Structure of Trypanosoma brucei flagellum accounts for its bihelical motion

Abstract

Trypanosoma brucei is a parasitic protozoan that causes African sleeping sickness. It contains a flagellum required for locomotion and viability. In addition to a microtubular axoneme, the flagellum contains a crystalline paraflagellar rod (PFR) and connecting proteins. We show here, by cryoelectron tomography, the structure of the flagellum in three bending states. The PFR lattice in straight flagella repeats every 56 nm along the length of the axoneme, matching the spacing of the connecting proteins. During flagellar bending, the PFR crystallographic unit cell lengths remain constant while the interaxial angles vary, similar to a jackscrew. The axoneme drives the expansion and compression of the PFR lattice. We propose that the PFR modifies the in-plane axoneme motion to produce the characteristic trypanosome bihelical motility as captured by high-speed light microscope videography.

Trypanosoma brucei has devastated the African continent for centuries by infecting humans and domestic animals and has hindered economic development in sub-Saharan Africa (1). Current sleeping sickness treatments are inadequate and the drugs used are highly toxic (2). In recent years, the motility of the T. brucei flagellum has been found to be essential for parasite survival, infection, and disease pathogenesis (3), and has emerged as a promising drug target (4). Flagella with similar structural organization and protein composition have also been found in euglenoids (5) and other kinetoplastid parasites including Leishmania spp. and Trypanosoma cruzi, which cause Leishmaniasis and Chagas disease, respectively (6).

The trypanosome flagellum is more complex than most other eukaryotic microtubule-based flagella (7–9) and is completely different from rotary-motor based bacterial flagella (10). Each T. brucei cell contains one flagellum that moves the cell body in an alternating right and left-handed twist resulting in bihelical motion (11) (Movie S1). The membrane-enclosed flagellum, composed of an axoneme, a paraflagellar rod (PFR) (12), and connecting proteins, is attached to the cell body (Fig. 1). PFR was identified as a lattice-like ultrastructure in T. brucei flagellum (13). This periodic and crystalline nature of the PFR was confirmed in T. brucei (14) and related species (15, 16). Monoclonal antibody screens (17) and proteomics studies (18–20) have identified at least 40 PFR proteins. Among them, PFR1 (73 kDa) and PFR2 (69 kDa), containing coiled-coil regions (21), are major structural components of the PFR (22). Depletion of these proteins results in failure of PFR assembly and cell motility defects (17, 23) (Fig. S1 and Movie S2). In the T. brucei pathogenic bloodstream form, ablation of PFR2 causes death of the parasite (18). These results demonstrate a critical role of the PFR in T. brucei motility and viability. We have employed cryoelectron tomography (cryo-ET) to determine the structure of a biochemically isolated T. brucei flagella (18). We describe here a model that explains how the structure and arrangement of the flagellar components produces the bihelical motion of the flagellum.

Fig. 1.

Trypanosoma brucei flagellum. (A) Diagram of a trypanosome cell (blue) with attached flagellum (yellow). (B) Slice of a tomographic reconstruction of an isolated straight flagellum. The flagellum, composed of an axoneme, a PFR (12), and connecting proteins, is attached to the cell body.

Fig. 1B is a projection through 30 slices (33 nm) of a tomogram of the T. brucei flagellum showing three components of a straight flagellum: the crystalline PFR, the axoneme, and the proteins connecting the PFR to the axoneme. The crystallinity of the PFR is demonstrated by the computed diffraction pattern of a single image (Fig. S2), which shows diffraction spots expected for a crystal. In other tomograms, we have observed flagella in bent conformations, similar to those observed during flagellar motion. Because these three flagellar components (PFR, axoneme, and connecting proteins) exhibit different periodicity and symmetry, we had to use different strategies to align and average each of them separately (Materials and Methods). The PFR has been found to contain multiple regions (24). We observe these distinct regions of the PFR in unaveraged, raw tomogram cross-sections (Fig. S3C and Movie S3). However, in our method of alignment, the crystallinity of the largest and most well-ordered portion of the PFR (the distal portion) strongly influences the overall PFR average. The proximal PFR region was not observed after the averaging because it does not possess such crystallinity.

The final averaged structure of the entire flagellum (Fig. 2A and Movie S3) was reconstituted from the averaged components of straight flagella.

Fig. 2.

