Evolving a New Electron Transfer Pathway for Nitrogen Fixation Uncovers an Electron Bifurcating-Like Enzyme Involved in Anaerobic Aromatic Compound Degradation

ABSTRACT Nitrogenase is the key enzyme involved in nitrogen fixation and uses low potential electrons delivered by ferredoxin (Fd) or flavodoxin (Fld) to reduce dinitrogen gas (N2) to produce ammonia, generating hydrogen gas (H2) as an obligate product of this activity. Although the phototrophic alphaproteobacterium Rhodopseudomonas palustris encodes multiple proteins that can reduce Fd, the FixABCX complex is the only one shown to support nitrogen fixation, and R. palustris Fix– mutants grow poorly under nitrogen-fixing conditions. To investigate how native electron transfer chains (ETCs) can be redirected toward nitrogen fixation, we leveraged the strong selective pressure of nitrogen limitation to isolate a suppressor of an R. palustris ΔfixC strain that grows under nitrogen-fixing conditions. We found two mutations were required to restore growth under nitrogen-fixing conditions in the absence of functional FixABCX. One mutation was in the gene encoding the primary Fd involved in nitrogen fixation, fer1, and the other mutation was in aadN, which encodes a homolog of NAD+-dependent Fd:NADPH oxidoreductase (Nfn). We present evidence that AadN plays a role in electron transfer to benzoyl coenzyme A reductase, the key enzyme involved in anaerobic aromatic compound degradation. Our data support a model where the ETC for anaerobic aromatic compound degradation was repurposed to support nitrogen fixation in the ΔfixC suppressor strain.

IMPORTANCE There is increasing evidence that protein electron carriers like Fd evolved to form specific partnerships with select electron donors and acceptors to keep native electron transfer pathways insulated from one another. This makes it challenging to integrate a Fd-dependent pathway such as biological nitrogen fixation into non-nitrogen-fixing organisms and provide the high-energy reducing power needed to fix nitrogen. Here, we show that amino acid substitutions in an electron donor for anaerobic aromatic compound degradation and an Fd involved in nitrogen fixation enabled electron transfer to nitrogenase. This study provides a model system to understand electron transfer chain specificity and how new electron transfer pathways can be evolved for biotechnologically valuable pathways like nitrogen fixation. KEYWORDS nitrogenase, Rhodopseudomonas palustris, ferredoxin, NAD 1 -dependent ferredoxin:NADPH oxidoreductase F erredoxins (Fds) and flavodoxins (Flds) are small protein electron carriers that transfer a single electron from an electron donor to an electron acceptor. In particular, Fds have specialized over evolutionary time to associate with specific partner proteins, allowing Fds to selectively shuttle electrons to specific pathways (1,2). Key factors such as the structure and charge of the Fd binding surface (3), regulation of Fd abundance (4), and the reduction potential of the Fd (5-7) affect which partner proteins interact with Fd. In theory, these properties can be altered to enable an Fd to interact with a new partner protein(s) to reroute electron flow, but this remains a challenge for rational design since these properties are ill-defined for many Fds. However, Fds can mediate electron transfer in biological reactions essential for growth, making it possible to select mutations that would allow an Fd and a new partner protein to interact (2,8).
One such reaction is biological nitrogen fixation, which is catalyzed by the enzyme nitrogenase. Nitrogenase uses large amounts of ATP and low potential electrons delivered by Fd or Fld to reduce atmospheric dinitrogen into ammonia, producing hydrogen gas as an obligate product of this activity (9). In the purple nonsulfur bacterium, Rhodopseudomonas palustris, electron transfer to nitrogenase requires the FixABCX complex, which couples the oxidation of NADH to the reduction of quinone and an Fd or Fld using flavin-based electron bifurcation (FBEB) (Fig. 1) (10)(11)(12). R. palustris with a deletion in fixA, fixB, fixC, or fixX has a severe growth defect under nitrogen-fixing conditions, despite R. palustris encoding multiple Fd-reducing enzymes that play a role in electron transfer to nitrogenase in other diazotrophs, including pyruvate:Fd oxidoreductase and Fd-NAD(P) 1 reductase (13)(14)(15)(16)(17)(18)(19)(20). R. palustris also encodes six 2[4Fe-4S] Fds, with the primary electron donor to nitrogenase being the Fd, Fer1 (Rpa4631) (11). The Fld, FldA (Rpa2117), can also act as an electron donor in the absence of Fer1 and plays a role under iron-limiting conditions (11). Since R. palustris encodes multiple Fds and Fd-reducing enzymes, R. palustris in which FixABCX is inactive could be leveraged to select for mutations that would enable a new electron transfer chain (ETC) to nitrogenase.
