Poc1 bridges basal body inner junctions to promote triplet microtubule integrity and connections

Basal bodies (BBs) and centrioles organize cilia and centrosomes, respectively, and are one of the most evolutionarily ancient structures of eukaryotic cells (Carvalho-Santos et al., 2011). They have diverse and essential functions in cell signaling, cell division, spermatogenesis, and development. These functions depend largely on the role of BBs in the nucleation and anchoring of cilia to the cell cortex (Carvalho-Santos et al., 2011; Wickstead and Gull, 2011). BBs are microtubule (MT)-based structures comprised of nine radially arranged triplet MTs (TMTs) creating a cylinder ∼450 nm in length and 200 nm in diameter. TMTs consist of one full, 13-protofilament MT (the A-tubule), and two partial MTs (the B- and C-tubules), which are assembled on the exterior side of the previous tubule. The positions where the tubules meet are referred to as junctions, and they are further specified as inner junctions if they face the BB lumen or outer junctions if they are on the BB exterior. BBs dock to the cell cortex and the ciliary axoneme is formed from doublet MTs (DMTs) that extend continuously from the A- and B-tubules of the BB TMTs. The resulting cilia have unique functions depending on whether or not they are motile. Immotile cilia primarily operate as a signaling and environment-sensing organelle. On the other hand, motile cilia contain axonemal dynein motor proteins that slide the cilia DMTs relative to each other and, when resisted by the Nexin-Dynein Regulatory Complex (N-DRC) and the BB, create a beat stroke that is important for fluid movement and cell motility. When DMTs of the cilium are continuous with the TMTs of the BB, the forces generated by ciliary beating are transmitted directly to the BB (Riedel-Kruse et al., 2007; Junker et al., 2022; Meehl et al., 2016). Thus, BBs must resist these forces both to maintain their structural integrity, as well as to transmit those forces to the cell. BB structural stability is thought to be achieved in two ways. First, BB TMTs are subject to myriad posttranslational modifications that confer stabilizing properties to the MT lattice (Wloga et al., 2017). Second, BBs contain hundreds of proteins that are associated with TMTs (Li et al., 2012; Kilburn et al., 2007; Le Guennec et al., 2020; Andersen et al., 2003; Keller et al., 2005). These proteins can stabilize the MT lattice of the TMT directly and/or connect neighboring TMTs, thereby promoting TMT organization and BB shape. How these molecules establish stabilizing structures, the identity of the full repertoire of these molecules, and how they reinforce the BB architecture remain poorly understood.

BBs from a wide range of organisms share the same structural features, and these features highlight how the nine TMTs of BBs are assembled and supported. Although species-specific differences exist, most commonly, the proximal region of the BB (∼100 nm) harbors the cartwheel, which is comprised of stacks of nine Sas6 homodimers (Allen, 1969; Dippell, 1968; Nakazawa et al., 2007). Sas6 forms radial spokes that emanate from a central hub (Kitagawa et al., 2011; van Breugel et al., 2011). The cartwheel is thought to, at least in part, establish the ninefold arrangement of TMTs (Nakazawa et al., 2007; Hilbert et al., 2016). Each cartwheel spoke terminates with a density called the pinhead that interacts with the A-tubule of each TMT. CEP135/Bld10 is an important component of the pinhead (Hiraki et al., 2007; Lin et al., 2013; Guichard et al., 2017). A/C linkers also reside in the proximal region of the BB (∼150 nm), where they connect the C-tubule of one TMT to the A-tubule of its adjacent TMT (Allen, 1969; Li et al., 2012). To date, molecules that comprise the A/C linkers have not been identified. In the medial ∼300 nm of the BB, termed the core region, an inner scaffold attaches to and reinforces the nine TMTs, along with unique core region A/C linking structures (Le Guennec et al., 2020; Allen, 1969). Recent advances in Ultrastructure Expansion Microscopy (U-ExM) led to the identification of components of the inner scaffold in human centrioles, namely, Fam161A, Poc1B, Centrin2, Poc5, HAUS6, γ-tubulin, and CCDC15 (Arslanhan et al., 2023; Le Guennec et al., 2020; Schweizer et al., 2021). However, the precise localization of these proteins in the scaffold architecture and how they link to the TMTs is not well understood. WDR90/Poc16 is a core region-localized protein that is important for the recruitment of proteins to the inner scaffold (Steib et al., 2020; Hamel et al., 2017). The combination of electron microscopy and super-resolution light microscopy has thus proved to be a powerful means to dissect structural and compositional features of the BB, but the structure–function relationships that promote BB stability in the face of cilia-generated forces are still lacking.

