Biocontrol and plant growth-promoting activity of rhizobacteria from Chinese fields with contaminated soils

The aim of this study was to inventory the types of plant growth-promoting rhizobacteria (PGPR) present in the rhizosphere of plants grown in soils contaminated with heavy metals, recalcitrant organics, petroleum sewage or salinity in China. We screened 1223 isolates for antifungal activity and about 24% inhibited Rhizoctonia solani or Sclerotinia sclerotiorum. Twenty-four strains inhibitory to R. solani, Gaeumannomyces graminis var. tritici and/or S. sclerotiorum and representing the dominant morphotypes were assayed for PGPR activity. Seven strains contained phlD, prnD, pltC or phzF genes and produced the antibiotics 2,4-diacetylphloroglucinol, pyrrolnitrin, pyoluteorin and phenazines respectively. Six strains contained acdS, which encodes 1-aminocyclopropane-1-carboxylic acid deaminase. Phylogenetic analysis of 16S rDNA and phlD, phzF and acdS genes demonstrated that some strains identified as Pseudomonas were similar to model PGPR strains Pseudomonas protegens Pf-5, Pseudomonas chlororaphis subsp. aureofaciens 30–84 and P. brassicacearum Q8r1-96. Pseudomonas protegens- and P. chlororaphis-like strains had the greatest biocontrol activity against Rhizoctonia root rot and take-all of wheat. Pseudomonas protegens and P. brassicacearum-like strains showed the greatest promotion of canola growth. Our results indicate that strains from contaminated soils are similar to well-described PGPR found in agricultural soils worldwide. Growth-promoting rhizobacteria in polluted soils


Introduction
Soil salinization and contamination by organic compounds (i.e. petroleum products and solvents) and heavy metals (i.e. Cd, Cr, Cu, Hg, Pb) are important environmental problems worldwide and have negative impacts on both human health and agricultural productivity (Jing et al., 2007). Soil pollution is of particular concern in developing countries where rapid industrialization has led to contamination of agricultural lands, especially in small farms that are sources of food for consumption by local populations. In a survey of contamination levels of heavy metals in agricultural soils in China, metal concentrations of Cd were significantly higher than in non-contaminated soils (Wei and Yang, 2010). Similarly, because of the accumulation of mine tailing outfalls, the content of Cu in the wetlands of Daye County, Hubei province, was 18 times higher than the background value (Guo et al., 2011). The build-up of pollutants in soils and their subsequent leaching into the ecosystem has cascading effects, adversely impacting plant growth and yield (Cheng, 2003b;Huang et al., 2005;Parida and Das, 2005). The uptake of soil pollutants by crop plants and the resultant entry of carcinogens, endocrine disruptors and neurotoxins into the food chain are major concerns for human and animal health (Cheng, 2003a;Cai et al., 2009). In addition, soil pollutants can positively or negatively influence crop susceptibility to pathogens and insects (Elmer, 2002;Poschenrieder et al., 2006). For example, salinity can predispose plants to infection by soilborne pathogens or increase the incidence and severity of disease (Triky-Dotan et al., 2005;Roubtsova and Bostock, 2009).
In contrast, exposure of wheat to a mild dose of cadmium induced resistance against Fusarium infection (Mittra et al., 2004).
Rhizobacteria are able to colonize and persist on roots and some, known as plant growth-promoting rhizobacteria (PGPR), are able to promote plant growth (Weller, 2007). PGPR improve plant growth directly by supplying nutrients (e.g. phosphorous, nitrogen), producing phytohormones [e.g. indoleacetic acid (IAA)], and/or decreasing the level of ethylene that is produced when the plant is stressed (Li et al., 2007;Lugtenberg and Kamilova, 2009). PGPR also antagonize plant pathogens and/or induce systemic resistance in the plant to disease Lugtenberg and Kamilova, 2009). Because plants lack effector-triggered immunity to many soilborne pathogens, they rely on PGPR for defence against pathogen attack .
Soil pollution, especially by heavy metals, can also decrease the metabolic activity, biomass and diversity of microorganisms in the bulk soil and the rhizosphere (Sandaa et al., 1999;Gremion et al., 2004;Jing et al., 2007) and limit the effectiveness of PGPR to protect roots against disease. However, PGPR with metal resistance and the ability to mitigate the toxic effects of soil pollutants are present in polluted soils (Huang et al., 2005;Reed and Glick, 2005;Jing et al., 2007). For example, Pseudomonas asplenii AC isolated from polycyclic aromatic hydrocarbon-contaminated soils significantly increased the root and shoot growth of canola in copper and creosote-contaminated soils (Reed and Glick, 2005). Salt-tolerant Pseudomonas chlororaphis TSAU13, isolated from the rhizosphere of wheat grown in a salinated soil (Egamberdieva and Kucharova, 2009), improved the shoot and root growth of common bean grown in soil amended with NaCl and protected cucumber and tomato against Fusarium solani (Egamberdieva, 2011).
Despite progress in understanding the crucial role of PGPR in helping plants adapt to environmental stress, major gaps remain in our knowledge of how soil pollution affects the population levels, diversity and activity of PGPR. Our study is part of an ongoing cooperative project in China and the United States with the goal of identifying new PGPR and characterizing their role in the promotion of soil and plant health. Our study focused on rhizobacteria isolated from Chinese soils polluted with heavy metals, recalcitrant organic compounds, petroleum sewage or excessive salinization. We have hypothesized that certain groups of PGPR and strains harbouring biocontrol and growth-promoting genes are cosmopolitan and conserved in soil microbiomes in agroecosystems worldwide regardless of the condition of the soil, and that plants enrich and support those PGPR best adapted to help them grow and resist diseases and other pests. Nothing is known about the composition of PGPR in soils from these polluted sites, thus making them ideal to test our hypothesis.
Our specific objectives were to: (i) isolate and identify PGPR, (ii) determine their biocontrol activity against takeall disease and Rhizoctonia root rot on wheat, (iii) determine their growth-promoting activity on rapeseed/canola, and (iv) characterize their growth-promoting traits and genes. Wheat and rapeseed/canola were selected for our study because China is the world's largest producer of both of these crops and they are often grown in rotation. Rhizoctonia root rot caused by Rhizoctonia solani AG-8 and take-all caused by Gaeumannomyces graminis var. tritici are two important soilborne diseases worldwide, against which wheat has no resistance Paulitz et al., 2002;Yang et al., 2011); however, biocontrol with PGPR is a viable approach to controlling them. Poor emergence is a chronic problem in rapeseed/canola production in both China and the United States and application of PGPR as seed treatments can greatly enhance shoot and root growth.
Here we report that strains of well-described PGPR species were among the common morphotypes from the rhizosphere of wheat and rapeseed grown in highly polluted soils in China. Strains of Lysobacter capsici, P. chlororaphis, Pseudomonas protegens and those related to P. brassicacearum had some of the greatest biocontrol and/or growth promoting activity. PGPR genes in these morphotypes included phlD, phzF, prnD, pltC and acdS, and traits included the ability to produce 2,4-diacetylphloroglucinol (2,4-DAPG), phenazines, pyrrolnitrin, pyoluteorin, cellulase, siderophores, protease and IAA.

