Cryo-EM reveals how the mastigoneme assembles and responds to environmental signal changes

Cilia and flagella (interchangeable terms henceforth) are evolutionarily conserved organelles that protrude from the surface of eukaryotic cells, forming membraned appendage-like structures. Motile cilia ubiquitously exist in eukaryotic cells and play central roles in cell motility, extracellular fluid transport (Ringers et al., 2020; Satir et al., 2014), and, likely, environmental signal reception as well (Bloodgood, 2010). Given their fundamental roles, cilia dysfunction can lead to a range of human diseases collectively referred to as ciliopathies, including hydrocephalus, polycystic liver and kidney disease, and retinal degeneration (Badano et al., 2006). The core region of a typical motile cilium is the “9 + 2” axoneme, characterized by nine periphery microtubule doublets (MTDs; Ma et al., 2019; Kubo et al., 2023) and two central microtubule singlets (Gui et al., 2022; Han et al., 2022; Klena and Pigino, 2022). Attached to each MTD are two rows of dyneins that produce mechanical forces required for ciliary beating and other microtubule-association protein complexes that regulate dynein activity, including radial spokes (RSs) and nexin-dynein regulatory complexes (N-DRCs; Lin and Nicastro, 2018). Ciliary motility is a highly regulated process, involving various extracellular signals. Recent progress in high-resolution cryo-electron microscopy (cryo-EM) structures has provided important insights into a mechanistic understanding of several key components (Gui et al., 2021; Gui et al., 2022; Han et al., 2022; Rao et al., 2021; Van den Hoek et al., 2022) and their regulatory roles in the axoneme. Nevertheless, the structures and functions of ciliary membrane proteins and other ciliary surface components are far less understood.

In the lower flagellates, there are hair-like structures on the surface of cilia called mastigonemes. They line up into two rows at the distal ends of the cilia (Nakamura et al., 1996). Mastigonemes were first observed in Chrysophyceae in 1965 (Bradley, 1965). In different species, there exist two different types of mastigonemes: the tubular mastigoneme and the non-tubular mastigoneme. The Chlamydomonas reinhardtii (CR) mastigoneme, belonging to the non-tubular type, was first reported in 1972 by Witman et al. as repeating units of a single glycoprotein (Witman et al., 1972). Bernstein successfully isolated and purified the mastigoneme protein and determined its molecular weight at ∼200 kD (Bernstein, 1995). Later, the monoclonal antibody against CR mastigonemes provided the first direct evidence of its location and function (Nakamura et al., 1996). It was found that under the action of the mastigoneme monoclonal antibody, the cell swimming velocity was reduced to 70–80% of the normal value. Upon the completion of CR genome sequencing, the MST1 gene was identified as the primary component of its mastigoneme (Merchant et al., 2007). Recent studies revealed that the mastigoneme is anchored on PKD2 (Liu et al., 2020), a homolog of human polycystin 2 (encoded by the PKD2 gene). PKD2 is a cation channel and a member of the transient receptor potential (TRP) channel protein superfamily (Wang et al., 2018). The PKD2 and mastigoneme form a stable complex anchored on doublet microtubules 4 and 8 (Liu et al., 2020). Their position is perpendicular to the ciliary beating plane. It was believed that the PKD2–mastigoneme arrays could perceive forces during ciliary beating and transfer signals to locally regulate the axoneme’s response. However, recent studies suggested that mastigonemes might not directly contribute to increasing the flagella’s effective area or enhancing the flagellar thrust (Amador et al., 2020). Hence, a more detailed examination of the CR mastigoneme’s structure and function is urgently needed.

Here, we determined the overall architecture of the CR mastigoneme by cryo-electron tomography (cryo-ET) and a 3.4-Å structure by single-particle cryo-EM. The mastigoneme is a polymer consisting of repeating MST1 subunits. Each MST1 monomer contains nine Immunoglobulin (Ig)-like folds (Igs), six Sushi domains (SDs), and a polyproline II helix (PPII) followed by a C-terminal helix. From our high-resolution cryo-EM map, we identified two modification sites (N1194 and N1643) within the SDs and a densely modified region on the PPII helix, aligning with the established knowledge that the mastigoneme is a glycoprotein polymer (Witman et al., 1972). Overall, the mastigoneme filament is centrosymmetric, non-polar, and has a spiral arrangement. This periodic configuration imparts flexibility and stability to the filament. Interestingly, the presence of multiple clustered disulfide bonds within this spiral configuration highlights its structural robustness, particularly crucial for proteins located on the cell surface and adapting to extracellular challenges. Although deficiencies in mastigoneme structure did not significantly alter cell swimming behavior, they did lead to compromised movement under conditions of heightened viscosity. These findings shed light on the essential role of mastigonemes in flagellar dynamics and indicate their participation in diverse responses to environmental stimuli.

