A colorimetric method to measure in vitro nitrogenase functionality for engineering nitrogen fixation

Biological nitrogen fixation (BNF) is the reduction of N2 into NH3 in a group of prokaryotes by an extremely O2-sensitive protein complex called nitrogenase. Transfer of the BNF pathway directly into plants, rather than by association with microorganisms, could generate crops that are less dependent on synthetic nitrogen fertilizers and increase agricultural productivity and sustainability. In the laboratory, nitrogenase activity is commonly determined by measuring ethylene produced from the nitrogenase-dependent reduction of acetylene (ARA) using a gas chromatograph. The ARA is not well suited for analysis of large sample sets nor easily adapted to automated robotic determination of nitrogenase activities. Here, we show that a reduced sulfonated viologen derivative (S2Vred) assay can replace the ARA for simultaneous analysis of isolated nitrogenase proteins using a microplate reader. We used the S2Vred to screen a library of NifH nitrogenase components targeted to mitochondria in yeast. Two NifH proteins presented properties of great interest for engineering of nitrogen fixation in plants, namely NifM independency, to reduce the number of genes to be transferred to the eukaryotic host; and O2 resistance, to expand the half-life of NifH iron-sulfur cluster in a eukaryotic cell. This study established that NifH from Dehalococcoides ethenogenes did not require NifM for solubility, [Fe-S] cluster occupancy or functionality, and that NifH from Geobacter sulfurreducens was more resistant to O2 exposure than the other NifH proteins tested. It demonstrates that nitrogenase components with specific biochemical properties such as a wider range of O2 tolerance exist in Nature, and that their identification should be an area of focus for the engineering of nitrogen-fixing crops.

www.nature.com/scientificreports/ ambitious and challenging. The new symbiotic relationship requires molecular signaling between the bacteria and plants to avoid an immune response, the formation of a nodule-like structure with a low-O 2 environment and the productive exchange of nutrients between the plant and the bacteria. On the other hand, the transfer of the N 2 fixation capability is complicated by the estimated number of required genes (ca. [10][11][12][13][14][15][16][17][18][19][20], the sensitivity of their products towards O 2 , the need to perform time consuming functional validations, and difficulty to troubleshoot pathway engineering in plants 6 . Diazotrophs harbor a protein complex called nitrogenase that converts nitrogen (N 2 ) into ammonia (NH 3 ) in an intricate process requiring a large amount of energy in the form of ATP and low potential electrons 7 . Nitrogenase has two protein components: an α 2 β 2 heterotetrameric dinitrogenase formed by the nitrogen fixation (nif) nifD and nifK gene products, and a nifH-encoded homodimeric dinitrogenase reductase. During N 2 to NH 3 reduction, one NifH homodimer binds to each αβ half of the NifDK protein and, in an ATP-dependent reaction, transfers the electrons needed to break the N 2 triple bond 8 . Nitrogenase requires a minimum of 16 ATP molecules and 8 electrons to convert one molecule of N 2 into two molecules of NH 3 9 . The electrons are funneled through three [Fe-S] clusters, starting at a [Fe 4 S 4 ] cluster bridging the two subunits of NifH, via the P-cluster ([Fe 8 S 7 ]) and finally to the iron molybdenum cofactor (FeMo-co, [MoFe 7 S 9 C-(R)-homocitrate]). The latter two clusters are located at each αβ half of the NifDK heterodimer 9 . All three metalloclusters are extremely O 2 -sensitive, which makes engineering nitrogenase in plants especially challenging.
