Carbon substrate re-orders relative growth of a bacterium using Mo-, V-, or Fe-nitrogenase for nitrogen fixation.

Biological nitrogen fixation is catalyzed by the molybdenum (Mo), vanadium (V) and iron (Fe)-only nitrogenase metalloenzymes. Studies with purified enzymes have found that the 'alternative' V- and Fe-nitrogenases generally reduce N2 more slowly and produce more byproduct H2 than the Mo-nitrogenase, leading to an assumption that their usage results in slower growth. Here we show that, in the metabolically versatile photoheterotroph Rhodopseudomonas palustris, the type of carbon substrate influences the relative rates of diazotrophic growth based on different nitrogenase isoforms. The V-nitrogenase supports growth as fast as the Mo-nitrogenase on acetate but not on the more oxidized substrate succinate. Our data suggest that this is due to insufficient electron flux to the V-nitrogenase isoform on succinate compared with acetate. Despite slightly faster growth based on the V-nitrogenase on acetate, the wild-type strain uses exclusively the Mo-nitrogenase on both carbon substrates. Notably, the differences in H2 :N2 stoichiometry by alternative nitrogenases (~1.5 for V-nitrogenase, ~4-7 for Fe-nitrogenase) and Mo-nitrogenase (~1) measured here are lower than prior in vitro estimates. These results indicate that the metabolic costs of V-based nitrogen fixation could be less significant for growth than previously assumed, helping explain why alternative nitrogenase genes persist in diverse diazotroph lineages and are broadly distributed in the environment.

Introduction

Biological nitrogen fixation has sustained ecosystem fertility and primary production throughout Earth history. This fundamental prokaryotic process reduces atmospheric dinitrogen (N2) to ammonia (NH3) in a metabolically costly reaction catalysed by the metalloenzyme nitrogenase (Nase). There are three phylogenetically and structurally distinct isoforms of Nase: the molybdenum (Mo‐) and ‘alternative’ vanadium (V‐) and iron (Fe‐) only Nases (Eady, 1996). These can be differentiated by a key active site metal atom (Mo, V, or Fe) in addition to their kinetics, stable isotope fractionation and reaction stoichiometry (Bishop et al., 1986; Robson et al., 1986; Eady, 1996; Gosse et al., 2010; McKinlay and Harwood, 2010b; Zhang et al., 2014; Mus et al., 2018). The metabolic costs of Mo‐based diazotrophy are well characterized, including the direct energy and reducing power requirements of N2 fixation and indirect costs related to oxygen protection (Andersen and Shanmugam, 1977; McKinlay and Harwood, 2010b; Großkopf and LaRoche, 2012; Inomura et al., 2016). The Mo‐Nase is considered the most efficient isoform and the primary form responsible for N2 fixation in the environment. However, alternative V‐ and Fe‐Nase genes are widespread (Betancourt et al., 2008; McRose et al., 2017) and active (Hodkinson et al., 2014; Zhang et al., 2016; Darnajoux et al., 2017, 2019; McRose et al., 2017) in diverse taxa and environments, prompting questions on their roles in ecosystem nitrogen and trace metal cycling. To better understand the persistence of Nase diversity and the controls on environmental N2 fixation, it is necessary to study how alternative N2 fixation affects the growth physiology of diazotrophs (Glazer et al., 2015).

Compared with the Mo‐Nase, less is known about the cellular costs of alternative V‐ and Fe‐Nase‐based diazotrophy. All Nase isoforms produce H2 as an obligatory byproduct of N2 reduction (Simpson and Burris, 1984; Harris et al., 2018a; Harris et al., 2018b). Early in vitro data indicated that the alternative V‐ and Fe‐Nases produce more H2 than the Mo‐Nase, resulting in greater energy and reducing power requirements (Eady, 1996). It has also been suggested that V‐ and Fe‐only dependent diazotrophy are more oxygen (O2) sensitive than Mo‐based diazotrophy (Robson et al., 1986; Krahn et al., 1996), possibly leading to greater indirect metabolic costs or more restrictive anaerobic lifestyles. These findings are consistent with observations that all diazotrophic organisms possess the Mo‐Nase (Dos Santos et al., 2012) and only use the alternative Nases under Mo limited conditions or genetic Mo‐Nase inactivation (Jacobson et al., 1986; Luque and Pau, 1991; Schneider et al., 1991; Gollan et al., 1993; Oda et al., 2005; Hamilton et al., 2011; Demtröder et al., 2019). However, the effects of these observations on organism fitness remain poorly characterized, and the assumption that Mo‐based diazotrophy is always superior is challenged by field experiments demonstrating that several diazotrophic isolates grew faster in the presence of vanadium than molybdenum (Betancourt et al., 2008).

