Phosphatidylethanolamine Is the Precursor of the Ethanolamine Phosphoglycerol Moiety Bound to Eukaryotic Elongation Factor 1A*

In addition to its conventional role during protein synthesis, eukaryotic elongation factor 1A is involved in other cellular processes. Several regions of interaction between eukaryotic elongation factor 1A and the translational apparatus or the cytoskeleton have been identified, yet the roles of the different post-translational modifications of eukaryotic elongation factor 1A are completely unknown. One amino acid modification, which so far has only been found in eukaryotic elongation factor 1A, consists of ethanolamine-phosphoglycerol attached to two glutamate residues that are conserved between mammals and plants. We now report that ethanolamine-phosphoglycerol is also present in eukaryotic elongation factor 1A of the protozoan parasite Trypanosoma brucei, indicating that this unique protein modification is of ancient origin. In addition, using RNA-mediated gene silencing against enzymes of the Kennedy pathway, we demonstrate that phosphatidylethanolamine is a direct precursor of the ethanolamine-phosphoglycerol moiety. Down-regulation of the expression of ethanolamine kinase and ethanolamine-phosphate cytidylyltransferase results in inhibition of phosphatidylethanolamine synthesis in T. brucei procyclic forms and, concomitantly, in a block in glycosylphosphatidylinositol attachment to procyclins and ethanolamine-phosphoglycerol modification of eukaryotic elongation factor 1A.

In addition to its conventional role during protein synthesis, eukaryotic elongation factor 1A is involved in other cellular processes. Several regions of interaction between eukaryotic elongation factor 1A and the translational apparatus or the cytoskeleton have been identified, yet the roles of the different post-translational modifications of eukaryotic elongation factor 1A are completely unknown. One amino acid modification, which so far has only been found in eukaryotic elongation factor 1A, consists of ethanolamine-phosphoglycerol attached to two glutamate residues that are conserved between mammals and plants. We now report that ethanolamine-phosphoglycerol is also present in eukaryotic elongation factor 1A of the protozoan parasite Trypanosoma brucei, indicating that this unique protein modification is of ancient origin. In addition, using RNA-mediated gene silencing against enzymes of the Kennedy pathway, we demonstrate that phosphatidylethanolamine is a direct precursor of the ethanolamine-phosphoglycerol moiety. Down-regulation of the expression of ethanolamine kinase and ethanolamine-phosphate cytidylyltransferase results in inhibition of phosphatidylethanolamine synthesis in T. brucei procyclic forms and, concomitantly, in a block in glycosylphosphatidylinositol attachment to procyclins and ethanolamine-phosphoglycerol modification of eukaryotic elongation factor 1A.
Eukaryotic elongation factor 1A (eEF1A) 2 is a member of the G-protein family and represents an essential component during protein synthesis by binding aminoacyl-tRNAs in a GTP-dependent reaction to the acceptor site of ribosomes during peptide chain elongation (1)(2)(3). Crystal structures of eEF1A in complex with components of the nucleotide exchange factor eEF1B have recently been reported (4,5). Besides its role during protein synthesis, eEF1A is involved in other cellular processes. It has long been proposed that eEF1A associates with and modulates microtubules and actin filaments in several cell types (1,6). However, only recently could it be demonstrated that the conventional role of yeast eEF1A in protein synthesis and its non-canonical role in cytoskeleton organization clearly are separate functions (7). Furthermore, it has been reported that eEF1A, at least in the protozoan parasite Trypanosoma brucei, has yet another role in mediating the specificity of mitochondrial tRNA import (8).
Almost 20 years ago, two groups independently showed that eEF1A from a human erythroleukemia cell line (9) and a murine lymphocyte cell line (10) is modified by ethanolamine-phosphoglycerol (EPG), which is covalently attached to two conserved glutamate residues in the polypeptide chain (Fig. 1). Subsequently, the same modification was found in plant (11), but not yeast (12), eEF1A. The discovery of the EPG modification was prompted by the observation that a 49-kDa cytosolic protein was labeled after incubation of mammalian cells in culture with tritiated ethanolamine (Etn) (9,10), an approach that was originally aimed at identifying glycosylphosphatidylinositol (GPI)-anchored proteins. Etn is a component of the GPI core structure consisting of ethanolamine-phosphate-6-mannose-␣1,2-mannose-␣1,6-mannose-␣1,4-glucosamine-␣1,6-myoinositol-1-phospholipid (13). The pathway of Etn incorporation into GPI-anchored proteins involves uptake into cells via choline/ethanolamine transporter, incorporation into phosphatidylethanolamine (PE) via common phospholipid biosynthetic pathways, transfer of the Etn moiety from PE onto a GPI precursor lipid, and attachment of the GPI to a polypeptide precursor in the endoplasmic reticulum (14 -16).