Using an R. palustris DfixC strain, we isolated a suppressor mutant that restored growth of this strain under nitrogen-fixing conditions. We found that two mutations in the suppressor strain were required to restore nitrogenase activity in the absence of a functional FixABCX complex (Fig. 1). One mutation was in the gene encoding Fer1, while the second was in the uncharacterized gene rpa0678. Protein modeling and genetic analysis revealed that the protein encoded by this gene is a homolog of a FBEB NAD 1 -dependent Fd:NADPH oxidoreductase (Nfn), and we found it is required for anaerobic aromatic compound degradation in R. palustris (Fig. 1). Because of its role in anaerobic aromatic compound degradation, we have renamed rpa0678 to aadN for anaerobic aromatic degradation, Nfn-like protein. The data here support a model where a new ETC for nitrogenase formed between components of two endogenous ETCs and provides a system that can be used to study the determinants of selective electron transfer.

RESULTS
A mutation in fer1 improves but is not sufficient for electron transfer to nitrogenase in the absence of FixC. As shown in Fig. 2A, the R. palustris DfixC strain (R. palustris DfixC) has a severe growth defect when grown under nitrogen-fixing conditions, FIG 1 Current model of electron transfer to nitrogenase (N 2 ase) and benzoyl-CoA reductase (BCR) in R. palustris. Electron transfer to BCR is incompatible with nitrogen fixation in the wild type, but the C38W substitution in AadN and T11I substitution in Fer1 enable the electron transfer pathway for benzoate degradation to support nitrogen fixation in DfixC*. The hypothesized activity of AadN shown in dotted lines is inferred based on its similarity to NAD 1 -dependent ferredoxin:NADPH oxidoreductase from Pyrococcus furiosus.
presumably because R. palustris DfixC is unable to generate enough reduced electron carrier (e.g., Fer1 or FldA) to support nitrogen fixation (10,11). To select for suppressor mutants of R. palustris DfixC, this strain was incubated for several weeks under nitrogen-fixing conditions supplemented with 20 mM acetate as a carbon substrate with light provided by a halogen light bulb. One of three replicate liquid cultures grew, from which a suppressor mutant strain of R. palustris DfixC was isolated, referred to here as R. palustris DfixC*. Deletion of fixA in R. palustris DfixC* did not disrupt the ability of the suppressor strain to grow in nitrogen-fixing conditions, confirming that the remaining Fix complex is not required in R. palustris DfixC* ( Fig. 2A). Genome sequencing revealed that DfixC* accumulated 18 mutations in 16 different genes ( Table 1). One of the mutations was in recQ, a DNA helicase involved in DNA repair, which may account for the large number of mutations found in the suppressor strain (21). While most of the mutations did not have an obvious connection to electron transfer, one of the mutations identified was in fer1, which encodes the primary electron donor to nitrogenase in R. palustris.