Here, we investigate the ability of the V‐ and Fe‐Nases to fulfill nitrogen requirements and enable competitive growth under photoheterotrophic conditions in a genetically tractable model organism, Rhodopseudomonas palustris CGA009 (Larimer et al., 2004). During photoheterotrophic growth, light is the source of energy and organic substrates are the source of carbon and reductant for biosynthesis. Reductant in cellular electron carrier pools (e.g. NAD(P)H and reduced ubiquinone) accumulates during oxidative processing of organic substrates (Fig. 1, Arrow B) when external electron acceptors are absent (Fig. 1, Arrow G). To maintain redox homeostasis and enable continued substrate assimilation, these reductants must be re‐oxidized through direct biomass synthesis (Fig. 1, Arrow D), Nase activity (Fig. 1, Arrow F), or re‐fixation of some of the CO2 produced during substrate assimilation (Fig. 1, Arrow E), such that electrons from the organic substrate are fully allocated between biomass, H2 and CO2. Under these conditions, Nase is important both for nutrient acquisition and as an electron sink (Muller, 1933; Hillmer and Gest, 1977; McKinlay and Harwood, 2010a, 2011; Farmer et al., 2014; Gordon and McKinlay, 2014; McCully and McKinlay, 2016). To determine how simultaneous nitrogen acquisition and electron balancing constraints influence growth based on each Nase isoform, we cultured wild type R. palustris and mutant strains known to utilize only a single isoform (hereafter Mo‐Nase, V‐Nase and Fe‐Nase strains, Oda et al., 2005) in diazotrophic photoheterotrophic media with organic substrates of varying oxidation states and assimilation pathways (Table 1) and measured growth, N2 and CO2 fixation, H2 production, biomass composition and global protein abundance patterns. Our results demonstrate and explain why, depending on the carbon substrate, the alternative V‐Nase isoform can support photoheterotrophic growth as fast as the Mo‐Nase isoform.

Figure 1

Electron balancing processes used by during diazotrophic, photoheterotrophic growth. During photoheterotrophic growth, consumes carbon substrates and converts them into biomass precursors and CO2 (‘central metabolism,’ arrows A and C, respectively). This process reduces cellular electron carriers (arrow B), which must be re‐oxidized to maintain intracellular redox homeostasis and thereby enable continued carbon substrate assimilation. The electron sinks that stabilize the cellular electron pool are re‐fixation of some of the CO2 produced during substrate assimilation (arrow E), synthesis of certain biomass precursors (arrow D), and Nase activity (arrow F). To achieve redox homeostasis, substrate‐derived electrons must be fully allocated between biomass, H2 (arrow I) and CO2 (arrow H). The cellular electron pool can also be re‐oxidized by extracellular electron acceptors like dimethyl sulfoxide (DMSO, arrow G).

Table 1

Oxidation states of biomass and carbon substrates.

Substrate	Formula	Oxidation statea
Succinate	C4H6O4	+0.5
Acetate	C2H4O2	0
Measured biomassb	CH1.75N0.18O0.54	−0.13
Published biomassc	CH1.8N0.18O0.38	−0.5
Butyrate	C4H8O2	−1

Average oxidation state of carbon in the compound calculated according to Dick (2014).

Average biomass composition of all strains on succinate and acetate.

The elemental composition of 42OL during photoheterotrophic growth with acetate and ammonium (Carlozzi and Sacchi, 2001).

Results and discussion

Alternative V‐Nase can support diazotrophic growth as fast as the Mo‐Nase on acetate

Previous studies have shown that photoheterotrophic growth of R. palustris on succinate, acetate, and butyrate is associated with increasing cellular electron burden and redox stress (McKinlay and Harwood, 2011). To determine how these substrates influence growth based on different Nase isoforms, we compared the growth of the wild type strain, which fixed nitrogen exclusively with the Mo‐Nase isoform in all three carbon substrates (see Supporting Information Discussion and Figs S1, S2, and S3), with that of mutant strains using exclusively the V‐ or Fe‐Nase. In our experiments, when provided with succinate as the carbon and electron source, the V‐Nase strain grew ~20% slower than the Mo‐Nase utilizing wild type strain (0.57 ± 0.02 day−1 vs. 0.70 ± 0.01 day−1, p < 10−5, errors are SE, Fig. 2). Similarly, in the mixed butyrate‐bicarbonate treatment, which combines a very reduced and oxidized carbon substrate, the V‐Nase strain grew ~20% slower than the wild type strain (0.66 ± 0.01 day−1 vs. 0.82 ± 0.02 day−1, p ~ 10−8). These growth rate patterns are consistent with prior observations (Oda et al., 2005) and the general view that nitrogen fixation with the Mo‐Nase isoform yields the fastest growth. In contrast, when provided with acetate, we found that the V‐Nase strain grew ~10% faster than Mo‐Nase utilizing wild type strain (0.80 ± 0.01 day−1 vs. 0.74 ± 0.01 day−1, p = 10−2). The V‐Nase strain also had a higher maximal optical density (OD660) than the wild type strain for growth on acetate but not succinate (Fig. S4). To our knowledge, these are the first data showing that, depending on the carbon substrate, diazotrophic growth with the alternative V‐Nase can rival growth based on the canonical Mo‐Nase isoform.

