Phosphatidylserine synthase 2 and phosphatidylserine decarboxylase are essential for aminophospholipid synthesis in T rypanosoma brucei

Summary Phosphatidylethanolamine (PE) and phosphatidylserine (PS) are ubiquitously expressed and metabolically interconnected glycerophospholipids in eukaryotes and prokaryotes. In Trypanosoma brucei, PE synthesis has been shown to occur mainly via the Kennedy pathway, one of the three routes leading to PE synthesis in eukaryotes, while PS synthesis has not been studied experimentally. We now reveal the importance of T. brucei PS synthase 2 (TbPSS2) and T. brucei PS decarboxylase (TbPSD), two key enzymes involved in aminophospholipid synthesis, for trypanosome viability. By using tetracycline‐inducible down‐regulation of gene expression and in vivo and in vitro metabolic labeling, we found that TbPSS2 (i) is necessary for normal growth of procyclic trypanosomes, (ii) localizes to the endoplasmic reticulum and (iii) represents the unique route for PS formation in T. brucei. In addition, we identified TbPSD as type I PS decarboxylase in the mitochondrion and found that it is processed proteolytically at a WGSS cleavage site into a heterodimer. Down‐regulation of TbPSD expression affected mitochondrial integrity in both procyclic and bloodstream form trypanosomes, decreased ATP production via oxidative phosphorylation in procyclic form and affected parasite growth.


