A Novel Endo-Hydrogenase Activity Recycles Hydrogen Produced by Nitrogen Fixation

Background Nitrogen (N2) fixation also yields hydrogen (H2) at 1∶1 stoichiometric amounts. In aerobic diazotrophic (able to grow on N2 as sole N-source) bacteria, orthodox respiratory hupSL-encoded hydrogenase activity, associated with the cell membrane but facing the periplasm (exo-hydrogenase), has nevertheless been presumed responsible for recycling such endogenous hydrogen. Methods and Findings As shown here, for Azorhizobium caulinodans diazotrophic cultures open to the atmosphere, exo-hydrogenase activity is of no consequence to hydrogen recycling. In a bioinformatic analysis, a novel seven-gene A. caulinodans hyq cluster encoding an integral-membrane, group-4, Ni,Fe-hydrogenase with homology to respiratory complex I (NADH : quinone dehydrogenase) was identified. By analogy, Hyq hydrogenase is also integral to the cell membrane, but its active site faces the cytoplasm (endo-hydrogenase). An A. caulinodans in-frame hyq operon deletion mutant, constructed by “crossover PCR”, showed markedly decreased growth rates in diazotrophic cultures; normal growth was restored with added ammonium—as expected of an H2-recycling mutant phenotype. Using A. caulinodans hyq merodiploid strains expressing β-glucuronidase as promoter-reporter, the hyq operon proved strongly and specifically induced in diazotrophic culture; as well, hyq operon induction required the NIFA transcriptional activator. Therefore, the hyq operon is constituent of the nif regulon. Conclusions Representative of aerobic N2-fixing and H2-recycling α-proteobacteria, A. caulinodans possesses two respiratory Ni,Fe-hydrogenases: HupSL exo-hydrogenase activity drives exogenous H2 respiration, and Hyq endo-hydrogenase activity recycles endogenous H2, specifically that produced by N2 fixation. To benefit human civilization, H2 has generated considerable interest as potential renewable energy source as its makings are ubiquitous and its combustion yields no greenhouse gases. As such, the reversible, group-4 Ni,Fe-hydrogenases, such as the A. caulinodans Hyq endo-hydrogenase, offer promise as biocatalytic agents for H2 production and/or consumption.


Introduction
Azorhizobium caulinodans is an obligate oxidative, microaerophilic bacterium originally isolated from stem-and root-nodules of the legume host plant Sesbania rostrata [1]. In legume nodules, endosymbiotic rhizobia, including A. caulinodans, fix atmospheric dinitrogen (N 2 ) yielding ammonium as utilizable N-source for the host plant. Unlike typical rhizobia which fix N 2 only endosymbiotically, A. caulinodans is also diazotrophic (able to grow on N 2 as Nsource in pure culture). Both processes are owed to molybdenumcontaining (Mo) dinitrogenase, an a 2 b 2 -tetrameric protein complex catalyzing directed electron-transfer. Metabolic electrons are tapped from pyruvate oxidation [2] and singly transmitted via flavo-and FeS-proteins ultimately to the Mo-dinitrogenase catalytic center, its iron-molybdenum cofactor (FeMo-co); the enzyme complex effectively operates an 8-electron reductive cycle of recursive single electron transfers [3,4]. At the FeMo-co center, the first two arriving electrons combine with hydrogen-ions to yield a molecule of H 2 . The bound H 2 is then displaced by N 2 , and the subsequent six, arriving electrons, together with hydrogenions, now reduce N 2 to yield two molecules of ammonium as coproduct: In vivo, H 2 yields (relative to 1:1 in vitro stoichiometry) may further increase as a function of Mo-dinitrogenase turnover [5]. As the substrate N 2 triple-bond is highly unreactive, the dinitrogenase catalytic cycle is kinetically limiting as an in vivo biochemical standard process. To accelerate catalysis and render such thermodynamically favorable, Mo-dinitrogenase is both oneelectron reduced and energetically charged by homodimeric dinitrogenase reductase, which harbors a bridging 4Fe-4S-center and two ATP binding sites, one per subunit. During single-electron transfer from dinitrogenase reductase to Mo-dinitrogenase, 2 ATP hydrolyze to yield 2 ADP and 2 orthophosphate (Pi). Thus, in the 8-electron dinitrogenase complex catalytic cycle: earning Mo-dinitrogenase complex activity distinction as the most ATP-consumptive metabolic reaction on a per substrate basis [5].
