Almost all major groups of prokaryotes include representatives that are able to grow autotrophically (
33). These organisms play an essential role in ecosystems by providing a continuous supply of organic carbon for heterotrophs. The Calvin-Benson-Bassham cycle (Calvin cycle) represents the most important extant autotrophic carbon fixation pathway (
43,
50). Despite its global significance, it is restricted to organisms with high-energy yield from a chemotrophic or phototrophic lifestyle. Microorganisms present in extreme environments, e.g., high temperature or anaerobic or acidic conditions, generally utilize different CO
2 fixation pathways (
17,
33). At present, there are three alternative pathways known: the reductive tricarboxylic acid (TCA) cycle, the reductive acetyl coenzyme A (CoA) pathway, and the 3-hydroxypropionate cycle (
4,
33).
It has been proposed that the first autotrophic pathway was akin to either the reductive TCA cycle or the reductive acetyl-CoA pathway (
11,
17,
35,
45,
58). The reductive TCA cycle has the characteristics of an autocatalytic cycle and leads to a complex cyclic reaction network from which other anabolic pathways could have evolved (
11,
58): e.g., the oxidative TCA cycle (
8,
45). Based upon biochemical and isotopic analyses, the reductive TCA cycle appears to operate in phylogenetically diverse autotrophic bacteria and archaea, including genera of anoxic phototrophic bacteria (
Chlorobium) (
14,
18,
28), sulfate-reducing bacteria (
Desulfobacter) (
48), microaerophilic, hyperthermophilic hydrogen-oxidizing bacteria (
Aquifex and
Hydrogenobacter) (
5,
49), and sulfur-reducing
Crenarchaeota (
Thermoproteus and
Pyrobaculum) (
5,
24,
46). The reductive TCA cycle is essentially the oxidative TCA cycle running in reverse, leading to the fixation of two molecules of CO
2 and the production of one molecule of acetyl-CoA (Fig.
1). Acetyl-CoA is reductively carboxylated to pyruvate, from which all other central metabolites can be formed. Most of the enzymes of the two pathways are shared, with the exception of three key enzymes that allow the cycle to run in reverse: ATP citrate lyase, 2-oxoglutarate:ferredoxin oxidoreductase, and fumarate reductase. 2-Oxoglutarate:ferredoxin oxidoreductase catalyzes the carboxylation of succinyl-CoA to 2-oxoglutarate, ATP citrate lyase the ATP-dependent cleavage of citrate to acetyl-CoA and oxaloacetate, and fumarate reductase the reduction of fumarate forming succinate. The presence of these enzyme activities in autotrophically grown bacteria and archaea is indicative of a functioning reductive TCA cycle (
5,
28,
46,
48,
49). However, activity of ATP citrate lyase has also been identified in heterotrophic and facultative autotrophic organisms: i.e., the sulfate reducers
Desulfobacter postgatei and
Desulfobacter hydrogenophilus, respectively (
37,
48). In
D. postgatei and
D. hydrogenophilus growing on acetate, ATP citrate lyase catalyzes citrate synthesis, so ATP can be formed via substrate-level phosphorylation, increasing the energy yield of the bacterium. ATP citrate lyase also exists in eukaryotes, where it plays a role in supplying cytosolic acetyl-CoA for biosynthesis of fatty acids and cholesterols (
15).
The only cultured bacteria or archaea for which a complete gene encoding ATP citrate lyase has been identified to date are the green-sulfur bacteria
Chlorobium tepidum and
Chlorobium limicola (
13,
29). The ATP citrate lyase of
Chlorobium is an α
4β
4-heteromeric enzyme encoded by two separate genes,
aclA (large α-subunit; 1,827 bp) and
aclB (small β-subunit; 1,197 bp), which are adjacent to one another in the genome (
13,
29). In
Hydrogenobacter thermophilus, the ATP-dependent cleavage of citrate is catalyzed by the combined action of two recently identified enzymes, citryl-CoA synthetase (homolog to succinyl-CoA synthetase) and citryl-CoA lyase (homolog to citrate synthase) (
2,
3), possibly representing the ancient state of ATP-dependent citrate cleavage enzyme.
