Mitochondrial outer membrane proteome of Trypanosoma brucei reveals novel factors required to maintain mitochondrial morphology

Here we present a comprehensive proteomic analysis of the mitochondrial OM of the procyclic form


INTRODUCTION
Trypanosomatids are unicellular parasites that cause devastating diseases in both humans and animals. They include Trypanosoma brucei, two subspecies of which cause human sleeping sickness, as well as Trypanosoma cruzi and Leishmania spp., which are responsible for Chagas disease and leishmaniasis, respectively. The treatment of these diseases is still in an unsatisfactory state and new drugs are urgently needed (1).
Besides their clinical importance, some trypanosomatids are highly accessible experimental model systems to investigate general biological processes. Moreover, trypanosomatids appear to have diverged from all other eukaryotes very early in evolution and therefore show many unique features, some of which may reflect primitive traits that were present in the universal ancestor of all eukaryotes (2).
Many of these features concern the mitochondrion. Its genome consists of two genetic elements, the maxi--and the minicircles, which are highly topologically interlocked and localized to a discrete region within the organelle (3). Many mitochondrial genes represent cryptogenes whose primary transcripts have to be processed by extensive RNA editing in order to become functional mRNAs (4). The mitochondrial genome lacks tRNA genes indicating that trypanosomatids, unlike most other eukaryotes, import all mitochondrial tRNAs from the cytosol (5). The mitochondrial outer membrane (OM) of trypanosomatids has an unusual protein translocase, termed ATOM (6), that shares similarity to the canonical protein import pore Tom40 (7) but also to the bacterial Omp85-like protein family that is involved in protein translocation (6,8,9).
Trypanosomatids, unlike most other eukaryotes, have a single continuous mitochondrion throughout their life and cell cycle (10,11). Its morphology changes from a complex network in procyclic cells to a single tube--like structure in the bloodstream form (12). Nothing is currently known about how the different morphologies of its mitochondrion are established and maintained.
The changes in organellar shape correlate with large functional differences between the mitochondrion of the procyclic and the bloodstream forms. Only organelles of the procyclic stage are capable of oxidative phosphorylation, whereas in the bloodstream form, energy is produced by substrate level phosphorylation (13--15).
Recently, a proteomic study of the whole T. brucei mitochondrion detected 401, 196 and 283 proteins that could be assigned to mitochondria with high, medium and low confidence, respectively (16). A follow up study analyzed mitochondrial membrane fractions and identified 202 proteins that contain one or more predicted transmembrane helices and that were associated with mitochondria with various levels of confidences (17). This added 65 new proteins to the previously defined mitochondrial proteome. Moreover, the proteomes of the respiratory complexes (18) and the mitochondrial ribosomes (19) have also been characterized. However, an inventory of the mitochondrial OM is still lacking. In fact, the way the mitochondria were isolated in the studies described above suggests that they are depleted for OM proteins (20).
The OM separates the organelle from the cytosol. Detailed knowledge about the OM proteome is therefore a prerequisite for a comprehensive understanding of how the cytosol and mitochondria communicate and how the organelle is integrated into the metabolism of its host cell. The OM is the first barrier imported proteins that tRNAs face while they are transported into the mitochondrion.
Knowing its proteome will therefore also help to understand the molecular mechanisms of these two processes.
Presently, only four mitochondrial OM proteins are known in trypanosomatids. These are the voltage--dependent anion channel (VDAC) that serves as a metabolite transporter (21) and three components of the mitochondrial protein import system. The latter include the trypanosomal SAM50 orthologue, which mediates insertion of beta barrel proteins into the OM (22), ATOM, the general mitochondrial preprotein translocase (6) and pATOM36, which may serve as a receptor for a subset of imported proteins (23). The situation is only marginally better outside the trypanosomatids and the only examples where global proteomic analyses of the mitochondrial OM have been performed are the two fungal species Saccharomyces cerevisiae (24) and Neurospora crassa (25) and the plant Arabidopsis thaliana (26). These studies detected 82 and 30 resident OM proteins, respectively, in the fungal species and 42 proteins in plants.
Here we present a comprehensive proteomic analysis of the mitochondrial OM of the procyclic form of T. brucei. To that end, we established a purification procedure allowing the isolation of a highly enriched OM fraction. To identify bona fide OM proteins, we employed label--free quantitative mass spectrometry to establish abundance profiles of several hundred proteins across four and six subcellular fractions including highly purified OMs. This allowed us to identify 82 proteins that could be localized to the mitochondrial OM with high confidence. Ablation of three trypanosomatid-specific proteins of unknown function affects mitochondrial morphology and thus defines the first factors controlling mitochondrial morphology in T. brucei. Based on a recent global RNAi study (27), 9 OM proteins are essential for normal growth under all tested conditions including the disease--causing bloodstream form of the parasite indicating that they could be novel potential drug targets.

