Trichomonas vaginalis: metronidazole and other nitroimidazole drugs are reduced by the flavin enzyme thioredoxin reductase and disrupt the cellular redox system. Implications for nitroimidazole toxicity and resistance


  • David Leitsch,

    Corresponding author
    1. Department of Specific Prophylaxis and Tropical Medicine at the Center for Physiology, Pathophysiology and Immunology, Medical University of Vienna, A-1090 Vienna, Austria.
    Search for more papers by this author
  • Daniel Kolarich,

    1. Department of Chemistry, University of Natural Resources and Applied Life Sciences, A-1190 Vienna, Austria.
    Search for more papers by this author
    • Present address: Core of Biomolecular Frontiers, Macquarie University, Sydney 2109, Australia.

  • Marina Binder,

    1. Department of Specific Prophylaxis and Tropical Medicine at the Center for Physiology, Pathophysiology and Immunology, Medical University of Vienna, A-1090 Vienna, Austria.
    Search for more papers by this author
  • Johannes Stadlmann,

    1. Department of Chemistry, University of Natural Resources and Applied Life Sciences, A-1190 Vienna, Austria.
    Search for more papers by this author
  • Friedrich Altmann,

    1. Department of Chemistry, University of Natural Resources and Applied Life Sciences, A-1190 Vienna, Austria.
    Search for more papers by this author
  • Michael Duchêne

    1. Department of Specific Prophylaxis and Tropical Medicine at the Center for Physiology, Pathophysiology and Immunology, Medical University of Vienna, A-1090 Vienna, Austria.
    Search for more papers by this author

*E-mail; Tel. (+43) 14277 64871; Fax (+43) 14277 64899.


Infections with the microaerophilic parasite Trichomonas vaginalis are treated with the 5-nitroimidazole drug metronidazole, which is also in use against Entamoeba histolytica, Giardia intestinalis and microaerophilic/anaerobic bacteria. Here we report that in T. vaginalis the flavin enzyme thioredoxin reductase displays nitroreductase activity with nitroimidazoles, including metronidazole, and with the nitrofuran drug furazolidone. Reactive metabolites of metronidazole and other nitroimidazoles form covalent adducts with several proteins that are known or assumed to be associated with thioredoxin-mediated redox regulation, including thioredoxin reductase itself, ribonucleotide reductase, thioredoxin peroxidase and cytosolic malate dehydrogenase. Disulphide reducing activity of thioredoxin reductase was greatly diminished in extracts of metronidazole-treated cells and intracellular non-protein thiol levels were sharply decreased. We generated a highly metronidazole-resistant cell line that displayed only minimal thioredoxin reductase activity, not due to diminished expression of the enzyme but due to the lack of its FAD cofactor. Reduction of free flavins, readily observed in metronidazole-susceptible cells, was also absent in the resistant cells. On the other hand, iron-depleted T. vaginalis cells, expressing only minimal amounts of PFOR and hydrogenosomal malate dehydrogenase, remained fully susceptible to metronidazole. Thus, taken together, our data suggest a flavin-based mechanism of metronidazole activation and thereby challenge the current model of hydrogenosomal activation of nitroimidazole drugs.


The microaerophilic protozoan parasite Trichomonas vaginalis is the causative agent of trichomonal vaginitis and urethritis (trichomoniasis), the most frequent non-viral sexually transmitted disease worldwide (Wendel and Workowski, 2007) with more than 170 million cases per year (Petrin et al., 1998). Trichomoniasis is reliably treated with the 5-nitroimidazole drug metronidazole, which is also in use against other microaerophilic protozoan parasites such as Entamoeba histolytica and Giardia intestinalis (Upcroft and Upcroft, 2001) as well as a number of bacterial pathogens including Helicobacter pylori and Bacteroides fragilis (Freeman et al., 1997). The activity of nitroimidazole drugs depends on reduction at the nitro group that results either in the formation of single-electron transfer reduction products, nitroradical anions or further reduced reactive intermediates, that is, nitrosoimidazoles or hydroxylamineimiadazoles (Moreno and Docampo, 1985). In aerobic cells, the reduction of nitroimidazoles does also occur (Pervez-Reyes et al., 1980; Viodéet al., 1999), but oxygen interferes by quickly reoxidizing the nitroradical anion to the parent compound. This redox cycling effect, also termed ‘futile cycle’ (Mason and Holtzman, 1975), generates reactive oxygen species, but the resulting oxidative stress is believed to be too insignificant to cause serious damage in the aerobic cell (Moreno and Docampo, 1985).

In microaerophilic parasites, reduction of metronidazole is suggested to be mainly carried out by reduced ferredoxin (Yarlett et al., 1986a; Vanacova et al., 2003), which is generated during the oxidative decarboxylation of pyruvate by pyruvate : ferredoxin oxidoreductase (PFOR). This hypothesis is based on studies on metronidazole activation in the trichomonad cattle parasite Tritrichomonas foetus (Lindmark and Müller, 1976; Moreno et al., 1983; Moreno et al., 1984; Kabickova et al., 1986), but also received support from studies in T. vaginalis (Lloyd and Pedersen, 1985; Kulda et al., 1993; Rasoloson et al., 2002; Land et al., 2002; Vidakovic et al., 2003) and G. intestinalis (Townson et al., 1996). In trichomonads, ferredoxin and PFOR are located in the hydrogenosome (Lindmark and Müller, 1973), a mitochondrion-derived organelle (Carlton et al., 2007). Hence, it has been proposed that hydrogenosomes are the site of metronidazole activation in T. vaginalis. Indeed, it was shown in several studies that isolated hydrogenosomes or hydrogenosomal extracts can reduce metronidazole (Chapman et al., 1985; Yarlett et al., 1985). Conflicting with the model of PFOR-mediated activation of metronidazole, however, is the observation that a ferredoxin knock-out mutant of T. vaginalis displayed only minimal amounts of PFOR but remained fully susceptible to metronidazole (Land et al., 2004). Rather recently, also an alternative hydrogenosomal pathway of metronidazole activation was suggested, that is, through the reduction of ferredoxin via NAD : ferredoxin oxidoreductase (Hrdy et al., 2005), which utilizes NADH generated by the hydrogenosomal decarboxylating malate dehydrogenase for this reaction. In other organisms metronidazole activation pathways without involvement of ferredoxin have been described, for example, reduction by the flavin enzyme thioredoxin reductase in E. histolytica (Leitsch et al., 2007), an enzyme of which a homologue also occurs in T. vaginalis (Coombs et al., 2004). Furthermore, in H. pylori, metronidazole reduction has been mainly attributed to the nitroreductases RdxA and FrxA (Goodwin et al., 1998; Kwon et al., 2000), and also nonenzymatic reduction of metronidazole by ferrous iron and thiols (Willson and Searle, 1975) as well as flavins (Clarke et al., 1980) was reported.

With regard to the mode of action of metronidazole, considerable uncertainty remains, especially as to whether the nitroradical anion or a further reduced metabolite (i.e. the respective 5-nitrosoimidazole or a 5-hydroxylamineimidazole) is the major toxic agent. In a number of in vitro studies, electrochemically generated metronidazole nitroradical anions were observed to cause single- and/or double-strand breaks in DNA (Zahoor et al., 1986; Edwards, 1993). Chemical reduction of metronidazole by dithionite (LaRusso et al., 1978), however, does not lead to double strand breaks but rather to covalent binding of reactive metronidazole intermediates reduced beyond the nitroradical stage to DNA (Lindmark and Müller, 1976) and proteins (West et al., 1982). Indeed, binding of metronidazole metabolites to proteins and DNA is assumed to be the mode of action of metronidazole in T. vaginalis (Ings et al., 1974), but specific targets have never been searched for in T. vaginalis or other microaerophilic parasites because the reactivity of metronidazole metabolites was anticipated to be indiscriminate. Our recent research in E. histolytica, however, revealed only five proteins that form covalent adducts with metronidazole and other nitroimidazoles (Leitsch et al., 2007), including thioredoxin and thioredoxin reductase. As the latter displayed nitroreductase activity with metronidazole and other nitroimidazoles, and as two other proteins bound are also known to, or at least assumed to, be involved in thioredoxin-mediated regulation, we suggested that the identified proteins are targeted by reactive nitroimidazole metabolites because they are located in spatial proximity to a source of these metabolites, that is, thioredoxin reductase. Furthermore, as covalent adduct formation of metronidazole with thioredoxin reductase and thioredoxin has a negative effect on protein function and as nitroimidazole treatment greatly diminishes intracellular non-protein thiol pools, we hypothesized that nitroimidazoles exert toxicity by disrupting the cellular redox balance (Leitsch et al., 2007).

