The Amitochondriate Eukaryote Trichomonas vaginalis Contains a Divergent Thioredoxin-linked Peroxiredoxin Antioxidant System*

Trichomonas is an amitochondriate parasitic protozoon specialized for an anaerobic lifestyle. Nevertheless, it is exposed to oxygen and is able to cope with the resultant oxidative stress. In the absence of glutathione, cysteine has been thought to be the major antioxidant. We now report that the parasite contains thioredoxin reductase, which functions together with thioredoxin and thioredoxin peroxidase to detoxify potentially damaging oxidants. Thioredoxin reductase and thioredoxin also reduce cystine and so may play a role in maintaining the cellular cysteine levels. The importance of the thioredoxin system as one of the major antioxidant defense mechanisms in Trichomonas was confirmed by showing that the parasite responds to environmental changes resulting in increased oxidative stress by up-regulating thioredoxin and thioredoxin peroxidases levels. Sequence data indicate that the thioredoxin reductase of Trichomonas differs fundamentally in structure from that of its human host and thus may represent a useful drug target. The protein is generally similar to thioredoxin reductases present in other lower eukaryotes, all of which probably originated through horizontal gene transfer from a prokaryote. The phylogenetic signal in thioredoxin peroxidase is weak, but evidence from trees suggests that this gene has been subject to repeated horizontal gene transfers from different prokaryotes to different eukaryotes. The data are thus consistent with the complexity hypothesis that predicts that the evolution of simple pathways such as the thioredoxin cascade are likely to be affected by horizontal gene transfer between species.

Trichomonas vaginalis is the protozoan parasite responsible for trichomoniasis in humans (1). This is the most common non-viral sexually transmitted infection, with an estimated Ͼ170 million cases occurring each year (2), and has been im-plicated as a major risk factor in predisposition to human immunodeficiency virus/AIDS (3). Indeed it has been suggested that successful treatment of trichomoniasis may be the most cost-effective means by which to reduce human immunodeficiency virus incidence (3). Current chemotherapy relies upon a single group of drugs, the 5-nitroimidazoles, and there are worrying signs that drug resistance may be emerging as a significant problem (4). Thus there is a need for new chemotherapeutic tools.
The parasite itself is an unusual eukaryote and has been considered to be one of the earliest branching organisms (5), although current evidence for this view is not compelling (6 -8). It lacks conventional mitochondria but possesses organelles termed hydrogenosomes that share common ancestry with mitochondria (9, 10) and appear to be adaptations for the parasite's existence in an environment containing only low oxygen concentrations. Trichomonads are fundamentally fermentative organisms, with oxygen apparently not making a significant contribution to energy metabolism (11). Nevertheless, the cells are exposed to oxygen in the natural environment. This was most elegantly demonstrated by the finding that T. vaginalis isolated from patients unresponsive to the standard chemotherapy treatment with metronidazole is drug-resistant in in vitro tests, but only if oxygen is present (4,(12)(13)(14).
The implications for T. vaginalis of exposure to oxygen have been pondered over many years. Despite a report of the beneficial effect of low oxygen concentrations (growth being significantly enhanced (15)), it is generally considered that oxygen provides problems rather than benefits. Some of the parasite's enzymes are inactivated by oxygen itself, notably key proteins of the hydrogenosomes, and various metabolites likely to arise from the metabolism of oxygen (such as hydrogen peroxide and hydroxyl free radical) are generally harmful to cells and so need to be countered. Most eukaryotes have glutathione as a key redox buffer and antioxidant, but trichomonads lack this and similar thiols (16). Thus cysteine has been considered the major cellular reducing agent and antioxidant, although T. vaginalis is able to generate thiols (propanethiol, methanethiol, and hydrogen sulfide, from the action of the unusual bacterial-like enzyme methionine-␥-lyase (17)), which have been postulated to have antioxidant roles (16). Nevertheless, it was believed that the organism relies heavily upon cytosolic NADH oxidase (reducing oxygen to water) and NADPH oxidase (reducing oxygen to hydrogen peroxide) to prevent permeation of oxygen to the hydrogenosomes (18). However, the generation of hydrogen peroxide by NADPH oxidase, and superoxide dismutase, poses the question as to how this and other reactive oxygen species (ROS) 1 are removed as T. vaginalis lacks catalase and glutathione-dependent peroxidase activities. An ascorbate peroxidase has been reported (19), but it seemed likely that another system must also exist.