Cell culture
Both procyclic wildtype T. brucei strain 427 and transgenic T. brucei strain 29--13 were used in this study. All cell lines were grown in SDM79, which was supplemented with 5% (427) or 10% (29--13) fetal calf serum (Sigma), respectively. For OM purification cells were harvested at late log phase corresponding to a density of 3.0--4.0 x 10 7 cell/ml.

Purification of the mitochondrial OM of T. brucei
Cells were lysed under isotonic conditions by N2--cavitation (28). Mitochondrial vesicles were isolated by differential centrifugation and subsequent Nycodenz step gradients as described (20).
Mitochondrial vesicles were used as a starting point to prepare a highly enriched mitochondrial OM fraction using a modified version of the procedure described in (29,30) (see Fig. 1A for an overview). 100 mg of mitochondrial vesicles, isolated from 4--5 x 10 11 cells, were diluted to 10 mg/ml in swelling buffer consisting of 5 mM potassium phosphate, pH 7.2 containing 5 mM EDTA and 1 mM PMSF. The vesicles were kept under hypotonic conditions for 25 min on ice and allowed to swell. Subsequently, they were homogenized by 20 strokes using a Dounce homogenizer (Wheaton) with a loose fitting Teflon pestle to dislodge the OM from the inner membrane (IM). To separate IM vesicles and residual mitochondria from the lower density OM, the mixture was loaded on top of a 0/15/32/60% (w/v) sucrose step gradient containing buffer A (10 mM MOPS, KOH pH 7.2 and 2.5 mM EDTA). Centrifugation was done for 1 hr at 2°C with 100,000 g and both the bottom fraction (32/60% sucrose interface) corresponding to the IM as well as the top fraction (15/32% sucrose interface) corresponding to crude OM were collected. The latter was adjusted to 50% (w/v) sucrose in buffer A and overlaid with two layers consisting of an equal volume of buffer A containing 32% (w/v) sucrose and of buffer A lacking sucrose, respectively. The flotation gradient was centrifuged for 5 hrs at 2°C with 240,000 g and the fraction at the 0/32% sucrose interphase corresponding to pure OM was collected. Approximately 300 µg OM per 100 mg mitochondrial vesicles were obtained.
Crude ER fraction was prepared using the supernatant resulting from centrifugation after the isotonic N2--cavitation of the cells (20,28). This supernatant was subjected to a clearing spin for 30 min at 4°C with 30,000 g in order to remove remaining mitochondrial vesicles and cell debris. From this cleared supernatant the crude ER fraction was harvested by ultracentrifugation (2 hrs at 4°C with 100,000 g).