The mutation in fer1 results in the substitution of threonine 11 for isoleucine (T11I). To determine whether the Fer1 T11I variant was required for the suppressor phenotype, the fer1 T11I allele in R. palustris DfixC* was replaced with either an in-frame deletion in fer1 or a wild-type fer1 allele. Both strains had a significantly lower growth rate than R. palustris DfixC* (Table 2; see also Fig. S1 in the supplemental material), indicating that the fer1 T11I allele is required for the suppressor phenotype. However, even in the absence of Fer1, the suppressor strain was still able to grow, albeit at a reduced rate, suggesting that other electron carriers can compensate in the absence of Fer1. To determine whether the fer1 T11I allele was sufficient to restore growth under nitrogenfixing conditions in the absence of an intact FixABCX complex, the fer1 T11I allele was introduced into R. palustris DfixC. As shown in Fig. 3, the fer1 T11I allele did not restore growth in R. palustris DfixC under nitrogen-fixing conditions, indicating other mutations in R. palustris DfixC* are required for electron transfer. This finding revealed that the fer1 T11I allele was necessary but not sufficient for the suppressor phenotype. Because many of the remaining mutations in R. palustris DfixC* had unknown or hypothetical functions (Table 1), we needed to broaden our search for genes involved in electron transfer to nitrogenase in R. palustris DfixC*.
Nitrogenase is an energetically expensive reaction, requiring eight low-potential electrons per catalytic cycle (9,22). We hypothesized that components of the ETC would be transcribed at higher rates to accommodate the demand for reducing power in R. palustris DfixC*. Therefore, transcriptome sequencing (RNA-seq) analysis was carried out to compare gene expression changes in R. palustris DfixC* compared to wildtype R. palustris under nitrogen-fixing conditions. We found expression of fldA had the highest change In gene expression and was upregulated 23-fold in R. palustris DfixC* compared to wild-type R. palustris (see Data Set S1 in the supplemental material). To test the role of FldA in the ETC used by R. palustris DfixC*, strains were constructed with in-frame deletions in fldA, and their growth rates under nitrogen-fixing conditions were measured ( Fig. 2B and Table 2). We found FldA was not required for R. palustris  a CGA753 contains a deletion in anfH and vnfH (see Table S2) and is the parent strain of DfixC. b Values are averages of three biological replicates grown in minimal medium lacking ammonium sulfate with 20 mM acetate.
DfixC* to grow under nitrogen-fixing conditions, and FldA was not redundant with Fer1 in R. palustris DfixC* since R. palustris DfixC* with a deletion in fer1 and fldA did not grow slower than R. palustris DfixC* with a deletion in fer1 (Fig. 2B). Instead, R. palustris DfixC* with a deletion in fer1 and fldA had a slightly higher growth rate compared to R. palustris DfixC* with a deletion in fer1, suggesting that the presence of FldA may have a slight inhibitory effect in the absence of Fer1 (Table 2). However, this inhibitory effect was not observed when Fer1 is present since the growth rates of R. palustris DfixC* and R. palustris DfixC* with a deletion in fldA were indistinguishable (Table 2). RNA-seq also showed that most other genes encoding enzymes known to reduce Fd were downregulated or showed relatively minor (less than 2-fold) changes in gene expression in R. palustris DfixC* (see Table S1).
A mutation in aadN is both necessary and sufficient for electron transfer to nitrogenase in the absence of FixC. To identify genes required for electron transfer to nitrogenase in R. palustris DfixC*, we used a random transposon mutagenesis strategy combined with a metronidazole enrichment (see Fig. S2) (23). Metronidazole is an antibiotic that is activated when reduced by low-potential electron carriers, specifically Fd and Fld, causing cell death (24). Transposon mutants that survive metronidazole enrichment likely have insertions that disrupt electron transfer. Using this approach, we identified one transposon mutant that survived the enrichment and grew similar to R. palustris DfixC* under non-nitrogen-fixing conditions but could not grow under nitrogen-fixing conditions. This mutant had a transposon insertion in aadN (rpa0678). In R. palustris DfixC*, aadN had a nonsynonymous mutation, encoding a variant of AadN in which cysteine 38 is substituted for tryptophan (C38W) ( Table 1). We found that replacing the aadN C38W allele in R. palustris DfixC* with wild-type aadN disrupted the ability of the strain to grow in nitrogen-fixing conditions ( Table 2). When the aadN C38W mutation was introduced into the parent strain, R. palustris DfixC, this mutation alone allowed R. palustris DfixC to grow under nitrogen-fixing conditions, indicating that the aadN C38W mutation is both necessary and sufficient to restore growth of R. palustris DfixC under nitrogen-fixing conditions. However, the growth rate of R. palustris DfixC aadN C38W was slower than R. palustris DfixC* ( Fig. 3 and Table 2). When the aadN C38W mutation was combined with the fer1 T11I mutation in R. palustris DfixC, the growth rate increased (Fig. 3 and Table 2). This suggests that the variants of AadN and Fer1 form a new ETC that can deliver electrons to nitrogenase in the absence of FixABCX (Fig. 1).
