The evolution of mitochondrial ADP and ATP exchanging proteins (AACs) highlights a key event in the evolution of the eukaryotic cell, as ATP exporting carriers were indispensable in establishing the role of mitochondria as ATP-generating cellular organelles. Hydrogenosomes, i.e. ATP- and hydrogen-generating organelles of certain anaerobic unicellular eukaryotes, are believed to have evolved from the same ancestral endosymbiont that gave rise to present day mitochondria. Notably, the hydrogenosomes of the parasitic anaerobic flagellate Trichomonas seemed to be deficient in mitochondrial-type AACs. Instead, HMP 31, a different member of the mitochondrial carrier family (MCF) with a hitherto unknown function, is abundant in the hydrogenosomal membranes of Trichomonas vaginalis. Here we show that the homologous HMP 31 of closely related Trichomonas gallinae specifically transports ADP and ATP with high efficiency, as do genuine mitochondrial AACs. However, phylogenetic analysis and its resistance against bongkrekic acid (BKA, an efficient inhibitor of mitochondrial-type AACs) identify HMP 31 as a member of the mitochondrial carrier family that is distinct from all mitochondrial and hydrogenosomal AACs studied so far. Thus, our data support the hypothesis that the various hydrogenosomes evolved repeatedly and independently.
Hydrogenosomes of anaerobic ciliates and chytridiomycnete fungi possess genuine mitochondrial-type AACs (Voncken, 2001; Vonken, et al., 2002; van der Giezen et al., 2002; Haferkamp et al., 2002). These AACs provide evidence for the hypothesis that the hydrogenosomes of anaerobic chytrids and ciliates are a kind of hydrogen-producing mitochondria that adapted to anaerobic environments (Akhmanova et al., 1998; Martin and Müller, 1998; Hackstein et al., 2001; Martin et al., 2001; Tielens et al., 2002). Surprisingly, all attempts to identify a mitochondrial-type AAC gene from the anaerobic, hydrogenosome-bearing flagellates Trichomonas vaginalis and Trichomonas gallinae so far have failed (Voncken et al., 2002). Accordingly, Dyall et al. (2000) convincingly showed that one of the most abundant membrane proteins of the hydrogenosomes of T. vaginalis, i.e. the hydrogenosomal membrane protein 31 (‘HMP 31′), is encoded by a gene that is clearly distinct from mitochondrial-type aac’s. Notwithstanding, this gene appeared to be a member of the mitochondrial carrier family (‘MCF′; Kuan and Saier, 1993; Palmieri et al., 2000). Therefore, Dyall et al. (2000) speculated as to whether that this protein could function as a hydrogenosomal ADP/ATP carrier, but van der Giezen et al. (2002) explicitly questioned such a function of HMP 31. Here we show that the homologous HMP 31 of the closely related T. gallinae (cf. Benchimol et al., 1997; Felleisen, 1997) specifically transports ADP and ATP – as do genuine mitochondrial-type AACs. However, we also demonstrate in this publication that HMP 31 from T. gallinae is biochemically distinct from all known mitochondrial-type AACs, and we analyse its phylogenetic position with respect to the mitochondrial-type hydrogenosomal AACs of anaerobic chytrids and ciliates.
