Comparative genomic analysis of four representative plant growth-promoting rhizobacteria in Pseudomonas

Background Some Pseudomonas strains function as predominant plant growth-promoting rhizobacteria (PGPR). Within this group, Pseudomonas chlororaphis and Pseudomonas fluorescens are non-pathogenic biocontrol agents, and some Pseudomonas aeruginosa and Pseudomonas stutzeri strains are PGPR. P. chlororaphis GP72 is a plant growth-promoting rhizobacterium with a fully sequenced genome. We conducted a genomic analysis comparing GP72 with three other pseudomonad PGPR: P. fluorescens Pf-5, P. aeruginosa M18, and the nitrogen-fixing strain P. stutzeri A1501. Our aim was to identify the similarities and differences among these strains using a comparative genomic approach to clarify the mechanisms of plant growth-promoting activity. Results The genome sizes of GP72, Pf-5, M18, and A1501 ranged from 4.6 to 7.1 M, and the number of protein-coding genes varied among the four species. Clusters of Orthologous Groups (COGs) analysis assigned functions to predicted proteins. The COGs distributions were similar among the four species. However, the percentage of genes encoding transposases and their inactivated derivatives (COG L) was 1.33% of the total genes with COGs classifications in A1501, 0.21% in GP72, 0.02% in Pf-5, and 0.11% in M18. A phylogenetic analysis indicated that GP72 and Pf-5 were the most closely related strains, consistent with the genome alignment results. Comparisons of predicted coding sequences (CDSs) between GP72 and Pf-5 revealed 3544 conserved genes. There were fewer conserved genes when GP72 CDSs were compared with those of A1501 and M18. Comparisons among the four Pseudomonas species revealed 603 conserved genes in GP72, illustrating common plant growth-promoting traits shared among these PGPR. Conserved genes were related to catabolism, transport of plant-derived compounds, stress resistance, and rhizosphere colonization. Some strain-specific CDSs were related to different kinds of biocontrol activities or plant growth promotion. The GP72 genome contained the cus operon (related to heavy metal resistance) and a gene cluster involved in type IV pilus biosynthesis, which confers adhesion ability. Conclusions Comparative genomic analysis of four representative PGPR revealed some conserved regions, indicating common characteristics (metabolism of plant-derived compounds, heavy metal resistance, and rhizosphere colonization) among these pseudomonad PGPR. Genomic regions specific to each strain provide clues to its lifestyle, ecological adaptation, and physiological role in the rhizosphere.


Background
Pseudomonas (sensu stricto) is a diverse genus that occupies many different niches and exhibits versatile metabolic capacity [1]. A number of pseudomonad strains function as plant growth-promoting rhizobacteria (PGPR). Such strains can protect plants from various soilborne pathogens and/or stimulate plant growth [2]. For example, Pseudomonas chlororaphis and Pseudomonas fluorescens are non-pathogenic biocontrol agents, while several strains of Pseudomonas aeruginosa and Pseudomonas stutzeri show strong plant growth-promoting activities. Some characteristic features associated with plant growth promotion have been studied at the molecular level. For example, effective PGPR show sufficient colonization of the rhizosphere [3,4]. Moreover, PGPR have certain biocontrol activities; for example, they can produce antibiotics that prevent infection by plant pathogens [2]. Such antibiotics include phenazine derivatives [5][6][7], pyoluteorin (Plt) [8,9], pyrrolnitrin (Prn) [10], hydrogen cyanide (HCN) [11], and so on. Some rhizobacteria directly promote plant growth in the absence of pathogens [12]. However, a comprehensive analysis of the characteristics of PGPR among different Pseudomonas species using a comparative genomics approach has not been reported yet.
To identify the shared characteristics of pseudomonad PGPR, we compared genomic information for GP72 with those of three other representative pseudomonad PGPR: the biological control agent Pf-5, the rhizobacterium M18, and the nitrogen-fixing strain A1501. There were 602 genes conserved among the four species. Comparison among these PGPR also revealed previously unknown common traits related to plant growth promotion. This comparative genomics analysis of different PGPR provides information about the genetic basis of diversity and adaptation. The results of this study also provide foundation knowledge to improve and exploit the plant growthpromoting activities of PGPR in agricultural applications via molecular techniques.

