Identification of a divergent cytochrome c oxidase complex in the mitochondrial proteome of Toxoplasma gondii

The mitochondrion of apicomplexan parasites is critical for parasite survival, although the full complement of proteins that localize to this organelle has not been defined. Here we undertake two independent approaches to elucidate the mitochondrial proteome of the apicomplexan Toxoplasma gondii. We identify 421 mitochondrial proteins, many of which lack homologs in the animals that these parasites infect, and most of which are important for parasite growth. We demonstrate that one such protein, termed TgApiCox25, is an important component of the parasite cytochrome c oxidase (COX) complex. We identify numerous other apicomplexan-specific components of COX, and conclude that apicomplexan COX, and apicomplexan mitochondria more generally, differ substantially in their protein composition from the hosts they infect. Our study highlights the diversity that exists in mitochondrial proteomes across the eukaryotic domain of life, and provides a foundation for defining unique aspects of mitochondrial biology in an important phylum of parasites.


127
To determine whether mtBirA* could label mitochondrial proteins, we incubated mtBirA*-128 expressing parasites in medium supplemented with 1 mM biotin for 1 day. We labelled parasites 129 with Oregon green-conjugated avidin and observed labelling in the mitochondrion of biotin-130 supplemented parasites, but not in untreated parasites ( Figure 1F). 131 132 To observe the extent of protein biotinylation in the treated mtAPEX and mtBirA* parasites, we 133 extracted proteins from RH strain wild type (WT), mtAPEX or mtBirA* parasites treated with 134 either biotin-phenol and H 2 O 2 or with biotin. We separated these by SDS-PAGE and probed with 135 horse radish peroxidase (HRP)-conjugated neutravidin to label biotinylated proteins. In WT 136 cells, we observed two major bands of the expected sizes of natively biotinylated proteins in 137 these parasites ( Figure 1G; (van Dooren et al., 2008)). In the biotin-phenol treated mtAPEX 138 parasites, we observed labelling of several additional proteins, whereas in biotin-supplemented 139 mtBirA* parasites, numerous proteins were labelled ( Figure 1G). These data indicate that 140 mtAPEX-and mtBirA*-mediated biotinylation is occurring in these parasites. 141

142
To determine the specificity of labelling, we extracted proteins from treated WT, mtAPEX and 143 mtBirA* parasites and subjected these to affinity purification using streptavidin-conjugated 144 magnetic beads. We separated purified proteins by SDS-PAGE and probed with antibodies marker (E.T. and G.v.D., unpublished). We did not detect mtHsp60 or cyt c in the streptavidin 148 bound fraction in WT parasites treated with biotin-phenol and H 2 O 2 , or with biotin ( Figure 1H). 149 We detected bound mtHsp60, but not bound cyt c, in proteins extracted from both biotin-phenol-150 treated mtAPEX and biotin-treated mtBirA* parasites. This is consistent with the mitochondrial 151 labelling that we observe being specific for the mitochondrial matrix. 152 153 Quantitative proteomics to elucidate the mitochondrial matrix proteome. Having established 154 two independent approaches for specifically labelling mitochondrial matrix proteins, we next 155 undertook a label-free quantitative proteomic analysis of biotinylated proteins in treated 156 mtAPEX and mtBirA* parasites. First, we generated 3 independent cell lysate pools of WT and 157 mtAPEX cells treated with biotin-phenol and H 2 O 2 , and WT and mtBirA* cells treated with 158 biotin. Biotinylated proteins were purified from these lysates using streptavidin beads, reduced, 159 alkylated, and trypsin-digested before being identified using mass spectrometry (MS). Triplicate 160 samples were then processed through our in-house quantitation pipeline to determine the relative 161 abundance of each protein identified in the mtAPEX or mtBirA* samples as compared to WT 162 controls. These data are represented on a volcano plot as a fold-change (log 2 value) vs 163 significance of the change (-log 10 p value) (Figure 2A gondii mitochondrial proteome, we undertook a series of in silico and experimental analyses.