Phylogenetic position of hmp 31 with respect to other hydrogenosomal AACs
Using PCR with primers directed against conserved regions of the hmp31 gene of T. vaginalis, we succeeded in cloning a homologous gene from T. gallinae (named tghmp 31). The deduced amino acid sequence of tghmp 31 is 91.8% identical to the corresponding protein from T. vaginalis (see Supplementary material for the alignments). Both hmp 31 and tghmp 31 are clearly members of the mitochondrial carrier family (MCF). Phylogenetic analysis using Mr Bayes confirmed the conclusion of Dyall et al. (2000) that HMP 31 does not belong to the cluster of genuine mitochondrial-type AACs (Fig. 1). Our analysis also revealed that HMP 31 is clearly different from the ADP/ATP carriers of the hydrogenosomes of anaerobic chytrids and ciliates, which cluster among the mitochondrial AACs of their aerobic relatives. Also, phylogenetic analysis using Neighbour-Joining (including all sequences with E < 1E-10 obtained by a blast search) and exact Maximum Likelihood generated the same result: hmp 31 and tghmp 31 do not branch among mitochondrial-type AACs, which form a monophyletic cluster that receives highest statistical support with all algorithms used (not shown). Consequently, hmp 31 could be either an orthologous, albeit rather divergent member of the mitochondrial AACs, or a paralogue, which might have evolved from a different subgroup of mitochondrial solute carriers. Members of this group encompass, for example, a peroxisomal transporter of Oryctolagus cuniculus, human Graves’ disease protein (GDC), LEU5 from the yeast S. cerevisiae, and BRITTLE-1 of maize, which transports ADP-glucose into plastids (see Experimental procedures for the accession numbers).
Expression of tghmp 31 in E. coli
The complete open reading frame encoding the TGHMP 31 protein has been cloned into the plasmid pET16b and expressed in the E. coli strain BL21 (DE3). After induction, a significant fraction of the transgenic TGHMP 31 was functionally incorporated into the cellular membrane of E. coli(Fig. 2) enabling the uptake of radioactive labelled adenine nucleotides. The kinetics of the uptake of [α32P]-labelled ADP and ATP into induced (living) E. coli cells showed that the recombinant protein transports ADP and ATP at similar rates as recombinant mitochondrial AACs (Fig. 3; see Haferkamp et al., 2002 for mitochondrial and hydrogenosomal AACs). Neither E. coli without a vector, nor E. coli carrying an empty (Fig. 3) or an uninduced vector, respectively, transport ADP or ATP (Fig. 3, cf. Krause et al., 1985; Möhlmann et al., 1998; Tjaden et al., 1998; Voncken et al., 2002).
Competition experiments with 20 different potential substrates reveal that TGHMP 31 has a high specificity for transporting ADP and ATP, just as mitochondrial AACs from A. thaliana(Table 1). Neither deoxynucleotides nor CoA can compete with the transport of radioactive labelled ADP. Moreover, the import of [α32P]ADP and [α32P]ATP by the recombinant TGHMP 31 into E. coli cells displayed typical Michaelis–Menten kinetics with apparent Km values of about 60 µM for ADP and about 135 µM for ATP (Table 2), very similar to the AAC1 and AAC2 isoforms from rat mitochondria (Haferkamp et al., 2002). The export of adenine nucleotides has been studied qualitatively using thin-layer chromatography of radioactive labelled adenine nucleotides, which were released from cells loaded with [α32P]ADP. IPTG-induced E. coli cells expressing TGHMP 31 release adenine nucleotides only in the presence of ADP or ATP, but not in the presence of Pi(Fig. 4). Therefore, a counter exchange mode for the transport of ADP and ATP has to be assumed that is characteristic for the mitochondrial AACs (Klingenberg, 1993).
Table 1. . Effects of various metabolites on [α32P]ADP uptake into E. coli cells expressing TGHMP 31 or mitochondrial AACs of Arabidopsis thaliana.
Rate of ADP transport
At AAC1 (%) of Control
At AAC2 (%) of Control
At AAC3 (%) of Control
(pmol · mg−1 protein · h−1)
(%) of Control
For the uptake experiments with TGHMP 31 effectors were given at a concentration of 250 µM. [α32P]-ADP was present at a concentration of 50 µM. Because of the different affinities of the plant AACs effectors were given at a concentration of 50 µM during uptake with At AAC 1–3. [α32P]-ADP was present at a concentration of 10 µM in A. t. experiments. n.m. = not measured. Data are the mean of three independent experiments. SE less than 9% of the mean values.
Table 2. . Km and Vmax values for ATP and ADP of TGHMP 31 determined on intact E. coli cells under different energy conditions (coupled and uncoupled).
