INTRODUCTION

Top Abstract Introduction Materials and Methods Results Discussion References

Rhizobitoxine [2-amino-4-(2-amino-3-hydropropoxy)-trans-but-3-enoic acid] is synthesized by the legume symbiont Bradyrhizobium elkanii (37) and the plant pathogen Burkholderia andropogonis (29). Because it induces foliar chlorosis of soybeans, rhizobitoxine has been regarded as a plant toxin (18, 36, 57). In terms of biochemical functions, rhizobitoxine inhibits -cystathionase in the methionine biosynthesis pathway (39, 57) and 1-aminocyclopropane-1-carboxylate (ACC) synthase in the ethylene biosynthesis pathway (59).

Recently, a beneficial role for rhizobitoxine in Rhizobium-legume symbiosis has been revealed. Using a rhizobitoxine mutant, Yuhashi et al. (60) found that rhizobitoxine production by B. elkanii enhances nodulation and competitiveness in the legume Macroptilium atropurpureum (siratro), probably via the inhibition of endogenous ethylene production in the host plant. Duodu et al. (7) reported that rhizobitoxine mutants formed fewer mature nodules on Vigna radiata (mung bean) than the wild-type strain. In addition, application of ethylene inhibitors to the rhizobitoxine mutants partly restored the nodulation phenotype. Therefore, rhizobitoxine is a nodulation enhancer rather than a phytotoxin for siratro and mung bean, although it is unlikely that rhizobitoxine exerts this positive effect in nodulation of soybean cultivars (28, 43, 60).

The biosynthetic pathway for rhizobitoxine has not been elucidated fully. Ruan et al. (43-45) obtained two Tn5-induced rhizobitoxine null mutants of B. elkanii USDA61 and isolated the rtxA gene, which is responsible for rhizobitoxine biosynthesis in culture and in planta. The N-terminal region of the amino acid sequence of rtxA has a motif that is homologous to an aminotransferase, whereas one similar to O-acetylhomoserine sulfhydrolase is found in the C-terminal portion (43, 44); however, there is some confusion about frameshift in rtxA genes. Proposed as a precursor of rhizobitoxine, serinol is abundant in soybean nodules formed by B. elkanii (23, 26, 30). A mutant with a Tn5 insertion in the portion of the rtxA gene corresponding to the N-terminal part of the protein was defective in serinol accumulation in soybean nodules, suggesting that the N-terminal part functions as an aminotransferase in serinol production, but this function has not yet been verified in pure culture.

Dihydrorhizobitoxine [O-(2-amino-3-hydroxypropyl) homoserine] has been found in cultures and nodules of B. elkanii (38), and it is less potent than rhizobitoxine as an inhibitor of ACC synthase and -cystathionase (59). Mitchell and Coddington (30) suggested that dihydrorhizobitoxine is an end product that lacks biological activity. However, in light of the homology with sulfhydrylase (44, 45), the C-terminal portion of the rtxA product may be involved in dihydrorhizobitoxine formation as an intermediate in rhizobitoxine biosynthesis. The absence of unequivocal, systematic determination of dihydrorhizobitoxine and serinol has complicated efforts to study the rhizobitoxine biosynthetic pathway, although two bioassay systems have been developed for rhizobitoxine (46, 59).

The aim of the present work was to investigate the rhizobitoxine biosynthetic pathway of B. elkanii in culture by an approach that combines mutagenesis of the rtxA gene and its flanking regions with unequivocal determination of the rhizobitoxine intermediates in culture by using liquid chromatography and mass spectroscopy. To this end, we chose the B. elkanii strain USDA94, which produces high concentrations of rhizobitoxine in culture; however, this strain is rather difficult to manipulate genetically because of increased resistance to antibiotics, particularly tetracycline.

	MATERIALS AND METHODS

Top Abstract Introduction Materials and Methods Results Discussion References

Bacterial strains, plasmids, and growth conditions. The bacterial strains and plasmids used in this study are listed in Table 1. B. elkanii cultures were grown aerobically at 30°C in HM salt medium (5) supplemented with 0.1% arabinose and 0.025% yeast extract (Difco, Detroit, Mich.) or in Tris-YMRT medium (25). Escherichia coli cells were grown at 30°C in Luria-Bertani medium (47). Antibiotics were added to media at the following concentrations: spectinomycin and streptomycin at 250 µg/ml and kanamycin at 150 µg/ml for B. elkanii, and tetracycline at 12.5 µg/ml, ampicillin at 100 µg/ml, and kanamycin at 100 µg/ml for E. coli.

