Novel, non-symbiotic isolates of Neorhizobium from a dryland agricultural soil

Semi-selective enrichment, followed by PCR screening, resulted in the successful direct isolation of fast-growing Rhizobia from a dryland agricultural soil. Over 50% of these isolates belong to the genus Neorhizobium, as concluded from partial rpoB and near-complete 16S rDNA sequence analysis. Further genotypic and genomic analysis of five representative isolates confirmed that they form a coherent group within Neorhizobium, closer to N. galegae than to the remaining Neorhizobium species, but clearly differentiated from the former, and constituting at least one new genomospecies within Neorhizobium. All the isolates lacked nod and nif symbiotic genes but contained a repABC replication/maintenance region, characteristic of rhizobial plasmids, within large contigs from their draft genome sequences. These repABC sequences were related, but not identical, to repABC sequences found in symbiotic plasmids from N. galegae, suggesting that the non-symbiotic isolates have the potential to harbor symbiotic plasmids. This is the first report of non-symbiotic members of Neorhizobium from soil.


INTRODUCTION
A group of α-proteobacteria from the Rhizobiales order, especially within the Rhizobiaceae and Bradyrhizobiaceae families, are collectively known as Rhizobia because they have the ability to establish root-nodule symbioses with legumes. Within these nodules, Rhizobia fix atmospheric nitrogen, and this fixed nitrogen is assimilated by the plant. This makes most legumes uniquely independent from the need of any exogenous nitrogen fertilizer, an important ecological and agricultural trait that is at the basis of any effort aimed at sustainable agriculture.
Since their first isolation from legume root nodules (Beijerinck, 1888), it has been known that Rhizobia are present in the soil, wherefrom they can colonize emerging roots of their legume host. However, they have rarely been isolated directly from the soil, given that the use of trap legume plants provides a facile method for Rhizobia isolation from root nodules. This is very convenient, especially because rhizobial soil populations have been often estimated as ranging between 10 2 -10 5 cells per gram of soil, depending on soil type and host plant (Singleton & Tavares, 1986). However, the use of legume trap plants allows a very limited glimpse at rhizobial populations in soil, for two reasons. First, legume-rhizobial symbioses are usually very specific, and a specific legume can only be nodulated by a specific type of Rhizobium, a phenomenon whose molecular bases have been intensively studied in recent years. Second, the genetic determinants for a successful symbiosis with a host legume are but a small fraction of the genetic complement of a Rhizobium. These determinants are often present in plasmids or mobile genomic islands that can be transferred, exchanged or lost (López-Guerrero et al., 2012;Andrews & Andrews, 2017). Therefore, it is possible that, for any given Rhizobium, a large non-symbiotic subpopulation co-exists with the symbiotic subpopulation and represents an ill-studied reservoir of genetic diversity. That this is indeed the case has been demonstrated in very few instances. In a pioneering study, Sullivan et al. (2002) were able to show that Mesorhizobium loti, the microsymbiont of Lotus spp. harbors its symbiotic determinants within a symbiotic genetic island integrated in its chromosome, and that this island can excise and be transferred to other cells. Furthermore, after inoculating a soil with a strain containing a marked symbiotic island, they were able to recover it in different genomic backgrounds, thus proving that the island undergoes cell-to-cell transfer in the soil, and that non-symbiotic M. loti strains are present in the soil that can receive the marked symbiotic island and thus acquire the ability to nodulate Lotus.
Despite the above, the direct isolation of Rhizobia from soil has received little attention, and even in those few cases, the interest was placed on symbiotically-competent Rhizobia. For example , Louvrier, Laguerre & Amarger (1995), in ground-breaking work, devised a semi-selective culture medium to enrich Rhizobium leguminosarum from soil. They were interested in isolating symbiotic strains that had not been selected by the plant host, in order to test the hypothesis that the different plant host species this bacterium colonizes, select specific genotypes among those present in the soil (Louvrier, Laguerre & Amarger, 1996). Likewise, Tong and Sadowsky optimized a Bradyrhizobium japonicum and B. elkanii-enriching medium (1994) that was later used to isolate symbiotic and non-symbiotic Bradyrhizobia from soil (Pongsilp et al., 2002). In our lab, we have recently built upon Laguerre and Amarger's observations by carrying out a population genomics study of genotype selection by the legume host in the same agricultural soil in Burgundy (Jorrín & Imperial, 2015). In this study, many non-symbiotic Rhizobia were isolated (Jorrín, 2016). The above results are in line with the recent identification of non-symbiotic Rhizobia as abundant components of plant microbiomes (Lundberg et al., 2012;Shakya et al., 2013;Chaparro, Badri & Vivanco, 2014;Ofek-Lalzar et al., 2014;De Souza et al., 2016) and the isolation of non-symbiotic Rhizobia from rhizospheric soil (Segovia et al., 1991;Sullivan et al., 1996;Van Insberghe et al., 2015;Jones et al., 2016) and the surface of roots (Sullivan et al., 1996).
In this work, we set out to directly isolate Rhizobia from a dryland agricultural soil in Southern Spain where no record of legume cultivation is available. A large fraction of those isolates was found to constitute a hitherto unsuspected, non-symbiotic clade within the recently described genus Neorhizobium (Mousavi et al., 2014), whose known members were, up to this work, legume symbionts: N. galegae, isolated as symbionts of the cold-climate legume Galega sp. (Lindström, 1989) as well as of many other legumes;