CR mastigonemes extend laterally from the membranes of flagella to form threadlike projections (Fig. 1 A). A recent study suggested that mastigonemes are positioned near the microtubule doublets 4 and 8 (Liu et al., 2020). During our recent cryo-EM work on the central apparatus (CA) structure (Han et al., 2022), we frequently observed mastigonemes on both intact flagella and the purified axoneme. Mastigonemes were clearly detected in the negative-stain images using the whole CR cell (Fig. 1 B). The lengths of the CR mastigonemes range from 600 to 900 nm and cluster at an average of 744 nm (Fig. 1 C).

We imaged isolated flagella with membranes as well as purified axonemes for cryo-ET analysis. We find that mastigonemes are attached to the axoneme, whether the membrane is intact or not (Fig. 1, D and E; and Video 1), suggesting that mastigonemes are exceptionally rugged structures on the axoneme, and the entire mastigoneme may contain additional structures between the axoneme and the membrane protein PKD2. A previous cryo-ET study suggested that the mastigoneme shows a helical structure composed of two protofilaments twisting around each other with a pitch of ∼40 nm (Liu et al., 2020). By contrast, both our cryo-ET reconstruction and high-resolution cryo-EM 2D suggested very different features (Fig. S1 A).

Our study involved the acquisition of two distinct datasets. Employing an iterative process of 2D/3D classification and realignment of filament segments, we meticulously curated 98,875 segments in dataset 1 and 70,074 segments in dataset 2, ultimately resulting in a final structure resolution of 3.4 Å and 3.9 Å, respectively. This was achieved using a single-particle approach (Fig. S1 B flowchart for dataset 1 and comprehensive dataset details in Table S1), facilitating the generation of both MST1 monomer and filament structures.

To avoid potential overfitting of wrong helical parameters, we simply used C1 symmetry (i.e., without applying any symmetry) during the refinement. We found that the refined filament is composed of multiple repeating units that clearly show equivalent density maps, suggesting they account for the same component. We then identified the individual domains of a single unit in the filament by matching the side-chain features to the protein sequences of the whole proteome using a de novo protein identification approach we previously developed (Han et al., 2022; Rao et al., 2021). Consistent with a previous study (Witman et al., 1972), our cryo-EM results confirmed that the filament is composed of a single protein, i.e., MST1, and excluded all other possibilities. By estimating the distances among individual domains and optimally matching the predicted MST1 topology, we successfully built atomic models of the MST1 subunits and the high-order assembly of the mastigoneme filament.

Our cryo-EM structure reveals that the overall architecture of mastigoneme appears as a polymer composed of repeating MST1 subunits (Video 2). From the N-terminus to the C-terminus, a single MST1 contains nine immunoglobulin (Ig) domains, six SDs, and a left-handed polyproline II helix (PPII) followed by a C-terminal α-helix (Fig. 2, A and B). MST1 is predicted to possess a cleavable 42-residue ER signal sequence at its N-terminus (Liu et al., 2020). Indeed, this region is missing from our cryo-EM map.

Each Ig domain we determined forms a compact structure that fits into a parallelepiped shape and is composed of two layers of antiparallel β-strands, connected by variable loops and/or short helices (Fig. S2 A). A common feature is that the core structure (β-strands B, C, E, and F) is embedded in an antiparallel curled β-sheet sandwich with a total of three to five additional strands and a characteristic intersheet angle (Bork et al., 1994). 3D superposition of the nine Ig domain structures shows that their overall conformations are similar (Fig. S2 B). The main structural variations among the nine Ig domains occur in the βC–βE loop region (Fig. S2 C).

