Carbon dioxide fixation as a central redox cofactor recycling mechanism in bacteria

Abstract

The Calvin-Benson-Bassham cycle (Calvin cycle) catalyzes virtually all primary productivity on Earth and is the major sink for atmospheric CO₂. A less appreciated function of CO₂ fixation is as an electron-accepting process. It is known that anoxygenic phototrophic bacteria require the Calvin cycle to accept electrons when growing with light as their sole energy source and organic substrates as their sole carbon source. However, it was unclear why and to what extent CO₂ fixation is required when the organic substrates are more oxidized than biomass. To address these questions we measured metabolic fluxes in the photosynthetic bacterium Rhodopseudomonas palustris grown with ¹³C-labeled acetate. R. palustris metabolized 22% of acetate provided to CO₂ and then fixed 68% of this CO₂ into cell material using the Calvin cycle. This Calvin cycle flux enabled R. palustris to reoxidize nearly half of the reduced cofactors generated during conversion of acetate to biomass, revealing that CO₂ fixation plays a major role in cofactor recycling. When H₂ production via nitrogenase was used as an alternative cofactor recycling mechanism, a similar amount of CO₂ was released from acetate, but only 12% of it was reassimilated by the Calvin cycle. These results underscore that N₂ fixation and CO₂ fixation have electron-accepting roles separate from their better-known roles in ammonia production and biomass generation. Some nonphotosynthetic heterotrophic bacteria have Calvin cycle genes, and their potential to use CO₂ fixation to recycle reduced cofactors deserves closer scrutiny.

At least 170 y ago, it became widely recognized that CO₂ could serve as the sole carbon source for plants and certain bacteria (autotrophic growth). Using experimental results collected by a number of scientists in the late 1700s and early 1800s, the German chemist von Liebig successfully argued that plants obtain carbon from CO₂ gas rather than from carbon in the soil (the prevailing theory at the time) (1). At this time, photosynthetic bacteria were still considered to be plants and thus were also recognized to fix CO₂ for growth (2).

Nearly a century later, it was proposed that CO₂ fixation also plays a role in maintaining redox balance in the group of anoxygenic phototrophs known as purple nonsulfur bacteria (PNSB). PNSB obtain energy from light and carbon from organic substrates to support growth under anaerobic conditions (photoheterotrophic growth). In 1933, Muller reported that CO₂ had to be supplied for PNSB bacteria to grow photoheterotrophically on butyrate, a substrate more reduced than biomass (Table 1) (3). Muller proposed a hypothesis that CO₂ is needed to accept excess electrons and allow butyrate to be oxidized to the redox state of biomass (Fig. 1) (3). This hypothesis went untested for nearly 50 y. In 1977 Hillmer and Gest demonstrated that Rhodobacter capsulatus would grow on butyrate without CO₂ if growth conditions permitted N₂ fixation and therefore expulsion of electrons as H₂ (Fig. 1) (4). Richardson et al. (5) later showed that adding other alternative electron acceptors, such as dimethyl sulfoxide, allowed R. capsulatus to grow photoheterotrophically on butyrate without added CO₂ (Fig. 1). These studies confirmed that the role of CO₂ fixation during photoheterotrophic growth on butyrate is to accept electrons because other electron acceptors could substitute for CO₂ and permit growth.

Table 1.

Oxidation/reduction values of R. palustris growth substrates and biomass

Compound	Formula	Oxidation/reduction value†
Malate	C4H6O5	+2
Succinate	C4H6O4	+1
Acetate	C2H4O2	0
Biomass‡	CH1.8N0.18O0.38	−0.5
Butyrate	C4H8O2	−2

^†Values were determined as described by Gottschalk (39) and in SI Methods.

^‡Based on the elemental composition of R. palustris 42OL (40).

Fig. 1.

Redox balancing strategies used by PNSB during anaerobic photoheterotrophic growth on a reduced carbon source. During photoheterotrophic growth of PNSB on butyrate, ATP is produced by cyclic photophosphorylation, in which electrons are repeatedly energized by light, allowing them to be cycled through a proton-pumping electron transport chain, rather than being transferred to a terminal electron acceptor. Thus, PNSB must use other means to dispose of extra electrons that are generated during use of a carbon source like butyrate, which is more reduced than biomass (Table 1). Three mechanisms are depicted: (A) CO₂ fixation via the Calvin cycle; (B) H₂ production via nitrogenase (N₂ase; green arrow); and (C) DMSO reduction by DMSO reductase (DMSOR; red arrow). Solid black lines, carbon metabolism; solid blue lines, energy metabolism; dotted lines, electron metabolism.