Bacterial strains and growth conditions
Bacterial strains used in this work are listed in Table 1. Rhizobial strains were grown in Yeast Mannitol Broth (YMB;Vincent, 1970) at 28 • C, either in liquid culture or on solid media supplemented with 1.5% agar. For long-term maintenance, strains were grown at 28 • C in YMB and preserved in 20% glycerol at −80 • C.

Soil
Tomejil soil is a dryland agricultural soil from the Las Torres-Tomejil Experimental Agricultural Station of the Instituto de Investigación y Formación Agraria y Pesquera de Andalucía (IFAPA) in Seville, Spain, where winter wheat is usually grown, cardoon occasionally, and there is no previous record of legume cultivation. A 10 m × 10 m plot (37 • 24 33.10 N, 5 • 34 51.91 W; 77 m above sea level) was selected and fenced for our research. For soil collection, portions of soil down to 30-50 cm depth were collected with a shovel from several points within the plot. Soil samples were maintained at 4 • C and −20 • C. Physicochemical characterization was done externally in Laboratorio de Edafología y Técnicas Analíticas Instrumentales, EUIT Agrícola, UPM, Madrid. Physicochemical properties of the soil are described in Table S1.
Positive fnrN isolates were characterized phylogenetically by PCR amplification and sequencing of the housekeeping genes 16S rDNA and rpoB. Presence of symbiotic genes was determined by amplification and sequencing (if present) of nodC and nifH.

DNA sequences
GenBank accession and Bioproject numbers for sequences and genomes, respectively, obtained in this work are listed in Table S2, together with accession numbers for reference sequences used in the analyses.

Direct isolation of Neorhizobium sp. from soil
A total of twenty-eight independent isolations were carried out from Tomejil soil (Fig. 1). After four days, pink-white colonies appearing on MNBP plates were chosen as putative fast-growing rhizobia. From those, we chose to discard all that were able to grow on LB agar. In our hands, this includes most Ensifer spp. and many non-rhizobial isolates. However, most Rhizobium spp. were unable to grow on the 5 g l −1 NaCl present in LB agar, and thus this step resulted in a further enrichment of Rhizobium spp. DNA from three hundred and forty-two of these putative Rhizobium spp. was amplified for the fnrN gene by PCR. This gene is important for microaerobic metabolism (Gutiérrez et al., 1997) and has been found in all sequenced Rhizobium spp. and Agrobacterium spp., but not in other Rhizobia. Ninety-nine of the DNAs resulted in good amplification of a single major band. However, two different band sizes were observed. While forty-six strains resulted in amplification of a PCR band of the expected size (279 bp), the remaining fifty-three strains amplified a larger fragment (353 bp, Fig. S1). In order to reduce the number of isolates for further characterization, a fragment of the rpoB gene was amplified and sequenced from all of them. The rpoB marker has been shown to be very effective in discriminating phylogenetically close bacteria (Mollet, Drancourt & Raoult, 1997;Khamis et al., 2003;Case et al., 2006), including Rhizobia (Jorrín, 2016). Surprisingly, rpoB sequences from the same fifty-three strains clustered with sequences from genus Neorhizobium (Fig. 2B). These sequences were compared and classified into groups (twenty in total) if they differed in at least one nucleotide (Fig. S2). In order to facilitate further studies, a representative strain of each these rpoB sequence types was chosen. Near complete 16S rDNA sequences were obtained from these representative strains and compared against those in databanks, confirming that they formed a diverse clade that was closely related to known members of the genus Neorhizobium, but distant from the Agrobacterium-Rhizobium group ( Fig. 2A).

