Coupling of the pathway of sulphur oxidation to dioxygen reduction: characterization of a novel membrane-bound thiosulphate:quinone oxidoreductase


  • Fabian H. Müller,

    1. Institute of Microbiology and Genetics, Darmstadt University of Technology, Schnittspahnstraße 10, D-64287 Darmstadt, Germany.
    Search for more papers by this author
  • Tiago M. Bandeiras,

    1. Instituto de Tecnologia Química e Biológica, Universidade Nova de Lisboa, Apt 127, Av. da República (EAN), 2781-901 Oeiras, Portugal.
    Search for more papers by this author
  • Tim Urich,

    1. Institute of Microbiology and Genetics, Darmstadt University of Technology, Schnittspahnstraße 10, D-64287 Darmstadt, Germany.
    Search for more papers by this author
  • Miguel Teixeira,

    1. Instituto de Tecnologia Química e Biológica, Universidade Nova de Lisboa, Apt 127, Av. da República (EAN), 2781-901 Oeiras, Portugal.
    Search for more papers by this author
  • Cláudio M. Gomes,

    1. Instituto de Tecnologia Química e Biológica, Universidade Nova de Lisboa, Apt 127, Av. da República (EAN), 2781-901 Oeiras, Portugal.
    2. Departamento de Química, Faculdade de Ciências e Tecnologia, Universidade Nova de Lisboa, 2825–114 Caparica, Portugal.
    Search for more papers by this author
  • Arnulf Kletzin

    Corresponding author
    1. Institute of Microbiology and Genetics, Darmstadt University of Technology, Schnittspahnstraße 10, D-64287 Darmstadt, Germany.
      E-mail; Tel. (+49) 6151 16-5242; Fax (+49) 6151 16-2956.
    Search for more papers by this author

E-mail; Tel. (+49) 6151 16-5242; Fax (+49) 6151 16-2956.


Thiosulphate is one of the products of the initial step of the elemental sulphur oxidation pathway in the thermoacidophilic archaeon Acidianus ambivalens. A novel thiosulphate:quinone oxidoreductase (TQO) activity was found in the membrane extracts of aerobically grown cells of this organism. The enzyme was purified 21-fold from the solubilized membrane fraction. The TQO oxidized thiosulphate with tetrathionate as product and ferricyanide or decyl ubiquinone (DQ) as electron acceptors. The maximum specific activity with ferricyanide was 73.4 U (mg protein)−1 at 92°C and pH 6, with DQ it was 397 mU (mg protein)−1 at 80°C. The Km values were 2.6 mM for thiosulphate (kcat = 167 s−1),  3.4 mM  for  ferricyanide  and  5.87 µM for DQ. The enzymic activity was inhibited by sulphite (Ki = 5 µM), metabisulphite, dithionite and TritonX-100, but not by sulphate or tetrathionate. A mixture of caldariella quinone, sulfolobus quinone and menaquinone was non-covalently bound to the protein. No other cofactors were detected. Oxygen consumption was measured in membrane fractions upon thiosulphate addition, thus linking thiosulphate oxidation to dioxygen reduction, in what constitutes a novel activity among Archaea. The holoenzyme was composed of two subunits of apparent molecular masses of 28 and 16 kDa. The larger subunit appeared to be glycosylated and was identical to DoxA, and the smaller was identical to DoxD. Both subunits had been described previously as a part of the terminal quinol:oxygen oxidoreductase complex (cytochrome aa3).


The dissimilatory oxidation of elemental sulphur (S°) and inorganic sulphur compounds is one of the most important energy-yielding processes in chemolithoautotrophic microorganisms, not only from solfataras and volcanic hot springs, but also from many other habitats. In addition, sulphur compound oxidation occurs in bioleaching, a process that is sometimes desired (metal recovery from low-grade ores) and sometimes not (acid mine drainage). Many different prokaryotes, both from the bacterial and from archaeal domains, thermophiles or mesophiles, are capable of oxidizing inorganic sulphur compounds. Also, many different enzymes have been described from a multitude of prokaryotes, which are (potentially) involved in these processes (reviewed, for example, in Kelly, 1982; Schönheit and Schäfer, 1995; Kelly et al., 1997; Friedrich, 1998; Amend and Shock, 2001; Friedrich et al., 2001).

Little is known about the sulphur oxidation pathways in Archaea, although they constitute a large proportion of the population of volcanic hot springs, and many of them have been termed as ‘sulphur-dependent’ (Stetter and Zillig, 1985). Our model organism, Acidianus ambivalens, belongs to the Sulfolobales order within the kingdom of crenarchaeota (Zillig et al., 1986; Fuchs et al., 1996). A. ambivalens is a chemolithoautotrophic acidophile growing optimally at 80°C and pH 2.5. It oxidizes S° to sulphuric acid under aerobic conditions, whereas under anaerobic conditions it uses hydrogen as the electron source for S° reduction to hydrogen sulphide (Zillig et al., 1985; 1986; Laska et al., 2003).

Usually, the first intermediate of S° oxidation is sulphite, which is further oxidized, sometimes indirectly, to the final product sulphate (Pronk et al., 1990; Kelly et al., 1997). The initial step in aerobically growing A. ambivalens cells is mediated by a soluble sulphur oxygenase reductase (SOR), a unique enzyme catalysing the simultaneous S° oxygenation and disproportionation to sulphite, thiosulphate (TS) and sulphide (Kletzin, 1989). The SOR or sor genes are found only in some thermophilic archaea and in the hyperthermophilic bacterium Aquifex aeolicus, but not in thiobacilli or in Paracoccus (Urich et al., 2004). The SOR from A. ambivalens typifies a new family of low-potential mononuclear non-haem iron proteins (Kletzin, 1994; Urich et al., 2004). It has not been solved whether TS is a primary product of the enzyme or whether TS formation results from the rapid non-enzymic reaction between S° and sulphite under the assay conditions (Kletzin, 1989). A. ambivalens does not grow on TS as sole sulphur-containing substrate in our hands (K. Lauber and A. Kletzin, unpubl.). The SOR does not couple S° oxidation with the aerobic electron transport chain and the fate of the products was not clear. Therefore, other enzymes have to be present to couple sulphite, TS or sulphide oxidation to proton transport across the membrane. A sulphite:acceptor oxidoreductase activity was described from membrane fractions, but the protein has not been purified yet (Zimmermann et al., 1999).