The activity of nitrogenase can be determined in vitro (e.g. using pure protein components), in vivo (e.g. in free-living cells) or in situ (e.g. bacteria associated to plants) using various techniques. One direct method is the quantitative measurement of the ammonia produced using its natural substrate N 2 10 , or by 15 N enrichment or 15 N natural abundance methods [11][12][13] . Furthermore, nitrogenase activity can be indirectly measured as nitrogenase can also reduce protons into H 2 14,15 , but also other double-and triple-bonded substrates such as acetylene, nitrite, nitrous oxide, and azide 16 . This promiscuity is often used by researchers to study nitrogenase in the laboratory. The most-used method is the acetylene reduction assay (ARA) in which acetylene is reduced to ethylene, which is detected using gas chromatography [17][18][19][20] . However, the ARA has several drawbacks. Firstly, the number of samples that can be measured is low due to the manual work involved. The manual steps include exchange of the gas phase in the reaction vial using inert argon to prevent N 2 reduction; injection of acetylene to start the reaction; incubation in a water-bath, injection of EDTA or NaOH to stop the reaction; and finally injection of gas from the vial headspace into the gas chromatograph. Secondly, acetylene reduction cannot be easily monitored in real time, and only the end concentration of ethylene after a defined time is determined. This second limitation of the ARA was recently overcome by the development of a viologen-based electron donor to nitrogenase 21 . In that method a reduced sulfonated viologen derivative (1,1′-bis(3-sulfonatopropyl)-4,4′-bipyridinium radical, hereafter referred to as S 2 V red ) replaces the function of sodium dithionite (DTH) as electron donor to NifH in vitro. Upon nitrogenase activity and turnover, this deeply violet-colored substrate is converted into an oxidized colorless form (S 2 V ox ) with greatly diminished absorbance at 600 nm (Fig. 1). The decrease in absorbance over time is therefore linear with nitrogenase activity. However, the exposure of the violet-colored S 2 V red to an oxidizing agent such as O 2 (or other reactive oxygen species), or to other natural electron acceptors, will lead to its conversion into the oxidized and colorless form (S 2 V ox ), limiting the use of S 2 V red to in vitro measurements under anaerobic conditions.
In this work, we have adapted the S 2 V red method to determine nitrogenase activity in 96-well microtiter plates with the aim to screen distinct NifH variants expressed in the mitochondria of the yeast Saccharomyces cerevisiae for functionality. The development of screening methods allows us to find Nif components with improved properties desirable for its expression in plant organelles. We demonstrate that: (1) the results obtained using S 2 V red are in accordance with those seen when using the standard ARA; (2) the method is compatible with NifH proteins originating from different prokaryotic origins; and (3) many samples and assay conditions can Figure 1. Schematic overview of nitrogenase activity and activity determination using S 2 V red . The flow of electrons from S 2 V red to protons (H + ) or N 2 via NifH and NifDK is shown. At NifDK, H 2 or NH 4 + (the dominating form at pH 7) is produced. ATP is regenerated from ADP and phosphocreatine by the action of creatine phosphokinase. Nitrogenase activity is determined by measuring the decrease in absorbance at 600 nm.

Results
Adaptation and use of S 2 V red for screening and activity determination of NifH proteins using 96 well microtiter plates. To confirm that S 2 V red -dependent electron donation to nitrogenase is not unique to NifH isolated from A. vinelandii (hereafter denoted as NifH Av ), as previously shown 21 , but is also suitable for functional screening of novel NifH variants, we combined the Hydrogenobacter thermophilus NifH protein previously isolated from S. cerevisiae 22 with NifDK Av . NifH variants expressed in yeast are hereafter denoted as ScNifH Xx where Sc and Xx indicates S. cerevisiae and the species from which the NifH sequence was obtained, respectively. Robust, although slower, oxidation of S 2 V red was observed using ScNifH Ht compared to NifH Av (Fig. 2). This was expected from the lower specific activity of ScNifH Ht when combined with NifDK Av determined by ARA in previous studies 22 . We next tested functionality of the S 2 V red assay using NifH Av and NifDK Av in 96 well microtiter plates sealed in the glove box (95% N 2 and 5% H 2 ). Several NifDK concentrations (0.05-0.2 µM) and NifH to NifDK molar ratios (2x-40x) were tested to identify conditions that provided a robust and linear rate constant (k obs ) over an extended time-period (to facilitate sample preparations and plate handling), and that minimized the amount of NifH and NifDK proteins required for the assay. Supplementary Fig. S1 shows the decrease in absorbance over time under the tested experimental conditions. As expected, higher NifDK concentrations resulted in a faster S 2 V red oxidation ( Supplementary Fig. S1d). Supplementary Figure S2 shows the corresponding k obs variation over time. The fast S 2 V red oxidation observed at high NifDK concentration resulted in lower k obs ( Supplementary Fig. S2c), presumably because the concentration of S 2 V red was already suboptimal by the time the plate was being scanned (after sealing in the glove box and transfer to the plate reader). The maximum activity observed under the conditions tested here (k obs of ca. 12 s −1 ) was obtained when the mixture contained 2 µM NifH Av and 0.05 µM NifDK Av (corresponding to a NifH:NifDK ratio of 40) (Fig. 2b), and was almost identical to that reported when performed under argon atmosphere and using cuvettes 21 . Importantly, the decrease in absorbance was linear under these conditions for more than 10 min ( Supplementary Fig. S1b) (R 2 = 0.9997, t 1 = 1 to t 2 = 10 min), which provided sufficient time for microtiter plate preparation, transfer and reading. These experimental conditions were therefore implemented for subsequent use of S 2 V red .

