Divergent Molecular Evolution of the Mitochondrial Sulfhydryl:Cytochrome c Oxidoreductase Erv in Opisthokonts and Parasitic Protists*

Background: The machinery for protein import into the mitochondrial intermembrane space has been studied in mammals, yeast, and plants. Results: Protist Erv homologues are conserved sulfhydryl:cytochrome c oxidoreductases with altered mechanisms. Conclusion: The composition and mechanism of the mitochondrial protein import machinery differs between eukaryotic lineages. Significance: The current knowledge on mitochondrial protein import cannot be generally transferred to all eukaryotes. Mia40 and the sulfhydryl:cytochrome c oxidoreductase Erv1/ALR are essential for oxidative protein import into the mitochondrial intermembrane space in yeast and mammals. Although mitochondrial protein import is functionally conserved in the course of evolution, many organisms seem to lack Mia40. Moreover, except for in organello import studies and in silico analyses, nothing is known about the function and properties of protist Erv homologues. Here we compared Erv homologues from yeast, the kinetoplastid parasite Leishmania tarentolae, and the non-related malaria parasite Plasmodium falciparum. Both parasite proteins have altered cysteine motifs, formed intermolecular disulfide bonds in vitro and in vivo, and could not replace Erv1 from yeast despite successful mitochondrial protein import in vivo. To analyze its enzymatic activity, we established the expression and purification of recombinant full-length L. tarentolae Erv and compared the mechanism with related and non-related flavoproteins. Enzyme assays indeed confirmed an electron transferase activity with equine and yeast cytochrome c, suggesting a conservation of the enzymatic activity in different eukaryotic lineages. However, although Erv and non-related flavoproteins are intriguing examples of convergent molecular evolution resulting in similar enzyme properties, the mechanisms of Erv homologues from parasitic protists and opisthokonts differ significantly. In summary, the Erv-mediated reduction of cytochrome c might be highly conserved throughout evolution despite the apparent absence of Mia40 in many eukaryotes. Nevertheless, the knowledge on mitochondrial protein import in yeast and mammals cannot be generally transferred to all other eukaryotes, and the corresponding pathways, components, and mechanisms remain to be analyzed.

Saccharomyces cerevisiae Erv1 (ScErv1) 3 and its mammalian homologue "augmenter of liver regeneration" (ALR) assist the Mia40-dependent oxidative trapping of proteins in the mitochondrial intermembrane space. Mia40 itself oxidizes a variety of cysteine-containing substrates resulting in substrate disulfide bond formation and protein folding. Reduced Mia40 subsequently transfers an electron pair to the flavoprotein Erv1/ ALR. Erv1/ALR serves as a redox switch, accepting one electron pair and transferring two single electrons separately to two molecules of cytochrome c. Accordingly, mitochondrial protein import is linked to the respiratory chain. Alternatively, Erv1/ALR was suggested to directly transfer electrons to molecular oxygen (1)(2)(3)(4)(5)(6). Even though the Mia40/ Erv1 system is essential in opisthokonts (a eukaryotic lineage including yeast and mammals), inactivation of Mia40 was reported to have no detectable deleterious phenotype in Arabidopsis (7). In addition, several protist lineages including apicomplexan and kinetoplastid parasites apparently have no Mia40 homologue at all, pointing to significantly altered mitochondrial import machineries (2,8,9). Nevertheless, recent experiments support the theory that mitochondrial protein import is functionally conserved in the course of evolution because Leishmania tarentolae marker proteins of all four mitochondrial compartments were efficiently imported into mitochondria from yeast, and opisthokont marker proteins were also successfully imported into isolated L. tarentolae mitochondria. These in organello experiments included the import of yeast and L. tarentolae homologues of Erv and small Tim proteins as substrates of the mitochondrial oxidative folding pathway (9).
The structures and mechanisms of Erv homologues from mammals (10 -13), yeast (14 -16), and Arabidopsis thaliana (17)(18)(19) have been studied in great detail. In contrast, very little is known about protist Erv homologues (2,8,9). Here, we compare Erv homologues from three different eukaryotic lineages, (i) L. tarentolae Erv (LtErv) as a member of the excavata/discicristata, (ii) Plasmodium falciparum Erv (PfErv) as a member of the chromalveolata, and (iii) ScErv1 as a member of the opisthokonta. We show that both parasite Erv homologues cannot complement the knock-out of ERV1 in yeast despite successful mitochondrial protein import. Furthermore, we characterize full-length wild type and mutant LtErv and compare its properties and enzymatic mechanism with related and non-related flavoproteins. Our findings indicate that the sulfhydryl:cytochrome c oxidoreductase activity of opisthokont Erv1/ALR is conserved in the course of evolution. On the one hand, Erv and non-related flavoproteins are intriguing examples of convergent molecular evolution resulting in similar enzyme properties. On the other hand, the exact mechanisms of Erv homologues from different eukaryotic lineages vary significantly. Considering the results from this and our previous study, the Mia/Erv pathway in opisthokonts does not seem to be strictly conserved throughout evolution. As a consequence, the knowledge on mitochondrial protein import into the intermembrane space of opisthokonts cannot be generally transferred to all other eukaryotes, and the corresponding pathways remain to be analyzed.

