Preferential Association of Endophytic Bradyrhizobia with Different Rice Cultivars and Its Implications for Rice Endophyte Evolution

Abstract

Plant colonization by bradyrhizobia is found not only in leguminous plants but also in nonleguminous species such as rice. To understand the evolution of the endophytic symbiosis of bradyrhizobia, the effect of the ecosystems of rice plantations on their associations was investigated. Samples were collected from various rice (Oryza sativa) tissues and crop rotational systems. The rice endophytic bradyrhizobia were isolated on the basis of oligotrophic properties, selective medium, and nodulation on siratro (Macroptilium atropurpureum). Six bradyrhizobial strains were obtained exclusively from rice grown in a crop rotational system. The isolates were separated into photosynthetic bradyrhizobia (PB) and nonphotosynthetic bradyrhizobia (non-PB). Thai bradyrhizobial strains promoted rice growth of Thai rice cultivars better than the Japanese bradyrhizobial strains. This implies that the rice cultivars possess characteristics that govern rice-bacterium associations. To examine whether leguminous plants in a rice plantation system support the persistence of rice endophytic bradyrhizobia, isolates were tested for legume nodulation. All PB strains formed symbioses with Aeschynomene indica and Aeschynomene evenia. On the other hand, non-PB strains were able to nodulate Aeschynomene americana, Vigna radiata, and M. atropurpureum but unable to nodulate either A. indica or A. evenia. Interestingly, the nodABC genes of all of these bradyrhizobial strains seem to exhibit low levels of similarity to those of Bradyrhizobium diazoefficiens USDA110 and Bradyrhizobium sp. strain ORS285. From these results, we discuss the evolution of the plant-bradyrhizobium association, including nonlegumes, in terms of photosynthetic lifestyle and nod-independent interactions.

INTRODUCTION

Plant infection and colonization by bradyrhizobia are found in not only leguminous plants but also nonleguminous species such as rice (1, 2). However, the symbiosis mechanisms governing the relationship between Bradyrhizobium bacteria and rice have been absolutely unclear although the evolution of the bradyrhizobium-rice association may have occurred earlier than the bradyrhizobium-legume symbiosis. In addition, rice is the most important food crop in Asia. High-yield rice production requires huge amounts of nitrogen fertilizers. Biological nitrogen fixation (BNF) from rice root-associated bacteria has a great potential to improve sustainable rice production. Rice-legume rotational cropping systems are useful for rice production since legumes can be planted after the rice season, and nitrogen from the legumes can be provided for rice. Bradyrhizobia are well recognized for their ability to fix atmospheric dinitrogen into ammonia in the nodules of legumes, thus providing ammonia to host plants. The bradyrhizobium-legume symbiosis has been reported to increase the legume productivity of agricultural farms (3–5).

Bradyrhizobium has been well defined as an oligotrophic bacteria which can survive under nutrient-deprived conditions (6, 7). In addition, a renewed interest in endophytic diazotrophs in gramineous plants has also arisen because of their occurrence mainly within plant tissues. The endophytic diazotrophs actually promote plant growth (8). However, the relationships of endophytic Bradyrhizobium with rice cultivars and/or habitats have never been elucidated. Thus, the host preference for endophytic bradyrhizobia and the promotion of rice growth are critical aspects to be determined before these bacteria are used in fields.

To obtain natural endophytic bradyrhizobia from rice, bradyrhizobial strains were isolated on the basis of the oligotrophic property. Siratro (Macroptilium atropurpureum) has been used as a standard host plant to screen the nod-containing Bradyrhizobium from homogenized rice tissues and soil samples (9). A selective medium (Bradyrhizobium japonicum selective medium, or BJSM) (10) was also applied for Bradyrhizobium isolation. In addition, to understand the mechanism of the plant-bradyrhizobium association, rice endophytic bradyrhizobia were tested with various types of legumes normally found in rice fields.

MATERIALS AND METHODS

Rice cultivars, soil sampling, and endophytic bradyrhizobial isolation.

The seven different rice (Oryza sativa L.) cultivars and rice rhizospheric soil samples used in this study are listed in Table 1. Rice roots and paddy soils were collected and immediately transferred to the laboratory in polyethylene boxes at 4°C to obtain endophytic bradyrhizobia. The roots and stems were chemically sterilized (in 3% sodium hypochlorite for 5 min, followed by soaking in 70% ethanol for 5 min). The rice roots were thoroughly rinsed with sterilized distilled water at least five times and cut into 4- to 5-cm-long sections. As a control to check superficial contamination for each individual plant, 200 μl of water from the final rinse was spread on plate count agar (PCA) medium. No individual root samples containing bacterial contaminations on PCA plates were used further for endophytic bradyrhizobial isolation (11). Rice root tissues were aseptically crushed in 1 ml of sterilized water by sterilized mortar and pestle. The root homogenates were used for further isolation of endophytic bacteria.

TABLE 1.

Rice cultivars used in this study, geographical origin, and relevant characteristics

Rice species and cultivar	Geographical origin	Relevant characteristics	Source
O. sativa L. subsp. indica
Khao Dowk Mali 105	Chon Buri, Thailand	Photosensitive, lowland rice	Sommai Lertnaa
Leuang Yai 148	Chiang Mai, Thailand	Photosensitive, lowland rice	Sommai Lertna
Lu Ni	Nakhon Ratchasima, Thailand	Photosensitive, floating rice	Sommai Lertna
Pathum Thani 1	Pathum Thani, Thailand	Nonphotosensitive, lowland rice	Sommai Lertna
Sang Yod Phatthalung	Phatthalung, Thailand	Photosensitive, lowland rice	Sommai Lertna
Kasalath	Japan	Nonphotosensitive, lowland rice	Tadashi Satob
O. sativa L. subsp. japonica
Nipponbare	Japan	Nonphotosensitive, lowland rice	Tadashi Sato

^aNakhon Ratchasima Rice Research Center, Phimai, Nakhon Ratchasima, Thailand.

^bGraduate School of Life Sciences, Tohoku University, Sendai, Japan.

