Trichomonas vaginalis Hmp35, a putative pore-forming hydrogenosomal membrane protein, can form a complex in yeast mitochondria.

An abundant integral membrane protein, Hmp35, has been isolated from hydrogenosomes of Trichomonas vaginalis. This protein has no known homologue and exists as a stable 300-kDa complex, termed HMP35, in membranes of the hydrogenosome. By using blue native gel electrophoresis, we found the HMP35 complex to be stable in 2 m NaCl and up to 5 m urea. The endogenous Hmp35 protein was largely protease-resistant. The protein has a predominantly beta-sheet structure and predicted transmembrane domains that may form a pore. Interestingly, the protein has a high number of cysteine residues, some of which are arranged in motifs that resemble the RING finger, suggesting that they could be coordinating zinc or another divalent cation. Our data show that Hmp35 forms one intramolecular but no intermolecular disulfide bonds. We have isolated the HMP35 complex by expressing a His-tagged Hmp35 protein in vivo followed by purification with nickel-agarose beads. The purified 300-kDa complex consists of mostly Hmp35 with lesser amounts of 12-, 25-27-, and 32-kDa proteins. The stoichiometry of proteins in the complex indicates that Hmp35 exists as an oligomer. Hmp35 can be targeted heterologously into yeast mitochondria, despite the lack of homology with any yeast protein, demonstrating the compatibility of mitochondrial and hydrogenosomal protein translocation machineries.

Trichomonas vaginalis is a deep-branching protist that lacks archetypal eukaryotic "aerobic" organelles, specifically mitochondria and peroxisomes. This microaerophilic human-infective parasite carries out fermentative carbohydrate metabolism within hydrogenosomes. Hydrogenosomes are bounded by double membranes and produce ATP by substrate level phosphorylation (1). Hydrogenosomes are also found in certain chytrids, ciliates, and fungi, in lineages that are phylogenetically distant to the Parabasalian lineage to which trichomonads belong (1)(2)(3)(4). Currently, several lines of evidence support a common endosymbiotic ancestry for hydrogenosomes and mitochondria, despite their distinct metabolic pathways (3,5,6). Although the origin of hydrogenosomes within ciliate and fungi lineages is debated, these lineages branch with mitochondria-containing groups and the hydrogenosomes confined therein exhibit strong similarity to ciliate and fungal mitochondria (7,8).
Trichomonad hydrogenosomes, on the other hand, are markedly less similar to mitochondria. These organelles lack a genome (9) which would have provided a means to investigate their endosymbiotic origin as has been elegantly and convincingly done for mitochondria (10). In lieu of this, we and others have attempted to define the relationship between trichomonad hydrogenosomes and mitochondria by examining the origin of their chaperonins, metabolic enzymes, and membrane proteins. Chaperonin genes, specifically heat shock protein (Hsp) 1 70, cpn60, and Hsp10 (11)(12)(13)(14), and the IscS enzyme, involved in FeS cluster formation (15) appear to have a mitochondrial origin. However, analyses of metabolic enzymes such as hydrogenase (16,17), which is typically found in anaerobic bacteria, present a far more ambiguous picture. Phylogenetic analyses of another metabolic enzyme, pyruvate:ferredoxin oxidoreductase, revealed a clustering of all eukaryotic sequences known to date but failed to reveal a relationship to any particular eubacterial group (6). Only one trichomonad hydrogenosomal membrane protein, Hmp31, has been analyzed to date, and it was shown to be a distant homologue of the ADP/ATP carrier, a member of the mitochondrial carrier family (18).
We have continued to study hydrogenosomal membrane proteins to gain further information on the nature of the endosymbiont that gave rise to the trichomonad hydrogenosome, and to further examine the possibility that the same endosymbiont gave rise to both mitochondrion and trichomonad hydrogenosomes. Organellar membrane proteins may originate from preexisting proteins from the endosymbiont that evolve to fulfill new functions in the emerging proto-organelle. For instance, the Toc75 protein pore translocase in chloroplasts has a homologue of unknown function in cyanobacteria that shows similar pore characteristics (19). Alternatively, "new" protein families may have emerged such as the mitochondrial carrier family proteins that translocate ATP and various solutes across the organellar membrane in mitochondria (20). Likewise, most of the protein translocases in mitochondria have no distinguish-able homologues in prokaryotes and appear to have evolved with organelles (21).
Here we describe a novel hydrogenosomal membrane protein, Hmp35, that has no known homologue in prokaryotes or eukaryotes to date. It is a relatively abundant membrane protein that exists as an integral membrane complex. The exceptional stability of the complex under stringent salt, urea, and protease treatment is similar to that of bacterial and organellar outer membrane protein pores. Although it has no primary sequence homology to those proteins, Hmp35 has a predominant predicted ␤-sheet structure similar to that found in the outer membrane proteins of Gram-negative bacteria, mitochondria, and plastids (22). Taken together, these data suggest a similar function for Hmp35. Finally, we demonstrate that despite the absence of a homologue in yeast or of a recognizable mitochondrial targeting sequence, the Hmp35 can be heterologously expressed in yeast and is targeted to yeast mitochondrial membranes in vivo.

EXPERIMENTAL PROCEDURES
Organisms and Strains-T. vaginalis T1 (gift from J. F. Alderete) and C1 (ATCC 30001) strains were used where indicated in this study. Saccharomyces cerevisiae MB2-22 wild-type strain (23) was used throughout in this study.
Sequencing of MP40 Proteins-Hydrogenosomal membrane proteins were size-separated by 15% SDS-PAGE and stained with Coomassie Blue dye R-250, and a gel slice containing 5 g of the 35-40-kDa proteins was excised, washed with 50% acetonitrile, and subjected to tryptic digest and peptide sequencing at the Harvard University Microchemistry Facility (Cambridge, MA).
Generation of Antisera against Endogenous Membrane Proteins and Recombinant Hmp35-Polyclonal antisera against 35-40-kDa hydrogenosomal membrane proteins (MP40) were raised in rabbits by 4 injections of 100 g each. Polyclonal rabbit anti-Hmp35 antisera were raised against purified recombinant His-tagged Hmp35 protein.
Screening of cDNA Library with anti-MP40 Antisera-10 5 phage from a previously described (18) T. vaginalis ZAP II unidirectional cDNA library were plated for induced expression screening as per manufacturer's instructions (Stratagene). Briefly, XL1-MRFЈ bacteria infected with 10 5 phage were grown in top agar at 42°C until plaques started appearing. Protein expression was induced by placing nitrocellulose membranes saturated with 10 mM isopropyl-1-thio-␤-D-galactopyranoside on the plates and further incubating at 37°C for 3 h. Duplicate membranes were applied for a further 5 h at 37°C. Nitrocellulose membranes with induced proteins were screened with anti-MP40 antisera, and the bound antibodies were detected by 125 I-labeled protein A. Positive phage were excised and sequenced to select those that matched the peptides obtained from sequencing endogenous MP40 proteins. A positive cDNA clone, CD40.31, with a 1.1-kb insert was selected for further analyses.
Construction and Screening of an EcoRI T. vaginalis C1 Genomic Library-A genomic library was constructed in the ZAP II vector (Stratagene) from T. vaginalis C1 genomic DNA completely digested with EcoRI. The library was screened with a 640-bp EcoRI/HindIII probe from the cDNA clone CD40.31, yielding a positive clone, MP40.1, with a 2-kb EcoRI fragment bearing the complete hmp35 open reading frame.
