Reversibility of citrate synthase allows autotrophic growth of a thermophilic bacterium

See allHide authors and affiliations

Science  02 Feb 2018:
Vol. 359, Issue 6375, pp. 563-567
DOI: 10.1126/science.aao2410

About-face for citrate synthase

Classically, it is thought that citrate synthase only works in one direction: to catalyze the production of citrate from acetyl coenzyme A and oxaloacetate in the tricarboxylic acid (TCA) cycle. The TCA cycle can run in reverse to cleave citrate and fix carbon dioxide autotrophically, but this was thought to occur only with alternative enzymes, such as citrate lyase. Now Nunoura et al. and Mall et al. have discovered thermophilic bacteria with highly efficient and reversible citrate synthase that requires reduced ferredoxin (see the Perspective by Ragsdale). This function is undetectable by metagenomics, but classical biochemistry filled in the gaps seen between the genome sequences and the phenotypes of the organisms. The direction of catalysis depends on the availability of organic versus inorganic carbon and reflects a flexible bet-hedging strategy for survival in fluctuating environments. In evolutionary terms, this capacity might predate the classical TCA cycle and is likely to occur in a wide range of anaerobic microorganisms.

Science, this issue p. 559, p. 563; see also p. 517


Biological inorganic carbon fixation proceeds through a number of fundamentally different autotrophic pathways that are defined by specific key enzymatic reactions. Detection of the enzymatic genes in (meta)genomes is widely used to estimate the contribution of individual organisms or communities to primary production. Here we show that the sulfur-reducing anaerobic deltaproteobacterium Desulfurella acetivorans is capable of both acetate oxidation and autotrophic carbon fixation, with the tricarboxylic acid cycle operating either in the oxidative or reductive direction, respectively. Under autotrophic conditions, the enzyme citrate synthase cleaves citrate adenosine triphosphate independently into acetyl coenzyme A and oxaloacetate, a reaction that has been regarded as impossible under physiological conditions. Because this overlooked, energetically efficient carbon fixation pathway lacks key enzymes, it may function unnoticed in many organisms, making bioinformatical predictions difficult, if not impossible.

Most organic carbon on Earth derives from biological CO2 fixation. Six different autotrophic pathways responsible for this process are known today (13). Each of these pathways is characterized by certain key enzymes, the genes for which can be confidently recognized in (meta)genomic databases, thus allowing an estimation of the autotrophic potential reaching from single species up to entire ecosystems. The reductive tricarboxylic acid cycle (rTCA cycle, or Arnon-Buchanan cycle) is among the most ancient metabolic processes (4, 5), probably emerged from a nonenzymatic precursor (6), and is present in bacteria belonging to various phylogenetic groups (1). It is a reversal of the oxidative tricarboxylic acid (oTCA) cycle, which provides redox equivalents and energy in organisms with a respiratory metabolism. The key step of the oTCA cycle is citrate synthesis from acetyl coenzyme A (CoA) and oxaloacetate, catalyzed by citrate synthase (CS). This reaction is regarded as one of the irreversible steps in the oTCA cycle (1, 2, 7). The current consensus is that in organisms using the rTCA cycle, CS is substituted either by a reversible adenosine triphosphate (ATP)–dependent citrate lyase (ACL) (8, 9) or by homologous enzymes catalyzing the same reaction in two steps, citryl-CoA synthesis and cleavage (fig. S1) (10, 11). The presence of ACL in a bacterium is therefore regarded as a key indication for autotrophic CO2 fixation via the rTCA cycle (12). Our study challenges this concept by demonstrating, through in vivo and in vitro experiments, that citrate cleavage in the autotrophic rTCA cycle can be catalyzed by CS, which was thought to function only in the oxidative direction.

