In contrast to the genomes of their closest relatives (other Rhodospirillaceae
), the three Azospirillum
genomes are larger and are comprised of not one, but seven replicons each (Figure S1
and ). Multiple replicons have been previously suggested for various Azospirillum
. The largest replicon in each genome has all characteristics of a bacterial chromosome, whereas the smallest is a plasmid. Five replicons in the genomes of A. lipoferum
Sp. 510 can be defined as “chromids” (intermediates between chromosomes and plasmids 
), whereas in A. brasilense
only three replicons are “chromids” (Tables S2
). While multiple replicons, and chromids specifically, are not unusual in proteobacteria 
, Azospirillum lipoferum
has the largest number of chromids among all prokaryotes sequenced to date 
indicating a potential for genome plasticity.
General features of Azospirillum genomes.
Comparisons among the three genomes reveal further evidence of extraordinary genome plasticity in Azospirillum
, a feature that has also been suggested by some experimental data 
. We found very little synteny between replicons of Azospirillum
species. The genetic relatedness among Azospirillum
strains is comparable to that of rhizobia, other multi-replicon alpha-proteobacteria (Table S4
). Surprisingly, we found substantially more genomic rearrangement within Azospirillum
genomes than within rhizobial genomes () that are suggested to exemplify genome plasticity in prokaryotes 
. This could be a consequence of many repetitive sequences and other recombination hotspots (Tables S4
), although the detailed mechanisms underlying such extraordinary genome plasticity remain incompletely understood.
Whole-genome alignments for Azospirillum and related multi-replicon rhizobial species.
Which genes does Azospirillum
share with its aquatic relatives, and what is the origin of its additional genes? To answer this question, we developed a robust scheme for detecting ancestral and horizontally transferred (HGT) genes () using bioinformatics tools, then classified most protein coding genes in the Azospirillum
genomes as ancestral or horizontally transferred with quantified degrees of confidence ( and Table S6
). Remarkably, nearly half of the genes in each Azospirillum
genome whose origins can be resolved appeared to be horizontally transferred. As a control, we subjected the genomes of other Rhodospirillaceae
to the same analysis, finding a substantially lower HGT level in aquatic species, while the number of ancestral genes in all organisms was comparable (). Horizontally transferred genes are frequently expendable, whereas ancestral genes usually serve ‘house-keeping’ functions and are conserved over long evolutionary distances 
. To further validate our classifications, we determined functional assignments of genes in each of the two categories using the COG database 
. The ‘ancestral’ set primarily contained genes involved in ‘house-keeping’ functions such as translation, posttranslational modification, cell division, and nucleotide and coenzyme metabolism (). The HGT set contained a large proportion of genes involved in specific dispensable functions, such as defense mechanisms, cell wall biogenesis, transport and metabolism of amino acids, carbohydrates, inorganic ions and secondary metabolites ( and Table S6
). This is consistent with the role of HGT in adaptation to the rhizosphere, an environment rich in amino acids, carbohydrates, inorganic ions and secondary metabolites excreted by plant roots 
Scheme for detecting ancestral and horizontally transferred genes.
Ancestral (red) and horizontally transferred (blue) genes in Azospirillum.
Functional categories for A. lipoferum 4B genes enriched in ancestral (top) and horizontally transferred (bottom) genes.
Such an extraordinary high level of HGT in Azospirillum
genomes leads us to hypothesize that it was a major driving force in the transition of these bacteria from aquatic to terrestrial environments and adaptation to plant hosts. This process was likely promoted by conjugation and transduction, as Azospirillum
hosts phages and notably a Gene Transfer Agent 
; this should have also resulted in loss of ancestral ‘aquatic’ genes that are not useful in the new habitat. Indeed, one of the defining features of Rhodospirillaceae
, photosynthesis (responsible for the original taxonomic naming of these organisms – purple bacteria) is completely absent from Azospirillum
. We have analyzed origins of genes that are proposed to be important for adaptation to the rhizosphere and interactions with the host plant 
. Consistent with our hypothesis, the majority of these genes were predicted to be horizontally transferred ( and Table S7
). It is important however to stress that plant-microbe interactions involve a complex interplay of many functions that are determined by both ancestral and horizontally acquired genes.
Figure 6 Proportion of ancestral (red) and horizontally transferred (blue) genes involved in adaptation of Azospirillum to the rhizosphere and its interaction with host plants (see Table S7 for details).
