Blocking variant surface glycoprotein synthesis alters endoplasmic reticulum exit sites/Golgi homeostasis in Trypanosoma brucei

The predominant secretory cargo of bloodstream form Trypanosoma brucei is variant surface glycoprotein (VSG), comprising ~10% total protein and forming a dense protective layer. Blocking VSG translation using Morpholino oligonucleotides triggered a precise pre‐cytokinesis arrest. We investigated the effect of blocking VSG synthesis on the secretory pathway. The number of Golgi decreased, particularly in post‐mitotic cells, from 3.5 ± 0.6 to 2.0 ± 0.04 per cell. Similarly, the number of endoplasmic reticulum exit sites (ERES) in post‐mitotic cells dropped from 3.9 ± 0.6 to 2.7 ± 0.1 eight hours after blocking VSG synthesis. The secretory pathway was still functional in these stalled cells, as monitored using Cathepsin L. Rates of phospholipid and glycosylphosphatidylinositol‐anchor biosynthesis remained relatively unaffected, except for the level of sphingomyelin which increased. However, both endoplasmic reticulum and Golgi morphology became distorted, with the Golgi cisternae becoming significantly dilated, particularly at the trans‐face. Membrane accumulation in these structures is possibly caused by reduced budding of nascent vesicles due to the drastic reduction in the total amount of secretory cargo, that is, VSG. These data argue that the total flux of secretory cargo impacts upon the biogenesis and maintenance of secretory structures and organelles in T. brucei, including the ERES and Golgi.

cells there are between 25 and 600 Golgi stacks per cell depending upon plant cell type. 10 In the baker's yeast Saccharomyces cerevisiae the "classic" Golgi stack structure is absent, and there are about 20 dispersed cisternae per cell. 11 In contrast, in procyclic T. brucei present in the insect vector, there is normally only one Golgi stack per cell. 12 These cells are coated in procyclin protein, which is less abundant than VSG, comprising about 1% total protein. 13 In BF T. brucei there are an increased number of Golgi bodies, with typically about 60% to 70% of cells in G1 with 2 Golgi bodies, rising to at least 4 Golgi in more than 60% of cells that have completed mitosis, but not yet initiated a cleavage furrow. 8,14 The increased number of Golgi bodies in bloodstream compared with procyclic form T. brucei could possibly be an adaptation allowing the cell to accommodate the vast amounts of VSG (10-fold higher than procyclin) that need to be processed through the secretory pathway to the cell surface in this life cycle stage. Hence, this small number of very active Golgi bodies per cell makes T. brucei an amenable system to investigate Golgi biogenesis.
Different models have been proposed for Golgi biogenesis, either relying upon de novo synthesis, or duplication of a pre-existing Golgi template. 15 In T. brucei a hybrid of the two models appears to exist, whereby the new Golgi is formed de novo at a fixed distance from the old Golgi, 12 but requiring components from the old Golgi. 16 However, in addition to these two models, studies in yeast have postulated that continuing vesicular traffic is essential for maintaining correct Golgi structure. 15,17 How the amount of secretory traffic affects Golgi maintenance and structure in T. brucei is unclear. We, therefore, attempted to investigate this by blocking synthesis of the major secretory cargo of the BF trypanosome, VSG.
VSG is central to T. brucei pathogenicity. 18,19 Surprisingly, VSG is essential even in vitro in BF T. brucei, and blocking its synthesis with RNAi results in cells arresting at a very precise cell-cycle stage, immediately before cell division. 20 As there is no re-initiation of S-phase in these stalled cells, this is compatible with VSG synthesis being monitored during the cell cycle, and a precise cell cycle checkpoint triggered in the absence of its synthesis. 21 However it is unclear what aspect of VSG (ie, RNA or protein) is being "sensed" to trigger this cell cycle checkpoint.