In the present study we extended our previous work to T. vaginalis in order to find supportive evidence for our hypothesis that, in the cell, nitroimidazole drugs do not form covalent adducts with proteins in an indiscriminate way but rather specifically with target proteins that are involved in the thioredoxin-mediated redox system. In analogy to our previous findings with E. histolytica, we further hypothesized that T. vaginalis thioredoxin reductase is both a key enzyme in nitroimidazole activation and a target of nitroimidazole drugs, and that nitroimidazole drugs disrupt the cellular redox balance by diminishing thioredoxin reductase activity and intracellular thiol levels. As, in contrast to E. histolytica, T. vaginalis can be readily adapted to high metronidazole concentrations in the laboratory, we received the opportunity to study metronidazole resistance and the role of thioredoxin reductase therein at the proteomic scale.


Proteomic analysis of T. vaginalis protein extracts by two-dimensional gel electrophoresis reveals 10 proteins to form adducts with metronidazole

As a first step we decided to check the T. vaginalis proteome for proteins that form covalent adducts with nitroimidazoles by comparing two-dimensional gel electrophoresis (2DE) profiles of metronidazole-treated and untreated cells. Before the experiment, cultures of T. vaginalis C1 (ATCC 30001) cells had been grown overnight in completely filled and sealed flasks. This results in very low oxygen tensions in the medium due to oxygen scavenging by the cells and due to reductants in the growth medium. Exponentially growing cells were treated for various periods of time (1, 2, 3 h) with different concentrations of metronidazole in the culture medium (10, 20, 50, 100 μM) followed by cell harvest and 2DE analysis of whole-cell soluble protein extracts. As described previously (Leitsch et al., 2007), covalent adducts of proteins with metronidazole could be readily identified on the 2D-gels as novel protein spots shifted to a more basic pI. A total number of only 10 proteins that bind to activated metronidazole were reproducibly detected on silver-stained and Coomassie-stained 2D-gels of whole-cell extracts, seven of which were identified (Fig. 1A and B, Table 1). Covalent adduct formation was also observed under strictly anaerobic and under microaerobic conditions (data not shown). Exposure to concentrations of more than 50 μM metronidazole for 2 h did not result in a higher number of affected proteins. The widths of the shifts depended on the protein affected, whereas the proportion of the shifted protein to the unshifted isoform depended on the concentration of metronidazole applied. Additional experiments showed that only a limited amount of each protein could be modified and shifted, as metronidazole concentrations higher than 50 μM did not lead to higher proportions of the shifted protein isoforms (data not shown). Spots were picked from Coomassie-stained gels and tryptically digested. Seven proteins were identified by liquid chromatography electrospray ionization quadrupole time-of-flight tandem mass spectrometry (LC-ESI-QTOF-MS/MS) (Table 1): thioredoxin reductase, thioredoxin peroxidase, thiol peroxidase, ribonucleotide reductase heavy subunit (annotated as ‘ribonucleotide reductase all α-domain containing protein’), cytosolic malate dehydrogenase, enolase and glucose 6-phosphate isomerase. As observed before in E. histolytica (Leitsch et al., 2007), some proteins were shifted only once whereas others were shifted twice or even more often (Fig. 1A). In order to confirm that the protein shifts were due to binding of metronidazole and not due to modifications brought about by a physiological response of the cell, T. vaginalis cells were treated under the same conditions with tinidazole, another 5-nitroimidazole that is also in use in the treatment of trichomoniasis (Narcisi and Secor, 1996). Tinidazole had already been found before to bind the same proteins as metronidazole in E. histolytica; however, the observed shifts in pI were narrower, presumably due to tinidazole's different pKa (Leitsch et al., 2007). Indeed, the result with E. histolytica was paralleled in T. vaginalis, as all protein shifts observed with metronidazole were also observed with tinidazole, but their pI intervals were smaller (Fig. 2).

Figure 1.

Seven proteins were identified that form covalent adducts with metronidazole.
A. The seven identified proteins that form covalent adducts with metronidazole: (1) thioredoxin reductase, (2) thioredoxin peroxidase, (3) thiol peroxidase, (4) ribonucleotide reductase heavy chain annotated as ‘all α domain containing protein’, (5) cytosolic malate dehydrogenase, (6) enolase, (7) glucose 6-phosphate isomerase. For reasons of perceptibility, the enolase spot images were taken from Coomassie stained gels because the spot density is very high in the vicinity of enolase. All other images were taken from silver-stained gels. Encircled spots are the native proteins whereas arrows mark metronidazole-modified isoforms. Note that (4) forms a chain of five metronidazole-modified isoforms that are not separately indicated by arrows.
B. For better orientation, the seven proteins are also indicated on a silver-stained 10% polyacrylamide 2D-gel.

Table 1.  Seven proteins that form adducts with nitroimidazoles were identified by LC-ESI-QTOF-MS/MS.
ProteinNr.Theoretical size (Da)Theoretical pIAccession numberNumber of shiftsMS/MS sequence coverage (%)
  1. The numbering refers to Fig. 1.

Thioredoxin reductase1323486.02XP_001316923244
Thioredoxin peroxidase2219536.31XP_001313356171
Thiol peroxidase3179826.05XP_001328156291
Ribonucleotide reductase4685976.40XP_001580988512
Malate dehydrogenase (cytosolic)5356986.02XP_001579239163
Glucose 6-phosphate isomerase7611175.94XP_001323512261
Figure 2.

The 5-nitroimidazole metronidazole and tinidazole bind the same proteins. The adduct formation of proteins with metronidazole and tinidazole is exemplified by thioredoxin reductase, thioredoxin peroxidase, thiol peroxidase, ribonucleotide reductase and glucose 6-phosphate isomerase. Cells were treated with 50 μM metronidazole or tinidazole for 2 h. Circles indicate the respective native proteins and arrows designate covalent adducts. Tinidazole leads to considerably narrower shifts towards basic pI values as compared with metronidazole.

As it was remarkable that none of the identified proteins were hydrogenosomal, we checked the hydrogenosomal compartment for overlooked metronidazole adducts by 2DE after isolating hydrogenosomes by differential centrifugation and subsequent Percoll gradient (Pütz et al., 2005). The identity of the hydrogenosomal fraction was confirmed by identifying a selection of four protein spots (Table S1), which contained PFOR, hydrogenosomal malate dehydrogenase, succinyl-coenzyme A (CoA) synthetase-2 and -3 α subunit, and acetyl-CoA hydrolase. These spots were highly prominent on 2D-gels of hydrogenosomes but completely absent on 2D-gels from hydrogenosome-free extracts (Fig. S1). Hydrogenosomal 2DE profiles of treated cells were almost indistinguishable from those of untreated cells and displayed no signs of protein degradation (Fig. S2A and B). Also high-density polyacrylamide gels (15%) were run in the second dimension in order to visualize proteins in the range of ferredoxin (c. 10–12 kDa). However, only one weakly expressed hydrogenosomal protein of about 20–22 kDa, which was reproducibly shifted after metronidazole treatment, could be detected on silver-stained 2D-gels (Fig. S2A and B). Due to its low abundance, however, the protein was not isolated and identified.

Purified recombinant T. vaginalis thioredoxin reductase displays nitroreductase activity

In our previous work on metronidazole action in E. histolytica, we showed that thioredoxin reductase is not just a target of nitroimidazoles, but that it also has nitroreductase activity and can reduce nitroimidazoles such as metronidazole and azomycin (Leitsch et al., 2007). Based on these observations, we had hypothesized that the small subset of modified proteins are vulnerable to covalent adduct formation with nitroimidazole metabolites because they are involved in the thioredoxin-mediated redox network and therefore located in spatial proximity to thioredoxin reductase that is a source of reduced nitroimidazole intermediates. As in T. vaginalis thioredoxin reductase was also among the modified proteins, we suspected it to be a nitroreductase as well and recombinantly expressed it in Escherichia coli BL21(DE3) cells.