Population densities and antagonistic activity of rhizobacteria from contaminated soils
The population sizes of total culturable aerobic, heterotrophic bacteria in the rhizosphere of wheat after three growth cycles in the greenhouse in the contaminated soils from five fields located in Hubei, Jiangxi and Jiangsu provinces of China ranged from 4.2 × 10 6 to 3.4 × 10 8 colony-forming units (cfu) g −1 fresh weight of root (Table 1). Populations recovered from rapeseed grown in the same soils ranged from 3.6 × 10 5 to 1.3 × 10 6 cfu g −1 fresh weight of root (Table 1). From the wheat and rapeseed rhizospheres, 1223 bacterial isolates (605 from wheat and 618 from rapeseed) were randomly selected from dilution plates of 1 /3 × King's medium B (KMB) and 1 /10 × tryptic soy agar (TSA). The percentage of isolates from wheat inhibitory to R. solani and from rapeseed inhibitory to Sclerotinia sclerotiorum varied among soils but averaged 24% and 23% respectively (Table 1). Isolates from the field contaminated with recalcitrant organic compounds in Hukou County, Jiangxi province were the least inhibitory to R. solani and S. sclerotiorum (6.7% and 15.1% respectively) as compared with isolates from the other soils.

Identity of the PGPR and their biocontrol and growth-promoting traits and genes
Twenty-four strains representative of the dominant morphotypes (Table 2) were screened for growth-promoting and biocontrol-related traits and genes. Of these 24 strains, 15, 11 and 20, respectively, inhibited the growth of G. graminis var. tritici ARS-A1, R. solani AG-8 C-1 and S. sclerotiorum MGSCF180002 (Table 3).
Results of traditional microbiological tests and analysis of partial 16S rDNA gene sequences (Table 2) revealed that five of the 24 strains were Gram-positive and closely related to Bacillus, Paenibacillus polymyxa and Microbacterium hydrocarbonoxydans. The rest of the strains represented two different clades of Gram-negative bacteria and were closely related to several species within the Pseudomonas fluorescens and Pseudomonas putida groups, and to Lysobacter capsici, Lysobacter enzymogenes, Serratia plymuthica and Delftia tsuruhatensis. We subjected strains of the P. fluorescens group to the multilocus typing scheme of Mulet and colleagues (2009;, which has proven to be useful for establishing phylogenetic relationships among pseudomonads and is based on sequences of 16S rDNA, gyrB and rpoD (Mulet et al., 2009;. The results of molecular typing identified strains TY1502, TY1508, KY4410 and TM1109 as P. protegens and strain KY5406 as P. chlororaphis (Fig. S1). Strains TY1205, TY1210, KM3113 and KY5404 were identified as members of the P. corrugata subgroup and were closely related but not identical to the wellstudied biocontrol species P. brassicacearum (Fig. S1).
All 24 strains produced one or more of the extracellular metabolites cellulase, protease and siderophores. Genes for the biosynthesis of the antibiotics 2,4-DAPG (phlD), phenazines (phzF), pyrrolnitrin (prnD) and/or pyoluteorin (pltC) were detected in strains of P. protegens, P. chlororaphis and/or S. plymuthica (Table 3) and the antibiotics themselves were detected in culture extracts analysed by thin-layer chromatography. These antibiotics are well known to function as major mechanisms of biocontrol of soilborne pathogens.
The amplified fragments of phlD, phzF, prnD and pltC from these strains were sequenced and used in phylogenetic analyses together with corresponding alleles previously deposited in GenBank. Phylograms inferred from phlD and pltC sequences revealed that both of these genes from strains TM1109, TY1502, TY1508 and KY4410 were closely related to their homologues from P. protegens (Figs S2 and S3). Strains TM1109, TY1502, TY1508 and KY4410 also carried a P. protegens-like prnD allele and therefore had a combination of antibiotic biosynthesis pathways similar to those of the model biocontrol agents P. protegens strains Pf-5 and CHA0 (Fig. S4). The second prnD-positive Pseudomonas strain, KY5406, also carried phzF (Table 3), a gene required for the production of phenazine antibiotics and highly conserved in all known phenazine producers. The prnD and phzF genes of KY5406 clustered with homologues from P. chlororaphis (Figs S4 and S5). Strain KY5406 also contained phzO (data not shown), a gene required for the biosynthesis of 2-hydroxylated phenazine compounds, indicating this strain is subsp. aureofaciens. Finally, two more prnD-positive strains, KM3407 and TM4307-1 (Table 3), carried prnD alleles similar to those present in pyrrolnitrin-producing strains of Serratia spp. (Fig. S4) and produced Prn in vitro (data not shown). In total, correlation of the antibiotic gene profiling with the taxonomic Crop and native plants and weeds were dug from the contaminated soil at each site, soil attached to the roots was removed, transported to Wuhan, and wheat or rapeseed was grown in the soil for three cycles. Population densities of total culturable, aerobic rhizosphere bacteria were determined by the end-point dilution assay and single colonies of bacteria were selected from plates of 1 /3 × KMB or 1 /10 × TSA at the end of the third growth cycle of wheat or rapeseed. placement of the strains by partial 16S rRNA gene sequence analysis identified the antibiotic-producing strains as members of P. protegens, P. chlororaphis subsp. aureofaciens and S. plymuthica. 