We measured hydrogen production in growing cultures to quantify nitrogenase activity. Hydrogen is an obligate product of nitrogenase activity (9). Nitrogenase activity can be determined by measuring hydrogen production in these strains because they do not express an uptake hydrogenase, do not encode any other hydrogenases, and accumulate hydrogen only under nitrogen-fixing conditions (25). Introduction of the aadN C38W allele into R. palustris DfixC was sufficient to allow hydrogen production, but Evolving Electron Transfer to Nitrogenase mBio it produced about 50% less hydrogen compared to wild-type R. palustris or R. palustris DfixC* (Table 3). However, when both the fer1 T11I and aadN C38W alleles were introduced into R. palustris DfixC, hydrogen production was restored to levels observed for R. palustris DfixC*, confirming that only these two mutations are required for electron transfer to nitrogenase in the absence of a functional FixABCX complex (Table 3).
AadN is a homolog of a Fd-reducing enzyme and is required for anaerobic aromatic compound degradation. While these data indicate that AadN C38W is required for electron transfer to nitrogenase in R. palustris DfixC*, the native function of AadN in R. palustris was unclear. To gain insight into the role of AadN in electron transfer, we used protein modeling to make predictions about the structure and activity of AadN (26,27). Although AadN is annotated as a sulfide dehydrogenase, sequence analysis of AadN revealed that it shares homology with both the large and small subunit of the enzyme NfnI from Pyrococcus furiosus (PfNfnI, Fig. 4A) (28). PfNfnI is an NAD 1 -dependent Fd:NADPH oxidoreductase (Nfn) that uses FBEB to balance NADP(H), NAD(H) and Fd pools to conserve energy and maintain redox balance (Fig. 4B) (28,29). While the large and the small subunit of PfNfnI are encoded by two separate genes, these subunits are fused in AadN (Fig. 4A) (30). We found the cofactor binding domains in PfNfnI are conserved in AadN and both the large and small subunits share 51% or more amino acid identity (Fig. 4A). This suggests that AadN may be able to carry out FBEB and use electrons from NAD(P)H to reduce Fd (Fig. 4B).
We also found that aadN is adjacent to genes involved in anaerobic degradation of aromatic compounds such as benzoate and 4-hydroxybenzoate (4-HB) (Fig. 4C) (31). Anaerobic degradation of these compounds requires the enzyme benzoyl coenzyme A (benzoyl-CoA) reductase, which carries out ATP-dependent electron transfer from a low potential Fd to reduce the aromatic ring of benzoyl-CoA to cyclohex-1,5-diene-1carbonyl-CoA (see Fig. S3) (32)(33)(34). In Thauera aromatica, benzoyl-CoA reductase is supplied with reducing power through a Fd:2-oxoglutarate oxidoreductase known as KorAB (35). While two strains of R. palustris encode korAB homologs in the benzoate degradation gene cluster, seven R. palustris strains that encode genes for anaerobic aromatic compound degradation lack korAB (36). If AadN plays a role in electron transfer to benzoyl-CoA reductase, we reasoned that strains lacking korAB should encode aadN. Among R. palustris strains, aadN is present in the genomes of the six strains lacking korAB but is not found in the two strains encoding korAB (Table 4). This suggests that R. palustris strains either use AadN or KorAB for electron transfer during aromatic compound degradation but not both ( Table 4). The only R. palustris strain lacking aadN and korAB was strain HaA2, which cannot degrade aromatic compounds (36).
