The Minimal Proteome in the Reduced Mitochondrion of the Parasitic Protist Giardia intestinalis

The mitosomes of Giardia intestinalis are thought to be mitochondria highly-reduced in response to the oxygen-poor niche. We performed a quantitative proteomic assessment of Giardia mitosomes to increase understanding of the function and evolutionary origin of these enigmatic organelles. Mitosome-enriched fractions were obtained from cell homogenate using Optiprep gradient centrifugation. To distinguish mitosomal proteins from contamination, we used a quantitative shot-gun strategy based on isobaric tagging of peptides with iTRAQ and tandem mass spectrometry. Altogether, 638 proteins were identified in mitosome-enriched fractions. Of these, 139 proteins had iTRAQ ratio similar to that of the six known mitosomal markers. Proteins were selected for expression in Giardia to verify their cellular localizations and the mitosomal localization of 20 proteins was confirmed. These proteins include nine components of the FeS cluster assembly machinery, a novel diflavo-protein with NADPH reductase activity, a novel VAMP-associated protein, and a key component of the outer membrane protein translocase. None of the novel mitosomal proteins was predicted by previous genome analyses. The small proteome of the Giardia mitosome reflects the reduction in mitochondrial metabolism, which is limited to the FeS cluster assembly pathway, and a simplicity in the protein import pathway required for organelle biogenesis.


Introduction
Mitochondria are eukaryotic organelles that are thought to have evolved from an alpha-proteobacterial endosymbiont about two billion years ago.The loss of bacterial autonomy and transition of the endosymbiont to a ''protomitochondrion'' were associated with a reduction in the number of genes in the endosymbiont genome; these genes were either transferred to the nuclear genome or lost.While the genome of the extant alpha-proteobacterium Rickettsia prowazekii contains 834 protein-coding genes [1], the largest number of genes (67 protein-coding genes) in a mitochondrial genome is found in Reclinomonas americana [2], with only three protein-coding genes present in the Plasmodium falciparum mitochondrial genome [3].Paradoxically, the reduction of the mitochondrial genome did not lead to a reduction of the organellar proteome [4].The acquisition of a mechanism for mitochondrial import at the earliest stage of the endosymbiont-toprotomitochondrion transition allowed the recruitment of the proteins of endosymbiotic origin that were now encoded in the nucleus, and the import of proteins of other origins [5].Contemporary mitochondrial proteomes contain hundreds of proteins, up to 1100 proteins in the mouse [6].
Mitosomes are the most highly reduced forms of mitochondria, having completely lost their genomes and dramatically reduced their proteomes.Mitosomes have also lost many of the typical mitochondrial functions, such as respiration, the citric acid cycle, and ATP synthesis.Biosynthesis of FeS clusters is the only mitochondrial function seen to be retained by at least some mitosomes [7].Mitosomes have become established independently in diverse groups of unicellular eukaryotes (protists); many of them once considered to be amitochondrial because they lack organelles with the expected mitochondrial morphology [8].
Organisms with mitosomes live under oxygen-limiting conditions, like the human intestinal parasites Giardia intestinalis [9] and Entamoeba histolytica [10], or are intracellular parasites like the microsporidians Encephalitozoon cuniculi and Trachipleistophora hominis [11,12] and the apicomplexan Cryptosporidium parvum [13].Mitosomes are tiny ovoid organelles enclosed by two membranes.Unlike mitochondria, the inner membrane of mitosomes does not form cristae.The morphology of the mitosome is reminiscent of the hydrogenosome, another form of mitochondrion that is present in some anaerobic protists, such as Trichomonas vaginalis.Unlike mitosomes, however, hydrogenosomes are metabolically active organelles that produce ATP by substrate level phosphorylation [14].
The limited knowledge of mitosomal proteomes has been gained mainly from analyses of genome sequences and localization studies of a few model mitosomal proteins [9,11,[15][16][17][18][19][20][21].The only published proteomics study that focused on mitosomes was that recently reported for the amoeba E. histolytica, identifying a unique sulfate activation pathway [22].To increase our understanding of the function and origin of these enigmatic organelles, we established a large-scale proteomic approach to analyze the mitosomes of Giardia intestinalis.This organism was selected because Giardia intestinalis is a common human intestinal pathogen, its genome sequence has been published [23,24], and it is considered to be among the most basal eukaryotes [25].Moreover, previous analysis of the G. intestinalis genome provided little new information pertaining to the putative mitosomal proteome [24], so there are substantial gaps in our knowledge of the structure and function of this essential organelle.Here, we quantitatively analyzed the presence of isobarically-tagged proteins in mitosome enriched fractions.This technique allowed us to discriminate the mitosomal proteins from those of contaminating cellular structures.Combined with an exhaustive bioinformatics analysis, this strategy identified 139 putative mitosomal proteins; 20 of which were experimentally confirmed to be localized in mitosomes.Our results revealed that the proteome of the G. intestinalis mitosome is selectively reduced and houses a single metabolic pathway for FeS cluster assembly, a novel diflavin protein with NADPH reductase activity, a minimal protein import machinery and proteins that may be important for the interaction of mitosomes with other cellular compartments.

