Evidence of a Reduced and Modified Mitochondrial Protein Import Apparatus in Microsporidian Mitosomes

Abstract

Microsporidia are a group of highly adapted obligate intracellular parasites that are now recognized as close relatives of fungi. Their adaptation to parasitism has resulted in broad and severe reduction at (i) a genomic level by extensive gene loss, gene compaction, and gene shortening; (ii) a biochemical level with the loss of much basic metabolism; and (iii) a cellular level, resulting in lost or cryptic organelles. Consistent with this trend, the mitochondrion is severely reduced, lacking ATP synthesis and other typical functions and apparently containing only a fraction of the proteins of canonical mitochondria. We have investigated the mitochondrial protein import apparatus of this reduced organelle in the microsporidian Encephalitozoon cuniculi and find evidence of reduced and modified machinery. Notably, a putative outer membrane receptor, Tom70, is reduced in length but maintains a conserved structure chiefly consisting of tetratricopeptide repeats. When expressed in Saccharomyces cerevisiae, EcTom70 inserts with the correct topology into the outer membrane of mitochondria but is unable to complement the growth defects of Tom70-deficient yeast. We have scanned genomic data using hidden Markov models for other homologues of import machinery proteins and find evidence of severe reduction of this system.

Microsporidia are a eukaryotic group highly adapted as obligate intracellular parasites (31, 50). They infect a diverse range of vertebrate and invertebrate animal hosts. In humans they are the cause of a number of diseases (e.g., gastroenteritis, encephalitis, and hepatitis), having their greatest impact on immunocompromised individuals, notably in children with human immunodeficiency virus (14, 31). Microsporidia are most closely related to fungi, although their high level of specialization as intracellular parasites obscured this relationship for a long time (18, 25, 30). Gene phylogenies now firmly connect these two groups, although it remains uncertain whether microsporidia are sisters to the fungi or represent a lineage derived from within fungal diversity (21, 28).

A clear adaptive response to parasitism in microsporidia has been a reduction in cellular complexity. This was first recognized at an ultrastructural level with the apparent lack of peroxisomes, flagella, stacked Golgi bodies, and mitochondria (31). This reductive evolution is mirrored at a genomic level, with microsporidia containing the smallest eukaryotic genomes described to date (28, 29). The complete genomic sequence from the human microsporidian parasite Encephalitozoon cuniculi reveals a genome of only ∼2.9 Mb containing approximately 2,000 genes, in contrast to the 6,000 genes found in the genome of the model fungus Saccharomyces cerevisiae. The minimal genome of E. cuniculi has been achieved through three mechanisms in concert: (i) gene loss, resulting in broad loss of biochemical pathways and capabilities, including much basic energy metabolism and numerous anabolic pathways; (ii) gene compaction with an average intergenic space of ∼130 bp; and (iii) gene shortening, with E. cuniculi genes being on average 14% shorter than their homologues in fungi such as S. cerevisiae (28, 45). Thus, microsporidian evolution has apparently been shaped by a very strong trend to eliminate superfluous molecular and biochemical complexity.

Despite earlier suppositions that microsporidia lacked mitochondria, genome and expressed sequence tag data from microsporidia suggested the presence of several proteins typically targeted to this organelle (3, 19, 20, 24, 28, 38). Immunolocalization of a mitochondrial Hsp70 to small double membrane-bound organelles in Trachipleistophora hominis provided strong evidence for the existence of a mitochondrion in microsporidia, albeit a simplified organelle that lacks cisternae (48). Annotation of genomic data from E. cuniculi provided compelling matches for only 22 proteins implicated in mitochondrial function, suggesting that the metabolism of this relict mitochondrion (or mitosome) is also significantly reduced compared to that of canonical mitochondria (28). Further, no mitochondrial genome has been retained; thus, biogenesis of this organelle is wholly dependent on nucleus-encoded proteins. Based on these 22 proteins, a major role for the mitosome is iron-sulfur cluster assembly (22, 28). No genes have been found for ATP synthesis via oxidative phosphorylation, suggesting loss of this activity in mitosomes (28, 46). While it is likely that further mitosome-targeted proteins will be identified, it is clear that compared to mitochondria from fungal relatives, which are known to import ∼1,000 proteins (40, 44), microsporidian mitosomes represent organelles with highly reduced proteomes, a feature consistent with other traits of cellular reduction.