The Km-values are determined with or without addition of the protonophore CCCP. It has been shown earlier that nucleotide uptake mediated by AAC isoforms from Rattus norvegicus and from Neocallimastix sp. L2 (hydrogenosomal AAC) are also influenced by CCCP due to its interference with the membrane potential (see Haferkamp et al., 2002). Km is given in [µM], Vmax is given in (nmol · mg−1 protein · h−1), E. coli cells were preincubated with 100 µM CCCP for 2 min for uncoupling. Data are the mean of three independent experiments.
63.5 ± 2.6
0.77 ± 0.26
ADP + CCCP
41.6 ± 0.5
0.12 ± 0.02
134.2 ± 6.9
0.41 ± 0.05
ATP + CCCP
47.2 ± 5.9
0.14 ± 0.02
Insensitivity of TGHMP 31 against bongkrekic acid (BKA)
All mitochondrial and hydrogenosomal AACs studied so far are highly sensitive against bongkrekic acid (‘BKA’; Winkler and Neuhaus, 1999; Haferkamp et al., 2002; Voncken et al., 2002). Notably, E. coli expressing transgenic TGHMP 31 was virtually insensitive against BKA (Table 3). In order to confirm this insensitivity in vitro, and to exclude the presence of BKA-sensitive mitochondrial-type AACs in the hydrogenosomes of T. gallinae, we studied the influence of BKA on the nucleotide transport activity of total hydrogenosomal membranes from T. gallinae, which were reconstituted in liposomes. Figure 5 shows that the transport of ADP and ATP by reconstituted hydrogenosomal membranes of T. gallinae was insensitive for BKA, whereas adenine nucleotide transport by liposomes containing mitochondrial membranes from Solanum tuberosum was highly sensitive. As expected, the reconstituted hydrogenosomal membranes of T. gallinae exhibited the same substrate specificity as recombinant TGHMP 31 expressed in E. coli (not shown; cf. Table 1).
Table 3. . Effect of bongkrekic acid and ADP on the uptake of [α32P]ADP into IPTG induced E. coli cells expressing tghmp 31 or aac2 (Arabidopsis thaliana).
Rate of ADP transport
(pmol · mg−1 protein · h−1)
(%) of Control
ADP uptake was measured at a concentration of 100 µM (tghmp 31) or 10 µM (aac2A. t). Bongkrekic acid was given at a concentration of 10 µM and preincubated for 5 min Unlabelled ADP (effector) was present at a concentration fivefold higher than the substrate. Uptake measurements were carried out for 14 min in the presence of lysozyme (2.5 mg ml−1), which is necessary to allow the penetration of BKA across the outer membrane of E. coli. Data are the mean of three independent experiments. SE is less than 8% of the mean values.
Our results indicate that HMP 31 of T. gallinae functions as a hydrogenosomal ADP/ATP carrier that is able to fulfil the same role as genuine mitochondrial AACs. In particular, we have shown that: (i) the ADP/ATP carrier activity is located in the hydrogenosomal membranes; (ii) that this ADP/ATP carrier is a highly active and (most likely) highly abundant membrane protein, just as its homologue HMP 31 in T. vaginalis (Dyall et al., 2000); (iii) that TGHMP 31 has the same substrate specificity and comparable kinetic properties as mitochondrial-type AACs from mitochondria and fungal hydrogenosomes (Haferkamp et al., 2002; Voncken et al., 2002), and (iv) that this ADP/ATP carrier is biochemically distinct from all mitochondrial and hydrogenosomal AACs studied so far because of its remarkable insensitivity to BKA. The latter observation, when using reconstituted hydrogenosomal membranes, clearly excludes the presence of significant amounts of BKA-sensitive mitochondrial-type AACs in the hydrogenosomes of T. gallinae (cf. Winkler and Neuhaus, 1999). Notably, the only known adenine nucleotide transporter, which is insensitive for BKA, is localized in the peroxisomes of the yeast Saccharomyces cerevisiae and it does not belong to the cluster of mitochondrial-type AACs (Palmieri et al., 2001; see below).