Identification of putative mitosomal proteins by isobaric tagging
Mitosome-enriched fractions were separated from a Giardia homogenate by preparative centrifugation using a discontinuous Optiprep (iodixanol) gradient [26].This method produced five dense organellar fractions (Fig. 1A).The mitosomal marker protein IscU was particularly enriched in fraction #4 and to a lesser extent in fraction #3 (Fig. 1B).Electron microscopy confirmed the presence of mitosomes in both fractions; however, co-fractionating vesicles of similar densities were also found (data not shown).To discriminate between putative mitosomal proteins and those of contaminating cellular structures, we compared the relative distribution of each protein in fractions #3 and #4.Because the mitosomal proteins necessarily co-fractionate (i.e.being contained within mitosomes) during gradient centrifugation, each of the bona fide mitosomal protein should display similar distribution ratios [27].To this end, the proteins of fractions #3 and #4 were digested in parallel with trypsin and each peptide population was labeled with a distinct iTRAQ reagent and then combined.The isobaric mass characteristics of the iTRAQ reagents means the differentially-labeled peptides from fractions #3 and #4 form a single peak in the MS scan for protein identification.MS/MS analysis of the iTRAQ-labelled peptides liberates the isotope-encoded reporter ions, the ratio of which can reflects the distribution of the protein across the two fractions.In our analysis, the pooled peptides were analyzed by tandem mass spectrometry after subsequent separation with isoelectric focusing and nano-liquid chromatography (nano-LC MS/MS).The iTRAQ ratio was then calculated for each protein, and the proteins were sorted according to the relative distributions in the fractions (Fig. 2).
Validating the methodology, mitosomal markers (IscS, IscU, [2Fe2S] ferredoxin, Cpn60, Hsp70 and glutaredoxin 5) [28] clustered together with similar iTRAQ ratios (Fig. 2).Proteins with ratios between the lowest and highest values for the markers were considered to be candidate mitosomal proteins.We also extended this window on both sides by half of the distance between the limiting markers and included all proteins in this extended window (Fig. 2).In total, we identified 638 proteins (Table S1), with 139 of these proteins meeting the defined criteria for mitosomal proteins (Tables 1-7).Each of the 139 mitosomal candidates was assigned to a probable function based on current annotations in the GiardiaDB, PSI BLAST searches in the NCBI nr database, and motif and domain searches in the Pfam database.Three additional bioinformatics tools were used to predict cellular localization (PsortII, TargetP 1.1 and SignalP 3.0), and two web-based programs were used to predict alpha-helical transmembrane region segments (TMHMM and Memsat3) (Tables S2-S4, summary is given in Tables 1-7).The candidate proteins were clustered into 13 groups according to their predicted functions (Tables 1-7, Fig. 3).The proteomic data confirmed the validity of 250 hypothetical genes predicted from the complete genome sequence of Giardia [24]; 40 of these formed the largest group of candidate mitosomal proteins.

Evolution-inspired orthology phylogenetic profiling
Previous phylogenetic analyses of known mitosomal proteins have generally confirmed their alpha-proteobacterial origin [28][29][30].On this premise, we compared the genomes of G. intestinalis and Rickettsia typhi using the orthology phylogenetic profile tool at GiardiaDB (http://www.orthomcl.org/cgi-bin/OrthoMclWeb.cgi) to identify proteins of alpha-proteobacterial ancestry in the G. intestinalis genome.The phylogenetic profiling yielded 106 candidate genes that were analyzed with the topology prediction algorithms described above (Table S5).Based on these analyses, six additional proteins: acetyl CoA acetyl transferase, CDPdiacylglycerol-glycerol-3-phosphate 3-phosphatidyltransferase, guanylate kinase, J-protein HesB, thioredoxin reductase, and thymidylate kinase were added to the set of candidate mitosomal proteins identified by our proteomics approach (Tables 1-7).

Experimental validation of protein subcellular localization
The cellular localization of the selected candidate proteins was observed by stable episomal expression in Giardia.To establish the morphology of subcellular localizations by this approach, we first observed the localization of five marker proteins: cytosolic enolase, two proteins from the endoplasmic reticulum (Hsp70 and protein disulfide isomerase 5), mitosomal Hsp70 and glutaredoxin (Fig. 4A).We added to these markers of known location, three proteins of untested location with iTRAQ ratios outside that of the mitosomal range: glutamate dehydrogenase, copine and peroxiredoxin.Glutamate dehydrogenase and copine were associated with cytoskeletal structures, while peroxiredoxin localizes to the endoplasmic reticulum network.This strategy was used to test the sub-cellular localization of 44 selected proteins.Of these 20 expressed fluorescent fusions that were found in the mitosomes (Tables 1-7).By way of example, four of these: VAP, Pam16, Cpn10 and unknown proteins GL50803_9296 and GL50803_ 14939 are shown in Fig. 4B.

