Strains were maintained on solid rich medium LB (1% peptone, 0.5% yeast extract, 1% NaCl), supplemented with appropriate antibiotics. Liquid cultures for inoculation were grown in LB at 30°C with shaking at 200 rpm (Encarnación et al., 1995). Antibiotics were added as follows: nalidixic acid, 20 μg/mL; streptomycin, 200 μg/mL; and fosfomycin, 200 μg/mL.

P. vulgaris Negro Jamapa plantlets were inoculated with Sinorhizobium americanum strain CFNEI73 or CCGM7, both of which form nodules and fix nitrogen (Toledo et al., 2003; Mora et al., 2014). When the nodules were crushed and applied to LB plates, where symbiotic rhizobia normally do not grow, single colonies appeared after 72 h of incubation at 29°C. Three colonies from each spot were washed exhaustively in 10 mM MgSO₄/0.01% Tween 40 and plated on LB medium, followed by incubation for 72 h at 29°C. To purify the strains, bacteria were grown overnight in liquid LB, serially diluted in 10 mM MgSO₄-0.01% Tween 40, plated on LB plates, and incubated as above. Individual colonies were isolated from these plates.

Plasmids were visualized by the Eckhardt technique with modifications, as previously reported (Hynes and McGregor, 1990).

Total DNA from strains was extracted from cultures grown in liquid LB medium; the cells were centrifuged, resuspended in 50 mM Tris/20 mM EDTA pH 8.0, and then lysed with proteinase K (2.5 mg/mL) and 10% SDS. After treatment with a mix of 1:1 phenol-chloroform, DNA was precipitated with absolute ethanol. Genomes were sequenced by Macrogen (Seoul, South Korea). For each strain, libraries of 3 kilobase pairs (kb) were prepared and run on an Illumina HiSeq2000 to obtain 100–base pair (bp) mated pair reads. Total read counts were 13,969,230 and 17,424,244 for CCGM10 and CCGM11, respectively; filtered read counts were 6,951,885 and 8,690,792, with genome coverage of 121 × and 152 ×. Assembly was performed with SOAPdenovo2 (Luo et al., 2012). Annotation was performed with RAST v4.0 (Aziz et al., 2008) with manual curation. Functions were assigned using the extended annotation of clusters of orthologous groups (COG) of COG 2014 database update (Galperin et al., 2015). Phages were detected using the PHASTER server (Arndt et al., 2016).

The rpoB genes of strains CCGM10 and CCGM11 were sent to BLAST server (http://www.ncbi.nlm.nih.gov/Blast.cgi) to detect closely related strains. The genomes of selected strains were compared with genomes of CCGM10 and CCGM11 using Proteinortho version 5, with BLAST parameters E < 1E-5; 30% identity, and 70% coverage. The organisms selected for comparison were A. pusense IRBG74; A. pusense HPC(L), also known as R. pusense; and A. fabrum C58, also designated A. tumefaciens.

The genomes of A. pusense strains CCGM10 and CCGM11 were registered at GenBank with the provisional accessions LNZW00000000 and MAPG00000000, respectively.

The main characteristics of the genomes of A. pusense strains CCGM10 and CCGM11 are listed in Table 1. The similarity between the strains was remarkable, with 93 and 92% of shared orthologs in the circular and linear chromosomes, respectively (Supplementary Table 1). Interestingly, some portions of the chromosomes were not shared with strain IRBG74; these segments mainly corresponded to genes with hypothetical functions (Supplementary Table 2). The sequence identity of shared orthologs in both strains was surprisingly high, 99.97%, and 5096 products had 100% identity at the amino acid level. Only about 142–205 genes in each chromosome were strain-specific. In addition, the plasmids were almost the same size (Figure 1A) and exhibited very high proportion of shared orthologs, ranging from 82 to 95%. These features indicate that the strains diverged very recently, mainly by insertion or deletion of genes.

A structural analysis of the genomes of these strains revealed very high conservation of both the linear and circular chromosomes, as shown in Figure 1B. The genomes were compared with that of the strains' closest relative, A. pusense IRBG74. The chromosomes were highly conserved, but the symbiotic plasmid of strain IRBG74 was not shared (not shown), and some regions were exclusive to the chromosomes of the reference strain. The genomes of A. pusense HPC(L) and that of A. fabrum C58, the model organism of the genus, were also included in the comparison.

