All extant eukaryotes are now considered to possess mitochondria in one form or another. Many parasites or anaerobic protists have highly reduced versions of mitochondria, which have generally lost their genome and the capacity to generate ATP through oxidative phosphorylation. These organelles have been called hydrogenosomes, when they make hydrogen, or remnant mitochondria or mitosomes when their functions were cryptic. More recently, organelles with features blurring the distinction between mitochondria, hydrogenosomes and mitosomes have been identified. These organelles have retained a mitochondrial genome and include the mitochondrial-like organelle of Blastocystis and the hydrogenosome of the anaerobic ciliate Nyctotherus. Studying eukaryotic diversity from the perspective of their mitochondrial variants has yielded important insights into eukaryote molecular cell biology and evolution. These investigations are contributing to understanding the essential functions of mitochondria, defined in the broadest sense, and the limits to which reductive evolution can proceed while maintaining a viable organelle.
Some eukaryotes including members of the obligate intracellular parasitic Microsporidia (e.g. Encephalitozoon), and microaerophilic human parasites from the Amoebozoa (e.g. Entamoeba), Diplomonadida (e.g. Giardia) and Parabasalia (e.g. Trichomonas) (for taxonomy see table 1 and Adl et al. 2005), were once thought to lack mitochondria because they separated from other eukaryotes before the mitochondrial endosymbiosis. As such, they were viewed as important model systems from which to infer the early stages of eukaryotic evolution and, to signify their importance, they were grouped together in a paraphyletic taxon called the Archezoa (Cavalier-Smith 1983, 1987). Detailed molecular cell biology and phylogenetic investigations over the past 15 years have now shown that all of the best-studied archezoans contain a double-membrane-bounded organelle of mitochondrial ancestry (Embley & Martin 2006). The first of these to be discovered was the Trichomonas hydrogenosome (Lindmark & Müller 1973), so-called because it makes molecular hydrogen through the activity of iron-[Fe]-hydrogenase. It was initially thought that the Trichomonas hydrogenosome might have originated from a separate endosymbiosis between the ancestor of Trichomonas and a Clostridium that contained this biochemical pathway (Müller 1980). The localization of proteins to Trichomonas hydrogenosomes that are typical of mitochondria, including chaperonins (Bui et al. 1996), the NADH-dehydrogenase module of complex I (Hrdy et al. 2004) and components of the mitochondrial iron–sulphur (Fe/S) cluster (ISC) biosynthesis pathway (Sutak et al. 2004), is now taken to support the simpler hypothesis that the hydrogenosomes of trichomonads are related to mitochondria (Embley & Martin 2006).
A similar approach, namely the immuno-localization of mitochondrial chaperonin 60 (Cpn60), was used to discover a tiny double-membrane-bounded organelle of cryptic function in Entamoeba histolytica, which was called a mitosome (Clark & Roger 1995; Tovar et al. 1999), or crypton (Mai et al. 1999), with the former name used most commonly. Mitosomes were subsequently discovered in the microsporidian Trachipleistophora hominis using the immuno-localization of mitochondrial Hsp70 (mtHsp70) (Williams et al. 2002) and in Giardia by immuno-localization of components of the mitochondrial ISC biosynthesis pathway (Tovar et al. 2003). Table 1 lists all of the experimentally localized proteins for the mitochondrial homologues discussed in this review and also incorporates current ideas (Adl et al. 2005) about the taxonomy of the diverse host organisms. The molecular cell biology investigations have been complemented by improved phylogenetic analyses to resolve the relationships of the former archezoans. These analyses have identified mitochondria-containing sister groups for Entamoeba and Microsporidia, in Dictyostelium and fungi, respectively, and have suggested that the deep branching positions of Giardia and Trichomonas are incorrect, or at best poorly supported (Embley & Martin 2006).
In summary, the prevailing view is that the Archezoa hypothesis has been falsified for the key taxa for which it was originally proposed. The discovery that all eukaryotes appear to contain an organelle related to mitochondria suggests that mitochondrial homologues are vital and defining features of eukaryotes. In that sense, the former Archezoa have played a central role in recasting hypotheses for eukaryogenesis (Embley & Martin 2006), and they have provided seminal examples of the diversity of mitochondrial form and function under different lifestyles. It is also now apparent that similarly diverse mitochondrial homologues can be found in other free-living and parasitic protists and fungi that were never grouped with archezoans.
2. Mitochondrial homologues: diverse in form and function
Hydrogenosomes, defined minimally as double-membrane-bounded organelles that make hydrogen, have also been described in diverse anaerobic ciliates and in anaerobic chytrid fungi isolated from the guts of herbivores such as Piromyces sp. and Neocallimastix sp. (Embley et al. 2003; Hackstein et al. 2008a). Among ciliates, anaerobic species with hydrogenosomes interleave among aerobes with mitochondria, suggesting that for this group at least, the transition from mitochondria to hydrogenosome is an easy one to make (Embley et al. 1995). Anaerobic ciliates with hydrogenosomes have been isolated from the cockroach gut (Nyctotherus ovalis), the gut of herbivores (Dasytricha ruminantium) and marine (Plagiopyla frontata, Metopus contortus) and fresh water sediments (Trimyema compressum, Trimyema sp.) (Fenchel & Finlay 1995; Boxma et al. 2005). In these habitats, they often harbour intracellular methanogenic archaebacteria that use the hydrogen produced by the ciliate host hydrogenosomes to reduce CO2 and to make ATP (Embley & Finlay 1994; Fenchel & Finlay 1995).
