Flagellum elongation is required for correct structure, orientation and function of the flagellar pocket in Trypanosoma brucei

African trypanosomes are protozoan parasites responsible for human tropical diseases, such as sleeping sickness. This neglected disease is fatal in the absence of treatment. No vaccine is available and current therapy relies on the use of toxic drugs, including arsenic components that lead to patient death in 2-10% of cases. Trypanosomes proliferate in the bloodstream of their host as extracellular parasites and escape the immune response by using a sophisticated process of antigenic variation. The surface of the trypanosome is covered with a dense coat made of a single type of protein (variant surface glycoprotein or VSG) that is periodically replaced. The genome contains ∼1000 different VSG genes and pseudogenes that can undergo recombination events to ensure rapid modification of this repertoire (Berriman et al., 2005).

Endocytosis and exocytosis exclusively take place at the flagellar pocket, a surface microdomain found at the posterior end of the cell. This restriction is due to the presence of a dense network of subpellicular microtubules, which prevents fusion-fission events at all other surface locations (Gull, 1999). The flagellar pocket is a depression of the plasma membrane from which the single flagellum emerges and represents barely 1% of the total cell surface (Overath and Engstler, 2004). Freeze-fracture studies demonstrated that membranes of the flagellar pocket and of the flagellum itself are closely opposed at the neck of the flagellar pocket (Yoshikawa et al., 1992). Nevertheless, antibodies that bind to the VSG coat are rapidly cleared by efficient endocytosis of VSG and bound antibodies, followed by recycling of the VSG proteins after antibody dissociation in endosomes (Engstler et al., 2004). The forward motility of the parasite in the bloodstream ensures efficient transfer of bound antibodies from over the cell surface to the posterior flagellar pocket via hydrodynamic flow (Engstler et al., 2007).

Little is known about flagellar pocket molecular identity and about its formation during the cell cycle. The new flagellum is constructed in the existing flagellar pocket, next to the mature flagellum (Absalon et al., 2008; Briggs et al., 2004; Davidge et al., 2006; Grasse, 1961). Although initial elongation of the new flagellum occurs within the same flagellar pocket, basal bodies migrate apart (Robinson et al., 1995) and two separate flagellar pockets are visible, each one with a single flagellum. A defined cytoskeletal structure called the flagellar pocket collar (FPC) is visible at the neck of the flagellar pocket and remains associated with the flagellum upon detergent extraction (Sherwin and Gull, 1989). A molecular component of the collar, termed BILBO1, has been identified recently and has been shown to be essential for both collar and flagellar pocket formation, but not for basal body duplication and flagellum growth (Bonhivers et al., 2008b). Absence of a flagellar pocket results in multiple cellular defects and is lethal in both procyclic and bloodstream trypanosomes.

These data raise the question of the coordination of these critical events of the trypanosome cell cycle: basal body duplication, flagellum construction and flagellar pocket formation. At least three hypotheses can be proposed: (1) the FPC duplicates and the new flagellum is then introduced in the new FPC; (2) the emergence of the new flagellum triggers the duplication of the FPC; or (3) both events take place and are coincidental. The availability of RNAi mutants, in which formation of the new flagellum can be halted (Absalon et al., 2008; Absalon et al., 2007; Davidge et al., 2006; Kohl et al., 2003), opens up the ability to evaluate the role of the flagellum in flagellar pocket duplication. We used procyclic trypanosome cell lines where expression of genes required for intraflagellar transport (IFT) had been silenced. IFT is a dynamic process that is required to construct flagella (Cole et al., 1998; Kozminski et al., 1993) and is conserved in almost all flagellated eukaryotes, from protists to humans (Kohl and Bastin, 2005). Our analysis demonstrates that when the new flagellum is too short or absent, a flagellar pocket is still made, showing that its formation is independent of flagellum growth. However, its structure, its orientation in the cell body and its functions are modified. We propose a model explaining the different steps of flagellum and flagellar pocket formation in trypanosomes.

As basal body duplication plays a major role in the trypanosome cell cycle (Robinson and Gull, 1991), it is important to first evaluate whether the basal body duplicates in the absence of a flagellum. Uniflagellated trypanosomes possess a mature basal body that is continuous with the flagellum and a shorter immature pro-basal body (Sherwin and Gull, 1989). Two sets of IFT mutants created by RNAi were used for this analysis. IFT88 and IFT172 (members of IFT complex B) are required for anterograde transport (from base to tip) and their silencing blocks flagellum assembly (Absalon et al., 2008; Kohl et al., 2003). IFT140 (IFT complex A) and DHC1b genes encode proteins that are essential for retrograde IFT (transport of material from tip to base of the flagellum). Their silencing initially leads to the formation of a short dilated flagellum that is filled with IFT particles as a result of a failure in recycling IFT material (Absalon et al., 2008). This results in a situation where cells cannot build a new flagellum or where a tiny abnormal new flagellum is formed but not elongated. In both cases, membrane extension still takes place, producing a flagellar `sleeve', which does not contain cytoskeletal elements (Absalon et al., 2008; Davidge et al., 2006).

