Microtubule doublets are double-track railways for intraflagellar transport trains

See allHide authors and affiliations

Science  06 May 2016:
Vol. 352, Issue 6286, pp. 721-724
DOI: 10.1126/science.aaf4594

Stay on the right track

Cilia contain a well-ordered array of microtubule doublets along their length. A longstanding question in cilium structure and function is why the microtubule arrangement in cilia is so complex. Stepanek and Pigino developed a time-resolved correlative fluorescence and three-dimensional electron microscopy method to show that the doublets provide directionality to intraflagellar transport. One microtubule in the pair moves cargoes up to the ciliary tip. Meanwhile, the other microtubule moves cargoes back to the cell body. These results explain why the axoneme is built out of microtubule doublets and suggest a mechanistic picture of how the logistics of bidirectional intraflagellar transport are regulated.

Science, this issue p. 721


The cilium is a large macromolecular machine that is vital for motility, signaling, and sensing in most eukaryotic cells. Its conserved core structure, the axoneme, contains nine microtubule doublets, each comprising a full A-microtubule and an incomplete B-microtubule. However, thus far, the function of this doublet geometry has not been understood. We developed a time-resolved correlative fluorescence and three-dimensional electron microscopy approach to investigate the dynamics of intraflagellar transport (IFT) trains, which carry ciliary building blocks along microtubules during the assembly and disassembly of the cilium. Using this method, we showed that each microtubule doublet is used as a bidirectional double-track railway: Anterograde IFT trains move along B-microtubules, and retrograde trains move along A-microtubules. Thus, the microtubule doublet geometry provides direction-specific rails to coordinate bidirectional transport of ciliary components.

The cilium is a conserved organelle that plays a fundamental role in signaling, sensing, and motility. Cilia have a complex microtubule-based structure that is not found in other cellular compartments. This structure includes nine peripheral microtubule doublets, each comprising a complete A-tubule and an incomplete B-tubule. However, the function of this distinctive conserved geometry has so far been unknown.

In addition to serving as platforms for periodically arranged axonemal proteins and protein complexes, ciliary microtubule doublets function as railways for intraflagellar transport (IFT), the process required for the assembly and disassembly of the cilium (1). Large protein complexes, known as IFT trains (2, 3), rapidly traverse up and down the cilium to move ciliary building blocks between the cell body and the ciliary distal tip (47), where the assembly of the cilium takes place. Electron microscopy (EM) shows that IFT trains move along the doublets while keeping contact with the ciliary membrane (2, 3, 8). It has thus far been unclear, however, how the IFT machinery is organized to avoid collisions between anterograde and retrograde trains.

In trypanosome flagella, anterograde IFT trains move at different speeds and can approach each other and fuse, similar to trains shunting, suggesting that anterograde trains might travel on a restricted set of axonemal microtubule doublets (9). No such interactions are observed between retrograde trains, nor do trains moving in opposite directions collide. To further analyze the dynamics and potential interactions of IFT trains in cilia, we imaged green fluorescent protein (GFP)–tagged trains in Chlamydomonas cells by means of total internal reflection fluorescence (TIRF) microscopy (fig. S1 and movies S1, S2, and S3). As expected, we observed anterograde-anterograde and retrograde-retrograde train interactions. Typically, a faster train caught up with a slower one moving in the same direction and both would then progress together at the same speed (fig. S1 and movies S2 and S3). Thus, it appears that IFT trains moving in the same direction often share the same microtubule rail. We never observed collisions between trains that traveled in opposite directions (supplementary text). Because oppositely directed trains passed by each other without changes in the direction of motion (fig. S1), a mechanism must exist to prevent collisions.

Although it is possible that trains could switch to another microtubule when they encounter a train moving in the opposite direction, this would be predicted to cause deceleration, which was not evident in our kymographs (see also the supplementary text). In trypanosomes, IFT trains avoid microtubule doublets that contact the paraflagellar rod (10), raising the possibility that IFT trains can recognize specific microtubule doublets. This could allow anterograde and retrograde trains to use different axonemal microtubules. However, the resolution required to test this hypothesis cannot be achieved with TIRF microscopy. Only EM offers sufficient resolution to see detailed interactions between IFT trains and microtubules. Previous EM work, comparing IFT trains from flagella of wild-type cells and from those of mutant cells with defective IFT, has suggested that anterograde and retrograde trains can be distinguished on the basis of ultrastructure and size (8). Long and short IFT trains have been proposed to be associated with anterograde and retrograde transport, respectively (8). However, the lack of dynamic information has precluded a definitive interpretation of the directionality of IFT trains.

