Redirection of Metabolism for Biological Hydrogen Production

Bacterial strains and growth conditions.The bacterial strains and plasmids used are listed in Table 1. All work was carried out with the R. palustris wild-type strain CGA009 or its derivatives. Strain CGA009 is defective in uptake hydrogenase activity (27). R. palustris was grown and manipulated aerobically in defined PM mineral medium (14) containing 10 mM succinate as a carbon and energy source and antibiotic additions as indicated. Otherwise, R. palustris strains were grown in defined medium (10 ml) anaerobically in light in sealed tubes (27 ml) with argon gas in the headspace (17 ml) and (NH₄)₂SO₄ supplied as a nitrogen source (non-nitrogen-fixing conditions) or with nitrogen gas in the headspace with (NH₄)₂SO₄ omitted (nitrogen-fixing conditions). Carbon sources were supplied as indicated in the text. Escherichia coli strains DH5α and S17-1 were grown at 37°C in Luria-Bertani (LB) medium. Where indicated, R. palustris was grown with 100 μg of gentamicin (Gm) per ml. E. coli was grown with 100 μg of ampicillin per ml or 20 μg of Gm per ml.

Strain constructions.DNA fragments (∼3 kb) spanning nifA plus flanking regions from R. palustris strains CGA571 and CGA574 were generated by PCR. The amplification products contained engineered XbaI cloning sites at both ends. These products were digested with XbaI and cloned into XbaI-digested pUC19 to generate pUC19-nifA*571 and pUC19-nifA*574. The XbaI fragments were then cloned into pJQ200KS to generate pJQ200KS-nifA*571 and pJQ200KS-nifA*574. This construct was mobilized from E. coli S17-1 into R. palustris CGA757 (ΔnifA; this strain cannot grow under nitrogen-fixing conditions) by conjugation. Colonies that contained plasmids that had undergone a single recombination to become inserted into the chromosome were identified by growth on PM mineral medium plus Gm. These colonies were spread on plates supplemented with 10% sucrose to counterselect against the sacB marker and incubated anaerobically under nitrogen-fixing conditions in order to select for strains that had undergone a double recombination to lose the sacB-containing vector. Gene replacements were confirmed by PCR and sequencing.

Description of the R. palustris GeneChip.A R. palustris custom-designed GeneChip was manufactured by Affymetrix according to the following specifications: 99.8% of R. palustris predicted open reading frame genes were represented by unique probe sets (16 probe pairs and their corresponding mismatch for each). In addition, the GeneChip contains information for all intergenic regions larger than 150 bp.

DNA microarray experiments.RNA was isolated as previously described (24). For cDNA synthesis, we used 10 μg of purified RNA, semirandom hexamer primers with an average G+C content of 75%, and Superscript II reverse transcriptase (Life Technologies). Fragmentation and labeling were performed according to the manufacturer's recommendations (Affymetrix). Samples were hybridized and scanned at the Center for Expression Arrays, University of Washington, according to the specifications provided by the manufacturer. The Affymetrix Microarray Suite was used for initial data acquisition and processing. Transcript data were further analyzed using Cyber-T (http://visitor.ics.uci.edu/genex/cybert) as previously described (30) with a P value threshold of 0.001, resulting in a corresponding posterior probability of differential expression no lower than 0.9975.

Nitrogenase activity and hydrogen measurements.Nitrogenase activity was measured by the acetylene reduction assay as described previously (24). Hydrogen was measured using a Hewlett Packard 5890 series II gas chromatograph equipped with a thermal conductivity detector and molecular sieve 13X column (80/100 mesh; inner diameter,1/4 in. by 8 ft) (24). Protein concentrations were determined using the Bio-Rad (Richmond, CA) protein assay kit.

Gene sequencing.In order to identify the mutation(s) responsible for the hydrogen-producing phenotype, we sequenced nifA and its promoter region, glnK1, draT1, draT2, and draG, in all strains. In addition, we sequenced regS, regR, cbbR, glnB, and glnK2 in strain CGA570 and ntrB, ntrC, glnK2, vnfA, and anfA in strain CGA572.

Microarray accession number.The transcriptome data have been deposited at http://www.ncbi.nlm.nih.gov/geo under accession number GSE5194.