Introduction
Glycerophospholipids are major building blocks of all biological membranes. The relative abundance of the different classes and their subclass and molecular species compositions not only modulate membrane characteristics, such as fluidity, curvature and membrane tension, but also affect properties of membraneassociated proteins (reviewed by Lee, 2004). The aminophospholipid classes, phosphatidylethanolamine (PE) and phosphatidylserine (PS), are present in membranes of most eukaryotes and prokaryotes (reviewed by Vance, 2008;Vance and Tasseva, 2013). PE has been shown to be involved in a wide range of biological processes, including cell division (Emoto et al., 1996) and protein folding (Bogdanov and Dowhan, 1998), and represents the ethanolamine donor for the synthesis of glycosylphosphatidylinositol (GPI) anchors (Menon and Stevens, 1992) and other protein modifications (Ichimura et al., 2000;Signorell et al., 2008a;Cullen and Trent, 2010). PS functions as an important precursor for some pools of PE (reviewed by Vance and Steenbergen, 2005), acts as a cofactor for enzymes involved in signaling pathways (Takai et al., 1979), and its exposure at the cell surface is an early event in apoptosis (Fadok et al., 1992) and serves as a critical cofactor in blood clotting (Bevers et al., 1983). The pathways for the synthesis of PS and PE are often coupled, but the contributions of the individual pathways and their interconnections differ considerably among eukaryotes (reviewed by Vance and Tasseva, 2013).
Decarboxylation of PS represents a main route for PE synthesis in many organisms (Borkenhagen et al., 1961; reviewed by Schuiki and Daum, 2009). It is the major pathway for PE production in bacteria and the major route in Saccharomyces cerevisiae, however it is also active in mammals and plants. In addition, in mammalian cells, plants and yeast, PE can be formed via the CDP-ethanolamine branch of the Kennedy pathway (Kennedy and Weiss, 1956; reviewed by Vance, 2008;Gibellini and Smith, 2010). A third pathway for PE synthesis in many eukaryotes involves head group exchange with PS (Suzuki and Kanfer, 1985; reviewed by Vance, 2008). PS decarboxylases (PSDs) in prokaryotes and eukaryotes consist of two different types. Type I PSDs are present in eukaryotic mitochondria and bacteria whereas type II PSDs comprise eukaryotic enzymes located in the endomembrane system (Golgi, endoplasmic reticulum (ER), vacuole/tonoplast) (reviewed by Schuiki and Daum, 2009). PSDs are usually transmembrane proteins that are active as heterodimers composed of an a-and a b-subunit. The subunits are generated from a proenzyme via autocatalytic cleavage at a conserved recognition motif close to the C-terminus (GSS/T), generating a long N-terminal (transmembrane) b-subunit and a short C-terminal asubunit Kitamura et al., 2002). PSDs contain an unusual pyruvoyl prosthetic group, which is generated at the amino terminus of the asubunit from the serine residue located at the cleavage site of the proenzyme. Although the two PSD subunits remain tightly associated after the cleavage, they are not covalently linked (reviewed by van Poelje and Snell, 1990;Schuiki and Daum, 2009).
Bacteria and mammalian cells contain a single type I PSD, while S. cerevisiae express a mitochondrial type I PSD (responsible for 80% of PSD activity) and a type II PSD localized in the Golgi (reviewed by Vance and Steenbergen, 2005;Schuiki and Daum, 2009). Yeast mutants lacking both PSD enzymes are auxotrophic for ethanolamine, which is used for PE synthesis via the CDP-ethanolamine pathway . Plants contain a mitochondrial psd1 and two type II PSDs localized in the endomembrane system (Nerlich et al., 2007). Interestingly, unique forms of PSD have been identified in protozoan parasites. An ER-localized type I PSD has been described in Plasmodium falciparum (Baunaure et al., 2003), representing the only known nonmitochondrial type I PSD so far, whereas a soluble PSD secreted into the parasitophorus vacuole has been described in Toxoplasma gondii (Gupta et al., 2012). PS can be synthesized by two different pathways. PS synthase (PSS) uses CDP-diacylglycerol (CDP-DAG) and serine as substrates, while PS synthase 1 (PSS1) and PS synthase 2 (PSS2) generate PS via calciumdependent head group exchange reactions using PC and PE, respectively, as substrates. PSS enzymes are membrane-bound and have been detected in membrane fractions of bacteria (Matsumoto, 1997) and in the outer mitochondrial and ribosomal fraction of yeast cells (Yamashita and Nikawa, 1997). Depletion of PSS causes a conditional lethal temperature-sensitive growth phenotype in Escherichia coli (Ohta and Shibuya, 1977) and renders yeast auxotrophic for ethanolamine/choline (Nikawa and Yamashita, 1981). In contrast, mammalian PSS1/2 enzymes have been localized to mitochondriaassociated membranes (MAMs) (Stone and Vance, 2000). MAMs represent a particular membrane fraction with distinct biochemical properties harboring proteins that are involved in the synthesis and transport of lipids and mediate import of PS into mitochondria for decarboxylation to PE (reviewed by Vance, 2014). Comparative schemes with biosynthetic connections between PC, PE, and PS in eukaryotes are shown in Supporting Information Fig. S1.
The pathways for de novo synthesis of PE have also been studied in the very early-branching eukaryote Trypanosoma brucei (reviewed by Serricchio and B€ utikofer, 2011;Farine and B€ utikofer, 2013). T. brucei is the causative agent of human African trypanosomiasis, also called sleeping sickness, and nagana a related livestock disease. The diseases are fatal unless treated and represent a major cause of poverty in sub-Saharan Africa (Fevre et al., 2008). Trypanosomes alternate between an insect vector, the tsetse, where they replicate as procyclic forms in the midgut and as epimastigotes in the salivary glands, and a mammalian host, where they replicate as long slender bloodstream forms (Vickerman, 1985).
It has been shown previously that the CDPethanolamine branch of the Kennedy pathway is essential for growth of T. brucei procyclic and bloodstream forms in culture (Gibellini et al., 2008;Signorell et al., 2008b;Gibellini et al., 2009;Signorell et al., 2009). In addition, experimental data in procyclic forms (Signorell et al., 2008b) together with bioinformatic analysis of the T. brucei genome suggested that two other possible routes may contribute to PE formation in T. brucei: decarboxylation of PS by putative TbPSD (Tb927.9.10080) and head group exchange with PS by putative TbPSS2 (Tb927.7.3760). However, the respective enzymes have not been characterized experimentally and their essentiality has not been investigated. In the present study, we characterize TbPSD as mitochondrial type I PSD. Its down-regulation affects cell growth, mitochondrial integrity and ATP production by oxidative phosphorylation. In addition, we demonstrate that base exchange between PE and PS is catalyzed by TbPSS2 and represents the only pathway for PS production in T. brucei. Expression of TbPSS2 is essential for normal growth of T. brucei in culture. membrane spanning regions (Phobius, TriTrypDB) and a domain between amino acids 135 to 412 belonging to the PSS pfam family (PF03034). The PSS family comprises base exchange enzymes that replace the existing head group of a phospholipid with L-serine, such as mammalian PSS1 and PSS2 (Finn et al., 2014). PSS enzymes are mechanistically not related to PS synthases from yeast (CHO1), which are members of the CDP-alcohol phosphatidyltransferase protein family. On the amino acid level, bioinformatics prediction tools (Pairwise Sequence Alignment, EMBOSS, EMBL-EBI) revealed 23.7% and 26.2% identity (37.1% and 39.4% similarity) between TbPSS2 and human PSS1 (isoform 1) and PSS2, respectively, but only 11.8% identity (19.8% similarity) between TbPSS2 and S. cerevisiae PSS, which is consistent with TbPSS2 representing a base exchange enzyme. This was confirmed by overexpressing TbPSS2 in E. coli (Supporting Information Fig. S2A) and assaying a membrane preparation for PSS activity in the presence of various substrates (Supporting Information Fig. S2B). Low PSS activity was observed in the presence of CDP-DAG as substrate, likely due to the action of endogenous E. coli PSS. In contrast, in the presence of PE, strong head group exchange activity with [ 3 H]serine was observed in TbPSS2-overexpressing membranes compared to control membranes.

Down-regulation of TbPSS2 inhibits growth of procyclic form trypanosomes
Essentiality of TbPSS2 for parasite growth was assessed using tetracycline-inducible RNAi-mediated down-regulation of TbPSS2. After 3 days of tetracycline induction, TbPSS2 mRNA levels showed efficient downregulation ( Fig. 1A) and parasite growth in the presence of tetracycline was reduced compared to uninduced cells after 6 days of culture, demonstrating that A. Growth of control (solid line) and TbPSS2-depleted (dotted line) procyclic form trypanosomes. The data points represent mean values from single measurements using two different clones. The inset shows a Northern blot analysis of TbPSS2 mRNA levels (left two lanes) in parasites incubated in the absence (2) or presence (1) of tetracycline for 2 days. rRNA levels are shown as loading controls (right two lanes). B. Subcellular localization of TbPSS2 in procyclic forms. Trypanosomes expressing 3xc-myc-tagged TbPSS2 were immunostained with anti-c-myc antibody. BiP and mtHSP70 were used as endoplasmic reticulum and mitochondrial marker, respectively. Scale bars indicate 10 mm. expression of TbPSS2 is essential for normal growth of T. brucei procyclic forms in culture.

