Intraflagellar transport is required for the maintenance of the trypanosome flagellum composition but not its length

Cilia and flagella (which are interchangeable terms) are found at the surface of eukaryotic cells where they perform multiple roles including cellular motility, sensory function, developmental signalling and cell morphogenesis (Drummond, 2012). They share a similar basic structure with an axoneme composed of nine doublets of microtubules surrounded by a specific membrane. During assembly of cilia and flagella, the incorporation of new subunits takes place at the distal end of the organelle. The construction relies on an evolutionary conserved process called intraflagellar transport (IFT), a bi-directional movement of multi-protein complexes between the membrane and the axoneme microtubules (Kozminski et al., 1993). IFT particles or trains (Pigino et al., 2009) are composed of proteins that can be separated into a complex A and a complex B (Cole et al., 1998; Piperno and Mead, 1997). IFT proteins are found along the length of cilia (mature or in construction), but a substantial proportion is also found at the basal body area and in the cell body (Absalon et al., 2008; Cole et al., 1998; Deane et al., 2001; Pazour et al., 2002). Inside the flagellum, kinesin II (a member of the kinesin-2 family of motors) transports IFT-A and IFT-B complexes, the inactive IFT dynein motor (Blisnick et al., 2014; Hao et al., 2011a; Signor et al., 1999) and axonemal precursors from the base to the tip of flagellum (Craft et al., 2015; Wren et al., 2013). At the tip of the flagellum, axonemal components are delivered for assembly whereas IFT trains are remodelled and returned to the base by the active dynein motor (Buisson et al., 2013; Pedersen et al., 2006).

In agreement with this model, inhibition of IFT blocks the construction of cilia in all species investigated to date, including algae (Kozminski et al., 1995; Pazour et al., 2000), nematodes (Haycraft et al., 2001; Signor et al., 1999), ciliates (Brown et al., 2003, 1999), mammals (Nonaka et al., 1998; Pazour et al., 2002), trypanosomes (Kohl et al., 2003), insects (Lee et al., 2008; Sarpal et al., 2003) or fish (Sun et al., 2004). However, video microscopy has revealed that IFT remains active after assembly of the organelle in many cell types (Buisson et al., 2013; Kozminski et al., 1993; Orozco et al., 1999), implying the existence of functions beyond construction. The temperature sensitive fla10 Chlamydomonas mutant (Huang et al., 1977) was instrumental in revealing a function for IFT in the maintenance of flagellum length in this organism. FLA10 encodes one of the motor subunits of the heterotrimeric IFT kinesin. At the permissive temperature, IFT takes place and flagella are assembled normally. Once the cells are shifted to the restrictive temperature, IFT ceases and flagella progressively shorten until they eventually disappear (Kozminski et al., 1995). This result was explained by elegant experiments using epitope-tagged tubulin revealing that the tip of the flagellum constantly depolymerizes and that incorporation of fresh tubulin compensates for that phenomenon (Marshall and Rosenbaum, 2001). Hence, IFT would be required to ensure a constant input of tubulin to balance this loss. This does not happen in fla10 cells grown at the restrictive temperature, supporting a role of IFT in the maintenance of flagellum length (Marshall and Rosenbaum, 2001). However, formal evidence in other organisms is missing, probably because of the difficulty of discriminating cilium assembly from maintenance in mammalian cells or in C. elegans.

Here, we used the flagellated protist Trypanosoma brucei as a model to investigate the role of IFT in flagellum maintenance. This organism is responsible for sleeping sickness in central Africa but also turned out to be an excellent model to study flagellum biology (Vincensini et al., 2011). It assembles a new flagellum at each cell cycle while conserving the old one (Sherwin and Gull, 1989a), providing the opportunity to compare mature and constructing flagella in the same cell. Its genome contains the full complement of IFT-A, IFT-B, kinesin and dynein genes with the only exception of the kinesin-associated protein (KAP) (Julkowska and Bastin, 2009). IFT has been imaged by tagging multiple IFT-B or dynein subunits and is active during both construction and maintenance. The expression of various components of the IFT-A and IFT-B complexes has been individually silenced using inducible RNA interference (RNAi), resulting in failure of flagellum construction (Absalon et al., 2008; Adhiambo et al., 2009; Blisnick et al., 2014; Davidge et al., 2006; Franklin and Ullu, 2010; Huet et al., 2014; Kohl et al., 2003). Usually, inhibition of IFT-B proteins (for example IFT88) blocks axoneme construction in agreement with a default in anterograde transport, whereas absence of IFT-A proteins (such as IFT140) leads to formation of short flagella filled with IFT material, indicative of defects in retrograde transport (Absalon et al., 2008).

In all IFT^RNAi mutants analysed thus far, flagella assembled before RNAi treatment remained present, a phenomenon which has been interpreted as the persistence of IFT proteins and trafficking (Kohl et al., 2003). Indeed, RNAi destroys mRNA and therefore impacts only on new protein synthesis whereas existing proteins disappear according to their own turnover rate (Bastin et al., 2000). However, efficient tools to investigate IFT were not available at the time of these studies. We have since produced a monoclonal antibody against the IFT-B protein IFT172 (Adhiambo et al., 2009) and set up conditions to reliably visualize IFT in trypanosomes (Buisson et al., 2013). Moreover, the development of vectors for in situ tagging now allows expression of fluorescent fusion proteins at their endogenous level, an essential condition to monitor IFT in mutant situations (Kelly et al., 2007). Therefore, it is now possible to examine IFT status in mature flagella in an attempt to decipher the role of IFT in flagellum maintenance.

Using transmission electron microscopy (TEM), immunofluorescence assays (IFA) and live video microscopy, we show here that IFT is absent or arrested in old flagella of trypanosome IFT88^RNAi (anterograde mutant) and IFT140^RNAi (retrograde mutant), respectively. In these conditions, the old flagellum does not shorten but the distribution of several of its proteins is drastically modified, revealing a key function for IFT in the maintenance of flagellum composition.