Desulfurella acetivorans is a sulfur-reducing thermophilic (growth optimum at 52° to 57°C) deltaproteobacterium capable of growing either heterotrophically with acetate as an electron donor and carbon source (13) or autotrophically with molecular hydrogen (14). In the heterotrophic route, acetate is oxidized via the oTCA cycle using CS and 2-oxoglutarate:ferredoxin oxidoreductase (15). Because the genome of D. acetivorans encodes neither ACL nor full sets of key enzymes for other known autotrophic pathways (table S1), the pathway of CO2 fixation could not be assigned yet. The genome contains pseudogenes for ACL with frame shifts in both the α and β subunits (region 484760-487933), which make it nonfunctional. Amplification of the D. acetivorans DNA fragment containing these pseudogenes and its sequencing (16) confirmed that acl was interrupted (Fig. 1A). Furthermore, the genome contains a gene for 4-hydroxybutyryl–CoA dehydratase, the characteristic enzyme of the archaeal 3-hydroxypropionate/4-hydroxybutyrate and dicarboxylate/4-hydroxybutyrate cycles (1719), but genes for other specific enzymes of these cycles are missing (table S1). Because the absence of key enzyme(s) of known CO2 fixation pathways in an organism capable of autotrophic growth can be regarded as an indication of a novel pathway, we decided to study CO2 fixation in D. acetivorans in more detail.

Fig. 1 ACL-independent citrate cleavage in D. acetivorans.

(A) ACL pseudogene in D. acetivorans genome, as compared with the corresponding fragment of the genome of a closely related species, D. multipotens, which possesses an intact acl gene. Numbers portray the pairwise nucleotides identity. bp, base pairs. (B and C) UPLC analysis of the CoA and CoA-esters formed from citrate in the reaction catalyzed by D. acetivorans cell extracts (B) and by CS and MDH from porcine heart (Sigma) (C). A260, absorbance at 260 nm.

In accord with the published data, D. acetivorans grew both heterotrophically with acetate/CO2 and autotrophically on H2/CO2 with minimal generation times of 5.7 and 4.9 hours to a density of 1.5 × 108 and 3.0 × 108 cells per ml, respectively. High activities of oTCA cycle enzymes were detected in extracts of cells grown under both conditions, with malate dehydrogenase (MDH) and CS activities being extremely high (Table 1). In contrast, no activities of key enzymes of known pathways of autotrophic CO2 fixation were detected (table S1). By ultra performance liquid chromatography (UPLC) analysis, we determined a citrate (20 mM)–, CoA (1 mM)–, and NADH (reduced form of nicotinamide adenine dinucleotide) (5 mM)–dependent formation of acetyl-CoA (Fig. 1B and fig. S2). Furthermore, a CoA- and citrate-dependent oxidation of NADH was observed that could be attributed to oxaloacetate formation followed by NADH-dependent reduction to malate by endogenous MDH (Table 1). However, none of these reactions were dependent on the presence of ATP, as would be expected if ACL or citryl-CoA synthetase/citryl-CoA lyase were operating.

Table 1 Enzymes of central carbon metabolism in D. acetivorans.

Activities were measured at 55°C. The number of biological repetitions (n) is indicated. ND, not determined; ADP, adenosine diphosphate; acc., accession number.

View this table:

To further elucidate the fate of citrate, we incubated cell extracts of D. acetivorans under anaerobic conditions in buffer containing [U-13C6]citrate in the presence of NADH and CoA. After 10 min of incubation, the reaction was stopped and subjected to 13C nuclear magnetic resonance (NMR) analysis. Though the specific 13C NMR signals for [U-13C6]citrate could still be detected, we also observed six 13C NMR multiplets as a result of product formation (Fig. 2 and table S2). On the basis of the chemical shifts, the 13C-13C coupling constants, and titration experiments, these signals were unequivocally attributed to [U-13C4]malate and [U-13C2]acetyl-CoA. Together with the UPLC data, these results provide firm evidence for the cleavage of citrate to acetyl-CoA and oxaloacetate, with the latter being further reduced to malate by MDH, which was highly active in our cell extracts (Table 1). The only possible enzyme in D. acetivorans that can catalyze an ATP-independent citrate cleavage to oxaloacetate and acetyl-CoA is CS. However, with a free-energy difference ∆G of –35.8 kJ mol–1 in the canonical direction of citrate formation [at pH 7, ionic strength 0.25, and 38°C (20)], the cleavage reaction has been regarded as impossible under physiological conditions. Even though the next reaction of the rTCA cycle—oxaloacetate reduction to malate (catalyzed by MDH)—is highly exergonic [∆G of–27.1 kJ mol–1 at pH 7, ionic strength 0.25, and 38°C (20)], the reversal of the CS reaction would require high substrate and low product concentrations. Using liquid chromatography–mass spectrometry (LC-MS), we found that the CoA/acetyl-CoA ratio in autotrophically grown D. acetivorans cells (93, table S3) was much higher than, for instance, in glucose-grown Escherichia coli [2.3 (21)], whereas the citrate concentration in D. acetivorans was typical for bacterial cells [1.4 mM compared to 2 mM in E. coli (21)]. Using these concentrations of metabolites, we calculated an equilibrium oxaloacetate concentration of 0.13 μM. Although it is low, this concentration is still in a physiological range, similar to, for example, the mitochondrial oxaloacetate concentration that is in the low micromolar range (22, 23). [U-13C4]malate and [U-13C2]acetyl-CoA could also be found when [U-13C6]citrate, CoA, and NADH were incubated with commercially available CS and MDH (from porcine heart, Sigma) (Fig. 1C and fig. S3), providing further evidence that the conversion of citrate in cell extracts is catalyzed by CS. The Michaelis constant (Km) values of CS and MDH measured in D. acetivorans cell extracts were close to those for the porcine enzymes (table S4). Apparently, the CS of D. acetivorans is not specifically adapted to citrate cleavage. This is also apparent from the high Km value for citrate, thus requiring the observed high specific activity of the enzyme.

Fig. 2 13C NMR analysis of [U-13C6]citrate cleavage catalyzed by cell extracts of autotrophically grown D. acetivorans.

The reaction mixture contained 2 mM NADH, 2 mM CoA, and 2 mM citrate. (A) Reaction after 0 min. (B) Reaction after 10 min of incubation. (C) Sample as in (B), plus 5 μg of [U-13C4]malate reference. Signals and couplings of [U-13C6]citrate, [U-13C4]malate, and [U-13C2]acetyl-CoA are indicated. For numerical values of chemical shifts and couplings, see table S2. Note that citrate is a heat stable compound and did not degrade nonenzymatically. ppm, parts per million.

In addition to citrate cleavage, we were able to detect activity of another characteristic enzyme of the rTCA cycle, fumarate reductase, in D. acetivorans cell extracts. This activity was measured as fumarate-dependent succinyl-CoA formation in an assay coupled with endogenous succinyl-CoA synthetase (Table 1). The identity of the product, succinyl-CoA, was confirmed by LC-MS analysis (fig. S4). The highest specific activity (9 nmol min−1 mg−1 protein) was observed with dithionite as the electron donor; lower activities could be detected when NADPH (reduced form of nicotinamide adenine dinucleotide phosphate) or NADH were used. No activity could be detected with reduced forms of benzyl viologen, methyl viologen, the menaquinone analogon 2,3-dimethyl-1,4-naphthoquinone, or flavin mononucleotide as electron donors. We assume that this reaction depends on a different, unidentified electron donor with low redox potential in vivo, possibly coupling fumarate reduction by H2 with the buildup of a proton-motive force, as was discussed for deltaproteobacterium Desulfuromonas acetoxidans (24).