General genome features and comparative genomics
The general features of the four PGPR genomes are summarized in Table 1. The assembled genome of GP72 had approximately 270-fold sequence coverage [28], with putative functions assigned to 83% of the genes. It is reasonable to assume that the vast majority of genes are important for cell metabolism. GP72, Pf-5, M18, and A1501 showed a wide range of genome sizes, ranging from 4.6 to 7.1 M, resulting in different numbers of proteincoding genes ( Table 1). The genome sequences and additional information related to each predicted gene, such as gene product annotation, KEGG orthology, gene ontology, and predicted subcellular location are available on the IMG database (https://img.jgi.doe.gov/cgi-bin/er/main.cgi) [39]. Predicted proteins were functionally categorized using the COGs database [40], and COGs categories were compared among the genomes of GP72, Pf-5, M18, and A1501 ( Figure 1). The COGs showed similar distributions among the four strains, except for the COGs K and L, which were quite different in A1501 compared with the other strains. The percentage of genes with COG K annotations, representing transcription clusters, was lower in A1501 than in the other three strains, mainly because it contained a smaller proportion of genes encoding transcriptional regulators. COG L represents proteins with functions in replication, recombination, and repair. There were 50 genes encoding transposases and their inactivated derivatives in A1501 (approximately 1.33% of the total genes with COG annotations), compared with 0.21% in GP72, 0.02% in Pf-5, and 0.11% in M18. The large number of transposases in A1501 indicated that this strain would be more suitable for transposition, providing clues to the genetic diversity within this species and its adaptability to changes in growth conditions. Global alignments provide a powerful tool to identify conserved and specific regions in the genome, which can reveal similar biological behaviors or adaptations to specific niches. We conducted BLASTN analysis using an online version of the Artemis Comparison Tool (WebACT) [41], comparing GP72, Pf-5, and M18 ( Figure 2). We excluded A1501 from this analysis, since the alignment analysis showed very low synteny (data not shown). According to alignments at current assemblies, the extent of conservation of regions among the different species of Pseudomonas was difficult to visualize by ACT, mainly because of multiple chromosomal rearrangements. Nevertheless, the genomes of Pf-5 and GP72 showed many regions with conserved sequences and conserved gene order, except for 10 major inversions. BLAST atlas, which provides a quick overview of genomic regions of gene conservation across many genomes [42], was used to compare the reference genome of GP72 to the other three query genomes ( Figure 3).
We established a phylogenetic tree of completely sequenced pseudomonads based on two housekeeping genes (gyrB and rpoD) ( Figure 4). The tree showed that GP72 was most closely related to P. fluorescens, and was more closely related to P. aeruginosa M18 than to other opportunistic pathogenic strains of P. aeruginosa. This was probably because the rhizosphere-originated M18 strain has evolved strain-specific genomic features, which benefit its environmental adaptability and competitiveness under certain conditions in the rhizosphere niche.
An in silico subtractive hybridization analysis using the mGenomeSubtractor web server identified specific and conserved proteins. In this analysis, proteins with homology (H) values of less than 0.42 or more than 0.81 are defined arbitrarily as specific or conserved, respectively [43]. The BLASTP-based homology value distribution of 6091 predicted CDSs from P. chlororaphis GP72 was individually compared with those of the other three subject genomes ( Figure 5A) to determine the degree of protein conservation between GP72 and each of the other genomes. Among the genes encoded in the GP72 genome, 3,544 had counterparts in the genome sequence of Pf-5. There were 999 genes conserved among the genomes of GP72, Pf-5, and M18. Comparison among GP72 and all of the other three strains (Pf-5, M18, and A1501) revealed 602 homologous genes and 994 CDSs that were strainspecific to GP72 ( Figure 5B). The number of homologs in each genome is shown in Figure 4. In addition, GP72 contained 463 CDSs that were identified as strain-specific (E-value <10 -5 ) when its genome was compared with those Gene annotations and comparisons were obtained from IMG database [39]. Numbers of conserved and specific genes in each strain determined by comparison to other PGPR genomes. Genes with homology (H) values less than 0.42 and more than 0.81 were arbitrarily defined as specific and conserved, respectively.
of 27 other completely sequenced Pseudomonas strains. Further analyses of these strain-specific CDS may provide clues to the phenotype and the specific environmental adaptations of each strain.