The highly reduced state of the microsporidian mitosome, requiring only a fraction of the protein diversity of other mitochondria, presents an interesting case for studying organelle biogenesis—particularly the machinery for protein import of nucleus-encoded proteins. Mitochondrial protein import has been best characterized in fungi, and in these systems most proteins are imported via four major import complexes: a TOM (translocase of the outer mitochondrial membrane), a SAM (sorting and assembly machinery), and one of two TIMs (translocase of the inner mitochondrial membrane), TIM23 or TIM22 (see Fig. 5A) (5, 36). These complexes are broadly conserved throughout fungi as well as animals (15). Mitochondrial proteins can take one of several routes to the mitochondrion via this apparatus (5, 36). Broadly, soluble matrix proteins are recognized at the TOM complex by the receptor protein Tom20 through the binding of N-terminal presequences with characteristic features (1, 5, 7, 8, 36). These proteins are passed through the pore protein Tom40 of the TOM to the TIM23 complex and then driven into the mitochondrial matrix by way of the presequence translocase-associated motor (PAM) complex, where their presequences are subsequently removed. Some membrane proteins can also be released into the inner membrane from the TIM23 complex. Mitochondrial proteins that apparently lack such an extension, notably including many of the membrane proteins, are recognized by internal sequence elements. Tom70 has a greater role in recognizing these internal signals and thus the import of hydrophobic proteins (4, 11, 32, 39, 47). Such hydrophobic proteins are often bound by cytosolic molecular chaperones (Hsp70 and/or Hsp90) en route to the mitochondrion, and Tom70 is known to independently bind to both the chaperone and the substrate protein (7, 23, 33, 52). While a measure of substrate overlap between Tom20 and Tom70 occurs, the division of responsibility between these two receptors has likely evolved in response to the wide range of substrate proteins that must be imported into mitochondria and the need to handle this complexity.

FIG. 5.

Schematics of the protein import machinery and pathways in yeast mitochondria (A) and E. cuniculi mitosome (B) based on identified homologues of the general fungal/animal pathways. Protein components of the yeast system were all represented by HMMs used to search the microsporidian genomic data and represent the major presequence-dependent and presequence-independent pathways. Homologues identified in E. cuniculi indicate a severely reduced import apparatus utilizing elements of the presequence-independent pathway.

For microsporidia little is known of the protein import apparatus for their relict mitochondrion, the mitosome. Has the very reduced organelle proteome, in concert with a genome-wide trend of the loss of redundant or superfluous genes, resulted in a smaller and/or derived import apparatus? In this study we have investigated the microsporidian mitosome protein import apparatus from E. cuniculi in order to evaluate how this apparatus has responded to the reduction in the number of proteins required to be imported and the overall radical reduction in the number and size of proteins encoded in the nuclear genome. A putative homologue of the outer membrane receptor protein Tom70 is of particular interest as the only receptor for the TOM complex and, given the known structure of Tom70 proteins, provides a highly informative example of how proteins can be shortened in the course of genome reduction.

MATERIALS AND METHODS

Analyses of HMMs.

Hidden Markov models (HMMs) were generated for the following mitochondrial import protein-related families: AlphaMPP, BetaMPP, Hsp70, MCP, Mdm10, Metaxin1, Metaxin2, OxaI, Pam16, Pam18, Sam35, Sam37, Sam50, Sec61gamma, SecA, SecE, SecG, SecY, Skp, SurA, TatC, Tim10, Tim13, Tim17, Tim18, Tim21, Tim22, Tim23, Tim44, Tim50, Tim54, Tim8, Tim9, Tim9-10, Tom20, Tom22, Tom40, Tom5, Tom6, Tom70, and Tom7. Models were based on known homologues in as broad a taxon sample as possible (protein training sets are available upon request). The building of HMMs and the search of the E. cuniculi genomic data were performed with the software package HMMER (16, 17), version 3.2.3. E. cuniculi genomic data were downloaded from the Genoscope web site with the EMBL file (http://www.genoscope.cns.fr/externe/sequences/banque_Projet_AD). Coding sequences were extracted with the EMBOSS program “coderet” (41). The HMM search was performed with the program “hmmpfam.” The initial candidate hits were extracted as matches with E values of <0.01 with in-house tools. Each candidate sequence was manually reviewed and searched against the NCBI Protein database (BLASTp) to eliminate proteins with strong matches to nontarget proteins.