9
Proteins targeted to the mitochondrial matrix typically harbor an N-terminal amphipathic a-helix 173 that facilitates import into the organelle (van Dooren et al., 2016). We examined the T. gondii 174 mitochondrial proteome for proteins predicted to contain such an N-terminal targeting domain 175 using the rules-based computational prediction tool MitoProt II (Claros and Vincens, 1996). 176 Approximately 40% of proteins in the proteome had a strongly predicted N-terminal targeting 177 sequences (probability of mitochondrial import >0.9), and a further ~20% had a moderately 178 predicted targeting sequence (probability of import 0.5 -0.9; Figure  of the seven enzymes predicted to function in coenzyme Q synthesis, and two of the three 202 mitochondrial proteins involved in heme synthesis ( Figure 2D). Additionally, we identified all 203 five ubiquinone-reducing dehydrogenases of the mitochondrial inner membrane, and all 204 currently predicted subunits of cytochrome c reductase (Complex III) and cytochrome c oxidase 205 (Complex IV) that are encoded on the nuclear genome ( Figure 2D). We were unable to identify 206 cytochrome b, CoxI and CoxIII, proteins encoded on the mitochondrial genome apicomplexan 207 parasites. As expected, we did not identify the two isoforms of cytochrome c, both predicted to 208 localize to the intermembrane space. We identified the a, b, g and d subunits of the F 1 209 component of ATP synthase, but not the e subunit. 12 'hypothetical' proteins, and a further 140 (33%) had no previously defined role or experimentally 241 determined localization in T. gondii (Supplementary File 1). We attempted to localize 37 242 proteins selected at random from this 'uncharacterized' protein data set by introducing a 243 hemagglutinin (HA) epitope tag at the 3' end of the native locus of genes encoding these 244 proteins. We then undertook immunofluorescence assays to determine the localization of the 245 proteins, co-labelling with anti-TgTom40 as a marker for the mitochondrion (van Dooren et al., and Cryptosporidium parvum, and the chromerid Vitrella brassicaformis. Using this approach, 259 we identified homologs for 71% of T. gondii mitochondrial proteins in P. falciparum, 61% in B. 260 bovis, 28% in C. parvum, and 83% in V. brassicaformis ( Figure 4A). 261

13
We were next interested in the extent of novelty in the T. gondii mitochondrial proteome when 263 compared to non-apicomplexan eukaryotes. We examined conserved orthology groupings of the 264 421 proteins in the mitochondrial proteome and identified 418 proteins that clustered into 412 265 separate orthology groups (http://orthomcl.org; (Chen et al., 2006)). We identified 86 proteins 266 that were unique to T. gondii and closely related coccidians such as Neospora caninum, 243 267 proteins with orthologs in non-apicomplexan eukaryotes, and a set of 89 proteins that were found 268 only in apicomplexans and/or chromerids ( Figure 4B; Supplementary File 1). 269

270
Novel drug targets against apicomplexans are likely to emerge from proteins which lack 271 homologs in animals. We therefore conducted an orthology analysis comparing the T. gondii 272 mitochondrial proteome to other apicomplexans, chromerids and animals. We found that 51% of importance for a gene's contribution to parasite fitness. The Sidik et al study found that most 287 genes that were important for parasite growth had phenotype scores of below -2, and most 288 dispensable genes had phenotype scores of greater than -2 (Sidik et al., 2016). Based on this, we 289 categorised proteins in the mitochondrial proteome as dispensable (phenotype score >-2), 290 important (-2 to -4), or critical (<-4) for parasite growth. Notably, 35% of proteins from the 291 mitochondrial proteome were critical, and 39% were important, for parasite growth ( Figure 4A; outside the apicomplexan/chromerid lineage, and no predicted function, we embarked on a broad 302 project to characterise the importance and role of these proteins. In the remainder of this 303 manuscript, we focus on one such protein, annotated as TGGT1_264040, which (for reasons that 304 will become apparent) we termed TgApiCox25. TgApiCox25 belongs to an OrthoMCL ortholog 305 grouping that is restricted to apicomplexans, contains no recognisable functional domains, and is 306 important for parasite fitness. It has a predicted molecular