Environmental adaptability Catabolism
PGPR can use a wide range of nutrients to colonize the rhizosphere successfully. The central metabolic pathways in GP72, such as the Entner-Doudoroff pathway, the pentose phosphate pathway, and the tricarboxylic acid cycle, are consistent with those reported for other Pseudomonas species [44]. Like other Pseudomonas strains, GP72 lacks 6-phosphofructokinase; therefore, it may not have a functional Embden-Meyerhof pathway. The genomes of GP72, Pf-5, M18, and A1501 contained genes encoding a fructose-specific IIA component, I-phosphofructokinase.
This enzyme is involved in the fructose dissimilation pathway, catalyzing the conversion of fructose to fructose-1,6diphosphate [45]. PGPR have a variety of genes related to catabolism and transport of plant-derived compounds, such as amino acids, fatty acids, nucleotides, organic acids, carbohydrates, and other exudates [46,47]. Amino acids are one of the major components of root exudates. Accordingly, there were at least 500 genes involved in amino acid transport and metabolism in the genomes of GP72, Pf-5, and M18, and more than 300 in the genome of A1501.
The ability to catabolize aromatic compounds in exudates is one strategy that could confer a selective advantage in the rhizosphere environment. Oxygenases play key roles in the chemical transformation of recalcitrant organic compounds [48,49]. P. putida modifies diverse aromatics to common intermediates, which feed into central pathways [50]. For example, P. putida KT2440 is able to use aromatic compounds including benzoate, phenylacetate, tyrosine, and vanillate, as the sole carbon and energy source [51]. There were 21 genes encoding dioxygenases (DOs) in the genome of GP72, 22 in M18, 21 in Pf-5, and 9 in A1501. The DOs in GP72, including benzoate 1,2-dioxygenase, anthranilate 1,2-dioxygenase, protocatechuate 3,4-dioxygenase, and catechol 1,2dioxygenase, were related to degradation of aromatic compounds. We compared the degradation pathways of aromatic compounds, including the three main pathways ( Table 2) and several peripheral pathways, among the four species (Additional file 1).
Genes encoding components of the 3-oxoadipate pathway, which is common in soil and plant-associated microorganisms [52], were present in the genomes of all four PGPR analyzed in this study. The pathway has two branches: one converting catechol and the other converting protocatechuate. Both branches produce two tricarboxylic acid cycle intermediates. Based on the comparative genomic analysis, the former branch may derive from the degradation of tryptophan [53], benzoate [54], salicylate [55], phenol [56,57], and so on, while the protocatechuate branch is generated from 4-hydroxybenzoate [58], and numerous lignin monomers such as vanillate [59] and quinate [60]. Analyses of aromatic compound catabolism not only reveal the broad metabolic activities of PGPR, but also provide insights into mediating the production of useful secondary metabolites such as phenazine [61], pyocyanin (PYO) [62], and C-1027 [63].
Some bacteria and fungi degrade tyrosine (Tyr) via the central intermediate homogentisate (2,5-dihydroxyphenylacetate). The reaction proceeds with conversion of Tyr into 4-hydroxyphenylpyruvate (HPP) (by tyrosine aminotransferase), and then formation of homogentisate (by HPP dioxygenase), which is degraded via the homogentisate central pathway [64]. The central pathway Figure 1 Comparison of COG categories among four pseudomonad PGPR. Functional classifications provided by the COG database [40] were used for functional comparisons among the genomes of P. chlororaphis GP72, P. fluorescens Pf-5, P. aeruginosa M18, and P. stutzeri A1501. The ordinate axis indicates the percentage of genes in each COG functional category relative to the genes of all COG categories. Comparison was based on 22 COGs categories: RNA processing and modification (A), chromatin structure and dynamics (B), energy production and conversion (C), cell cycle control, cell division, chromosome partitioning (D), amino acid transport and metabolism (E), nucleotide transport and metabolism (F), carbohydrate transport and metabolism (G), coenzyme transport and metabolism (H), lipid transport and metabolism (I), translation, ribosomal structure and biogenesis (J), transcription (K), replication, recombination and repair (L), cell wall, membrane, envelope biogenesis (M), cell motility (N), posttranslational modification, protein turnover, chaperones (O), inorganic transport and metabolism (P), secondary metabolites biosynthesis, transport and catabolism (Q), general function prediction only (R), function unknown (S), signal transduction mechanisms (T), intracellular trafficking, secretion and vesicular transport (U), defense mechanisms (V).
In a few organisms, phenylethylamine, an intermediate of phenylalanine degradation, can be converted into phenylacetaldehyde by quinohemoprotein amine dehydrogenase, and then transformed into phenylacetate by phenylacetaldehyde dehydrogenase [72][73][74]. The corresponding genes were predicted in the genomes of GP72 and Pf-5, but they were not located in a single operon.
Phenylacetyl-CoA is derived from various substrates such as phenylalanine, lignin-related aromatic compounds, and environmental contaminants, and can be degraded to succinyl-CoA and acetyl-CoA [75,76]. Based on the genomic comparison at the 60% identity threshold, we found that the phenylacetate degradation pathway was present in GP72 and Pf-5. However, this pathway was not detected in M18 or A1501 at the same identity level, indicating potentially different evolutionary directions in specific niches.
Five putative phenylpropionate dioxygenases and related ring-hydroxylating dioxygenases of unknown specificity can also participate in aromatic compound catabolism [77].
Plant-derived substances not only serve as important carbon and energy sources for rhizosphere bacteria, but also influence bacterial behaviors [78,79]. For example, the ratio of rhizospheric carbon:nitrogen (C:N) can alter the nutritional status of Rhizoctonia solani, making the fungus a pathogen [80]. Tomato root exudates promote Figure 4 Phylogenetic relationships among completely sequenced Pseudomonas species. Phylogenetic tree for members of the genus Pseudomonas was constructed based on aligned concatenated sequences of gyrB and rpoD using the neighbor-joining method with 1000 bootstrap replicates. Analysis was carried out using Phylip 3.67 software and the tree was plotted using iTOL software. Colors on the phylogenetic tree indicate membership in Pseudomonas phylogenetic groups according to NCBI taxonomy. Completely sequenced species in the genus Pseudomonas include P. aeruginosa (yellow), P. brassicacearum (olive), P. entomophila (purple), P. fluorescens (green), P. fulva (blue), P. mendocina (pink), P. putida (navy), P. stutzeri (magenta), and P. syringae (cyan). In this research, the tree branch of P. chlororaphis, whose draft genome sequence was reported recently, is shown in red. Bar chart associated with nodes indicates numbers of genes conserved between GP72 and the corresponding organism. Conserved genes were determined using mGenomeSubtractor.
germination of spores of the tomato root pathogen F. oxysporum f. sp. radicis-lycopersici, whereas the biocontrol agent P. fluorescens WCS365 delays this process [81]. Microarray analyses showed that root exudates affected the transcriptome of P. aeruginosa PAO1 by influencing genes encoding enzymes related to alginate biosynthesis and twitching motility [82]. Therefore, the production of plant-derived exudates could alter the composition of rhizospheric microorganism communities. Further research is required to investigate the molecular mechanisms underlying changes in community structure.