In order to obtain insight into this "sensing" mechanism, we attempted to block VSG synthesis at the level of translation, while leaving VSG mRNA intact. In order to investigate the connection between amount of secretory cargo and homeostasis of secretory organelles in BF T. brucei, we emptied the secretory pathway of its predominant cargo (VSG) in an inducible fashion. We characterised the secretory structures and organelles in cells where VSG synthesis had been inducibly blocked. We show that blocking VSG synthesis at the level of translation produces an equivalent arrest to that derived after ablating VSG mRNA. This argues that VSG synthesis or transport is being "sensed" rather than VSG transcript. In addition, stopping VSG synthesis leads to a reduction in the number of endoplasmic reticulum exit sites (ERES) and Golgi. Both the endoplasmic reticulum (ER) and Golgi showed distortions in their morphology, with distended cisternae particularly at the trans-face of the Golgi. This specific membrane accumulation argues that post-Golgi secretory vesicles only bud off from these structures if there is adequate cargo to fill them. These results support a model that there is a direct connection between amount of secretory cargo and secretory organelle homeostasis in T. brucei.

| Morpholino mediated block of VSG mRNA translation triggers a pre-cytokinesis arrest
We have previously shown that the RNAi mediated ablation of VSG mRNA in BF T. brucei triggers a specific pre-cytokinesis cell cycle arrest, which is characterised by the absence of re-initiation of S phase in the stalled cells. 20 In order to determine if this unique cell cycle checkpoint was sensed at the protein or the RNA level, we attempted to block VSG synthesis by preventing its translation without affecting VSG mRNA levels. We investigated this using Morpholino antisense oligonucleotides (Morpholinos) designed to bind the region downstream of and including the start codon of VSG221 mRNA, with Morpholinos targeting α-tubulin mRNA serving as a positive control (Figure S1A, Supporting Information). As a negative control, anti-VSG and anti-tubulin Morpholinos were designed, each including 5 mismatched nucleotides which would disrupt Morpholino binding to the target mRNA ( Figure S1A). BF T. brucei HNI (221+) cells expressing VSG221 were transfected with both anti-VSG and anti-tubulin Morpholinos. 22 As expected for a perturbation disrupting VSG synthesis, transfection of anti-VSG221 Morpholinos resulted in T. brucei arresting in cell growth immediately after the transfection, comparable to as observed after the induction of VSG221 RNAi in T. brucei VB1.1 ( Figure 1A). 20 Transfection of cells with mismatched anti-VSG221 Morpholinos did not trigger this growth arrest. As a control, disruption of tubulin translation also resulted in the expected perturbation of cell growth. The cell cycle arrest phenotype triggered by transfection with Morpholinos was transient, with cells recovering 21 hours after electroporation with anti-VSG221 Morpholinos and after 36 hours for cells transfected with anti-tubulin Morpholinos ( Figure 1A). In contrast, the cell cycle arrest induced by VSG221 RNAi was relatively more stable.
Trypanosoma brucei transfected with the anti-tubulin Morpholinos become spherical and were multi-nucleated after 12 hours, due to re-initiation of S-phase in the stalled cells ( Figure 1B). This is consistent with the FAT phenotype generated after blocking tubulin synthesis using anti-tubulin RNAi. 23  Transfection of T. brucei with anti-tubulin Morpholinos resulted in the majority of the cells (55%) stalling before cell division as multinucleated cells that had undergone re-initiation of S-phase ("others" FIGURE 1 Blocking translation of VSG mRNA using antisense Morpholino oligonucleotides triggers a specific pre-cytokinesis cell cycle arrest. A, Perturbation of VSG synthesis with Morpholinos results in a growth arrest. Growth curves of Trypanosoma brucei after electroporation with antitubulin (tub) or anti-VSG221 Morpholinos with water (H 2 O) shown as a negative control. Morpholinos were designed against either the WT sequences or had mismatches (mis) introduced to disrupt base-pairing with mRNA. In comparison, T. brucei VB1.1 in the presence (+) or absence (−) of tetracycline to induce VSG221 RNAi for the time indicated in hours (h) is shown. Error bars indicate the SD for biological replicates (n = 4 for anti-tubulin Morpholinos and H 2 O, n = 3 for anti-VSG