Horizontal Transfer of Symbiosis Genes within and Between Rhizobial Genera: Occurrence and Importance

Rhizobial symbiosis genes are often carried on symbiotic islands or plasmids that can be transferred (horizontal transfer) between different bacterial species. Symbiosis genes involved in horizontal transfer have different phylogenies with respect to the core genome of their ‘host’. Here, the literature on legume–rhizobium symbioses in field soils was reviewed, and cases of phylogenetic incongruence between rhizobium core and symbiosis genes were collated. The occurrence and importance of horizontal transfer of rhizobial symbiosis genes within and between bacterial genera were assessed. Horizontal transfer of symbiosis genes between rhizobial strains is of common occurrence, is widespread geographically, is not restricted to specific rhizobial genera, and occurs within and between rhizobial genera. The transfer of symbiosis genes to bacteria adapted to local soil conditions can allow these bacteria to become rhizobial symbionts of previously incompatible legumes growing in these soils. This, in turn, will have consequences for the growth, life history, and biogeography of the legume species involved, which provides a critical ecological link connecting the horizontal transfer of symbiosis genes between rhizobial bacteria in the soil to the above-ground floral biodiversity and vegetation community structure.


Introduction
Approximately 70% of the ca. 19,300 species in the Fabaceae (Leguminosae, the legume family) can fix atmospheric nitrogen (N 2 ) via symbiotic bacteria (general term 'rhizobia') in root nodules [1,2]. Rhizobia reduce atmospheric N 2 to ammonia (NH 3 ) through the enzyme nitrogenase, and this NH 3 , as ammonium (NH 4 + ), is transported to plant cells where it is assimilated into amino acids via the glutamine synthetase/glutamate synthase (GS/GOGAT) pathway [3,4]. The ability to fix N 2 can give legumes an advantage under low soil nitrogen (N) conditions if other factors are favourable for growth [5,6]. Also, legume N 2 fixation can be a major input of N into a wide range of natural and agricultural ecosystems [7][8][9][10].
searching the Institute for Scientific Information (ISI) Web of Science, using each legume genus known to nodulate partnered with each of the rhizobia genera, and each of the rhizobia genera partnered with 'horizontal gene transfer' and 'lateral gene transfer' as key words. Further searches were carried out on the literature quoted in the selected papers and on those papers listed as quoting the selected papers in ISI Web of Science. Only data for plants sampled under field conditions, for plants grown in soils taken from the field, or for plants supplied field soil extracts, were used. Bacteria isolated from legume nodules were accepted as rhizobia on the basis of the criteria described previously [17]. All cases of 'authenticated' rhizobia for which core and symbiosis gene sequences were presented were studied further, and the cases for which the authors considered there was incongruence between core and symbiosis gene sequences are discussed. The core and symbiosis genes that showed incongruence are given in Tables. Representative data are presented for Glycine max and Phaseolus vulgaris because of the large number of publications reporting incongruence between core and symbiosis gene sequences for these two species. In some studies, incongruence was tested statistically but, in most cases, this was determined from a visual assessment of phylogenetic trees of the two sets of genes. For example, Mesorhizobium strains isolated from New Zealand endemic Sophora spp. had diverse concatenated (gene sequences aligned head to tail) glnII-recA-rpoB gene sequences but similar concatenated nodA-nodC gene sequences (Figure 1), and, here, the housekeeping and symbiosis genes were considered incongruent [17]. Subsequent statistical analysis of these sequences indicated that the housekeeping and symbiosis genes were incongruent. Comparative maximum likelihood phylogenetic analysis using housekeeping and symbiotic gene clusters from Mesorhizobium strains isolated from New Zealand endemic Sophora spp. Sequence alignment, alignment editing, and phylogenetic analysis were performed using MEGA7 [24]. The phylogenetic trees were built using the GTR model with G + I substitutions for the housekeeping genes (1083 bp) and Tamura 3-parameter model with G + I substitutions for the symbiosis genes (869 bp). The possibility of concatenation was investigated using the partition-homogeneity test with PAUP [25,26]. Each concatenation was investigated for 1000 replicates. All housekeeping genes were congruent with each other (p = 0.015), and both symbiosis genes were congruent with each other (p = 0.02). The housekeeping genes were shown not to be congruent with the symbiosis genes (p = 0.001). Bootstrap values after 500 replicates are expressed as percentages; values less than 50% are not shown.  Comparative maximum likelihood phylogenetic analysis using housekeeping and symbiotic gene clusters from Mesorhizobium strains isolated from New Zealand endemic Sophora spp. Sequence alignment, alignment editing, and phylogenetic analysis were performed using MEGA7 [24]. The phylogenetic trees were built using the GTR model with G + I substitutions for the housekeeping genes (1083 bp) and Tamura 3-parameter model with G + I substitutions for the symbiosis genes (869 bp). The possibility of concatenation was investigated using the partition-homogeneity test with PAUP [25,26]. Each concatenation was investigated for 1000 replicates. All housekeeping genes were congruent with each other (p = 0.015), and both symbiosis genes were congruent with each other (p = 0.02). The housekeeping genes were shown not to be congruent with the symbiosis genes (p = 0.001). Bootstrap values after 500 replicates are expressed as percentages; values less than 50% are not shown. The scale bar indicates the fraction of substitutions per site. M: Mesorhizobium, S: Sophora, FS: Field Site.