To understand how the BB architecture enables force resistance, we studied the conserved BB stability and disease-related protein, Poc1 (Keller et al., 2009; Pearson et al., 2009a; Beck et al., 2014; Li et al., 2021; Hua et al., 2023). Poc1 is necessary for centriole integrity and enables BBs to resist mechanical forces from beating cilia. Studies in the ciliate Tetrahymena thermophila revealed that in the absence of Poc1, BBs assemble the stereotypical ninefold symmetric BB structure that appears normal, but elevated cilia forces induced by increased temperature cause the BBs to disassemble (Pearson et al., 2009a; Meehl et al., 2016). Temperature impacts ciliary beat frequency, thus higher temperature imposes more forces on the basal body (Jing et al., 2017; Junker et al., 2022). Therefore, Poc1 loss provides a unique opportunity to understand the structural features of BBs that specifically promote BB stability. Identifying the precise localization of Poc1 within the BB structure is central to this goal. One of the human homologs of Poc1, human Poc1B, localizes to the BB core region and is proposed to be a component of the inner scaffold (Le Guennec et al., 2020). However, in Tetrahymena, a significant portion of the Poc1 protein population localizes to the BB proximal end (Pearson et al., 2009a). Consistent with a proximal localization of Poc1, poc1Δ mutants in Tetrahymena partially disrupt A/C linkers, leading to the hypothesis that Poc1 is an integral component of the A/C linkers, or at least stabilizes them (Meehl et al., 2016; Li et al., 2019). In addition, Poc1 is a known MT-binding protein (Hames et al., 2008). Therefore, how and if Poc1 contributes to structural features unique to different BB domains is unclear.

Poc1 is a WD40 domain-containing protein comprised of a seven-bladed β-propeller at its N-terminus that forms a doughnut-shaped structure. WD40 domains commonly provide scaffolds for the assembly of protein complexes where components are recruited at different faces of the doughnut (Xu and Min, 2011). In addition, Poc1 contains a short N-terminal extension before the WD40 domain (37 amino acids in TtPoc1), and a longer, C-terminal extension after the WD40 domain (292 amino acids in TtPoc1), which likely exits the WD40 motif from the same β-propeller blade (Hudson and Cooley, 2008). Although the WD40 domain alone is sufficient to rescue the BB loss phenotype in the Tetrahymena poc1Δ mutant, Poc1’s C-terminal extension is conserved, suggesting it plays important roles as well (Pearson et al., 2009a; Fourrage et al., 2010). Most of the C-terminal extension is predicted to be unstructured in an unbound, apo state, but the very end of the C-terminal extension contains a predicted helix-forming region that has coiled-coil forming propensity. This helical region in human Poc1B directly interacts with the Fam161A inner scaffold protein (Roosing et al., 2014). Thus, Poc1 may use its WD40 domain as well as its N- and C-terminal extensions for its BB functions.

To understand how BBs resist ciliary forces, we used cryogenic electron tomography (cryoET) and subtomogram averaging to solve the structure of the Tetrahymena BB at different longitudinal positions of the BB. Focused refinement of subregions of the TMT structure reached a subnanometer resolution. This approach, using wild-type and poc1Δ BBs, enabled us to identify the location of Poc1 at the inner junctions of the TMTs, as well as to understand the role of Poc1 in the establishment of stabilizing structures of the BB. We propose that Poc1 seals MTs together on the luminal side of the BB and stabilizes the cartwheel pinheads at the BB proximal end and the inner scaffold in the BB central core. This promotes TMT interconnections throughout most of the BB length, which are critical for the organelle’s shape maintenance and force resistance. Together, these findings identify the TMT inner junctions as a key location from which multiple BB stabilizing strategies are organized.

To identify the 3D high-resolution structure of Tetrahymena BBs using cryoET, we developed a procedure to gently isolate BBs while preserving overall BB structure (Fig. 1, A and B; and Video 1). Subtomogram averages were generated on TMTs from the BB proximal 150 nm, the central core 300 nm, and the distal 100 nm at 19–27 Å resolution (Fig. 1, C–E and Fig. S1; and Table 1). As in other organisms, the proximal region contained pinhead densities attached to the A-tubules (protofilaments A03-A04), as well as A/C linkers between the A- and C-tubules (protofilaments A06-08, and C07-10, Fig. 1 C and Fig. S2 A). The core region contained an inner scaffold (Fig. 1 D and Fig. S2 B). Finally, the subtomogram average from the BB distal region lacked these MT interconnecting structures and consisted of only DMTs. Some loss of the C-tubule was observed in the subtomogram average in the core region and at the distal end of the BB. While nearly all TMTs in the proximal region were complete TMTs, the core region showed only 9.3% complete TMTs. Given that TMTs are apparent in the core region in thin section EMs of Tetrahymena BBs, the C-tubule loss likely occurred as a result of the isolation and/or freezing process (Giddings et al., 2010; Allen, 1969). This suggests that the core region C-tubule is less stable than the A- and B-tubules. In summary, the 3D domain architecture of Tetrahymena BBs is comparable with other organisms (Fig. S2).