Two markers were used to examine basal body duplication. The monoclonal antibody mAb22 detects an as yet unidentified antigen localised at the region of the mature basal body and the pro-basal body within a structure called the tripartite attachment complex, which links the kinetoplast to the basal body (Bonhivers et al., 2008a). We also used a rabbit antiserum recognising the trypanosome basal body component protein (TBBC), a 164 kDa coiled-coil-rich protein (Dilbeck et al., 1999). Previous data indicate that the anti-TBBC antibody labels only one spot (Absalon et al., 2007; Dilbeck et al., 1999), suggesting that it could be a differential marker of the mature basal body or of the pro-basal body. Control cells were labelled with mAb22, anti-TBBC and DAPI to reveal the nucleus and the kinetoplast, which is always linked to the basal body (Robinson and Gull, 1991), thus providing a marker of basal body position. Cells with one flagellum exhibited two spots for mAb22, one corresponding to the fibres of the mature basal body (continuous with the flagellum, which is clearly visible by phase contrast imaging) and one corresponding to the fibres of the immature pro-basal body (Fig. 1Aa). Both spots were found in close proximity to the kinetoplast, in agreement with mAb22 localisation at the proximal region (Absalon et al., 2007; Bonhivers et al., 2008a). By contrast, the same cells displayed only a single spot for TBBC, localised at the distal region of the basal body and in continuity with the axis of the flagellum (Fig. 1Aa). When cells duplicated their basal bodies, four spots were stained with mAb22 and only two with the anti-TBBC antiserum (Fig. 1Ab). This pattern was highly reproducible and demonstrates that TBBC can be used as a marker of the older basal body that is associated with the flagellum.

Double staining experiments were performed on cell lines where expression of genes encoding either a complex A protein (IFT140) (Fig. 1B,C) or a complex B protein (IFT88 or IFT172) (supplementary material Fig. S1) had been silenced by RNAi. Inhibition of IFT blocks formation of the new flagellum but does not affect the existing flagellum (Absalon et al., 2008; Absalon et al., 2007; Davidge et al., 2006; Kohl et al., 2003). When formation of the new flagellum was inhibited, two basal body complexes could be detected: one associated with the old flagellum and one close to the posterior kinetoplast (Fig. 1B). Both displayed two spots for mAb22 and one for TBBC, exactly as in control cells, demonstrating that replication of the basal body took place normally (Fig. 1B). Once such a cell divides, it produces two daughters: one with a flagellum and one without a flagellum. Basal body duplication goes on in cells without a flagellum, as shown by the presence of a second basal body complex, stained with mAb22 and the anti-TBCC (Fig. 1Ca-a′). These cells fail to divide but can complete S phase and undergo mitosis, thus becoming multinucleated (Kohl et al., 2003). They are also able to duplicate their basal bodies and acquire TBBC normally, as shown by the presence of two spots for mAb22 always associated with a single spot for TBBC (Fig. 1Cb-b′). However, basal body migration is severely limited, as previously reported (Absalon et al., 2007).

Electron microscopy analysis of induced IFT RNAi cells confirmed the presence of multiple basal bodies with triplet microtubules (at the proximal end) and typical extensions (Fig. 1D). However, the transition region of basal bodies appears shorter in the absence of a flagellum (203±78 nm in IFT88^RNAi cells compared with 377±37 nm in wild-type controls, n=18), suggesting that elongation of the new basal body is not complete. In conclusion, IFT is only required for flagellum construction and possibly basal body elongation, but not for basal body duplication and acquisition of the TBBC protein.

As basal bodies duplicate, we asked whether a flagellar pocket could be formed in these conditions. The presence and structure of the flagellar pocket was investigated by transmission electron microscopy (Fig. 2). In wild-type cells, the basal body of the flagellum is found underneath the base of the flagellar pocket, with its proximal part made of triplet microtubules found within the cell body (Fig. 2A). The transition region is made of nine doublets of microtubules surrounded by the flagellar membrane and is found in the lumen of the flagellar pocket. When the flagellum exits the flagellar pocket, it now contains a complete axoneme made of nine doublets of microtubules with dynein arms and radial spokes, and a central pair of microtubules. A large extra-axonemal structure called the PFR is only assembled after the flagellum exits the flagellar pocket (Bastin et al., 1996). The flagellum membrane is in close contact with that of the collar of the flagellar pocket and specific cytoskeletal structures made of electron-dense material are visible (Fig. 2A, arrows). On electron micrographs, the flagellar pocket exhibits a typical flask shape and the flagellum is positioned asymmetrically, defining a larger and a thinner lumen side found at the anterior and posterior ends, respectively. Vesicles corresponding to endosomes or exosomes are frequently seen in the cytoplasm in close proximity to the flagellar pocket.

The ultrastructure of the flagellar pocket was analysed in IFT88^RNAi, IFT172^RNAi, IFT140^RNAi and DHC1b^RNAi cells at induction times where assembly of the new flagellum was inhibited (Fig. 2B-E). The flagellar pocket was visible in all induced samples but its shape was considerably modified, with variable profiles. Cells that possessed only a short basal body often presented an abnormally small, but recognisable flagellar pocket, without a clear asymmetry (Fig. 2B,C). The structure of the flagellar pocket was also investigated in IFT140^RNAi and DHC1b^RNAi mutant cells that still possessed a short flagellum where microtubules elongated beyond the basal body and often presented excessive amounts of IFT-like material (Fig. 2D, arrowheads). These exhibited flagellar pocket of various shapes: either deeper than usual (Fig. 2D) or distorted (Fig. 2E).