Iron-sulfur cluster assembly
Proteins involved in FeS cluster assembly formed the most prominent functional group within the predicted mitosomal proteins.These included components required for the formation of transient FeS clusters on the molecular scaffold (IscS, IscU, Nfu) (Fig. S1) and components that have been proposed to transfer the transient FeS clusters to target apoproteins, including IscA (Fig. S2), the monothiol glutaredoxin 5, chaperone Hsp70 and its co-chaperones the J-protein HscB (Fig. S3) and nucleotide exchange factor GrpE (Fig. S4).The identification of the FeS cluster assembly machinery in the mitosomal proteome is consistent with the ability of the mitosome-enriched fraction to catalyze the formation of FeS clusters on a ferredoxin apoprotein [9].However, when we compared the FeS cluster machinery of Giardia mitosomes to that of S. cerevisiae and Trypanosoma brucei mitochondria, we found that several mitochondrial components were absent from the mitosomes (Table 8).
A striking deviation from other eukaryotes is the absence of frataxin in Giardia mitosomes.Frataxin is invariably present in eukaryotes that contain the ISC-type FeS cluster assembly machinery.The presence of frataxin in mitosomes was found in E. cuniculi [17], and genes encoding frataxin are present in the genomes of C. parvum and the diplomonad Spironucleus vortens, a close relative of Giardia.We failed to identify frataxin in the genomes of three G. intestinalis strains in the GiardiaDB, using either BLAST searches or the motif search tool.
Two IscA-like proteins, IscA1 (Isa1) and IscA2 (Isa2) are present in virtually all mitochondria [31] and are thought to act as scaffold proteins for transient FeS clusters [32][33][34] and/or serve as iron donors [35].Interestingly, the Giardia mitosome contains only a single IscA-2 type protein (Fig. S2), while IscA-1 is absent.The same situation was found in hydrogenosomes of Trichomonas vaginalis (Table 8).No genes encoding IscA were found in the genomes of other organisms with mitosomes.The observed distributions of IscA therefore suggest that IscA-1 was lost in mitosomes and hydrogenosomes together with a specific set of mitochondrial FeS proteins, while IscA-2 was retained in Giardia mitosomes to function either in the maturation of specific FeS protein(s) or as an iron transporter [35].
The mitosomes did not contain Ind1 or Iba57.In mitochondria, these proteins are required for the formation of FeS clusters on specific substrates.Ind1 is a P-loop NTPase that is required for the maturation of FeS proteins of the multi-subunit respiratory complex I [36,37].Homologues of Ind1 are also present in the hydrogenosomes of T. vaginalis (Table 8), which contain a highly reduced form of complex I with only two FeS catalytic subunits [38].The selective absence of Ind1 in the mitosomes of Giardia (Table 8) is thus consistent with the absence of complex I and highlights the specific role of Ind1 in the biogenesis of this respiratory complex.Iba57 forms a complex with the scaffold protein IscA (Isa1p and Isa2p in yeast), which plays a specific role in [4Fe4S] cluster assembly of aconitase-type proteins and the functional activation of mitochondrial radical-SAM FeS proteins [39].As in the case of Ind1, the absence of Iba57 likely reflects the absence of the respective substrate proteins in mitosomes.

Pyridine nucleotide-driven electron transport in mitosomes
The formation of FeS clusters requires reducing equivalents, which are provided by a short electron chain consisting of the [2Fe2S] ferredoxin and ferredoxin:NADP + reductase (FNR) [40].The presence of this chain has been predicted in the mitosomes of C. parvum and E. cuniculi; however, [2Fe2S] ferredoxin, but not FNR, was found in Giardia mitosomes (Table 1).We identified a distinct protein with a possible redox activity named GiOR-1 (GL50803_91252), which is currently annotated in the GiardiaDB as an inducible nitric oxide synthase.This protein consists of a flavodoxin-like FMNbinding domain that is connected to a cytochrome p450 reductaselike domain, including a FAD binding pocket and an NADP(H) binding site (Fig. S5).These two domains are present in the C-termini of various oxidoreductases, such as cytochrome p450 reductase and nitric oxide synthase, and serve as electron donors (Fig. S5).GiOR-1 does not contain an N-terminal domain that determines the specific functions of known oxidoreductases.
The architecture of GiOR-1 resembles that of the recently identified protein Tah18 in Saccharomyces cerevisiae [41,42].Tah18 was shown to form a complex with Dre2 in the cytosol, where it participates in cytosolic FeS cluster assembly [43].Under oxidative stress, the Dre2-Tag18 complex was destabilized, and Tah18 relocalized from the cytosol to the mitochondria.This behavior has been shown to be associated with apoptotic events.Searches for a Dre2 homologue in Giardia were unsuccessful.However, we identified a second paralogue of Tah18 named GiOR-2 (GL50803_15897, Fig. S5).The expression of tagged GiOR-1 and GiOR-2 in G. intestinalis confirmed that the GiOR-1 is localized to the mitosome, but GiOR-2 was found in numerous vesicles that did not correspond to mitosomes (Fig. 4C).To assess the oxidoreductase activity of GiOR-1, recombinant GiOR-1 was produced in Escherichia coli and isolated as a yellow protein, which is expected for diflavin oxidoreductases.GiOR-1 efficiently transferred electrons from NADPH to dichlorophenolindolphenol, whereas an about 30 fold lower activity was measured using NADH as the electron donor (Table 9).Low specific activities were observed also with methyl viologen and oxygen as electron acceptors (Table 9).No activity was observed when GiOR-1 was assayed with G. intestinalis mitosomal ferredoxin as a possible native electron acceptor.These results suggest that GiOR-1 does not act directly as a ferredoxin reductase in mitosomes, however, its ability to utilize NADPH as an electron donor indicates that pyridine nucleotides are involved in mitosomal electron transport.