Previous experiments have shown that that depletion of IFT88 by RNAi (denoted as IFT88^RNAi) blocks axoneme construction in agreement with a default in anterograde transport, whereas absence of IFT140 (denoted IFT140^RNAi) leads to formation of short flagella filled with IFT material, indicative of defects in retrograde transport (Absalon et al., 2008; Kohl et al., 2003). Here, we investigated the fate of IFT proteins and their trafficking in cells with long flagella that had been assembled prior to RNAi.

The presence of IFT particles in the old flagella of IFT88^RNAi and IFT140^RNAi cells was examined by TEM in non-induced cells and in samples induced for 2 days. In these conditions, less than 10% of the cells still construct a new flagellum and when it happens, this flagellum is very short, rarely extends beyond the flagellar pocket and misses the paraflagellar rod (PFR) (Absalon et al., 2008; Kohl et al., 2003). Therefore, sections through the flagellum in these samples will overwhelmingly correspond to the old flagellum. IFT particles are recognized on flagellum cross-sections as electron-dense granules found between the membrane and the axoneme (Absalon et al., 2008). We examined multiple cross-sections of non-induced RNAi mutants. IFT particles showed a very specific location, being found almost exclusively close to axoneme doublets three and four (Fig. 1A,B,F), and seven and eight (Fig. 1A,E), as previously reported (Absalon et al., 2008). In non-induced control IFT88^RNAi cells, more than half of the flagellum cross-sections showed an IFT particle (n=224) (Fig. 1I). After 2 days of RNAi induction, this value dropped to less then 30% of sections (n=222) (Fig. 1C,D,I). In the case of IFT140^RNAi, 65% of the cross-sections through flagella in control cells presented IFT particles (n=177) (Fig. 1J). After 2 days of induction, ∼30% of the sections still contained a single IFT particle (n=261). In contrast, close to 60% of sections were characterized by the presence of a large amount of electron-dense material (termed ‘pluri-density’) that could correspond to IFT material (Fig. 1G,H,J). These results suggest that in the absence of IFT88, the number of IFT particles is reduced in the old flagellum whereas in IFT140^RNAi cells, excessive material is visible.

To firmly correlate the modification in abundance of the electron-dense structures with IFT, an immunofluorescence assay (IFA) was performed on methanol-fixed trypanosomes, using double labelling with MAb25, a monoclonal antibody that recognizes a protein found all along the axoneme (Pradel et al., 2006) and with a monoclonal antibody raised against IFT172, a classic IFT marker (Absalon et al., 2008; Blisnick et al., 2014). MAb25 produced staining along the whole length of the flagellum in both mature flagella (white arrows) and flagella undergoing construction (yellow arrows, Fig. 2A,C). The anti-IFT172 gave punctuate staining along the length of the flagellum and also lit up its base in non-induced IFT88^RNAi and IFT140^RNAi cells (Fig. 2A,C). In IFT88^RNAi cells induced for 2 days, the majority of cells failed to assemble a new flagellum as revealed by little or no MAb25 signal (yellow arrowhead). Cells that retained the old flagellum only showed an IFT172 signal at the basal body area and very little, if any, signal along the axoneme (Fig. 2B). This result is consistent with the reduced number of IFT trains detected by TEM.

A very different result was obtained when IFT140^RNAi cells were stained with the same antibodies (Fig. 2D). As expected, induced cells constructed only a short axoneme (short line as MAb25 signal; yellow arrowhead in Fig. 2D) that accumulated IFT172 in agreement with the retrograde transport defect (Absalon et al., 2008). In contrast to IFT88^RNAi, a strikingly increased signal for IFT172 was observed along the length of the old flagellum (white arrowheads), especially at the distal region (circles, Fig. 2D). This is consistent with the excessive number of visible IFT trains observed by TEM (Fig. 1G,H). We conclude that in the absence of anterograde transport, entry of IFT proteins in the old flagellum is restricted, whereas in the absence of retrograde transport, IFT-B complex proteins accumulate in the persisting flagellum, presumably because of a failure in train recycling.

IFT is a dynamic process with IFT particles travelling rapidly in both directions inside the flagellum. To monitor IFT in the old flagellum of IFT88^RNAi and IFT140^RNAi cells, they were nucleofected with a plasmid allowing expression of YFP::IFT81 (another component of IFT-B complex) from its endogenous locus (Bhogaraju et al., 2013). Western blot analysis confirmed the expression of YFP::IFT81 at the expected size of 113 kDa. Its total amount was not modified upon IFT88 or IFT140 knockdown (Fig. S1). Observation of live cells in non-induced IFT88^RNAi and IFT140^RNAi cultures revealed classic IFT trafficking in both new and old flagella. In both cell lines, YFP::IFT81 was concentrated at the flagellar base (Fig. 3A, asterisk) and moved rapidly in both anterograde and retrograde directions in the flagellum (Movies 1 and 3). The movement of IFT trains can be followed on still images (coloured arrowheads on Fig. 3A). Kymograph analysis was performed (an example is shown on the last panel on Fig. 3A) to determine the speed and the frequency of anterograde trains. In the case of IFT88^RNAi speed was 2.27±0.72 µm/s and train frequency was 0.9/s (mean±s.d., n=271). For IFT140^RNAi cells, it was 2.66±0.89 and 0.79/s (n=233). These values are within the usual range for IFT in trypanosomes (Bhogaraju et al., 2013; Blisnick et al., 2014; Buisson et al., 2013; Huet et al., 2014). In contrast, observation of live IFT88^RNAi trypanosomes induced for 2 days showed a dramatic reduction in trafficking of IFT trains inside the flagellum (Fig. 3B) (Movie 2). IFT trafficking was not visible, neither on videos nor on kymographs (n=28). In induced live IFT140^RNAi cells, a bright YFP signal was present along the length of the flagellum especially in the distal region whereas the pool of IFT material at the basal body area was drastically reduced (Fig. 3C). Two different types of profiles could be described, with two thirds of the cells (n=16) showing abundant YFP::IFT81 signal along the whole flagellum length with a brighter signal at the distal tip whose intensity varied from one cell to another, with substantial (Movie 4) to spectacular accumulations (Movie 5). In the remaining one third of cells, a continuous gradient of YFP::IFT81 from the distal to the proximal end of flagellum was observed (white bracket, Fig. 3C; Movie 6). No IFT movement could be detected. These results formally show that IFT88 is essential for the entry of other IFT-B proteins whereas IFT140 is required for the return of IFT particles in already assembled flagella. They also demonstrate that IFT proteins undergo turnover in the mature flagellum, in contrast to what was initially thought (Kohl et al., 2003). This turnover is relatively slow as it takes hours to be detected, explaining why it had been missed in photobleaching experiments that were focused on a short acquisition period because the IFT cycle in the flagellum takes ∼20 s (Buisson et al., 2013).

