Identification of the glycosylphosphatidylinositol-specific phospholipase A2 (GPI-PLA2) that mediates GPI fatty acid remodeling in Trypanosoma brucei

The biosynthesis of glycosylphosphatidylinositol (GPI)-anchored proteins (GPI-APs) in the parasitic protozoan Trypanosoma brucei involves fatty acid remodeling of the GPI precursor molecules before they are transferred to protein in the endoplasmic reticulum. The genes encoding the requisite phospholipase A2 and A1 activities for this remodeling have thus far been elusive. Here, we identify a gene, Tb927.7.6110, that encodes a protein that is both necessary and sufficient for GPI-phospholipase A2 (GPI-PLA2) activity in the procyclic form of the parasite. The predicted protein product belongs to the alkaline ceramidase, PAQR receptor, Per1, SID-1, and TMEM8 (CREST) superfamily of transmembrane hydrolase proteins and shows sequence similarity to Post-GPI-Attachment to Protein 6 (PGAP6), a GPI-PLA2 that acts after transfer of GPI precursors to protein in mammalian cells. We show the trypanosome Tb927.7.6110 GPI-PLA2 gene resides in a locus with two closely related genes Tb927.7.6150 and Tb927.7.6170, one of which (Tb927.7.6150) most likely encodes a catalytically inactive protein. The absence of GPI-PLA2 in the null mutant procyclic cells not only affected fatty acid remodeling but also reduced GPI anchor sidechain size on mature GPI-anchored procyclin glycoproteins. This reduction in GPI anchor sidechain size was reversed upon the re-addition of Tb927.7.6110 and of Tb927.7.6170, despite the latter not encoding GPI precursor GPI-PLA2 activity. Taken together, we conclude that Tb927.7.6110 encodes the GPI-PLA2 of GPI precursor fatty acid remodeling and that more work is required to assess the roles and essentiality of Tb927.7.6170 and the presumably enzymatically inactive Tb927.7.6150.