Thiosulphate-oxidizing enzymes have been studied from different bacteria. The multisubunit, periplasmatic multienzyme complexes termed as thiosulphate oxidase multienzyme complex (TOMES from P. versutus) or Sox complex (P. pantotrophus, P. denitrificans and other bacteria) seemed to be rather similar in structure and protein composition (Kelly et al., 1997; Friedrich et al., 2001). The Sox complex from P. pantotrophus oxidizes sulphite, sulphide and S° besides TS (Rother et al., 2001). The Sox complex from another isolate oxidizes tetrathionate (TT) as well (Mukhopadhyaya et al., 2000). One of the characteristics of this complex is that it binds TS covalently via a cysteinyl residue. Both sulphur atoms are oxidized to sulphate without formation of free intermediates. The electrons are transferred to c-type cytochromes (Quentmeier and Friedrich, 2001). It has been postulated that this complex constitutes a widespread mechanism in many bacteria regardless whether the organisms grow under aerobic or anaerobic conditions (Friedrich et al., 2001). However, Sox or TOMES complexes are only known so far from microorganisms growing at neutral or near-neutral pH (Kelly et al., 1997).

An alternative reaction for TS oxidation with free TT as the product has been described from studies with other microorganisms, mostly thiobacilli, in what has been described as the ‘S4 intermediate sulphur oxidation pathway’ (Trudinger, 1961a; Kelly, 1988; Meulenberg et al., 1993):

2 S-SO32 −   +  2 acox → O3S-S-S-SO32 −   +  2 acred(1)

The TT is further oxidized after hydrolysis via various intermediates to sulphate. The known TS-oxidizing and TT-forming enzymes are termed as TS dehydrogenase (EC number They use either c-type cytochromes or artificial electron acceptors (ac) as co-substrates and/or contain themselves c-type haems. Several of them have been characterized from the periplasmatic or the ‘soluble’ fractions of TS-grown cells of both neutrophilic and acidophilic Thiobacillus species (summarized by Visser et al., 1997; see also Nakamura et al., 2001). None of them was membrane bound. The subunit composition and the sizes of the enzymes differed considerably, suggesting that they do not stem from a common and homologous enzyme family (Visser et al., 1997). To our surprise, there was no sequence information about TS dehydrogenases available to test this hypothesis, whereas the deduced amino acid sequences of the Sox complexes and of subunits thereof have been well analysed from many microorganisms and genome sequences (Friedrich et al., 2001).

Here, we describe the purification of a membrane-bound, tetrathionate-forming thiosulphate:quinone oxidoreductase (TQO) and simultaneously the first enzyme coupling sulphur compound oxidation with quinone reduction in Archaea. In addition, this is the first time that a TT-forming and TS-oxidizing enzyme is characterized on the molecular level at all. In the course of our studies we could also show that the quinol:oxygen oxidoreductase of A. ambivalens is simpler and composed of less subunits than previously thought (Purschke et al., 1997).

Results and discussion

The TQO is a membrane protein

Extracts of aerobically grown A. ambivalens cells were analysed for the presence of TS-oxidizing activity (TQO). A specific activity of 1.5 U mg−1 protein was measured in the membrane fraction with ferricyanide as artificial electron acceptor. No activity was found in the soluble fraction (Table 1). This is in contrast to TS dehydrogenases from thiobacilli, which had been found in the soluble or periplasmatic fractions (summarized in Visser et al., 1997).

Table 1.  Purification table of the TQO from cell extracts.
FractionTotal protein (mg)Specific activity
[U (mg protein)−1]
Total activity (U)Yield (%)Purification factor
  • a

    . Specific activity with DQ: 7.3 mU (mg protein)−1.

  • b

    . Specific activity with DQ: 248 mU (mg protein)−1.

  • c

    . F1 and F2 represent different fractions from the same gel permeation chromatography run.

  • d

    . Specific activity with DQ: 397 mU (mg protein)−1.

Soluble fraction16 850 0     0
Membrane fraction11 500 1.5a17 000  
Solubilized membrane fraction 7 800 2.418 625100 1
Q-Sepharose flow-trough   66817.211 887 63.8 7.4
Hydroxyapatite eluate    6041.5b 2 490 13.417.3
Gel permeation eluate F1c    4749.9d 1 746  9.420.8
Gel permeation eluate F2c    1247.9   560  3.020.0

The enzyme was purified from the dodecyl maltoside (DM)-solubilized membrane fraction in a three-step chromatographic procedure using Q-Sepharose, hydroxyapatite and gel permeation chromatography. The final step resulted in a 21-fold enrichment with a recovery of 9.4% of the total activity and a yield of 59 mg of nearly pure protein (Table 1; Fig. 1). With this fraction most of the experiments were performed. An additional DEAE Sepharose column resulted in a further reduction of contaminating bands (Fig. 1); however, the protein losses were high. The TQO was separated from the fully active terminal quinol:oxygen oxidoreductase in the first chromatographic step (Q-Sepharose, Fig. 1).

Figure 1.

A. Coomassie-stained SDS polyacrylamide gel (12% acryl amide; Schägger and von Jagow, 1987) of fractions of the purification stages of the TQO. M, molecular mass marker; Qseph, Q-Sepharose column flow-through; HTP MF and SF, active hydroxyapatite eluate, main and side fraction; S200, gel permeation chromatography eluate, main and side fraction; DEAE, DEAE-Sepharose column flow-through; aa3, purified terminal quinol:oxygen oxidoreductase (cytochrome aa3); DoxA and DoxD, subunits of the TQO.
B. SDS gel of the terminal oxidase preparation (15% acryl amide; reproduced from Purschke et al., 1997; with permission).

The TQO consists of two different subunits and contains bound quinones

SDS–PAGE analysis of the purified protein showed that the TQO consists of two subunits with apparent molecular masses of 16 and 28 kDa (Fig. 1). The 28 kDa but not the 16 kDa band could be stained with Alcian blue indicating significant glycosylation of the subunit (not shown). The apparent molecular mass of the native TQO determined by gel permeation chromatography was 102 kDa.

