and plasmidic markers and of the symbiotic properties of different isolates and strains.
Our analysis of a chromosomal region containing the 16s rDNA plus the IGS nested between 16s and 23s rRNA genes allows an easy differentiation among Rhizobium isolates from the same geographical origin. Because the restriction patterns of amplified DNAs harbouring only the 16s rRNA gene were identical for all the rhizobia tested, divergence between isolates and strains laid on the IGS part analysis. This result is not surprising as 16s rDNA sequences are highly conserved within a species: two divergent R. meliloti strains differed only by a single nucleotide on a 260-bp 16s rDNA segment [30] and differences between 16s rRNA sequences may be insufficient to give satisfactory resolution within bacterial genera [3 11. Thus, we confirmed that intergenic and non-coding rDNA regions are excellent targets to discriminate closely related species among rhizobia and relatives [13,32]. Groups I and II were B. Brunei rt al. / FEMS Micmhiolog~ Ecology 19 (I 9961 71-82 already found by two other chromosomal analyses with isoenzymes [lo] and with repetitive DNA dispersed in the genome [33]. Our results compare well with those just mentioned as we used some identical strains. The two ribosomal groups I and II we detected, should correspond to divisions A and B, as defined by Eardly et al. [lo]: indeed, division A represented by strains M29, L5-30 and 2011, parallels ribosomal group I, whereas division B including strains M3 and M104, matches ribosomal group II. The classification of R. meliloti in two groups is therefore validated by a combination of different techniques as required for a polyphasic taxonomic approach. However, groups I and II will be formally stated as individual species only when differences by DNA/DNA hybridization and by phenotypic traits are well documented [34]. Although chromosomally encoded characteristics are crucial for establishing bacterial taxonomy, symbiotic performances which are plasmid-encoded among R. meliloti [35] are important for a practical use of rhizobia. So it was interesting to chose a pSym marker since plasmidic DNA could evolve differently than chromosomal DNA. A plasmidic marker might have been more informative than a chromosomal one for screening symbiotic properties. Nevertheless, the discrimination power between the two molecular markers was slightly better with the amplified ribosomal intergenic region than those of n$KD genes (five versus four profiles), and divisions I and II detected by rzifK-D amplified DNA analysis matched those established by ribosomal PCR-RFLPs, except for one strain M29 which had a highly divergent nif region.
The occurrence of groups I and II was confirmed by nodulation and fixation abilities on M. polymorpha. Profile R2 was consistent with isolates efficient on M. polymorphu whereas profile Rl corresponded to isolates which elicit rudimentary and non efficient nodules. Such properties were shared by reference strains from different geographical origins: strains genetically close to RI type were not efficient whereas those close to R2 type were efficient in fixing nitrogen on M. polymorpha. The strain M29. characterized by a Rl profile with a broadly divergent rzif region, induced non-efficient nodules on M. pol.ymorpha and thus behaved like the other strains belonging to group I. Yet, M29, which was selected as inoculant for M. rigid&n [36], behaved in a particular way: its nodulation activity could be unstable after long storage at 4°C (> 3 months) (L.A. Materon. personal communication, unpublished results). Erratical nodulation on M. rigid& for group II rhizobia, was also noted. A similar phenomenon was obtained by Brockwell et al. [8] on one line of M. rigid&z. This could be the result of variability due to experimental conditions used for plant tests. It may be also due to a mixture of nodulating and non-nodulating plants grown from seeds collected from local M. rigidula plants, as shown on other leguminous plants [37]. Further data are thus required on M. rigid& for determining if the low ability of nodulation is correlated to rhizobia from group II. Yet all molecular polymorphism's detected were not related to symbiotic properties. Within group I, Nl and N2 groupings, and within group II. R2, R4 and R.5, could not be related to the nodulation and fixation test results, but could correspond to other biological properties. The correlation between genotypic and symbiotic characterisations was thus consistent only when the deeper dichotomy I and II of the ribosomal and n{f dendrograms were compared with the nodulation behaviors on M. polymorpha. M29 was still an exception when we considered only its nif grouping which was excluded from groups I and II although its ribosomal profile belonged to group I. The ribosomal profiles we have observed have probably no particular significance in the nodulation process on M. polymorpha, they are merely markers of genetic lineages. Therefore, it would be interesting to study nodulation processes of these two lineages as they could show differential adaptation to this medic.
Seventy-four R. meliloti strains originating from four of the five annual medics we used (M. trutrcatulcl, M. polymorpha, M. orbicularis and M. rigid&) were included in both divisions A and B as defined by Eardly et al. [lo]. Similarly, isolates trapped with the Combaillaux M. polymorpha plants, belonged to the two different groups I and II. Although group II rhizobia were able to nodulate efficiently the perennial lucerne Medicago satioa and some other medics, they were not extracted from them. Group II rhizobia could be a lower competitor than those of group I on M. satirsa and the medics tested except M. polymorpha. This suggestion of a difference in competition ability is not in contradiction with characterization by MLEE [IO] for strains collected from M. rigidulu and M. satitla. When M. rigidula was the plant origin. 23 isolates fell into division A and ten in division B, whereas when M. satitla was the plant origin, the difference of distribution was more clearly marked: 93 M. satira rhizobia belonged to division A while only 14 were found in division B. In contrast, isolates from M. trzmcatula, M. orbicularis and M. polynorpha were equally distributed between both divisions A and B [lo]. One important point to consider for characterizing groups I and II seems to be their nodulation behaviors on medics and not only their plant origin as non effective strains could be isolated from M. polyzorpha.
In this study, the divergence between groups I and II cannot reflect any geographic difference, as they originated from a very restricted area of ten-gram soil. They are two sympatric groups in the Combaillaux sampling site but their distribution areas could be different as their multiplication into the soil could depends on the occurrence of the plant host. The presence of R. meliloti with R2 backgrounds could be favoured by the distribution of M. polymorpha species. whereas rhizobia with Rl backgrounds could be better promoted by other medics or luceme.