EXPERIMENTAL PROCEDURES
Bioinformatics Analyses and Molecular Modeling-BLAST searches and multiple sequence alignments were performed as described previously (9). Partial molecular models of PfErv and LtErv were generated based on the crystal structures of homodimeric rat ALR (PDB entry 1OQC) (10) and Erv from A. thaliana (PDB entry 2HJ3) (19). Alignments between the templates and the parasite Erv homologues were optimized manually in the Swiss-PDB Viewer (spdbv). Computations of the models were carried out at the Swiss-Model server using the "project" mode (20,21). Very similar models were obtained for both templates. A variety of automated analyses at the server indicated a high quality of the models, and force field energies (calculated with the GROMOS96 implementation of spdbv) ranged without further energy minimization from Ϫ3.2 MJ/mol per subunit of PfErv to Ϫ4.8 MJ/mol per subunit of LtErv.
Cloning and Site-directed Mutagenesis-Untagged fulllength PFERV was PCR-amplified from a P. falciparum cDNA library (strain 3D7) with perfect match primers (Metabion) that were designed based on the Plasmo DB annotation PFA0500w. The PCR product was cloned into the vector pDrive using the PCR cloning kit from Qiagen. LTERV was cloned previously (9). Constructs encoding N-terminally MRGS(H) 6 GS-tagged PfErv and LtErv were subcloned by PCR using the BamHI and HindIII restriction sites of the Escherichia coli expression vector pQE30 (Qiagen). Cysteine point mutations of LTERV C17S , LTERV C63S , LTERV C66S , LTERV C300S , and LTERV C304S in pQE30 were introduced by PCR using mutated primers and the QuikChange site-directed mutagenesis kit from Stratagene. A construct encoding C-terminally LE(H) 6 -tagged PfErv I2V was PCRcloned via the restriction sites NcoI (causing the point mutation) and XhoI of pET28 (Novagen). The same restriction sites were used to clone untagged PFERV I2V into the yeast expression vectors pYX143 and pYX232. Untagged LTERV was subcloned from pDrive into plasmids pYX143 and pYX232 via the BamHI and XhoI restriction sites. As positive controls for complementation assays, intron-free SCERV1 was subcloned from pYX232 into pYX143 using the restriction sites EcoRI and XhoI. Correct insertions and sequences of all wild type and mutant constructs were confirmed by sequencing both strands.
Cell Culture and Western Blot Analyses-Unless otherwise indicated, S. cerevisiae cultures were grown at 30°C in YP (1% yeast extract, 2% peptone) medium or in synthetic (S) minimal medium supplemented with or without the bases adenine and uracil and the amino acids lysine, leucine, histidine, and tryptophan (22). Media contained either 2% glucose (YPD, SD), galactose (YPGal or SGal), or lactate (SLac). The GAL-MIA40 strain expresses Mia40 under control of the galactose-induced and glucose-repressed GAL10 promoter (23). For expression of LTERV in this genetic background, the strain was transformed with plasmid pYX232-LTERV and selected on SD agar plates without tryptophan. Positive clones were grown in liquid SLac medium containing 0.5% galactose. For depletion of Mia40, the culture was shifted to tryptophan-free SLac medium containing 0.5% glucose for 30 h. For the interaction studies the plasmid pYX232-LTERV was transformed in a strain expressing Mia40 with a C-terminal His 8 tag (24).
L. tarentolae promastigotes were cultured at 27°C according to standard protocols (25) in brain heart infusion medium supplemented with hemin as described previously (9). P. falciparum strain 3D7 was cultured at 37°C according to standard protocols (26) in RPMI medium containing 0.45% (w/v) Albumax II and human type A erythrocytes at a hematocrit of 3% under a low-oxygen atmosphere (5% CO 2 , 5% O 2 , 90% N 2 , and 80% humidity). For Western blot analyses P. falciparum cultures were centrifuged (3 min, 1500 ϫ g), and erythrocytes were lysed in 0.02% saponin as described previously (27). Purified parasites were directly boiled in SDS-PAGE Laemmli buffer. The content and redox state of ScErv1, LtErv, and marker proteins were evaluated by reducing and non-reducing SDS-PAGE and Western blot analyses as described previously (9,28,29). For the generation of PfErv antibodies, rabbits were immunized with a synthetic peptide (NH 2 -CQDIQTIIDKYKTVD-COOH) corresponding to the C terminus of PfErv (Pineda Antibody Service). Antibodies were subsequently purified from sera by affinity chromatography according to established protocols (9). Approximately 2 ϫ 10 8 purified malaria parasites were loaded per lane.
Complementation Assays-A YPH501-derived yeast ⌬erv1 knock-out strain (genotype ura3-52, lys2-801 amber , ade2-101 ocre , trp1-⌬63, his3-⌬200, leu2-⌬1, erv1::His3) with a copy of SCERV1 on a URA3-containing pRS426 plasmid under control of the MET25 promoter (30) was transformed with the TRP1-containing plasmids pYX232, pYX232-SCERV1, pYX232-LTERV, and pYX232-PFERV or a plasmid-free control according to a standard protocol by Ref. 31. Transformants were selected on SD agar plates without uracil, histidine, and tryptophan, and six clones of each strain were incubated in 1 ml of SD liquid medium without histidine and tryptophan overnight. Liquid cultures were subsequently diluted to an absor-bance of 0.1 in sterile water, and 10-l drops of each dilution were spotted onto SD agar plates without histidine and tryptophan containing 50 mg/liter uracil and 1 g/liter 5-fluoroorotic acid for negative selection and plasmid shuffling (32). As a control, 10-l drops of each diluted culture were also spotted on SD agar plates without uracil, histidine, and tryptophan. In analogy, the same ⌬erv1 knock-out strain with an episomal copy of SCERV1 was transformed with the LEU2-containing plasmids pYX143, pYX143-SCERV1, pYX143-LTERV, and pYX143-PFERV or a plasmid-free control. After selection on SD agar plates without uracil, histidine, and leucine, six clones of each strain were incubated in 1 ml of SGal liquid medium without histidine and leucine for several days. Liquid cultures were subsequently diluted to an absorbance of 0.1 in sterile water, and 10-l drops of each dilution were spotted onto SD and SGal agar plates without histidine and leucine containing 50 mg/liter uracil and 1 g/liter 5-fluoroorotic acid for negative selection and plasmid shuffling. All plates were incubated for up to 3 weeks at 24, 30, and 37°C. Strains were checked for ScErv1 and LtErv by Western blot analyses.
