Chlamydomonas is a unicellular photosynthetic eukaryote with an ovoid cell body that possesses two flagella (approximately 12 μm in length each) found at the apical end of the cell (Figure 1A,B). Chlamydomonas reinhardtii is a well-established model organism for studying fundamental processes such as photosynthesis, motility, responses to light, and cell-cell recognition. Chlamydomonas displays numerous biological and technical advantages for the study of eukaryotic flagella. First, it can be grown synchronously and large amounts of flagella can be easily purified for biochemical analyses. Second, forward genetics allows the generation of many mutant strains 14 that can be easily crossed for exhaustive characterization. Furthermore, Chlamydomonas cells exhibit complex swimming behaviors in response to various light stimuli, allowing dissection of flagellar beating regulatory pathways. Finally, flagella are not essential for the survival of Chlamydomonas but play key roles in gamete recognition, hence permitting investigation of sensory processes.

Trypanosoma brucei is a kinetoplastid protist that possesses one flagellum (approximately 22 μm in length) attached along the length of the cell body (Figure 1A’,B’). It is well known for being responsible for sleeping sickness in Africa. It proliferates in mammalian blood and is transmitted via the bite of a tsetse fly where it undergoes a complex development in the midgut and the salivary glands 17 . The flagellum remains present during the whole cell cycle and throughout the life cycle. The most extensively studied stage of this parasite is derived from the peritrophic space of the insect vector midgut, and is called the procyclic stage. The cell is 20 to 25 μm long and 3 to 5 μm wide and exhibits a slightly helical shape. Trypanosoma are attractive models to study cilia and flagella since they can be easily cultivated in laboratories and are genetically tractable (RNAi, endogenous tagging, imaging, and so on).

The flagellum of Trypanosoma and Chlamydomonas is of the 9 + 2 configuration (Figure 1E,F,C’,D’), as confirmed by the conservation of components of the dynein arms, the central pair, or the radial spokes 18 19 . In Chlamydomonas, the flagella emerge at the surface through pores in the cell wall (Figure 1B), whereas in trypanosomes, the flagellum arises from a cell surface invagination called the flagellar pocket (Figure 1B’) that is also the unique site for endocytosis and exocytosis 20 . Nevertheless, there are some significant differences such as the presence of an extra-axonemal structure called the paraflagellar rod (PFR) in the Trypanosoma flagellum (Figure 1C’,D’). This unique structure is made of filaments finely organized and can be subdivided into three distinct regions called the proximal, intermediate, and distal domains 21 . The PFR is composed of unique proteins and is required for cell motility, although its actual contribution to flagellum beating remains enigmatic 22 23 24 . Some ultrastructural differences have been noted at the base of the flagellum: the transition zone is longer in trypanosomes (approximately 400 nm) (Figure 1B’) and does not contain the central structure encountered in Chlamydomonas (Figure 1B). Some differences have also been reported at the distal tip where the central pair of microtubules is linked to the membrane by a cap in Chlamydomonas 25 , whereas a discrete electron-dense structure more distant from the membrane is found in Trypanosoma 26 .

The coordination between cell division and flagellum formation is rather divergent between Chlamydomonas and Trypanosoma (Figure 2). In Chlamydomonas, a vegetative cell divides within a single cell wall to produce 4, 8, 16, or more cells from a unique parent. During mitosis flagella are resorbed and basal bodies ensure their function as centrioles, organizing the mitotic spindle 27 . Mitosis is closed, meaning that the nuclear membrane does not break down. Instead, microtubules cross the nuclear envelope through its pores to assemble the mitotic spindle. The new flagella are assembled once the cell has fully divided (Figure 2A). In the trypanosome cell division cycle, two distinct S phases need to be coordinated: one for the mitochondrial DNA contained within the kinetoplast (trypanosomes possess a single mitochondrion) and one for nuclear DNA (Figure 2B). The process begins with the S phase of the mitochondrial DNA immediately followed by basal body maturation and duplication 28 29 . A tripartite filament system connecting the duplicated DNA and a specific membrane region of the mitochondrion is duplicated and linked to the duplicated basal bodies 30 . The old flagellum remains in place and the new flagellum invades the flagellar pocket and connects its tip to the old flagellum by a transmembrane mobile junction called the flagella connector (FC) 31 32 . It has been proposed that the FC guides the positioning of the new flagellum. Then, mitosis occurs leaving one of the two nuclei positioned between the two kinetoplasts, to finally produce two ‘clusters’ of cytoplasmic organelles ready for division. Cleavage furrow ingression is unidirectional, from the anterior to the posterior end of the dividing cell, between the old and the new flagellum. The length of the new flagellum determines the point where cell cleavage initiates and hence the length of the daughter cell (Figure 2B) 33 .