Typical movement of IFT trains could not be identified in induced IFT140^RNAi cells. However, trafficking could still take place but be hidden by the intense fluorescent signal. To address this, fluorescence recovery after photobleaching (FRAP) experiments were performed in IFT140^RNAi cells expressing endogenously tagged YFP::IFT81. Cells were induced and a region of interest in the middle of the flagellum was selected for photobleaching. Only cells with a long flagellum were used to make sure this visualized flagellum indeed corresponded to an old flagellum. The fluorescence recovery was evaluated by continuously recording videos of live cells for 45 s and by kymograph analysis. Two different situations were observed. In 41% (n=15) of the videos, the photo-bleached region remained dark and exhibited no detectable recovery, with the fluorescent signal of the region of interest remaining at the background level. Typical IFT trafficking was clearly absent from both video images (Movie 7) and from analysis of the kymograph (data not shown). In the other videos, the photo-bleached region showed very weak recovery ranging from 2 to 9% of the initial signal (after 30 s, Fig. 3D; Movie 8). Kymograph analysis failed to detect typical IFT trafficking, even after enhancing brightness and contrast (Fig. 3D, right-most panel). However, a weak progressive return of fluorescence was detected in the periphery of the region of interest (indicated by yellow arrows on the kymograph, Fig. 3D), more suggestive of a diffusion-like process. These experiments confirm that IFT proteins present in the old flagellum of induced IFT140^RNAi cells are indeed arrested and do not undergo IFT.

We conclude that in the absence of IFT88 or IFT140, the typical IFT train trafficking along the old flagellum is lost. The silencing of IFT88 reveals that IFT trains cannot enter in the old flagellum whereas the silencing of IFT140 demonstrates an accumulation of IFT particles along the length of the flagellum in agreement with a default in train recycling. In both situations, flagella do not display active IFT. This allows us to address the biological significance of IFT trafficking in the mature flagellum.

In the green algae Chlamydomonas, IFT trafficking is essential for the maintenance of flagellum length (Kozminski et al., 1995). In the absence of IFT, the axoneme disassembles progressively from its distal end. The average length of the flagellum appears reduced in trypanosome IFT RNAi mutants (Absalon et al., 2008, 2007; Kohl et al., 2003). However, the situation is complex because trypanosomes maintain their flagellum during the whole cell cycle. It is therefore of prime importance to discriminate the consequences of IFT RNAi silencing on construction and maintenance. We therefore examined the cell cycle and the length of the new and the old flagellum during the course of RNAi induction in IFT88^RNAi cells and IFT140^RNAi cells. We stained the axoneme with the MAb25 antibody and used DAPI to visualize nuclear and kinetoplast DNA (n>100). At the onset of the cell cycle, control trypanosomes possess a single nucleus (N) and a single mitochondrion whose genetic material is concentrated in an organelle named a kinetoplast (K) that is physically linked to the basal body (Robinson and Gull, 1991). The progression of trypanosomes in the cell cycle can be followed by DNA staining as individuals with one kinetoplast and one nucleus (1K1N) are in the G1/S phase, those with two kinetoplasts and one nucleus (2K1N) are in G2/M and cells with two kinetoplasts and two nuclei (2K2N) are about to divide (Sherwin and Gull, 1989a) (see cartoons in Fig. 4A). Flagellum elongation (the growing flagellum is shown in yellow on all cartoons) starts immediately after basal body maturation and continues at a constant rate of 3–4 µm/h throughout the rest of the cell cycle (Bastin et al., 1999).

Trypanosome cultures are not synchronized, and cells in all phases of the cell cycle are present. For each cell line, we measured the length of the old and the new flagellum in non-induced cells and early after induction at a time when IFT is aborted in the old flagellum (24 h for IFT88^RNAi and 48 h for IFT140^RNAi). In non-induced 1K1N cells, the length of the old flagellum measured using the MAb25 staining was on average 15 µm for IFT88^RNAi cells (Fig. 4B) and 18 µm for IFT140^RNAi cells (Fig. 4F). This is a bit shorter compared with wild-type cells (19.5 µm) and is probably explained by RNAi leakiness (Absalon et al., 2007). At the 2K1N stage, the new flagellum can be at different stages of elongation resulting in a large heterogeneity in length as shown by the high dispersion (open red dots, Fig. 4C,G). This dispersion is due to the heterogeneity of this cell cycle stage where the flagellum elongates from 1–2 µm to 15 µm (Sherwin and Gull, 1989a; Subota et al., 2014). The length of the old flagellum (closed red dots) was similar to cells with one flagellum (Fig. 4C,G). The 2K2N stage is a shorter phase preceding cytokinesis where the length of the new flagellum (open red dots) reaches ∼80% of that of the old flagellum (Fig. 4D,H) (Robinson et al., 1995; Subota et al., 2014). The length of the old flagellum (closed red dots) was similar to what was observed for the two preceding stages of the cell cycle (Fig. 4D,H). One day after initiation of RNAi in the IFT88^RNAi strain, construction of the new flagellum was impeded as revealed in 2K1N and 2K2N cells (Fig. 4C,D, open blue dots), in agreement with the essential role of IFT in flagellum construction (Kohl et al., 2003). In contrast, the length of the old flagellum (closed blue dots) was unchanged despite the fact that IFT was absent in these flagella (Fig. 4C,D). These results are summarized in the cartoons presented at Fig. 4E. IFT arrest emerges more slowly in IFT140^RNAi cells and consequences were examined after 2 days of induction. It revealed interesting data with a mixture of two populations of 1K1N cells with different flagellum lengths. About half of the cells exhibited the same length as non-induced controls whereas the other half displayed much shorter flagella (<5 µm) (Fig. 4F, blue dots). Examination of 2K1N and 2K2N stages (Fig. 4G,H, closed blue dots) showed that these shorter cells are the daughters that inherited a flagellum that was constructed at too short a length due to inhibition of retrograde IFT (Absalon et al., 2008) (Fig. 4G,H, open blue dots). The other half inherited the flagellum that was assembled prior to RNAi inhibition and these showed no reduction in length despite the fact that IFT is arrested (see cartoons at Fig. 4I).