Although the homologs of mastigoneme are limited in distribution to green algae, the domain organization could provide some clues to evolutionary analysis. Ig-like folds are widely distributed in hundreds of proteins of diverse functions, including antibodies, cell adhesion molecules, the giant muscle kinase titin, etc. (Kelly et al., 2021; Liu and May, 2012; Van Buul et al., 2007). However, a tandem of nine Ig folds is unusual. Another well-known Ig-fold protein in CR is Fus1, which exists on the external surface of gametes and plays an essential role in an early step of the fusion process during CR fertilization. Fus1 has seven continuous Ig fold domains (Misamore et al., 2003). This assembly has been found in all members of the invasin/intimin family of proteins, which are used for bacterial adhesion to their host cells (Hamburger et al., 1999). In humans, the Roundabout (Robo) receptor family is one of the few proteins with eight continuous Ig-like folds in its ectodomain. The transmembrane Robo receptors mediate various neuronal responses, including neurogenesis, migration, and axonal guidance (Barak et al., 2019). Another example is filamin, which is composed of 24 Ig-like repeat modules (Jain et al., 2018). Studies found that the Ig-like folds make the filamin protein mechanosensory (Chen et al., 2013; Jahed et al., 2014). Secondly, ephrin receptors, the largest subfamily of receptor tyrosine kinases, are membrane-bound. They can bind to the ephrin ligands to send information bidirectionally, thereby regulating cell communication, cell attraction, cell repulsion, and other activities (Himanen and Nikolov, 2003). The SDs in mastigonemes are highly similar to the cysteine-rich domain in the ephrin receptor ectodomain (Seiradake et al., 2013). In addition, the extracellular domain of the ephrin receptor contains two fibronectin type III (FnIII) domains, which is also a member of the Ig superfamily (Koide et al., 2012). The structural compositions of the ephrin receptor and mastigoneme are also similar. The structural model suggests that mastigonemes may play an important role in cell–cell or cell–environment communications.

Following the nine Ig domains is a tandem of six SDs. SDs exist in a variety of complement and adhesion proteins. The SD is an evolutionarily conserved protein domain and usually forms short consensus repeats. In our mastigoneme structure, the six SDs provide major interfaces for interactions among mastigoneme subunits. The linker between Ig₉ and SD_a (IS linker; Fig. 2 A) serves as a tether that gathers multiple regions and offers numerous additional interaction areas.

Following the SD repeats, the PPII helix, which appears in the central region of the MST1 monomeric structure, folds back and extends over 22 nm to contact the Ig-fold region. Left-handed PPII helices in recent studies have emerged as a highly attractive structural component as they ubiquitously exist in fibrillar proteins, like collagen and Amelogenin, and are also abundant in many other folded and unfolded proteins (Chebrek et al., 2014). It is worth noting that we provide the first PPII helix structure determined by cryo-EM. Near the C-terminus of the MST1 sequence, there are 57 prolines in ∼100 amino acids, leading to a proline-rich region that forms a long polyproline helix in our cryo-EM structure (Fig. 2 B). Each PPII helical turn is composed of three proline residues, which form a threefold rotational symmetry visualized from the top view (Fig. S3 B). In contrast to a typical α-helix, which shows a 1.5-Å rise along the helical axis per residue, the PPII helix is extended ∼3 Å along the helical axis per residue. As a result of this special geometry organization, the regular pattern of intrachain hydrogen bonds does not exist in the PPII helix structure we determined. Though the conformational freedom will be partially limited by the pyrrolidine ring, the PPII helix is still more relaxed than a typical α-helix and β-strand. In the mastigoneme, the PPII helix is located on the outer surface and interacts with SD_a–b, serving as “rubber bands” that bind all the rest of mastigoneme components, simultaneously providing tenacity, flexibility, and mobility in the filament structure to make it a dynamically stable system required for the high-frequency beating. To test that hypothesis, we employed 3D variability analysis in cryoSPARC (Punjani et al., 2017). Our analysis revealed that the PPII helix is indeed the most flexible region, showing a significant lateral movement by up to 20 Å with respect to the SD domains (Video 3), while the key interaction interfaces keep invariant during this movement.

The PPII helix is distinct from the common secondary structures, such as α-helix or β-strand. The left-handed helix has long been known as a part of collagen. The typical PPII helix is an extended left-hand helix defined by the torsional angle cluster with the distribution maximum at −75 to 145°. The overall shape is like a triangular prism. It has three residues per turn, and the rotational symmetry is about threefold (Chebrek et al., 2014). PPII helices are involved in transcription, cell motility, self-assembly, elasticity, and other functions (Kay et al., 2000; Siligardi and Drake, 1995). However, PPII helices are not always assigned in experimentally solved structures. Here, with the long PPII helix in the mastigoneme, we were able to build a molecular model useful for future structural studies. The other interesting finding is that there are additional densities on the PPII helix surface, which are attributed to glycosylation. Glycans on the surface may protect CR against protease cleavage and minimize damage from the outer environment. Studies on the archaeal filaments suggest that glycosylation protects these extracellular appendages in extreme environments (Wang et al., 2019). Similarly, we speculate that glycosylation on the PPII helix may help CR resist environmental challenges.

