The bacterial strains used in this study are listed in Table 1. Bradyrhizobium sp. strain DOA9 and its ΩrhcN derivative were grown in YM medium (Vincent, 1970) at 28°C. Escherichia coli strains were cultured at 37°C in LB medium (Sambrook et al., 2001). The media were supplemented with antibiotics at the following concentrations when appropriate: 200 μg/mL streptomycin (Sm), 20 μg/mL nalidixic acid (Nal), and 50 μg/mL kanamycin (Km).

The single cross-homologous recombination technique was used to construct the insertion mutant in the structural type III secretion system gene (rhcN) of strain DOA9. For this purpose, an internal fragment (263-bp) of rhcN was amplified by polymerase chain reaction (PCR) using the following primers: RhcN.D9p.int.f (5′-CATCTCGTCGACTCGCAGCAAAGGATGTCGATAC-3′) and RhcN.D9p.int.r (5′-GAGCAGTCTAGACCCGACTGACACTTCCTGCATG-3′). The PCR product was then digested by Xbal and SalI and cloned onto the corresponding sites of the plasmid pVO155-Sm-npt2-gfp (Wongdee et al., 2016). This plasmid, which is non-replicative in Bradyrhizobium strains, is a derivative of the plasmid pVO155 (Oke and Long, 1999) and carries the promoterless gusA gene, a constitutively expressed gfp, and a Km- and a Sm/spectinomycin-resistance genes (Wongdee et al., 2016). The resulting plasmid was introduced into E. coli S17-1 by electroporation (15 kv/cm, 100 Ω, and 25 μF) and was transferred into the Bradyrhizobium sp. strain DOA9 by biparental mating, using the protocol described by Giraud et al. (2010). The transconjugants were spread on YM medium supplemented with an antibiotic mixture of Sm, Km, and Nal, and the successful insertion of the plasmid in rhcN was checked by PCR.

The GFP-labeled DOA9 strain was constructed by introducing the replicative plasmid pMG103-npt2-Sm-npt2-gfp harboring a constitutive gfp gene by electroporation (15 kv/cm, 100 Ω, and 25 μF). Plasmid-containing strains were selected on YM plates supplemented with Sm.

DOA9 and its derivatives were grown for 5 days as previously described and used as inoculum. Sterilization and germination of the seeds of all plants tested (Supplementary Table S1) were performed as described by Teamtisong et al. (2014). Plants were sown in Leonard’s jar filled with sterilized vermiculite (Somasegaran and Hoben, 1994). All the plants were watered with BNM medium (Ehrhardt et al., 1992) and grown under the following controlled environmental conditions: 28 ± 2°C with a 16 h light/8 h dark cycle at light intensities of 300 μE/m²S and with 50% humidity. Five days after planting, each seedling was inoculated with 1 mL of a 5 day old inoculum (log-phase) after washing and adjusting the optical density to 600 nm to 1 (approximately 10⁹ cells). The experiment was conducted with five replicates with the following treatments: control (no inoculation), GPF-labeled DOA9, and ΩrhcN. To check nodulation and nitrogen fixation abilities, five plants were taken at 21 day post-inoculation (dpi) and the number of nodules and nitrogenase activity were assessed using the acetylene reduction assay (ARA). Briefly, all the nodules on each plant were collected and placed in headspace bottles with 10% acetylene, and incubated at 28°C for 1 h. Gas chromatography was used to measure the peak height of ethylene and acetylene with 1 mL gas samples from the bottles by using a PE-alumina packed column with injector at 150°C, an oven temperature of 200°C, and ionization detector (FID) at a temperature of 50°C (Somasegaran and Hoben, 1994).