Notably, A. caulinodans diazotrophic cultures, as with other aerobic diazotrophic bacteria, do not evolve significant H 2 . Rather, H 2 produced by Mo-dinitrogenase complex activity is efficiently recycled as respiratory electron donor to O 2 (as preferred electron-acceptor), thus recouping by oxidative phosphorylation ATP invested in H 2 production as part of the dinitrogenase catalytic cycle: ð ÞO 2 z 4 ð ÞADP,P i ?H 2 Oz 4 ð ÞATP which represents some 25% of total ATP invested in N 2 fixation [6]. H 2 production has long been associated with N 2 fixation in pure diazotrophic cultures of both fermentative and oxidative bacteria as well as by endosymbiotic rhizobia in legume nodules [7]. Endogenous H 2 recycling, both in diazotrophic bacterial cultures [8] as well as in certain symbiotic nodules, among those, garden pea [9], has been presumed owed to a respiratory Ni,Fehydrogenase activity highly conserved among disparate aerobic diazotrophic bacteria [10,11]. As studied in archetypal aerobic bacteria such as Ralstonia eutropha, orthodox respiratory hydrogenase is a heterodimeric protein comprising a bimetallic Ni,Fecatalytic subunit and a 4Fe-4S-center subunit, which complex with an integral-membrane b-type cytochrome, linking Ni,Fe-hydrogenase H 2 -oxidizing activity to cellular respiration and oxidative phosphorylation [12].
To the contrary, as we report here for A. caulinodans diazotrophic cultures open to the environment, endogenous H 2 is not recycled via orthodox respiratory exo-hydrogenase activity but instead via a novel respiratory endo-hydrogenase complex, presumably reflecting the need to sequester endogenous H 2 by metabolic channeling.

Results
A. caulinodans exo-hydrogenase deletion mutants lose chemoautotrophy but retain diazotrophy To study the metabolic role of the orthodox respiratory exohydrogenase activity for H 2 recycling in diazotrophic culture, A. caulinodans haploid strain 66081 carrying an in-frame hupDSL2 allele (Table 1), a result of perfect gene-replacement, was constructed by ''crossover PCR'' mutagenesis [13] (Fig. 1; Methods). To verify its hupSL deletion genotype, strain 66081 genomic DNA served as template for diagnostic PCR analysis. Using haploid genomic oligodeoxynucleotides HupSL-Prox and HupSL-Dist (Table 1; Fig. 1) as primer-pair, a single, novel 2.3 kbp DNA fragment was amplified from the 66081 genome, as template, and then sequenced on both strands. Strain 66081 therefore carries the in-frame hupDSL2 allele, arisen by perfect gene-replacement. When tested in chemoautotrophic liquid batch cultures (under 20% H 2 as sole energy source, 5% CO 2 as sole Csource, 2% O 2 , bal. N 2 ) with ammonium added as sole N-source, whereas parental strain 61305R (virtual wild-type; Methods) grew, strain 66081(hupDSL2) did not. Therefore, HupSL exo-hydrogenase activity is required for respiration with exogenous hydrogen. When tested in diazotrophic liquid batch cultures (Methods), strains 61305R and 66081(hupDSL2) both proved fully proficient (able to grow on N 2 as sole N-source; Nif + phenotype), in comparison to Nif 2 strains 60107R(nifA) and 60035R(nifD), both deficient [14]. Strain 66081(hupDSL2) also grew as wild-type when plated on solid, defined medium lacking added-N, thus requiring use of atmospheric N 2 (Methods). Therefore, orthodox respiratory HupSL exo-hydrogenase activity was not material to growth, nor, by presumption, endogenous H 2 recycling in A. caulinodans diazotrophic cultures.
