Stringent Expression Control of Pathogenic R-body Production in Legume Symbiont Azorhizobium caulinodans

ABSTRACT R bodies are insoluble large polymers consisting of small proteins encoded by reb genes and are coiled into cylindrical structures in bacterial cells. They were first discovered in Caedibacter species, which are obligate endosymbionts of paramecia. Caedibacter confers a killer trait on the host paramecia. R-body-producing symbionts are released from their host paramecia and kill symbiont-free paramecia after ingestion. The roles of R bodies have not been explained in bacteria other than Caedibacter. Azorhizobium caulinodans ORS571, a microsymbiont of the legume Sesbania rostrata, carries a reb operon containing four reb genes that are regulated by the repressor PraR. Herein, deletion of the praR gene resulted in R-body formation and death of host plant cells. The rebR gene in the reb operon encodes an activator. Three PraR binding sites and a RebR binding site are present in the promoter region of the reb operon. Expression analyses using strains with mutations within the PraR binding site and/or the RebR binding site revealed that PraR and RebR directly control the expression of the reb operon and that PraR dominantly represses reb expression. Furthermore, we found that the reb operon is highly expressed at low temperatures and that 2-oxoglutarate induces the expression of the reb operon by inhibiting PraR binding to the reb promoter. We conclude that R bodies are toxic not only in paramecium symbiosis but also in relationships between other bacteria and eukaryotic cells and that R-body formation is controlled by environmental factors.

Contributions of the reb operon to R-body formation. To identify roles of the reb operon in R-body formation, we generated a praR deletion (ΔpraR) mutant, a deletion mutant with deletion of a region from AZC_3781 to AZC_3787 (ΔAZC_3781-7), and a ΔpraR ΔAZC_3781-7 double mutant and observed phenotypes of stem nodules at 14 days postinoculation (dpi) with these mutant and wild-type (WT) bacteria. The nitrogen-fixation-defective (Fix Ϫ ) phenotype of the stem nodules carrying bacteria with the praR deletion was suppressed by the second deletion in the AZC_3781-7 region ( Fig. 2A), as observed previously (8), whereas the AZC_3781-7 deletion did not affect reb operon expression levels (Fig. 2B). Transmission electron microscopy (TEM) observations showed that R bodies were produced in many ΔpraR bacterial cells in shrunken host cells, and many R-body-containing bacterial cells had collapsed appearances (Fig. 2C). R bodies were not observed in stem nodules harboring the double mutant (Fig. 2C), suggesting that genes that are essential for R-body formation are located in the region from AZC_3781 to AZC_3787.
To observe the dynamics of host-bacterium interactions in single nodules harboring the ΔpraR mutant, pathogenic roles of R bodies were observed at an early stage of nodule development (7 dpi). Some normal-shape host cells contained bacterial cells that lacked R bodies ( Fig. 3A and B), and nuclei in these normal-shape host cells were intact (Fig. 3C). In contrast, R bodies were observed in bacteria within shrunken host  cells ( Fig. 3A and D), in which nuclei were collapsed (Fig. 3E), indicating that R-body production is associated with host cell death in the nodules. Requirements of the non-reb-homologous genes AZC_3784, AZC_3785, and AZC_3787 and the reb-homologous genes reb AZC1 , reb AZC2 , reb AZC3 , and reb AZC4 for R-body production were determined in ΔpraR mutant derivatives with the deletions of these genes. In these experiments, AZC_3784 was essential for R-body formation and AZC_3785 and AZC_3787 were not (see Fig. S3 in the supplemental material). In addition, both reb AZC3 and reb AZC4 and either reb AZC1 or reb AZC2 were essential for R-body production (see Fig. S4 in the supplemental material).
In investigations of the contributions of RebR to nodule formation, deletion of rebR from the WT strain did not alter the phenotypic expression of stem nodules, but abolished the Fix Ϫ phenotype of nodules harboring the ΔpraR mutant so that the ΔpraR ΔrebR mutant produced the Fix ϩ phenotype (Fig. 4A) observed in the ΔpraR ΔrebR cells in the stem nodules (Fig. 4B). However, in symbiotic stem nodules, reb operon expression in the ΔpraR mutant was more than 10-fold higher than that in the ΔpraR ΔrebR mutant and was about 2 orders of magnitude higher than that in the WT strain, whereas levels of reb operon expression were similar in the WT   Total RNAs were isolated from bacteria residing in stem nodules and from bacterial cultures after growth to an OD 600 of approximately 1.0 at 38°C. Expression levels of the reb operon were estimated using quantitative RT-PCR and were normalized to 16S rRNA. Data are presented as means Ϯ standard deviations of three replicate cultures and plants and are expressed relative to mRNA levels in free-living WT cultures. Statistical analyses were carried out for stem nodules and free-living cultures, respectively. Different letters indicate significant differences (P Ͻ 0.05; Tukey-Kramer). and ΔrebR mutant strains (Fig. 4C). In contrast, in the free-living state, reb operon expression in the ΔpraR mutant was similar to that in the ΔpraR ΔrebR mutant. These results indicate that although PraR predominantly represses the expression of the reb operon, RebR acts an activator of the operon under conditions of symbiosis.