Submitochondrial Localization Studies in Yeast-Yeast strains were grown on SD liquid medium without negative selection by 5-fluoroorotic acid. Mitochondria were purified, and proteolytic susceptibility assays with proteinase K in the presence or absence of a hypotonic buffer and/or detergent were performed as described previously (28). The membrane association of LtErv and marker proteins was analyzed by carbonate extraction and Western blotting as described (9) with yeast mitochondria replacing L. tarentolae mitochondria.
Interaction (Pulldown) Studies-Purified yeast mitochondria (1 mg) containing wild type or His-tagged Mia40 were solubilized in 250 l of buffer (20 mM Tris/HCl, 80 mM KCl, 30 mM imidazole, pH 7.4, 1 mM phenylmethylsulfonyl fluoride) containing 0.5% (w/v) Triton X-100 at 4°C. After a clarifying spin, the solubilized material was incubated with 30 l of nickelnitrilotriacetic acid agarose (Ni-NTA) beads for 1 h at 4°C under gentle mixing. After 3 washing steps with the solubilization buffer containing 0.05% (w/v) Triton X-100, bound proteins were eluted with Laemmli buffer containing 300 mM imidazole for 5 min at 95°C. Samples were analyzed by SDS-PAGE and immunodecoration with antibodies.
Heterologous Expression and Protein Purification-Histagged wild type and mutant LtErv was expressed in E. coli strain SHuffle T7 Express (New England Biolabs). Overnight cultures were diluted into 6 liters of Luria-Bertani medium supplemented with 100 g/ml ampicillin to an initial absorbance at 600 nm of ϳ0.02. The cultures were grown at 25°C to an absorbance of 0.3-0.4 and were subsequently shifted to 16°C. Expression was induced at an absorbance of 0.5-0.7 by adding 0.1 mM isopropyl-␤-D-1-thiogalactopyranoside. After 17-18 h, cultures were transferred to an ice-water bath and harvested by a single centrifugation step at 4°C using a Beckman JS-4.2 rotor (15 min, 4000 ϫ g). Bacteria were rapidly resuspended on ice in 60 ml of disruption buffer containing 0.25% (v/v) Tween 20, 20 mM imidazole, 300 mM NaCl, 50 mM sodium phosphate, pH 8.5 (at 4°C). The suspension was supplemented with DNase and lysozyme (1 mg/ml) and stirred for 1 h on ice. Afterward, the suspension was sonicated and centrifuged at 4°C (30 min, 10,000 ϫ g). The obtained pellet was resuspended again in disruption buffer followed by another sonication and centrifugation cycle (the second extraction was found to significantly improve the overall yield). The supernatants were loaded on a column containing 0.5-0.7 ml of Ni-NTA that had been equilibrated with disruption buffer. The column was washed with 10 -14 ml of 75 mM imidazole, 300 mM NaCl, 50 mM sodium phosphate, pH 8.0. Afterward, LtErv was eluted twice with 0.5-0.7 ml of elution buffer containing 400 mM imidazole, 300 mM NaCl, 50 mM sodium phosphate, pH 8.0.
LtErv was further purified by gel filtration chromatography on a HiLoad 16/60 Superdex 75 prep grade column that was connected to an ÄKTA-FPLC system (GE Healthcare). The protein was eluted in buffer containing 300 mM NaCl, 50 mM sodium phosphate, pH 7.8, with or without 2 mM dithiothreitol (DTT). FPLC fractions were detected photometrically, and protein-containing fractions were analyzed by standard reducing and non-reducing SDS-PAGE and Western blotting against the His tag.
Non-enzymatic Characterization of LtErv-Concentrations of purified wild type and mutant LtErv were determined according to Bradford with bovine serum albumin as a standard (33). In addition, protein concentrations and extinction coefficients were evaluated based on FAD absorbance: Spectra of 800-l samples of ϳ20 M LtErv in elution buffer were recorded at 25°C on a thermostatted Jasco V-650 UV-visible spectrophotometer. Afterward 20 l of a 10% SDS-solution were added, and spectra were recorded over time until no further changes were observed (34). The spectra were corrected for the absorbance of elution buffer with or without 0.25% SDS. The exact concentration of FAD in the cuvette was determined with reference spectra of 12-57 M FAD (Sigma) in elution buffer containing 0.25% SDS.
To confirm the integrity of purified LtErv, proteolytic digests of the final single SDS-PAGE band were analyzed by HPLC-MS/MS on a LTQ-Orbitrap mass spectrometer at a core facility of the Heidelberg University. The sequence coverage with trypsin and elastase was 61 and 80%, respectively. C-terminal peptides with complete y series were detected in both digests indicating that not only the N terminus (detected by Western blotting) but also the C terminus of purified LtErv was unaltered.