TbPSS2 co-localizes with the ER marker BiP
Localization of TbPSS2 in T. brucei procyclic forms was studied by expressing an inducible N-terminally 3xcmyc-tagged form of TbPSS2. Immunofluorescence microscopy revealed good co-localization of 3xc-myc-TbPSS2 with the ER marker BiP, whereas no colocalization was observed with mitochondrial mtHSP70 (Fig. 1B).
TbPSS2 is the only route for PS synthesis in T. brucei To study the importance of TbPSS2 on PS synthesis in procyclic forms, RNAi-suppressed parasites were labeled with [ 3 H]serine and lipids were analyzed by one-and two-dimensional thin layer chromatography (TLC) in combination with radioisotope scanning, fluorography and lipid phosphorus determination. Incubation of trypanosomes with [ 3 H]serine is expected to not only label PS and PE (via decarboxylation of [ 3 H]PS), but also the major sphingolipids, inositol phosphorylceramide (IPC) and sphingomyelin (SM), via serine palmitoyltransferase (Signorell et al., 2008b). Since degradation of [ 3 H]serine-labeled sphingolipids via sphingosine-1-phosphate lyase may result in the formation of [ 3 H]ethanolamine-phosphate, which is prevalent in Leishmania (Zhang et al., 2007), which in turn can be used for [ 3 H]PE synthesis via the Kennedy pathway, an inhibitor of serine palmitoyltransferase, myriocin, was used to block de novo synthesis of sphingolipids. As shown in Fig. 2A, and consistent with the Leishmania results (Zhang et al., 2007), the addition of myriocin completely blocked de novo formation of IPC and SM in T. brucei procyclic forms in culture, resulting in labeling of PS and PE only. These conditions were subsequently used to study [ 3 H]serine-labeling in parasites after down-regulation of TbPSS2. The results show that RNAi against TbPSS2 for 3 days inhibited incorporation of [ 3 H]serine into newly synthesized PS by 94.8 6 1.0% (mean value 6 standard deviation from three independent Phosphatidylserine metabolism of T. brucei 415 experiments) in intact parasites (Fig. 2B) and by 91.3 6 1.0% (mean value 6 standard deviation from three independent experiments) in digitonin extracts from TbPSS2-depleted parasites (Fig. 2C). Interestingly, [ 3 H]PE formation was not affected by downregulation of TbPSS2 ( Fig. 2B) (remaining at 92.5 6 14.6% of control levels; mean value 6 standard deviation from three independent experiments) and no labeling of PE was observed in digitonin extracts from control and TbPSS2-depleted cells (Fig. 2C). In addition, quantification by two-dimensional TLC and lipid phosphorus determination of total phospholipids after down-regulation of TbPSS2 for 4 days revealed a reduction in the steady-state levels of PS from 3.0 6 1.1% of total phospholipids in control parasites to 0.9 6 0.2% in TbPSS2-depleted cells (mean values 6 standard deviations from four independent experiments), indicating decreased activity of TbPSS2 resulting in lower PS levels. This finding is consistent with the observation that PS could no longer be labeled with [ 3 H]serine in vivo in trypanosomes after down-regulation of TbPSS2 (Supporting Information Fig. S3). Together, these results identify TbPSS2 as the only route for de novo PS synthesis in T. brucei.
In addition, to study whether the N-terminally tagged TbPSS2 used for sub-cellular localization (see Fig. 1B) is functional, digitonin extracts from parasites expressing (inducible) 3xc-myc-TbPSS2 were assayed for enzyme activity. The results show increased (i.e. 2.2 6 0.3-fold; mean value 6 standard deviation from three independent experiments) [ 3 H]PS formation in extracts from TbPSS2 over-expressing parasites compared to extracts from the same cell line before induction of 3xc-myc-TbPSS2 expression (Fig. 2D), indicating that tagged TbPSS2 is catalytically active.

Tb927.9.10080 encodes a type I PS decarboxylase
Using prediction programs and amino acid sequence comparison tools, we identified Tb927.9.10080 as a candidate gene encoding putative T. brucei type I PSD (TbPSD). The sequence revealed a predicted mitochondrial targeting sequence (iPSORT prediction; Bannai et al., 2002) and a single transmembrane domain in the N-terminal part of the protein (SMART; Schultz et al., 1998). The pfam domain PS Decarboxylase (PF02666) was located between amino acids 125 and 348 (Finn et al., 2014). Amino acid sequence alignments of TbPSD with E. coli, P. falciparum, T. gondii, Arabidopsis thaliana and human PSDs showed between 19.2 and 24.6% identity and revealed the presence of a putative proteolytic PSD cleavage recognition motif, WGSS (amino acids 316-319) (Fig. 3A). The same sequence motif is also present in putative PSD homologs of other kinetoplastids, including Leishmania major (LmjF.35.4590) and Crithidia fasciculata (CFAC1.300097400). The predicted proteolytic cleavage motif of TbPSD is identical to that of P. falciparum PSD, rather than the typical GST motif of higher eukaryotes ( (Dowhan, 1997;Voelker, 1997; reviewed by Schuiki and Daum, 2009). To analyze whether TbPSD is proteolytically processed in vivo, we expressed a C-terminally 3xHA-tagged copy of Tb927.9.10080 in T. brucei procyclic and bloodstream forms. Analysis by SDS-PAGE followed by immunoblotting revealed two bands corresponding to the expected sizes of the tagged proenzyme (i.e. before proteolytic processing) and the asubunit after proteolytic cleavage at the predicted recognition site (Fig. 3B). To further study whether the GSS motif of TbPSD is the cleavage site, we expressed a Cterminally 3xHA-tagged copy of TbPSD in which Ser318 was replaced by alanine. As shown in Fig. 3, changing the cleavage site from GSS to GAS inhibited processing of the proenzyme, confirming the GSS motif as proteolytic cleavage site of TbPSD.