Plasmid Construction-A 969-bp fragment was generated by PCR from the genomic clone MP40.1 using the primers MP40F and MP40R (Table I) to introduce a 5Ј BamHI and a 3Ј SalI restriction enzyme site, respectively, for subsequent ligation into the vector pQE30 (Qiagen) to yield the expression construct pEP40.17 with an in-frame hexahistidine tag situated at the N terminus of the hmp35 open reading frame.
For construction of the yeast transformation plasmid pRS313-hmp35, the genomic DNA clone MP40.1 was amplified with the PCR primers MP40B and MP40S (Table I) to generate the open reading frame of hmp35 with a 5Ј BamHI and a 3Ј SalI site, respectively, to allow cloning into the pRS313 vector.
For generation of the pHmp35H plasmid, we used the previously described (18) plasmid construct pHmp31-(HA) 2 which carries a neomycin phosphotransferase (neo) cassette that allows selection of transformants in T. vaginalis (25). The hmp35 open reading frame was amplified from the genomic clone MP40.1 using the forward primer Nhmp35H (Table I) to introduce an NdeI restriction site at the 5Ј end of the PCR product and the reverse primer BAhmp35H (Table I) to introduce a hexahistidine codon at the 3Ј end, followed by a BamHI restriction site. Following restriction digest with NdeI and BamHI, the purified PCR product was introduced into the corresponding sites in the restricted pHmp31-(HA) 2 plasmid.
Modification of Cysteine Residues with AMS-200 g of isolated hydrogenosomes were initially boiled for 5 min in SM, 50 mM NaCl, 0.5% SDS and cooled to 37°C. Controls with His-tagged Hmp35 recombinant protein (3 g per assay) were processed in parallel. Each sample was then incubated for 1 h at 37°C with or without 5 mM EDTA or 10 mM Tris [2-carboxyethyl] phosphine hydrochloride (TCEP), or 5 mM H 2 O 2 , or a mixture of these reagents as indicated. All samples were precipitated with 10% trichloroacetic acid, resuspended in 100 l of alkylating solution (100 mM iodoacetamide, 100 mM Tris, 100 mM Tris, 10 mM EDTA, pH 9.5), and incubated for 5 min at 37°C. The reaction was stopped by trichloroacetic acid precipitation, and all samples were resuspended in 50 l of 10 mM TCEP, 100 mM Tris, 0.5% SDS, 10 mM EDTA, pH 9.5, and incubated for 1 h at 44°C. Some samples were further treated with 4-acetamide-4-maleimidylstilbene-2,2-disulfonic acid (AMS) at a final concentration of 25 mM, and all samples were incubated for 90 min at 25°C. Finally, all samples were trichloroacetic acid-precipitated and resuspended in reducing Laemmli sample buffer for 12% SDS-PAGE separation and Western analysis.
Blue Native Gel Electrophoresis of Organelles-Hydrogenosomes were solubilized at a protein concentration of 1 mg/ml for 30 min on ice in n-dodecyl maltoside or Triton X-100 at indicated concentrations in the presence of 20 mM MOPS, 0.2 M or 0.5 M NaCl, 1 mM MgCl 2 , 10% glycerol, 2 mM phenylmethylsulfonyl fluoride (PMSF), pH 8.0. Sodium carbonate-extracted hydrogenosomal membranes were solubilized at a protein concentration of 0.1 mg/ml for 30 min on ice in 0.5% n-dodecyl maltoside or 0.5% Triton X-100 in the presence of 20 mM MOPS, 0.5 M NaCl, 1 mM MgCl 2 , 10% glycerol, 2 mM PMSF, pH 8.0. Insoluble material was removed by centrifugation at 100,000 ϫ g for 15 min at 4°C.

Cysteine-rich Hydrogenosomal Membrane Protein Hmp35
Denatured samples were generated by heating hydrogenosomes (1 mg/ml protein concentration) at 95°C for 5 min in 0.5% SDS, 20 mM MOPS, 1 mM MgCl 2 , 10% glycerol, 2 mM PMSF, pH 8.0. The solubilized proteins were analyzed by blue native electrophoresis on a 6 -16% linear polyacrylamide gradient (26). Solubilization of mitochondria (2.5 mg/ml protein concentration) was performed in 0.5% digitonin as described previously (27). Protease Treatment of Hydrogenosomes-Hydrogenosomes (200 g of total protein) were incubated in SM buffer for 30 min at 37°C in the presence of TCEP and/or 10 mM EDTA as indicated and treated with 0.25 mg/ml proteinase K (Roche Applied Science) for 30 min at 0°C. The digestion was inhibited with 2 mM phenylmethylsulfonyl fluoride (PMSF) for 10 min on ice. Samples were trichloroacetic acid-precipitated and resuspended in Laemmli sample buffer for SDS-PAGE. A similar protease treatment was performed on hydrogenosomes initially solubilized at 1 mg/ml protein concentration in 1% Triton X-100, 0.1 M NaCl, 20 mM MOPS, 10% glycerol, 2.5 mM MgCl 2 , pH 8.0, and samples were processed for SDS-PAGE. Sodium carbonate-extracted hydrogenosomal membranes were resuspended at 0.1 mg/ml protein concentration in SM and treated with 0.1 mg/ml or 0.25 mg/ml proteinase K for 20 min on ice. Following inhibition with 2 mM PMSF, the membranes were recuperated by centrifugation at 100,000 ϫ g and resuspended in SM for trichloroacetic acid precipitation before resuspension in Laemmli sample buffer for SDS-PAGE.
Transformation of Yeast-The S. cerevisiae wild-type strain MB2-22 was transformed with the plasmid pRS313-hmp35 under the manipulation of the histidine marker. Standard genetic techniques were used for growth and for transformation of yeast strains (28).
Crude Fractionation of T. vaginalis 35H Transformants-Cells from cultures of the 35H transformants were broken in a cell disruptor (Energy Service Co.) and fractionated by centrifugation at 12,000 ϫ g into a crude cytosolic fraction and a crude organellar fraction as described previously (18).
Crude Fractionation of Yeast Cells and Isolation of Mitochondria-Clones from pRS313-hmp35-transformed yeast cells (h35) and from the wild-type MB2-22 strain (wt) were grown in the presence of lactate. The harvested cells were converted to spheroplasts by incubation in 0.6 M sorbitol, 20 mM HEPES-KOH, pH 7.4, supplemented with zymolase-20T (Seikagaku Corp.), at a concentration of 3 mg/g cells, and subjected to a cell disruptor to break the cellular wall. After an initial 1000 ϫ g spin to discard cellular debris, a crude organellar fraction was separated from the cytosolic supernatant fraction by pelleting the sample at 10,000 ϫ g. For further analyses, highly purified mitochondria were obtained by the Nycodenz gradient procedure (29).
