Are the local adjustments of the relative spatial frequencies of the dynein arms and the beta-tubulin monomers involved in the regulation of the "9+2" axoneme?

The “9+2” axoneme is a highly specific cylindrical machine whose periodic bending is due to the cumulative shear of its 9 outer doublets of microtubules. Because of the discrete architecture of the tubulin monomers and the active appendices that the outer doublets carry (dynein arms, nexin links and radial spokes), this movement corresponds to the relative shear of these topological verniers, whose characteristics depend on the geometry of the wave train. When an axonemal segment bends, this induces the compressed and dilated conformations of the tubulin monomers and, consequently, the modification of the spatial frequencies of the appendages that the outer doublets carry. From a dynamic point of view, the adjustments of the spatial frequencies of the elements of the two facing verniers that must interact create different longitudinal periodic patterns of distribution of the joint probability of the molecular interaction as a function of the location of the doublet pairs around the axonemal cylinder and their spatial orientation within the axonemal cylinder. During the shear, these patterns move along the outer doublet intervals at a speed that ranges from one to more than a thousand times that of sliding, in two opposite directions along the two opposite halves of the axoneme separated by the bending plane, respecting the polarity of the dynein arms within the axoneme. Consequently, these waves might be involved in the regulation of the alternating activity of the dynein arms along the flagellum, because they induce the necessary intermolecular dialog along the axoneme since they could be an element of the local dynamic stability/instability equilibrium of the axoneme. This complements the geometric clutch model [Lindemann, C., 1994. A “geometric clutch” hypothesis to explain oscillations of the axoneme of cilia and flagella. J. Theor. Biol. 168, 175-189].     

Elastic extension and jump of the flagellar nexin links: a theoretical mechanical cycle.

The functions of the nexin links of a flagellar axoneme have not been clearly demonstrated. Taking into account both the elastic (Hookean) characteristics and the possible jump of the nexin links, we calculated the sliding to bending conversion of a theoretical model in a tip-ward direction step by step, according to the essential principles proposed by the geometric clutch hypothesis [Lindemann, 1994: J Theoret Biol 168:175-189]: the activity of the dynein arms depends on the transverse forces induced by the axonemal curvature. In our calculations, however, the transverse forces that are involved in the regulation of the activities of the dynein arms were due to the extension of the nexin links located upstream of a given abscissa. This allowed us to define a bent segment as the axonemal portion at whose proximal and distal ends the nexin links were relaxed, and as fully extended as possible, respectively. The model creates an apparent disorder in the orientation of the nexin links as already observed [Bozkurt and Wooley, 1993: Cell Motil Cytoskeleton 24:109-118; Wooley, 1997: J Cell Sci 110:85-94]. We propose that the nexin links are involved in a mechanical cycle, whose 3 stages are (1) rapid extension, (2) slow relaxation, and (3) stand-by. The rapid extension is compatible with the mechanical interactions between the nexin links and the inner dynein arms with which they form the dynein regulatory complex. This was correlated with the oscillating properties of the nexin links along the axoneme that allow them to be one of the regulatory elements of the local ATPase activity of the dynein arms.     

Inner dynein arms but not outer dynein arms require the activity of kinesin homologue protein KHP1(FLA10) to reach the distal part of flagella in Chlamydomonas

Inner dynein arms, but not outer dynein arms, require the activity of KHP1(FLA10) to reach the distal part of axonemes before binding to outer doublet microtubules. We have analyzed the rescue of inner or outer dynein arms in quadriflagellate dikaryons by immunofluorescence microscopy of p28(IDA4), an inner dynein arm light chain, or IC69(ODA6), an outer dynein arm intermediate chain. In dikaryons two strains with different genetic backgrounds share the cytoplasm. As a consequence, wild-type axonemal precursors are transported to and assembled in mutant axonemes to complement the defects. The rescue of inner dynein arms containing p28 in ida4-wild-type dikaryons progressively occurred from the distal part of the axonemes and with time was extended towards the proximal part. In contrast, the rescue of outer dynein arms in oda2-wild-type dikaryons progressively occurred along the entire length of the axoneme. Rescue of inner dynein arms containing p28 in ida4fla10-fla10 dikaryons was similar to the rescue observed in ida4-wild-type dikaryons at 21 degrees C, whereas it was inhibited at 32 degrees C, a nonpermissive temperature for KHP1(FLA10). In contrast, rescue of outer dynein arms in oda2fla10-fla10 dikaryons was similar to the rescue observed in oda2-wild-type dikaryons at both 21 degrees and 32 degrees C and was not inhibited at 32 degrees C. Positioning of substructures in the internal part of the axonemal shaft requires the activity of kinesin homologue protein 1.     