Indeed, by measuring NADPH consumption in the assay buffer (OD340), nitroreductase activity of recombinant thioredoxin reductase (recTvTrxR) was confirmed with the nitrofuran drug furazolidone as substrate (5870 ± 270 nmol min−1 mg−1). Furazolidone was assayed because it is a potent antitrichomonal drug and similarly toxic to T. vaginalis as metronidazole (Narcisi and Secor, 1996). Unfortunately, the assay was heavily disturbed when conducted with nitroimidazoles due to the strong absorption of nitroimidazoles at this wavelength (Leitsch et al., 2007), which did not allow nitroimidazole concentrations higher than 100 μM to be used. Thus, an alternative approach (Leitsch et al., 2007), by which nitroradical anion formation is measured via reduction of cytochrome c (OD550) in the reaction buffer was chosen. Cytochrome c is either directly reduced by nitroradical anions or indirectly by superoxide generated by redox cycling of nitroradical anions. This alternative method allowed the determination of the nitroreductase activity of recTvTrxR with the 5-nitroimidazole metronidazole, ronidazole, tinidazole and ornidazole, the 2-nitroimidazole azomycin and 4-nitroimidazole (Table 2) with substrate concentrations as low as 100 μM. The reduction rate of furazolidone (6056 ± 784 nmol min−1 mg−1) was almost equal to that measured before by determining NADPH consumption (OD340), which argues for the validity of the approach. With the exception of ronidazole, the 2-nitroimidazole azomycin was clearly more efficiently reduced than the 5-nitroimidazoles. Under the assay conditions applied, 4-nitroimidazole, having by far the lowest midpoint reduction potential of all drugs tested, was not reduced. In general, nitroreduction rates strongly depended on midpoint redox potentials (Table 2), nitroheterocycles with higher potentials being more efficiently reduced. This is especially true with furazolidone as nitrofurans have much higher midpoint redox potentials (around −250 mV) than nitroimidazoles. When the assay was performed in anaerobic reaction buffer with azomycin and metronidazole, reduction of cytochrome c also occurred, although at a lower rate (Table 2). Furazolidone was more slowly reduced as well (1022 ±  90 nmol mg −1 min−1). We assume that this effect is due to a lower concentration of available oxidized cytochrome c, because after incubation of the assay buffer in an anaerobic jar we observed that considerable reduction had also occurred in the absence of any additives. It is also possible that the respective nitroradicals reduce cytochrome c less efficiently than superoxide that is formed by redox cycling.

Table 2.  Recombinant Trichomonas vaginalis thioredoxin reductase (recTvTrxR) exerts nitroreductase activity with nitroimidazoles.
NitroimidazoleMidpoint redox potential (mV)ConcentrationNitroreductase activity of TvTrxR (nmol min−1mg−1)Turnover (min−1)
  • a. 

    Taken from Smith and Edwards (1995).

  • b. 

    Taken from Schmid and Schmid (1999).

  • Reduction of nitroimidazoles was determined by measuring cytochrome c reduction by nitroradical anions or by superoxide generated by redox cycling of nitroradical anions respectively. In all reactions 5 μg recTvTrxR (= 150 nM) were applied. All measurements were performed at least three times. Midpoint redox potentials of the nitroimidazoles are given. For measurements under anaerobic conditions, the assay buffer had been incubated in an anaerobic jar overnight prior to the assay.

Metronidazole−486a100 μM251
1 mM542.2
Ronidazole−466b100 μM552.2
1 mM2068.2
Ornidazole−467a100 μM110.4
1 mM602.4
Tinidazole−464a100 μM210.8
1 mM823.3
Azomycin−418a100 μM441.8
1 mM29912
4-Nitroimidazole−527a100 μM00
1 mM00
Metronidazole (anaerobic buffer)−4861 mM200.8
Azomycin (anaerobic buffer)−4181 mM1787

Nitroimidazole treatment reduces thioredoxin reductase activity in T. vaginalis cell extracts

It had been previously shown that metronidazole-modified thioredoxin reductase from E. histolytica displayed diminished activity (Leitsch et al., 2007). In order to confirm that nitroimidazole treatment has a negative effect on the disulphide reductase activity of T. vaginalis thioredoxin reductase, we measured the thioredoxin-mediated reduction of dithionitrobenzene (DTNB) to thionitrobenzene (TNB) by hydrogenosome-free extracts of untreated or nitroimidazole-treated cells and recombinant T. vaginalis thioredoxin (recTvTrx). The gene for a previously studied thioredoxin (CAD47836, XP_001311336), which had been confirmed to be reduced by thioredoxin reductase (Coombs et al., 2004), was chosen for PCR amplification and recombinant expression. Prior to measurements, hydrogenosomes had been completely removed by centrifugation at 20 000 g for 10 min. The purity of the hydrogenosome-free extract was indicated by the total absence of PFOR activity. Thioredoxin reductase activity in untreated extracts amounted to 126 ± 24 nmol min−1 (mg protein)−1. Treatment for 2 h with 50 μM metronidazole, ornidazole or tinidazole, respectively, led to an approximate 50% decrease of thioredoxin-stimulated reduction of DTNB as compared with the untreated sample (Fig. 3A). When the same concentration of ronidazole and azomycin was applied, thioredoxin reductase activity in cell extracts was lowered even further (to 7% and 19% respectively). Importantly, these two nitroimidazoles were those most strongly reduced by thioredoxin reductase. Exposure of the cells to the nitroimidazoles under strictly anaerobic conditions (18% CO2, 0% O2) but also under microaerobic conditions (8% CO2, 5% O2) gave similar results. Metronidazole treatment decreased the thioredoxin reductase activity to 40% [52 nmol min−1 (mg protein)−1] under anerobic conditions and to 50% under microaerobic conditions [62 nmol min−1 (mg protein)−1] respectively. This indicates that the observed decreases in thioredoxin reductase activity are not due to oxidative damage inflicted by reactive oxygen species that are generated by redox cycling, but possibly due to adduct formation.

Figure 3.

Nitroimidazoles impair thioredoxin reductase activity and deplete intracellular thiol pools.
A. Nitroimidazoles diminish thioredoxin reductase activity. Thioredoxin reductase activity was measured using extracts of untreated and nitroimidazole-treated cells. Cells were treated with 50 μM of the respective nitroimidazole for 2 h.
B. In congruence with the impairment of thioredoxin reductase activity, peroxidase activity, which depends on thioredoxin, is also diminished.
C. Nitroimidazoles deplete non-protein thiol levels in T. vaginalis. Cells were treated with 50 μM of the respective nitroimidazole for the time indicated (100% = untreated).All values were determined in at least three independent experiments.

Nitroimidazole treatment slows down the detoxification of hydrogen peroxide

An impairment of the disulphide reducing activity of thioredoxin reductase must have corresponding consequences for enzymes that depend on reduction by thioredoxin for their activity. Consequently, we reasoned that H2O2 detoxification, which is carried out by thioredoxin-dependent peroxidases (Coombs et al., 2004), would be negatively affected as well. The breakdown of H2O2 in T. vaginalis ultimately depends on NADPH used by thioredoxin reductase for the reduction of thioredoxin, which, in turn, reduces the catalytically active cysteines of peroxidases, for example, thioredoxin peroxidase and thiol peroxidase. Thus, H2O2 removal was determined in the presence of hydrogenosome-free extract and recTvTrx by measuring NADPH consumption. Peroxidase activity in extracts of untreated cells amounted to 32 ± 3 nmol min−1 (mg protein)−1. The previously observed 50% decrease in the activity of disulphide reductase upon metronidazole treatment was mirrored by a congruent drop of H2O2-stimulated NADPH consumption to 55% of the normal rate when hydrogenosome-free extracts of metronidazole-treated cells were used (Fig. 3B). The fact that H2O2-stimulated NADPH consumption did not drop to a lower level indicates that metronidazole may not directly affect the activities of thioredoxin peroxidase and thiol peroxidase, although both enzymes were found to form adducts with metronidazole (Fig. 1A, Table 1). The decreases in peroxidase activity in ronidazole- and azomycin-treated cells (to 3% and 15% respectively) (Fig. 3B) were also similar to the respective effects of these nitroimidazoles on thioredoxin reductase activity (Fig. 3A).

The activities of two other enzymes found to form adducts with nitroimidazole metabolites, enolase [335 ± 33 nmol min−1 (mg protein)−1] and cytosolic malate dehydrogenase [4400 ± 1240 nmol min−1 (mg protein)−1], however, were either unaffected or only insignificantly decreased after nitroimidazole treatment (data not shown).