PraR and RebR directly control the reb operon. Following a systematic evolution of ligands by exponential enrichment (SELEX) analysis, Frederix et al. (18) proposed that PraR of R. leguminosarum binds the consensus palindrome CAAC-N5-GTTG. In the present SELEX analysis using N-terminally His 6 -tagged PraR (His 6 -PraR) from A. caulinodans, the consensus PraR sequence of A. caulinodans was also CAAC-N5-GTTG (see Fig. S5A in the supplemental material). However, no sequence on the promoter region of the reb operon (reb promoter) matched this consensus sequence perfectly or with a base substitution, whereas four sequences matched with two or three substitutions, and these were examined as candidates for PraR binding sites. Subsequent electrophoretic mobility shift assays (EMSAs) revealed that His 6 -PraR binds strongly to a sequence designated PraR-bs-A and weakly to the sequences of PraR-bs-B and -C ( Fig. S5B and C). Meanwhile, SELEX analysis using N-terminally His 6 -tagged RebR (His 6 -RebR) revealed that RebR potentially binds to a consensus palindrome, GT(A/G)(A/ C)C-N4-G(T/G)(T/C)AC (Fig. S5D). A sequence (designated RebR-bs) matching this consensus palindrome was present on the reb promoter, and the EMSA revealed that His 6 -RebR binds to this sequence (Fig. S5E). The positions of PraR-bs-A, -B, and -C and RebR-bs on the reb promoter are shown in Fig. 1.
To investigate whether PraR and RebR actually bind to the reb promoter, we carried out EMSAs using a double-stranded DNA (dsDNA) probe that covers the intergenic noncoding region upstream of the reb operon (reb promoter dsDNA probe) as shown in Fig. 5A. The molecular weight (MW) of the reb promoter dsDNA probe increased with the increasing quantity of His 6 -PraR, whereas addition of His 6 -RebR to the promoter probe increased not only the MW of the probe but also the amount of stacked probe in the wells of the gel in a concentration-dependent manner (Fig. 5B). Even in the presence of His 6 -PraR, the addition of His 6 -RebR to the promoter probe increased the MW of the probe (Fig. 5C), indicating that PraR does not interfere the binding of RebR to the RebR-bs.
To confirm the involvement of PraR and RebR in reb operon expression, the significance of the promoter sequences was assessed in stem nodules that were formed after inoculation with mutants carrying base substitutions of P reb (PraR-bs-A Ϫ ) and P reb (RebR-bs Ϫ ) mutants and after inoculation with the P reb (PraR-bs-A Ϫ RebR-bs Ϫ ) double mutant (Fig. 6A). Stem nodules harboring the P reb (PraR-bs-A Ϫ ) mutant had the Fix Ϫ phenotype, whereas those harboring the P reb (PraR-bs-A Ϫ RebR-bs Ϫ ) double mutant had restored nitrogen-fixing activity (Fig. 6B). Accordingly, R bodies were observed in P reb (PraR-bs-A Ϫ ) mutants but not in P reb (PraR-bs-A Ϫ RebR-bs Ϫ ) double mutants in stem nodules (Fig. 6C). In further experiments, expression levels of the reb operon in P reb (PraR-bs-A Ϫ ) mutants and P reb (PraR-bs-A Ϫ RebR-bs Ϫ ) double mutants were about 2 orders of magnitude and several times higher than those in the WT strain, respectively, whereas reb operon expression levels in the P reb (RebR-bs Ϫ ) mutant were similar to Pathogenicity of R-body-Producing Rhizobium ® those in the WT strain (Fig. 6D). The consistency of phenotypic expression between deletion mutants of praR and rebR (ΔpraR, ΔrebR, and ΔpraR ΔrebR mutants in Fig. 4) and corresponding promoter sequence mutants strongly indicates that PraR and RebR directly control the expression of the reb operon. Identification of environmental factors that abolish reb repression by PraR. In the experiments described above, symbiotic R-body production was observed in ΔpraR and P reb (PraR-bs-A Ϫ ) mutants but was not present in the WT strain under symbiotic or free-living conditions, suggesting that R-body production is subjected to environmental conditions. Thus, we investigated environmental factors that induce RebR-dependent activation of the reb operon in the ΔpraR mutant and then identified factors that attenuate PraR-dependent repression of the reb operon in WT cells grown under the identified favorable environmental conditions.