DTT:Cytochrome c Electron Transferase Assay-The enzymatic activity of purified wild type and mutant LtErv was analyzed according to a previously published assay with DTT as electron donor and cytochrome c as electron acceptor (15). DTT was from Fluka, and equine heart cytochrome c and bovine erythrocyte superoxide dismutase 1 (SOD) were purchased from Sigma. Stock solutions of 2 mM DTT and 1 mM cytochrome c were prepared freshly before each set of measurements in assay buffer containing 50 mM sodium phosphate, pH 8.0, supplemented with 300 mM NaCl and 2.5 mM EDTA. Initial reaction velocities were measured at 25°C in assay buffer with 2 M LtErv, 40 M equine heart cytochrome c, and variable concentrations of DTT ranging from 5 M to 1 mM. In one set of experiments S. cerevisiae cytochrome c from Sigma was used for comparison. Alternatively, variable concentrations of heart cytochrome c ranging from 2 to 40 M were analyzed with 0.4 M LtErv and 0.1 mM DTT. All assays were performed with or without 75 units/ml SOD to discriminate between Erv-mediated specific electron transfer and nonspecific electron transfer due to flavoprotein-dependent superoxide anion formation (13). Furthermore, activities were corrected for the spontaneous reduction of cytochrome c by DTT. All kinetics were monitored at 550 nm and evaluated by non-linear (hyperbolic) regression analyses in Sigmaplot 10.0 (Systat) using an extinction coefficient of 29 mM Ϫ1 cm Ϫ1 for reduced equine cytochrome c.

Kinetoplastid Erv Homologues Have Altered Domain
Architectures-Based on our previous in silico analyses (9), we generated a multiple sequence alignment for Erv homologues from kinetoplastid and apicomplexan parasites (Fig. 1A). Although LtErv shares a higher sequence similarity with yeast mitochondrial Erv1 than endoplasmic reticulum Erv2 (E values of 1e Ϫ24 and 1e Ϫ16 , respectively), the difference is far less obvious for PfErv (E values of 4e Ϫ18 and 6e Ϫ18 , respectively). Using different bioinformatics programs listed at ExPASy, no N-terminal targeting signal was predicted for the Erv homologues from parasitic protists in contrast to Erv2 from yeast. Thus, as previously confirmed for LtErv (9), Erv homologues from kinetoplastid and apicomplexan parasites are probably located in the mitochondrial intermembrane space and not in the endoplasmic reticulum. Moreover, the acquisition of an Erv homologue in the endoplasmic reticulum seems to be a more recent, presumably fungi-specific event in the course of evolution (35).
The Erv domain (36) is conserved in all homologues in Fig.  1A, except for helix ␣5, which might be replaced by a different structural element e.g. in Erv homologues from apicomplexan parasites. Accordingly, stable and reliable molecular models for dimeric PfErv and LtErv including the FAD binding site were generated based on the structure of rat ALR (Fig. 1, B-D). The most striking feature of Erv homologues from kinetoplastid parasites is an additional C-terminal domain of almost 200 amino acids (Fig. 1, A and C). Using this sequence as an input for BLAST searches against the genomes of kinetoplastid parasites yielded Erv homologues only. Analogous searches against non-Kinetoplastida revealed no significant hits (the lowest E values were 0.002 for hypothetical proteins). Thus, the domain is specific for these Erv homologues, and we therefore termed it Kinetoplastida-specific second (KISS) domain of Erv. Bioinformatics predictions for the KISS domain suggest an ␣␤␤␣␣ secondary structure with elongated loops between each of the secondary structure elements. Of note, the KISS domain is rather variable and differs not only between Leishmania and Trypanosoma homologues but also between Trypanosoma spp. For example, a highly charged segment after the first predicted ␣ helix in Leishmania spp. and Trypanosoma cruzi presumably forms another ␣ helix that is missing in the Trypanosoma brucei homologue (see deletion between residues VKS and VLEL in Fig. 1A). In contrast, the N-terminal linker region between the domains as well as the C-terminal arm of the KISS domain are highly conserved. In summary, Erv homologues from apicomplexan and kinetoplastid parasites most likely adopt the typical Erv/ALR fold resulting in FAD binding and homodimer forma-tion. The C-terminal KISS domain is a special feature of Erv homologues from Kinetoplastida.
Protist Erv Homologues Have Altered Cysteine Motifs-The proteins in Fig. 1 can be grouped according to their cysteine patterns and deviations at the N and C termini. Four different disulfide bond-forming cysteine pairs can be distinguished for Erv homologues (Fig. 1) (2): (i) a proximal or active site cysteine pair, (ii) a distal or shuttle cysteine pair, (iii) a structural disulfide bond, and (iv) a clamp disulfide bond. (i) The cysteine residues in the Y(P/A)CXXC motif at the active site are the only strictly conserved residues in addition to the FAD binding site (Fig. 1, A-D). The disulfide bond between the former residues was shown to be essential for the oxidase activity of ScErv1 (37). (ii) The distal cysteine pair is highly variable. Erv homologues from alveolates including the apicomplexan parasites have a CX(D/E)X 2 C motif instead of an opisthokont-specific CRXC motif at the N-terminal arm of the protein (Fig. 1, A and B). In contrast, the distal cysteine pair of Erv homologues from kinetoplastid parasites is located close to the C terminus in a CXVYC motif (Fig. 1, A and C). This architecture is analogous to Erv homologues from plants (18) and Erv2 from yeast (14) (Fig. 1A). (iii) The structural disulfide bond connecting helix ␣4 and the following loop cannot be formed in PfErv (Fig. 1, A and B) because the first cysteine residue is missing, and the second residue is poorly conserved in Plasmodium. Because both structural cysteine residues are found in ciliate Erv homologues (Fig. 1A), the disulfide bond was presumably lost during the evolution of apicomplexan parasites. Thus, the structural disulfide bond is not a fundamental prerequisite for proper Erv folding. (iv) The N-terminal cysteine residue of the disulfide clamp connecting both subunits in ALR is conserved among Kinetoplastida, ciliates, and most opisthokonts (Fig. 1, A, C, and D), but is absent in apicomplexan parasites, plants, and yeast (Fig. 1,  A and B). The second residue is even less conserved, suggesting that the clamp disulfide bond was either lost or acquired several times throughout evolution. Please note that the cysteine residue in the CP motif before the Erv/ALR domain might form alternative intermolecular disulfide bonds, e.g. in LtErv (Fig. 1, A and C) and similar proteins, provided that the N-terminal arm has a high flexibility. Noteworthy, alternative inter-or intramolecular disulfide patterns that are poorly conserved throughout evolution are relatively common among redox proteins and can be also found, for example, in peroxiredoxins (38) and glutaredoxins (39).