Isolation of mutant strains of R. palustris that produce hydrogen in the presence of ammonium.When grown in light with organic compounds present, anoxygenic photosynthetic bacteria, the metabolic group to which R. palustris belongs, generate ATP by cyclic photophosphorylation and use carbon compounds to make cell biomass (photoheterotrophic growth). Carbon substrates that are electron rich relative to cell material cannot be assimilated unless an external electron sink, such as carbon dioxide, nitrate, or dimethyl sulfoxide, is available to dissipate excess reducing equivalents (22). Another process to dissipate excess reducing power is nitrogen fixation (12). Consistent with this, we noticed that R. palustris grew on the electron-rich carbon compound cyclohexanecarboxylate if it was able to generate hydrogen as part of the process of nitrogen fixation (Table 2 and Fig. 1). However, wild-type R. palustris failed to grow when incubated with cyclohexanecarboxylate or other reduced carbon compounds under non-nitrogen-fixing conditions, presumably because ammonium sulfate present in the growth medium repressed the synthesis of nitrogenase and therefore hydrogen production (Table 2). This observation suggested that we should be able to obtain mutant strains that produce hydrogen constitutively by selecting for growth on cyclohexanecarboxylate in the presence of ammonium. We inoculated R. palustris wild-type cells into anaerobic culture tubes containing mineral medium with cyclohexanecarboxylate as a source of carbon, ammonium sulfate as the source of nitrogen, and light as the source of energy. After a long incubation period of several months, growth was observed in four culture tubes, and four mutant strains were purified from these tubes and retested for the ability to grow with cyclohexanecarboxylate in the presence of ammonium (Table 3). Each of the strains produced hydrogen from abundant carbon compounds, including acetate and p-coumarate, a plant lignin monomer (Table 4), under conditions in which wild-type cells do not produce hydrogen (Table 2). Each of the mutants also expressed nitrogenase constitutively (Table 5). Because the mutants were supplied with argon rather than nitrogen gas in the headspace of culture tubes during growth, we conclude that nitrogenase in these cells catalyzed the synthesis of hydrogen without accompanying ammonia production. Many reports have shown that hydrogen is the only product formed by nitrogenase in the absence of nitrogen gas (1, 2, 9). Apparently, these mutant strains had overcome the regulatory barriers that prevent nitrogenase gene expression.

• Open in new tab
• Download powerpoint

FIG. 1.

Metabolic route leading to hydrogen production by R. palustris strain CGA009. Carbon sources that are electron rich (highly reduced) relative to cell material can be degraded and used to support cell growth only if cells are able to produce hydrogen via nitrogenase. The theoretical stoichiometries for nitrogenase reactions for hydrogen production in the presence and absence of nitrogen gas are indicated.

Nitrogenase genes and a small number of other genes are highly expressed in the hydrogen-producing strains.We analyzed the transcriptomes of each of the four hydrogen-producing strains grown with acetate in light in the presence of ammonium sulfate, a condition that normally represses nitrogenase synthesis. As expected, the structural genes nifHDK, encoding the two subunits of molybdenum nitrogenase, were among those most highly expressed by the hydrogen-producing strains. Most of the 30 genes that surround nifHDK on the R. palustris chromosome were also expressed at high levels in the mutant strains (Fig. 2). Many of these are homologous to genes shown in other bacteria to be involved in the synthesis and assembly of two complex metalloclusters that are part of nitrogenase (28).

• Open in new tab
• Download powerpoint

FIG. 2.

Expression levels of genes in the molybdenum nitrogenase gene cluster in the hydrogen-producing strains grown with acetate and ammonium relative to those of the wild type grown under the same conditions. Expression levels of genes from wild-type cells grown with nitrogen gas and acetate (nitrogen-fixing conditions [NF]) relative to wild-type cells grown with ammonium and acetate are also shown.