These data confirm that IFT is essential for the construction of the new flagellum as expected, but reveal that IFT is dispensable for maintenance of flagellum length. Indeed, the length of the mature flagellum is not reduced in the absence of IFT trafficking in both mutant cell lines. Given that IFT trafficking requires energy, this raises the question of the role of IFT in the old flagellum. One possible explanation for the persistence of this machinery could be the involvement of IFT in the turnover or the maintenance of flagellar components. To address this question, the fate of different categories of proteins was investigated: (1) skeletal proteins composing the axoneme and the PFR, and (2) non-structural proteins.

Flagella are complex organelles made of hundreds of proteins with variable structure, function and localization (Broadhead et al., 2006; Oberholzer et al., 2011; Pazour et al., 2005; Subota et al., 2014). To determine the role of IFT trafficking in the maintenance of structural proteins within the old flagellum, we used either endogenous tagging of several typical axoneme or PFR markers or IFA with available antibodies in both IFT88^RNAi cells and IFT140^RNAi cells. First, the hydin axonemal central pair protein (Dawe et al., 2007; Lechtreck and Witman, 2007) was tagged with GFP, revealing its presence not only all along the flagellum but also in the cell body, as shown by IFA with an anti-GFP antibody (Fig. S2) or by live video microscopy (data not shown). After RNAi induction, cells that had retained the old flagellum were still positive for hydin with no detectable modification of signal intensity in both cell lines (Fig. S2). A similar result was obtained when using a polyclonal antibody against the intermediate dynein chain (DNAI1) of the outer dynein arm (Duquesnoy et al., 2009): the flagellum signal was unchanged in the old flagella after 2 days of RNAi induction (data not shown).

We next turned our attention to the PFR, a large axonemal structure found alongside the axoneme (Portman and Gull, 2010). One of its main components, the adenylate kinase B protein (AKB) (Pullen et al., 2004) was tagged with YFP and its presence and distribution was investigated in control and induced IFT88^RNAi and IFT140^RNAi cell lines (Fig. S3). In the non-induced situation, AKB was found in the flagellum but only after its emergence from the cell body at the flagellar pocket level and double staining with the MAb25 axoneme marker confirmed a slightly off set positioning, in agreement with a PFR location (Fig. S3A,C). This pattern was not modified in the old flagella of induced samples (Fig. S3B,D). This was confirmed using a monoclonal antibody against PFR2, one of the major PFR proteins (Kohl et al., 1999) that showed the same fluorescence location and intensity in old flagella of induced samples (data not shown). These data indicate that the presence of structural axoneme and PFR components is not altered in the absence of IFT, in agreement with the apparently normal ultrastructure observed during the TEM analysis (Fig. 1). Therefore, we conclude that IFT is not necessary for the maintenance of these constituents after construction.

These results raise the question of the role of IFT in flagellum motility. Trypanosomes grow in suspension and swim actively in liquid medium (Heddergott et al., 2012). The movement of IFT140^RNAi cells was quantified in non-induced control conditions and after 3 days of RNAi induction. Non-induced cells exhibited an average speed of 9.8±2.8 µm/s (mean±s.d., n=380) (Fig. S3E), in the expected range for procyclic trypanosomes in culture (Brasseur et al., 2013). Microscope observation revealed a striking reduction in motility that was confirmed by tracking analysis with an average speed at 2.8±1.7 µm/s (n=987, P<0.0001) (Fig. S3F). Barely 1% of the cells displayed the control speed of 9 µm/s. At this stage, the population is composed of a mixture of trypanosomes that inherited either the flagellum that was assembled before RNAi or the short flagellum that was constructed after RNAi was initiated (Kohl et al., 2003). To discriminate the contribution of each population, immunofluorescence was carried out with the MAb25 axonemal marker. It revealed that ∼50% of the induced cells still possessed a flagellum longer than 10 µm (n=125). Given that cell tracking revealed that 99% of induced IFT140^RNAi trypanosomes failed to show wild-type motility movement, these data demonstrate that flagellar beating is significantly reduced in the absence of active IFT. As seen above, this is not due to visible structural defects in the peripheral doublet organization, dynein arms or central pair structure (Fig. 1). This motility phenotype could possibly be explained by discrete structural perturbations (dynein regulatory complex, radial spokes) or by modification of non-structural elements.