Nitroimidazoles sharply decrease non-protein thiol levels in T. vaginalis

It is known that nitroimidazole metabolites are thiol-active compounds that form adducts with protein cysteines and free thiols (West et al., 1982). Indeed, nitroimidazoles deplete intracellular thiol levels in E. histolytica (Leitsch et al., 2007), so we reasoned that a similar effect of nitroimidazole treatment would be observable in T.  vaginalis. Upon treatment with 50 μM of the respective nitroimidazole drug for different periods of time (1, 2, 3 h), T. vaginalis C1 cells were disrupted and the non-protein thiol content was determined by measuring the reduction of DTNB to TNB in diluted cell extracts (OD412) (Fig. 3C). The thiol content of untreated cells amounted to 7.15 ± 1.23 nmol (mg protein)−1. All 5-nitroimidazoles tested depleted intracellular thiol levels more effectively than the 2-nitroimidazole azomycin. After 3 h, thiol levels in 5-nitroimidazole-treated cells had decreased to less than 20%. Ronidazole and metronidazole, however, led to a faster decrease in thiol levels, with less than 20% already reached after 1 h. In azomycin-treated cells, thiol levels never dropped below 30% (2 h) and even recovered to some extent (40%) after 3 h. Although this increase was not very pronounced, it was reproducible and statistically significant (Table S2). Remarkably, decreases in thiol levels below 20% coincided well with the detachment of cells from the culture flasks and loss of motility. We observed that ronidazole clearly had the strongest impact on cell motility with practically all cells rendered immotile after only 60 min of incubation with the drug. Most of the metronidazole-treated cells were also immotile after 1 h, whereas cells treated with azomycin, tinidazole and ornidazole were either still attached to the bottom of the culture flasks or at least fairly motile. Finally, however, tinidazole- and ornidazole-treated cells also lost motility (after 2–3 h), whereas azomycin-treated cells remained motile until the end of the experiment.

Generation and characterization of metronidazole-resistant T. vaginalis C1 cells

Metronidazole resistance in T. vaginalis can be induced in vitro and has been attributed to the inability to reduce and thereby activate nitroimidazole prodrugs to toxic intermediates (Kulda, 1999). As T. vaginalis thioredoxin reductase is not only a target but also an activator of nitroimidazole drugs, we reasoned that its expression should be downregulated in metronidazole-resistant cells. Thus, in order to assess whether thioredoxin reductase has an important role in nitroimidazole activation we generated a metronidazole-resistant C1 line and followed the development of resistance at the proteomic level.

Resistance to metronidazole in T. vaginalis C1 was induced as described previously (Kulda et al., 1993) by growing cells in the presence of ever-increasing concentrations of the drug over a period of several months. The cultures were grown in completely filled and sealed culture flasks but not under strictly anaerobic conditions. In accordance with previous reports (Kulda et al., 1993; Rasoloson et al., 2002), stages of mild- and intermediate-level metronidazole resistance occurred before high-level metronidazole resistance was attained. Only subtle changes in the 2DE protein expression pattern of mildly resistant cells (20–50 μM metronidazole) were observed when the cells were grown under low oxygen tension, including the appearance of two lactate dehydrogenases (XP_001307557, XP_001320862) comigrating in one spot and another new spot in the range (in terms of molecular weight and pI) of hydrogenosomal malate dehydrogenase that was not identified (Fig. 4A, Fig. S3, Table S3). Remarkably, when mildly resistant cells were grown under more aerobic conditions (half filled vented culture flasks in an aerated incubator), upregulation of several other proteins was detected (Fig. 4B, Fig. S3). Three of the spots corresponded to superoxide dismutases (XP_001276944, XP_001324547, XP_001317169, XP_001317169) whereas one spot contained two ‘tryparedoxin peroxidases’ (XP_001298652, XP_001326926) and two thioredoxin peroxidases (XP_001313356, XP_001316602) (Table S3). The designation ‘tryparedoxin peroxidase’ is very probably a misnomer generated by automated gene annotation because T. vaginalis possesses neither tryparedoxin nor trypanothione reductase. It can therefore be assumed that both ‘tryparedoxin peroxidases’ in T. vaginalis constitute another two thioredoxin peroxidases. The remaining spots corresponded to a protein of unknown function with a flavodoxin-like domain (XP_001319260). The expression levels of the two previously described lactate dehydrogenases were clearly elevated (Fig. 4A). In intermediately (100–200 μM) and highly metronidazole-resistant cells (> 1 mM), the above-described proteins remained expressed under normal growth conditions as well (completely filled and sealed culture flasks) (Fig. 4A, Fig. S3). Very much to our surprise, the expression of thioredoxin reductase remained unchanged even in highly resistant cells (Fig. 4A). On the other hand, in perfect accordance with previous reports (Kulda et al., 1993; Brown et al., 1999; Rasoloson et al., 2002), the expression of PFOR and hydrogenosomal malate dehydrogenase had decreased to very low levels in intermediately resistant cells and even to minimal levels in highly resistant cells (Fig. 4A, Fig. S3). Accordingly, enzyme activities of PFOR and hydrogenosomal malate dehydrogenase were undetectable or very low (Table 3). Also in good agreement with previous reports (Cerkasovováet al., 1986; Kulda et al., 1993; Rasoloson et al., 2002), the loss of PFOR and hydrogenosomal malate dehydrogenase was obviously compensated by a roughly threefold lactate dehydrogenase activity (Table 3). Interestingly, however, our resistant cell line also displayed sharply decreased levels of succinyl-CoA synthetase (XP_001328129, XP_001330450), a hydrogenosomal protein whose expression level had been found to remain unchganged in other highly metronidazole-resistant T. vaginalis strains (Brown et al., 1999; Rasoloson et al., 2002). In addition, although the total loss of PFOR and hydrogenosomal malate dehydrogenase expression has been suggested to result in ‘full anaerobic resistance’ (Kulda, 1999), our highly resistant line was not able to divide in the presence of 250 μM metronidazole when cultured strictly anaerobically and finally perished after a duration of 5–7 days.

Figure 4.

Differential expression of proteins in metronidazole-susceptible and metronidazole-resistant T. vaginalis C1 cells.
A. With increasing metronidazole resistance, the expression of the hydrogenosomal enzymes, PFOR, succinyl CoA synthetase α-subunit (SCS) and hydrogenosomal malate dehydrogenase (hyd MDH) is being downregulated and, eventually, almost abolished. In contrast, the lactate dehydrogenases XP_001307557, XP_001320862 (LDH) are only expressed in metronidazole-resistant cells. Expression of thioredoxin reductase (TrxR) remains unchanged.
B. In metronidazole-resistant cells, several antioxidant proteins (1–4) are upregulated. (1), ‘tryparedoxin peroxidases’ (XP_001298652, XP_001326926) and two thioredoxin peroxidases (XP_001313356, XP_001316602), all comigrating, (2), iron-containing superoxide dismutases (XP_001317169, XP_001317169), comigrating, (3), iron superoxide dismutase (XP_001324547), (4), iron-containing superoxide dismutase (XP_001276944). (5), a protein of unknown function with a flavodoxin-like domain was also upregulated (XP_001319260).

Table 3.  Key enzymatic activities in metronidazole-susceptible and metronidazole-resistant T. vaginalis C1 cells.
Key enzyme activities in cell extractsPFORhyd MDH nmol min−1 (mg protein)−1LDH
  1. Enzyme activites of PFOR, hydrogenosomal malate dehydrogenase (hyd MDH) and lactate dehydrogenases (LDH) were determined in at least three independent experiments each.

Susceptible1438 ± 5201075 ± 1771630 ± 348
Resistant062 ± 34480 ± 516

The doubling time of our highly resistant line was considerably longer (550 ± 60 min) as compared with susceptible cells (260 ± 40 min), and became even more extended under aerobic conditions (860 ± 105 min). In contrast, the growth rate of metronidazole-susceptible cells remained almost equal under these aerobic conditions (310 ± 30 min). Thus, even though metronidazole-resistant cells express higher levels of antioxidant enzyme proteins than metronidazole-susceptible cells (Fig. 4B), they are more vulnerable to oxygen. These results confirm observations made before with other resistant strains (Ellis et al., 1994; Rasoloson et al., 2001). However, in addition, we also observed that metronidazole-resistant cells required supplementation of the growth medium with cysteine for survival whereas metronidazole-susceptible cells did not.

Thioredoxin reductase activity in metronidazole-resistant cells is minimal but can be restored by addition of FAD to cell extracts

It was surprising to us that thioredoxin reductase, although being an activator of nitroimidazole drugs, was at least equally abundant on 2D-gels of metronidazole-resistant cells as on those of metronidazole-susceptible cells. However, our observation that highly metronidazole-resistant cells strictly required cysteine in the growth medium suggested that they are sensitive to disulphide stress although thioredoxin reductase is being expressed. When we measured the thioredoxin-reducing activity in cell extracts of our highly resistant line (Fig. 5), we noticed that it was almost absent (less than 5% of the normal level), although cells had been grown for several passages without metronidazole prior to the assay. Remarkably, thioredoxin-reducing activity could be restored to more than 60% when 3 μM FAD was added to the reactions (Fig. 5). Higher concentrations of FAD only very slightly increased the thioredoxin reduction rate (data not shown) and the addition of riboflavin and FMN had no effect at all. When performing the assay with extracts of metronidazole-susceptible cells, we found that the addition of 5 μM FAD did not increase thioredoxin-reducing activity (data not shown). Consequently, we attribute the partial restoration of thioredoxin-reducing activity in metronidazole-resistant cells to the replenishment of the enzyme with its FAD-cofactor.