Data analysis
Mass spectrometric data of two independent experiments were processed separately using the software MaxQuant (version 1.2.0.18) (32,33). For protein identification, spectra were correlated with the Trypanosoma brucei protein database (TriTrypDB; www.tritrypdb.org; version 3.1) containing 9,826 protein entries using Andromeda (33). All searches were performed with tryptic specificity allowing up to two missed cleavages. Oxidation of methionine and acetylation of protein N--termini were considered as variable modification. No fixed modifications were considered. Raw data were recalibrated using the "first search" option of Andromeda with the full database employing a mass error of 20 ppm for precursor ions and 0.5 Da for fragment ions. Mass spectra were searched using the default settings of Andromeda. The mass tolerance for precursor and fragment ions was 6 ppm and 0.5 Da, respectively. A false discovery rate of 1% was applied on both the peptide and protein level. For retrieving information about protein abundance, the label--free protein quantification option in MaxQuant was enabled using default settings and the "match between runs"--option with a retention time window of two minutes. Only razor and unique peptides were considered for quantification. Protein intensity values were normalized to the total ion current (TIC) of the pure mitos and the OM/ER fraction of experiment 1 and 2, respectively. To establish protein profiles, intensity values of a given protein were plotted against the different subcellular fractions and normalized to one. For statistical analysis, protein abundance profiles were hierarchically clustered using Euclidian distances and complete linkage in the R software environment. Only proteins identified and quantified in both datasets were considered for clustering analysis; mass spectra of single peptide identifications are shown in Suppl. Figure S1.

Miscellaneous
Protein concentrations were determined using the BCA protein assay reagent (Pierce). Protein samples were analyzed by conventional SDS--PAGE and immunoblotting using the Odyssey Clx infrared imaging system (Li--Cor, Biosciences). Proteins were C--terminally c--MYC--tagged; SAM35 was C--terminally tagged with the HA--epitope using a pLew100--based construct (34,35). Inducible RNAi cell lines of POMP9, 14 and 40 were prepared using pLew100--based stem--loop constructs

Purification of the mitochondrial OM from T. brucei
The mitochondrial OM purification strategy for T. brucei was adapted according to protocols from yeast (29,30) and is outlined in Fig. 1A. In the first step, procyclic T. brucei cells are lysed under isotonic conditions using N2--cavitation (28). After DNA digestion and differential centrifugation, a crude mitochondrial fraction (crude mitos) is obtained and further purified by Nycodenz gradient sedimentation yielding pure mitochondria (pure mitos), which have an intact OM (20). This fraction is subjected to a swelling step under hypotonic conditions and extensively homogenized. The aim of this step is to disrupt the OM and to dislodge it from the IM. The resulting homogenate is resolved on a sucrose gradient yielding two main fractions. The denser, more abundant one is enriched for IM markers, whereas the less dense, minor fraction corresponds to crude OM. The latter is collected and subjected to a sucrose flotation gradient again resulting in two fractions, one consisting of pure OM and the other of both ER and OM (OM/ER). Starting from 4 x 10 11 cells, approximately 300 µg of pure OM fraction was obtained.
Selected fractions of the purification (Fig. 1A, grey boxes) were analyzed by silver staining (Fig. 1C) and on immunoblots using a panel of antisera directed against marker proteins for submitochondrial and different subcellular compartments (Fig. 1B). The results show that the previously characterized trypanosomal OM proteins, VDAC (21), SAM50 (22) and ATOM (6), are strongly enriched in the pure OM fraction. A quantification of the immunoblots indicates that the enrichment factors for the three proteins between whole cells and the pure OM fraction are between 50 to 60--fold (it should be noted that the three last lanes of the immunoblots in Fig. 1B contain 20--fold less protein than all other lanes). Markers for the intermembrane space (IMS), the IM and the matrix are strongly depleted in the pure OM fraction, indicating that there is only very little contamination with components from the other mitochondrial subcompartments. Moreover, the pure OM fraction is essentially free of cytosol, glycosomes and components of the cytoskeleton.
In the case of the ER, both BiP, a component of the ER lumen, and a tagged transmembrane subunit of the trypanosomal glycosylphospatidylinositol transamidase complex (GPI16:HA) (38) were analyzed. Both ER proteins are depleted in the pure OM fraction but to a lesser extent than the markers for the other subcellular compartments. Interestingly, the most dense fraction recovered from the sucrose flotation gradient, termed OM/ER, is enriched for the OM as well as for both ER markers.
Finally, the pure OM fraction was also subjected to electron microscopy using uranyl acetate for negative staining. This analysis revealed a population of single membrane--bounded vesicles of similar morphology but different diameters consistent with a highly enriched mitochondrial OM fraction (Suppl. Fig. S2) (39).