The motility phenotype suggests that other flagellar components could be altered in the absence of IFT. Proteomics studies revealed a large number of flagellar proteins that are not strongly associated to the axoneme or the PFR (Oberholzer et al., 2011; Subota et al., 2014). The distribution of three such proteins was investigated. First, the protein kinase A regulatory subunit (PKAR) has been shown to localize to the flagellum matrix and to contribute to flagellum motility (Oberholzer et al., 2011). Endogenous tagging has been used to determine the presence and location of PKAR. In non-induced IFT88^RNAi and IFT140^RNAi cells, the protein is probably located to the PFR, as shown by the slightly offset signal compared to the MAb25 axoneme marker (Fig. 5A,C). After 2 days of RNAi induction, the signal was decreased in the old flagella of IFT88^RNAi cells, a phenomenon accompanied by increased signal in the cytoplasm (Fig. 5B). In IFT140^RNAi cells, the signal for YFP::PKAR in the old flagellum as assessed by IFA using anti-GFP antibodies remained unchanged, but additional cytoplasmic signal was observed after induction (Fig. 5D). These results suggest that IFT could be involved in the distribution of PKAR between the cytoplasm and the flagellum.

We next investigated the location of the KIF9B protein, a kinesin molecular motor involved in the assembly of the PFR that is present along the axoneme and at the base of both old and new flagella (Demonchy et al., 2009). IFT88^RNAi and IFT140^RNAi cells were transfected to express a GFP::KIF9B fusion protein from its endogenous locus. Live-cell analysis revealed the presence of GFP::KIF9B at the flagellum base (circles on figures) and along the flagellum in both cell lines as expected (94% of cells, n=78) (Fig. 6A,C). Surprisingly, KIF9B was found concentrated at the distal tip of the old flagellum in the absence of IFT in both IFT88^RNAi and IFT140^RNAi cells (59% of cell, n=164) (arrowheads, Fig. 6B,D). The signal appeared as a nicely focused spot at the far end of the axoneme and was confirmed by IFA analysis with an anti-GFP antibody (not shown). We conclude that the correct distribution of KIF9B at the base of the flagellum depends on IFT in mature flagella.

FLAM8 is a protein concentrated at the distal tip of the axoneme whose amount increases during flagellum construction and after cell division (Subota et al., 2014). A rabbit anti-FLAM8 antibody confirmed the distal tip localization in the mature flagellum of control non-induced IFT88^RNAi and IFT140^RNAi cells (Fig. 7A,C). A different picture was obtained 24 h after triggering RNAi: the staining at the tip was less intense and had the tendency to stretch in the proximal direction, and multiple discrete spots appeared along the flagellum (arrowheads, Fig. 7B,D). One-dimensional projection of the flagellum based on the MAb25 marker confirmed the modified distribution pattern (Fig. 7, right panels). These results were reproduced with the anti-GFP antibody in IFT88^RNAi and IFT140^RNAi cells expressing YFP::FLAM8. The distal tip signal was reduced and shifted along the axoneme after 24 h of induction (Fig. S4). We conclude that FLAM8 requires IFT activity to be maintained at the distal tip of the flagellum after assembly. Overall, these results demonstrate that IFT is not necessary for flagellum length maintenance but is required for the correct distribution of several non-structural proteins.

Knockdown of any individual IFT gene by inducible RNAi inhibits the construction of the new flagellum but the old flagellum persists during the course of RNAi induction (Kohl et al., 2003). We show here, by a combination of TEM, IFA and video microscopy, that IFT is absent or arrested in the old flagellum (Fig. 8). This is explained if IFT proteins are slowly exchanged between the flagellar and the cytoplasmic compartment. In the case of anterograde mutants, a reduced import of IFT trains would not be sufficient to compensate for those that exit, leading to a progressive depletion of IFT. For retrograde mutants, a progressive imbalance would emerge between trains that get in and trains that can be recycled, leading to excessive IFT material that cannot traffic anymore.

This loss of IFT activity in mature flagella is not accompanied by flagellum shortening (Fig. 8). Although a reduction of the average flagellum length is observed in trypanosome populations where RNAi has been induced, it is actually explained by the emergence of daughter cells that assembled a new flagellum that was too short. Cells that inherited the long old flagellum remain present, but their proportion goes down progressively as they are diluted by daughter cells with a shorter flagellum.

To our knowledge, this paper is the first report of a situation where IFT is active in growing and mature flagella but is essential only for construction and not for maintenance of length. This could be explained if the flagellum does not undergo tubulin turnover, presumably because it is never disassembled (see below). Three other situations had been encountered so far in the literature regarding the presence and involvement of IFT in flagellum formation and maintenance. First, IFT is active in growing and mature flagella and is essential for construction and maintenance of length. This has formally been proven only in Chlamydomonas but often considered as the norm. This makes sense if the organelle undergoes tubulin turnover at its distal end as encountered in Chlamydomonas (Marshall and Rosenbaum, 2001) and in C. elegans (Hao et al., 2011b). Second, IFT is active and essential only for flagellum construction, but becomes absent once the organelle has matured. This has been reported recently for the flagellum of mouse spermatozoa. IFT proteins are highly abundant during early stages of flagellum construction but are not detected in mature spermatozoa (San Agustin et al., 2015) and inhibition of IFT following absence of the KIF3A motor has severe consequences on flagellum formation (Lehti et al., 2013). This situation is by no means universal for all spermatozoa: for example, substantial amounts of two subunits of the heterotrimeric IFT kinesin II are present in the flagellum of mature spermatozoa in sea urchin and sand dollar (Henson et al., 1997). Third, IFT is absent and obviously not required for flagellum construction when it takes place in the cytoplasm, as reported for Plasmodium (Briggs et al., 2004; Sinden et al., 1976) and Drosophila (Han et al., 2003; Sarpal et al., 2003). These multiple situations further highlight the need to use different model organisms to study the breath of diversity in cilia and flagella (Morga and Bastin, 2013).