Rice (Oryza sativa L. ssp. indica cv. Pathum Thani) seeds were surface sterilized by soaking in 70% (v/v) ethanol for 30 s, then washed twice with 10% (v/v) hydrogen peroxide for 10 min, washed again in 3% (v/v) sodium hypochlorite for 1 h. Finally, the seeds were washed in sterilized water five times and germinated overnight on YM medium containing 0.8% (v/v) agar. The germinated seeds with 1 cm of root were soaked in DOA9 cell suspension (OD₆₀₀ = 1) overnight. The rice seedlings were transferred to metal net in 80 mL tubes containing 15 mL Hoagland’s plant growth medium (Hoagland and Arnon, 1950) in presence of 0.1 mM NH₄NO₃. The experiment was conducted with three technical replicates with the following treatments: control (no inoculation), GFP-labeled DOA9, and ΩrhcN. Plants were grown under the following controlled environmental conditions: 28 ± 2°C with a 16 h light/8 h dark cycle at light intensities of 300 μE/m²S and with 50% humidity.

Two and 3 weeks after inoculation with the DOA9 tagged strains and ΩrhcN containing gfp and Sm^r, the rice roots were washed in normal saline solution (0.85% of NaCl) to remove debris and any bacteria not firmly attached to the root surface. Next, the bacterial cells remaining on the surface of the roots were harvested by vortex with normal saline containing 0.3% (v/v) tween-80 for 30 s. The washing solution containing the epiphytic bacteria was then spread on YM plates containing Sm. For the enumeration of endophytic bacterial, root tissues were surface sterilized with 70% (v/v) ethanol for 5 min followed by 3% (v/v) sodium hypochlorite for 5 min and rinsed with sterilized water five times. The surface sterilized roots were macerated with a sterilized mortar and pestle and diluted in saline solution prior to being spread on medium as described above (Piromyou et al., 2015a). The epiphyte and endophyte population counted on agar plates are expressed as colony-forming unit per gram of root fresh weight.

The sterilized legume seeds were germinated and transferred into tubes containing BNM medium (50 mg seeds/mL). Plants were maintained in controlled environmental conditions as described above for 5 days. The root exudates were obtained from plant medium after filtration using a 0.2 μm filter syringe. Root exudates were stored at -20°C until use.

The mid-log phase culture of DOA9 was washed and the OD₆₀₀ was adjusted to approximately 0.4 with YM supplemented with 1/3 (v/v) of the root exudates or purified flavonoids (20 μM of naringenin or genistein dissolved in DMSO). The sterilized BNM medium and DMSO were used as negative controls. The bacterial cells were cultured at 28°C for 24 h and then collected by centrifugation (4,000 × g for 10 min, 4°C) and immediately frozen in liquid nitrogen and stored at -80°C for further total RNA isolation.

Total bacterial RNA was extracted from induced cells using the RNeasy^® Mini Kit (QIAGEN, United States) according to the manufacturer’s protocol. Total RNA was treated with RNase-free DNase I (NEB) for 30 min at 37°C. cDNA was synthesized from 500 ng total RNA using High Capacity cDNA Reverse Transcription Kits (Applied Biosystems^TM) according to the manufacturers’ protocols. Thirty nanograms of cDNA was subjected to PCR amplification using specific primers of rhcN (rhcN.RT.D9.f: 5′-CATTGGCGATATGGTAGGCT-3′; rhcN.RT.D9.r: 5′-GGACAAGTGTGAACCGTCCT-3′) and nopX (nopX.RT.D9.f: 5′-CATCAACCCGAACAACACAG-3′; nopX.RT.D9.r: 5′-GGCTCGATAGACAAGGTCCACAA-3′) genes. PCR amplification was performed using QuantStudio 3 Real-Time PCR System Mix (Applied Biosystems^TM) and the following program: an initial denaturation step at 95°C for 2 min; 35 cycles at 95°C for 30 s, 55°C for 30 s, and at 72°C for 1.5 min; and a final extension step at 72°C for 10 min. The relative gene expression was analyzed using the comparative Ct method (-ΔΔCT) normalized to the endogenous housekeeping gene (16S rRNA) using PBA338F (5′-ACTCCTACGGGAGGCAGCAG-3′) and PRUN518R (5′-ATTACCGCGGCTGCTGG-3′). Three biological replicates were pooled and analyzed. At least three replicate PCR amplifications were performed for each sample.