A family of peroxidases has been discovered in recent years that do not use glutathione as the reductant but instead are dependent on reduction by a small protein known as thioredoxin (Trx), which is itself reduced by thioredoxin reductase (TrxR) (20 -22). These Trx-dependent peroxidases, now commonly known as peroxiredoxins (designated TrxP), are seemingly ubiquitous in eukaryotes with there being a number of isoforms localized in different cellular compartments (22)(23)(24)(25)(26). It has been shown that peroxiredoxins reduce hydrogen peroxide and alkyl hydroperoxides and therefore constitute a major cellular protection system against the devastating consequences of oxidative damage (20,26).
Peroxiredoxin-linked detoxification may occur in all eukaryotes but has been perceived to be of special relevance to some parasites, including helminths, trypanosomatids, and the malaria parasite Plasmodium falciparum, as a crucial means of detoxifying peroxides in the apparent absence of enzymes such as catalase and glutathione peroxidases (26 -30). Such metabolism may be of particular importance to amitochondriate eukaryotes, such as the protozoa Entamoeba and Giardia as well as Trichomonas, because they all lack glutathione (30,31). Interestingly, Entamoeba histolytica possesses a peroxiredoxin-linked detoxification system that is functionally similar to the bacterial AhpF/AhpC system in that it does not involve thioredoxin as an intermediate electron acceptor (32,33). Giardia lamblia has also been reported to contain TrxR (34), but the way in which the enzyme functions has not yet been addressed.
This study was undertaken both to elucidate whether a peroxiredoxin cascade involving TrxR and Trx is a fundamental antioxidant mechanism of T. vaginalis and to investigate whether the biochemical characteristics of the components of the cascade provide optimism that it may be a drug target. Because the complexity hypothesis (35) predicts that horizontal gene transfer (HGT) between species should particularly affect simple systems like peroxiredoxin cascades, we also investigated the evolutionary origins of the Trichomonas genes.

EXPERIMENTAL PROCEDURES
Growth and Harvesting of Parasites-A clonal cell line (G3) of T. vaginalis was routinely grown axenically in modified Diamond's medium and harvested as previously described (17). Parasites were either stored as cell pellets at Ϫ70°C for the Western blot analyses and genomic DNA isolation or suspended in TRIzol reagent (Invitrogen) for the RNA preparations and analyses.
Cloning of TRXR, TRX, and TRXP of T. vaginalis-Genes apparently corresponding to TRXR, TRX, and TRXP were identified in an EST data base of T. vaginalis G3. 2 Total RNA was prepared from T. vaginalis G3 using the SV Total RNA Isolation System (Promega). The 5Ј RACE system (Invitrogen) was used to clone the 5Ј-ends of the TRX, TRXR, and TRXP mRNAs. Nested gene-specific primers were designed based on EST sequences. Those used for first strand cDNA synthesis and amplification of cDNA, respectively, were NT3 and NT4 (for TRXR), NT5 and NT6 (for TRX), and NT7 and NT15 (for TRXP) as detailed in Table I. PCR was performed using Pfu DNA polymerase (Promega) and the following conditions: 94°C for 5 min; 30 cycles of 94°C for 30 s, 66°C for 30 s, and 72°C for 3 min; then 72°C for 7 min. PCR products were cloned into vector pGEM-T (Promega) and sequenced. The sequence of the 5Ј-end of each mRNA, including the translation initiation codon, was determined by analyzing the sequence of at least two PCR products from each of two independent 5Ј RACE experiments.