In the early 1990s the groups of Kaplan and Tabita showed that CO₂ was required for photoheterotrophic growth even on substrates that are more oxidized than biomass (Table 1) (6, 7). R. sphaeroides mutants lacking ribulose 1,5-bisphosphate carboxylase (Rubisco), the key CO₂-fixing enzyme of the Calvin-Benson-Bassham cycle (Calvin cycle), were incapable of photoheterotrophic growth on succinate and malate. Photoheterotrophic growth was restored when the alternative electron acceptor dimethyl sulfoxide was supplied (6, 7). The fact that PNSB grow on oxidized organic substrates as the sole carbon source suggests that CO₂ derived from the oxidation of these substrates can support the necessary amount of electron-accepting CO₂ fixation. However, the reason for the requirement of CO₂ fixation for photoheterotrophic growth on substrates more oxidized than biomass and the quantitative contribution of CO₂ fixation to growth under these conditions were unclear.

Here we used ¹³C-metabolic flux analysis (¹³C-MFA) to quantify the significance of CO₂ fixation as an electron-accepting process during photoheterotrophic growth. The PNSB Rhodopseudomonas palustris was provided with acetate, a compound slightly more oxidized than biomass (Table 1), as its sole carbon source. Our results show cells oxidized 22% of the carbon in acetate to CO₂, and most of this then reentered metabolism via the Calvin cycle. This degree of Calvin cycle flux enabled cells to reoxidize almost half of the reduced cofactors generated during acetate oxidation. We also examined the effects of H₂ production by R. palustris on Calvin cycle flux and gene expression.

Results

Wild-Type Cells Have a Large Metabolic Flux from Acetate Through the Calvin Cycle.

R. palustris wild-type strain CGA009 was grown anaerobically in light in mineral medium containing NH₄ ⁺, a condition in which H₂ production via nitrogenase is repressed. To determine metabolic flux distributions, we first confirmed that cells convert acetate carbon and electrons entirely into biomass and CO₂ (Table 2). We then manually constructed a metabolic network from the R. palustris CGA009 genome sequence (8) (Table S1). We grew R. palustris with acetate labeled with ¹³C in one of its carbons and measured mass isotopomer distributions in proteinaceous amino acids (Table S2). This information along with measurements of CO₂ efflux (Table 2) and biomass composition (Table S3) (both relative to acetate uptake) were applied to the metabolic network using 13C-Flux software (9) to solve for intermediary metabolic fluxes. Details of the fitting solutions are described in the supplementary materials. The complete set of net and exchange flux values are listed in Table S4.

Table 2.

Conversion of acetate to biomass, CO₂, and H₂ by R. palustris

Strain	g (h)	Yield (mol • mol acetate−1)			C recovery (%)	e− recovery (%)‡
		YBiomass†	YCO2	YH2
Wild-type	8.4 ± 0.6	1.75 ± 0.15	0.11 ± 0.02	0	93 ± 8	98 ± 9
NifA*	9.4 ± 0.6	1.57 ± 0.08	0.34 ± 0.04	0.41 ± 0.05	96 ± 5	99 ± 5

Cultures were grown in minimal medium with 20 mM sodium acetate and NH₄ ⁺ as the nitrogen source. Values are averages from 8 to 10 biological replicates ± SD.

^†Moles of R. palustris biomass were determined from the elemental composition of R. palustris 42OL (40): CH_1.8N_0.18O_0.38, which has a molecular weight of 22.426 g/mol.

^‡Electron recovery was based on available hydrogen as described by Gottschalk (39) and in SI Methods.