Mass spectrometric analysis of subcellular fractions
A eukaryotic cell contains many different membranes of which the mitochondrial OM is of minor abundance. Moreover, the OM is known to be tightly associated with the IM and to interact with parts of the ER (40). This makes it difficult, if not impossible, to prepare OM that is free of contaminants, the most likely sources of which are the IM and the ER. In order to identify bona fide OM proteins and to distinguish them from contaminants, we employed protein abundance profiling by high resolution mass spectrometry (41--43). Two independent experiments were performed to characterize the mitochondrial OM proteome. In the first, crude and pure OM fractions as well as pure mitos and OM/ER fractions were analyzed (Fig. 1A, see *). In the second experiment the analysis was extended by including the crude ER and the crude IM fraction (Fig. 1A, see °). All fractions were analyzed by high resolution LC/MS to provide a comprehensive dataset for label--free quantitative protein profiling. Table 1  Tables S1A--D and S2A--D). Altogether, more than 100,000 MS/MS spectra were acquired resulting in the identification of 2,142 unique proteins with a false discovery rate of ≤ 1%. In experiment 2, the inclusion of the crude ER and the crude IM fraction led to the identification of a further 587 proteins that were not found in experiment 1. Protein identification numbers obtained for the pure mitos and pure OM fraction from the two independent experiments were highly consistent with a variation of 4% and 11%, respectively. Moreover, the corresponding overlaps of proteins identified in these fractions were 67.6% and 59.1%, underscoring the high consistency of our data.

Localization of proteins by correlation of abundance profiles
To further distinguish between OM constituents and co--purified contaminants, we relied on quantitative feature analysis in MS survey scans using the MaxQuant algorithm (32,33). This allows to determine intensity values of all proteins identified in the four and six different subcellular fractions obtained during the two OM purifications.
Altogether, 1062 proteins were quantitatively followed through both experiments (Suppl . Table S3) and the corresponding normalized abundance profiles were calculated. Figure 2 depicts the abundance profiles of marker proteins of the mitochondrial OM (VDAC, SAM50 and ATOM) and IM (COX 4 and CYT C1) as well as the ER (BiP and GPI16). Notably, MS--based protein abundance profiles of subcellular and suborganellar marker proteins were highly reproducible between the two independent experiments (Fig. 2) and essentially congruent with the ones obtained by immunoblotting (Fig. 1B). OM proteins characteristically exhibited low intensities in crude ER, pure mitos and IM fractions, while showing a distinct maximum in pure OM fractions. Notably, the measured relative intensity of OM proteins typically increased by more than two--fold between crude and pure OM fractions, thereby facilitating the reliable identification of even minor OM constituents. Contaminants of the OM mainly derived from the ER and the IM (Fig. 1B). Such components could be well distinguished from the OM proteome based on their abundance profiles exhibiting distinct maxima in either the crude ER or the crude IM fraction as exemplarily shown for the marker proteins GPI16, BiP and COX 4, CYT C1, respectively (Fig. 2).
Through this statistical approach, a cluster of 83 putative OM proteins was obtained. However, since it contained the inner membrane localized ADP/ATP carrier, which is the most abundant mitochondrial membrane protein and therefore the most obvious contaminant of the pure OM fraction, this protein was excluded from the cluster resulting in an OM proteome of 82 proteins (Fig.   3A, Suppl. Table S3). All cluster components are listed in Tables 2 and 3, including information about their accession number, name, predicted protein domains, molecular weight and number of putative transmembrane domains. They comprise all four previously known mitochondrial OM proteins: VDAC (21), SAM50 (22), ATOM (6) and pATOM36 (23). Using BLAST, we found 38 proteins of the OM cluster that show similarities to proteins of known function in other eukaryotes and two trypanosomatid--specific components of the OM protein import system (6,23). The function of 42 proteins (51% of the OM proteome) is unknown. These proteins were termed POMPs for present in the outer mitochondrial membrane proteome and numbered. 16 POMPs have known protein domains (Table 2) and 33 proteins (40% of the total proteome) are conserved in trypanosomatids only (Table 2, asterisks).
The power of the protein abundance profiling approach is illustrated by the fact that, except for the ADP/ATP carrier mentioned above, the OM proteome lacks all the common contaminants such as the highly abundant ribosomal proteins, eukaryotic translation elongation factor 1a and tubulins.
The same applies to proteins known to reside in other submitochondrial compartments.
Interestingly MICU1 (44) the regulatory partner of the recently characterized mitochondrial Calcium uniporter MCU (45,46) was also found in the OM cluster. MICU1 is peripheral membrane protein that interacts with inner membrane protein MCU. Our results suggest the exciting possibility that MICU1 unlike assumed before might not be localized in the IM but together with MCU may form contact sites between the two mitochondrial membranes.
It should be considered in this context that our OM cluster might also contain dually localized OM proteins as long as their second localization is outside mitochondria.
Even though it was not the aim of our study, it is worth to note that our quantitative proteomics survey further allows to define a second cluster of 282 proteins that constitutes the mitochondrial IM proteome (Fig. 3A). Its components showed highest enrichment in the crude IM fraction and significant, but lower intensity values in the pure mitochondrial fraction (see Fig 2, experiment 2).
Interestingly, 37% of these proteins are not found in the least stringent list of IM proteins that has been published before (17).
To demonstrate the consistency of data obtained from both experiments, the median and the middle 50% of the profiles of all the 82 and 282 proteins present in the OM and IM cluster are depicted in Figure 3B and 3C, respectively. Notably, the OM and IM proteome comprise only 4 and 13% of all the proteins identified in this study. Moreover, these constituents represent only a minor fraction of all the proteins identified in highly purified OM and crude IM fractions by high performance LC/MS (see Table 1).
In summary, the data presented here underscore the effectiveness of our quantitative proteomics strategy for gaining comprehensive and accurate information about submitochondrial proteomes.