Figure 5.

Thioredoxin reductase activity in extracts of metronidazole-resistant C1 cells. Thioredoxin reductase activity is almost absent but can be restored to a great extent upon addition of FAD.

Extracts of metronidazole-resistant cells cannot reduce free flavins

It has been shown in vitro that not only flavin enzymes but also free flavins can reduce nitroimidazoles (Clarke et al., 1980). Thus, we assumed that levels of reduced flavins would be considerably lowered in metronidazole-resistant T. vaginalis. We addressed this issue by determining the reduction rates of FAD, FMN and riboflavin with extracts of resistant and susceptible cells under aerobic conditions (Table 4). In the presence of oxygen, reduced flavins are rapidly re-oxidized, so that even low flavin substrate concentrations (10 μM) could be used for the assay. Indeed, all three flavins were readily reduced by extracts of metronidazole-susceptible cells when NADPH was present. In contrast, no reduction of flavins was detected when using extracts of metronidazole-resistant cells. As even FAD, although re-establishing thioredoxin-reducing activity in metronidazole-resistant cells (Fig. 5), did not result in NADPH consumption rates above background, it is obvious that flavin reduction is mainly catalysed by (an) enzyme(s) different from thioredoxin reductase. In fact, we could measure some flavin reductase activity with purified recombinant thioredoxin reductase when using FAD as substrate (96 ± 40 nmol min−1 mg−1), but it is doubtful whether this low activity is of any physiological relevance.

Table 4.  Flavin reductase activity is lost in metronidazole-resistant cells.
Flavin reductase activity in cell extractsRiboflavin (10 μM)FAD (10 μM)
nmol min−1 (mg protein)−1
FMN (10 μM)
  1. Reduction of flavins (10 μM each) was measured in extracts of metronidazole-susceptible and -resistant cells. All values were determined in at least three independent experiments.

Susceptible62 ± 356 ± 18112 ± 3

Bipyridyl-treated T. vaginalis C1 cells almost entirely lack PFOR and hydrogenosomal malate dehydrogenase but remain fully susceptible to metronidazole

The (near) absence of thioredoxin reductase and flavin reductase activities in metronidazole-resistant cells indicates that reduced flavins, either protein-bound or free, could play an important role in nitroimidazole reduction in T. vaginalis. However, as shown in other studies, we also found PFOR and hydrogenosomal malate dehydrogenase to be nearly absent in our metronidazole-resistant line (Cerkasovova et al., 1986; Kulda et al., 1993; Rasoloson et al., 2002), that is, those enzymes that have so far been suggested to be directly (PFOR) or indirectly (PFOR again, hydrogenosomal malate dehydrogenase) responsible for the activation of nitroimidazole drugs in trichomonads. This unresolved issue prompted us to directly address the role of PFOR and hydrogenosomal malate dehydrogenase in nitroimidazole toxicity. It had been previously shown in T. foetus that several passages with the iron chelator 2,2′-bipyridyl (bipyridyl) resulted in drastically decreased expression levels of PFOR, hydrogenosomal malate dehydrogenase and ferredoxin (Vanacova et al., 2001). Very interstingly, thus treated T. foetus are even more susceptible to metronidazole than normal cells (Sutak et al., 2004). Thus, it was our goal to test whether bipyridyl-treated T. vaginalis would be still susceptible to metronidazole as well. Preliminary experiments with varying concentrations of bipyridyl in the growth medium showed that 50 μM were well tolerated and that cells were still able to grow well and thrive. Nevertheless, after four subcultures in presence of 50 μM of bipyridyl PFOR was almost absent and the levels of hydrogenosomal malate dehydrogenase were strongly reduced (Fig. 6). After nine subculturing steps, hydrogenosomal malate dehydrogenase was also barely detectable anymore (Fig. 6). Even after nine passages with 50 μM of bipyridyl, however, the cells remained fully susceptible to metronidazole and were rendered immotile within 1–2 h when treated with 50 μM metronidazole. This was also reflected by the occurrence of shifts of the identified proteins on 2D-gels (Fig. S4). When we determined the enzymatic activities of PFOR, hydrogenosomal malate dehydrogenase and lactate dehydrogenase from cells routinely grown with 50 μM bipyridyl (Table 5), we found a metabolic profile very similar to that of metronidazole-resistant cells: PFOR activity was absent, malate dehydrogenase activitiy was sharply decreased and lactate dehydrogenase activity was several-fold upregulated. In contrast to metronidazole-resistant cells, however, thioredoxin reductase was fully functional [153 ± 7 nmol−1 min−1 (mg protein)−1]. Thus, when summarizing our results with iron-depleted T. vaginalis, it seems that PFOR and hydrogenosomal malate dehydrogenase are not necessary for nitroimidazole activation and nitroimidazole toxicity.

Figure 6.

Pyruvate : ferredoxin oxidoreductase and hydrogenosomal malate dehydrogenase are almost lost upon extended subculture with the iron-chelator bipyridyl. The sections of 2D-gels where PFOR and hydrogenosomal malate dehydrogenase (hyd MDH) locate to are shown. After several subcultures in the presence of bipyridyl (here shown with four and nine subcultures respectively), the expression of these enzymes is strongly downregulated and, eventually, almost abolished.

Table 5.  Key enzymatic activities in normally grown and iron-depleted T. vaginalis C1 cells respectively.
Key enzyme activities in cell extractsPFORhyd MDH
nmol min−1 (mg protein)−1
  1. Enzyme activites of PFOR, hydrogenosomal malate dehydrogenase (hyd MDH) and lactate dehydrogenases (LDH) were determined in at least three independent experiments each.

Normal iron1438 ± 5201075 ± 1771630 ± 348
Iron-depleted0251 ± 285149 ± 1302


Covalent adduct formation with sulphhydryl groups and nitroimidazole toxicity

In this study we present evidence that covalent adduct formation with proteins and non-protein thiols could considerably contribute to the toxicity of nitroimidazole drugs in T. vaginalis (Fig. 7A and B). Covalent adduct formation of nitroimidazoles with proteins, cysteine and DNA had already been demonstrated in vitro (LaRusso et al., 1978; West et al., 1982; Wislocki et al., 1984; Kedderis et al., 1988) and was shown to depend on the transfer of two or more electrons to the drug's nitro group, resulting in the formation of nitrosoimidazole or hydroxylamine intermediates respectively. As the reduction of the nitro-group can occur in single-electron transfer steps, the nitroradical anion can be an intermediate for the formation of the nitrosoimidazole. As nitroimidazole radical anions are easily reoxidized by oxygen (Mason and Holtzman, 1975; Pervez-Reyes et al., 1980), the formation of nitrosoimidazoles from nitroimidazoles hardly occurs in aerobic organisms. This redox cycling effect very probably saves the aerobic cell from severe damage because nitrosoimidazoles deplete glutathione levels and are highly toxic to mammalian and bacterial cells (Ehlhardt et al., 1988; Ehlhardt and Goldman, 1989; Mulcahy et al., 1989; Bérubéet al., 1992).

Figure 7.

Model of activation and action of nitroimidazoles in T. vaginalis.
A. After entry into the cell by passive diffusion (1), nitroimidazoles are reduced either by flavin enzymes, for example, thioredoxin reductase (TrxR), or by reduced free flavins. Oxygen interferes with nitroreduction by reoxidizing the nitroradical anion before the respective nitrosoimidazole (2) can be formed. Nitroimidazoles or further reduced metabolites form covalent adducts with thiol groups, either free or protein-bound. Despite diminished overall nitrosoimidazole formation in the cell under high oxygen tensions, thioredoxin reductase is eventually covalently bound because it is a source of reactive nitroimidazole metabolites and therefore in close proximity to elevated concentrations of the latter. Under low oxygen tensions, reduction of nitroimidazole beyond the nitroradical stage is much more efficient, leading to the depletion of intracellular thiol pools (3). Covalent adduct formation with nitroimidazole metabolites also diminishes thioredoxin reductase activity, which is likely to impair the cell's oxidative defence because peroxidases depend on thioredoxin for their function. In the presence of elevated concentrations of oxygen (4), this might lead to the accumulation of H2O2, which results in oxidation of thiols and proteins, and in the generation of highly toxic hydroxyl radicals. Under low oxygen tensions, the depletion of thiols also leads to oxidative damage in the cell and, possibly, to the loss of essential thiols such as coenzyme A.
B. Physiological processes in T. vaginalis that are known or can be assumed to depend on thioredoxin-mediated reduction. Thioredoxin is required in the antioxidative defence as it recycles peroxidases (Coombs et al., 2004; this study) and methionine sulphoxide reductases (Boschi-Muller et al., 2005) to their catalytically active state. Besides numerous thioredoxin/thiol peroxidases, the T. vaginalis genome contains a gene for a ‘possible methionine sulphoxide reductase’ (XP_001307970) and a ‘methionine sulphoxide reductase family protein’ (XP_001300955) respectively. For dNTP synthesis, thioredoxin is required as a reducer of ribonucleotide reductase (Nordlund and Reichard, 2006). The assimilation of sulphate depends on the enzyme phosphoadenosine phosphosulphate reductase (PAPS reductase), which, in turn, is reduced by thioredoxin (Lillig et al., 1999). In the T. vaginalis genome a PAPS reductase family protein gene has been annotated (XP_001582281). Thioredoxins have also been suggested to aid in protein folding in various ways, for example, by reduction of chaperones (Berndt et al., 2008). Moreover, there is a large number of enzymes that are assumed to interact with thioredoxin, for example, malate dehydrogenase and enolase (Anderson et al., 1998; Wong et al., 2003; Hara et al., 2006), two proteins that have been identified as targets of nitroimidazoles in this study (Fig. 1). Finally, thioredoxin was also suggested to be required for iron-sulphur cluster synthesis in E. coli (Ding et al., 2005).