Morpholinos and n = 2 for VSG221 RNAi). The scale bar indicates 5 μM. B, Microscopy analysis of T. brucei 12 hours after electroporation of anti-tubulin Morpholinos (αTub-Morph) or anti-VSG221 Morpholinos (αVSG221-Morph) with no Morpholinos (water) shown as a control. Trypanosoma brucei VB1.1 cells with VSG221 RNAi induced for 12 hours (h) are included for comparison. Images of cells visualised with differential interference contrast (DIC) or with DNA stained with DAPI (blue) are shown. C, Cell cycle analysis shows that blocking VSG synthesis with Morpholinos is comparable to induction of VSG RNAi. Cells were karyotyped 12 hours after electroporation of anti-tubulin (αTub) or anti VSG221 (αVSG221) with water (No Morph) serving as a negative control. Anti-sense Morpholinos were either against the WT sequence or mis-matched (mis). The percentage (%) of cells at the 1K1N, 2K1N or 2K2N stage of the cell cycle is shown, with "others," including multi-nucleated cells which had reinitiated S-phase. In comparison, T. brucei VB1.1 before or after the induction of VSG221 RNAi for 12 hours (h) is shown. Error bars indicate the SD from multiple biological replicates (n = 4 for anti-tubulin Morpholinos and the no Morpholino control, n = 3 for anti-VSG221 Morpholinos and n = 2 for T. brucei VB1.1). (n > 300 cells were counted for each treatment in each biological replicate). D, Electroporation of anti-VSG221 Morpholino oligonucleotides does not lead to reduction in VSG221 mRNA. Northern blot analysis shows VSG221 transcript 12 hours after cells were transfected with either antitubulin (αTub) or anti-VSG221 (αVSG) antisense Morpholinos. In parallel, water (H 2 O) was electroporated as a negative control. RNA from T. brucei VB1.1 after the induction of VSG221 RNAi for 12 hours is shown in comparison. An image of the gel stained with ethidium bromide (EtBr) is shown to indicate loading in Figure 1C). In contrast, although the induction of VSG RNAi also prevented cell division, there was no re-initiation of S phase, resulting in an accumulation of 2K2N cells to 67% of the population ( Figure 1C). 20 Essentially equivalent results (52% 2K2N) were obtained after transfection with anti-VSG221 Morpholinos. Northern blot analysis showed that although induction of VSG221 RNAi in T. brucei VG1.1 resulted in the ablation of VSG221 mRNA to nearly undetectable levels after 12 hours, transfection of the various Morpholinos, as expected did not affect the level of VSG221 transcript ( Figure 1D). It therefore appears that with regards to this checkpoint response, it is levels of VSG protein, VSG protein synthesis or VSG transport that are "sensed" during trypanosome progression through the cell cycle, rather than the VSG mRNA itself.

| Blocking synthesis of VSG results in a reduced number of ER exit sites and Golgi
As the most abundant protein in BF T. brucei, VSG makes up the vast majority of the total secretory cargo. Along with other secretory cargo, including various surface receptors, VSG is recruited to COPII vesicles at the ERES characterised by TbSec23/ TbSec24 heterodimers, 24 and is subsequently transported through the Golgi cisternae to eventually be transported to the cell surface by secretory vesicles. 25,26 Each BF trypanosome typically has two ERES:Golgi junctions, which are replicated as the cell progresses through the cell cycle. 14 This culminates in cells with normally about four ERES and Golgi immediately after mitosis. 8,14 We, therefore, investigated if blocking VSG synthesis, and therefore, emptying the secretory pathway of a significant amount of its cargo, affected the total number of ER exit sites and Golgi.
We investigated this in BF T. brucei SL221 TbSec24.1::Ty1 where the ER exit sites were visualised using Ty1 epitope tagged TbSec24.1, 24 while the Golgi were visualised with an antibody against the GRASP marker for the mid-Golgi. 12 The localisation of TbSec23.2 to the ERES has previously been demonstrated to a high degree of resolution using immunogold electron microscopy. 24 In order to confirm that TbSec24.1 indeed colocalised with TbSec23.2, we epitope As expected, immunofluorescence microscopy analysis showed that the ERES and Golgi were located immediately adjacent to each other as non-overlapping spots within the cell, 12,14,24 and this subcellular location was not disrupted after blocking VSG synthesis for 8 hours (Figure 2). This allowed quantitation of the number of ERES and Golgi after the induction of a VSG synthesis block.