Given its proximity to genes required for anaerobic aromatic compound degradation, its similarity to PfNfnI, and its importance in electron transfer to nitrogenase in DfixC*, we hypothesized that AadN plays a role in electron transfer to benzoyl-CoA reductase. To test this, we looked at the ability of R. palustris DaadN to metabolize aromatic compounds. Benzoate and 4-HB are converted to benzoyl-CoA and are reductively de-aromatized by benzoyl-CoA reductase, but cyclohexane carboxylate (CHC) enters the same degradation pathway after this de-aromatization step (see Fig. S3) (36). When benzoate or 4-HB were provided as sole carbon sources, R. palustris DaadN had a growth defect, indicating that a All strains were grown in minimal medium lacking ammonium sulfate and supplemented with 20 mM acetate. b CGA753 contains a deletion in anfH and vnfH (see Table S2) and is the parent strain of DfixC. c Values are averages of three biological replicates calculated by subtracting H 2 measured from uninoculated samples.
Evolving Electron Transfer to Nitrogenase mBio aadN is required for degradation of benzoate and 4-HB (Fig. 4D). We found that AadN is not required to grow on CHC because both wild-type R. palustris and R. palustris DaadN were able to grow when CHC was provided as the sole carbon source (Fig. 4D). We also found that the C38W substitution in AadN did not disrupt the ability of R. palustris aadN C38W to grow on aromatic carbon sources (Fig. 4D), and R. palustris DfixC aadN C38W and R. palustris DfixC fer1 T11I aadN C38W grow under nitrogen-fixing conditions with benzoate as a carbon source (see Fig. S4), indicating that the C38W variant is able to support both nitrogen fixation and anaerobic aromatic compound degradation simultaneously. In tandem with evidence from protein sequence analysis, this suggests that AadN plays a role in electron transfer to benzoyl-CoA reductase.  The native ETC for anaerobic aromatic compound degradation is insulated from nitrogen fixation. Since R. palustris DfixC is unable to grow under nitrogen-fixing conditions, the ETC for anaerobic aromatic compound degradation and nitrogen fixation are likely insulated from each other. To probe the insulation of these two ETCs, we grew R. palustris strains with an in-frame deletion in fixC or aadN under nitrogen-fixing conditions with benzoate as a sole carbon source. We found that although aadN was required for growth on benzoate under non-nitrogen-fixing conditions (Fig. 4D), R. palustris DaadN was able to grow with benzoate as a carbon source under nitrogen-fixing conditions (Fig. 4E), suggesting that FixABCX may be able to complement the loss of aadN and support benzoate degradation under nitrogen-fixing conditions. However, R. palustris DfixC was unable to grow under nitrogen-fixing conditions with benzoate, indicating that the ETC for benzoate degradation cannot sustain electron transfer to nitrogenase (Fig. 4E). These data support a model in which electron transfer for benzoate degradation is insulated from nitrogen fixation, and the C38W substitution in AadN overcomes the apparent insulation to allow AadN to function in electron transfer to nitrogenase.

DISCUSSION
In this study, we identified mutations that would restore electron transfer to nitrogenase in R. palustris in the absence of FixABCX. We hypothesized that changes in Fd or Fld would be required to alter the flow of electrons in the cell, enabling the formation of a new ETC from existing components. While we found that a single amino acid substitution in Fer1 was important for the new ETC, it was not sufficient to support electron transfer to nitrogenase in the absence of FixABCX. We found a mutation in an Nfn-like gene we termed aadN was sufficient for electron transfer, but when combined with the mutation in fer1, electron transfer to nitrogenase was more efficient. While changing the properties of a Fd played an important role in making the new ETC more efficient, a single change in a Fd-reducing enzyme had a larger role in the formation of the new ETC. Therefore, it is likely that changes in both the Fd-reducing enzyme and the Fd will be required to optimize electron transfer through engineered pathways.