Molecular chaperones in the mitosomal matrix: protein folding and assembly
A single mitosomal Hsp70, three J-protein co-chaperones and the nucleotide exchange factor GrpE were identified in the mitosomes.The J-proteins included HscB, an orthologue of yeast Jac1 (Fig. S3)that has a predicted role in FeS cluster biogenesis [44], and Pam18/Tim14, which is required for translocation of proteins across the mitochondrial inner membrane [45].The third J-protein also contains an N-terminal DnaJ domain (type III family); however, its function cannot be inferred from domain structure or phylogenetic profiling.We also identified the chaperonins Cpn60 and Cpn10 (Fig. S6), that function in folding and assembly of newly-imported proteins [46,47] (Table 1).
Protein translocation and processing Mascot score, Mascot total ion score for the identified protein.Coverage, number of unique peptides per identified protein.MiD, mitosomal distribution.Proteins are marked ''Y'' if their distributions in fractions #3 and #4 of the Optiprep gradient (measured by the iTRAQ ratio) were within the range between Cpn10 and IscU and the window that extended in both directions by half of the distance between these markers.Proteins with ratios outside of this range are indicated with ''N''.TargetP and PsortII were used to predict the subcellular location of Giardia proteins.S, secretory; N, non-secretory; M, mitochondrial; O, other.Exp.ver., experimental verification of protein localization using the pONDRA expression vector.The recombinant tagged proteins were localized by fluorescence microscopy.M, mitosome; ER, endoplasmic reticulum; O, other; ?inconclusive.MEMSAT3 and TMHMM were used to predict transmembrane domains.SGP, predicted soluble proteins are marked with number sign (#).Asterisk (*) is used where no data were available.(**) transformed Giardia did not express the recombinant tagged protein.doi:10.1371/journal.pone.0017285.t001

Protein import
We identified four components that are potentially involved in transporting proteins across the mitosomal membranes: a homologue of a mitochondrial Tom40, which would form a general import pore in the outer mitosomal membrane, and the three essential components of the PAM (presequence translocaseassociated motor) complex: Pam18, Pam16 (Fig. S7) and mHsp70.Pam18 and Pam16 form an intimate complex that anchors a population of the matrix Hsp70 to the inner membrane and regulates its activity to drive protein translocation across the inner  membrane [48].Typically, it functions together with a TIM complex that forms the translocation pore for protein passage across the membrane.In representative organisms from all lineages of eukaryotes, the TIM complex is built from one or two proteins of the Tim17/22/23 family [49].Surprisingly, we find no evidence for a member of this protein in our proteomics data, and sensitive hidden Markov model searches detected no related sequences in the Giardia genome (unpublished, see Methods).In eukaryotes, the Sec61 channel catalyzes protein transport across the endoplasmic reticulum [2], while a highlyrelated protein called SecY is the translocation channel in the inner membrane of bacteria, including the alpha-proteobacteria from which mitochondria are derived.Interestingly, Reclinomonas americana encodes a bacterial-type SecY protein translocation channel in its mitochondrial genome [50], and our proteomics analysis detected what appeared to be contamination of the mitosomal membranes with GiSecY/Sec61.We expressed a tagged version of this protein in Giardia but it localized to the endoplasmic reticulum, as expected for a cognate Sec61, rather than to the mitosomes.The nature of the mitosomal inner membrane protein translocation channel remains unknown, and yet must exist given that at least 17 of the proteins detected in the mitosomal proteome are likely to reside in the matrix.We suggest that Tim23/17/22 protein(s) have been secondarily lost from Giardia, given that the these proteins appear to be derived from components of the ancestral endosymbiont [48] and are present in all other groups of eukaryotes including other members of the Excavata [51], particularly T. vaginalis (TrichDB, http:// trichdb.org/trichdb;our unpublished data).Because there is evidence to suggest that T. vaginalis and G. intestinalis share a common ancestor [44,52], the absence of a Tim23 homologue in Giardia likely reflects the overall simplification of the organelle than a primitive trait.Why has the TIM complex been replaced?In addition to a reliance on ATP hydrolysis mediated by the PAM motor, the TIM complex is powered by the membrane potential through its physical association with the respiratory complexes III and IV [53,54].Giardia mitosomes do not generate a large membrane potential, as shown by their inability to accumulate the routinely used mitochondrial probes that are sensitive to the membrane potential (e.g., mitotrackers, JC-1, our observations).Perhaps any membrane potential that is present, is insufficient to support the function of a TIM23 translocase.

Interaction of mitosomes with other cellular compartments
In the Giardia mitosomes, we identified a VAMP (vesicleassociated membrane protein)-associated protein, VAP (Table 6).VAPs are involved mainly in membrane trafficking and lipid metabolism.They provide membrane anchors for various lipid binding proteins on the surfaces of the endoplasmic reticulum and Golgi complex [53] and physically interact with SNARE proteins, with FFAT-motif containing lipid transport proteins and microtubules.Like other VAPs, the Giardia VAP protein contains an Nterminal domain that includes the VAP consensus sequence [55], a central coiled-coil domain and a C-terminal transmembrane domain with the putative dimerization motif GxxxG (Fig. S8).The presence of a VAP protein has not been reported in mitochondria or other mitosomes so far.In Giardia, GiVAP was found within the set of hypothetical proteins with distribution value corresponding to mitosomal proteins (Table 6) and its mitosomal localization was experimentally confirmed (Fig. 4B).

Hypothetical proteins
The set of mitosomal candidates contains 40 proteins annotated as hypothetical proteins.We selected six proteins with high mitochondrial score (Tables 6-7) for the verification of their subcellular localization.Three proteins were confirmed to reside in mitosomes (Table 6, Fig 4): (i) putative VAP (GL50803_15985) that is discussed above, (ii) hypothetical protein GL50803_14939 that contains two predicted transmembrane domains (residues 13-35 and 102-124), and (iii) a putative soluble globular protein GL50803_9296.The latter two proteins seem to be unique for giardia as no orthologues were identified in available databases.Two other hypothetical proteins (GL50803_16596 and GL50803_4768) were observed in the cytosol and in association with kinetosomes, respectively (Table 6, data not shown).The cellular localization of hypothetical protein GL50803_12999 remains inconclusive.Although the protein co-localized with IscU in some vesicles, it was not observed in typical rod-like structure between nuclei (data not shown).