One of the Nine Doublet Microtubules of Eukaryotic Flagella Exhibits Unique and Partially Conserved Structures

The axonemal core of motile cilia and flagella consists of nine doublet microtubules surrounding two central single microtubules. Attached to the doublets are thousands of dynein motors that produce sliding between neighboring doublets, which in turn causes flagellar bending. Although many structural features of the axoneme have been described, structures that are unique to specific doublets remain largely uncharacterized. These doublet-specific structures introduce asymmetry into the axoneme and are likely important for the spatial control of local microtubule sliding. Here, we used cryo-electron tomography and doublet-specific averaging to determine the 3D structures of individual doublets in the flagella of two evolutionarily distant organisms, the protist Chlamydomonas and the sea urchin Strongylocentrotus. We demonstrate that, in both organisms, one of the nine doublets exhibits unique structural features. Some of these features are highly conserved, such as the inter-doublet link i-SUB5-6, which connects this doublet to its neighbor with a periodicity of 96 nm. We also show that the previously described inter-doublet links attached to this doublet, the o-SUB5-6 in Strongylocentrotus and the proximal 1–2 bridge in Chlamydomonas, are likely not homologous features. The presence of inter-doublet links and reduction of dynein arms indicate that inter-doublet sliding of this unique doublet against its neighbor is limited, providing a rigid plane perpendicular to the flagellar bending plane. These doublet-specific features and the non-sliding nature of these connected doublets suggest a structural basis for the asymmetric distribution of dynein activity and inter-doublet sliding, resulting in quasi-planar waveforms typical of 9+2 cilia and flagella.     

Direction of force generated by the inner row of dynein arms on flagellar microtubules

Our goal was to determine the direction of force generation of the inner dynein arms in flagellar axonemes. We developed an efficient means of extracting the outer row of dynein arms in demembranated sperm tail axonemes, leaving the inner row of dynein arms structurally and functionally intact. Sperm tail axonemes depleted of outer arms beat at half the beat frequency of sperm tails with intact arms over a wide range of ATP concentrations. The isolated, outer arm-depleted axonemes were induced to undergo microtubule sliding in the presence of ATP and trypsin. Electron microscopic analysis of the relative direction of microtubule sliding (see Sale, W. S. and P. Satir, 1977, Proc. Natl. Acad. Sci. USA, 74:2045-2049) revealed that the doublet microtubule with the row of inner dynein arms, doublet N, always moved by sliding toward the proximal end of the axoneme relative to doublet N + 1. Therefore, the inner arms generate force such that doublet N pushes doublet N + 1 tipward. This is the same direction of microtubule sliding induced by ATP and trypsin in axonemes having both inner and outer dynein arms. The implications of this result for the mechanism of ciliary bending and utility in functional definition of cytoplasmic dyneins are discussed.     

Geometric Clutch model version 3: the role of the inner and outer arm dyneins in the ciliary beat

The Geometric Clutch model of ciliary and flagellar beating uses the transverse force (t-force) that develops between the outer doublets of the axoneme as the regulator for activating and deactivating the dynein motors and organizing the flagellar beat. The version of the model described here adds detail to the formulations used in the two previous versions as follows: (1) In place of two opposing sets of dyneins, the new model has four sets of dyneins, corresponding to two sets on each side of the axoneme acting in series. (2) The four sets of dyneins are each subdivided into two ranks representing inner and outer arm dyneins. (3) The force produced by each dynein is governed by a force-velocity relationship that is independently specified for the inner and outer arms. Consistent with the original model, the new version of the Geometric Clutch model can simulate both the effective and recovery stroke phases of the ciliary beat using a single uniform algorithm. In addition, the new version can operate with the outer arms disabled. Under this condition, the simulation exhibits a beat pattern similar to the original but the beat frequency is reduced to approximately one third. These results are contingent on using force-velocity relationships for the inner and outer arms similar to those described by Brokaw [1999: Cell Motil. Cytoskeleton 42:134-148], where the inner arms contribute most of the driving force at low shear velocities. This constitutes the first examination of the effects of the force-velocity characteristics of dynein on a cilia-like beat in a theoretical framework.     