Homology analysis of plasmid genes revealed that many had homologs in rhizobia and agrobacteria, but also with more distant organisms (Figure 1C). Most of genes in plasmid E were related to those from Agrobacterium and Rhizobium. Plasmid D harbored numerous genes with homologs in Roseomonas mucosa (with relatives isolated from human blood; Han et al., 2003) and Ochrobactrum anthropi, a member of the Brucellaceae that is found in natural ecosystems but also able to infect humans (Daxboeck et al., 2002). More than half of the genes on plasmid C were related to those from Rhizobium sp. H41, isolated from eroding rocks (Xi et al., 2014). Plasmid B, which is absent from strain CCGM11, is an intact temperate prophage of about 37 kb, harboring 47 genes, some shared with a prophage present in the circular chromosome of A. tumefaciens C58 and related to phage AmM-1 from Aurantimonas sp., a psychrotolerant bacterium belonging to the order Rhizobiales found in the deep ocean (Goldstein et al., 1982; Yoshida et al., 2015). Some genes assigned to plasmid A had homologs in Rhizobium sp. H41.

In light of our detection of the prophage in plasmid B, we analyzed the remaining replicons and identified additional intact prophages (Supplementary Table 3). The circular chromosome of strains CCGM10 and CCGM11 harbored three prophages: two related to phages RR1 and 16-3, also present in rhizobia, and vB_PmaS from the marine bacterium Paracoccus marcusii. In the putative linear chromosome of both strains, we found an intact prophage related to RC1 from Rhodobacter capsulatus. All prophages belonged to the family Caudovirales.

Next, we analyzed the functions of genes in the plasmids of strain CCGM10 (Table 1 and Supplementary Table 4). Plasmid E was enriched in genes encoding transcriptional regulators and factors involved in signal transduction and amino acid or ion transport. Among these genes were those encoding proteins related to DNA uptake and folding (as HU), dnaE and ligD; central metabolism (e.g., fabG, ilvB, and arcCD), and ubiquinone synthesis. Plasmid D was especially enriched with genes involved in transcription, replication, and intracellular trafficking. Other notable genes encoded factors associated with secretion system IV, chemotaxis (cheBRW and methyl-accepting proteins mcp), transcriptional regulators and sensors, and genes for the redox response involving flavodoxin and glutathione. In addition, the plasmid contained a 500 bp segment homologous to the T-DNA segment of plasmid pTi.

Plasmid C was rich in genes involved in trafficking and ion transport, and it also harbored genes related to defense (toxin-antitoxin pairs) and central metabolism, such as arcACD, argF, speD, and the universal stress protein uspA. Plasmid A contained some genes involved in chemotaxis and glycerol utilization. In addition, we identified clusters of genes involved in conjugal plasmid transfer and replication (repABC family) in plasmids pC, pD, and pE.

In the circular chromosome, we identified the following genes involved in nitrogen fixation: fixGHS, fixNOQP, nifURS, nodLN, and fixR. A. fabrum C58 harbored homologs of these genes. However, some other genes not present in the C58 genome, such as the fixLJK cluster, were present in plasmid D. FixLJ are transcriptional regulators of nifA, the general activator of nitrogen fixation; FixK is regulator of the fixNOQP operon (encoding Cbb3, the cytochrome used by symbiotic rhizobia for respiration inside the nodules). These genes may represent the remnant of a complete gene cluster related to nitrogen fixation.

Several clues point to bean seeds as the origin of this pair of A. pusense strains, which were isolated from nodules but lack genes for nodulation. First, we have isolated 10 other rhizobial strains from nodules of non-inoculated plants; seeds from plants inoculated with two selected strains, CCGM1 or CCGM7, yielded positive signals for specific rhizobial gene markers (Mora et al., 2014). Second, some of the seed rhizobia strains did not have nodulation genes, but they are nonetheless able to inhabit nodules. Third, seed rhizobia have been identified as pairs of highly related strains differing in only one plasmid, as in this A. pusense pair. No A. pusense strains have been used previously in our greenhouse. It is possible that the A. pusense strains originally harbored a symbiotic plasmid that has been lost, but is still present in close relatives such as IRBG74. As with other agrobacteria, it will be of interest to define the precise symbiotic role of our isolates (Mrabet et al., 2006).

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.