The absorption spectrum of the aerobically purified yellow-coloured enzyme showed sharp and narrow peaks at 280 and 330–333 nm and a very broad absorption shoulder between 400 and 600 nm (Fig. 2). Maxima were visible at 307–311 and 512–517 nm in a reduced minus oxidized difference spectrum and a minimum at around 405 nm; however, the amplitude of the signal was low. The absorption and the fluorescence emission spectra of the solution obtained by organic extraction of cofactors were nearly identical to that of caldariella quinones (CQ) isolated from Sulfolobus acidocaldarius cells (Fig. 2, fluorescence date not shown; Anemüller and Schäfer, 1990; Lemos et al., 2001). Mass spectra of the quinone fractions isolated from the purified enzyme, from the membrane fraction and from active fractions of the purification intermediates showed peaks at 630, 628 and, in lower intensity, 598 (not shown) consistent with the masses of the caldariella quinones CQ-6(12H) and CQ-6(10H) and the sulfolobus quinone SQ-6(12H) (Thurl et al., 1986). The fingerprint of fragment masses also matched that of a mixture of the three isolated quinones. It was concluded from the results that the same mixture of quinones present in the membranes of aerobically grown cells (Thurl et al., 1986; Trincone et al., 1989) is also found in the isolated TQO.

Figure 2.

Left. UV/Vis spectra of TQO, of quinone extracted from purified TQO and of CQ (Anemüller and Schäfer, 1990). Inset: reduced, oxidized and reduced minus oxidized difference spectrum of isolated TQO.
Right. CD spectrum of TQO at a concentration of 100 µg ml−1 smoothed with a program supplied by the manufacturer of the instrument.

A ratio of 4 mol quinones per mol TQO holoenzyme was found in active fractions after the hydroxyapatite column and 2 mol quinones per mol after the gel filtration column when using the specific absorption coefficient of ɛ = 1137 M−1 cm−1 for CQ at 460 nm (T. M. Bandeiras, C. M. Gomes and M. Teixeira, unpubl. data). These data suggest a loss of the cofactor during the purification procedure and were consistent with a decrease in absorption (not shown). The amount of trace metals determined by X-ray fluorescence spectroscopy was 0.09 mol Ca, 0.05 mol Fe, 0.06 mol Cu and 0.07 mol Zn/mol of holoenzyme respectively. All other metals were below the limits of detection. These results suggest that no metals are present as additional redox-active cofactors besides CQ and SQ.

Ferricyanide and a quinone can be used as electron acceptors

TQO activity was routinely measured at 80°C and pH 6 with ferricyanide as electron acceptor. The suboptimal reaction conditions (see below) were chosen for the lower background.   The   TS   oxidation   rate   was   445 nmol ml −1 min−1 with 9 µg of TQO per ml corresponding to a specific activity of 49 U (mg protein)−1(Fig. 3A). The background reaction rate was 10 nmol ml −1 min−1, 2.2% of the total TS oxidation rate. A specific activity of 397 mU (mg protein)−1 was obtained with decylubiquinone (DQ) as electron acceptor (Fig. 3B), thus suggesting that the native CQ or SQ is the physiological electron acceptor (Thurl et al., 1986; Trincone et al., 1989). The velocity of DQ reduction increased when higher amounts of TQO were used, suggesting that DQ reduction is TQO dependent (Fig. 3B). Horse heart cytochrome c was not reduced; in addition, no c-type cytochromes have been found in Acidianus, Sulfolobus and relatives so far. Thus the TQO represents the first enzyme of a novel EC class because of the electron acceptor quinone, for which the number is proposed.

Figure 3.

A. Kinetics of TQO enzyme activity (control: buffer) with ferricyanide as electron acceptor measured at 420 nm and 80°C. Arrows indicate the addition of TS or sulphite.
B. Kinetics of TQO enzyme activity with decyl ubiquinone as electron acceptor measured at 291 nm and 80°C, various amounts of enzyme as indicated and with a reaction inhibited with sulphite.
C. Time-course of TS consumption (▴, with TQO; •, with buffer) and TT formation (◆, with TQO; ×, with buffer) measured at 60°C and suboptimal substrate concentrations; the concentrations of TS and TT were determined at given time points with the combined cyanolytic determination method (Kelly et al., 1969).
D. Kinetics of back reaction with reduced methylene blue as electron donor, measured anaerobically at 663 nm and 80°C. Arrows indicate the addition of TT and TQO or buffer to the reaction mixture.

The activity measured with DQ as electron acceptor depended on the order of addition of the compounds. Activity was observed only when DQ or enzyme was added last but not when DQ was added before TS (not shown). In addition, the DQ was further reduced when adding a trace of sodium borohydride after the reaction had completed. These results were interpreted that the quinone binds rather tightly to the enzyme, when solubilized from the membrane, thus preventing the normal substrate turnover by saturation of the binding site.

Tetrathionate is the product of the TQO reaction

The TS consumption and TT production were followed discontinuously in a different experiment (Fig. 3C). The stoichiometry of the conversion of TS into TT was 1 ± 0.04 mM TS into 0.48 ± 0.14 mM TT at the end of the reaction. The amount of TT produced was at any time point of the TQO reaction approximately the half of the amount of TS consumed. No tetrathionate formation was observed in the absence of TQO. The suboptimal reaction conditions (60°C, 2 mM TS and 1 mM ferricyanide) were chosen to demonstrate the kinetics of the reaction, which proceeded to fast to resolve the TT formation under optimal conditions.

The TQO reaction could be reversed when using reduced methylene blue as electron donor (Fig. 3D). The specific activity was 4.8 U (mg protein)−1 with this assay. These results and the other activity measurements showed that the TQO reversibly catalyses the reaction depicted in Eq. 1 most probably with CQ/SQ and not cytochrome c as the physiological electron acceptor.