The biosynthesis of glycosylphosphatidylinositol (GPI)anchored proteins (GPI-APs) in the parasitic protozoan Trypanosoma brucei involves fatty acid remodeling of the GPI precursor molecules before they are transferred to protein in the endoplasmic reticulum. The genes encoding the requisite phospholipase A2 and A1 activities for this remodeling have thus far been elusive. Here, we identify a gene, Tb927.7.6110, that encodes a protein that is both necessary and sufficient for GPI-phospholipase A2 (GPI-PLA2) activity in the procyclic form of the parasite. The predicted protein product belongs to the alkaline ceramidase, PAQR receptor, Per1, SID-1, and TMEM8 (CREST) superfamily of transmembrane hydrolase proteins and shows sequence similarity to Post-GPI-Attachment to Protein 6 (PGAP6), a GPI-PLA2 that acts after transfer of GPI precursors to protein in mammalian cells. We show the trypanosome Tb927.7.6110 GPI-PLA2 gene resides in a locus with two closely related genes Tb927.7.6150 and Tb927.7.6170, one of which (Tb927.7.6150) most likely encodes a catalytically inactive protein. The absence of GPI-PLA2 in the null mutant procyclic cells not only affected fatty acid remodeling but also reduced GPI anchor sidechain size on mature GPI-anchored procyclin glycoproteins. This reduction in GPI anchor sidechain size was reversed upon the re-addition of Tb927.7.6110 and of Tb927.7.6170, despite the latter not encoding GPI precursor GPI-PLA2 activity. Taken together, we conclude that Tb927.7.6110 encodes the GPI-PLA2 of GPI precursor fatty acid remodeling and that more work is required to assess the roles and essentiality of Tb927.7.6170 and the presumably enzymatically inactive Tb927.7.6150.
Glycosylphosphatidylinositol (GPI)-anchored proteins (GPI-APs) are almost ubiquitous in eukaryotes (1). The first complete GPI structure was determined for the variant surface glycoprotein (VSG) of the bloodstream form (BSF) of Trypanosoma brucei, the causative agent of human and animal African trypanosomiasis (2). Subsequent structures for rat Thy-1 antigen (3), human erythrocyte acetylcholinesterase (4), yeast Gas1p (5) and a myriad of other examples, reviewed in (1), have established the conserved and species/tissue-specific features of GPI anchors.
All GPI membrane anchors are based on a common underlying structure of ethanolamine-P-6Manα1-2Manα1-6Manα1-4GlcNα1-6PI (EtN-P-Man 3 GlcN-PI), where the amino group of the ethanolamine residue is in amide linkage to the α-carboxyl group of the C-terminal amino acid. This common core can then be decorated with carbohydrate sidechains and additional ethanolamine phosphate groups in a species-and tissue-specific manner (1). The phosphatidylinositol (PI) component can be diacyl-PI, a lyso-acyl-PI, or an alkylacyl-PI (with or without an additional ester-linked fatty acid on the 2-position of the D-myo-inositol residue) and in some cases, like in yeast and some protozoa, an inositolphospho-ceramide (1).
The biosynthesis of GPI anchors is quite well understood in T. brucei and other protozoan parasites as well as in yeast and mammalian cells (1). As first noted in (6), the PI species of GPI anchors are different in lipid composition from the bulk PIs of their resident cells. The pathways in T. brucei and mammalian cells are summarized in (Fig. 1). Some of the key differences between GPI biosynthesis in T. brucei and mammalian cells lie in their inositol-acylation and lipid remodeling reactions. The latter is responsible for the aforementioned atypical PI compositions of GPI anchors.