The unique situation reported here for trypanosomes could be explained by closer examination of the biology of the flagellum in this organism. In Chlamydomonas and C. elegans, IFT is required to bring tubulin to compensate for the continuous turnover that takes place at the distal end of mature flagella (Craft et al., 2015; Hao et al., 2011b; Marshall and Rosenbaum, 2001). Two pieces of evidence indicate that there is little or no turnover in the skeletal elements of the mature flagellum of Trypanosoma brucei. First, tyrosinated tubulin, a marker of recently assembled tubulin, is only encountered at the tip of the new flagellum where immuno-gold staining or IFA displayed a clear gradient from tip to base but virtually no signal in the old flagellum (Sherwin and Gull, 1989b; Sherwin et al., 1987). This, however, cannot be investigated directly because tagged tubulin is not incorporated in the trypanosome flagellum (Bastin et al., 1996; Sheriff et al., 2014). Second, inducible expression of an epitope-tagged PFR2 protein, a major component of the PFR (a lattice-like structure associated to the axoneme) (Portman and Gull, 2010) has revealed strong incorporation at the distal region of the new flagellum with only a discrete incorporation all along the old flagellum without any polarity (Bastin et al., 1999). This could well explain the different consequences of the absence of IFT in new and old flagella. When the new flagellum is assembled, tubulin and other flagellar proteins require IFT to be transported to the distal tip; hence IFT inhibition blocks flagellum formation. In the old flagellum, tubulin does not need to be replaced and therefore the loss of IFT does not result in flagellum shortening. This absence of shortening could be due to the sophisticated architecture of the trypanosome flagellum that contains a PFR physically connected to the axoneme and that is attached along most of the length of the cell body via another cytoskeletal structure called the flagellum attachment zone (Sunter and Gull, 2016). Accordingly, when trypanosomes produce developmental stages with a shorter flagellum length, this is always achieved by mean of an asymmetric division with the construction of a shorter new flagellum and not by shortening an existing flagellum (Ooi and Bastin, 2013; Rotureau et al., 2011; Sharma et al., 2008).

If IFT remains active in trypanosome flagella after construction, it is likely to fulfil other functions than assembly. One that springs to mind is the maintenance of flagellum components. TEM images show that the structure of the axoneme and of the associated PFR appears normal in the old flagellum of both induced IFT88^RNAi and IF140^RNAi cells, what is confirmed at the molecular level by the unmodified distribution of hydin, AKB or PFR1 and PFR2. Similar results were obtained in Chlamydomonas using metabolic labelling where most axoneme proteins show limited turnover once the flagellum has reached its final length, although the actual contribution of IFT to this process could not formally be demonstrated (Song and Dentler, 2001). Investigation of a temperature-sensitive mutant for the IFT dynein heavy chain in this same organism revealed a twofold and fourfold reduction in anterograde and retrograde IFT frequency, respectively, but flagella maintained their length for ∼1 day (Engel et al., 2012), in contrast to the fla10 mutant where IFT is arrested and flagella are resorbed in 5 h. 2D-difference gel electrophoresis (DIGE) comparative analysis of the composition of such flagella with controls has revealed expected accumulation of several IFT proteins but also perturbations in the amounts of some membrane and matrix proteins, indicating likely functions for IFT in the control of their distribution in mature flagella (Engel et al., 2012).

However, the loss of flagellum beating in the absence of IFT indicates that these flagella are modified. We could identify modifications of the distribution of three non-structural proteins. The amount of the regulatory subunit of protein kinase A (PKAR) that is normally located in the matrix of the flagellum (Oberholzer et al., 2011) is reduced in the flagellum whereas its concentration in the cytoplasm increases. This is reminiscent of the Chlamydomonas aurora-like kinase (CALK) that translocates from the cell body to the flagellum during gamete activation (Pan and Snell, 2000). CALK is resolved as two bands on SDS-PAGE gels, both requiring the activity of the FLA10 kinesin II subunit, one for exclusion from the flagellum and the other one for efficient translocation (Pan and Snell, 2003). Here, we propose that the PKAR shuttles between the cytoplasm and the flagellum in a process that is dependent on IFT. The cytoplasmic and flagellar compartments are separated with a diffusion barrier at the transition zone. This gate would restrict the entry of cytoplasmic molecules on the basis of their size (Kee et al., 2012) and/or shape (Breslow et al., 2013). The PKAR protein (57 kDa) could be too large to enter by diffusion and would require energy to cross the gate. Preliminary FRAP analysis showed that flagellar YFP::PKAR diffuses within the flagellum without displaying typical IFT movement (L. Vincensini and P.B., unpublished data). This suggests that IFT could contribute to the concentration of PKAR in the flagellum, perhaps through a transient association to cross the barrier between the cytoplasm and the flagellar compartment, followed by release and free diffusion inside the flagellum. This can be compared with some membrane ciliary proteins in mammalian cells (Ye et al., 2013) or even some structural components such as DRC4 in Chlamydomonas (Wren et al., 2013). In the absence of IFT, the PKAR is still produced in the cytoplasm but it cannot be transported efficiently in the flagellum, hence resulting in an inversion of the flagellum-to-cytoplasm signal ratio.

FLAM8 is a large protein found at the distal tip of the axoneme whose amount increases during flagellum construction, reaching about half of the concentration found in the mature flagellum at the time of cell division (Subota et al., 2014). We propose that FLAM8 is an IFT cargo that uses anterograde transport to reach the tip where it remains apparently stationary. After cell division, IFT is still active and more FLAM8 can be transported until a plateau is reached when the flagellum becomes mature. In the absence of IFT, FLAM8 proteins present in the old flagellum relocate towards the base, perhaps by diffusion following its concentration gradient (Fig. 8). In a normal situation, anterograde IFT would return these proteins to the distal tip, hence contributing to both constitution and maintenance of the FLAM8 distal location. Such a principle could be applied to other distal tip proteins (Liu et al., 2010; Saada et al., 2014; Woodward et al., 1995) or IFT-independent mechanisms could also exist, as recently shown for the localization of EB1 at the end of flagellar microtubules in Chlamydomonas (Harris et al., 2016). The motor action of IFT could hence define a micro-domain at the distal end of the flagellum. This could be important given that trypanosomes swim with the flagellum forward and use the flagellum to attach to the epithelium of the salivary glands during infection in the tsetse fly (Tetley and Vickerman, 1985). Signalling molecules could also be concentrated depending on IFT (Eguether et al., 2014).