Rhizobia Associated with the Caesalpiniodeae
There are reports of phylogenetic incongruence between core and symbiosis genes for Ensifer, Cupriavidus, Burkholderia, Rhizobium, and Devosia associated with legumes in the Mimosoid clade of the sub-family Caesalpinioideae (Table 1). Table 1. Reported cases of phylogenetic incongruence between core and symbiosis genes for rhizobia associated with legumes in the sub-family Caesalpinioideae. All species have indeterminate nodules.
Five separate studies indicated that horizontal transfer of symbiosis genes had occurred between Ensifer spp. associated with species in the Caesalpinioideae [27,28,[34][35][36]. For Ensifer associated with the Mexican native Leucaena leucocephala sampled in Panxi, China, gene sequences indicated that symbiotic genes of strains associated with introduced plants were transferred into indigenous strains in the soil [28]. Ensifer isolates from Vachellia macracantha sampled within the plant's native range in Peru showed nifH and nodC sequences closely related to other American rhizobial strains, which adds support to the use of symbiotic genes as valuable indicators of geographical origin [35]. Gene sequences indicated diverse origins for the housekeeping genes nifH and nodA for seven Ensifer isolates representative of 73 isolates from Vachellia jacquemontii sampled within its native range in the Thar Desert of India [34]. The authors suggested that the stressful desert conditions, and stressful conditions in general, may favour frequent horizontal gene transfer. Alternatively, rather than promote its occurrence, stressful environments may represent a situation where the positive consequences of horizontal gene transfer in terms of natural selection are more significant, and thus horizontal gene transfer becomes more apparent within the rhizobial population.
Ten Cupriavidus rhizobia strains isolated from five Mimosa spp. in southern Uruguay showed symbiosis and housekeeping gene sequence phylogenies that were not congruent [29]. The strains separated into two groups of five strains on their 16S rRNA sequences, and one strain selected from each of these groups differed substantially in its recA and gyrB sequences. However, both the nodA and nifH sequences for the ten strains grouped together in a cluster. Also, Rhizobium altiplani Br 10423 T isolated from Mimosa pudica in Distrito Federal in central Brazil had nifH and nodC sequences closely related (identical for nodC) to those of Rhizobium mesoamericanum CCGE 501 T [32].
Mimosa is a genus of ca. 550 species native to the Americas, South Asia, and Africa including Madagascar [1,2]. The evidence indicates that Cupriavidus is the main rhizobial symbiont of endemic Mimosa in southern Uruguay, but Burkholderia and Rhizobium/Ensifer are the main symbionts of Mimosa in central and southern Brazil and central Mexico, respectively [29,[37][38][39]. In contrast with findings for Mimosa Cupriavidus symbionts in Uruguay, the symbiosis gene sequences for Burkholderia in Brazil and Rhizobium/Ensifer in Mexico were largely congruent with their respective 16S rRNA and housekeeping gene sequences [37,38]. This indicates that these symbiosis genes diverged over a long period within Burkholderia and Rhizobium/Ensifer without substantial horizontal transfer between species. Similarly, it was concluded that nodC and nifH sequences for Burkholderia isolated from the Piptadenia group (Piptadenia, Parapiptadenia, Pseudopiptadenia, Pityrocarpa, Anadenanthera, and Microlobius) (Mimoseae) have evolved mainly through vertical transfer, with rare occurrence of horizontal transfer [40]. Outside South America, Burkholderia strains isolated from Mimosa diplotricha and M. pudica in Yunan province in subtropical China showed diverse 16S rRNA sequences but grouped together, along with Burkholderia phymatum STM815 T , on nodA sequences [31]. Also, Burkholderia caribensis TJ182 isolated from the invasive M. diplotricha in Taiwan and characterized on 16S rRNA sequence, grouped with Cupriavidus strains on nodA sequence, indicating that the nodA gene had been transferred from Cupriavidus to Burkholderia [30]. However, this transfer was not confirmed [41]. In both studies, it was concluded that it was likely that the Burkholderia 'travelled' with their invasive hosts from South America to South East Asia [30,31].