In recent years, the ε subdivision of the proteobacteria has received increased attention as 16S rRNA genes phylogenetically associated with ε-proteobacteria have been frequently obtained from a variety of environments, including hydrothermal vents, the oxic-sulfidic interfaces of anoxic marine basins (Cariaco Basin and Black Sea), and oilfields, indicating their importance in these systems (
22,
23,
32,
34,
38,
44,
55,
57). Based on the physicochemical environment, it has been speculated that these ε-proteobacteria could potentially be autotrophs, which is further supported by the recent isolation of several chemolithoautotrophic ε-proteobacteria (
1,
9,
26,
30,
36,
53,
54,
60). However, no information exists what pathway of autotrophic carbon fixation these bacteria might use. Recently, genes coding for ATP citrate lyase have been identified on two fosmids also containing the 16S rRNA of the main ε-proteobacterial epibionts of
Alvinella pompejana—a polychaete living on sulfide structures at deep-sea hydrothermal vents on the East Pacific Rise—suggesting that these bacteria use the reductive TCA cycle either for autotrophic CO
2 fixation or for assimilation of organic carbon (
10). On these fosmids, the genes corresponding to
aclA and
aclB were adjacent to each other in an arrangement identical to that of
Chlorobium.
MATERIALS AND METHODS
Bacteria and growth conditions.
T. denitrificans (DSM 1251) was grown autotrophically in 2-liter glass bottles at 25°C under denitrifying conditions according to Timmer-Ten Hoor (
56). Thiosulfate served as an electron donor, nitrate as an electron acceptor, and sodium bicarbonate as the sole carbon source. The average doubling time was 20 h.
Candidatus Arcobacter sulfidicus can at present only be cultivated in a special growth apparatus, resulting in the fact that no transferable pure cultures of
Candidatus Arcobacter sulfidicus are maintained on a routine basis (
60). For the present investigation, we initiated new enrichments from purified cell suspensions that were harvested previously from a purified cell culture (
60) and stored in our laboratory. The purified cells were grown autotrophically in a flowthrough system at 25°C under microaerophilic conditions using H
2S as an electron donor, O
2 as an electron acceptor, and CO
2 as a carbon source (
60). Since the organism is growing in an open system, it is difficult to determine the actual growth rate, but CO
2 fixation experiments demonstrated a fixation rate of 1.04 × 10
−6 nmol of C cell
−1 h
−1 (
60).
Chloroflexus aurantiacus OK-70-fl (DSM 636) (
51),
Desulfobacterium autotrophicum (DSM 3382) (
31),
Desulfobacter hydrogenophilus (DSM 3380) (
59), and
Ralstonia eutropha JPM 134 (DSM 4058) (
16) were grown as described in the references. The biomass of anaerobic and microaerophilic organisms was harvested under the exclusion of oxygen.
Escherichia coli TOP10 (Invitrogen, Carlsbad, Calif.) was grown on Luria-Bertani medium. Kanamycin was added to a final concentration of 50 μg/ml.
Syntheses.
Malonyl-CoA and 3-hydroxypropionate were synthesized as described elsewhere (
21) and were stored at −20°C.
Preparation of cell extracts.
Cell extracts were prepared using a mixer mill (type MM2; Retsch, Haare, Germany). Four ml buffer (100 mM Tris HCl, pH 7.8, 3 mM 1,4-dithioerythritol [DTE], 250 μl protease inhibitor cocktail for bacterial cells [Sigma-Aldrich, Saint Louis], and 1 mg of DNase I per 10 ml of buffer) was added per 1 g of wet cells. Aliquots (0.6 ml) of the suspension were placed in 1.5-ml plastic microcentrifuge vials; for anaerobic preparations, 1-ml stoppered glass vials were used and 1 g of glass beads (diameter, 0.1 to 0.25 mm) was added to the solution. From
Candidatus Arcobacter sulfidicus, only a small amount of cells was available. In this case, 20 mg of cells was resuspended in 100 μl of buffer (see above). The suspension was placed in a 0.5-ml microcentrifuge vial, and 100 mg glass beads was added. The ice-cold solution was treated in the mixer mill for 7 min at 100% intensity (30 Hz). Following a centrifugation step (10 min, 12,000 ×
g, 4°C), the supernatant was used for enzyme tests. Protein concentration in cell extracts was determined by the method of Bradford (
6) using bovine serum albumin as a standard.