KIF9B is present along the axoneme and located at the base of both old and new flagella. In the absence of KIF9B, trypanosomes assemble an apparently normal axoneme but the construction of the PFR is defective: large ‘blocks’ of PFR materials alternate with regions where only the axoneme is present (Demonchy et al., 2009). It has been suggested that the motor activity of KIF9B somehow contributes to the transport of PFR proteins during construction of the structure. We propose that KIF9B walks towards the plus end of flagellum microtubules using its motor activity, perhaps transporting PFR components or other elements necessary for PFR assembly. Once KIF9B reaches the tip of the flagellum, it cannot return by itself and needs retrograde IFT to be recycled to the base. KIF9B would therefore be an IFT cargo but only for retrograde transport. In both mutants where retrograde transport is abolished (directly in IFT140^RNAi and because of lack of trains in the first place in the IFT88^RNAi strain), KIF9B would still be able to reach the tip of the flagellum because this would rely solely on its own activity but would not be able to return, resulting in the fluorescent tip signal observed (Fig. 8). This process emerges relatively late (2–3 days after RNAi induction), possibly either because of slow displacement of KIF9B or because only a proportion of KIF9B is motile. This is compatible with the absence of GFP::KIF9B movement in short video acquisitions (Demonchy et al., 2009).

Overall, these data demonstrate that IFT is necessary to create and maintain micro-domains within the flagellum by concentrating or removing specific proteins. IFT activity and association to various cargoes therefore could allow the generation of molecular diversity inside cilia by concentrating (or excluding) sensors, structural elements or other motors. This is relevant for multicellular organisms that produce several types of cilia but also in protists that assemble different types of flagella either in the same cell (Dawson and House, 2010) or in different life cycle stages (Gluenz et al., 2010; Rotureau et al., 2012). In addition to protein distribution, IFT could contribute to the transport of other molecules. The lipid composition of flagella appears very different from the rest of the cell membrane (Serricchio et al., 2015) and the fact that IFT trains are in tight contact with the flagellar membrane (Absalon et al., 2008; Kozminski et al., 1993) is intriguing. In conclusion, IFT appears to be a key player in the composition and maintenance of cilia and flagella, in addition to its well-characterized role in construction. Investigating these functions in different species promises to reveal exciting new features.

Cell lines used for this work were derivatives of T. brucei strain 427 and cultured in SDM79 medium supplemented with hemin and 10% fetal calf serum (Brun and Schonenberger, 1979). The IFT88^RNAi (Kohl et al., 2003) and IFT140^RNAi (Absalon et al., 2008) cell lines have been described previously. These cell lines express complementary single-stranded RNA corresponding to a fragment of the IFT88 or the IFT140 gene from two tetracycline-inducible T7 promoters facing each other in the pZJM vector (Wang et al., 2000) integrated in the rDNA locus of 29-13 cells that express the T7 RNA polymerase and the tetracycline repressor (Wirtz et al., 1999). Addition of tetracycline (1 µg/ml) to the medium induces expression of sense and anti-sense RNA strands that can anneal to form double-stranded RNA (dsRNA) and trigger RNAi.

Genecust Europe (Luxembourg) synthesized the insert PKAR (GeneDB number Tb927.11.4610), RSP9 (Tb927.8.810) and AKB (Tb927.10.830) that were then cloned in the p2675 plasmid allowing YFP tagging at the N-terminal end of proteins of interest (Kelly et al., 2007). The gene fragment PKAR (1–476 nucleotides relative to the start codon), RSP9 (1–496 nucleotides) and AKB (1–476 nucleotides) were digested with HindIII and ApaI, respectively, and cloned into the corresponding sites of the p2675 vector. Each p2675 construct was linearized with a specific enzyme allowing recombination within the PKAR (SgrdI), RSP9 (BbsI) or AKB (NsiI) gene. The eGFP::Hydin fusion contains 1–1200 bp of the hydin gene in frame downstream of the eGFP gene in the vector pPCPFReGFP (Adhiambo et al., 2009). In this vector, the fusion gene is flanked by PFR1 and PFR2 intergenic sequences. The vector was linearized by EcoRI and integrated in the hydin endogenous locus, resulting in a reconstructed complete eGFP::Hydin fusion. Plasmids p2675YFPIFT81 (Bhogaraju et al., 2013), pPCPFReGFPKIF9B (Demonchy et al., 2009) and p3329FLAM8 (Subota et al., 2014) have been described previously.

IFT88^RNAi and IFT140^RNAi cell lines were transfected with the above plasmids by Nucleofector technology (Lonza, Italy) as described previously (Burkard et al., 2007). Transfectants were grown in media with the appropriate antibiotic concentration and clonal populations were obtained by limiting dilution. Cell lines were characterized either by phase contrast or direct fluorescence observation (eGFP or YFP). Cell culture growth was monitored daily with an automatic Muse cell analyser (Merck Millipore, Paris).

Cultured parasites were washed twice in SDM79 medium without serum and spread directly onto poly-L-lysine-coated slides. The slides were air-dried for 10 min, fixed in methanol at −20°C for 5 min and rehydrated for 20 min in PBS (Sherwin et al., 1987). For immunodetection, slides were incubated with primary antibodies diluted in phosphate-buffered saline (PBS) with 0.1% bovine serum albumin (BSA) for 1 h at room temperature or at 37°C in a humid chamber. Slides were washed in PBS and appropriate subclass-specific secondary antibodies coupled to Alexa Fluor 488 or Alexa Fluor 594 (Invitrogen), and Cy3 or Cy5 (Jackson ImmunoResearch, West Grove, PA) were added using a 1:400 dilution in PBS with 0.1% BSA. After 45 min incubation, slides were washed in PBS then 4′,6-diamidino-2-phenyl-indole (DAPI) (2 μg/μl) staining of the nucleus and the kinetoplast was performed. Slides were mounted using ProLong antifade reagent (Invitrogen). Samples were observed with a DMI4000 Leica microscope, and images were acquired with a Horca 03G (Hamamatsu, Hamamatsu City, Japan) camera. In the case of RNAi mutants, all IFA signals were normalized using the signal obtained in non-induced controls as a reference.