Solubility screening of NifH variants targeted to the mitochondria of yeast.
To obtain proofof-concept of S 2 V red suitability to quickly screen the function of engineered Nif proteins, we used a library of 35 mitochondria-targeted NifH proteins in yeast (Supplementary Tables S1 and S2). The bulk of the NifH variants originated from a library that had previously been tested in Nicotiana benthamiana 22 . The NifH variants were expressed as N-terminally TwinStrep (TS)-tagged proteins with a Cox4 mitochondria targeting signal 23,24 . The purpose of the TS-tag was to enable equal detection of the distinct NifH variants and their subsequent isolation using Strep-tag affinity chromatography (STAC). The NifH proteins were co-expressed with NifM, NifU and NifS using galactose inducible promoters. NifM has been proposed to be necessary for proper folding of the NifH polypeptide 25 , whereas NifU and NifS provide NifH with its [Fe 4 S 4 ] cluster 7 . All the accessory proteins originated from A. vinelandii and were equipped with a Su9 leader sequence 24,26 . The functionality of the Cox4 and the Su9 sequences for mitochondrial targeting of each respective protein in yeast has been shown previously 22,27,28 .
With the assumption that Nif components must accumulate as soluble proteins to be relevant for engineering nitrogenase in eukaryotes, we first tested the NifH variant solubility. Accumulation of ScNifH Xx variant  Isolation and activity measurements of soluble ScNifH variants. Three of the nine soluble ScNifH variants (from H. thermophilus, M. marburgensis and M. infernus) had already been purified in our laboratory 22 . The remaining ScNifH proteins (together with ScNifH Ht that was reisolated for further work described in this study) were isolated using STAC under anaerobic conditions ( Fig. 4a-c, Supplementary Fig. S4). The yield varied from 9-35 mg ScNifH per 100 g cell paste for six of the variants (Table 1), in line to what was previously reported for ScNifH Mm and ScNifH Mi . Only ScNifH Lb was isolated at much lower level. This variant was excluded from further analysis. These ScNifH proteins all presented color and an UV-vis absorbance spectrum characteristic of [Fe-S] cluster containing proteins (Fig. 4d, Supplementary Fig. S5). Iron quantification suggested that six ScNifH variants had a similar amount of bound [Fe-S] cluster as NifH Av purified from its native host 29 (Table 1), while ScNifH Mi and ScNifH Mm were isolated with much lower Fe content. We then performed nitrogenase assays with the different ScNifH proteins using S 2 V red as the electron donor to NifH, and NifDK Av as its electron acceptor (Fig. 4e). Three variants, namely those originating from H. thermophilus (ScNifH Ht ), G. sulfurreducens (ScNifH Gs ) and D. ethenogenes (ScNifH De ), accelerated S 2 V red oxidation when the ScNifH:NifDK Av ratio was increased, which is expected from a functional and NifDK Av -compatible NifH variant. These ScNifH proteins showed acetylene reduction activities consistent with those obtained using S 2 V red (Fig. 4e,f). Interestingly, ScNifH Rs (and to some extent ScNifH Ra ) could act as reductase for NifDK Av during ARA (i. e. when using DTH as electron donor), but not in the S 2 V red assay (Fig. 4f). This divergence could potentially originate from different reduction potential requirements among the NifH variants, as S 2 V red has a potential of − 0.40 V vs Normal Hydrogen Electrode (NHE) 21 and DTH has a potential of − 0.66 V vs NHE 30 . Whether this discrepancy indicates a different mechanism or requirement for NifH Rs activity remains to be investigated in future studies.

Inhibition of ScNifH by O 2 .