Bioinformatic identification of a novel respiratory endohydrogenase gene-cluster A. caulinodans ORS571 genome fragments were then assembled and screened for additional hydrogenase genes. Previously identified, and localized to the same polycistronic operon carrying the hupSL genes, were the hupUV genes encoding a cytoplasmic sensory hydrogenase activity [15].  1). From these genome-contigs, we designed synthetic oligodeoxynucleotide primers, carried out PCR amplification and nucleotide sequencing, and assembled the complete sequence for a novel, tightly organized, seven-gene hyqRBCEFGI operon (GenBank accession: FJ378904; Fig. 2). In the presumed hyq operon, the distal hyqGI genes encode a canonical heterodimeric Ni,Fe-hydrogenase. The hyqBCEF genes all specify integral-membrane proteins orthologous to the Escherichia coli hyf genes, whose syntax in labeling the A. caulinodans hyq genes, including hyqGI, has thus been conserved. (Note the A. caulinodans hyq operon however lacks both E. coli hyfA and hyfD genes.) The E. coli hyf operon encodes hydrogenase-4 [18], an integral-membrane complex representative of the H 2 -evolving or group-4 hydrogenases, previously identified in and restricted to anaerobic bacteria [11]. and CRD were mixed, thermally denatured, allowed to partially renature, and used as combination PCR template/primer. As synthetic primers B and C share a complementary 21 bp extension sequence (angled line), the ARB Watson and CRD Crick strands (and vice versa) may partially reanneal via this 21 bp linker sequence In the third line, when such occurs, the resulting, partially-reannealed ARB(21 bp)CRD spliced DNA fragment which carries 59-overhangs and free 39-ends on both strands is a template for the thermostable DNA polymerase elongation reaction, producing a finished ARD duplex DNA fragment which may then be further amplified by PCR in the presence of added A and D primers. As verified by DNA sequencing analysis, finished, amplified ARD duplex fragments carry a genetic crossover which fuses (via the 21 bp linker sequence) in-frame the ''start'' codon of the proximal hupS gene to the ''stop'' codon of the distal hupL gene. Primers A and D may be extended with genome non-complementary elements to facilitate molecular cloning of resulting ARD fragments (Table 1). In vivo using homologous genetic recombination, wild-type loci are then exchanged for recombinant ARD crossover DNA fragments, which yield in-frame, translational deletion alleles of target genes of interest (Methods). doi:10.1371/journal.pone.0004695.g001 As well, the group-4 integral-membrane hydrogenases include multiple subunits homologous to those of respiratory NADH : quinone dehydrogenase (NADH-DH), an integral membrane complex whose active site for NADH oxidation faces the cytoplasm [11,18,19]. Such homology extends to both NADH-DH L 0 (integral-membrane) and L 1 (membrane-associated) subcomplex proteins. By analogy to NADH-DH, and to conceptually distinguish the two A. caulinodans cell membrane-associated, respiratory hydrogenases, we therefore denote the presumed HyqBCEFGI complex as endo-hydrogenase and the HupSL+cytb complex as exo-hydrogenase, so as to distinguish relative orientations of substrate oxidative sites: the former presumably facing the cell interior (cytoplasm), the latter facing the cell exterior (periplasm). As the A. caulinodans genome encodes bona fide respiratory complex I in an unlinked 15-gene operon (AZC_1667 to AZC_1681) [20], the hyq genes and their encoded endo-hydrogenase complex, while similar, are indeed structurally and functionally distinct.

A. caulinodans Hyq endo-hydrogenase activity facilitates growth in diazotrophic cultures open to the environment
To test metabolic role(s) for the presumptive Hyq endohydrogenase complex, an A. caulinodans hyq operon deletion (hyqDRI7) allele was constructed using crossover PCR, in which the hyqR start-and hyqI stop-codons were fused in-frame by a 21 bp linker sequence ( Fig. 2; Methods). In both strain 61305R(virtual wild-type) and 66081(hupDSL2), the wild-type hyq + operon was then swapped for this hyqDRI7 allele by homologous recombination (Methods). As with strain 66081, the resulting haploid strain 66132 was verified by combined PCR and DNA sequencing analyses. Using haploid genomic Hyq-Prox and Hyq-Dist oligodeoxynucleotides (Table 1; Fig. 2) as primer-pair and strain 66132 genomic DNA as template, a single 2.2 kbp DNA fragment was amplified by PCR and then sequenced on both strands using either Hyq-Prox or Hyq-Dist as DNA sequencing primers. Accordingly, strain 66132 proved a true hyqDRI7 haploid arisen by perfect gene-replacement. Similarly, starting with strain 66081(hupDSL2), strain 66203 proved a true double-mutant haploid, carrying both hupDSL2 and hyqDRI7 alleles.