In summary, structural prerequisites for electron transfer, including the proximal active site cysteine pair and a distal shuttle cysteine pair for the interaction with protein substrates, are conserved among Erv homologues. However, the distal cysteine residues are located either close to the N or the C terminus and are found in highly variable motifs regarding the type and number of spacing residues. Other cysteine residues forming a structural or a clamp disulfide bond are not strictly conserved among Erv homologues.
Protist Erv Homologues Cannot Complement a ⌬erv1 Yeast Strain-Even though protein import into the mitochondrial intermembrane space is functionally conserved throughout evolution, the import mechanisms are presumably altered (9). To address the question of whether protist Erv homologues can replace ScErv1 in the yeast model system, we cloned and subcloned the genes encoding PfErv and LtErv (Table 1) and performed complementation assays using alternative vector/promotor systems and culture conditions as summarized under "Experimental Procedures". In contrast to a positive control, none of the conditions resulted in a successful complementation ( Fig. 2A). Thus, the genes encoding PfErv and LtErv cannot rescue yeast in the absence of endogenous ScErv1. Whether the The N and C termini are missing in the molecular models. The distance between the thiol groups of residues Cys 101 and Cys 129 in PfErv is almost 18 Å. Putative interactions between the active site and distal cysteine residues are indicated by dotted arrows. D, the structure of rat ALR (10) served as a template for the models and is shown for comparison (PDB entry 1OQC). Please note the additional intermolecular head-to-tail disulfide clamp between relatively conserved cysteine residues before and after the Erv/ALR domain. unsuccessful complementation is due to the essential roles of ScErv1 for protein import (1)(2)(3)(4)(5) or for maturation of cytosolic iron-sulfur clusters (40) remains to be studied. Moreover, a negative result from complementation assays can obviously have alternative causes such as protein mistargeting. We, therefore, performed the following controls.
Because of the availability of a specific antibody (9), we checked whether LtErv localizes to the intermembrane space of yeast mitochondria (Fig. 2B). LtErv was clearly enriched in mitochondrial subcellular fractions (data not shown). In contrast to Tom70 (a marker protein of the outer membrane), LtErv was protected from proteolytic degradation as long as the outer mitochondrial membrane was intact. Under hypotonic conditions, the outer membrane ruptured, and LtErv was degraded like ScErv1 and the intermembrane space marker protein Tim50. Matrix proteins including the marker Hep1 were still protected unless the inner mitochondrial membrane was permeabilized by a detergent (Fig. 2B). We, therefore, conclude that LtErv was correctly targeted to the intermembrane space of yeast mitochondria. Of note, the presented in vivo localization data confirm our previous conclusions on the functional conservation of mitochondrial protein import from in organello experiments (9) (Fig. 2C) (even though analogous in organello import experiments with PfErv could not be performed because of technical difficulties).
Despite the absence of striking hydrophobic patches, LtErv was shown to be tightly membrane-associated in L. tarentolae mitochondria (9). To analyze a potential membrane association in yeast mitochondria, we performed analogous carbonate extractions (Fig. 2D). In contrast to soluble ScErv1 and Hep1, LtErv was predominantly found in the pellet fraction indicating a membrane association as determined in Leishmania. Next, we checked whether S. cerevisiae Mia40 plays a role for the import of LtErv into yeast mitochondria. Depletion of Mia40 did not affect the mitochondrial levels of LtErv but resulted in reduced levels of known import substrates of Mia40 including ScErv1, Ccs1, and Tim13 (Fig. 2E) (41)(42)(43)(44)(45). Other mitochondrial proteins, such as the matrix proteins Tim44 and Hep1, were not affected by the depletion of Mia40 (Fig. 2E). These results suggest that LtErv is imported into the intermembrane space of yeast mitochondria independent of Mia40. Endogenous ScErv1 interacts with Mia40 (29). When we performed pulldown experiments using a His-tagged form of Mia40 and mitochondrial lysates containing LtErv, LtErv was not co-purified with Mia40-His 8 , whereas a fraction of ScErv1, used as control protein, was co-isolated. Thus, an interaction between LtErv and Mia40 was not observed, in contrast to the interaction with ScErv1 (Fig. 2F). Two possible explanations are that LtErv either does not bind at all to Mia40 or that the interaction cannot be observed in the presence of competing ScErv1 due to a significantly lower affinity. At present, preliminary experiments revealed no indication that LtErv is able to oxidize Mia40 in an Erv-catalyzed Mia40 oxidation assay in vitro. In conclusion, the inability of parasite Erv homologues to complement a ⌬erv1 yeast strain is not due to protein mislocalization but is most likely due to different mechanistic properties. Moreover, the in vivo import of LtErv into yeast mitochondria appears to   JANUARY 25, 2013 • VOLUME 288 • NUMBER 4

JOURNAL OF BIOLOGICAL CHEMISTRY 2681
be Mia40-independent, pointing to an alternative import mechanism.