Vesicular material was often present in the lumen of the flagellar pocket (Fig. 2D-E), a feature rarely observed in procyclic wild-type trypanosomes. Moreover, some vesicles were detected within the short flagella of retrograde transport mutants (Fig. 2D, see below). Quantitative analysis performed on a total of 146 images from all four induced mutants revealed that 50-80% of the flagellar pocket sections presented an abnormal profile (Fig. 2G). These values correspond to the actual percentage of non-flagellated cells in the culture at the induction time where the analysis was performed (data not shown). Accordingly, the lowest number of aberrant flagellar pockets was found in the IFT140^RNAi induced sample, in agreement with the lower proportion of non-flagellated cells in this population (50% compared with 75-80% for the other cell lines).

These multiple structural defects could be directly linked to failure in flagellum elongation but could also result from the morphogenetic defects observed in non-flagellated cells (Kohl et al., 2003). We therefore examined the status of the flagellar pocket in PF16^RNAi and in PF20^RNAi mutants where the expression of central pair proteins is inhibited, leading to cell paralysis. These mutants share the same morphogenetic defects (reduced basal body migration, short cell body, loss of polarity, poor elongation of the FAZ filament) but assemble a flagellum of normal length (Absalon et al., 2007). Transmission electron microscopy revealed that the flagellar pocket of these two mutants was undistinguishable from wild-type cells (Fig. 2F and data not shown), supporting a direct role of flagellum elongation in flagellar pocket formation. These data indicate that the flagellar pocket is formed, but in the absence of correct flagellum elongation, it cannot adopt its normal shape and length.

The above data do not give a global view of the structure and positioning of the flagellar pocket. We extracted trypanosomes with detergent to remove the plasma membrane and examined the resulting cytoskeletons by scanning electron microscopy. Stripping the membrane of wild-type cells exposed the subpellicular microtubules with their typical helical pattern, the flagellum and the flagellum attachment zone filament (FAZ) (Fig. 3A). This treatment revealed the FPC, which appeared as a horseshoe structure around the flagellum (Fig. 3A). However, it could also be that the FPC is a ring with a portion hidden by the flagellum. In any case, the posterior part of this structure always appeared to be lifted on top of the flagellum and in a perpendicular orientation relative to the main axis of the cell (Fig. 3A). However, this localisation could be a consequence of the detergent treatment. We therefore examined untreated cells using scanning electron microscopy; strikingly, the shape of the FPC could be recognised underneath the plasma membrane at the point of emergence of the flagellum from the flagellar pocket (Fig. 3B). Its posterior part appeared as an arch positioned above the flagellum at its point of emergence from the flagellar pocket and was found in an orientation very similar to that observed in detergent-extracted samples. These data demonstrate the defined orientation of the FPC and validates the detergent treatment methodology for investigation.

Analysis of cells from the IFT140^RNAi mutant with an abnormally short flagellum revealed the presence of a FPC structure around the short flagella (Fig. 3Ca-c). Two main differences were noted compared with controls: (1) the shape of the FPC was more irregular and its thickness was frequently reduced, and (2) its orientation relative to the main axis of the cell was extremely variable (Fig. 3Ca-c). The defect in FPC orientation was also visible on non-extracted cells (Fig. 3D), demonstrating that it was not due to detergent treatment. Non-flagellated cells could not be analysed by these methods because it was not possible to identify unambiguously the FPC in the absence of a flagellum (not shown).

The mis-orientation of the FPC in cells with a short flagellum could be a secondary consequence of the loss in cell polarity typical of non-flagellated cells (Kohl et al., 2003). Alternatively, it could be incorrectly positioned as soon as the new basal body is unable to assemble a flagellum, for example, as a result of poor basal body migration (Absalon et al., 2007). To evaluate the specificity of this mis-orientation phenotype, we analysed the positioning of the FPC at early stages of RNAi knockdown in intact cells or in detergent-extracted cytoskeletons (Fig. 4; supplementary material Fig. S2). In control cells (wild-type or non-induced samples), as the new basal body is duplicated, the new flagellum is assembled in close proximity to the existing flagellum. An FPC is seen at the base of each flagellum (Fig. 4A; supplementary material Fig. S2A), indicating that this structure duplicates rapidly upon formation of the new flagellum. In all the cells examined (n>100), the orientation of the FPC associated with the new flagellum was parallel to that of the FPC of the old flagellum and perpendicular to the main cell axis (Fig. 4A; supplementary material Fig. S2A). Next, we analysed cells at early stages of IFT protein deprivation (between 48 and 72 hours after tetracycline addition) that still possessed an old flagellum but failed to assemble a normal-length new flagellum. In contrast to control cells, the defined relative orientation of old and new FPC was already lost in 33% of IFT140^RNAi mutant cells (n=30) at this stage (supplementary material Fig. S2B). These values rose to 67% in IFT88^RNAi mutant cells possessing an old flagellum of normal length and a short new flagellum (n=15). This higher proportion could be related to the fact that IFT88^RNAi cells produce extremely short flagella compared with IFT140^RNAi cells, which often fail to emerge from the flagellar pocket (Absalon et al., 2008). Similar defects in FPC positioning were observed in extracted (supplementary material Fig. S2B) and in intact cells (Fig. 4B). Previous work demonstrated that cells with a normal length old flagellum but with an abnormally short new flagellum exhibit a normal cytoskeleton (Kohl et al., 2003). Therefore, these results reveal that the defined orientation of the FPC is already lost before the emergence of the morphogenetic defects at the cellular level and is more likely to be related to failure in elongating the new flagellum.