We are aware that the proposed scenario conflicts with the current notion of nitroimidazole toxicity in T. vaginalis and other microerophilic organisms, which suggests nitroimidazole radical anions to be the actual toxic agent, for example, by causing DNA-strand breaks (Edwards, 1993). Such DNA damage could indeed be demonstrated in vitro by applying electrochemical, but not chemical, reduction of nitroimidazoles (LaRusso et al., 1978; Zahoor et al., 1986; Zahoor et al., 1987). Still, in our opinion, DNA strand breaks occurred much too slowly to account for the rapid demise of nitroimidazole-treated T. vaginalis. Furthermore, DNA damage was inversely proportional to the reduction rate, which is the exact opposite of the situation in vivo, because higher doses of nitroimidazoles result in higher toxicity. It is also not clear how nitroradical anions could reach the DNA in the nucleus of T. vaginalis without having before reacted with lipids, proteins or other compounds.

Identified targets of nitroimidazole drugs in T. vaginalis

We detected 10 and identified seven cytosolic proteins in T. vaginalis that form adducts with metronidazole (Fig. 1, Table 1), including thioredoxin reductase that had already been identified as a target in E. histolytica (Leitsch et al., 2007). On 2D-gels of hydrogenosomal extracts from treated cells, only one weakly expressed protein could be repeatedly observed, which seemed to be shifted in terms of its pI. Covalent adduct formation with 5-nitroimidazoles was confirmed by treatment with tinidazole, which shifted the same proteins, although less wide with regard to pI (Fig. 2). This effect is arguably caused by the different pIs of metronidazole and tinidazole. As these observations were fully congruent with those made in our previous study in E. histolytica (Leitsch et al., 2007), nitroimidazole adduct formation was not directly confirmed by mass spectrometry as described with recombinantly expressed E. histolytica thioredoxin and thioredoxin reductase (Leitsch et al., 2007). With the exception of glucose 6-phosphate isomerase, all identified T. vaginalis proteins that are shifted on 2D-gels upon nitroimidazole treatment have been reported to, or can be assumed to be involved in thioredoxin-mediated redox regulation. Evidently, thioredoxin reductase has to interact with thioredoxin in order to re-establish the two reactive thiol groups of thioredoxin (Coombs et al., 2004), and ribonucleotide reductase is paradigmatic for an enzyme that depends on reduction by thioredoxin (Nordlund and Reichard, 2006). Thioredoxin peroxidase and thiol peroxidase require reduction by thioredoxin at their active site for the degradation of hydrogen peroxide (Coombs et al., 2004; Camier et al., 2007), whereas enolase and malate dehydrogenase have been identified as interaction partners of thioredoxin in plants (Anderson et al., 1998; Wong et al., 2003; Hara et al., 2006). We suggest that, as in E. histolytica, these seven proteins are rendered vulnerable to the attack of reactive nitroimidazole metabolites because they are in spatial proximity (or, in the case of thioredoxin reductase, identical) to a source of activated nitroimidazoles in the cell, that is, thioredoxin reductase. Of course, as hydrogenosomal thioredoxin reductases also exist in T. vaginalis (Mentel et al., 2008), similar processes could take place in the hydrogenosome as were observed in the cytosol. Still, as educible from 2DE, the effects of metronidazole treatment on the protein expression profiles of hydrogenosomes were very small (Fig. 2).

In nitroimidazole-treated cells, thioredoxin reductase displayed a considerably diminished capability to reduce thioredoxin (Fig. 3A), possibly due to covalent adduct formation. This effect was more pronounced with nitroimidazoles with a higher reduction rate, that is, the 2-nitroimidazole azomycin and the 5-nitroimidazole ronidazole, than with the other nitroimidazoles assayed. Diminished thioredoxin reductase activity consequently leads to reduced activities of those enzymes that depend on thioredoxin for their function. In this study we showed that the breakdown of H2O2 by peroxidases is slowed down congruently with the decrease in thioredoxin reductase activity. A similar effect on ribonucleotide reductase activity is likely as ribonucleotide reductase in T. vaginalis is a class I-like enzyme (Nordlund and Reichard, 2006), which directly depends on reduction by thioredoxin for activity. Of course, also a direct effect of nitroimidazole adduct formation with ribonucleotide reductase cannot be ruled out. However, our observation that enolase and cytosolic malate dehydrogenase activities were, if at all, only minimally affected by nitroimidazole binding does not allow such a conclusion to be drawn at present.

We assume that the contribution of the described impairments of thioredoxin reductase to nitroimidazole toxicity strongly depends on the ambient oxygen concentration (Fig. 7A). For example, under low oxygen tensions azomycin is less toxic to T. vaginalis than metronidazole, although it leads to a stronger decrease in thioredoxin reductase activity (Fig. 3A). It seems therefore unlikely that thioredoxin reductase is the prime target of nitroimidazoles in the (near) absence of oxygen. Under these conditions, nitroimidazole toxicity could be rather based on the efficient depletion of intracellular thiol pools, which most probably leads to a breakdown of the redox equilibrium in the cell and to enhanced unspecific disulphide formation between proteins. Moreover, certain non-protein thiols, for example, CoA, are of vital importance and their depletion might have an immediate effect on the cell. In this context, it is interesting to note that the non-protein thiol content of the cell corresponded well with cell motility and that 5-nitroimidazoles lowered the non-protein thiol levels more than the 2-nitroimidazole azomycin. Probably, activated 5-nitroimidazoles form covalent adducts with thiol groups more avidly than 2-nitroimidazoles.

Under microaerobic or aerobic conditions the situation could be different. It is known that susceptible T. vaginalis strains are almost equally sensitive to metronidazole under aerobic as under anaerobic conditions, although the uptake of 14C-metronidazole label, which we interpret predominantly as metronidazole adduct formation, is decreased to only 5–10% (Müller and Gorrell, 1983). The sensitivity to metronidazole under aerobic conditions, however, is very probably not due to reactive oxygen species generated by redox cycling of the metronidazole radical anion, because reduction rates of metronidazole and other 5-nitroimidazoles are so low that hardly any oxygen consumption can be observed (Moreno et al., 1984). This brings thioredoxin reductase into focus again because we observed that covalent adduct formation also takes place under microaerobic conditions, although somewhat slower (data not shown). Thus, it is possible that under microaerobic and aerobic conditions the impaired antioxidant defence mechanisms in the metronidazole-treated cell can make up for the reduced extent of overall covalent adduct formation. Surely, a diminished thioredoxin reductase activity would be a decisive disadvantage for T. vaginalis when encountering the host defence, which includes the generation of oxidative stress. In addition to antioxidative defence, several other physiological processes that are known to require thioredoxin could occur in T. vaginalis, including dNTP synthesis by ribonucleotide reductase, sulphate reduction and redox control of chaperones (Fig. 7B). In how far a disturbance of these processes contributes to metronidazole toxicity remains to be resolved.