LtErv Is a Functional Sulfhydryl:Cytochrome c Oxidoreductase-Initial experiments revealed that recombinant LtErv and PfErv cannot be purified from E. coli using standard expression/ purification protocols. Even though we were unable to purify soluble N-or C-terminally tagged PfErv (despite a variety of tested E. coli strains, tags, and expression protocols), we established and optimized an expression/purification protocol for membrane-associated N-terminally His-tagged full-length LtErv (Fig. 3, A and B) (see also "Experimental Procedures" for details). The purified protein was subsequently analyzed in a DTT:cytochrome c electron transferase assay (Fig. 3, C and D). The activity in the assay was comparable to ScErv1, which served as a positive control (data not shown). Depending on the purification, apparent k cat values in the presence of SOD were ϳ1 s Ϫ1 , and the estimated apparent K m value for DTT was ϳ1-2 mM. A similar K m app value for DTT was determined for ALR in an oxidase assay, and a k cat app value of 4.5 s Ϫ1 was reported for a short splice variant of ALR in stopped-flow measurements using 2 mM DTT and variable cytochrome c concentrations (13). Furthermore, the reported K m app value for cytochrome c of ALR around 10 M (13) is similar to the value of LtErv around 3 M (Fig. 3D). In the absence of SOD the apparent activity of LtErv was significantly higher, in particular, at increasing DTT concentrations (Fig. 3C). Thus, in agreement with previous reports on flavoproteins (2,13), a background reaction due to an unspecific oxidase activity of LtErv cannot be neglected under aerobic conditions in vitro, and the addition of SOD (to remove superoxide anion as an electron donor) is a prerequisite for the correct analysis of the electron transferase activity. Nevertheless, a comparison of the activities in Fig. 3C and the low K m app value in Fig. 3D support the theory that cytochrome c, not O 2 , is the physiological electron acceptor of Erv in Kinetoplastida (2) in accordance with previous observations on ALR (13). Thus, LtErv is primarily an electron transferase (EC 1.8.2) and not an oxidase (EC 1.8.3), and the enzymatic activity of opisthokont Erv homologues is presumably conserved in other eukaryotic lineages. Of note, LtErv transferred electrons to equine and yeast cytochrome c, suggesting that the interaction with the second substrate might be either strictly conserved throughout evolution or that an electron might tunnel from reduced Erv to cytochrome c as soon as both proteins get close enough. In summary, we established the purification of full-length LtErv and confirmed that it is a functional sulfhydryl:cytochrome c oxidoreductase.

Protist Erv Homologues Form Intermolecular Disulfide Bonds in Vitro and in Vivo-A second band at ϳ70 kDa was detected by non-reducing SDS-PAGE upon purification of recombinant
LtErv. Under reducing conditions, the 70-kDa band was converted to the single 36-kDa band of full-length monomeric LtErv (Fig. 4A). We, therefore, conclude that the 70-kDa band corresponds to disulfide-bridged dimeric LtErv. Next, we asked whether the band is also formed in E. coli and in L. tarentolae. A second band at the corresponding size was indeed observed by Western blot analyses of E. coli extracts and purified L. tarentolae mitochondria (Fig. 4B). To analyze the covalent dimerization of PfErv, a peptide antibody was generated and purified. Although a dimer could not be unambiguously assigned, Western blot analyses of P. falciparum cell extracts revealed several potential mixed disulfide bonds between PfErv and other proteins (Fig. 4C). In E. coli extracts, a band at ϳ35 kDa was detected in addition to the band of monomeric recombinant PfErv around 18 kDa. Because the former band was lost upon reduction, PfErv also most likely forms disulfide-bridged homodimers. In summary, we identified (potential) intermolecular disulfide bonds for PfErv and LtErv in vitro and in vivo.
The Enzymatic Mechanism of LtErv-To decipher which cysteine residues shuttle the electron pair from the substrate at the protein surface to the active site, we generated several LtErv mutants and determined their ability to form disulfide bridges in E. coli by Western blot analyses (Fig. 5). Long film exposures revealed two alternative LtErv dimer bands. The lower dimer band was specifically lost when either the more N-terminal residue of the active site cysteine pair (Cys 63 ) or the more C-terminal residue of the distal cysteine pair (Cys 304 ) was mutated. Thus, the C-terminal arm of one subunit forms a disulfide bond with the active site of a second subunit via Cys 304Ј and the socalled interchange residue Cys 63 (Figs. 1C and 5). Another intermolecular disulfide bond resulting in the upper dimer band was formed between Cys 300 of two subunits, suggesting that this residue usually interacts with the substrate thiol group resulting in substrate oxidation.
Next, we purified selected cysteine mutants and wild type LtErv and compared their spectra and activities. The flavin absorbance maxima of LtErv (Table 2) shifted to 450, 376, and 271 nm upon denaturation (Fig. 6A). Thus, FAD (and not FMN) was liberated from LtErv. Calibration curves from two independent expression and purification experiments revealed flavin-to-protein ratios (M FAD to M LtErv) of 1.0 and 0.98. Thus, the recombinant protein was saturated with FAD. The spectral properties of LtErv C304S , LtErv C300S/C304S , and wild type enzyme were identical (Fig. 6B), whereas LtErv C300S had an additional maximum at longer wavelengths (Fig. 6C, Table 2). The positions of the maxima in the LtErv C66S spectrum resembled free FAD (Fig. 6, A and D, Table 2). In contrast, the spectrum of LtErv C63S revealed a pronounced shoulder at longer wavelengths (Fig. 6E).