Initially, we constructed a reb operon-lacZ transcriptional fusion (reb-lacZ) on chromosomes of the WT and ΔpraR strains, namely reb-lacZ and reb-lacZ ΔpraR strains, respectively. During these manipulations, the reb-lacZ ΔpraR strain expressed ␤-galactosidase in the free-living state at room temperature (around 26°C) but not at 38°C, suggesting that activation by RebR is temperature dependent. Accordingly, when reb-lacZ and reb-lacZ ΔpraR strains were grown at various temperatures between 26 and 41°C, ␤-galactosidase activity was highly induced below 35°C in the reb-lacZ ΔpraR strain, but not in the reb-lacZ strain (Fig. 7A). Moreover, expression levels of the reb operon were about 30-fold higher at 26°C than at 38°C in the ΔpraR strain, whereas activation at 26°C was not observed in either the ΔrebR or ΔpraR ΔrebR mutant (Fig. 7B). Under free-living conditions, R bodies were observed in up to 10% of ΔpraR cells grown at 26°C, but not in those grown at 38°C (Fig. 7C). Similarly, ΔpraR cells in stem nodules (symbiotic state) failed to produce R bodies when plants were grown at 38°C, and reb operon expression was lower than that in ΔpraR cells under symbiotic conditions at 30°C (see Fig. S6 in the supplemental material). Taken together, these data indicate that activation of the reb operon by RebR is temperature dependent under both free-living and symbiotic conditions. However, binding of RebR to the reb promoter was not affected by temperature (see Fig. S7 in the supplemental material).
In subsequent experiments, the effects of host-derived tricarboxylic acid (TCA) cycle, nitrogen, and oxygen metabolites on reb expression were investigated in reb-lacZ and reb-lacZ ΔpraR strains. In the absence of praR (the reb-lacZ ΔpraR strain), all organic acids except for citrate promoted reb expression under nitrogen-sufficient conditions at either 21 or 3% oxygen (Fig. 8). However, even in the presence of praR (reb-lacZ strain), reb expression was induced in medium containing 2-oxoglutarate (2OG) with sufficient nitrogen sources (Fig. 8). Moreover, induction was increased with 2OG concentrations even in the presence of succinate as a carbon source (Fig. 9A), suggesting that 2OG induces reb irrespective of its metabolism as a carbon source, although these effects of 2OG were observed at 26°C but not at 38°C (Fig. 9B). In agreement, TEM observations showed that R bodies were produced in WT cells grown in the presence of 2OG at 26°C but not at 38°C (Fig. 9C). Taken with the absence of a response to 2OG in ΔrebR mutant cells (Fig. 9D), these observations indicate that rebR is essential for 2OG-mediated induction of the reb operon. Moreover, the ΔpraR mutants did not respond to 2OG, further suggesting that 2OG derepresses the reb operon by attenuating PraRdependent repression. Western blotting of an rgs-his 6 -praR transformant (19) showed that expression levels of RGS-His 6 -PraR protein were not affected by 2OG (Fig. 9E). However, binding of PraR to the reb promoter dsDNA probe decreased with increasing 2OG concentration, whereas RebR binding was impervious to 2OG ( Fig. 9F; Fig. S7). These observations strongly suggest that 2OG derepresses the reb operon directly by concentration-dependently inhibiting PraR binding to the reb promoter.