A physical model of microtubule sliding in ciliary axonemes.

Ciliary movement is caused by coordinated sliding interactions between the peripheral doublet microtubules of the axoneme. In demembranated organelles treated with trypsin and ATP, this sliding can be visualized during progressive disintegration. In this paper, microtubule sliding behavior resulting from various patterns of dynein arm activity and elastic link breakage is determined using a simplified model of the axoneme. The model consists of a cylindrical array of microtubules joined, initially, by elastic links, with the possibility of dynein arm interaction between microtubules. If no elastic links are broken, sliding can produce stable distortion of the model, which finds application to straight sections of a motile cilium. If some elastic links break, the model predicts a variety of sliding patterns, some of which match, qualitatively, the observed disintegration behavior of real axonemes. Splitting of the axoneme is most likely to occur between two doublets N and N + 1 when either the arms on doublet N + 1 are active and arms on doublet N are inactive or arms on doublet N - 1 are active while arms on doublet N are inactive. The analysis suggests further experimental studies which, in conjunction with the model, will lead to a more detailed understanding of the sliding mechanism, and will allow the mechanical properties of some axonemal components to be evaluated.     

Flagellar arrest behavior predicted by the Geometric Clutch model is confirmed experimentally by micromanipulation experiments on reactivated bull sperm.

The central tenet of the Geometric Clutch hypothesis of flagellar beating is that the internal force transverse to the outer doublets (t-force) mediates the initiation and termination of episodes of dynein engagement. Therefore, if the development of an adequate t-force is prevented, then the dynein-switching necessary to complete a cycle of beating should fail. The dominant component of the t-force is the product of the longitudinal force on each outer doublet multiplied by the local curvature of the flagellum. In the present study, two separate strategies, blocking and clipping, were employed to limit the development of the t-force in Triton X-100 extracted bull sperm models. The blocking strategy used a bent glass microprobe to restrict the flagellum during a beat, preventing the development of curvature in the basal portion of the flagellum. The clipping strategy was designed to shorten the flagellum by clipping off distal segments of the flagellum with a glass microprobe. This limits the number of dyneins that can contribute to bending and consequently reduces the longitudinal force on the doublets. The blocking and clipping strategies both produced an arrest of the beat cycle consistent with predictions based on the Geometric Clutch hypothesis. Direct comparison of experimentally produced arrest behavior to the behavior of the Geometric Clutch computer model of a bull sperm yielded similar arrest patterns. The computer model duplicated the observed behavior using reasonable values for dynein force and flagellar stiffness. The experimental data derived from both blocking and clipping experiments are fully compatible with the Geometric Clutch hypothesis.     

Evidence for axonemal distortion during the flagellar beat of Chlamydomonas

In order to understand the working mechanism that governs the flagellar beat it is essential to know if the axoneme undergoes distortion during the course of the beat cycle. The rapid fixation method employed by Mitchell was able to preserve the waveform of Chlamydomonas flagella much as it appears during normal flagellar beating [Mitchell, Cell Motil Cytoskeleton 2003;56:120-129]. This conservation of the waveform suggests that the stress responsible for the production of bending is also trapped by the fixation procedure. Longitudinal sections of these well-preserved flagella were used to document variations in the relative axonemal diameter. Sections aligned to the plane of bending, showing both the central pair microtubules and outer doublets, were examined for this purpose. Micrographs were selected that continuously showed both the outer doublets and the central pair from a straight region to a curved region of the flagellum. Axoneme diameters measured from these select micrographs showed an increase in relative diameter that averaged 39 nm greater at the crest of the bent region. This constituted a 24% increase in the axoneme diameter in the bends. The transverse stress acting across the axoneme during bending was calculated from the Geometric Clutch computer model for a simulated Chlamydomonas-like flagellar beat. If we assume that this is representative of the transverse stress acting in a real flagellum, then the Young's modulus of the intact axoneme is approximately 0.02 MPa. The possibility that the distortion of the axoneme during the beat could play a significant role in regulating dynein function is discussed.     

On the contribution of dynein-like activity to twisting in a three-dimensional sliding filament model.