TGHMP 31 is quite distinct from mitochondrial-type AACs at the level of sequence similarity (some 25–30% identity, and 40–45% similarity). Phylogenetic analysis shows that HMP 31 does not belong to the cluster of mitochondrial-type AACs, which radiate from a single node. This node is highly supported by all types of phylogenetic reconstruction (Fig. 1; see also Dyall et al., 2000; Löytynoja and Milinkovitch, 2001; van der Giezen et al., 2002; Voncken et al., 2002). However, it is impossible to answer the question as to whether HMP 31 is a divergent, but orthologous member of the mitochondrial-type ADP/ATP carrier family or a paralogue, which evolved from a different member of the MCF by changing its substrate specificity. Thus, it is possible that Trichomonas once might have possessed a mitochondrial-type AAC that had been replaced by HMP31 in the course of its evolution into an anaerobic, mucus-dwelling parasite. However, it has to be stressed that the adaptation to anaerobic niches per se does not require the loss of mitochondrial type AACs (Voncken, 2001; Haferkamp et al., 2002; van der Giezen et al., 2002; Voncken et al., 2002) as revealed by the existence of hydrogenosomal, but unequivocally mitochondrial-type AACs of anaerobic ciliates and chytrids, which became inhabitants of anaerobic, gastro-intestinal environments in the course of their evolution (Fig. 1).
We realize that the phylogenetic position of the host, Trichomonas, remains hotly debated since the phylogenetic analysis of a number of proteins led to rather controversial results (e.g. Roger, 1999; Philippe et al., 2000; Keeling and Palmer, 2000; Henze et al., 2001; Stechmann and Cavalier-Smith, 2002; Emelyanov, 2003; Baldauf, 2003). Particularly, the deep branching of Trichomonas seen in some studies has been attributed to Long Branch Attraction (LBA) (Philippe, 2000; Philippe et al., 2000). Of course, we cannot exclude that LBA plays a role in the deep branching of HMP31 as well, but the distinctiveness of HMP31 from the mitochondrial AACs is not only based on the consistent results of all methods of phylogenetic reconstruction used, including Maximum Likelihood, Neighbour Joining, and Puzzle (not shown) or paup/Fitch-Margoliash (Dyall et al., 2000). Notably, HMP 31 is also clearly distinct from the AACs of the hydrogenosomes of anaerobic chytrids and ciliates (Fig. 1). Importantly, this phylogenetic oddity of HMP31 is also supported by experimental evidence: in contrast to the hydrogenosomal AACs of anaerobic chytrids (Haferkamp et al., 2002; Voncken et al., 2002), TGHMP31 is insensitive to BKA (Fig. 5; Table 3), a specific inhibitor for all mitochondrial-type AACs studied so far (Winkler and Neuhaus, 1999), and also studies with isolated hydrogenosomes from Tritrichomonas foetus have revealed a peculiar insensitivity against atractyloside, another specific inhibitor for mitochondrial-type AACs (Cerkasov et al., 1978).