The flux map indicates that acetate is mainly assimilated via the glyoxylate shunt (Fig. 2A, red). The distribution of fluxes then seems to be primarily determined by the demand for biosynthetic precursors, with little flux through the lower TCA cycle (Fig. 2A, blue) or pyruvate dehydrogenase (Fig. 2A, PDH/POR). An exception to this trend is the flux through the Calvin cycle. Calvin cycle flux was quantified from the unique metabolite labeling patterns that it generated, such as fully labeled serine (Fig. S1). Flux through the CO₂-fixing enzyme of the Calvin cycle, Rubisco, was considerable at 29.6 mol% of the acetate uptake rate [0.58 μmol × mg dry cell weight (DCW)⁻¹ × h⁻¹; Fig. 2A]. The magnitude of the CO₂-fixing Rubisco flux was surprising because no CO₂ or NaHCO₃ was provided to the cultures. The sum of the flux through all CO₂-producing reactions was 43.8 mol% of the acetate uptake rate, indicating that approximately 22% of the acetate carbon was released as CO₂. Rubisco then fixed 67% of this CO₂ into cell material. This high Rubisco flux makes R. palustris metabolism efficient with respect to carbon utilization. According to the flux distribution, only 6% of acetate carbon is expelled as CO₂ (11.9 mol% of the acetate uptake rate), leaving 94% of the acetate carbon for incorporation into cell material (Fig. 2A).

Fig. 2.

Central metabolic fluxes in wild-type and NifA* R. palustris strains during photoheterotrophic growth on acetate. Both strains were grown with NH₄ ⁺ as a nitrogen source. Under these conditions the wild type does not synthesize nitrogenase or produce H₂ (A), but the NifA* strain does synthesize nitrogenase and produce H₂ (B). Net fluxes values are listed as mole percentage of the acetate uptake rate ± 90% confidence intervals. Actual acetate uptake rates are below the uptake arrow. Gray values indicate that the confidence interval was larger than the mean. For brevity, biosynthetic fluxes and exchange fluxes (reverse fluxes) that could be determined are shown in Table S4. Net flux direction is indicated by an enlarged arrowhead. Flux magnitudes are indicated by arrow thickness. Fluxes that have a value of less than 5 are given a dashed arrow. The names of flux reactions not mentioned in the text are listed in Table S1.

Whereas most of the acetate carbon was used for biosynthesis, only approximately half (54%) of the reduced redox cofactors generated during acetate oxidation were oxidized in biosynthetic reactions (Table 3). Thus, to maintain redox balance and metabolic flow, R. palustris must recycle the remaining redox cofactors by some other means. The Calvin cycle can maintain cellular electron carriers in the oxidized reduction state after fixing CO₂ by oxidizing NADH via GAPDH. At 59.5 mol% of the acetate uptake rate (Fig. 2A), GAPDH flux was 44% of the total flux through reactions that generate reduced cofactor (Table 3). Thus, the Calvin cycle fulfilled a major role in maintaining redox balance by recycling reduced redox cofactors. Approximately 2% of reduced cofactors were oxidized by a proposed pyruvate–ferredoxin oxidoreductase flux (Fig. 2 and Table 3, PDH/POR).

Table 3.

Redox accounting based on fluxes through redox reactions

Reaction	Cofactor made	Wild type	NifA*
Biosynthesis NADH†	NADH	12.2	10.7
Biosynthesis NADP+†	NADP+	−72.3	−63.2
αKDH	NADH	1.9	1.8
IDH	NADPH	8.7	7.7
SDH	QOH	38.4	39.0
MDH	NADH/QOH	75.6	76.8
MEnz	NADPH	0.0	0.0
PDH/POR	NAD+/ Fdox	−2.4	−4.0
GAPDH (Calvin cycle)‡	NAD+	−59.5	−19.0
OPPP	NADPH	0.0	1.0
H2	Fdox	0.0	−45.3
Total cofactor reduction§		136.8	137.0
Total cofactor oxidation§		−134.2	−131.5
Redox balance (%)¶		98	96

Values are fluxes as mole percentages of acetate uptake rate. Italicized reactions are the same as those in Fig. 2 and in Table S1. Positive values indicate flux through reactions that form reduced redox cofactors (e.g., NADH), and negative values indicate flux through reactions that form oxidized redox cofactors (e.g., NAD⁺). QOH, quinol; Fd_ox, oxidized ferredoxin.

^†NADH and NADP⁺ formation expected from biosynthesis based on the R. palustris biomass composition (Table S3).

^‡NAD⁺ is assumed to be made by GAPDH rather than NADP⁺, on the basis of ref. 20.