The spectra can be mechanistically interpreted as follows (Fig. 6F); in LtErv C304S , LtErv C300S/C304S , and wild type enzyme the active site cysteine pair formed a disulfide bond in the proximity of FAD. In accordance with the molecular model in Fig.  1C, removal of the interchange residue Cys 63 (Fig. 5) trapped an interaction between reduced residue Cys 66 and FAD (Fig. 6, E  and F). Because a fraction of Cys 66 was deprotonated at the slightly alkaline pH, a charge-transfer complex (CTC) between the thiolate and FAD was partially formed as predicted previously (2). Furthermore, the rather minor spectral changes in LtErv C66S (Fig. 6D) indicate that the electron density of FAD is also slightly influenced by Cys 66 in the fully oxidized protein.
The pronounced CTC of LtErv C300S and the additional maximum at longer wavelengths are interpreted as a delocalization of conjugated electron pairs between the C-terminal cysteine residues of one subunit and the active site of a second subunit (Fig. 6F). In the absence of Cys 300 and of a physiological electron acceptor, this species was trapped (Fig. 6C). Thus, the mutant probably mimics a reaction intermediate of the reductive halfreaction (2). Because of an oxidase activity of LtErv (Fig. 3C), the CTC in LtErv C300S and LtErv C66S was slowly lost over time (data not shown). A semiquinone radical, therefore, had to be formed as predicted for the catalytic mechanism (2) in accordance with observations on a short splice variant of human ALR (13) and A. thaliana Erv1 (17). A similar spectrum to the one of LtErv C300S in Fig. 6C was also recorded for ScErv1 C30S (16) in agreement with an essential intermolecular disulfide bond between Cys 33 and Cys 130Ј in vivo (15,16).
Regarding the electron transferase activity of LtErv, replacement of the interchange residue in LtErv C63S completely abolished the enzymatic activity, whereas mutations of the distal cysteine residues in the C-terminal arm had either no or intermediate effects (Fig. 6G). Thus, DTT cannot directly reduce Cys 66 or FAD, and the proximal active site cysteine pair is a prerequisite for the electron transferase activity of LtErv. However, DTT is small enough to bypass the distal cysteine residues and to directly interact with the interchange residue Cys 63 dur-  (Table 1) results in a protein with a theoretical molecular mass of ϳ13 kDa. The band labeled with an asterisk could, therefore, be either a nonspecific cross reaction or proteolytically processed PfErv. Ab, affinity purified antibody. D, dimer; D?, potential dimer; M, monomeric full-length LtErv; ME, 2-mercaptoethanol; PS, preimmune serum; S, serum.
ing catalysis. The lower activity and the higher K m app value of the mutant LtErv C300S (Fig. 6G, Table 2) might be explained by a trapped conformation of the C-terminal arm partially blocking the active site of the second subunit. The results are not only in perfect agreement with the molecular model in Fig. 1C and the spectral analyses in Fig. 6 but also with DTT oxidase activities measured for the short splice variant of human ALR lacking the distal cysteine residues (as well as mutants without the clampforming cysteine residues) (13). Moreover, similar results were obtained for oxidase activities of ScErv1 double mutants without the N-terminal distal cysteine pair Cys 30 /Cys 33 or the active site cysteine pair Cys 130 /Cys 133 (37). In summary, we unraveled the disulfide bond patterns for recombinant LtErv and gained first insights into its catalytic mechanism.
Model of LtErv Catalysis-Based on the results in Figs. 4 -6, we propose the following model for the sulfhydryl:cytochrome c oxidoreductase activity of LtErv (Fig. 7). During the reductive half-reaction electrons are transferred from the sulfhydryl substrate to the distal shuttle cysteine disulfide bond at the C-terminal arm of LtErv. After a conformational change, Cys 304 from one subunit interacts with the interchange cysteine residue Cys 63Ј of another subunit. The electron pair is subsequently delocalized between the shuttle cysteine residues and the active site of two subunits (Figs. 6C and 7) before Cys 300 restores the disulfide bond with Cys 304 at the C-terminal arm. As a result, a protein species with a thiolate-FAD CTC is formed (Figs. 6E  and 7), and the disulfide bond formation at the C-terminal arm might trigger the reversal of the conformational change. During the predicted oxidative half-reaction, two molecules of cytochrome c consecutively accept a single electron from the flavin of LtErv (Fig. 7). In summary, in analogy to A. thaliana Erv1 (18,19), LtErv requires a C-terminal shuttle cysteine pair in contrast to ScErv1 and ALR. Because of the position of the shuttle arm, the additional KISS domain, and the altered cysteine motifs, the mechanisms of kinetoplastid and opisthokont Erv homologues differ significantly, providing a plausible explanation for the results in Fig. 2.
Flavoenzymes Are Intriguing Examples of Convergent Molecular Evolution-The spectra of LtErv C304S , LtErv C300S/C304S , and wild type enzyme are similar to other oxidized flavoproteins such as glutathione reductase (46) or thioredoxin reductase (47,48) despite completely different primary structures and protein folds (2,35,36). Analogous to the spectrum of borohydride-reduced glutathione reductase (46), the delocalization of the negative charge in LtErv C63S results in a thiolate-FAD CTC shoulder at longer wavelengths (Fig. 6E). Similar spectra were recorded for mutants of the interchange residue Cys 130 in ScErv1 (16) and Cys 145 in human ALR (49,50). Thus, a CTC similar to many other flavoproteins also seems to be a conserved feature of the Erv family.