The case of the anaerobic ciliate N. ovalis is of particular interest, as its hydrogenosomes still retain a mitochondrial genome which has been partially sequenced (Boxma et al. 2005), providing a direct link between the two types of organelle. The partial hydrogenosomal genome encodes some of the genes for complex I, as well as ribosomal (r)RNA, tRNA and ribosomal proteins: consistent with the retention of a capacity for organelle protein synthesis (Boxma et al. 2005). Other components of mitochondrial complex I and complex II are encoded by the nuclear genome, suggesting that the hydrogenosomes of Nyctotherus may have retained a partial electron transport chain (ETC), consistent with the detected membrane potential (Boxma et al. 2005). The [Fe]-hydrogenase of Nyctotherus is probably nuclear-encoded since it has a mitochondrial-like targeting signal. The gene itself was initially described as a fusion protein comprising an [Fe]-hydrogenase linked to the 24 and 51 kD subunits of mitochondrial complex I (Hackstein et al. 1999), but subsequent phylogenetic analyses provided no support for this hypothesis (Horner et al. 2000). In most analyses, the Nyctotherus 51 kD subunit clustered with the 51 kDa-like component of the NAD(H)-dependent [Ni–Fe] hydrogenases from bacteria rather than complex I-associated sequences (Horner et al. 2000). Phylogenetic analyses do provide weak evidence that the Nyctotherus [Fe]-hydrogenase component of the fusion protein might have a separate origin to other eukaryotic [Fe]-hydrogenases (Horner et al. 2000; Embley et al. 2003; Hampl et al. 2008). Interestingly, when the [Fe]-hydrogenase of the fresh water anaerobic ciliate Trimyema sp. was analysed, it was found to form a monophyletic group with the Nyctotherus sequence (Embley et al. 2003). This suggests that a gene for [Fe]-hydrogenase was already present in the common ancestor of these disparate ciliate lineages, helping to explain the apparent facility whereby different anaerobic ciliates have been able to make hydrogenosomes from mitochondria (Embley et al. 1995).
Mitosomes are double-membrane-bounded organelles that have so far been found in parasitic members of the lineages Amoebozoa, Microsporidia, Diplomonads and Apicomplexa (Tovar et al. 1999; Williams et al. 2002; Tovar et al. 2003; Keithly et al. 2005). Based upon the published biochemical studies over many years (Müller 1988, 2003) and more recent genomic studies (Katinka et al. 2001; Abrahamsen et al. 2004; Loftus et al. 2005; Morrison et al. 2007), it appears unlikely that mitosomes have a role in cellular ATP generation. Indeed, mitosomes are largely defined by what they lack; they are remnant mitochondria that lack ATP-generating pathways and do not make hydrogen. Mitosomes are morphologically simple and lack cristae and are typically smaller than mitochondria, varying in size from approximately 50 nm to approximately 500 nm in diameter depending on the species (Scheffler 2008; Tachezy & Smid 2008). Their small size and simple morphology along with the absence of mitochondrial marker enzymes undoubtedly contributed to past failures to recognize mitosomes as mitochondrial homologues. Interestingly, however, there are references to mitochondrial-like organelles lacking cristae for Microsporidia and for Giardia lamblia, in the historical literature (Cheissin 1965; Vávra 1976).
The number of mitosomes within each cell also varies between species. Cryptosporidium parvum contains a single mitosome (Keithly et al. 2005) but E. histolytica (Leon-Avila & Tovar 2004) and Giardia (Regoes et al. 2005) may contain between 25 to 150 mitosomes per cell (Tachezy & Smid 2008). An average cell of the microsporidian T. hominis contains around 28 mitosomes (Williams et al. 2002) but its smaller relative Encephalitozoon cuniculi, typically contains less than 10 mitosomes per cell (Tsaousis et al. 2008). Very little is known about how these organelles are inherited during host cell division (Regoes et al. 2005; Tsaousis et al. 2008), but, given the small number of mitosomes in species such as C. parvum and E. cuniculi, a mechanism must exist to ensure the reliable segregation of mitosomes between daughter cells.
Blastocystis is an anaerobic parasite commonly found in the human gut that belongs to a diverse eukaryotic group called the stramenopiles or heterokonts, which also includes the brown algae, diatoms and oomycetes (Perez-Brocal & Clark 2008). The mitochondrion-like organelles (MLO) of Blastocystis spp. resemble the hydrogenosomes of the anaerobic ciliate Nyctotherus hydrogenosome, in that they have retained an organelle genome (Perez-Brocal & Clark 2008; Stechmann et al. 2008; Wawrzyniak et al. 2008). The characterized circular-mapping MLO genomes (Perez-Brocal & Clark 2008; Wawrzyniak et al. 2008) are DNA molecules of between 27–29 kbp that are A + T rich (approx. 80%) and gene dense. Each contains 45 genes of which 27 are open reading frames, including genes for ribosomal proteins and some subunits of mitochondrial complex I. The rest of the genes encode structural ribosomal RNAs and tRNAs (Perez-Brocal & Clark 2008; Wawrzyniak et al. 2008). These data, in combination with EST data (Stechmann et al. 2008) have been interpreted (Perez-Brocal & Clark 2008) to suggest that a proton-pumping complex I may function in the Blastocystis MLO: this would be consistent with the previously detected organelle membrane potential (Zierdt et al. 1988; Hamblin et al. 2008). By contrast, and unlike the mitochondrial genomes of related stramenopiles, the MLO genomes lack genes for cytochromes and ATPase subunits, suggesting that Blastocystis lacks mitochondrial complexes III and IV (Zierdt et al. 1988; Perez-Brocal & Clark 2008).
Blastocystis also has an iron-[Fe]-hydrogenase of the same type as used by Trichomonas hydrogenosomes to make hydrogen (Stechmann et al. 2008). Antiserum to the Blastocystis [Fe]-hydrogenase was used to localize the protein to the Blastocystis MLO (Stechmann et al. 2008), suggesting that it resembles the Nyctotherus hydrogenosome in this feature too. However, in a recent biochemical study of another strain of Blastocystis (Lantsman et al. 2008), no hydrogenase activity could be detected in fractions enriched for the MLO or in cell-free extracts. The reasons for this discrepancy are unknown: it could be down to strain differences, to differences in growth conditions or simply that detecting hydrogen production by Blastocystis will need a more sensitive assay (Lantsman et al. 2008). For example, mass spectrometry was needed to detect Giardia cytosolic hydrogen production (Lloyd et al. 2002), which is typically 10-fold less than that produced by Trichomonas under the same conditions. As a consequence of the failure to detect hydrogen production, the Blastocystis organelle is generally referred to as an MLO rather than as a hydrogenosome (Lantsman et al. 2008).