Origin of mitosomes and perspectives
Mitosomes are thought to have evolved several times in different eukaryotic lineages through the reduction of ancestral mitochondria.For example, microsporidians are intracellular parasites allied with Fungi; whereas Fungi typically possess fully equipped mitochondria with large proteomes (.850 proteins) [11,56], only twenty to thirty proteins have been identified from genome analysis of E. cuniculi as having similarity to bona fide mitochondrial proteins of Saccharomyces cerevisiae [15,18,21].Apicomplexan parasites related to Plasmodium also include organisms with mitosomes, such as Cryptosporidium parvum and Cryptosporidium hominis.Based on genomic analyses, 37-54 proteins have been predicted to reside in these mitosomes [19], of which 18 were detected by mass spectrometry in whole C. parvum sporozoites [25].
An intriguing question concerns the nature of the mitochondrial progenitor from which mitosomes of G. intestinalis have evolved.Giardia is a member of the Excavate group, which has recently been re-considered to belong to the basal groups of eukaryotes based on its mechanism of cytochrome c and c1 biogenesis   [25,57].These and other data have placed the root of eukaryotes between Excavata and Euglenozoa, a group of protists that includes trypanosomatids [58].In this respect, there is an apparent simplicity in the protein import machinery of the Giardia mitosomes that deserves attention (Fig. 5).The proteomics analysis detected in mitosomes a protein recently shown to be Tom40, the protein translocation channel across the outer membrane [25,59].The current model for the evolution of the TOM complex posits that Tom40 was derived from a beta-barrel protein in the endosymbiont's outer membrane, perhaps of an usher or autotransporter type protein translocase [60].Because two other proteins: Tom7 and Tom22, have been found in representative species of all other eukaryotic groups [58], the model further suggests that the first TOM complex was composed of Tom40, Tom22 and Tom7.Our proteomics finds no evidence of Tom7 or Tom22 in mitosomes, and sensitive hidden Markov model searches likewise fail to find any proteins encoded in the Giardia genome with similarity to Tom7 or Tom22 [25,57].Whether reflecting a secondary gene loss or the ancestral condition, GiTom40 would appear to be a selectively simple protein translocase.In addition to Tom40, mitochondria contain one other member of the mitochondrial porin family, the voltage- dependent anion channels (VDAC), which serve to exchange metabolites [61].The absence of VDAC in Giardia mitosomes might reflect the disappearance of many of the metabolic pathways, and the concomitant decrease in metabolite flux across the outer membrane.It is likely that the Giardia Tom40, in addition to importing proteins, exchanges ions and small metabolites across the outer mitosomal membrane as has been demonstrated for the yeast Tom40 in mutants lacking VDAC [44,[62][63][64].
Another surprising result, one that can only be explained by a secondary gene loss, is the absence of the outer membrane protein Sam50 in Giardia.Sam50 is a component of the SAM (sorting and assembly machinery) complex, which is required for the assembly of both Tom40 and VDAC [48,65].The apparent absence of Sam50 from the Giardia genome and from our proteomics data is unique among eukaryotes.A putative Sam50 homologue has been predicted in the genomes of all eukaryotes, including trypanosomatids [58,65] and mitosome-and hydrogenosome-containing protists (C.parvum, E. cuniculi, E. histolytica and T. vaginalis) [66].Numerous phylogenetic and functional analyses indicate that Sam50 was derived from the Omp85/BamA protein present in the outer membrane of the ancestral, alpha-proteobacterial endosymbiont and it must, therefore, have been present in the earliest mitochondria [44].It is not clear how Giardia Tom40 is assembled within the outer membrane without the assistance of the SAM complex.It is known that even in yeast Tom40 mediates the import of new molecules of Tom40 into mitochondria [67] and it is tempting to speculate that the Giardia Tom40 is capable of mediating its own import and membrane insertion, given the highly simplified nature of the TOM complex in mitosomes.
Our proteomics data support the hypothesis that ISC assembly is an important and possibly the only biosynthetic function of Giardia mitosomes.Previous phylogenic analyses have indicated that the ISC assembly machinery was obtained from the alphaproteobacterial endosymbiont; nearly complete ISC assembly machinery is present from trypanosomatids to higher eukaryotes.Therefore, the absence of certain components, such as IscA-1, Iba57, and Ind, in the mitosomal machines (Table 8) is apparently due to a secondary loss of specific target proteins.Noteworthy, we did not identify any proteins that would carry FeS clusters in Giardia mitosomes, except for components of the FeS cluster assembly machinery itself.It seems likely then that the main role of mitosomes could be to export preassembled FeS clusters, or other compounds that are essential for the biogenesis of FeS proteins, to other cellular compartments.In mitochondria, the export of these enigmatic compounds is dependent on the membrane ABC ''halftransporter'' Atm1 [68] and sulfhydryl oxidase Erv1 [69].In the mitosome-enriched fraction, we identified four ABC half-transporters by mass spectrometry, and another candidate was predicted based on phyletic profiling of the G. intestinalis genome (Table 2).However, compared to other Atm1 homologues, these candidates lack the x-loop with the conserved arginine, which is essential for known Atm1 transporters (Fig. S9).No protein with homology to Erv1 was found by proteomics or by analysis of the Giardia genome.Another remaining question pertains to the source of ATP that is required for the multiple processes identified in mitosomes including FeS cluster assembly and export, organelle division, protein import and protein folding.In E. histolytica, it has been shown that a mitochondrial carrier family (MCF) protein localizes to mitosomes and exchanges ATP and ADP across the inner membrane, effecting the import ATP into mitosomes [20].E. cuniculi mitosomes contain a distinct bacterial nucleotide transporter that may fulfill the same function [23,24].However, our proteomic analysis did not revealed a candidate nucleotide transporter in the mitosomes of Giardia leaving open the question of ATP acquisition.
In conclusion, using iTRAQ-based mass spectrometry and bioinformatics we identified 139 candidate mitosomal proteins.Mitosomal localization was confirmed experimentally for 20 of 44 proteins tested, suggesting the complete mitosomal proteome of Giardia to be of the order of 50-100 proteins.Previous genome analyses failed to predict any of the novel mitosomal proteins identified here [70]; only by combining quantitative mass spectrometry and bioinformatics were these novel proteins identified.The small proteome of the G. intestinalis mitosome indicates a marked reduction in mitochondrial metabolic activity and reduced requirements for organelle biogenesis.These do not mirror the reductions seen in the mitosomal proteome of Cryptosporidium, supporting the view that lineage-specific reductions produce organelles with distinct metabolic pathways and specific ''short-cut'' pathways for biogenesis.Our findings provide new insight into aspects of mitochondrial evolution and the basis from which to begin reconstructing the details of precisely how these organelles are built and replicated to support Giardia growth and division.