Genotypic and genomic characterization of Neorhizobium sp. isolates
Since all previously characterized Neorhizobium isolates had been obtained from nitrogenfixing legume root nodules, we tried to amplify nodC and nifH, markers related to symbiosis and nitrogen fixation, respectively, from our Tomejil isolates. Negative results were obtained in all cases, suggesting that these isolates were non-symbiotic. Since direct tests of symbiotic ability with legume plants were hampered by the fact that the known hosts of Agrobacterium tumefaciens NCPPB 2437
We first searched for fnrN genes. This was important because no fnrN gene has been described in Neorhizobium spp. isolates, and it was absent from the Neorhizobium spp. genomic sequences available in databanks. Using the well-characterized R. leguminosarum fnrN gene for BLAST comparisons, we were unable to find any relevant hit against genomic sequences from Tomejil isolates, despite the fact that the isolates were chosen because they showed clear amplification of a single (although of different size, see above) band with fnrN primers (Fig. S1A). Amplified PCR bands from some of the isolates were sequenced and compared with databanks. Part of the amplified region showed similarity with genes encoding a poly (3-hydroxybutyrate) depolymerase from Rhizobia, including Neorhizobium galegae (but not N. alkalisoli or N. huautlense). This gene, together with an upstream ORF, was present in genome assemblies of the Tomejil strains and contained sequences partially complementary to those of the fnrN primers that may explain the successful amplifications observed (Fig. S1B).
We then tried to find nod and nif genes by running BLAST searches of known Neorhizobium nod and nif genes against draft genome sequences, with negative results in all cases. Since nod and nif genes are harbored on large megaplasmids in N. galegae (Österman et al., 2014), we searched for similar plasmids in the genomes of Tomejil strains. All five genomes contained a set of repABC sequences characteristic of rhizobial plasmids Table 3 Presence of repABC regions in the genomes of Tomejil strains. DNA regions similar to the 3,628 bp region containing repABC genes from the symbiotic megaplasmid from N. galegae bv. orientalis HAMBI 540 were located in genome sequences by BLAST, extracted and compared by multiple alignment (ClustalW).

Genome
Number of (Cevallos, Cervantes-Rivera & Gutiérrez-Ríos, 2008;Pinto, Pappas & Winans, 2012), highly similar (85-86%) to the repABC cluster from the 1.8 Mb megaplasmid from N. galegae bv. orientalis (Table 3) and that are probably responsible for the replication a large plasmid in each of these strains. This is supported by the fact that repABC homologues are present in large contigs (83-523 kb) in the Tomejil strains, despite the draft level quality of their genome sequences (27-69 contigs, Table 2). Multiple sequence alignment of these repABC sequences showed that plasmids from the Tomejil strains were highly related, and separated from the symbiotic plasmids from N. galegae bv. orientalis and N. galegae bv. officinalis (Fig. 3). The completely sequenced strain N. galegae bv. officinalis HAMBI 1140 harbors an additional megaplasmid (175 kb) that appeared to be highly unrelated to both N. galegae symbiotic plasmids and plasmids from Tomejil strains (Table 3, Fig. 3). Finally, it was interesting to note that, among the Tomejil genomes, only strain T25_13 harbored a complete set of conjugal genes (tra and trb genes). In the remaining strains, these genes were absent altogether (T6_25, T20_22, T25_27) or incomplete (T7_12, where sequences similar to just trbG, trbI, and trbL were present) (Table S3). In order to ascertain the phylogenetic relationships among the sequenced Tomejil strains, and between this clade and the type strains of the three Neorhizobium species: N. galegae, N. alkalisoli, N. huautlense, we pulled out the complete sequences of the genes atpD, glnII, recA, rpoB and thrC from the genome assemblies. These genes had been used successfully to define the Neorhizobium genus and its species (Mousavi et al., 2014). Figure  4 shows a maximum-likelihood phylogenetic tree derived from a multiple alignment of the concatenated genes, rooted by using the Agrobacterium tumefaciens Ach5 sequences. There was very strong support for two clades within the Neorhizobium genus; one formed by N. alkalisoli and N. huautlense, and the other by N. galegae and the Tomejil strains. Four of the Tomejil strains (T7_12, T20_22, T25_13, T25_27) formed a well-supported group clearly separated from the fifth strain, T6_25, and all five, in turn, clearly separated from N. galegae. Genome-wide comparisons were carried out by calculating pairwise Average Nucleotide Identities (Richter & Rosselló-Mora, 2009;Richter et al., 2016) with the above genomes. Both ANIm and ANIb scores were calculated, and the ANIb matrix is shown in Table S4. A distance matrix was generated from the ANIb matrix using 100-ANIb (Chan et al., 2012) and Euclidean distances were calculated. These were represented in the dendrogram shown in Fig. 5. The dendrogram faithfully reproduced the topology of the multilocus phylogenetic tree (Fig. 4). Using a 95% ANI value as a widely accepted delimiter for genomic species (Richter & Rosselló-Mora, 2009), our results are consistent with Tomejil strains representing at least a clearly differentiated genospecies within Neorhizobium, and probably two, represented by strains T6_25 on one hand, and T7_12, T20_22, T25_13, and T25_27, on the other.