Enzyme assays.
Routinely, enzyme assays (0.5-ml assay mixture) were performed in stoppered 0.5-ml glass cuvettes or 1-ml glass vials. The assay temperature was 25°C. Reactions involving pyridine nucleotides were followed spectrophotometrically at 365 nm: ε
365 nm(NAD(P)H) = 3.4 × 10
3 M
−1 cm
−1. Reactions involving methyl viologen (MV) or benzyl viologen (BV) were followed spectrophotometrically at 578 nm: ε
578 (MV) = 9.8 × 10
3 M
−1 cm
−1, and ε
578 (BV) = 8.6 × 10
3 M
−1 cm
−1 (
12).
(i) ATP citrate lyase (reductive TCA cycle).
ATP citrate lyase activity was determined by coupling the reaction to malate dehydrogenase activity, which reduces the produced oxaloacetate with NADH. The citrate-, CoA-, and MgATP-dependent oxidation of NADH was monitored. The assay mixture contained 100 mM MOPS (morpholinepropanesulfonic acid)/NaOH, pH 8.3, 5 mM DTE, 5 mM MgCl2, 3 mM ATP, 0.5 mM CoA, 0.4 mM NADH, 1 U malate dehydrogenase (Sigma-Aldrich), and 3 mM d-citrate. The reaction was started by the addition of citrate.
(ii) 2-Oxoglutarate:acceptor oxidoreductase and pyruvate:acceptor oxidoreductase (reductive TCA cycle).
The 2-oxoglutarate- or pyruvate- and CoA-dependent reduction of oxidized BV was monitored. The following anaerobic assay mixture was used with N2 gas as headspace: 100 mM Tris-HCl, pH 7.8, 4 mM DTE, 2 mM MgCl2, 0.5 mM CoA, 2 mM BV, and 3 mM 2-oxoglutarate or pyruvate. Dithionite was added by syringe from 5 mM stock solution until a faint bluish color was obtained. The reaction was started by the addition of 2-oxoglutarate or pyruvate. Two viologen dyes were reduced per 2-oxoacid transformed.
(iii) Fumarate reductase (reductive TCA cycle).
The anaerobic assay mixture contained 100 mM Tris-HCl, pH 7.8, 4 mM DTE, 2 mM MgCl2, 2 mM reduced BV, and 1 mM fumarate. Reduced viologen dye was obtained by adding dithionite (50 mM solution) until an absorption of 1.5 at 578 nm was obtained. The reaction was started by the addition of fumarate. Two viologen dyes were oxidized per fumarate transformed.
(iv) Malate dehydrogenase (reductive and oxidative TCA cycle).
The oxaloacetate-dependent oxidation of NADPH or NADH was monitored (
24).
(v) Isocitrate dehydrogenase (reductive and oxidative TCA cycle).
The isocitrate-dependent reduction of NADP
+ was monitored (
24).
(vi) Fumarate hydratase (reductive and oxidative TCA cycle).
Fumarate consumption was monitored spectrophotometrically at 240 nm (ε
240 = 2,440 M
−1 cm
−1) (
24).
(vii) 2-Oxoglutarate dehydrogenase and pyruvate dehydrogenase (oxidative TCA cycle).
In the oxidative TCA cycle, 2-oxoglutarate dehydrogenase and pyruvate dehydrogenase replace 2-oxoglutarate:acceptor oxidoreductase and pyruvate:acceptor oxidoreductase. The 2-oxoglutarate- or pyruvate- and CoA-dependent reduction of NAD+ or NADP+ was monitored. The test mixture contained 100 mM Tris-HCl, pH 7.8, 5 mM DTE, 5 mM MgCl2, 1 mM NAD+ or NADP+, 0.5 mM CoA, 0.5 mM thiamine pyrophosphate, and 5 mM 2-oxoglutarate or pyruvate. The addition of 2-oxoglutarate or pyruvate started the reaction.
(viii) Malonyl-CoA reduction to 3-hydroxypropionate (3-hydroxypropionate cycle).
The malonyl-CoA-dependent oxidation of NADPH was monitored (
24).