The NifDK Av compatible ScNifH variants were assayed for their sensitivity to O 2 , which represents a major barrier to engineer nitrogenase in plants. As S 2 V red itself is oxidized by O 2 , and  Supplementary Table S2 for full species names) in soluble (a) and total (b) protein extracts was determined using antibodies detecting the TwinStrep-tag (α-Strep). Two different exposure times (s.e., short exposure (above) and l.e., long exposure (below)) are shown for the analysis of total extracts.  31,32 , we were not able to design an experiment using S 2 V red to study the effect of O 2 on the activity. We therefore measured the ScNifH variants capacity to support acetylene reduction upon exposure to O 2 following a previously reported method 31 . In short, DTH present in the buffer of the isolated ScNifH protein was first removed using a desalting column inside an anaerobic glove box. The ScNifH protein was then added to anaerobic buffer in a glass vial containing argon in the headspace (representing t = 0). Then, O 2 was injected into the headspace to a final concentration of 20% and the vial was incubated with rigorous shaking. At distinct time points, ScNifH was extracted using a Hamilton syringe and transferred to an open vial containing anaerobic buffer supplemented with DTH to quench the O 2 . Finally, NifDK Av and an ATP-regenerating mixture was added , right Y-axis, blue bars). Mean and SD is shown. n = 3 technical replicates (ARA) and n = 4 (S 2 V red ).

NifM-dependency for ScNifH solubility and functionality. Co-expression of the A. vinelandii nifM
gene with nifH Av is required for the accumulation of functional NifH Av protein in the mitochondria of S. cerevisiae 33 . Several of the NifH sequences in this study were selected because of the absence of a nifM orthologue in the organism's genome. To test whether NifM was required for the solubility of the eight ScNifH variants, we compared their accumulation in total and soluble protein extracts when co-expressed with NifU Av and NifS Av , but not NifM Av , in yeast. Surprisingly, six of the eight ScNifH variants (Roseiflexus sp., R. albus, M. marburgensis, M. infernus, Firmicutes bacterium, and D. ethenogenes) showed no obvious decrease in solubility when NifM Av was absent (Fig. 6a). To test whether functionality could be affected although solubility was not, we isolated ScNifH De from the yeast strain not expressing NifM Av (Fig. 6b, Supplementary Fig. S6a). The UV-vis spectrum suggested no apparent difference in [Fe-S] cluster content ( Supplementary Fig. S6b), and the specific activity was similar to ScNifH De protein isolated from cells co-expressing NifM Av (Fig. 6c). Seven of our eight ScNifH variants contained a proline residue at the site corresponding to Pro 259 (when including the methionine) in A. vinelandii (Fig. 6d, Supplementary Fig. S7), which is thought to be the target of NifM prolyl isomerase activity 25 . The NifH protein from Firmicutes bacterium is shorter and terminates before this proline. Interestingly, the only genome of the eight selected NifH variants that contained a gene with high similarity to NifM Av was G. sulfurreducens (Supplementary Table S2). ScNifH Gs was also the variant that was least soluble when NifM Av was not co-expressed (Fig. 6a). The only other protein that showed reduced solubility in the absence of NifM Av was ScNifH Ht . The genome of H. thermophilus harbors a gene encoding a hypothetical protein  (Supplementary Fig. S8). Therefore, the direct action of NifM with regards to NifH is not clear and to some extent in disagreement with the published literature 7 , and should be the topic of future studies.

Discussion
The transfer of prokaryotic nitrogenase activity into cereals could generate crops suited to grow well under limited nitrogen fertilizer. Although there are excellent reports on engineering of nitrogenase in heterologous (non-N 2 -fixing) bacterial hosts [34][35][36][37][38][39][40] , our experience is that it is very difficult to directly translate and transfer that knowledge to a eukaryotic system and expect comparable results, even in a relatively simple, unicellular eukaryote such as yeast 41 . A major challenge arises from the extremely complex biochemical requirements of the nitrogenase enzyme and its stepwise maturation involving several inter-dependent gene products 7 . Additionally, from a metabolic point of view, nitrogenase requires high levels of energy and reducing power in an environment that is low in O 2 to protect its metalloclusters from oxidative damage.
