Identification of the Hydrogen Uptake Gene Cluster for Chemolithoautotrophic Growth and Symbiosis Hydrogen Uptake in Bradyrhizobium Diazoefficiens

The hydrogen uptake (Hup) system of Bradyrhizobium diazoefficiens recycles the H2 released by nitrogenase in soybean nodule symbiosis, and is responsible for H2-dependent chemolithoautotrophic growth. The strain USDA110 has two hup gene clusters located outside (locus I) and inside (locus II) a symbiosis island. Bacterial growth under H2-dependent chemolithoautotrophic conditions was markedly weaker and H2 production by soybean nodules was markedly stronger for the mutant of hup locus I (ΔhupS1L1) than for the mutant of hup locus II (ΔhupS2L2). These results indicate that locus I is primarily responsible for Hup activity.

Soybean bradyrhizobia have two lifestyles: as symbiotic bacteroids that fix atmospheric nitrogen in host plants or as free-living soil bacteria. As a symbiont, B. diazoefficiens synthesizes a hydrogen uptake (Hup) system that recycles the H 2 formed as a byproduct of nitrogenase activity (4). This symbiotic hydrogen oxidation increases nitrogen fixation efficiency, thereby enhancing the productivity of the legume host (3,6). As free-living cells, Bradyrhizobium diazoefficiens Hup + strains have the ability to grow chemolithoautotrophically by using H 2 as an electron donor (7).
Two sets of hup genes have been identified in B. diazoefficiens: a large cluster outside the symbiosis island (hup locus I, genome position 7,620,025-7,645,755) and a small cluster inside the symbiosis island (hup locus II, genome position 1,888,916-1,902,575) (6,11). A transcriptome analysis previously showed that several hup genes located outside the symbiosis island were up-regulated during H 2 -dependent chemolithoautotrophic growth, whereas several hup genes located inside the symbiosis island were up-regulated in symbiotic bacteroids (1,6). These findings imply that hup locus I plays an important role in chemolithoautotrophic growth, while symbiotic hup locus II may contribute to Hup activity in the nodules. In the present study, we constructed hupSL deletion mutants in order to clarify the contribution of each hup gene cluster during the chemolithoautotrophic growth and nodulation of B. diazoefficiens USDA110.

Materials and Methods
The strains and plasmids used in this study are listed in Table 1. The HM salt medium for the preculture and Hup medium for chemolithoautotrophic growth were described previously (14). Antibiotics were added to the media for B. diazoefficiens USDA110 and Escherichia coli strains as described previously (13).
We generated hupS 1 L 1 and hupS 2 L 2 deletion mutants as follows. A DNA fragment containing hupFDCL 1 S 1 V was isolated from brp15657, a plasmid from the pUC18 clone library of B. diazoefficiens USDA110 sequences (11), and inserted into the HindIII site of pK18mob. The resultant plasmid, pK18mob-hupI, was digested with ApaI and ligated with the ApaI/EcoRI adaptor, yielding pK18mob-hupIad. The omega cassette isolated from pHP45Ω was inserted at the EcoRI site of pK18mob-hupIad, yielding pK18mob-hupIomega. Triparental mating using pRK2013 was performed as described previously (13). A similar strategy was used to construct the hupS 2 L 2 deletion mutant. Briefly, hupKHFDCL 2 S 2 was isolated from brp07423, and inserted into pK18mob, yielding pK18mob-hupII. pK18mob-hupII was ligated with the EcoO109I/BamHI adaptor, resulting in pK18mob-hupIIad. The Tc r cassette was isolated from p34S-Tc and inserted into the BamHI site of pK18mob-hupIIad, yielding pK18mob-hupIITc. The double crossover events of these deletion mutations were verified by a Southern blot analysis.
Soybean seedlings (Glycine max cv. Enrei) grown in a plant box in a growth chamber were inoculated with B. diazoefficiens as described previously (8,9). The nodulated roots were transferred into a 300-mL bottle 30 d later, and a 0.5-mL gas sample from the head space of the bottle was injected into a GC-2014 gas chromatograph (Shimadzu, Kyoto, Japan) as described previously (13). The flow rate of the carrier gas (N 2 ) was 30 mL min -1 .

Results
Wild-type USDA110 and the ΔhupS 2 L 2 mutant grew on Hup agar medium under chemolithoautotrophic growth conditions ( Fig. 2A). However, the ΔhupS 1 L 1 mutant showed markedly weaker growth than that of the wild type on this medium ( Fig.  2A). The height of plants inoculated with ΔhupS 1 L 1 appeared to be lower than those inoculated with the wild type and ΔhupS 2 L 2 mutant (Fig. 2B); however, no significant differences were observed in plant dry weights or fresh nodule weights ( Table 2). H 2 was not produced from the nodulating roots of the wild type or ΔhupS 2 L 2 mutant (Fig. 2C), indicating that Hup activity compensated for the production of H 2 via nitrogenase. In contrast, H 2 was produced by the ΔhupS 1 L 1 nodules (6.4 μmol h -1 g fresh nodule weight -1 ) (Fig. 2C), indicating that Hup activity was lower than the production of H 2 via nitrogenase. These results suggest that hup genes outside the symbiosis island are the primary cluster involved in chemolithoautotrophic growth and Hup activity in the nodules.

Discussion
In the present study, the mutation of hupS 2 L 2 did not change nodule H 2 production from that by wild-type USDA110 (Fig. 2C), even though some genes on hup locus II were up-regulated in symbiotic bacteroids (1,6). Hup locus I contains a complete set of the hup-hyp-hox cluster, and is missing from hup locus II (11). Thus, hupS 2 L 2 in locus II may not be fully induced without the hup gene assemblage, resulting in no or weak Hup activity by the hupS 2 L 2 genes. On the other hand, the hup gene cluster outside the symbiosis island, which we identified as the primary hup gene cluster contributing to Hup activity in free-living and symbiotic cells, is located on a typical genomic island (trnM element) of B. diazoefficiens USDA110. The trnM element is likely acquired in the USDA110 lineage after the divergence of strains USDA110 and USDA6 T because B. japonicum USDA6 T completely lacks this element (10,11,12).
Hup genes were also found in the symbiosis island of the USDA6 T genome even though USDA6 T was previously reported to exhibit no Hup activity (10,11,12). The hup genes in USDA6 T on the symbiosis island had 99-100% amino acid sequence identity to the corresponding genes in hup locus II of USDA110. In contrast, USDA6 T hup genes had only 43-83% amino acid sequence identity to genes in hup locus I of USDA110, which is similar to the homology (45-83%) between hup genes in loci I and II of USDA110. These results suggest that hup genes on locus II of USDA110 and hup genes in USDA6 T do not contribute to the Hup activities of these strains and appear to be derived from the acquisition of symbiosis islands. Therefore, our results imply the horizontal gene transfer of the primary hup cluster via the genomic