By comparing our cryoET structures to those previously reported for other organisms, several consensus features can be identified. The proximal end structure of the TMT is highly conserved in Tetrahymena, Chlamydomonas, and Paramecium (Fig. S2 A; [Li et al., 2019; Klena et al., 2020]). The positions and shape of the pinheads are similar among Tetrahymena, Paramecium, and Chlamydomonas BBs. In contrast, the A/C linkers are more variable in structure. The A/C linker in Chlamydomonas has an X-shaped configuration, but this is not the case in Tetrahymena, which lacks one leg of the X-shaped structure. Structural differences in the A/C linker are also observed between the Paramecium and Tetrahymena BBs, with the Paramecium A/C linkers appearing less dense. However, the degree to which these differences are species-specific or are due to the difference in resolution and/or flexibility is unclear (Fig. S2 A). Side views of the TMTs from the BB lumen show high structural conservation between organisms. In particular, the structure and periodicity of the pinhead and densities of the A–B inner junction are similar; all exhibit 8 nm longitudinal periodicity, consistent with the periodicity of tubulin heterodimers in the MT lattice (Fig. S2 A, bottom panels).

The core region and inner scaffold structure are more variable than the proximal region (Fig. S2 B). Tetrahymena and Paramecium TMTs have primary attachment points to the inner scaffold at protofilament A03 and at the A–B and B–C inner junctions. However, the A03 attachment site is missing in the Chlamydomonas structure. In addition, the shape of the B–C inner junction attachment site is divergent compared with Tetrahymena and Paramecium. In Tetrahymena and Paramecium, the densities at the B–C inner junction are shorter and hold the scaffold more closely to the TMT inner walls, whereas in Chlamydomonas BBs, the B–C inner junction attachment densities create an elongated V-shape that holds the scaffold further away from the TMT walls. Despite these differences, the A–B inner junction attachment structure is nearly identical between all three species. We cannot rule out that the other attachments are simply more flexible in Chlamydomonas leading to their reduced density, or that they are invisible in this lower-resolution structure. Regardless, the A–B inner junction is the most consistent attachment site in all three organisms compared here. In summary, the structure of the TMTs is broadly conserved throughout the BB, with most differences found in the core region. Within the core region, the attachment site of the inner scaffold at the A–B inner junction (the so-called “stem”; Le Guennec et al., 2020) is the most conserved attachment. The molecular composition of these structures and the mechanism of how they contribute to BB assembly and stability remain poorly understood.

Tetrahymena poc1Δ cells assemble BBs and cilia that function under normal ciliary force conditions (growth at 30°C) but disassemble at high force (growth at 38°C; Fig. S3 A; Pearson et al., 2009a). Inhibition of ciliary beating using NiCl₂ rescues BB loss in poc1Δ cells at high temperatures, verifying that the disassembly is due to ciliary forces (Fig. S3 A; Meehl et al., 2016). To understand how the Poc1 stability protein localizes in the BB architecture, Ultrastructure Expansion Microscopy imaged with Structured Illumination Microscopy (U-ExM-SIM) was used to determine Poc1 localization within BBs (Gambarotto et al., 2019). In cross-sectional views, inducible GFP:Poc1 expressed in poc1Δ cells localizes to the interior TMT walls of BBs, creating a peak of ring-shaped intensity just interior to the peak of ring-shaped tubulin signal (Fig. 2 A, top images). Longitudinal views of BBs revealed that GFP:Poc1 localizes through the length of the BB to include the proximal and core regions, with a slight reduction of signal between the two domains (Fig. 2 A, bottom images, arrowhead). Similar localization was observed using endogenously expressed C-terminal tags of Poc1, with the exception that clear rings were not observed in cross-sectional views, suggesting either that the C-terminus is located interior to the N-terminus in the BB, or that it is dynamic, resulting in diffuse signal (Fig. S3 B). Previous immunoEM studies showed localization of Poc1 through the length of the BB, consistent with the localization pattern observed here (Pearson et al., 2009a). In summary, Tetrahymena Poc1 localizes to the BB proximal and core regions, indicative of functions in both BB regions.