To obtain the endophytic bradyrhizobia from rice tissues, the bradyrhizobial strains were isolated using three different protocols (Fig. 1). First, 1 ml of root homogenate was directly inoculated onto germinated siratro (M. atropurpureum) seeds grown in plastic growth pouches. Then, after 1 month, each nodule was crushed and streaked on arabinose-gluconate (AG) agar plates (12). Second, oligotrophic-based isolation was carried out to reduce growth of other microbial contaminants. One milliliter of the root homogenate was added to 50 ml of sterilized water for 1 month, and then 0.1 ml of solution was spread on BJSM plates (10). Finally, 0.1 ml of root homogenate was directly spread on BJSM plates (Fig. 1). In order to select different strains of bacteria, BOX-A1R PCR was carried to screen for redundant strains (13). Bacterial strains showing different BOX-A1R PCR patterns were selected for bacterial identification by 16S rRNA gene analysis (Table 2).

FIG 1.

Scheme for the isolation of endophytic bradyrhizobial strains. Strain names that include a “P” (e.g., SUT-PR9) represent photosynthetic bradyrhizobial strains. The numbers in parentheses represent the numbers of bacterial strains isolated from each procedure.

TABLE 2.

Strains and plasmids used in this study, sampling sites, and relevant characteristics

Strain or plasmid	Province of origin (coordinates)	Relevant characteristic(s) (isolate source)	Source or reference
Strains
Bradyrhizobium sp.
SUT-R3	Chiang Mai, Thailand (19°13′11.3″N, 98°50′51.7″E)	Rice root isolate (rice with crop rotation)	This study
SUT-PR9	Uttaradit, Thailand (17°39′1.9″N, 100° 8′34.2″E)	Rice root isolate/paddy soil isolate (rice with crop rotation)	This study
SUT-PR48	Surin, Thailand (14°39′41.3″N, 103°17′09.5″E)	Rice root isolate (rice with crop rotation)	This study
	Prachuap Khiri Khan, Thailand (12°36′43.4″N, 99°51′40.5″E)	Rice root isolate (rice with crop rotation)
SUT-R55	Lampang, Thailand (18°49′42.5″N, 99°56′37.2″E)	Rice root isolate (rice with crop rotation)	This study
SUT-PR64	Khonkaen, Thailand (16°14′13.5″N, 102°31′31.1″E)	Rice root isolate (rice with crop rotation)	This study
SUT-R74	Chiangrai, Thailand (19°22′31.7″N, 99°30′5.9″E)	Rice root isolate/paddy soil isolate (rice with crop rotation)	This study
DOA1	Lopburi, Thailand	A. americana nodule isolate	27
DOA9	Lopburi, Thailand	A. americana nodule isolate	27
SUTN9-2	Lampang, Thailand (18°49′42.5″N, 99°56′37.2″E)	A. americana nodule isolate	27
WD16	Kashimadai City, Miyagi Prefecture, Japan	Rice root isolate (rice monoculture)	Ryo Shinoda, unpublished data
RP5	Kashimadai City, Miyagi Prefecture, Japan	Rice root isolate (rice monoculture)	Ryo Shinoda, unpublished data
RP7	Kashimadai City, Miyagi Prefecture, Japan	Rice root isolate (rice monoculture)	Ryo Shinoda, unpublished data
SUT-R3GUS		SUT-R3 marked with mTn5SSgusA20 (pCAM120); Smr Spr	This study
SUT-PR9GUS		SUT-PR9 marked with mTn5SSgusA20 (pCAM120); Smr Spr	This study
SUT-PR48GUS		SUT-PR48 marked with mTn5SSgusA20 (pCAM120); Smr Spr	This study
SUT-R55GUS		SUT-R55 marked with mTn5SSgusA20 (pCAM120); Smr Spr	This study
SUT-PR64GUS		SUT-PR64 marked with mTn5SSgusA20 (pCAM120); Smr Spr	This study
SUT-R74GUS		SUT-R74 marked with mTn5SSgusA20 (pCAM120); Smr Spr	This study
Escherichia coli S17-1		pro recA RP4-2 (Tcs::Mu) (Kms::Tn7), Mob+	48
Plasmids
pRK404		Broad-host-range vector	49
pCAM120		mTn5SSgusA20 in pUT/mini-Tn5	Shin Okazakia; 50
pBjGroEL4::dsRed2		mTn5dsRed2 in pUT/mini-Tn5	1
pRK2013		Helper plasmid	23

^aDepartment of International Environmental and Agricultural Science, Graduate School of Agriculture, Tokyo University of Agriculture and Technology, Tokyo, Japan.

Phenotypic characteristics and nitrogen fixation assay.

The color and morphology of bradyrhizobial colonies were observed during cultivation on HM (HEPES-morpholineethanesulfonic acid [MES]) agar medium (14). In order to detect photosynthetic pigments from the bradyrhizobial strains, cells were grown aerobically at 30°C for 7 days under a 12-h light/12-h dark cycle. To determine bacterial pigment production, cell pellets were extracted in the dark with cold acetone-methanol (7:2 [vol/vol]) for 30 min (15). Absorbance of the supernatant was observed at wavelengths ranging from 350 to 850 nm. Another method to determine the bacterial pigment production, as described by Chaintreuil et al. (2), was also conducted. The bradyrhizobia were grown at 30°C on YEM modified agar medium (g/liter; K₂HPO₄, 0.66; sodium glutamate, 2.5; yeast extract, 2; NaCl, 0.05; MgSO₄·7H₂O, 0.1; CaCl₂·2H₂O, 0.04; FeCl₃, 0.004; MnSO₄, 0.01) for 7 days under aerobic conditions on 15-h light/9-h dark cycle. A lawn of bacteria was analyzed using a Beckman DU40 spectrophotometer (Cary WinUV scan). Absorption spectra were obtained by scanning over wavelengths ranging from 350 to 900 nm with medium speed. Photosynthetic pigment peaks were detected at 800 and 870 nm.

Symbiotic abilities of putative bradyrhizobial strains were determined in Leonard's jars containing sterilized vermiculite, and after that 1 ml of each bacterial strain, equivalent to 10⁷ cells, was inoculated onto germinated Aeschynomene indica seed. After 30 days, plants were harvested, and the roots were used to measure nitrogenase activity by means of an acetylene reduction assay (ARA) (16).