(ix) Reduction of 3-hydroxypropionate to propionyl-CoA (3-hydroxypropionate cycle).
The 3-hydroxypropionate-, CoA-, and MgATP-dependent oxidation of NADPH was followed (
24).
(x) Carbon monoxide dehydrogenase (reductive acetyl-CoA pathway).
The anaerobic assay mixture contained 100 mM Tris-HCl, pH 7.8, 4 mM DTE, 2 mM MgCl2, 4 mM MV, and CO gas (1-bar headspace). Dithionite was added by syringe from 5 mM stock solution until a faint bluish color was obtained. The reaction was started by the addition of protein. In control experiments, CO was replaced by N2 gas.
(xi) RubisCO.
The carboxylation of ribulose 1,5-bisphosphate with
14CO
2 was followed by measuring the substrate-dependent incorporation of
14C from [
14C]bicarbonate into acid-stable products (
24).
Ralstonia eutropha, a facultative autotroph expressing ribulose 1,5-bisphosphate carboxylase/oxygenase (RubisCO) activity also while growing on fructose as an organic C source, served as a positive control (
5).
DNA extraction, PCR amplification, cloning, and sequencing of ATP citrate lyase.
Chromosomal DNA was extracted with the “SoilMaster DNA extraction kit” (Epicentre, Madison, WI) according to the provided protocol. Degenerate primers were designed based on an alignment of known ATP citrate lyase genes. An ∼1,000-bp fragment of the gene for the ATP citrate lyase α-subunit (
aclA) was directly amplified from the purified DNA of
Candidatus Arcobacter sulfidicus and
Thiomicrospira denitrificans in a 35-cycle PCR at an annealing temperature of 42°C with degenerate primers (F2, 5′-TGCATAGCAATHGGNGGNGA-3′; and R5, 5′-CCGATAGANCCRTCNACRTT-3′). A second ∼340-bp fragment of the gene for the ATP citrate lyase β-subunit (
aclB) was amplified from both cultures using primers 892F and 1204R and conditions according to Campbell et al. (
10). PCR products were purified from agarose gels using a QIAquick gel extraction kit (QIAGEN, Inc., Chatsworth, CA) and cloned using the TOPO TA cloning kit for sequencing (Invitrogen, Carlsbad, CA) as described by the manufacturer. Colonies were picked, and the plasmid DNA was purified with the QIAprep spin miniprep kit (QIAGEN) as described by the manufacturer. Plasmids were sent to the BioResource Center at Cornell University (Ithaca, NY) for sequencing. DNA sequencing was performed using the Applied Biosystems Automated 3730 DNA analyzer. The big dye terminator chemistry and AmpliTaq-FS DNA polymerase were used.
Phylogenetic analysis.
Sequences were loaded into MacVector (Accelrys Inc., San Diego, CA), and aligned using the ClustalW program implemented in MacVector. Aligned sequences were imported into PAUP, version 4.0b10, for phylogenetic analysis (
52). Models for use in minimum-evolution and maximum-likelihood searches were chosen using the likelihood ratio test in the Modeltest program, version 3.5 (
42). The chosen substitution model corresponded to the GTR + G model: six classes of substitution and unequal base frequencies (GTR; general time reversible), the proportion of invariant sites set at 0, and the evolutionary rate of sites varying according to a gamma distribution (G). The specific model parameters are available upon request. Trees were inferred using distance analysis with minimum evolution as the optimal criterion and the distance measure set to maximum likelihood, parsimony analysis, and maximum-likelihood analysis. Heuristic searches under the minimum-evolution and maximum-parsimony criterion were performed using 1,000 random-addition replicates with tree bisection-reconnection branch swapping. Bootstrapping of minimum-evolution and maximum-parsimony analyses used 1,000 bootstrap replicates of 10 random addition replicates each. Heuristic searches under the maximum-likelihood criterion were performed using five random addition replicates with tree bisection-reconnection branch swapping.
Nucleotide sequence accession number.
The nucleotide sequences have been deposited in GenBank. The accession numbers of the section of aclA coding for the large ATP citrate lyase α-subunit and aclB coding for the small ATP citrate lyase β-subunit of T. denitrificans are AY885676 and AY885677, respectively. The accession number of the section coding for the ATP citrate lyase α-subunit of Candidatus Arcobacter sulfidicus is AY885678 .