^§Total reduction: sum of fluxes through reactions forming reduced cofactors. Total oxidation: of fluxes through reactions forming oxidized cofactors.

^¶Total oxidation ÷ total reduction × 100%.

Calvin Cycle Fluxes Are Diminished when Cells Produce H₂.

H₂ production by nitrogenase, which produces H₂ along with NH₃ as an obligatory aspect of its catalytic cycle (10), enables photoheterotrophic growth of PNSB on highly reduced substrates (4). We used ¹³C-MFA to investigate the effect of H₂ production on carbon fluxes from acetate and to determine how the Calvin cycle responds to an alternative means of electron disposal. For this purpose we decided to use an R. palustris nifA* mutant, which expresses nitrogenase genes constitutively and thus produces H₂ via nitrogenase in NH₄ ⁺-containing mineral medium, so that we could compare non–hydrogen-producing (wild-type) and hydrogen-producing (NifA*) cultures grown under the same conditions.

NifA is a σ⁵⁴-dependant enhancer-binding protein that is the master transcriptional regulator of nitrogenase genes (11). Previously described nifA* mutants have single point mutations in the Q-linker region that is between the NifA N-terminal GAF domain and a central AAA⁺ domain preceding the C-terminal helix-turn-helix DNA-binding domain (Fig. 3) (12). These mutations apparently enable the protein to adopt an active conformation in NH₄ ⁺-grown cells. Because our original nifA* point mutants can revert to the wild-type phenotype with a single nucleotide change, we established a stable nifA* mutant by deleting 48 bp of the Q-linker, for use in our ¹³C-MFA experiments. This NifA* strain, CGA676, produced H₂ at comparable levels to the previously described nifA* point mutant, CGA584 (12).

Fig. 3.

Diagram of NifA structural domains and sites of NifA* mutations. A 16-aa NifA Q-linker region deletion strain (CGA676) produces H₂ in the presence of NH₄ ⁺. This deletion mutant is much more stable than the NifA* point mutants that were previously described (12) and are shown below the NifA diagram. NifA and its domains are drawn to scale.

Central metabolic flux distributions of the H₂-producing NifA* strain were remarkably similar to those in the non–H₂-producing wild type with the exception of fluxes through the Calvin cycle (Fig. 2). Although a similar amount of acetate (23%) was oxidized to CO₂ as in the wild type, the NifA* strain’s Rubisco flux was approximately one fifth of that of the wild type (specific in vivo activity of 0.11 μmol × mg DCW⁻¹ × h⁻¹), and it thus refixed only 13% of CO₂. The low Rubisco flux during H₂ production resulted in a 3-fold increase in net CO₂ production compared with the wild type, leaving 82% of the carbon for biosynthesis (Fig. 2).

Similar to the wild type, nearly half of the reduced redox cofactors (47%) generated during acetate oxidation were used for biosynthesis during H₂ production by the NifA* strain (Table 3). Unlike the wild type, the Calvin cycle played a lesser role in maintaining redox balance during H₂ production. GAPDH flux was one third of the wild-type flux (Fig. 2), recycling only 14% of reduced redox cofactors (Table 3). H₂ production (0.45 ± 0.01 mol% of the acetate uptake rate) played a larger role in maintaining redox balance, recycling 34% reduced redox cofactors. Thus the NifA* strain used a combination of CO₂ fixation and H₂ production to recycle approximately half of its redox cofactor (one sixth and one third, respectively). As in the wild type, there was minor participation (3%) by the PDH/POR flux in maintaining redox balance. The difference between wild-type and NifA* GAPDH flux accounted for 89% of the electrons needed to produce the observed H₂ (Table 3). In other words, most of the electrons for H₂ production were diverted away from the Calvin cycle. The remaining electrons for H₂ production are accounted for by the NifA* strain's lower NADPH requirement to support its lower growth yield (Table 2). H₂ production had little effect on fluxes through other dehydrogenases or oxidoreductases (Table 3).

Calvin Cycle Flux Is Essential for Growth in the Absence of H₂ Production.