With respect to inositol-acylation, in mammalian cells and yeast the 2-position of the inositol ring is acylated with predominantly palmitate (C16:0) in an acyltransferase reaction catalyzed by PIG-W (GWT1 in yeast) using acyl-CoA as donor substrate and GlcN-PI as acceptor substrate to yield GlcN-(acyl)PI (7,8). In T. brucei, there is no orthologue of PIG-W/ GWT1 and the acyltransferase has not been identified. However, it is known that the donor substrate is not acyl-CoA but most likely an endogenous endoplasmic reticulum (ER) membrane lipid (9, 10) and its earliest acceptor is Man 1 GlcN-PI, producing Man 1 GlcN-(acyl)PI (11). Further, the fatty acid transferred is more heterogeneous in T. brucei (a mixture of C14:0, C16:0, C18:0, C18:1 and C18:2) (12,13). In mammalian cells and BSF T. brucei, the inositol-acyl chain can be removed ‡ These authors contributed equally to this work. * For correspondence: Michael A. J. Ferguson by an inositol-deacylase (PGAP1 or deAc2) but this occurs after (14) or before (9)(10)(11) transfer of the GPI precursor to protein, respectively. The expression of deAc2 is tightly regulated and does not occur in procyclic form (PCF) trypanosomes (15). Lipid remodeling of GPIs was first described in T. brucei (16). In BSF cells, the diacyl-PI moiety of EtN-P-Man 3 GlcN-PI (glycolipid A 0 ) is sequentially acted upon by (i) an unidentified GPI-PLA2 to produce glycolipid θ, (ii) a myristoyltransferase (TbGup1) to produce glycolipid A'' (17), (iii) an unidentified GPI-PLA1 to produce glycolipid θ 0 , and (iv) an unidentified myristoyltransferase to yield the mature GPI precursor (glycolipid A) bearing a dimyristoyl-PI lipid that is competent for transfer to protein. In PCF cells, inositol de-acylation does not occur and fatty acid remodeling only proceeds as far as the action of GPI-PLA2 (18,19).
In mammalian cells, lipid remodeling starts in the ER through the selection of sn-1-alkyl-2-acyl-(acyl)PI forms of GlcN-(acyl)PI (20). The unsaturated sn-2-acyl chain is then exchanged for a saturated (predominantly C18:0) fatty acid through the sequential action of a GPI-PLA2 (called PGAP3) (21) and an acyltransferase reaction dependent on PGAP2 (22). The resulting, generally fully saturated, alkylacyl-PI GPI anchor then associates with liquid-ordered membrane domains (known as lipid rafts) for forward transport to the plasma membrane (21).
The mammalian fatty acid remodeling GPI-PLA2 enzyme (PGAP3) belongs to the alkaline ceramidase, PAQR receptor, Per1, SID-1, and TMEM8 (CREST) superfamily of proteins (23). Another CREST superfamily member, TMEM8A/ PGAP6, was also shown to be a GPI-PLA2 acting on certain GPI-APs to facilitate their release from mammalian cells in a soluble form (24,25). The CREST superfamily includes several lipases that share a core structure of seven predicted transmembrane domains and five conserved residues (three histidine, one aspartic acid, and one serine) (23).
In this article, we identify three CREST superfamily genes in T. brucei that are in tandem array on chromosome 7. We provide direct evidence that one of these genes encodes the hitherto elusive GPI-PLA2 of T. brucei GPI fatty acid remodeling, and we speculate on the possible functions of the other two. The GPI-PLA2 of fatty acid remodelling in T.brucei