Growth kinetics of four haploid strains, 61305R and its descendants 66081(hupDSL2), 66132(hyqDRI7) and 66203 (hupDSL2, hyqDRI7) were analyzed in liquid batch cultures with defined media. For all strains, batch cultures were started in defined medium with 40 mM succinate added as sole C-and energy-source, and 0.5 mM ammonium added as N-source. All starter cultures grew exponentially up to viable cell counts of ,1610 8 ml 21 at which growth arrested due to ammonium limitation (as when 5 mM ammonium was then added, exponential growth rapidly resumed for at least two additional cell doublings). These N-limited, static cultures were one-thousandfold diluted into the same defined growth medium with or without added 5 mM ammonium, placed in sealed 30 ml vials, sealed and subcultured with continuous sparging (6 ml min 21 ) using a defined gas mixture (2% O 2 , 5% CO 2 , bal. N 2 ). Samples were periodically withdrawn and plated on rich medium for viable cell counts (Methods). Whether in the presence and absence of added ammonium, strains 61305R and 66081(hupDSL2) grew similarly ( Table 2). In contrast, while both strains 66132(hyqDRI7) and 66203(hupDSL2, hyqDRI7) grew similarly in the presence of added ammonium, cell doubling-times slowed 21% in the absence of ammonium, relative to strains 61305R and 66081 (Table 2).
These growth experiments were repeated, except that cultures were sparged with an H 2 -enriched defined gas mixture (2% O 2 , 5% CO 2 , 20% H 2 , bal. N 2 ). With the inclusion of H 2 at saturating metabolic levels in the sparge gas, strains 61305R and 66132(hyq-DRI7) when cultured without added ammonium both grew much faster. In one experiment, the diazotrophic cell doubling-time for strain 61305R was cut from 7.2 to 4.16 hr, that for 66132 was cut from 8.8 to 5.0 hr, both in the presence of 20% H 2 (Table 2). Therefore, while A. caulinodans growth rates in liquid diazotrophic cultures were otherwise limited by N 2 fixation ( Table 2), acceleration of oxidative phosphorylation with added 20% H 2 as respiratory electron-donor also accelerated growth, as is generally characteristic of microaerophilic bacteria [21]. Nevertheless, even with exogenous H 2 added in sparge gases at levels sufficient to yield maximum growth-rate enhancement, the 21% growth deficit observed for 66132(hyqDRI7) versus parental 61305R persisted. Therefore, cell bioenergetic role(s) for Hyq endo-hydrogenase and HupSL exo-hydrogenase activities are not entirely synonymous. By analogy to NADH-DH complex activities, Hyq endo-hydrogenase activity might also be membrane proton-motive and/or electrogenic, translocating multiple ions such a H + , K + and/or Na + (see Discussion).