Incorporation of tritiated ethanolamine into protein has also been observed for another lipid-modified protein, microtubule-associated protein 1 light chain 3 (17). This protein, together with its yeast homologue Atg8, is modified by PE, which renders these proteins membrane-bound (17,18). The attachment of PE to a C-terminal glycine residue in Saccharomyces cerevisiae Atg8 occurs in a ubiquitin-like conjugation reaction (18). Recently, PE modification of Atg8 was shown to be linked to Atg8 function in membrane tethering and hemifusion during autophagosome formation (19).
At present, no information is available on the biosynthesis of EPG and its attachment to eEF1A, nor on its role in eEF1A function. Possible pathways for EPG synthesis include (i) the sequential addition of individual EPG components to the glu-tamate side chains of eEF1A, (ii) the pre-assembly of EPG, followed by its transfer to eEF1A, or (iii) the attachment of a larger structure containing the EPG moiety to eEF1A, followed by modification reactions on the protein. We favor this last possibility and hypothesize that eEF1A is modified with the phospholipid PE, followed by removal of the hydrophobic acyl or alkyl chains by appropriate enzymes. As a model system to address these questions, we chose T. brucei procyclic culture forms, because they readily take up tritiated Etn as a potential label for the EPG modification and because mutant cell lines can easily be generated to express tagged proteins, or downregulate the expression of endogenous proteins using RNA interference (RNAi). In addition, as in most eukaryotic cells, the postulated substrate of the EPG modification, PE, represents a major phospholipid class in T. brucei (20,21).
In mammalian cells and yeast, PE biosynthesis has been studied in great detail (22,23). PE can be synthesized from Etn by phosphorylation to ethanolamine phosphate (Etn-P), which is subsequently activated to CDP-ethanolamine (CDP-Etn), and transferred onto diradylglycerol to form PE. This reaction sequence involving a CDP-activated intermediate has originally been delineated by Kennedy and Weiss (24) and is commonly referred to as the Kennedy pathway. A similar reaction sequence involving a CDP-activated intermediate is also responsible for the synthesis of phosphatidylcholine (PC) (24). The individual reactions for PE and PC synthesis by the Kennedy pathway have been localized to the cytosol (25). Alternatively, PE can be generated by decarboxylation of phosphatidylserine (PS) or by head group exchange with PS. PS decarboxylation represents the major route for PE synthesis in mammalian cells and occurs in the Golgi and mitochondria (22,26,27). In T. brucei, the pathways for the synthesis of PE, or other phospholipid classes, have not been studied in detail (28,29). Based on our hypothesis that PE may be the precursor of the EPG modification of eEF1A, we studied if blocking PE biosynthesis in T. brucei procyclic (insect) forms using RNAi against enzymes of the Kennedy pathway affects attachment of EPG to eEF1A.

EXPERIMENTAL PROCEDURES
Unless otherwise specified, all reagents were of analytical grade and were from Merck (Darmstadt, Germany), Sigma-Aldrich (Buchs, Switzerland) or MP Biomedicals (Tägerig, Switzerland Trypanosomes and Culture Conditions-The T. brucei EP/GPEET procyclin null mutant ⌬procyclin#1 (30), and the derived procyclic cell line expressing tagged eEF1A, were cultured at 27°C in DTM supplemented with 15% heat-inactivated fetal bovine serum (Invitrogen). T. brucei 29-13 procyclic forms (31) (obtained from Paul Englund, John Hopkins University School of Medicine) were cultured at 27°C in SDM-79 containing 15% heat-inactivated fetal bovine serum, 25 g/ml hygromycin, and 15 g/ml G418 to maintain constitutive expression of the T7 RNA polymerase and the tetracycline repressor. Derived RNAi strains, including the RNAi strain against T. brucei eEF1A (8) (a kind gift of André Schneider, University of Fribourg), were cultured in the presence of an additional 2 g/ml puromycin. The expression of double-stranded RNA was induced by the addition of 1 g/ml tetracycline. T. brucei 427 bloodstream forms were cultured at 37°C and 5% CO 2 in HMI-9 containing 10% heat-inactivated fetal bovine serum. Differentiation of bloodstream forms to procyclic forms was induced by the addition of 6 mM cis-aconitate and transferring the cells to 27°C in SDM-79 containing 15% heat-inactivated fetal bovine serum (32).