The NifA* strain maintained a low level of Calvin cycle flux, even though it disposed of most of its excess electrons by H₂ production. To assess the physiological importance of this low Calvin cycle flux, we deleted the genes encoding both type I and type II Rubiscos in the NifA* and wild-type strains. Consistent with previous work with R. sphaeroides (6, 7) and R. palustris (13), the Rubsico mutant of wild-type R. palustris (CGA669) was unable to grow photoheterotrophically in NH₄ ⁺-containing medium (Table 4). However, the Rubisco mutant could grow photoheterotrophically in NH₄ ⁺-containing medium when H₂ production was permitted by introducing the CGA676 nifA* mutation to form strain CGA679. H₂ production seemed to compensate for the lack of Calvin cycle flux in expelling excess electrons, because the H₂ yield increased 1.3-fold in the NifA* Rubisco mutant strain compared with the NifA* strain with functional Rubisco enzymes (Table 4). However, the NifA* Rubsico strain had a 1.7-fold longer generation time compared with the NifA* strain (Table 4). The slower growth rate may be explained by the fact that the specific H₂ productivity was the same for both strains, indicating that nitrogenase activity did not increase despite the extra redox burden in the absence of Calvin cycle flux (Table 4). Thus, H₂ production can maintain redox balance in the absence of the Calvin cycle, but the growth rate is likely limited by the slow rate at which electrons are expelled as H₂ via nitrogenase.

Table 4.

Effect of Rubisco mutations on growth and H₂ production

Strain	Genotype	g (h)	H2 yield (mol × mol ace−1)	Sp. H2 productivity (mmol × g DCW−1 × h−1)†	H2:biomass (mol × g protein−1)
CGA009	Wild type	7.4 ± 0.2	None detected	―	―
CGA669	ΔcbbLS::KmR ΔcbbM	No growth	―	―	―
CGA676	nifA*	8.3 ± 0.8	0.43 ± 0.04	1.12 ± 0.08	66 ± 5
CGA679	nifA* ΔcbbLS::KmR ΔcbbM	12.8 ± 0.8‡	0.58 ± 0.03‡	1.02 ± 0.07	92 ± 3‡
CGA679 complemented	nifA* ΔcbbLS::KmR ΔcbbM (pBBPcbbLSX)	9.8 ± 1.4	0.44 ± 0.08	0.98 ± 0.04	68 ± 11

Cultures were grown in minimal medium with 20 mM sodium acetate and NH₄ ⁺ as the nitrogen source. Values are averages ± SDs from 3 to 5 biological replicates.

^†Specific H₂ productivity. Determined using the Monod model as described in ref. 21.

^‡Significantly different from CGA676 characteristic (Student's t test, two-tailed, equal variance, P < 0.05).

Additional Features of R. palustris Metabolism Revealed by 13C-MFA.

There are some general features of the ¹³C-metabolic flux maps outside of our main objectives that are worth mentioning briefly. First, R. palustris used a citramalate-dependent pathway (14) to synthesize approximately 94% of its isoleucine, whereas the rest was made by the traditional threonine-dependent pathway (Fig. S2). Second, flux between pyruvate and acetyl-CoA was reversible, suggesting that R. palustris has an active pyruvate ferredoxin oxidoreductase in addition to pyruvate dehydrogenase, which is generally considered to unidirectionally decarboxylate pyruvate to acetyl-CoA (Fig. 2 and Fig. S1). Third, there was little flux through NADPH-generating reactions (Table 3), suggesting that cells generated nearly 90% of their NADPH by oxidizing other redox cofactors (e.g., NADH oxidation via membrane-bound transhydrogenase encoded by RPA4180-2).

Electron Flux Distribution Is Regulated at the Level of Transcription.

Our data show that electrons are diverted away from the CO₂-fixing Calvin cycle toward nitrogenase to produce H₂. However, the flux maps do not indicate whether nitrogenase simply out-competes the Calvin cycle for electrons or whether regulatory mechanisms are involved in the redistribution of electrons. We used quantitative PCR (qPCR) to compare expression levels of nifD, encoding the nitrogenase α-subunit, and several Calvin cycle genes in the wild-type and NifA* strains. As expected, nifD expression levels were much higher (450-fold) in the NifA* strain compared with the wild type because NH₄ ⁺ does not repress nitrogenase expression in the NifA* strain (Fig. 4). Similar to the Calvin cycle flux trend, Calvin cycle gene expression levels were lower in the H₂-producing NifA* strain than in the non–H₂-producing wild type. Thus, the shift of electrons from CO₂ fixation to H₂ production is controlled at the transcriptional level. We also compared the transcript levels of genes encoding central metabolic enzymes in five different nifA* mutants vs. the wild-type strain, all grown on acetate with NH₄ ⁺ (12). Like the flux maps, the only consistent changes in transcripts encoding central metabolic enzymes in response to H₂ production were for those encoding Calvin cycle enzymes, which were down-regulated (Table S5). A similar decrease in Calvin cycle transcript levels was also observed in wild-type cultures grown under conditions permitting N₂ fixation, and therefore H₂ production, relative to non–H₂-producing cultures (Table S5).