In contrast, strain 66081(hupDSL2) showed only a slight increase in growth rate in diazotrophic culture, and strain 66203(hupDSL2, hyqDRI7) showed no detectable increase in growth rate, both in response to added 20% H 2 ( The A. caulinodans hyq operon is strongly and specifically induced in diazotrophic cultures To assess growth conditions in which the hyq operon was genetically expressed, A. caulinodans strains using b-glucuronidase activity to report hyq operon transcription were constructed. The E. coli uidA + gene, encoding b-glucuronidase, was amplified by PCR and, using standard in vitro molecular cloning techniques, the resulting 1.8 kbp uidA + coding sequence was inserted in-frame into the 21 bp crossover linker sequence of pHyqDRI7 yielding plasmid pHyqRU5 (Table 1; Methods). Derived from both A. caulinodans 57100 and 60107R(nifA) as parent, hyq merodiploid strains 66205 and 66210, both carrying an upstream in-frame fusion hyqR::uidA + hyqDBI operon in tandem with the downstream hyq + operon, were isolated and verified by both PCR and DNA sequencing analyses (Methods). Merodiploid hyq reporter strain 66205 was first tested for bacterial colony appearance on solid defined media supplemented with X-Gluc (Methods) as chromogenic b-glucuronidase substrate. When inoculated onto defined medium also supplemented with 5 mM ammonium and cultured in fully aerobic conditions, strain 66205 colonies were white, lacking any evidence of b-glucuronidase activity. When the same petri plates were incubated under a reduced O 2 atmosphere (2% O 2 , 5% CO 2 , bal. N 2 ), strain 66205 colonies appeared light-blue, or partially induced. When 0.5 mM L-glutamine was added to solid culture medium, 66205 colonies were again completely white when incubated under 2% O 2 indicating the hyq operon was strongly repressed. When 66205 was cultured diazotrophically (in the absence of added ammonium and L-glutamine) under reduced O 2 , colonies were dark blue, indicative of strong hyq operon expression.
To obtain quantitative data for both strains 66205 and 66210, liquid batch cultures were pre-grown aerobically in defined medium with 0.5 mM ammonium as N-source to cell titers of ,1610 8 ml 21 (at which available ammonium was exhausted) and physiologically shifted to diazotrophic culture conditions (Methods) for 12 hr at 29uC. Cells were then harvested and bglucuronidase specific activities were measured in cell-free extracts (Methods). These results (Table 3) corroborated visual inspections of bacterial plate cultures supplemented with chromogenic X-Gluc. The hyq operon was specifically and strongly expressed in diazotrophic culture but strongly repressed either in the presence of added ammonium under air or in the presence of added 0.5 mM L-glutamine under reduced O 2 . As strain 66210 was only weakly induced (Table 3), hyq operon induction specifically required NIFA transcriptional activation.
When the presumed hyq operon control region (immediately upstream from the hyqR coding sequence) was analyzed, the  genetic signatures of an orthodox A. caulinodans nif operon were apparent (Fig. 2). A NifAbox element, serving as cis-acting site for NIFA transactivation, and a s54 N site, serving as cis-acting site for RNA polymerase complexed with s54 N initiation factor, were both present and strategically positioned [22]. Thus, the A. caulinodans hyq operon control region likely binds NIFA, which activates hyq transcription via RNA polymerase ? s54 N in response to both limiting physiological O 2 and absence of fixed-N, as is observed for nifA autoregulation [22]. Accordingly, the A. caulinodans hyq operon is constituent of the nif regulon.
Archetypal for the group-4 hydrogenases is E. coli hydrogenase-3, encoded by hycGE. This heterodimeric Ni,Fe-hydrogenase anchors an integral-membrane formate-hydrogen lyase complex, oxidizing formate to CO 2 and reducing 2H + to H 2 , all cellinternal, under strict, fermentative conditions [19]. In E. coli, a second group-4 hydrogenase (hydrogenase-4), encoded by the hyf operon, seems coupled to yet another fermentative formate dehydrogenase activity [18]. In Rhodospirillum rubrum a distinct, but related, group-4 hydrogenase activity is coupled to COdehydrogenase activity [23,24]. In all cases these group-4 hydrogenases are active under anaerobic, strictly fermentative physiological conditions and so have been termed H 2 -evolving, simultaneously oxidizing either formate or CO to yield CO 2 and reducing H + -ions to yield H 2 (gas), all as fermentative endproducts.
From multiple protein sequence alignments, the A. caulinodans HyqBCEFGI hydrogenase is a constituent member of the group-4 hydrogenases. However, as A. caulinodans, like all rhizobia, is an obligate oxidative (aerobic) bacterium and does not ferment, any metabolic role for H 2 evolution is not obvious. Indeed, we have identified unlinked A. caulinodans genes encoding both formate-and CO-dehydrogenase activities; the former are orthologous to the aerobic, respiratory formate dehydrogenase of facultative bacteria such as E. coli; the CO-dehydrogenase genes are orthologous to the soluble, NAD-linked activities of obligate aerobic bacteria (data not presented).