A more extreme case of incongruence between core and symbiosis genes was reported from India for Devosia natans isolated from the aquatic 'water mimosa' Neptunia natans. Here, the 16S rDNA sequences indicated that two strains isolated from N. natans were Devosia, but the nifH and nodD sequences were most closely related to those of Rhizobium tropici CIAT899 T [33]. This finding indicates that the horizontal transfer of symbiosis genes has occurred across genera from Rhizobium to Devosia at some stage.

The Inverted Repeat-Lacking Clade (IRLC)
Ensifer, Mesorhizobium, and Rhizobium are the main rhizobial symbionts of legumes in the IRLC, and there are several examples of phylogenetic incongruence between core and symbiosis genes for Mesorhizobium and Rhizobium associated with IRLC species (Table 2). Table 2. Reported cases of phylogenetic incongruence between core and symbiosis genes for rhizobia associated with legumes in the inverted repeat-lacking clade (IRLC) of the legume sub-family Papilionoideae. All species have indeterminate nodules.
Finally, within the IRLC, the pasture legume Biserrula pelecinus was introduced into western Australia from the Mediterranean region in 1994 and, as indigenous rhizobial populations in western Australia do not nodulate this legume, the seed was inoculated with M. ciceri sv biserrulae strain WSM1271 [63,64]. In 2000, Mesorhizobium strains, including WSM2073 and WSM2075 with 16S rRNA, dnaK and GS11 phylogenies different from strain WSM1271, were isolated from nodules of B. pelecinus grown in Western Australia and shown to nodulate the legume, although the bacteria were largely ineffective with regard to N 2 fixation [63,64]. Where tested, these strains had identical sequences for the symbiosis insertion regions with WSM1271, indicating that they had obtained their symbiosis genes via horizontal transfer of a symbiosis island from the inoculant within the space of six years. This quick transfer in the field was not a single event, as it was also demonstrated for commercially grown B. pelecinus inoculated with WSM1497 and resulted in three different Mesorhizobium lineages with identical symbiosis genes to the inoculant [64]. The ability of rhizobial strains that do not fix N 2 to produce nodules on B. pelecinus could result in decreased yield of the crop.

Ammopiptanthus nanus, Ammopiptanthus mongolicus
Ensifer arboris and Neorhizobium galegeae characterized on 16S rRNA and concatenated recA-atpD-rpoB-thrC sequences aligned with Ensifer meliloti ATCC9930 T on nifH and nodC sequences [88]. Phyllobacterium giardinii characterized on 16S rRNA and concatenated recA-atpD-rpoB-thrC sequences aligned with R. leguminosarum sv. viciae USDA 2370 T on nifH and nodC sequences [88]. Rhizobium/Agrobacterium radiobacter characterized on 16S rRNA and concatenated recA-atpD-rpoB-thrC sequences aligned with E. fredii USDA205 T on nifH and nodC sequences [88] Anagyris latifolia Mesorhizobium isolates with diverse 16S-23S rDNA ITS, 16S rRNA and glnII sequences had identical nodC sequences closely related to Mesorhizobium tianshanense USDA 3592 T [89] Four studies carried out on species within the Crotalarieae found evidence of horizontal transfer of symbiosis genes between different genera of rhizobia. Rhizobium isolated from Aspalathus sp. grown in the Cape Fynbos biome in South Africa and characterized on 16S rRNA and housekeeping gene sequences had nifH and nodA,B,C sequences closely related to those of Mesorhizobium [65]. Methylobacterium nodulans ORS2060 T isolated from Crotalaria podocarpa in Senegal, grouped with Bradyrhizobium spp. on nodA sequences [66]. Microvirga lotonidis WSM3557 T and Microvirga zambiensis WSM3693 T isolated from Listia angolensis in Zambia had identical nodA sequences which clustered with strains of Bradyrhizobium, Burkholderia, and Methylobacterium [67]. Burkholderia, isolated from Rafnia triflora in the Core Cape subregion of South Africa and characterized on 16S rRNA and housekeeping gene sequences, had a nifH sequence closely related to those of Ensifer spp. [68].