The full-length TRX coding region was amplified from EST 144 using primers NT10 and NT11 and the PCR conditions detailed above for the 5Ј RACE. The primers contain restriction sites to facilitate cloning of the PCR product into the expression vector pET21aϩ. The PCR product was digested with NdeI and XhoI and ligated to NdeI-and XhoIdigested pET21aϩ to generate pBP1. In the same way and using the same PCR conditions, the full-length TRXP coding region was amplified from EST450 using primers NT12 and NT28 and cloned into pET21aϩ to give pBP24. The TRXR coding region was amplified from total RNA by reverse transcriptase-PCR. First strand cDNA was synthesized using Superscript II reverse transcriptase (Invitrogen) and oligo(d(T) 15 ) as a primer, according to the manufacturer's instructions. The TRXR was amplified using primers NT1 and NT2 and the Expand High Fidelity PCR system (Roche Applied Science) (94°C for 2 min, 25 cycles of 94°C for 15 s, 55°C for 30 s, and 72°C for 2 min, 72°C for 7 min) and cloned into pGEM-T. The 0.92-kb NdeI and NotI fragment containing the full-length TRXR sequence was ligated into pET21aϩ previously digested with the same enzymes to give pBP3. The sequence of each construct was confirmed on both strands by the University of Glasgow MBSU Sequencing Facility. Sequence analysis was carried out using Vector NTI software (Informax). Plasmids were introduced into Escherichia coli strain BL21/DE3 for protein expression. Recombinant proteins have a C-terminal 6ϫ histidine tag to facilitate purification using Ni 2ϩ -nitrilotriacetic acid-agarose.
When ambiguous sites and indels were removed for the broadest taxonomic sampling for TrxR (74 taxa), a dataset of 268 sites was left for analysis. A second alignment of 284 selected sites was also used for a reduced taxa set (38 taxa). Phylogenetic analyses of these two datasets recovered essentially identical topologies for the shared taxa. The programs TREE-PUZZLE 5.0 (38) and MRBAYES (39) were used to perform protein maximum likelihood (ML) and Bayesian analyses. To assess support for relationships we used bootstrapping of protein max- 1 The abbreviations used are: ROS, reactive oxygen species; TRXR, gene encoding thioredoxin reductase; TrxR, protein encoded by TRXR; rTrxR, recombinant TrxR (equivalent nomenclature is used for thioredoxin (Trx) and thioredoxin peroxidase (TrxP); TrxP is also known as peroxiredoxin); DTNB, 5,5Ј-dithiobis(2-nitrobenzoic acid); E-64, N-trans-epoxysuccinyl-L-leucine-4-guanidinobutylamide; HGT, horizontal gene transfer; EST, expressed sequence tag; RACE, rapid amplification of cDNA end; ML, maximum likelihood.  imum likelihood distances using the script PUZZLEBOOT (available from the author Andrew Roger at hades.biochem.dal.ca/Rogerlab/) in combination with SEQBOOT and NEIGHBOUR. 3 These analyses used the JTT substitution matrix with site-rate heterogeneity expressed by a gamma correction, including a fraction of potentially invariable sites. The parameters for the gamma correction were estimated using TREE-PUZZLE 5.0. The same analytical approaches were used to analyze the TrxP dataset comprising 47 taxa and 175 aligned positions.
To investigate the support for alternative phylogenetic positions for the T. vaginalis TrxR and TrxP sequences, for example for their monophyly with the diplomonad parasites G. lamblia and Spironucleus barkhanus sequences, we performed constrained parsimony analyses in PAUP4.0b8. The MP trees for each hypothesis were then evaluated for their maximum likelihood values using the Shimodara Hasegawa test (41) implemented in the p4 software (Peter Foster, The Natural History Museum, London).