Subfractionation of Purified Mitochondria-250 g of total protein of mitochondria were used for each of the following sub-localization treatments. Intact mitochondria were treated with 0.1 mg/ml proteinase K for 30 min at 4°C. Mitochondria were osmotically shocked at 250 g/ml in 20 mM HEPES-KOH, pH 7.4, for 30 min at 4°C to generate mitoplasts in the presence or absence of 0.1 mg/ml proteinase K. Following the protease or mock treatment, samples were treated with 2 mM PMSF for 10 min on ice to inhibit proteinase K and spun at 16,000 ϫ g to separate the supernatant from the pellet. Both fractions were trichloroacetic acid-precipitated and resuspended in equal volumes of Laemmli sample buffer. Mitochondria were separated into integral membrane protein and soluble fractions by incubation for 30 min on ice at 250 g/ml in 0.1 M Na 2 CO 3 , pH 11.5, followed by ultracentrifugation at 100,000 ϫ g. Both fractions were resuspended in equal volumes of Laemmli sample buffer, and proteins were separated by 10% Tris-Tricine SDS-PAGE.
Miscellaneous-Following separation by PAGE, proteins were electroblotted onto PVDF membranes for Western analysis. Hexahistidine tags were detected with anti-His tag mouse monoclonal antibody. Detection of bound primary antibodies was carried out by using horseradish peroxidase-conjugated antibodies raised against mouse or rabbit IgG (The Jackson Laboratories), followed by enhanced chemiluminescence with the ECLϩ detection system (Amersham Biosciences). DNA and protein sequences were assembled and analyzed using the MacVector program (Accelrys).

Isolation and Characterization of Hmp35, a Hydrogenosomal
Membrane Protein-To pursue the identification of hydrogenosomal membrane proteins from T. vaginalis, we targeted a number of proteins in the 35-40-kDa range (Fig. 1A, lane 3, MP40) that are of intermediate abundance. The MP40 proteins were microsequenced on a small scale and isolated on a larger scale to raise rabbit polyclonal antibodies, anti-MP40. A positive cDNA clone, CD40.31, was isolated by screening an expression T. vaginalis cDNA library. Four of the peptide sequences obtained from the MP40 proteins exactly matched stretches in the translated CD40.31 sequence, confirming that a gene encoding one of the MP40 proteins had been cloned ( Fig. 2A). This gene was named hmp35, for hydrogenosomal membrane protein 35. From Northern and Southern analyses performed with the CD40.31 insert (data not shown), the hmp35 gene appeared to be unique. A probe from the CD40.31 clone was used to screen a T. vaginalis genomic DNA library to isolate a clone, MP40.1, that bore the complete hmp35 gene sequence. The gene is flanked by a double initiator sequence 18 bases up- stream of the initial methionine residue of the predicted protein, which is within the expected range for T. vaginalis genes (30). The conceptual translation of the hmp35 gene predicts a protein of 34.6 kDa consisting of 318 amino acids ( Fig. 2A).
To investigate the properties of the Hmp35 protein, we raised rabbit polyclonal antibodies (anti-Hmp35) against purified recombinant His-tagged Hmp35. Western analysis of sodium carbonate-extracted hydrogenosomes with these antibodies showed that the Hmp35 protein is found in the membrane pellet fraction and not in the soluble fraction, thus confirming that Hmp35 is present in the hydrogenosome exclusively as an integral membrane protein (Fig. 1B, lanes 2 and 3). The protein, with a calculated mass of 34.6 kDa, migrated at an apparent molecular mass of 39 kDa on SDS-PAGE (Fig. 1B).
Analysis of the Hmp35 Protein Sequence-The deduced amino acid composition of Hmp35 is shown in Fig. 2A. The predicted N-terminal sequence of Hmp35 bears no homology to the cleavable presequence (MAQPAEQILIAT) that had been found previously (18) in the only other characterized hydrogenosomal membrane protein Hmp31. The Hmp35 protein has an unusually high percentage of charged amino acids (24%), with a correspondingly high proportion of lysine residues (11%), resulting in a basic hydrophilic protein (Fig. 2B, top panel) with a high isoelectric point (pI) of 9.4. In comparison, a yeast mitochondrial membrane protein, Tom40, has only 17% charged residues and has a pI of 5.2. However, the Hmp35 protein has discrete areas of hydrophobicity, which are predicted by the von Heijne transmembrane algorithm to form 4 or 5 potential transmembrane domains (Fig. 2B, both panels). No transmembrane helices were predicted from the TMPRED program. Secondary structure computation using the PHD program predicted the protein structural composition as 47% ␤-sheet, 3% ␣-helical, and the rest mixed, suggesting that the protein has predominant ␤-sheet structure. Interestingly, Hmp35 has 14 cysteine residues, representing 4.4% of the total amino acid content, with nine of them clustered at the Cterminal third of the protein ( Fig. 2A). Seven of these residues are found in close proximity to each other in the 221-265 amino acid region, in the conformation CX 6 CCX 2 CX 9 HX 15 CCXHX 2 C, where C is a cysteine residue, H is a histidine residue, and X is any residue.
Hmp35 Is a Novel Protein-Data base searches performed using the Hmp35 protein sequence with the BLASTp and tBLASTn algorithms revealed no significant similarity to known proteins. However, by using the WU-BLAST 2.0 program (31), we found that the 154 -318 region of Hmp35 had 46% similarity to a 167-amino acid region in a 35-kDa basic membrane lipoprotein, BmpC, from Borrelia burgdorferi, a bacterium of the spirochaete family. This protein is part of a family of outer membrane lipoproteins found in various bacteria.
To find if the residues in the cysteine-rich region in Hmp35 were part of a known motif, we searched the data base using the BLAST program with the short nearly exact matches algorithm using only the 221-265-amino acid region. The subregion from 221 to 231 ( Fig. 2A) had 83% similarity, with all cysteine residues conserved, to a zinc binding region from the anaphasepromoting complex subunit 11 protein found in a number of eukaryotes. This short 84 -140-amino acid cysteine-rich protein has a RING finger motif (32). The RING (for really interesting new gene) finger is a conserved cysteine-rich domain of 40 -60 residues that binds two zinc atoms in a "cross-brace" manner (33). Two variants of this motif have been described, the C3HC4 and the C3H2C3 types, neither of which exactly match the Hmp35 cysteine-rich region, as computed by the ExPASy PROSITE pattern and profile searches. However, the number and distribution of cysteine and histidine residues in the 221-265 region of Hmp35 are reminiscent of those patterns, and this region may represent a related but distinct motif that binds zinc.
Analysis of Cys Residues in Hmp35-To determine whether the cysteine residues in endogenous Hmp35 are engaged in intermolecular disulfide bonding, we analyzed the migration of Hmp35 by SDS-PAGE under non-reducing conditions. The test hydrogenosomal sample was treated with the alkylating agent N-ethylmaleimide before separation on SDS-PAGE to prevent artifactual oxidation of free cysteines during handling. We found that the migration of the Hmp35 protein was the same in the absence or presence of the reducing agent dithiothreitol (DTT), suggesting that the endogenous protein does not form intermolecular disulfide bridges with itself or with other proteins (Fig. 3A). As a control, Western analysis was performed using antisera against ferredoxin, a protein with three cysteine residues that coordinate iron (34).