Lemaire and co-workers specifically assessed the degree of horizontal transfer of nodulation genes within rhizobia genera of a range of legumes endemic to the Cape Fynbos biome in South Africa [65]. It was concluded that Mesorhizobium strains isolated from Aspalathus spp., Argyrolobium spp. (Genisteae, Table 3), Otholobium spp., and Psoralea spp. (Psoraleeae, Table 4), and Burkholderia isolated from Podalyria calyptrata ( Table 3) show high degrees of horizontal transfer of nodulation genes among closely related species. In associated studies, a Mesorhizobium isolate from Psoralea sp., characterized on 16S rRNA and housekeeping gene sequences, aligned closely to Ensifer on nifH sequence, and a Mesorhizobium isolate from Psoralea oligophylla aligned closely to Burkholderia on nodA sequence (Table 4) [68]. Also, for Burkholderia isolated from P. calyptrata, different branching patterns were found among numerous isolates for recA and nodA phylogenies [83]. In a separate study of Burkholderia isolates from Hypocalyptus spp. (Hypocalypteae) and Cyclopia spp., P. calyptrata and Virgilia oroboides (Podalyrieae) sampled in the Cape Floristic Region of South Africa, phylogenies inferred from nifH and nodA sequences were incongruent and, generally, phylogenies inferred from nifH and nodA sequences were incongruent with those from 16S rRNA and recA sequences [79]. These findings confirm that horizontal transfer of symbiosis genes is common in South African Burkholderia, which is in contrast with findings for South American Burkholderia rhizobial symbionts [37,40].
There is also strong evidence that horizontal transfer of symbiosis genes has occurred between different Bradyrhizobium spp. associated with Cytisus spp. [69][70][71], Genista versicolor [72] and Lupinus spp. [73][74][75]77,78] in the tribe Genisteae (Table 3). In particular, most Bradyrhizobium isolates from native Lupinus spp. (and native Genisteae species in general [90]) in Europe form a distinct lineage, 'Clade II', on the basis of their nodA gene sequences [75,81]. Also, Bradyrhizobium isolates from native Cytisus villosus in Morocco had diverse 16S rRNA and housekeeping gene sequences, but all showed similar nifH and nodC sequences which were closely related to those of Bradyrhizobium japonicum sv. genistearum [71]. These findings indicate that horizontal transfer of symbiosis genes has played a role in the development of specific relationships between Genisteae spp. and Bradyrhizobium spp. over wide areas. Bradyrhizobium isolated from invasive Cytisus scoparius in the United States had housekeeping genes similar to indigenous Bradyrhizobium, but their nodC, nifD, and nifH sequences were highly similar or identical to those of a Bradyrhizobium strain from Spain [70]. It appears, therefore, that indigenous North American Bradyrhizobium had acquired symbiosis genes from Bradyrhizobium symbionts of European C. scoparius via horizontal gene transfer.
Microvirga and Ochrobactrum, which are rare as rhizobial symbionts, can nodulate specific Lupinus spp. [17,67,76]. The nifD and nifH sequence for Microvirga lupini Lut6 T isolated from Lupinus texensis aligned closely to Rhizobium etli CFN 42 T , while its nodA sequence was placed in a clade that contained strains of Rhizobium, Mesorhizobium, and Ensifer [67]. The nifH sequence for Ochrobactrum lupini LUP21 T isolated from Lupinus honoratus in Argentina showed 99.6% similarity to two Mesorhizobium strains, while its nodD sequence showed 86.4% similarity to R. etli CFN 42 T [76]. Thus, evidence is strong that M. lupini Lut6 T and O. lupini LUP21 T obtained their symbiosis genes from other, more common rhizobial genera.
Incongruence between housekeeping and symbiosis genes indicate that horizontal gene transfer of symbiosis genes has occurred between Mesorhizobium strains from Coronilla varia (Loteae) grown in Shaanxi province, China [80], Bradyrhizobium isolates from Ornithopus spp. (Loteae) sampled in Europe and western Australia [75,81], and Ensifer strains from Tephrosia spp. (Millettieae) in the Indian Thar Desert [82]. New Zealand endemic Sophora spp. are nodulated by diverse Mesorhizobium spp. with similar symbiosis genes ( Figure 1) [85,86,91,92]. Generally, Mesorhizobium isolates from the same field site grouped together on housekeeping gene sequences [85,86]. This apparent link between housekeeping gene sequences and field site in association with almost identical symbiosis genes is consistent with the proposal that horizontal transfer of symbiosis genes to Mesorhizobium strains adapted to local soil conditions has occurred, but this requires further testing. The relationship between New Zealand endemic Sophora species and Mesorhizobium with particular symbiosis genes is highly specific and contrasts with findings for Sophora alopecuroides and Sophora. flavescens sampled in China, which are nodulated by Ensifer, Mesorhizobium, Phyllobacterium, and Rhizobium with a wide range of nodC and nifH gene sequences [87,93]. The data indicate that symbiosis genes of rhizobia associated with S. alopecuroides and S. flavescens are primarily maintained by vertical transfer, but there is also evidence for occasional horizontal gene transfer of symbiosis genes within and between rhizobia genera associated with these legume species [87,93].