In cells where VSG synthesis had been blocked for 8 to 12 hours, there was a significant reduction in the number of ERES ( Figure 3). In T. brucei, the kinetoplast (K), that is, the mitochondrial DNA, replicates before the nuclear (N) DNA, meaning that these can be used to follow trypanosome progression through the cell cycle. 27 The mean  The number of Golgi is reduced in cells arrested after the induction of VSG221 RNAi. A, The number of Golgi were quantitated using an antibody against the GRASP Golgi marker in Trypanosoma brucei SL221TbSec24.1::Ty1 cells which were also co-stained with a Ty1 antibody to visualise the TbSec24.1 marker for the ERES. Only Golgi foci associated with TbSec24.1 were counted to avoid counting of "phantom Golgi". 14 VSG221 RNAi was induced for the time indicated in hours (h). Data are grouped according to cell cycle stage (1K1N, 2K1N, 2K2N). Error bars show the SD from 3 biological replicates (n~500 cells per time point per biological replicate). The significance of the difference with uninduced cells (*P < .05) was determined using paired t tests. B, Data in panel (A) presented as the percentage of cells in each cell cycle stage with a given number of Golgi foci. Error bars show the SD from 3 biological replicates (n~500 cells per time point per replicate). Significance of the difference (*P < .05, **P < .01, ***P < .001) of the change in cells with a given number of foci was determined using one-way ANOVA followed by Tukey post-hoc comparing the different points of VSG RNAi within each cell cycle stage cells with 2 foci (8.4% AE 4.7% to 38.1% AE 3.3% at 8 hours). These results are all compatible with the hypothesis that the number of ERES foci within the cell is subject to requirement, and is dictated by the amount of secretory cargo. In wild type (WT) BF T. brucei, less than 1% of the cells had ERES, but no Golgi. The percentage of cells without Golgi increased after induction of a VSG synthesis block, with this increase becoming statistically significant 12 hours after induction (P < .05). In 1K1N cells the increase was from 0.5% AE 0.3% to 4.8% AE 2.0% Golgi per cell, in 2K1N cells from undetectable to 9.2% AE 3.3% and in 2K2N cells from undetectable to 6.6% AE 2.6%. GRASP is one of the first components recruited during assembly of the Golgi in T. brucei, 28 and cells without GRASP signal are unlikely to have assembled a new Golgi body. Assembly of Golgi bodies de novo in yeast is dependent on COPII export from the ER, which nucleates the formation of a new Golgi body through fusion with vesicles originating from preexisting Golgi. 29 Given the asynchronous nature of cells where VSG RNAi has been induced, the small proportion of cells where Golgi were not visible could be daughters of cells with reduced Golgi that did not inherit a Golgi body after cytokinesis.

| Distortion of the ER and dilation of Golgi cisternae in the presence of a VSG synthesis block
Staining BF T. brucei with antibodies against BiP (binding immunoglobulin protein) allows for the visualisation of the ER, which forms a continuous network throughout the cell. 30,31 In cells where VSG synthesis had been blocked, the organisation of this ER network was disturbed, and the ER appeared distorted irrespective of cell cycle stage ( Figure 5). This could be a consequence of decreased vesicular traffic FIGURE 5 Blocking VSG synthesis results in distortion of the ER. Representative immunofluorescence microscopy images of Trypanosoma brucei VB1.1 cells either before or after the induction of VSG221 RNAi for 8 hours (h). Trypanosomes are visualised with differential interference contrast (DIC) and with the DNA stained with DAPI, or the ER visualised using an anti-BiP antibody. Representative cells at different stages of the cell cycle (1K1N,  2K1N and 2K2N) are shown at equivalent exposure times. Scale bar is 5 μM out of the ER. In other experimental systems, structural distortion of the ER can be a consequence of ER stress, which typically occurs after the accumulation of misfolded protein. [32][33][34] The Golgi in BF T. brucei appears as a stack of 5 to 6 flattened cisternae ( Figure 6A; Figure S3A). However, after the induction of VSG221 RNAi for 8 hours, gross distension can be observed in the last 2 or 3 cisternae