The iron nitrogenase reduces carbon dioxide to formate and methane under physiological conditions: A route to feedstock chemicals

Abstract

Nitrogenases are the only known enzymes that reduce molecular nitrogen (N₂) to ammonia. Recent findings have demonstrated that nitrogenases also reduce the greenhouse gas carbon dioxide (CO₂), suggesting CO₂ to be a competitor of N₂. However, the impact of omnipresent CO₂ on N₂ fixation has not been investigated to date. Here, we study the competing reduction of CO₂ and N₂ by the two nitrogenases of Rhodobacter capsulatus, the molybdenum and the iron nitrogenase. The iron nitrogenase is almost threefold more efficient in CO₂ reduction and profoundly less selective for N₂ than the molybdenum isoform under mixtures of N₂ and CO₂. Correspondingly, the growth rate of diazotrophically grown R. capsulatus strains relying on the iron nitrogenase notably decreased after adding CO₂. The in vivo CO₂ activity of the iron nitrogenase facilitates the light-driven extracellular accumulation of formate and methane, one-carbon substrates for other microbes, and feedstock chemicals for a circular economy.

Iron nitrogenases produce the feedstock chemical formate from environmental CO₂ and offer opportunities for a green economy.

INTRODUCTION

Nitrogenase is the key enzyme of biological nitrogen fixation (BNF) and catalyzes the ambient reduction of elemental dinitrogen (N₂) to ammonia (NH₃) to build life’s central metabolites, e.g., nucleotides and amino acids. Three nitrogenase isoforms are known in nature, which are distinguished by the heterometal content of their active site cofactor. The molybdenum (Mo) nitrogenase (encoded by nifHDK) is present in all diazotrophs, organisms that fix N₂. In contrast, only a few diazotrophic microorganisms harbor in addition to the Mo also vanadium (V; encoded by vnfHDGK) or iron (Fe; encoded by anfHDGK) nitrogenases (1). Mo nitrogenases have a higher N₂ reduction efficiency and are preferably expressed, whereas the alternative V or Fe nitrogenases are believed to function as fail-safe enzymes under Mo starvation (2).

All three nitrogenases consist of a reductase component (NifH₂, VnfH₂, and AnfH₂) and a catalytic component [Nif(DK)₂, Vnf(DGK)₂, and Anf(DGK)₂; Fig. 1] (3–8). The catalytic component harbors two [Fe₈S₇] clusters (P-clusters) and two active site cofactors. The latter is a [MFe₇S₉C-(R)-homocitrate] cluster for the Mo and Fe nitrogenase (where M is either Mo or Fe) and a [VFe₇S₈C(CO₃)(R)-homocitrate] cluster for the V nitrogenase. On the basis of the contained heterometal, the clusters are termed FeMoco, FeVco, or FeFeco. The reductase component contains two adenosine triphosphate (ATP) binding sites and a [Fe₄S₄] cluster.

Fig. 1. The nitrogenase composition and activities of R. capsulatus.

(A) Architecture of the Mo [left, Protein Data Bank (PDB): 7UTA] and Fe (right, PDB: 8OIE) nitrogenase. (B) Electron flow through the Mo nitrogenase. Electrons are delivered by the [Fe₄S₄] cluster of the reductase component and transferred via the [Fe₈S₇] cluster (P-cluster) to the FeMoco. The same scheme applies to the Fe nitrogenase. (C) Product spectrum of Mo (top) and Fe (bottom) nitrogenase for the reduction of N₂ and CO₂.

In the ATP-bound state, the reductase component transiently associates with the catalytic component, enabling the transfer of a low-potential electron from the [Fe₄S₄] cluster via the P-cluster to the active site cofactor. Subsequently, both molecules of ATP are hydrolyzed, and the reductase component dissociates (9, 10). This way, the active site cofactor accumulates electrons, eventually used to reduce substrates (11). Despite the similar architecture, different reaction stoichiometries are observed for the reduction of N₂ (at 1 atm) by the three nitrogenase isoforms (Eqs. 1 to 3) (12):

Mo nitrogenase:

$$\begin{array}{c}{\mathrm{N}}_2+10 {\mathrm{H}}^{+}+20 \text{MgATP}+10 e^{-}\to \\ 2 {\text{NH}}_3+2 {\mathrm{H}}_2+20 \text{MgADP}+20 {\mathrm{P}}_i\end{array}$$ (1)

V nitrogenase:

$$\begin{array}{c}{\mathrm{N}}_2+18 {\mathrm{H}}^{+}+36 \text{MgATP}+18 e^{-}\to \\ 2 {\text{NH}}_3+6 {\mathrm{H}}_2+36 \text{MgADP}+36 {\mathrm{P}}_i\end{array}$$ (2)

Fe nitrogenase:

$$\begin{array}{c}{\mathrm{N}}_2+20 {\mathrm{H}}^{+}+40 \text{MgATP}+20 e^{-}\to \\ 2 {\text{NH}}_3+7 {\mathrm{H}}_2+40 \text{MgADP}+40 {\mathrm{P}}_i\end{array}$$ (3)

The N₂ reduction requires a low reduction potential of the active site cofactors (E_1/2/V versus normal hydrogen electrode FeMoco, −0.59 V; FeVco, –0.38 V; and FeFeco, −0.40 V) (13). Hence, nitrogenases have the reducing power to convert a variety of small molecules and exhibit diverse promiscuous activities (1, 14).

The reduction of CO₂ by nitrogenase was first described for the Mo isoform, which converts CO₂ to carbon monoxide (CO) at a rate of ~0.09 nmol × [nmol Nif(DK)₂ × min]^–1 (15). Later, formate (HCOO⁻) was identified as the main product of the reaction, which is formed at 100-fold higher rates {9.8 nmol × [nmol Nif(DK)₂ × min]^–1} than CO (16). More recently, also the alternative nitrogenases were found to reduce CO₂. Intriguingly, the V nitrogenase performs carbon-carbon coupling reactions during CO₂ reduction, releasing short-chain hydrocarbons such as ethene and ethane besides CO (17, 18). The Fe nitrogenase was shown to convert CO₂ to HCOO⁻ and methane (CH₄) (19, 20). Notably, the electron flux of the Fe nitrogenase toward the reduction of CO₂ was found to be equivalent to the electron flux going into N₂ in vitro (21).

CO₂ is a common metabolite in bacteria (22). Thus, under physiological conditions, nitrogenases are constantly exposed to intracellular CO₂ that likely interferes with the reduction of N₂. However, the competing reduction of N₂ and CO₂ by nitrogenases has not been investigated in vitro or in vivo so far. Therefore, the impact of CO₂ on nitrogenase activity and thus N₂ fixation in diazotrophic organisms remains elusive. Moreover, CO₂ is the primary waste product of modern fossil fuel–based economies and the greatest contributor to human-made climate change (23). Therefore, its removal and utilization from the atmosphere is a major societal challenge driving the need for CO₂ reduction processes to establish a carbon-neutral economy (24). Given their unique reactivities, nitrogenases offer alternative approaches to convert CO₂ directly into biofuels and feedstock chemicals such as methane, ethene, or propene (25).