Of course, as radioactive metronidazole was also found to be bound to DNA in metronidazole-treated T. vaginalis (Ings et al., 1974), it is possible that covalent adduct formation with DNA also has a role in nitroimidazole toxicity. However, metronidazole-treated T. vaginalis become immotile and die very quickly when treated with nitroimidazoles. This is difficult to reconcile with DNA damage whose consequences are less acute and which rather results in either apoptosis in multicellular organism or in growth inhibition in unicellular organisms. Other targets in T. vaginalis could be hydrogenosomal iron-sulphur proteins, for example, PFOR, ferredoxins or hydrogenases, whose oxidation by nitroimidazole radicals could result in a breakdown of the cellular metabolism. Still, 4-nitroimidazole that is also reduced to the respective nitroradical anion in intact hydrogenosomes (Yarlett et al., 1987) is non-toxic to T. vaginalis. Moreover, iron-depleted cells that almost totally lack PFOR are similarly sensitive to metronidazole as normal cells (Fig. 6, Fig. S4).

The role of thioredoxin reductase and free flavins in nitroimidazole activation and implications for metronidazole resistance in T. vaginalis

Our data suggest an involvement of the flavin enzyme thioredoxin reductase and of free reduced flavins in nitroimidazole activation. Purified recombinant T. vaginalis thioredoxin reductase showed strong nitroreductase activity with the antitrichomonal nitrofuran drug furazolidone and also reduced nitroimidazoles (Table 2). Of course, due to their low redox potentials, the reduction rates of nitroimidazoles by thioredoxin reductase are rather low (roughly two orders of magnitude lower as compared with furazolidone). However, it is known that nitroimidazoles are poorly reduced in vivo as well (Moreno et al., 1984) and in electron spin resonance experiments for the measurement of steady-state concentrations of nitroradicals, 10–50 times higher concentrations of metronidazole than of the nitrofuran drug nitrofurantoin are needed (Moreno et al., 1984; Chapman et al., 1985; Lloyd and Pedersen, 1985). Moreover, the metronidazole nitroradical anion has a much longer half-life than nitrofuran nitroradicals (Edwards, 1993). Hence, in the electron spin resonance experiments reduction of metronidazole must have been very much slower than that of nitrofurans as well.

An important role of thioredoxin reductase in nitroimidazole activation is indicated by the fact that it is a rather strongly expressed protein (0.3% of the total cellular protein as visualized on Coomassie-stained 2D-gels of whole-cell extracts) and that it is largely inactive in highly metronidazole-resistant cells (less than 5% of the normal rate). As the expression levels of thioredoxin reductase remained unchanged in resistant cells and as the addition of FAD re-established more than 60% of the original activity (Fig. 5), it is probable that the loss of thioredoxin reductase activity in metronidazole-resistant cells occurs due to a lack of FAD cofactor. In addition to the near to total loss of thioredoxin activity no reduction of FAD, FMN and riboflavin seems to occur in metronidazole-resistant cells (Table 4), whereas extracts of susceptible cells readily reduce these flavins. This finding indicates that levels of reduced flavins in the metronidazole-resistant cell line must be very low. This might be highly relevant with regard to nitroimidazole toxicity as nitroimidazole activation by flavin enzymes (West et al., 1982; Clarke et al., 1982; Kedderis et al., 1988) and by reduced flavins is well documented (Clarke et al., 1980).

Nitroradical anion formation from metronidazole and other 5-nitroimidazoles by microsomal fractions or flavin enzymes was repeatedly demonstrated (Pervez-Reyes et al., 1980; Lloyd and Pedersen, 1985; Viodéet al., 1999) and reduction of nitroimidazoles by reduced flavins and flavin enzymes was shown to proceed to the nitroso-intermediate and beyond under microaerobic/anaerobic conditions (Clarke et al., 1980). A cell with diminished flavin enzyme activity (as described here for thioredoxin reductase), lowered flavin levels, and/or diminished flavin reductase activity can therefore be expected to be more tolerant to nitroimidazole drugs. However, such changes must also have grave consequences for the cell's physiology, as evident with our metronidazole-resistant line. For example, highly resistant cells grow considerably slower than susceptible cells, depend on high concentrations of cysteine in the medium and are more vulnerable to oxygen. These impairments very likely result from the near loss of thioredoxin reductase activity, which leads to diminished peroxidase activities and to a higher sensitivity to disulphide stress. Rather than having a direct role in metronidazole resistance, the upregulation of antioxidative enzymes as observed in metronidazole-resistant T. vaginalis (Rasoloson et al., 2001; this study) could therefore be a consequence of decreased thioredoxin reductase activity. It is plausible that the cell reacts to diminished thioredoxin-mediated reduction by upregulating peroxidases in order to maintain at least some peroxidase activity. The expression of peroxidases, in turn, could be coupled to the expression of superoxide dismutases. However, also an alternative explanation can be given: flavin reductases and flavins are suggested to be involved in the reduction of ferric to ferrous iron in E. coli (Coves and Fontecave, 1993; Woodmansee and Imlay, 2002; Crossley et al., 2007) and evidently riboflavin plays an influential role in iron acquisition in microaerophilic bacteria such as Campylobacter jejuni (Crossley et al., 2007) and H. pylori (Worst et al., 1998). Thus, it is conceivable that the loss of flavin reductase activity in metronidazole-resistant T. vaginalis also has a distorting impact on iron metabolism. Depletion of intracellular iron levels could affect the expression of numerous proteins, including iron superoxide dimutases. As T. vaginalis only possesses iron-containing superoxide dismutases, the cell could upregulate their expression when iron is scarce in order to preserve some superoxide dismutase activity. Higher levels of superoxide dismutase, in turn, could lead to increased peroxiredoxin expression as has been demonstrated in E. histolytica (Wassmann et al., 1999). In addition, the expression levels of iron-sulphur proteins, and other proteins that either directly or indirectly depend on them, are known to be affected by low (ferrous) iron levels (Imlay, 2006). Such a connection has been observed in a wide range of organisms, including E. coli (McHugh et al., 2003), Saccharomyces cerevisiae (Puig et al., 2005) and T. foetus (Vanacova et al., 2001). Also iron-depleted T. vaginalis sharply downregulate the expression of PFOR and hydrogenosomal malate dehydrogenase (Fig. 6). Nevertheless, iron-restricted T. foetus (Sutak et al., 2004) and T. vaginalis (this study) remain fully susceptible to nitroimidazoles, which clearly contradicts the current model of nitroimidazole activation in trichomonads. Based on this observation, we suggest that the loss of iron-sulphur enzymes and other hydrogenosomal pathways in metronidazole-resistant T. vaginalis is not a cause, but probably a result of metronidazole resistance, that is, a severe disruption of the cellular flavin metabolism.

It is questionable whether a T. vaginalis strain with the characteristics of our metronidazole-resistant C1 could occur naturally, mainly due to its dependence on high concentrations of cysteine and due to its low oxygen tolerance. Indeed, metronidazole-resistant clinical isolates display fundamental differences as compared with metronidazole-resistant cell lines generated in vitro. The former were found to activate the drug but also to have defective oxygen-scavenging mechanisms, leading to redox cycling of metronidazole (Lloyd and Pedersen, 1985; Yarlett et al., 1986b). Consequently, these strains lose most of their resistance when treated with metronidazole under anaerobic conditions (Müller and Gorrell, 1983). However, clinical isolates that displayed diminished drug reduction have been described (Clackson and Coombs, 1982) and at least one of the better studied metronidazole-resistant clinical isolates, CDC 85, is resistant to treatment level concentrations of metronidazole (50–100 μM) even under anaerobic conditions (Yarlett et al., 1986b). Interestingly, in this strain, ‘thiol reductase activity’, which is likely to be identical with thioredoxin reductase activity, amounts to less than 20% as compared with a metronidazole-sensitive isolate (Ellis et al., 1994).


The observations we made with regard to the effects of nitroimidazole treatment appear to be very similar in T. vaginalis and E. histolytica (Leitsch et al., 2007). In both parasites the flavin enzyme thioredoxin reductase can act as a nitroreductase, capable of reducing metronidazole and other nitroheterocyclic drugs. In addition, in both parasites a comparably small number of proteins, of which the most are known or expected to be involved in the thioredoxin-mediated redox network, form adducts with reactive nitroimidazole metabolites. Upon nitroimidazole treatment, thioredoxin reductase from either parasite displays diminished activity and cellular non-protein thiol levels are depleted.

Furthermore, we assume that the mechanisms of metronidazole resistance, as described here with T. vaginalis, could be similar to those in microaerophilic/anaerobic bacteria. The loss of PFOR activity and the concomitant upregulation of lactate dehydrogenases activity were also observed in metronidazole-resistant B. fragilis and Clostridium perfringens (Diniz et al., 2004; Narikawa et al., 1991; Sindar et al., 1982). Moreover, the loss of PFOR expression has, if at all, only minimal effect on metronidazole susceptibility in B. fragilis (Diniz et al., 2004) and H. pylori (Debets-Ossenkopp et al., 1999). To which extent the other observations made with nitroimidazole action in T. vaginalis do also apply for bacteria, remains to be resolved.