Expression vector construction.

E. cuniculi genomic DNA, kindly provided by E. S. Didier (Tulane University, Louisiana), was harvested from purified spores grown in tissue culture using the QiaAmp mini-DNA extraction kit (Qiagen, California). By using PCR the complete EcTom70 open reading frame was amplified with in-frame restriction sites with the forward primer 5′-GTA TCT AGA ATG GTG GGG AGG AAG ACT TCG-3′ and either the reverse primer 5′-GAC TGT CGA CTC AAA ACG TCT TAA GAA CAT CAG-3′, removing the stop codon for ligation in front of green fluorescent protein (GFP)-S65T, or 5′-ACT GGG ATC CAA ACG TCT TAA GAA CAT CAG G-3′ for expression without GFP. PCR fragments were ligated into a modified version of the yeast expression vector p416 MET25 HDEL (37), and sequences were confirmed by DNA sequencing. Plasmids were transformed into haploid yeast strain MH272 and the Δtom70/tom71 double deletion strain (NCY 0404) (11) and plated onto uracil-deficient selective medium.

In vivo localization.

Yeast cells expressing EcTom70-GFP fusion proteins were stained with Mitotracker Red CM-H2XRos (Molecular Probes) and imaged with a Zeiss Axioplan2 microscope and an AxioCam MRm digital camera. For protein immunoblot analyses, membrane-associated proteins were separated from soluble proteins by mechanical disruption of cells (twice for 2 min each in a bead beater with silica beads in breaking buffer [BB; 0.6 M sorbitol, 20 mM K⁺ morpholineethanesulfonic acid, pH 6.0], 0.2 mg/ml phenylmethylsulfonyl fluoride, and protease inhibitors) followed by ultracentrifugation (16,000 × g, 10 min, 4°C). Pelleted membrane and soluble fractions were precipitated with trichloroacetic acid (TCA). Mitochondria were isolated from whole cells according to published methods (13), quantified by protein content, and stored as frozen aliquots in bovine serum albumin (5 mg/ml). Trypsin shaving of intact mitochondria was performed on 25-mg aliquots of purified mitochondria, and shaved mitochondria were washed in BB (pH 7.4) and pelleted (16,000 × g, 10 min, 4°C). Mitochondria were resuspended in 100 μl of one of the following: (i) BB, pH 7.5; (ii) BB (pH 7.5), trypsin (100 μg/ml), and trypsin soybean inhibitor (800 μg/ml); (iii) BB (pH 7.5) and trypsin (100 μg/ml); and (iv) a hypotonic buffer (50 mM morpholinepropanesulfonic acid, pH 7.4) and trypsin (100 μg/ml). Suspensions were incubated for 20 min at 4°C. Trypsin soybean inhibitor was then added to an 800-μg/ml concentration to mixtures 3 and 4, and proteins were precipitated by TCA prior to immunoblot analysis. Sodium carbonate extraction of proteins from mitochondria was performed by resuspension of mitochondrial aliquots (25 mg) in 100 μl Na₂CO₃ (0.1 M, pH 11), incubation for 30 min at 4°C, and then ultracentrifugation (100,000 × g, 30 min, 4°C) to pellet membranes. Pelleted and supernatant fractions were TCA precipitated prior to immunoblot analysis.

Yeast growth assays.

Yeast cells were grown to mid-logarithmic phase (optical density at 600 nm of 0.6) in selective minimal medium (SD^−URA) and diluted to an optical density at 600 nm of 0.2, and then 5-μl aliquots were serially diluted fivefold and spotted onto YPAD (yeast extract 1%, peptone 2%, adenine 0.1%, glucose 2%) and YPEG (yeast extract 1%, peptone 2%, ethanol 3%, glycerol 3%) plates. Plates were incubated at 25°C, 30°C, or 37°C for 2 to 4 days until colonies were visible and then photographed.

RESULTS

Structural conservation and divergence of E. cuniculi Tom70.