Here, we study the competing reduction of N₂ and CO₂ by the Mo and Fe nitrogenase of Rhodobacter capsulatus to assess the impact of CO₂ reduction on nitrogen fixation. We find marked selectivity differences between the two isozymes. The Mo nitrogenase is very selective for N₂ fixation, and the Fe nitrogenase is more promiscuous toward CO₂ reduction. In vivo, R. capsulatus depending on the Fe nitrogenase is inhibited by CO₂ and forms HCOO⁻ and CH₄. Thus, the Fe nitrogenase functions as a N₂ and CO₂ reductase under physiological conditions.

RESULTS

To comprehensively characterize the reduction of CO₂ particularly in competition experiments of N₂ and CO₂ by the Mo and Fe nitrogenase, we extended our recently established plasmid-based expression and purification system of the Fe nitrogenase in R. capsulatus by the Mo nitrogenase (8). After purifying both nitrogenase enzymes, we confirmed their full H⁺ and N₂ reduction activity in vitro (fig. S1). The nitrogenase-specific activity followed the expected hyperbolic trend for H₂ and NH₃ formation under an increasing ratio of reductase to catalytic component because higher relative concentrations of the reductase component lead to more association events with the catalytic component and thus an increased electron flux (26). Both isozymes reached the activity plateau at the expected reductase component excess (Mo nitrogenase: ~20:1, Fe nitrogenase: ~40:1). They exhibited the previously observed NH₃ formation rates under high electron flux conditions {Mo nitrogenase: 107 ± 6 nmol × [nmol Nif(DK)₂ × min]^–1 and Fe nitrogenase: 36.4 ± 1.0 nmol × [nmol Anf(DGK)₂ × min]^–1; fig. S1} (27).

After confirming that the purified Mo and Fe nitrogenase were fully active for N₂ and H⁺ reduction, we proceeded by analyzing their ability to reduce CO₂ in vitro (Fig. 2). Under a CO₂ atmosphere, the Mo nitrogenase formed mainly H₂, followed by HCOO⁻ with formation rates of 168.7 ± 1.6 nmol × (nmol Nif(DK)₂ × min)⁻¹ and 58 ± 2 nmol × (nmol Nif(DK)₂ × min)^–1, respectively (Fig. 2A). Besides H₂ and HCOO⁻, the Mo nitrogenase also formed traces of CO at a rate of 0.40 ± 0.03 nmol × [nmol Nif(DK)₂ × min]^–1, but we could not detect any hydrocarbons (Fig. 2B). This follows the overall behavior of the Mo nitrogenase of Azotobacter vinelandii (16). The higher obtained HCOO⁻ formation activity {58 ± 2 nmol × [nmol Nif(DK)₂ × min]^–1} for the Mo nitrogenase of R. capsulatus compared to the reported value {9.8 nmol × [nmol Nif(DK)₂ × min]^–1} for the Mo nitrogenase of A. vinelandii might be due to the higher partial pressure of CO₂ (1.2 atm versus 0.45 atm) used in our experiments (16).

Fig. 2. In vitro CO₂ reduction by the Mo and Fe nitrogenase.

Product formation rates of the Mo (A and B) and Fe nitrogenase (C and D) for in vitro activity assays conducted under a 1.2-atm CO₂ atmosphere. Plotted are specific activities for the formation of H₂ and HCOO⁻ [(A) and (C)] and CO and CH₄ [(B) and (D)] versus the molar excess of reductase component over catalytic component. Each dot represents an activity assay. n = 3 independent experiments. Solid lines represent the nonlinear fit of the data.

In contrast to the Mo nitrogenase, the main product of the Fe nitrogenase was HCOO⁻ {60 ± 5 nmol × [nmol Anf(DGK)₂ × min]^–1}, followed by H₂ {34 ± 6 nmol × [nmol Anf(DGK)₂ × min]^–1; Fig. 2C}. The fact that less electrons go into H₂ than in HCOO⁻ experimentally confirms that the HCOO⁻ formation does not proceed via reductive elimination but via direct hydride transfer or an associative pathway as predicted by Khadka et al. (16) using density functional theory calculations. In addition, we found that the Fe nitrogenase produced CO and CH₄ at rates of 0.19 ± 0.03 nmol × [nmol Anf(DGK)₂ × min]^–1 and 0.026 ± 0.003 nmol × [nmol Anf(DGK)₂ × min]^–1, respectively (Fig. 2D). The formation of CO from CO₂ by the Fe nitrogenase has previously not been reported, although the activity for CO formation is an order of magnitude higher than the CH₄ formation. Both the Fe and Mo nitrogenase exhibit the same specific activity for HCOO⁻ formation [~60 nmol × (nmol×min)^–1] despite the more than twofold higher electron flux of the Mo nitrogenase. The same HCOO⁻ formation rate could indicate a common rate-limiting step for both nitrogenases in reducing CO₂ to HCOO⁻, independent of the electron flux. Moreover, the ratio of reductase to catalytic component of the Fe nitrogenase required for maximum substrate reduction rates decreased from ~40:1 for N₂ reduction to ~20:1 for CO₂ reduction (Fig. 2, C and D). This result might indicate distinct rate-limiting steps for CO₂ and N₂ reduction of the Mo and Fe nitrogenase. For N₂ reduction under high electron flux, the hydrolysis of ATP and thus the release of the reductase component are regarded to be the rate-limiting step (28). Since we observe lower reductase to catalytic component ratios to be sufficient for maximal activity for HCOO⁻ formation by nitrogenases, a step different from ATP hydrolysis might be rate limiting for CO₂ reduction. In conclusion, the Mo and Fe nitrogenase of R. capsulatus show distinct product profiles for the reduction of CO₂. In contrast to the Mo isoform, the Fe nitrogenase can reduce CO₂ directly to CH₄ and unexpectedly uses most electrons to form HCOO⁻, while the Mo nitrogenase mainly forms H_2. The rate of HCOO⁻ formation is similar for both nitrogenase enzymes and, notably, seems to be independent of the electron flux.

To gain a deeper understanding of the observed activities, we analyzed the in vitro electron flux under Ar, N₂, and CO₂ to evaluate the efficiency of the Mo and Fe nitrogenase in directing electrons into the different products (Fig. 3, A, B, E, and F) (16, 20). Under an Ar atmosphere, the Mo nitrogenase shows a more than twofold higher electron flux toward the formation of H₂ than the Fe isoform. Changing the headspace to N₂, we observed a drop in the Mo nitrogenase electron flux of 47%, whereas the overall electron fluxes of the Fe nitrogenase dropped by only 17%. This result follows previously observed trends for the nitrogenase enzymes from A. vinelandii (29). Similarly, when changing the reaction atmosphere from Ar to CO₂, we also observed a drop in the overall electron flux for both isoenzymes (Mo nitrogenase: 60% drop; Fe nitrogenase: 42% drop; Fig. 3, A and E). Intriguingly, previous studies have demonstrated that the Mo nitrogenase electron flux does not change between Ar and N₂ atmospheres in a cyclic voltammetry setup that includes electron mediators such as methyl viologen (30). This implies that unaccounted reduction reactions cause the apparent drop in electron flux by switching from Ar to N₂. Recent work by Tanifuji et al. (31) suggested that releasing NH₃ from the active site cofactor of the Mo nitrogenase requires replenishment of the belt sulfides by reducing SO₃²⁻, thereby causing a previously unaccounted electron flux. Thus, the observed drop of electron flux for the two isozymes when switching to a CO₂ atmosphere indicates that this mechanism could also apply to the CO₂ reduction reaction and might imply that CO₂ has a similar binding mode as N₂.