Hmp35 Forms an Intramolecular Disulfide Bridge-To investigate whether any of the cysteine residues in Hmp35 formed intramolecular disulfide bridges, we used the cysteinemodifying reagent AMS. This reagent adds 500 daltons in molecular mass to the protein per modified cysteine. We adapted the iodoacetamide/AMS trapping approach used for determination of the number of disulfide bridges in soluble proteins (35,36). The hydrogenosomes were treated with a reducing (TCEP) or oxidizing agent (H 2 O 2 ) or left untreated, and then free thiol groups were blocked with the alkylating agents N-ethylmaleimide or iodoacetamide. Finally, after a reduction step to break any induced or endogenous disulfide bridges, the proteins were treated with AMS to modify any resulting free cysteines. By calculating the difference in migration of the protein as determined by Western analysis after this blocking treatment, the number of AMS moieties added can be computed and hence the number of disulfide bridges determined. Because Hmp35 is an integral membrane protein, we performed the first step of denaturing hydrogenosomes with SDS to ensure full accessibility of Hmp35 to the various reagents.
To test whether the alkylation and AMS procedures were efficient, we used reduced or oxidized recombinant His 6 -Hmp35 under denaturing conditions. When the recombinant protein was initially reduced, then alkylated with iodoacetamide before further reduction and treatment with AMS, we found that the protein migrated at the same position as the unmodified protein at 40 kDa (Fig. 3B, lanes 1, 2, and 4), indicating that no free cysteines were available for modification by AMS. When the recombinant protein was initially oxidized prior to the alkylation, reduction and AMS treatment, the relative migration of the protein increased by 7 kDa (Fig. 3B,  lanes 3 and 5), suggesting that 14 AMS molecules had been added. This result implies that all the cysteine residues in the recombinant protein had been efficiently modified by AMS, and hence have potential for inter/intramolecular disulfide bond formation (Fig. 3B, lanes 4 and 5). The presence of EDTA in either case did not affect the number of AMS molecules added, as might be expected because the recombinant protein had initially been completely denatured in 8 M urea, and no divalent cations were added to this in vitro assay.
We found that in SDS-denatured hydrogenosomal protein, the AMS treatment increased the apparent molecular mass of the endogenous Hmp35 protein from 39 to 40 kDa (Fig. 3B,  lanes 6 and 7). This 1-kDa increase would correspond to the addition of two AMS moieties to two cysteine residues. This implies that one disulfide bridge is present, most likely intramolecularly, in Hmp35. When the proteins had been reduced prior to the trapping, the migration was unaffected as the endogenous disulfide bridge was reduced and the iodoacetamide reagent efficiently alkylated all the cysteine residues, blocking any reaction with AMS (Fig. 3B, lane 8). The initial oxidizing step did not induce any further disulfide bonds (Fig.  3B, lane 9), unlike in the case of the oxidized recombinant (Fig.  3B, lanes 4 and 5) where all the cysteine residues had been oxidized. This suggests that the endogenous protein is in a form or structure that is incapable of forming extra disulfide bonds even when denatured in SDS. However, when EDTA was included during the oxidizing treatment, the migration of the protein increased by a further 1 kDa (Fig. 3B, lanes 9 and 13). This suggests that a divalent cation may be bound by the endogenous Hmp35 protein, which prevents the induction of further disulfide bonds, unless it is chelated by EDTA during oxidation.
Hmp35 Forms Part of a High Molecular Weight Complex in Hydrogenosomal Membranes-To assess whether the Hmp35 protein is in a stable protein complex in hydrogenosomal membranes, we adapted the technique of BN-PAGE that has been widely used for analyses of protein complexes in mitochondria. Isolated hydrogenosomes were solubilized in two non-ionic, non-denaturing detergents, n-dodecyl maltoside and Triton X-100, and the cleared lysates were subjected to BN-PAGE. Western analysis performed with anti-Hmp35 showed that the Hmp35 protein migrates mostly within a discrete band of ϳ300 kDa in both detergents (Fig. 4A), suggesting that it forms a stable complex in the hydrogenosome. The slight difference in migration between the two detergents used could be due to different micelle sizes.
To investigate whether this complex is held together by ionic interactions, we tested the effect of sodium chloride on its  1 and 2) for 15 min at room temperature and resuspended in Laemmli sample buffer with or without 50 mM dithiothreitol (DTT) at a final protein concentration of 1 g/l. As a control for artifactual oxidation, hydrogenosomes were mock-treated with ethanol (lanes 3 and 4) prior to resuspension at a final protein concentration of 1 g/l in Laemmli sample buffer with or without 50 mM DTT. 20 g of protein was size-separated by 8 -20% SDS-PAGE linear gradient, blotted onto PVDF, and immunodecorated with anti-Hmp35 or anti-ferredoxin (Fd) polyclonal antisera. Molecular mass markers are indicated in kDa. B, modification of cysteine residues with AMS. Isolated hydrogenosomes (200 g of total protein) were denatured by boiling in 0.5% SDS and incubated for 1 h at 37°C with or without 5 mM EDTA (chelating agent), or 10 mM TCEP (reducing agent), or 5 mM H 2 O 2 (oxidizing agent), or a mixture of these reagents as indicated (lanes 7-13). As controls, 3 g of purified recombinant His 6 -Hmp35 protein were subjected to the same reducing or oxidizing treatment under denaturing conditions prior to AMS modification (lanes [2][3][4][5]. After incubation at 37°C, all samples were alkylated to block any free cysteines and then reduced. All samples except for those in lanes 1, 6, and 10 were treated with 25 mM AMS. The protein samples (40 g per lane) were size-separated by 12% SDS-PAGE, transferred to PVDF, and immunodecorated with anti-Hmp35 antisera. Protein sizes were estimated using a calibration curve plotted according to the migration of known standards.
stability. The complex migrated at 300 kDa at all concentrations tested up to 2 M (Fig. 4B, lanes 1-6), without any monomer detected unless the hydrogenosomes were boiled in SDS prior to loading on the gel (Fig. 4B, lane 7). This resistance of the complex to salt shows that ionic interactions are not involved in maintaining its components together.
To determine what part of this complex consists of integral membrane proteins, we solubilized the membrane fraction of sodium carbonate-extracted hydrogenosomes and separated the stable complexes by BN-PAGE. Western analysis with anti-Hmp35 antibodies revealed that Hmp35 still migrated at 300 kDa in n-dodecyl maltoside and Triton X-100 (Fig. 4C). These  1 and 2) and 0.5 or 1% Triton X-100 (tx100, lanes 3 and 4). B, the HMP35 complex is resistant to high sodium chloride concentrations. Purified hydrogenosomes (50 g of protein per lane) were solubilized with 0.5% n-dodecyl maltoside in the presence of up to 2 M NaCl (lanes 1-6). A denatured sample was included on the gel to determine the position of migration of the Hmp35 monomer (lane  1-10). Half of each sample was treated with 0.25 mg/ml proteinase K (lanes 2, 4, 6, 8, and 10) for 30 min on ice, whereas the remaining half was mock-treated (lanes 1, 3, 5, 7, and 9). Two hydrogenosomal samples were denatured at 95°C in 0.5% SDS with or without 10 mM EDTA and split into 2 aliquots prior to addition or not of proteinase K (lanes 11-14). Total hydrogenosomal protein solubilized in 1% Triton X-100 was preincubated for 30 min at 37°C with increasing concentrations of TCEP and with 10 mM EDTA as indicated (lanes 15-24). Half of each sample was treated with 0.25 mg/ml proteinase K (lanes 16, 18, 20, 22, and 24) for 30 min on ice, whereas the remaining half was mock-treated ( lanes 15, 17, 19, 21, and 23). Proteins (40 g of protein per lane) were separated by 15% SDS-PAGE, transferred to PVDF, and detected with anti-Hmp35 antisera. B, the HMP35 complex is protease-resistant. Isolated hydrogenosomes (250 g of total protein per 250 l of buffer) were incubated in isotonic buffer with or without 0.25 mg/ml proteinase K (lanes 1, 2, 7, and 8). A second series was solubilized in 0.16% n-dodecyl maltoside in the presence or absence of 0.25 mg/ml proteinase K (lanes 3, 4, 9, and 10). A third series was boiled in 0.5% SDS for 5 min prior to results show that the HMP35 complex consists essentially of proteins that are not sodium carbonate-extractable and are therefore integral membrane proteins.