The monoclonal antibody MAb25 (IgG2a) (Pradel et al., 2006), which specifically recognizes the axoneme protein Trypanosoma brucei (Tb)SAXO1 (Dacheux et al., 2012), was used as flagellum marker. The IFT complex B IFT172 protein was recognized using anti-IFT172 mouse monoclonal antibody diluted 1:200 (Blisnick et al., 2014). GFP was observed directly or upon fixation by immunofluorescence using a 1:200 dilution of a rabbit anti-GFP antibody (cat. no. AG455, Invitrogen). FLAM8 was detected using a rabbit polyclonal serum (1:1000 dilution) that was kindly provided by Paul McKean (University of Lancaster, Lancaster, UK).

GFP and YFP visualization on live trypanosomes was performed as described previously (Buisson et al., 2013). IFT videos were acquired using a Zeiss inverted microscope (Axiovert 200; Jena, Germany) equipped with an oil immersion objective (magnification 100× with numerical aperture 1.4) and a spinning disk confocal head (CSU22; Yokogawa, Tokyo, Japan). Images were acquired using Volocity software with an electron-multiplying charge-coupled device camera (C-9100; Hamamatsu) operating in streaming mode. A sample of parasites (6×10⁶ to 8×10⁶ cells/ml) was taken directly from the culture and trapped between slide and coverslip. The samples were kept at ambient temperature and used for no longer than 30 min. Images were analysed using ImageJ (National Institutes of Health, Bethesda, MD). The NeuronJ plugin was used to measure the flagellum length. Kymograph analysis (Buisson et al., 2013) and cell tracking for motility (Brasseur et al., 2013) were performed exactly as described previously.

YFP::IFT81 expression in the IFT140^RNAi cell line was first observed directly with a DMI4000 Leica microscope using an EL6000 (Leica) or LED light source (Spectra X, Lumencor) for excitation to confirm correct protein expression and localization. For FRAP analysis, a spinning disk confocal head (UltraView Vox, Perkin Elmer) was used. Images were acquired using Volocity software with an EMCCD camera (ImagEM X2, Hamamatsu) operating in streaming mode. A parasite sample from induced IFT140^RNAi expressing YFP::IFT81 was taken directly from the culture grown at 6×10⁶ to 8×10⁶ cells/ml and trapped between slide and coverslip. The samples were kept at room temperature and a region of interest was defined within the flagellum. Images were recorded continuously for 5 s before bleaching and then recovery was monitored for 45 s by continuously recording videos with a 100 ms exposure.

For TEM, cells were fixed directly in medium during 10 min at room temperature in 2.5% glutaraldehyde (glutaraldehyde 25% stock solution, EM Grade, Electron Microscopy Sciences). Centrifugation was carried out at 4°C for 5 min at 580 g. The supernatant was discarded and the pellet was fixed in 2.5% glutaraldehyde and 4% PFA (paraformaldehyde 32% stock solution, EM Grade, Electron Microscopy Sciences) in 0.1 M cacodylate buffer (pH 7.2) and contrasted with OsO4 (2%) (Osmium tetroxide 4% Aqueous solution, Electron Microscopy Sciences) in cacodylate buffer. After serial dehydration with ethanol solutions, samples were embedded in Agar 100 (Agar Scientific, UK) and left to polymerize at 60°C for 2 days. Ultrathin sections (50–70-nm thick) were collected on Formvar-carbon-coated nickel grids using a Leica EM UC6 ultra-microtome and stained with uranyl acetate (2%, w/v) (Uranyl Acetate dehydrate, Electron Microscopy Sciences) and lead citrate (80 mM, buffer made in-house). Observations were made on a Tecnai BioTWIN 120 cryo electron microscope (FEI) and images were captured with a MegaView II camera (Arecont Vision, France) and processed with AnalySIS and Adobe Photoshop CS4 (San Jose, CA).

Cells were washed in PBS and pellets were boiled in Laemmli loading buffer before SDS-PAGE separation, loading 20 μg of total cell protein per lane. Proteins were transferred to polyvinylidene fluoride membranes overnight at 4°C under low voltage (25 V). Membranes were blocked with 5% reconstituted skimmed milk in PBS and 0.1% Tween 20 (PBST) and incubated with primary antibodies diluted in 1% reconstituted skimmed milk and PBST. A mixture of two anti-GFP monoclonal antibodies (cat. no. 11814460001, Roche, Life Science) was used to detect YFP and eGFP (1:200 dilution) whereas the anti-BIP was diluted 1:1000 (Bangs et al., 1993) and served as loading control. Three membrane washes were performed with PBST for 5 min. Species-specific secondary antibodies coupled to horseradish peroxidase (GE Healthcare) were diluted 1:20,000 in PBST containing 1% skimmed milk and incubated for 1 h. Detection was carried out by using an enhanced chemiluminescence assay and the imaging system Pxi (Ozyme, France).

We thank Brice Rotureau for help in quantitative analysis of cell motility. We acknowledge Derrick Robinson and Paul McKean for generous gifts of antibodies, and Raphaël Demonchy for preliminary observations of the KIF9B phenotype. We thank Jean-Yves Tinevez (Imagopole) and Gérard Pehau-Arnaudet (Ultrapole) for training and valuable advises for light and electron microscopy, respectively.