mass of 25 kDa, and we confirmed its  Electrons are ultimately used to reduce O 2 , with the electron transport chain simultaneously 334 generating a proton gradient across the inner membrane. This proton gradient is then used to 335 drive the F-type ATP synthase, a rotary motor that phosphorylates ADP to form ATP, the energy 336 currency of cells. Defects in any of the processes involved in oxidative phosphorylation will lead 337 to defects in mitochondrial O 2 consumption. To test whether TgApiCox25 has a role in oxidative 338 phosphorylation, we established an assay to measure O 2 consumption by the parasite using a 339 Seahorse XFe96 extracellular flux analyzer. We grew rTgApiCox25 parasites in the absence of 340 ATc, or presence of ATc for 1-3 days then used the XFe96 analyzer to measure basal 341 mitochondrial O 2 consumption rates (mOCR) in extracellular parasites. This revealed a 342 significant, ~80% depletion in basal mOCR upon TgApiCox25 knockdown ( Figure 5D), 343 concomitant with knockdown of protein levels ( Figure 5A). 344 345 Treatment with the protonophore FCCP uncouples OCR from ATP synthesis and enables the 346 determination of maximal mOCR in parasites. We found that maximal mOCR was also depleted Nevertheless, we can use ECAR measurements as a general indication of parasite metabolic 367 activity. ECAR levels in WT parasites was approximately 40 mpH/min/1.5 x 10 6 parasites, and 368 slightly less in rTgApiCox25 parasite grown the absence of ATc. Growth of TgApiCox25 369 parasites for 2 or 3 days in ATc resulted in a slight increase in ECAR ( Figure 5E), indicating that 370 parasites remained metabolically active upon TgApiCox25 knockdown. As a control for non-371 metabolically active parasites, we treated WT parasites with the translation inhibitor 372 cycloheximide for 24 hr, which would be expected to deplete key metabolic enzymes in the 373 parasite. XFe96 measurements revealed that both mOCR and ECAR were depleted upon 374 cycloheximde treatment ( Figure 5E), consistent with a general loss of parasite metabolism 375 leading to simultaneous defects in mOCR and ECAR. We conclude that parasites remain 376 metabolically active in the absence of TgApiCox25. 377 378 Next, we asked whether knockdown of TgApiCox25 led to general defects in mitochondrial 379 morphology. We performed immunofluorescence assays labelling the mitochondrion in to mass spectrometry-based protein identification. Using this approach, we identify 12 proteins, 418 including TgApiCox25, that were enriched in the TgApiCox25-HA immunoprecipitation 419 compared to the TgTom40-HA immunoprecipitation ( Figure 6C; Table 1; Supplementary File 5). 420 Of these 12 proteins, three are annotated as being canonical components of cytochrome c oxidase 421 (COX, also known as Complex IV of the mitochondrial electron transport chain; Figure 6C; 422 Table 1) (Table  428 1). All of the proteins detected were identified in the mitochondrial proteome (Table 1;  Similarly, immunoprecipitation of TgCox2a-FLAG with anti-FLAG antibodies co-purified 443 TgApiCox25, but not TgAtpB or TgTom40 ( Figure 7B). Together, these data indicate that indicating the predicted molecular mass of the protein (Table 1). We experimentally localized 27 previously uncharacterised proteins from the T. gondii 507 mitochondrial proteome, finding that 22 of these localised to the mitochondrion (Figure 3). From 508 this, we estimate that ~80% of the 'uncharacterized' proteins from the proteome localize to the 509 mitochondrion. Our findings suggest a low false positive rate in the proteins identified from the 510 APEX proteome alone (~5%), but a higher false positive rate in the mtBirA* proteome (~20%). 511 Based on these analyses, we consider the 213 proteins from the mtAPEX and shared 512 mtAPEX/mtBirA* proteomes to be 'likely' mitochondrial proteins, and the 208 proteins found in 513 the mtBirA* proteome alone to be 'possible' mitochondrial proteins. we identified in the T. gondii COX complex had homologs in chromerids (Table 1), indicating 545 that even conserved mitochondrial processes in this group of organisms have a considerable 546 degree of novelty. Together, these data indicate that much of the mitochondrial biology in T. 547 gondii was present in the free-living ancestor that they share with chromerids. 