RFT1 Protein Affects Glycosylphosphatidylinositol (GPI) Anchor Glycosylation*

The membrane protein RFT1 is essential for normal protein N-glycosylation, but its precise function is not known. RFT1 was originally proposed to translocate the glycolipid Man5GlcNAc2-PP-dolichol (needed to synthesize N-glycan precursors) across the endoplasmic reticulum membrane, but subsequent studies showed that it does not play a direct role in transport. In contrast to the situation in yeast, RFT1 is not essential for growth of the parasitic protozoan Trypanosoma brucei, enabling the study of its function in a null background. We now report that lack of T. brucei RFT1 (TbRFT1) not only affects protein N-glycosylation but also glycosylphosphatidylinositol (GPI) anchor side-chain modification. Analysis by immunoblotting, metabolic labeling, and mass spectrometry demonstrated that the major GPI-anchored proteins of T. brucei procyclic forms have truncated GPI anchor side chains in TbRFT1 null parasites when compared with wild-type cells, a defect that is corrected by expressing a tagged copy of TbRFT1 in the null background. In vivo and in vitro labeling experiments using radiolabeled GPI precursors showed that GPI underglycosylation was not the result of decreased formation of the GPI precursor lipid or defective galactosylation of GPI intermediates in the endoplasmic reticulum, but rather due to modifications that are expected to occur in the Golgi apparatus. Unexpectedly, immunofluorescence microscopy localized TbRFT1 to both the endoplasmic reticulum and the Golgi, consistent with the proposal that TbRFT1 plays a direct or indirect role in GPI anchor glycosylation in the Golgi apparatus. Our results implicate RFT1 in a wider range of glycosylation processes than previously appreciated.

Trypanosoma brucei is a human and animal parasite endemic in sub-Saharan Africa causing sleeping sickness in humans and nagana in livestock. The life cycle of the extracellular living parasite comprises stages in the midgut and salivary gland of the tsetse fly vector and in the blood of the mammalian host. Alternating between the two organisms, the parasite not only adapts its energy metabolism to the respective environment but also its cell surface protein coat, which is a crucial determinant of the parasite's virulence. Apart from being a deadly pathogen affecting the socio-economic development in endemic areas, the parasite has emerged as an interesting model organism for basic research. Two proliferative stages of T. brucei, bloodstream trypomastigotes (bloodstream form parasites) and the insect-stage procyclic trypomastigotes (procyclic form parasites), can easily be cultured in vitro. Biological features such as trans-splicing (1), RNA editing (2), and antigenic variation (3) were first described in Trypanosomatids and only later also found in other eukaryotes. In addition, T. brucei was one of the first organisms in which glycosylphosphatidylinositol (GPI) 3 anchoring of cell surface proteins was described and extensively explored (4,5). Protein-linked GPI anchors consist of the conserved core structure ethanolamine-HPO 4 -Man␣1-2Man␣1-6Man␣1-4GlcN␣1-6-myo-inositol phospholipid with the amino group of ethanolamine linked to the C terminus of the protein (4,5). A wide variety of linear and branched glycosyl substituents and additional ethanolamine phosphate moieties can be attached to this core, depending on the protein to which the anchor is attached and the organism in which it is synthesized. The best-studied and most abundant GPI-anchored proteins of T. brucei are the variant surface glycoproteins (VSGs) in bloodstream form parasites (6) and the procyclins in procyclic forms (7)(8)(9). Although the variant surface glycoprotein GPI core is modified by rather simple galactosyl side chains (5), the GPI anchors of procyclins are the largest and most complex anchors known, comprising large branched N-acetyllactosamine (Gal␤1-4GlcNAc) and lacto-N-biose (Gal␤1-3GlcNAc)-containing side chains often capped with ␣2-3-linked sialic acid residues (10,11). Based on their C-terminal amino acid sequences containing di-or pentapeptide tandem repeats, procyclins are divided into two classes: EP (rich in Glu-Pro repeats) and * The work was supported by Swiss National Science Foundation Grants Sinergia CRSII3_141913 and 149353 (to P. B.), National Institutes of Health Grant R21NS093457 (to A. K. M.), and GlycoPar-EU FP7 Marie Curie Initial Training Network Grant GA. 608295 (to A. A. S.). The authors declare that they have no conflicts of interest with the contents of this article. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health. 1 Both authors contributed equally to this work. 2  GPEET (rich in Gly-Pro-Glu-Glu-Thr repeats) procyclins (11). Two of the three subclasses of EP procyclins, EP1 and EP3, contain a single N-glycosylation site (11,12), whereas EP2 and GPEET procyclins are not N-glycosylated. Interestingly, EP1 and EP3 procyclins are modified exclusively by a triantennary Man 5 GlcNAc 2 moiety (11), transferred to protein by oligosaccharyltransferase TbSTT3B, which is expressed in procyclic forms (13) and specifically uses mature Man 9 GlcNAc 2 -PPdolichol (mDLO) for transfer to N-glycosylation sites (14). Due to the lack of a Golgi ␣-mannosidase in procyclic form trypanosomes, Man 9 GlcNAc 2 glycans can only be trimmed to triantennary Man 5 -GlcNAc 2 that are not further modified (15). Fig. 1A shows a schematic representation of a typical N-glycosylated EP procyclin. RFT1 was first described in Saccharomyces cerevisiae as a protein "requiring fifty-three," i.e. human p53, in a screen of mutants that could be rescued by heterologous expression of p53 (16). Only later, yeast RFT1 (Rft1p) was found to play an essential role in protein N-glycosylation (17). The multi-pass transmembrane protein was reported to be localized in the endoplasmic reticulum (ER) of yeast and human cells (17,18), although no corresponding localization data have been published. The accumulation of the dolichol-linked oligomannose intermediate Man 5 GlcNAc 2 -PP-dolichol (M5-DLO) in S. cerevisiae cells depleted of Rft1p but having intact O-glycosylation and GPI anchoring suggested that RFT1 is a flippase enabling translocation of M5-DLO across the ER membrane (17). However, this interpretation was challenged by subsequent biochemical studies, where flipping of M5-DLO was assayed in vitro using proteoliposomes containing Triton X-100-extracted yeast ER membrane proteins (19,20). Flipping of M5-DLO occurred robustly in the absence of Rft1p; e.g. when proteoliposomes were reconstituted with fractionated ER membrane proteins, M5-DLO translocation activity was found in fractions devoid of Rft1p (19). Similar experiments using sealed microsomes confirmed these findings (21). Hence, it was postulated that S. cerevisiae Rft1p has only an indirect involvement in the translocation of M5-DLO (19 -21).
More recently, the role of RFT1 was revisited using T. brucei as a model organism. By complementation of yeast lacking RFT1 function, T. brucei RFT1 (TbRFT1; Tb927.11.11670) was shown to represent a functional homolog of S. cerevisiae Rft1p (22). TbRFT1 null procyclic trypanosomes grew nearly normally and had normal steady-state levels of mDLO and reduced but still significant N-glycosylation, indicating robust M5-DLO flippase activity. Nevertheless, TbRFT1 null parasites had 30 -100-fold greater steady-state levels of M5-DLO when compared with wild-type trypanosomes. Fluorophore-assisted carbohydrate electrophoresis analysis of N-glycans released from N-linked glycoproteins showed that all N-glycans in the TbRFT1 null cells originate from mDLO, indicating that the M5-DLO excess is not used for glycosylation. Together, these results suggested that rather than facilitating M5-DLO flipping, RFT1 appears to promote conversion of M5-DLO to mDLO by another mechanism, possibly by acting as an M5-DLO chaperone (22).
We now report that the lack of TbRFT1 in T. brucei procyclic forms not only affects N-glycosylation but also GPI anchor gly-cosylation. In addition, we unexpectedly localize TbRFT1 to both the ER and the Golgi. These results suggest that RFT1 has a pleiotropic influence on protein glycosylation.