1-aminocyclopropane-1-carboxylic acid (ACC) deaminase and IAA production by PGPR are important determinants in plant-bacterial interactions, and secretion of these compounds can induce plants to produce longer and more vigorous roots (Penrose and Glick, 2003). Delftia tsuruhatensis TM3205 and Pseudomonas strains KM3113, KY5404, TY1210, TY1205, KY3201 and TY3101 produced IAA in culture. PCR analysis demonstrated that the aforementioned Pseudomonas strains were positive for acdS, a gene required for the production of ACC deaminase (Table 4), and the sequences of those strains closely resembled those from other fluorescent pseudomonads (Fig. S6). All acdS sequences from the Chinese strains formed a distinct clade within the Pseudomonas group and clustered tightly with acdS from Pseudomonas sp. 2-3 (GenBank Accession No. EU520401). All acdS-positive strains grew in defined medium supplemented with ACC as a sole nitrogen source.

Biocontrol and growth-promoting activity of the rhizobacterial strains
A subset of strains with the highest antagonistic activity and representative of the dominant morphotypes was selected to determine their biocontrol activity against take-all and Rhizoctonia root rot of wheat under greenhouse conditions in sterile, pasteurized and raw soils from Wuhan, China and Quincy, Washington. Most of the isolates provided some levels of reduction in root disease as compared with the pathogen-inoculated controls (Tables 5-7). However, strains identified as L. capsici, L. enzymogenes, P. protegens and P. chlororaphis provided the best and most consistent disease suppression across experiments and pathogens. Pseudomonas protegens TM1109 was especially effective in biocontrol of both take-all and Rhizoctonia root rot. These results indicate that indigenous rhizobacteria from the contaminated soil have the capacity to provide protection of plants against soilborne fungal pathogens. Surprisingly, application of S. plymuthica TM4307-1, which highly antagonized the pathogens in vitro (Table 3), resulted in significantly higher levels of Rhizoctonia root rot (Table 6). We speculate that TM4307-1 produces yet unidentified phytotoxic metabolite(s) since treatment of both wheat and canola with this bacterium resulted in significant stunting and bleaching of the seedlings and root damage. Twenty-two of the 24 isolates were tested for the ability to promote canola root growth in growth pouch assays. The two S. plymuthica strains were eliminated from this assay because they enhanced disease when applied to wheat and caused stunting. Thirteen strains increased root growth significantly (P = 0.05) in at least one of the three experiments (Table 4). Pseudomonas protegens and P. brassicacearum-like strains demonstrated the most consistent capacity to promote root growth and, on average, increased root length in canola by 41% to 42%, respectively, as compared with the non-treated controls. All P. brassicacearum-like isolates produced IAA and carried acdS. In contrast, strains of P. protegens had neither trait. Other tested Pseudomonas strains (i.e. KY3201, TY3101, Ty3116w, Ty5511 and KM2402) increased root length consistently but not always significantly, whereas strains of Bacillus, Paenibacillus and Lysobacter had no growth promotion activity. These results indicate that rhizobacteria from the contaminated soil have the capacity to directly enhance plant growth.

Tolerance to abiotic stressors
Based on the results from the biocontrol and growth promotion assays, 10 of the 24 strains were selected and tested for tolerance to two heavy metals and high levels of salt by growing them in LB broth supplemented with CdCl 2, CuCl2 and NaCl. All 10 strains tolerated high levels of salt and grew well in up to 5-7% NaCl. Similarly, these 10 strains grew in the presence of heavy metals and demonstrated minimal inhibitory concentrations (MICs) to Cu 2+ of > 4 mM and Cd 2+ of > 1 mM. Overall, P. protegens TM1109 was the most tolerant to these stressors (Table 8).