Transport
Consistent with the abundance of genes related to metabolism of plant-derived substances, the four PGPR contained many putative transport genes related to substrate uptake and excretion (Table 3). GP72 and Pf-5 contained similar numbers of transport genes. In bacteria, secretion systems play an important role in transport or translocation of effectors for adaptation to their natural surroundings. The genomes of these four PGPR contained type I, type II, type IV, type V, and type VI secretion systems, as well as the chaperone-usher secretion system and the twin-arginine translocation system. M18 also contained the Type III (flagellar/pathogenesis) secretion system, a key virulence factor in pathogenic Pseudomonas [83].

Defense pathways
Previous studies showed that GP72 resists streptomycin up to a concentration of 100 μg ml -1 , and tolerates salt (5% NaCl solution). Both of these resistances are stronger than those of P. chlororaphis strain 30-84 [7]. Figure 5 Homology analysis between P. chlororaphis GP72 genome and three subject genomes. The mGenomeSubtractor arbitrarily defines CDSs with homology (H) values less than 0.42 as strain-specific, and those with H values greater than 0.81 as conserved [43]. (A) Histogram of BLASTP-based homology value distribution of 6091 predicted CDSs from P. chlororaphis GP72 compared individually with those of three other genomes: P. fluorescens Pf-5, P. aeruginosa M18, and P. stutzeri A1501. (B) Numbers of conserved and specific genes in GP72 compared with three other PGPR strains. Total numbers of conserved and specific genes are shown above columns. Antibiotic resistance assays showed that GP72 displays resistance to penicillin, spectinomycin, streptomycin, and tetracycline. Here, our genomic analyses confirmed different kinds of defenses in the four PGPR, including resistance/tolerance to heavy metals, temperature stress, osmotic stress, oxidative stress, and multiple drugs. Many essential trace elements contain metal ions that are important components of the active sites of many enzymes. As such, they play a vital role in many biological processes, including photosynthetic and respiratory pathways. However, most heavy metals are toxic at higher concentrations. For example, copper ions can damage the cytoplasmic membrane of E. coli by catalyzing harmful redox reactions [84]. In many regions, agricultural soils are heavily contaminated with various heavy metals originating from chemical fertilizers and industrial processes. Consequently, certain soil bacteria have developed resistance to toxic metals, either via active efflux mechanisms to pump the toxic metals out [85], or by enzymatic detoxification to convert a toxic ion into a harmless one [86,87]. Our genomic analysis revealed many genes related to heavy metal resistance (summarized in Table 4).
The four pseudomonad PGPR studied contained at least two different copper resistance systems, which resemble those identified in the plant growth-promoting endophytic bacterium P. putida W619 [15]. One system is periplasmic detoxification encoded by copABCDGcopRS, which is wellcharacterized in the plasmid pPT23D from P. syringae pv. tomato strain PT23.2 [88]. This system is also widely distributed in other Pseudomonas species [15,88,89]. Another copper resistance system is the cytoplasmic detoxification system cue, which maintains a strict quota of cellular copper in other organisms [90,91]. The Cu(I)-responsive transcriptional regulator CueR [91] activates expression of a copper-translocating P-type ATPase (CopA) [92], a periplasmic multicopper oxidase (CueO) [93], and a copper chaperone (CopZ) [90] under mild copper stress. CopA exports Cu(I) from the cytoplasm to the periplasm, and then Cu(I) is converted into the less toxic Cu(II) form by CueO. A third copper resistance strategy in the genome of GP72 consisted of cusFABBC (MOK_00020-00016) and copRS (MOK_00012-00013). The cus operon is related to periplasmic detoxification, and is exclusively found in Gramnegative bacteria [94,95]. Therefore, the mechanism for copper resistance in P. chlororaphis is very complex, and has not been completely characterized yet. Further research is required to clarify the details of this system. Pseudomonas spp. have arsenic-resistance genes (arsRB, arsCH and arsC) that are dispersed throughout the genome. The chromosomal ars operon was characterized in P. aeruginosa. A homologous ars operon was detected in some, but not all, Pseudomonas species, indicating that some other mechanisms are involved in arsenic resistance in pseudomonads [96]. Our genomic analysis indicated that the GP72 genome lacked a homologous gene encoding an arsenite-and antimonite-stimulated ATPase (ArsA). However, a previous study showed that ArsB could export arsenite ions in the absence of ArsA in E. coli [97]. Since ArsB was predicted in the genome of GP72, we can assume that this strain also shows arsenic resistance. The czcABCRD operon encoding a cation-proton antiporter, which is responsible for cobalt, zinc, and cadmium resistances [98], was predicted in the genomes of GP72 and Pf-5. Other genes found in their genomes may also be related to heavy metal resistance, such as homologs of chrA and chrB genes involved in chromate resistance [99], and homologs of genes encoding siderophores that participate in metal homeostasis in P. aeruginosa [100].