16S rRNA gene, housekeeping gene, and phylogenetic analyses.

The genomic DNA of bradyrhizobial strains was prepared from the purified bacterial strains grown in HM broth (17). The 16S rRNA gene was amplified using the primer pair fD1 and rP2 (18). The DNA primers for housekeeping genes and other DNA primers used are listed in Table 3. Gene fragments were amplified using a Go Taq Flexi DNA polymerase kit (Promega, Mannheim, Germany). The PCR products were purified using a Wizard SV gel and PCR cleanup system (Promega, Germany) and sequenced using the same primers as those for the PCR. DNA sequencing was carried out by Macrogen, Inc. (Seoul, South Korea). The DNA sequences were generated, and the most closely related sequences were obtained from the NCBI database. The nucleotide sequences were aligned using the Clustal W program, and the phylogenetic trees of the 16S rRNA gene and housekeeping gene sequences were constructed by the maximum-likelihood method using PhyML (19). The confidence levels were estimated for 1,000 replicates. For comparison, phylogenetic trees were also reconstructed by the distance neighbor-joining method (20) using the MEGA, version 4.1, package (21). DNA sequences of each gene from related strains in the family Bradyrhizobiaceae, of other rhizobia, and of outgroups were obtained from the NCBI database.

TABLE 3.

Primers used in this study

Identification method and target gene	Primer name (sequence)	GenBank accession no. and/or description	Reference	Annealing temp (°C)
16S rRNA	fD1	Universal primer for 16S rRNA gene from bacteria	46	55
	rP2
dnaK	TSdnaK4	FJ347237	47	53
	TSdnaK2
glnB	TSglnBF	FJ347346	47	53
	TSglnBR
recA	recA41F	BAC51020	48	53
	recA640R
BOX-A1Ra	BOX-A1R		13	63
nodA	nodAF28	NC_004463	27	45
	nodAR627
	nodAF25-ORS285	AF284858	27	46
	nodAR584-ORS285
nodB	nodBF26	NC_004463	27	47
	nodBR625
	nodBF73-ORS285	AF284858	27	53
	nodBR622-ORS285
nodC	nodCF195	NC_004463	27	45
	nodCR1394
	nodCF197-ORS285	AF284858	27	53
	nodCR1196-ORS285
pufM	pufM-F (5′-CGAGATCATCGGCCTCAACATGTTGG-3′)	Designed from pufM of Bradyrhizobium sp. BTAi1 (CP000494) and Rhodopseudomonas palustris (AB015977)		62
	pufM-R (5′-GAAGCCCATCGTCCAGCGCCAGAAC-3′)
bchL	bchL-F (5′-AAGCAYGACTCSACCTTCAC-3′)	Designed from bchL of Bradyrhizobium sp. BTAi1(CP000494) and Rhodopseudomonas palustris BisB5 (CP000283)		53
	bchL-R (5′-CGTTGAACTTGTCGATCTGG-3′)

^aRepetitive sequence-based PCR for bacteria.

GUS and DsRed-tagging of endophytic bradyrhizobia and monitoring of root colonization.

Bradyrhizobial strains were cultured in HEPES-MES (HM) medium at 30°C (14). The Escherichia coli strains were cultured at 37°C in Luria-Bertani (LB) medium (22). The medium was supplemented with the following antibiotics when appropriate: 30 μg/ml gentamicin for E. coli and 200 μg/ml streptomycin for Bradyrhizobium. Bradyrhizobial strains were tagged with mTn5SSgusA20 (pCAM120) (Table 2) and pBjGroEL4::DsRed2 (1) by triparental mating on HM agar plates using pRK2013 as a helper plasmid (23). The cell slurry was mixed with HM medium, and then 100 μl of cell suspension was plated on HM agar plates containing 100 μg/ml streptomycin, 100 μg/ml spectinomycin, and 50 μg/ml polymyxin B. The colonies formed were selected as transconjugants.

Seed coats of cultivated rice (Table 1) were removed, and seeds were surface sterilized with 3% NaHPO₄ for 5 min and shaken in 1% hydrogen peroxide solution for 60 min. Seeds were then washed five times with sterilized distilled water with shaking (15 min each). Surface-sterilized seeds were gnotobiotically germinated on a water-agar plate. After 3 days, rice seedlings were transferred into 80-ml tubes containing 10 ml of nitrogen-free plant nutrient solution (8). Three days after the seedlings were transferred into the growth medium, β-glucuronidase (GUS)-tagged bacteria were inoculated into the growth medium to a density of 10⁷ to 10⁸ cell/ml. Roots of 3-day-old seedlings were examined for bacterial colonization by GUS staining. For the detection of GUS expression by endophytes in plant tissue (24), the roots were embedded in 5% agarose gel, and 90-μm-thick sections were prepared using a vibratome (Microm HM 650V; Thermo Scientific, United Kingdom). The plant samples were then directly observed under a light microscope. Because of the possibility of the blue GUS stain diffusing and interfering with visualization of bacteria in plant tissues, DsRed-labeled strains were also used to confirm true bacterial localization. Then, bradyrhizobial localizations were observed under a confocal microscope (Nikon Confocal Microscope).

Plant experiments.

A Leonard's jar assembly was filled with sterilized vermiculite, and nutrient solutions were applied through a wick to provide nutrients to the plants (25). The whole apparatus was autoclaved (25 min at 120°C) prior to the transplantation of seedlings. Surface-disinfected rice seeds were germinated on sterilized filter sheets in a petri dish. Uniformly germinated seeds were transplanted into the Leonard's jar under aseptic conditions (three plants per Leonard's jar). One milliliter of 5-day-old inoculum (approximately 10⁷ CFU/ml) was inoculated onto the seedling at 2 days after transplanting. This experiment was conducted as three replicates per single bacterial isolate. Plants were grown under controlled environmental conditions of 28 ± 2°C on a 12-h light/12-h dark cycle at a light intensity of 180 μE/m²s and 50% humidity. Plant dry weights were measured after 28 days.

Enumeration of endophytic bradyrhizobia.