The examples discussed above illustrate that reductive evolution of mitochondria to MLO, hydrogenosome or mitosome is common across the eukaryotic tree. Some of this reductive evolution is convergent, the occurrence of multiple anaerobic ciliate lineages with hydrogenosomes interleaved among aerobic lineages with mitochondria (Embley et al. 1995), and the similarities between the Nyctotherus (Boxma et al. 2005) and Blastocystis organelles (Perez-Brocal & Clark 2008; Stechmann et al. 2008) provide particularly striking examples of this trend. The organelles described so far are probably just the tip of a very large iceberg of mitochondrial homologue diversity that remains to be discovered. Intracellular methanogens have been described in a variety of otherwise poorly characterized anaerobic protists, suggesting that hydrogenosomes may be relatively common in nature (Fenchel & Finlay 1995). The protozoological and microbial ecology literature is also littered with descriptions of protists living in low-oxygen environments, and which contain uncharacterized double-membrane-bounded organelles (Fenchel & Finlay 1995; Hampl & Simpson 2008).
3. All mitochondrial homologues need to make or import atp
Canonical mitochondria can be classified into two different types: aerobic mitochondria, which use oxygen as the final acceptor during aerobic respiration, and anaerobic mitochondria, present in facultative anaerobes, which can use organic or inorganic compounds, such as nitrate or fumarate, as final acceptors during anaerobic respiration (Tielens et al. 2002). Both pathways make ATP for the organelle to sustain its own metabolism, to import proteins and to import and export solutes. The excess of ATP produced by mitochondria is exported to the cytosol using members of the mitochondrial carrier family (MCF) (Kunji 2004). Some ATP is generated by the Trichomonas hydrogenosome by substrate level phosphorylation involving acetyl CoA released by decarboxylation of pyruvate by the enzyme pyruvate:ferredoxin oxidoreducatase (PFOR) (Hrdy et al. 2008). The MLO of Blastocystis probably makes ATP in the same way, but using the PFOR homologue pyruvate: NADP(+) oxidoreductase, to decarboxylate pyruvate (Hamblin et al. 2008; Lantsman et al. 2008). Based upon biochemical studies (Reeves 1984; Müller 1988, 2003) and genomic sequences (Katinka et al. 2001; Abrahamsen et al. 2004; Loftus et al. 2005; Morrison et al. 2007), it appears unlikely that the mitosomes of Microsporidia, Giardia, Entamoeba and Cryptosporidium can make ATP. However, at least according to the best studied mitochondrial models, mitosomes must require ATP to import nuclear encoded proteins and to support the activities of their Cpn60 and mtHsp70 ATPases (Tovar et al. 1999; Williams et al. 2002; Riordan et al. 2003; Regoes et al. 2005; Tovar et al. 2007; Tsaousis et al. 2008). So how do these organelles acquire the ATP they need?
(a) Mitochondrial ATP transport proteins
The genomes of eukaryotes with canonical mitochondria typically contain between 35 and 55 members of the MCF family of transport proteins (Kunji 2004). For example, the human genome codes for about 48 MCF proteins, about half of which have been functionally characterized. MCF proteins support mitochondrial metabolic energy generation, DNA replication and amino acid metabolism by linking biochemical pathways in the mitochondrial matrix with those in the cytosol (Kunji 2004). The number and diversity of MCF proteins thus closely mirrors important facets of mitochondrial metabolic diversity, and the functional characterization of hydrogenosomal and mitosomal MCF proteins is an important part of investigating how these organelles function. A specialized member of the MCF, the mitochondrial ADP/ATP carrier (AAC), located in the inner mitochondrial membrane, exchanges mitochondrial-generated ATP with cytosolic ADP on an equimolar basis—so the adenine nucleotide pool of the mitochondrial matrix does not change (Kunji 2004). It is also thought to be an MCF protein, possibly an ATP-Mg/Pi carrier, that replenishes adenine nucleotide pools during mitochondrial division (Kunji 2004). In the classical endosymbiosis theory, insertion of a host-nuclear-encoded MCF ATP/ADP transporter into the protomitochondrion is viewed as a key step that allowed the host cell to harvest ATP from the enslaved endosymbiont (John & Whatley 1975).
The hydrogenosomes of the anaerobic ciliate N. ovalis (Hackstein et al. 2008b) and the anaerobic chytrid fungus Neocallimastix (van der Giezen et al. 2002; Voncken et al. 2002) contain functionally characterized mitochondrial-type AAC. These proteins share strong structural and functional similarities to mitochondrial AAC and cluster with AAC in phylogenetic trees (van der Giezen et al. 2002; Hackstein et al. 2008b). The Neocallimastix AAC was able to functionally complement a yeast mitochondrial mutant showing conservation of function and targeting signals. Moreover, when hydrogenosomal membranes were fused with liposomes, ADP/ATP transport was inhibited by bongkrekic acid, a known inhibitor of mitochondrial ADP/ATP transporters (van der Giezen et al. 2002). The Trichomonas vaginalis genome contains five genes annotated as coding for MCF proteins, none of which cluster with the classic AAC (Carlton et al. 2007). One of these genes codes for an abundant protein, named Hmp31 (for hydrogenosomal membrane protein), in the hydrogenosome-enriched membrane fraction of T. vaginalis (Dyall et al. 2000). Hmp31 was targeted to the inner membrane of Saccharomyces cerevisiae mitochondria as measured using radiolabelled Hmp31 in isolated mitochondria (Dyall et al. 2000). An Hmp31 homologue from Trichomonas gallinae was also shown to localize to the hydrogenosomal membrane fraction (Tjaden et al. 2004). The T. gallinae Hmp31 protein showed substrate specificity and comparable kinetic properties to mitochondrial-type AAC from mitochondria and fungal hydrogenosomes (Tjaden et al. 2004). However, unlike the classic AAC, the T. gallinae Hmp31 protein was not sensitive to bongkrekic acid (Tjaden et al. 2004). Several ESTs from a Blastocystis cDNA library appear to code for putative MCF proteins, but none of these have been localized or functionally characterized (Stechmann et al. 2008).