To identify the precise localization of Poc1 in the cryoET structure, we searched for Poc1’s WD40 domain in the BB proximal region subtomogram average, focusing on the BB luminal side of the TMTs. We expected Poc1’s density to be shaped like a doughnut due to the seven-bladed β-propeller WD40 predicted structure. Indeed, a doughnut-shaped density was identified at the A–B inner junction (Fig. 2 B). Focused refinements on the A–B and B–C inner junctions substantially improved the resolution of the structure in these regions. Both inner junctions show a WD40 domain density where seven β-propeller blades can be resolved (Fig. 2, C and D). Rigid body fitting of the AlphaFold predicted Tetrahymena Poc1 WD40 domain (UniProt ID Q229Z6) shows it fits well in the density (Fig. 2, C and D; and Video 2; see Materials and methods).

To determine whether these WD40 densities were indeed the Poc1 protein, we prepared BBs from Tetrahymena poc1Δ cells using the BB isolation procedure described above (Video 3). The subtomogram average of poc1Δ BBs revealed a variety of TMT defects compared with the wild-type structure, including varying degrees of B- and C-tubule loss throughout the length of the BB. However, focusing on the subclass of complete TMTs from poc1Δ BBs showed that the WD40 densities at the A–B and B–C inner junctions were lost when compared with the wild-type structure, indicating they are indeed the Poc1 protein (Fig. 2, E and F). In both inner junctions, Poc1 displays 8 nm longitudinal periodicity (Fig. 2 F). In the A–B inner junction, Poc1 is the only protein found, whereas in the B–C inner junction, it is interspersed with an unknown protein that is retained in the absence of Poc1 (Fig. 2 F). Thus, Poc1 localizes to the A–B and B–C inner junctions at the proximal end of Tetrahymena BBs (Fig. 2, G and H).

Based on our highest resolution structures and by using the AlphaFold predicted model of Poc1 and previously determined α- and β-tubulin structure (PDB ID 8G2Z), we built a pseudo-atomic model for the A–B and B–C inner junctions, showing details of how Poc1 links the A- and B-tubules and the B- and C-tubules at the junctions (Fig. 2, C and D; and Video 2). In this model, the WD40 motif of Poc1 is tilted 45° relative to the longitudinal direction of the BB. The WD40 domain connects protofilaments A01 and B10 by arranging multiple copies of Poc1 protein every 8 nm longitudinally, effectively sealing the inner junctions of the TMT. Our model predicts that a loop in propeller blade two from Poc1’s WD40 domain binds to helix 12 of α-tubulin from protofilament A01. Blades three and four face protofilament B10, potentially making contact with an α-β-tubulin heterodimer in protofilament B10 at its longitudinal intradimer interface. In the A–B inner junction, a density that appears to be the C-terminal end of Poc1 exits the WD40 domain at blade number seven. It forms a long coil, horizontally traversing the BB-luminal wall of the B-tubule, binding to protofilaments B09 and B08. A similar binding mode of the C-terminus exiting the WD40 domain is observed at the B–C inner junction. This long coil observed in the average density map is consistent with the predicted structure of the extended C-terminal tail of Poc1. In this model, the C-terminus of Poc1 at the A–B inner junction can be traced to approximately residue Q402 near protofilament B09. Residues 403–634 are not visible. The C-terminus of Poc1 at the B–C inner junction can be traced to residue Q375 near protofilament C10. A similar extended coil density that may be the Poc1 N-terminus exits the WD40 domain at the A–B inner junction, horizontally traverses a gap in the junction, and binds the BB-luminal wall of the A-tubule at protofilament A02 (Fig. 2 C and Video 2). However, because a similar density was not observed at the B–C inner junction, we did not model it and more work should be done to understand if this is indeed the N-terminus of Poc1. Together, the subnanometer resolution structure refinement coupled with structural modeling reveals details of how Poc1 recognizes BB TMTs through the WD40 domain and potentially its N- and C-termini.

Consistent with the in vivo BB instability found in poc1Δ cells grown in high-force conditions, isolated poc1Δ BBs displayed varying degrees of structural disintegration in vitro as assessed by negative stain EM, despite not having been exposed to high-force conditions, indicating their intrinsic structural instability (Fig. S3 C, compare white arrowhead to red arrowhead). The cryoET and subtomogram averaging of poc1Δ BBs showed partial B-tubules and missing C-tubules in the majority of BBs, inconsistent with prior EM-tomography studies showing that poc1Δ BB TMTs in situ at normal force are mostly complete (Meehl et al., 2016). This suggests that the tubule loss observed in the subtomogram averages likely occurred during BB isolation and freezing, as was observed for the wild-type BB in the core region. Subtomogram classification of poc1Δ TMTs in the proximal region showed substantial structural heterogeneity in different parts of the TMT. Only 8% of poc1Δ subtomograms contained complete TMTs with full A-, B-, and C-tubules. In contrast, 99% of wild-type BBs contained complete TMTs in the proximal region. This highlights that the poc1Δ structure is more labile than the wild-type structure, and differences in poc1Δ compared with wild-type BBs reveal the parts of the TMT that are sensitive to Poc1 loss, whether under forces from cilia or from BB isolation.