We next examined whether the localisation of a molecular component was also affected. BILBO1 is a structural protein of the FPC that is essential for its formation (Bonhivers et al., 2008b). In control cells possessing a single flagellum, an antibody against BILBO1 identified a region found precisely around the point where the flagellum exits the main cell body, corresponding to the FPC (Fig. 5A, middle cell). Soon after exit of the new flagellum from the flagellar pocket, a second region of labelling was detected upon elongation of the new flagellum (Fig. 5A, left cell). Kinetoplast and basal body segregation then occurs and two signals were clearly visible, each associated with a flagellum (Fig. 5A, right cell). Knocking down IFT gene expression first led to the formation of shorter flagella (arrowheads), which possess an apparently normal amount and arrangement of the BILBO protein at the expected flagellar pocket position (Fig. 5B). However, once flagella were not visible, the BILBO1 signal was less defined and sometimes less intense (Fig. 5C). These data were confirmed upon double staining with the anti-axoneme marker mAb25 (supplementary material Fig. S2) and indicate that BILBO1 is still synthesised in the absence of a flagellum and localises to the expected flagellar pocket position; however, its localisation seems modified, a feature that could be related to the thinner aspect of the FPC in electron microscopy (Fig. 3).

We next asked whether membrane flagellar pocket proteins could still be routed properly in cells with a modified flagellar pocket. CRAM (cysteine-rich acidic membrane) protein is a flagellar pocket transmembrane protein that is predominantly expressed in procyclic trypanosomes (Lee et al., 1990). Induced and non-induced IFT^RNAi cells were analysed by IFA with anti-CRAM antibodies following paraformaldehyde fixation and permeabilisation with detergent (Fig. 6A-C). In control trypanosomes, the presence of CRAM was limited to the posterior region where the flagellar pocket is localised (Fig. 6A). By contrast, the CRAM signal was not restricted to the flagellar pocket area in both IFT88^RNAi and IFT140^RNAi induced cells (Fig. 6B,C). As CRAM is a transmembrane protein, this localisation could correspond to material present at the surface of the cell or/and within vesicles in the cytoplasm. To solve the issue, induced and non-induced cells were fixed and stained with the anti-CRAM antibodies without permeabilisation. Surface-bound antibodies were visualised with Protein AG coupled to gold particles, and samples were analysed by scanning electron microscopy (Fig. 6D-G). Control cells failed to reveal significant staining either on the cell body or on the flagellum surface (Fig. 6D), as expected from the restricted localisation of CRAM within the flagellar pocket (Yang et al., 2000), which is not accessible without detergent permeabilisation. Silencing of IFT140 produces cells with a short dilated flagellum, with excessive IFT material, and other cells without a flagellum (Absalon et al., 2008). Remarkably, in induced IFT140^RNAi cells, the short dilated flagella exhibited strong CRAM staining at their surface: on average 18 gold particles per μm² of flagellar surface (Fig. 6E-G), which corresponds to a density ∼30-fold higher than in normal flagella of non-induced samples (0.6 gold particles per μm² of flagellar surface) (Fig. 6D). The old flagellum of induced IFT140^RNAi cells maintained a low level of gold particles similarly to wild-type cells, indicating that CRAM localisation was only modified in the new flagellum (not shown). Moreover, the `flagellar sleeve' was also stained with the anti-CRAM antibody in cells with (Fig. 6E) or without (Fig. 6F) a flagellum. The frequency of gold particles on the rest of the cell surface remained very low, exactly as in control cells (Fig. 6D-F), revealing that the abundant staining detected in permeabilised cells by immunofluorescence (Fig. 6B,C) actually corresponds to CRAM protein present in intracellular compartments. In conclusion, a flagellar pocket membrane protein reaches the flagellum membrane in IFT140^RNAi cells and is found in the cytoplasm, revealing defects in the routing of flagellar pocket proteins.

The accumulation of CRAM in cytoplasmic compartments, as well as its presence on the membrane of short flagella indicates defects in surface protein trafficking. Furthermore, the presence of numerous vesicles in the lumen of the flagellar pocket (Fig. 2D-F) or in the short flagella (Fig. 2D) of IFT^RNAi cells indicates possible defects in vesicular trafficking. We therefore examined by transmission electron microscopy the presence, nature and distribution of cytoplasmic vesicles around the flagellar pocket area in DHC1b^RNAi, IFT140^RNAi and IFT88^RNAi induced cells and compared them with that in wild-type cells (Fig. 7).