The hydrogenosomes of Trichomonas (Müller 2003) and the Blastocystis MLO (Hamblin et al. 2008; Lantsman et al. 2008) can generate ATP by substrate level phosphorylation, but mitosomes appear to depend on the cytosol for the ATP and other solutes that they need to function. Relatively few proteins have been experimentally localized to mitosomes so far, but all mitosomes appear to contain at least one protein, typically either Cpn60 or Hsp70 or both, that requires ATP for its activity (Embley & Martin 2006). In principle, an MCF AAC could operate in reverse along a concentration gradient to exchange mitosomal ADP for cytosolic ATP (Kunji 2004). The genome of C. parvum contains genes for MCF proteins annotated as ADP/ATP transporters (Abrahamsen et al. 2004), but these have not been functionally characterized or localized to the organelle. The E. histolytica mitosome contains a single type of MCF protein that, in principle, could be used to fuel its mitosome. The Entamoeba MCF protein is resistant to bongkrekic acid and it transports ATP and ADP using a novel mechanism that does not appear to require a membrane potential (Chan et al. 2005). Phylogenetic analyses confirmed that the Entamoeba AAC is distinct from archetypal mitochondrial AACs (Chan et al. 2005).
A gene for an MCF protein was found in an EST library for the microsporidian Antonospora locustae (Williams et al. 2008b). The Antonospora MCF was targeted to the mitochondria of Saccharomyces, suggesting conservation of targeting signals, and it showed high specificity for ATP and ADP when expressed in Escherichia coli (Williams et al. 2008b). Although the Antonospora ATP/ADP transporter was not shown to actually localize to a mitosome in Antonospora, the data are consistent with at least one microsporidian retaining an MCF. By contrast, the genome of the microsporidian E. cuniculi lacks any genes for MCF proteins (Katinka et al. 2001). It does, however, contain four genes for bacterial-type nucleotide transporters (named EcNTT1-4) similar to those used by intracellular pathogens, including Rickettsia and Chlamydia, to steal ATP from their host cells (Winkler & Neuhaus 1999). As proliferating microsporidia recruit host mitochondria near their plasma membrane, it has previously been proposed that microsporidians use host-derived ATP to supplement their energy budget (Weidner et al. 1999). The presence in E. cuniculi of NTT homologues provides a potential means of achieving this goal (Katinka et al. 2001). Consistent with this hypothesis, three of the EcNTT were expressed on the surface of the parasite when living inside rabbit kidney cells (Tsaousis et al. 2008). The fourth transporter (EcNTT3) colocalized with the mitosomal marker protein mtHsp70, consistent with it being targeted to E. cuniculi mitosomes (Tsaousis et al. 2008) (figure 1). Thus, the E. cuniculi mitosome appears to be using the same strategy as the parasite itself to acquire ATP. Heterologous expression in E. coli demonstrated that all four EcNTT were able to transport ADP and ATP in contrast to bacterial pathogens where typically only one paralogue will mediate ATP uptake (Audia & Winkler 2006; Haferkamp et al. 2006). Adenine nucleotide efflux (figure 1) was observed for EcNTT3 when the external substrate was removed, indicating that it is able to equilibrate nucleotide pools across a concentration gradient. Efflux was stimulated by the addition of external cold ATP or ADP, confirming that EcNTT3 is an exchanger of adenine nucleotides and thus potentially capable of importing cytosolic ATP, in exchange for mitosomal ADP, into the E. cuniculi mitosome (Tsaousis et al. 2008).
The genome of the diplomonad parasite Giardia (Morrison et al. 2007) appears to lack any genes for MCF proteins or NTT-like transporters, so how its mitosomes acquire the ATP needed for ISC assembly (Tovar et al. 2003) and protein import (Dolezal et al. 2005; Regoes et al. 2005; Šmíd et al. 2008) is unknown.
4. Protein targetting and protein import
(a) Presence of mitochondrial-like targeting signals
Even in species that have a mitochondrion with an associated organelle genome, the vast majority of organelle proteins are encoded by the host nuclear genome and must be targeted to the mitochondrion and its individual compartments using specific targeting signals (see also Lithgow & Schneider 2009). Many of the nuclear-encoded mitochondrial matrix proteins are synthesized with N-terminal cleavable leader sequences that are recognized by receptors on the mitochondrial surface (Neupert & Herrmann 2007). Typical pre-sequences contain 10 to 80 amino acid residues, many of which are positively charged, hydrophobic and hydroxylated (Dyall & Dolezal 2008). The N-terminus may often form an amphiphilic α-helix (Pfanner & Geissler 2001). Many mitochondrial pre-sequences contain an arginine residue at the −2 position from the cleavage site, and this is also a common feature of the relatively few, mostly for Trichomonas but also for Entamoeba, Giardia and Neocallimastix, experimentally verified N-terminal cleaved pre-sequences on hydrogenosomal or mitosomal proteins (but see Smutna et al. 2009). The targeting sequences of hydrogenosomal or mitosomal proteins are generally shorter than the mitochondrial versions: most predicted Giardia mitosomal pre-sequences and a majority of Trichomonas hydrogenosomal pre-sequences have a length of between four and 21 amino acids (Šmíd et al. 2008). A full list of experimentally verified, or predicted, hydrogenosomal or mitosomal targeting sequences is given by Dyall & Dolezal (2008) and Šmíd et al. (2008). The positive charges on mitochondrial pre-sequences are drawn through the TIM23 inner membrane channel aided by the membrane potential (Dyall & Dolezal 2008), after which the pre-sequence is normally cleaved by a mitochondrial-processing peptidase (MPP) (Šmíd et al. 2008). After cleavage, the protein is refolded into its native structure aided by the Cpn60/Cpn10 complex (Dyall & Dolezal 2008).