TbPSD localizes to the mitochondrion
Analysis by immunofluorescence microscopy showed good co-localization of TbPSD-3xHA with MitoTracker, a mitochondrial membrane potential-dependent red-fluorescent dye, in both procyclic and bloodstream forms (Fig.  4). Co-staining with other organelle markers (TbGRASP for Golgi, BiP for ER, Trypanopain for lysosome) showed no co-localization with TbPSD (results not shown). Mitochondrial localization was also observed for the uncleaved S318A TbPSD-3xHA mutant (Fig. 4), demonstrating that correct localization of TbPSD is independent of proteolytic processing of the proenzyme into its subunits.

Down-regulation of TbPSD inhibits trypanosome growth
Essentiality of TbPSD for T. brucei parasites in culture was studied by tetracycline-inducible RNAi-mediated gene silencing. Down-regulation of TbPSD mRNA was analyzed by Northern blotting after 3 days of RNAi induction and showed efficient depletion of TbPSD mRNA in both procyclic and bloodstream form parasites ( Fig. 5A and B). After 4 days of RNAi against TbPSD, procyclic forms showed a growth defect compared to control parasites (Fig. 5A), while growth of TbPSDdepleted bloodstream forms was only slightly affected (Fig. 5B).

TbPSD depletion affects mitochondrial integrity and ATP production via oxidative phosphorylation
To study possible effects of TbPSD depletion on mitochondrial integrity, procyclic and bloodstream form trypanosomes were analyzed by fluorescence microscopy. Mitochondria were stained using two different markers: the membrane potential-dependent dye MitoTracker Red and an antibody against mitochondrial heat shock protein 70 (mtHSP70). While control cells showed the typical branched mitochondrial network in procyclic forms and the tubular mitochondrial structure in bloodstream forms, trypanosomes depleted of TbPSD after 6 days of down-regulation revealed abnormal staining ( Fig. 6A and B). In both life cycle forms, down-regulation of TbPSD resulted in the appearance of distinct and brightly fluorescent spots, an indication of mitochondrial fragmentation.
Subsequently, we measured mitochondrial ATP production in control and TbPSD-depleted procyclic forms in vitro using digitonin-solubilized membranes. Succinate was used as substrate to measure ATP formation via oxidative phosphorylation whereas 2-ketoglutarate was used to determine ATP synthesis via substrate level phosphorylation (Allemann and Schneider, 2000). The results showed a decrease in ATP production via oxidative phosphorylation of more than 75% in TbPSDdepleted mitochondria isolated from parasites after 6 days of RNAi compared to controls (Fig. 6C). In contrast, ATP production via substrate level phosphorylation was not affected (Fig. 6D).

PE and PS levels are not affected by down-regulation of TbPSD
It has previously been suggested that decarboxylation of PS may contribute to PE formation in T. brucei procyclic forms (Signorell et al., 2008b). The availability of RNAi parasites against TbPSD now allowed us to experimentally address the importance of this pathway in both procyclic and bloodstream forms. H]PS) in parasites after RNAi against TbPSD were unchanged compared to control uninduced cells (5.5 6 1.3 vs. 6.0 6 1.0 and 15.6 6 2.8 vs. 14.5 6 2.0 respectively; mean values 6 standard deviations from three independent experiments). In addition, we saw no major differences in total phospholipid composition between control and TbPSD-depleted parasites after analysis of lipid extracts by two-dimensional TLC and phosphorus quantification (Supporting Information Fig. S4). Based on our findings that RNAi against TbPSD induces changes in mitochondrial structure, affects ATP production (Fig. 4) and reduces procyclic parasite growth (Fig. 5) (Fig. 2C and D).
Interestingly, we observed that the amounts of [ 3 H]PS formed during labeling with [ 3 H]serine decreased during a subsequent 4 h chase in the absence of [ 3 H]serine (Fig. 7B). In contrast, labeling of [ 3 H]PE remained unchanged during the chase (Fig. 7B)  A, B. Mitochondrial structure and membrane potential were analyzed by immunofluorescence in procyclic (panel A) and bloodstream form (panel B) parasites. Mitochondria from control and TbPSD-depleted trypanosomes (induced with tetracycline for 6 days) were live-stained with MitoTracker, then cells were fixed on slides and co-stained with an antibody against a mitochondrial marker protein (mtHSP70). DNA was stained with 4 0 ,6 0 -diamidino-2-phenylindole (in blue). Scale bars indicate 10 mm. C, D. ATP production by oxidative phosphorylation (C) or at substrate level (D) was measured in digitonin extracts of control and TbPSDdepleted trypanosomes (induced with tetracycline for 6 days). Antimycin and atractyloside were added to inhibit complex III (panel C) and ADP/ATP translocase (panel D), respectively. Data are from two independent experiments performed in duplicates. Error bars indicate means 6 standard deviations. The asterisk indicates significant difference with control (*P < 0.0001, unpaired student's t test  (Rontein et al., 2001) and Plasmodium parasites (Elabbadi et al., 1997) are capable of decarboxylating serine to ethanolamine via serine decarboxylase, we investigated if a putative T. brucei serine decarboxylase activity may convert [ 3 H]serine to [ 3 H]ethanolamine, which would then enter the Kennedy pathway to form [ 3 H]PE. However, using previously published protocols (Elabbadi et al., 1997), we were unable to detect such an activity in T. brucei procyclic forms. In addition, we found no candidate gene in the T. brucei genome with homology to plant serine decarboxylase. It should be noted that the gene responsible for the observed serine decarboxylase activity in Plasmodium has not been reported.