Results and Discussion
Procyclins of TbRFT1 Null Cells Exhibit Reduced Apparent Molecular Masses-Based on our previous observation that the lysosomal marker protein p67 is underglycosylated in TbRFT1 null mutants (22), we investigated whether a similar glycosylation phenotype could also be observed for the major surface coat protein of T. brucei procyclic forms, EP procyclin. EP procyclins are encoded by three different genes, EP1-3, with EP1 and EP3 proteins each containing a single N-glycosylation site, whereas EP2 and the other subclass of procyclin, GPEET, are not N-glycosylated (11,23,24). The relative abundance of EP1-3 and GPEET varies among trypanosome strains and culture conditions (11,12,25) and during tsetse infection (26,27). In the strain used in this study, T. brucei Lister 427, GPEET represents the predominant surface protein, but EP is also expressed (25). Analysis by SDS-PAGE and immunoblotting revealed a smaller apparent molecular mass of EP in TbRFT1 null cells when compared with WT trypanosomes (Fig. 1B). Treatment of EP from WT cells with protein N-glycosidase F (PNGase) to release protein N-glycans ( Fig. 1A) reduced its apparent molecular mass. However, the apparent size of de-N-glycosylated EP from WT trypanosomes was larger than that of EP from TbRFT1 null cells, which was unaffected by PNGase treatment (Fig. 1B). Together, these results indicate that the altered molecular mass of EP is not only due to altered N-glycosylation in TbRFT1 null parasites but involves other modifications caused by lack of TbRFT1. TbRFT1 Null Procyclins Have Truncated GPI Anchor Side Chains-Because all EP isoforms, as well as GPEET, are GPIanchored (23,28), we hypothesized that decreased glycosylation of the GPI anchor side chain may contribute to the observed phenotype. The procyclin GPI anchors are modified by a large heterogeneous, branched side chain comprising poly-N-acetyllactosamine and lacto-N-biose units that may be capped with sialic acid residues (10, 11) (Fig. 1A). Partial or complete loss of this glycan moiety would lead to reduced apparent molecular masses of GPEET and EP. Hence, we analyzed the procyclin GPI anchors by in vivo labeling of trypanosomes with [ 3 H]ethanolamine ([ 3 H]Etn), which gets incorporated into the GPI core structure (23,25). SDS-PAGE and fluorography showed that GPEET, migrating with an apparent molecular mass of 22-29 kDa, was readily labeled with [ 3 H]Etn in WT trypanosomes (Fig. 1C). Labeling of GPEET was also observed in TbRFT1 null parasites; however, the protein migrated with a lower molecular mass when compared with WT trypanosomes (Fig. 1C). To eliminate molecular mass differences present in the protein part of GPEET, extracts were treated with Pronase, which has been shown to digest the entire protein portion of GPEET down to the C-terminal glycine residue (25). The resulting GPI anchor was again analyzed by SDS-PAGE and fluorography and showed a clear difference in molecular mass between WT and TbRFT1 null parasites (Fig. 1C).
To confirm that the GPEET polypeptide itself is not truncated in TbRFT1 null parasites, its mass was analyzed by MALDI-TOF-MS after sequential treatment with aqueous hydrofluoric acid (to remove the GPI anchor, leaving ethanolamine attached to the C-terminal amino acid (Fig. 1A)) and mild trifluoroacetic acid (to cleave Asp-Pro bonds within the EP sequence ( Fig. 1A)) (12). The results in Fig. 2, A and B, show that both WT and ⌬Tbrft1 cells express the same GPEET fragments GP-4 and GP-13, representing proteins lacking 4 and 13, respectively, N-terminal amino acids. Furthermore, the masses are consistent with the presence of ethanolamine linked to Gly, further corroborating that ⌬Tbrft1 cells are not defective in the transfer of GPIs to proteins. In addition, a series of EP procyclin C-terminal fragments was also detected.