Sections through the flagellar pocket region of wild-type trypanosomes displayed the usual organelles, including endosomes and exosomes, the Golgi complex, the flagellum or the kinetoplast (Fig. 7A-C). We noticed the frequent presence of a long and narrow cisternal compartment adjacent to the flagellar pocket membrane (Fig. 7B-C, white arrows), which was regularly found in flagellar pocket sections (60%, n=72). This compartment was more frequently seen at the anterior face of the flagellar pocket, close to the Golgi, and we named it the `flat compartment'. Although its nature and function are still unknown, it could represent a morphological marker of the flagellar pocket environment.

Numerous large and small vesicles are present in the vicinity of the flagellar pocket of the IFT^RNAi mutants. Furthermore, some unusual vesicles were observed close to the flagellar pocket and the Golgi complex. Some vesicles appeared to have two membranes (Fig. 7D,G,H, arrowheads) and others looked bent (Fig. 7D,G,H, arrows). Quantification was performed on a total of 150 sections from induced IFT88^RNAi, IFT140^RNAi and DHC1b^RNAi cells. These vesicles were encountered in almost 30% of the sections in anterograde and retrograde transport mutants (Fig. 7I). This is quite close to the frequency of flagellar pocket with vesicles in their lumen (27%, n=146 for four different IFT^RNAi mutants) (Fig. 2G).

As noted above, vesicles were found within short flagella in the IFT140^RNAi and the DHC1b^RNAi mutants that contained IFT-like material (Fig. 2D; Fig. 7E,F). In IFT88^RNAi and IFT172^RNAi cells, a few sections (7/51) showed vesicles in the lumen of the flagellar sleeve albeit without IFT-like material (Fig. 7G). Some, but not all, of these small vesicles exhibited a disk-like shape and were reminiscent of exocytic vesicles (Grunfelder et al., 2003).

As the Golgi complex and the basal body are linked (He et al., 2004), Golgi location was analysed in non-flagellated cells. It was found close to the flagellar pocket with the same frequency in DHC1b^RNAi, IFT88^RNAi and IFT172^RNAi (50-60% sections, n=28-51) cells as in wild-type cells (64%, n=72), showing that a flagellum is not required for Golgi association to the flagellar pocket or basal body region. The flat compartment identified above (Fig. 7A-C) was equally present in the vicinity of the modified flagellar pocket in IFT^RNAi cells (Fig. 7I).

We next examined the status of several organelles associated with the endocytic network using specific markers. First, the p67 protein was used as a marker of lysosomes that localise between the kinetoplast and the nucleus (Alexander et al., 2002). This pattern was reproduced in control non-induced cells (supplementary material Fig. S3A) but was altered in non-flagellated IFT140^RNAi or IFT88^RNAi cells where p67 was found in multiple foci at various locations in the cytoplasm (supplementary material Fig. S3B). By contrast, cells that retained their flagellum showed a normal staining pattern. The endoplasmic reticulum was analysed with an antibody against the chaperone protein BiP, which normally decorates the periphery of the nucleus and most of the cytoplasm (Bangs et al., 1993): a staining pattern reproduced in non-induced populations (supplementary material Fig. S3C). Staining was apparently normal in non-flagellated cells, but turned out to be stronger in multinucleated cells (supplementary material Fig. S3D). To ensure that these modifications were not due to general disorganisation of mutant cells, the distribution of organelles unrelated to the endocytic network such as the mitochondrion or the glycosomes was analysed. An antiserum recognising the glycolytic enzyme aldolase and the mitochondrion tracer MitoTracker (not shown) were used for the analysis. In both cases, normal staining patterns were observed in flagellated and non-flagellated cells (supplementary material Fig. S3E,F).

These multiple phenotypes raise the question whether the flagellar pocket is competent for endocytosis in IFT^RNAi mutants. Receptor-mediated endocytosis cannot be traced easily in procyclic trypanosomes but fluid-phase endocytosis can be monitored by incubating cells with the lipophilic dye FM4-64 (Hall et al., 2005). Induced or non-induced cells were incubated at 4°C with FM4-64 and subsequently heated at 27°C, which should result in internalisation through the flagellar pocket and accumulation at the posterior region. This was observed in non-induced trypanosomes (quantification at Fig. 8C). In induced IFT172^RNAi and IFT88^RNAi cells, those cells that still possessed a flagellum showed a similar bright signal, demonstrating that the flagellar pocket of the old flagellum remains functional (Fig. 8A-D). Strikingly, the vast majority of non-flagellated cells presented much less (but still detectible) fluorescence (Fig. 8A-C), which was often more dispersed and not restricted to the flagellar pocket area. Dividing cells were analysed at early stages of RNAi, when they possess an old flagellum but do not grow a new flagellum (Fig. 8D). FM4-64 accumulation occurred normally and the tracer concentrated around two separate zones, each found in the vicinity of a kinetoplast. This shows that despite the absence of a normal flagellum, a new flagellar pocket can be formed, which is competent for endocytosis, at least before the cell undergoes cytokinesis. Nonflagellated cells exhibit a much-reduced capacity to capture tracers, showing that the presence of the flagellum is required for maintaining a fully functional flagellar pocket.