Verification of OM localization of 7 POMPs
Despite the virtual absence of obvious contaminants, we decided to verify the localization of 7 POMPs. Protease treatment is the standard assay to establish the intramitochondrial localization of a protein and the only proteins sensitive to proteolytic cleavage in intact mitochondria are OM proteins. Transgenic cell lines were produced expressing c--MYC--tagged versions of the proteins and mitochondria were isolated under isotonic conditions, which retain the integrity of the OM (28).
Isolated organelles were incubated with proteinase K, which is expected to selectively cleave OM proteins and to leave proteins of the other submitochondrial fractions intact. However, the protease shaving assay is only informative when performed with mitochondria that have an intact OM. The intactness of the OM was assayed using immunoblots that determine the ratio of the matrix-localized heat shock protein 70 (mHSP70) to the intermembrane space protein cytochrome c (CYT C). Mitoplasts are devoid of CYT C since their OM is disrupted (Fig 4A, first lane) which results in a high mHSP70/CYT C ratio. In intact mitochondria, however, CYT C remains in the intermembrane space, yielding a much lower mHSP70/CYT C ratio. The graph in Fig. 4A shows that isolated mitochondria from all the cell lines used for the protease shaving experiments have a low mHSP70/CYT C ratio when compared to mitoplasts indicating that their OM is still intact. Fig. 4B shows that all tested POMPs can be efficiently digested by proteinase K in intact mitochondria confirming that they are bona fide mitochondrial OM proteins. The same immunoblots were also decorated with an antiserum against the IM mitochondrial carrier protein MCP--5 (47), which serves as a loading control. In addition to the 7 POMPs, we also verified the OM localization of the trypanosomal SAM35 orthologue.