To determine the degree of GPI underglycosylation in ⌬Tbrft1 cells, butan-1-ol extracts (rich in procyclins) were analyzed by GC-MS. As expected, although the WT sample yielded a composition of Man:Gal:GlcNAc:Sia of 1.0:1.4:0.4: 0.2, ⌬Tbrft1 cells showed an ϳ7-fold reduction in the overall GPI sugar content, resulting in a ratio of Man:Gal:GlcNAc:Sia of 1.0:0.2:0.1:ϳ0 (Sia were not detectable). A similar reduction in the overall sugar composition of another ⌬Tbrft1 mutant (i.e. B1 cells (22)) was also observed by GC-MS (not shown). Collectively, these results show that trypanosomes lacking TbRFT1 express procyclins with truncated GPI anchor side chains containing fewer poly-N-acetyllactosamine/lacto-Nbiose repeats and sialic acids, thus explaining the observed reductions in apparent molecular masses after SDS-PAGE.
Procyclin GPI Anchor Size Is Restored by Ectopically Expressed TbRFT1-To study whether the observed differences in EP and GPEET molecular masses between WT and TbRFT1 null mutants are indeed due to the lack of TbRFT1, we generated add-back mutants by expressing HA-tagged or untagged copies of TbRFT1 in the ⌬Tbrft1 background. If functional, these ectopically expressed proteins are expected to restore the molecular masses of EP and GPEET to wild-type sizes. Immunoblotting using antibodies against the HA epitope demonstrated that RFT1-HA was expressed in the respective clones (Fig. 3A). In addition, the results showed that in TbRFT1 null parasites expressing HA-TbRFT1 or untagged TbRFT1, the apparent molecular mass of EP was comparable with wild-type EP procyclin (Fig. 3B), indicating that both tagged and untagged TbRFT1 are functional and restored EP glycosylation as well as GPI glycan maturation. A similar result was obtained by analyzing [ 3 H]Etn-labeled EP and GPEET using SDS-PAGE and fluorography (Fig. 3C).
GPI Precursor Synthesis and in Vitro GPI Galactosylation Are TbRFT1-independent-To study whether the lack of TbRFT1 affects the formation of PP1, the GPI precursor added to protein in the ER (29), trypanosomes were cultured in the presence of [ 3 H]Etn, which becomes incorporated into all ethanolaminecapped GPI precursors. Analysis of extracts from WT, TbRFT1 null, and TbRFT1 addback parasites by TLC revealed that PP1 FIGURE 1. Schematic representation of procyclin glycosylation and analysis of GPI-anchored proteins in TbRFT1 null cells. A, most procyclin isoforms (except EP2 and GPEET) are modified by a homogeneous triantennary Man 5 GlcNAc 2 glycan near the N terminus. The GPI anchor of all procyclins is modified with several N-acetyllactosamine or lacto-N-biose repeats, which may be capped with sialic acids depending on the presence of blood sialoglycoconjugates. These repeats are linked to the middle mannose of the GPI anchor core via two consecutive galactose residues (dotted box), which are probably added in the ER as suggested in Fig. 2. The C-terminal regions of EP procyclins consist of 22-30 Glu-Pro repeats. PI, phosphatidylinositol. B, immunoblotting analysis of EP procyclins isolated from WT and TbRFT1 null (⌬rft1) cells. Denatured proteins were treated with (ϩ) or without (Ϫ) PNGase F to remove N-glycans and then separated by SDS-PAGE. After electrotransfer to membranes, EP was visualized by enhanced chemiluminescence using anti-EP antibody and HRP-conjugated anti-mouse IgG. C, [ 3 H]ethanolamine labeling and fluorography of GPI-anchored proteins from WT and TbRFT1 null (⌬rft1) cells. Trypanosomes were grown in the presence of [ 3 H]ethanolamine, and GPI-anchored proteins were extracted from the delipidated protein pellet using 9% butan-1-ol. Extracts incubated in the absence (Ϫ) or presence (ϩ) of Pronase to remove the protein portions of GPEET and EP were separated by SDS-PAGE and analyzed by fluorography.
was the major [ 3 H]Etn-labeled lipid irrespective of the presence or absence of TbRFT1 (Fig. 4A).