Identification of putative TbGPI-PLA2 genes
While the lipid remodeling reactions in GPI biosynthesis are different between T. brucei and mammalian cells (1), as indicated in (Fig. 1), both processes include GPI-specific phospholipase A2 reactions. We reasoned that the T. brucei GPI-PLA2(s) might have amino acid sequence similarity to PGAP3 and/or PGAP6, the two known mammalian GPI-PLA2s.

Tb927.7.6110 is necessary and sufficient for TbGPI-PLA2 activity
Since GPI biosynthesis is not essential to PCF cells (27), we made a Tb927.7.6110/6150/6170 locus null mutant by gene replacement, as indicated in (Fig. 3A). Confirmatory Southern blot data are in (Fig. S1). The resulting null mutant was viable with similar growth kinetics to wild-type cells.
Using the procedure described in (28) the procyclins purified from wild-type PCF cells and the Tb927.7.6110/6150/ 6170 −/− null mutant were analyzed by negative ion matrixassisted laser desorption ionization-time of flight (MALDI-ToF) following aqueous HF dephosphorylation and mild acid treatment. This showed that both the wild-type and null mutant cells were expressing EP-procyclins, predominantly the EP3 form, and not GPEET-procyclins (Fig. S2). We, therefore, probed Western blots of wild-type and mutant PCF cell lysates with anti-EP procyclin antibodies. This showed that deletion of the Tb927.7.6110/6150/6170 locus had a significant effect on the mature procyclin profile (Fig. 3B, compare lanes 1 and 2). In the null mutant, the procyclins adopted a more polydisperse appearance ranging from substantially lower to higher apparent MW compared to wild-type procyclins. The deletion of the Tb927.7.6110/6150/6170 locus did not, however, affect cell surface expression of the EP-procyclins as judged by immunofluorescence microscopy (Fig. S3A).
Activation of an ectopic Tb927.7.6150 gene had no effect on the procyclin SDS-PAGE pattern (Fig. 3C, lane 5), whereas activation of an ectopic Tb927.7.6110 gene reverted the procyclin SDS-PAGE pattern to that of the wild-type (Fig. 3C, lane 3). Activation of an ectopic Tb927.7.6170 gene had an unexpected effect on the procyclin SDS-PAGE pattern (Fig. 3B, lane 9); that is, the generation of higher apparent MW procyclins (discussed later).
The compositions (by accurate mass) and structures (deduced from the MS 2 product ion spectra) of the wild-type PIs were as expected from previous reports (12). Thus, the wild-type PI structures were all lyso-PI structures with a mixture of fatty acids ester-linked to the inositol ring (Figs. 4B and S4A; Table 1). In contrast, the major released PI structures from the Tb927.7.6110/6150/6170 −/− null mutant were mostly diacyl-PIs with the same range of fatty acids attached to the inositol ring (Figs. 4C and S4B; Table 1), although some minor lyso-PI species identical to the wild-type PIs were also observed ( Fig. 4C and Table 1). Taken together, these data show unambiguously that one or more genes in the Tb927.7.6110/6150/6170 locus is/are required for GPI-PLA2 activity in PCF cells.
From these data, we conclude that the Tb927.7.6110 gene alone is necessary and sufficient for TbGPI-PLA2 activity in PCF cells. Further, due to its structural similarity with mammalian PGAP6, we postulate that it directly encodes TbGPI-PLA2 activity.
Effects of Tb927.7.6110/6150/6170 ablation and Tb927.7.6110 and Tb927.7.6170 re-expression on procyclin GPI glycan structure The procyclin preparations used to analyze the nitrous acid-released PI species ( Fig. 4 and Table 1) were also subjected to neutral monosaccharide composition analysis by gas chromatography-mass spectrometry (GC-MS) and, following dephosphorylation with aqueous HF and permethylation, by MALDI-ToF and methylation linkage analysis by GC-MS.
The GPI-PLA2 of fatty acid remodelling in T.brucei In the monosaccharide composition analysis, a reduction in Gal and GlcNAc relative to Man was observed in the procyclins from the Tb927.7.6110/6150/6170 −/− null mutant compared to wild-type procyclins (Fig. 5, A and B). This reduction in Gal : Man ratio was largely recovered in the Tb927.7.6110 and Tb927.7.6170 add-back procyclins (Fig. 5, C and D). We can use the Gal : Man ratio as an indicator of how large, on average, the GPI anchor glycan sidechains are. Each EPprocyclin molecule contains a Man 5 GlcNAc 2 N-linked glycan and an ethanolamine-P-6Manα1-2Manα1-6Manα1-4GlcN GPI glycan core (11). Of the latter, only two of the three Man residues are detected by the GC-MS method employed (because of the acid stability of the 6-linked phosphate group) (2). This means that there are seven detectable Man residues per molecule of procyclin. Based on this, we estimate the average number of Gal residues in the GPI anchor sidechains of the wild-type and add-back procyclins as 12 to 14 and in the Tb927.7.6110/6150/6170 −/− null mutant as 6 to 7 (Fig. 5E).
We focused on comparing the GPI anchor glycan sidechains in the wild-type and Tb927.7.6110/6150/6170 −/− null mutant procyclins. The positive ion MALDI-ToF spectra of the aq. HF-released and permethylated GPI glycans were consistent with a smaller average size of GPI glycan sidechain in the mutant procyclins (Fig. 6).
The GC-MS methylation linkage analysis results for these same preparations were normalized to the areas for the non- The GPI-PLA2 of fatty acid remodelling in T.brucei reducing terminal-Man residues arising from the common Man 5 GlcNAc 2 N-linked glycans. As expected, the relative 3,6disubstituted Man signals, also arising from the common Man 5 GlcNAc 2 N-linked glycans, were very similar. On the other hand, the relative amounts of Gal and GlcNAc derivatives were significantly reduced for the Tb927.7.6110/ 6150/6170 −/− null mutant sample, particularly those corresponding to terminal-Gal, 3-substituted Gal, 3,6-disubstituted Gal and 3-substituted GlcNAc residues ( Fig. 7 and Table S1). These data suggest that the Tb927.7.6110/6150/6170 −/− null mutant GPI glycan sidechains are principally deficient in lacto-N-biose (-3Galβ1-3GlcNAcβ1-) repeats.