DNA isolation and manipulations. Isolation of plasmid DNA, restriction enzyme digestion, DNA ligation, bacterial transformation of E. coli, and Southern hybridization were performed as described by Sambrook et al. (47). Total DNA of B. elkanii was prepared as described previously (21). To isolate the DNA regions flanking rtxA, we used a cosmid library of B. elkanii USDA94 in the pLAFR1 vector (12) that had been constructed previously (61). E. coli HB101 cells containing the library were plated on Luria-Bertani agar (1.5%) with tetracycline. The B. elkanii USDA94 rtxA gene was PCR amplified as described previously (60) and was used as a colony hybridization probe. The probe was labeled by using digoxigenin-dUTP random priming (Boehringer Mannheim, Mannheim, Germany), and hybridization signals were detected by using the digoxigenin nucleic acid detection kit (Boehringer Mannheim). Cosmid clones from colonies showing a positive hybridization signal were isolated, digested with EcoRI, and subjected to electrophoresis in 0.8% agarose-TAE (47) to make a restriction map of the regions flanking rtxA of B. elkanii USDA94.

DNA sequencing. The B. elkanii USDA94 genome cosmid clones pRTN2, pRTF1, and pRTS1, which cover a 40- to 45-kb DNA region containing rtxA, were digested into 4- to 8-kb fragments by using several restriction enzymes. These DNA fragments were ligated into pBluescript SK(+) (Takara Shuzo Co., Ltd., Kusatsu, Japan) and transformed into E. coli JM109. The cloned plasmid DNAs were isolated and digested into 0.5- to 4-kb fragments by sonication. The 1- to 2-kb DNA fragments were purified by 1% agarose gel electrophoresis, blunt ended by using KOD polymerase (Toyobo Inc., Tokyo, Japan), ligated into HincII-digested pUC118 (Takara Shuzo Co.), and transformed into E. coli DH5. Consequently, more than 300 plasmids containing portions of rtxA and its flanking regions were isolated.

Construction of a B. elkanii mutant with a large deletion of rtxA and its flanking region. For the construction of a large deletion mutant for the putative rhizobitoxine gene, the 13.4-kb ApaI-NotI fragment from pRTF1 was cloned into pBSIISacI. The resulting plasmid (pBS13.4) was digested by SacI, and a 6.4-kb fragment was isolated. This 6.4-kb fragment was blunt ended and then ligated with the 2.1-kb  cassette from pHP45 (42). The resulting plasmid (pBS3.6::) was double digested with ApaI and NotI, and a 5.7-kb ApaI-NotI fragment containing noeE, the  cassette, a partial open reading frame (ORF3), and ORF4 was cloned into pSUP202 (51), yielding pSUP3.6::. pSUP3.6:: was introduced into B. elkanii USDA94 by triparental mating using pRK2013 as a helper plasmid (9). Crossover mutants were selected by screening for resistance to streptomycin and spectinomycin, and a double-crossover mutant was identified by Southern hybridization with the 5.7-kb ApaI-NotI fragment from pSUP3.6:: as a probe. The relevant characteristics of the double-crossover mutant USDA94rtx:: 1 and plasmids used for the construction are listed in Table 1.

Construction and complementation of pRTF1 derivatives with kanamycin cassette insertion. To examine the function of rtxA and its flanking region in rhizobitoxine biosynthesis, we used the genome priming system kit (GPS-1; New England BioLabs, Inc., Beverly, Mass.) according to the manufacturer's instructions to randomly insert a kanamycin cassette into these sequences. After mutation of 0.08 µg of pRTF1 and 0.02 µg of pGPS1.1 and transformation, kanamycin-resistant colonies were isolated, and plasmids were purified, digested with EcoRI, and electrophoresed to broadly specify the sites of insertion and eliminate mutants with multiple cassettes. The exact insertion points were determined by DNA sequencing using outward primers of the kanamycin cassette, as described in the manufacturer's instructions. These pRTF1 insertion mutant derivatives were introduced into B. elkanii USDA94rtx::1 by triparental mating with pRK2013 (9). Transconjugants were selected by screening for kanamycin resistance and were assayed for serinol, dihydrorhizobitoxine, and rhizobitoxine production.