Assembly of the GPI core structure and attachment to protein occurs in the ER (30 -32). At present, it is unclear whether the glycan modification at the central mannose residue starts before or after GPI attachment to protein in the ER, or on protein-bound GPI anchors in the Golgi. It is known that galactose is added in vitro to GPI anchor precursors in bloodstream form T. brucei (33), suggesting the presence of ER-resident galactosyltransferases that are able to act on GPI precursors. To date, only two T. brucei glycosyltransferases involved in GPI processing are known: GT8 and GT3. However, both seem to localize to the Golgi, with GT8 mediating the transfer of the first GlcNAc moiety to the terminal digalactose moiety of the immature GPI anchor (34,35) and GT3 attaching a galactose residue to GlcNAc (36). To study whether decreased galactosylation in the ER may cause underglycosylation of the GPI anchors in TbRFT1 null cells, we pulse-labeled crude membrane preparations from TbRFT1 add-back and TbRFT1 null parasites with [ 3 H]GDP-mannose ([ 3 H]GDP-Man) in the presence of UDP-GlcNAc and chased the labeled GPI precursors with nonradiolabeled GDP-Man and UDP-Gal. Tunicamycin was added to the assays to inhibit formation of the N-glycan precursor dolichyl-pyrophosphate GlcNAc (Dol-PP-GlcNAc). TLC analysis of 3 H-labeled GPI lipids after extraction with chloroform/ methanol/water followed by butan-1-ol-water partitioning showed that additional more polar species were formed during the chase with UDP-Gal (Fig. 4B, green arrows). However, we observed no differences in the galactosylation pattern between wild-type and TbRFT1 null cells, indicating that galactosylation of GPI precursors in the ER is not affected by lack of TbRFT1.
The ER lumenal mannose donor dolichol-phosphate mannose (Dol-P-Man) is one of the most prominent radiolabeled lipid species after pulse labeling membranes with [ 3 H]GDP-Man (Fig. 4B). As expected, Dol-P-Man as well as the early GPI intermediate Man␣1-4GlcN␣1-6-myo-inositol phospholipid (Man 1 -GlcN-PI) almost completely disappeared during the chase. Interestingly, extracts from both pulse-labeled and chase-labeled ⌬Tbrft1 membranes contained additional polar mannose-containing lipids (Fig. 4B, blue arrows). A comparison with extracts from membranes labeled in the absence of tunicamycin (Fig. 4C) shows that compounds with similar R f values are also formed by WT membranes as long as N-glycan synthesis is enabled. Treatment of the labeled lipids with mild acid led to degradation of Dol-P-Man as well as the unknown polar species from both wild-type and mutant extracts (Fig. 4, C and D). Conversely, the acid-sensitive lipids were not cleaved by the GPI-specific phospholipase D, whereas the known GPI intermediates were readily hydrolyzed (Fig. 4D). Together, these results suggest that the unknown species made by mutant membranes are tunicamycin-insensitive dolichollinked oligomannose species that most likely originate from a pool of preformed (early) intermediates. We conclude that side-chain glycosylation of newly synthesized GPI precursors occurs normally in ⌬Tbrft1 membranes.
Possible Roles of TbRft1 in GPI Anchor Modification-The defects in GPI anchor maturation of TbRFT1 null cells can be interpreted in two ways. TbRFT1 may have a direct role in GPI anchor glycan modification that is independent of its function in N-glycosylation. Alternatively, the defect may result from incomplete N-glycosylation of a glycosyltransferase involved in GPI anchor modification, leading to decreased activity and thus incomplete glycan modification.