Annotation of E. cuniculi genomic data identified an open reading frame with sequence similarity to Tom70 (28). This putative protein, EcTom70, is 477 residues long, 23% shorter than S. cerevisiae Tom70 (617 residues), and shares only 12% sequence identity with this fungal homologue. Given the size difference and low identity, EcTom70 is not reliably recovered by BLAST searches using fungal or animal Tom70s. To scrutinize the identity of this putative microsporidian Tom70, we used HMMs as search tools to see if this E. cuniculi sequence could be recovered. Using 35 diverse Tom70s from fungi and animals, an HMM was developed and used to screen the E. cuniculi genomic data. EcTom70 was recovered as the best match to the Tom70 model (E value, 4.00e⁻⁰⁵), supporting its identity as a Tom70 homologue.

The Tom70 from S. cerevisiae (ScTom70) contains an N-terminal transmembrane domain that anchors this protein in the outer mitochondrial membrane, with the remainder of the protein exposed to the cytosol (11). The crystal structure of this cytosolic portion of ScTom70 has been determined and consists of 26 α-helices (51). Twenty-two of these helices contribute to 11 tetratricopeptide repeat (TPR) motifs; this motif is a 34-residue motif consisting of two antiparallel α-helices. The first three of these TPRs and a seventh helix form an N-terminal “clamp” domain implicated in chaperone binding, and the remaining 19 helices create a C-terminal region (referred to as the “core” domain) that selectively binds mitochondrial preproteins (9, 11). A 27-residue region linking the clamp and core domains appears to provide a flexible interdomain loop and was unresolved in the crystal structure; however, the two domains contact via helices A7 and A25-26 arranged in antiparallel orientation (51).

Comparison of EcTom70 to ScTom70 in multiple sequence alignments, including homologues from diverse fungi and animals, allowed assessment of equivalent structures in the microsporidian protein. EcTom70 is predicted to contain a single transmembrane domain at the N terminus (using the TMPred algorithm [26]) (Fig. 1A). Prediction of repeated motifs (by REP algorithm [http://www.embl-heidelberg.de/]) (2) identifies seven likely TPR motifs that correspond in position to ScTom70 TPR1, -2, -4, -8, -9, -10, and -11, with a similar helix-turn-helix structure predicted (with weak similarity to an ankyrin repeat) at the position equivalent to TPR7. Manual inspection of the alignment and prediction of α-helices by JPRED (12) indicates that paired helices corresponding to TPR3 and TPR5 might form TPR structures equivalent to these regions of EcTom70. Further, helices A7, A8, A25, and A26 in ScTom70, which contribute to clamp-core interactions, are also represented by helical segments in the EcTom70 protein sequence (Fig. 1A). Thus, several of the key structural features of ScTom70 are predicted for the EcTom70 sequence. Notable differences are the apparent loss of TRP6 by a deletion that retains only a short helical region (Fig. 1A) and reduction of the two linker regions, (i) the region linking the transmembrane domain with the clamp domain and (ii) that linking the clamp and core domains. Together these changes contribute to the overall shortening of EcTom70 without substantial change to the predicted structure of the protein.

FIG. 1.

(A) Schematic of the conserved domain structure of Tom70s from yeast (S. cerevisiae) and E. cuniculi. The three functional domains consist of (i) the transmembrane anchor (TM), (ii) the clamp domain, and (iii) the core domain. TPRs determined by crystal structure (ScTom70) or predicted by REP (EcTom70) are shown in yellow, α-helices are shown in green, and helix-turn-helix motifs are shown in orange. (B) Sequence-structural alignment of TPR clamp domains that bind chaperones Hsp70 and/or Hsp90. Conserved residues involved in electrostatic interactions with the terminal EEVD motif of chaperones are shown in red. Conserved residues implicated in TRP packing interactions between α-helices are shown in bold black. H represents predicted helical segments. Protein accession numbers and domain residue numbers are as follows: E. cuniculi Tom70, XP_955641, 37 to 144; S. cerevisiae Tom70, ABN58618, 99 to 200; Neurospora crassa Tom70, P23231, 136 to 237; Homo sapiens Tom70, AAH65555, 113 to 220; Drosophila melanogaster, NP_609536, 90 to 197; H. sapiens HopTPR2A, 1ELR_A, 4 to 113; H. sapiens HopTPR1, 1ELW_A, 4 to 106; H. sapiens CHIP, AAD33400, 26 to 128; H. sapiens FKBP52, 1P5Q_A, 147 to 264. (C) EcTom70 clamp domain (residues 34 to 143) modeled on the structure of human Hop TPR1 clamp domain in complex with an Hsp70 peptide (EEVD) (42) (PDB 1ELW). The bound EEVD peptide is shown in thick-stick mode. The five conserved clamp domain residues highlighted red in panel B are shown in narrow-stick mode with the conserved residue identity shown before the slash and the EcTom70 residue after the slash. (D) Hsp70 and Hsp90 alignment of C termini representing diverse eukaryotes. The conservation of the terminal EEVD is almost ubiquitous with the exception of microsporidian E. cuniculi.