Fig. 3. Electron flux analysis of the Mo and Fe nitrogenase.

Left: Electron flux toward the formation of H₂, NH₃, HCOO⁻, CO, and CH₄ in vitro under Ar, N₂, and CO₂ atmospheres for the Mo nitrogenase (A and B) and Fe nitrogenase (E and F). Right: Electron flux distribution of in vitro competition assays of Mo (C and D) and Fe nitrogenase (G and H) under mixed atmospheres of CO₂ and N₂. All data presented as means ± SD from replicates (n = 3).

We also found that the flux of electrons toward H₂ formation increased for the Mo nitrogenase while being suppressed for the Fe nitrogenase when we changed the headspace gas from N₂ to CO₂ (Fig. 3, A and E). This resulted in the Fe nitrogenase having a 2.75-fold higher efficiency for CO₂ reduction, with 66% of the total electron flow ending in CO₂ reduction products in the case of the Fe nitrogenase. These results contradict the previously described ratios for the Fe nitrogenase of A. vinelandii, where only 31 ± 0.3% of the electrons are directed into HCOO⁻ and the majority into H₂ formation; this difference might originate from the lower used CO₂ partial pressure of 0.45 atm in the A. vinelandii study (20). In comparison, the Mo nitrogenase directs only 24% of the electron flux to HCOO⁻ under the same conditions. The Fe nitrogenase exhibited a similar electron flux toward HCOO⁻ formation (116 ± 9 e⁻ × min^–1) under a CO₂ atmosphere as toward NH₃ (124 ± 5 e⁻ × min⁻¹) under a N₂ atmosphere, which emphasizes the ability of the Fe nitrogenase to function as a N₂ and a CO₂ reductase (Fig. 3E).

Intrigued by the similar activity of the Fe nitrogenase for CO₂ and N₂ reduction, we wondered whether CO₂ is a competitor for N₂ reduction for the Fe nitrogenase and investigated the substrate specificity in more detail. For this purpose, we prepared in vitro activity assays under mixed atmospheres of CO₂ and N₂ and compared the Mo and Fe nitrogenase electron flux for the individual reduction products (Fig. 3, C, D, G, and H). For the Mo nitrogenase, the electron flux toward NH₃ formation decreased only by 10% when the CO₂ concentration was increased from 0 to 75% (Fig. 3C). Consequently, NH₃ remains the main product in all assays, where N₂ is present in the headspace. HCOO⁻ formation was first observed under an atmosphere of 75% CO₂ in N₂ (3.7% of total electron flux) and increased to 25.8% of the total electron flux under 100% CO₂. We observed the same trend for the electron flux toward CO formation that started at 75% CO₂ and increased fivefold in the absence of N₂ (Fig. 3D). In contrast to the Mo isoform, the Fe nitrogenase electron flux toward NH₃ formation decreased markedly from 50% in the absence of CO₂ to only 6% under 75% CO₂ (Fig. 3G). Moreover, the formation of HCOO⁻ by the Fe nitrogenase already started at 50% CO₂, and HCOO⁻ became the main product of the reaction under an atmosphere of 75% CO₂. Formation rates of CO and CH₄ increased accordingly with rising concentrations of CO₂ in the headspace (Fig. 3H). Notably, we observed similar trends for both nitrogenases in CO₂ titration experiments conducted under an Ar atmosphere. However, the onset of CO₂ reduction was earlier (50% CO₂ for the Mo nitrogenase and 25% CO₂ for the Fe nitrogenase) compared to the results obtained in N₂ (fig. S2). In conclusion, we have demonstrated that the Mo nitrogenase is more selective for reducing N₂, and the Fe nitrogenase is more promiscuous for reducing CO₂ even in the presence of N₂. Despite the difference in selectivity, the Mo nitrogenase readily reduces CO₂ in the absence of N₂; thus, we speculate that a structural feature beyond the difference of active site metalloclusters (i.e., FeMoco and FeFeco) is responsible for the discrimination of CO₂ in the presence of N₂. Hence, the Fe nitrogenase might be a well-suited model system to study substrate selectivity as it is naturally more promiscuous than the Mo nitrogenase.

Next, we tested whether CO₂ would influence the diazotrophic growth of R. capsulatus expressing either the Mo or the Fe nitrogenase. For this purpose, R. capsulatus strains relying on the Mo nitrogenase (MM0335, encoding both nitrogenases, the Mo and Fe nitrogenase, but only expressing the Mo nitrogenase in the presence of Mo) or the Fe nitrogenase (MM0057, ΔnifD::SpR, ΔmodABC) were used to characterize their growth under an N₂ atmosphere with increasing CO₂ concentrations (Fig. 4). Both strains were derived from strain BS85, a strain with an interrupted and thus unfunctional Mo nitrogenase (ΔnifD::SpR) (32). Strain MM0335 was obtained by reverting the nifD interruption. For strain MM0057, the molybdenum transporter gene modABC was deleted to prevent the repression of the Fe nitrogenase in the presence of Mo traces in the medium. In the absence of CO₂, growth of R. capsulatus strains depending on the Mo nitrogenase showed a 23% lower doubling time (T_d) than the Fe nitrogenase–dependent strain (Mo: T_d = 6.65 ± 0.19 hours versus Fe: 8.6 ± 0.2 hours), which is in line with the higher N₂ fixing activity and efficiency of the Mo nitrogenase (compare Fig. 4, A and B to Fig. 4, C and D) (33, 34). Increasing the CO₂ concentration in the culture headspace gradually to 5, 10, and 20% CO₂ caused the T_d of the Mo nitrogenase–dependent strain to increase slightly (12%) to 7.42 ± 0.1 hours (Fig. 4B). In contrast, the Fe nitrogenase–dependent strains exhibited a clear dose response to rising concentrations of CO₂. The T_d of the Fe nitrogenase–expressing strain increased strongly already under 10% CO₂ and (T_d = 11.6 ± 0.3) reached a T_d of 15.0 ± 0.5 hours (a 74% increase) under 20% CO₂ (Fig. 4D). In conclusion, our findings suggest that CO₂ is a competitor to N₂ for in vivo nitrogenase activity and appears to affect diazotrophic growth for Fe nitrogenase–dependent strains.

Fig. 4. Effect of CO₂ on diazotrophic growth of R. capsulatus.

Diazotrophic growth curves of R. capsulatus strains expressing the Mo (A) and Fe (C) nitrogenases with increasing CO₂ concentrations. (B and D) Corresponding doubling times (T_d) in the exponential growth phase. Each dot represents the values from biological replicates (n = 6). Data in (B) and (C) are presented as means ± SD: *P < 0.05 and ****P < 0.0001; ns, not significant; a.u., arbitrary units.