The direct binding of Poc1 to the TMTs suggested that Poc1 is required to promote closure of the B- and C-tubules against the A- and B-tubules, respectively, which in turn promotes TMT stability. Indeed, U-ExM-SIM of poc1Δ BBs in whole cells showed loss of entire TMTs (Fig. 3 A). Consistent with these observations and as discussed above, only 8% of poc1Δ TMT subtomograms contained complete TMTs. Our cryoET and subtomogram averaging therefore provides the opportunity to dissect this MT loss in more detail. Classification of the poc1Δ subtomograms showed various degrees of loss of the B- and C-tubule protofilaments (Fig. 3, C and D; and Fig. S4 A). No defects in A-tubule protofilament number were observed. Protofilament loss from the B- and C-tubules occurs closest to the inner junctions. Classification focused on the B-tubule shows loss of the B10 protofilament that is immediately adjacent to the inner junction in most classes, as well as protofilaments B9 and down in protofilament numbering as the severity of the structural disintegration increases (Fig. 3 C, classes B3-6). When all B-tubule protofilaments are present, two subclasses accounting for approximately half of the subtomograms were observed (Fig. 3 C, classes B1 and B2). The major difference between these two subclasses resides in densities at the A–B inner junction both in the B-tubule MT lumen and on the BB luminal side, where different proteins structurally distinct from Poc1 appear to keep the B-tubule attached to the A-tubule. In all classes, densities for the C-tubule are weak, indicating that C-tubule loss precedes B-tubule defects (Fig. 3 C, all classes). Classification focused on the C-tubule reveals, as with the B-tubule, subclasses with protofilament loss from the inner junction and/or flexibility of those protofilaments, indicated by a smeared signal (Fig. 3 D, classes C4 and C5). Three classes of complete TMTs were identified and combined to generate the poc1Δ complete TMT subtomogram average used in Fig. 2 (Fig. 3 D, classes C1-3). In summary, both U-ExM-SIM and focused classification of TMT subtomograms showed that Poc1 in the proximal region promotes protofilament attachment at the inner junctions, thereby stabilizing TMTs directly.

We next asked whether Poc1 stabilizes additional structures of the BB proximal end. A/C linkers were often disrupted in poc1Δ mutants even though detectable cartwheel formation and general BB assembly were unperturbed (Meehl et al., 2016). In both wild-type and poc1Δ complete TMT subtomogram averages, A/C linkers are present, but their density is smeared (Fig. 2, E and F). This indicates flexibility or heterogeneity of the structure, even in wild-type BBs, making it difficult to assess quantitatively or qualitatively the differences between these structures in wild-type versus poc1Δ BBs. Thus, consistent with previous reports, A/C linkers can form in the absence of Poc1 protein (Meehl et al., 2016).

The pinhead links the cartwheel structure to the TMTs. A primary component of the pinhead is the α-helical protein, Bld10/CEP135 (Hiraki et al., 2007; Lin et al., 2013; Guichard et al., 2017). To determine whether Poc1 impacts the localization of Tetrahymena Bld10, Bld10:mCherry levels were measured in both wild-type and poc1Δ cells (Fig. 4 A). No difference was detected in the amount of Bld10 localized to wild-type versus poc1Δ BBs at normal cilia force conditions (30°C). However, under high-force conditions created by shifting cells to 38°C for 12 h, Bld10 levels were reduced by ∼60% at BBs in poc1Δ cells (Fig. 4 A). Thus, Poc1 is not required for Bld10 assembly at the BB but does stabilize Bld10 in the face of elevated ciliary force.

The loss of Bld10 at elevated force in poc1Δ BBs led us to ask whether the structure of the pinhead was impacted by the loss of Poc1. Indeed, the pinhead density was reduced in poc1Δ complete TMTs compared with wild-type TMTs (Fig. 4 B). Thus, despite not being a primary component of the pinhead, Poc1 loss causes the pinhead to be unstable.