In recent years, multidrug resistance has reached alarming levels, especially in the field of medicine [101]. Such resistance mechanisms have been fully described by Alekshun and Levy [102]. Some PGPR strains contain a broad spectrum of putative multidrug resistance genes, including genes related to well-developed efflux systems [103], penicillin-binding protein-mediated resistance [104], and enzymes that degrade antibiotics (Additional file 2). Efflux systems contribute significantly to resistance to multiple antimicrobial compounds. This is a very important mechanism to enhance biological fitness [105]. Like other Pseudomonas species, GP72 contained 37 putative ABC transporters, which potentially participate in the uptake or efflux of toxic metabolites and other drugs. Some secondary transport system genes were also present in GP72 (Table 3); there were 35 genes encoding RND family members, three genes for MATE family members, and four genes for SMR family members. All pseudomonad PGPR contained genes encoding the efflux pumps TtgABC and TtgDEF (toluene tolerance genes). These enzymes prevent the accumulation of toluene and other related aromatics, such as phenol [106]. Genes encoding an MexEF-OprN efflux pump, a member of the RND family, were also present in the genomes of GP72, Pf-5, M18, and A1501, but the order of the efflux pump genes in the genome differed among the four strains. The efflux pump operon is upregulated by MexT under nitrosative stress and chloramphenicol stress [107]. Overexpression of this system can decrease the production of several secondary metabolites such as PYO, elastase, and rhamnolipids [108]. AcrB (homologous to MOK_00261 in the GP72 genome), which also belongs to the RND family [109], plays a role in pumping out basic dyes (such as acriflavine), most antibiotics (except aminoglycosides), and detergents (such as bile salts, Triton X-100, and SDS) [110]. In conclusion, the genomic data indicated that these PGPR harbor genes that can confer resistance to multiple drugs, including penicillin, aminoglycosides, fluoroquinolones, trimethoprimsulfamethoxazole, lipid A, and acriflavine.
Bacteria that inhabit the rhizosphere of plants can use plant-derived compounds as nutrients; however, they must   [111] and other biological processes. ROS show antimicrobial activities [112], as they can damage proteins, nucleic acids, and cell membranes. Rhizospheric bacteria produce several enzymes to resist oxidative stress [113]. Genes encoding these enzymes have already been identified in the genomes of Pf-5 and A1501 [17,18]. Putative ROS-detoxifying enzymes in GP72 included 11 peroxidases, five catalases, two superoxide dismutases, and 19 glutathione S-transferases. There was no significant difference in the numbers of these enzymes among the four PGPR ( Figure 6). Genes encoding regulators of the oxidative stress response, including the two-component regulator GacS/GacA [114], SoxR, and OxyR [113,115] were present in the genomes of GP72, Pf-5, M18, and A1501. However, a homolog of SoxR in P. aeruginosa did not function as a key regulatory player in the bacterial oxidative stress response [116]. Exopolysaccharides such as alginate [117] and polyhydroxyalkanoates (PHAs) [118] are important for tolerance to oxidative stress under ambient pressure. For instance, PHA accumulation enhances the survival of pseudomonads under salinity stress, oxidative stress, and cold-shock [119,120]. Additionally, a pyrroloquinolinequinine (PQQ) synthase expressed in E. coli improves its resistance to photodynamically produced ROS [121]. Rhizosphere bacteria usually survive in a changeable environment; therefore, they have evolved several traits related to adaptation [122]. The genomes of GP72, Pf-5, M18, and A1501 contained homologs of genes related to tolerating cold-shock, including cspACDG, which is constitutively expressed at 37°C [123]. In P. aeruginosa cells, a temperature increase from 30 to 45°C enhances production of 17 proteins, including the heat-shock proteins DnaK and GroEL [124]. A chaperone system formed by DnaK, DnaJ, and GrpE proteins modulates the heat-shock response in E. coli [125] (Additional file 2). As an opportunistic pathogen, P. aeruginosa has evolved to survive in diverse stressful environments. A microarray analysis showed that P. aeruginosa synthesizes osmoprotective compounds, such as hydrophilins and osmoprotectants, to cope with osmotic stress [126,127]. Glycine betaine (GB), a major osmoprotectant for many bacteria [128], can accumulate via de novo synthesis or via absorption from the environment [126]. Mutant analyses and 13 C NMR studies confirmed GB catabolism in P. aeruginosa [129]. Previously, it was shown that GP72 shows strong osmotic stress tolerance [7]. The genomic analysis in this study showed that the genomes of GP72, Pf-5, and M18 contained at least one complete gene set required for conversion of GB to glycine; this gene set included gbcAB, dgcAB, and soxGADB. In contrast, A1501 contained only a homolog of the betAB operon, which encodes a system for oxidation of choline to GB under osmotic stress conditions [130]. Osmoregulated periplasmic glucans are highly branched oligosaccharides found in the periplasm of Gram-negative bacteria. They are probably produced in response to periplasmic osmolality, which is controlled by the products of mdoD and mdoG [131]. Enteric bacteria can modulate their cytoplasmic osmolality through mobilizing K + , glutamate, and other compatible solutes, such as trehalose, proline, and GB [132]. K + first responds to osmotic upshifts via the transporters Trk and Kdp, possibly acting as a putative osmoregulatory second messenger [126,133]. The genes related to osmotic stress tolerance are listed in Additional file 2.
We found that resistance genes were present in the genomes of all four PGPR, although some genes showed low similarity to others. These results indicated that PGPR may undergo long-term evolution to adapt to specific ecological niches. To adapt to changeable environments, each pseudomonad PGPR strain has a complex array of regulatory networks, including sigma factors, transcriptional regulators, and a variety of two-component transcriptional regulators.