DISCUSSION
In this study, we demonstrated that reb-driven pathogenicity is associated with R-body production by A. caulinodans and suggest that R-body production is a widespread trait among bacterial pathogens that carry reb operons. Although the reb operon is predominantly repressed by PraR, we identified biological and environmental factors that derepress reb gene expression and thus R-body production under freeliving conditions. R bodies are rolled up at neutral pH, but reportedly unroll to form needle-shaped structures at low pH (10). Recombinant R bodies from Escherichia coli act as pistons that puncture spheroplasts of E. coli at low pH (20). In paramecia, the role of R bodies in the killer trait follows release of R-body-containing bacteria from killer paramecia and ingestion by sensitive paramecia. Subsequently, internalized bacteria enter acidified food vacuoles, and R bodies are unrolled and penetrate the phagosomal membrane to deliver lethal toxins to the cytoplasm (9, 21). This scenario may also be applicable to interactions of A. caulinodans and S. rostrata, in which the peribacteroid space (mi- croenvironment surrounding bacteroids) is progressively acidified during nodule morphogenesis (22), likely triggering the conformational change of R bodies into the needle-shaped structure that penetrates membranes. The present series of mutant analyses showed that essential genes for R-body formation in A. caulinodans include both reb-homologous and non-reb-homologous genes. Although the roles of the proteins encoded by the reb-homologous genes remain poorly understood, they are likely to be components of R bodies (5). Among non-reb-homologous genes, AZC_3784 was found in the reb operon and was also essential to R-body formation. In C. taeniospiralis, RebC is encoded by a non-rebhomologous gene that may be involved in the assembly of R bodies (5). Similarly, the AZC_3784 protein is not a component of R bodies but is likely involved in their assembly, although it may lack homology to rebC from C. taeniospiralis. Taken together, these observations warrant further compositional analyses of R bodies and investigations of the molecular mechanisms of R-body formation.
R bodies were frequently produced by ΔpraR mutants in the symbiotic state, but were observed in fewer than 10% of free-living cells, even at the optimum temperature (26°C) for reb operon expression. These results suggest that R-body formation is more strongly regulated (suppressed) in the free-living state-probably at the translation level or at the R-body assembly level-and that A. caulinodan R bodies play more important roles in the symbiotic state.
Although R-body formation was not observed in WT bacterial cells in the symbiotic state, the present environmental factors that derepress the praR regulatory system have significant implications for the understanding of reb operon evolution during microsymbiosis of A. caulinodan with S. rostrata. It is widely accepted that virulence genes are regulated by temperature in many pathogenic bacteria and are upregulated in mammalian bacterial pathogens at around host body temperature (37°C) (23,24). In contrast, most plant-pathogenic bacteria express virulence genes at ambient temperatures that are generally lower than their optimal growth temperatures (25). For example, Agrobacterium tumefaciens mediates the formation of crown galls at temperatures below 32°C, and the VirA/VirG two-component system regulates the expression of virulence genes according to temperature (26,27). In agreement, the reb operon was reb-lacZ reb-lacZ ∆praR reb-lacZ reb-lacZ ∆praR reb-lacZ reb-lacZ ∆praR  expressed in the ΔpraR mutant of A. caulinodans at temperatures below 35°C and within the optimal range for the growth of the host plant (around 30°C). Moreover, because regulation by PraR was derepressed by 2OG in the present free-living WT cells at 26°C, the reb operon may also be induced during symbiosis in host nodules, wherein the 2OG is accumulated in the host plant cells, although we have not found the conditions under which 2OG actually accumulates in the host plant cells. Plant 2OG levels reflect cellular C/N status and may play a signaling role in the coordination of C and N metabolism (28). Alterations in the activities of nitrogen fixation by bacteria and ammonia assimilation by plant cells may lead to 2OG accumulation in host cells. To elucidate the roles of reb operon in the symbiotic state, we need to conduct more investigations to estimate the conditions wherein 2OG is accumulated in the host plant cells.
Although PraR homologues are widespread among Alphaproteobacteria (8), the roles of PraR have not been well characterized. In particular, the praR homologue phrR was originally identified in the acid-tolerant rhizobium Sinorhizobium medicae WSM419 as a gene that is induced at low pH (29). However, praR expression is not pH sensitive in A. caulinodans and Rhizobium leguminosarum (8,18). Moreover, R. leguminosarum PraR directly represses the expression of the quorum-sensing genes rhiR and raiR and the biofilm formation genes rapA2, rapB, and rapC (18,30), whereas the homologous genes in A. caulinodans are not controlled by PraR (8). Hence, although PraR homologues are widely distributed, the roles of PraR have diversified during the evolution of Alphaproteobacteria. The ubiquity of chemical and environmental factors that regulate praR expression in the Alphaproteobacteria, such as the effects of A. caulinodans factors, also requires investigation in the context of the evolution of praR regulatory systems. RebR belongs to a novel subfamily of the Crp-Fnr superfamily, and all Crp-Fnr members carry putative DNA-binding helix-turn-helix domains on their C terminus and ligand-binding domains on their N terminus (17). Various intracellular and exogenous signals activate Crp-Fnr members via their ligand-binding domains, including 2OG and temperature (17). In A. caulinodans, however, binding activity of RebR to the reb operon was not affected by 2OG and temperature, indicating that in addition to 2OG and temperature, as yet unidentified factors are involved in the activation of reb operon expression via RebR.