DISCUSSION
In this communication, we present evidence that
Candidatus Arcobacter sulfidicus and
T. denitrificans, two chemolithoautotrophic organisms that belong to the ε subdivision of the proteobacteria, use the reductive TCA cycle for autotrophic carbon fixation. All enzymes of the reductive TCA cycle could be measured in these organisms, including the key enzymes ATP citrate lyase and 2-oxoglutarate:ferredoxin oxidoreductase. Based on the doubling time of 20 h, a specific CO
2 fixation rate of 48 nmol min
−1 mg cell protein
−1 follows for
T. denitrificans (for calculation of specific CO
2 fixation rates of growing cultures see reference
21). Since the reductive TCA cycle including pyruvate synthesis contains three CO
2-fixing enzymes, the minimal specific activities of the enzymes of the pathway to sustain autotrophic growth should be 16 nmol min
−1 mg cell protein
−1. The measured specific activities of all enzymes of the cycle were higher than this minimal activity and can therefore account for the growth rate of autotrophically growing cells. Unfortunately, we cannot make the same argument for
Candidatus Arcobacter sulfidicus, since the organism can presently only be consistently grown in a flowthrough system and thus we were not able to determine its actual growth rate. However, based on the measured CO
2 fixation rate of 1.04 × 10
−6 nmol of C cell
−1 h
−1 for an exponentially growing culture of
Candidatus Arcobacter sulfidicus (
60), we can calculate a specific CO
2 fixation rate of 110 nmol min
−1 mg cell protein
−1. The minimal specific activities of the enzymes should therefore be 37 nmol min
−1 mg cell protein
−1. The measured specific activities of all enzymes of the reductive TCA cycle in
Candidatus Arcobacter sulfidicus were higher than this minimal activity. However, although we used purified cells of
Candidatus Arcobacter sulfidicus for our experiments, we cannot totally exclude the growth of minor contaminants, which might have contributed to the measured enzyme activities or sequence data presented in this communication. Yet, several lines of evidence indicate that the results presented here can be attributed to
Candidatus Arcobacter sulfidicus. First, possible contaminants constituted at most 0.1% of the total cell counts (
60). Assuming we measured the enzymes of the contaminating organism(s), their enzyme activities should be at least 1,000 times higher than the measured ones, which would result in an activity of 45,000 nmol min
−1 mg cell protein
−1 for ATP citrate lyase. To our knowledge, the highest so far measured activity of ATP citrate lyase in crude extract was 1,300 nmol min
−1 mg cell protein
−1 (
D. hydrogenophilus) (
48). Second, based on the stable carbon isotopic composition of the total biomass formed in the reactor, we have previously determined an isotopic discrimination value of around 13 ‰ (
60), which is in line with the conclusion that the dominant organism in our enrichment is using the reductive TCA cycle for autotrophic carbon fixation. A possible contaminant would not be expected to shift the stable carbon isotopic composition to that extent. Third, the obtained ATP citrate lyase sequences clearly cluster with sequences obtained from other ε-proteobacteria, which indicates that we sequenced the fragments of a ε-proteobacterium.
Another severe problem in performing physiological experiments with Candidatus Arcobacter sulfidicus is that for further analyses the cells have to be freed of the filamentous sulfur produced during their metabolism, resulting in a significant loss of cells and hence a low biomass available for enzyme measurements. As part of this work, we successfully established an efficient cell breakage protocol for measuring enzyme activities in samples of only 10 to 20 mg wet weight, which will also prove useful for other experiments where the amount of available biomass is an issue.