Rhizosphere colonization
A confocal laser scanning microscopy analysis showed that P. fluorescens WCS365 and P. chlororaphis PCL1391 are able to effectively colonize the tomato rhizosphere [134], and the major traits for niche competition were identified [135]. Development of new genetic approaches such as in vivo expression technology (IVET) together with "omic" technologies has provided opportunities to identify genes required for rhizosphere competence, and to elucidate the genetic mechanisms of plant-microbe interactions [14,136]. Pseudomonad PGPR show certain competitive Figure 6 Numbers of predicted enzymes with roles in oxidative stress response. Predicted proteins with roles in the oxidative stress response found in P. chlororaphis GP72, P. fluorescens Pf-5, P. aeruginosa M18, and P. stutzeri A1501. Four types of enzymes (glutathione S-transferase, peroxidase, catalase, and superoxide dismutase) were compared among the four species.
colonization traits, such as motility and the ability to attach to the root surface.
First, motility is a major trait for the competitive tomato root-tip colonization of P. fluorescens, based on chemotaxis [137,138]. We found genes related to chemotaxis and motility in the genomes of the four PGPR; GP72 contained 14 genes responsible for various aspects of chemotaxis, including genes encoding a two-component system (CheA/CheY). The activity of the histidine kinase CheA can be regulated by methyl-accepting chemoreceptor proteins (MCPs) during chemotaxis [139]. Swimming behavior can be initiated when the phosphorylated CheY binds to the flagellar switch protein, Flim [140]. In the present study, we found 28 genes encoding MCPs and 40 genes associated with flagella biosynthesis, including the flg and fli operons.
The second trait of competitive colonization is attachment to the root surface. In this study, several genes involved in attachment were predicted in the PGPR genomes (Additional file 3). The functions of some genes have been confirmed experimentally in certain Pseudomonas species, including genes associated with type IV pili and twitching motility [141], genes for biosynthesis of alginate [142], hemolysin [135,143], filamentous hemagglutinin [144], and lipopolysaccharide O-antigen [145], and genes for other enzymes or factors involved in adhesion [135]. For instance, twitching motility, a type of flagella-independent surface motility mediated by type IV pili, is a mechanism of rapid bacterial colonization [141]. As well as the common type IV pilus assembly proteins, we identified a second set of genes in P. chlororaphis GP72 that were previously reported to play roles in the biogenesis of the Flp subfamily of type IVb tight adherence (Tad) pili [146]. However, tad genes were not found in the genomes of the other three PGPR at the 60% identity threshold, when compared with the genome of GP72. Tad pili are an essential and conserved host-colonization factor in Bifidobacterium species [147]. Therefore, we can speculate that the tad genes are probably derived from organisms outside of the genus Pseudomonas. In strain P. putida KT2440, a series of rap genes (root-activated promoters) were identified during maize root colonization by IVET [148]. Some of the promoters isolated by rap fusions responsible for adhesion were present in the genomes of GP72, Pf-5, and M18, such as secB (rap 1-2 fusion) [135] and algD (rap 2-45 fusion) [142]. The genetic locus aggA, which is involved in agglutination and adherence [149], was also predicted in the genomes of GP72 and Pf-5. Espinosa-Urgel et al. [143] characterized several mus (mutants unattached to seeds) loci in P. putida, and confirmed that mutants of these loci show impaired attachment to corn seeds. The genome of GP72 contained four mus loci: mus-13, mus-21, mus-24, and mus-27, with possible functions as a carbon starvation protein, transporter, calcium-binding protein, and hemolysin, respectively. Genes involved in competitive rhizosphere colonization have been well studied in P. fluorescens. These include xerC, which encodes a site-specific recombinase. xerC is a homolog of sss in P. chlororaphis PCL1391; sss plays a role in phase variation caused by DNA rearrangements [3,150]. The nuo operon encodes subunits of NADH: ubiquinone oxidoreductase, which is related to ATP-dependent rotation of flagella [145]. Some of the genes isolated by IVET and identified to play roles in plant-microbe interactions [14,151] were present in the genomes of GP72, Pf-5, M18, and A1501 when compared with the genome of P. fluorescens SBW25 (data not shown). However, it remains to be confirmed whether these genes specifically contribute to rhizosphere competence.
GP72, Pf-5, and A1501 lacked virulence factors found in plant pathogens, such as the type III secretion system, phytotoxins, and exoenzymes associated with cell wall degradation. Homologs of genes encoding phytotoxins produced by P. syringae (coronatine, syringomycin, syringopeptin, tabtoxin, and phaseolotoxin) [152] were also absent from the genomes of GP72, Pf-5, M18, and A1501. Their genomes did not contain genes related to the biosyntheses of cellulases, pectinases, or pectin lyases, which play roles in the degradation of cell wall components. Therefore, the lack of these genes can result in efficient rhizosphere colonization and improvement of plant growth.