We investigated the resistance of the HMP35 complex to urea to determine whether hydrogen bonds were involved in maintaining it. The complex was not completely disassembled until it was exposed to 6 M urea (Fig. 4D, lane 7) but was maintained at a high molecular weight up to the 5 M urea treatment (Fig. 4D, lanes 1-6). The complex decreased in size by 12-kDa increments as the urea concentration was increased from 1 to 3 M, suggesting the loss of small proteins. When the urea concentration was increased from 3 to 4 M, a decrease of 32 kDa was noted, and a further 20 kDa as 5 M urea was used. These data suggest that the core of the complex is fairly stable in urea and that hydrogen bonds may be involved in anchoring small proteins to the core of the complex.
Endogenous Native Hmp35 Protein Is Resistant to Protease-We treated intact hydrogenosomes, solubilized hydrogenosomes, and isolated hydrogenosomal membranes, respectively, with proteinase K and tested the effect on Hmp35. When intact hydrogenosomes were exposed to 0.25 mg/ml proteinase K, the endogenous Hmp35 protein appeared to be unaffected (Fig. 5A, lanes 1-10). We tested the potential effect of EDTA, which would chelate any divalent cations that may create a stable domain within the protein, and also used the reducing agent TCEP, which would reduce the intramolecular disulfide bond within the endogenous Hmp35 protein (Fig. 3B). Neither of these reagents had any effect on the protease resistance of the Hmp35 protein in intact hydrogenosomes (Fig. 5A, lanes  4 -10). To rule out the possibility that the protein may be inherently resistant to protease due to its primary or secondary structure, we treated SDS-denatured hydrogenosomes with proteinase K and found that under these conditions Hmp35 was completely digested (Fig. 5A, lanes 11-14). Taken together, these data show that the Hmp35 protein does not have any major protease-sensitive cytosol-exposed domain. When the protein was solubilized under non-denaturing conditions, it was still largely resistant to 0.25 mg/ml proteinase K (Fig. 5A,  lanes 15-24). However, some of the Hmp35 protein was cleaved into fragments of 17 and 18 kDa (Fig. 5A, lanes 15-24) and may be protruding out of the complex core. Because these fragments are not seen in the assay performed on intact hydrogenosomes (Fig. 5A, lanes 1-4), it is likely that the protruding part of the complex is inside the hydrogenosome or within the membrane bilayer. As the reducing agent TCEP was added, these fragments got smaller until they disappeared at 50 mM TCEP (Fig.  5A, lanes 19 -24). There is also a reduction in the amount of full-length Hmp35 detected where 20 and 50 mM TCEP were included in the assay (Fig. 5A, lanes 22 and 24). Hence, the intramolecular disulfide bond detected in endogenous Hmp35 (Fig. 3B) may have a small but detectable effect on the integrity of the complex and its resistance to protease. Thus, the Hmp35 protein may be forming an exceptionally stable structure that defies protease digestion. The efficiency of the proteinase K to digest solubilized hydrogenosomal proteins was verified by Western analysis of the samples in Fig. 5A with antibodies against the membrane protein Hmp31. This protein was found to be completely sensitive to the proteinase K when solubilized under non-denaturing conditions (data not shown (18)). The HMP35 complex was found to be unaffected by proteinase K in either intact hydrogenosomes (Fig. 5B, lanes 1 and 2) or solubilized non-denatured hydrogenosomal protein (Fig. 5B, lanes 3  and 4). Thus, the HMP35 complex in its integrity is proteaseresistant. As a control for efficient digestion, we verified the effect of the protease on Hsp70, a soluble matrix protein, on intact hydrogenosomes, where it is completely protected (Fig.  5B, lanes 7 and 8), and on solubilized non-denatured hydrogenosomal protein, where it is completely digested (Fig. 5B,  lanes 9 and 10). Finally, when isolated sodium carbonate-extracted hydrogenosomal membranes were treated with proteinase K, we saw no discernible effect on the Hmp35 protein (Fig.  5C), although the Hmp31 integral membrane protein was digested. This observation, together with data from Fig. 5A, indicates that the protruding part of the Hmp35 protein within the complex (Fig. 5A, lane 16) is membrane-embedded. The recombinant protein His 6 -Hmp35 was found to be sensitive to 0.1 mg/ml proteinase K even under non-denaturing conditions (data not shown).
Purification of the HMP35 Complex-In order to purify the HMP35 complex from hydrogenosomes, we created 35H, a T. vaginalis transformant that expresses a C-terminally hexahistidine-tagged Hmp35 protein. This fusion protein is expressed from an episomal plasmid, pHmp35H (Fig. 6A), which also contains the neomycin phosphotransferase gene that confers resistance to the drug geneticin and provides for selection (18,25). The Hmp35-His 6 protein was efficiently targeted to the organellar pellet as shown by a crude cell fractionation and Western analysis with an anti-hexahistidine antibody (Fig. 6B,  lanes 1-3). Within purified 35H hydrogenosomes, the tagged protein was exclusively found in the membrane fraction of a sodium carbonate extraction, showing that the tag did not affect its sub-localization (Fig. 6B, lanes 4 -6).