Consequently, one might postulate that the hydrogenosomal ADP/ATP transporters of Trichomonas evolved before the radiation of the genuine mitochondrial-type AACs, which seems to parallel the advent of aerobic mitochondria. This distinctiveness of the hydrogenosomal ADP/ATP transporters of Trichomonas supports the hypothesis that all eukaryotes once possessed a proto-organelle that evolved into a hydrogen and ATP-generating organelle in anaerobic environments, while it evolved into a mitochondrion in aerobic niches (Martin and Müller, 1998). Moreover, the peculiar properties of HMP 31 clearly identify it as distinct from the hydrogenosomal AACs of anaerobic chytrids and ciliates (Haferkamp et al., 2002) corroborating the evidence in favour of multiple, independent origins of hydrogenosomes (Hackstein et al., 2001; Voncken et al., 2002)
Trichomonas gallinae was cultured as described earlier (Voncken et al., 2002). Hydrogenosomes were isolated by differential centrifugation as described by Drmota et al. (1996). Mitochondria from potato tubers were isolated with the aid of isopycnic Percoll gradients (Haferkamp et al., 2002). Reconstitution of mitochondrial and hydrogenosomal membrane proteins in proteoliposomes and uptake experiments were performed as described by Möhlmann et al. (1997) and Heimpel et al. (2001). ADP/ATP uptake and back exchange studies were performed as described earlier (Haferkamp et al., 2002; Voncken et al., 2002). To investigate to which degree incorporated nucleotides are metabolized by the E. coli cells, the cells were disrupted after preloading with ADP. Thin layer chromatography revealed that ATP and ADP are present at a ratio of about 0.5. SDS-PAGE in 15% polyacrylamide gels was performed according to Laemmli (1970). Immunodetection with an anti-his antibody was carried out as described in the Qiagen protocol.
DNA manipulations were performed essentially as described in Sambrook et al. (1989). tghmp 31 was isolated by PCR from first-strand cDNA of T. gallinae and inserted into pGEM-T-easy (Promega, Mannheim). The primers used were constructed on the basis of the sequence of HMP 31 of T. vaginalis (Dyall et al., 2000). The gene encoding TGHMP 31 was inserted into bacterial expression vector pET16b (Novagen, Heidelberg, Germany), and introduced into the E. coli strain BL21 (DE3). AACs from A. thaliana were cloned into the same vector and studied after induction in E. coli as described earlier (Haferkamp et al., 2002). Escherichia coli expressing TGHMP 31 were fractionated into inclusion body, cytoplasmic membrane and soluble fraction according to Sambrook et al. (1989).
An alignment of representative sequences from the MCF, including all the hydrogenosomal ones, was constructed with T-coffee (Notredame et al., 2000), and the most reliably aligned regions (145 positions) were selected with the program Gblocks (Castresana, 2000). Phylogenies were subsequently derived using either Maximum Likelihood as calculated by the program mrbayes (Huelsenbeck and Ronquist, 2001) or Neighbour Joining (Saitou and Nei, 1987). The JTT model (Jones et al., 1992) of amino-acid displacement with 4 gamma-distributed rates was used as a model for protein sequence evolution. The tree displayed in Fig. 1 was derived by running the Mr Bayes program for 500 k iterations and subsequently calculating the branch-lengths of the consensus tree with puzzle (Strimmer and von Haeseler, 1996). The support values for the partitions within the tree were calculated from the frequency of that branch in the Bayesian Monte Carlo simulation (leftmost) and a bootstrap of the Bayesian analysis in which for 100 of the 50% samples of the alignment the most dominant tree was selected (middle) from 30 k iterations. These 100 samples were also compared to generate a bootstrap (rightmost) for trees calculated with Neighbour Joining algorithm, using puzzle (Strimmer and von Haeseler, 1996) with gamma-distributed rates. A phylogenetic Neighbour Joining analysis (Saitou and Nei, 1987) using all sequences with E < 1E-10 obtained by a blast search was performed in parallel. This analysis produced similar results with respect to the relative phylogenetic positions of the mitochondrial AACs and (TG)HMP 31, i.e. outside of the veritable mitochondrial ATP/ADP translocases, but at a similar distance to these as the Brittle, the peroxisomal carrier and the Graves disease protein clusters. The sequences used for the MrBayes analysis are indicated below together with their identifiers:
The sequence of tghmp 31 has been deposited in the genebank under AF503503.
We thank Susanne van Weelden and Marion Schmitz for culturing T. gallinae and for the preparation of the hydrogenosomal fractions, and Markus Wahl for the help with the reconstitution experiments. The help of Alan Schwartz with the English phrasing is gratefully acknowledged. Work in the laboratory of J.T. was sponsored by the Deutsche Forschungsgemeinschaft (grants TJ 5/1–1, TJ 5/1–2).