Discussion
Plants have evolved along with a rhizosphere microbiome that contributes to the growth and health of the plant (Pieterse et al., 2014). Beneficial microbes that directly promote growth or defend roots are recruited by the roots from the bulk soil, making soil type an important determinant of the composition of the rhizosphere microbiome. However, the plant also strongly modulates the shape of the microbial community because both the quality and quantity of root exudates are regulated by the plant genotype (Bais et al., 2006;Hartmann et al., 2009). Our Table 3. Biocontrol related genes and traits of Chinese strains and in vitro inhibition of soilborne pathogens a .
results support the hypothesis that certain groups of PGPR and biocontrol and growth-promoting genes are cosmopolitan and conserved in soil microbiomes worldwide. Driven by rhizodeposition, the plant enriches and supports PGPR best adapted to colonize the roots, antagonize pathogens, stimulate growth and/or initiate pattern-triggered immunity. We tested our hypothesis by characterizing culturable PGPR in the rhizospheres of wheat and rapeseed grown in soils from polluted fields, and assessed their growth-promoting and biocontrol abilities. After the most abundant bacterial morphotypes were isolated, the only other criterion we used for selecting the rhizobacteria was the ability to inhibit G. graminis var. tritici, R. solani or S. sclerotiorum. This was done because in vitro inhibition of plant pathogens is a characteristic commonly found in many PGPR strains (Haas and Défago, 2005) even though antibiosis does not predict biocontrol or growthpromoting activity. The 24 strains described here were shown to be most closely related to species known to  have members with PGPR activity. Common among these strains were genes and traits that are well known to contribute to biocontrol and growth-promoting activity (Haas and Défago, 2005;Glick et al., 2007;Loper et al., 2007;Lugtenberg and Kamilova, 2009). For example, phzF, detected in P. chlororaphis subsp. aureofaciens KY5406, is a key gene in the biosynthesis of phenazine-1carboxylic acid (PCA) and serves as a marker for the capacity of PGPR to produce one or more phenazines, a large class of broad-spectrum antibiotics produced by a variety of bacteria (Mavrodi et al., 2006;. Yang and colleagues (2011) showed that phzF-containing and PCAproducing Pseudomonas spp. colonized wheat grown in Hebei and Jiangsu provinces, China, and that these strains suppressed take-all. In the Pacific Northwest of the United States, PCA-producing pseudomonads are abundant on the roots of dryland wheat and suppress take-all and Rhizoctonia root rot (Thomashow and Weller, 1988;Mavrodi et al., 2012). Phenazine-producing PGPR also contributed to the natural suppressiveness of soils in the Châteaurenard region of France to Fusarium wilt (Mazurier et al., 2009). phlD is a key gene in the  biosynthesis of the broad-spectrum antibiotic 2,4-DAPG in Pseudomonas spp. and 2,4-DAPG producers are responsible for the natural suppression of take-all of wheat in the United States and black root rot of tobacco in soils of the Morens region of Switzerland . 2,4-DAPG also induces systemic resistance in plants to pathogens (Weller et al., 2012). prnD and pltC are genes that encode for the biosynthesis of pyrrolnitrin and pyoluteorin, antibiotics that are highly active against Rhizoctonia and Pythium species respectively (Loper et al., 2007). The gene acdS encodes ACC deaminase, which promotes plant growth and development by decreasing ethylene levels in the plant. ACC deaminaseproducing PGPR convert the ethylene precursor ACC to α-ketobutyrate and ammonia and can relieve stress resulting from pathogens, polyaromatic hydrocarbons, heavy metals, salt and drought (Glick et al., 2007).
To validate the biocontrol activity of our strains, we used the Rhizoctonia root rot-and take-all-wheat pathosystems. Strains of P. protegens, P. chlororaphis subsp. aureofaciens and L. capsici provided the most consistent disease suppression and members of all three species are biocontrol agents of a broad range of diseases. For example, P. protegens Pf-5 produces a wide arsenal of antibiotics and toxins and approximately 6% of its genome is devoted to the production of secondary metabolites (Paulsen et al., 2005;Loper et al., 2007). Strain CHA0, isolated from a tobacco black root rot suppressive soil, is another notable P. protegens with exceptional biocontrol activity (Ramette et al., 2011). Our four P. protegens strains, TM1109, TY1502, TY1508 and KY4410, contained the three antibiotic biosynthesis genes, phlD, prnD and pltC, and produced the antibiotics that are characteristic of Pf-5 and CHA0.
To test our strains for the ability to directly promote plant growth, we used the canola root elongation assay, a model system commonly employed to demonstrate direct growth-promoting activity by PGPR (Patten and Glick, 2002). The most consistent and effective growth promoters were strains of Delftia and Pseudomonas that produced IAA and, in the case of P. fluorescens-and P. brassicacearum-like strains, harboured acdS. Of these species, only P. fluorescens and P. brassicacearum have been reported to have direct growth-promoting ability (Belimov et al., 2007;Glick et al., 2007). Members of these two species also are biocontrol agents (Loper et al., 2012). Pseudomonas brassicacearum has commonly been isolated from Arabidopsis thaliana, Brassica napus  and wheat (Ross et al., 2000). Pseudomonas brassicacearum strain Am3, with ACC deaminase activity, promoted root elongation of Indian mustard and increased root and shoot biomass of rapeseed and pea (Belimov et al., 2001;Safronova et al., 2006). A mutation in acdS resulted in a loss of growthpromoting activity in P. brassicacearum (Belimov et al., 2007).
We tested 10 strains for tolerance to heavy metals and high levels of salt by growing them in broth supplemented with CdCl2, CuCl2 and NaCl. Unfortunately, there are no records of how long the sampled fields have been polluted but discussions with local inhabitants indicated that some fields may have been in that condition for decades. Not surprising, all of the PGPR strains tested tolerated high levels of salt and demonstrated MICs to Cu 2+ and Cd 2+ on par or higher than those previously reported for strains adapted to survival in the presence of acute levels of heavy metal contamination of soil (Thomas et al., 2008;Altimira et al., 2012). Thus, our results confirm that PGPR species have adapted not only to a wide variety of cropping systems and crops but also to adverse environmental conditions. Of particular note are our strains of P. protegens, which appeared to be especially well adapted to a wide variety of pollutants. Perhaps not surprisingly, analysis of the genome of P. protegens Pf-5 identified a considerable collection of efflux systems that provide protection against a range of toxic metabolites (Loper et al., 2007). Evidence is accumulating that strains of P. protegens form the cornerstones of communities of microbial defenders of roots in rhizosphere microbiomes globally.
In conclusion, Chinese farmers often must utilize all arable land, even polluted fields, because of the challenges of feeding a huge population. We inventoried PGPR in contaminated soils from sites representative of polluted fields supporting agricultural production. These sites were especially interesting because their microbiology previously had not been studied. Our findings demonstrate that well-described PGPR species and biocontrol and growth-promoting genes are common in the rhizosphere microbiome of plants grown in polluted Chinese soils. PGPR from such soils may be the best source of strains for commercial development given that in China nearly 50 million acres of arable land suffer from some type of soil pollution (Zhou, 2013). With the world population expected to be 9-11 billion by 2050, the challenge is to identify crop cultivars and cropping systems that make fuller use of indigenous and introduced PGPR for pest protection and growth promotion since the amount of arable land and water available for food production will likely continue to decrease.