Protein synthesis was blocked in ⌬procyclin#1 parasites by the addition of cycloheximide (50 g/ml, final concentration) to the culture medium 30 min before the start of the experiment. Protein synthesis was measured by incorporation of [ 3 H]serine into protein.
Construction of Epitope-tagged T. brucei eEF1A-To express a hemagglutinin-tagged variant of eEF1A (HA-eEF1A), the annotated T. brucei eEF1A (TEF1) gene (GeneDB accession number Tb10.70.5670) was amplified by PCR with flanking HindIII and BamHI restriction sites (primers TbEF1A, supplemental Table S1), cloned into the TOP10FЈ vector with a TOPO TA Cloning Kit (Invitrogen), and subsequently subcloned between HindIII and BamHI restriction sites into the pCorleone vector (a kind gift of Isabel Roditi, University of Bern) (33). The vector was linearized with HindIII, and an oligonucleotide coding for the HA tag and flanked by two HindIII sites (primers HA, supplemental Table S1) was cloned into the same vector at the N terminus of the T. brucei eEF1A gene. Before transfection into ⌬procyclin#1, the vector (pAShaEF) was linearized with NotI and SalI. RNAi-mediated Gene Silencing-Putative T. brucei ethanolamine kinase (GeneDB accession number Tb11. 18.0017) and putative T. brucei ethanolamine-phosphate cytidylyltransferase (Tb11.01.5730) were down-regulated by RNAi-mediated gene silencing using stem loop constructs containing a puromycin resistance gene. Cloning the gene fragments into the vector pALC14 (a kind gift of André Schneider, University of Fribourg) was performed as described previously (34), using PCR products obtained with primers Tb0017 (spanning nucleotides 261-810 of Tb11.18.0017) and Tb5730 (spanning nucleotides 51-585 of Tb11.01.5730) (supplemental Table S1), resulting in plasmids pAS0017 and pAS5730, respectively. For transfection of T. brucei procyclic forms, the vectors were linearized with NotI.
Stable Transfection of Trypanosomes-T. brucei procyclic forms were transfected with pAS0017, pAS5730, or pAShaEF and selected for antibiotic resistance by addition of 10 g/ml blasticidin S HCl (Invitrogen) for ⌬procyclin#1 cells or 2 g/ml puromycin for 29-13 RNAi cells to the culture medium. Clones were obtained by limiting dilution.
RNA Isolation and Northern Blot Analysis-Total RNA for Northern blotting was prepared by the standard acidic guanidium isothiocyanate method (35). Total RNA (10 g) was separated on formaldehyde agarose gels and transferred to GeneScreen Plus nylon membranes (PerkinElmer Life Sciences). The 32 P-labeled probes were made by random priming of the same PCR products used as inserts in the stem-loop vector (Prime-a-Gene Labeling System, Promega, Madison, WI). Hybridization was performed overnight at 60°C, and the membrane was analyzed by autoradiography using BioMax MS film and a TransScreen-HE intensifying screen (Kodak). Ribosomal RNA was visualized on the same gel by ethidium bromide staining to control for equal loading.
TLC-One-dimensional TLC was performed on Silica Gel 60 plates. To separate phospholipids, CM extracts were run in solvent system 1, composed of chloroform:methanol:acetic acid: water (25:15: 4:2, by volume). For separation of GPI precursors, CMWbut phases were run in solvent system 2, composed of chloroform:methanol:water (4:4:1, by volume). Ethanolamine metabolites were separated in solvent system 3, composed of 25% ammonium hydroxide:methanol:0.6% NaCl in water (1:10: 10, by volume). Radioactivity was detected by scanning the airdried plate with a radioisotope detector (Berthold Technologies, Regensdorf, Switzerland) and quantified using the Rita Star software provided by the manufacturer. Alternatively, the plate was sprayed with En 3 hance (PerkinElmer Life Sciences) and exposed to MXB film at Ϫ70°C. On all TLC plates, appropriate lipid standards were run alongside the samples.
Lipid Phosphorous Determination-Phospholipid fractions were scraped from TLC plates and digested by boiling in perchloric acid, and the released inorganic phosphate was reacted with ammonium molybdate and quantified photometrically (40). Each determination was accompanied by a series of inorganic phosphate standards. The assay was linear between 0 and 200 nmol of phosphate per tube.