In this study we have developed an important part of the nitrogenase engineering process, namely the analysis of NifH protein functionality in a high throughput assay. While the ARA is very precise, it requires training to generate consistent results, it is rather time-consuming and would be difficult to scale up for screening large numbers of samples and/or conditions. We optimized the S 2 V red assay and showed that it fulfilled many of our main objectives, most importantly to be fast and simple to use, to require a lower amount of purified proteins (corresponding to about half of that used in the ARA as the reaction volume is smaller), and to not depend on expensive or sophisticated equipment. We also believe that this method is easily adaptable to automated robotic systems as the reactions are performed in microtiter plates. In addition, the S 2 V red assay has two important advantages over the ARA. Firstly, activities can be monitored in real-time, which means that it is possible to directly study the effect of various effector molecules or reaction components on nitrogenase functionality. Secondly, the reduction potential of S 2 V red (used as the electron donor to NifH) is much closer to that of ferredoxin or flavodoxin, the physiological reductants of nitrogenase 42,43 than DTH. NifF for example, a flavodoxin in the diazotrophic free-living model-bacteria A. vinelandii donating electrons to NifH, harbors a flavin mononucleotide (FMN) cofactor with a redox potential in the semiquinone/hydroquinone state of − 0.483 V vs NHE 44 . The corresponding potentials for S 2 V red is − 0.40 V vs NHE, compared to − 0.66 V vs NHE for DTH 21,30 . However, it is important to note that other reported flavodoxins and ferredoxins have lower reduction potentials, for example  45 and ferredoxin in A. vinelandii (− 619 mV) 46 , and that S 2 V red would not be a suitable electron donor to study nitrogenases requiring such strong reductants. Whether this could explain the lack of nitrogenase activity when combining ScNifH Rs and ScNifH Ra with NifDK Av in the S 2 V red is not clear, as it is also possible that other steric or charge factors prelude productive electron transfer from S 2 V red to these NifH variants. Other important drawback with using S 2 V red is that it is not commercially available, and that it cannot be used directly with yeast extracts as it is effectively oxidized by unknown molecule(s) in the lysate (data not shown). Therefore, Nif proteins must be purified prior to the activity assay. Solving this limitation would further expand the use of the S 2 V red .
Regarding the functional assessment of Nif proteins expressed in yeast and plants, we have observed that many of the essential Nif components have poor solubility, especially NifH and NifB 22,24,27,47 . This is critical as the structural components (NifH and NifDK) are needed at very high levels during nitrogen fixation. In N 2 -fixing A. vinelandii for example the NifH concentration within the cell can reach up to 100 μM 48 , whereas in K. oxytoca about 40% of the total protein is NifHDK 49 . For this, a simple protein solubility study is always the first experiment to perform before initiating more complex analyses 22,27 . From the 35 mitochondrial-targeted NifH variants expressed in this study, we obtained nine that were soluble in yeast mitochondria. The phyla from where these nine NifH variants originated were diverse, and so was their mechanisms of nutrition and relationship to oxygen. The only common factor we could observe was a bias towards coming from thermophilic organisms, as has been observed and discussed previously in works from our laboratory 22,24,27 .
To expand the analysis of these soluble NifH variants and to see if we could identify properties that would facilitate their functionality in future crops, we tested two aspects that are sought after for eukaryotic nitrogenase engineering; 1) simplification of the nitrogenase genetic machinery by minimizing the number of genes needed to transfer, and 2) identification of Nif components with better functionality in an environment containing oxygen. In this work, that meant (1) the identification of a NifH variant that did not depend on NifM for solubility and functionality, and (2) one NifH variant whose [Fe 4 S 4 ] cluster was more resistant towards O 2 .
Although NifM is just one protein, each gene fewer to transfer will make the engineering of nitrogenase in plants less complex. When the K. pneumoniae NifH protein was expressed in Escherichia coli in the absence of NifM, the protein was much less stable and completely inactive 50 . This was in agreement with the low levels of NifH Kp polypeptide and dinitrogenase reductase activity detected in nifM − strains of the native host 51 . Work in yeast has shown that NifM co-expression was required for homodimer formation and polypeptide stability of Rhizobium meliloti NifH 52 , and in tobacco NifM was required to prevent NifH aggregation in the mitochondria 53 .
While not many studies have investigated how NifM acts on NifH, sequence analysis suggests NifM to be a member of the rotamase family (PF00639) containing a PPIC-type PPIASE domain 54 . Prolyl isomerases (also known as peptidylprolyl isomerases or PPIases) are enzymes that accelerate protein folding by catalyzing the cis-trans isomerization of prolyl peptide bonds. This annotation is consistent with work identifying Pro 259 in NifH from A. vinelandii as the prime target for NifM action 25 . In this work, seven of the final eight variants contained a proline residue at the site corresponding to Pro 259 in A. vinelandii. The only exception was NifH from Firmicutes bacterium, but this protein was significantly shorter than the other NifH variants and therefore lacking this proline. However, the only two NifH variants that showed NifM-dependent solubility (ScNifH Gs and ScNifH Ht ) corresponded to those that originated from organisms containing genes with some similarity to nifM Av . Therefore, our work suggests that presence of a nifM homologue in the organism genome is a better indicator of NifM-dependency than presence of a proline at a site corresponding to Pro 259 in A. vinelandii.