It was suggested that mastigoneme is a glycoprotein polymer (Witman et al., 1972). Our cryo-EM structure reveals multiple modification motifs (Burda and Aebi, 1999), such as NYT1194–1196 at SD_a, NAT1643–1645 at SD_d, and NRS 1924–1926 at the PPII helix (Fig. S3 A). Moreover, additional densities are associated at nearly the entire surface of the PPII helix in our high-resolution cryo-EM map, suggesting that this region is heavily glycosylated (Fig. S3 B). The PPII helix contains only three additional residues, Ser, Asn, and Arg, besides the prolines and one alanine. Although we cannot unambiguously determine the identity of the posttranslational modifications due to the flexibility and limited resolution of these putative glycans, the potential modification sites are attributed to the residues Ser, Asn, and Arg in the PPII helix.

The CR mastigoneme was previously proposed to comprise two protofilaments twisting around each other (Liu et al., 2020). However, with a cryo-EM structure determined at high resolution and an almost complete atomic model built, we show a different mastigoneme architecture which is centrosymmetric and translationally repeats along the filament axis (Fig. 2 C and Video 2). Different from actin and microtubule filaments, the mastigoneme structure is not built by individual protofilaments. The filament is non-polar, and in either direction, a mastigoneme dimer rises to ∼19.2 nm along with a 90° twist to repeat itself (the basic symmetry operator; Fig. 2 D). Two adjacent MST1 dimers form a 58-nm-long tetramer, which is the minimally required unit showing a translational symmetry from the cryo-EM 2D averages (Fig. S1 A). By repeating the basic symmetry operation four times (toward either direction as it is non-polar), the fifth MST1 dimer will simply reproduce the starting dimer by a 76.8-nm shift along the filament axis (Fig. 2 D and Video 2). Using these geometric parameters and the estimated length of the mastigoneme filament, we estimate that there are ∼100 MST1 subunits per filament.

Along the filament axis, the mastigoneme has thinner regions (∼6 nm in diameter) and thicker regions (∼10 nm in diameter; Fig. 2 D). This width is approximately the same as that of intermediate filaments (IFs), despite the fact that the overall architecture of the mastigoneme is porous and appears looser than IFs. IFs play roles in cell adhesion as well as providing mechanical support and stability to cells and tissues (Block et al., 2015). In addition to the size similarity, the non-polarity of the mastigoneme filament also resembles the IFs, implying that threadlike projections of mastigonemes may affect the mechanical behavior of the CR flagella and, eventually, the physical movement of cells.

In the mastigoneme filament, we identified three different types of MST1 pairing. Based on the relative positions of MST1 subunits and areas of interaction interfaces, we define them as dimeric pairing (Fig. 3 A), supporting pairing (Fig. 3 B), and anti-parallel pairing (Fig. 3 C). The two subunits of the dimeric MST1 tightly pack together with their SD_a–c domains wrapping around each other in a centrosymmetric manner, which is further enhanced by the hydrophobic interactions between their Ig6 domains in a back-to-back manner (Fig. 3 A). This MST1 dimer serves as the main building block of the mastigoneme filament.

In addition to the dimeric MST1 interfaces, we also identified other interaction patterns that support the filament assembly. One notable pattern is that two MST1 subunits interact with each other in a centrosymmetric manner around the C-termini of their PPII helices (Fig. 3 B). In this pattern, the N-terminal Ig_1-2 region of one MST1 subunit hands on the SD_a-b domains of the other MST1 (Fig. 3 B). We identified three notable interaction sites, which may play a key role in the overall organization of adjacent MST1 subunits: F202^A (phenylalanine 202 in subunit A) inserting into a hydrophobic pocket of the other subunit B, a salt bridge between R206^A and E1282^B, and hydrophobic interactions around F126^A, P148^A, and L1341^B. Another interaction pattern is the anti-parallel arrangement between adjacent MST1 subunits, including the PPII helix end attached to the outer surface of SD_f^B, the last two Ig domains Ig^A_8-9 clipping the PPII helix, and the Ig^A_1-2 contacting SD^B_b-c (Fig. 3 C).

Our thorough analysis of the high-resolution cryo-EM structure of the mastigoneme has revealed an interesting feature: the SD tandem of MST1 harbors an intricate array of disulfide bonds, exhibiting a well-organized spiral-like pattern with an average spacing of ∼11 Å (Fig. 4 A). These disulfide bonds are abundantly distributed throughout the entire filament, with a predominant concentration on the inner aspect of the mastigoneme (Fig. 4 B, horizontal view and top view). The pronounced clustering of disulfide bonds in the ladder-like spiral arrangement may impart structural robustness to the mastigoneme.