Discussion
The identity of the GPI-PLA2 enzyme responsible for GPI precursor fatty acid remodeling in T. brucei (Fig. 1) has been elusive for many years, but mammalian PGAP6 (24,25) provided an inroad to identify potential GPI-PLA2 encoding genes in T. brucei. Three PGAP6-related genes were identified (Fig. 2) in a tandem array with homologous intergenic regions, making individual gene knockouts difficult to achieve. We deleted all three genes in PCF T. brucei to make a Tb927.7.6110/6150/6170 −/− null mutant and observed a biochemical phenotype consistent with the deletion of GPI-PLA2 activity: Namely, the presence of sn-1,2-diacylglycerol in place of sn-1-acyl-2-lyso-glycerol in the PI component of the GPI membrane anchors of the procyclin surface glycoproteins. Through a tetracycline-inducible add-back approach, we were able to show that only one of the three genes, Tb927.7.6110, was necessary and sufficient to reverse the mutant phenotype to that of wild-type. From this, we conclude that Tb927.7.6110 encodes the GPI-PLA2 of GPI precursor fatty acid remodeling, a biochemical activity originally E B A D C Figure 4. Analysis of procyclin PI species. A, schematic of the GPI anchor structure core of procyclin (centre) showing the Lacto-N-biose (-3Galβ1-3GlcNAc β1-) and LacNAc (-6Galβ1-4GlcNAc β1-) repeats in the GPI glycan side chains. The schematic also shows the expected liberation of PI structures from GPI anchored procyclin after nitrous acid deamination (bottom) and the expected GPI glycan core released after aq. HF dephosphorylation (top). B-E, the negative ion ES-MS spectra of the PI species released upon nitrous acid deamination from wild-type (B), Tb927.7.6110/6150/6170 −/− null mutant (C), Tb927.7.6110 + add back (D) and Tb927.7.6170 + add back (E) procyclins. The PI species were observed as [M-H] − ions and the major [M-H] − precursor ions were subjected to ES-MS 2 using collision-induced dissociation (Fig. S4). Refer to Table 1 for the fatty acid compositions of various PI ions. The minor ions between m/z 850 to 900 in (C) and (E) correspond to lyso-PI species, similar to those from wild-type procyclins.
described by Masterson et al. (16) in BSF trypanosomes. While we cannot exclude the formal possibility that PCF and BSF parasites use different GPI-PLA2 activities, the polysomeassociated transcript levels of Tb927.7.6110 (and also of Tb927.7.6150 and Tb927.7.6170) are very similar in PCF and BSF cells (Fig. S5), suggesting that the GPI-PLA2 encoded by Tb927.7.6110 is most likely expressed at similar levels and used in GPI precursor fatty acid remodeling in both lifecycle stages.
We had expected that deletion of GPI-PLA2 activity might result in slightly increased mobility of the procyclins on SDS-PAGE because a more hydrophobic (diacyl-PI containing) GPI anchor lipid should bind more SDS than the native (acyl-lyso-PI containing) GPI lipid. However, we were surprised at the observed increase in mobility and increase in dispersion (Fig. 3). Since the apparent MW heterogeneity of procyclins stems primarily from the heterogeneity of their GPI anchor carbohydrate sidechains (12,28), we compared the carbohydrate contents and GPI glycan structures of the wild-type and Tb927.7.6110/6150/6170 −/− null mutant PCF cells. The data showed an approximate halving in Gal content in the mutant (Fig. 5) and a reduction in the size profile of GPI glycans released from the procyclins (Fig. 6), apparently from the selective loss of lacto-N-biose (-3Galβ1-3Galβ1-) repeat units  The GPI-PLA2 of fatty acid remodelling in T.brucei (Fig. 7). We, therefore, interpret the change in SDS-PAGE pattern of the procyclins in the Tb927.7.6110/6150/6170 −/− null mutant as a combination of a reduction in SDS-binding capacity and overall GPI glycan sidechain size. The change in overall GPI sidechain size in the Tb927.7.6110/6150/6170 −/− null mutant is interesting. One explanation could be that with an extra fatty acid now present on the GPI anchor, the procyclins and the GPI anchor core present different aspects of the membrane. This might, in turn, reduce the accessibility of GPI anchor to some of the GPI sidechain processing glycosyltransferases, some or all of which are in the Golgi apparatus, reviewed in (29), reducing overall glycosylation. Another, not mutually exclusive, explanation could be that, with the additional fatty acid present, the transit time of the procyclins through the Golgi might be reduced, leading to less processing. Yet another possibility is that Tb927.7.6110 not only encodes GPI-PLA2 activity but also couples newly synthesized procyclins to their GPI-processing glycosyltransferases in some way. This might also explain the odd phenotype of the Tb927.7.6170 add-back where there is no restoration of GPI-PLA2 activity (Fig. 4) yet its overexpression causes a notable change in procyclin apparent MW range by SDS-PAGE (Fig. 3). It is conceivable, therefore, that the GPI side chain modification is influenced in some way by the independent interaction of Tb927.7.6110 and Tb927.7.6170 proteins with downstream GPI processing enzymes. Interestingly, a similar inference to this was made recently in BSF T. brucei for TbGPI2, a component of the UDP-GlcNAc: PI α1-6 GlcNAc transferase complex that catalyzes the first committed step in GPI biosynthesis in T. brucei (30).
While Tb927.7.6170 does not encode a GPI precursor fatty acid remodeling GPI-PLA2 activity, it might encode a similar activity, such as GPI-PLA1. The latter does not act in PCF cells but is required to complete fatty acid remodeling in BSF cells (Fig. 1). The obligate requirement for GPI-PLA2 action and Gup1-mediated myristoylation of the sn-2-position prior to GPI-PLA1 action in the BSF GPI fatty acid remodeling pathway (11,16) make this difficult to assess. Indeed, our attempts to make comparable Tb927.7.6110/6150/6170 −/− null and conditional add-back cell lines in BSF T. brucei have been unsuccessful.
Another potential role for the Tb927.7.6170 gene product could be as a distinct GPI-PLA2 for the process of BSF GPI anchor fatty acid proof-reading, as described in (31,32). In this process, it is postulated that the myristic acids of the sn-1,2dimyristoylglycerol component of the VSG GPI anchor are removed and replaced by fresh myristic acid in the endosomal compartment of the cell, requiring GPI-PLA2 and GPI-PLA1 activities.
The enzymes of GPI fatty acid remodeling were thought to be potential drug targets for human and animal African The GPI-PLA2 of fatty acid remodelling in T.brucei trypanosomiases (33). However, these studies with myristate analogs may, in retrospect, have been confounded by the essentiality of trypanosome protein N-myristoyltransferase (NMT) (34)(35)(36). More work is required to assess the essentiality of GPI-PLA2 (encoded by Tb927.7.6110) in BSF T. brucei, and the role and essentiality of its relative encoded by Tb927.7.6170 and of their presumably enzymatically inactive relative encoded by Tb927.7.6150.