In summary, Poc1 stabilizes the proximal region of BBs. First, Poc1 stabilizes TMT tubule connectivity by binding directly to the MT protofilaments in the A–B and B–C inner junctions, sealing the tubules together. Second, Poc1 promotes stability of the pinhead structure that connects the TMTs to the cartwheel. This ultimately organizes and stabilizes TMT interconnections in the BB proximal region.

Given that Poc1 localizes in the BB core region by U-ExM-SIM, we studied its contribution to the BB core structure, function, and molecular composition. Tetrahymena homologs of the centriole and BB core region localizing proteins, Fam161A and Poc16/WDR90, were identified using BLAST analyses of the human proteins. Fam161A is an inner scaffold protein that binds to MTs and interacts with Poc1B in human centrioles and yeast two-hybrid studies (Roosing et al., 2014; Le Guennec et al., 2020). Poc16 also localizes in the core region but is not thought to be an inner scaffold protein (Steib et al., 2020). Endogenous, mCherry-tagged Fam161 and Poc16 localized to Tetrahymena BBs. Both proteins colocalized with Poc1 in the core region (Fig. 5 A). In addition to its localization in the core region, Fam161A localized to the BB distal end, at what is assumed to be the transition zone at the cilium base (Fig. 5 A; and Fig. S5, A and B). The distal localized population of Fam161A did not colocalize with Poc1. Thus, like Poc1, Fam161A and Poc16 localize to the core region in Tetrahymena BBs.

The core region is characterized by the inner scaffold, which is evident in our cryoET subtomogram average of Tetrahymena BBs (Fig. 5, B and C). The overall structure of the inner scaffold is consistent with other organisms previously reported (Li et al., 2019; Klena et al., 2020; Le Guennec et al., 2020). In Tetrahymena, the inner scaffold repeat is an elongated structure attached to the luminal side of the TMT. It binds to the TMTs at multiple sites, as described in Fig. S2 B; this includes protofilament A03 of the A-tubule, the A–B inner junction, and the B–C inner junction. The inner scaffold exhibits 8 nm longitudinal periodicity. Laterally, it connects to the neighboring repeats near the A-tubule and the B–C inner junction of the TMT. Backfitting the core region TMT structure onto BBs revealed that, in Tetrahymena, the inner scaffold is comprised of stacked rings that line the luminal circumference of the BB (Fig. 5 C and Fig. S4 E). The density for the inner scaffold in wild-type BBs is not as well defined as the density for the TMTs themselves, indicating that the inner scaffold structure in the subtomogram average is heterogeneous or flexible. Nonetheless, the structure shows that the inner scaffold forms an interwoven meshwork in the lumen of the BB core region that holds the TMTs together in a manner that promotes structural integrity in the core region.

To determine whether Poc1 localizes to the A–B and B–C inner junctions in the core region, as found for the proximal region, we asked whether core region inner junctions also displayed WD40 densities. Focused refinement on the A–B inner junction revealed that, indeed, a density with seven β-propeller blades is present and an Alphafold model of Tetrahymena Poc1 can be docked into the structure (Fig. 5 D and Video 4). The density was absent from the subclass of complete TMTs from poc1Δ BBs in the core region, validating that it is indeed the density for Poc1 (Fig. 5 E). This suggests Poc1 localizes to the A–B inner junction in the BB core region, as in the proximal region (Fig. 5 E and Fig. 2 C). Unlike in the proximal region, the BB core region B–C inner junction does not contain a Poc1 WD40 density. Rather, it is comprised of bulky, non-tubulin densities that are distinct from the Poc1-containing density found at the B–C inner junction at the BB proximal region (Fig. S4, B–D). This bulky density bridges a larger area between the C-tubule protofilament and the B-tubule than Poc1’s WD40 domain would be able to do on its own, suggesting a different protein or set of proteins occupies this space. Moreover, a comparison of the TMTs of poc1Δ BBs to wild-type TMTs at the B–C inner junction in the core region revealed no structural differences or lost density at the current resolution (Fig. S4 D). This suggests that Poc1 is not a component of the B–C inner junction in the core of the BB. In summary, within the BB core region, Poc1 occupies the A–B inner junction, a major site of attachment of the inner scaffold to the TMTs (Fig. 5, F and G).