SDS-PAGE and Immunoblotting-Extracted proteins were separated on 12% polyacrylamide gels under reducing conditions (41). For detection of 3 H-labeled proteins, gels were fixed, stained with Coomassie Brilliant Blue, soaked in Amplify (GE Healthcare), dried, and exposed to MBX films at Ϫ70°C. For immunoblotting, proteins were transferred onto Immobilon P polyvinylidene difluoride membranes (Millipore, Bedford, MA) by semi-dry blotting. Mouse monoclonal antibody against eEF1A (␣-EF, Upstate, Lake Placid, NY) was used at a dilution of 1:5000. Mouse monoclonal antibody against HA (␣-HA, Covance, Berkeley, CA) was used at a dilution of 1:3000. Mouse monoclonal antibody against EP procyclin (␣-EP), generously provided by Terry W. Pearson (University of Victoria), was used at a dilution of 1:2500. Primary antibodies were detected with secondary rabbit anti-mouse IgG conjugated to horseradish peroxidase (Dako, Baar, Switzerland) at a dilution of 1:5000 and using an enhanced chemiluminescence detection kit (Pierce).
Immunoprecipitation-HA-eEF1A from trypanosomes lyzed in 0.1% Nonidet P-40 was immunoprecipitated with Anti-HA Affinity Matrix (Roche Applied Science) according to the manufacturer's instructions and boiled in sample loading buffer for SDS-PAGE.
Mass Spectrometry-Immunoprecipitated HA-eEF1A was subjected to SDS-PAGE, followed by in-gel reductive alkylation and trypsin digestion. Identification of peptide masses was done at the FingerPrints Proteomics Facility, Wellcome Trust Biocenter, University of Dundee (Dundee, Scotland), using an ABI 4700 matrix-assisted laser desorption ionization-time-offlight/time-of-flight (MALDI-Tof/Tof) mass spectrometer. The tryptic peptides were further analyzed by liquid chromatography tandem mass spectrometry (nano-LC-MS/MS) using a Dionex Ultimate LC, equipped with a Pepmap C18 column (75 m ϫ 15 cm), coupled to an ABI Q-Trap 4000 mass spectrometer. The peptide-resolving part of the nano-LC gradient was from 1% to 40% acetonitrile in 0.1% formic acid over 20 min at 300 nl/min.
Chemical Treatment of eEF1A-Immunoprecipitated [ 3 H]Etn-labeled HA-eEF1A was incubated in 50 l of 70% formic acid at 37°C for 4 or 22 h. The reaction was stopped by diluting the sample with 150 l of water. After drying in a SpeedVac, peptides were resuspended in sample loading buffer and analyzed by SDS-PAGE, followed by immunoblotting or fluorography. Alternatively, SpeedVac-dried SDS fractions of [ 3 H]Etn-labeled ⌬procyclin#1 parasites were incubated with 50 l of 70% formic acid at 37°C for 4 or 22 h, diluted with 150 l of water, and analyzed as above.

T. brucei eEF1A Is Labeled with [ 3 H]
Etn-Alignment of the deduced protein sequence of T. brucei eEF1A with carrot and mouse eEF1A shows 76% identity with both sequences (Fig. 2). In particular, the two glutamate residues shown to be modified with EPG in plant, Glu 289 and Glu 362 (11), and mammalian eEF1A, Glu 301 and Glu 374 (9,10), are conserved in T. brucei eEF1A (Glu 289 and Glu 362 ). The original identification of EPG bound to eEFA1 in mammalian cells was prompted by the observation that eEF1A was labeled with [ 3 H]Etn (9, 10). We used the same approach in T. brucei and found that incubation of procyclic forms lacking the genes for procyclins (⌬procy-clin#1) with [ 3 H]Etn resulted in labeling of a single strong band at 49 kDa (Fig. 3A, left panel). The reason to use this particular mutant cell line was to avoid possible misidentification of labeled bands with the multiple forms of procyclins, which readily incorporate large amounts of [ 3 H]Etn into their GPI anchors (36,42). In addition, we found that a band of similar molecular mass is also recognized by a monoclonal