Biocontrol activities
Biocontrol activities are important mechanisms by which PGPR suppress plant pathogens. The main biocontrol strategy is the production of a spectrum of antibiotics [2]. The antibiotics produced by the biocontrol agents GP72, Pf-5, and M18 are listed in Table 5. Phenazines are versatile secondary metabolites produced by P. fluorescens, P. chlororaphis, and Pseudomonas aureofaciens [153]. These compounds play critical roles in the biological control activities of Pseudomonas spp. [5]. Previous studies showed that GP72 can completely suppress various phytopathogens, mainly because of the production of PCA and 2-OH-PHZ. Clusters of phenazine-compound biosynthetic genes were present in the genomes of both GP72 and M18, but the genes differed between the two species. The GP72 genome contained phzO, encoding an aromatic monooxygenase [154] that converts PCA to 2-OH-PHZ, whereas M18 contained two phz gene clusters and one set of modified phzMS genes. phzM and phzS encode a putative S-adenosylmethionine-dependent N-methyltransferase and a putative flavin-dependent hydroxylase, respectively. They participate in the conversion of PCA to PYO in P. aeruginosa. PYO is a virulence factor to cystic fibrosis patients infected by pathogenic pseudomonads [155]. However, M18 does not produce detectable levels of PYO at 28°C, mainly because of the temperature-dependent expression of phzM and its regulatory genes lasI and ptsP. The biocontrol activity of M18 is, therefore, not attributed to PYO but to PCA [33], and it shows lower pathogenicity than other closely related strains. Previous studies showed that some plant pathogens are more strongly inhibited by 2-OH-PHZ than by PCA [154]. Another important antibiotic is Plt, which is produced by both Pf-5 and M18. Strains with the ability to produce the insect toxin 'Fit' (P. fluorescens insecticidal toxin) [156] show potent insecticidal activity [157]. The fitD gene encoding the cytotoxin in GP72 showed 84% identity to that in Pf-5. The amino acid sequence of Fit shared 77% amino acid identity with the insect toxin Mcf (makes caterpillars floppy) produced by the entomopathogen Photorhabdus luminescens [157].
Fluorescent pseudomonads can produce pyoverdin (Pvd), a fluorescent siderophore, and chelate Fe(III) efficiently under low-iron conditions to improve their biocontrol activity [158,159]. The fluorescent pseudomonads GP72, Pf-5, and M18 contained the complete Pvd biosynthetic gene cluster. In addition, Pf-5 and M18 contained genes encoding another siderophore, Pch, which has antifungal activity [160]. GP72 lacked these genes, but it contained putative genes for synthesis of achromobactin (Acr), a temperature-regulated secondary siderophore. The related biosynthetic gene clusters in GP72 included acsFDECBA, yhcA, and acrABCD, which are responsible for the biosynthesis of Acr, permease, and a specific outer membrane receptor, respectively [161,162]. Siderophores can bind metals other than iron [163] and, therefore, can play roles in sequestering toxic metals including aluminum, cobalt, copper, and lead [100]. GP72 contained a locus (MOK_02694) encoding a nickel-uptake substrate-specific transmembrane protein, adjacent to the acr operon. Acr in GP72 may be involved in metal transport, signaling pathways, or antimicrobial activities. The comparative genomic analysis indicated that there was no homology of the acr operon between Pf-5 and M18. As well as producing their own siderophores, Pseudomonas can also use siderophores produced by other microorganisms. For example, A1501 may obtain iron via heterologous siderophores, since it lacks pathways for siderophore biosynthesis [18]. Genes involved in the uptake of soluble Fe(III) complexes, that is, those encoding putative outer membrane receptors, were present in the genomes of the four pseudomonads: 31 genes in GP72, 45 in Pf-5, 36 in M18, and 24 in A1501. The variable iron acquisition systems among Pseudomonas reflect their large capacity for niche colonization, providing insights into how their biocontrol abilities can be improved. Therefore, the availability of complete genome sequences provides an excellent opportunity to explore the diversity and evolution of biosynthetic pathways in different species/strains [153,164].

Direct plant-growth promotion
Rhizobacteria can directly promote plant growth, and some strains have been developed as 'biofertilizers'. The mechanisms underlying plant-growth promotion include nitrogen fixation, increased nutrient availability, production of phytohormones, and so on [165]. The biofertilizers Azotobacter [12] and P. stutzeri [18], both of which belong to the Pseudomonadaceae, are able to fix nitrogen. The genome of A1501 contains a cluster of 59 genes specific to nitrogen fixation, and the nif operon shows a high degree of similarity to that in the genome of Azotobacter vinelandii. Therefore, we compared A1501 with GP72, Pf-5, and M18 at a threshold of 30% identity to screen for putative genes related to nitrogen fixation. The analyses revealed 13, 13, and 14 homologous genes in GP72, Pf-5, and M18, respectively (Additional file 4); however, these three strains lacked the nitrogenase complex-encoding genes nifDK [166,167]. We conducted a similar screen for denitrification genes; of 45 genes in A1501, 7 homologs were found in the genome of Pf-5, and 21 in the genome of GP72. We can speculate that the low identities may be because of the relatively distant evolutionary relationship, as shown in the phylogenetic analysis ( Figure 4). GP72 and M18 contained several genes involved in denitrification: narL and narX, which encode a two-component regulatory system; narGHJI, which encodes respiratory nitrate reductase [168]; and nor genes, which are involved in nitric oxide metabolism. Previous studies reported that P. fluorescens and P. chlororaphis produce N 2 O as the only detectable gaseous product of denitrification [169], while P. stutzeri emits only N 2 , and P. aeruginosa produces both N 2 and N 2 O [170]. Thus, the denitrification process can accommodate large quantities of anthropogenic nutrients, converting nitrate into nitrogen. This could decrease nitrate accumulation and counteract eutrophication in the environment [171].