First, we considered the possibility that the appearance of truncated GPI anchors may be caused by a glycosylation defect occurring in the Golgi. To study whether TbRFT1 is present exclusively in the ER, as suggested for yeast (17), we analyzed the localization of the functional HA-tagged copy of TbRFT1 in the TbRFT1 null background using immunofluorescence microscopy. Co-staining of TbRFT1-HA with an antibody against the ER lumenal chaperone BiP confirmed the expected localization of TbRFT1 in the ER, where it was predominantly found in the perinuclear region (Fig. 5A). Interestingly, however, co-staining of TbRFT1-HA with an antibody against the Golgi resident protein TbGRASP (37) also showed co-localization (Fig. 5A). In Ͼ70% of parasites examined (n Ͼ 100), TbRFT1-HA co-stained with a spot located between the nucleus and the kinetoplast and representing the Golgi. To exclude the possibility that the observed co-localization of TbRFT1 with TbGRASP is unspecific, we analyzed the localization of an HA-tagged membrane-bound member of the EMC family of proteins, TbEMC3 (Tb927.10.4760), which has been shown in yeast to localize to the ER (38). The results showed Ͻ35% co-localization of TbEMC3-HA with TbGRASP (Fig.  5B). Although we cannot completely exclude the possibility that a portion of TbRFT1 is localized to the Golgi as a result of saturation of the retention system for ER membrane proteins, we consider it unlikely for the following reasons. (i) TbRFT1-HA shows a similar dual localization when expressed by a different, tetracycline-inducible expression vector in T. brucei 427 wild-type cells (Fig. 5C); (ii) overexpression of another ER-localized membrane protein, TbEPT (T. brucei ethanolamine phosphotransferase), using the same tetracycline-inducible vector showed no Golgi localization and was specifically targeted to the perinuclear ER (39); and (iii) there is no precedent in trypanosomes for mislocalization of ER proteins on overexpression of a single ER resident. Dual localization of trypanosome proteins in both the ER and the Golgi is not unique to TbRFT1 but has been reported before (40). In contrast, no co-localization was observed between TbRFT1-HA and cathepsin L, a marker for the T. brucei lysosome (41) that also localizes between the nucleus and the kinetoplast (Fig. 5D). The role of TbRFT1 in glycosylation in the Golgi remains speculative. In a recent publication, RNAi-mediated silencing of the nucleotide sugar transporter TbNST4 responsible for import of UDP-GlcNAc, GDP-mannose, and UDP-GalNAc into the Golgi resulted in production of underglycosylated EP procyclin in T. brucei procyclic forms (42). In addition, defective forms of GT3 (36) or GT8 (35, 43) resulted in impaired GPI glycan maturation and reduced protein N-glycosylation. It is possible that a lack of TbRFT1 affects Golgi resident proteins involved in glycosylation, by directly interacting with these proteins or by affecting their glycosylation status.
To study the second possibility, we analyzed GPEET procyclin in cells grown in the presence of tunicamycin during 10 days to inhibit N-linked glycosylation. The results show that the molecular mass of GPEET was reduced by tunicamycin treatment to that of GPEET in ⌬Tbrft1 cells (Fig. 6A). Binding of FITC-labeled concanavalin A to parasites followed by analysis by FACS demonstrated that tunicamycin effectively inhibited N-glycosylation (Fig. 6B).
Concluding Remarks-We report several new findings that implicate RFT1 in a wider range of glycosylation processes than previously demonstrated. Procyclic form trypanosomes lacking TbRFT1 not only have decreased protein N-glycosylation but also produce underglycosylated GPI anchors. The defect is not associated with the synthesis of the anchor but with GPI processing/maturation steps that likely occur in the Golgi apparatus. A role for TbRFT1 in the Golgi would be consistent with our observation that TbRFT1-HA is dually localized in the ER and Golgi. Our observations open new questions regarding the enigmatic function of TbRFT1 and its orthologs in yeast and mammalian cells.