Experimental procedures
Cultivation of trypanosomes T. brucei brucei Lister strain 427 procyclic form (PCF) parasites, that maintain T7 polymerase and tetracycline repressor protein under G418 and hygromycin antibiotic selection, were used in this study (37). This genetic background will be referred to from hereon as wild-type. The cells were cultivated in SDM-79 medium supplemented with 15% fetal bovine serum (FCS), Glutamax, and hemin, and containing 15 μg/ml G418 and 50 μg/ml hygromycin at 28 C in a 5% CO 2 incubator.

DNA isolation and manipulation
Plasmid DNA was purified from Escherichia coli DH5α competent cells using a Qiagen Miniprep kit. Extraction and purification of DNA from gels was performed using Qiaquick kits (Qiagen). Custom oligonucleotides were obtained from Thermo Fisher. T. brucei genomic DNA was isolated from  Table S1). C, the comparison of PMAA signals (n = 1) of wild-type and Tb927.7.6110/6150/6170 −/− null mutant samples, normalized to non-reducing-terminal mannose (t-Man) in each sample.

Generation of gene replacement constructs
A full list and descriptions of primers used in this study are summarized in (Table S2). A construct containing a puromycin acetyltransferase (PAC) gene flanked at the 5 0 end with 500 bp of Tb927.7.6170 5 0 UTR followed by 135 bp of T. brucei actin gene 5 0 UTR, and at the 3 0 end with 293 bp actin 3 0 UTR followed by 500 bp of Tb927.7.6110 3 0 UTR was synthesized by Genscript. The same construct containing a blasticidin-S deaminase (BSD) gene in place of PAC was generated by Gibson assembly (NEB) using primers ZJ1-ZJ4 (Table S2). The first copy of the Tb927.7.6170, Tb927.7.6150, and Tb927.7.6110 locus was replaced with the PAC drug resistance construct to generate a single knock out (sKO) mutant for these three genes. The second copy of the Tb927.7.6170, Tb927.7.6150, and Tb927.7.6110 locus was replaced by the BSD resistance construct to generate the Tb927.7.6170, Tb927.7.6150, Tb927.7.6110 −/− double knockout (dKO) mutant. The identity of all constructs was confirmed by DNA sequencing.

Generation of T. brucei add-back constructs
To generate the add-back constructs for three genes (Tb927.7.6110, Tb927.7.6150, Tb927.7.6170), their individual open reading frames (ORFs) were synthesized by Genscript. The individual ORFs were then amplified by PCR using primers ZJ6-ZJ10 (Table S2) and cloned individually into a pLEW100v5 vector (pLEW100v5 was a gift from George Cross; Addgene plasmid # 24011) using Gibson assembly, placing the genes under a tetracycline-inducible promoter. The identity of all constructs was confirmed by DNA sequencing.

Transformation of T. brucei procyclic form cells
The gene replacement constructs, and ectopic expression (add back) constructs were purified using the Qiagen Miniprep kit, linearized by NotI (NEB) restriction digestion, precipitated and washed twice with 70% ethanol, and redissolved in sterile water. The precipitated linearized DNA was used to electroporate T. brucei PCF as described in (37). The generation of sKO and dKO (null) mutants was confirmed by Southern blot.