Equally important is the identification of more O 2 -resistant nitrogenase components, as O 2 -sensitivity is likely to be the major barrier to overcome to obtain a functional plant nitrogenase 5 . Active NifH could be expressed in the cytosol of anaerobically cultured yeast, while only the mitochondria could produce active protein under aerobic conditions 33 . This has been explained by the low O 2 concentration in the mitochondria of actively respiring cells. Whether it is possible to obtain similarly O 2 -depleted conditions in plant mitochondria is not known. One scenario would be to limit nitrogenase expression to plant cells in hypoxic niches 55 . In any case, the identification of more O 2 -tolerable nitrogenase components would be a breakthrough for the engineering of nitrogenase in crops. In this regard, we were surprised to see increased resistance towards O 2 by NifH from G. sulfurreducens. As all NifH proteins contain a [Fe 4 S 4 ] cluster, we assumed that the variants tested in this study would also show similar O 2 -susceptibility. Whether the [Fe 4 S 4 ] cluster in ScNifH Gs is less exposed, or whether it is stabilized by other means, is not known but these are interesting questions for future work. Importantly, this study shows that Nif proteins with better properties for expression in eukaryotic cells exist in Nature, and that their identification could pave the way for the engineering of N 2 fixing crops.

Materials and methods
NifH library design and assembly. The majority of the nifH genes originated from a previously published gene set 22 . To this nifH library the genes for expression of NifH originating from L. boryana, Frankia alni and D. ethenogenes were added. All yeast codon-optimized DNA sequences and their corresponding protein products can be found in Supplementary Table S2. The nifH genes originating from the previously published gene set 22 were amplified as cox4-ts-nifH gene fusions by PCR using primers #2584 (5´-AAT TTT TGA AAA TTC GAA TTC CTC TTG ACC ATG CTT TCAC-3') and #2585 (5´-GAA GAA TTG TTA ATT AAG AGC TCG GGG AAA TTC GAG CTGG-3´). These primers include 15 bp overhangs complementary to the pESC-HIS yeast expression plasmid (#217451, Agilent Technologies) when digested with SacI and EcoRI, and allowed for the insertion by an exonuclease and ligation-independent (ELIC) method 56  The gene encoding for mitochondria-targeted SU9-NifM Av was amplified from plasmid pN2XJ165 22 using primers #2478 (5´-CTC TAC AAA TCT ATC TCT CTC GAG ATG GCC TCC ACT CGTG-3´) and #2479 (5´-ATT ATG  GAG AAA CTC GAG TTA ACC ATG TGC TAA GTT TTCC-3´) and inserted into pESC-TRP (#217453, Agilent Technologies) digested with XhoI by ELIC. The pESC-URA plasmid for expression of mitochondria-targeted Su9-NifU Av and Su9-NifS Av has been previously described 33 . All DNA digestions were performed using enzymes from New England Biolabs. PCR amplifications were carried out using Phusion Hot Start II High-Fidelity DNA Polymerase (ThermoFisher Scientific). ELIC products with their corresponding digested target vectors were transformed using a molar ratio 1:4 (vector:insert) into chemically competent E. coli DH5α and selected on solid LB (Lysogenic broth) media supplemented with appropriate antibiotics. Plasmid preparations were performed using Qiaprep Spin Miniprep kit (QIAGEN) and correct cloning was confirmed by Sanger sequencing (Macrogen). Plasmids were transformed into S. cerevisiae W303-1a (MATa leu2-3,112 trp1-1 can1-100 ura3-1 ade2-1  his3-11,15) according to the lithium acetate method 57 , and selected and grown in synthetic drop-out medium with the appropriate auxotrophic selection 24 .