The role of the clamp domain in fungal and animal Tom70 proteins is to bind to molecular chaperones Hsp70 and/or Hsp90, which aid in delivering hydrophobic mitochondrial proteins to the import complex (52). This chaperone-receptor interaction is mediated by binding of the C terminus of the chaperones, characterized by the sequence EEVD, within a groove formed by the three TPRs of the clamp (6, 42, 52). Electrostatic interactions of the terminal aspartate form a dicarboxylate anchor within the clamp. In PSI-BLAST searches of the Protein Data Bank (http://www.rcsb.org/pdb/home/home.do), TPR1 to TPR3 of EcTom70 recovered only significant matches (6.8e⁻²) to other chaperone binding proteins, such as Tom70s, Hop, CHIP, and FKBP52 (52). A sequence-structural alignment of Tom70 clamps from E. cuniculi and select fungi and animals was generated (Fig. 1B). Contact residues in the known structures that interact with the EEVD motif, in particular the dicarboxylate anchor, are shown in red (Fig. 1B). These residues are conserved in the clamp sequences, indicating the importance of this dicarboxylate interaction (42, 52). Further, residues important for packing interactions of the TPRs are also conserved (Fig. 1B, bold) (42). While the residues for packing of the clamp domain are conserved in EcTom70, those residues that specifically interact with the terminal aspartate are not (Fig. 1B and C). These data suggest that while EcTom70 would likely form the clamp structure, it would be unable to form a dicarboxylate anchor with the EEVD peptide of chaperones.

To investigate the implications of this unusual clamp domain, the genome sequence for E. cuniculi was searched for molecular chaperones that could interact with TPR-clamp domains. No chaperones were identified with C-terminal EEVD, and a search of all E. cuniculi open reading frames equal to or greater than 40 codons revealed no such terminal sequence encoded in E. cuniculi. The E. cuniculi genome encodes three isoforms of Hsp70 and one of Hsp90, but none have a C-terminal sequence of EEVD. The only candidate with a similar motif is EcHsp90 (accession no. NP_584635), with the C-terminal sequence EEVQ. The presence of a cytosolic chaperone ending in EEVD (either Hsp70, Hsp90, or both) is virtually ubiquitous in eukaryotes (Fig. 1D) (6, 42), implying a critical function for this chaperone motif. Thus, the microsporidian E. cuniculi is exceptional: the binding site of the clamp domain of EcTom70 is modified, and the C terminus of chaperone EcHsp90 ends in the sequence EEVQ. Notably, loss of a negative charge in the chaperone C terminus coincides with loss of conserved positively charged residues in the clamp binding pocket (Fig. 1C and D).

EcTom70 targets mitochondria in yeast.

Microsporidia represent an experimentally intractable group, with no systems currently available to genetically transform or manipulate these parasites. We therefore used S. cerevisiae as a model fungus in which to express EcTom70 and test for targeting and function consistent with a role in mitochondrial protein import. EcTom70 was fused with the reporter protein GFP, and fluorescence microscopy showed that EcTom70 colocalized with the mitochondrial stain MitoTracker in yeast (Fig. 2), with no detectable localization in other cell compartments. The protein is present in membrane extracts from yeast, specifically in mitochondrial membranes (Fig. 2). To test if EcTom70 is targeted to the yeast mitochondrial outer membrane, with cytosolic orientation consistent with TOM function, we analyzed the topology of EcTom70. Mitochondria were purified from yeast cells expressing EcTom70 and treated with trypsin, with immunoblotting being used to test for exposure of EcTom70-GFP to protease degradation. Under these conditions both EcTom70-GFP and endogenous ScTom70 are degraded by trypsin shaving of intact mitochondria (Fig. 3A). With the same treatment, the intermembrane space protein cytochrome b₂ (Cyb2) is not degraded, implying that the outer membrane is intact and, thus, EcTom70-GFP, like ScTom70, is attached to the cytosolic face of the outer membrane. Only with disruption of the outer mitochondrial membrane by mild osmotic shock is Cyb2 susceptible to trypsin (Fig. 3A). Sodium carbonate extraction of mitochondrial membranes shows that EcTom70 is anchored as an integral membrane protein, not as a peripheral membrane protein (Fig. 3B). This was also the case for endogenous ScTom70 and the integral membrane protein porin, whereas Hsp70, a peripheral component of the mitochondrial inner membrane and matrix, was enriched in the soluble fraction.