Our data reveal that in the absence of elongation of the new flagellum upon silencing of distinct components of the IFT machinery, a flagellar pocket develops close to the basal body and kinetoplast complex. However, this flagellar pocket exhibits a different structure compared with that in control cells or with the flagellar pocket associated with the old flagellum. First, transmission electron microscopy showed the presence of a shorter flagellar pocket in cells without a flagellum or of modified shape in cases where the flagellum was abnormally short. Second, the orientation of the FPC relative to the long axis of the cell was much more variable compared with the well-defined positioning observed in control cells. Moreover, the structure of the collar consistently looked thinner or more irregular, as revealed upon detergent extraction. Third, the BILBO1 protein, the only molecular marker of the FPC, showed partially modified staining in cells with a short or no flagellum.

Flagellum replication and flagellar pocket formation are intertwined processes that can be summarised as follows (Fig. 9). After basal body duplication, the new flagellum is constructed in the large anterior side of the existing flagellar pocket (Absalon et al., 2008; Briggs et al., 2004; Grasse, 1961). The new flagellum contains a significant amount of IFT-like amorphous material (Absalon et al., 2008; Grasse, 1961), a situation reminiscent of the short new flagella of Chlamydomonas (Dentler, 2005). Flagellar microtubules elongate and the new flagellum grows in the same flagellar pocket as the old flagellum. However, when the new flagellum exits from the cell body, it possesses its own flagellar pocket. Analysis of control trypanosomes failed to identify cells with two flagella emerging from the same flagellar pocket, suggesting that the new flagellum acquires its own FPC before or as it emerges from the cell body. Transmission and scanning electron microscopy of detergent-extracted cells suggest that the collar of the new flagellar pocket is thinner compared with that of the old flagellum. At this stage, the new flagellum is in a posterior position relative to the old one, indicating that rotation of its basal body must have taken place prior to formation of the flagellar pocket associated with the new flagellum. Elongation continues, the basal bodies migrate apart extensively, and the new flagellar pocket acquires associated organelles, such as the Golgi complex (He et al., 2004) or the lysosomes (Alexander et al., 2002). Endocytic/exocytic vesicles become visible (Jeffries et al., 2001; Morgan et al., 2001) and FM4-64 uptake reveals endocytosis activity (Hall et al., 2005).

In cells that cannot assemble a new flagellum (no anterograde IFT), duplication of the basal body complex takes place normally, at the expected anterior position. It is followed by formation of a new but shorter flagellar pocket with a less developed or structured collar. Basal body rotation and initial migration to a posterior position occur, showing that the flagellum is not required for these processes. However, complete migration of the new basal body towards the posterior end is rarely observed (Absalon et al., 2007). In mutants of retrograde IFT, a short, dilated, flagellum is formed associated with a flagellar pocket of an apparently normal size but of drastically modified shape. This implies at least two contributions of the flagellum to flagellar pocket formation: (1) elongation, possibly by providing a supporting anchor and (2) shape definition, possibly by bringing specific proteins to the contact area between the flagellar pocket and the flagellum membrane. Movement of flagellar membrane proteins by IFT has been shown in other organisms (Huang et al., 2007; Qin et al., 2005) and could be envisaged for proteins linking the FPC to the flagellum membrane. Thus, flagellum elongation could be required for formation of a fully functional FPC, as suggested by common phenotypes observed in IFT^RNAi and BILBO1^RNAi mutants (Bonhivers et al., 2008b), such as accumulation of vesicles containing CRAM and disturbed p67 labelling.

Another phenotype is the randomisation of flagellar pocket orientation in both types of IFT mutants. In normal cells, detergent treatment reveals the FPC as a horseshoe-like structure perpendicular to the axis of the flagellum, with its posterior part slightly lifted above its point of emergence from the cell body (Fig. 9). This configuration is visible by scanning electron microscope observation of whole cells, confirming that this defined positioning is not an artefact of detergent extraction. Given the role of forward motility in shifting surface-bound ligands to the flagellar pocket (Engstler et al., 2007), such a configuration could restrict movement beyond the flagellar pocket and favour penetration in the flagellar pocket. This precise orientation is lost in both types of IFT mutant cells and the flagellar pocket is found in any position. The disorientation of the new flagellar pocket is already detected in cells that still possess the old flagellum, and could be due to the failed basal body migration typical of IFT mutants (Absalon et al., 2007).