It has been shown (Hines, M., and J. J. Blum, Biophys.J., 1984, 46:559-565) that passive moment-bearing links do not contribute appreciable twist resistance to an axoneme nor do they cause appreciable twisting in response to internal shear forces. We now examine the contribution of active moment-bearing links such as dynein arms to the generation of twist within an axoneme. The dynein model used causes distal sliding of the adjacent doublet by a force dependent on the angle of attachment of the arms. Attachment of the arms occurs at a specified angle relative to the angle of minimum potential energy. The steady state shape consistent with the forces applied by the attached dyneins is computed. It is shown that the twist generated in an active region is counterclockwise as viewed from tip to base and therefore accumulates at the end of the axoneme. For realistic forces and twist resistances, cumulative twist should not exceed a few degrees.     

Rib72, a conserved protein associated with the ribbon compartment of flagellar A-microtubules and potentially involved in the linkage between outer doublet microtubules

Ciliary and flagellar axonemes are basically composed of nine outer doublet microtubules and several functional components, e.g. dynein arms, radial spokes, and interdoublet links. Each A-tubule of the doublet contains a specialized "ribbon" of three protofilaments composed of tubulin and other proteins postulated to specify the three-dimensional arrangement of the various axonemal components. The interdoublet links hold the doublet microtubules together and limit their sliding during the flagellar beat. In this study on Chlamydomonas reinhardtii, we cloned a cDNA encoding a 71,985-Da polypeptide with three DM10 repeats, two C-terminal EF-hand motifs, and homologs extending to humans. This polypeptide, designated as Rib72, is a novel component of the ribbon compartment of flagellar microtubules. It remained associated with 9-fold arrays of doublet tubules following extraction under high and low ionic conditions, and anti-Rib72 antibodies revealed an approximately 96-nm periodicity along axonemes, consistent with Rib72 associating with interdoublet links. Following proteolysis- and ATP-dependent disintegration of axonemes, the rate of cleavage of Rib72 correlated closely with the rate of sliding disintegration. These observations identify a ribbon-associated protein that may function in the structural assembly of the axoneme and in the mechanism and regulation of ciliary and flagellar motility.     

High-frequency vibration in flagellar axonemes with amplitudes reflecting the size of tubulin

Flagellar axonemes of sea urchin sperm display high-frequency (approximately 300 Hz) vibration with nanometer-scale amplitudes in the presence of ATP (Kamimura, S., and R. Kamiya. 1989. Nature (Lond.). 340:476-478). The vibration appears to represent normal mechanochemical interaction between dynein and microtubules because the dependence of the frequency on MgATP concentration is similar to that of the axonemal motility, and because it is inhibited by micromolar concentrations of vanadate. In this study a two-dimensional photo-sensor was used to characterize this phenomenon in detail. Several new features were revealed. First, the vibration was found to be due to a back-and-forth movement of the doublet microtubules along the axonemal length. Two beads attached to different parts of the same axoneme vibrated in unison, i.e., synchronized exactly in phase. This suggested that the outer doublet can be regarded as a stiff rod in vibrating axonemes. Second, evidence was obtained that the amplitude of the vibration reflected the number of active dynein arms. Third, under certain conditions, the vibration amplitude took stepwise values of 8 x N + 4 nm (N = 0, 1, 2, 3, or 4), indicating that the amplitude of microtubule sliding was limited by the size of tubulin dimer (8 nm) or monomer (4 nm). To explain this phenomenon, a model is presented based on an assumption that the force production by dynein is turned off when dynein is subjected to tensile force; i.e., dynein is assumed to be equipped with a feedback mechanism necessary for oscillation.     

Dynein heavy chain isoforms and axonemal motility

The translocation of dynein along microtubules is the basis for a wide variety of essential cellular movements. Dynein was first discovered in the ciliary axoneme, where it causes the directed sliding between outer doublet microtubules that underlies ciliary bending. The initiation and propagation of ciliary bends are produced by a precisely located array of different dyneins containing eight or more different dynein heavy chain isoforms. The detailed clarification of the structural and functional diversity of axonemal dynein heavy chains will not only provide the key to understanding how cilia function, but also give insights applicable to the study of non-axonemal microtubule motors.     