Southern blotting
The DNA probes used in this study were digoxigenin (DIG)labeled. These were generated using a PCR DIG Probe Synthesis Kit (Roche) according to the manufacturer's protocol. ZJ13 and ZJ14 primers were used for the synthesis of the PAC resistance cassette probe, ZJ15 and ZJ16 primers were used for the synthesis of the BSD resistance cassette probe, ZJ17 and ZJ18 primers were used for the synthesis of the Tb927.7.6170 ORF probe. For analysis of T. brucei mutants, 5 μg of genomic DNA from the clones were digested solely with XhoI (NEB) overnight at 37 C. RNAase (Sigma) was added to the reaction to avoid RNA contamination for later detection. Endonuclease-digested gDNA was separated on a 0.8% (w/v) agarose gel for 4 h at 40 V in TAE buffer. The gel was washed with 0.25 M HCl with mild agitation (30 rpm) for 10 min to depurinate the DNA followed by 15 min denaturation with 0.5 M NaOH and 20 min neutralization with buffer 1 M Tris-HCl, 1.5 M NaCl, pH 7.5. After these steps, the DNA samples were transferred to a positively charged nylon membrane (Roche) through reverse capillary action overnight with 10 × saline-sodium citrate buffer (SSC) buffer. Transferred DNA fragments were covalently cross-linked to the membrane by UV cross-linking in a CL-100 (UVP) UV crosslinker at 1200 mJoules. The membrane was pre-incubated with 20 ml DIG Easy Hyb Granules solution (Roche) at 42 C for 1 h in a hybridization oven (Techne Hybridization Oven). The DIGlabelled probe was denatured for 5 min at 100 C followed by a rapid cool down on ice prior to mixing with 20 ml DIG Easy Hyb Granules solution. The membrane was incubated overnight with this hybridization solution with 20 μl denatured probe. Following the hybridization, the membrane was washed twice at low stringency (42 C, 5 min, 1 × SSC with 0.01% (w/v) SDS) and high stringency (65 C, 15 min, 0.5 × SSC with 0.01% (w/v) SDS) sequentially. The blot was then developed with DIG wash and block buffer set (Roche) following the manufacturer's instructions. The blots were equilibrated in 1 × wash buffer for 5 min at room temperature before being blocked in blocking buffer for 30 min. The anti-DIG AP-conjugate antibody (Roche) was then added to the blocking buffer with a 1:10,000 dilution and the blot was incubated further for 30 min followed by twice washing with 1 × wash buffer (15 min). The membrane was placed in a plastic folder and chemiluminescent substrate CSPD (Roche) was applied and incubated for detection. The blots were exposed to Amersham Hyperfilm ECL film for 1 to 30 min and developed with a KODAK film developer. For stripping the blot, 0.4 M NaOH was applied twice to the membrane and for 5 min at 42 C in the hybridization oven. The blot was then washed three times with 1× SSC buffer for 10 min at 42 C before re-probing.  (Table S2). The individual-specific gene transcript levels are presented as fold change relative to wild-type controls, determined using the comparative CT (ΔΔCT) method using β-tubulin as an endogenous control for normalization. All reactions were carried out in technical triplicates.

Extraction and purification of procyclins
Procyclins were purified from 10 10 cells by organic solvent extraction and octyl-Sepharose chromatography as previously described (30). Briefly, the cells were extracted three times with chloroform/methanol/water (10:10:3, v/v). The pellet obtained after the delipidation process was extracted twice with 9% butan-1-ol in water. The pooled supernatants were back-extracted twice with an equal volume of 9% water in butan-1-ol and the lower 9% butan-1-ol in water phase containing procyclins was recovered and dried under N 2 stream.
Solvent-extracted procyclins were used for nitrous acid deamination and monosaccharide composition analysis. While for analyzing the procyclin by MALDI-ToF and permethylated GPI glycans, the extracted procyclins were further purified using octyl-Sepharose 4B (Sigma) chromatography. Briefly, the extracted procyclins, dried and redissolved in buffer A (5% propan-1-ol in 0.1 M ammonium acetate) were applied to 0.5 ml of octyl-Sepharose 4B, packed in a disposable column, and pre-equilibrated with buffer A. The column was washed with buffer A followed by buffer B (5% propan-1-ol). The procyclins were then eluted in buffer C (50% propan-1-ol), concentrated, and dried until further use.