On the basis of our in vitro results, we hypothesized that the strong inhibitory effect of CO₂ on the diazotrophic growth of Fe nitrogenase–dependent strains might arise from the high promiscuity of the Fe nitrogenase for the reduction of CO₂. To address this hypothesis, we first monitored CO₂ reduction products (HCOO⁻ and CH₄) during diazotrophic growth in the absence and presence of 20% CO₂ (Fig. 5, A and E). As demonstrated in our in vitro data, HCOO⁻ is the main product of CO₂ reduction by nitrogenases. Consequently, we detected high concentrations of HCOO⁻ (4.54 ± 0.13 mM) in the culture medium for the Fe nitrogenase strain. Because of the membrane permeability of HCOO⁻, it is released into the culture medium, allowing for easy quantification of metabolically derived HCOO⁻ and thus nitrogenase CO₂–reducing activity (35). To account for non-nitrogenase–derived HCOO⁻, we compared the Mo and Fe nitrogenase–dependent strains to controls supplemented with NH₃, which suppresses nitrogenase expression. Intriguingly, only the Fe nitrogenase–dependent strain exhibited significantly elevated HCOO⁻ concentrations in the culture medium compared to the NH₃ controls (Fig. 5A). Although Mo nitrogenase–derived HCOO⁻ formation is possible, the difference under Mo nitrogenase–expressing and repressed conditions is insignificant. However, we observe a significant increase of HCOO⁻ for the Fe nitrogenase–expressing strain compared to the Mo counterpart even under conditions without supplemented CO₂, revealing the Fe nitrogenase–catalyzed HCOO⁻ formation under physiological conditions (Fig. 5A). HCOO⁻ formation by the Fe nitrogenase plateaus at the time when the culture is entering the stationary phase (fig. S3), which resembles the pattern known from N₂ fixation by nitrogenases (36, 37). This observation adds another line of evidence that HCOO⁻ is indeed derived from a nitrogenase-catalyzed CO₂ reduction. To exclude the interference of the formate dehydrogenase (FDH) with our results, we repeated the same experiments with FDH-deficient but otherwise identical strains (fig. S4). FDH of R. capsulatus is a Mo-dependent enzyme with a Mo cofactor at the active site. Hence, FDH should only be active under Mo-containing conditions, i.e., exclusively in the Mo nitrogenase–dependent strain. The deletion of FDH did not increase or decrease the HCOO⁻ concentrations, neither for the Fe nor the Mo nitrogenase–dependent strain. Thus, we conclude that the observed differences in HCOO⁻ formation were solely based on the respective nitrogenase isoform.

Fig. 5. In vivo CO₂ reduction by nitrogenase-expressing R. capsulatus strains.

(A) HCOO⁻ concentration in R. capsulatus culture supernatant expressing the Mo nitrogenase or the Fe nitrogenase under diazotrophic growth conditions in the absence or presence of 20% CO₂ after 6 days of growth. (B) Representative ¹H NMR (300.0 MHz, D₂O) spectra of R. capsulatus culture supernatant incubated under 20% CO₂ or ¹³CO₂. Shown are the resonance peaks of H¹²COO⁻ (red) and H¹³COO⁻ (teal). (C) Fraction of ¹³C-labeled HCOO⁻ in the culture supernatant of the Fe nitrogenase strain and the respective Δfdh strain grown with 20% CO₂ (black) or 20% ¹³CO₂ (99%-¹³C enriched, white) in the culture headspace. The ratio between ¹²C- and ¹³C-labeled HCOO⁻ was determined by NMR spectroscopy. (D) HCOO⁻ accumulation over time in the culture medium of R. capsulatus ΔanfHDGK strain expressing the Fe nitrogenase from a plasmid (orange) or carrying an empty plasmid (gray). The cultures were cultivated under an atmosphere of 8% CO₂ in Ar using glutamate as the N source. (E) Amount of CH₄ measured in the culture headspace of R. capsulatus cells after 9 days of growth determined by GC-FID. Each dot represents the individual values from biological replicates (A and C to E: n = 3). Data in (A), (C), and (E) are presented as means ± SD: **P < 0.01 and ****P < 0.000. Solid lines in (D) represents the nonlinear fit of the data.

To confirm that CO₂ is the substrate of the observed reactions, we performed ¹³CO₂ labeling experiments (Fig. 5, B and C). For this purpose, the Fe nitrogenase strains with and without the FDH gene [wild type (WT) or Δfdh] were cultivated under an atmosphere of N₂ supplemented with 20% CO₂ or 20% ¹³CO₂. After 6 days, extracellular HCOO⁻ was quantified by ¹H nuclear magnetic resonance (NMR) spectroscopy using the characteristic down field shifted signal at δ 8.44 parts per million (ppm) (Fig. 5B). Here, the H¹³COO⁻ signal is split by ¹J(¹³C,¹H) coupling, which enables the relative quantification of the H¹³COO⁻ and H¹²COO⁻ signal intensity. Both Fe nitrogenase–dependent strains showed high ¹³C-labeling ratios of 41.4% for the WT and 44.0% for the Δfdh strain, confirming our expectations that the WT strain cannot produce a functional FDH under Mo starvation. Unlabeled HCOO⁻ likely originates from metabolically derived CO₂, which is directly reduced by the Fe nitrogenase. In summary, these results confirm CO₂ as the origin of nitrogenase-derived HCOO⁻.

To unambiguously confirm that the Fe nitrogenase is responsible for the in vivo reduction of CO₂, we created knockout strains, lacking both nitrogenases and fdh. Next, we complemented the strain either with an anfHDGK expression plasmid or an empty control plasmid and supplemented the growth media with glutamate as the N source. Growing these two strains under an atmosphere of 8% CO₂ in Ar, cultures of the anfHDGK complemented strain produced even higher amounts of HCOO⁻ (13.3 ± 0.4 mM). In contrast, the control strain with an empty plasmid produced only 2.18 ± 0.12 mM HCOO⁻ under identical conditions (Fig. 5D).

Analogously to the formation of HCOO⁻, we observed a CO₂ concentration–dependent formation of CH₄ (Fig. 5E). Notably, we could only detect CH₄ in the headspace of the Fe nitrogenase but not in the Mo nitrogenase strains, which stands in accordance with our in vitro results and previous literature (19). We could also detect significant amounts of CH₄, beyond the background of the nitrogenase repressed culture, without the addition of any CO₂. Thus, physiological CO₂ concentrations are sufficient for the formation of CH₄ and HCOO⁻ in Fe nitrogenase–expressing diazotrophs and in the presence of N₂.