Bacterial strains, soilborne pathogens and growth media
Bacterial strains used in this study are listed in Table 2. Pseudomonas isolates were grown in KMB or one-thirdstrength KMB ( 1 /3 × KMB) agar or broth at 28°C as described by Mavrodi and colleagues (2012). Other strains were grown in TSA or tryptic soy broth (TSB) or these media at one-tenth strength ( 1 /10 × TSB or TSA)  at 28°C. When needed to inhibit fungi, these media were supplemented with cycloheximide (100 μg ml −1 ). All strains were stored in 60% glycerol at −80°C. The fungal wheat root patho-gens Gaeumannomyces graminis (Sacc.) von Arx and Olivier var. tritici Walker isolate ARS-A1  and R. solani Kühn [teleomorph Thanatephorus cucumeris (Frank) Donk] AG-8 isolate C-1 (Huang et al., 2004), and the rapeseed/canola pathogen S. sclerotiorum (Lib.) de Bary isolate MGSCF180002 (from rapeseed) were grown at room temperature on one-fifth-strength homemade potato dextrose agar ( 1 /5 × PDA) . In vitro inhibition assays of pathogens by PGPR were conducted on full-strength PDA.

Sample collection
Soil was sampled from five polluted sites in China: farmland contaminated with heavy metals in Daye County, Hubei Province; a riverside field at a recalcitrant organic sewage outlet in Gold Sand Bay Industrial Park, Hukou County, Jiangxi Province; farmland beside a petroleum products sewage outlet in Jinmen City, Hubei Province; and saline-alkali fields on two separate seaside farms in Qidong City (near Dongyuan and Yinyang respectively), Jiangsu Province. At each site, plants were dug to a depth of 30 cm from five random locations. Two kilograms of soil surrounding the roots of plants from each location was placed in plastic bags, transported to the laboratory and stored at 4°C until processed as described below.

Wheat and rapeseed greenhouse cycling assays in contaminated soils
Soil from each location was sieved, put into three square pots (6.5 cm high × 7 cm wide) and 10 ml of a metalaxyl (Syngenta, Wilmington, DE, USA) solution (75 mg l −1 ) was added to prevent damping-off disease caused by indigenous Pythium spp. Each pot served as a replicate. Six surfacesterilized seeds of wheat (cv. Zhengmai 9023) or rapeseed (cv. Xiangyou 571) were pre-germinated for 48 h, sown in each pot and then covered with a 1.5 cm-thick layer of sterile vermiculite. Each pot was then given 30 ml of water. The pots were covered with a plastic bag for 4 days and incubated in a greenhouse (15-18°C; dark/light cycle, 12 h). Each pot received 30 ml of water two times weekly and diluted (1:3, vol/vol) Hoagland's solution once weekly. After 3 weeks, the seedlings were removed from the pots, roots were excised and soil and roots from all of the pots of the same treatment were mixed and added back to the pots and sown as before with wheat or rapeseed to begin a second growth cycle. At the end of the third growth cycle, roots from two randomly chosen plants from each pot were selected for enumeration of populations of total culturable aerobic, heterotrophic rhizobacteria (total bacteria) by the end-point dilution assay as described below.