To purify the HMP35 complex, we solubilized membranes from sonicated 35H and wild-type (wt) hydrogenosomes in 0.5% n-dodecyl maltoside in the presence of buffered 1 M NaCl and various protease inhibitors, and we incubated the resulting precleared lysate with nickel beads to isolate bound proteins. The wash buffers used, which included up to 4 M urea and 1 M NaCl, provided stringent conditions that would keep the complex largely intact (Fig. 4, B and D) but would clear nonspecifically bound proteins from the beads. The beads were divided into 2 aliquots that were eluted differently. One aliquot was eluted with a buffered detergent solution that contains 4 M urea supplemented with 0.25 M imidazole (Fig. 6C, lanes 2, 3, 6,  and 7). Under these conditions, the HMP35 complex should maintain a size of at least 200 kDa (Fig. 4D, lane 5), and the imidazole should specifically release nickel-bound hexahistidine-tagged proteins from the beads. The second aliquot was treated for 30 min at room temperature with a buffered detergent solution that contains 6 M urea and no imidazole. Under these conditions, the HMP35 complex is unstable (Fig. 4D, lane  7) and thus would disassemble into its constituents. However, because no imidazole was included in the buffer, the Hmp35-His 6 protein would stay bound to the beads. The resulting eluates were applied to a blue native gel, which was destained to reveal any major protein complexes stained by Coomassie addition or not of proteinase K (lanes 5, 6, 11, and 12). The protease treatment was performed for 30 min on ice followed by 10 min of inhibition with 2 mM PMSF. The samples were split into 2 aliquots, and 50 g of total protein was analyzed by 8 -20% SDS-PAGE linear gradient (top panels) or by 6 -16% BN-PAGE (bottom panels). Following electrophoresis, proteins were transferred to PVDF and detected with antisera against Hmp35.  (Fig. 6C, lanes 1-5) or transferred to PVDF for Western analysis with anti-hexahistidine antibody (Fig. 6C, lanes 6 -9). A large 300 -330-kDa band was detected specifically in the 4 M urea, 0.25 M imidazole eluate from 35H hydrogenosomes (Fig.  6C, lanes 3 and 7). No monomer-sized Hmp35-His 6 protein or any species smaller than 300 kDa was detected by the antihexahistidine antibody. This suggests that the purified HMP35-His 6 complex in the eluate was mostly of high molecular weight. When the complex was eluted with 6 M urea, no protein complex was seen in the destained gel (Fig. 6C, lane 5) nor was any protein detected by the anti-hexahistidine antibody (Fig. 6C, lane 9). This confirmed that most of the Hmp35-His 6 remained bound to the beads and that treatment with 6 M urea would not release an intact high molecular weight complex. The 4 M urea, 0.25 M imidazole eluates were separated by SDS-PAGE followed by silver staining to reveal the individual constituent proteins. The HMP35-His 6 complex appeared to consist of a major 40-kDa protein, a 35-and 12-kDa protein of moderate intensity, and lesser proteins of 32 and of 25-27 kDa (Fig. 6D, lane 2). The band at 65-kDa is likely to be a nonspecific protein because it also appears in the wt eluate (Fig. 6D, lane 1). The 6 M eluate from 35H hydrogenosomes reveals a similar pattern when analyzed on SDS-PAGE but has a reduced amount of the 40-kDa protein (Fig. 6D, lane 4). The 6 M eluate from wt hydrogenosomes does not contain these bands (Fig. 6D, lane 3), confirming that they are likely to originate from the HMP35-His 6 complex. Western analysis of the SDS-PAGE-separated proteins from the eluates with anti-hexahistidine antibody and with anti-Hmp35 antisera confirmed that the major 40-kDa band contained Hmp35-His 6 (Fig. 6E, both  panels, lane 2). The latter antisera also detected the lesser 35-kDa band (Fig. 6D, lane 2, and Fig. 6E, bottom panel, lane  2), which could be a degradation product of Hmp35. The 40-kDa protein in the 6 M eluate did not react with the anti-histidine antibody (Fig. 6E, top panel, lane 4), confirming that these elution conditions did not release Hmp35-His 6 protein from the beads. However, the 40-kDa protein was detected by the anti-Hmp35 antibody (Fig. 6E, bottom panel, lane 4) and is likely to be endogenous non-histidine-tagged Hmp35. Because this untagged protein was purified from a complex containing Hmp35-His 6 , the present data show that at least two Hmp35 monomers can self-associate. Moreover, from the relative amounts of the individual proteins in the HMP35-His 6 complex (Fig. 6D, lane  2), it appears that its major constituent is Hmp35 itself.
Hmp35 Protein Can Be Expressed in Yeast and Targeted to Mitochondrial Membranes in Vivo-Given evidence for a common ancestry for hydrogenosomes and mitochondria, we tested whether Hmp35, which has no known homologue in yeast, could be targeted to its mitochondria. To do so, we generated a yeast transformant, h35, with the hmp35 gene in the pRS313 plasmid under histidine selection. The Hmp35 protein was expressed stably in the h35 cells (Fig. 7A, lane 4) as shown by Western analysis with polyclonal anti-Hmp35 antisera. The h35 cells were subjected to a crude separation into a cytosolic and an organellar fraction. Western analysis showed that the Hmp35 protein was found almost exclusively in the organellar fraction (Fig. 7A, lanes 5 and 6), suggesting an efficient targeting mechanism to mitochondria from the cytosol.
Mitochondria were further purified from the crude organellar fraction of h35 cells to confirm that the Hmp35 protein was present in the organelles (Fig. 7C, top panel, lane 1). The sub-localization of the Hmp35 protein was investigated, by Western analysis, in relation to the control markers Tom20, an outer mitochondrial integral membrane protein, cytochrome b 2 , a soluble intermembrane space protein, Tim54, an inner mitochondrial membrane protein, Tim44, a peripheral inner membrane protein, and Cpn10, a soluble matrix protein. The relative positions of these markers in the mitochondrion are shown in Fig. 7B. To determine whether Hmp35 had any cytosolic-exposed domains, intact h35 mitochondria were treated with proteinase K. Under these conditions, Tom20 (37), which has a cytosolic domain, was sensitive to protease treatment, but Hmp35 was found to be resistant, thus suggesting that it is not significantly exposed to the cytosol (Fig. 7C, lanes 2 and 3).
Mitoplasts were generated by osmotically shocking the h35 mitochondria under hypotonic conditions that cause the outer membrane to rupture selectively. This treatment leaves the inner membrane and its matrix contents intact, while still retaining the ruptured outer membrane, and releases the contents of the intermembrane space (IMS) into the supernatant after centrifugation, where the mitoplasts are recuperated in the pellet. After this treatment, the marker for the IMS, cytochrome b 2 (Fig. 7C, lane 5), was found in the supernatant fraction as expected (38), but Hmp35 was found in the mitoplast pellet fraction (Fig. 7C, lane 4), showing that it was not present in soluble form in the IMS. When mitoplasts are treated with proteinase K, soluble IMS proteins, peripheral membrane proteins, and integral membrane proteins with domains protruding into the intermembrane space are digested in addition to outer membrane-exposed proteins such as Tom20. Upon this treatment, Tom20, cytochrome b 2 , and the inner mitochondrial membrane marker Tim54 (39), which protrudes into the IMS, were all digested, but the inner membraneprotected proteins Tim44 (23) and Cpn10 (40) were not (Fig.  7C, lanes 6 and 7). Under similar conditions, Hmp35 is sensitive to the protease (Fig. 7C, lane 6) suggesting that it is present between the two membranes.
To determine whether Hmp35 is present as an integral membrane protein in h35 mitochondria, the latter were sodium carbonate-extracted. It was found that the Hmp35 protein was distributed in both the integral membrane fraction and the soluble fraction (Fig. 7C, lane 9). The other markers for this fractionation were all distributed as expected (Fig. 7C, lane 9), with the integral membrane proteins Tom20 and Tim54 in the pellet, and the soluble proteins cytochrome b 2 and Cpn10 and the peripheral membrane protein Tim44 in the soluble fraction. Taken together, these localization experiments suggest that panel, H indicates purified hydrogenosomes; S indicates soluble fraction from sodium carbonate-extracted hydrogenosomes, and Mp indicates membrane protein pellet from the sodium carbonate extraction. The fractions were loaded such that soluble fraction ϩ membrane protein pellet ϭ purified hydrogenosomes. C, BN-PAGE showing the purification of the HMP35-His 6 complex from 35H hydrogenosomes (Hyd). Solubilized 35H and wild-type (wt) hydrogenosomes (1 mg of total protein each) were incubated with nickel beads, which were then washed and divided into 2 aliquots. One aliquot was incubated in solution D to release His-tagged proteins bound directly to the immobilized nickel ions (lanes 2, 3, 6, and 7). The 2nd aliquot was incubated in solution E, which provides conditions where His-tagged proteins should not be released from the beads but where the HMP35 complex would disassemble (lanes 4, 5, 8, and 9). Samples eluted from 0.1 mg of starting material were loaded per lane. Lanes 1-5 were destained to reveal Coomassie-stained proteins, and lanes 6 -9 were transferred to PVDF, and proteins were immunodecorated with antihexahistidine antibody. Lane 1 shows molecular mass markers, given in kDa. D, the HMP35-His 6 complex appears to consist mostly of a 40-kDa protein. Eluates from 0.1 mg of starting material described in C above were separated by 10 -16% SDS-PAGE to visualize the constituent proteins of the HMP35-His 6 complex by silver staining. Molecular mass markers are indicated in kDa. E, the Hmp35 protein self-associates within the HMP35-His 6 complex. The eluates shown in D were separated by 10 -16% SDS-PAGE, transferred to PVDF, and immunodecorated with either anti-hexahistidine tag antibody (top panel) or with anti-Hmp35 (bottom panel).
the Hmp35 protein is present in yeast mitochondria as both a peripheral and an integral membrane protein from the trans side of the outer membrane or the cis side of the inner membrane.