DISCUSSION

In this study, we analyzed the competing reduction of N₂ and CO₂ by the Mo and Fe nitrogenase. Under CO₂, both enzymes generate HCOO⁻ at similar rates, an unexpected result considering the twofold higher overall electron flux of the Mo over the Fe nitrogenase. Our results demonstrate that the Fe nitrogenase directs two-thirds of its electron flux into the reduction of CO₂. This exceeds the electron flux observed for the formation of NH₃ under an N₂ atmosphere, leading to our conclusion that the Fe nitrogenase acts as N₂ and CO₂ reductase, simultaneously. In support of this hypothesis, the Fe nitrogenase reduces more CO₂ than N₂ in an atmosphere of 75% CO₂ and 25% N₂. In contrast, the Mo isoform is highly selective for N₂ and reduces detectable amounts of CO₂ only in the absence of N₂. The lower selectivity of the Fe nitrogenase might be of physiological relevance, as we observe CO₂-dependent deceleration of diazotrophic growth in R. capsulatus strains relying on the Fe nitrogenase. Metabolically derived CO₂ is sufficient to drive the Fe nitrogenase–catalyzed HCOO⁻ and CH₄ formation in vivo (Fig. 5, A and E). These findings challenge our overall understanding of nitrogenases as sole N₂- and H⁺-reducing enzymes by showing CO₂ to be a competing substrate for BNF under physiological conditions.

The Fe nitrogenase–derived CO₂ reduction products (i.e., CH₄ and HCOO⁻) are known one-carbon substrates for methanotrophs, as well as methanogens, acetogens, formate-hydrogen lyase-containing taxa, and aerobic formatotrophs (e.g., using the Calvin or serine cycle) (38, 39). In addition, the formed CH₄ and HCOO⁻ is excreted by R. capsulatus into the environment and could therefore influence the local microbial community. The amount of HCOO⁻ and the rate of its excretion by the Fe nitrogenase–expressing strains would be sufficient to support the growth of HCOO⁻-dependent syntrophs and methanogens. For instance, the 4.5 mM HCOO⁻ observed in the Fe nitrogenase–expressing cultures exceeds concentrations required for the cultivation of the syntrophic organism Moorella sp. strain AMP (40). Furthermore, HCOO⁻ was secreted during the exponential growth at a rate of 53 μM/hour (fig. S3). This is a similar rate as found in a steady-state syntrophic consortia, where the syntrophic organism Desulfovibrio alaskensis G20 donates lactate-derived HCOO⁻ to the methanogen Methanococcus maripaludis at a rate of ~50 μM/hour (41). Whether this activity is observed in nature depends on the expression of the Fe nitrogenase, which is expected to be a fail-safe enzyme in Mo-depleted environments. However, this assumption might be too simplistic for BNF in the environment. Several studies have demonstrated remarkable activities of alternative nitrogenases in the environment contributing up to 50% of the total BNF (42). These activities were observed in soil, anaerobic sediments, salt marshes, leaf litter, decaying wood, microbial mats, boreal cyanolichens, cryptogams, and termites (43, 44). These environmental observations are also reflected in laboratory studies, where the Fe nitrogenase provided a fitness advantage for colonies on Mo containing agar plates (45). Here, the diffusion of Mo was likely limiting the Mo nitrogenase activity. In summary, the Fe nitrogenase–catalyzed HCOO⁻ and CH₄ release likely occurs in the environment and affects the local composition of bacterial communities.

Astonishing is the HCOO⁻ concentration achieved by the Fe nitrogenase. In the absence of N₂, we observed an accumulation of more than 13 mM HCOO⁻ in the culture supernatant of Fe nitrogenase–expressing R. capsulatus strains. The ability to use light-derived ATP to drive the irreversible reduction of CO₂ to HCOO⁻ differentiates the Fe nitrogenase from FDH. FDH catalyzes a reversible reaction that is limited by the change in Gibbs free energy for the formation of HCOO⁻ from CO₂, which depends on the reduction potential of the electron donor. In case of the reduced form nicotinamide adenine dinucleotide, this results in a HCOO⁻ concentration of around 10 μM HCOO⁻ [pH 6.8, ionic strength of 0.25 M, c(CO₂) = 2.86 mM, and concentrations of other reactants set to 1 mM] (46). Stronger reducing agents such as ferredoxins can shift the equilibrium concentration of HCOO⁻ to higher concentrations. In contrast, the Fe nitrogenase reduces CO₂ irreversibly by coupling the reaction to ATP hydrolysis.

Because our model organism, R. capsulatus, is a phototrophic organism, ATP is produced from light energy. Hence, R. capsulatus or similar phototrophic organisms could serve as a chassis for the sustainable, light-driven in vivo reduction of CO₂ by the Fe nitrogenase. This approach opens alternative pathways for converting carbon waste into value-added compounds such as HCOO⁻ and hydrocarbons. The use of HCOO⁻ as a feedstock chemical for microorganisms was conceptualized by Bar-Even and coworkers as the formate bioeconomy (47–49). Recent highlights in the field include bioplastic formation in Escherichia coli using CO₂ and HCOO⁻ as sole carbon and energy sources (50). This emphasizes that a sustainable and fully biological processes for carbon-neutral chemical production can be designed using Fe nitrogenase–driven HCOO⁻ formation. Moreover, improving nitrogenases’ ability to reduce CO₂ directly to CH₄, C₂H₄, and C₂H₆ would allow the direct carbon-neutral synthesis of fuels and bulk chemicals. Thus, nitrogenases potentially offer solutions for transitioning into a circular economy.

MATERIALS AND METHODS

Chemicals

Unless noted otherwise, all chemicals were purchased from Carl Roth GmbH+ Co. KG (Karlsruhe, Germany), Thermo Fisher Scientific Inc. (Waltham, USA), Sigma-Aldrich (St. Louis, USA), or Tokyo Chemical Industry Deutschland GmbH (Eschborn, Germany) and were used directly without further purification. Gases were purchased from Air Liquide Deutschland GmbH (Düsseldorf, Germany).

Strains, plasmids, and growth conditions

All strains used in this study are listed in table S1. R. capsulatus is cultivated in R. capsulatus minimal medium [RCV; containing 30 mM dl-malic acid, 0.8 mM MgSO₄, 0.7 mM CaCl₂, 0.05 mM sodium ethylenediaminetetraacetic acid (Na₂EDTA), 0.03 mM thiamine hydrochloric acid, 9.4 mM K₂HPO₄, 11.6 mM KH₂PO₄, 120 μM FeSO₄, 45 μM B(OH)₃, 9.5 μM MnSO₄, 0.85 μM ZnSO₄, and 0.15 μM Cu(NO₃)₂ at a pH set to 6.8; omitting an N-source for diazotrophic growth or supplemented with 10 mM glutamate as an N-source. Mo nitrogenase–expressing strains were supplemented with 10 mM molybdenum] in sealed 22-ml glass vials or on complex peptone-yeast (PY) medium [containing peptone (10 g/liter), yeast extract (0.5 g/liter), and agar-agar (15 g/liter) after autoclaving 2 mM MgCl₂, 2 mM CaCl₂, and 80 μM FeSO₄ are added.] agar plates under anaerobic and phototrophic growth conditions (illuminated by 6 × 60–W light bulbs or custom-built light-emitting diode panels (λ: 850, 420LED and 470 nm, ~60-μmol photons m^–2 s⁻¹) at 30°C adapted from Katzke et al. (51) The culture medium was supplemented to a final concentration of kanamycin sulfate (50 μg/ml; Km) or streptomycin sulfate (20 μg/ml; Sm) when appropriate.