Experimental Procedures

Cell culture and determination of growth rates

When not indicated otherwise, T. vaginalis C1 (ATCC 30001) cells were grown axenically 37°C in 25 ml polysterene culture flasks (Falcon, Becton-Dickinson) completely filled with TYM medium (Diamond, 1957). Strictly anaerobic (18% CO2, 0% O2) and microaerobic (8% CO2, 5% O2) conditions were generated by growing the cells in vented culture flasks in an anerobic jar with Anaerocult A or Anaerocult C respectively (Merck Chemicals). Growth rates of metronidazole-susceptible and metronidazole-resistant C1 were determined by starting cultures with inoculates of 50 000 cells ml−1 and counting cells in a Bürker-Türk chamber after 24 h. The cells were either grown in completely filled and sealed culture flasks (‘normal’ conditions) or in half-filled, vented culture flasks in an aerated incubator (‘aerobic’ conditions). Growth rates were determined in six independent experiments. E. coli BL21(DE3) was grown in LB medium with appropriate antibiotics. Agar plates contained 15 g l−1 of agar.

Cell harvest and preparation of cell lysates for 2DE and image analysis

Cells were harvested by centrifugation at 700 g at room temperature for 5 min and washed twice with PBS to remove residual serum components. Preparation of cell lysates for two-dimensional electrophoresis 2DE (Leitsch et al., 2005) and purification of hydrogenosomes (Pütz et al., 2005) were performed as described previously. 2DE was performed as previously described (Leitsch et al., 2005). About 100 μg protein was applied when 2D-gels were to be silver-stained (Blum et al., 1987); 500–1000 μg were applied when 2D-gels were to be stained with Coomassie brilliant blue R-250. After staining, gels were scanned with an Epson 1680 Pro scanner and spots were quantitatively and qualitatively analysed with the Melanie 2D-gel analysis software (GeneBio, Geneva, Switzerland).

Protein identification and mass spectrometry of intact proteins

The excised 2DE spots were destained, digested with trypsin and analysed by LC-ESI-QTOF-MS/MS as described previously (Kolarich et al., 2006). For protein identification, the MS/MS data were subjected to database search against the NCBI genome database (Carlton et al., 2007) using the Mascot search engine ( and Protein Global Server 2.1 (Waters-Micromass, Manchester, UK).

Expression of recombinant hexahistidine-tagged T. vaginalis thioredoxin (recTvTrx) and thioredoxin reductase (recTvTrxR) in E. coli

The genes for thioredoxin reductase and thioredoxin were amplified from genomic DNA. The primers for thioredoxin reductase (XP_001316923, CAD47837) were 5′-TAC GTA CGC ATA TGT CTG CTC AAG CAT TCG ATC TCG TTA TCA TTG GC-3′ (forward) and 5′-TCA TCC AGG AAT TCT TAG TGA TGG TGA TGG TGA TGG TCA CTG AGA TAT CTC TCA GCA AGG-3′ (reverse), those for thioredoxin (CAD47836) 5′-TAC GTA CGC ATA TGT CCG ATC CAA TTG TTC ACT TCA AT-3′ (forward) and 5′-TCA TCC AGC TCG AGT TAG TGA TGG TGA TGG TGA TGT TTG AAC TTT TCA ATA TCA GCT TTG AT-3′ (reverse). Forward primers include an NdeI restriction site, whereas the reverse primers bear either an EcoRI (thioredoxin reductase) or a XhoI (thioredoxin) restriction site and a sequence encoding a hexahistidine tag for convenient protein isolation. PCR fragments were ligated into the pET-17b vector (Novagen). The plasmid sequences were confirmed on both strands by using T7 and pET reverse primers (GATC Biotech, Constance, Germany). The plasmids were transformed into E. coli BL21(DE3) (for expression of thioredoxin reductase) or BL21-AI cells (for expression of thioredoxin). Transformants were selected on 20 μg ml−1 ampicillin. Expression of recombinant proteins was induced by addition of 0.5 mM IPTG (and 0.2% l-arabinose when thioredoxin was expressed). Three hours after induction, cells were harvested and disrupted by vigorous grinding in a mortar. Subsequently, recombinant proteins were purified via Ni-NTA spin columns (Qiagen).

Assaying nitroreductase activity of T. vaginalis thioredoxin reductase

When not indicated otherwise, all measurements were performed under aerobic conditions at 37°C in a Perkin Elmer Lambda 25 UV/VIS spectrophotometer. Reduction of furazolidone (10–100 μM) by recTvTrxR was measured by determining NADPH consumption at λ = 340 nm (Δε340 = 6.2 mM−1 cm−1). The reaction buffer contained 100 mM Tris/HCl pH 6.2, 1 mM EDTA and 0.2 mM NADPH. Reduction of nitroimidazoles was measured in a modified assay (Viodéet al., 1999; Cenas et al., 2006) via reduction of cytochrome c at λ = 550 nm (Δε550 = 20 mM−1 cm−1), either directly by reduced nitro compounds or indirectly by superoxide radical anions that are generated when nitroradicals transfer an electron to molecular oxygen in the reaction buffer. In any case, nitro compounds had been previously reduced by recTvTrxR. Therefore, it was assumed that reduction of one nitro group by recTvTrxR subsequently resulted in one reduced cytochrome c molecule. Reaction mixtures contained 100 mM Tris/HCl pH 6.2, 1 mM EDTA, 0.5 mM NADPH, 50 μM cytochrome c, 5 μg ml−1 recTvTrxR (= 150 nm) and varying concentrations of nitroimidazoles. Furazolidone and all nitroimidazoles were obtained from Sigma. For measurements under anaerobic conditions, the reaction buffer was incubated ON in an anaerobic jar (with Anaerocult A) prior to experiments.

Enzyme assays with hydrogenosome-free extracts

For measurements of the influence of nitroimidazole treatment on thioredoxin-reducing activity, cells were either treated for 2 h with 50 μM of the respective nitroimidazole or left untreated. After harvest, hydrogenosome-free extracts were obtained by taking up the cells in Tris buffer (100 mM Tris/HCl pH 7.5, 1 mM EDTA) and by disrupting them in a Dounce homogenizer. Afterwards, the extracts were centrifuged at 20 000 g for 10 min in order to remove the hydrogenosomes. Thioredoxin-reducing activity of cell extracts was measured by determining reduction of DTNB (OD412) upon addition of recombinant T. vaginalis thioredoxin (30 μg ml−1). All measurements were performed under aerobic conditions at 37°C. When not explicitly stated otherwise, 40 μg of extract was used per reaction (100 mM Tris/HCl pH 7.5, 1 mM EDTA, 1 mM DTNB, 0.5 mM NADPH). Hydrogen peroxide-induced NADPH consumption, either with or without thioredoxin (30 μg ml−1), was measured at λ = 340 nm as described before for purified peroxidases (Kim et al., 2005) but with altered buffer composition: 100 mM Tris/HCl pH 7.5, 1 mM EDTA, 0.2 mM NADPH and 4 mM H2O2. Enolase activity was measured as described previously (Garchow et al., 2006), but EDTA was omitted and reactions were buffered in 100 mM Tris/HCl pH 7.5. The activities of cytosolic NADH-dependent malate dehydrogenase and lactate dehydrogenase were measured along the lines described in Wu et al. (1999) using 5 μg ml−1 cell extract and different buffer compositions: 100 mM Tris/HCl pH 7.5, 0.2 mM NADH, 1 mM EDTA, 10 mM oxaloacetate for malate dehydrogenase activity and 100 mM Tris/HCl pH 7.5, 0.2 mM NADH, 1 mM EDTA, 2 mM pyruvate for lactate dehydrogenase activity. Flavin reductase activity was measured by determining the consumption of NADPH (0.2 mM) upon addition of flavins (10 μM) at λ = 340 nm.

Other enzyme assays

The activities of PFOR and hydrogenosomal malate dehydrogenase were measured with hydrogenosome-enriched fractions (obtained by centrifuging at 20 000 g, 20 min) as described in Lindmark and Müller (1973) and Hrdy et al. (1993) respectively, but in case of the latter, 1 mM NAD+ and 100 mM Tris buffer pH 7.5 were used.

Determination of non-protein thiol levels

Cells were incubated for 1, 2 or 3 h, with 50 μM of the nitroimidazole drug indicated. Measurement of non-protein thiols was performed as described (Leitsch et al., 2007).


The presented work was supported in part by Grant P15960 from the Austrian Science Fund. We are also grateful to Iain Wilson for critical reading of the manuscript.