Figure 2

A. R. palustris growth rates. The wild type strain, which uses the Mo‐Nase isoform, grows fast in all treatments. The V‐Nase strain grows faster than the Mo‐Nase utilizing wild type strain when provided with the more reduced carbon substrate, acetate, but not with the more oxidized substrate, succinate. It also grows faster than the Mo‐Nase mutant strain on acetate and butyrate + bicarbonate. B. Growth rate differences are linked to electron balance. The addition of an external electron acceptor, dimethyl sulfoxide (DMSO), diminishes the growth advantage of the V‐Nase strain on the more reduced carbon substrate acetate. In the absence of DMSO, the growth rates are significantly different within each carbon treatment (p < 0.05). However, in the presence of DMSO, the growth rates of the Mo‐ and V‐Nase strains are not significantly different. The error bars show the mean ± SE. C. R. palustris protein abundance dendrograms. Hierarchical clustering of all detected proteins and of the proteins annotated to COG H (coenzyme transport and metabolism) mimics the observed growth rate trends, demonstrating that, physiologically, on acetate, the wild‐type strain is more similar to the V‐Nase strain than the Mo‐Nase strain. From an energy acquisition perspective (COG C), the wild‐type strain is more similar to the V‐Nase strain than the Mo‐Nase strain on both succinate and acetate.

Metabolic conditions influence the rate of N

To understand why the V‐Nase isoform enables growth comparable to the Mo‐Nase isoform on some carbon substrates but not others, we measured biomass composition (Fig. 3), CO2 and N2 fixation rates (Fig. 4), and global protein abundance patterns on acetate and succinate for all strains. Despite indistinguishable V‐Nase protein abundance (Fig. S1; log2 abundance difference < 0.3, p > 0.3 for VnfHDKG), the V‐Nase strain reduces N2 approximately 30% more slowly on succinate than on acetate (Fig. 4, p = 0.31). Its biomass N:C ratio is comparable to the wild type strain on acetate but is substantially lower on succinate (Fig. 3). This is notable because, in our experiments, higher biomass N:C ratios are associated with faster growth (Fig. 3), suggesting that the growth limiting process is either nitrogen acquisition itself and/or a process correlated with nitrogen acquisition, such as electron availability and balance. Importantly, there were no statistically significant differences in Nase abundance between growth on succinate and acetate for all strains (Fig. S1), demonstrating that the observed growth rate trends are due to differences in the rate and efficiency of the enzyme and its integration into metabolism rather than Nase abundance. The different growth pattern and N2 reduction rate of the V‐Nase strain on acetate and succinate relative to the Mo‐Nase utilizing wild type strain demonstrate that the carbon substrate used for growth can alter the relative rate of nitrogen acquisition by the V‐Nase isoform.

Figure 3

R. palustris biomass composition. The fastest‐growing strains are also the most nitrogen rich. The wild type and V‐Nase strains adjust their growth rates depending on the carbon substrate, whereas the Fe‐ and Mo‐Nase strains grow at similar rates in all treatments. Error bars show the mean ± SE.

Figure 4

A. Rates of N2 and CO2 fixation and H2 production of strains growing on succinate (left) and acetate (right). For H2 production, the error bars represent the standard error of the slope when data from multiple biological replicates (n ≥ 2) are plotted together. For CO2 and N2 fixation rates, the error bars represent the standard error of multiple biological replicates (n ≥ 2). Different letters represent treatments that are significantly different at p < 0.05. B. Electron flux through Nase and Rubisco. The fraction of the electron flux through Nase versus Rubisco remains approximately constant despite differences in total flux and in the partitioning of electrons through Nase to proton versus N2 reduction. C. Growth rate versus nitrogen fixation rate and total electron flux. The nitrogen fixation rate and the reductive electron flux through Nase and Rubisco scale with the growth rate. Error bars show the mean ± SE. Data for the Fe‐Nase strain are provided in the Supporting Information.