Relative abundance of trypanosomal OM proteins
The OM proteome was further assessed by calculating log10 intensity plots of all proteins uniformly identified in pure mitos fractions to provide an estimate of their relative abundance. The abundance distribution of the 982 proteins, including 69 OM proteins, was found to span approximately five orders of magnitude ( Fig. 5A and Suppl. Table S4). The remaining 13 putative OM proteins including also the trypanosomal SAM35 orthologue could not be detected in pure mitos fractions and are therefore not included in the intensity plots. However, based on the intensity values observed in pure OM fractions, a two orders of magnitude lower abundance of SAM35 compared with SAM50 is estimated. These findings underscore the need for effective suborganellar fractionation to detect even minor, but functionally important OM constituents. Among the four known OM proteins, SAM50 exhibits the lowest abundance, whereas VDAC represents the most abundant OM constituent. Interestingly, ATOM was found to be a mitochondrial OM protein of high abundance, while pATOM36 (23) is approximately by a factor of 10 less abundant ( Figure 5B).
Among the seven new OM constituents whose localization was confirmed by protease treatment (see Fig. 4), POMP10 was found to be of similar abundance as ATOM, while POMP25 was found to be approximately three orders of magnitude less abundant than POMP10 (Fig. 5C).

The trypanosomal OM proteome contains mainly unstudied proteins
Two proteomic studies characterizing the total and the membrane proteome of the T. brucei mitochondrion have been published before (16,17). However, the previously detected 662 proteins that could be localized to mitochondria with high and medium confidence included only 14 proteins (17%) that are present in our OM proteome. This is consistent with the fact that the previous proteomic analyses were performed with hypotonically isolated organelles that are depleted for the OM (20). Of all experimentally identified OM proteins, an estimated 55 have not previously been associated with the mitochondrion and to our knowledge have not been studied before.
A significant fraction of the OM proteins (30 proteins, 36.6% of the OM proteome) contains at least one predicted transmembrane helix. The relative large proportion of OM proteins that lack apparent transmembrane helices is expected since the OM purification protocol in our study does not select for integral membrane proteins but detects peripheral membrane proteins as well.
Moreover, some OM proteins are beta--barrel membrane proteins, which lack classical membrane spanning domains.

Conserved features of mitochondrial OM proteomes
The mitochondrial OM proteome has been analyzed from two fungal and one plant species. In Neurospora crassa, 30 proteins were localized to the mitochondrial OM (estimated coverage 65%) (25). Orthologues of virtually all of these proteins were also found in the much larger mitochondrial OM proteome of S. cerevisiae which consists of 82 different proteins (estimated coverage 85%) (24). In the plant A. thaliana, 42 proteins were assigned to the OM (estimated coverage 88%) (26).
The mitochondrial OM proteome of T. brucei as determined in this study consists of 82 different proteins. Fungi, plants and trypanosomes belong to three different eukaryotic supergroups (48).
Having defined the OM proteomes from these species therefore allows to identify a set of proteins that are likely found in the mitochondrial OM of all eukaryotes (Table 4). These universally conserved OM proteins are VDAC, Tom40, SAM50, SAM35 and the GEM1 GTPase. VDAC is the most abundant OM proteins in all species investigated and is responsible for metabolite transport across the OM. Interestingly, all systems appear to have several different VDAC isoforms. In T. brucei, initially only one VDAC was found (21) but more elaborate bioinformatic analysis revealed the presence of two additional highly diverged VDAC--like proteins whose functions are unknown (49).
Tom40 is the general protein import channel. Also the T. brucei mitochondrial OM contains a protein that shows limited similarity to Tom40 (7). The protein was termed ATOM and is unique in that it shares also similarity to the bacterial Omp85--like protein family (6,8,9). SAM50 and SAM35 are components of the highly conserved beta--barrel protein insertion machinery (50). GEM1 appears to be involved in the regulation of ER/mitochondria contact sites and mitochondrial morphology, respectively (40). Its universal presence in the OM proteomes suggests that ER/mitochondria contact sites are of importance in all eukaryotes.
The sizes of the OM proteomes of unicellular T. brucei and S. cerevisiae are identical and a global comparison reveals many shared features. The fraction of OM proteins attributed to the functional groups of "lipid metabolism", "signaling/G--proteins" and "protein folding/turnover" are of similar size in both organisms. Within the group of "lipid metabolism", the fatty acyl CoA synthetase subunit 4, the fatty aldehyde deydrogenase and the squalene synthase are found in both OM proteomes. In the group of "signaling/G--proteins", a number of small GTPase of the rab subfamily are present in both the T. brucei and the yeast OM proteome. In the "protein folding/turnover" group, the CAAX prenyl protease is detected in both proteomes. Moreover, components of the ubiquitin protein degradation system, although not the same ones, are found in the OM of both species. All in all, we find that the orthologues of 17 different OM proteins of T. brucei, corresponding to 20.7% of the OM proteome, are also present in the yeast mitochondrial OM (Table   4). Except for VDAC and GEM1 as well as the components of the protein import system, the function of these conserved OM proteins is unknown.