FIG. 2.

Localization of EcTom70-GFP to the mitochondrion of S. cerevisiae in live cells costained with mitochondrial marker MitoTracker and immunostaining of EcTom70-GFP with GFP antisera on Western blots of total soluble protein (Sol), total membrane proteins (Memb), and total mitochondrial proteins (Mito). DIC, differential interference contrast.

FIG. 3.

Membrane topology of EcTom70-GFP in yeast mitochondria determined by trypsin shaving (A) and sodium carbonate extraction (B). (A) Trypsin-shaved isolated yeast mitochondria immunostained for EcTom70-GFP, ScTom70, and intermembrane space protein Cyb2 (with GFP, ScTom70, and Cyb2 antisera, respectively). EcTom70-GFP and ScTom70 are susceptible to trypsin without mitochondrial perturbation, whereas Cyb2 is degraded only with osmotic disruption of the outer membrane. (B) Sodium carbonate (pH 11)-treated mitochondrial membranes subsequently pelleted (P) from solubilized proteins (S), Western blotted, and immunostained with either GFP, ScTom70, mtHsp70, or porin antisera. EcTom70-GFP, ScTom70, and membrane protein porin were enriched with membranes (P), and matrix protein mtHsp70 was enriched with the supernatant (S).

EcTom70 does not complement ScTom70/Tom71 knockouts.

Though EcTom70 is assembled correctly in the mitochondrial outer membrane, it cannot replace the function of ScTom70. S. cerevisiae contains two paralogues of Tom70 (ScTom70 and ScTom71), and so the Δtom70/tom71 strain was used for complementation tests (11). These cells show a strong growth defect at 37°C on nonfermentable media (YPEG). When EcTom70 was expressed in Δtom70/tom71 cells to test for complementation, these cells showed growth equivalent to that of the Δtom70/tom71 cells under all growth conditions (Fig. 4).

FIG. 4.

Serial dilution growth assay of wild-type and ΔScTom70/Tom71 yeast cells transformed with plasmids expressing EcTom70 and ScTom70. Cells were incubated at 37°C on YPEG plates for 2 to 4 days.

Further Tom/Tim homologues identified from E. cuniculi genomic data.

Given that microsporidia are most closely related to fungi and therefore diverged within the fungal-animal lineage, the ancestral microsporidia most likely possessed the common Tom and Tim proteins seen in both fungi and animals today. However, previous sequence similarity searches (BLAST) identified putative homologues of only three such proteins in E. cuniculi—Tom70, Tom40, and Tim22 (10, 28). Since many TOM and TIM proteins show weak conservation of primary sequence but greater conservation of structural features, HMM searches have proven a superior tool for identifying diverse TOM/TIM homologues in other lineages (15). Therefore, HMM searches were employed to comprehensively screen the E. cuniculi genomic data for candidate proteins of the mitochondrial protein import machinery. HMMs were generated for 26 known homologues of mitochondrial import proteins, focusing on those proteins characterized in fungi (see Materials and Methods).