Experimental Procedures
Materials-Unless otherwise stated, all reagents were of analytical grade and purchased from Sigma-Aldrich (Buchs, Switzerland) or Merck (Darmstadt, Germany). Restriction enzymes were from Fermentas (St. Leon-Rot, Germany), and antibiotics were from Sigma-Aldrich, InvivoGen (Nunningen, Switzer-  A, functional TbRFT1-HA was co-stained with the ER marker protein BIP (upper panels) or the Golgi marker TbGRASP (lower panels). The merged channels (composite) show overlap (yellow color) of the BIP and TbRFT1-HA signals mainly in the perinuclear zone, with some weaker signal distributed throughout the rest of the cells. Co-staining with the Golgi marker TbGRASP shows that the brightest spots of the HA signal co-localize with the Golgi signal. Some areas were zoomed for better visibility (panels on the right). DIC, differential interference contrast. B, trypanosomes expressing TbEMC3-HA were co-stained with TbGRASP (upper panels). The merged channels (composite) show little overlap (yellow color) of the signals. Golgi-stained areas were zoomed for better visibility (panels on the right). The lower panels show co-staining of TbRFT1-HA with TbGRASP done in parallel with the staining shown in the upper panels. Again, TbRFT1 co-localized with TbGRASP. DNA was stained with DAPI (blue). C, co-staining of TbRFT1-HA (RFT-HA) with the Golgi marker TbGRASP in cells transiently expressing TbRFT1-HA under the control of a tetracycline operator. Tetracycline (1 g/ml) was added to the growth medium for 20 h prior to preparation of slides. D, functional TbRFT1-HA was co-stained with the lysosomal marker protein cathepsin L (CatL) and DAPI as indicated. The composite shows little overlap (yellow color) of the signals. Cathepsin L-stained areas were zoomed for better visibility (panels on the right). land), or Invitrogen (Basel, Switzerland Trypanosome Cultures-T. brucei strain Lister 427 procyclic forms were cultured at 27°C in SDM-79 containing 5% heatinactivated fetal bovine serum. TbRFT1 knock-out trypanosomes (⌬Tbrft1, TbRFT1 null) (22) were grown under the same conditions, but in the presence of 1 g/ml G418 for the single allele knock-out and an additional 10 g/ml blasticidin for the double allele knock-out clones. TbRFT1 addback mutants in ⌬Tbrft1 double allele knock-out cells were selected and grown in the same medium, with an additional 2 g/ml puromycin. T. brucei strain Lister 427 29-13 (TetR T7RNAP) procyclic forms for tetracycline-inducible gene expression were cultured at 27°C in SDM-79 containing 10% heat-inactivated fetal bovine serum, 25 g/ml hygromycin, and 15 g/ml G418.

H]Etn Labeling of GPI Precursors and GPI-anchored
Proteins-For in vivo radiolabeling of procyclins, procyclic form trypanosomes were cultured in the presence of [ 3 H]Etn (2 Ci/ml) during 16 h to a density of ϳ1.5 ϫ 10 7 cells/ml, as described before (25). Cells were counted, harvested by centrifugation, and washed twice in cold TBS (10 mM Tris-HCl, pH 7.5, 144 mM NaCl). Bulk lipids from up to 2.5 ϫ 10 8 cells were extracted from the cell pellets using 2 ϫ 10 ml chloroform/ methanol (2:1, v/v), and GPI precursors and free GPIs were extracted using 3 ϫ 5 ml of chloroform/methanol/water (10: 10:3, v/v/v) (25). GPI-anchored proteins were extracted from the remaining protein pellet using 2 ϫ 1 ml of 9% (v/v) butan-1-ol in water during 2 h on ice, followed by 10 min of centrifugation at 17,000 ϫ g. The resulting supernatants were pooled, dried under a stream of nitrogen, and dissolved in electrophoresis sample buffer containing 2.5% (w/v) SDS. Butan-1-ol-insoluble material was further extracted with 0.1% (w/v) Triton X-100 in 20 mM Tris (pH 7.4) for 10 min at 95°C (25). Radioactivity in the butan-1-ol and Triton X-100 extracts was determined by liquid scintillation counting of small aliquots. For the analysis of [ 3 H]Etn-labeled GPI precursors, all cultures were adjusted to 1.5 ϫ 10 7 cells/ml prior to incubation with 4 Ci/ml [ 3 H]Etn during 4 h. After washing with TBS and extraction of bulk lipids using 2 ϫ 10 ml chloroform/methanol (2:1, v/v), GPI precursors were extracted by 3 ϫ 5 ml of chloroform/methanol/water (10:10:3, v/v/v). The resulting supernatants were pooled, dried under a stream of nitrogen, and partitioned between 0.5 ml of butan-1-ol and water. After a second extraction of the resulting water phase with 0.5 ml of water-saturated butan-1-ol, butanol phases were pooled and dried using a SpeedVac apparatus. Dry GPI lipids were resuspended in 50 l of chloroform/methanol/water (10:10:3, v/v/v) and separated by TLC as described below.