Conclusions
We analyzed plant growth-promoting traits by a comparative genomics analysis of four representative pseudomonad PGPR strains. The genes that were conserved among the different Pseudomonas species have provided clues to the common characteristics of pseudomonad PGPR, such as rhizosphere competence traits (nutrient catabolism and transport, resistance to various environmental stresses, and rhizosphere colonization). The strain-specific genes differentiated each strain on the basis of its lifestyle, specific ecological adaptations, and physiological role in the rhizosphere. The recently reported genome of P. chlororaphis, together with other sequenced strains of different species of pseudomonad PGPR, provides insights into the genetic basis of diversity and adaptation to specific environmental niches. Comparative genomic analyses, combined with certain IVET-based analyses, can reveal many genetic factors related to plant growth promotion. First, the strong adaptability of PGPR to their environment is related to putative genes involved in catabolism and transport of plant-derived compounds and resistance to various environmental stresses (heavy metals, ROS, cold-, heat-, or osmotic-shock, and multiple drugs). These genes were very common in the genomes of PGPR, especially those of P. chlororaphis and P. fluorescens, and provide the foundation for rhizosphere fitness. Second, we compared genes involved in rhizosphere colonization. Some related genes showed low similarity between P. chlororaphis GP72 and the other three strains, including biosynthetic genes for the O-antigen and type IV pilus assembly. Hence, GP72 may have stronger rhizosphere competence than the other three strains. Third, we analyzed genes related to biocontrol activities, namely those encoding production of antifungal metabolites such as PCA and Plt. The genomic information indicated that the secondary metabolites differ markedly among the four PGPR. For example, GP72 contained putative gene clusters for biosynthesis of the siderophore Acr, whereas the other strains contained gene clusters for biosynthesis of different siderophores. Some rhizobacteria cannot produce antifungal compounds, but promote plant growth in the absence of pathogens. One such strain was P. stutzeri A1501, which fixes nitrogen. Therefore, the metabolic pathways, transporters, and regulators related to cell metabolism provide directions to improve plant growth-promoting activities. Genetic modification may accelerate the commercialization of PGPR as biocontrol agents, which could further contribute to sustainable development of agriculture.

Methods
Medium and growth conditions for P. chlororaphis GP72 P. chlororaphis GP72 (deposited in China General Microbiological Culture Collection Center; collection number 1748), isolated from green pepper rhizosphere in eastern China, was incubated at 28°C in King's medium B [177].

Genome sequencing and annotation
The genome of P. chlororaphis GP72 was sequenced using the Illumina GAIIx platform and assembled using VELVET 1.1.07. The genome of GP72 was automatically annotated using the RAST server [178], and proceeded with manual curation and comparative analysis using the IMG/ER system (https://img.jgi.doe.gov/cgi-bin/er/main.cgi) [179]. The genome sequence is available at the IMG database [39]. Information of COGs [40], combined with that from the Conserved Domain Database, was also used in the comparisons. The metabolic pathways were examined using KAAS (KEGG Automatic Annotation Server) [180] and the MetaCyc database [181].

Nucleotide sequence accession number
This whole genome shotgun project has been deposited in DDBJ/EMBL/GenBank under the accession number AHAY00000000.

Genome comparisons
The genome sequence of GP72 was aligned against sequences of other Pseudomonas genomes from NCBI's Entrez database and the IMG database. Pair-wise alignments were performed using WebACT (http://www. webact.org/WebACT/home) [41]. BLAST atlases [42] were generated using the CBS DTU online tool, GeneWiz browser 0.94 server (http://www.cbs.dtu.dk/services/ gwBrowser/). Strain-specific and conserved genes were identified using the mGenomeSubtractor web server (http://bioinfo-mml.sjtu.edu.cn/mGS/) [43]. The conserved CDSs were identified using a homology (H) value cut-off of 0.42 at E-value <10 -5 . Comparative genomic analyses of GP72, Pf-5, M18, and A1501 were conducted using the tool set available at the IMG website; genes homologous to those in GP72 were computed with an E-value < 10 -2 and at 60% identity; BLAST comparisons between PGPR and P. stutzeri A1501 were screened at the 30% identity threshold.

Phylogenetic analysis
The phylogenetic relationships among completely sequenced Pseudomonas were determined by a multilocus sequence analysis using a concatenated data set of gyrB and rpoD genes. Multiple-sequence alignments were carried out with Clustal W (http://www.genome.jp/tools/clustalw/) [182]. Evolutionary distances were calculated using the neighbor-joining method [183] with 1000 bootstrap replicates, using Phylip 3.67 software (http://evolution.genetics. washington.edu/phylip.html). The phylogenetic tree was generated using interactive tree of life (iTOL) software [184].