To overcome the technical limitations of static EM, we developed a correlative light and EM (CLEM) approach. Because the currently available CLEM techniques (11) do not provide sufficient spatiotemporal resolution to analyze IFT trains, which move at 2.5 to 4 μm/s, we developed a novel method that allows millisecond resolution in correlative TIRF and three-dimensional (3D) EM (Fig. 1 and fig. S2). Briefly, a Chlamydomonas cell in gliding position was rapidly fixed with glutaraldehyde during time-lapse TIRF imaging of GFP-tagged IFT trains in the cilium (Fig. 1, B and C, and movie S4). In this way, the position of each fixed IFT train could be cross-referenced to its direction of movement just before fixation (Fig. 1D). In addition to anterograde trains (shown in green in Fig. 1D) and retrograde trains (shown in magenta in Fig. 1D), we also observed trains that did not move in the cilium during TIRF acquisition (shown in yellow in Fig. 1D). The same cell was then prepared for electron tomography, using a modification of the flat embedding protocol (8, 12). A reconstruction of the 3D structure of the entire cilium was then obtained by stitching together several tomograms (Fig. 1E). This provided an overview of the complete cilium and offered sufficient spatial resolution to observe the structural details of each train and its interaction with microtubules. Last, correlation of light microscopy (LM) movies and electron tomograms (materials and methods and fig. S2) allowed us to identify anterograde and retrograde trains unambiguously in a complete flagellum (Fig. 1, F and H) and describe their ultrastructure.

Fig. 1 Time-resolved CLEM of complete flagella.

(A) Diagram of a Chlamydomonas cell with flagella in the gliding position. The area in the rectangle is enlarged in (G). (B) Kymograph showing fixation during live-cell imaging of Chlamydomonas (IFT27-GFP strain). Magenta arrowheads indicate the time point of glutaraldehyde addition. (C) TIRF microscopy image of the positions of IFT trains after fixation. (D) Each IFT train was color-coded according to the direction of movement: anterograde, green; retrograde, magenta; and standing, yellow. Intensity and contrast were adjusted separately for each fluorescent particle. (E) Longitudinal section through the cilium 3D reconstruction, assembled from 12 tomograms. (F) Overlay of fluorescence and EM images on the section shown in (E). The area in the rectangle is enlarged in (H). (G) Diagram of anterograde and retrograde trains traveling along microtubule doublets. The plane of the virtual slice in (H) is marked. (H) Virtual slice through the tomogram, containing an anterograde IFT train (m, membrane; mtd, microtubule doublet; cp, central pair microtubule; ift, IFT train). Throughout, plus signs indicate flagellar tips. EM virtual slice thickness, 7 nm; vertical scale bar, 2 s (B); horizontal scale bars, 2 μm [(B) to (F); the scale bar in (C) also applies to (B)] and 50 nm (H).

In contrast to previous studies (8), our CLEM analysis revealed that anterograde and retrograde trains are of similar length (Fig. 2, B and D, and fig. S3, A to H). We measured a mean length of 233 nm (SD = 73 nm, n = 50) for anterograde trains and 209 nm (SD = 41 nm, n = 27) for retrograde trains (Fig. 2 and fig. S4). Nevertheless, anterograde and retrograde trains could be distinguished by their distinctive appearance (Fig. 2, A to D, and fig. S3, A to H). The morphology of the anterograde trains corresponded to that of those previously described as short trains (8) and has often been observed in electron micrographs of IFT trains (2, 13). The anterograde trains appeared as compact electron-dense structures with clearly defined boundaries. In most of the 3D reconstructions of anterograde IFT trains, the main body of the train had repeating units, possibly IFT complexes, with a periodicity that ranged between 8 and 16 nm (Fig. 2B and fig. S3A). A characteristic feature of these trains was a straight plate-like structure at the interface between the train and the microtubule (Fig. 2, A and B; fig. S3, A to D; and movie S5). The structure of the retrograde trains appeared less condensed and less regular (Fig. 2, C and D; fig. S3, E to H; and movie S6), although in some tomograms, we observed repeating structures (fig. S3, E and F) with a similar periodicity to that previously reported for so-called long trains (8). The ultrastructure of these trains has not been reported in the literature, probably because it is not distinctive enough to be recognized as related to IFT without a CLEM approach.

Fig. 2 Morphology of anterograde, retrograde, and standing trains.

(A) Anterograde train in cross-sectional view and (B) longitudinal view. (C) Retrograde train in cross-sectional view and (D) longitudinal view. (E) Standing train in cross-sectional view and (F) longitudinal view. Arrowheads show the borders of IFT trains. Scale bars, 50 nm [(A) to (E)] and 100nm (F). Virtual slice thicknesses, 105 nm [(A), (C), and (E)], 70 nm [(B) and (D)], and 10 nm (F). Cross-sectional views [(A), (C), and (E)] are shown in proximal-to-distal orientation. Proximal (–) and distal (+) regions [(B), (D), and (F)] are indicated in (B). The position of the glass slide is indicated in (A).

We also identified another class of IFT train, which differed from anterograde and retrograde trains in both motility and morphology. These static trains appeared as bright standing particles along the cilium in TIRF microscopy (Fig. 1, B to D) and were 650 nm long in EM images (SD = 106 nm, n = 9) (fig. S4), with a regular 40-nm periodicity (Fig. 2F; fig. S3, I to K; and movie S7). Their ultrastructure and size clearly resembled the morphology of the long trains described previously (8). Our findings can thus be used to classify anterograde, retrograde, and standing IFT trains based on the correlation of ultrastructure and dynamics.