To enumerate the endophytic bradyrhizobia, the rice roots were surface sterilized as described above. The surface-sterilized roots were macerated with a sterilized mortar and pestle and diluted in 0.85% (wt/vol) saline solution prior to being spread on yeast extract-mannitol (YM) plates containing streptomycin (300 μg/ml) and 5-bromo-4-chloro-3-indolyl-β-d-glucuronide (X-Gluc; 10 μg/ml). After 7 days of incubation at 28°C, the number of blue colonies was counted to determine bacterial population densities (CFU) in rice root tissues.

Plant nodulation tests.

Bradyrhizobial strains were grown for 5 days in HM broth. All plants were grown in a growth chamber under controlled environmental conditions of 28 ± 2°C on a 12-h light/12-h dark cycle at a light intensity of 180 μE/m²s and 50% humidity. Mung bean (Vigna radiata), soybean (Glycine max cv. SJ5), and siratro (M. atropurpureum) seeds were sterilized (26). The seeds of Aeschynomene americana (a local Thai variety), A. indica (African ecotype), and Aeschynomene evenia (provided by Eric Giraud) were sterilized as described in Noisangiam et al. (27). A nodulation test was performed in sterilized plastic pouches for five replicates (26). The root nodules from the tested plants were enumerated at 28 days postinoculation (p.i.).

Statistical analyses.

The experimental data were statistically analyzed according to Steel and Torrie (28), and means were compared by Duncan's multiple range test (29).

RESULTS

Bacterial isolation.

The putative rice endophytic bacteria including bradyrhizobia were isolated from two different systems of agricultural practice, rice monoculture and rice/legume crop rotational systems, along with the bacteria isolated according to the procedures depicted in Fig. 1. Sixty-two bacterial colonies were obtained from the samples from the monoculture system from the central part of Thailand, while 91 colonies were isolated from the samples from the crop rotation system from three areas of Thailand (see Table S1 in the supplemental material). Five isolates were obtained from root nodules of siratro, while 41 isolates were obtained from oligotrophic-based isolation. In addition, 107 isolates were obtained directly from BJSM culture. A total of 153 bacterial colonies were screened to remove the redundant strains using BOX-A1R PCR. The remaining 98 strains showing different BOX-A1R fingerprints (data not shown) were chosen for 16S rRNA identification. From all of the isolation strategies, a total of six different strains of putative rice endophytic bradyrhizobia were obtained (Fig. 1), while the other isolates were classified into four groups, including Alphaproteobacteria (Agrobacterium sp., Methylobacterium sp., and Rhodopseudomonas sp.), Firmicutes, Flavobacteria, and Betaproteobacteria (Ralstonia sp., Burkholderia sp., and Acidovorax sp.) (see Fig. S1 in the supplemental material). In addition, rice endophytic bradyrhizobial strains (DOA1, DOA9, and SUTN9-2) isolated from A. americana L. (weeds grown in rice fields) (27) were also used in this study, and three strains, RP5, RP7, and WD16, from O. sativa L. subsp. japonica cv. Nipponbare in Japan were also used as putative Japanese rice endophyte representatives (R. Shinoda, T. Okubo, and K. Minamisawa, unpublished data).

To enhance the opportunity to acquire rice endophytic bradyrhizobia, sterilized rice root homogenates were applied using three procedures. First, to obtain nod-dependent bradyrhizobia, a nodulation test with siratro was conducted. Only isolate SUT-R55 was obtained from root nodules of siratro. Second, in order to avoid growth of other contaminants, the oligotrophic isolation approach (2% [vol/vol] homogenate solution was incubated in sterilized water for 1 month) was applied to rice tissue prior to testing on BJSM and siratro nodulation. Most of the six bradyrhizobial strains, SUT-R3, SUT-PR9, SUT-PR48, SUT-R55, and SUT-PR64, can be isolated from this procedure. Finally, three strains, SUT-R3, SUT-R55, and SUT-R74, were also directly obtained from homogenate and grown on bradyrhizobial selective (BJSM) agar plates. The rice rhizospheric soils were also used for bradyrhizobium isolation. The strains obtained from these soil samples were SUT-PR9 and SUT-R74, which were cultivated in BJSM but not found by the siratro nodulation test (Fig. 1).

In addition, most of putative rice endophytic bradyrhizobia from Thailand were found exclusively in samples from the crop rotation system, whereas no bradyrhizobial isolates were obtained from samples from the monoculture system (Table 2; see also Table S1 in the supplemental material).

Phenotypic characteristics and phylogenetic analysis of 16S rRNA and housekeeping genes.

To determine the relationships of bradyrhizobial strains among the members of the Bradyrhizobiaceae, the nucleotide sequences from various reference Bradyrhizobium members and Rhodopseudomonas species were used to construct a phylogenetic tree. Mesorhizobium loti ATCC 33669 and Bosea thiooxidans DSM9653 were chosen as outgroup strains to root the phylogenetic tree (Fig. 2). On the basis of 16S rRNA gene sequence similarity, the phylogenetic tree could be divided into four major clusters. Cluster 1 comprises bradyrhizobial strains SUT-PR64, SUT-PR48, SUT-PR9, RP5, and RP7. Only strains SUT-PR9, RP5, and RP7 separated slightly from strains in the photosynthetic bradyrhizobia (PB) group. Bradyrhizobial strains SUT-PR48 and SUT-PR64 were closely related to Bradyrhizobium sp. strain ORS278, Bradyrhizobium sp. strain ORS285, and Bradyrhizobium sp. strain ORS375. The bradyrhizobial members in the cluster 2 belonged to various groups of Bradyrhizobium species (Bradyrhizobium sp., B. yuanmingense, B. japonicum, B. liaoningense, and B. diazoefficiens), Bradyrhizobium sp. strain S23321 (nonsymbiotic strain), and Rhodopseudomonas members. Three strains from rice tissues, including SUT-R3, SUT-R55, SUT-R74, and A. americana isolated strains (SUTN9-2, DOA1, and DOA9) belonged to this cluster, with strong bootstrap support (99%) separation from cluster 1. Cluster 3 of the phylogenetic tree contained only Bradyrhizobium elkanii species. Finally, strain WD16 was assigned to cluster 4 of the phylogenetic tree. A phylogenetic tree based on sequences of dnaK, recA, and glnB was also constructed. The taxonomic positions of the bradyrhizobial strains in the combination tree (see Fig. S2 in the supplemental material) were almost concordant with their taxonomic positions in the 16S rRNA gene tree (Fig. 2). In the combination tree, the strains were clearly separated into two clusters. In order to determine the correlation between phenotypic characteristics and phylogenetic positions, photosynthetic pigment production in bradyrhizobial strains was considered.