TbPSD is functional in E. coli
To study whether Tb927.9.10080 encodes a functional PS decarboxylase, the full-length form was recombinantly expressed in E. coli. [ 3 H]PS was then added to a membrane preparation and incubated for 10 min. Subsequently, lipids were extracted and separated by HPTLC and radiolabeled lipids visualized by fluorography. As seen in Fig. 8A by the ADH1 promoter were introduced into the S. cerevisiae psd1Dpsd2Ddpl1D strain. The constructs include (1) a full length TbPSD, (2) N-terminal deletion mutants of TbPSD (D48 and D83) and (3) a chimera TbPSD where the N-terminal 139 amino acids of the yeast PSD1 mitochondrial targeting sequence are fused to D83TbPSD. The yeast mutant strains are devoid of functional PSD enzymes and require ethanolamine for growth. Expression of the various constructs of TbPSD in the mutant strain failed to suppress the growth defect in the absence of exogenous ethanolamine (Supporting Information Fig. S5). The results indicate that none of the TbPSD forms were functional when expressed in yeast.
In addition, we expressed V5-tagged full length TbPSD and Sc-TbPSD chimera using pYES2.1 vector under the strong GAL1 promoter in psd1Dpsd2Ddpl1D strain. Similar to the findings above, the V5-tagged constructs also failed to suppress the growth defect (Supporting Information Fig. S6). Immunoblot analysis of cell extracts containing TbPSD-V5 fusion and Sc-TbPSD-V5 chimera using anti-V5 antibody show that TbPSD proenzymes can be detected, however the signals were weak compared to a yeast PSD1-V5 control (Supporting Information Fig. S6). Interestingly, the processed alpha subunits were barely detected, indicating poor processing of the pro-enzyme into its active form. Together, these results indicate that TbPSD constructs were not functional in yeast because they were poorly expressed and not effectively processed.

Discussion
Although the multiple pathways for PE biosynthesis are conserved among organisms, their contributions vary significantly. Several eukaryotes and most prokaryotes use decarboxylation of PS as main route for PE synthesis, while mammalian cells use both PS decarboxylation and the CDP-ethanolamine branch of the Kennedy pathway for PE formation. T. brucei parasites are unique in that the Kennedy pathway represents the only pathway for de novo production of PE (reviewed by Farine and B€ utikofer, 2013). Its disruption, by depletion of any of the four enzymes catalyzing the individual reactions, leads to inhibition of growth and parasite death (Gibellini et al., 2008;Signorell et al., 2008b;Gibellini et al., 2009;Signorell et al., 2009). The presence of PS decarboxylase and PS synthase activities in T. brucei has been suggested in several reports (Menon et al., 1993;Rifkin et al., 1995;Signorell et al., 2008b), but TbPSD activity was later shown to be absent in bloodstream trypanosomes (Gibellini et al., 2009;Richmond et al., 2010). However, the enzymes were not characterized biochemically and the contributions of these pathways to PE formation and their importance for parasite viability were not investigated. We now report the identification and characterization of TbPSS2 and TbPSD. By using tetracycline-inducible RNAi against the individual enzymes, we demonstrate that their expression in T. brucei is essential for normal parasite growth in culture.
Our results using in vivo and in vitro [ 3 H]serine labeling demonstrate that depletion of TbPSS2 in procyclic form trypanosomes completely blocks de novo formation of PS and reduces cellular PS levels, identifying TbPSS2 as the sole route for PS synthesis in T. brucei. PSS enzymes belong to two categories: in bacteria and yeast, PSS catalyzes PS formation using serine and CDP-DAG as substrates, whereas in mammals and plants PS is formed via head group exchange (reviewed by Schuiki et al., 2010;Vance and Tasseva, 2013). Thus, our identification of a base exchange enzyme in the ancient eukaryote T. brucei is remarkable and is in accordance with recent results from T. brucei CDP-DAG knockout trypanosomes showing that PS production is largely unaffected after knocking down CDP-DAG synthesis (Lilley et al., 2014). Comparisons of the primary sequence of TbPSS2 with its predicted homologs in other kinetoplastids, including T. cruzi and Leishmania spp., indicate that PS synthesis in these parasites is also catalyzed by a head group exchange enzyme (Ramakrishnan et al., 2013).
Mammalian PSS1/2 enzymes have been localized to mitochondria-associated membranes (MAMs) (Stone and Vance, 2000). Enzyme activities were found in the ER and enriched in MAMs, while immunofluorescence studies using c-myc-tagged PSS1/2 showed colocalization with an ER marker (Stone and Vance, 2000). Consistent with the results in mammalian cells, using immunofluorescence microscopy we localized a C-terminally 3xc-myc-tagged copy of TbPSS2, which was shown to be catalytically active, in the ER of T. brucei procyclic forms. At present, it is not known whether MAMs are also present in T. brucei parasites.
Incubation of T. brucei parasites with [ 3 H]serine has been shown to not only label sphingolipids and PS, but also PE (Menon et al., 1993;Signorell et al., 2008b). Conversion of PS to PE has been observed in many organisms before and has been attributed to PSD enzymes (Kanfer and Kennedy, 1964;Butler and Morell, 1983;Vance and Vance, 1986;Elabbadi et al., 1997;Choi et al., 2012). Our results demonstrate that T. brucei parasites express a mitochondrial type I PSD enzyme, TbPSD, which undergoes similar proteolytic processing as its mammalian (Kuge et al., 1996) and yeast (Horvath et al., 2012) homologs. RNAimediated down-regulation of TbPSD affected mitochondrial morphology in both procyclic and bloodstream forms, reduced ATP production via oxidative phosphorylation in procyclic trypanosomes and reduced parasite growth in culture. When expressed in E. coli, but not in S. cerevisiae, TbPSD was active. However, as reported before for T. brucei bloodstream forms (Gibellini et al., 2009) H]PE in T. brucei is unlikely formed via PS followed by its decarboxylation to PE, but by a different mechanism involving soluble factors that are lost upon extraction of parasites with digitonin. The presence of a serine decarboxylase activity in T. brucei could explain our findings, however, we were unable to detect such an activity using published procedures (Elabbadi et al., 1997). In addition, mining the T. brucei predicted proteome with plant-type serine decarboxylases revealed no reasonable sequences. Since TbPSD is essential in T. brucei, we hypothesize that it may be involved in limited and locally restricted synthesis of PE in the mitochondrion, perhaps by producing a subset of PE molecular species important for mitochondrial structure and function, whereas bulk mitochondrial PE may be generated via the Kennedy pathway. Alternatively, we can't exclude the possibility that TbPSD acts on a different substrate from PS, such as phosphatidylthreonine as recently identified in T. gondii (Arroyo-Olarte et al., 2015).