What was the source of horizontally transferred genes? A large proportion of genes that we assigned as HGT show relatedness to terrestrial proteobacteria, including representatives of Rhizobiales
(distantly related alpha-proteobacteria) and Burkholderiales
(beta-proteobacteria) () that are soil and plant-associated organisms. In the absence of fossil data, it is nearly impossible to determine the time of divergence for a specific bacterial lineage; however, a rough approximation (1–2% divergence in the 16S rRNA gene equals 50 Myr 
) suggests that azospirilla might have diverged from their aquatic Rhodospirillaceae
relatives 200–400 Myr (Table S8
). This upper time limit coincides with the initial major radiation of vascular plants on land and evolution of plant roots, to 400 Myr 
. Grasses, the main plant host for Azospirillum
, appeared much later, about 65–80 Myr 
, which is consistent with reports that azospirilla can also colonize plants other than grasses.
Taxonomic distribution of the best BLAST hits for predicted HGT in Azospirillum.
Using a global proteomics approach we have found that many HGT genes including nearly 1/3 of those that are common to all three Azospirillum
genomes were expressed under standard experimental conditions and under nitrogen limitation, a condition usually encountered in the rhizosphere of natural ecosystems ( and Table S9
Proportion of ancestral (red) and horizontally transferred (blue) genes in the proteomics data for A. lipoferum 4B.
Genes that differentiated the Azospirillum
species from one another and from their closest relatives are implicated in specializations, such as plant colonization. Azospirillum
and closely related Rhodospirillum centenum
both possess multiple chemotaxis operons and are model organisms to study chemotaxis 
. Interestingly, operon 1, which controls chemotaxis in R. centenum 
, plays only a minor role in chemotaxis of A. brasilense 
. All three Azospirillum
species possess three chemotaxis operons that are orthologous to those in R. centenum
; however, they also have additional chemotaxis operons that are absent from their close aquatic relative (Figure S2
and Tables S6
). Additional chemotaxis operons have been acquired by azospirilla prior to each speciation event yielding 4, 5 and 6 chemotaxis systems in A. brasilense
Sp245, A. lipoferum
4B and Azospirillum
sp. 510, respectively. These stepwise acquisitions have made the latter organism an absolute “chemotaxis champion”, with 128 chemotaxis genes, more than any other prokaryote sequenced to date (data from MiST database 
). Recent analysis showed the prevalence of chemotaxis genes in the rhizosphere 
. We have determined that the dominant chemotaxis genes in this dataset belong to a specific chemotaxis class F7 
and Table S11
). Strikingly, it is this F7 system that is shared by all Azospirillum
and is predicted to have been transferred horizontally into their common ancestor.
Cellulolytic activity may be crucial to the ability of some azospirilla to penetrate plant roots 
. All Azospirillum
genomes encode a substantial number of glycosyl hydrolases that are essential for decomposition of plant cell walls (). The total number of putative cellulases and hemicellulases in azospirilla is comparable to that in soil cellulolytic bacteria (Table S12
) and most of them are predicted to be acquired horizontally (Table S6
). We tested three Azospirillum
species and found detectable cellulolytic activity in A. brasilense
Sp245 (). The A. brasilense
Sp245 genome contains three enzymes encoded by AZOBR_p470008, AZOBR_p1110164 and AZOBR_150049 () that are orthologous to biochemically verified cellulases. We propose that these and other horizontally transferred genes (e.g.
glucuronate isomerase, which is involved in pectin decomposition) contributed to establishing A. brasilense
Sp245 as a successful endophyte 
. Interestingly, another successful endophytic bacterium, Herbaspirillum seropedicae
, lacks the genes coding for plant cell wall degradation enzymes 
indicating that endophytes may use very different strategies for penetrating the plant.
Glycoside hydrolases in Azospirillum with a potential to degrade the plant cell wall.
Cellulolytic activity of A. brasilense Sp245 cells.
Phylogenetic trees for thiamine synthetase (left) and cellulase (right).
Attachment, another function important for plant association by Azospirillum
, was also acquired horizontally. Type IV pili is a universal feature for initiating and maintaining contact with the plant host 
. The genome of A. brasilense
Sp245 lacks genes coding for Type IV pili, but encodes a set of genes for TAD (tight adhesion) pili that are known to be HGT prone 
. In our analysis, TAD pili were confidently predicted to be a result of HGT (Table S6
). We show that a mutant deficient in TAD pili had a severe defect in attachment and biofilm formation () suggesting a role for TAD in plant-microbe association.
TAD pili in A. brasilense are required for biofilm formation.
Horizontal gene transfer has been long recognized as a major evolutionary force in prokaryotes 
. Its role in the emergence of new pathogens and adaptation to environmental changes is well documented 
. While other recent studies indicate that HGT levels in natural environments may reach as much as 20% of a bacterial genome 
, our data suggest that HGT has affected nearly 50% of the Azospirillum
genomes, in close association with dramatic changes in lifestyle necessary for transition from aquatic to terrestrial environments and association with plants. Emergence of these globally distributed plant-associated bacteria, which appear to coincide with radiation of land plants and root development, likely has dramatically changed the soil ecosystem.