Even using 26 HMMs, only seven candidate proteins were identified from E. cuniculi with significant matches. These include recovery of Tom70, Tom40, and Tim22 and the mitochondrial matrix chaperone Hsp70. In addition to these previously annotated gene products, candidates were identified for an essential component of the TIM23 complex in the mitochondrial inner membrane, Tim50; a J protein of the PAM complex, Pam16; and the major component of the SAM complex, Sam50. While some of the smaller proteins (e.g., the small Toms and tiny Tims) may be difficult to recognize because of the relative simplicity of these short sequences, the absence of matches for many of the import proteins is conspicuous given that they are readily recovered from other fungal and animal genomes. For example, Tom20 and Tom22 are conserved proteins that occur in all fungi and animals (34, 35). To validate the negative outcomes in E. cuniculi, we searched the UniProt database with the Tom20 and Tom22 models and recovered numerous strong matches to new sequences from diverse fungal and animal taxa (e.g., for Tom20 and Tom22 matches were found in basidomycete fungus Ustilago maydis [9.20e⁻⁵⁴ and 4.20e⁻²⁵, respectively], the starfish Nematostella vectensis [1.50e⁻⁵⁶ and 1.70e⁻³⁴, respectively], and even the ancestral lineage of choanoflagellate Monosiga brevicollis [1.20e⁻¹⁷ and 8.40e⁻¹³, respectively]). Thus, failure to identify any candidates from E. cuniculi likely indicates their absence and indicates that the import apparatus for mitochondrial proteins in E. cuniculi is heavily reduced in complexity (Fig. 5).

DISCUSSION

The discovery in anaerobic eukaryotes of mitosomes that represent organelles derived from mitochondria provides a fascinating opportunity to examine reductive evolution of an organelle (49). The mitosomes of microsporidia provide an especially useful model due to microsporidia having diverged from within the fungal and animal lineages, both of which serve as well-studied models for mitochondrial biogenesis and biology. Organelle biogenesis is heavily dependent on the import of proteins encoded in the nucleus and translated in the cytoplasm. In the case of mitosomes, which now lack a genome, they are wholly dependent on this process. Common features of this organelle protein import machinery shared by fungi and animals allow us to predict the features of this machinery that likely occurred in the ancestral mitochondria of microsporidia. Thus, from examination of the existing machinery in microsporidia we can infer the nature of any change that has occurred during organelle reduction. Given that the microsporidian E. cuniculi has demonstrated a very strong trend for genomic and biochemical reduction and that the function and proteome of the mitosome are also apparently heavily reduced (28), we speculate that the protein import machinery of the mitosome might also show great change.

Modification of E. cuniculi Tom70.

The protein receptors of the TOM complex are the first points of discrimination for proteins to be imported to mitochondria (5, 36). Tom70 serves as one of the two major receptors in fungi and animals, with a dual binding capacity for mitochondrial preproteins and, independently, molecular chaperones bound to these preproteins (9, 11, 52). A crystal structure of the Tom70 from yeast shows that TPRs are central to this structure and assemble as two semi-independent domains (51). Protein sequence similarity was used to identify a putative Tom70 homologue from E. cuniculi genomic data (28), and we have substantiated this identification by screening the E. cuniculi data with a Tom70 HMM and recovering “EcTom70” with a high confidence value. Targeting of heterologously expressed EcTom70 to the yeast mitochondrion, where it correctly inserts as an integral protein of the outer membrane, provides further strong evidence that this is a genuine Tom70 homologue and that it likely functions in the microsporidian mitosome. Moreover, it indicates that the requirements for targeting proteins to the outer membrane and membrane insertion are conserved between fungi and microsporidia.

The reduced size of EcTom70 compared to that of yeast is consistent with overall shortening of proteins observed in E. cuniculi (28). The elimination of the two linker regions between the structurally conserved domains in EcTom70 likely reflects preservation of protein functional regions under a strong pressure to reduce size. The further loss of one of eight TPRs of the core domain might also reflect a tolerable loss without affecting the function of this region in binding to mitosomal preproteins. Despite these losses, all other major structural features identified in the yeast Tom70 structure appear to have been conserved. Lack of function of EcTom70 in the yeast TOM complex might reflect divergence of this more compact Tom70 and loss of EEVD-mediated chaperone binding. We note, however, that human Tom70, which can bind the EEVD sequences, is also unable to complement yeast Tom70 mutants (52).

Loss of TOM/TIM homologues.

Our use of HMM to search E. cuniculi data shows that candidates for Tom70, Tom40, and Tim22 are identified with high confidence, and additional new components of the TIM complex (Tim50 and Pam16) and the SAM complex (Sam50) were discovered. These results offer further evidence of a conserved import system, albeit in relatively minimal form (Fig. 5).