Normally, cells produce hydrogen as an unintended consequence of nitrogen fixation during ammonia deprivation, an environmental cue that alters the expression levels of about 4% of the genome (24). The transcriptomes of the hydrogen-producing strains allowed us to identify genes that might be important for hydrogen production, as opposed to adaptation to ammonium starvation. Only 21 genes that are physically distant from the molybdenum nitrogenase cluster showed increased expression in at least three of the hydrogen-producing mutant strains compared to the wild type. A subset of 18 of these genes also had increased expression levels in the wild type under nitrogen-fixing (and hydrogen-producing) conditions (Table 6). Some of these likely encode proteins that indirectly facilitate nitrogenase activity, for example, by supplying iron needed for enzyme function (RPA2380 to RPA2388). Of more direct relevance to hydrogen production are genes that enhance the ability of R. palustris to convert light to ATP or that are involved in efficient channeling to nitrogenase of electrons required for the reduction of protons into molecular hydrogen. RPA3012 to RPA3013 encode light-harvesting II polypeptides that function to absorb light energy and transfer it to the bacteriochlorophyll-containing reaction center where it is converted into a proton gradient that is used to generate ATP (16, 29). RPA1928 codes for a ferredoxin predicted to function as an electron carrier. It may act in concert with the RPA1927 protein in this capacity. Electron transfer can be rate-limiting for nitrogenase activity (11). The intracellular pathways of transfer from electron-donating substrate to the nitrogenase have not yet been studied in R. palustris but bear investigation as, in addition to RPA1928, four other ferredoxin genes, all in the nitrogenase gene cluster, are expressed at high levels in the hydrogen-producing strains. Understanding the individual roles of these ferredoxins in the transfer of electrons to nitrogenase will be important for developing a complete understanding of the route to hydrogen production from electron-donating organic compounds in whole cells. In addition to genes of known or suspected function, the hydrogen-producing R. palustris strains expressed several genes (RPA2156, RPA4714, and RPA4827) of unknown function to high levels.

Single-amino-acid changes in the enhancer-binding protein NifA are sufficient to elicit nitrogenase expression and hydrogen production.The R. palustris genome codes for a set of regulatory proteins that overlap with those known to control nitrogen fixation in other bacteria in response to intracellular nitrogen status (6, 16, 21, 24, 37, 38). To identify the mutations responsible for constitutive hydrogen production, we sequenced genes predicted to be involved in these regulatory networks (Materials and Methods). We identified four different single point mutations in the regulatory gene nifA in the four hydrogen-producing strains (Fig. 3). NifA is an RNA polymerase sigma 54-dependent transcriptional activator that is required for nitrogenase gene expression in R. palustris (Y. Oda and C. S. Harwood, unpublished results). It contains a predicted central AAA⁺ ATPase motif, a carboxy-terminal DNA binding domain, and an amino-terminal GAF domain. The sigma 54 interaction domain is in the central portion of the protein and overlaps the AAA⁺ motif. GAF domains of NifA proteins from other bacteria sense nitrogen status directly as well as through interactions with other proteins (3, 19, 20). In proteins related to NifA, the amino-terminal domains modulate ATPase activity (5, 20, 33). All the mutations that we identified are located in a linker region between the regulatory domain and the AAA⁺ domain (Fig. 3). It is likely that these mutations cause structural changes in NifA that relieve a repressive effect of the GAF domain on the AAA⁺ ATPase domain, rendering the protein competent to activate gene expression constitutively.

• Open in new tab
• Download powerpoint

FIG. 3.

Proposed effects of nifA mutations on transcriptional activation. The mutations are S213P for strain CGA570, Q209P for strain CGA571, L212R for strain CGA572, and M202K for strain CGA574. A proposed configuration for NifA under noninducing conditions is shown on the left. We propose that the four mutations indicated in the linker region cause NifA to assume a conformation that mimics the active protein, as shown on the right. At this writing, no NifA structure has been solved. HTH, helix-turn-helix.

To determine whether the amino acid changes identified in NifA were sufficient to cause the hydrogen production phenotype, we introduced two of the mutated nifA genes, one encoding the Q209P change and the other encoding the M202K change, by homologous recombination into the chromosome of an R. palustris strain that has a nifA deletion mutation (Table 1). Both recombinant strains behaved similarly to the original nifA mutants and grew with cyclohexanecarboxylate as a carbon source (doubling times [in hours], 18 ± 3 for strain CGA581 and 17 ± 1 for strain CGA584). They also had nitrogenase activity (nitrogenase activity [in nmol C₂H₄ formed/min/mg protein], 74 ± 13 for strain CGA581 and 75 ± 23 for strain CGA584) and produced hydrogen when grown on acetate in the presence of ammonium (hydrogen production [in μmol H₂/mg protein], 95 ± 36 for strain CGA581 and 101 ± 6 for strain CGA584). We carried out a transcriptome analysis of one of the recombinant strains (CGA584) and found that it had elevated levels of gene expression that were similar in magnitude to those of the original hydrogen-producing mutant (Table 6; see Table S1 in the supplemental material). This indicates that the mutations that we identified in nifA are sufficient to activate all genes necessary for hydrogen production. We cannot exclude the possibility that the strains may also have additional mutations that contribute to the differences observed in growth, nitrogenase activity, and hydrogen production among the isolated mutant strains.