T. brucei flagellum components. (A) Flagellum cross-section. The axoneme is radially colored: central pair complex (yellow), radial spokes (light green), microtubule doublets (blue). The dashed red line (bisecting the central pair) represents the plane of bending of an axoneme. The PFR is offset from that plane (dashed purple line through the middle of the PFR). (B) The diagonal (shown by a horizontal white arrow) of the crystallographic unit cell of the PFR (green) repeats with the same spacing (56 nm) as that of the connecting proteins (red and pink) along the axoneme axis. (C) The layered nature of the PFR. Skeletonized (27) PFR density is shown in gray, the crystallographic unit cells are shown as green parallelograms, the connecting proteins are represented as spheres, and the microtubules are cylinders. (D) T. brucei axoneme internal components.

The averaged PFR density was derived from five different tomograms (Fig. 2 A and B). The PFR is a three-dimensional protein lattice that has crisscrossing densities and a large proportion of empty volume, both of which are reminiscent of the crystal structure of tropomyosin (25, 26). The linear densities are parallel to the crystallographic unit cell axes and may correspond to one or more parallel coiled-coil bundles of the major PFR proteins. The diagonal of the crystallographic PFR unit cell repeats at 56-nm intervals in the direction parallel to the axoneme (Fig. 2B). We used a skeletonization algorithm (27) to provide a simplified representation of the densities in the PFR lattice (Fig. 2C). A green parallelogram, passing through the highest PFR densities, represents four crystallographic unit cells in a plane (Fig. S4). Fig. 2C shows the layered nature of the PFR and its spatial relationship to the axoneme (Movie S3). The distance between adjacent layers corresponds to the c crystallographic unit cell spacing (22 nm). Note that none of the crystallographic axes is parallel or perpendicular to the axoneme axis.

The axoneme average (Fig. 2D and Movie S3) was derived from a single tomogram that had the best-preserved structure. It contains the characteristic nine outer microtubule doublets (28–30) (blue) arranged around the central microtubule pair (yellow) similar to those found in sea urchin sperm (8). All nine microtubule doublets, with their associated structures, were aligned and averaged assuming that they were structurally equivalent, thereby compensating for the distortions caused by the limited range of tilt angles (Materials and Methods). In our annotation, microtubule doublet 1 (Fig. 2A and Movie S3) is distal to the PFR (31) and radial spokes (light green) connect outer microtubule doublets to the central pair. Typically, axonemes bend in a plane bisecting the two microtubules of the central pair (32) with a frequency of 10–20 Hz (33). Here, cryo-ET provides a direct observation, in the same frozen specimen, of this orthogonal relationship between the central pair and the bending plane (Fig. S3 A and B). In our tomograms, the plane bisecting the two microtubules of the central pair (Fig. 2A, red dashed line) and the plane through the connecting proteins (Fig. 2A, purple dashed line) intersect at an angle consistent with previous measurements (34–36). To statistically validate our observation, we measured the angle that the PFR makes with the perpendicular bisector of the central pair for a total of eight different flagella (Fig. S3 B–H, in addition to the tomogram for Fig. 2). We found it to be 20 ± 7°. Some features of the axoneme (Fig. 2D), such as the nine doublets, the radial spokes, and the central microtubule pair, appear similar to axonemes from other organisms (7, 8, 37–40). However, structures such as the outer dynein arms could have been partially removed during the extraction with 1 M KCl.

Two rows of connecting proteins (red and pink in Fig. 2 A–C) between the PFR and the axoneme were identified. We observe substantial connections to doublets 5 and 6, but not to 4 and 7 as previously visualized and reported (3, 41). The connections to doublets 4 and 7 are not as bulky as those to doublets 5 and 6. They are longer and perhaps more flexible, and therefore not readily visible in the raw tomograms. Indeed it was shown that connections to doublets 4 and 7 are thin linear structures (13, 14, 16, 24, 42). Due to the difficulty in visualizing the connectors to doublets 4 and 7, they were not chosen for averaging from our reconstructions. The connecting proteins which we do observe (to doublets 5 and 6) repeat every 56 nm in straight flagella, a distance corresponding to seven tubulin dimers along the axoneme (Fig. 2B). This distance is the same as the PFR repeat along the axoneme and also similar to periodic attachments seen in the Euglena axoneme (5). Each of the two rows of connecting proteins was averaged along the length of the flagella. The averages of the connecting proteins in the two rows are similar in size (Fig. 2B and Movie S3) and are schematically represented by spheres in Fig. 2C. It is likely that each connecting density is a complex of several proteins.