Disulfide bonds introduce covalent linkages into the polypeptide chain, which significantly influence protein folding and stability. This is especially crucial for cell surface proteins, where disulfide bonds are abundant, providing stability against unfolding and dissociation in the extracellular environment (Botos et al., 2016; Feige et al., 2018). In our effort to comprehend the functional role of mastigonemes, an integral component of flagella, we examined their stability under varying concentrations of DTT (a redox reagent) and captured corresponding cryo-EM images (Fig. 4 C). Our investigation revealed a clear association between higher DTT concentrations and a reduction in mastigoneme numbers, accompanied by filament disassembly. It is worth noting that as DTT concentrations increased, there was a decline in the number of intact mastigonemes (Fig. 4, C and D). Moreover, the rise in DTT concentration corresponded to an increase in the number of single filaments per micrograph, coupled with a simultaneous decrease in the occurrence of multifilament clusters (Fig. 4 E). Yet, lower DTT concentrations maintained a relatively stable frequency of intact filaments per image, suggesting the quaternary structure of the assembled filament is stable and the mastigoneme can tolerate environmental redox changes to a certain level.

Early studies have suggested that mastigoneme could influence the hydrodynamic performance of flagella by increasing their effective areas or enhancing adhesion (Namdeo et al., 2011). However, recent work employing optical-tweezers-based flow velocimetry analysis suggested that mastigonemes might not enhance swimming (Amador et al., 2020). To shed light on this matter, we conducted a comprehensive investigation of cell behaviors in cc125, cw15, and mst1 (mst1 deficient) groups under diverse conditions, including mating behavior, phototaxis, and pH shock yielding consistent results with a recent study (Amador et al., 2020; Liu et al., 2020). Across multiple parameters, no significant differences were observed among the cc125, cw15, and mst1 groups.

Careful analysis of recorded videos capturing free cell swimming (Fig. 5 A) revealed a distinct staggering-like behavior in mst1 cells compared with wild-type cells (cc125 and cw15; Video 4 and Fig. S4 A). Despite no significant changes in flagellar waveform and frequency, extensive screenings under various conditions—including photo shock (Fig. 5 B), oxidative, nitrogen or carbon-deficient conditions, different pH levels, and swimming with fluorescent beads—failed to elicit discernible alterations in most cases.

To investigate the potential role of numerous disulfide bonds in regulating flagellar motility, we conducted experiments measuring various swimming parameters at different concentrations of dithiothreitol (DTT; Fig. 5, C and D). To ensure minimal intracellular responses due to DTT uptake, all measurements were promptly performed immediately after adding different DTT concentrations to the minimal media (Sagar and Granick, 1953). The mst1 deficient cells exhibited a slightly higher proportion of stable cells compared to the wild-type cells in 1 mM DTT (Fig. 5 C, P = 0.45). However, the addition of DTT had no significant impact on cell swimming velocity, with no more than 20 mM DTT demonstrating the highest effect within a short duration of 30 s (Fig. 5 D). These findings provide valuable insights into the potential involvement of disulfide bonds in flagellar motility regulation and suggest that DTT does not exert a pronounced effect on cellular swimming behavior within the tested conditions.

Furthermore, our investigations revealed notable alterations in cell swimming in response to increased viscosity. Low viscosity did not significantly affect cell movement, but exposure to 20% Ficoll 400 led to a noticeable decrease in cell motility, particularly in the mst1 group (Fig. 5 F and Video 5). Moreover, further increasing viscosity resulted in a complete blockage of cell movement in all tested strains. All cells displayed a comparable swimming pattern (Fig. 5 E). These results underscore the critical role of viscosity in governing flagellar behavior and highlight the potential implications for cellular motility studies.