FIG 2.

Neighbor-joining tree based on sequences of 16S rRNA genes showing classification of the photosynthetic bradyrhizobia (PB) and nonphotosynthetic bradyrhizobia (non-PB) strains isolated from rice tissues. Bootstrap values are expressed as percentages of 1,000 replications. The bar represents 1 estimated substitution per 100 nucleotide positions. Evolutionary distances were computed using the Kimura two-parameter method and are shown in units representing the number of base substitutions per site.

All of the strains formed typical slow-growing Bradyrhizobium colonies on HM agar plates (4 to 7 days). From a total of nine strains, strains SUT-PR48 and SUT-PR64 could synthesize the pink/orange pigments when cultured on yeast-mannitol agar and broth, whereas pigment formation could not be detected from strains DOA1, DOA9, SUTN9-2, SUT-R3, SUT-PR9, SUT-R55, SUT-R74, and Japanese bradyrhizobial strains (see Fig. S3 in the supplemental material). In order to confirm photosynthetic characteristics on the basis of photosynthetic pigment production, extraction of pigment using the acetone-methanol method was carried out. The results also showed the same trend as those obtained from colony color formation analysis (see Fig. S4). In addition, the light-harvesting complex I (puf) and bacteriochlorophyll biosynthesis (bch) genes were used to determine the photosynthetically related genes from every isolate. Degenerate primers were designed from both of the photosynthetic bacteria Bradyrhizobium sp. and Rhodopseudomonas palustris. The pufM gene was perfectly conserved in Bradyrhizobium strains ORS285, BTAi1, SUT-PR48, and SUT-PR64. However, the pufM bands were still detected from nonphotosynthetic Bradyrhizobium strain S23321. In contrast, the specific band of the pufM gene could not be detected in strains USDA110, SUT-R3, SUT-PR9, SUT-R55, SUT-R74, WD16, RP5, and RP7. The bchL gene was amplified from Bradyrhizobium strains BTAi1, ORS285, SUT-PR48, and SUT-PR64, whereas the specific band of the bchL gene could not be detected in Bradyrhizobium strains USDA110, S23321, SUT-R3, SUT-PR9, SUT-R55, SUT-R74, WD16, RP5, and RP7 (see Fig. S5 in the supplemental material).

The results from phylogenetic tree analyses were mainly congruent with the pigment production characteristics, as described before. On the basis of the phylogenetic tree and phenotypic properties, endophytic bradyrhizobial strains might be classified into two main groups: PB and nonphotosynthetic bradyrhizobia (non-PB) groups. Nevertheless, the strains SUT-PR9, RP5, RP7, and WD16 separated slightly from the PB cluster. To confirm again the photosynthetic property of this group, strain SUT-PR9 was chosen as the representative of strains RP5 and RP7 to determine pigment production by another method, that described by Chaintreuil et al. (2). The photosynthetic pigment at 800 and 870 nm was not detected in strain SUT-PR9 (see Fig. S6 in the supplemental material). Therefore, strains SUT-PR9, RP5, RP7, and WD16 were classified as non-PB strains (see Fig. S6).

Promotion of growth of rice.

To examine whether the isolated strains are rice endophytes, they were tagged with a GUS reporter gene and observed under light microscopy (see Fig. S7 in the supplemental material). The effects of bradyrhizobial strains on rice biomass in Leonard's jar assemblies were compared among three rice cultivars. In this experiment, NH₄NO₃ as a nitrogen source for plants at 0, 0.1, and 1 mM was added to N-free medium. Among three concentrations of nitrogen, only 0.1 mM NH₄NO₃ showed significant promotion of rice growth with the bradyrhizobial inoculation treatments compared to growth of the uninoculated control (data not shown). Therefore, 0.1 mM NH₄NO₃ was further used to determine rice growth promotion (Fig. 3).

FIG 3.

The effects of putative rice endophytic bradyrhizobia on rice biomass in Leonard's jar experiments were compared among three rice cultivars (O. sativa L. subsp. indica cv. Pathum Thani 1, O. sativa L. subsp. indica cv. Kasalath, and O. sativa L. subsp. japonica cv. Nipponbare). Within each rice cultivar, means labeled with different letters are statistically different at a P value of <0.05.

Promotion of rice growth by Thai bradyrhizobial strains was significantly higher than growth of the uninoculated control, except for strains SUT-R55 and SUT-R74, which were not able to promote the growth of rice compared with growth of the uninoculated control. There was a tendency in the promotion of rice growth that strains SUT-PR9 and SUT-PR64 could promote higher rice biomasses than other strains. In addition, strains SUT-PR9 and SUT-PR64 could enhance rice biomass (cultivar Pathum Thani 1) to levels 40% and 50% higher, respectively, than the level of the uninoculated control. Among the Japanese bradyrhizobial strains, only strain RP7 was able to promote a higher rice biomass (cultivar Pathum Thani 1) than the uninoculated control but without a difference that was statistically significant (32% higher than that of the uninoculated control). Generally, Thai bradyrhizobial strains could promote rice growth better than Japanese bradyrhizobial strains, especially when strains were inoculated with Thai rice cultivars, whereas the promotion of rice growth of cultivar Kasalath could not clearly be observed with any bradyrhizobial strain. Nevertheless, almost all of the bradyrhizobial strains (except WD16) could promote the growth of the rice cultivar O. sativa Nipponbare to a level that was significantly higher than that with the uninoculated control.