Horizontal transfer of symbiosis genes has occurred across genera in the case of Agrobacterium sp. strain IRBG74 originally isolated from the aquatic legume Sesbania cannabina and subsequently shown to effectively nodulate Sesbania sesban and seven other Sesbania species [21]. Housekeeping gene sequences identified strain IRBG74 as a close relative of the plant pathogen Agrobacterium radiobacter (=Agrobacterium tumefaciens). However, it did not contain vir genes but harboured a sym-plasmid containing nifH and nodA genes with sequences similar to those of Ensifer spp. which nodulate S. cannabina. In a separate study, Rhizobium/Agrobacterium and Ensifer isolates from S. cannabina with diverse housekeeping gene sequences had similar nifH and nodA sequences, and one Rhizobium/ Agrobacterium strain showed highly similar nifH and nodA sequences to strain IRBG74 [84].
Within the Thermopsideae, Ammopiptanthus nanus and Ammopiptanthus mongolicus sampled across nine sites in three regions of China were nodulated by Ensifer, Neorhizobium, Pararhizobium, and Rhizobium [88]. For strains characterized to species level on 16S rRNA and housekeeping gene sequences, Ensifer arboris and Neorhizobium galegeae strains aligned with Ensifer meliloti ATCC9930 T on nifH and nodC gene sequences, Phyllobacterium giardinii strains aligned with R. leguminosarum sv. viciae USDA 2370 T on nifH and nodC gene sequences, and R./A. radiobacter strains aligned with E. fredii USDA 205 T on nifH and nodC gene sequences [88]. For Anagyris latifolia grown in soil samples collected from within natural populations of the legume growing in the Canary Islands, Mesorhizobium isolates with diverse 16S-23S rDNA ITS, 16S rRNA and glnII gene sequences had identical nodC sequences closely related to Mesorhizobium tianshanense USDA 3592 T [89].
For P. vulgaris, strains of R. leguminosarum sv. phaseoli, R. gallicum, and Pararhizobium giardinii, isolated from plants grown in France, and R. etli, isolated in Mexico and Belize, had highly similar nodC sequences, and a strain characterized as Rhizobium on its 16S rRNA sequence aligned with E. meliloti on its nodC sequence [103]. R. lusitanum P1-7T isolated from P. vulgaris in Portugal had nifH and nodD sequences similar to R. tropici CIAT 899 T and D. neptuniae LMG 21357 T [104]. In a related study, a Pararhizobium giardinii strain characterized on 16S rRNA and housekeeping gene sequences aligned with Ensifer spp. on nodC sequence [107].
Data are presented for three Vigna crop species (Table 4). For Vigna. angularis (adzuki bean), grown in the sub-tropical region of China, Bradyrhizobium, and E. fredii were major, and Rhizobium, Mesorhizobium, and Ochrobactrum were minor rhizobial symbionts [108]. Here, 16S rRNA, housekeeping and nodC gene phylogenies were congruent, except that one Rhizobium strain aligned with E. fredii strains on nodC sequences. In a related study on Bradyrhizobium isolates from V. radiata (mungbean) and V. unguiculata grown in subtropical China, 16S rRNA, housekeeping and nodC gene phylogenies were mainly congruent [116]. Similarly, for V. radiata grown in three agro-ecological regions in Nepal, Bradyrhizobium 16S-23S RNA IGS, 16S rRNA, nodA, nodD1, and nifD sequences were mainly congruent, but five Bradyrhizobium strains aligned with E. meliloti on nodA sequences [109]. In contrast, overall phylogenies for core and nodulation genes for Bradyrhizobium, isolated from V. unguiculata at a range of sites in Botswana and Roodeplaat, South Africa, were incongruent [110]. It was concluded that horizontal gene transfer has significantly influenced the evolution of V. unguiculata root-nodule bacteria in these African countries [110]. Also, for M. vignae BR3299 T isolated from V. unguiculata in north-east Brazil, the nifH gene sequence aligned with Mesorhizobium and Rhizobium strains, which supports the proposal that M. vignae BR3299 T obtained its nifH gene via horizontal transfer from Rhizobium [111].