Whereas straight flagella were used to determine the average structure, bent flagella can suggest a model for their motion. Fig. 3 A–F show top and side views of tomogram slices of the PFR from three differently bent flagella. The PFR averages are shown in different colors. The crystallographic unit cell lengths of bent and straight PFR agree within 5%: a = 52 ± 2, b = 46 ± 2, and c = 22 ± 1 nm (a total of 7,000 unit cells went into these averages) (Fig. 3G, Fig. S5, and Movie S3). Because the linear PFR protein densities are parallel to the crystallographic unit cell axes rather than to the axoneme axis, the PFR proteins are not compressed or bent (Fig. 3) when the axoneme bends. Rather, the PFR densities pivot as if on hinges at the corners of the crystallographic unit cell with a scissor-like motion, resulting in interaxial angular variations from 85° to 112° (Fig. 3 G and H).

Fig. 3.

The T. brucei flagellum acts like a biological jackscrew. Side views (A, C, and E) and top views (B, D, and F) of a tomogram of a PFR which is bent inward (A and B), straight (C and D), and bent outward (E and F). Tomogram slices are shown in gray. Aligned and averaged subtomograms are shown in color. (Scale bars: 100 nm.) (G) Top view of the crystallographic unit cell axes. While the axial distances stay constant, the γ angle changes as the PFR bends. (H) Jackscrew models of the PFR representing the 85° angle of the inwardly bent (expanded) crystallographic unit cell; 106° angle of the straight crystallographic unit cell; 112° angle of the crystallographic unit cell bent outward (compressed). Connecting proteins are represented by pink spheres.

The organization and structural flexibility of the individual flagellar components and their relationship to each other suggest a mechanism for the movement of the trypanosome powered by the beating axoneme. It is well accepted that the axoneme powers flagellar movement (32, 43, 44). Bending of the axoneme makes the connecting proteins bend along with it, causing their spacing to shorten or lengthen from the 56-nm spacing in straight flagella (Figs. 3H and 4A). This change in spacing of the connecting proteins in turn compresses or stretches the PFR along the axoneme direction (Fig. 3). Our observations suggest that each crystallographic PFR unit cell is analogous to an automobile jackscrew (Fig. 3H and Movie S4). The crystallographic unit cell axes, corresponding to the rigid arms of the jackscrew and composed of the coiled-coil structural proteins, remain constant in length, whereas the interaxial angles vary. The connecting proteins stretch and compress the PFR hinges and thus correspond to the screw in a jackscrew (Figs. 3H and 4, and Movie S4). The change in the crystallographic unit cell angles is analogous to the results of the action of the screw in a jackscrew. The coordinated action of the unit cells, illustrated by the parallelograms in Fig. 3G, produce an overall stretching and contraction of the PFR in a manner analogous to an expansion gate (Fig. S6). Interestingly, using a flexible protein crystal as a part of a cellular mechanical device has been observed before (45). The actin-based acrosomal bundle, proposed as a biological spring (46), is another example of an intracellular 3D protein crystal serving a biomechanical function.

Fig. 4.

Transformation of the normal axoneme (purple) planar motion into the observed bihelical wave motion of T. brucei. (A) The relative sliding motion of the microtubule doublets will tend to compress and expand the connecting proteins (pink) (X1 > X2 > X3). (B) The presence of the PFR (green) that is offset 20° resists this tendency, leading to a bihelical twist of the attached flagellum and T. brucei cell.