548 549 Many (~40-50%) mitochondrial proteins in T. gondii lack apparent orthologs in animals and 550 other eukaryotes ( Figure 4B-C). Surprisingly, proteins found only in coccidians, or restricted to 551 apicomplexans and chromerids, were just as likely to be important for parasite growth as proteins 552 conserved across eukaryotic evolution ( Figure 4D). This suggests that many derived or unique 553 To understand the functions of apicomplexan-specific mitochondrial proteins that are important 559 for parasite growth, we have commenced a project to characterize these proteins. In this study, 560 we describe the characterisation of TgApiCox25, demonstrating that this is a component of the T. 561 gondii COX complex. TgApiCox25 is important for parasite growth and mitochondrial oxygen 562 consumption ( Figure 5). Knockdown of TgApiCox25 also leads to defects in the integrity of the 563 COX complex. In particular, TgApiCox25 knockdown leads to a depletion of TgCox2a 564 abundance, and also results in the appearance of a smaller, ~400 kDa TgCox2a-containing 565 complex ( Figure 7E-F). These data imply an important role for TgApiCox25 in the assembly 566 and/or stability of the COX complex. It remains unclear whether loss of TgApiCox25 leads to 567 loss of a ~200 kDa module from the complex, or whether TgApiCox25 knockdown leads to 568 defects in COX assembly, with the ~400 kDa complex representing an assembly intermediate.  (Table 1), while none of these have clear homologs in ciliates 597 (http://orthomcl.org), the eukaryotic lineage that is the sister taxon to myzozoans. This 598 suggests either a high degree of novelty in the proteins that comprise the T. gondii/myzozoan 599 COX complex, or that the sequences of ApiCox proteins have diverged to the extent that they are 600 no longer easily recognisable by sequence comparisons. Notably, similarity searches that 601 incorporate secondary structure information suggest that ApiCox25 and ApiCox23 may have 602 homology to Cox6a and Cox4, respectively, from animals. A priority in the field is to establish 603 the structure of the COX complex in myzozoans, which will reveal whether ApiCox proteins  Sigma clone M2), and rabbit anti-AtpB (1:500; Agrisera, catalog number AS05 085). 872 Horseradish peroxidase (HRP)-conjugated anti-mouse, anti-rat and anti-rabbit antibodies (Santa 873 Cruz) were used at 1:5,000 dilution. For probing for mouse antibodies on immunoprecipitation 874 western blots, HRP-conjugated anti-mouse TrueBlot ULTRA antibodies (eBioscience) were used 875 at 1:5,000 dilution. Neutravidin-HRP (Life Technologies) was used to detect biotinylated 876 proteins on membranes at 1:10,000 dilution. was removed and the plate washed twice in sterile water, before drying. 100 µl of the parasite 888 suspensions (1.5 x 10 6 parasites) were seeded into wells of the coated plate, and the plate was 889 centrifuged at 50 g for 3 min. An additional 75 µL of supplemented base medium was added to 890 each well following centrifugation. Parasites were kept at 37°C in a non-CO 2 incubator until the 891 start of experiment. Parasite oxygen consumption rates (OCR) and extracellular acidification 892 rates (ECAR) were measured using an Agilent Seahorse XFe96 Analyzer at 3 minute intervals. 893 To determine the maximal OCR, parasites were treated with 20 µM oligomycin A, B and C mix 894 (Sigma) to inhibit ATP synthase, then subsequently treated with 1 µM carbonyl cyanide-4-895 (trifluoromethoxy)phenylhydrazone (FCCP; Sigma). To determine the non-mitochondrial OCR, 896 parasites were treated with 10 µM antimycin A (Sigma) and 1 µM atovaquone (the minimal 897 concentration that preliminary experiments indicated is sufficient to maximally inhibit 898 mitochondrial OCR). The mitochondrial OCR (mOCR) was calculated by subtracting the non-899 mitochondrial OCR from the basal and maximal OCR values. A minimum of 4 wells were used 900 Table 1. Summary of the features of proteins identified in proteomic analysis of the TgApiCox25 complex. Similarity searches were performed using HMMER (https://www.ebi.ac.uk/Tools/hmmer/). The accession numbers listed were derived from http://EuPathDB.org (apicomplexan and chromerid species) or www.ncbi.nlm.nih.gov (all others). Abbreviations: Plasmodium falciparum (Pf), Cryptosporidium parvum (Cp), Vitrella brassicaformis (Vb), Saccharomyces cerevisiae (Sc), Homo sapiens (Hs), Arabidopsis thaliana (At).