Cell culture and fractionation
G. intestinalis strain WB (American Type Culture Collection) was grown in TYI-S-33 medium supplemented with 10% heatinactivated bovine serum and 0.1% bovine bile [9].Trophozoites were freeze-detached, washed in PBS and collected by centrifugation.Cells were then resuspended in hypotonic buffer (12 mM MOPS, pH 7.4) and incubated for 15 minutes.The cells were then pelleted at 6806 g for 15 minutes, resuspended in the same buffer with DNase I (40 mg/mL) and homogenized by 10 passages through a 25G needle.After homogenization, the isotonicity was immediately restored with the addition of an equal volume of 500 mM sucrose in MOPS buffer.The homogenate was then treated with trypsin (200 mg/mL) for 10 minutes at 37uC to release the organelles from the cytoskeleton.Proteases inhibitors were then added (5 mg/mL of soybean trypsin inhibitor, leupeptin and TLCK), and the homogenate was diluted and centrifuged for 20 minutes at 27606 g to remove cellular debris.The collected supernatant was centrifuged using a Beckman rotor Ti 50 at 20,000 rpm for 30 minutes.After centrifugation, the pellet was collected and washed in SM buffer (250 mM sucrose and 12 mM MOPS, pH 7.4).Next, the pellet was resuspended in 0.5 mL of SM buffer and layered onto a discontinuous density OptiPrep (Axis-Shield PoC AS, Oslo, Norway) gradient, which consisted of 1 ml each of 15%, 20%, 25%, 30% and 50% OptiPrep diluted in 12 mM MOPS buffer.The gradient was centrifuged for 24 h in a Beckman SW 40 rotor at 120,0006 g at 4uC.Fractions (1 mL each) were collected, washed and analyzed by immunoblot using a polyclonal rabbit anti-IscU antibody [71,72].S2-S3).doi:10.1371/journal.pone.0017285.g003

Bioinformatics
Bioinformatics searches based on simple pair-wise alignment Psi-BLAST and hidden Markov models (HMMs) were applied to verify the automatic protein annotations and estimate their functions.Protein sequences (,1000 residues) were submitted (i) against a 90% redundancy reduced NCBI nr database for 8 iterations at an e-value cutoff of 10 23 and (ii) against Pfam 23.0 A+B database of families represented by multiple sequence alignments and hidden Markov models at an e-value of 0.044 (http://pfam.sanger.ac.uk/search).Where noted in the text, tailored HMM libraries were used to search for components of the protein import machinery [16].

Transformation of G. intestinalis and immunofluorescence
Selected genes were amplified by PCR from genomic Giardia DNA and inserted into the pONDRA plasmid [73].Table S6 contains a list of primers that were used for subcloning of genes into expression vector.Cells were transformed and selected as described previously [16].G. intestinalis cells expressing the recombinant proteins fused to a hemagglutinin tag (HA) at the C-terminus were fixed and stained for immunofluorescence microscopy with a mouse monoclonal anti-HA antibody.A secondary AlexaFluor-488 (green) donkey anti-mouse antibody was used.