This is the first report that members of the ε subdivision of the proteobacteria use the reductive TCA cycle for autotrophic carbon fixation, and it confirms earlier results for
Candidatus Arcobacter sulfidicus that indicated that this organism is using a carbon fixation pathway other than the Calvin cycle for autotrophic CO
2 fixation (
60). It should be noted that
T. denitrificans belongs to the ε subdivision of the proteobacteria (
39), whereas the
Thiomicrospira species belonging to the γ subdivision of the proteobacteria have been shown to use the Calvin cycle for autotrophic carbon fixation (
7). Thus our data reemphasize the need to reclassify
T. denitrificans. Based on the presence of the ATP citrate lyase gene, it was previously postulated that bacteria of the ε subdivision of the proteobacteria might use the reductive TCA cycle for autotrophic growth, but no biochemical data were presented (
10). Although indicative for this pathway, the presence of this gene is not sufficient to unequivocally demonstrate that the reductive TCA cycle is used as an autotrophic pathway. For example, the obligate organotrophic bacterium
D. postgatei uses the reductive TCA cycle for acetate metabolism (
37) and the facultative autotroph
D. hydrogenophilus has been shown to use ATP citrate lyase in both directions, depending on whether it grows autotrophically or with acetate as the carbon source (
48).
Based on our data,
Candidatus Arcobacter sulfidicus and
T. denitrificans—like
C. limicola and
C. tepidum—use ATP citrate lyase for the cleavage of citrate. In contrast,
H. thermophilus, a member of the
Aquificales, has recently been shown to use citryl-CoA synthetase and citryl-CoA lyase for the same reaction (
2,
3). It has been speculated that the reaction sequence present in
H. thermophilus might represent an ancient form of the reductive TCA cycle and that the use of ATP citrate lyase like in
Chlorobium was a later innovation (
2,
3). This scenario is in line with 16S rRNA phylogeny, which places
H. thermophilus at the base of the bacterial domain (
40). In this regard, it is interesting to consider the phylogenetic relationship between δ- and ε-proteobacteria, which are currently the only subdivisions within the proteobacteria to have members using the reductive TCA cycle. It has been hypothesized that ε-proteobacteria, together with δ-proteobacteria, represent the most ancient subdivision of the proteobacteria and that together they might have shared a common ancestor with a group containing green sulfur bacteria and
Aquificales (
19), taxonomic groups encompassing several species which use the reductive TCA cycle for autotrophic carbon fixation. The finding of the reductive TCA cycle in ε-proteobacteria as described here is in line with this hypothesis. However, the database of prokaryotic
acl sequences is very limited at this point and the exact reconstruction of the evolutionary history of ATP citrate lyase has to await further sequence data.
Our data lend support to the argument that autotrophic carbon fixation through the reductive TCA cycle is widespread and possibly more important in different environments than previously thought. This seems to be especially true for hydrothermal environments, where autotrophic ε-proteobacteria probably contribute substantially to primary production (
10,
53). ε-Proteobacteria have been identified as a major, if not dominant, component of microbial communities in deep-sea hydrothermal vents, ranging from black smoker chimney walls and associations with invertebrates to the shallow subsurface (
20,
22,
23,
41,
44). The habitat of these microorganisms seems to be confined to the zone where hydrothermal fluids mixes with ambient seawater, creating conditions that are characterized by temperatures in the range from ∼20 to ∼60°C and microaerobic to anaerobic conditions. Since organisms living under these conditions are potentially energy limited, it seems fitting that these microbes are using a carbon fixation pathway other than the Calvin cycle, which requires 9 ATP to synthesize one triose phosphate molecule compared to 5 ATP for the reductive TCA cycle (
33).
At present, the reductive TCA cycle is the only confirmed autotrophic pathway in ε-proteobacteria and it is not known whether other autotrophic ε-proteobacteria might use a different carbon fixation pathway or not. For example, it is known that autotrophic members of the δ-proteobacteria, which represent a sister group of ε-proteobacteria, use either the reductive TCA cycle or the reductive acetyl-CoA pathway for CO
2 fixation (
47,
48). Current work aims at identifying the carbon fixation pathway in other autotrophic ε-proteobacteria, including recent isolates from hydrothermal environments. This will not only result in a better understanding of the carbon fixation pathways used by ε-proteobacteria, but also result in an increased database of
acl sequences, which is needed to reconstruct the evolution of ATP citrate lyase. However, it is interesting to note that the phylogeny based on presently available
aclAB sequences is in congruence with 16S/18S rRNA phylogeny, indicating the potential usefulness of this gene not only as a functional but also a phylogenetic marker. The origin of ATP citrate lyase in eukaryotes also remains an unresolved and interesting question, and in this context it will be particularly helpful to obtain
acl sequences from deep-branching microbial eukaryotes.