of the trans-Golgi ( Figure 6A,B). The proportion of Golgi stacks with dilated cisternae increases to over 80% in a time dependent manner over a period of 8 to 24 hours ( Figure 6C). There was no evidence for a significant increase in cisternae per stack ( Figure S3B). VSG constitutes the bulk of secreted protein in the BF trypanosome, so blocking its synthesis would be expected to radically  35,36 Appearance of mature TbCatL, therefore, provides a kinetic measure of forward trafficking from the ER. VSG221 RNAi was induced for an 8 hours window, during which period VSG synthesis was expected to fall dramatically, while synthesis of other proteins would remain relatively unaffected. 21 Aliquots of cells were removed at 2-hour intervals and subjected to pulse radiolabelling. This was followed by specific immunoprecipitation of VSG221, cytoplasmic Hsp70 and TbCatL polypeptides from cell extracts ( Figure 7B)

| Lipid biosynthesis is largely unaffected in cells arrested after induction of VSG RNAi
As we saw distortion of the ER and dilation of the trans-cisternae of the Golgi after the induction of a VSG synthesis block, we also inves-   Figure S4A). However, the [ 3 H]-myristate incorporation into the protein pool does decrease over this period ( Figure S4A). As expected, this is primarily due to the reduction in biosynthesis of newly formed VSG protein in the ER, and subsequent GPI-anchor addition via a transamidase reaction. 38 Nevertheless, myristate incorporation into protein is not completely abolished after 24 hours induction of VSG RNAi, as there are several other less abundant GPI-anchored proteins such as transferrin receptor that will likely still undergo GPI addition, and direct myristylation of cytosolic proteins will still be ongoing. 39 In addition, pre-existing cell surface VSG will still undergo endocytosis, early endosomal sorting, and post-translational myristate-exchange prior to being trafficked back to the cell surface. 40 Uptake and utilisation of [ 3 H]mannose was next investigated.
Mannose can be incorporated into glycoproteins primarily via N-glycosylation, but also in GPI-anchors, both of which are important for the formation of mature VSG. It is, therefore, unsurprising that incorporation of [ 3 H]mannose into protein decreased significantly, and in keeping with the decreased rate of VSG synthesis ( Figure S4B), as well as decreased protein synthesis in general. 21    This lipidomic analysis suggests that despite lipid biosynthesis in general not being altered significantly during VSG RNAi, there is a disproportionate increase in SM levels, suggesting an accumulation, most likely due to a lack of catabolism as the rate of synthesis does not seem to increase ( Figure 8A). We, therefore, show here that the cell-cycle arrest induced by a block in VSG synthesis has very striking characteristics. Cells do not grow, but are metabolically active with no major changes in lipid or GPI-anchor biosynthesis, but certain phospholipids (ie, SM) accumulate, presumably as a result of not being turned over (catabolised).

| DISCUSSION
In summary, we show that blocking VSG synthesis at the level of its translation produces an equivalent cell cycle arrest to that derived after ablating VSG transcript. Blocking synthesis of the major secretory protein of BF African trypanosomes also results in striking changes to both the number and morphology of secretory structures and organelles. There is a reduction in number of both ERES and Golgi bodies. This is compatible with the hypothesis that the maintenance of these structures and organelles is linked to the amount of secretory cargo that passes through them. Induction of a VSG synthesis block also resulted in distorted ER and Golgi morphology, with an apparent increasing accumulation of membrane, particularly in the cisternae at the trans-face of the Golgi. This could argue that secretory vesicles only bud off from the Golgi if there is sufficient cargo to fill them. However, despite the severe changes in morphology of secretory organelles, the secretory pathway of the stalled cells still appeared to be functional, as transport of endogenous TbCatL to the lysosome was unaffected. Furthermore, there were no observable gross changes in lipid biosynthesis.