The assembly of the flagellum is a huge commitment for the cell, as this requires the correct production and assembly of more than 500 proteins 18 19 34 , both in time (right moment of the cell cycle) and in space (in a defined compartment). Assembly of the axoneme 35 36 and also of the PFR 37 takes place at the distal end of the growing flagellum. Since the flagellum does not possess any ribosomes, all the components needed for its construction must first be synthesized in the cytoplasm and then imported into the flagellum before reaching the distal tip either by transport or by diffusion. In 1993, an active transport of ‘rafts’ was discovered within the flagellum of Chlamydomonas and termed intraflagellar transport (IFT) 38 . It was first observed by differential interference contrast (DIC) microscopy in paralyzed flagella of live cells. It was proposed that these rafts could correspond to electron-dense structures sandwiched between the flagellar membrane and the axonemal outer doublet B identified by electron microscopy in the late 1960s 39 . IFT was not observed in the thermosensitive Fla10 mutant 40 maintained at the restrictive temperature and the number of particles detected by electron microscopy dropped significantly, supporting the proposal that these electron-dense structures indeed correspond to the transported granules detected by DIC 41 . These were termed IFT particles and later renamed as IFT trains 42 . Fla10 is a kinesin motor member of the heterotrimeric kinesin-2 complex composed of two motor subunits (FLA10 and FLA8) and a kinesin-associated protein (KAP) possibly involved in cargo binding 43 . Immunogold experiments revealed that FLA10 localizes to the particles 41 .

This discovery raised the question of the identity of the molecules involved in this transport. Cole and co-workers were the first to purify the IFT particles from the flagellum matrix of Chlamydomonas using sucrose density gradients 44 45 . Two distinct complexes (A and B) were identified: the IFT-A complex is a 550-kDa tetramer containing at least five subunits of 144, 140, 139, 122, and 43 kDa, whereas IFT-B complex is a 750-kDa complex containing at least 11 subunits ranging from 20 to 172 kDa (Table 1) 45 . A metagenomic analysis revealed that most IFT genes are conserved in ciliated and flagellated species 1 46 with the exception of Plasmodium that assembles its flagella in the cytoplasm 47 . In most species, IFT proteins are found in the flagellum but are mostly concentrated at its base and also found in fairly high abundance in the cytoplasm 48 49 . Mutant analysis demonstrated that kinesin-2 is responsible for anterograde movement 41 , whereas retrograde trafficking is powered by a specific type of dynein motor 50 51 52 . IFT genes are conserved in all trypanosomatid genomes (Table 1) with the exception of the KAP that is missing 53 , suggesting that kinesin-2 is more likely to function as a homodimer as reported for the osm-3 kinesin in Caenorhabditis elegans, and not as a heterotrimer as observed in other species 54 . In contrast, the Chlamydomonas genome appears not to contain osm-3 homologues, indicating that only heterotrimeric kinesin-2 is present in this organism.

Proteins have been grouped according to their category. When inhibition of a given gene blocked flagellum formation, it was described as an anterograde phenotype (A) and when it stopped retrograde transport, resulting in the formation of shorter flagella filled with IFT material, it was described as a retrograde phenotype (R). Ambiguous phenotypes are shown with a question mark and a hyphen indicates that the gene is conserved but has not been studied in that organism. A cross indicates that the gene is missing from the genome. Experiments were carried out by RNAi knockdown except where indicated. The asterisk denotes an insertion mutant whereas names of the mutant gene obtained by forward genetics are shown in italics

IFT plays a key role in the construction of the flagellum as its inactivation blocks flagellum formation in all species studied so far. Inactivation of any single IFT gene is sufficient to inhibit flagellum assembly, indicating that the integrity of the particle is required for efficient IFT. This is supported by many experiments using mutant, RNAi, or knockout approaches in different organisms: Chlamydomonas 41 , mouse 55 , C. elegans 56 , Tetrahymena 57 , Trypanosoma 33 , zebrafish 58 , Leishmania 59 , and Xenopus 60 .