From bacterial genome searches, orthologous hyq operons are evident in two additional rhizobial species, R. leguminosarum and B. japonicum both previously classified phenotypically as H 2 recyclers [9,25]. In the non-symbiotic but very closely related species Xanthobacter autotrophicus Py2 [26], an orthologous hyq operon is also present, as is the complete nif regulon, implying X. autotrophicus Py2 is also diazotrophic. All such bacteria carrying the hyq operon are obligate oxidative, in which any H 2 -evolving hydrogenase activities would seem not only superfluous but antithetical.
Metabolic roles for the group-4 hydrogenases are not definitive. All show integral-membrane components with homology to NADH : quinone dehydrogenase (respiratory complex I), which functions as unidirectional NADH oxidant and membrane quinone pool reductant [11,18,19]. Included in this homology are the heterodimeric Ni,Fe-hydrogenase subunits of group-4 hydrogenases (the A. caulinodans HyqGI proteins) which are closely related to the 49 kDa (Nqo4) and 20 kDa (Nqo6) subunits of the Thermus thermophilus (hyperthermophile) respiratory NADH-DH L 1 sub-complex, whose crystal structure has been solved by X-ray diffraction at atomic resolution [27]. By structural analogy and genetic homology to the NADH-DH L 1 sub-complex then, the homologous HyqGI heterodimeric endo-hydrogenase, with its active site facing cell-internally, likely interacts with the integralmembrane, L 0 -homologous HyqBCEF sub-complex and together function as H 2 oxidant and membrane quinone reductant. Like both NADH-DH complex and E. coli hydrogenase-4, A. caulinodans Hyq endo-hydrogenase activity is presumably proton-motive, energy-conserving [18] and thus likely drives aerobic respiration. From multiple protein alignments including sequences identified in the four aerobic bacteria (A. caulinodans, B. japonicum, R. leguminosarum, X. autotrophicus), together with the E. coli HyfGI proteins, the endo-hydrogenase peripheral HyqG large-subunit carries the Ni,Fe-hydrogenase catalytic center and the HyqI smallsubunit carries the (N2) proximal 4Fe-4S center as likely electrondonors to membrane-bound quinones.
Given this inferred organization and integral-membrane orientation of the Hyq endo-hydrogenase complex (for which we as yet lack direct experimental evidence), one implication is obvious: the Hyq endo-hydrogenase might physically interact with Mo-dinitrogenase so as to channel evolved H 2 as substrate for membrane-driven respiration and oxidative phosphorylation. Coupled respiration would allow quantitative recovery of ATP consumed by Mo-dinitrogenase complex activity in H 2 synthesis (and activation of N 2 reduction to ammonium) [3,4]. The group-4 hydrogenases of anaerobes, while quite possibly active in vivo in H 2 evolution during strictly fermentative metabolism, nevertheless remain capable of H 2 oxidation, albeit slowly [11]. Because Modinitrogenase complex activity has exceedingly slow in vivo turnover (,10 sec 21 ), any directly coupled Hyq endo-hydrogenase H 2 oxidizing activity might operate at correspondingly very slow rates in vivo.
H 2 -oxidative endo-hydrogenase activity necessitates that H + ions be membrane-translocated else deplete the cell membrane protonmotive force. Any endo-hydrogenase driven, vector H + translocation, an energy-requiring process, would be necessarily slow by comparison with exo-hydrogenase activity, uncoupled from H + translocation, and thus relatively fast. (In the latter case, as H + ions are produced external to the cell membrane, they in principle contribute directly to the cell membrane proton-motive force.) Thus, exo-hydrogenase activity is kinetically preferred as agent for exogenous H 2 oxidation. By contrast, endo-hydrogenase activity, via metabolic channeling, might confer an increased efficiency of endogenous H 2 recycling, thus mitigating energy loss, were such H 2 to escape to the environment by simple diffusion.