Erv and thioredoxin reductase both require a (flexible) arm with a distal cysteine pair to shuttle electrons between cysteine residues of a bulky protein substrate and the active site with a flavin cofactor serving as an electron switch (2). The reductive half-reaction of Erv (Fig. 7) (2) and the oxidative half-reaction of thioredoxin reductase (47,48), therefore, share significant similarities despite completely different protein folds. As a result, the spectrum of LtErv C300S with a delocalized electron pair between the C-terminal arm of one subunit and the active site of another subunit (Fig. 6C) resembles the spectra of reduced thioredoxin reductase isoforms (47,48). Accordingly, analogous reaction intermediates are formed (Fig. 7) (47,48) irrespective of the position of the distal shuttle cysteine pair at the N or C terminus and the inverse flow of electrons in Erv and thioredoxin reductase. In summary, thioredoxin reductase and Erv homologues are bona fide examples for convergent molecular evolution that is presumably driven by the selection pressure of fundamental principles in redox catalysis and flavin biochemistry.
Divergent Molecular Evolution of Erv-We recently described three different scenarios for the evolution of the Mia40/Erv1 system and oxidative protein import into the mito-  chondrial intermembrane space; primitive eukaryotes either contained the ancestors of (i) Erv and Mia40, (ii) Erv alone, or (iii) Erv and an unknown component Z (9). Because Mia40 cannot be identified in silico in many eukaryotic lineages (8,9), the first scenario is rather unlikely unless Mia40 was lost or its primary structure was significantly altered in the course of evolution. Considering (a) the conserved electron transferase activity with cytochrome c as an electron acceptor (Fig. 6G), (b) the conserved import of LtErv in organello and in L. tarentolae (9), and (c) the successful Mia40-independent import of LtErv in S. cerevisiae (Fig. 2), scenarios (ii) and (iii) might be favored. The tight membrane association of full-length LtErv in L. tarentolae (9) and S. cerevisiae (Fig. 2D) might be responsible for trapping LtErv in the intermembrane space of both organisms. Thus, following an established nomenclature (51), LtErv could be a special class III intermembrane space protein. Because LtErv was also membrane-associated in E. coli, the protein might have a high affinity for lipids such as cardiolipin.
Four different types of non-related shuttle arms are found in Erv homologues from opisthokonts, plants, alveolates, and Kinetoplastida. Two of them are located at the N terminus, and two of them are found at the C terminus (Fig. 8) (2,12,15,16,18,19). Thus, potential alternative machineries for the import of proteins into the mitochondrial intermembrane space might have evolved independently in different eukaryotic lineages. The position of the shuttle arm does not seem to be a prerequisite for the interaction with Mia40 because opisthokonts and plants both have a Mia40/Erv1 couple despite opposing arm topologies (Fig. 8). However, like the protist Erv homologues ( Fig. 2A), the Arabidopsis Erv homologue could not complement Erv1 in yeast (7), suggesting a complementarity of the lineage-specific shuttle arm and cysteine motifs of the Mia40/ FIGURE 6. Analysis of the enzymatic mechanism of LtErv. A, UV-visible spectrum and determination of the extinction coefficient of wild type LtErv are shown. The eluate is shown in the inset. Spectra were recorded before (orange line) and after the addition of SDS (black line). The exact concentration of liberated FAD in the cuvette was determined with reference spectra of pure FAD yielding calibration curves for the absorbances at 450 and 376 nm (closed and open symbols, respectively). B, shown are normalized spectra of LtErv C304S and LtErv C300S/C304S (black lines) compared with wild type enzyme (orange line). C-E, shown are normalized spectra of freshly purified LtErv 300S (blue line), LtErv C66S (green line), and LtErv C63S (red line) compared with wild type enzyme (orange line). F, shown are schematic interpretations of the spectra in panels B-E. G, shown is the sulfhydryl:cytochrome c oxidoreductase activity of wild type and mutant LtErv in the presence of SOD. Protein properties are summarized in Table 2.
Erv couple. The exact compositions and mechanisms of the mitochondrial machineries for oxidative protein import in parasitic protists are so far completely unknown (2,8,9). Notably, because of the conservation of the cysteine residues in kinetoplastid homologues (Fig. 1A), the enzymatic mechanism of LtErv depicted in Fig. 7 could be also valid for other kinetoplastid parasites including highly important pathogens. Because mitochondrial protein import is expected to be essential for parasite survival, the alternative mechanisms of kinetoplastid and mammalian Erv catalysis are not only interesting from an evolutionary viewpoint but might also be attractive for inhibition studies.
Conclusion-We showed that neither apicomplexan nor kinetoplastid Erv can replace essential Erv1 in yeast. Likely reasons are altered cysteine patterns and protein architectures that might result in incompatible protein-protein interactions with Mia40. Furthermore, a purification protocol for recombinant full-length wild type and mutant LtErv was established, and first insights into the mechanism of kinetoplastid Erv catalysis were gained. On the one hand, the characterization of LtErv revealed that the sulfhydryl:cytochrome c oxidoreductase activity of opisthokont Erv homologues is conserved in the course of evolution, whereas the exact enzymatic mechanism is significantly altered in different eukaryotic lineages. The results are in agreement with previous in silico analyses suggesting that the Mia40/ Erv1 pathway in opisthokonts is not a general feature in different eukaryotic lineages (2,8,9). On the other hand, the comparison of Erv properties with non-related flavoproteins  Representative homologues without similarities at the N and C termini (Fig. 1A) from four eukaryotic lineages are shown. The protein architectures suggest an early divergent molecular evolution that might point to significantly different import machineries for mitochondrial intermembrane space proteins.