The ‘classic’ MPP typically consists of two subunits: an α-subunit implicated in substrate binding and release and a β-subunit that possesses catalytic activity (Šmíd et al. 2008). The processing peptidases of Giardia and Trichomonas have recently been functionally characterized (Brown et al. 2007; Šmíd et al. 2008). The Giardia MPP is so far unique in that it lacks the α-subunit and functions as a monomeric β-subunit (βGPP) (Šmíd et al. 2008). The recombinant Giardia βGPP was able to process in vitro the N-terminal extensions of Giardia mitosomal ferredoxin and of two scaffold proteins (IscU and IscA) involved in mitosomal ISC biosynthesis (Šmíd et al. 2008). The MPP of T. vaginalis was also initially thought to function as a single β-subunit MPP (Brown et al. 2007), but subsequent work has demonstrated that it functions most efficiently as a heterodimer comprising a divergent α-subunit and a β-subunit (Šmíd et al. 2008). The T. vaginalis MPP processed the N-terminal extensions of Trichomonas hydrogenosomal ferredoxin, adenylate kinase, Hsp70 and IscU.
Targeting signals for other types of mitochondrial proteins, for example for members of the MCF targeted to the inner mitochondrial membrane, are less well characterized. These do not use an N-terminal cleaved leader sequence but instead rely on internal targeting signals to guide them to their final location (Pfanner & Geissler 2001). Intriguingly, the T. vaginalis MCF protein Hmp31 does have a cleavable 12 amino acid N-terminal extension, but this is not necessary for targeting Hmp31 to the hydrogenosomal membrane (Dyall et al. 2000). Trichomonas Hmp31 was efficiently targeted to the inner membrane of yeast mitochondria, and, in the reciprocal experiment, hydrogenosomes were able to import yeast AAC (Dyall et al. 2000). These data strongly suggest that elements of the inner membrane protein-targeting machinery are conserved between yeast mitochondria and Trichomonas hydrogenosomes. Mentel et al. (2008) recently demonstrated that other hydrogenosomal proteins possess internal targeting signals. One of the two hydrogenosomal thioredoxin reductase (TrxR) isoforms, TrxRh1, carried an N-terminal extension resembling known hydrogenosomal targeting signals and for this protein the N-terminal extension was necessary to import the protein into hydrogenosomes (Mentel et al. 2008). The second hydrogenosomal TrxR isoform, TrxRh2, had no N-terminal targeting signal but was nonetheless also efficiently targeted to hydrogenosomes. Moreover, the α-subunit of succinyl CoA synthetase, lacking its normal N-terminal extension, was still efficiently imported into hydrogenosomes, indicating that this extension is not required for import of this major hydrogenosomal protein. These data indicate the presence of additional targeting signals within the mature subunits of several hydrogenosome-localized proteins (Mentel et al. 2008).
Analysis of an EST library for Blastocystis also revealed a large number of putative proteins with N-terminal extensions, indicative of mitochondrial-like targeting pre-sequences (Stechmann et al. 2008). Two of these proteins: [Fe]-hydrogenase (Stechmann et al. 2008) and the α-subunit of succinyl CoA synthetase (Hamblin et al. 2008) localized to the MLO of Blastocystis. Based upon the EST library, Blastocystis also has a putative MPP (Stechmann et al. 2008). An EST library for the anaerobic ciliate N. ovalis contained sequences with putative N-terminal pre-sequences for proteins that typically function in mitochondria, such as pyruvate dehydrogenase (Boxma et al. 2005). The N-terminal sequences of C. parvum mtHsp70, IscU and IscS can direct green fluorescent protein to yeast mitochondria (LaGier et al. 2003; Slapeta & Keithly 2004), and mtHsp70 and Cpn60 are known to localize to the Cryptosporidium mitosome (Riordan et al. 2003). The genome of Cryptosporidium (Abrahamsen et al. 2004) contains putative α- and β-subunits of an MPP, but these have not been functionally characterized.
Mitochondrial-like pre-sequences have also been detected on E. histolytica pyridine nucleotide transhydrogenase, Cpn60 and mtHsp70 (Tovar et al. 1999; Bakatselou et al. 2003). The targeting signal of Cpn60 was shown to be required for targeting the protein into the Entamoeba mitosome. Deletion of the first 15 amino acids of Cpn60 led to an accumulation of the truncated protein in the cytoplasm of Entamoeba. This mutant phenotype was reversed by the replacement of the deleted amino acids with a mitochondrial targeting signal from Trypanosoma cruzi Hsp70 (Tovar et al. 1999). Consistent with the cleavage of the leader sequence form Cpn60, there is evidence for a putative MPP on the Entamoeba genome sequence (Loftus et al. 2005). When the Entamoeba MCF was expressed in yeast, it was targeted to the inner membrane of mitochondria, suggesting that it retains targeting signals recognized by the yeast inner mitochondrial membrane import pathway (Chan et al. 2005).
The first protein shown to be targeted to a microsporidian mitosome, T. hominis mtHsp70, lacks any detectable N-terminal targeting sequence, and it was not targeted to mammalian mitochondria (Williams et al. 2002). Other proteins, experimentally verified to localize to T. hominis or E. cuniculi mitosomes, namely T. hominis cysteine desulphurase (abbreviated Nfs or IscS), and E. cuniculi Hsp70, IscS, IscU and frataxin, also lack detectable N-terminal targeting signals (Goldberg et al. 2008; Tsaousis et al. 2008). Consistent with the absence of cleaved leader sequences on E. cuniculi mitosomal proteins, its genome does not contain genes for an MPP (Katinka et al. 2001). It appears from these data that E. cuniculi has dispensed with cleaved N-terminal leader sequence-based targeting for mitosomal matrix proteins, and relies on internal signals to guide these proteins to their destination.
In yeast, some proteins, for example mitochondrial glycerol-3-phosphate dehydrogenase (G3PDH), destined for the inter-membrane space carry an N-terminal leader sequence that is cleaved by another type of protease: the inner membrane peptidase (IMP) (Burri et al. 2006). This peptidase consists of two catalytic subunits (Imp1 and Imp2) and a regulator (Som1). Antonospora protein extracts appear to contain two sizes of G3PDH, consistent with the processing of a 40 amino acid N-terminal extension (Burri et al. 2006). By contrast, E. cuniculi G3PDH does not appear to be processed. Consistent with this processing, there is a gene (found as an EST) for an IMP (IMP2) on the A. locustae (Burri et al. 2006), but not the E. cuniculi (Katinka et al. 2001), genome. Although the location of Imp2 inside A. locustae is currently unknown, it was shown to localize to yeast mitochondria in heterologous targeting experiments (Burri et al. 2006).