antibody against eEF1A (Fig. 3A, right panel). The band at 37 kDa likely represents a degradation product of eEF1A. Further evidence that the 49-kDa labeled protein is eEF1A was obtained from experiments using a T. brucei 29-13 procyclic cell line expressing an RNAi construct designed to knock down the expression of eEF1A (8). The results show that, after induction of RNAi by the addition of tetracycline to the culture medium, the 49-kDa band is no longer labeled with [ 3 H]Etn; in contrast, incorporation of label into GPI-anchored EP procyclin, which migrates as a broad band at ϳ42 kDa, continued (Fig. 3B, left panel). Immunoblotting revealed that induction of RNAi against eEF1A resulted in complete disappearance of the 49-and 37-kDa bands that are recognized by eEF1A antibody in control cells (Fig. 3B, right panel). In addition, we found that when protein synthesis, as measured by incorporation of [ 3 H]serine into total protein, is inhibited by the addition of cycloheximide to the culture medium (Fig. 3C, left panel), incorporation of label into eEF1A is blocked, indicating that the attachment of Etn to eEF1A occurs during, or shortly after, protein synthesis (Fig.  3C). Furthermore, the 49-kDa [ 3 H]Etn-labeled band was also seen in T. brucei 427 bloodstream forms and during differentiation of bloodstream to procyclic forms in culture (Fig. 3D). The additional 3 H-labeled band at 55 kDa in T. brucei 427 bloodstream forms represents the variant surface glycoprotein, which is replaced during differentiation to procyclic forms by the EP/GPEET procyclins (43). Together, these results strongly indicate that the 49-kDa [ 3 H]Etn-labeled band in T. brucei procyclic and bloodstream forms represents eEF1A.
T. brucei eEF1A Is Modified with EPG at Glu 362 -To study the chemical nature of the Etn bound to T. brucei eEF1A, a HA-tagged form of eEF1A was transfected into T. brucei ⌬pro-cyclin#1 and isolated by immunoprecipitation using anti-HAcoated beads. The results show that isolated HA-tagged eEF1A is also labeled with [ 3 H]Etn and is recognized by antibodies against eEF1A and HA (Fig. 4A). Analysis of immunoprecipitated HA-eEF1A by SDS-PAGE and Coomassie staining revealed two bands migrating very close to each other at 50 kDa (Fig. 4B). The bands were cut out individually and prepared for MS analysis by in-gel reduction and alkylation and digestion with trypsin. The resulting tryptic peptides were analyzed by MALDI-Tof/Tof MS. Both bands were thus identified as being T. brucei eEF1A, with sequence coverage of 50 and 42%, respectively. Both samples also exhibited ions at m/z 1020.465, consistent with being the [MϩH] ϩ ions of the tryptic peptide FAEIESK (823.420 Da, amino acids 360 -366 of the eEF1A sequence, see Fig. 2) plus a predicted mass increase of 197.045 Da resulting from the attachment of EPG to a glutamate residue. In contrast, the second tryptic peptide predicted to possibly contain an EPG modification at Glu 289 , SIEMHHEQLAEATPGDNVGFN-VK (amino acids 279 -301 of the eEF1A sequence, see Fig. 2), was only found in unmodified form, i.e. with [MϩH] ϩ 2523.193 Da. At present, we don't know why HA-eEF1A runs as a doublet on SDS-PAGE.
To confirm that peptide FAEIESK was modified and to localize the modification, the same tryptic peptide samples were analyzed by nano-LC-MS/MS. In both samples, a peptide eluting at 13.3 min produced an [Mϩ2H] 2ϩ ion at m/z 510.8 that gave a product ion spectrum that clearly localized an EPGmodified glutamic acid residue (E*) in the sequence FAE*IESK. The spectrum is consistent with a facile elimination of glycerol phosphate, leaving an aminoethene modified Glu 362 residue, as indicated in Fig. 5.