Examples of horizontal transfer of symbiosis genes within and between rhizobial genera were also found for non-crop species with determinate nodules. For Desmodium spp. (Desmodieae) growing in Panxi, Sichuan, China, nodC sequences for one Pararhizobium, one Rhizobium/Agrobacterium, and two Rhizobium isolates characterized on 16S rRNA and housekeeping gene sequences were identical to those for Ensifer strains isolated from L. leucocephala in China (Panxi) and Brazil [96]. It was concluded that horizontal transfer of nodC genes had occurred between the different genera, but that further work was required to determine the direction of gene transfer [96]. Evidence was also found for lateral transfer of symbiosis genes between rhizobial genera associated with G. soja ('wild soybean'). Here, Ensifer and Rhizobium isolates characterized on 16S rRNA and housekeeping gene sequences formed a single Ensifer lineage on nifH and nodA sequences [102].
Across two studies, diverse Mesorhizobium isolated from seven Lotus species in the Canary Islands and characterized on 16S rRNA, atpD, and recA gene sequences clustered together close to M. loti on nodC sequences [113,114]. Similarly, diverse Mesorhizobium (16S rRNA) isolated from L. tenuis grown in three soils of the Salado River Basin Buenos Aires Province, Argentina, clustered together in a large clade on nifH and nodC sequences, again close to M. loti [115]. An Aminobacter (Phyllobacteriaceae) strain was also reported to nodulate L. tenuis and have a nodC sequence similar to the Mesorhizobium strains. This is the first report of an Aminobacter rhizobial strain, however, and requires verification. Similarly, Geobacillus (Phylum Firmicutes), Paenibacillus (Firmicutes), and Rhodococcus (Actinobacteria) were reported as rhizobial symbionts of L. corniculatus [117]. It was stated that these bacterial species had similar nodA gene sequences to Mesorhizobium isolates from the same plants and that they had obtained their nodA gene via horizontal transfer from the Mesorhizobium. This report also needs to be verified using authentification experiments and whole-genome sequencing.

Recombination of Symbiotic Islands
Incongruence between rhizobium core and symbiosis genes is strong evidence that horizontal transfer of symbiosis genes has occurred between strains, but it tells little about when the transfer took place. However, there are several cases linked to rhizobia associated with invasive species or use of rhizobial inoculum on crop species where the transfer of symbiosis genes between rhizobial species has been monitored. Here, we briefly consider the recombination of symbiotic islands and focus on two studies in which symbiosis genes were shown to transfer between a Mesorhizobium strain used as a crop inoculant and indigenous soil Mesorhizobium. The structure and physiology of R. leguminosarum symbiotic plasmids is considered in a separate paper in this Special Issue of Genes [118].
Firstly, in an important early study, the chromosomal symbiotic element of M. loti strain ICMP 3153 used as inoculum on L. corniculatus in New Zealand was shown to have transferred to indigenous Mesorhizobium strains and was fully functional in the recipient strains [19,20]. Subsequent work on M. loti strain R7A, a derivative of strain ICMP 3153, has shown that the symbiosis island is a single 502 kb Integrative Conjugative Element (ICEMlSym R7A ) which integrates into a phenylalanine transfer RNA (phe-tRNA) gene [119][120][121] (Figure 2). The recombination reaction is driven by the recombination directionality factor S (RdfS) on the attachment sites attL S and attR S , resulting in the excised ICEMISym R7A . The attB S and attP S sites facilitate the reintegration of the ICE into the chromosome, driven by integration factor S (IntS). In a separate study, the full genome sequence was reported for M. ciceri strain CC1192, which is the sole inoculant used on C. arietinum in Australia [122]. The nod, nif, and fix genes appear to be located on a 419 kb symbiosis island integrated within the chromosome of strain CC1192. This symbiosis island shows a similar structure to that of M. loti R7A strain. Rhodococcus (Actinobacteria) were reported as rhizobial symbionts of L. corniculatus [117]. It was stated that these bacterial species had similar nodA gene sequences to Mesorhizobium isolates from the same plants and that they had obtained their nodA gene via horizontal transfer from the Mesorhizobium. This report also needs to be verified using authentification experiments and wholegenome sequencing.

Recombination of Symbiotic Islands
Incongruence between rhizobium core and symbiosis genes is strong evidence that horizontal transfer of symbiosis genes has occurred between strains, but it tells little about when the transfer took place. However, there are several cases linked to rhizobia associated with invasive species or use of rhizobial inoculum on crop species where the transfer of symbiosis genes between rhizobial species has been monitored. Here, we briefly consider the recombination of symbiotic islands and focus on two studies in which symbiosis genes were shown to transfer between a Mesorhizobium strain used as a crop inoculant and indigenous soil Mesorhizobium. The structure and physiology of R. leguminosarum symbiotic plasmids is considered in a separate paper in this Special Issue of Genes [118].