Because bradyrhizobial isolates from A. americana nodules (DOA1, DOA9, and SUTN9-2) were phylogenetically intermixed in rice bradyrhizobia (Fig. 2), the inoculation of rice cultivar O. sativa Pathum Thani 1 with strains DOA1 and SUTN9-2 was also tested; results showed that these strains could also significantly promote a rice biomass higher than that of the uninoculated control. In contrast, strain DOA9 did not effectively promote rice growth.

The bacterial strains capable of promoting the highest rice growth in each rice cultivar were selected. Strains SUT-R3 and SUT-PR64 represented non-PB and PB strains, respectively. Strain RP7 was selected as a Japanese rice endophytic bradyrhizobia, and strain SUTN9-2 was chosen as a representative isolate derived from A. americana. In order to confirm the compatibility of rice cultivars and endophytic bradyrhizobia, these selected strains were tested with four more Thai rice cultivars (see Fig. S8 in the supplemental material). For rice cultivar O. sativa Khao Dowk Mali 105, strains SUTN9-2, SUT-R3, and SUT-PR64 could promote significantly better rice growth than that of the uninoculated control and RP7. Likewise, for rice cultivar O. sativa Leuang Yai 148, both strains SUTN9-2 and SUT-PR64 had a positive effect on promoting the growth of rice. In the case of rice cultivar O. sativa Lu Ni, strains SUT-R3 and SUT-PR64 could promote rice growth significantly higher than the growth of the uninoculated control. The same trend was seen with rice cultivar O. sativa Sang Yod Phatthalung as strain SUT-R3 could promote a rice biomass significantly higher than that of uninoculated control. In contrast, the promotion of rice growth derived from RP7 inoculation did not affect any of the Thai rice cultivars (Fig. 3; see also Fig. S8 in the supplemental material).

Endophytic bradyrhizobium population from different rice cultivars.

The density of the bacterial population in the tissue of rice cultivar Pathum Thani 1 was significantly higher than densities in tissues of cultivars O. sativa Kasalath and Nipponbare (Fig. 4A). The putative Thai rice endophytic bradyrhizobial populations in rice root were around 2.5 log₁₀ CFU/g root (fresh weight) in rice cultivar O. sativa Pathum Thani 1 while the populations in O. sativa Kasalath and Nipponbare were around 1.0 log₁₀ CFU/g root (fresh weight) at 7 days p.i.

FIG 4.

Enumeration of bradyrhizobia in rice root tissues. (A) Putative Thai rice endophytic bradyrhizobial populations in rice roots of three rice cultivars (O. sativa L. subsp. indica cv. Pathum Thani 1, O. sativa L. subsp. indica cv. Kasalath, and O. sativa L. subsp. japonica cv. Nipponbare) were determined at 7 days p.i. (dai). (B) The putative Thai rice endophytic bradyrhizobial population in rice root of O. sativa L. subsp. indica cv. Pathum Thani 1 at each day p.i. Means labeled with different letters are statistically different at a P value of <0.05.

Since most of the putative Thai rice endophytic bradyrhizobial strains had a better response to O. sativa Pathum Thani 1 than other cultivars, the bacterial colonization of this cultivar was further determined (Fig. 4B). The bradyrhizobial populations among Thai strains were not significantly different at 7 days p.i. In addition, bradyrhizobial densities were not statistically different at 30 days p.i., with populations in rice root around 3.5 to 5 log₁₀ CFU/g root (fresh weight), except for the densities of SUT-R55. The proliferation of all strains showed the highest numbers at 30 days p.i., and then the numbers declined slightly when plant age increased, with densities around 2 to 4 log₁₀ CFU/g root (fresh weight) at 60 days p.i. (Fig. 4B). Moreover, all strains still persisted in rice root tissues (approximately 1.5 log₁₀ CFU/g root, fresh weight) at 90 days p.i., and the bacterial cell numbers of strain SUT-PR9 were still significantly higher than those of all tested strains.

In addition, six more rice cultivars were also tested in terms of bacterial colonization in rice root tissues (as shown in Fig. S8 in the supplemental material and including two Japanese cultivars). Four bacterial strains with the highest level of rice growth promotion, SUTN9-2, SUT-R3, SUT-PR64, and RP7, were also used. For rice cultivar O. sativa Khao Dowk Mali 105, the population densities of SUTN9-2, SUT-R3, and SUT-PR64 were significantly higher than the population density of RP7 at 7 days p.i. The cell numbers of both SUTN9-2 and SUT-PR64 strains were higher in rice cultivars O. sativa Leuang Yai 148 and Lu Ni than the number of RP7 cells. In the case of O. sativa Sang Yod Phatthalung, all tested strains could increase their cell densities by around 2.5 log₁₀ CFU/g root (fresh weight), except for Japanese strain RP7, which clearly showed a small bacterial population (see Fig. S9 in the supplemental material).

Since strains SUT-PR9 and SUT-PR64 promoted the highest rice biomass in the Leonard's jar experiment, both strains were tagged with GUS/DsRed reporter genes to determine their localization in planta. Both strains were capable of the rapid colonization of rice root in cortical cell layers and of xylem occupation of rice roots at 3 days p.i. (Fig. 5A and B). However, the colonization of strain SUT-PR9 was especially dense at the surface and root hair zones of rice root at 7 days p.i. (Fig. 5C and D), while strain SUT-PR64 was able to heavily colonize and invade the rice tissues from epidermal cells to endodermal cells at 7 days p.i. (Fig. 5E and F).

FIG 5.

The localization of bradyrhizobial strains SUT-PR9 and SUT-PR64 at 3 days p.i. (A and B) The localization in rice roots of DsRed-tagged SUT-PR9 and SUT-PR64, respectively. (C and D) The localization in rice roots of GUS-tagged SUT-PR9 at 7 days p.i. (E and F) The localization in rice roots of GUS-tagged SUT-PR64 at 7 days p.i. Bacterial localization inside rice root tissues is indicated as follows: ep, epidermis; en, endodermis; ex, exodermis; co, cortex; scl, sclerenchyma.

Nodulation test and A. indica growth promotion.