Mo‐Nase strain grows slowly compared with the wild type strain

We also grew a Mo‐Nase mutant strain. This strain grew ~10% more slowly than the wild type strain when succinate was the carbon source (0.65 ± 0.01 day−1 vs. 0.70 ± 0.01, p ~ 3 × 10−3, Fig. 1). On acetate, it grew ~25% more slowly than the V‐Nase strain (0.60 ± 0.02 day−1 vs. 0.80 ± 0.01 day−1, p < 10−9) and ~20% more slowly than the wild type strain (0.74 ± 0.01 day−1, p ~ 10−9). This result is surprising because the wild type strain uses exclusively the Mo‐Nase to fix nitrogen (see Supporting Information Discussion), and the Mo‐Nase strain mutations are in proteins that are not used by the wild type strain under these conditions (the alternative nitrogenase structural protein genes vnfH and anfH, Fig. S1). The Mo‐Nase mutant strain has a somewhat lower abundance of several Mo‐Nase accessory proteins (including NifW, NifS2 and NifA; Fig. S3) and higher abundance of glutamate and ammonium transporters (Fig. S5) relative to the wild type strain, possibly indicating altered intracellular fixed nitrogen availability. We cannot exclude the possibility that there are additional random mutations in the Mo‐Nase strain compared with the wild type strain, although similar growth phenotypes and gene expression profiles were previously observed for these strains during growth on succinate at 30°C (Oda et al., 2005). Thus, the reasons for the different growth rates of these strains in our experiments are unclear. From a mechanistic perspective, the slow CO2 re‐fixation rate of the Mo‐Nase strain on acetate is notable, which could be due to its lower, more oxidized, cellular [NADH]/[NAD+] ratio (Fig. S6) and lower Form II Rubisco abundance on acetate (Fig. S5). This is consistent with measurements of particularly high bacteriochlorophyll a concentrations in the Mo‐Nase strain (Figs. S5 and S7) and a lower ratio of the more efficient light‐harvesting 4 (LH4) complex to the light‐harvesting 2 (LH2) complex (Table S1, Fixen et al., 2019), which could reflect an insufficient rate of reverse electron flow to regenerate NADH for continued carbon assimilation (Hädicke et al., 2011). However, it remains unclear how mutations in the V‐Nase vnfH and Fe‐Nase anfH genes would cause this effect. Regardless of the underlying mechanism, the slow growth of the Mo‐Nase strain on acetate demonstrates that the alternative nitrogenase genes are well integrated into the genome and possibly regulatory pathways of R. palustris CGA009 and that other metabolic constraints can be more important determinants of fitness than Mo‐ versus V‐Nase isoform usage.

Diazotrophic growth on succinate and acetate are differentiated by electron availability

Previous in vitro studies have shown that electron availability, which can be manipulated by varying the dinitrogenase reductase versus dinitrogenase protein ratio, can influence the efficiency of Nase, including the specific N2 reduction rate and the H2:N2 ratio (Eady, 1996; Harris et al., 2019). For R. palustris, the key determinants of whether it is limited by electron acceptors or donors are (i) the oxidation state of the substrate relative to the biomass (Table 1) and (ii) the pathway by which substrate is assimilated into biomass (McKinlay and Harwood, 2010a, 2011). Acetate is assimilated via the glyoxylate shunt (indicated in our data by high expression of isocitrate lyase in acetate‐grown cells, Fig. S8), resulting in rapid electron availability for energy acquisition and carbon fixation in R. palustris (Alsiyabi et al., 2019). In contrast, succinate is assimilated through the oxidative TCA cycle, which is associated with greater CO2 loss (McKinlay and Harwood, 2011) and reduced electron flow through the energy acquiring electron transport chain (Alsiyabi et al., 2019). Proteins related to branched‐chain amino acid metabolism, whose synthesis can re‐oxidize cellular electron carriers for R. palustris during photoheterotrophic growth on succinate but not on acetate (McCully et al., 2019), are more abundant in this treatment (Fig. S8). The similar abundance changes in a wide range of proteins related to carbon metabolism and electron balance (Fig. S8) demonstrate that, despite differences in Nase isozyme usage and growth rates, the metabolic shift from diazotrophic growth on succinate to acetate is broadly consistent between strains and compatible with a greater cellular electron burden during growth on acetate (McKinlay and Harwood, 2011).

Excess H

H2 is an obligatory byproduct of biological N2 fixation. In vitro experiments in the 1980s and 1990s led to the conclusion that the alternative V‐Nase produces approximately threefold more H2 per N2 reduced than the Mo‐Nase (Table 2; literature compilation in Table S2; Bishop et al., 1986; Dilworth et al., 1993; Eady, 1996; Schneider et al., 1997; Gosse et al., 2010; McKinlay and Harwood, 2010b; Mus et al., 2018). As the H2:N2 ratio increases, Nase supplies less fixed nitrogen relative to its contribution as an electron sink. This observation underpins the long‐standing paradigm that use of the alternative Nases always results in slower diazotrophic growth than the Mo‐Nase (Schubert and Evans, 1976; Eady, 1996).

Table 2

H2:N2 ratios in .

	Succinate	Acetate	Literaturea
Wild type strain	1.4 ± 0.3	0.9 ± 0.2	1
Mo‐Nase strain	1.1 ± 0.2	1.0 ± 0.2	1
V‐Nase strain	1.7 ± 0.6	1.4 ± 0.5	3
Fe‐Nase strainb	6.9 ± 0.8	3.9 ± 0.4	7.5–9

The V‐Nase strain has a lower H2:N2 ratio than reported previously. Errors are SEs. The differences between the wild type, Mo‐Nase and V‐Nase strains are not statistically significant.

See e.g. Eady (1996) and Gosse and colleagues (2010).

Fe‐Nase ratio was calculated by adjusting 15N2‐fixation rates measured in serum vials to Balch Tubes by 1.5, the average difference between rates in serum vials and Balch tubes for the other strains (see Supporting Information Experimental Procedures).