Herein, we demonstrated the roles of R bodies in the pathogenicity of bacteria that harbor the reb operon, although the ensuing roles in nodule symbiosis and the related evolutionary implications remain uncharacterized. Because bacterial genomes are plastic, endosymbionts may become pathogenic after acquiring the reb operon if they do not suppress its expression. Although we did not determine whether R bodies threaten biodiversity or ecosystems, this possibility may require solutions in the future. Unlike obligate endosymbionts of paramecia, A. caulinodans can be cultured in vitro and genetic manipulation techniques have been established in this bacterium, warranting further use of A. caulinodans as a model for studies of R-body/reb genes.

MATERIALS AND METHODS
Bacterial strains and culture conditions. The bacterial strains used in this study are listed in Table S1 in the supplemental material. A. caulinodans strains were grown in tryptone-yeast extract (TY) medium (31)  In some experiments, BD medium was further supplemented with disodium 2OG. To grow A. caulinodans strains under aerobic conditions, test tubes containing medium were sealed with butyl rubber septums, and the contained air was replaced with N 2 gas with 3% O 2 . Before inoculation into BD medium, bacterial cells were cultured overnight in TY medium and were washed twice in 10 mM potassium phosphate buffer (pH 7.0). Unless otherwise noted, initial optical density at 600 nm (OD 600 ) values of cultures were adjusted to 0.1 or 0.02 for growth at 26 or 38°C, respectively, and OD 600 values were approximately 1.0 after 24 h of incubation.
Construction of deletion and substitution mutants. The plasmids and primers used for strain construction are listed in Tables S1 and S2 in the supplemental material, respectively.
To construct A. caulinodans deletion mutants of AZC_3784, AZC_3785, AZC_3787, and rebR genes, two DNA fragments containing upstream and downstream regions of each gene were amplified from the WT genomic DNA by PCR using appropriate primer pairs and were then directionally cloned into a suicide vector, pK18mobsacB (32), using the In-Fusion cloning kit (Clontech, Mountain View, CA). The linearization of pK18mobsacB was performed by inverse PCR using the PrimeSTAR Max (TaKaRa-Bio, Shiga, Japan) with primer pair Tp73-Tp74. The resulting plasmids were conjugated into the WT or ΔpraR (8) strains via E. coli S17-1(pir) (33), and gene deletions were introduced by allelic exchange.
To construct deletion mutants of the reb AZC1 , reb AZC2 , reb AZC3 , and reb AZC4 genes, a series of plasmids were constructed as follows. DNA fragments containing the WT AZC_3781-7 region with its upstream and downstream regions were amplified by PCR and cloned into the linearized pK18mobsacB. Genes on plasmids containing the WT region were deleted by inverse PCR using the PrimeSTAR Mutagenesis Basal kit (TaKaRa-Bio). To introduce double deletions, second inverse PCRs were conducted using plasmids harboring single mutations. Constructed plasmids were conjugated into the ΔpraR ΔAZC_3781-7 mutant, and deletion mutants were obtained after allelic exchange.
To construct mutants with base substitutions in PraR-bs-A and/or RebR-bs, a series of plasmids were constructed as follows. A DNA fragment containing the WT reb promoter region was amplified by PCR and cloned into the linearized pK18mobsacB. An XbaI site was generated within the PraR-bs-A on the plasmid containing the reb operon by inverse PCR using the PrimeSTAR Mutagenesis Basal kit. Similarly, an EcoRI site was generated within the RebR-bs on the plasmid containing the reb promoter. The resulting plasmids were conjugated into the WT strain, and mutants were obtained after allelic exchange.
To construct strains that express the reb-lacZ fusion gene, two fragments containing rebR and AZC_3789 and a lacZ fragment were amplified by PCR from the WT genomic DNA and the plasmid pTA-MTL (34), respectively. Fragments were then cloned into the linearized pK18mobsacB in the direction of the rebR, lacZ, and AZC_3789 fragments using the In-Fusion cloning kit. The plasmid containing reb-lacZ was conjugated into the WT and ΔpraR strains, and strains with lacZ at the position immediately downstream of the rebR open reading frame (ORF) were obtained after allelic exchange.