Conspicuously absent from the predicted import machinery in E. cuniculi are the other outer membrane receptors Tom20 and Tom22. Tom20 is specific to the fungus-animal lineage, is a well-conserved protein throughout this group, and is reliably recovered with HMMs throughout fungal and animal lineages (34). Tom20 is chiefly responsible for recognizing presequences at the N terminus of many mitochondrial proteins. Tom22 assists the function of Tom20, passing precursor proteins on to the translocation pore, Tom40 (reviewed in reference 27). Tom22 is widely conserved throughout eukaryotes and is reliably recovered with similarity searches (35). Tom70, on the other hand, has a greater role in importing membrane proteins, since it binds hydrophobic stretches, e.g., those characteristic of mitochondrial transmembrane solute carrier proteins. Microsporidian mitosomal proteins have shown a tendency for loss of N-terminal targeting extensions and a greater reliance of internal signals for targeting, based on heterologous targeting to yeast mitochondria (10). It is conceivable that EcTom70 is sufficient for recognition of the remnant mitosomal proteins and facilitates translocation through the EcTom40 channel.

A further reduction of the mitosomal import machinery is indicated by identification of only one member of the Tim23/Tim17/Tim22 family of proteins. In E. cuniculi, only the HMM representing Tim22 identified a homologue, and therefore this protein is designated EcTim22. Ancient duplications of the inner membrane pore protein are represented widely throughout eukaryotes by paralogues Tim22, Tim23, and Tim17 (15, 43). In yeast and other fungi and in animals Tim23 and Tim17 together contribute to presequence-mediated import across the inner membrane, while Tim22 contributes to a separate complex dedicated to insertion of membrane proteins such as the carrier proteins of the inner membrane (Fig. 5) (5, 36). The specialization of two TIM complexes early in eukaryotic evolution enabled diverse mitochondrial protein traffic. The presence of only a single inner membrane pore in E. cuniculi suggests that microsporidia have reversed this specialization in the face of reduced protein traffic and overall cellular reduction. We suggest that the protein called EcTim22 is probably assisted by EcTim50 to form a TIM23 complex: Tim50 both regulates the inner membrane barrier and binds to emergent proteins from the TOM complex and hence is essential for TIM function. A single J-protein, EcPam16, is likely responsible for regulation of the ATP-dependent role of mtHsp70 during final passage across the inner membrane (Fig. 5). As is the case in trypanosomatids (43), the single TIM complex in E. cuniculi mitosomes might drive translocation of proteins into the matrix and assembly of inner membrane proteins, too.

Identification of a Sam50 homologue, for insertion of β-barrel proteins (e.g., Tom40), reflects the essential function of the SAM complex in mitosomes, as in mitochondria (5, 36). It is unclear whether failure to identify further SAM complex proteins (Sam35, Sam37, and Mdm10) is due to poor conservation of these proteins or to a capacity of Sam50 to act alone. Another lone complex component is Erv1, of the MIA (mitochondrial intermembrane space import and assembly) machinery (5, 36). Erv1, identified as a likely mitosomal protein in the genome annotation of E. cuniculi, is implicated in promoting the sequential formation of intramolecular disulfides in imported intermembrane space proteins. These molecules provide further evidence of relicts of a mitochondrial-type protein import system.

The skeletal form of the mitosomal protein import machinery identified here might reflect the difficulty in identifying homologues of many of the import proteins of fungi and animals, particularly several of the small proteins. Presently few genomic data exist for microsporidia aside from E. cuniculi, limiting the opportunity to look more broadly for mitosomal homologues in this group. It is possible, however, that our observations reflect an import machinery that has been reduced in response to dramatic reductionism seen throughout microsporidian biology. It is notable that, of the subset of import machinery that the HMMs have identified, all major essential functions are represented, although only once rather than in duplicate specialist form as seen in fungi and animals (Fig. 5). Reduction and change in EcTom70 and apparent loss of Tom20 and one of the inner membrane pore complexes are all consistent with conversion to a minimal apparatus. Moreover, use of Tom70 as the principal outer membrane receptor is supported by the loss of presequences from many mitosomal proteins and a greater role of internal signals for organelle import. If such changes hold across microsporidia, these insights offer new approaches to tackling microsporidia as human and animal pathogens. A peptide (ending in EEVQ) that specifically targets and blocks the EcTom70 clamp domain, for instance, could offer scope for perturbing mitosomal protein import as an antimicrosporidial chemotherapeutic strategy.