Preparation of recombinant proteins and enzyme assay
The coding region of GiOR-1 and [2Fe2S]ferredoxin was subcloned into pET42b and pET3a (Invitrogen), respectively and  expressed in Escherichia coli BL21.The bacteria were induced with 0,5 mM IPTG (isopropyl-b-D-thiogalactopyranoside) and grown at 37uC in LB medium.For expression of GiOR-1, the LB medium was supplemented with 250 mM flavin adenine dinucleotide (FAD) and 250 mM flavin mononucleotide (FMN), whereas the LB medium supplemented with 400 mM ferric ammonium citrate was used for expression of ferredoxin.After induction, the cells were incubated overnight at 4uC.The harvested cells were homogenized, and soluble fraction was obtained by centrifugation at 250,0006 g, 1 h, 4uC.The his-tagged GiOR-1was affinity purified under native conditions using a Ni-NTA column (Qiagen) according to manufacture's protocol.Ferredoxin was isolated by gel filtration chromatography using a BioLogic HR system (BioRad).Enzyme activity of GiOR-1 was assayed spectrophotometrically at 25uC in anaerobic cuvettes under nitrogen atmosphere.The activity was monitored as a rate of NADPH or NADH (0,25 mM) oxidation in the presence of dichlorophenol-indolephenol (0,1 mM) or ferredoxin at 340 nm, or as a rate of methyl viologen (2 mM) reduction at 600 nm.NADPH oxidase activity was measured under aerobic conditions at 340 nm.The enzymatic activity was determined in phosphate buffer (100 mM KH 2 PO 4 / KOH, 150 mM NaCl, pH 7,4).Protein concentration was determined according to Lowry method.Table S1 Complete list of proteins identified by LC MS/MS in mitosomal fractions labelled by iTRAQ reagents.(PDF)

Supporting Information
Table S2 List of Giardia proteins within the mitosomal distribution range (MiD) identified by LC MS/MS.(PDF) Table S3 Identification of protein families using PfamA+B databases.(PDF) Table S4 Predictions of cellular localization.(PDF) Table S5 Orthology phylogenetic profililng.Genomes of G. intestinalis and Rickettsia typhi were compared using orthology phylogenetic profile tool at GiardiaDB.(PDF) Table S6 List of primers that were used for subcloning of genes into expression vector pONDRA to investigate subcellular localization of corresponding gene products in G. intestinalis.(PDF)

Figure 1 .
Figure 1. Isolation of mitosome-rich fractions.(A) Trophozoites were disrupted and centrifuged to remove unbroken cells, nuclei and cytoskeletal residue.The high-speed pellet was resuspended in sucrose buffer, layered onto an Optiprep density gradient, and centrifuged overnight.Five distinct fractions were obtained.(B) Fractions were collected and analyzed by SDS-PAGE and Western blot.The mitosomal marker GiIscU was detected in fractions #3 and #4 using a polyclonal rabbit antibody.(C-D) Electron microscopy of subcellular fractions.Fraction #3 (C) contains numerous vesicles of variable sizes, while fraction #4 (D) contains vesicles of more homogeneous sizes.Arrows indicate mitosomes.doi:10.1371/journal.pone.0017285.g001

Figure 2 .
Figure 2. iTRAQ ratios define protein subcellular localization.Proteins in fractions #3 and #4 isolated on the Optiprep gradient were labeled with the iTRAQ-114 and iTRAQ-115 reagents, respectively, analyzed by LC MS/MS, and sorted according to the iTRAQ ratios.Mitosomal marker proteins (red diamonds) fall into a narrow range of iTRAQ ratios.Green diamondsdindicate the zone of proteins considered as mitosomal candidates (mitosomal distribution, MiD).doi:10.1371/journal.pone.0017285.g002 Mascot score, Mascot total ion score for the identified protein.Coverage, number of unique peptides per identified protein.MiD, mitosomal distribution.Proteins are marked ''Y'' if their distributions in fractions #3 and #4 of the Optiprep gradient (measured by the iTRAQ ratio) were within the range between Cpn10 and IscU and the window that extended in both directions by half of the distance between these markers.Proteins with ratios outside of this range are indicated with ''N''.TargetP and PsortII were used to predict the subcellular location of Giardia proteins.S, secretory; N, non-secretory; M, mitochondrial; O, other.Exp.ver., experimental verification of protein localization using the pONDRA expression vector.The recombinant tagged proteins were localized by fluorescence microscopy.M, mitosome; ER, endoplasmic reticulum; O, other; ?inconclusive.MEMSAT3 and TMHMM were used to predict transmembrane domains.SGP, predicted soluble proteins are marked with number sign (#).Asterisk (*) is used where no data were available.(**) transformed Giardia did not express the recombinant tagged protein.doi:10.1371/journal.pone.0017285.t002 Mascot score, Mascot total ion score for the identified protein.Coverage, number of unique peptides per identified protein.MiD, mitosomal distribution.Proteins are marked ''Y'' if their distributions in fractions #3 and #4 of the Optiprep gradient (measured by the iTRAQ ratio) were within the range between Cpn10 and IscU and the window that extended in both directions by half of the distance between these markers.Proteins with ratios outside of this range are indicated with ''N''.TargetP and PsortII were used to predict the subcellular location of Giardia proteins.S, secretory; N, non-secretory; M, mitochondrial; O, other.Exp.ver., experimental verification of protein localization using the pONDRA expression vector.The recombinant tagged proteins were localized by fluorescence microscopy.M, mitosome; ER, endoplasmic reticulum; O, other; ?inconclusive.MEMSAT3 and TMHMM were used to predict transmembrane domains.SGP, predicted soluble proteins are marked with number sign (#).Asterisk (*) is used where no data were available.(**) transformed Giardia did not express the recombinant tagged protein.doi:10.1371/journal.pone.0017285.t005

Figure 5 .
Figure 5. Schematic representation of protein import pathway in the mitosome of G. intestinalis.Components identified in mitosome are highlighted by color.Components that are known to participate in the protein import into mitochondria of animals and fungi are shown in grey colour.OM, outer membrane; IMS, intermembrane space; IM, inner membrane; TOM, translocase of outer membrane; SAM, sorting and assembly machinery; TIM, translocase of inner membrane; PAM, presequence translocase-associated motor; VAP, VAMP (vesicle-associated membrane protein)associated protein; VDAC, voltage-dependent anion channel; MPP, mitochondrial processing peptidase.doi:10.1371/journal.pone.0017285.g005

Table 1 .
Putative mitosomal proteins classified by predicted function: Iron-sulfur cluster assembly, chaperones, redox mechanism and protein translocation and processing.