Although blocking VSG synthesis triggers a specific precytokinesis checkpoint, it is unknown what aspect of VSG is "sensed" during trypanosome progression through the cell cycle. We, therefore, investigated here if VSG mRNA depletion, or alternatively a reduction in total amount of VSG protein or protein synthesis was key in triggering this checkpoint. We show that blocking VSG synthesis at the level of its translation, while leaving levels of VSG mRNA unaffected, produced an equivalent cell cycle arrest to that induced by the ablation of VSG mRNA. This eliminates mRNA as the target being "sensed" in this pathway, and argues that it is the total amount of VSG protein or its synthesis or trafficking to the cell surface that is being monitored to trigger this intriguing cell cycle arrest.
After inducing VSG RNAi for only 8 hours, there was an observed reduction in the number of both ERES and Golgi bodies.
There is likely to be an interplay between the amount of cargo and ERES. Overexpression of secretory cargo has been shown to result in more ERES in other experimental systems, as in B cells stimulated to produce large amounts of secreted immunoglobulins, there is an almost 4-fold increase in the number of ERES. 42 The ERES are juxtaposed to the Golgi in T. brucei, 3,43 and are duplicated at the same FIGURE 9 Lipidomic analysis of Trypanosoma brucei in the presence or absence of VSG RNAi for 16 hours. Survey ESI-MS in positive ion mode (600-1000 m/z) of lipid extracts from T. brucei VG1.1 that had been cultured in the presence or absence of VSG RNAi for 0 or 16 hours (h). The red arrows indicate the species which increase significantly upon induction of VSG RNAi, which were confirmed as being choline-phosphate containing SM or PC species ( Figure S3C and Table S1) time during T. brucei cell division. 12,14,28 After blocking VSG synthesis, in addition to a reduction in the number of ERES, we also saw a significant decrease in the number of Golgi bodies.
In a model supported by plant experimental systems, it has been proposed that both ERES and Golgi biogenesis is linked to the amount of bulk secretory traffic in the cell. 44,45 Here, it was proposed that Golgi bodies can be generated directly from ERES, and in this scenario the number of Golgi in the cell is directly affected by the number of ERES on the ER surface. This would allow the cell to dynamically respond to variations in the synthesis of different secretory proteins, allowing it to modify its subcellular secretory machinery subject to requirement. Our data is compatible with this model.
Although during T. brucei cell division the new Golgi does not appear to be an exclusive product of the ERES, 12,28 there could still be feedback between the two subcellular structures. The Golgi is clearly a very dynamic structure, and its structure and function has been argued to rely on the integrity of export from the ER. 46 Similarly, when mammalian cells were depleted of Golgi bodies using laser nanosurgery, there was a reduction in the number of ERES and reduced ER export. 47 This argues for an interplay between the two structures. Although Golgi biogenesis in T. brucei is likely to involve both de novo as well as templated organellar biogenesis, as has been argued to be the case in plants, 12,16,48 our results suggest that the total amount of secretory cargo also plays an additional modulatory role on Golgi number, which has also been postulated to be the case in plants. 44 Our data are, therefore, compatible with a model whereby ERES and Golgi biogenesis and maintenance are stimulated by the amount of secretory cargo, which could give us additional insight into Golgi biology in T. brucei. 15 When the secretory pathway is emptied of VSG through the induction of a synthesis block, the stalled cells had Golgi with an unchanged number of cisternae per stack. This is compatible with a proposed "stable cisternae" model, whereby each cisterna contains a stable complement of Golgi enzymes. 49 However, the stalled cells showed distortion of both the ER and the Golgi. The size of the ER can be controlled subject to requirement in different types of eukaryotic cells, 50 and both the ER and Golgi can expand enormously in B cells producing vast amounts of secreted immunoglobulins. 42,51 Distortion of the ER lumen can also be produced by ER stress. [32][33][34] This can occur when the unfolded protein response (UPR) is triggered, and the ER is clogged with either misfolded or overexpressed protein. [52][53][54] For example, in mice where a genetic perturbation results in misfolded major histocompatibility complex Class I accumulating in the ER, the ER-Golgi compartment becomes an expanded and distorted tubular network. 