(b) Conservation of mitochondrial import pathways
The machinery used by yeast mitochondria to import or insert proteins into the outer membrane, inter-membrane space, inner membrane or mitochondrial matrix is based on a number of major import complexes (Pfanner & Geissler 2001; Dyall & Dolezal 2008). The translocase of the outer mitochondrial membrane (TOM) complex provides the general import pore through the outer mitochondrial membrane. The sorting and assembly machinery (SAM) complex is involved in the correct insertion of β-barrel proteins such as Tom40 and porin into the outer membrane; Sam50 and Sam35 are essential components of the yeast SAM pathway (Dyall & Dolezal 2008). Two translocase of the inner mitochondrial membrane (TIM) complexes, namely TIM22 and TIM23, are responsible for protein translocation across, or insertion into, the inner mitochondrial membrane (Neupert & Herrmann 2007). There are additional complexes: the mitochondrial inter-membrane space import (MIA) complex, which inserts proteins into the inter-membrane space, and the OXA complex, which is mainly used for proteins encoded by mitochondrial genomes. Each of the import complexes comprises multiple components, many of which are conserved in fungi, animals and plants, suggesting that the last common ancestor of these species already had a sophisticated protein import machinery. The results of genomic sequence analyses and the small number of functional studies already suggest that species with hydrogenosomes and mitosomes have dramatically reduced the complexity of their organelle protein import machineries (Dyall & Dolezal 2008), raising intriguing questions as to how protein import functions and what are the driving forces behind its reductive evolution. In the following discussion, we focus mainly on the major outer (TOM) and inner membrane (TIM22, TIM23) import pathways, and on proteins for which there are at least some experimental data that they function in hydrogenosomes or mitosomes.
The TOM complex forms the general import pore through which proteins can cross the outer mitochondrial membrane. In yeast, it comprises the Tom40 pore and a variety of accessory proteins including a Tom20 receptor that recognizes N-terminal pre-sequences, and a Tom70 receptor that recognizes internal signal sequences. Homologues of the main Tom40 pore can be found on the fully sequenced genomes of C. parvum (Abrahamsen et al. 2004), E. cuniculi (Katinka et al. 2001), E. histolytica (Loftus et al. 2005), G. lamblia (Morrison et al. 2007) and T. vaginalis (Carlton et al. 2007), although some are not annotated as such. Tom40 shares a porin-3 motif with porin (also called the voltage-dependent anion channel), suggesting that they are ancient paralogues from the same broad ‘mitochondrial porin’ protein family (Pusnik et al. 2009). As the sequences from parasites are often very difficult to identify as specific types of ‘mitochondrial porin’, all we can say at the moment is that it seems possible (if not likely) that each species has a mitochondrial porin in the outer membrane of its mitochondrial homologue. However, experimental data will be needed to confirm this inferred cellular localization and to determine whether any of these proteins actually function as general import pores (Dagley et al. in press).
The localized E. cuniculi mitosomal proteins all lack an N-terminal leader sequence (Goldberg et al. 2008; Tsaousis et al. 2008), consistent with the absence of an MPP (Katinka et al. 2001), and suggesting that they are imported using internal targeting signals. In yeast, the Tom70 protein is part of the outer membrane machinery that recognizes proteins with internal targeting signals (Pfanner & Geissler 2001). The genome of E. cuniculi contains a gene for a putative Tom70 (Katinka et al. 2001) that can be imported into the mitochondria of S. cerevisiae (Waller et al. 2009). The EcTom70 inserted with the correct topology into the outer membrane of yeast mitochondria but was unable to complement the growth defects of Tom70-deficient yeast.
The TIM22 and TIM23 complexes provide the main inner mitochondrial membrane translocases in yeast. The TIM22 complex is used to insert inner mitochondrial membrane proteins such as the MCF AAC (Neupert & Herrmann 2007). In yeast, the passage of the substrate through the twin pores formed by Tim22 is dependent on the membrane potential (Rehling et al. 2003). Putative Tim22 orthologues have been annotated for E. cuniculi (Katinka et al. 2001) and A. locustae (Williams & Keeling 2005) and potentially could be used to insert the experimentally localized E. cuniculi mitosomal ADP/ATP transport proteins (Tsaousis et al. 2008) and the putative A. locustae mitosomal MCF ADP/ATP transporter (Williams et al. 2008b). MCF proteins have been identified and functionally characterized for Entamoeba (Chan et al. 2005) and Trichomonas (Dyall et al. 2000), including their heterologous import into yeast mitochondria, presumably via the TIM22 complex. The failure to identify Tim22 orthologues on the genomes of Entamoeba (Loftus et al. 2005) or Trichomonas (Carlton et al. 2007) is thus particularly surprising.
In yeast, the TIM23 complex is the main inner membrane import pathway for matrix or inner membrane proteins carrying N-terminal leader sequences. It comprises the paralogues Tim23 and Tim17, which are also homologous with Tim22, and a number of accessory proteins including Tim50, which regulates the opening of the Tim23/Tim17 channel, and Tim21, which regulates module docking (Dyall & Dolezal 2008). In yeast, the translocation of proteins through the TIM23 complex is ATP dependent and requires a membrane potential. Cryposporidium parvum (Abrahamsen et al. 2004) and T. vaginalis (Carlton et al. 2007) have genes for putative Tim17 and Tim23, so an inner membrane translocation pore could potentially be present for the organelles of these species. However, genes for Tim17 or Tim23 were not identified for Entamoeba (Loftus et al. 2005) or Giardia (Morrison et al. 2007), and this lack, together with the apparent absence of genes for Tim22, makes it difficult to understand how proteins are imported across the inner membrane of the mitosomes of these species.