Structural modifications of the flagellar pocket were accompanied by several defects: reduced fluid-phase endocytosis, different vesicular profile in the cytoplasm, presence of vesicles in the short flagella and mislocalisation of CRAM that leaked out of the flagellar pocket and on to the surface of the short flagella of IFT140^RNAi cells. In wild-type cells, CRAM is abundant at the flagellar pocket close to the base of the flagellum (Lee et al., 1990). Dissection studies revealed that the short cytoplasmic tail of CRAM is essential for retention in the flagellar pocket. In its absence, the truncated protein is allowed to diffuse along the length of the flagellum, indicating the existence of a molecular filter at the junction between the flagellum and the flagellar pocket (Yang et al., 2000). This molecular filter is clearly less or not active in induced IFT140^RNAi cells, because CRAM is allowed to diffuse in the flagellar compartment and on the surface of the flagellar sleeve. This suggests that proteins localised to the flagellum or flagellar pocket filter area are missing or non-functional. An alternative explanation could be a direct involvement of retrograde IFT in recycling CRAM to the base of the flagellum in normal cells, a feature that is not functional in retrograde mutants. By contrast, the presence of CRAM at the rest of the cell surface is rarely observed, indicating that the physical distinction between the pellicular membrane and the flagellum and flagellar pocket membranes (Bastin et al., 2000) is still operating.

Endocytosis and exocytosis exclusively take place at the flagellar pocket. Fluid-phase endocytosis still takes place in non-flagellated cells, in agreement with the presence of a flagellar pocket. However, it is much less efficient, which could be attributed to a combination of several factors: (1) a smaller flagellar pocket surface; (2) a lower density of proteins involved in endocytosis; (3) the absence of forward motility; and (4) the randomised orientation of the flagellar pocket. Reduced endocytosis probably results in supplementary problems in the balance with exocytosis, for example if the modified flagellar pocket is not able to cope with the arrival of exocytic vesicles. This could explain the formation of the flagellar sleeve (Davidge et al., 2006) that regularly looks like a succession of vesicles in all nine IFT^RNAi mutants examined to date (Absalon et al., 2008) (our unpublished data). The CRAM signal was much more apparent in the cytoplasm, where it could accumulate at the level of the ER (Yang et al., 2000). This could be due to the imbalance of vesicle trafficking, if the same amount of CRAM is produced but unable to properly insert in the flagellar pocket, therefore accumulating in the exocytic pathway.

The presence of vesicles in the flagellar pocket could be related to the emergence of vesicles with a double membrane found in the Golgi and flagellar pocket regions, as suggested by their similar frequency (∼30%) (Figs 2 and 7). The origin of these vesicles is not clear at this stage, but they seem to arise from the Golgi area where deformed vesicles have also been identified. Bending of such vesicles would lead to formation of double-membrane vesicles, which in turn could fuse with the flagellar pocket membrane and hence deliver a single membrane vesicle in the flagellar pocket lumen. Excessive vesicles could be targeted for autophagy and bent vesicles might represent the initial stage of this process. The proportion of sections showing concentric vesicles indicating autophagy rose to 5-10% in IFT^RNAi mutants compared with 1% in wild-type controls. However, a similar increase was noted for the axonemal mutants PF16^RNAi and PF20^RNAi, which display flagellar pockets of normal size and shape (Branche et al., 2006), suggesting that this phenomenon is not specific to defects in flagellar pocket function.

Alternatively, flagellar pocket vesicles could arise from direct vesicle budding from the flagellar pocket membrane, possibly related to the plasma membrane `blebs' observed upon silencing of WCB, a protein containing calcium-binding C2 domains that is associated to the subpellicular microtubules of the corset (Baines and Gull, 2007). Intriguingly, some flagellar proteins contain C2 domains (Avidor-Reiss et al., 2004), although none of the IFT proteins studied here present such a motif.

In conclusion, elongation of the new flagellum is not required for flagellar pocket formation but is central to flagellar pocket structural development and orientation. IFT^RNAi mutants illustrate two levels of this contribution: flagellar pocket elongation and flagellar pocket shape definition, with important functions in the molecular filter that separates the membrane of the flagellum from that of the flagellar pocket. Identification of the molecular components of the flagellar pocket, such as the recent discovery of BILBO1, should contribute to better understanding of this remarkable and critical structure of trypanosomes. These results could be significant for other organisms, because the base of primary cilia in numerous mammalian cell types is positioned in an invagination of the cell surface that is remarkably similar to the flagellar pocket (Han et al., 2008; Jensen et al., 2004; Sorokin, 1968).

All cell lines used for this work were derivatives of strain 427 of T. brucei and cultured in SDM79 medium supplemented with hemin and 10% fetal calf serum. Cell lines DHC1b^RNAi, IFT88^RNAi (Kohl et al., 2003), IFT20^RNAi (Absalon et al., 2007), IFT140^RNAi and IFT172^RNAi (Absalon et al., 2008) express double-stranded RNA from two tetracycline-inducible T7 promoters facing each other in the pZJM vector (Wang et al., 2000), transformed in 29-13 cells that express the T7 RNA polymerase and the tetracycline repressor (Wirtz et al., 1999). RNAi was induced by addition of 1 μg tetracycline per ml of medium and fresh tetracycline was added at each cell dilution.