An Olympus Fluoview FV1000 confocal laser scanning microscope was used to investigate nodule development, bacteroid differentiation, and rice colonization. For this purpose, 40–50 μm thick sections of fresh nodules and rice roots were prepared using a VT1000S vibratome (Leica Nanterre, France). Nodule sections were stained for 20 min with 0.01% calcofluor (to stain the plant cell wall) and with 30 μM propidium iodide (PI) in 50 mM Tris–HCl pH 7.0 buffer to identify dead cells (Haag et al., 2011). Calcofluor was excited at 405 nm and detected with a 460–500 nm emission filter. The GFP-labeled DOA9 and ΩrhcN strains were detected after excitation with the 488 nm laser line and emission signal collection at 490–522 nm, while the PI used to identify dead cells was excited with the 535 nm laser line and emission signals were collected at 617–636 nm. Confocal images were reconstructed with NIS elements software (Nikon), and images were colored and prepared for publication with Adobe Photoshop software. To detect β-glucuronidase (GUS) activity in the nodules elicited by the rhcN mutant the same protocol than Bonaldi et al. (2010) was used.

Two complementary approaches were used to identify putative effector genes: (i) a TBlastN search of the genome for known Nops identified in rhizobia (Kimbrel et al., 2013; Staehelin and Krishnan, 2015), the identified homologs were considered as possible candidates when they matched the following parameters (% of identity ≥40% over 80% of the length of the sequence) and (ii) a search for tts-boxes on the chromosome and plasmid of DOA9 genome (GenBank accession numbers DF820425 and DF820426, respectively) using a hidden Markov model initially trained with the sequences of 26 confirmed tts-boxes from Sinorhizobium (Ensifer) sp. strain NGR234, B. diazoefficiens USDA110, and B. elkanii USDA61 (Eddy, 1998; Marie et al., 2004; Zehner et al., 2008; Okazaki et al., 2009). To be considered as a candidate, the gene had to contain a tts-box with a score ≥9 located on the same strand in its upstream region (up to 1 kb).

The experiment data were subjected to analysis of variance (ANOVA). When confirming a statistically significant value in the F-test (p ≤ 0.05), a post hoc test (Duncan’s multiple-range test at p ≤ 0.05) was used as a multiple comparison procedure (Duncan, 1955) using SPSS1 software for WINDOWS^TM, Version 14.0.

Sequence genome analysis revealed the presence of a T3SS gene cluster organized in several operons on the pDOA9 plasmid (Figures 1A,B). The main putative operon preceded by a tts-box contained most of the genes coding for components of the secretion machinery in the order nopB, rhcJ, nolU, rhcL, rhcN, rhcU, rhcQ, rhcR, rhcS, rhcT, and rhcU. This genetic organization is similar to that found in other Bradyrhizobium or Sinorhizobium strains and a very high level of AA identity was observed with the corresponding homologous proteins. However, unlike these strains, a large intergenic region of 650 bp was observed between rhcS and rhcT also containing a tts-box (Figure 1A), raising the possibility that the rhcT and rhcU genes are transcribed independently.

As described in other rhizobia, we could assume that this main cluster is under the dual control of TtsI and NodD since a ttsI homolog was found at the vicinity of the main operon preceded by a nod-box (Figure 1A). However, it is to highlight the presence of a gene encoding a putative transposase between this nod-box and ttsI. To confirm that the presence of this putative transposase does not interfere with the formation of the T3SS secretion apparatus, we analyzed the expression of two structural T3SS genes (rhcN and nopX) in the presence of two flavonoids (genistein or naringenin) previously reported to induce T3SS genes in rhizobial strains (Krause et al., 2002; Okazaki et al., 2010). As shown in Supplementary Figure S1, both flavonoids strongly induced nopX and rhcN expression, showing that despite this transposase, the T3SS genes in DOA9 strain are classically regulated by flavonoids, most probably via the control of NodD and TtsI.