Protein import into the matrix is driven by the PAM complex (pre-sequence translocase-associated motor), which in yeast comprises Pam16, Tim17/Pam17, Tim18/Pam18, Tim44/Pam44 and Mge1. The mtHsp70 also plays a part in this process (Pfanner & Geissler 2001). The Cpn60/Cpn10 complex is thought to be involved in protein folding following their translocation into the matrix (Dyall & Dolezal 2008). Putative Pam18 homologues have been localized to the mitosome of Giardia and to the hydrogenosome of Trichomonas (Dolezal et al. 2005), and a putative Pam16 has been annotated for E. cuniculi (Waller et al. 2009). An mtHsp70 has been localized to the hydrogenosomes of Trichomonas (Bui et al. 1996) and to the mitosomes of C. parvum (Slapeta & Keithly 2004), E. cuniculi (Tsaousis et al. 2008), T. hominis (Williams et al. 2002) and E. histolytica (Tovar et al. 2007). Cpn60 and Cpn10 have been localized to the mitosomes of E. histolytica (Tovar et al. 1999, 2007), Cpn60 to Giardia (Regoes et al. 2005) and C. parvum (Riordan et al. 2003) and Cpn60 and Cpn10 to the hydrogenosomes of T. vaginalis (Bui et al. 1996).
5. Iron–sulphur cluster biosynthesis: an essential common function of mitochondrial homologues?
The biogenesis of ISCs is the only essential biosynthetic function of yeast mitochondria (Lill & Muhlenhoff 2008) and is required for the maturation of mitochondrial and cytosolic Fe/S proteins of diverse functions. In humans, defects in Fe/S protein biogenesis cause haematological and neurological disorders (Lill & Muhlenhoff 2008) such as Friedreich's ataxia. In yeast, the synthesis of an ISC occurs inside mitochondria on the scaffold proteins Isu1/Isu2 (yeast nomenclature) (Muhlenhoff et al. 2003). This step requires a cysteine desulphurase (Nfs, also abbreviated IscS) complex (Nfs1–Isd11) (Zheng et al. 1993; Wiedemann et al. 2006) and frataxin (Yfh1) (Gerber et al. 2003), as the sulphur and iron donors, respectively. The electrons needed for the reaction are supplied by ferredoxin (Yah1) (Muhlenhoff et al. 2003). ISC transfer from Isu1/Isu2 to target mitochondrial apoproteins is thought to be facilitated by a dedicated chaperonin system comprising mtHsp70 (Ssq1) and the DnaJ-like cochaperone Jac1 (Lill 2009). This mitochondrial pathway for making ISCs is collectively called the ISC assembly system. Core components of the ISC machinery used to make ISCs inside mitochondria appear to have been inherited from the mitochondrial endosymbiont (Tachezy et al. 2001; Emelyanov 2003), making it potentially one of its aboriginal functions. Recent work suggests that ISC assembly may be an important function of diverse mitochondrial homologues too.
Trichomonas hydrogenosomes contain core components of the ISC machinery, including Hsp70, ferredoxin, Nfs, Isu and frataxin (Hrdy et al. 2008), and isolated hydrogenosomes catalyze the enzymatic assembly and insertion of ISCs into apo-ferredoxin (Sutak et al. 2004). When expressed in the mitochondria of a frataxin-deficient S. cerevisiae strain, T. vaginalis frataxin also partially restored defects in heme and ISC biosynthesis (Dolezal et al. 2007). The localization of Isu and Nfs to tiny double-membrane-bounded organelles provided the first evidence for a mitosome in Giardia (Tovar et al. 2003). High-speed fractions obtained using differential centrifugation and enriched in Giardia mitosomes were also able to add ISCs to apo-ferredoxin, suggesting a complete pathway for ISC assembly is present in the organelle (Tovar et al. 2003). Interestingly, however, no frataxin gene was detected on the Giardia genome (Morrison et al. 2007).
There is also published evidence that the mitosomes of two species of microsporidia, E. cuniculi and T. hominis, contain key components of the ISC machinery. Using indirect immuno-fluorescence analyses (IFA) of parasite-infected rabbit kidney cells, and antisera raised against recombinant E. cuniculi or T. hominis proteins, the mitosomes of E. cuniculi were shown to contain frataxin, Nfs, Isu and mtHsp70 (Goldberg et al. 2008; Tsaousis et al. 2008) and potentially ferredoxin (Williams et al. 2008a). By contrast, using the same IFA approach, the mitosomes of T. hominis appear to contain mtHsp70 and Nfs but the main pools of Isu and frataxin were apparently cytosolic (Goldberg et al. 2008). The IFA data cannot exclude the possibility that there is some minor pool of Isu and frataxin in the T. hominis mitosome or of Nfs in the cytosol, but the apparent disjunct distribution of major pools raises intriguing questions as to how the separated T. hominis ISC components can perform their usually tightly coordinated function (Goldberg et al. 2008). Functional data from the same study already demonstrate that ThIsu accepts sulphur from Saccharomyces mitochondrial Nfs1, because it can complement a yeast Isu1/2 mutant. How the persulphide moiety generated by the cysteine desulphurase activity of mitosomal ThNfs might be delivered to a cytosolic ThIsu scaffold is also problematic. However, even in well-studied model systems like yeast (Lill & Muhlenhoff 2008), there are still uncertainties about what essential product of mitochondrial Fe/S biosynthesis is actually exported from mitochondria to support cytosolic Fe/S biosynthesis (Netz et al. 2007; Lill 2009).
Bacteria possess three distinct systems to generate Fe/S proteins. The ISC assembly system is used for ‘housekeeping’ proteins, the sulphur-mobilization (SUF) machinery probably repairs Fe/S proteins and the nitrogen-fixation (NIF) machinery is used primarily to form specific ISCs in the nitrogenase of nitrogen-fixing bacteria (Tachezy & Smid 2008). Entamoeba histolytica is the only eukaryote known to have lost key components of the ISC machinery (Loftus et al. 2005) and to have replaced it by lateral gene transfer of a NifS cysteine desulphurase and NifU scaffold protein, from bacteria related to Campylobacter and Helicobacter (Ali et al. 2004; van der Giezen et al. 2004). Heterologous expression of EhNifS and EhNifU successfully complemented, under anaerobic but not aerobic conditions, the growth defect of an E. coli strain, in which both the ISC and SUF operons were deleted. These data suggest that EhNifS and EhNifU are necessary and sufficient for ISCs of non-nitrogenase Fe/S proteins to form under anaerobic conditions (Ali et al. 2004). The cellular localization of EhNifS and EhNifU in Entamoeba is currently unknown, but is crucial to testing the hypothesis that all mitochondrial homologues make ISCs.