[␣-32 P]dATP-labeled DNA probes were prepared from agarose gel-purified restriction endonuclease fragments using Prime-IT II random primer kit (Stratagene) and purified on Microspin S-200 HR columns (Amersham Biosciences). The probes used were: TRXR, a 0.8-kb XhoI fragment of EST 610; TRXP, a 0.5-kb EcoRI/XhoI fragment of EST 450; TRX, a 0.8-kb PruII fragment of EST 144; ␣-actin, a 1.2-kb EcoRI/XhoI fragment of EST 197. Hybridizations were performed at 42°C overnight in 5ϫ SSPE, 5ϫ Denhardt's solution, 50% formamide, 0.5% SDS, and 100 g/ml denatured salmon sperm DNA. Filters were washed twice for 10 min at room temperature in 2ϫ SSC, 0.5% (w/v) SDS, and twice for 30 min at 55°C in 0.1ϫ SSC, 0.1% SDS. Storage phosphor screens were exposed to the labeled filters and scanned using a Typhoon 8600 Imager (Amersham Biosciences). Levels of mRNA were quantified using Im-ageQuaNT image analysis software (Amersham Biosciences). Filters were stripped with boiling 0.1% SDS, rinsed with 2ϫ SSC, and reprobed. The T. vaginalis ␣-actin mRNA levels were used to normalize TRX, TRXR, and TRXP mRNA levels. Normalized mRNA levels in the treated cultures were expressed relative to the control culture, which was given an arbitrary value of 1.0 unit.
Production of Recombinant TrxR, Trx, and TrxP-Single colonies of BL21/DE3 harboring pBP1, pBP3, and pBP24, respectively, were grown in Luria Bertani medium, the expression of the recombinant proteins was induced with 1 mM isopropyl thio-␤-D-galactoside, and recombinant proteins purified using Ni 2ϩ -nickel-nitrilotriacetic acid-agarose (Qiagen) according to the manufacturer's recommendations using a BioCad fast protein liquid chromatography system. The eluted proteins (rTrxR, rTrx, and rTrxP) were stored at 4°C with 0.02% azide. They remained highly active, as assessed by the enzymatic analysis, for more than 2 months. rTrxR concentration was assessed by using the molar extinction coefficient of 11,300 M Ϫ1 cm Ϫ1 at 453 nm (42). rTrx concentration was determined by using the molar extinction coefficient of 13,700 M Ϫ1 cm Ϫ1 at 280 nm (43). rTrxP concentration was determined using the BCA protein assay microtitre plate method (Pierce), with bovine serum albumin as standard.
Enzyme Assays-TrxR activity was measured by monitoring the oxidation of NADPH at 340 nm in a reaction mixture comprising 0.1 M potassium phosphate, pH 7.0, 5 mM EDTA, 0.2 mM NADPH, 200 g/ml insulin (bovine pancreas, 28.1 USP units/mg, Sigma), 12.5 M rTrx, and 44 pmol of rTrxR. The reaction was initiated by addition of the rTrx (44,45). The K m value for rTrx was estimated by varying the rTrx concentrations from 0.07 to 22 M. rTrxP activity was measured by monitoring the oxidation of NADPH at 340 nm in a reaction mixture comprising 0.1 M potassium phosphate, pH 7.0, 5 mM EDTA, 0.2 mM NADPH, 12.5 M rTrx, 74 pmol of rTrxR, 3 nmol of rTrxP, and 20 M peroxide (hydrogen peroxide, tert-butylhydroperoxide or cumene hydroperoxide). The reaction was initiated by the addition of the peroxide. NADPH oxidase activity was determined by monitoring the oxidation of NADPH at 340 nm in reaction mixtures comprising 0.1 M potassium phosphate, pH 7.0 or 9.0, 5 mM EDTA, 0.2 mM NADPH, and 148 pmol of rTrxR. 5,5Ј-Dithiobis(2-nitrobenzoic acid) (DTNB) reductase activity was measured by monitoring the production of thionitrobenzoate at 412 nm in a reaction mixture comprising 0.1 M potassium phosphate, pH 7.0, 5 mM EDTA, 0.2 mM NADPH, 5 mM DTNB, and 44 pmol of rTrxR. Specific activity for the reaction was calculated according to Holmgren (46) using the molar extinction coefficient at 412 nm as 2 ϫ 13,600 M Ϫ1 cm Ϫ1 as 1 mol of NADPH yields 2 mol of thionitrobenzoate. Cystine reductase activity of the TrxR/Trx couple was measured by monitoring the oxidation of NADPH at 340 nm in a reaction mixture comprising 0.1 M potassium phosphate, pH 7.0, 5 mM EDTA, 0.2 mM NADPH, 12.5 M rTrx, 44 pmol of rTrxR and varying concentrations of L-cystine (5-50 M). All assays were performed at 37°C. Kinetic calculations were performed using the computer program Grafit (Erithacus Software).