Fig. 4.

Effect of H₂ production on Calvin cycle gene expression. Relative levels of transcripts between wild-type R. palustris that is not producing H₂ (CGA009; WT) and a mutant (CGA676; NifA*) that produces H₂ in the presence of NH₄ ⁺. The two strains were grown under identical conditions with 20 mM sodium acetate with NH₄ ⁺ present. (A) nifD, encoding the α-subunit of nitrogenase; (B) cbbL, encoding the Rubisco type I large subunit; and (C) cbbM, encoding Rubisco type II. Data are averages from duplicate cultures, with error bars representing the range. Transcript levels are shown relative to fixJ transcript levels.

RegSR Does Not Transcriptionally Control the Calvin Cycle in Response to H₂ Production.

In other PNSB, the redox-sensing two-component regulatory system RegBA/PrrBA modulates gene expression of both Calvin cycle operons (reviewed in refs. 15 and 16). We therefore examined the effect of deleting the R. palustris regBA homologs, regSR, on Calvin cycle gene expression. For these experiments we used wild-type R. palustris CGA010 (a derivative of CGA009) and its preexisting ΔregSR mutant, CGA2023. We used the presence and absence of NH₄ ⁺ to respectively repress or activate H₂ production by nitrogenase in these strains. When not producing H₂, the ΔregSR mutant had 3.1-fold higher cbbL expression than the wild type and negligible increases in cbbM and cbbP expression (Fig. 5). The ΔregSR mutation had no effect on the expression of the three Calvin cycle genes or on nifD expression during H₂ production (Fig. 5). Most importantly, regSR was not required for the drastic difference in Calvin cycle cbbL, cbbM, and cbbP expression observed between H₂-producing cultures and non–H₂-producing cultures (Fig. 5), indicating that this two-component system is not central to integrating redox metabolism in R. palustris as it is in other PNSB (15, 16).

Fig. 5.

Calvin cycle gene expression in a ΔregSR strain. Relative levels of transcripts between wild-type R. palustris (CGA010; WT) and a ΔregSR mutant derivative (CGA2023; ΔSR). H₂ production was accomplished by supplying N₂ as the sole nitrogen source. Acetate was the carbon source. Data are averages from duplicate cultures, with error bars representing the range. (A) nifD, encoding the α-subunit of nitrogenase; (B) cbbL, encoding the Rubisco type I large subunit; (C) cbbM, encoding Rubisco type II; and (D) cbbP, encoding phosphoribulokinase. Transcript levels are shown relative to fixJ transcript levels.

Discussion

CO₂ Fixation Is a Major Aspect of Photoheterotrophic Metabolism.

Our results illustrate why, and to what extent, CO₂ fixation is required during photoheterotrophic growth on substrates, like acetate, that are more oxidized than R. palustris biomass. During the oxidation of acetate, approximately twice as much reduced redox cofactor is produced than is needed for biosynthesis. CO₂ fixation plays a major and obligatory role as an electron-accepting process to recycle the majority of these excess reduced cofactors under non–N₂-fixing conditions. The NifA* strain, which synthesizes nitrogenase constitutively, recycled approximately half of its electrons using a combination of CO₂ fixation and H₂ production. Because no other electron acceptors were available, our results suggest that R. palustris growing on acetate can only devote approximately half of its reduced cofactor toward biosynthesis, and the rest must be recycled by other means (e.g., CO₂ fixation and H₂ production). These ¹³C-labeling experiments highlight two underappreciated facts: (i) that photoheterotrophic growth is associated with a challenge in maintaining redox balance, and (ii) that the CO₂-fixing Calvin cycle is not only at the heart of photoautotrophic metabolism but can be a central aspect of photoheterotrophic metabolism as well. Our results are in agreement with theoretical work predicting that photoheterotrophic growth is associated with a large Rubisco flux (17).