The IC138 and IC140 intermediate chains of the I1 axonemal dynein complex bind directly to tubulin

Dyneins are minus end directed microtubule motors that play a critical role in ciliary and flagellar movement. Ciliary dyneins, also known as axonemal dyneins, are characterized based on their location on the axoneme, either as outer dynein arms or inner dynein arms. The I1 dynein is the best-characterized subspecies of the inner dynein arms; however the interactions between many of the components of the I1 complex and the axoneme are not well defined. In an effort to elucidate the interactions in which the I1 components are involved, we performed zero-length crosslinking on axonemes and studied the crosslinked products formed by the I1 intermediate chains, IC138 and IC140. Our data indicate that IC138 and IC140 bind directly to microtubules. Mass-spectrometry analysis of the crosslinked product identified both α- and β-tubulin as the IC138 and IC140 binding partners. This was further confirmed by crosslinking experiments carried out on purified I1 fractions bound to Taxol-stabilized microtubules. Furthermore, the interaction between IC140 and tubulin is lost when IC138 is absent. Our studies support previous findings that intermediate chains play critical roles in the assembly, axonemal targeting and regulation of the I1 dynein complex.     

The Oligomeric Outer Dynein Arm Assembly Factor CCDC103 Is Tightly Integrated within the Ciliary Axoneme and Exhibits Periodic Binding to Microtubules

CCDC103 is an ∼29-kDa protein consisting of a central RPAP3_C domain flanked by N- and C-terminal coiled coils. Defects in CCDC103 lead to primary ciliary dyskinesia caused by the loss of outer dynein arms. This protein is present along the entire length of the ciliary axoneme and does not require other dynein or docking complex components for its integration. Unlike other known dynein assembly factors within the axoneme, CCDC103 is not solubilized by 0.6 m NaCl and requires more chaotropic conditions, such as 0.5 m KI. Alternatively, it can be extracted using 0.3% sarkosyl. CCDC103 forms stable dimers and other oligomers in solution through interactions involving the central domain. The smallest particle observed by dynamic light scattering has a hydrodynamic diameter of ∼25 nm. Furthermore, CCDC103 binds microtubules directly, forming ∼9-nm diameter particles that exhibit a 12-nm spacing on the microtubule lattice, suggesting that there may be two CCDC103 units per outer arm dynein repeat. Although the outer dynein arm docking complex is necessary to form arrays of dyneins along microtubules, it is not sufficient to set up a single array in a precise location on each axonemal doublet. We propose that CCDC103 helps generate a high-affinity site on the doublets for outer arm assembly, either through direct interactions or indirectly, perhaps by modifying the underlying microtubule lattice.     

Mutations in the "dynein regulatory complex" alter the ATP-insensitive binding sites for inner arm dyneins in Chlamydomonas axonemes

To understand mechanisms of regulation of dynein activity along and around the axoneme we further characterized the "dynein regulatory complex" (drc). The lack of some axonemal proteins, which together are referred to as drc, causes the suppression of flagellar paralysis of radial spoke and central pair mutants. The drc is also an adapter involved in the ATP-insensitive binding of I2 and I3 inner dynein arms to doublet microtubules. Evidence supporting these conclusions was obtained through analyses of five drc mutants: pf2, pf3, suppf3, suppf4, and suppf5. Axonemes from drc mutants lack part of I2 and I3 inner dynein arms as well as subsets of seven drc components (apparent molecular weight from 29,000 to 192,000). In the absence of ATP-Mg, dynein-depleted axonemes from the same mutants bind I2 and I3 inner arms at both ATP-sensitive and -insensitive sites. At ATP-insensitive sites, they bind I2 and I3 inner arms to an extent that depends on the drc defect. This evidence suggested to us that the drc forms one binding site for the I2 and I3 inner arms on the A part of doublet microtubules.     

Dynein as a microtubule translocator in ciliary motility: current studies of arm structure and activity pattern

The dynein arms of ciliary doublet microtubules cause adjacent axonemal doublets to slide apart with fixed polarity. This suggests that there is a unique mechanochemistry to the dynein arm with unidirectional force generation in all active arms and also that not all arms are active at once during a ciliary beat. Negative stain and thin-section images of arms in axonemes treated with beta, gamma methylene adenosine triphosphate (AMP-PCP) show a consistent subunit construction where the globular head of the arm interacts with subfiber B of doublet N+1. This interpretation differs from that provided by freeze etch and STEM interpretations of in situ arm construction and has implications for the mechanochemical cycle of the arm. A computer model of the arms in relation to other axonemal structures has been constructed to test these interpretations. Attachment of the head of the arm subfiber B is directly demonstrable in splayed axonemes in AMP-PCP. About half of the doublets in an axoneme show such attachments, while half do not. This might imply that about half the doublets in an axoneme are active at any given instant and can be identified as such. This information may be useful in probing questions of how active arms differ biochemically from inactive arms and of how microtubule translocators in general become active.     