To evaluate this hypothesis, we measured H2 production in each treatment (Table 2; Fig. 4). During growth on succinate, our data are compatible with the classical view that excess H2 production by the V‐Nase strain leads to slower growth compared with the Mo‐Nase utilizing strains. For example, it has a lower biomass N:C ratio despite a higher abundance of the primary electron donor to Nase, the ferredoxin protein Fer1 (log2 abundance difference of 0.53 ± 2.56 relative to the wild type strain, p < 0.01; Fixen et al., 2018). These data indicate that, on succinate, nitrogen acquisition is more limiting to growth than electron balance. In contrast, when grown on acetate, the V‐Nase strain produced a similar amount of excess H2 (Table 2) compared with the wild type strain, yet grew just as fast, demonstrating that excess H2 production by the V‐Nase isoform is not always disadvantageous for growth. The strong correlation between growth rate and both N2 fixation rate and the total electron flux through Nase and Rubisco (Fig. 4C) for V‐ and Mo‐based nitrogen fixation suggests that the tradeoff between nitrogen acquisition and electron balancing, which must be co‐optimized for optimal growth, depends on broader metabolic conditions like electron availability. Taken together, our data show that the physiological impact of excess H2 production on growth varies depending on the metabolic conditions.

In our experiments, H2 production directly reflects Nase activity because there is no functional uptake hydrogenase in R. palustris CGA009 (Rey et al., 2006; confirmed with proteomics, see Supporting Information). Our direct measurements of N2 reduction and H2 production indicate that the V‐Nase H2:N2 ratio can be as low as ~1.5, half of the canonical 3:1 ratio (Table 2). In fact, recent evidence has confirmed that all three forms of Nase use an analogous catalytic mechanism with the same minimum 1 H2: 1 N2 stoichiometry for N2 binding to the active site (Lukoyanov et al., 2015, 2016; Harris et al., 2018b; Harris et al., 2019). Many of the measurements that led to the canonical 3:1 ratio for V‐Nase were conducted at 30°C (Bishop et al., 1986; Dilworth et al., 1993) but there are hints that the H2:N2 ratio may be temperature dependent and that it is lower at 19°C, the temperature where our growth experiments were performed (Table S2; Miller and Eady, 1988). Additionally, as temperatures decrease, the V‐Nase isoform's in vitro N2 (Miller and Eady, 1988) and in vivo acetylene (Darnajoux, 2015) reduction rates decrease less than that of the Mo‐Nase. This could help explain why Oda and colleagues (2005) observed a larger gap between V‐ and Mo‐Nase based growth rates on succinate (~30%) than observed here (~20%). Thus, as suggested by Miller and Eady (1988) over three decades ago, temperature could be a key parameter modulating the relative favourability of fixing nitrogen with the Mo‐ or V‐Nase isoforms.

Fe‐Nase strain grows slowly because of excess H

Unlike the V‐Nase strain, the Fe‐Nase strain grew slowly on all three carbon substrates (0.45 ± 0.01 day−1 on succinate, 0.46 ± 0.01 day−1 on acetate and 0.48 ± 0.01 day−1 on butyrate‐bicarbonate). It produced much more H2 than the other strains, accumulating up to 20% v/v H2 in the culture headspace. Although it has a comparable total electron flux through Nase (Fig. S9), it has the lowest biomass N:C ratios of all strains and a much lower N2 reduction rate (Fig. 3, Fig. S9C), demonstrating strong nitrogen limitation. In addition, the Fe‐Nase strain has the lowest intracellular [NAD(P)H]/[NAD(P)+] ratios (Fig. S6) and expresses more electron acquisition proteins than other strains, like formate dehydrogenase and the electron bifurcating protein complex FixABC used by organisms to maximize energy conservation (Fig. S5; also see Oda et al., 2005). These data indicate that the Fe‐Nase isoform over‐allocates electron flux to H2 production at the expense of nitrogen acquisition, resulting in slower growth of the Fe‐Nase strain relative to the other strains. In the environment, it is possible that this stark growth disadvantage could be mitigated by the presence of a functional uptake hydrogenase. Indeed, R. palustris CGA009 has a non‐functional uptake hydrogenase due to a frameshift mutation in HupV (Rey et al., 2006), and early chemostat experiments with Azotobacter vinelandii showed similar growth rates on the Mo‐ and Fe‐Nases when an uptake hydrogenase was present (Bishop et al., 1986). Taken together, our data show that the effect of excess H2 production on growth depends on its magnitude as well as the tradeoffs between electron availability, balancing and nitrogen acquisition during growth under different metabolic conditions.

Fixed relative electron fluxes through Nase and Rubisco

During diazotrophic, photoheterotrophic growth, the allocation of substrate‐derived electrons between biomass, H2 and CO2 reflects the total electron fluxes through Nase and Rubisco (Fig. 1). Nase functions in both nitrogen acquisition and electron balance, while the primary role of CO2 fixation during photoheterotrophic growth is as an electron sink (Hallenbeck et al., 1990; Falcone and Tabita, 1991; Hädicke et al., 2011; Farmer et al., 2014; Gordon and McKinlay, 2014). Total electron flux through Nase and Rubisco increases with the growth rate for the wild type, Mo‐Nase and V‐Nase strains, but is always high for the Fe‐Nase strain (Fig. 4C and Fig. S9). The addition of the exogenous electron acceptor dimethyl sulfoxide (DMSO, 40 mM) removes the growth rate differences between the Mo‐Nase and V‐Nase mutant strains (Fig. 2B), indicating the use of the V‐Nase isoform as an electron sink can promote growth.