Hydrogen has elicited considerable interest as potential renewable energy source for human civilization. If hydrogen is to be produced at scale as part of a sustainable cycle, external energy source(s) are then required. Solar energy represents an obvious energy coupling source, in principle allowing photoelectron transport and H 2 -evolving hydrogenase activities to operate as an integrated biocatalytic process in photosynthetic membranes. Accordingly, the reversible group-4 hydrogenases, such as the A. caulinodans Hyq endo-hydrogenase, offer particular promise as biocatalytic agents for hydrogen production and/or consumption.

Bacterial strains and culture conditions
Azorhizobium caulinodans ORS571 wild-type (strain 57100), originally isolated from Sesbania rostrata stem-nodules [1], was cultured in both rich (SYPC) and miminal, defined media as previously described [14]. As 57100 wild-type is NAD auxotrophic, defined growth media must be supplemented with nicotinate (or similar) as precursor. However, nicotinate serves strain 57100 as both anabolic (for NAD production) and catabolic (as both utilizable C-and N-source) supplement. When strain 57100 is cultured in media with limiting primary C-and/or N-sources, nicotinate is rapidly catabolized and exhausted cultures quickly become NAD-limited for growth [28]. Accordingly, to eliminate nicotinate catabolism as a metabolic variable, all experiments reported herein employ A. caulinodans 61305R, a 57100 derivative carrying an IS50R insertion in the (catabolic) nicotinate hydroxylase structural gene, as ''virtual'' wild-type; 61305R only uses nicotinate as anabolic substrate and thus requires minimal (1 mM) nicotinate supplementation in all defined media [29].

Genetic constructions
A.
caulinodans in-frame translational fusion mutants. Precise, in-frame deletion mutagenesis of the A. caulinodans hupSL genes was carried out by ''crossover PCR'' as previously described [13]. In the first step, separate ,1 kbp genomic fragments immediately proximal to hupS and distal to hupL coding sequences were PCR amplified [13]. These two, ,1 kbp amplified genomic fragments shared an artificial, complementary ''crossover'' sequence introduced by 21 bp extension of PCR primers ''B'' and ''C'' (Fig. 1). In a secondround PCR, the two amplified DNA fragments were purified, mixed, and used as combination primer-template. A ,2 kbp DNA fragment was then produced when non-homologous template strands annealed via complementary 21 bp extensions; when an upstream coding-strand annealed to a downstream non-coding strand via the 21 bp crossover extension, 39-ends on both annealed strands were extended by thermostable DNA polymerase yielding a contiguous ,2 kbp DNA fragment in which the 21 bp crossover sequence fused the ,1 kbp upstream and downstream sequences. In this second-round PCR, terminal ''A'' and ''D'' oligodeoxynucleotides ( Fig. 1) were also included as primers such that, by standard recursive PCR, this ,2 kbp crossover DNA fragment was further amplified. By design, the 21 bp crossover within the ,2 kbp DNA fragment fuses in-frame an upstream target gene's translational ''start'' codon with a downstream target gene's ''stop'' codon yielding a translational (e.g., hupDSL) fusion (Fig. 1).
The PCR amplified, crossover DNA fragment carrying the inframe ,2 kbp hupDSL2 fusion allele was verified by DNA sequencing and introduced into the EcoRI site of plasmid pSUP202 (Table 1) by standard molecular cloning; E. coli strain MH3000 (Table 1) was subject to electroporation with recombinant plasmids, and transformed bacterial colonies were selected for tetracycline (Tc) resistance. In this manner, recombinant plasmid pHupDSL2 was identified (Table 1), purified, and reintroduced by electroporation into E. coli SM10 (Table 1), proficient as donor for bacterial conjugation, again selecting for Tc-resistance. To allow plasmid conjugal transfer, E. coli SM10/ pHupDSL2 as donor was mixed with A. caulinodans 61305R as recipient and plated overnight on SYPC solid medium at 37uC. Conjugal cell mixtures were then selectively plated on solid ORSMM (to counter-select E. coli) supplemented with Tc (10 mg/ ml) at 37uC. As parental plasmid pSUP202 cannot stably replicate in A. caulinodans, transconjugants that are stably Tc-resistant arise after homologous, single recombination events in which the entire plasmid and the target genome are cointegrated [22]. Accordingly, A. caulinodans 61305R hupSL merodiploids were then isolated and confirmed by PCR and DNA sequencing analyses; such strains carried both genomic hupS + L + and hupDSL2 alleles bridged by the integrated SUP202 sequence. To then isolate haploid genereplacement strains, merodiploids were subcultured absent Tc selection in rich GYPC liquid medium and then plated with Tc added at very low (0.125 mg/ml) levels sufficient to 50% inhibit growth of parental wild-type. Pinpoint colonies were identified, retested, and a Tc-sensitive phenotype verified. These putative haploid derivatives arose by a second, single homologous recombination (disintegration) event within the merodiploid, segregating the hupSL alleles. By PCR analysis with Hup-Prox and Hup-Dist as oligodeoxynucleotide primer-pair (Table 1; Fig. 1), resulting haploid strains showed either hupS + L + or hupDSL2 alleles.