LC/MS analysis of serinol, dihydrorhizobitoxine, and rhizobitoxine. We simultaneously determined the serinol, dihydrorhizobitoxine, and rhizobitoxine concentrations in cultures of B. elkanii by using liquid chromatography and mass spectrometry (LC/MS) to quantitate their phenylthiocarbamyl derivatives. A 15-ml aliquot of a stationary-phase culture of B. elkanii grown in Tris-YMRT medium was centrifuged at 10,000 × g for 10 min. The resulting supernatant was loaded on a Dowex 50 column (H+ type; resin size, 50 to 100 mesh; column volume, 5 ml; Muromachi Chemicals, Tokyo, Japan). The column was washed with 10 column volumes of deionized water. Serinol, dihydrorhizobitoxine, and rhizobitoxine were eluted with 3 column volumes of 2 M NH₄OH, and evaporated in vacuo. Pellets were dissolved in 500 µl of deionized water, and 10 nmol of aminoethoxyvinylglycine (a structural analogue of rhizobitoxine) was added as an internal standard before phenylthiocarbamyl derivatization.

Conversion of serinol and various compounds to dihydrorhizobitoxine. A 500-ml aliquot of a stationary-phase culture of B. elkanii USDA94 in HM medium was centrifuged at 10,000 × g for 10 min at 20°C. The cells were washed twice with 300 ml of 50 mM potassium phosphate buffer (KP; pH 6.8), resuspended in 10 ml of 50 mM KP (pH 6.8), and aliquoted into Eppendorf tubes (1.5 ml). Then the amount of cells was adjusted to 60 mg (wet weight)/tube. After centrifugation, the cells were resuspended in 1 ml of 20 mM KP (pH 6.8) containing 1 mM homoserine, O-acetylhomoserine, cysteine, cystathionine, homocysteine, or methionine. O-Acetylhomoserine was synthesized from L-homoserine and acetic anhydride according to the method of Nagai and Flavin (33); the other compounds were purchased from Wako Pure Chemical Industries Ltd. (Osaka, Japan), Nacalai Tesque Inc. (Kyoto, Japan), and Sigma Chemical Co. The cell suspensions were incubated with shaking at 30°C for 1 h in the dark. After centrifugation, the supernatants were analyzed by LC/MS as described above.

	RESULTS

Top Abstract Introduction Materials and Methods Results Discussion References

Selection of pLAFR1 cosmids containing rtxA from B. elkanii USDA94 library. In previous work, the rtxA gene of B. elkanii USDA94, a high rhizobitoxine producer, was PCR amplified by using two primers designed from published rtxA sequence of B. elkanii USDA61, a low rhizobitoxine producer in culture (60). Using a PCR-derived fragment of rtxA in B. elkanii USDA94 as a probe (Fig. 1), we carried out colony hybridization against the pLAFR1 cosmid library of B. elkanii USDA94 (61). Consequently, we were able to align seven independent cosmids containing rtxA in light of their EcoRI restriction sites (Fig. 1). The identity of the 4.3-kb EcoRI fragment containing rtxA in the seven cosmids was verified by Southern hybridization with the PCR-derived fragment of rtxA from B. elkanii USDA94 as a probe (Fig. 1).

View larger version (14K): [in this window] [in a new window]   FIG. 1.   Physical map of the flanking regions of rtxA. (A) Seven independent overlapping pLAFR1 cosmids containing rtxA from a genomic library of B. elkanii USDA94. (B) In the 28,401 bp of sequence obtained, an fixGHIS cluster, noeE, nodPQ, rtxA, and six ORFs were found. Restriction sites: A, ApaI; B, BamHI; E, EcoRI; H, HindIII; P, PstI. C, probe used for colony and Southern hybridization.

Nucleotide sequence around the rtxA gene. The DNA encompassing the rtxA gene of B. elkanii USDA94 was sequenced, and we identified 14 ORFs, all of which were in the same orientation. Of these, seven ORFs upstream of rtxA are homologous to fixGHIS, noeE, and nodPQ and appear to be involved in symbiotic functions. Another four ORFs are downstream of rtxA, suggesting that they are involved in rhizobitoxine biosynthesis. A potential promoter for sigma 70 was found 0.5 kb upstream of rtxA, although the other potential promoters recognized by NodD, FixK, and sigma 54 were not found in the DNA regions by their consensus sequences (15). Detailed descriptions of each gene and ORF follow.