The types of leguminous plants selected on the basis of their frequent presence in rice field environments belonged to the genus Aeschynomene, including A. americana and A. indica. In addition, the commercial leguminous plants V. radiata and G. max, normally used in rice-legume crop rotation systems in Thailand, were also included. The results of nodulation tests with the various leguminous plants are summarized in Table 4. The PB strains, including SUT-PR48 and SUT-PR64 (Fig. 2), could produce up to 100 nodules per plant of A. indica and A. evenia, which belong to cross-inoculation (CI) group 3 (nodulated only by PB), at 28 days p.i. In contrast, nodulation of A. indica and A. evenia could not be detected when they were inoculated with non-PB group strains (27), with the exception of inoculation with strain SUT-PR9, which formed small amounts of nodules (approximately 5 nodules per plant) in A. indica and A. evenia. Only strain SUT-R3 formed nodules in V. radiata. Strain SUT-R55 could effectively nodulate A. americana, M. atropurpureum, and V. radiata, while SUT-R74 could not nodulate any leguminous plant tested. B. diazoefficiens USDA110 had the ability to form root nodules with M. atropurpureum, G. max, and V. radiata, while Bradyrhizobium sp. BTAi1 could form root/stem nodules only with A. indica and A. evenia.

TABLE 4.

Plant nodulation test in this study

Bradyrhizobium group and strain	Nodulation on:a
	A. americana	A. indicab	A. evenia	M. atropurpureum	G. max	V. radiata
PB strains
SUT-PR48	−	+	+	−	−	−
SUT-PR64	−	+	+	−	−	−
BTAi1	−	+	+	−	−	−
Non-PB strains
SUT-PR9	−	+/−	+/−	−	−	−
WD16	−	+/−	+/−	−	−	−
RP5	−	+/−	+/−	−	−	−
RP7	−	+/−	+/−	−	−	−
SUT-R3	−	−	−	−	−	+
SUT-R55	+	−	−	+	−	+
SUT-R74	−	−	−	−	−	−
USDA110	−	−	−	+	+	+

^aNodulation is indicated as follows: +, more than 50 nodules/plant, reddish nodules; +/−, fewer than 10 nodules/plant, white nodules; −, no nodules detected.

^bAfrican ecotype.

Among the Thai isolated strains, SUT-PR9 was used as a non-PB representative although it was closely related to PB strains. Therefore, A. indica, as an appropriate plant for PB groups, was used to evaluate symbiotic efficiency. Total plant dry weights derived from PB strains SUT-PR48 and SUT-PR64 were significantly higher than those from inoculations with non-PB strains (Fig. 6), while SUT-PR9 showed low efficiency for A. indica growth promotion and nitrogen fixation compared with PB strains SUT-PR48 and SUT-PR64.

FIG 6.

A. indica growth promotion and nitrogen fixation. Means labeled with different letters are statistically different at a P value of <0.05.

To detect the nodulation genes among bradyrhizobial isolated strains, PCR amplification of nodA, nodB, and nodC was carried out. The primer sets were designed from B. diazoefficiens USDA110 and Bradyrhizobium sp. ORS285. No specific bands representing nodA, nodB, and nodC expression were obtained from putative endophytic bradyrhizobial strains (see Fig. S10 in the supplemental material).

DISCUSSION

Prior to our having an understanding of the behavior of bradyrhizobia as an endophyte in rice, the advantages of oligotrophically based isolation were used to reduce the growth of other contaminants while the bradyrhizobia still survived under nutrient-deprived (oligotrophic) conditions. The incubation of as little as 1 ml of root homogenate in a large amount of water (50 ml) for 1 month mimicked the oligotrophic condition. Thereby, almost all of rice oligotrophic endophytic bradyrhizobia in this study were obtained except SUT-R74. The proportion of bradyrhizobial strains and other bacterial endophytes detected in rice tissues using the culture-dependent approach was smaller (6 bradyrhizobial strains from 98 bacterial strains). It seemed that the putative endophytic bradyrhizobia may not be the dominant species in cultivated rice. This observation may support a previous report which revealed that Proteobacteria are the major endophytic bacteria in rice tissues. Among Alphaproteobacteria (approximately 30% relative abundance of all subdivisions), Bradyrhizobium is found only in rice root, with 10% relative abundance of total Alphaproteobacteria (30).

Bradyrhizobial isolation was successfully obtained from the rice-legume crop rotation system. In contrast, endophytic bradyrhizobia could not be obtained from the monoculture system (2 to 3 crops/year). The data are congruent with those of Guong et al. (31), which revealed that Bradyrhizobium sp. and Herbaspirillum sp. instantly colonize the interior of rice roots when rice is grown in rotation with a legume crop. Moreover, the bacterial community detected in the rice-legume crop rotation system was significantly different from and present in a greater concentration than that in rice from monoculture. Therefore, appropriate crop rotation provides a feasible practice to maintain the equilibrium of the soil microbial environment. These observations suggest the hypothesis that agricultural practice conditions may drive the evolution of the bradyrhizobium-host association. This hypothesis may also gain more supportive evidence since A. americana-nodulating bradyrhizobial strains (DOA1, DOA9, and SUTN9-2) have rice endophytic properties. Furthermore, the bradyrhizobial strains nodulating A. americana in this study have never been detected in paddy soil (Table 2), but they can survive during a period without the genuine host and reform nodules the following year. Their cells may proliferate in rice tissues even though A. americana cannot grow well each year from December to July of the following year (tentatively 7 months). Actually, A. americana is a leguminous weed grown in rice fields for only 5 months a year (from July to December). Therefore, these results may be part of the evidence that confirms the bradyrhizobium-host evolution.

Plant growth promotion properties of rhizobia have been previously reported for various cereals such as wheat, maize, and rice (32–35). Most of the Thai isolates and especially strain SUTN9-2 significantly increased rice biomass. On the other hand, the growth of Thai rice was not affected by Japanese strains (strains RP5, RP7, and WD16) (Fig. 3). Moreover, among the three rice cultivars, O. sativa L. subsp. indica cv. Pathum Thani 1 responded positively only to putative Thai rice endophytic bradyrhizobia, while a similar phenomenon was not found in Japanese rice cultivars. These results implied that the rice cultivar is one of the factors that control the compatibility of the rice-bacterium association. In addition, this phenomenon reflects the bradyrhizobial host preference of Thai endophytic bradyrhizobial strains for Thai rice cultivars (see Fig. S9 in the supplemental material). This scenario is confirmed by a previous report that rice genotypes significantly affected the bacterial association in rice tissues (36).