To further demonstrate that Glu 362 is the only site modified with EPG, [ 3 H]Etn-labeled eEF1A was treated with mild acid. Formic acid has been shown to selectively and efficiently cleave Asp-Pro bonds in proteins (44); in addition, it also cleaves Asp-Gly bonds, albeit at much lower efficiency (45). Because T. brucei eEF1A contains a single Asp-Pro bond located between the two potential EPG modification sites and only two Asp-Gly bonds, formic acid treatment was used to cleave [ 3 H]Etn-labeled eEF1A into distinct peptides (Fig. 6A). Analysis by SDS-PAGE, followed by fluorography and immunoblotting, shows that treatment of endogenous or HA-tagged eEF1A with mild acid for 22 h generated a single labeled polypeptide of ϳ12 kDa (Fig. 6, B and D). Immunoblotting using ␣-EF shows that the 49-kDa band of eEF1A completely disappears during formic acid treatment and the band at ϳ37 kDa accumulates (Fig. 6C). Similarly, immunoblotting using ␣-HA shows the disappearance of the 50-kDa HA-eEF1A doublet and a transient accumulation of the 37-kDa band; after prolonged treatment with formic acid, the antibody detects a product of ϳ14 kDa (Fig. 6E). These results, in particular the absence of 3 H-labeled fragments of 35 or 21 kDa, are consistent with the predicted cleavage pattern of eEF1A by formic acid (Fig. 6A) and demonstrate that T. brucei eEF1A contains a single EPG-modified site at Glu 362 that is labeled with [ 3 H]Etn. Down-regulation of PE Biosynthesis-To test our working hypothesis that PE is the precursor of the EPG modification of eEF1A, we studied the effects of down-regulating possible PE biosynthetic pathways by RNAi on EPG attachment in T. brucei procyclic forms. Searching the T. brucei genome for homologues encoding enzymes involved in PE synthesis revealed candidate sequences for all three enzymes of the Kennedy pathway and, in addition, candidate sequences for PE/PS head group exchange and PS decarboxylation (supplemental Fig. S1). Subsequently, we targeted the first two putative enzymes of the Kennedy pathway, ethanolamine kinase (Tb11.18.0017) and ethanolamine-phosphate cytidylyltransferase (Tb11.01.5730), for RNAi using tetracycline-inducible expression of doublestranded RNA. The nucleotide sequences of the constructs used showed no significant homology to any other sequences in the trypanosome genome. Transfection of T. brucei procyclic forms with these constructs and selection by resistance to puromycin resulted in several mutant clones, two each of which were selected for all subsequent experiments.
Down-regulation of ethanolamine kinase or ethanolaminephosphate cytidylyltransferase by RNAi resulted in a complete disappearance of the respective RNA (Fig. 7, top). After 3 days of culture in the presence of tetracycline, growth of the individual clones slowed down and, in the case of ethanolamine-phosphate cytidylyltransferase, ceased after 7 days of induction (Fig. 7, bottom). To study the effects of RNAi on PE biosynthesis, mutant clones were incubated in the presence of [ 3 H]Etn and the incorporation of radioactivity into PE, or other lipids, was determined by TLC analysis. The results show that essentially all radioactivity in the lipid fraction co-migrated with a PE standard (Fig. 8A, left panels). After down-regulation of ethanolamine kinase or ethanolaminephosphate cytidylyltransferase by RNAi for 3 days, the amount of labeled PE decreased to Ͻ30% of control levels (Fig. 8A, right panels). A block in PE synthesis at ethanolamine-phosphate cytidylyltransferase is expected to result in accumulation of the immediate biosynthetic precursor of the reaction, Etn-P. Thus, we analyzed the 3 H-labeled acid-soluble metabolites and found that control trypanosomes show small amounts of radioactivity co-migrating with Etn-P and CDP-Etn standards, whereas RNAi cells after downregulation of ethanolamine-phosphate cytidylyltransferase show an accumulation of [ 3 H]Etn-P, and a lack of CDP-[ 3 H]Etn (Fig. 8B).  Although a decrease in [ 3 H]Etn incorporation into PE is consistent with a decreased rate of PE synthesis by the Kennedy pathway, it may not reflect a decrease in cellular PE content, because, at least in mammalian cells and yeast, substantial amounts of PE are generated by decarboxylation of PS (22,23). Thus, we determined the total PE content in T. brucei procyclic forms after down-regulation of ethanolamine kinase or ethanolamine-phosphate cytidylyltransferase by RNAi. The results show that knocking down the two enzymes has a dramatic effect on cellular PE levels. After 5 days of incubation in the presence of tetracycline, ethanolamine kinase knockdowns showed a 38% reduction in total PE, whereas ethanolaminephosphate cytidylyltransferase knockdowns had Ͻ20% of PE compared with cells incubated in the absence of tetracycline (Fig. 8C). In contrast, the PC content was unaffected during down-regulation of PE biosynthesis (results not shown), indicating that the two enzymes that were targeted by RNAi are only involved in PE synthesis and show no significant overlapping substrate specificity with the corresponding enzymes of the Kennedy pathway for PC biosynthesis. Furthermore, these results demonstrate that PC in T. brucei is not synthesized by methylation of PE, which is in agreement with the apparent lack of genes for the corresponding methyltransferases in the T. bru-cei genome. Furthermore, the dramatic reduction of cellular PE after down-regulation of ethanolamine-phosphate cytidylyltransferase demonstrates that the Kennedy pathway provides the majority of PE in T. brucei procyclic forms.