Mastigoneme in algae can be divided into two types: fine (non-tubular) hairs and stiff (tubular) hairs. In CR, mastigoneme are fine hairs (Hoek et al., 1995). Previous studies suggested that mastigoneme filaments are anchored onto PKD2, which are further linked to the axoneme. The CR PKD2 homolog in mammals may play a sensory role, leading to a series of downstream responses in cells (Liu et al., 2020). How the mastigoneme plays a role together with PKD2 remains largely unclear. Our structure clearly shows that MST1, the major component of the mastigoneme filament, is composed of tandems of Ig-like and SDs, followed by a heavily glycosylated PPII helix. Ig and SDs in cellular surfaces usually play important roles in cell–cell interactions and communications as well as extracellular signal reception (Biermann et al., 2010; Ratcliffe et al., 2001). Thus, we speculate that the key structural features we identified in the CR mastigoneme may respond to changes in environmental conditions such as redox concentrations. These responses may be converted into mechanical signals through structural changes of the entire mastigoneme assembly or interfaces among MST1 subunits, which could be remotely transmitted to and sensed by PKD2 to regulate the influx of environmental ions and subsequent cellular activities of CR. The porous nature of the mastigoneme amplifies the filament’s surface area, potentially augmenting the likelihood of stochastic interactions with chemical signals (e.g., redox agents) and enhancing its sensitivity to mechanical perturbations in the extracellular milieu, thereby stabilizing the cellular structure during motility. Furthermore, the observed disparity in gravitaxis between wild-type and mst1 cells may be attributable to mechanical alternations incurred during vertical locomotion (Fig. S4 B). Analogously, the mastigoneme might parallel the function of airplane flaps, crucial not throughout the flight but particularly significant for mechanical equilibrium during takeoff and landing.

As a microalgal species, CR experiences low Reynolds numbers, implying the predominance of viscous forces over inertial forces in its locomotion (Wu and Libchaber, 2000). As a typical puller-type microswimmer in suspensions, each CR cell utilizes its two flagella to rapidly generate bending waves that create water flow over the cell surface (Rafaï et al., 2010). The distinctive mastigoneme structure, characterized by clustered disulfide bonds and a spiral-like arrangement, may serve as robust entities capable of detaching to shield the cell from environmental challenges. Additionally, the mastigoneme’s response to viscosity variations could impact flagellar behavior, providing insights into CR’s adaptability to diverse environmental conditions for efficient cellular motility.

CR strains used in this study included the wild-type strains cw15 and cc124, which are available from the Chlamydomonas Resource Center (https://www.chlamycollection.org). The mst1 strains were kindly provided by Dr. Karl Lechtreck’s lab at the Department of Cellular Biology, University of Georgia, Athens, GA, USA. Cells were cultured in TAP (tris-acetate-phosphate) medium under continuous aeration and illumination for 3 d at 25°C to reach a cell density of ∼2.5 × 10⁷–4 × 10⁷ cells/ml.

To isolate mastigonemes, we adopted previously published protocols (Craige et al., 2013; Mitchell and Smith, 2009). Briefly, the flagella were excised from CR cell bodies following the dibucaine procedure. The purified flagella were resuspended into HMDEKP buffer (30 mM HEPES, 5 mM MgSO₄, 1 mM DTT, 0.5 mM EGTA, 25 mM KCl, 1 mM PMSF, pH 7.4). Mastigonemes were also co-isolated from central apparatus samples in a previously described method (Han et al., 2022) for cryo-EM imaging or simply imaged on the entire flagella.

Cryo-EM grids of mastigonemes were prepared using Vitrobot Mark IV (Thermo Fisher Scientific). A 3-µl aliquot of the mastigoneme sample was applied to each Quantifoil holy carbon grid (R2/1, 300 mesh gold). After a 4-s incubation at 8°C and 100% humidity, each grid was blotted and then plunged into liquid ethane at −170°C.

Data collection was automated by SerialEM software (Mastronarde, 2005). The first dataset (isolated sample) was collected on a 300-kV Titan Krios microscope (Thermo Fisher Scientific) equipped with a Bioquantum Energy Filter and a K2 Summit direct electron detector (Gatan) at Yale CCMI Electron Microscopy Facility. A total of 10,047 movies were recorded using beam-tilt–induced image-shift protocol (five images for each stage movement) with a total dose of 39.2 e/Ų and the defocus ranging from −1.2 to −2.5 µm. The second dataset (entire flagella) was collected on a 200-kV Krios microscope equipped with a Bioquantum Energy Filter and a K2 Summit direct electron detector (Gatan) at Yale CCMI Electron Microscopy Facility. A total of 15,034 movies were recorded. All details are in Table S1.

For all datasets, motion correction was performed by MotionCor2 (Zheng et al., 2017), CTF was estimated by Gctf (Zhang, 2016), and particles were picked by Gautomatch. The processing was streamlined using modified scripts (https://github.com/JackZhang-Lab/). Particles were extracted and imported into cryoSPARC v3 (Punjani et al., 2017) for all subsequent steps of data processing.