The TQO reaction followed Michaelis-Menten kinetics under standard assay conditions. An apparent Km value of 2.6 mM TS was calculated from the Lineweaver–Burk plot using substrate concentrations below 4 mM. The apparent Vmax was 78 U (mg protein)−1, and a TS turnover rate (kcat) of 167 s−1 was obtained, giving a specificity constant of 64.000 s−1 M−1. Apparent Km values of 3.4 mM and 5.87 µM were calculated for the electron acceptors ferricyanide and DQ, when using concentrations below 1.2 mM and 32 µM, respectively, supporting the conclusion that a quinone is the native electron acceptor. The Km and kcat values for TS are not very favourable; however, they are within the same order of magnitude of what has been described for enzymes from Thiobacillus ferrooxidans and T. neapolitanus, which show the same catalytic activity (summarized in Visser et al., 1997; see also Nakamura et al., 2001). In contrast, considerable differences occurred in the Vmax where the TS dehydrogenases fell into two classes, one slow [100–300 U (mg protein)−1] and the other fast [1200–4500 U (mg protein)−1]. The A. ambivalens TQO seems to be one of the slow-class enzymes from the data presented here.

TQO activity is inhibited by sulphite

The reaction product TT did not inhibit the TQO activity up to a concentration of 1 mM, and only slightly at higher concentrations (Table 2); neither did sulphate. The thiol-binding reagent N-ethylmaleiimide, zinc ions and the reductant titanium citrate showed moderate inhibition. The TQO activity was strongly inhibited by sulphite, with an apparent Ki of 5 µM (Fig. 3A, Table 2). Metabisulphite, dithionite and Triton X-100 were also found to be strong inhibitors (Table 2). It was interesting to note that the TQO was inhibited by sulphite and other intermediate oxyanions of sulphur, although it is thought that one of the primary products of S° oxidation is sulphite. The sulphite inhibition, however, has been seen with some of the TS dehydrogenases described to date (summarized in Visser et al., 1997). This has been attributed rather to a tight fit of sulphite into the TS binding site because of the similar sizes and surface charges of the molecules than to an effect of the reducing powers of sulphite (Eo′ = −527 mV; Thauer et al., 1977) or dithionite. This interpretation is supported by the observation that the strong reductant titanium citrate is only moderately inhibitory both in the oxidized and reduced form (Table 2). It was also interesting to note that the detergent Triton X-100 was an inhibitor. In fact, several attempts failed to solubilize the TQO from membrane fractions and resulted in inactive enzyme when using detergents other than DM (data not shown).

Table 2.  Inhibition of the TQO activity by various substances.
SubstanceConcentrationInhibition as % activity of control
  • a

    . With ferricyanide and with DQ as electron acceptors.

  • b

    . As prepared anaerobically (Zehnder and Wuhrmann, 1976).

  • c

    . Air-oxidized.

  • d

    . Assay performed at 60°C because of turbidity at higher temperatures.

Control: gel permeation buffer100
Sulphate 1 mM 96
Tetrathionate 1 mM 96
10 mM 50
Sulphite50 µma  0a
10 µM 12
 5 µM 52
 1 µM 88
Metabisulphite 1 mM  0
Dithionite 1 mM  0
Titanium citrate reducedb 1.0 mM 35
 0.5 mM 77
Titanium citrate oxidizedc 1.0 mM 46
 0.5 mM 92
N-ethylmaleiimide 1 mM 46
Zn2+ 1 mM 63
 5 mM 18
Triton X-100d 1%  0

pH and temperature dependence

The optimum of the TQO activity analysed at 80°C in two different buffer systems, BisTris/citric acid/HCl buffer and citric acid/phosphate (McIlvaine) buffer, was around pH 5 in both cases (Fig. 4), which is similar compared to most of the TS dehydrogenases (Visser et al., 1997). The pH optimum is thus between the optimal pH of the growth medium (2.5) and the cytoplasm (6.5; Moll and Schäfer, 1988). The pH range of activity was much broader in BisTris than in McIlvaine buffer. The rate of the non-enzymic background reaction increased with decreasing pH, especially at pH 4.5 and below, consistent with the lower chemical stability of the substrate TS (Roy and Trudinger, 1970, p. 18). TS tends to disproportionate at elevated temperature and low pH into sulphite, sulphate and S°; in contrast, sulphite incubated with S° reacts rapidly to TS at 80°C and pH 6 (Roy and Trudinger, 1970; Kletzin, 1989).

Figure 4.

pH (left) and temperature dependence (right) of TQO activity. Inset: arrhenius plot of temperature dependence. ▴, BisTris/citric acid/HCl buffer; ▪, McIlvaine buffer; straight lines, enzymic reaction (left axes); dashed lines, non-enzymic thiosulphate oxidation (right axes).

Increasing enzyme activity was observed in the range between 20 and 92°C; a maximum was not observed because of technical reasons (Fig. 4; Zillig et al., 1986). The rate of the non-enzymic background reaction increased slightly with the temperature (not shown). No significant discontinuities were observed in the Arrhenius plot of this process (Fig. 4). An activation energy of 37.7 kJ mol−1 was calculated from the plot.

The 28 and 16 kDa subunits are identical to DoxA and DoxD from the A. ambivalens quinol:oxygen oxidoreductase

The N-terminal amino acid sequence of the 28 kDa subunit of the TQO was determined to be MERVTIIGxI. Database searches showed that this was identical to the N-terminus of the 168 aa DoxA subunit previously described as part of the terminal oxidase (calculated mass: 18.7 kDa, observed apparent mass: 28 kDa; Figs 1 and 5; Purschke et al., 1997). The N-terminus of the 16 kDa subunit was blocked. A MALDI-TOF analysis of the trypsin-digested fragments resulted in a fingerprint where 19 out of 23 fragments matched the calculated pattern of DoxD, another 184 aa subunit described as part of the terminal oxidase (calculated mass: 20.4 kDa, observed apparent mass: 16 kDa; Figs 1 and 5).

Figure 5.