The currently accepted model for IFT mostly relies on studies of Chlamydomonas 61 and is summarized in Figure 3. First, IFT-A and IFT-B complexes, kinesin-2, cytoplasmic dynein 2, and axonemal precursors are produced in the cytoplasm and gather at the flagellum base. Second, once inside the flagellum, the active kinesin-2 transports IFT-A and IFT-B complexes, inactive IFT dynein, and axonemal precursors from the base of the flagellum to the tip. Third, kinesin-2 reaches the distal end of the B microtubule, where axonemal cargo proteins and IFT particles are released into the ciliary tip compartment. Following remodeling of the IFT train, complex A binds to the active IFT dynein. Fourth, the IFT-B complex associates with the IFT-A complex, and the active IFT dynein transports all components, including kinesin-2 back from the tip to the cell body. The IFT cycle is completed when IFT components are returned at the base of the flagellum, where they can be recycled or targeted for destruction.

Producing at least 500 proteins at about the same time and same place is a sophisticated feat of engineering. This includes proteins constituting the basal body, the transition zone, the IFT particles, and the axoneme (and the PFR in trypanosomatids), as well as membrane elements. In Chlamydomonas, deflagellation stimulates the transcription of all flagellar genes 62 . This is accompanied by a stimulation of synthesis of flagellar proteins through an increase in the level of translatable mRNAs. The stimulation of the production of mRNA could be related to the presence of response elements called ‘tub boxes’ found in the promoter region of several flagellar genes 63 . Monitoring of some IFT proteins during the normal cell cycle in Chlamydomonas has brought more insight about the timing of this process 64 . After cell synchronization, it was found out that transcripts for IFT27, IFT46, IFT140, and FLA10 were upregulated during the S/M phase before the construction of the flagellum (Figure 4). Another study has shown that mRNAs for tubulin and other axonemal components such as radial spoke and outer or inner dynein arms were overexpressed during flagellar regeneration 65 .

Until recently, little was known about the expression of mRNA encoding flagellar proteins in T. brucei due to difficulties of reliably synchronizing cells in culture. However, a recent study overcame this limit, allowing the investigation of the gene expression profile during the cell cycle 66 . Procyclic T. brucei cells were collected in log phase and treated by elutriation in order to separate cells by density and size. The larger cells were collected and placed for 1 hour in culture, and then a second elutriation centrifugation was performed in order to select the smallest cells that had just divided. These were returned in culture and proceeded with good synchronization through the cell cycle that was completed in 9 hours 66 . An RNA-seq profile using Solexa sequencing (Illumina, San Diego, CA, USA) was performed at four stages: early G1 (cells with one flagellum), late G1 (maturation and duplication of the basal body), S phase (construction of the new flagellum), and G2/M phase (elongation of the new flagellum). A total of 546 genes showed cell cycle-dependent fluctuations, with peaks at specific time points (Figure 5). Since many of them encode components of flagellum structures, we analyzed the list in detail and grouped genes according to their relation to basal body, IFT, membrane and matrix, axoneme, and PFR (Additional file 1: Table S1). Analysis of gene expression profiles revealed that most of the basal body mRNAs were upregulated when the basal body was duplicating (late G1). This phase was preceded by a peak in mRNA for IFT and membrane proteins prior to the initiation of flagellum construction. The mRNAs for axoneme components were mainly produced when the new flagellum elongates, whereas the PFR mRNA increased later on (Figure 5). This is consistent with the fact that this structure is the last one to be assembled in the flagellum 67 . Therefore, the profile of mRNA production correlates with the successive steps of flagellum formation, suggesting that trypanosomes produce the right amount of transcripts exactly when needed. Protein translation is expected to follow the RNA dynamics of production although direct evidence is missing. These observations are in agreement with the fact that the actual amount of flagellar proteins available in the cytoplasm is very low 23 . This situation is quite different from Chlamydomonas where a pool of non-assembled material is available in the cytoplasm and sufficient to support construction of two half-length (or one full-length) flagella 68 .

Construction of the flagellum follows a strict hierarchy: maturation of the basal body, docking to the membrane, formation of the transition zone, and then elongation of the axoneme. In trypanosomes, the first detectable event in the cell cycle is the maturation of the probasal body that elongates and docks to the membrane via the transition fibers, becoming competent for nucleating the new flagellum. This maturation process is concomitant with the formation of a new probasal body alongside each mature basal body 28 . Such a cell possesses two basal body complexes assembled at the same time, but the mature basal body, which bears the old flagellum, is always at least one generation older than the one that bears the new flagellum. In Chlamydomonas, the existing flagella are disassembled at mitosis, but old and new basal bodies differ slightly in their protein composition. The docking of the basal body is independent of IFT since it takes place apparently normally in all IFT mutants 16 .