Enumeration and isolation of rhizosphere bacteria
Roots with adhering rhizosphere soil of two plants from the same pot were excised and placed in a 50 ml screw-cap tube with 10 ml of sterile distilled water, vortexed (1 min) and sonicated in an ultrasonic cleaner (1 min) to dislodge rhizobacteria . The dilution endpoint assay with 1 /10 × TSB plus cycloheximide was used to determine the population density of total bacteria as described by Mavrodi and colleagues (2012). The 96-well microplates were incubated at room temperature in the dark, and after 72 h, each well was scored for growth. At the end of the 3rd growth cycle, rhizobacteria were isolated by dilution plating of the root washings onto 1 /3 × KMB agar and 1 /10 × TSA, each with cycloheximide. Plates were incubated at room temperature for 2 to 3 days or until colonies were visible. Colonies of different morphotypes were re-streaked onto 1 /3 × KMB and 1 /10 × TSA plates and stored in 60% glycerol at −80°C.

Pathogen inhibition in vitro
Inhibition of soilborne pathogens by rhizobacteria was tested on PDA by two approaches. In the first, aliquots (20 μl) from overnight broth cultures were introduced into a hole in the agar cut with a 5 mm cork borer, 1 cm from the edge of the Petri dish. A 5 mm plug from a culture of R. solani AG-8 or S. sclerotiorum grown on 1 /5 × PDA was placed in the centre of the plate. Plates were incubated at 28°C and scored after 4 days by measuring the distance between the edges of the bacteria and the fungal mycelium. Four isolates were tested on each plate and each isolate was tested three times. In a second approach, aliquots (2 μl) from overnight cultures were spotted twice, 1 cm from the edge of a plate of PDA, and a plug of R. solani AG-8 was placed in the centre. For tests against G. graminis var. tritici, the plug of the fungus was placed in the centre of the plate 24 h before the bacteria were spotted. The zone of inhibition was measured 5 days later. Each isolate was tested four times.

Morphological and physiological characteristics of isolates
Selected isolates were characterized based on morphological and biochemical tests as described by Zhao and He (2012). Morphological characteristics included colony morphology, pigmentation, cell shape and Gram stain reaction. Biochemical and physiological characterization included tests for starch hydrolysis, gelatin liquefaction, catalase production, pectin decomposition, indole production and production of phenazine-1-carboxylic acid, 2,4-DAPG, pyrrolnitrin and pyoluteorin. Cellulase activity was determined on carboxymethylcellulose (CMC) medium (per 1 l: peptone, 10 g; yeast extract, 10 g; sodium CMC, 10 g; NaCl, 5 g; KH2PO4, 1 g; agar, 15 g; pH 7.0). A 10 μl aliquot of an overnight bacterial culture was spotted on sterile filter paper (5 mm diameter, 0.5 mm thick) on the surface of CMC agar. To visualize the zone of hydrolysis, plates were incubated at 28°C for 48 h, flooded with an aqueous solution of 0.1% Congo red for 1 h and washed with 1 M NaCl (Ghose, 1987). Protease activity, demonstrated by casein degradation, was indicated by the appearances of zones of clearing around colonies grown on milk agar (per 1 l: dry skim milk, 100 g; agar, 20 g) after incubation at 28°C for 3 days. Siderophores were detected on CAS (chrome azurols) agar plates (Shin et al., 2011). Antibiotics (phenazine-1-carboxylic acid, pyrrolnitrin, pyoluteorin and DAPG) were isolated from bacterial cultures grown for 5 days on KMB agar by a modification of procedures described by Kraus and Loper (1992). Briefly, cells and spent agar (c. 15 ml) were macerated with a mortar and pestle in an equal volume of acetone. The acetone was evaporated in a hood overnight and the residual aqueous phase was extracted twice with equal volumes of ethyl acetate. The ethyl acetate was recovered, dried and the residue was dissolved in 50 μl of methanol. Samples (5 μl) of the extracts, standards of the four antibiotics, and positive control extracts from strains P. protegens Pf-5 and P. brassicacearum Q8r1-96 were spotted on silica gel (GHLF) thin layer chromatography plates (Analtech, Newark, DE, USA). The plates were developed with chloroform/acetone (9:1) and observed under UV light (254 nm).

Sequence and phylogenetic analysis
Sequence data were assembled and analysed using Geneious Pro v. 7.1.5 (Biomatters, Auckland, New Zealand). The 16S rDNA similarity searches and taxa assignments were carried out using the Classifier and SeqMatch tools available through the Ribosomal Database Project (http:// rdp.cme.msu.edu). For antibiotic, ACC deaminase, gyrB and rpoD gene sequences, database searches for similar DNA and protein sequences were carried out with NCBI's BLAST network service (http://www.ncbi.nlm.nih.gov/ BLAST). Sequences were aligned and neighbour-joining phylogenies were inferred using the Molecular Evolutionary Genetics Analysis software (MEGA) v. 5.2.2 (Tamura et al., 2007). DNA and protein distances were corrected by using the Kimura two-parameter (Kimura, 1980) and Jones-Taylor-Thornton (Jones et al., 1992) models of evolution, respectively, and bootstrap re-sampling with 1000 replicates was used to assess the reproducibility of clades in the inferred phylogenies.