Cells were washed, settled on poly-L-lysine-coated slides and fixed in 4% paraformaldehyde for 10 minutes (for anti-aldolase or anti-CRAM). Fixed cells were permeabilised with 1% Nonidet P-40 for 10 minutes and samples were rinsed to remove excess detergent. Blocking was performed by incubation for 45-60 minutes in PBS containing 1% bovine serum albumin. Alternatively, cells or detergent-extracted cytoskeletons were washed, settled on poly-L-lysine-coated slides, fixed in methanol at –20°C for 5 minutes and rehydrated in PBS (Mab22, anti-TBBC, anti-BiP, anti-p67). In all cases, slides were incubated with primary antibodies for 45-60 minutes. Monoclonal mAb22 (dilution 1:5) is an IgM that detects an as yet unidentified antigen found at the exclusion zone of the tripartite attachment complex that links the proximal zone of both the mature and the pro-basal body to the kinetoplast (Bonhivers et al., 2008a; Pradel et al., 2006). TBBC presence and localisation was analysed with a rabbit polyclonal antiserum (dilution 1:100) (Dilbeck et al., 1999). The anti-BILBO1 antibody was used as a marker of the FPC (IgM, dilution 1:2 in PBS after blocking in 0.5% BSA for 10 minutes, and revealed by a FITC-conjugated secondary antibody 1:100 in PBS, Sigma). Antibodies against p67 (1:200) and BiP (1:500) were used as markers for lysosomes (Alexander et al., 2002) and the endoplasmic reticulum (Bangs et al., 1993), respectively. An antiserum raised against aldolase (dilution 1:4000) was used as marker of glycosomes (kind gift from Paul Michels, Christian de Duve Institute, Brussels, Belgium). Slides were washed in PBS and incubated with the appropriate secondary antibodies coupled to Alexa Fluor 488 (Invitrogen), Cy3 or Cy5 (Jackson). Slides were stained with DAPI for visualisation of kinetoplast and nuclear DNA content.

For detergent treatment, cells were settled on poly-L-lysine-coated slides and exposed to various concentrations of Nonidet P-40 in spindle stabilisation buffer made of 4 M glycerol, 10 mM PIPES (pH 6.5), 10 mM MgCl₂ and 5 mM EGTA (Roobol et al., 1984).

For FM4-64 uptake, induced and non-induced RNAi cell lines were grown at a density of 1-5×10⁶ cells per ml. A total of 10⁷ cells were washed and incubated with 40 μM FM4-64 for 20 minutes at 4°C or at 27°C and treated as described (Hall et al., 2005).

Slides were observed with a DMR Leica microscope and images were captured with a Cool Snap HQ camera (Roper Scientific). Alternatively, slides were also viewed on a DMI4000 Leica microscope and images were acquired with a Retiga-SRV camera (Q-Imaging). Images were analysed using the IPLab Spectrum 3.9 software (Scanalytics & BD Biosciences) or Image J (NIH).

Cell fixation, embedding and sectioning for transmission electron microscopy of whole cells from wild-type and induced PF20^RNAi, DHC1b^RNAi, IFT88^RNAi and IFT172^RNAi samples was carried out as described previously (Branche et al., 2006). Cells from the IFT140^RNAi cell line were treated similarly except that tannic acid was not included and that dehydrated samples were embedded in Epon resin followed by polymerisation for 48 hours at 60°C. For scanning electron microscopy of whole cells, samples were prepared and analysed as described previously (Absalon et al., 2007). For immunogold detection with the anti-CRAM antibody, 10⁷ cells were washed twice before fixation for 45 minutes in 4% paraformaldehyde and 0.1% glutaraldehyde, followed by two 5 minute washes and neutralisation in PBS containing 50 mM NH₄Cl for 30 minutes. Coverslips were incubated in PBS containing 1% BSA for blocking and incubated with the anti-CRAM (dilution 1:150) in PBS with 0.1% BSA for 45 minutes. After three washes in the same buffer, bound antibodies were labelled by addition of Protein AG conjugated to gold particles of 20 nm diameter (1:70) in PBS with 1% BSA. Coverslips were then prepared for scanning electron microscopy analysis according to standard procedures (Absalon et al., 2007), except that metallisation was carried out with carbon instead of gold-palladium.

Cells were treated with 0.1-1% Triton X-100 at 4°C in PBS for 10 minutes to strip the plasma membrane and visualise the FPC. Samples were washed twice in PBS, fixed in glutaraldehyde and processed for scanning electron microscopy in standard conditions (Absalon et al., 2007).

We thank Gwéndola Doré for assistance in plasmid DNA preparation and in preliminary immunofluorescence experiments; Mark Field and Belinda Hall for setting up the FM4-64 assay and for various discussions, Derek Nolan for discussions about the flagellar pocket phenotype, Markus Engstler for critical reading of the manuscript, Etienne Pays, Jay Bangs and Paul Michels for providing several antibodies. We acknowledge the electron microscopy departments of the Muséum National d'Histoire Naturelle and of the Pasteur Institute for providing access to their equipment, as well as advice from Marie-Christine Prévost and Stéphanie Guadagnini. This work was funded by the CNRS, by a Programme Protéomique et Génie des Protéines (D.R. and P.B.), an ACI grant in Development Biology, by a GIS grant on rare genetic diseases (P.B.) and by Sanofi-Aventis. S.A. was funded by a MRT fellowship, by an FRM doctoral fellowship and by a Pasteur-Weizman fellowship. J.B. is funded by a MRT fellowship.