The deduced amino acid sequence of rtxA in B. elkanii USDA94. The deduced amino acid sequence of rtxA (803 amino acid residues) in B. elkanii USDA94 was 95% similar to that of B. elkanii USDA61, a low rhizobitoxine producer (44, 45). The 346 N-terminal residues had 24% identity and 40% similarity to the aminotransferase of Methanobacterium thermoautotrophicum (52). The 443 C-terminal residues had 41% identity and 56% similarity to the O-acetylhomoserine sulfhydrylase of Leptospira meyer (2). O-Acetylhomoserine sulfhydrylase synthesizes sulfur-containing amino acids from O-acetylhomoserine and sulfide. Generally, the enzyme shows O-alkylhomoserine synthase activity from O-acetylhomoserine and alcohol as well (31, 32), whose reaction mode resembles that of dihydrorhizobitoxine synthesis from O-acetylhomoserine and serinol. These amino acid homologies of rtxA product in B. elkanii USDA94 are similar to those of B. elkanii USDA61, although the rtxA gene was formally separated into rtxA and rtxB genes in B. elkanii USDA61 because of a sequencing error (44, 45). The predicted amino acid sequences for rtxA suggested that their possible enzymatic functions might be involved in serinol formation (1) and dihydrorhizobitoxine synthesis.

Deduced amino acid sequence of ORF1 (rtxC). The deduced amino acid sequence of ORF1 (352 amino acid residues) had 19% identity and 31% similarity to the fatty acid desaturase of Pseudomonas syringae (62) (Fig. 2). Although the predicted amino acid sequence of ORF1 indicated low similarity to other desaturases, alignment of these sequences revealed the presence of two regions conserved among membrane-bound desaturases (49, 62): a copper-binding signature and eight histidine residues. Analysis of the deduced secondary structure showed two potential transmembrane regions in the ORF1 product (56) (Fig. 2). Rhizobitoxine possesses a double bond between C-3 and C-4 (37). Therefore, we hypothesized that the ORF1 product catalyzes the introduction of the carbon double bond into dihydrorhizobitoxine. We consequently designated ORF1 as rtxC, which begins 32 bp downstream of rtxA and lacks an upstream promoter-like sequence. Therefore, rtxA and rtxC (at least) probably form an operon.

View larger version (82K): [in this window] [in a new window]   FIG. 2.   Alignment of the amino acid sequence of rtxC product with those of various desaturases. PsDES, P. syringae desaturase (U27310); AtDES, Arabidopsis thaliana desaturase (Al022198); EgDES, Euglena gracilis -8 fatty acid desaturase (AF139720); CrDES, Chlamydomonas reinhardtii desaturase (AB007640). Conserved among membrane-bound desaturases, the copper-binding signature () and eight histidine residues () of rtxC product were highly homologous to those of fatty acid desaturases from several sources. Putative membrane-spanning regions were deduced by hydropathy analysis and are indicated by asterisks.

Deduced amino acid sequences of ORF2, ORF3, and ORF4. ORF2 began 192 bp downstream of rtxC, and its deduced amino acid sequence (207 amino acid residues) had 45% identity and 61% similarity to the amidotransferase subunit of Pseudomonas aeruginosa PAO1 (53). ORF3 began 555 bp downstream of ORF2, and the deduced amino acid sequence of ORF3 (588 amino acid residues) lacked homology to any known protein. ORF4 began in the termination codon of ORF3 (TGA-TG). The deduced amino acid sequence of ORF4 (444 amino acid residues) had 34% identity and 53% similarity to the glutamine synthetase of Mycobacterium tuberculosis (6).

Deduced amino acid sequences of the remaining ORFs. The deduced amino acid sequence of ORF5 had 84% similarity to the transposase of Rhizobium sp. strain NGR234 (11), and that of ORF6 had 79% similarity to the ATP-binding helper protein of Rhizobium sp. strain NGR234. The insertion sequence (IS) element NGR IS5 belongs to the IS21/IS1162 family. ORF5 and ORF6 probably form an IS element belonging to the IS21/IS1162 family, which has been identified as IS1631 in Bradyrhizobium japonicum (17).