Hmp35 Forms a Stable Protein Complex in Mitochondria-To assess whether Hmp35 formed a high molecular weight complex in mitochondria, we performed BN-PAGE on digitonin-solubilized organelles from both wild-type and h35 yeast cells. The respective sizes of the TOM complex (Fig. 7D, lanes 3  and 4) at 400 kDa (41), the TIM23 complex (Fig. 7D, lanes 5 and  6) at 90 kDa (42), and of the TIM22 complex (Fig. 7D, lanes 7  and 8) at 250-300 kDa were similar in h35 mitochondria and in wild-type mitochondria, showing that the three major protein translocase complexes were unaffected in h35 mitochondria. Western analysis with anti-Hmp35 antisera showed that Hmp35 was present in a stable discrete complex of 300 kDa in h35 mitochondria (Fig. 7D, lanes 1 and 2). A smaller band was not detected by the antisera, suggesting that most of the Hmp35 protein is associating into this complex, despite the sodium carbonate extraction procedure showing that the protein was both peripherally and integrally attached to the membranes (Fig. 7C, lanes 8 and 9). Coimmunoprecipitation experiments performed with antibodies described above and against several other yeast mitochondrial membrane proteins known to be in large complexes were attempted, and all failed to immunoprecipitate Hmp35 (data not shown).

FIG. 7. Hmp35 is targeted to yeast mitochondria in vivo.
A, Hmp35 is expressed in yeast cells where it is specifically targeted to the crude mitochondrial pellet. Lactate-grown yeast h35 clones transformed with pRS313.hmp35 were converted to spheroplasts and subjected to a cell disruptor to break the cellular wall. As a control, wild-type (wt) yeast cells were subjected to the same treatment. B indicates broken cells after a 1000 ϫ g preclearing spin (lanes 1 and 4); S indicates supernatant after a 10,000 ϫ g spin to pellet crude mitochondria (lanes 2 and 5); P indicates crude mitochondrial pellet (lanes 3 and 6). The fractions were loaded such that supernatant ϩ crude mitochondrial pellet ϭ broken cells, and proteins were size-separated by 12% SDS-PAGE, followed by transfer to PVDF and immunodecoration with anti-Hmp35 antisera. B, scheme representing the relative positions of characterized mitochondrial protein markers. Integral membrane proteins are shown in white, peripheral membrane proteins in gray, and soluble proteins in black; OM indicates outer mitochondrial membrane; IM indicates inner mitochondrial membrane; IMS indicates intermembrane space. C, localization of Hmp35 within purified h35 mitochondria. M indicates intact mitochondria, untreated; M* indicates intact mitochondria treated with 100 g/ml proteinase K; MP indicates mitoplasts, untreated; MP* indicates mitoplasts treated with 100 g/ml proteinase K; Na 2 CO 3 indicates mitochondria subjected to sodium carbonate extraction. The protease treatment was inhibited with 2 mM PMSF, and samples were processed into a pellet (P), and a supernatant (S) after a 16,000 ϫ g spin in all cases except for the sodium carbonate extraction where centrifugation was performed at 100,000 ϫ g. Following these treatments, mitochondrial proteins were solubilized in Laemmli sample buffer and size-separated by 10% Tris-Tricine PAGE, electroblotted onto PVDF, and immunodecorated with antisera against Hmp35, Tom20, cytochrome b 2 , Tim54, Tim44, and Cpn10 as shown. D, Hmp35 associates into a stable 300-kDa complex in h35 mitochondria. Mitochondria isolated from wild type or h35 yeast strains were solubilized with 0.5% digitonin, and proteins were separated by BN-PAGE and transferred to PVDF membranes for immunodecoration with antisera against Hmp35, Tom40 (TOM complex), Tim23 (TIM23 complex), and Tim54 (TIM22 complex). Molecular mass markers are indicated in kDa.

DISCUSSION
We have characterized Hmp35, a novel membrane protein from hydrogenosomes of T. vaginalis, that has no known homologues. This protein has a putative RING-like, zinc-binding, cysteine-rich motif and forms an intramolecular disulfide bond. We demonstrate the application of blue native gel electrophoresis for hydrogenosomal complex separation and show that the Hmp35 protein is part of an integral membrane complex of about 300 kDa, HMP35, which is resistant to high concentrations of protease, urea, and salt. We have purified this complex, which appears to consist mostly of oligomerized Hmp35. Despite the Hmp35 protein not having any homologues in yeast, we show that it can be expressed heterologously in vivo in yeast and is targeted efficiently to mitochondria where it is integrated into membranes and forms a 300-kDa complex.
Hmp35 is the second membrane protein that has been reported from trichomonad hydrogenosomes after Hmp31, a member of the mitochondrial carrier family (18). The N terminus sequence of Hmp35 has no sequence similarity to the N-terminal cleavable presequence defined in Hmp31, which had been shown to be functional for targeting a matrix protein to the hydrogenosome, yet unnecessary for targeting mature Hmp31 (18). Thus, it appears that the Hmp31 N-terminal presequence is not conserved among membrane proteins as in the case for the N-terminal presequence on matrix-destined proteins (43).
Hmp35 is the first protein from hydrogenosomes to be described that has a putative zinc finger motif. Although most zinc finger protein motifs have been defined in transcription factors, there are examples of these motifs found in organellar membrane proteins. For example, subunit Vb of the cytochrome c oxidase complex, which is found in the inner mitochondrial membrane, has three conserved cysteine residues that coordinate one zinc atom (44). Other notable examples are the peroxins, which are integral membrane protein translocases in peroxisomes. The peroxins Pex2 (45), Pex10 (46), and Pex12 (47) are cysteine-rich proteins that have most of the cysteine residues clustered at the C-terminal end in a RING motif. The RING domain in Pex12 was found to be essential for its function in peroxisome biogenesis by interaction with another peroxin, Pex5, which conducts matrix protein import (47). These peroxins have no primary sequence similarity with each other, but the RING finger motif is conserved in a similar position. The arrangement of the seven cysteine residues at the C-terminal end of Hmp35 raises a parallel to these peroxins, although they are not arranged in a defined RING motif.