The preferred bending plane of the axoneme (34) makes an approximately 20° angle with the PFR (the angle between the dashed lines in Fig. 2A and Movie S3). A constraint on the bending of the axoneme, like that imposed by the PFR having this 20° offset, would induce a rotation, twisting the flagellum into a right- or left-handed helix (Fig. 4B and Movie S5). Fig. S7 shows this twist measured geometrically along a slightly bent flagellum in a single projection image. Fig. S8 shows the twist measured in 3D along the axoneme axis from a tomogram. Both methods give a twist of about 15° per 800 nm in slightly bent flagella. It is possible that flagella were bent passively during freezing. However, no matter what caused these flagella to bend, we observe a consistent relationship of bend to twist (Figs. S7 and S8). It is thus likely that the isolated flagella are constrained to bend in the same way they would in the motile trypanosome. We have observed long, straight flagella over a distance of up to 15 μm. On the other hand, bent flagella occur only over short distances (less than 1 μm) on the grid, possibly because the combination of bend and twist would eventually cause them to come out of the plane of the thin ice layer. Movies S1 and S5 include high-speed light microscopy showing the bihelical motion of the T. brucei flagella with a frequency of 19 ± 3 Hz (11). In contrast, the sea urchin sperm axoneme bends up and down with a simple sinusoidal motion with a similar frequency (32, 33) (Movie S5). At the end of each half-cycle of the T. brucei axoneme beat, the direction of the twist reverses, generating a helix of the opposite hand (Fig. 4B and Movies S1 and S5) (11). Thus, the normal planar motion of an axoneme is transformed into the observed bihelical wave movement characteristic of T. brucei. Depletion of the PFR assembly (Fig. S1) of the T. brucei flagella by RNAi (17, 23) changes the bihelical motion into sinusoidal planar motion with a frequency of 18 ± 1 Hz (Movie S2). The RNAi experiments reinforce the critical role of the PFR for the bihelical wave motion of the organism. Whereas other structures, including the cell body, may influence the bihelical motion, the relative orientation of the PFR to the axoneme was observed to be consistent, unlike the relationship of flagellum to the cell body (34).

The trypanosome flagellum is the most complex microtubule-based nanomachine that has been studied by cryo-ET. Our observations support a molecular mechanism by which the properties of the paraaxonemal structures shape the motion of T. brucei. We propose a straightforward model for energy transduction, from the beating axoneme through the connecting proteins to the deformable PFR crystal, resulting in efficient cell propulsion of this trypanosome.

Materials and Methods

Isolation of T. brucei flagella was performed as previously described (18, 47). Approximately 2 × 10⁹ procyclic 29.13 cells (48) were extracted sequentially with buffers containing 1% Nonidet P-40 and 1 M KCl. Isolated flagella were stored at 4 °C. Quantifoil Copper 200 mesh R 2/2 grids were washed overnight with ethyl acetate. Grids were pretreated with 15-nm gold nanoparticles for tomogram alignment. To prepare a frozen, hydrated grid, 2.5 μL of flagella sample was applied to the grid, blotted, and plunged into liquid ethane using a Vitrobot (FEI). The optimal ice thickness is around 300–400 nm, slightly larger than the diameter of the flagellum, as observed in our tomograms. Imaging was performed on 200-kV microscopes JEM2200FS and JEM2100 equipped with Gatan 4,096 × 4,096 pixel CCD cameras. The JEM2200FS has a field emission gun and an in-column energy filter, whereas the JEM2100 has a LaB₆ gun. Depending on the microscope and imaging conditions, the effective magnifications varied between 11,100× and 16,500×. Tilt series were collected using SerialEM (49) targeted at 6–10 μm underfocus. Each 120° tilt series contained 60 images. Electron dose per tomogram ranged from 60 to 74 electrons/Ų as typically used for cryo-ET (50). Tomograms were reconstructed using IMOD (51). Subvolumes enclosing segments of the PFR, the connecting proteins, and the axoneme were extracted from the reconstructed tomograms. Initial models for the PFR and for the connecting proteins were created by aligning and averaging two adjacent subvolumes (52). More distal segments were then aligned to and averaged with the initial model iteratively. A final realignment of each subvolume to the model was performed to create a final average. The PFR from bent flagella, recognized by visual inspection, were similarly aligned to each other. For the axoneme, 288-nm segments, representing three 96-nm repeat lengths, were extracted along each microtubule doublet. The segments overlapped each other by 96 nm. These segments were first aligned along their length by cross-correlation. The aligned segments were then averaged, after which the nine averaged doublets were then rotationally and translationally aligned and averaged with each other, mitigating the effect of the missing wedge. Finally, the original axoneme was reassembled by properly rotating and translating the microtubule doublet average back into the original volume in the nine original orientations. The separately averaged subtomograms of the PFR, of the connecting proteins, and of the axoneme were fitted back into an original tomogram density map. Density visualization was performed using Chimera (53).