Preparation of barley and oat grain inoculum
Inocula of G. graminis var. tritici and R. solani AG-8 were prepared as described by Yang and colleagues (2011). Briefly, oat or barley grains (250 g) and water (350 ml) were combined in a flask and autoclaved on each of two consecutive days. Sterilized grains were inoculated with pieces of agar cut from 1 /5 × PDA plates of R. solani or G. graminis var. tritici. After 21 days at room temperature, the colonized grains were tested for contamination, dried under a stream of sterile air and stored at 4°C. Prior to use, inoculum was fragmented and sieved into sizes, with particles of 0.25-0.5 mm added to soil.

Bacterial treatment of wheat seed
For biocontrol studies, seeds were coated with bacteria by methods similar to those of Yang and colleagues (2011). Strains were inoculated onto plates of KMB or TSA and incubated at room temperature for 24-48 h. Bacteria were scraped into 1.0 ml H2O, washed by centrifugation (twice for 3 min at 14 000 r.p.m.) and then suspended in H2O. The concentration of the cell suspension was adjusted based on optical density (600 nm), and an aliquot was mixed with a 2% solution of methylcellulose (MC) and deionized H2O. Wheat (130 seeds) was added into the mixture, shaken for 3 min and seeds were dried under a stream of sterile air. To determine the final cfus per seed, 10 seeds were placed in a tube with 10 ml of sterile water, vortexed (1 min) and sonicated (1 min). Population size as determined by the end-point dilution assay ranged from 10 4 to 10 7 cfu seed −1 .

Biocontrol activity of rhizobacteria
The biocontrol activity of strains was determined by several approaches, and against both Rhizoctonia root rot and takeall, in autoclaved, pasteurized (60°C, 30 min) and raw soil. For pot experiments, soil from a farm at Huazhong Agricultural University, Wuhan, China, was sieved, autoclaved and amended with barley grain inoculum of R. solani (1%, w/w). Each plastic pot (10 cm deep × 6 cm wide) was filled with a 5 cm-thick layer of sterile vermiculite followed by 8 g of infested soil. Four treated wheat seeds (cv. Yangmai 158) were sown on the surface of the soil and covered with a 1.5 cm layer of sterile vermiculite. Controls included: control 1 (CK1), soil not amended with inoculum and sown to nontreated seeds; control 2 (CK2), soil amended with inoculum and sown to seed coated with MC; and control 3 (CK3), soil amended with inoculum and sown to non-treated seed. Each pot was given 10 ml of 2.5 mg ml −1 metalaxyl (Syngenta Wilmington, DE, USA) to prevent Pythium damping-off. Pots were covered with a plastic sheet to reduce evaporation, incubated at room temperature (22°C) overnight and then transferred to a greenhouse (17 ± 5°C). The plastic was removed when the tips of the shoots were visible. Each pot was watered twice weekly with water and once with 1 /3 × Hoagland's solution (macro-elements only). Treatments were replicated five times and arranged in a randomized complete block design; each pot served as a replicate. After 3 to 4 weeks, the seedlings were removed from the pots, washed free of soil, and the plants were evaluated for disease severity on a scale of 0 to 8, where 0 = no lesions evident and 8 = seedling dead or nearly so (Huang et al., 2004). The length of the seedling shoot also was measured. Biocontrol studies of Rhizoctonia root rot and take-all were also conducted using the tube assay as described by Yang and colleagues (2011) with pasteurized or raw Shano sandy loam (Quincy virgin) from a non-cropped site near Quincy, WA, USA. Plastic tubes (2.5 cm diameter, 16.5 cm long) with a hole in the bottom were hung in plastic racks (200 per rack). Each tube had a cotton ball placed in the bottom, followed by a 6.5 cm-thick column of sterile vermiculite, and then soil (10 g), with or without inoculum of R. solani at 1.0 % (w/w) or G. graminis var. tritici at 0.7% (w/w). A 1.5 cm-thick layer of vermiculite was placed over the infested soil, followed by three bacteria-treated wheat seeds (cv. Louise) and finally another 1.5 cm-thick topping of vermiculite. Each tube then received 10 ml of water with metalaxyl. Racks of tubes were covered with plastic sheets until shoots emerged and were incubated in a growth chamber (15-18°C; dark/light cycle 12 h). Each cone was watered twice weekly with water and once weekly with 1 /3 × Hoagland's solution. Seedlings were harvested after 3-4 weeks and washed under a stream of water. The severity of Rhizoctonia root rot and take-all was evaluated on a scale of 0-8 .

Detection of IAA
IAA production was determined colorimetrically (Patten and Glick, 2002;Egamberdieva et al., 2008). Strains were incubated with shaking for 48 h at 28°C in KMB broth alone or supplemented with 500 μg per millilitre of tryptophan. Cultures were centrifuged at 13 000 r.p.m. for 10 min and a 1 ml aliquot of the supernatant was mixed vigorously with 4 ml of Salkowski reagent. The mixture was incubated at room temperature for 25 min and the absorbance of the pink colour was measured spectrophotometrically at 540 nm. The concentration of IAA was determined by using a calibration curve of pure IAA (0, 20, 40, 60, 80 and 100 μg ml −1 ) as a standard. Readings of < 1 μg ml −1 per OD600 unit were considered as negative for IAA production.