The Effects of Growth Conditions upon the Expression of TrxR, Trx, and TrxP in T. vaginalis-Parasites were grown in 25 ml of medium in tightly capped universal tubes with little gas phase, except for the "aerobic" cultures, which were grown in 500-ml tissue culture vessels with loose caps in a normally aspirated incubator. Under these aerobic conditions, growth over 18 h was reduced by 62% compared with control anaerobic cultures. The standard modified Diamond's medium was varied by removal of ascorbate (normally present at 5.7 mM) or addition of L-cysteine, L-serine, or L-methionine to 10 mM. None of these significantly affected growth over 18 h. Cultures were initiated at 10 5 parasites/ml, and incubation was for l8 h at 37°C, whereupon the parasites were harvested, washed, and stored as pellets at Ϫ70°C until analysis (17).

Isolation and Characterization of TRXR of T. vaginalis-The
identification of an EST from T. vaginalis with strong sequence similarity to TRXR simplified the isolation of the full-length cDNA. The initiation methionine was confirmed by using 5Ј RACE on mRNA. TRXR of T. vaginalis (accession number AJ507831) is predicted to encode a protein (TrxR) of 304 amino acids, with a subunit molecular mass of 32.4 kDa. Thus the enzyme is similar in size to low molecular weight TrxRs (designated L-TrxR) characteristic of bacteria, plants, and many lower eukaryotes. These proteins differ fundamentally from the high molecular weight TrxRs (designated H-TrxR) of humans and some other eukaryotes, including P. falciparum (29,47). Analysis of the predicted amino acid sequences of the T. vaginalis TrxR confirmed that it indeed belongs to the L-TrxR group and is distinct from both the H-TrxR group and the AhpFs of bacteria. An alignment of the TrxR of T. vaginalis with other L-TrxRs shows that it possesses the key active site cysteine residues (boxed in Fig. 1 TrxR, Trx, and TrxP of T. vaginalis Function as a Peroxiredoxin Cascade-Soluble recombinant TrxR, Trx, and TrxP of T. vaginalis (designated rTrxR, rTrx, and rTrxP) were produced in large quantities (ϳ30 mg/l) in E. coli with a C-terminal 6ϫ histidine tag. Denaturing SDS-PAGE analysis confirmed the predicted sizes of rTrx, rTrxR, and rTrxP (13.3, 33.7, and 23.0 kDa) and showed that they had a high degree of purity ( Fig. 2A).
rTrxR showed activity as a thiol reductase, whereby electrons were passed from NADPH to recombinant Trx (rTrx) with insulin as the ultimate electron acceptor (Table II). T. vaginalis TrxR also reduced the dithiol DTNB, albeit at a low rate (Table  II). This is in contrast to the E. coli enzyme, which lacks such activity (46). The apparent K m of rTrxR toward NADPH (in the DTNB assay) was 2.1 M and toward rTrx (in the insulin assay) was 2.0 M (Table II). In contrast, the K m toward rTrx of E. coli was Ͼ100 M. Using the TrxR/Trx couple, cystine could also act as ultimate acceptor (Table II), the rate increasing with the concentration of cystine up to 50 M. The reduction of both insulin and cystine by reduced Trx is non-enzymatic, and the linearity of the reduction rate up to 50 M cystine indicates that in this case the cystine concentration is limiting. Nevertheless, the finding that cystine is reduced by the thioredoxin system suggests that it may be responsible for maintaining cysteine (the major redox buffer of Trichomonas) in its reduced form. Neither insulin nor cystine stimulated NADPH oxidation in the absence of added rTrx. The addition of hydrogen peroxide, cumene hydroperoxide, or tert-butyl hydroperoxide to a mixture of NADPH, rTrxR, rTrx, and rTrxP resulted in a rapid oxidation of the NADPH (Fig. 2B). No activity toward the hydroperoxides was apparent if rTrx was omitted, which confirms that the T. vaginalis TrxR system is functionally as well as structurally different from the alkyl hydroperoxidases (Ahp) of many prokaryotes (48,49) and the enzyme of E. histolytica (32,33). Low activity toward hydrogen peroxide (4.6 Ϯ 1.2 mol/min/mg of protein) was detected when rTrxR and rTrx were used in the absence of TrxP. rTrxR had low NADPH oxidase activity, which was maximal at pH 9.0 (Table II). The activity at pH 7.0 was only 35% of that at pH 9.0.