Expression analysis and solubility screening of ScNifH Xx variants. Small-scale yeast protein extracts were prepared from yeast grown in galactose induction media as previously described 24 . The Yeast-Buster protein extraction reagent (Merck) was used to prepare total and soluble yeast protein extracts. First, galactose-induced yeast was pelleted for 10 min at 3000×g. YeastBuster mixture supplemented with 25 μg/ml DNAse I and 1 mM phenyl-methylsulfonyl fluoride (PMSF) was added to the yeast pellets at a ratio of 9 μl per OD × ml in Eppendorf tubes, and then incubated on a Eppendorf shaker for 20 min at room temperature to lyse the cells. This sample was then divided in two equal parts. For total extracts, the resulting YeastBuster lysate was added to 2 × Laemmli buffer at a 1:1 (v/v) ratio. For soluble extracts, the YeastBuster lysate was centrifuged in a benchtop centrifuge at maximum speed for 20 min at 4 °C before the supernatant was added to 2 × Laemmli buffer at a 1:1 (v/v) ratio. Both samples were prepared for SDS-PAGE by heating for 5 min at 95 °C. Following SDS-PAGE, proteins were either stained using Coomassie brilliant Blue R-250 (Sigma) or transferred to nitrocellulose membranes (Protran Premium 0.45 µm, GE Healthcare) membranes for immunoblotting. Nitrocellulose membranes were stained with Ponceau S (Sigma) to ensure equal loading control and successful transfer. The membranes were blocked with 5% non-fat milk in TBS-T (20 mM Tris-HCl pH 7.5, 150 mM NaCl, 0.02% Tween-20) for 1 h at room temperature before incubation with primary antibodies overnight at 4 °C. Polyclonal antibodies detecting NifM Av (used at 1:2,000 in 5% BSA), NifU Av (used at 1:2000 in 5% BSA) and NifS Av (used at 1:1,000 in 5% BSA), were raised against purified preparations of the corresponding A. vinelandii proteins (generated in house). Strep-tag II antibody ("Strep-MAB", IBA Lifesciences, 1:2000 in 5% BSA) was used for detection of all ScNifH Xx variants. Secondary antibodies (Sigma) were diluted 1:20,000 in TBS-T supplemented with 2% non-fat milk and incubated for 2 h at room temperature. Membranes were developed using enhanced chemiluminescence and images were recorded digitally (iBright FL1000, ThermoFisher).

S. cerevisiae growth and NifH variants purification.
The growth of yeast cultures, galactoseinduced Nif expression and STAC-purification of soluble ScNifH Xx variants followed the procedure previously described 27 . Cell pellets from 4 l fermenters stored in liquid N 2 (typically 200-220 g) were resuspended in lysis buffer (100 mM Tris-HCl pH 8.8, 200 mM NaCl, 10% glycerol, 2 mM DTH, 1 mM PMSF, 1 μg/ml leupeptin, 5 μg/ml DNAse I) at a ratio of 1:2 (w/v) inside an anaerobic glovebox (Coy Laboratories). Total extracts (TE) were prepared by lysis of the cell suspensions under anaerobic atmosphere using an EmulsiFlex-C5 homogenizer (Avestin Inc.) operating at 20,000 psi. The TE was transferred to centrifuge tubes equipped with sealing closures (Beckman Coulter) and centrifuged at 50,000×g for 1 h at 4 °C (Avanti J-26 XP). The supernatant was filtered using filtering cups with a pore size of 0.2 μm (ThermoFisher), rendering cell-free extract (CFE) of soluble proteins that was loaded at 2.5 ml/min into a 5 ml Strep-Tactin XP column (IBA LifeSciences) attached to an ÄKTA FPLC (GE Healthcare) at O 2 -levels below 1 ppm in anaerobic chambers operating at 16 °C (MECAPLEX or MBraun). The column was washed overnight using about 120 ml wash buffer (100 mM Tris-HCl pH 8.0, 200 mM NaCl, 10% glycerol, 2 mM DTH). Strep-Tactin XP column-bound proteins were eluted with 15 ml washing buffer supplemented with 50 mM biotin (IBA LifeSciences). The elution fraction was concentrated using centrifugal filters with 30 kDa cutoff (Amicon, Millipore), loaded into PD-10 desalting columns (GE Healthcare) equilibrated with wash buffer to remove biotin and DTH, and then used to UV-Vis absorption spectrum analysis (see section below). The desalted eluate was supplemented with 2 mM DTH, further concentrated using centrifugal filters and finally snap-frozen as protein pellets in cryovials (Nalgene) and stored in liquid N 2 .