To test the function of the rTCA cycle in D. acetivorans in vivo, we grew autotrophic cultures in the presence of 0.2 mM [1-13C]pyruvate, which was added in four 0.05 mM portions during the exponential growth phase. The cells were harvested and hydrolysed under acidic conditions. Four of the obtained amino acids—Ala, Asp, Glu, and Pro—were then converted into tert-butyl-dimethylsilyl derivatives, followed by gas chromatography–mass spectrometry analysis to determine 13C enrichments and positions in the molecules (table S5). Alanine, reflecting its precursor pyruvate, displayed ~6% 13C excess. Incorporation into aspartate, reflecting its precursor oxaloacetate, and into glutamate and proline, reflecting the precursor 2-oxoglutarate, was lower (1 to 2% 13C excess) but still significant. A more detailed MS analysis of masses comprising all carbon atoms of the original amino acids (i.e., Ala, Asp, Glu, and Pro at a mass/charge ratio of 260, 418, 432, and 286, respectively, with fragments having lost one or two carbon atoms from the respective precursor amino acids) revealed some positional assignment of the 13C label (fig. S5). Thus, comparison of 13C excess in Ala260 with that in Ala232 (still carrying C-2 and C-3 of the original Ala molecule) immediately revealed that the 13C label in Ala must have been entirely located at C-1, which is lost in the fragment, as expected from its origin from the [1-13C]pyruvate precursor. The apparently identical 13C enrichments in Glu432/Glu104 and Pro286/Pro184 suggested that the 13C label is located in positions 2 to 5 in glutamate and proline. The lower 13C excess in Asp390 (carrying C-2, C-3, and C-4 of Asp) in comparison with Asp418 (comprising all carbon atoms of Asp) indicated that the label must be present in C-1 of Asp. As shown in fig. S6, these 13C-distributions can be predicted by shuffling the [1-13C]pyruvate tracer via phosphoenolpyruvate (PEP) carboxylation into [1-13C]oxaloacetate, which enters the rTCA cycle and is converted into 2-oxoglutarate via the symmetric intermediates fumarate and succinate (fig. S6A). On the other hand, pyruvate assimilation via [1-13C]oxaloacetate followed by reactions of the oTCA cycle could not explain 13C incorporation into Glu and Pro, as the 13C label becomes lost during the conversion of isocitrate into 2-oxoglutarate (table S6 and figs. S5 and S6B). Together, these results clearly proved the existence of a functional rTCA cycle under in vivo conditions.

The operation of the rTCA cycle was also confirmed by inhibitor analysis of autotrophically growing D. acetivorans cultures. Two inhibitors of the cycle, fluoroacetate [which is converted through the CS reaction to the aconitase inhibitor fluorocitrate in the cell (25)] and glyoxylate [pyruvate:ferredoxin oxidoreductase inhibitor (26)], suppressed growth of D. acetivorans (fig. S7).

We designate this novel CS-dependent version of the rTCA cycle as “reversed oTCA cycle” (roTCA). The functioning of the roTCA cycle requires reduced ferredoxin, which is probably synthesized through electron bifurcation. Indeed, the D. acetivorans genome contains genes for a NAD-dependent ferredoxin:NADPH oxidoreductase (NfnAB) (27). Furthermore, the direct reduction of ferredoxin in a hydrogenase reaction is not implausible during growth at high H2 concentration (80% in our experiments).

The activities of the enzymes catalyzing the key reactions of the TCA cycle (Table 1 and supplementary text) were nearly the same for D. acetivorans grown organoheterotrophically in medium with acetate or lithoautotrophically with hydrogen. However, the same set of TCA cycle enzymes has to function in either the oxidative or reductive direction, depending on the growth conditions (Fig. 3, A and B). The question then arises as to how the regulation of carbon flux can occur. Two obvious possibilities are that either the presence of H2 shifts the flux toward the roTCA cycle or that exogenous acetate drives the TCA cycle in the oxidative direction. Isotopologue profiling of Desulfurella metabolism during growth with or without H2 in the presence of CO2 and [U-13C2]acetate revealed that acetate is mostly being oxidized under both conditions (tables S7 and S8 and figs. S8 and S9). These data disprove the regulatory effect of H2 and suggest that the presence of acetate itself leads to increased intracellular acetyl-CoA concentration, triggering the direction of the cycle. The detected high activities of CS and MDH (Table 1) are not required for growth under heterotrophic conditions. However, the high activities allow fast switching from heterotrophic to autotrophic growth and reflect the adaptation of D. acetivorans to fluctuating acetate concentrations characteristic for the D. acetivorans natural environment, where both active acetate formation and oxidation take place (28).

Fig. 3 A reversible TCA cycle in D. acetivorans and its efficiency in comparison with other autotrophic CO2 fixation pathways.