The architecture of outer dynein arms in situ

Outer dynein arms, the force generators for axonemal motion, form arrays on microtubule doublets in situ, although they are bouquet-like complexes with separated heads of multiple heavy chains when isolated in vitro. To understand how the three heavy chains are folded in the array, we reconstructed the detailed 3D structure of outer dynein arms of Chlamydomonas flagella in situ by electron cryo-tomography and single-particle averaging. The outer dynein arm binds to the A-microtubule through three interfaces on two adjacent protofilaments, two of which probably represent the docking complex. The three AAA rings of heavy chains, seen as stacked plates, are connected in a striking manner on microtubule doublets. The tail of the alpha-heavy chain, identified by analyzing the oda11 mutant, which lacks alpha-heavy chain, extends from the AAA ring tilted toward the tip of the axoneme and towards the inside of the axoneme at 50 degrees , suggesting a three-dimensional power stroke. The neighboring outer dynein arms are connected through two filamentous structures: one at the exterior of the axoneme and the other through the alpha-tail. Although the beta-tail seems to merge with the alpha-tail at the internal side of the axoneme, the gamma-tail is likely to extend at the exterior of the axoneme and join the AAA ring. This suggests that the fold and function of gamma-heavy chain are different from those of alpha and beta-chains.     

Pathology of the cytoskeleton of the human sperm flagellum: axonemal and peri-axonemal anomalies

A quantitative ultrastructural study was performed on 56 ejaculates showing anomalies of the sperm axonemal complex. The anomalies comprised either the absence of one, or more often several, axonemal structures, or defective elongation of the doublets. Several characteristics relating to the extent and superimposition of the various anomalies could be described and enabled the definition of 6 groups of anomalies. In decreasing order of frequency these were: absence of the doublets and peripheral junctions, absence of the central complex, of the outer dynein arms, of the central junctions, of both dynein arms, and absence of the inner dynein arms and peripheral junctions. Some anomalies caused total immobility, whereas others caused abnormal movement patterns. Abnormalities of the peri-axonemal structures were found in each group. The various light microscopic characteristics of each of the 6 groups represented 6 seminal profiles which should permit their detection during a routine semen analysis. Several specific associations of axonemal and/or peri-axonemal anomalies would suggest some morphogenetic links between them. Relationships between the absence of doublets or the absence of the central complex and disturbances of microtubular polymerization are discussed. Finally, the study has provided new data on the composition of the axoneme.     

Casein kinase I is anchored on axonemal doublet microtubules and regulates flagellar dynein phosphorylation and activity

Flagellar dynein activity is regulated by phosphorylation. One critical phosphoprotein substrate in Chlamydomonas is the 138-kDa intermediate chain (IC138) of the inner arm dyneins (Habermacher, G., and Sale, W. S. (1997) J. Cell Biol. 136, 167-176). In this study, several approaches were used to determine that casein kinase I (CKI) is physically anchored in the flagellar axoneme and regulates IC138 phosphorylation and dynein activity. First, using a videomicroscopic motility assay, selective CKI inhibitors rescued dynein-driven microtubule sliding in axonemes isolated from paralyzed flagellar mutants lacking radial spokes. Rescue of dynein activity failed in axonemes isolated from these mutant cells lacking IC138. Second, CKI was unequivocally identified in salt extracts from isolated axonemes, whereas casein kinase II was excluded from the flagellar compartment. Third, Western blots indicate that within flagella, CKI is anchored exclusively to the axoneme. Analysis of multiple Chlamydomonas motility mutants suggests that the axonemal CKI is located on the outer doublet microtubules. Finally, CKI inhibitors that rescued dynein activity blocked phosphorylation of IC138. We propose that CKI is anchored on the outer doublet microtubules in position to regulate flagellar dynein.