Next, we addressed the hypothesis that anterograde and retrograde trains avoid collisions by using different microtubule doublets. We first numbered the doublets (14) and then analyzed the positions of IFT trains in the flagellum. Anterograde and retrograde trains could share the same microtubule doublet (Fig. 3, A and I), leaving open the question of how collisions between trains that move in opposite directions are avoided. To answer this question, we carried out a more detailed analysis of the localization of the trains on the microtubule doublet. Only anterograde trains were associated with the B-microtubule (Fig. 3, A, B, D, E, F, and I, and fig. S5), and only retrograde trains were associated with the A-microtubule (Fig. 3, A, C, D, G, H, and I, and fig. S5). To confirm these observations and obtain more detailed 3D images of the connections between IFT trains and microtubules, we performed the same CLEM technique on cross sections of the cilium. With this approach, we observed clear connections (possibly representing the positions of the motors) between A-microtubules and retrograde trains and between B-microtubules and anterograde trains (Fig. 3, E to H). Thus, collisions between anterograde and retrograde IFT trains are prevented by direction-specific usage of the two microtubules in a doublet (fig. S6).

Fig. 3 Anterograde and retrograde trains use different microtubules in the same doublet.

(A) Segmentation of an axoneme, showing anterograde and retrograde trains moving simultaneously on microtubule doublet 9. IFT trains are not shown on doublets 2 to 8 for clarity. (B) Average of 50 anterograde train positions on the doublet. The A- and B-microtubules are indicated. (C) Average of 27 retrograde train positions on the doublet. (D) Overlay of (B) and (C), showing the position of anterograde (green) and retrograde (magenta) trains with respect to the microtubules of the doublet. (E to I) Virtual slices through a cross-sectional tomogram, showing [(E) and (F)] anterograde trains (black arrowheads) connecting to the B-microtubule, [(G) and (H)] retrograde trains (black arrowheads) connecting to the A-microtubule, and (I) anterograde trains (green arrowheads) and retrograde trains (magenta arrowheads) stopped next to each other on the same doublet. Scale bars, 250 nm (A), 25 nm [(B) to (H)], and 50 nm (I). Virtual slice thickness, 7 nm. Cross-sectional views [(B) to (I)] are shown in proximal-to-distal orientation.

It is well known that anterograde and retrograde IFT trains use different molecular motors to move along microtubules. Whereas retrograde trains use dyneins, anterograde trains use kinesin-II motors (3, 15, 16). Our data therefore suggest that these specific motors must recognize A- and B-microtubules, respectively. One possibility could be that they recognize differences in the arrangement of α- and β-tubulin. However, this possibility seems to be ruled out by recent findings that A- and B-microtubules have the same tubulin arrangement, know as a B-lattice (17). An alternate possibility could be that IFT trains recognize specific posttranslational modifications on A- versus B-microtubules, similar to the mechanisms controlling the selective transport of cargos along axonal versus dendritic microtubules in mammalian axons (18). The human IFT anterograde motor kinesin-II has exhibited a differential response to posttranslational modifications of tubulin in a reconstituted in vitro system (19). Furthermore, differences have been reported in the posttranslational modification of A- and B-microtubules (20). This might contribute to the efficiency of IFT by providing optimized tracks for kinesin and dynein motors in addition to the IFT train segregation. Nevertheless, the effect of such modifications on the control of IFT in vivo remains unclear. Alternatively, a specific molecular machinery to direct trains to the correct microtubule could be present at the base and tip of the axoneme. More complex motor regulation could be necessary in the sensory cilia of Caenorhabditis elegans, where the B-microtubules do not extend beyond the middle segment of the flagellum. There, the kinesin-II moves only in the middle segment, and an additional motor, OSM-3, is required to bring cargo to the tip along the A-microtubules (21, 22). Regardless of the mechanisms involved in the recognition of the microtubules by the motors, our work highlights the critical role played by microtubule doublets in the assembly of cilia.

Supplementary Materials

Materials and Methods

Supplementary Text

Figs. S1 to S6

References (2330)

Movies S1 to S7

References and Notes

  1. Acknowledgments: We thank J. Rosenbaum and D. Diener for providing the Chlamydomonas IFT27-GFP strain; P. Kiesel, the MPI-CBG EM facility, J. Meissner, the MPI-CBG LM facility, V. Geyer, and A. Kopach for technical support; I. Patten for comments and corrections to the manuscript; and J. Howard, S. Diez, J. Brugués, T. Hyman, M. Zerial, P. Tomancak, F. Jug, and I. Sbalzarini for fruitful discussion and comments. This work was supported by the Max Planck Society and a fellowship of the Dresden International Graduate School for Biomedicine and Bioengineering (GS97), granted by the German Research Foundation to L.S. Supporting data are provided in the supplementary materials.
View Abstract

Navigate This Article