Protein quantification, UV-Vis absorption spectrum and iron measurements. The concentrations of purified ScNifH Xx variants were measured using the BCA protein assay (Pierce) in combination with iodoacetamide to eliminate the interfering effect of DTH 58 . ScNifH Xx UV-Vis absorption spectra were recorded after removal of the DTH from the protein samples. The DTH-free protein samples were further diluted in wash buffer and transferred to Q6 spectroscopy cuvettes with sealing closures. Absorption (280 nm to 800 nm) was recorded using a UV-2600 spectrophotometer (Shimadzu). For recording of the air exposed ScNifH Xx samples, the sealing closure was removed, and the protein sample was carefully exposed to air using a pipette equipped with a gel loading tip. Iron content of as isolated ScNifH Xx  Nitrogenase activity determination by S 2 V red assay in cuvette. Nitrogenase activity of ScNifH Ht expressed in yeast mitochondria and isolated by STAC 22 was determined following a recently described spectrophotometric method 21 . Activity assays were performed in cuvette (600 µl final reaction volume) in the presence of ATP regenerating mixture (6.7 mM MgCl 2 , 5 mM ATP, 30 mM phosphocreatine, 0.2 mg/ml creatine phosphokinase, 1.3 mg/ml bovine serum albumin (BSA) in 100 mM MOPS pH 7.0) and 0.5 mM 1,1′-bis(3sulfonatopropyl)-4,4′-bipyridinium (S 2 V red ). Nitrogenase activity was determined from the decrease in absorbance at 600 nm upon addition of 0.4 µM of NifDK Av and increasing concentrations of NifH Av or ScNifH Ht . Absorbance was recorded using a USB 400-ISS-UV/VIS spectrophotometer (Ocean Optics) using a cuvette with a path length of 0.2 cm. Nitrogenase activity calculations were performed as previously described 21 .
Nitrogenase activity determination by S 2 V red assay in 96-well microtiter plates. Nitrogenase activity determined by the S 2 V assay were scaled down to 200 µl reaction volume to be performed in a 96-well microtiter plate. Except for during the optimization of the method, the assay was performed using a mixture of 2 µM NifH Av and 0.05 µM NifDK Av (corresponding to a NifH:NifDK ratio of 40:1). Other reaction conditions such as buffer composition, ATP regenerating mixture and S 2 V red concentration were identical to those described in the section above. The plate was prepared under anaerobic conditions inside a glovebox (Coy Laboratories) and sealed using PCR plate sealing films. Absorbance reading was performed using an absorbance plate reader (SPECTROstar Nano, BMG LABTECH) operating at 30 °C. The absorbance was recorded for 1 h, with measurements taken every 30 s. Nitrogenase activity calculations were performed in Excel (Microsoft) using an molar extinction coefficient of 9925 M −1 cm −1 at 600 nm 21 , and a path length of 5 mm. The slope was calculated using the range in which the decrease in absorbance was linear, normally for at least 10 min. The final calculation can be expressed in simplified form as k obs (s −1 ) =|m|/ ((ε•l) • [NifDK Av ]), where |m| is the absolute value of the slope of the decrease in absorbance at 600 nm, ε is the molar extinction coefficient of S 2 V red at 600 nm in M −1 cm −1 (9925), l is the path length in cm (0.5) and [NifDK Av ] is the molar concentration of NifDK Av in the assay (normally 0.05e −6 ). Oxygen sensitivity assays. ScNifH Xx samples and NifH Av were passed through PD-10 desalting columns (GE Healthcare) equilibrated with 100 mM MOPS (pH 7.5) to remove DTH. Oxygen sensitivity was tested inside an anaerobic glovebox (Coy Laboratories) as previously described 31 with slight modifications. First, 2.5 ml of 17.4 μM NifH in 100 mM MOPS (pH 7.5) was prepared in a 13 ml sealed glass vial. The atmosphere in the headspace was exchanged for argon. Diluted NifH sample was removed (t 0 ) before pure O 2 was injected (0.2 atm final) using a 250 μl gastight syringe (Hamilton). The vial was incubated at room temperature with shaking (800 rpm) on a thermomixer (Eppendorf) with an adaptor for 13 ml vials. Air exposed NifH samples were removed after 1, 2, 5 and 10 min. Fifty ul was transferred to open 9 ml vials (three technical replicates) containing 150 μl of 100 mM MOPS (pH 7.5) supplemented with 4 mM DTH. Finally, 200 ul ATP mixture supplemented with 5 µg of NifDK Av was added before nitrogenase activity was measured following the protocol for ARA as described above.