The plasmids used are listed in table S2. All primers used in this study were purchased from Eurofins Genomics (Ebersberg, Germany) and are listed in table S3. Genetic modifications of R. capsulatus were done with the sacB method described previously (52). Plasmids for modification of the respective loci (pMM0092 for the deletion of fdhABCD, pMM0152 for the insertion of modABC, and pMM0205 for the insertion of nifD) were introduced into the respective R. capsulatus strain via conjugation as described by Katzke et al. (51), selecting for the Km resistance conferred by the suicide vector. Single-recombinant clones derived from single colonies of the previous step were passaged three times in liquid PY medium, growing each passage for 24 hours at 30°C and moderate shaking under chemoheterotrophic conditions. The final passage was spread on a PY agar plate containing 5% (m/V) sucrose. The plate was incubated for 72 hours at 30°C under a closed atmosphere that was anaerobised using Oxoid AnaeroGen 3.5L sachets (Thermo Fisher Scientific Inc., Waltham, USA) and illumination by six 60-W krypton lamps (Osram Licht AG, Munich, Germany). Single colonies of R. capsulatus growing on the sucrose-containing agar plate were screened for Km and sucrose sensitivity on PY plates containing Km (50 μg/ml) or 5% (m/V) sucrose, respectively. Colonies that could tolerate sucrose but were not growing on Km-containing agar plates were further investigated via colony polymerase chain reaction (PCR) to check the targeted genomic locus (fdhABCD deletion: P19, P20/21; modABC insertion: P22, P23/24; nifD insertion: P25, P26/27). Last, the purified PCR products were analyzed by Sanger sequencing (Microsynth Seqlab GmbH, Göttingen, Germany) to identify successful knockout clones. Plasmids for the recombinant expression of nitrogenase were introduced into R. capsulatus via conjugation.

PCRs were conducted with Q5 High-Fidelity DNA Polymerase (New England Biolabs, Ipswich, USA), PCR purifications with the Monarch PCR & DNA Cleanup Kit (New England Biolabs, Ipswich, USA), extraction of genomic DNA with the Monarch Genomic DNA Purification Kit (New England Biolabs, Ipswich, USA), Gibson assemblies with the NEBuilder HiFi DNA Assembly Master Mix (New England Biolabs, Ipswich, USA), restriction digest ligation cloning with T4 DNA Ligase (New England Biolabs, Ipswich, USA), and Golden Gate cloning with the NEBridge Golden Gate Assembly Kit (New England Biolabs, Ipswich, USA) according to the instructions provided by the manufacturer. The successful assembly of desired vectors was verified by Sanger sequencing through Microsynth Seqlab GmbH (Göttingen, Germany).

For the generation of plasmid pMM0156, pK18mobSacB was linearized by PCR using Primers P1 and P2. The nifHDK operon was PCR-amplified from R. capsulatus genomic DNA (strain B10S) with P3 and P4. The purified fragments were cloned via Gibson assembly, and the reaction mix was used to transform chemo-competent DH5α cells. The outgrowth was plated on a LB agar plate containing Km (50 μg/ml), X-gal (20 μg/ml), and 100 μM isopropyl-ß-d-thiogalacto-pyranoside (IPTG) for selection.

The plasmid pMM0092 was generated from PCR amplicons from two homologs regions [~500 base pair (bp)] from R. capsulatus genomic DNA (strain B10S) up- and downstream of fdhABCD that were amplified with the primers P7, P8, P9, and P10. Plasmid pK18mobSacB was linearized via PCR with the primers P5 and P6. The primers introduced terminal Bsa I cutting sites with the required 4-bp overhangs. All amplicons were purified and assembled by Golden Gate Cloning. The reaction mix was used to transform chemo-competent DH5α cells. For selection, the outgrowth was plated on LB agar plates containing Km (50 μg/ml), X-gal (20 μg/ml), and 100 μM IPTG.

The plasmid pMM0205 was generated by introducing the modABC operon into pMM0106. The modABC gene was PCR-amplified from genomic DNA (strain B10S) with primers P11 and P12. The amplicon was purified and then cloned into plasmid pMM0106 (pK18mobSacB with Bsa I cutting sites) by Golden Gate Cloning. The reaction mix was used to transform DH5α cells. The outgrowth was plated on LB agar plates containing Km (50 μg/ml) for selection.

To generate pMM0132, the nifHDK operon was PCR-amplified from genomic DNA (B10S) using the primers P13 and P14. After purification, the amplicon was cloned into pOGG024-Km^R by Golden Gate Cloning. Chemo-competent DH5α cells were transformed with the reaction mix, and the outgrowth was plated on LB agar plates supplemented with Km (50 μg/ml) for selection.

The plasmid pMM0190 was generated from pMM0132 by introducing an N-terminal 6x His-Tag at anfH and a C-terminal StrepTag II at nifD by restriction-free cloning using primers P15, P16, P17, and P18 (53).

Protein production and purification

The Fe nitrogenase and Mo nitrogenase were purified by affinity chromatography. The Fe nitrogenase was purified according to established protocols (8). For purification of the Mo nitrogenase, the R. capsulatus expression strain (MM0480) was inoculated on PY agar plates containing Km (50 μg/ml) to select for the expression plasmid (pMM0207). The plates were incubated phototrophically for 48 hours at 32°C under an Ar atmosphere. Obtained cell mass was used to inoculate liquid cultures in N₂-flushed RCV-Mo medium containing 30 mM dl-malic acid, 0.8 mM MgSO₄, 0.7 mM CaCl₂, 0.05 mM sodium Na₂EDTA, 0.03 mM thiamine hydrochloric acid, 9.4 mM K₂HPO₄, 11.6 mM KH₂PO₄, 5 mM serine, 1 mM Fe(III) citrate,45 μM B(OH)₃, 9.5 μM MnSO₄, 0.85 μM ZnSO₄, 0.15 μM Cu(NO₃)₂, 10 μM Na₂MoO₄, and Km (25 μg/ml) at a pH set to 6.8. The cultures were cultivated phototrophically at 32°C for 24 hours. Subsequently, the liquid cultures were used to inoculate 800 ml of N₂-flushed RCV-Mo medium to an optical density of 0.1 at 660 nm (OD₆₆₀) for protein production. Protein purification was initiated when the cultures reached an OD₆₆₀ of ~3.0. The isolation of the Mo nitrogenase catalytic and reductase components was performed analogously to the purification procedure that we established previously for the Fe nitrogenase (8). For SDS–polyacrylamide gel electrophoresis (SDS-PAGE), protein samples were denatured at 98°C in Pierce Lane Marker Reducing Sample Buffer (Thermo Fisher Scientific) for 10 min. The sample tubes were centrifuged (17,000g, 1 min), and the supernatant was loaded on a 4 to 20% Mini-PROTEAN TGX Stain-Free Gel (Bio-Rad Laboratories Inc.) including PageRuler Plus Prestained Protein Ladder (Thermo Fisher Scientific) as a molecular weight reference. The electrophoresis was run at a constant voltage of 180 V for 30 min. Bands were visualized by staining with GelCode Blue Safe Protein Stain (Thermo Fisher Scientific).