Approximately 2,000 particles were manually extracted from the micrographs using a 360 × 360 box size for the initial two-dimensional (2D) classification. 20 high-quality 2D averages were then chosen for automatic particle picking by Gautomatch, yielding a total of 119,702 particles. Several rounds of 2D classification were conducted to isolate the highest-quality particles. Given the overlapping regions in the extracted filaments, there was a potential bias in the initial reference reconstruction due to strong signal interference. Our attempts to achieve a reliable 3D reconstruction using previously reported helical parameters were unsuccessful, indicating unknown mastigoneme filament parameters. To circumvent these challenges and ensure accurate reconstruction, multiple rounds of ab-initio reconstruction were undertaken with varying parameters. The most promising initial models were chosen for subsequent high-resolution 3D refinement. Several rounds of heterogeneous refinement were performed on the top candidate models. The model of the highest resolution was chosen for additional refinement, targeting a near-atomic resolution map. A variety of refinement techniques were employed to progressively improve the maps. These included homogeneous refinement, local alignment refinement using multiple masks for distinct regions, optimization of high-order attenuation parameters for individual optical groups, and local CTF refinement, similar to what we described previously (Ton et al., 2023). After the refinement, particle coordinates were recentered, extended along the filament’s primary axis, and re-extracted from the original micrographs to produce a new particle set using the multicurve fitting approach as we previously developed for structure determination of large, flexible motors bound to microtubules (Rao et al., 2021; Chai et al., 2022). Ultimately, most regions of the mastigoneme filament were refined to a resolution around 3.4 Å, enabling atomic model construction with side chains assigned in Coot (Emsley and Cowtan, 2004; see Fig. 2).

Most of the regions were refined at a 3.4-Å resolution or better, which allowed us to de novo build atomic models with side chains assigned in Coot. For the slightly worse regions, we were able to build backbone models with the residues assigned on the basis of the relative positions among the large residues (such as Try and Arg) of each domain. For the regions that showed a clear backbone with low-quality side-chain density, we coarsely assigned the residues using the predicted domain model from Phyre2 (Kelley et al., 2015; before AlphaFold), which were further refined using models of individual domains predicted by AlphaFold2 (Jumper et al., 2021). For regions that were solved at a resolution with clear secondary structures, we fitted the predicted models into the density as rigid bodies in Chimera. All models at a resolution of better than 4 Å were automatically refined by Refmac (Murshudov et al., 2011) and manually checked in Coot. The process was repeated until all parameters were reasonably refined. The figures and movies were created by ChimeraX (Goddard et al., 2018).

Cryo-ET grids of CR flagella and axoneme (with 1% IGEPAL CA-630 on ice for 10 min to solubilize flagellar membrane) were prepared with Vitrobot Mark IV or manual plunger. Tomographic datasets were collected on the 300-kV Titan Krios equipped with a K2 detector. SerialEM was used for automatic data collection under the bidirectional scheme at a 3° interval and tilt angles ranging from −51° to 51°. Each of the final tilt series contained five movie stacks with a pixel size of 1.33 Å, defocus at −5 µm, and an accumulative dose of 120 e⁻/Ų. All cryo-ET datasets were processed with IMOD (Kremer et al., 1996).

About 20 μl of cells at a concentration of 1 × 10⁶ cells/ml suspended in a minimal medium (Sagar and Granick, 1953) was dropped on the glass slide, carefully covered with a coverslip, and the swimming tracks and swimming velocity were measured by tracking the moving cells recorded with bright-field microscopy (Nikon ECLIPSE Ti-U) with a high-speed camera (PCO.edge; PCO Tech). To rule out the effect of photo shock induced by white light on the flagellar waveform, red light (a white light through a red filter λ = 620 nm) was used to record the swimming using the same procedures. High-speed camera recordings were taken for 30 s (using a 20× objective at 350 frames/s) and 7 s (using a 100× oil immerse objective at 500 frames/s; Zhao et al., 2021). Subsequently, the videos were processed using Fiji (Schindelin et al., 2012) with the Trackmate plugin (Tinevez et al., 2017).

To assess the impact of different concentrations of DTT on cell swimming, we analyzed the ratio of cells with mean speeds >50 μm/s to those with mean speeds <50 μm/s. Treatment with various DTT concentrations (1, 5, and 20 mM) did not result in significant changes in cell swimming behavior. The number of cells tested for each DTT group was as follows: n = 65 for 1 mM DTT, n = 67 for 5 mM DTT, and n = 89 for 20 mM DTT. Error bars indicate the standard deviations for the means.