Interplay Between CO₂ Fixation and H₂ Production as Redox-Balancing Mechanisms.

Because the NifA* strain is capable of both CO₂ fixation and H₂ production, we were able to examine the interplay between these two redox-balancing processes. In the NifA* strain electrons were shifted away from CO₂ fixation toward H₂ production (Table 3). Although Calvin cycle flux decreased significantly in response to H₂ production, it did not shut off completely. When Calvin cycle flux was prevented in the NifA* Rubisco strain, CGA679, the growth rate was severely impaired (Table 4). Interestingly, the specific H₂ production rate did not vary when Rubisco activity was eliminated (Table 4), suggesting that there is a bottleneck preventing an increase in the H₂ production rate. Possibilities are a rate limitation in electron transfer to nitrogenase, nitrogenase concentration, or the rate of nitrogenase activity. It seems likely that the NifA* Rubisco strain's growth rate is limited by the rate at which redox balance can be achieved by expelling electrons as H₂. When the Calvin cycle is also available, the nifA* mutant may be able to adjust electron distribution between the two redox-balancing mechanisms to maximize its growth rate.

We observed that the shift from CO₂ fixation to H₂ production was accompanied by a decrease in Calvin cycle gene transcription (Figs. 4 and 5). It was previously reported that in R. sphaeroides, Rubsico transcript levels varied in response to other factors affecting the intracellular redox state, such as feeding substrates of differing redox state (18) and adding the electron acceptor dimethyl sulfoxide (7). Although we have excluded the RegSR system as being central to controlling Calvin cycle fluxes in R. palustris in response to redox changes due to H₂ production, there are a number of other regulatory mechanisms that could be involved. For example, high NADH levels can stimulate Calvin cycle phosphoribulokinase activity (19, 20), which in turn can stimulate Calvin cycle gene expression through the transcriptional regulator CbbR (15).

Alternative Roles of the Calvin Cycle and Nitrogenase in Nature.

Use of CO₂ fixation for redox balance is not restricted to photoheterotrophic metabolism. For example, succinate-producing fermentative bacteria fix CO₂ via C₃-carboxylating enzymes as part of a redox-balancing mechanism (21). It is possible that there are niches, for example, microaerobic environments, where it is advantageous for nonphotosynthetic, obligately heterotrophic bacteria to expend some ATP and use the Calvin cycle to maintain redox balance. Complete Calvin cycle operons have been annotated in heterotrophic bacteria for which autotrophic growth has not been demonstrated, including Dechloromonas aromatica str. RCB (22), Methylibium petroleiphilum PM1 (23), and Burkholderia xenovorans (24). Rubisco genes have also been detected in environmental isolates related to Arthrobacter, Bacillus, and Streptomyces (25). Calvin cycle activity by such bacteria may help to explain observations in which adding acetate to soil in the dark can stimulate CO₂ fixation (26).

As proposed at least as early as 1962 by Ormerod and Gest (27), H₂ production by PNSB may also have an underappreciated role in maintaining redox balance. In PNSB mutants having constitutive nitrogenase activity, nitrogenase expression decreased in response to the addition of the electron acceptor dimethyl sulfoxide or the expression of Calvin cycle genes from a plasmid (28, 29). These results suggest that there are mechanisms in place to control the contribution of nitrogenase toward redox balance in response to other electron acceptors. However, the use of nitrogenase for redox balance is likely an auxiliary or specialized mechanism because the large ATP requirement of nitrogenase would present challenges for its use in maintaining redox balance in a nonphotosynthetic heterotroph.

Methods

Chemicals, Bacteria, and Culture Conditions.