Multiple alignment of the combined A. ambivalens DoxD and DoxA amino acid sequences with homologues. Conserved residues are shown with black shadow, and conservative exchanges are shown in grey. Predicted helical regions are shown with cylinders, and transmembrane helices are shown with the central part in light grey and the helix caps shaded. *, stop codons of the genes; |, end of the DoxA sequence; ♯, conserved cysteine residue. Accession numbers (GenBank/SWISSPROT): Acidianus ambivalens DoxD and DoxA, CAA70827 and CAA70828; S. tokodaii DoxD and DoxA, NP_377837 and NP_377838; Sulfolobus solfataricus DoxD and DoxA, NP_343149 and NP_343148; A. ambivalens DoxD2 and DoxA2, CAC86936 and CAC86935; Acidithiobacillus ferrooxidans-fused DoxDA and DoxDA2 translated from the genome sequence, NC_002923; Bacteroides thetaiotaomicron-fused DoxDA, NP_809428; B. fragilis-fused DoxDA translated from the genome sequence, NC_003228; Neisseria meningitides DoxX, CAB83537; Mesorhizobium loti DoxX, NP_103189; S. solfataricus DoxD2 and DoxA2, NP_343280 and NP_343281; S. tokodaii DoxX, NP_376777.

Both subunits migrate abnormally in SDS gels as previously observed (Purschke et al., 1997). This anomalous behaviour is frequently observed in very hydrophobic (37.5% for DoxD and 32% for DoxA) and for glycosylated proteins. The high apparent molecular mass of DoxA (28 kDa) was most probably the result of the unspecified glycosylation. The DoxD subunit was not glycosylated and migrated at a lower molecular mass than calculated. Extensive proteolytic processing of the subunit could be excluded as the MALDI-TOF fragments covered the amino acid positions 5–179 out of a total of 184. This would not leave space for signal sequences being cleaved off during the process of integration of the protein into the membrane and to explain the low apparent mass.

On the basis of the molecular masses of the subunits, of the native enzyme and of the glycosylation, we suggest an α2β2-tetramer for the quaternary structure of the enzyme, corresponding to a calculated mass of 96 kDa when using the sequence-deduced mass of the DoxD subunit and the experimentally determined apparent molecular mass of the DoxA subunit. This is in good agreement with the experimentally determined mass of the holoenzyme.

Topology prediction showed that the DoxD subunit most probably forms four membrane-spanning helices and DoxA one helix (Fig. 5). The transmembrane helix of DoxA was predicted to be localized at the N-terminus with most of the protein oriented to the outside of the cytoplasmatic membrane. This is also that part of the subunit that is expected to be glycosylated.

The average alpha helix content was 46% as calculated from CD spectra in the range between 190 and 250 nm (Fig. 2). The average helix length was 12 residues; the beta sheet contents was 11.6%. The high average length of the helices is in accordance with the predicted transmembrane helices. Secondary structure prediction was performed on the soluble parts of the protein separately. When combing the results of both predictions, a higher beta sheet (25%) and lower alpha helix content (33%) were observed (Fig. 5). The discrepancy probably results from the less accurate prediction of the programs for membrane proteins.

The doxD and doxA genes form a bicistronic operon in the A. ambivalens genome, physically separated from the doxBCEF operon encoding the catalytic and the other subunits of the terminal quinol:oxygen oxidoreductase (Purschke et al., 1997). Subunits with apparent molecular masses of 56 (DoxB) and 61 kDa (DoxC) appeared in our preparation of the terminal oxidase but none with masses of 20 and 28 kDa (the 95 kDa band was probably a contamination; Fig. 1). These results support the conclusion that both enzymes form separate entities and that the previously observed co-purification (Fig. 1; Anemüller et al., 1994; Purschke et al., 1997) was the result of a different procedure.

The DoxB and DoxC subunits migrate at a higher molecular mass than calculated from the deduced amino acid sequence (Fig. 1) and also than previously reported (Purschke et al., 1997). Alcian blue staining showed that both subunits were probably glycosylated (not shown), thereby explaining the differences.

In consequence, the models of the function of the subunits of the terminal oxidase have to be revised. The haem and a3-bearing DoxB subunit is the only one with low but significant similarity to other terminal oxidases. None of the sequences of what had been thought to be auxiliary subunits (DoxA, -C, -D, -E and -F) allowed a meaningful alignment with subunits of known terminal oxidases. DoxA had been proposed to be a ‘pseudosubunit II’ and DoxC a ‘pseudosubunit III’ (Purschke et al., 1997). It became obvious with the results presented here, that DoxA and DoxD do not belong to the quinol oxidase, which consists mainly of DoxB and DoxC. DoxE did not appear in our preparations, whereas DoxF has not been observed so far (Purschke et al., 1997). The reduction of subunits of the quinol oxidase found here makes it more similar to other bacterial oxidases although the exact function of the DoxC and DoxE subunits remains obscure.

DoxD and DoxA represent a novel enzyme family of membrane-bound oxidoreductases

Several other doxD/doxA genes or gene pairs were identified with significant similarity when comparing the deduced amino acid sequences of the A. ambivalens doxD and doxA genes to public databases. The amino acid sequences of two of them were highly similar to DoxD and DoxA from A. ambivalens (from Sulfolobus solfataricus and S. tokodaii, each 74% identity respectively; Figs 5 and 6). It can be assumed from the degree of similarity that they also encode TQO proteins. Two other homologous doxDA gene pairs with lower similarity in the amino acid sequences were also found in A. ambivalens and S. solfataricus. They were 70% identical to each other and 37% identical to A. ambivalens DoxDA (termed as DoxD2 and DoxA2; Figs 5 and 6). It is not known whether these genes encode proteins with TQO activity or a different function. They branched next to the three TQOs in a phylogenetic dendrograms both of the DoxD (Fig. 6) and of the DoxA domains calculated separately (not shown).

Figure 6.

Phylogenetic dendrogram of DoxD subunit or domain amino acid sequences calculated from the alignment shown in Fig. 5, using maximum likelihood tree reconstruction. The alignment was shortened to the conserved residues of the DoxD subunit or domain before calculation. For explanation of the subunits and accession numbers, see legend of Fig. 5.