To identify putative effectors that are possibly translocated by this T3SS machinery, we combined two in silico searches: (i) a TBlastN search of the genome for Nops previously identified in other rhizobia and (ii) a search of the tts-box motif using a hidden Markov model. After eliminating the Nops proteins predicted to correspond to components of the secretory apparatus (NopA, NopX, and NopB) (Staehelin and Krishnan, 2015), this analysis retrieved 20 putative translocated effectors. Of these, six were located on the chromosome (Figure 1C). However, all six candidates remain doubtful because homologous proteins for all of them can be identified in photosynthetic Bradyrhizobium strains that lack a T3SS gene cluster (Supplementary Table S2). Interestingly, among the other putative effectors found on the plasmid, three correspond to new effectors not previously identified in other rhizobia. They were preceded by a tts-box and all contained the small ubiquitin-like modifier (SUMO) protease domain of the C48 peptidase [ubiquitin-like protease 1 (Ulp1)] family. Several effectors containing this functional domain have already been characterized in pathogenic and symbiotic bacteria (Hotson et al., 2003; Hubber et al., 2004; Rodrigues et al., 2007; Tsurumaru et al., 2015), reinforcing the hypothesis that these three effector candidates are bona fide new effectors.

To analyze the role of the T3SS in DOA9 strain during symbiosis, we constructed a T3SS mutant by inserting the non-replicative plasmid pVO155-Sm-npt2-gfp into the 5′-region of rhcN. The rhcN gene was selected as target because it encodes an ATPase that is indispensable for the functioning of the T3SS injectisome. Furthermore, the pVO155 plasmid used for inactivation of the gene contains a constitutive expressed GFP that enables monitoring of the bacteria inside the nodules. The symbiotic performance of the WT and ΩrhcN mutant was compared on several legume species belonging to Genistoid (Crotalaria juncea), Dalbergioid (Aeschynomene americana, A. afraspera, Arachis hypogaea, and Stylosanthes hamata), and Millettioid tribes (Macroptillium atropurpureum, Vigna radiata, Indigofera tinctoria, and Desmodium tortuosum) by analyzing several symbiotic parameters (number of nodules per plant, nitrogen fixation estimated using the ARA, plant dry weight, and cytological aspect of the nodules).

Depending on the plant species, the ΩrhcN mutant displayed contrasted responses compared to the WT strain. The responses were classified in three categories. In category 1, non-responsive phenotype (called T3SS-no effect group), the plants inoculated with the mutant displayed the same number of nodules and the same nitrogen fixing capacity as the plants inoculated with the WT strain. Furthermore, no cytological differences were observed between the mutant and WT nodules. These plants were A. americana, A. afraspera, I. tinctoria, and D. tortuosum (Figure 2).

In category 2, the rhcN mutation had a positive effect on one of the symbiotic parameters analyzed (Figure 3). This group comprised A. hypogaea, V. radiata, C. juncea, and M. atropurpureum. Inoculation with the ΩrhcN mutant led to an almost 65% increase in the number of nodules on A. hypogea compared with WT (Figure 3F). This result suggests that in DOA9, T3SS compromises either the infection or nodule organogenesis process in this species. Figure 3 shows that the higher number of nodules is not correlated with an increase in N₂ fixation. We assume that the plant compensates for the small number of nodules elicited by the WT strain by stimulating their expansion, as can be seen in Figures 3B,C. The same mechanism has been reported in Medicago and is related to plant N demand (Laguerre et al., 2012). The deleterious effect of the T3SS on the establishment of the symbiosis was more marked in the cases of C. juncea and V. radiata. Indeed, whereas the WT strain only elicited bumps or small completely necrotic nodules in C. juncea and V. radiata, respectively, the ΩrhcN mutant formed perfectly developed nodules on these two plants (Figures 3H–U). This suggests that some effectors translocated by the T3SS could activate plant defense reactions in these two species, thereby preventing infection and the development of nodules. Similar stimulation of plant immunity probably also occurred in M. atropurpureum but to a lesser extent, since the nodules induced by the WT strain were well formed, but displayed brown necrotic areas. The bacteroids in these areas were dead as revealed by the red PI staining, whereas the ΩrhcN nodules were perfectly normal and the host cells were filled with viable bacteroids, as revealed by the green of GFP tagging cells (Figures 3V–AB). Nevertheless, while the T3SS mutation made it possible to restore nodule formation capacity in V. radiata and C. juncea, only very weak nitrogenase activity was detected, and no real benefit for plant growth was observed (Figures 3N,U and Supplementary Table S3). This shows that other restrictions exist between DOA9 and these two plant species resulting in inefficient symbiosis.