Unique features of the OM proteome of T. brucei
The comparative analysis of the trypanosomal and the yeast mitochondrial OM proteomes also uncovers striking differences. Proteins categorized in the functional groups of "protein import" and "dynamics/morphology" are strongly underrepresented in the mitochondrial OM proteome of T.
brucei, whereas the proteins associated with metabolite transport are overrepresented (Fig. 6A).
The latter is mainly due to the presence of three ABC transporters that are not found in the OM proteome of yeast.
The paucity of recognizable protein import factors in the trypanosomal OM proteome is in line with the fact that even the normally conserved protein import pore of the OM (6) and a receptor--like component of the protein import system (23) are highly diverged in trypanosomes. The absence of orthologues of the other protein import factors is therefore not surprising. However, the number and the kind of proteins that need to be imported into the trypanosomatid mitochondrion are comparable to other unicellular organisms indicating that the protein import system, while different, will be of similar complexity than in other eukaryotes. We therefore expect that some of the POMPs are trypanosomatid--specific components of the mitochondrial protein import system.
The virtual absence of proteins that belong to the group "mitochondrial dynamics/morphology" in the T. brucei OM proteome is surprising since the main components of the mitochondrial fission and fusion machineries are conserved (51). However, it is in line with the fact that mitochondrial dynamics in trypanosomatids shows unique features.

Ablation of POMP9, 14 and 40 affects mitochondrial morphology.
The single mitochondrion of T. brucei divides in two only during cytokinesis to allow its transmission to the daughter cells (52,53). In trypanosomatids, mitochondrial fusion has never been observed and mitochondrial fission must be coordinated with the cell cycle. As in other systems, fission requires the dynamin--like protein, DLP1, which is the single member of the dynamin protein family found in trypanosomatids (53,54). Moreover, ablation of DLP1 in T. brucei caused the accumulation of cells that are blocked in cytokinesis suggesting that mitochondrial division acts as a checkpoint for cell division (53). Maintaining a single mitochondrion at all times, both the regulated fission prior to cytokinesis and the morphology change during the life cycle are expected to require as yet unknown factors associated with the trypanosomal mitochondrial OM.
We therefore expect that some of the POMPs will be such factors. As a first test for this prediction, we prepared inducible RNAi cells directed against POMP9, POMP14 and POMP40. RNAi--mediated ablation shows that all three proteins are essential for normal growth in procyclic cells (data not shown). Mitochondrial morphology was analyzed by immunofluorescence microscopy using anti--ATOM antiserum (Fig. 7). Individual ablation of all three proteins dramatically altered the network-like mitochondrial structure in procyclic trypanosomes. The strongest phenotype was observed for the POMP40 RNAi cells. Their mitochondria collapsed from a highly branched structure to an essentially single straight tubule, reminiscent of the mitochondrial shape observed in the bloodstream form (55). Also for POMP9 and POMP14, induction of RNAi causes a collapse of the network. However, the phenotype is less severe and a more condensed form of the network appears to be maintained. In conclusion, we report POMP9, POMP14 and POMP40 as being the first factors known to regulate mitochondrial morphology in T. brucei. These results illustrate the value of the OM proteome to identify novel factors that directly or indirectly control mitochondrial morphology.