In the genome sequences of two Bacteroides species (B. thetaiotaomicron and B. fragilis), doxDA genes were found with both subunit genes fused to a single ORF (75% identity to each other and 30% identity to A. ambivalens DoxDA). More interestingly, two other fused doxDA genes were found in close vicinity to each other in the genome sequence of the sulphur- and iron-oxidizing acidophilic bacterium Acidithiobacillus ferrooxidans (70% identity to each other; Figs 5 and 6). The gene duplication suggests an important role in the metabolism of the bacterium, although it remains to be demonstrated whether the proteins catalyse TS oxidation or not. A TS-oxidizing enzyme activity has been characterized previously from Acidithiobacillus ferrooxidans (Silver and Lundgren, 1968); however, neither the subunit composition nor the quinone content had been reported, nor is it clear from the procedure given whether the enzyme is membrane bound or soluble. For these reasons, it cannot be differentiated whether the enzyme is a TQO or a TS dehydrogenase and it is likewise impossible at present to link the observed activity to the findings deduced from the genome sequence.

We also performed searches in the Acidithiobacillus  ferrooxidans  and  in  the Sulfolobus  spp.  genomes  with the Paracoccus pantotrophus-deduced amino acid sequences of the soxABCDEF and soxXYZ genes encoding the Sox complex (Friedrich et al., 2000; 2001) but did not find homologues with the exception of a weak similarity with SoxF. The result suggests that Acidithiobacillus ferrooxidans might be the first well-studied mesophilic bacterium, which could use the TQO for TS oxidation and does not utilize the Sox or TOMES complexes as many other neutrophilic, sulphur- and TS-oxidizing bacteria do. It also suggests that the sox genes do not occur in the Sulfolobales as observed earlier (Friedrich et al., 2001).

The A. ambivalens DoxD amino acid sequence (but not DoxA) is the namesake of a large superfamily in the PFAM database comprising 147 hypothetical proteins (as of April 2004; PF04173). All of the ‘DoxD’ proteins in the database have a similar size and four predicted transmembrane helices. However, the multiple alignment shows that they share only a few conserved residues with the DoxD subunits or domains described here except for those transmembrane regions (Fig. 5). We therefore concluded that these ‘DoxX’ proteins might have a remote common phylogenetic origin with DoxD, but have evolved into at least two paralogous protein families with different functions not related to the TQO or terminal oxidases. A partner gene with similarity to doxA is usually missing with those doxX genes. This hypothesis is supported by the dendrogram (Fig. 6). Actually, the A. ambivalens DoxD is to date the only protein in the superfamily for which biochemical data are available.

The multiple alignment showed many conserved residues but only a single conserved cysteine, which is predicted to be located on the outside of the membrane (often missing in DoxX proteins; Fig. 5). This cysteine would be a putative TS binding residue; however, the relative insensitivity of enzyme activity against inhibition by the thiol-modifying reagent N-ethylmaleiimide speaks against this hypothesis. A predicted quinone-binding site (Fisher and Rich, 2000) was not identified. A comparison of the TQO amino acid sequences to those of TS dehydrogenases was not possible, as no nucleotide or amino acid sequence was found in the literature and in sequence database searches. Their sizes and subunit composition are in some cases similar, in others not, but they all have been purified from soluble fractions, cytoplasmatic and/or periplasmatic, suggesting that they belong to different enzyme families and that the TQO described here is of a novel type (the older literature has been summarized by Visser et al., 1997; see also Nakamura et al., 2001).

Coupling of sulphur compound oxidation to dioxygen reduction

The observation that pure TQO can reduce DQ and that CQ and SQ were found to be non-covalently bound to the protein strongly suggests that the latter electron-proton carrier may be the physiological partner of the enzyme. Furthermore, the previously reported co-purification of TQO and the terminal quinol:oxygen oxidoreductase also indicates that these two enzymes may operate closely in vivo, eventually forming a supercomplex. The standard redox potentials of the various compounds (TS/TT couple Eo′ = 25 mV, ferri-/ferrocyanide Eo′ = 356 mV, reduced/oxidized methylene blue Eo′ = 11 mV; Thauer et al., 1977), reduced/oxidized DQ Eo′ = 70 mV (Rich, 1984) and reduced/oxidized CQ Eo′ = 103 mV (Anemüller and Schäfer, 1990) showed that the reactions are thermodynamically possible.

To test these hypotheses we measured oxygen consumption in the intact membrane fraction upon the addition of TS. This substrate could drive cyanide-sensitive oxygen consumption at a maximal rate of 6.7 nmol O2 min−1 (mg total protein)−1(Fig. 7A). This result is interpreted considering that CQ is shuttling electrons between the TQO and the terminal quinol oxidase, which is the only terminal oxygen reductase present in the A. ambivalens respiratory chain (Fig. 7B). This reaction will thus enhance the transmembrane proton gradient and subsequent ATP formation.

Figure 7.

Thiosulphate driven oxygen consumption in intact membranes and schematic interpretation of events.
A. Oxygen consumption traces at 70°C. Arrows indicate addition of homogenized intact membrane extract (1; 2.32 mg total protein) in 47 mM BisTris buffer, pH 6; sodium thiosulphate (2; 10 mM); KCN (3; 18 µM).
B. Schematic cartoon illustrating the interaction between TQO and the terminal oxygen reductase, highlighting the role of the caldariella quinone (CQ) in the process.


We describe in this article the catalytic properties of the first TQO known from archaea and also from acidophilic and from (hyper-)thermophilic microorganisms. The description of the TQO expands the knowledge on the sulphur oxidation pathway in A. ambivalens. It is also the first TS-oxidizing enzyme that is an integral membrane glycoprotein, and also the first one that contains quinones and uses them as electron acceptors. The TQO represents a novel type of enzyme in the TS-oxidizing pathway different from the various TT-forming TS dehydrogenases and from the well-studied Sox system and the TOMES complex, which does not yield free TT as product.

The oxidation pathway of TT formed by the TQO has not yet been investigated. TT is not stable in the presence of strong reductants like H2S and is reduced to yield TS in vitro especially at high temperatures (Xu et al., 1998; 2000). Therefore, the H2S formed by the SOR might be able to re-reduce tetrathionate made by the TQO and thus feed electrons indirectly from S° disproportionation into the quinone pool.

Interestingly, homologous genes were found not only in different Sulfolobales but also in the acidophilic iron and sulphur oxidizer Acidthiobacillus thiooxidans and in two Bacteroides species. This points to a more widespread distribution of this pathway in mesophilic bacteria and raises the yet unanswerable question whether this is a general feature of the acidophilic sulphur oxidizers.