The A–B inner junction stem attachment is the most consistent structure linking the inner scaffold to TMTs across organisms and potentially serves as the major inner scaffold-TMT attachment site (Fig. S2 B). Given Poc1’s localization at the A–B inner junction stem, we hypothesized that Poc1 influences the localization and stability of the Fam161A inner scaffold protein at BBs in vivo. Endogenously tagged Fam161A:mCherry localizes to BBs in poc1Δ cells under normal force conditions. However, this localization is decreased relative to wild-type cells (Fig. 6 A, top panels). This is consistent not only with human Poc1B direct binding to Fam161A but also shows that Fam161A loading in BBs is not entirely dependent on Poc1, suggesting there are redundant mechanisms to recruit Fam161A to BBs. When cells were shifted to high ciliary force conditions, Fam161A levels at the BB decreased by ∼60% relative to poc1Δ cells at normal force. A similar decrease in Fam161A levels was also observed in wild-type cells after temperature shift, suggesting that Fam161A is not stable in BBs at elevated ciliary force. Because Fam161A localizes to both the core and presumed transition zone BB regions (Fig. 5 A), we asked whether one or both Fam161A populations are decreased in high ciliary force conditions. In both wild-type and poc1Δ cells, Fam161A was preferentially lost in the core region (Fig. S5 B). This is consistent with Fam161A’s colocalization with Poc1 only in the core region.

To determine whether Poc1 influences Poc16 localization to BBs, we visualized Poc16:mCherry in wild-type and poc1Δ cells. Poc16:mCherry intensity at BBs was unchanged between wild-type and poc1Δ cells in normal force conditions (Fig. 6 A, bottom panels). However, Poc16 levels decreased by ∼70% in poc1Δ BBs at elevated ciliary force but were unchanged in wild-type BBs. These results suggest that Poc1 is partially required for Fam161A, but not Poc16, binding in the core region of the BB under normal force conditions, and that Poc1 stabilizes both Fam161A and Poc16 at BBs under high ciliary force conditions.

Because Fam161A was decreased in the core region in poc1Δ cells, we asked whether the inner scaffold is compromised in the poc1Δ BB structure. Indeed, a near complete loss of the inner scaffold was observed in the poc1Δ TMT structure from the core region, even when comparing the subclass of complete TMTs from both wild-type and poc1Δ BBs (Fig. 6 B). The only remaining density appears at the A03 attachment site and is blurry compared with the MTs. This indicates decreased occupancy and/or increased flexibility of the molecules in that region. Therefore, in accordance with our fluorescence data, Poc1 stabilizes the inner scaffold.

To understand how the A–B inner junction attachment specifically was impacted by the loss of Poc1, we characterized the structure of the stem at a higher resolution. In wild-type BBs, the Poc1 WD40 motif density in the A–B inner junction directly binds another protein of unknown identity with a barrel-shaped tertiary fold (Fig. 6 C, left). This second protein attaches to the elongated horizontal density of the inner scaffold. In the absence of Poc1, the barrel density is missing, suggesting that Poc1 directly recruits and/or is necessary for the stable binding of this barrel-shaped protein at the A–B inner junction to form the attachment site of the scaffold (Fig. 6 C, right). We conclude that Poc1 stabilizes and organizes the inner scaffold by using its WD40 domain to create the base of the A–B inner junction attachment site, which then anchors the scaffold to the TMTs.

To understand how an intact inner scaffold contributes to the structural integrity of the BB, we asked whether there were gross differences in the shape of BBs from wild-type versus poc1Δ cells. Mapping the TMT subtomogram averages from wild-type and poc1Δ BBs onto individual BBs revealed that poc1Δ BBs, compared with wild-type BBs, were significantly flattened on the EM grid (Fig. 6 D). The observed flattening was likely induced during BB isolation in centrifugation steps or by the compression force applied to the sample during blotting and freezing the EM grid, which was previously observed in centrioles from CHO cells (Greenan et al., 2018; Le Guennec et al., 2020). This suggests that Poc1 and the inner scaffold bolster a sturdy architecture that is resistant to mechanical forces, including those introduced by cryoET sample preparation.

We previously showed that BBs bend in response to the cilia beat stroke, suggesting that BBs balance rigidity with plasticity to ensure they do not disassemble (Junker et al., 2022). To understand how the BB architecture protects itself from ciliary forces, we used U-ExM-SIM to assess how BBs bend in response to ciliary beating. Under high force conditions, wild-type BBs exhibit a gradual bend, whereas dramatic fracturing of the BB through the core region was observed in the absence of Poc1, suggesting that poc1Δ BBs lack the structural integrity that can integrate the bending without destructive deformation (Fig. 6 E; Junker et al., 2022). To determine whether BBs that show structural defects have intact core structures, we used U-ExM-SIM to visualize Poc16 in BBs in high-force conditions. In wild-type cells, Poc16 colocalizes with the tubulin walls of BBs as has been previously reported (Fig. 6 F; Steib et al., 2020). However, in poc1Δ cells, Poc16 is sparsely found in the core region and is no longer closely associated with the TMTs. Regions lacking Poc16 display poor or absent TMT structure (Fig. 6 F and Fig. S5 C). We conclude that Poc1 contributes to BB structural integrity by promoting the assembly and/or stability of TMT interconnecting features in the BB core region.