Canola root elongation assay
A gnotobiotic canola root elongation assay was used to determine the ability of strains to promote growth (Egamberdieva et al., 2008). An aliquot of an overnight culture was centrifuged, the cell pellet was washed twice and suspended in 1 ml of sterile 0.03 M MgSO 4. The suspension was adjusted to an OD600 of 0.15. Surface-sterilized spring canola seeds (cv. InVigor 8440) were soaked for 1 h in 700 μl of a bacterial suspension. Depending on the isolate, the concentration ranged from 10 4 to 10 6 cfu seed −1 . Growth pouches were made using 1-gallon Ziploc plastic bags and sterile germination paper soaked with 50 ml of 1 /5 × Hoagland's solution. Twenty-five seeds of a single bacterial treatment were distributed along the germination paper and the paper was placed inside of the bag. Individual bags were then hung vertically, covered with aluminium foil for 1-2 days and incubated at 20°C in a dark/light cycle of 12 h. The length of the primary roots was measured 5 days after sowing. Roots of seeds that failed to germinate 2 days after sowing were not measured. This study was conducted five times with three representative experiments shown.

Tolerance of bacteria to abiotic stress
The tolerance of selected strains to copper, cadmium and NaCl was determined in Luria-Bertani (LB) broth amended with CuCl2 to give concentrations from 3 to 6 mM; CdCl2 to give concentrations from 0.4 to 2.0 mM; and NaCl to give concentrations from 0% to 8%. Each well of a 96-well microplate was filled with 200 μl of LB supplemented with the heavy metal or salt solutions and inoculated with 2 μl of an overnight culture of each isolate (OD600 = 0.1). Each isolate at each concentration of heavy metal or salt was replicated eight times. The microplates were incubated at room temperature in the dark for 72 h and assayed spectrophotometrically at 600 nm using a model 680 microplate reader (Bio-Rad). An OD600 > 0.1 was scored as positive for growth. The MIC was defined as the lowest concentration that completely inhibited the growth of the isolate.

Statistical analysis
Comparisons of means of root disease ratings were performed using standard analysis of variance (ANOVA) followed by Kruskal-Wallis all pairwise comparisons (P ≤ 0.05) (Statistix 8.1, Analytical Software, Tallahassee, FL, USA). Mean comparisons of plant height and root length among treatments were determined by ANOVA followed by using either the Fisher's protected least significant difference test (P = 0.05) or the Kruskal-Wallis all pairwise test (P = 0.05). Bacterial population sizes were log transformed before analysis.

Supporting information
Additional Supporting Information may be found in the online version of this article at the publisher's web-site: Fig. S1. Neighbour-joining phylogeny inferred from concatenated housekeeping genes 16S, gyrB and rpoD. The concatenated housekeeping gene data set contained 2911 characters. Evolutionary distances were estimated using the Kimura two-parameter model of nucleotide substitution. Indels were ignored in the analysis and reproducibility of clades was assessed by bootstrap resampling with 1000 pseudoreplicates. The branch lengths are proportional to the amount of evolutionary change. The scale bars indicate substitution per site. Subgroups within the P. fluorescens lineage are defined according to Mulet and colleagues (2010). Fig. S2. Neighbour-joining phylogeny inferred from data for aligned 578-bp fragments of phlD sequences of fluorescent Pseudomonas spp. Indels were ignored in the analysis, and evolutionary distances were estimated using the Kimura twoparameter model of nucleotide substitution. Bootstrap values greater than 60% are indicated at the nodes, and the branch lengths are proportional to the amount of evolutionary change. Clades indicated by capital letters correspond to BOX-PCR genotypes of 2,4-diacetylphloroglucinol-producing Pseudomonas spp. as defined by De La Fuente and colleagues (2006). Fig. S3. Neighbour-joining phylogeny inferred from data for aligned 390-bp fragments of pltC sequences of fluorescent Pseudomonas spp. Indels were ignored in the analysis, and evolutionary distances were estimated using the Kimura twoparameter model of nucleotide substitution. Bootstrap values greater than 60% are indicated at the nodes, and the branch lengths are proportional to the amount of evolutionary change. Fig. S4. Neighbour-joining phylogeny inferred from data for aligned 745-bp fragments of prnD sequences of pyrrolnitrinproducing species. Indels were ignored in the analysis, and evolutionary distances were estimated using the Kimura twoparameter model of nucleotide substitution. Bootstrap values greater than 60% are indicated at the nodes, and the branch lengths are proportional to the amount of evolutionary change. Clades C1 through C2.4 correspond to distinct groups of pyrrolnitrin-producing bacteria as defined by Costa and colleagues (2009). Fig. S5. Neighbour-joining phylogeny inferred from data for aligned 391-bp fragments of phzF sequences of fluorescent Pseudomonas spp. Indels were ignored in the analysis, and evolutionary distances were estimated using the Kimura twoparameter model of nucleotide substitution. Bootstrap values greater than 60% are indicated at the nodes, and the branch lengths are proportional to the amount of evolutionary change. Clades A1 through A5 correspond to distinct groups of phenazine-producing bacteria as defined by Mavrodi and colleagues (2010). Fig. S6. Neighbour-joining phylogeny inferred from data for aligned 714-bp fragments of acdS sequences of ACCdeaminase-producing bacteria. Indels were ignored in the analysis, and evolutionary distances were estimated using the Kimura two-parameter model of nucleotide substitution. Bootstrap values greater than 60% are indicated at the nodes, and the branch lengths are proportional to the amount of evolutionary change. Clades A through C correspond to distinct groups of ACC-deaminase-producing bacteria as defined by Blaha and colleagues (2006).