RNAi-mediated gene silencing
Putative TbPSD mRNA was down-regulated in procyclic and bloodstream trypanosomes by RNAi-mediated gene silencing using stem-loop constructs containing a puromycin resistance. A 374 bp fragment containing the last 361 bp of the TbPSD ORF and the first 13 bp of the TbPSD 3 0 UTR was amplified by PCR using primers PSD_RNAi_Fw GCGCCCAAGCTTGGATCCCTCTGTCCTTCCATTAAACG and PSD_RNAi_Rv CTAGGCTCTAGACTCGAGTTGTGCA-TACTGCCTACTTG (restriction sites underlined) and cloned into the tetracycline-inducible vector pMS14  resulting in plasmid pLF1610RNAi. For PSS2 RNAi in procyclic forms, a 550 bp fragment of the TbPSS2 ORF was amplified using primers fwd3760 GTGAAGCTTGGATCCACGATTATTGTGAGGGATTG and rev3760 GTGAGATCTGAGCTCACAAAAGGTTCACCACAC TC and cloned into the tetracycline-inducible vector pALC14 resulting in plasmid pJJ3760RNAi. Selection of the gene sequence for RNAi was done with RNAit, a prediction algorithm designed to prevent potential cross-talk and hence off-target effects (Redmond et al., 2003). Before transfection into trypanosomes, plasmids were linearized using NotI enzyme.

Stable transfection of trypanosomes
Trypanosomes were harvested at mid-log phase, washed once in phosphate-buffered saline (PBS; 137 mM sodium chloride, 2.7 mM potassium chloride, 10 mM disodium phosphate, 2 mM monopotassium phosphate, pH 7.4) and suspended in 100 ml TbBSF buffer (90 mM sodium phosphate, 5 mM potassium chloride, 0.15 mM calcium chloride, 50 mM HEPES, pH 7.3) (Schumann Burkard et al., 2011) previously mixed with 10 mg of linearized plasmid (pLF1610HA, pLF1610RNAi, pJJ3760RNAi and pLF3760cmyc). Electroporation was performed in 100 ml Nucleocuvette using Lonza 4D Nucleofector System (pulse code FI-115, "Primary Cell P3" solution). Recombinant clones were obtained by limited dilutions and selected with 1.75 mg ml 21 phleomycin for pLF1610RNAi or 2 mg ml 21 puromycin for the other vectors. Proper integration of the constructs was confirmed by PCR using primers binding upstream of the recombination sites and at the end of the inserted genes. Expression of tagged proteins or generation of double-stranded RNA was induced by addition of 1 mg ml 21 tetracycline to the culture medium.