The mitochondrial OM proteome contains essential proteins
Recently, a global high--throughput analysis was performed, which mapped the fitness costs associated with the inducible RNAi--mediated ablation of mRNAs encoded by essentially all trypanosomal ORFs (27). The effect the RNAi has on growth of T. brucei was measured under four different conditions: in procyclic forms after 3 days of induction, in the bloodstream forms 3 and 6 days after induction and in differentiating cells.
Nine proteins from the T. brucei OM proteome are essential under all tested conditions and two proteins each are essential for the bloodstream and the procyclic form, respectively. Eight OM proteins appear to be specifically required for the differentiation of the bloodstream to the procyclic form (Fig. 6B, Supp. Table S5). Preliminary validation experiments using individual RNAi cell lines against some OM proteins indicate that the global analysis is probably biased towards false negatives suggesting that the number of essential OM proteins may even be higher (data not shown).
The nine proteins that are essential under all conditions are of special interest since they are required for core essential functions. Five of them, namely ATOM, pATOM36 and POMP6, POMP14 and POMP22 are specific for trypanosomatids. The function of these three POMPs is not known, but the only essential function of mitochondrial OM proteins described so far is protein import. This suggests that these POMPs may represent as yet unknown factors of the mitochondrial protein import machinery that are specific for trypanosomatids. Moreover, the trypanosomal mitochondrion lacks tRNA genes and therefore needs to import all of its tRNAs from the cytosol (5).
Mitochondrial translation was shown to be essential not only for the procyclic but also for the bloodstream form of T. brucei (56) indicating that the same must be true for mitochondrial tRNA import. Thus, we expect that also the machinery for mitochondrial tRNA import is encoded by POMPs that encode core essential functions. It is even possible, as has been suggested for the mitochondrial IM (57,58), that some protein import factors may also be required for tRNA import.
Interestingly, of the three newly discovered factors that are required to maintain normal mitochondrial morphology (Fig. 7), only POMP14 but not POMP9 and POMP40 are required for core essential functions as defined by (27)(Suppl. Table S5).

Concluding remarks
Trypanosoma brucei is a unicellular eukaryote that causes devastating diseases in humans and animals. It has a unique evolutionary history resulting in a mitochondrion showing many distinct features when compared to other eukaryotes. Several of these features are linked to the OM, which forms the interface between the mitochondrion and the cytosol. They include unusual protein and tRNA import systems as well as a unique mitochondrial shape that is highly regulated. The 82 OM proteins that were identified in our study will be a treasure trove to gain insight into these processes. Most of the 82 OM proteins have never been identified before and 33 are specific for trypanosomatids. The latter are of special interest since a comparison with the yeast OM proteome reveals a strong underrepresentation of protein import factors and orthologues of proteins that regulate the shape of the mitochondrion. This suggests that many of the trypanosomal OM proteins of unknown function recovered in the present study may either be required for protein import or for maintenance of mitochondrial morphology. In line with the latter prediction, we used RNAi to identify three OM proteins of unknown function as the first candidate factors for regulation of mitochondrial morphology in T. brucei.
Finally, it is worth noting that ATOM, pATOM36 and the three POMPs 6, 14 and 22, are essential for the bloodstream form of T. brucei. These proteins might therefore represent excellent novel drug targets since they are specific for trypanosomatids and essential for the disease--causing form of the parasite. Studying their function will not only be of interest for basic science but is likely to provide novel targets for future drug development.      for growth of either bloodstream (0--0--1--0) or procyclic forms (1--1--0--0) only, or (iii) for differentiation of bloodstream to procyclic forms only (1--1--1--0). The binary code in parentheses is described in (27). For details see Supp. Table S5.     Aminoacyl--tRNA hydrolase ----1 1 1 ATOM of T. brucei has similarity to both the bacterial Omp85--like protein family and to eukaryotic Tom40. 2 The putative small GTPases identified in T. brucei all belong to the rab subfamily but cannot be categorized further with adequate levels of confidence.