Firstly, in an important early study, the chromosomal symbiotic element of M. loti strain ICMP 3153 used as inoculum on L. corniculatus in New Zealand was shown to have transferred to indigenous Mesorhizobium strains and was fully functional in the recipient strains [19,20]. Subsequent work on M. loti strain R7A, a derivative of strain ICMP 3153, has shown that the symbiosis island is a single 502 kb Integrative Conjugative Element (ICEMlSym R7A ) which integrates into a phenylalanine transfer RNA (phe-tRNA) gene [119][120][121] (Figure 2). The recombination reaction is driven by the recombination directionality factor S (RdfS) on the attachment sites attLS and attRS, resulting in the excised ICEMISym R7A . The attBS and attPS sites facilitate the reintegration of the ICE into the chromosome, driven by integration factor S (IntS). In a separate study, the full genome sequence was reported for M. ciceri strain CC1192, which is the sole inoculant used on C. arietinum in Australia [122]. The nod, nif, and fix genes appear to be located on a 419 kb symbiosis island integrated within the chromosome of strain CC1192. This symbiosis island shows a similar structure to that of M. loti R7A strain.  [20,[119][120][121]. The recombination is initiated by IntS (integration factor S) at the attachment sites attB and attP to integrate the ICEMlSym R7A in the chromosome and produce the attachment sites attL and attR. The recombination directionality factor S (RdfS) together with IntS stimulates excision of the ICE and forms the attachment sites attP and attB. phe-tRNA = phenylalanine transfer RNA. The ICEMlSym R7A is coloured blue, and the remaining chromosome is coloured grey. The schematic diagram is not drawn to scale.  [20,[119][120][121]. The recombination is initiated by IntS (integration factor S) at the attachment sites attB and attP to integrate the ICEMlSym R7A in the chromosome and produce the attachment sites attL and attR. The recombination directionality factor S (RdfS) together with IntS stimulates excision of the ICE and forms the attachment sites attP and attB. phe-tRNA = phenylalanine transfer RNA. The ICEMlSym R7A is coloured blue, and the remaining chromosome is coloured grey. The schematic diagram is not drawn to scale.
The second example is the chromosomal symbiotic element of M. ciceri sv biserrula strain WSM1271, which was used as inoculum on B. pelecinus in western Australia. Again, the symbiotic island was shown to have been transferred from the inoculum to indigenous Mesorhizobium strains, but, here, the recipient strains were ineffective or poorly effective on N 2 fixation of the host plant [63,64]. In this case, the symbiosis ICE exists as three separate chromosomal regions when integrated in their host (alpha, beta, and gamma in Figure 3) [121,123]. These regions occupy three different recombinase attachment sites that do not excise independently but recombine in the host chromosome to form a single contiguous region prior to excision and conjugative transfer. Nine additional tripartite ICEs were identified in diverse mesorhizobia, and transfer was demonstrated for three of them [123]. Single-part ICE and tripartite ICEs appear to be widespread in the Mesorhizobium genus [122,123]. The second example is the chromosomal symbiotic element of M. ciceri sv biserrula strain WSM1271, which was used as inoculum on B. pelecinus in western Australia. Again, the symbiotic island was shown to have been transferred from the inoculum to indigenous Mesorhizobium strains, but, here, the recipient strains were ineffective or poorly effective on N2 fixation of the host plant [63,64]. In this case, the symbiosis ICE exists as three separate chromosomal regions when integrated in their host (alpha, beta, and gamma in Figure 3) [121,123]. These regions occupy three different recombinase attachment sites that do not excise independently but recombine in the host chromosome to form a single contiguous region prior to excision and conjugative transfer. Nine additional tripartite ICEs were identified in diverse mesorhizobia, and transfer was demonstrated for three of them [123]. Single-part ICE and tripartite ICEs appear to be widespread in the Mesorhizobium genus [122,123]. . Comparative genome analysis using the circular BLASTN alignment in BRIG (BLAST Ring Image Generator, [124] of WSM2073 and WSM2075 against WSM1271. The three ICEMcSym 1271 regions are indicated and summarised from Haskett et al. [123]. It highlights the transfer of the symbiosis island from the Biserrula pelecinus inoculant strain WSM1271 to non-symbiotic recipient strains that turn into poorly effective symbionts. ICEMcSym 1271 consists of three separate regions (alpha, beta, and gamma) when integrated in the chromosome but excises as one circular plasmid and re-integrates in the recipient chromosome [123]. . Comparative genome analysis using the circular BLASTN alignment in BRIG (BLAST Ring Image Generator, [124] of WSM2073 and WSM2075 against WSM1271. The three ICEMcSym 1271 regions are indicated and summarised from Haskett et al. [123]. It highlights the transfer of the symbiosis island from the Biserrula pelecinus inoculant strain WSM1271 to non-symbiotic recipient strains that turn into poorly effective symbionts. ICEMcSym 1271 consists of three separate regions (alpha, beta, and gamma) when integrated in the chromosome but excises as one circular plasmid and re-integrates in the recipient chromosome [123].