Sodium [1-¹³C] and [2-¹³C]acetate were purchased from Cambridge Isotope Laboratories. All experiments were conducted with R. palustris wild-type strains CGA009 or CGA010 and their derivatives (Table S6). CGA009 is defective in uptake hydrogenase activity, whereas the uptake hydrogenase mutation is repaired in CGA010 (30). CGA010 uptake hydrogenase was not functional in the studies herein owing to insufficient nickel in the growth medium. R. palustris was grown anaerobically in front of a 60-W light bulb at 30 °C in 16.5-mL volumes of defined mineral medium (31) in sealed 27-mL anaerobic culture tubes (Bellco) without added NaHCO₃ or CO₂. Cultures were grown in minimal medium either with (NH₄)₂SO₄ under an Ar headspace (i.e., PM) or without (NH₄)₂SO₄ under a N₂ headspace (i.e., NFM). Sodium acetate or disodium succinate was supplied at a final concentration of 20 mM or 10 mM, respectively. For comparisons between the nifA* parent, CGA676, and its Rubisco mutant derivatives, cultures were grown aerobically in the dark in PM, before being adapted and tested in anaerobic PM in the light. Escherichia coli strains DH5α and S17-1 were grown at 37 °C in Luria-Bertani medium. Ampicillin, kanamycin, and gentamcyin were supplied at 100, 100, and 20 μg/mL, respectively, for E. coli, and kanamycin and gentamycin were each supplied at 100 μg/mL for R. palustris where appropriate.

Strain Construction.

In-frame deletion mutants of R. palustris were generated using primers and plasmids listed in Table S6 as previously described (30), with the modifications described in SI Methods. To complement the cbbLS deletion, cbbLSX were expressed from a plasmid downstream of the promoter region for RPA0944, which encodes GAPDH, as described in SI Methods.

Analytical Techniques.

Acetate was quantified as previously described (32) using a Varian HPLC equipped with a UV detector. H₂ was quantified using gas chromatography as previously described (12). Total CO₂ [i.e., the sum of headspace CO₂ (g) plus dissolved CO₂ (aq)] was quantified according to previously described methods (33), with details in SI Methods.

R. palustris Biomass Composition.

Measurements were made for strains CGA009 and CGA676 in PM as previously described (21), with details in SI Methods.

¹³C-Labeling Experiments.

CGA009 and CGA676 were grown in PM containing [1-¹³C] or [2-¹³C]acetate, such that there were three biological replicates of each strain for each substrate isotopomer. Cultures were transferred once in midexponential phase to avoid light-limiting conditions at high cell densities and to dilute the unlabeled inoculum to below 1% of the harvested biomass. Upon reaching a density between 0.3 and 0.4 OD₆₆₀, cultures were chilled in ice water and centrifuged at 4 °C. Cell pellets were lysed by sonication, soluble protein hydrolyzed, and amino acids purified, quantified, and derivatized with tert-butyldimethylsilyl groups, as previously described (21). Derivatized amino acids were analyzed by gas chromatography–mass spectrometry using an Agilent 5973 GC-MSD equipped with a DB5-MS column (30 m × 0.25 mm × 0.25 μm) as previously described (21).

Metabolic Flux Analysis.

Metabolic fluxes were determined using 13C-Flux software (9). Flux models were fit to experimental data using the 13C-Flux program, Donlp2 (P. Spellucci, Technische Universität Darmstadt, Germany). Mass isotopomer distributions were corrected for the natural abundances of amino acid atoms other than carbon and all atoms in the derivatization groups using previously described software (34). Datasets resulting from experiments using either [1-¹³C] or [2-¹³C]acetate were fit against a single metabolic model as previously described (35). To determine whether multiple flux solutions existed, ≈100 different starting values were used for Donlp2 runs. Confidence intervals for individual fluxes were determined as previously described (36). The isotopomer measurements used to determine specific fluxes are given in Table S7, with examples illustrated in Fig. S1.

Microarray and qPCR.

Microarray analyses were carried out with an Affymetrix R. palustris Custom GeneChip as previously described (12) for CGA009 grown in NFM with 20 mM acetate and were deposited at http://www.ncbi.nlm.nih.gov/geo under accession nos. GSM159274 and GSM159275. For qPCR experiments, cultures were grown to a density of 0.3–0.5 OD₆₆₀ with unlabeled acetate. RNA for microarrays and qPCR was isolated as previously described (37). cDNA was generated as previously described (12). The qPCR reactions were set up as previously described (38), except R. palustris chromosomal DNA was used to generate standard curves, and transcript levels were normalized to transcript levels of fixJ. We also made use of previously reported microarray data (12).