Similarly, a haploid 61305R derivative carrying a complete hyqRBCEFGI in-frame deletion allele was isolated using the same crossover PCR technique. Recombinant plasmid pHyqDRI7 carried a ,2 kbp hyqDRI7 allele in which the identical 21 bp linker fused in-frame the hyqR ''start'' codon with the hyqI ''stop'' codon (Table 1; Fig. 2). After gene replacement, haploid strain 66132 carrying the hyqDRI7 allele was isolated and verified by PCR and DNA sequencing analyses.
A. caulinodans Hyq transcriptional reporter strains. To construct hyq merodiploid transcriptional reporter strains, a 1.8 kbp fragment carrying the E. coli uidA + coding sequence was amplified by PCR using synthetic oligodeoxynucleotide primers extended with 6 bp SalI endonuclease recognition sequences ( Table 1). As the 21 bp linker sequence used to construct in-frame translational fusions includes a SalI recognition sequence, plasmid pHyqDRI7 was partially digested with SalI endonuclease, a 9.8 kbp DNA fragment was isolated, mixed with SalI digested, amplified 1.8 kbp uidA + DNA fragment and recombinant plasmids were recovered by standard molecular cloning techniques. After electroporation of E. coli MH3000, and selection for Tc-resistance, uidA + recombinant plasmids were identified by plating candidate strains on minimal media supplemented with (0.1 mg ml 21 ) 5bromo-4-chloro-3-indolyl-b-D-glucuronide (XGluc), a chromogenic b-glucuronidase substrate, and screening for blue colonies. From PCR and DNA sequencing analysis, recombinant plasmid pHyqRU5 carrying the hyqR::uidA + in-frame translational fusion allele was isolated (Table 1). Plasmid pHyqRU5 was introduced to E. coli SM10, and this strain was employed as conjugal donor with A. caulinodans 61305R and 60107R, and Tc-resistant derivatives were selected and isolated. Merodiploid strains 66205 and 66210 (Table 1) carrying both upstream hyqR::uidA + hyqDBI and downstream hyq + operon were identified and verified by PCR and DNA sequencing analyses.
Physiological growth and b-glucuronidase activity measurements. Starter cultures of A. caulinodans strain 61305R and its derivatives were aerobically cultured in minimal defined NIF liquid medium [14] supplemented with: ammonium (0.5 mM) as sole, limiting N-source and 1 uM nicotinate at 37uC until growth arrested (cell densities ,1610 8 cells ml 21 ). For kinetic measurements of diazotrophy, arrested starter cultures were diluted one-thousandfold in NIF medium (with 1 uM added nicotinate) into 30 ml serum vials, sealed with silicone rubber septa, sparged continuously (6 ml min 21 ) with defined gas mixtures (e.g. 2% O 2 , 5% CO 2 , bal. N 2 ), and incubated at 29uC. At least three times per cell-doubling period, culture samples were removed, serially diluted, plated on rich GYPC medium [14], incubated 48 hr at 37uC, and colonies were counted, in triplicate. b-glucuronidase activity was measured with as chromogenic substrate 5-bromo-4-chloro-3-indolyl-b-D-glucuronide (X-Gluc) [30]; total protein concentrations were determined in a folin phenol reagent assay [31]. All induction experiments were conducted in triplicate and were repeated until the standard error in b-GUS activities was below 15%.