Establishment of a complementation system for rhizobitoxine biosynthesis. The efficiency of homologous recombination in B. elkanii is lower than that in B. japonicum (22). Therefore, to evaluate their functions in rhizobitoxine biosynthesis, we adopted a shortcut strategy in which cosmids mutagenized by insertion of a kanamycin cassette complement a B. elkanii USDA94 mutant lacking a putative DNA region for rhizobitoxine biosynthesis. We first constructed the large-deletion mutant USDA94rtx::1, which lacks a 9.8-kb region (nodQP, rtxA, rtxC, ORF2, and truncated ORF3) of the B. elkanii USDA94 chromosome (Fig. 3A). We could not delete the entire ORF3 sequence or ORF4 because of the absence of appropriate restriction sites.

View larger version (19K): [in this window] [in a new window]   FIG. 3.   Serinol, dihydrorhizobitoxine, and rhizobitoxine production by B. elkanii USDA94rtx::1 complemented with pRTF1 cosmid derivatives created by insertion of a kanamycin cassette. (A) Large-deletion mutant USDA94rtx::1, which lacks a 9.8-kb region (nodQP, rtxA, rtxC, ORF2, and truncated ORF3) of B. elkanii USDA94. , 9.8-kb deleted SacII fragment. (B) The insertion point of each kanamycin cassette is indicated by the arrowheads. The N and C domains of rtxA show the regions where their deduced amino acid sequences are homologous to aminotransferase and O-acetylhomoserine sulfhydrylase, respectively (see the text). P, putative promoter (according to the sequence). For serinol production, +, ++, and +++ indicate 0 to 50, 50 to 100, and >100 µM, respectively; for dihydrorhizobitoxine production, +, ++, +++, and ++++ indicate 0 to 2, 2 to 5, 5 to 10, and >10 µM, respectively; for rhizobitoxine production, +, ++, +++, and ++++ indicate 0 to 0.5, 0.5 to 1.0, 1.0 to 10, and >10 µM, respectively. ND, not detected.

View larger version (17K): [in this window] [in a new window]   FIG. 4.   LC/MS chromatograms of cultures of B. elkanii USDA94 (wild-type), USDA94rtx::1, and USDA94rtx::1(pRTF1-F1). (A) PITC derivatives of serinol were detected at m/z 227 and eluted at a retention time of 3.8 min. (B) PITC-dihydrorhizobitoxine was detected at m/z 463 and eluted at a retention time of 10.4 min. (C) PITC-rhizobitoxine was detected at m/z 461 and eluted at a retention time of 10.4 min. The large-deletion mutant USDA94rtx::1 ceased to produce serinol, dihydrorhizobitoxine, and rhizobitoxine. However, these compounds were again produced after introduction of pRTF1-F1 into the mutant.

Production of rhizobitoxine, dihydrorhizobitoxine, and serinol in USDA94rtx::1 complemented by various pRTF1 derivatives. To examine the functions of rtxA and its associated ORFs in rhizobitoxine biosynthesis, we constructed 12 independent pRTF1 derivatives in which the kanamycin cassette was inserted into this DNA region and then analyzed the serinol, dihydrorhizobitoxine, and rhizobitoxine in the culture supernatants of the pRTF1 derivatives (Fig. 3B). The derivatives were named by adding the abbreviations C3, C1, D8, D2, D5, D3, E26, E9, E10, E2, E8, and F6 to the designation pRTF1 (Table 1), and the corresponding positions of kanamycin cassette insertions are shown in Fig. 3. Mutants with insertions of the cassette at positions C3 and C1, which are located 1.3 and 0.6 kb upstream of rtxA, continued to produce serinol (C3, 142 µM; C1, 116 µM), dihydrorhizobitoxine (C3, 7 µM; C1, 6 µM), and rhizobitoxine (C3, 4 µM; C1, 5 µM) in culture. However, the D8 insertion (located 0.2 kb upstream of rtxA) stopped dihydrorhizobitoxine and rhizobitoxine production and reduced serinol production (24 µM). Because we found a putative promoter sequence (5'-TTGAAA-cgcacctaacgtcaagttg-TACGAT-3') 0.5 kb upstream of rtxA, the loss of or decrease in the ability to produce these compounds probably is due to a polar effect of the D8 insertion downstream of the rtxA promoter.