For experiments involving increased viscosity, appropriate volumes of 10, 20, and 30% wt/vol Ficoll 400 solution were added to a reduced liquid culture volume to achieve a total mixed volume of 1 ml (Wilson et al., 2015). The Ficoll solution was carefully pipetted into the liquid culture to ensure complete mixing. Regarding the influence of viscosity on cell movement, the mst1 mutant group (n = 74 in 10% Ficoll, n = 41 in 20% Ficoll, and n = 36 in 30% Ficoll) exhibited greater sensitivity compared with the cc125 (n = 100 in 10% Ficoll, n = 58 in 20% Ficoll, and n = 43 in 30% Ficoll) and cw15 (n = 74 in 10% Ficoll, n = 58 in 20% Ficoll, and n = 43 in 30% Ficoll) strains.

Statistical analysis was performed using ordinary one-way ANOVA among the three groups, followed by unpaired t-tests to compare the mst1 group with cw15 or cc125, respectively. Statistical significance is denoted as *P < 0.05, and error bars represent standard deviations for the means.

The cells were cultured to the mid-log phase. The cell number is near 3 × 10⁶ cells/ml. Cells were harvested by centrifugation at 600 g for 5 min at room temperature. The cells were washed twice in phototaxis buffer (5 mM HEPES, pH 7.4, 0.2 mM EGTA, 1 mM KCl, 0.3 mM CaCl₂) and suspended at 2 × 10⁶ cells/ml in the Ep tube. The EP tube was placed and shaken for 30 min under light for recovery. Then, cell suspension was transferred into 100-μl glass tubes (2-000-100; Drummond), and its upper and lower ends were sealed with dental wax. The tubes were placed upright under red-light illumination for 30 min and then given white light illumination for 30, 60, 90, 120, and 150 min. Then the tubes were cut into halves, and the cell number on each side was measured. The degree of gravitaxis index (GI) was calibrated as GI = (n_l − n_u)/(n_l + n_u),where n_u represents the cell number of the upper part and n_l is the lower part, respectively. If the cell floats to the upper part, the GI value will be <0, which means the cells will have negative gravitaxis. Otherwise, the GI value will be negative (Yoshimura et al., 2003).

In the supplemental materials, we provide a comprehensive set of figures, tables, and videos to enhance the understanding of our research findings. Fig. S1 presents cryo-EM 2D class averages of MST1 and a cryo-EM flowchart. Fig. S2 delves into the structural analysis of MST1 Ig domains. In Fig. S3, we directly detect glycosylation modifications in the cryo-EM map. Fig. S4 explores the influence of mastigonemes on cell swimming behavior and viscosity sensitivity. Table S1 offers detailed cryo-EM data collection, refinement, and validation statistics. For a visual perspective, Video 1 showcases mastigoneme visualization via cryo-ET slices. Video 2 unveils the precise structural details of the MST1 subunit through cryo-EM. Video 3 provides insights into the 3D viabilities of PPII helices, revealing substantial lateral displacement. Videos 4 and 5 illustrate the free swimming behavior of cc125, cw15, and mst1 strains in both minimal medium and 20% Ficoll 400 minimal medium, highlighting noteworthy observations related to cell motility under varying conditions.

Cryo-EM maps and atomic coordinates have been deposited in the Electron Microscopy Data Bank (EMDB) and the Protein Data Bank (PDB) under the accession codes EMD-41674/PDB 8TX1, EMD-41679/PDB 8TXB, EMD-41680/PDB 8TXC, and EMD-41696. All locally refined maps used in the final model building have been deposited as additional maps associated with these codes. Source data are provided in this paper.

We express our gratitude to K. Lechtreck for generously sharing the cell lines and engaging in valuable discussions. We thank J. Lin, S. Wu, and K. Zhou at Yale University for their valuable technical support on electron microscopy. We wish to extend our gratitude to L. Han for his contribution to the data collection process. Furthermore, we express our appreciation to W. Zheng for her assistance with the structure refinement.

This work was supported by start-up funds from Yale University and the National Institutes of Health grant R35GM142959 awarded to K. Zhang, S10OD023603 awarded to F. Sigworth at Yale University, and Rudolf J. Anderson Fellowship awards to Y. Wang.

Author contributions: Y. Wang and K. Zhang conceptualized and designed the experiment. Y. Wang, J. Yang, and Y. Yang prepared the samples for structural studies. J. Yang and K. Zhou collected the cryo-EM data. Y. Wang, Y. Yang, and K. Zhou determined the cryo-EM structures, built the model, and analyzed the structures. F. Hu and Y. Wang executed the C. reinhardtii movement and gravitaxis assays. Y. Wang and K. Zhang, with important inputs from K. Huang, analyzed the results of cell movement. Y. Wang and K. Zhang wrote the paper.