DISCUSSION
Direct isolation of Rhizobia from soil without resorting to trapping them inside their legume host is complicated by the fact that Rhizobial populations in soil can be small (Singleton & Tavares, 1986) and because of the paucity of selective characters that can be used. This is especially true for non-symbiotic variants lacking any symbiotic marker. As a result, very few studies have aimed at isolating non-symbiotic Rhizobia from soil. Groundbreaking work by Sullivan and Ronson with Mesorhizobium loti (Sullivan et al., 1995) did not only reveal the existence of an abundant population of non-symbiotic variants in the soil, but also showed that the genetic determinants for symbiosis could be readily transferred to these variants within the soil. Some of these non-symbiotic variants were further characterized and suggested to be representatives of species hitherto undescribed. Using a semi-selective culture medium, Tong & Sadowsky (1994) were able to enrich soybean-specific Rhizobia of the species Bradyrhizobium japonicum and B. elkanii. Using this medium, the same group later described that about half of the Bradyrhizobia isolated directly from Thai soils were non-symbiotic, although they did not characterize them further (Pongsilp et al., 2002). Louvrier, Laguerre & Amarger (1995) were also able to enrich members of the genus Rhizobium using a specific medium. Although they focused on symbiotic isolates (Louvrier, Laguerre & Amarger, 1996) they also isolated a fraction of non-symbiotic Rhizobia (G Laguerre, pers. comm., 2011). Our group had previously used the Rhizobium semi-selective medium, together with the fnrN gene marker, to isolate Rhizobia from soil (Jorrín, 2016) in order to study, at the genomic level, the legume host selection of specific R. leguminosarum genotypes (Jorrín & Imperial, 2015). This phenomenon had been described by the Amarger-Laguerre (Laguerre et al., 2003) and the Young (Young & Wexler, 1988) groups. During this study, we were able to isolate a number of non-symbiotic members of the genus Rhizobium (Jorrín, 2016;Jorrín & Imperial, 2015). Therefore, it was surprising that the same enrichment methodology resulted in the isolation of a large (over 50%) proportion of non-symbiotic members of the genus Neorhizobium from soil samples from the IFAPA Tomejil Experimental Station in Carmona (Seville, Spain), first because non-symbiotic Neorhizobium had not been previously isolated, and second because our genetic screening with fnrN should have left them behind (although the different size of the PCR amplified bands was already an indication that these were atypical members of the genus Rhizobium at best). The serendipitous reason why these particular Neorhizobium were chosen as putative Rhizobium has been presented above. Clearly, they grow well on MNBP semi-selective medium, with cultural characteristics similar to those of Rhizobium. However, it is possible that, among the two hundred and forty-three colonies that tested negative for fnrN amplification, other Neorhizobium that do not have a high enough conservation of the poly (3-hydroxybutyrate) depolymerase region exist in the Tomejil soil.
At any rate, Neorhizobium appears to be at least as abundant in the particular sampled soil as Rhizobium and, given the high rates of horizontal gene transfer that can take place in soil (Sullivan et al., 2002), they may become relevant in the establishment of symbioses with native legumes after receiving the appropriate symbiotic genes. Acquisition of these genes might be facilitated by the fact that N. galegae has been shown to harbor symbiotic genes on a megaplasmid (Radeva et al., 2001;Österman et al., 2014). In view of their size, nucleotide composition and conservation, it is possible that these plasmids are chromids, as defined by Harrison et al. (2010), although it has yet to be shown that they carry core genes. As shown above, all the sequenced Tomejil strains contain a set of repABC genes characteristic of plasmids from the Rhizobiales (Cevallos, Cervantes-Rivera & Gutiérrez-Ríos, 2008;Pinto, Pappas & Winans, 2012) in large contigs, suggesting that these strains also harbor a megaplasmid (or chromid) and are, thus, probably able to receive similar, large plasmids containing symbiotic genes.
A final consideration is whether the existence of populations of non-symbiotic Neorhizobium in soils reflects a normal situation in agricultural soils. This would require a large screening that is beyond the scope of this work. However, our preliminary studies suggest that in at least two other soils from Southern Spain that we have tested, members of Neorhizobium are not present in detectable numbers, and that the predominant isolates resulting from our semi-selective screening are Rhizobium spp. (A Soenens & J Imperial, 2017, unpublished data). This would then shift the question to why the Tomejil soil harbors a Neorhizobium population, a question that would require a better understanding of the ecology of this group in soil.

CONCLUSIONS
In conclusion, our work has allowed, for the first time, the isolation and identification of non-symbiotic members of the genus Neorhizobium from soil. Genotypic and genomic characterization of these isolates suggests that they are representatives of one, or perhaps two, new genospecies within Neorhizobium. It also suggests that soils harbor a large diversity of Rhizobia in the form of non-symbiotic variants that have traditionally escaped characterization and that may play an important role in the biology of these organisms.