In vitro nitrogenase assays

The specific nitrogenase activities were measured for the formation of hydrogen (H₂), NH₃, carbon monoxide (CO), methane (CH₄), and formate (HCOO⁻) under varying atmospheres of N₂, argon (Ar), and CO₂. Assays were set up under an Ar atmosphere by adding the nitrogenase reductase component to an anaerobic solution of 50 mM tris (pH = 7.8), 10 mM sodium dithionite, 3.5 mM ATP, 7.87 mM MgCl₂, 44.59 mM creatine phosphate, and creatine phosphokinase (0.20 mg/ml; catalog number: C3755; Sigma-Aldrich St. Louis, USA). The reaction vials were sealed with butyl rubber stoppers, and the headspace was exchanged to 1.2-atm N₂, Ar, or CO₂. For mixed atmospheres, amounts of CO₂ were added to N₂ or Ar as indicated. Following a 10-min incubation at 30°C, the reactions were started by adding 0.1 mg of nitrogenase catalytic component to a final volume of 700 μl. Reactions were allowed to proceed at 30°C and moderate shaking at 200 rpm for 9 min before quenching with 300 μl of 400 mM sodium ethylenediaminetetraacetic acid solution (pH = 8.0).

In vivo nitrogenase assays

R. capsulatus strains were inoculated from a glycerol stock on PY agar plates containing Sm (20 μg/μl). If the strain carried a nitrogenase expression plasmid, then the agar contained Km (50 μg/ml) instead of Sm. Plates were incubated phototrophically for 48 hours at 30°C under an Ar atmosphere. Obtained cell mass was used to inoculate liquid cultures in N₂-flushed RCV medium. For R. capsulatus strains depending on the Mo nitrogenase for diazotrophic growth, 10 μM N₂MoO₄ was added. Subsequently, the headspace of the culture flasks was exchanged to 1.2-atm N₂ to allow for diazotrophic growth. For the complementation experiment with R. capsulatus strains MM0262 and MM0263 (Fig. 5C), the RCV medium was flushed with Ar instead of N₂ and was supplemented with Km (25 μg/ml) and 10 mM glutamine. Moreover, the culture headspace contained Ar instead of N₂. In all cases, the bacteria were cultivated phototrophically for 24 hours at 30°C. Subsequently, RCV cultures (6 ml) were inoculated to an OD₆₆₀ of 0.1. The headspace was exchanged to 1.2-atm Ar, N₂, or 8% CO₂ in Ar. For mixed atmospheres, CO₂ was added as indicated. The liquid cultures were cultivated phototrophically at 30°C, samples for HCOO⁻ quantification were taken at the indicated time points, and the growth was monitored at 660 nm.

Diazotrophic growth curves

R. capsulatus strains were inoculated from freshly grown PY agar plates into N₂-flushed RCV medium (containing 4.7 mM K₂HPO₄ and 5.8 mM KH₂PO₄ at a pH set to 6.8) liquid cultures (6 ml). For R. capsulatus strains depending on the Mo nitrogenase for diazotrophic growth, 10 μM Na₂MoO₄ was added. The headspace of the culture flasks was exchanged to 1.2-atm N₂, and the bacteria were cultivated phototrophically for 48 hours at 30°C. Subsequent inoculation steps were performed strictly anaerobically by working under an Ar atmosphere. The grown cultures were used to inoculate N₂-flushed RCV medium cultures (50 ml) at an OD₆₆₀ of 0.5, which were cultivated phototrophically under 1.2-atm N₂ atmosphere for 48 hours at 30°C. The bacteria were then used to inoculate the growth assay cultures (6 ml of N₂-flushed RCV medium) in 22-ml glass vials to an OD₆₆₀ of 0.1. The headspace of the cultures was exchanged to 1.2-atm N₂, and they were cultivated phototrophically at 30°C. Samples for OD₆₆₀ measurements were taken at the indicated time points by piercing the septum of the cultivation vials with a syringe. The OD₆₆₀ was measured with an Infinite 200 PRO plate reader (Tecan Group Ltd., Männedorf, Switzerland). The doubling time T_d was determined for the exponential growth phase by plotting ln[OD₆₆₀/OD₆₆₀(t = 0)] versus the time of growth using the GraphPad Prism 9 software (Dotmatics, Boston, USA). The data points were interpolated by a linear regression model to determine the growth rate r. T_d was calculated with Eq. 4

$$T_{\mathrm{d}}=\frac{\text{ln}(2)}{r}$$ (4)

Quantification of CO, CH₄, and H₂

Amounts of in vivo or in vitro formed CO, CH₄, and H₂ were determined via headspace analysis using a Clarus 690 GC system [gas chromatography–flame ionization detector/thermal conductivity detector (GC-FID/TCD); PerkinElmer Inc., Waltham, USA] with a custom-made column circuit (ARNL6743) that was operated with the TotalChrom v.6.3.4 software (PerkinElmer Inc., Waltham, USA). The headspace samples were injected by a TurboMatrixX110 (PerkinElmer Inc., Waltham, USA) autosampler, heating the samples to 45°C for 15 min before injection. The samples were then separated on a HayeSep column (7′ HayeSep N 1/8″ Sf; PerkinElmer Inc., Waltham, USA), followed by a molecular sieve (9′ Molecular Sieve 13× 1/8″ Sf; PerkinElmer Inc., Waltham, USA) kept at 60°C. Subsequently, the gases were detected with a FID (at 250°C) and a TCD (at 200°C). The quantification of all substrates was based on a linear standard curve derived from measuring varying amounts of CO, CH₄, and H₂ under identical conditions. The results were plotted using the GraphPad Prism 9 software (Dotmatics, Boston, USA).

Quantification of NH₃

NH₃ was quantified with fluorescence-based assay as described by Corbin (54). Sample (100 μl) was combined with 1 ml of a solution containing 2 mM o-phthalaldehyde, 10% (v/v) ethanol, 0.05% (v/v) β-mercaptoethanol, and 0.18 M potassium phosphate buffer (pH = 7.3) and incubated at 25°C for 2 hours in the dark. Each sample (50 μl) was transferred into individual wells of a black Nunc F96 MicroWell plate (Thermo Fisher Scientific Inc., Waltham, USA), and fluorescence at 485 nm was monitored with an Infinite 200 PRO plate reader (Tecan Group Ltd., Männedorf, Switzerland) in fluorescence top reading mode using an excitation wavelength of 405 nm. The quantification of ammonia was based on a linear standard curve that was derived from measuring varying amounts of NH₄Cl under identical conditions. Samples incubated under an argon atmosphere instead of dinitrogen were used to correct for background signal. The results were plotted using the GraphPad Prism 9 software (Dotmatics, Boston, USA).