Effect of homoserine-like compounds on dihydrorhizobitoxine production in B. elkanii USDA94. The C-terminal portion of the rtxA product is homologous to O-acetylhomoserine sulfhydrylase, and the results of our disruption experiments suggested that its enzymatic function is involved in dihydrorhizobitoxine synthesis. However, O-acetylhomoserine is located within the methionine biosynthetic pathway (31, 32). To identify possible substrates for this putative dihydrorhizobitoxine synthase, we investigated the dihydrorhizobitoxine production by B. elkanii USDA94 cell suspensions in the presence of various compounds in the methionine biosynthetic pathway (Fig. 5). Dihydrorhizobitoxine production was increased dramatically by the addition of these compounds, in particular sulfur-containing compounds such as methionine.

View larger version (26K): [in this window] [in a new window]   FIG. 5.   Effect of homoserine, O-acetylhomoserine, and sulfur-containing amino acids on dihydrorhizobitoxine production in B. elkanii USDA94. B. elkanii cells were incubated in 20 mM potassium phosphate buffer (pH 6.8) containing homoserine, O-acetylhomoserine, or sulfur-containing amino acids at 30°C for 1 h in the dark. Values are means and standard errors from two separate experiments.

Comparison of the rtx regions of B. elkanii and B. japonicum. The 410-kb DNA region related to symbiosis in B. japonicum USDA110 was sequenced and analyzed by Göttfert et al. (15). Interestingly, they found rtxA-like genes in this region, even though B. japonicum could not produce rhizobitoxine (13, 27). Therefore, we compared the sequence of rtxA and its flanking regions of B. elkanii USDA94 with those of B. japonicum USDA110 (Fig. 6). The comparison of the 8,641-bp DNA sequence from rtxA to ORF4 shows that this region of the B. japonicum USAD110 gene is 79% homologous to that of the B. elkanii USDA94 gene. In contrast, the comparison of DNA upstream of rtxA and downstream of ORF4 in these two species of Bradyrhizobium revealed no noteworthy similarities. Furthermore, the ORFs corresponding to rtxA, rtxC, ORF2, ORF3, and ORF4 were also found in B. japonicum, and their order was well conserved between the two species. The DNA sequences of rtxC, ORF2, and ORF4 of B. japonicum showed 93, 87, and 89% similarity to those of B. elkanii, respectively. However, the ORFs corresponding to rtxA and ORF3 were fragmented into three ORFs in B. japonicum. In particular, the fragmentation of the C domain of the rtxA product appeared to abolish the ability to synthesize dihydrorhizobitoxine.

View larger version (21K): [in this window] [in a new window]   FIG. 6.   Comparison of the rtx gene regions of B. elkanii USDA94 and B. japonicum USDA110. The nucleotide sequence of the rtx gene region of B. japonicum USAD110 (8,641 bp) is 79% homologous to that of B. elkanii USDA94. The amino acid sequences of rtxC, ORF2, and ORF4 of B. japonicum USAD110 show 96, 87, and 89% identity to those of B. elkanii USDA94, respectively. The rtxA and ORF3 regions of B. japonicum USAD110 are interrupted and fragmented compared with those of B. elkanii USDA94. Numbers indicate lengths (in base pairs) of genes and intergenic regions.

	DISCUSSION

Top Abstract Introduction Materials and Methods Results Discussion References

Here we demonstrated that at least the rtxA and rtxC genes are responsible for rhizobitoxine biosynthesis in free-living B. elkanii, in light of results from mutagenesis experiments and the determination of concentrations of rhizobitoxine intermediates in culture by using LC/MS. The biosynthetic route to rhizobitoxine and the biological activities of the various genes are summarized in Fig. 7.

View larger version (78K): [in this window] [in a new window]   FIG. 7.   Proposed biosynthetic pathway of rhizobitoxine and relevant metabolism.

Whether dihydrorhizobitoxine is an end product (30) or intermediate (44) of rhizobitoxine biosynthesis has been a source of discussion. Our work indicates that the rtxC gene is responsible for dihydrorhizobitoxine desaturation at the final step of rhizobitoxine biosynthesis, and this gene product catalyzes the conversion of dihydrorhizobitoxine to rhizobitoxine by creating a double bond between C-3 and C-4. Therefore, we confirmed that dihydrorhizobitoxine is a key intermediate in rhizobitoxine biosynthesis (Fig. 7).