The TQO is also the first known entry point for electrons from sulphur compound oxidation into the electron transport chain in A. ambivalens. The TQO consists of two subunits, which had been thought to be constituents of the terminal oxidase. Therefore, the previous model of the terminal oxidase had to be revised, as there are only two major (DoxBC) and one minor subunit left for the entire enzyme (DoxE; Figs 1 and 7). The TQO and the terminal oxidase might still form a loose aggregation in the membrane, transferring electrons rapidly either directly via the bound CQ molecule or indirectly via free CQ. The most interesting questions remaining open are how the quinone is bound to the enzyme, how the redox reaction takes place without other redox active centres besides CQ and what the reaction mechanism is. The same questions have not been solved for the TS dehydrogenases from thiobacilli; therefore, the A. ambivalens enzyme is an interesting model for future studies not only for the archaeal enzymes but also for the haem-containing counterparts.

Experimental procedures

Organism, cell growth, membrane extract preparation and protein purification

Acidianus ambivalens (DSM 3772) was grown aerobically as described (Teixeira et al., 1995). About 285 g of cells (wet mass) were suspended in 285 ml of 5 mM potassium phosphate buffer, pH 6.5, and broken at 6000 psi using a French press. The lysate was centrifuged for 1 h at 10 000 g and 4°C to remove unbroken cells and S° particles. The resulting supernatant was ultracentrifuged for 12 h at 138 000 g and 4°C in a 45 Ti rotor (42 000 r.p.m.; Beckman). The green sediment (membrane fraction) was resuspended in 40 mM potassium phosphate buffer, pH 6.5. The membrane proteins were solubilized by gentle stirring with 1 g of DM per gram of protein for 12 h at 4°C. The resulting suspension was centrifuged at 138 000 g for 5 h at 4°C. The green supernatant, corresponding to the solubilized membrane fraction, was again centrifuged for 10 min at 31 500 g before loading on a 400 ml Q-Sepharose fast-flow column. This and all subsequent chromatographic steps were performed on a Pharmacia HiLoadTM system at 4°C (Amersham Pharmacia). After elution with 600 ml of low-ionic-strength buffer at 10 ml min−1 (40 mM potassium phosphate pH 6.5, 0.1% DM), a linear gradient of 10 column volumes from 0 to 100% of high-ionic-strength buffer (40 mM potassium phosphate pH 6.5, 1 M NaCl, 0.1% DM) was applied at the same flow rate. The TQO did not bind to the column and was recovered from the flow-though. The terminal quinol:oxygen oxidoreductase was eluted at ≈150 mM NaCl. The pooled TQO fractions were applied directly to an 85 ml ceramic hydroxyapatite HTP column, equilibrated with the same low-ionic-strength buffer. Proteins were eluted with a linear gradient of 7 column volumes from 0.04 to 1 M potassium phosphate (pH 6.5, 0.1% DM), at a flow rate of 0.75 ml min−1. The TQO was eluted at around 0.9 M potassium phosphate. The active fractions were concentrated by ultrafiltration to 4 ml total volume over a 100 kDa cutoff membrane, and loaded onto a 330 ml gel permeation Superdex S-200 column, equilibrated and eluted at a flow rate of 0.5 ml min−1 with 50 mM potassium phosphate buffer pH 6.5, containing 150 mM NaCl and 0.1% DM. The purified enzyme fraction was split into aliquots (1 ml) and stored at −70°C. Most of the experiments were performed with this preparation.

Additional chromatographic steps were performed to determine whether the TQO could be further purified or whether the number of bands on SDS gels could be reduced. Two millilitres of the Gel permeation eluate were equilibrated to low-ionic-strength buffer by repeated concentration and dilution using Nanosep ultrafiltration units (10 kDa cutoff; Pall Biosciences), loaded onto a 6 ml DEAE fast-flow column and in a parallel experiment onto a 2 ml Mono S cation exchange column (both Pharmacia). The columns were eluted with a gradient from 0 to 100% of high-ionic-strength buffer as described above. The TQO did not bind to either of the columns but was eluted with the flow-through.

Enzymic assays

TQO activity was measured spectrophotometrically by following the TS-dependent reduction of ferricyanide at 420 nm (ɛ = 1020 M−1 cm−1) and 80°C in a temperature-controlled cuvette holder. The standard reaction mixture (1 ml final volume) contained 50 mM BisTris buffer [Bis-2-hydroxyethyl-imino-tris(hydroxymethyl)-methane; Roth], 1.25 mM potassium ferricyanide, 1.25 mM citric acid at pH 6, adjusted with HCl, 10 mM TS, and protein solution (modified after Trudinger, 1961b). The cuvettes were placed into the cuvette holder for 2 min to allow for temperature equilibration. The reaction was started either by addition of 100 µl TS (from a freshly prepared 100 mM solution in water) or of protein solution respectively. Alternatively, a phosphate citrate buffer was used (McIlvaine buffer, after Dawson et al., 1986). All activity measurements were corrected for the non-enzymic TS oxidation rate. One unit of activity (U) was defined as the oxidation of 1 µmol of TS per minute. For kinetic studies, the amounts of TS or ferricyanide were varied. Several substances (Table 2) were tested for inhibitory effects on TQO activity. Nine micrograms of TQO were incubated in 1 ml of the standard reaction mixture with the respective substance in the designated concentrations for 1 min at 80°C before starting the reaction by addition of 10 mM TS.

Alternatively, TQO activity was measured with DQ as artificial electron acceptor at 291 nm (ɛox-red = 5033 M−1 cm−1). The cuvette and the reaction buffer (1 ml final volume; 50 mM BisTris, pH 6, 0.1% DM) were made anaerobic by flushing with N2/H2 (95:5) in an anaerobic chamber. One microlitre of a DQ solution (31 mM in ethanol) and 5 µl of 2 M TS in H2O were added after equilibration to 80°C. The reaction was started by addition of the protein solution containing 9 µg of enzyme. For control, the order of addition of substrates and protein was permuted in every possible way. For measuring activity with horse heart cytochrome c as the electron acceptor, 40 µl of a 1 mM solution (in 40 mM HEPPS buffer, pH 7) were added to the standard reaction mixture containing 10 mM TS. The reaction was started by addition of 9 µg of TQO. The absorbance change was recorded at 80°C and 550 nm.