Down-regulation of PE Synthesis Blocks EPG Attachment to eEF1A-Because PE has been shown to be the donor of the Etn residue for GPI anchor assembly in trypanosomes (46), incorporation of [ 3 H]Etn into GPI-anchored EP procyclin was expected to be severely affected after down-regulation of PE synthesis and, thus, was used as positive control for a limited availability of PE for protein modification. The results show that, during labeling of procyclic T. brucei RNAi cells with [ 3 H]Etn in the absence of tetracycline, significant amounts of radioactivity were incorporated into eEF1A and EP procyclin (Fig. 9A). In contrast, after RNAi-mediated down-regulation of PE synthesis, labeling of eEF1A and EP procyclin was partly (RNAi against ethanolamine kinase) or completely (RNAi against ethanolamine-phosphate cytidylyltransferase) inhibited (Fig. 9A). Immunoblotting of the same extracts shows that RNAi against the two enzymes did not reduce the amounts of eEF1A (Fig. 9B). In contrast, protein levels of EP were decreased in the induced cells compared with control uninduced trypanosomes (Fig. 9C). In addition, two bands with lower molecular masses were detected in cells after down-regulation of ethanolamine-phosphate cytidylyltransferase (Fig. 9C); these likely represent precursor forms of EP procyclin.  To further demonstrate that the reduction of PE synthesis had a major effect on GPI anchor biosynthesis and attachment to protein, T. brucei procyclic forms after RNAi were labeled with [ 3 H]Etn or [ 3 H]myristate, and the GPI precursors and free GPI anchors were analyzed by TLC and SDS-PAGE followed by fluorography. The results show that the major GPI precursor lipid, named PP1 (42), was labeled with [ 3 H]Etn in control cells but not in cells after induction of RNAi against ethanolaminephosphate cytidylyltransferase (supplemental Fig. S2A). In contrast, when trypanosomes were labeled with [ 3 H]myristate, the amount of 3 H-labeled GPI precursor lipids increased considerably (supplemental Fig. S2B). Similarly, induced cells showed a depletion of [ 3 H]Etn-labeled and an accumulation of [ 3 H]myristate-labeled free GPI anchors (supplemental Fig. S2C). Together, these results are consistent with a depletion of Etncontaining GPI lipids and free GPIs and a concomitant accumulation of GPI lipids and free GPIs lacking the terminal Etn group in cells after RNAi against ethanolamine-phosphate cytidylyltransferase.

DISCUSSION
Our results demonstrate that eEF1A in the parasitic protozoa, T. brucei, is modified with EPG. Labeling experiments, in Etn-P, ethanolamine-phosphate; CDP-Etn, CDP-ethanolamine. C, to determine the total PE content of RNAi clones B1 and B5 cultured for 5 days in the absence (white bars) or presence (black bars) of tetracycline, phospholipids from 4 ϫ 10 8 trypanosomes were extracted with CM and separated by TLC using solvent system 1. After visualization of lipids by iodine vapor, spots co-migrating with a PE standard run on the same plate were scraped and quantified by lipid phosphorous analysis. The values represent means Ϯ S.E. from three independent experiments. combination with chemical and mass spectrometric analyses, show that, in analogy to mammalian cells (9 -11) and plants (11), endogenous and HA-tagged eEF1A in T. brucei procyclic and bloodstream forms can be labeled with [ 3 H]Etn at an EPG moiety attached to the side chain of a glutamate residue. This amino acid, Glu 362 in the T. brucei eEF1A primary sequence, is strictly conserved between the predicted amino acid sequences of mammalian, plant, and yeast cells. In contrast, in the yeast S. cerevisiae, the corresponding Glu 372 has been reported not to be modified with EPG (12). The reason of this lack of EPG addition to eEF1A in S. cerevisiae is unknown. Mammalian cells (9, 10) and plants (11) contain a second EPG modification attached to Glu 301 or Glu 289 , respectively, of eEF1A (see also Fig. 2). In contrast, we found that the corresponding Glu 289 of T. brucei eEF1A is not labeled with [ 3 H]Etn. In addition, an EPG-modified tryptic peptide covering the second potential EPG site could not be detected by mass spectrometry. Therefore, it is likely that Glu 289 of T. brucei eEF1A is not modified with EPG, which distinguishes it from all previously identified EPG-modified eEF1As. We were unable to identify an amino acid sequence motif that predicts EPG attachment to one or both glutamate residues in eEF1A of the different organisms. Point mutations at and around the potential modification sites are currently underway to define the requirements for EPG attachment to eEF1A.