Occurrence and Importance of Horizontal Transfer of Rhizobial Symbiosis Genes
Andrews and Andrews [17] reviewed the literature on legume-rhizobia symbioses in field soils and related genotypically characterized rhizobia (genus level) to the taxonomy of the legumes (species level) from which they were isolated. Symbioses were described for approximately 450 legume species over 255 separate studies, and phylogenetic incongruence between rhizobium core and symbiosis genes was reported in 73 (~30%) of these studies (Tables 1-4). In the current review, we have listed examples of phylogenetic incongruence between rhizobium core and symbiosis genes for strains of 14 of the currently accepted 15 genera of rhizobia. The exception is Allorhizobium, for which no sequences for symbiosis genes are available for comparison with those in the databases. Thus, horizontal transfer of symbiosis genes between rhizobial strains is of common occurrence and is not restricted to specific rhizobial genera.
Phylogenetic comparisons of gene sequences indicated that horizontal transfer of symbiosis genes was more common within than between genera. In several cases such as Bradyrhizobium associated with native Genisteae species in Europe [90] and Mesorhizobium associated with Sophora spp. throughout New Zealand [85,86], within-genus horizontal transfer of nod genes has occurred across many bacterial species and is associated with legume divergence over a wide range of habitats and over long time periods. In 27 studies (~35% of those listed here), gene sequences indicated that horizontal transfer of symbiosis genes had occurred between different rhizobia genera. In most cases, this involved gene transfer between the common alphaproteobacterial genera Bradyrhizobium, Ensifer, Mesorhizobium, and Rhizobium. However, evidence is strong that these genera provided the symbiosis genes for the less common rhizobia genera Devosia [33], Methylobacterium [66], Microvirga [67,111], and Ochrobactrum [76], which thrive under particular environmental conditions. It seems certain that within-and between-genera horizontal transfer of symbiosis genes to indigenous soil bacteria has aided the diversification and establishment of legumes in different habitats.
The data indicate that horizontal transfers of symbiosis genes between alpha-and beta-proteobacteria and between the beta-proteobacteria Burkholderia and Cupriavidus are of rare occurrence. Only two cases (in one study) were found for horizontal transfer of symbiosis genes between alpha-and beta-proteobacteria. Specifically, in the Core Cape subregion of South Africa, Burkholderia isolated from R. triflora had a nifH sequence closely related to those of Ensifer spp., and a Mesorhizobium isolate from P. oligophylla aligned closely to Burkholderia on nodA sequence [68]. There are no confirmed reports of horizontal transfer of symbiosis genes between Burkholderia and Cupriavidus, which may at least in part be related to the genera not overlapping extensively in their biogeographic ranges [17,29,38]. For Burkholderia, findings indicate that within-genus horizontal transfer of symbiosis genes is rare for South American species but common in South African species [37,40,65,79]. The reasons for this regional difference are not clear, but the patterns across different rhizobial genera indicate that 'lateral transfer of symbiosis traits is an important evolutionary force among rhizobia of the Cape Fynbos biome' [65]. The potential rapidity of these evolutionary-scale changes in bacterial genomes was illustrated by the example from western Australia, where populations of indigenous Mesorhizobium, originally unable to nodulate the pasture legume B. pelecinus, were isolated from nodules and displayed symbiosis genes obtained from the commercial inoculant within six years [63,64].
In relation to the legume host, horizontal transfer of symbiosis genes to rhizobia and non-rhizobial bacteria that are adapted to local soil conditions is likely to increase the likelihood of establishment of a successful legume-rhizobia symbiosis in these soils as long the recipient bacteria can induce functional N 2 -fixing nodules on the legume. The ecological success of the transfer is enhanced by the strong selection the plant exerts towards efficient infection and nodulation and the action of error-prone DNA polymerases that accelerate adaptation to symbiosis after gene transfer [22].
By considering all of the examples described here, we can conclude that horizontal transfer of symbiosis genes between rhizobial strains is of common occurrence, is widespread geographically, is not restricted to specific rhizobial genera, and occurs within and between rhizobial genera. The transfer of symbiosis genes to bacteria adapted to local soil conditions can allow these bacteria to become rhizobial symbionts of previously incompatible legumes growing in these soils. This, in turn, will have consequences for the growth, life history, and biogeography of the legume species involved, which provides a critical ecological link connecting the horizontal transfer of symbiosis genes between rhizobial bacteria in the soil to the above-ground floral biodiversity and vegetation community structure.