The iodoacetamide/AMS trapping experiment allowed us to determine that 2 cysteines out of the 14 in Hmp35 were engaged in intramolecular disulfide bonding in the endogenous protein. Proteins targeted to hydrogenosomes are post-translationally imported after synthesis in the cytosol (48). The cytosol is a reducing environment, such that precursor Hmp35 in the cytosol is likely to be in the reduced form. This implies the existence of a redox system in hydrogenosomes that is capable of oxidizing residues after import. In prokaryotes, the most studied disulfide bond formation is the Dsb system that consists of oxidation/isomerization proteins to form and regulate disulfide bonds (49). In mitochondria, it has been proposed that protein-disulfide isomerase (50) in the outer membrane could act similarly. Our finding that the inclusion of EDTA during the oxidation step of the trapping assay allows an extra disulfide bond to be formed suggests that divalent cations may be coordinated to the Hmp35 protein, presumably through some of the numerous cysteines. We were unable to force formation of extra disulfide bridges by oxidation of the endogenous protein in the absence of EDTA, although oxidation of most if not all of the cysteine residues was possible on the recombinant protein. This suggests that the cysteine-rich Hmp35 protein is unlikely to serve as defense against oxidative stress.
We found that Hmp35 was part of a high molecular mass protein complex of around 300 kDa, the first to be described in hydrogenosomes. The HMP35 complex stability in up to 5 M urea and 2 M salt is comparable with that of known pore proteins. For instance, in yeast mitochondria, the general im-FIG. 8. Secondary structure prediction profile for Hmp35. The secondary structure for Hmp35 (GenBank™ accession number AY312361), S. cerevisiae Tom40 (GenBank TM accession number NP013930), and Mycobacterium smegmatis MspA (GenBank TM accession number CAB56052) was predicted by the Garnier-Osguthorpe-Robson method using the programs Peptide Structure and Plotstructure (GCG Wisconsin Package, Accelrys). T indicates turn; H indicates ␣-helix; and B indicates ␤-sheet. Hmp35 and Tom40 are predicted to form 14 ␤-sheets each, and MspA has 10 potential ␤-sheets. The ␤-sheets in Hmp35 and Tom40 are separated by turns such that some of them may be arranged as anti-parallel sheets to form a ␤-barrel that could act as a pore. port pore, which contains the core proteins Tom22 and Tom40 in a 400-kDa complex, was shown to be resistant up to 1.5 M NaCl and 6 M urea and to sodium carbonate extraction (51). The general import pore forms the core of the outer membrane translocon that allows proteins from the cytosol to enter mitochondria. It has been proposed that the stability of the general import pore has evolved to withstand the forces involved in protein translocation into the matrix of mitochondria (51). The remarkable stability demonstrated for HMP35 suggests that this complex plays a role in hydrogenosomal translocation.
The endogenous non-denatured Hmp35 protein and HMP35 complex were found to be remarkably protease-resistant, whereas the recombinant non-denatured protein was completely protease-sensitive. It is possible that the resistance arises from a stable structure or a tight conformation between the individual components of the endogenous complex. Alternatively, the Hmp35 protein could be protected by another protein as in the case of bacterial secretins, which are proteaseresistant through protection by pilot proteins (52). One point to note is that the Hmp35 protein expressed heterologously in mitochondria was protease-sensitive, although it was in a complex and membrane-inserted. This implies that there may be a chaperone or other helper protein, present in hydrogenosomes but missing in mitochondria, that confers protease resistance upon the Hmp35 protein. Recently, the major porin from Mycobacteria, MspA, has been described as one of the most stable transmembrane proteins (53). This porin exists as a tetramer that is stable in heat up to 90°C in 2% SDS and in 7.6 M urea and at all pH values. In addition, the purified tetramer was found to be protease-resistant over extended periods and at high concentrations. It has been proposed that the extreme stability of this tetramer is due to the predominantly ␤-structure of the MspA protein, which may form a ␤-barrel within the tetramer to form the pore. Likewise, Tom40, the major constituent of the urea/salt/carbonate-resistant mitochondrial general import pore (51) is also predicted to have a ␤-barrel structure that forms a voltage-gated ion channel through which proteins are imported into mitochondria (22). Interestingly, Hmp35 is also predicted to be a predominantly ␤-structured protein with a 47% ␤-sheet content, which is comparable with the 45% predicted and the 48 -52% determined experimentally for MspA (53). The Toc75 protein translocase found in chloroplasts likewise exists as a ␤-barrel (54). Furthermore, the ␤-barrel structure is found in all outer membrane proteins in Gramnegative bacteria and a large number of outer membrane proteins in organelles of symbiotic origin, mitochondria, and chloroplasts. In contrast, membrane proteins in plasma or endoplasmic reticulum membranes are ␣-helical (55). As illustrated in Fig. 8, the secondary structure prediction profile for Hmp35 strongly resembles that of known pore-forming proteins. These data, together with the complex stability, indicate that the Hmp35 protein complex has a pore-forming function either as a protein translocase or as a solute carrier. Further experimentation addressing this issue awaits the development of new biochemical techniques for the hydrogenosome, such as the capacity to distinguish between the membranes, and the generation of productive protein translocation intermediates.
We had shown previously that a trichomonad hydrogenosomal membrane protein, Hmp31, could be targeted in organello into isolated yeast mitochondria. The Hmp31 protein has up to 52% similarity to yeast mitochondrial carrier family proteins and therefore was likely to have conserved membrane-targeting signals recognized by the mitochondrial protein translocation machinery (18,56). Recently, a related mitochondrial carrier family protein, the hydrogenosomal ADP/ATP carrier from the fungus Neocallimastix, which has Ͼ90% similarity with its yeast counterparts, was successfully imported in vivo into yeast mitochondria and restored an ADP/ATP carrier-deficient strain (8). In the present study, we show that a hydrogenosomal protein with no homology to any reported protein from yeast was expressed stably in yeast and was targeted to its mitochondria. It is unlikely that the Hmp35 protein was loosely associated to mitochondria because sodium carbonate extraction showed that at least half of it was integrated into the membrane. Interestingly, we found that the Hmp35 protein is present in a 300-kDa complex in mitochondria, which is of similar size to the endogenous HMP35 complex. The inability to coimmunoprecipitate this complex with mitochondrial proteins could be due to Hmp35 possibly being able to form a pore or complex by itself.
The successful import of hydrogenosomal Hmp35 into yeast mitochondria reiterates the compatibility between the protein translocation machineries of the two organelles which we and others have reported before (18,58). Our finding implies conservation of signals at three levels as follows: recognition by the cytosolic chaperone during protein synthesis in the cytosol, recognition by the TOM receptor for import into mitochondria, and finally, conserved membrane insertion signals (57). This finding has important evolutionary implications as it suggests that hydrogenosomal Hmp35 was present in mitochondriates and subsequently lost, although the protein import machinery of mitochondria can still recognize its targeting signals.
Trichomonad hydrogenosomal proteins reported to date have been shown to be homologous to either aerobic or anaerobic prokaryotic proteins or to mitochondria-derived proteins. Hmp35, on the other hand, has no known homologue in any kingdom even though a wide range of prokaryotes, Archaea and Metazoa, have now been sequenced. This raises the possibility that this abundant membrane protein arose during the creation of the trichomonad hydrogenosome. As genome sequence data becomes available for chytrids, fungi and ciliates that contain hydrogenosomes, it will be important to determine whether Hmp35 homologues exist in these lineages. The ubiquity of Hmp35 in hydrogenosome-containing lineages would imply a single origin for this organelle, although its presence solely in trichomonad hydrogenosomes would support polyphyletic origins.