Table 2 .
Putative mitosomal proteins classified by predicted function: transporters and proteins known to operate in endoplasmic reticulum and tramsport vesicles.

Table 3 .
Putative mitosomal proteins classified by predicted function: protein modification, cztosceletal and motor proteins.
Mascot score, Mascot total ion score for the identified protein.Coverage, number of unique peptides per identified protein.MiD, mitosomal distribution.Proteins are marked ''Y'' if their distributions in fractions #3 and #4 of the Optiprep gradient (measured by the iTRAQ ratio) were within the range between Cpn10 and IscU and the window that extended in both directions by half of the distance between these markers.Proteins with ratios outside of this range are indicated with ''N''.TargetP and PsortII were used to predict the subcellular location of Giardia proteins.S, secretory; N, non-secretory; M, mitochondrial; O, other.Exp.ver., experimental verification of protein localization using the pONDRA expression vector.The recombinant tagged proteins were localized by fluorescence microscopy.M, mitosome; ER, endoplasmic reticulum; O, other; ?inconclusive.MEMSAT3 and TMHMM were used to predict transmembrane domains.SGP, predicted soluble proteins are marked with number sign (#).Asterisk (*) is used where no data were available.(**) transformed Giardia did not express the recombinant tagged protein.doi:10.1371/journal.pone.0017285.t003

Table 4 .
Putative mitosomal proteins classified by predicted function: various metabolic processes, lipid metabolism.
Mascot score, Mascot total ion score for the identified protein.Coverage, number of unique peptides per identified protein.MiD, mitosomal distribution.Proteins are marked ''Y'' if their distributions in fractions #3 and #4 of the Optiprep gradient (measured by the iTRAQ ratio) were within the range between Cpn10 and IscU and the window that extended in both directions by half of the distance between these markers.Proteins with ratios outside of this range are indicated with ''N''.TargetP and PsortII were used to predict the subcellular location of Giardia proteins.S, secretory; N, non-secretory; M, mitochondrial; O, other.Exp.ver., experimental verification of protein localization using the pONDRA expression vector.The recombinant tagged proteins were localized by fluorescence microscopy.M, mitosome; ER, endoplasmic reticulum; O, other; ?inconclusive.MEMSAT3 and TMHMM were used to predict transmembrane domains.SGP, predicted soluble proteins are marked with number sign (#).Asterisk (*) is used where no data were available.(**) transformed Giardia did not express the recombinant tagged protein.doi:10.1371/journal.pone.0017285.t004

Table 5 .
Putative mitosomal proteins classified by predicted function: miscellaneous.

Table 6 .
Putative mitosomal proteins classified by predicted function: miscellaneous -continued; hypothetical proteins.Mascot score, Mascot total ion score for the identified protein.Coverage, number of unique peptides per identified protein.MiD, mitosomal distribution.Proteins are marked ''Y'' if their distributions in fractions #3 and #4 of the Optiprep gradient (measured by the iTRAQ ratio) were within the range between Cpn10 and IscU and the window that extended in both directions by half of the distance between these markers.Proteins with ratios outside of this range are indicated with ''N''.TargetP and PsortII were used to predict the subcellular location of Giardia proteins.S, secretory; N, non-secretory; M, mitochondrial; O, other.Exp.ver., experimental verification of protein localization using the pONDRA expression vector.The recombinant tagged proteins were localized by fluorescence microscopy.M, mitosome; ER, endoplasmic reticulum; O, other; ?inconclusive.MEMSAT3 and TMHMM were used to predict transmembrane domains.SGP, predicted soluble proteins are marked with number sign (#).Asterisk (*) is used where no data were available.(**) transformed Giardia did not express the recombinant tagged protein.doi:10.1371/journal.pone.0017285.t006
Mascot score, Mascot total ion score for the identified protein.Coverage, number of unique peptides per identified protein.MiD, mitosomal distribution.Proteins are marked ''Y'' if their distributions in fractions #3 and #4 of the Optiprep gradient (measured by the iTRAQ ratio) were within the range between Cpn10 and IscU and the window that extended in both directions by half of the distance between these markers.Proteins with ratios outside of this range are indicated with ''N''.TargetP and PsortII were used to predict the subcellular location of Giardia proteins.S, secretory; N, non-secretory; M, mitochondrial; O, other.Exp.ver., experimental verification of protein localization using the pONDRA expression vector.The recombinant tagged proteins were localized by fluorescence microscopy.M, mitosome; ER, endoplasmic reticulum; O, other; ?inconclusive.MEMSAT3 and TMHMM were used to predict transmembrane domains.SGP, predicted soluble proteins are marked with number sign (#).Asterisk (*) is used where no data were available.(**) transformed Giardia did not express the recombinant tagged protein.doi:10.1371/journal.pone.0017285.t007