Cultivated rice production is frequently found in combination with several aquatic legumes belonging to the genus Aeschynomene (A. americana and A. indica) during the rainy season (from August to October). Among these aquatic species, A. indica (CI group 3) forms stem and/or root nodules with PB strains (2, 37). These data were congruent with our results; the PB strains SUT-PR48 and SUT-PR64 could form nodules with A. indica (African ecotype) (Fig. 6). A previous report showed that photosynthesis plays an important role in the capacity to establish stem nodules within the dalbergioid group (38). Moreover, the dalbergioid group occurred earlier on earth than phaseolid species (tentatively, 1.5 million years) (39). Perhaps, the symbiotic evolution of the PB group was established on dalbergioid species, and the PB still preserve host specificity. Our results revealed that PB strains formed symbioses with both aquatic leguminous plants and nonleguminous species such as rice (2, 40). Interestingly, strain SUT-PR9 showed low efficiency in nodulation and nitrogen fixation with A. indica (Fig. 6). Moreover, the photosynthetic pigments of SUT-PR9 were not detected at levels comparable to those of a PB group strain (see Fig. S6 in the supplemental material). It seems that the phenotypic properties of SUT-PR9 are still ambiguous. Perhaps the ability to form nodules without Nod factor, rather than being driven solely by the plants, has also been dependent on a specific single bacterial evolution/mutation (41). The Calvin-Benson-Bassham (CBB) cycle from Bradyrhizobium species plays an important role in chemoautotrophic growth (42), and it is important for efficient symbiosis with A. indica (43) by controlling the oxygen tension of the early stage of symbiosis (44). In addition, the ancestors of B. japonicum and A. indica symbionts were photosynthetic free-living bacteria (37). If this scenario is true, it should be hypothesized that strains SUT-PR48 and SUT-PR64 still preserve some ancestral properties of the PB group. In contrast, SUT-PR9 may partially lose its capacity for photosynthetic pigment production and establish symbiosis with A. indica. Perhaps strain SUT-PR9 is an intermediate between PB and non-PB strains, as suggested by the evolutionary branches among PB strains (Fig. 2). Furthermore, the photosynthesis-related genes were perfectly conserved among PB strains (see Fig. S5 in the supplemental material). Even though the light-harvesting complex (pufM) gene can still be detected in the nonphotosynthetic strain S2332, the PCR product of the bacteriochlorophyll biosynthesis (bchL) gene was not detected. However, the bchL gene can be found in the strain S23321 genome. This might be because the bchL primer in this experiment was designed based on the sequences of both the photosynthetic Bradyrhizobium sp. and its closest photosynthetic relative, Rhodopseudomonas palustris. Moreover, the bchL gene of strain S23321 (GenBank accession number BAL75154.1) is not absolutely conserved in photosynthetic strains. Thus, bchL amplification in strain S23321 could not be seen because of the incompatibility of the designed primer.

A preliminary experiment by Okubo et al. (45) showed that a photosynthetic gene cluster was detected in the genome of nonsymbiotic Bradyrhizobium strain S23321 that displayed high similarity to the A. indica symbionts, whereas the phylogenetic tree of the 16S rRNA gene was closely related to that of B. diazoefficiens USDA110. These observations may also support our hypothesis that environmental conditions and/or habitat have been continuously driving the evolution of the plant-bradyrhizobium interaction.

In contrast, root nodules were found from A. americana only with the non-PB group (27). Similar to our findings, non-PB strain SUT-R55 can form effective nodules on A. americana, M. atropurpureum, and V. radiata. Therefore, this is the first report to demonstrate that these non-PB strains are also capable of forming a natural endophytic association with rice. Interestingly, strain SUT-R74 displayed endophytic properties only with rice and was not able to form a symbiotic relationship with any tested leguminous plants. We predict that SUT-R74 may be one of the nonnodulating bradyrhizobial strains. This result was similar to a previous report by Okubo et al. (45) that Bradyrhizobium sp. S23321 was not able to form nodules with M. atropurpureum. When the PCR amplifications of nodA, nodB, and nodC were carried out, none of specific bands of nodA, nodB, and nodC were obtained from any rice endophytic bradyrhizobial strain (see Fig. S10 in the supplemental material). These data were congruent with a previous report which found that divergent nod-containing Bradyrhizobium strains could induce root nodules on A. americana and M. atropurpureum (27). Thus, the rearrangement of nodulation gene sequences may have evolved after the arrival of legumes on the planet.

In conclusion, our findings have provided evidence that the rice-legume crop rotational system (commercial legumes and/or weed legumes) is one way to maintain putative endophytic bradyrhizobial strains in ecosystems. Among Thai rice cultivars, the Thai bradyrhizobial strains could promote rice growth better than Japanese strains. In addition, among the three rice cultivars (Pathum Thani 1, Kasalath, and Nipponbare), cultivar Pathum Thani 1 responded positively only to putative Thai rice endophytic bradyrhizobia. In contrast, this phenomenon was not found in Japanese rice cultivars. It seems that rice cultivars are one of the factors that control the compatibility of the rice-bradyrhizobial association. In addition, our work provides the evidence that non-PB strains are also capable of forming a natural endophytic association with rice. Interestingly, strains SUT-PR9, WD16, RP5, and RP7 displayed non-PB phenotypes but were genotypically close to PB strains. In addition, none of these endophytic bradyrhizobial strains seems to contain typical nodulation genes. The evolution of the plant-bradyrhizobium association may have started from a photosynthetic lifestyle, and then the strains lost their photosynthetic apparatus during evolution in a nod-independent manner. Genome comparisons of non-PB strains and such nonendophytic Bradyrhizobium species as B. diazoefficiens USDA110 should be performed in order to gain a deeper understanding of the mechanism of the Bradyrhizobium-rice association. This would lead to the use of oligotrophic rice endophytes for rice growth promotion in the future.