The pathways for the assembly of the EPG modification and its attachment to eEF1A have not been studied in detail before. We now show using RNAi against the first two enzymes of the Kennedy pathway that PE is donor of the Etn moiety for EPG in T. brucei eEF1A. We found that down-regulation of ethanolamine kinase or ethanolamine-phosphate cytidylyltransferase causes a severe reduction in total cellular PE content, demonstrating that the Kennedy pathway is the major route for PE synthesis in T. brucei procyclic culture forms. Importantly, inhibition of this pathway results in a partial (RNAi against ethanolamine kinase), or complete (RNAi against ethanolamine-phosphate cytidylyltransferase) block in incorporation of [ 3 H]Etn into eEF1A, demonstrating that PE is a direct precursor of the EPG modification. This conclusion is supported by our findings that Etn or Etn-P cannot serve as precursors, because a block in PE synthesis by RNAi against ethanolamine-phosphate cytidylyltransferase completely prevents [ 3 H]incorporation into eEF1A. These results are in contradiction to a previous report showing that Etn is directly added to eEF1A in a cell-free system (47). In addition, experiments targeting the third step in PE synthesis by the Kennedy pathway showed that RNAi against a putative ethanolamine-phosphotransferase also led to a reduction in [ 3 H]incorporation into eEF1A, indicating that CDP-Etn is not a direct precursor for EPG (results not shown).
In addition, we show that a block in PE synthesis by the Kennedy pathway also results in a reduction, or lack, of [ 3 H]Etn incorporation into GPI-anchored EP procyclin. This result is consistent with a previous report showing that PE is the donor of the Etn moiety linking the GPI anchor to protein in procyclic form trypanosomes and yeast (46,48), and, thus, inhibition of PE synthesis is expected to inhibit GPI anchor attachment to procyclins. To our knowledge, this is the first time that GPI anchoring of proteins has been blocked by inhibition of an enzyme that is not directly involved in the GPI biosynthetic pathway.
Recent advances in the structure determination of eEF1A have clarified its role during protein translation elongation (5). In contrast, although eEF1A contains a number of (post-translational) modifications, their roles in eEF1A function are largely unknown. Amino acid modifications of eEF1A include phosphorylation, lysine methylation, and methyl esterification at the C terminus. Similarly, nothing is known about the role of EPG in eEF1A function. The available three-dimensional structures of yeast eEF1A (4,5) and the high homology between eEF1As from different organisms predict the EPG-modified glutamate residues in mammalian cells, plants, and T. brucei to be located at the surface of the third polypeptide domain of eEF1A. This allows us to propose the following mechanism for EPG attachment to eEF1A. During or shortly after protein synthesis, eEF1A is modified with PE at one or more of the conserved glutamate residues in the third domain. This modification occurs at, or near, the site of PE synthesis and temporarily renders eEF1A membrane-bound. Interestingly, plant eEF1A has been reported to incorporate 14 C-labeled fatty acids, however, no direct evidence was provided that eEF1A was modified with an entire PE molecule (11). Similar experiments in T. brucei procyclic forms to label eEF1A with [ 3 H]myristate or [ 3 H]palmitate, or by incubating the cells with [ 3 H]pyruvate, which is incorporated into the fatty acid pool in T. brucei procyclic forms (49), showed no incorporation of radioactivity into eEF1A (results not shown). However, this is not surprising, because we were unable to detect a PE-modified form of eEF1A by mass spectrometry, and, thus, this form may represent a minor fraction of total eEF1A only. After addition of PE to eEF1A, the acyl and/or alkyl (or alkenyl) chains of PE will be removed by the concerted action of esterases of the phospholipase A-type and/or glyceryl ether-cleaving enzymes, respectively, releasing eEF1A to the cytosol. Such a pathway offers interesting implications for eEF1A function, because a small fraction of eEF1A may associate with membranes and could be involved in processes other than its canonical role in protein synthesis at the ribosome, such as tRNA import into mitochondria (8), actin-binding (7), association with sphingosine kinases (50) or other signaling molecules (51), or binding to peroxisomes (52).