TRX and TRXP Expression in T. vaginalis Is Modulated by Oxidative Stress-Northern blots with total and poly(A) ϩ RNA revealed a single transcript with probes specific for each of TRXR, TRX, and TRXP of T. vaginalis (not shown). The length of the mRNAs for the genes corresponded well with the size of the cDNA clones isolated, the position determined for the addition of the poly(A) ϩ tails, and the sites mapped for transcription initiation. Western blots using the antibodies raised against the recombinant proteins revealed that they all recog- nized the recombinant proteins themselves and those raised against rTrxR and rTrx detected just a single protein of the expected size, of about 32 kDa and 12 kDa, respectively, in a T. vaginalis-soluble fraction (Fig. 3). However, the anti-rTrxP serum detected two proteins (20 and 22 kDa) in the parasitesoluble fraction (Fig. 3). The 22-kDa protein corresponds to the expected size of the 2-Cys peroxiredoxin, whereas the second protein is possibly due to cross-reactivity with a second peroxiredoxin present in the parasites. This is not unusual as many organisms contain several peroxiredoxins that differ in their precise substrate specificity and often are located in different cellular compartments (24). A second peroxiredoxin gene was not found in the T. vaginalis EST dataset, but analysis of the genome sequence that is now available (www.tigr.org/tdb/e2k1/ tvg/) reveals other peroxiredoxins. No proteins were detected  when duplicate blots were probed with pre-immune rTrx, rTrxR, or rTrxP antisera (not shown).
T. vaginalis was cultured under conditions that could alter the oxidative stress to which the parasite was exposed and gene expression assessed via measuring mRNA levels by Northern blotting and protein levels by Western blotting. The conditions chosen were: (i) aerobic, exposure to greater oxygen concentrations through the use of shallow cultures with a large surface area/volume ratio; (ii) minus ascorbate, removal of the antioxidant ascorbate from the medium; (iii) plus cysteine, addition of an extra 10 mM cysteine as a redox buffer and precursor of hydrogen sulfide; (iv) plus serine, addition of an extra 10 mM serine as a precursor of cysteine; (v) plus methionine, addition of an extra 10 mM methionine as a precursor of methanethiol. The mRNA and protein levels of TrxR were not greatly affected (Fig. 3A), although there was a reduction of ϳ50% in the mRNA and protein levels when the medium was supplemented with cysteine (Fig. 3A, lane 4). In contrast, both TrxP and Trx mRNA and protein levels (especially of the smaller isoenzyme  of TrxP) were elevated when the parasites were grown either exposed to more oxygen or in medium lacking ascorbate (Fig. 3,  B and C, lanes 2 and 3). These conditions presumably increase the parasite's exposure to oxidative stress such that the two components of the thioredoxin system are required at elevated concentrations. The TrxP mRNA levels were reduced a little by addition of cysteine (Fig. 3B, lane 4) but not the other amino acids to the medium (not shown), and this is in agreement with the suggestion that cysteine acts as a major thiol redox-buffer in T. vaginalis. Interestingly, addition of cysteine had differential effects upon the two isoenzymes of TrxP, with the larger protein apparently being totally down-regulated, whereas there was a slightly increased expression of the smaller isoenzyme (Fig. 3B, lane 4). The level of mRNA and protein for Trx was slightly decreased when cysteine was added but significantly increased by addition of serine and methionine (Fig. 3C,  lanes 4 -6, respectively).