Analysis of procyclins by MALDI-ToF
The solvent-extracted procyclins (approx. 600 pmol) were dried and subjected to dephosphorylation using 50 μl of icecold 50% aqueous hydrogen fluoride (aq. HF) for 24 h at 0 C to cleave the GPI anchor ethanolamine-phosphate bond. Dephosphorylated samples were further subjected to mild acid treatment with 50 μl of 40 mM trifluoroacetic acid at 100 C for 20 min to cleave Asp-Pro bonds and thus remove the N-glycosylated N-termini. The samples were dried and redissolved in 5 μl of 0.1% trifluoroacetic acid and the aliquots were co-crystallized with α-cyano-4-hydroxycinnamic acid matrix (10 mg/ml in 50% acetonitrile, 0.1% trifluoroacetic). The samples were analyzed by linear-mode negative-ion MALDI-ToF (Autoflex Speed MALDI-ToF MS system by Bruker).

Deamination of extracted procyclin and ES-MS analysis
The procyclins were deaminated as described (12) with minor modifications. Briefly, 15% of the extracted procyclin preparation was dried and deaminated with 50 μl of 0.3 mM sodium acetate buffer (pH 4.0) containing 250 mM sodium nitrite for 1 h at room temperature. The samples were replenished with a further 50 μl of 0.3 mM sodium acetate buffer (pH 4.0) containing 250 mM sodium nitrite and incubated for another 2 h at 37 C. The deaminated procyclins were extracted three times with water-saturated butan-1-ol (i.e., butan-1-ol containing 9% water). The upper butan-1-ol phase containing the released phosphatidylinositol (PI) moieties was pooled in a fresh tube and dried under N 2 . To analyze the released PI components, the samples were redissolved in 100 μl of chloroform/methanol (1:1, v/v) and infused into an LTQ Orbitrap Velos Pro mass spectrometer (Thermo Scientific) using static infusion nanoflow probe tips (M956232AD1-S, Waters). Data were collected in negative ion mode for ES-MS and ES-MS 2 . The negative ion spray voltage was 0.8 kV, the temperature of ion transfer tube was 275 C and collision-induced dissociation (CID) was used for MS 2 fragmentation, using 25% to 35% collision energy.

Permethylation and analysis of GPI glycans by MALDI-ToF MS
The octyl-Sepharose purified procyclin samples were subjected to dephosphorylation and permethylation as described in (30). Briefly, the procyclins were treated with 100 μl of icecold 50% aqueous hydrogen fluoride (aq. HF) for 24 h at 0 C to cleave the GPI anchor ethanolamine-phosphate-mannose and inositol-phosphate-acylglycerol phosphodiester bonds and release the GPI glycan. After removing the aq. HF by freeze drying, the released GPI glycans were resuspended in 100 μl water, centrifuged at 16000 x g for 10 min, and obtained in the supernatant. The permethylation of released GPI glycans was performed using the sodium hydroxide method, as described earlier (38) to obtain the GPI glycans bearing a fixed positive charge in the form of an N-trimethyl-glucosamine quaternary ammonium ion. Aliquots (10%) of permethylated GPI-glycan were analyzed using positive ion MALDI-ToF MS (Autoflex Speed MALDI-ToF MS system by Bruker). The permethylated GPI glycans samples were co-crystallized with a matrix of 2,5-dihydroxybenzoic acid (20 mg/ml 30% acetonitrile and 0.1% TFA) and analyzed in reflectron positive ion mode.

Methylation linkage analysis
Methylation linkage analysis of GPI glycans was performed as described in (30). Briefly, 80% of the permethylated GPI The GPI-PLA2 of fatty acid remodelling in T.brucei glycan samples were subjected to acid hydrolysis, NaB[ 2 H] 4 reductions, and acetylation to generate partially methylated alditol acetates (PMAAs) which were analyzed by GC-MS.
Supporting information-This article contains supporting information (40). Conflicts of interest-The authors declare that they have no conflicts of interest with the contents of this article.