Northern blot analysis
Total RNA was extracted from 4 x 10 7 procyclic or bloodstream form parasites using the Total SV RNA extraction Kit (Promega). 20 lg of RNA were loaded and run on a 1% agarose gel. The gel was stained with ethidium bromide and rRNA amounts were assessed as loading control. After transfer onto a Hybond-N 1 nylon membrane (Amersham Pharmacia Biotech) using 10x SSC buffer (150 mM trisodium citrate, pH 7.0, containing 1.5 M sodium chloride), RNA was cross-linked by UV irradiation. The membrane was probed with a [ 32 P]-labeled 374 bp probe of the TbPSD ORF/3'UTR or a [ 32 P]-labeled 550 bp probe of the TbPSS2 ORF, generated using the Prime-a-Gene labeling system (Promega). Detection was done by autoradiography using BioMax MS films (GE Healthcare, Buckinghamshire, UK) in combination with intensifying screens.
Alternatively, MitoTracker-stained and paraformaldehyde-fixed cells were used for co-localization studies using the following immunofluorescence protocol.
For immunofluorescence microscopy, trypanosomes were collected and washed before being allowed to adhere onto a microscopy slide (Thermo Scientific) for 10 min. Cells were fixed with 4% paraformaldehyde, washed 33 with cold PBS and permeabilized with 0.2% Triton X-100 in PBS. Blocking was performed with 2% bovine serum albumin in PBS for 30 min followed by 45 min incubation of first antibody diluted in blocking solution. The following antibodies were used: mouse monoclonal a-HA.11 16B12 (Covance, 1:250 dilution), mouse a-c-myc (Santa Cruz Biotech, 1:200 dilution) or rabbit a-c-myc (Bethyl Laboratories, 1:200 dilution), mouse a-mtHSP70 (kindly provided by Paul Englund, Baltimore, 1:1000 dilution) and rabbit a-BiP (kindly provided by Jay Bangs, Buffalo, 1:2000 dilution). After washing, secondary antibodies goat anti-mouse and antirabbit AlexaFluor 594 and 488 (Invitrogen, 1:800 dilution in blocking solution) were added for 45 min in either combinations. After washing and air-drying, cells were mounted with Vectashield containing DAPI. Fluorescence microscopy was performed with a Leica DMI6000 B inverted microscope using a 603 oil objective. Pictures were acquired, processed and 3D-deconvolved with the Leica LAS AF Version 2.1.0 software (Leica Microsystems CMS GmbH).

[ 3 H]Serine labeling and lipid analysis
Procyclic form trypanosomes (5-10 3 10 7 ) were pre-treated with 2.5 mM myriocin for 30 min and labeled with 30 lCi [ 3 H]serine for 4 h. For pulse/chase experiments, parasites labeled for 4 h were collected, washed once and resuspended in fresh medium without radioactivity. The chase was conducted for 2 to 6 h. Lipids were extracted according to the work by Bligh and Dyer (1959) and separated by onedimensional thin layer chromatography (TLC) (Silica gel 60 Phosphatidylserine metabolism of T. brucei 423 plate, Merck) using solvent system I (chloroform: methanol: acetic acid: water (25:15:4:2, by vol.)). Products from the aqueous phase were resolved using solvent system II (0.6% NaCl in water: methanol: 25% ammonium hydroxide (10:10:1, by vol.)). Radioactive standards were run on each plate alongside the samples to be analyzed. Radioactivity was detected by scanning the dried plate with a radioisotope detector (Berthold Technologies) and quantified using the Rita Star software provided by the manufacturer. For two-dimensional TLC followed by lipid phosphorus quantification or fluorography, 4 3 10 8 procyclic trypanosomes were labeled overnight with 60 lCi [ 3 H]serine, lipids were extracted and separated with solvent system III (chloroform:methanol:25% ammonium hydroxide:water (90:74:12:8, by vol.)) for the first dimension and solvent system IV (chloroform:methanol:acetone:acetic acid:water (40:15:15:12:8, by vol.)) for the second dimension. Radioactive plates were dried, sprayed with EN 3 HANCE Spray (PerkinElmer) and exposed to BioMax MS films (GE Healthcare, Buckinghamshire, UK) for 3 days at 2758C. For quantification of lipid using phosphorus determination, individual spots were scraped from TLC plates and quantified as described before (Signorell et al., 2008a).

Preparation and labeling of membranes-containing digitonin extracts
Digitonin extracts were prepared as described elsewhere (Schneider et al., 2007). Procyclic form trypanosomes (1 3 10 8 cells) were harvested, washed once in SBG buffer (150 mM Tris-HCl pH 7.9, 20 mM glucose, 20 mM NaH 2 PO 4 ) and suspended in SoTE buffer (20 mM Tris-HCl pH 7.5, with 600 mM sorbitol and 2 mM EDTA) containing a final concentration of 0.025% (wt/vol) digitonin. After 5 min of incubation on ice, mitochondria were isolated by centrifugation at 6200g. After removal of the supernatant, membranes from digitonin extract were resuspended in 500 ll SoTE and labeled with 15 lCi [ 3 H]serine for 4 h. Lipids were extracted and analyzed as described above.
ATP production assay ATP production in digitonin extracts was measured as previously described (Allemann and Schneider, 2000). About 5 mM of succinate or 2-ketoglutarate and 67 lM of ADP were added to membranes isolated by digitonin extraction from 9.5 3 10 7 procyclic trypanosomes and incubated for 30 min at room temperature. Antimycin (inhibitor of complex III) and atractyloside (inhibitor of ADP/ATP translocase) were pre-incubated with digitonin extracts for 10 min at a concentration of 2.7 and 43 lM, respectively. ATP concentrations were determined using ATP Bioluminescence Assay Kit CLS II (Roche, Basel, Switzerland).