TrxR of T. vaginalis Is Highly Divergent-Phylogenetic analysis recovered the fungal, plant, and Entamoeba and Dictyostelium TrxR sequences as a strongly supported cluster, which also contained the TrxR sequences from the bacterial genus Chlamydia (Fig. 4A). The integrity of this cluster is supported by the presence of two indels, which were not part of the alignment used to make the tree, in all of these sequences. The TrxR sequences from the two diplomonad parasites G. lamblia and S. barkhanus, which lack both indels, were recovered with moderate bootstrap support at the base of this cluster. The precise position of the T. vaginalis TrxR was not strongly resolved by our analyses, although it also lacks the indels characteristic of the majority of eukaryotic sequences and clearly does not cluster with them (Fig. 4A). In most trees, the T. vaginalis sequence fell at the base of the proteobacterial clade (as in Fig. 4A). However, Shimodaira Hasegawa tests (41) of competing trees suggest that the data cannot significantly reject a relationship between the T. vaginalis and the G. lamblia/S. barkhanus sequences at the 95% level (0.24 versus a cut off level of 0.05), even though this relationship does not appear in the best tree (Fig. 4A).
TrxP of T. vaginalis Is Also Highly Divergent-Phylogenetic analysis of this short protein produced a tree where most relationships were only poorly supported (Fig. 4B). Most of the eukaryotic sequences, including the G. lamblia and S. barkhanus sequences, formed a weakly supported cluster. Notably, there was no support from unconstrained analyses to suggest that the T. vaginalis TrxP is a part of this cluster, and it tended to branch at the base of a poorly resolved radiation comprising intermingled eukaryotic and bacterial sequences (Fig. 4B). DISCUSSION The results show that T. vaginalis contains a thioredoxinlinked redox system that may be a principal means whereby these parasites cope with oxygen and its metabolites to which they are naturally exposed. The activity data presented for the recombinant TrxR, Trx, and TrxP provide strong evidence that the three proteins of T. vaginalis could function together in vivo to reduce hydroperoxides resulting from oxidative stress. The increased expression of both Trx and, in particular, TrxP when T. vaginalis was experimentally exposed to oxidative stress is fully consistent with a role in enabling the parasite to withstand such challenges. The finding that addition of cysteine to the growth medium resulted in somewhat lower levels of TrxR, Trx, and TrxP supports the idea that this amino acid itself contributes to the maintenance of an adequate reducing environment in the cell (16) and so acts synergistically with the Trx-linked system. Indeed the two systems may well be integrated, because rTrxR and rTrx could regenerate cysteine from cystine and so may play a role in maintaining the levels of this key redox buffer.
It is noteworthy that, not only were the levels of the Trx and TrxP proteins themselves increased in response to environmental changes, but so were the levels of mRNA encoding them. This provides additional evidence that the parasites are able to respond to exogenous stresses by adapting their transcriptional machinery (50).
It is also interesting that the Trx levels were enhanced when the medium was supplemented with serine or methionine, even though the level of TrxP remained unchanged. Serine was included in the experiments with T. vaginalis because it is a possible source of cysteine via serine acetyltransferase and cysteine synthase (30). T. vaginalis has been shown to contain the latter enzyme, the expression of which does vary according to the availability of exogenous sulfur amino acids and cysteine in particular. 4 If serine is used as a precursor for cysteine, sulfide needs to be provided. Sulfide could be provided by the action of methionine ␥-lyase on homocysteine (17,30), alternatively the 3Ј-phosphoadenosylsulfate reductase pathway may be the source as in bacteria and fungi (51). It has been shown that 3Ј-phosphoadenosylsulfate reductase in E. coli is dependent on thioredoxin (52), and if the same is true for T. vaginalis then it is conceivable that the increased Trx level when additional serine was included in the medium reflected an up-regulation of the cysteine de novo synthetic pathway from serine. Certainly, however, the elevated levels of Trx resulting from the addition of methionine and serine to the medium suggest that Trx has multiple functions in T. vaginalis and is not only involved in the reduction of TrxP. Notably, Trx has been implicated in many other cellular events in bacteria, yeast, and mammalian cells, including the reduction of methionine sulfoxide reductase and ribonucleotide reductase and the modulation of the activity of various transcription factors (52)(53)(54)(55)(56).