Mass Spectrometry Analysis-For the analysis of procyclin C-terminal polypeptides, total procyclins were purified by sequential extraction with chloroform/methanol/water (10: 10:3, by volume) and 9% (v/v) butan-1-ol as described above. Butan-1-ol extracts were then dried, dephosphorylated with 48% anhydrous hydrofluoric acid for 24 h at 0°C, hydrolyzed with mild (40 mM) trifluoroacetic acid, and then analyzed by negative-ion MALDI-TOF on an ABI Voyager DE-STR instrument using sinapinic acid as the matrix (12). To determine the total monosaccharide composition of procyclins, samples were mixed with 200 pmol of scyllo-inositol (as internal standard) and analyzed by GC-MS as described elsewhere (46).
Enzyme Treatments-For PNGase treatment, parasites were lysed by boiling 10 min at 100°C in denaturing buffer (0.5% SDS, 40 mM DTT) and incubated with PNGase F (New England Biolabs) according to the manufacturer's instructions in a buffer containing 1% Nonidet P-40 during 1 h at 37°C. For Pronase treatment, 3 H-labeled GPI extracts were dried under a stream of nitrogen, re-dissolved in a buffer containing 20 mM Tris (pH 7.5) and 5 mM CaCl 2 , and incubated with Pronase (0.3 mg/ml) during 16 h at 37°C.
Cell-free GPI Glycosylation Analysis-Membranes for cellfree labeling of GPI precursors were collected from trypanosome cultures grown to a density of 10 7 cells/ml by hypotonic lysis in water containing 0.1 mM tosyl-L-lysyl-chloromethane hydrochloride (TLCK) and 1.25 g/ml leupeptin. Pulse-chase radiolabeling of GPI precursors with [ 3 H]GDP-Man was performed essentially according to the protocol developed by Masterson et al. (47) as described by Leal et al. (48). If indicated, experiments were performed in the presence of 0.4 mg/ml tunicamycin in the assay buffer to inhibit the formation of dolichollinked N-glycan precursors. In pulse-chase experiments, membranes from 1.5 ϫ 10 8 cells were labeled with 3 Ci of [ 3 H]GDP-Man and 1 mM UDP-GlcNAc for 8 min at 37°C, followed by a 75-min chase with 2 mM non-radioactive GDP-Man. For the analysis of GPI galactosylation, 8 mM UDP-Gal was added during the chase with GDP-Man. GPI lipids were extracted as described (47). Butan-1-ol extracts were separated by TLC on Silica Gel 60 plates (Merck Millipore) using chloroform/methanol/water (10:10:3, v/v/v) as solvent system and analyzed using a radioactivity TLC scanner (Berthold Technologies, Regensburg, Switzerland). For autoradiography, TLC plates were treated with EN 3 HANCE TM spray (Perkin Elmer) prior to exposure to film at Ϫ70°C.
Flow Cytometry-Trypanosomes were grown in the presence or absence of tunicamycin (1 g/ml) during 10 days. Approximately 1 ϫ 10 7 parasites were harvested by centrifugation at 4°C for 10 min at 1500 ϫ g in 15-ml centrifuge tubes, washed twice in ice-cold SDM-79, and resuspended in 200 l of SDM-79 containing 0.5 mM MnCl 2 . Concanavalin A-FITC conjugate was added to a final concentration of 1.5 g/ml. After 1 h of incubation in the dark on ice, the cells were diluted with ice-cold PBS to a volume of 5 ml, pelleted, resuspended in a final volume of 2 ml (final concentration 5 ϫ 10 6 /ml PBS), and passed through a cellfilter cap prior to analysis by FACSCalibur (BD Biosciences). Data were analyzed using flow cytometry software FlowJo.
Author Contributions-P. G. designed, performed, and analyzed the experiments shown in Figs. 1C and 3-6 and prepared all figures. A. G. S. generated RFT1 addback-strains and performed the experiment shown in Fig. 1B. Y. C. L. and A. A. S. developed the protocol for in vitro galactosylation of GPI precursors. A. A. S. designed the scheme in Fig. 1A and designed and performed mass spectrometry analyses presented in Fig. 2. P. B. conceived and coordinated the study, and P. B. and P. G. wrote the paper. A. A. S. and A. K. M. provided scientific and experimental advice and critically revised the paper.