Our approach also uncovered a new role for an uncharacterized Nfn homolog. Nfn homologs are found in all domains of life, but the physiological role of many of these homologs is unknown (30,37). Sequence homology revealed that AadN is related to Nfn and is part of an uncharacterized family of Nfn enzymes known as pattern B Nfns, in which the two subunits of Nfn are fused (30). Pattern A Nfns, including PfNfnI, ligate two [4Fe-4S] clusters, one [2Fe-2S] cluster, two FAD cofactors, and have binding sites for NADPH and NAD 1 (28,38). We found that each of these substrate and cofactor binding sites were conserved in AadN, suggesting that AadN carries out FBEB using NADPH and NAD 1 to reduce Fd. However, not all Nfn homologs show Nfn bifurcating activity (39), and further structural and enzymatic analysis will be required to determine whether AadN carries out FBEB using the same substrates as Nfn. We showed that AadN is required for aromatic compound degradation and likely plays a role in electron transfer to benzoyl-CoA reductase. To the best of our knowledge, this is the first proposed role for a pattern B Nfn, and this discovery provides evidence that an Nfn-like enzyme can supply reducing power for anaerobic aromatic compound degradation. Our results also implicate the Nfn enzyme family in electron transfer to nitrogenase. Some diazotrophs do not appear to encode any Fd-or Fld-reducing enzymes known to be involved in electron transfer to nitrogenase (40). Our evidence that an Nfn homolog can supply reducing power to nitrogenase may help illuminate the ETCs for nitrogen fixation in some of these diazotrophic organisms.
The insulation of nitrogen fixation and anaerobic aromatic compound degradation highlights the complicated nature of electron transfer insulation. The key enzyme in anaerobic aromatic compound degradation, benzoyl-CoA reductase, requires low potential electrons delivered by a Fd (41). The benzoate degradation gene cluster in R. palustris encodes a Fd known as BadB, and a homolog of BadB in T. aromatica has been shown to have a Evolving Electron Transfer to Nitrogenase mBio very low midpoint potential of 2587 mV (31,34). Based on thermodynamics alone, the ETC for anaerobic aromatic compound degradation is predicted to be compatible with nitrogenase. However, we found that even when grown with benzoate, R. palustris DfixC could not grow under nitrogen-fixing conditions, indicating that the ETC for benzoyl-CoA reductase cannot support electron transfer to nitrogenase. The electron transfer insulation we observe could be due, in part, to the inability of BadB to interact with nitrogenase. However, our results indicate that the insulation between these two pathways must also be due to the inability of AadN to reduce Fer1 since an amino acid substitution in AadN restores some activity to nitrogenase and is further facilitated by an amino acid substitution in Fer1. It is unclear how these variants enable electron transfer to nitrogenase, but it is likely that they facilitate interaction between AadN and Fer1. The C38W amino acid substitution in AadN could enable electron transfer to nitrogenase by disrupting posttranslational regulation of AadN, affecting the stability of AadN, or altering the Fd binding site of AadN. The threonine residue at position 11 in Fer1 is adjacent to a cysteine predicted to coordinate one of the [4Fe-4S] clusters in Fer1. Many other low-potential 2[4Fe-4S] Fds encode isoleucine at this position and similar threonine to isoleucine substitutions in other Fds have been shown to lower the reduction potential of Fds (5,34,42). This suggests that the T11I substitution in Fer1 lowers its reduction potential, although it is unclear why this would facilitate interaction with AadN. Further characterization of how these amino acid substitutions alter the properties of Fer1 and AadN is needed to understand how these proteins can form a new electron transfer pathway to nitrogenase.
In summary, this study illustrates the potential of using a selection strategy to enable a new electron transfer pathway to nitrogenase. Given that electron transfer to nitrogenase is a major hurdle to engineering non-nitrogen-fixing organisms to fix nitrogen, this approach could be useful in evolving nonnative ETCs to be compatible with nitrogenase.