Only one species, S. hamata, was attributed to category 3, corresponding to a negative effect of the rhcN mutation. The number of nodules and nitrogenase activity of nodules induced by the ΩrhcN strain were reduced by approximately 30% compared with the wild-type strain (Figures 4A–G). These data suggest that the T3SS of strain DOA9 also plays a positive role in the establishment of the symbiosis, most probably by translocating effectors that weaken the plant immune system.

We also examined the effect of root exudates on the expression of rhcN and nopX genes to determine whether the absence of phenotype observed for the rhcN mutant in some legume species could be linked with the absence of formation of the T3SS apparatus. As expected, a significant level of expression of the two T3SS genes was detected in the five species displaying a positive or negative phenotype with the ΩrhcN strain (Supplementary Figure S1). In contrast, of the four species that displayed no phenotype with the rhcN mutant, the root exudates of two of them (I. tinctoria and D. tortuosum) displayed a similar level of expression to the control without inducers raising the possibility that the T3SS machinery is not formed during the early steps of the interaction between the DOA9 strain and these two legumes species. We also took advantage of the promoterless gusA reporter gene present in the insertional pVO155 plasmid used to construct the rhcN mutant to determine whether the T3SS is expressed in mature nodules as already reported in B. diazoefficiens USDA110 nodulating soybean (Zehner et al., 2008). As shown in Supplementary Figure S2, for the three tested species (A. americana, M. atropurpureum, and I. tinctoria), a GUS activity was detected in the infected nodule cells, in contrast to the WT nodules, in which no activity was observed (Supplementary Figure S2). Altogether, these data suggest that T3SS genes of DOA9 strain are active during the early and/or late stages of symbiosis on the different species tested.

We also investigated the effect of the T3SS mutation on the ability of DOA9 to colonize the surface of rice roots and to infect deep root tissue. These observations were made on a Thai rice (cultivar Pathum Thani 1) at the second and third week post-inoculation (wpi) (Figure 5F). The total population density of WT and ΩrhcN strains on the surface of the rice roots was almost the same (around 8 log₁₀ CFU/g of root fresh weight), and this level of population remained constant at 2 and 3 wpi. When the roots were surface sterilized to conserve only the endophytic bacteria cells, the estimated population of both WT and mutant strains was also very close (around 4 log₁₀ CFU/g root fresh weight). We also took advantage of the GFP tag added on the WT and ΩrhcN strains to monitor invasion of the roots. As can be seen in Figures 5A–C,E, both WT and ΩrhcN cells were attached to the root hairs and to the surface of the root epidermis. Analysis of root sections revealed endophytic bacterial cells in the intercellular space of the root cortex and endodermis. Taken together, these data suggest that the T3SS mutation did not affect the ability of the DOA9 strain to colonize and infect the rice root tissue intercellularly. This absence of effect of the T3SS mutation was not related to the absence of T3SS genes expression considering that a GUS activity could be detected at the surface of the rice roots inoculated with the rhcN mutant (Figure 5D).