At the early stage of flagellum formation, a large amount of electron-dense material is observed by transmission electron microscopy (TEM) in the short flagellum of both Chlamydomonas and Trypanosoma, prior to microtubule elongation (Figure 6). The identity of this material remains to be determined. It could correspond to tubulin and other axoneme precursors before their assembly or to IFT material. This hypothesis is supported by immunofluorescence assays in Trypanosoma showing a bright signal for IFT proteins in short flagella before axoneme markers can be detected (T Blisnick, unpublished data). Similarly, immunofluorescence assay (IFA) with an anti-IFT52 antibody and live microscopy analysis of GFP::IFT27 expressing cells data show that a high concentration of IFT protein is present at early stages of flagellum formation in Chlamydomonas 36 69 . Immunoelectron microscopy indicates that IFT52 is associated with the periphery of transitional fibers, which extend from the distal portion of the basal body to the cell membrane and demarcate the entry of the flagellar compartment 70 .

There is very little information about the way IFT trains are assembled within the flagellum. The IFT-A and IFT-B complexes can be purified from cell bodies in Chlamydomonas suggesting that they are preassembled in the cytoplasm 71 . In Trypanosoma, as in Chlamydomonas, IFA or biochemical fractionation indicate that a large amount of IFT protein is present in the cytoplasm 48 49 . Quantification experiments revealed that the cell body contains up to 50-fold more IFT material than the flagellum 48 . Nevertheless, conventional IFT trains have never been visualized elsewhere other than in the flagellum compartment. In Trypanosoma, GFP::IFT52 is found at the flagellum base and traffics in the flagellum, but is also very abundant in the cytoplasm. Photobleaching an area of the cytoplasm resulted in rapid recovery but no train movement could be detected (J Buisson, unpublished data).

We propose that IFT train formation takes place when the local concentration of IFT complexes is sufficiently high. This might only be achieved at the very early phase of flagellum formation, when IFT proteins appear to be highly concentrated in the short flagellum (Figure 6). Dynein is not required for this process since long trains can be incorporated in the short flagellum of the fla14 mutant, that present a mutation in a dynein light chain 42 . In the future, it will be interesting to produce in vitro the IFT complexes 72 73 and the two different motor complexes, to monitor their ability to constitute trains according to their respective concentration and the nature of the environment.

In vivo visualization of IFT particles is essential to understand the mechanisms responsible for flagellum growth and maintenance. This can be achieved by two methods: direct observations by DIC (so far only achieved in Chlamydomonas) and by the use of IFT proteins fused to fluorescent markers such as GFP. When viewed in DIC, IFT trains in Chlamydomonas leave traces on kymographs that appear twice as large (0.12 μm) for anterograde trains compared to retrograde ones (0.06 μm) 74 . However, these should be viewed as approximations given the limited resolution of light microscopy. More recent electron tomographic analysis of IFT trains in situ 42 discriminated two populations. An electron-opaque population gathers around a 250 nm size and exhibits an approximate 16 nm periodicity, and a less electron-opaque type with a 700 nm mean length presents an approximate 40 nm periodicity. Longitudinal sections of fla14 flagella only showed long trains of low electron density and/or presenting a 40 nm periodicity 42 51 . Therefore, long trains are likely to correspond to anterograde particles and short trains represent the retrograde IFT trains. Hence, DIC appears to underestimate the actual size of the IFT trains. In Trypanosoma, analysis of the traces left by GFP::IFT52 from live cells suggests that the trains are at least 400 nm in length on the anterograde transport direction and 250 nm in the retrograde direction. These must be considered as approximations due to the limited resolution of light microscopy and the relatively long exposure time 75 . In the original publication, Kozminski and co-workers reported IFT rates of 2.0 μm/s ^-1 in the anterograde direction and 3.5 μm/s ^-1 in the retrograde direction. However, some variability has been observed between different experiments (Table 2). In Trypanosoma, IFT is sensitive to temperature (Table 2). Hence, fluctuations in the reported IFT speed could be related to experimental conditions, particularly because it is difficult to regulate or measure temperatures when observing IFT with oil immersion lenses and high intensity illumination.

Asterisks indicate the population of slower anterograde trains. DHC, dynein heavy chain; DIC, differential interference contrast; IFT, intraflagellar transport; KAP, kinesin-associated protein; nd, no data.