Genes for NifS and NifU were also identified in a cDNA library for the related anaerobic amoeba Mastigamoeba balamuthi, suggesting that gene transfer predates the diversification of Entamoeba and Mastigamoeba (Gill et al. 2007). By contrast, Dictyostelium discoideum, an aerobic relative of Entamoeba, contains genes for the mitochondrial ISC system, suggesting that transfer post-dates the last common ancestor of these two species. Intriguingly, Acanthamoeba castellani, another relative of Dictyostelium and Entamoeba, is reported to contain genes for both ISC and NIF components, but the proteins have so far not been localized or functionally characterized (van der Giezen 2009).
6. Summary and future prospects
Over the past 15 years, we have gone from a perspective whereby some eukaryotes were held to be primitively without mitochondria to one where it seems reasonable to suggest that organelles related to mitochondria are essential components of all eukaryotic cells. Some of these mitochondrial homologues, the mitosomes of parasites, no longer make ATP and have undergone dramatic reductive evolution at every level. For example, it is difficult to understand how Entamoeba and Giardia mitosomes can import proteins through their inner membrane because they appear to have lost key components of the complexes used for this purpose in model systems. Studying protein import into mitosomes directly is needed to better understand how they achieve this and to reveal the limits to which reductive evolution can proceed while maintaining a viable organelle. Interesting progress has already been made in this area, for example the discovery that the mitochondrial peptidase of Giardia functions, uniquely among eukaryotes, as a monomer (Šmíd et al. 2008).
Some of the work needed will be based upon proteomics investigations of isolated organelles. Identifying parasite orthologues of functionally characterized proteins can be difficult because the parasite versions are often highly divergent. Using computers trained on targeting sequences of model organisms to try and predict the cellular locations of parasite proteins often suffers from the same limitations. High-throughput proteomics using mass spectrometry has proved invaluable for investigating the structures of model mitochondria and their functions in health and disease (Mathy & Sluse 2008). New highly sensitive mass-spectrometry methods that use differential isotope tagging to identify protein location (Lilley & Dunkley 2008), and which do not require large amounts of highly purified organelles—which can be difficult to isolate even from yeast, never mind obligate intracellular parasites like Microsporidia—are now available to facilitate proteomics studies of mitosomes and hydrogenosomes. The potential value of this work is illustrated by proteomics studies on Trichomonas hydrogenosomes that have already found novel hydrogenosomal proteins, including peroxidases and [Fe]-hydrogenase maturases (Henze 2008). So far, there have been no proteomics studies of mitosomes from Entamoeba, Giardia or Microsporidia, or of any MLO or non-Trichomonas hydrogenosomes. Such work, together with detailed bioinformatics, will provide clues to the structure and functions of different organelles including lineage-specific differences and innovations.
The discovery of organelles such as the Blastocystis MLO and the Nyctotherus hydrogenosome provided important missing links between classic mitochondria, hydrogenosomes and mitosomes and examples of intermediate stages of reductive evolution. The key ATP-generating pathways in classic mitochondria are the TCA cycle and the ETC, both of which are inhibited in the absence of oxygen. Although the data for Blastocystis and Nyctotherus are only partial (Boxma et al. 2005; Perez-Brocal & Clark 2008; Stechmann et al. 2008), it already suggests that complexes of the ETC components have been lost, presumably because of the decreased selection pressure to retain genes for aerobic respiration in low-oxygen environments. The constraints of living under low-oxygen conditions have probably also been major drivers for the reductive evolution of the mitosomes of Cryptosporidium, Entamoeba, Giardia and the hydrogenosomes of Trichomonas. By contrast, microsporidia live inside eukaryotic cells that have perfectly good aerobically functioning mitochondria. It is not an anaerobic lifestyle that is driving the reductive evolution of microsporidia and their mitosomes, but obligate intracellular parasitism.
At present, the only common metabolic function identified for a diverse sample of mitochondria, hydrogenosomes and mitosomes is ISC assembly.
The biogenesis of ISCs is an essential biosynthetic function of yeast mitochondria, and defects in human Fe/S protein biogenesis cause severe haematological and neurological disorders (Lill & Muhlenhoff 2008). Quite why ISC biosynthesis might need to be contained within the organelle is not certain (Lill 2009). For organelles containing Fe/S proteins such as complex I or ferredoxin, it might simply be easier to insert ISCs into apo-proteins in situ. Another reason could be that cellular compartmentalization limits the toxicity of ferrous iron and sulphide, which are needed for Fe/S assembly (Tachezy & Smid 2008). In yeast and humans, the products of mitochondrial ISC biogenesis are also needed to make essential non-mitochondrial Fe/S proteins by a cytosolic iron–sulphur protein assembly system (Netz et al. 2007). These Fe/S proteins include Rli1, which is essential for a functional translation apparatus (Kispal et al. 2005). The component exported from yeast mitochondria (by an ISC export machinery) has not been fully characterized; it appears to contain sulphur but not iron (Lill & Muhlenhoff 2008). It will be interesting to investigate whether hydrogenosomes and mitosomes are vital for the species that contain them because they also carry out this function.
Work done in our laboratory on this topic was supported by European Union Marie Curie Training Fellowships to K.H. and A.V.G., and by a Wolfson Research Merit Award from the Royal Society to T.M.E.
↵† Present address: School of Life Sciences, Södertörn University, 141 89 Huddinge, Sweden.
↵‡ Present address: Center for Comparative Genomics and Evolutionary Bioinformatics, Department of Biochemistry and Molecular Biology, Dalhousie University, Halifax, Nova Scotia, Canada.
One contribution of 12 to a Theme Issue ‘Evolution of organellar metabolism in unicellular eukaryotes’.
- © 2010 The Royal Society