Ruan et al. (44, 45) isolated the rtxA (formally rtxA and rtxB) gene from USDA61 of B. elkanii and observed that rtxA mutants do not accumulate serinol in nodules and do not produce rhizobitoxine in culture or nodules. Those authors speculated that rtxA is involved in serinol formation and dihydrorhizobitoxine synthesis in light of DNA homology and results from other studies (44, 45). In our work, determination of serinol and dihydrorhizobitoxine concentrations in culture revealed that the N-terminal region of rtxA product is responsible for serinol formation and that the C-terminal portion is involved in dihydrorhizobitoxine biosynthesistwo crucial steps in rhizobitoxine biosynthesis. However, it is still unclear whether the predicted protein of 90 kDa mediated both activities simultaneously.

Serinol probably is a precursor of dihydrorhizobitoxine, because the one rtxA gene has two functions: serinol formation and dihydrorhizobitoxine biosynthesis. However, the substrates of the homoserine moiety in dihydrorhizobitoxine biosynthesis remain ambiguous. The addition of methionine and its intermediates (including O-acetylhomoserine, cysteine, cystathionine, and homocysteine) dramatically increased dihydrorhizobitoxine production in culture (Fig. 5). This suggests that the sulfur-containing intermediates are candidate precursors for dihydrorhizobitoxine as well as O-acetylhomoserine, although substrate specificity ultimately should be determined by using the purified enzyme derived from the rtxA gene. If so, it would be interesting if -cystathionase is subject to feedback inhibition by rhizobitoxine in B. elkanii (Fig. 7).

In B. japonicum, most genes concerned with nodulation and symbiotic nitrogen fixation are clustered within an approximately 410-kb region on the 8.7-Mb chromosome (15, 20). In Mesorhizobium loti strain ICMP3153, a symbiotic cluster, termed the symbiosis island, can be transferred to other strains when it is integrated into a phenylalanine-specific tRNA gene (54). Further, this island structure is well conserved on a similar chromosome of another M. loti strain, MAFF303099 (19). Therefore, it is generally accepted that symbiosis genes in rhizobia have evolved by horizontal gene transfer and genomic rearrangements thereafter.

The rtxA and rtxC genes of B. elkanii were located in a DNA region that includes nodulation and symbiotic nitrogen fixation genes (Fig. 1). Interestingly, the rtx cluster and noeE gene upstream of the cluster were almost completely conserved in B. japonicum USDA110 (Fig. 6) (15), which does not synthesize rhizobitoxine (24, 25, 27). A possible explanation is that the loss of the ability to synthesize rhizobitoxine is due to the fragmentation in B. japonicum USDA110 of the C-terminal region of the rtxA gene (Fig. 6). The transfer of cosmids containing B. elkanii rtxA and rtxC genes to B. japonicum would help answer the question of whether the loss of production of rhizobitoxine in B. japonicum was brought about only by the fragmentation of rtxA gene or by other mechanisms as well.

Experiments using hypernodulation legume mutants (40) and ethylene inhibitor applications (35) indicate that detection of ethylene by host legumes is involved in the control of nodulation. Ethylene has been reported to reduce nodulation of several legumes, with the exception of soybeans (Glycine max) (35). B. japonicum preferentially nodulates soybean cultivars in a multistrain environment (28). Therefore, perhaps after an ancestor of bradyrhizobia acquired rhizobitoxine biosynthetic genes as well as various symbiotic genes, B. japonicum lost the ability to synthesize rhizobitoxine in the absence of selection pressure because of the ethylene insensitivity of soybean nodulation. The partial collapse of the rtx region in B. japonicum USDA110 supports this idea (Fig. 6).

So far, the ability to synthesize rhizobitoxine is confined to the slow-growing B. elkanii (25, 27, 37) and Burkholderia andropogonis (29). The question arises whether fast-growing rhizobia other than Bradyrhizobium spp. produce another inhibitor for ethylene biosynthesis of host plants, because it could enhance nodulation. To test this possibility, we sought potential enzymes and compounds for reducing ethylene biosynthesis from the entire genome of the fast-growing M. loti represented in a database (http://www.kazusa.or.jp/en/) and identified the ACC deaminase gene as a candidate. The ACC deaminase gene is located within a 611-kb symbiosis island (downstream of the nifDK genes) on the 7.0-Mb chromosome (19). The plant growth-promoting Pseudomonas spp. possess ACC deaminase and reduce the amount of plant ethylene by degrading ACC into -ketobutylate and ammonia (14). It is, therefore, a fascinating hypothesis that rhizobia have two strategies for fulfilling nodulation enhancement by ethylene inhibition in host plants: rhizobitoxine biosynthesis in the slow-growing bradyrhizobia and ACC deaminase in fast-growing rhizobia.