55 Similarly, disruption of disulphide bond formation of mouse proinsulin leads to a 2.9-fold increase in dilated ER cisternae and a 4.5 to 5.8-fold increase in Golgi associated and pre-Golgi intermediate structures. 56 We do not know if the distorted ER in our arrested cells is a direct response to the drastic reduction in the amount of secretory cargo trafficking through this organelle, or the consequence of a novel ER stress response resulting from the relative emptying of its predominant cargo. ER stress appears to be sensed differently in T. brucei compared with other eukaryotes. In BF T. brucei chemical induction of ER stress, or disruption of the ER translocon by RNAi does not result in induction of the expected response to ER stress, which is the UPR. 57,58 In procyclic form T. brucei persistent ER stress leads to a spliced leader RNA silencing pathway (SLS) and programmed cell death. 59 We have no evidence that this occurs after blocking VSG synthesis in BF T. brucei. 21 After blocking VSG synthesis we see grossly distended cisternae in the Golgi bodies, and particularly at the trans-face. We think that it is most likely that the drastic reduction in the amount of secretory cargo passing through this compartment is directly leading to a reduction in the number of secretory vesicles leaving the trans-cisternae, as they do not bud off in the absence of VSG cargo to fill them. This could result in membrane accumulation in the distal cisternae. In

| Northern blot analysis
For Northern blot analyses total isolated RNA (1-1.5 μg per sample) was mixed with 10 mM MOPS, pH 7.0; 1 mM NaOAc; 0.5 mM EDTA 50% (v/v) formamide; 17.5% (v/v) formaldehyde; 60 μg/mL −1 ethidium bromide and denatured for 15 minutes at 65 C. The RNA was resolved on 1.5% formaldehyde-agarose gel in MOPS buffer. RNA was transferred onto Hybond-XL membrane (GE Healthcare) using standard protocols. Blots were hybridised with random primed probes that had been radiolabeled with [ 32 P]-dCTP using Amersham Ready-To-Go DNA labelling Beads (-dCTP) (GE Healthcare) following the instructions of the manufacturer. An 803-bp DNA probe targeting VSG221 was PCR amplified from T. brucei 90-13 genomic DNA using primers from. 20 Membranes were washed to an end stringency of 0.1 × SSC at 65 C. The blots were imaged with a Personal Molecular Imager FX (Bio-Rad). Post-acquisition analyses were carried out with ImageJ (National Institutes of Health).

| Microscopy analysis
The T. brucei examined by thin section transmission electron microscopy (TEM) were processed as described previously, 20 and examined in a FEI Tecnai 12 transmission electron microscope.

| Trypanosoma brucei metabolic labelling and lipid and protein analysis
Pulse-chase radiolabelling of log-phase cultured BF T. brucei with [ 35 S]methionine/ cysteine [Perkin Elmer, Waltham, Massachusetts], and subsequent immunoprecipitation of labelled polypeptides was performed as described previously. 35,36 Briefly, cells were harvested, washed and transferred to labelling media at 10 7 mL −1 . After a 15 minute pre-incubation at 37 C, radiolabel was added to 200 μCi mL −1 and incubation was continued for an additional 15 minutes.
For the experiments involving lipid or protein analysis, midlogarithmic WT BF T. brucei or T. brucei VG1.1 cells 20 where VSG221 RNAi had been induced with tetracycline (1 μg/mL) were centrifuged (800 g for 10 minutes) and washed in minimal essential media (fattyacid, glucose or serine free), before suspension in the same media at 1 × 10 7 cells mL −1 . Total protein synthesis was inhibited in WT cells by pre-incubating for 10 minutes at 37 C with cycloheximide (60 μg/ mL). Cells were labelled for 1 hour at 37 C with 50 μCi mL −1 of [ 3  MgSO 4 , 100 mM Na 2 HPO 4 / NaH 2 PO 4 , 100 mM glucose) prior to samples taken for either lipid or protein analysis as previously described. 68 Lipids were extracted with organic solvents, dried and partitioned between butanol/ water. The desalted lipids (2 × 10 6 cell equivalents per lane) were separated by HPTLC using silica 60 HPTLC plates and a chloroform:methanol:water (10:10:3 v/v) solvent system. Radiolabeled lipids were detected by fluorography at −70 C, after spraying with En 3 hance and using Kodak XAR-5 film with an intensifying screen. GPI intermediates were identified using various enzyme and chemical digests as previously described. 69 For experiments involving radiolabel incorporation into total, pro-