MATERIALS AND METHODS
Reagents, bacteria, and culture methods. All R. palustris strains were grown in defined mineral medium (non-nitrogen-fixing medium) containing 12.5 mM Na 2 HPO 4 , 12.5 mM KH 2 PO 4 , 7.6 mM (NH 4 ) 2 SO 4 , 0.1 mM Na 2 S 2 O 3 Á5H 2 O, 0.015 mM p-aminobenzoic acid, and 1% of a mineral salt solution (see Table S3) (43). Mineral medium lacking ammonium sulfate was used for nitrogen-fixing conditions. Media were prepared using an anaerobic chamber (atmosphere: 98% N 2 , 2% H 2 , ,10 ppm O 2 ) as described previously (11). Liquid cultures were supplemented with 20 mM acetate, and agar plates were supplemented with 10 mM succinate as carbon sources. Where indicated, liquid cultures were grown with 5.7 mM benzoate, 4-hydroxybenzoate, or cyclohexanecarboxylate and were supplemented with 10 mM HCO 3 (44). Plates were incubated in GasPak EZ anaerobe container systems at 30°C (Becton Dickinson). Plates were placed within 10 in. of a 60-W light bulb and liquid cultures were placed within 5.5 in. of the light bulb, which provides 30 mmol of photons m 22 s 21 (General Electric). Where applicable, R. palustris was grown with 100 mg/mL gentamicin and 200 mg/mL kanamycin, and Escherichia coli strains were grown in lysogeny broth at 37°C supplemented with gentamicin (20 mg/mL). For metronidazole enrichment, metronidazole was added to a final concentration of 50 mM.
Genetic manipulation of R. palustris. For each gene of interest, a corresponding pJQ200SK-derived deletion or allelic exchange vector was created (see Table S2) (45). Deletion vectors included ;1 kb of sequence upstream of the start codon and 1 kb of sequence downstream of the stop codon of the gene. Allelic exchange vectors contained 1 kb of sequence upstream and downstream of the point mutation of interest. Construction was carried out as described previously (11). Vectors were mobilized into R. palustris by conjugation using E. coli S17-1 (46). Gene deletions were confirmed by PCR (see Table S2). All allelic exchange strains were confirmed by Sanger sequencing (GENEWIZ, South Plainfield, NJ).
RNA extraction, cDNA library preparation, and sequencing. R. palustris cells were harvested from 10 mL of nitrogen-fixing medium after cultures had grown to an optical density (660 nm) of 0.4. Cells were incubated on ice for 10 min and harvested by centrifugation. The cell pellet was frozen in liquid nitrogen and stored at 280°C. Cell pellets were thawed and resuspended in 1 mL of QIAzol lysis reagent (Qiagen, Hilden, Germany) and homogenized using a BioSpec Products BeadBeater-24 (Bartlesville, OK) at maximum rpm for 1 min at 4°C and then allowed to cool for 1 min on ice. This cycle was repeated four times. Total RNA was isolated using the miRNAeasy minikit (Qiagen), and DNA was removed with TURBO DNase (Invitrogen, Carlsbad, CA). RNA was purified and concentrated using the RNeasy MiniElute Cleanup kit (Qiagen). cDNA library construction and library sequencing were performed at GENEWIZ, LLC (South Plainfield, NJ). rRNA was depleted using the Ribo-Zero rRNA removal kit (Illumina, San Diego, CA). cDNA was prepared using the NEBNext Ultra RNA Library Prep kit and sequencing reactions, image analysis, and base calling were performed on an Illumina HiSeq 2500 instrument (Illumina).

Evolving Electron Transfer to Nitrogenase mBio
Differential gene expression analysis. Quality base calling in sequencing data were analyzed using the FastQC application (v 0.11.8; https://www.bioinformatics.babraham.ac.uk/projects/fastqc/) and TrimGalore (v0.6.2) was used to remove adapter sequences, process, and validate all reads using the default parameters. Analysis was performed on the Avadis software package (v3.1.1; Strand Life Sciences, Bengaluru, India). Reads were aligned to the published genome of R. palustris CGA009 and differentially expressed genes were identified using the DESeq2 package (47) in R version 3.6 using the default parameters.