Quantification of formate via GC-MS

Formate quantification via GC–mass spectrometry (GC-MS) was performed according to (55). In brief, 0.1 ml of sample was mixed with 0.5 ml of 100 mM pentafluorobenzylbromide solution and 0.1 ml of 500 mM potassium phosphate buffer (pH 6.8). The combined sample was mixed vigorously and then incubated for 1 hour at 60°C while shaking at 500 rpm. Samples were allowed to cool down to room temperature, and 1 ml of 100 μM 1,3,5-tribromobenzene solution in n-hexane was added. The samples were again mixed vigorously before separating the two phases by centrifugation (15 min, 1400g). The organic phase (300 μl) was transferred into a 1.5-ml short thread vial (VWR, Radnor, USA) and analyzed on a Clarus 690 gas chromatograph (split = 1:20), equipped with an Elite-1 column that was coated with dimethylpolysiloxane (60 m, 0.32 mm inner diameter, 5.0-μm film thickness; PerkinElmer Inc., Waltham, USA). The gas chromatograph was connected to a Clarus SQ8 T mass spectrometer (PerkinElmer Inc., Waltham, USA), and the system was operated with the TurboMass GC-MS Software version 6.1.2 (PerkinElmer Inc., Waltham, USA). The column’s initial temperature (50°C) was held for 3 min and then increased to 280°C at a rate of 30°C/min. The injection port and ion source were kept at 220°C. Helium was used as the carrier gas at a flow rate of 1.5 ml/min. Mass spectra were obtained by positive-ion electron ionization (EI) mode scanning every 0.1 s from 40 to 600 m/z. Selected ion recording (SIR) was measured every 0.1 s for the molecular peak ion of the derivative of HCOO⁻ at 226 mass/charge ratio (m/z), and the base peak ion of 1,3,5-tribromobenezene at 314 m/z. The ionization energy of the EI condition was 70 eV. A linear standard curve [R² (coefficient of determination) ≥ 0.998] for HCOO⁻ was generated by plotting the ratio of the peak at 226 and 314 m/z against the concentration of HCOO⁻. The detection limit of HCOO⁻ was 0.03 mM. The results were plotted using the GraphPad Prism 9 software (Dotmatics, Boston, USA).

Purification of FDH

FDH from Pseudomonas sp. 101 used for enzymatic HCOO⁻ quantification was purified according to published protocols (56). In brief, E. coli BL21 DE3 cells expressing N terminally His-tagged FDH from plasmid pTE1390 were inoculated on LB agar plates containing Sm (20 μg/ml). The plates were incubated at 37°C for 24 hours and then used to inoculate liquid cultures in terrific broth medium containing Sm (20 μg/ml). Following 16 hours of incubation at 30°C and moderate shaking (80 rpm), the cells were harvested by centrifugation (15 min at 6000g, 10°C) and then resuspended in 30 ml of binding buffer [20 mM tris, 500 mM NaCl, and 5 mM Imidazole (pH 7.9)] per gram of pellet. Next, the cell suspension was supplemented with bovine pancreatic deoxyribonuclease I (0.2 mg/ml) and incubated for 20 min on ice before lysing the cells by sonication. The lysate was centrifuged (20 min at 4347g, 4°C), and the supernatant was loaded on 3 ml of Protino nickel nitrilotriacetic acid (Ni-NTA) agarose resin in a gravity flow column. Subsequently, the column was washed with 60 ml of washing buffer [20 mM tris, 500 mM NaCl, and 20 mM imidazole (pH 7.9)] before eluting the protein with elution buffer [20 mM tris, 500 mM NaCl, and 250 mM imidazole (pH 7.9)]. The elution fraction was buffer-exchanged to 100 mM Na₂HPO₄ buffer (pH 7.0) using an Amicon Ultra-15 Centrifugal Filter Unit (molecular weight cutoff = 30 kDa; Merck Millipore, Billerica, USA). The obtained protein solution was flash-frozen in liquid N₂ and stored at −70°C until further use. Protein yields were determined using the Quick Start Bradford 1x Dye Reagent (Bio-Rad Laboratories Inc., Hercules, USA) according to the instructions by the manufacturer, and the purity was assessed via SDS-PAGE.

Quantification of formate via FDH assay

The FDH assay reaction mix contained 10 mM nicotinamide adenine dinucleotide (NAD⁺) and purified FDH (0.1 mg/ml) dissolved in 100 mM sodium phosphate buffer (pH 7.0). The reaction mix (97 μl) was added to each well of a 96-well plate, and absorbance at 340 nm (A₃₄₀) was monitored using a CLARIOstar Plus plate reader (shaking plate for 30 s at 400 rpm before each measurement and assay temperature set to 30°C; BMG labtech, Ortenberg, Germany). Once A₃₄₀ remained stable, 3 μl of a sample was added to each well using a VIAFLO 96 pipetting station (Integra, Biebertal, Germany), and monitoring of A₃₄₀ was continued until it remained constant. The A₃₄₀ value was used to quantify HCOO⁻ based on a standard dilution series of sodium formate ranging from 2.5 to 30 mM that was quantified identically. The results were plotted using the GraphPad Prism 9 software (Dotmatics, Boston, USA).

¹³CO₂-labeling experiment

Cultures of R. capsulatus strains were prepared as described for the in vivo activity assays. Following inoculation at OD₆₆₀ of 0.1 and exchange of the culture headspace to N₂, 20% (v/v) ¹³CO₂ (99.0 atom % ¹³C) or ¹²CO₂ (¹³C at natural abundance) was added to the culture vial headspace. The cultures were cultivated phototrophically at 30°C for 6 days. The cultures (5 ml) were centrifuged (4347g, 15 min), and 4 ml of the supernatant was lyophilized with an Alpha 1-4 LSCplus freeze dryer (Martin Christ Gefriertrocknungsanlagen GmbH, Osterode am Harz, Germany). The remaining solids were dissolved by vortexing in 1 ml of D₂O containing 3-(trimethylsilyl)propionic-2,2,3,3-acid sodium salt D₄ (8.33 mg/ml) as an internal standard. The solution was centrifuged (16,000g, 10 min), and 0.6 ml was transferred into glass tubes. NMR spectra were measured at 298 K with an Avance II 300 MHz (¹H 300 MHz) NMR spectrometer (Bruker, Billerica, USA). The spectra were calibrated to the residual signal of HOD (δ = 4.79 ppm). The ¹³C/¹²C ratio of HCOO⁻ in the culture supernatant was derived by dividing the NMR peak area of H¹³COO⁻ [¹H NMR (300.0 MHz, D₂O): δ 8.44 (d, ¹J(H,¹³C) = 195 Hz, 1H] ppm) by the peak area of H¹²COO⁻ [¹H NMR (300.0 MHz, D₂O): δ 8.44 (s, 1H) ppm].

Statistics

The statistical analysis was performed in the GraphPad Prism 10 software. Error bars represent the means ± SD. Comparison between more than two groups was analyzed by ordinary one-way analysis of variance (ANOVA) using the Tukey’s multiple comparisons test. P values of <0.05 were considered significant.

Supplementary Materials

This PDF file includes:

Figs. S1 to S4

Tables S1 to S3

References