(A) roTCA cycle during growth on CO2, H2, and elementary sulfur (S0). (B) oTCA cycle during growth on acetate and S0. (C) ATP costs and estimated ΔG′° values for the synthesis of acetyl-CoA via the known autotrophic CO2 fixation pathways (red,anaerobic; brown, microaerobic; blue, aerobic; see table S10 for details). Note that comparison of the pathways by their ATP costs does not take into account the costs for ferredoxin reduction and thus overestimates the energetic efficiency of anaerobic pathways. Enzymes: 1, citrate synthase; 2, malate dehydrogenase; 3, fumarase; 4, fumarate reductase (roTCA cycle) or succinate dehydrogenase (oTCA cycle); 5, succinyl-CoA synthetase; 6, 2-oxoglutarate synthase; 7, isocitrate dehydrogenase; 8, aconitase; 9, pyruvate synthase; 10, pyruvate carboxylase; 11, acetate:succinyl-CoA CoA-transferase; 12, acetyl-CoA synthetase/acetate kinase + phosphate acetyltransferase.

What makes this newly discovered variant of the rTCA cycle so special? First, the roTCA cycle requires one less ATP molecule per synthesized acetyl-CoA compared with the classical ACL-dependent cycle (13). In bioenergetic terms, the roTCA cycle seems to be the most efficient pathway of autotrophic CO2 fixation known today (Fig. 3C). Still, the modified pathway is associated with an additional energy demand, as an organism using it must increase the amount of some catalysts to cope with their low substrate concentrations [note the very high activities of CS and MDH (Table 1)]. The high activities of the enzymes involved in thermodynamically unfavorable reactions ensure efficient coupling of the roTCA cycle to biosynthetic reactions. Anabolism is highly exergonic with a thermodynamic efficiency of ~40%, thus providing an additional driving force for the autotrophic pathway. Nevertheless, the required low product-to-substrate ratio and high enzymatic activities may render this strategy infeasible for some organisms or under certain growth conditions. Second, the roTCA cycle can hardly be recognized bioinformatically. It makes (meta)genome-based bioinformatic predictions of the autotrophic potential of an anaerobic organism (or microbial community) difficult, if not impossible. Third, many bacteria using the rTCA cycle to fix CO2 also possess CS genes (29, 30), and conversion of the ACL version of the rTCA cycle into the roTCA cycle appears to be quite easy. Notably, another Desulfurella species, D. multipotens, possesses an intact acl gene (Fig. 1A) but apparently uses the roTCA cycle, as can be judged from the ACL and CS activity measurements and ATP independence of citrate cleavage (table S9). The usage of the roTCA cycle could be a widespread trait of anaerobic autotrophic organisms. Considering a wide distribution of its enzymes in anaerobes, the cycle could have evolved convergently in different microbial groups. Fourth, CS is structurally simpler than ACL, which consists of several domains, with one of them being homologous to CS (3). Therefore, the roTCA cycle may even predate the modern ACL-rTCA cycle, with an ancestral TCA cycle being fully reversible.

Our results show that unexpected discoveries are possible, even when studying well-known metabolic pathways. If the highly endergonic citrate synthase reaction can be reversed in vivo, which other apparently unidirectional metabolic reactions may be reversed under certain conditions? Despite the many advances of the omics era, these features cannot be identified bioinformatically but require classical biochemical studies for their discovery.

Supplementary Materials

Materials and Methods

Supplementary Text

Figs. S1 to S9

Tables S1 to S11

References (3149)

References and Notes

  1. ACL has been found only in bacteria using the rTCA cycle for autotrophic CO2 fixation and is thus regarded as its key enzyme. Two other characteristic enzymes of the cycle, 2-oxoglutarate:ferredoxin oxidoreductase and fumarate reductase, are present in many nonautotrophic anaerobes.
  2. Materials and methods are available as supplementary materials.
Acknowledgments: We thank G. Fuchs, Albert-Ludwigs-Universität Freiburg, for his support, discussions, and suggestions during this work and for critical reading of the manuscript. This work was supported by the Deutsche Forschungsgemeinschaft (grant BE 4822/5-1 and Heisenberg Fellowship BE 4822/1-2 to I.A.B.). We also thank the Hans-Fischer Gesellschaft (Munich) for financial support. All data to understand and assess the conclusions of this research are available in the main text, supplementary materials, and the GenBank database ( The genome sequences of D. acetivorans and D. multipotens are deposited under GenBank accession numbers CP007051.1 and PRJNA262210, respectively. Amino acid sequences of candidate enzymes are available under their respective GenBank accession numbers (see Table 1 and table S1).
View Abstract

Navigate This Article