Transposon mutagenesis and metronidazole enrichment. Cultures of E. coli BW20767 (48) and R. palustris DfixC* were grown to mid-log phase, washed with minimal medium twice, mixed at equivalent concentrations, and then plated on minimal medium agar supplemented with 10 mM succinate, 0.2% yeast extract, and 0.5% Casamino Acids. Plates were incubated overnight at 30°C. After incubation, all visible biomass was transferred from the plate into liquid minimal medium supplemented with 20 mM acetate and kanamycin for 6 h. The Tn5 mutant pool was then pelleted and transferred to nitrogen-fixing medium with 20 mM acetate and kanamycin overnight. Metronidazole was added, and the culture was allowed to incubate at 30°C for 8 h. The culture was washed with minimal medium twice and plated on minimal medium agar with 10 mM succinate as a carbon source and kanamycin. Roughly 200 individual clones were isolated and screened for their ability to grow under nitrogen-fixing conditions.
Inverse PCR. Transposon mutants were grown in liquid minimal medium with 20 mM acetate to stationary phase, and genomic DNA was purified using a Yeast/Bact genomic DNA purification kit (Qiagen). 1 mg of genomic DNA was digested with the restriction enzyme AatII overnight at 37°C to generate fragments of genomic DNA which were, on average, 1,500 bp (New England Biolabs). The digestion product was then treated with Antarctic phosphatase for 1 h at 37°C (New England Biolabs). PCR products were purified using the Zymogen Clean and Concentrator PCR-cleanup kit (Zymo Research, Irvine, CA), and the recovered DNA fragments were ligated together to form closed circular DNA using T4 DNA ligase (New England Biolabs). The library of circular DNA fragments was used as a PCR template with forward and reverse primers specific to the transposable element and amplified using Phusion High-Fidelity DNA polymerase (New England Biolabs). PCR products were separated by electrophoresis on a 1% agarose gel and purified using the Zymoclean Gel DNA recovery kit (Zymo Research, Irvine, CA). The purified DNA fragments were sequenced using Sanger sequencing (GENEWIZ) using primers in (see Table S2).
Hydrogen measurements. H 2 was quantified using a Shimadzu GC-2014 gas chromatogram equipped with a thermal conductivity detector and a 60/80 molecular sieve 5-Å column (6 feet by 1/8 in.; Supelco). H 2 standards were measured in triplicate. Samples of headspace taken from growing cultures were measured in biological triplicate and technical duplicate. Samples were also taken from the headspace of uninoculated tubes containing the same medium used for growing cultures. The amount of H 2 found in the headspace of uninoculated tubes was subtracted from the amount of H 2 measured in growing cultures. For growing cultures, the headspace was sampled at an optical density of 0.4 to 0.55 at 660 nm. Cultures were vortexed briefly before sampling. H 2 production in each sample was normalized to the optical density (660 nm) of the culture at sampling time.
Protein sequence analysis. Specific protein-protein alignments were generated using the Constraint Based Alignment Tool (COBALT; NCBI) using the default parameters. Protein domains were identified using InterPro v.86.0 using the default parameters (26). Homologs of KorAB were identified using JGI/IMG-M using R. palustris BisB5 KorA as a bait sequence. Candidate homologs of korA had .80% amino acid identity to KorA and were adjacent to genes involved with anaerobic benzoate or 4-hydroxybenzoate degradation. AadN was used as a bait sequence to identify homologs among the selected R. palustris strains. Homologs of AadN had .90% amino acid identity and were adjacent to anaerobic benzoate or 4-hydroxybenzoate degradation genes.
Statistical analysis. Doubling times of different strains in each growth experiment were compared using analysis of variance (ANOVA; P ANOVA , 0.001). Welch's t test was used to compare the mean doubling times of individual strains. Similarly, normalized H 2 accumulation was compared using ANOVA, followed by a Welch's t test. All statistical analyses were performed in R version 4.1.1.
Data availability. Genome sequencing data for R. palustris DfixC* has been deposited in the NCBI Sequence Read Archive in BioProject PRJNA858464. RNA-seq reads have been deposited on the NCBI Gene Expression Omnibus in BioProject PRJNA858255. KorA and AadN sequences can be found using accession numbers WP_011501953.1 and WP_011156245, respectively.

SUPPLEMENTAL MATERIAL
Supplemental material is available online only.