Recombinant expression and enzyme activity of TbPSS2 in E. coli
The TbPSS2 ORF was obtained by PCR from T. brucei gDNA, using 5 0 -CGCGGATCCGATGGCCGGTAAGCTCAA C-3 0 and 5 0 -CCGCTCGAGTCGCCAAAAGATTATGTA-3 0 primers. The PCR product was sub-cloned into pET32b expression vector, expression was induced in E. coli (BL-21 strain) at 158C overnight with 0.5 mM isopropyl b-D21-thiogalactopyranoside. E. coli membranes were isolated and PSS activity assays conducted, using an empty vector expressing E. coli as a control. PSS activity was assayed as follows. 100 mM potassium phosphate (pH 7, 5), 20 mM MgCl 2 , 1% (wt/vol) Triton X-100, 50 mg membrane protein and 0.05 mCi [ 3 H]serine (ARC, specific activity 30.5 Ci mmol 21 ) with or without 300 lM substrate (PE, PC, PS, PI, DAG, CDP-DAG; all dipalmitoyl) in a final volume of 200 ll. After sonication reactions were incubated at 308C for 1 h, and terminated by the addition of 750 ll methanol: chloroform (2:1, by vol). Lipids were extracted in the organic lower layer after making the extract biphasic by the addition of 250 ll of water and chloroform , allowing the incorporation of the radiolabel to be quantitated by liquid scintillation counting.
The truncated soluble form of TbPSD was expressed in the same manner as the full length enzyme but then purified using Ni 21 affinity purification. Various amounts of protein were assayed in the same way as above except reactions were performed at 308C for 60 min.

Construction of TbPSD vectors and complementation of the yeast psd1Dpsd2Ddpl1D strain
Constitutive expression vectors harboring a full length TbPSD and truncated TbPSDs (lacking the first 48 or 83 amino acids) were made using the pBEVY-U vector (Miller, 1998). Briefly, specific primers for the individual constructs were generated and used to amplify DNA from pALC14 vector harboring TbPSD using PCR with Phusion High-Fidelity DNA polymerase (New England Biolabs Inc.). The TbPSD cDNA constructs also contained 5 0 and 3 0 flanking sequences that are homologous to those of the vector. The constructs were introduced into a linearized pBEVY-U vector (by digestion with KpnI and EcoRI) through ligation reactions by In-Fusion HD Cloning Kits (Clontech Laboratories Inc.) to yield pBEVY-U-TbPSD, pBEVY-U-D48T bPSD, and pBEVY-U-D83TbPSD. To construct pBEVY-U-Chimera-D83TbPSD vector harboring a chimera of ScPSD1-TbPSD in which 139 amino acids of the yeast N-terminus PSD1 mitochondrial targeting sequence were fused to D83TbPSD, the DNA fragment of the N-terminus ScPSD1 sequence was amplified by PCR and introduced into the linearized pBEVY-U-D83TbPSD (by KpnI digestion) through the ligation reaction by the In-Fusion HD Cloning kits. All TbPSD sequences in the newly made constructs were confirmed by DNA sequencing. For complementation, plasmids were transformed into an ethanolamine auxotrophic strain, HKY44 (MATalpha psd1-D1::TRP1 psd2-D1::HIS3 dpl1D::KanMX trp1 ura3 his3 lys2 leu2) (Storey et al., 2001) by using Yeastmaker yeast transformation kit (Clontech Laboratories Inc.). The transformants were selectively grown on uracile dropout synthetic glucose (SC-U) plates supplemented with 2 mM ethanolamine. The transformants were replica plated into the SC-U plates supplemented without ethanolamine and the plates with ethanolamine.
A galactose inducible pYES2.1-V5 vector was used to express a full length TbPSD, a V5-tagged full length TbPSD and a V5-tagged chimera of ScPSD1-TbPSD in yeast psd1Dpsd2Ddpl1D strain. Specific primers were generated so that proteins could be expressed as untagged or C-terminal V5 fusion forms. The PCR constructs were amplified from vector templates pBEVY-U-TbPSD and pBEVY-U-Chimera-D83TbPSD. The constructs were ligated into a linearized pYES2.1-V5 TOPO vector (Invitrogen) and transformed into ethanolamine auxotrophic PTY44 strain. The transformants were selectively grown on uracile dropout synthetic galactose (SG-U) plates supplemented with 2 mM ethanolamine. The transformants were replica plated into the SG-U plates supplemented without ethanolamine and the plates with ethanolamine.

TbPSD activity assay in trypanosomes
About 100 million trypanosomes were suspended in 500 ml of extraction buffer solution (50 mM Tris-HCl, pH 8, 10 mM 2-mercaptoethanol, 1 mM EDTA, 0.25 M sucrose, 0.5 mM PMSF, Sigma Protease inhibitor cocktail) and the cell free extracts were obtained by sonication (30 sec burst at 12% amplitude, 3 times with 30 second cooling intervals) followed by centrifugation at 1000g for 5 min. PSD activity of the extracts (in various amounts ranging from 50 to 437 mg) was measured using 0.2 mM phosphatidyl[1 0 -14 C]serine (400 cpm nmol 21 ) as the substrate, and the reaction product was trapped as 14 CO 2 on 2 M KOH-impregnated filter paper, as described previously (Trotter et al., 1993). No detectable enzyme activities were found in T. brucei cell extracts with this method.