Strikingly, despite large differences (up to ~1.75‐fold) in growth rate and biomass nitrogen content between strains (Fig. 3), the relative electron fluxes through Nase and Rubisco remain constant, averaging ~60% Nase for all strains (Table 3). This means that the electron supply from the substrate is allocated at fixed proportions to Nase and Rubisco irrespective of growth rates and the extent of biomass nitrogen limitation (Fig. 3). This observation suggests that, as long as both proteins are present, cellular metabolites like ATP or electron carriers, involved in energy metabolism rather than nutrient status, control the relative rates of carbon and nitrogen fixation in photoheterotrophic R. palustris. It is unclear why biomass nitrogen limitation, as is the case for the Fe‐Nase strain on both carbon substrates, does not result in the reallocation of electrons from Rubisco to Nase. This result suggests that, for photoheterotrophy, there could be selective pressures to maintain a carbon fixation flux even at the expense of nitrogen acquisition and growth rate.

Table 3

Percentage of the reductive electron flux allocated to H2 and Nase.

	Succinate		Acetate
	%H2 of Nase	%Nase of total	%H2 of Nase	%Nase of total
Wild type strain	32 ± 8	63 ± 11	23 ± 6	64 ± 8
Mo‐Nase strain	27 ± 6	63 ± 7	25 ± 5	58 ± 12
V‐Nase strain	36 ± 13	56 ± 24	32 ± 12	59 ± 21
Fe‐Nase strain	70 ± 8	55 ± 10	56 ± 6	45 ± 7

The proportion of the electron flux through Nase that is allocated to H2 production varies among strains, but the relative electron flux through Nase rather than Rubisco remains fixed (~60%). Errors are SEs.

Growth rate differences are associated with widespread proteomic changes beyond Nase isoform

The Nase isoforms are often described in a hierarchy, where the alternative V‐ and Fe‐Nases act as back‐up enzymes to the Mo‐Nase to sustain diazotrophic growth, albeit at slower rates, when Mo becomes limiting (e.g. Oda et al., 2005; Hamilton et al., 2011; Darnajoux et al., 2017). From a regulatory perspective, this is compatible with our data: the wild type strain uses the Mo‐Nase isoform under all the Mo‐replete conditions we tested, regardless of the carbon substrate. However, our data show that the hierarchy in Nase usage is not necessarily related to the isoforms' abilities to support fast growth under different conditions, because the V‐Nase can support comparably fast growth on acetate. Conversely, fixing nitrogen exclusively with the Mo‐Nase does not always yield fast growth, as is the case for the Mo‐Nase mutant strain on acetate, which appears to suffer from an as yet undetermined metabolic deficiency.

To better understand how changes in Nase isoform usage are integrated with the broader physiology of N2‐fixing R. palustris, we examined the global protein expression patterns of all four strains in both carbon substrates. Interestingly, despite exclusive use of the Mo‐Nase isoform by the wild type strain when grown on acetate, both its growth rate and proteome are more similar to that of the V‐Nase strain than the Mo‐Nase strain (Fig. 2 and Fig. S10, Table S4). The abundance of proteins related to coenzyme transport and metabolism (COG H) and energy production and conversion (COG C) in the wild type strain are also more similar to the V‐Nase strain than the Mo‐Nase strain on acetate (Fig. 2C). This shows that the observed differences in growth rates (Fig. 2) are associated with widespread proteomic changes and that many metabolic levers control nitrogen acquisition and diazotrophic fitness. Together the data imply that the effect of Nase isoform usage on growth is complex and may depend on R. palustris' ability to efficiently integrate a specific isoform into broader metabolic constraints under given growth conditions, rather than the Nase isoform properties alone. These results add nuance to the traditional view that use of the Mo‐Nase is always more favorable for growth than the use of the V‐Nase.

Conclusions

Contrary to the paradigm that diazotrophic growth with the Mo‐Nase is inherently better than with the other Nase isoforms, we show that, for the bacterium R. palustris, use of the alternative V‐Nase can support similar growth rates as use of the canonical Mo‐Nase during N2 fixing photoheterotrophic growth on acetate. The V‐Nase strain can efficiently reduce N2 and generate nitrogen replete biomass. At 19°C, its H2:N2 stoichiometry, around ~1.5, is only slightly higher than for the Mo‐Nase strain (~1.0). During growth on acetate, excess H2 production does not lead to slower V‐Nase based growth compared with the Mo‐Nase. From an evolutionary perspective, our results suggest that using the alternative V‐Nase might not be as physiologically detrimental for organisms as has been assumed, decreasing the selective pressure to remove genes for this isoform from diazotroph genomes. Our results show that the effect of Nase isoform usage on physiology, growth rate and H2 yield is more complex than previously recognized, and that our understanding of why alternative nitrogenases persist in nature requires consideration of broader metabolic constraints, including carbon substrates, environmental conditions, and metabolic integration of diazotrophy with phototrophy, heterotrophy and carbon fixation.