The TrxR of T. vaginalis clearly belongs to the L-TrxR group of enzymes, which differ fundamentally from those in mammals (H-TrxR) (29), and yet it is not enzymatically identical to the E. coli TrxR. Notably, the T. vaginalis TrxR transferred electrons to the parasite's own Trx with high specificity (Trx of E. coli being a very poor substitute) and reduced DTNB, a model substrate for H-TrxR, which is not used by E. coli TrxR (46).
Phylogenetic analyses of the eukaryotic L-TrxRs show that they arise from within the bacterial domain, suggesting that eukaryotes gained this type of TrxR through horizontal gene transfer from a bacterium (29). The eukaryotic L-TrxR enzymes are most closely related to those from proteobacteria, but there was no evidence from the tree that the nearest neighbors are the alpha-proteobacteria. Thus, there is no strong support for the eukaryotic enzyme originating with the mitochondrial endosymbiont, which is widely held to have been a member of the alpha-proteobacteria. The Trichomonas L-TrxR does not cluster with the other eukaryotes in the best trees we found (Fig. 4A), and its position outside of the main eukaryotic cluster is supported by the observation that it lacks the two indels that occur in most of the sampled eukaryotic L-TrxRs. However, its position is only weakly supported by the data, and thus there is no strong evidence that it gained its TrxR separately from other eukaryotes. Notably, the G. lamblia and S. barkhanus sequences also lack these indels, but they branch with moderate support at the base of the main eukaryotic cluster, suggesting that the HGT of L-TrxR may have occurred just once and that indels were acquired subsequent to the divergence of T. vaginalis and diplomonads. An interesting feature of the main eukaryotic cluster is that it also contains the three indel-containing TrxR sequences from the eubacterial genus Chlamydia. At face value, the tree topology suggests that the Chlamydia species acquired their genes for TrxR from a eukaryote.
TrxP of T. vaginalis is similar to the TrxR of the parasite in being quite divergent from the homologous proteins in other eukaryotes. TrxP clearly belongs to the 2-Cys class of TrxPs, but a phylogenetic analysis revealed that the tree for eukaryotes is not exclusive of eubacterial sequences. The tree is thus consistent with repeated horizontal gene transfers from prokaryotes to eukaryotes of genes for TrxP and T. vaginalis having acquired the gene for TrxP separately from the other eukaryotes sampled.
Horizontal gene transfer from prokaryotes to eukaryotes has received increasing attention lately (57)(58)(59), with the thioredoxin system being suggested as an example of a simple enzyme substrate system for which HGT from prokaryote to eukaryote is likely to have occurred (35). Not only are our trees consistent with this hypothesis, but in the case of T. vaginalis we have also shown experimentally that components with potentially different evolutionary origins can function together.
The results of this study suggest that the Trx-linked system provides a crucial means to detoxify ROS in T. vaginalis. It remains to be discovered whether or not the system is essential for the survival of T. vaginalis in its host, but the finding that S. cerevisiae requires either TrxR or glutathione reductase for survival (60) and that TrxR is essential for Drosophila, which lacks glutathione reductase (61), provide strong support for the idea that T. vaginalis, which lacks glutathione reductase, requires TrxR. If this is the case, then the TrxR of the parasite would represent a potential therapeutic target of some promise, because it differs fundamentally from the mammalian counterpart. Novel therapies against the parasite would certainly be most welcome, especially if parasite lines resistant against the 5-nitroimidazoles become more widespread (40).