For the measurement of the TT formation, a discontinuous assay was used. Seventy microgram of TQO were incubated at 60°C in a mixture containing 35 mM BisTris buffer (pH 6), 1 mM citric acid, 1 mM potassium ferricyanide and 2 mM sodium TS (total volume: 10 ml). TS and TT concentrations in the assay mixture were determined by cyanolysis (Kelly et al., 1969) in samples taken after different time intervals.

The TT reduction to TS was measured spectrophotometrically under anaerobic conditions at 80°C and 663 nm in a buffer containing 50 mM BisTris (pH 6) and 0.1 mM reduced methylene blue as electron donor (ɛ663 = 11 000 M−1 cm−1). Methylene blue was reduced with small amounts of titanium citrate (max concentration: 0.5 mM; Zehnder and Wuhrmann, 1976) before the addition of 5 mM TT with a gas tight Hamilton syringe. The reaction was started by addition of the enzyme (9 µg).

Activity measurements with a 5300 Biological Oxygen Monitor (Yellow-Springs Instruments) were performed at 70°C in a volume of 1.3 ml. The reaction mixture contained 47 mM BisTris buffer (pH 6) and 2.3 mg of total protein from the membrane fraction. Ten millimolar TS was added after 3 min and 18 µM KCN after 6.7 min of the total incubation time.

Analytical procedures

Proteins were separated by SDS–PAGE as described (Schägger and von Jagow, 1987) using protein molecular mass markers ♯SM0431 from Fermentas. Proteins were stained with colloidal Coomassie brilliant blue (Roti Blue; Roth). Alcian blue staining was used to demonstrate protein glycosylation in SDS gels (Wardi and Michos, 1972). The protein concentration was determined by a one-step biuret assay for protein in the presence of detergent (Watters, 1978). The apparent molecular mass of the native enzyme was determined by gel permeation chromatography on a superose 6 HR 10/30 column (Pharmacia) at 0.4 ml min−1 in the same gel permeation buffer as described above. Glutamate dehydrogenase (318 kDa), lactate dehydrogenase (140 kDa), bovine serum albumin (67 kDa) and ovalbumin (45 kDa; all Serva) were used as molecular mass standards.

The N-terminal sequence of the 28 kDa subunit was determined by automated Edman degradation (Esplora) after electrophoretic separation and transfer of the protein to a polyvinylidene difluoride membrane. The 16 kDa subunit was identified using MALDI-TOF mass fingerprinting followed by  MSFIT  and  MASCOT  analysis  of  the  results  (Esplora). A metal analysis of the purified TQO was performed by X-ray fluorescence spectroscopy (TXRF-analysis) with an EXTRAIIA instrument (Atomika Instruments) at the Institute of Inorganic and Analytical Chemistry in Frankfurt/Main (Germany).

UV/visible spectra were measured at room temperature in a Beckman DU 640 spectrophotometer. CQ was a kind gift of S. Anemüller and G. Schäfer (Lübeck, Germany). Fluorescence spectra of the native protein and of extracted CQ were recorded on a fluorescence spectrofluorimeter (Varian Cary Eclipse; Lemos et al., 2001). Mass spectra of extracted CQ fractions were recorded on a Bruker-Franzen Esquire LC mass spectrometer with electron spray ionization in the negative mode at the mass spectroscopy unit of the Clemens Schöpf Institute of Organic Chemistry and Biochemistry at the Darmstadt University of Technology. CD spectra were recorded using a J810 spectropolarimeter (Jasco Instruments) at the Eduard Zintl Institute of Inorganic and Physical Chemistry of the Darmstadt University of Technology. The spectra were obtained with a cuvette of 0.2 cm path length, a scanning speed of 50 nm min−1 and a bandwidth of 2 nm. Scans were collected at 1 nm intervals with a response time of 0.25 s and accumulated 10 times. The CD spectra were corrected by subtracting the spectrum of the buffer solution. The spectra between 190 and 250 nm were used to calculate the alpha helical and beta sheet content of the isolated TQO using the programs Contin&Sol;Ll, cdsstr and selcon3 included in the CDPro software package and the reference protein set containing 43 proteins (average values of the results obtained from three programs are given in the work by Sreerama and Woody, 2000).

Sequence analysis

The N-terminal amino acid sequence of the 28 kDa subunit and both DoxA and DoxD sequences were compared to the public databases using blastp and tblastn for the identification of homologues. The Acidithiobacillus ferrooxidans doxDA genes were identified in the yet unpublished genome sequence kindly provided by The Institute of Genome Research (TIgr; The Bacteroides fragilis doxDA gene was identified with the blast tool at the Sanger Institute ( The DoxD and DoxA amino acid sequences were merged when necessary before alignment using pileup, clustalw and hmmeralign included in the Wisconsin package (Accelrys). The phylogenetic analyses were performed with the resulting multiple alignment (the whole alignment and the DoxD and DoxA domains separately) using the programs distances, growtree, paupsearch, and paupdisplay from the GCG package. The secondary structure prediction programs psipred, memsat (for single sequences) and dsc (for single sequences and the multiple alignment) available via the expasy server ( were used to predict helical and β-sheet regions. The PFAM database ( was queried according to the user instructions provided at the website.


We want to acknowledge the help of V. Yatseev and W. Haase (Eduard Zintl Institute of Inorganic and Physical Chemistry, Darmstadt) for support in recording the CD spectra and of J.J. Veith (Clemens Schöpf Institute of Organic Chemistry, Darmstadt) for the mass spectra of the extracted quinones. This work has been supported by grants from the Deutsche Forschungsgemeinschaft to A.K. (Kl885/3-1, 3-2 and 3-3), by the Fundação Ciência e Tecnologia, Portugal (BIO99/36560) to M.T. and by travel grants from the Deutscher Akademischer Austauschdienst (DAAD; to A.K.) and the Instituto Cooperação CientÌfica Internacional (to C.M.G.). T.B. is a recipient of an FCT PhD fellowship (BD 3133/00).