Experimental procedures

Bacterial cultures

Rhodopseudomonas palustris strains CGA009 (wild type), CGA753 (‘Mo‐Nase strain,’ Mo‐nitrogenase only, genotype: ΔvnfH ΔanfH), CGA766 (‘V‐Nase strain,’ V‐nitrogenase only, genotype: ΔnifH nifD::Tn5 ΔanfA) and CGA755 (‘Fe‐Nase strain,’ Fe‐nitrogenase only, genotype: ΔnifH ΔvnfH) were grown in batch cultures at 19°C and ~90 μmols photons m−2 s−1 under anaerobic photoheterotrophic conditions in defined nitrogen‐fixing medium with 2.5 μM Fe, 100 nM Mo, 10 μM V, Wolfe's vitamin solution, 0.0005% yeast extract and either 10 mM succinate, 20 mM acetate, or a mixture of 10 mM butyrate and 5 mM sodium bicarbonate (Oda et al., 2005; Zhang et al., 2014). We used genetic mutants because Mo limitation is difficult to achieve, and even residual Mo can result in the presence of some Mo‐nitrogenase (Jacobitz and Bishop, 1992; Kimble and Madigan, 1992; Bellenger et al., 2008, 2011; Sippel et al., 2017). In R. palustris CGA009, the V‐ and Fe‐Nase isoforms are regulated by nitrogen status rather than metal availability, allowing these mutants to grow despite the presence of ample Mo (Oda et al., 2005). Where applicable, the final concentration of dimethyl sulfoxide (DMSO, Sigma Life Science, Bio Reagent Grade) was 40 mM. The culture media were prepared in an anaerobic glove box (COY Laboratory Products) in lieu of the Hungate technique, leading to a headspace composition of N2 with a background level of ~2% v/v H2. Oxygen was removed by microwaving, vacuum sonicating, and allowing the medium to sit in glove bag for several days until dissolved O2 was below the detection limit of our O2 probe (Hach HQ40d portable meter with LDO101 rugged optical dissolved oxygen probe, range 0.1–20 mg L−1 corresponding to 1%–200% saturation). Bacterial growth was monitored by optical density (OD660) using a Spectronic 20D+ (Thermo Fisher Scientific) or a Genesys 20 visible spectrophotometer (Thermo Fisher Scientific) and converted to cell density using the empirically observed relationship cells ml−1 = 2.29 × 109 × OD660.

Biomass composition

Cultures were subsampled anaerobically during exponential growth (OD660 ~ 0.35–0.75). Bacterial biomass was collected either on combusted glass fiber filters, washed with 10 mL milli‐Q water, and dried overnight at 60°C, or by pelleting and freeze‐drying, which yielded statistically indistinguishable results. A vario ISOTOPE select (Elementar Isoprime) was used to quantify total carbon, nitrogen, and biomass δ13C and δ15N. Measurements were calibrated with an in‐house aminocaproic acid standard (ACROS) and validated using the USGS 40 standard. Biomass CHNO composition was measured using a TruSpec Micro (LECO) and was, on average, CH1.75N0.18O0.54.

Analytical techniques

Headspace H2 and CO2 were sampled periodically throughout growth and in early stationary phase. In early experiments, samples were diluted 10‐fold and H2 and CO2 were quantified simultaneously using gas chromatography with thermal conductivity and flame ionization detectors (TCD and FID Peak Performer 1 series, Peak Laboratories) using a methanizer to convert CO2 into CH4 and UHP argon as a carrier gas. In later experiments, H2 and CO2 were measured simultaneously in undiluted samples using gas chromatography with a thermal conductivity detector (GC‐8AIT TCD, Shimadzu Instruments; detector temperature = 150°C, column temperature = 100°C) and the Restek ShinCarbon ST column with UHP nitrogen as a carrier gas. Dissolved CO2 was quantified by injecting 500 or 200 μl of culture into a sealed, N2‐flushed 27 ml anaerobic culture tube resp. 10 ml amber vial containing 0.1 ml of 12 N HCl, vortexing for 30 s, and allowing the sample to equilibrate for at least 15 min before the headspace was sampled (McKinlay and Harwood, 2010a). Total CO2 was calculated as the sum of the measured headspace CO2 and dissolved inorganic carbon.

N

N2 and CO2 fixation rates were measured in dual 15N2 H13CO3 − tracer experiments. As described in Mohr and colleagues (2010), 15N2 enriched medium was prepared by replacing the headspace of 30 ml serum vials, which had been pre‐filled with 25 ml of anaerobic nitrogen‐fixing medium, with Ar gas, degassing with a vacuum handpump, and injecting an overpressure of 15N2 gas (98%+, Cambridge Isotope Laboratories, Andover, MA, USA). The bottles were shaken for >5 days, the overpressure was removed, and they were allowed to equilibrate until the experiment. The 0.5 M 13C‐bicarbonate stock was prepared in 5 ml amber serum vials by adding the bicarbonate stock (99%, Aldrich Chemistry), flushing with N2, evacuating the headspace with a handpump, and filling the vial completely with sterile growth medium. Pre‐grown cultures (10 ml in 27 ml Balch Tubes; OD660 ~0.3–0.5) were spiked with 1 ml of the 15N2 saturated medium, 100 μl of the 13C‐bicarbonate stock, and 10% 15N2 gas was added to the headspace. Every hour for 3 h, cells were filtered onto combusted GF/F filters (Whatman, GE Healthcare Life Sciences) and analyzed for their 15N13C content by isotope ratio mass spectrometry at the Stable Isotope Facility (UC Davis). Using the measured stoichiometry, electron fluxes through Nase and Rubisco were calculated by assuming six electrons per N2 reduced, two electrons per H2 produced, and 4.13 electrons per CO2 fixed.

Proteomics

R. palustris cell pellets were extracted by heating (95°C, 20 min) and vortexing in a reducing and denaturing SDS(1%)/Tris(200 mM, pH 8.0)/DTT(10 mM) buffer, and cysteine thiols alkylated with 40 mM iodoacetamide. Proteins were purified by a modified eFASP (enhanced filter‐aided sample preparation) protocol (Erde et al., 2014), using Sartorius Vivacon 500 concentrators (30 kDa nominal cutoff). Proteins were digested with MS‐grade trypsin (37°C, overnight), and peptides were eluted from the concentrator dried by vacuum centrifugation. For quantitative analysis, peptides were isotopically labelled at both N‐ and C‐termini using the diDO‐IPTL methodology (Waldbauer et al., 2017). Briefly, C‐termini were labelled with either oxygen‐16 or ‐18 by enzymatic exchange in isotopic water of >98 atom% enrichment. N‐termini were labelled with either un‐ or di‐deuterated formaldehyde via reductive alkylation using sodium cyanoborohydride. Peptide extracts from each sample were split and aliquots labelled separately with CD2O/16O and CH2O/18O; the latter were pooled to serve as a common internal standard for quantification. Aliquots of the 16O‐labelled peptides and 18O‐labelled internal standard were mixed 1:1 v/v and analysed by LC–MS for protein expression quantification.

For LC–MS analysis, peptide samples were separated on a monolithic capillary C18 column (GL Sciences Monocap Ultra, 100 μm I.D. × 200 cm length) using a water‐acetonitrile +0.1% formic acid gradient (2%–50% AcN over 180 min) at 360 nl min−1 using a Dionex Ultimate 3000 LC system with nanoelectrospray ionization (Proxeon Nanospray Flex source). Mass spectra were collected on an Orbitrap Elite mass spectrometer (Thermo) operating in a data‐dependent acquisition mode, with one high‐resolution (120 000 m/Δm) MS1 parent ion full scan triggering Rapid‐mode 15 MS2 CID fragment ion scans of selected precursors. Proteomic mass spectral data were analysed using MorpheusFromAnotherPlace (MFAP; Waldbauer et al., 2017), using the predicted proteome of R. palustris CGA009 as a search database. Precursor and product ion mass tolerances for MFAP searches were set to 20 ppm and 0.6 Da respectively. Static cysteine carbamidomethylation and variable methionine oxidation, N‐terminal (d4)‐dimethylation and C‐terminal 18O2 were included as modifications. The false discovery rate for peptide‐spectrum matches was controlled by target‐decoy searching to <0.5%. For Nase proteins with high homology between isoforms, the spectra were additionally manually curated to ensure that peptide assignments were unique and not due to contamination. Protein‐level relative abundances and standard errors were calculated in R using the Arm postprocessing scripts for diDO‐IPTL data (Waldbauer et al., 2017; http://github.com/waldbauerlab).

Proteomics analysis

The log2 effect was calculated as the log2 difference over the sum of the errors for both treatments and used to calculate the Bonferroni corrected p‐value for significant differences from the Excel standard normal distribution function (norm.s.dist, cumulative = false). COG assignments were taken from UniProt on March 13, 2019, and hierarchical clustering was performed using the Morpheus Software from the Broad Institute (https://software.broadinstitute.org/morpheus) with the one minus Spearman Rank Correlation metric, average linkage method, and clustering by strain and carbon substrate treatment.

Appendix S1. Supporting Information figures, discussion, experimental procedures and Tables S1 and S4.

Click here for additional data file.

Table S2. This table summarizes the H2:N2 ratios available in the literature. Published H2:N2 ratios are highly variable. They appear to be lowest at lower temperatures, high Fe protein:MoFe resp. VFe resp. FeFe protein ratios and higher N2 partial pressures. This table was generated from the references in Eady (1996) and by a literature review of the 497 publications which, as of May 23, 2019, had cited it. Except for Bishop, Hawkins and Eady (1986), no data from experiments with an uptake Hase are included. Only papers that measured H2 production and N2 fixation rates under the same atmospheric composition are included.

Click here for additional data file.

Table S3. This excel document contains the data used in this publication.

Click here for additional data file.