Poc1, via its WD40 domain, seals the inner junctions of the A–B and B–C tubules at the BB proximal end, and the A–B tubules at the BB central core. By linking MT protofilaments at these sites, Poc1 stabilizes TMTs directly. Moreover, Poc1 promotes the stability of the pinhead and the inner scaffold, which interconnect TMTs and organize the BB architecture, thereby imparting structural integrity to the BB in the face of cilia beating. Although conventional EM shows that poc1Δ cells assemble BBs with apparently normal TMTs, the weak connections between TMTs, revealed in this study, render BBs unstable when ciliary forces are elevated. We establish a high-resolution structure–function relationship to explain the role of Poc1 in the mechanics of BB stability.

Poc1 localizes to the A–B and B–C inner junctions at the proximal end of BB TMTs (Figs. 1 and 2). The structures of wild-type A–B and B–C inner junctions identified WD40 domains with seven β-propeller repeats, consistent with Poc1’s predicted structure (Fig. 2 C and Video 2). Although we did not identify other doughnut-shaped densities likely to be Poc1, we cannot exclude other, non-inner junction Poc1 localizations. Poc1’s localization at the inner junctions rectifies several observations. It was proposed that Poc1 may be an A/C linker protein, but Poc1 is not exclusively localized to the BB proximal end (Li et al., 2019). It also resides in the core region, which does not harbor proximal A/C linkers and the previously proposed doughnut structure for Poc1 in the Chlamydomonas A/C linker remains present in the Tetrahymena poc1Δ structure (Fig. S3 D). Finally, the POC1 gene is absent in C. elegans, whose BBs are comprised of singlet MTs rather than TMTs, and therefore have no inner junctions (Keller et al., 2009). Together, Poc1’s specific localization to inner junctions along the length of the BB fits existing studies.

Tetrahymena Poc1 localization is different from human centrioles. In RPE-1 cells, Poc1B was found to localize along the BB length like in Tetrahymena when ciliated, but when unciliated and in U2OS cells, Poc1B is only present in the central core (Arslanhan et al., 2023; Le Guennec et al., 2020; Pearson et al., 2009a). These differences in Poc1 localization may reflect differences in BBs that template motile cilia versus primary cilia, or that exist in a non-ciliated state. The above studies focused on Poc1B, and it is possible that the human Poc1A ortholog preferentially occupies the proximal end of human centrioles and Poc1B resides at the central core region. These distinct localization patterns could exist in a wild-type state, even though Poc1A and Poc1B can compensate for one another when knocked out individually (Venoux et al., 2013). Regardless, Poc1 promotes BB integrity through TMT stabilization and interconnection via its binding to the inner junctions, which is likely to be a unifying principle across phylogeny.

TMT loss is a primary defect in poc1Δ BBs (Meehl et al., 2016). In poc1Δ BBs, we observed missing or partially missing B- and C- tubules (Fig. 3). However, complete TMTs are formed in poc1Δ BBs and BB assembly rates are unaltered in poc1Δ cells, suggesting TMT formation itself is not a primary defect in poc1Δ cells (Pearson et al., 2009b; Meehl et al., 2016). We interpret the difference in TMT structure observed here as a state of TMT loss via disassembly, rather than incomplete TMT assembly or maturation. However, interconnecting structures between TMTs may not be completely assembled in poc1Δ cells. The TMT structure loss is akin to the BB disassembly observed in poc1Δ cells upon high mechanical force from beating cilia. Here, the observed tubule loss in the tomograms likely was introduced by isolation and freezing required for cryoET. Multiple states of TMT disassembly were identified by classification of the TMTs from poc1Δ BBs (Fig. 3, C and D). Except for the complete TMT subclass (8% of subtomograms), all other subclasses show significant structural defects at the A–B and B–C inner junctions. These defects manifest as the loss of protofilaments close to the inner junctions, to varying extents, and changes to MT curvatures, even in less severe TMT disassembly states. These subclasses could represent a temporal view of TMT disassembly, wherein protofilament loss begins at the inner junction and spreads toward the exterior side of the TMT in the subclasses with more severe defects. These data suggest that Poc1 serves to maintain a tight association of the B- and C-tubules with their neighboring tubules to seal the inner junctions and stabilize TMTs.

The proximal end of BBs contains A/C linker and cartwheel structures that interconnect TMTs. Although defects in A/C linkers in poc1Δ BBs were previously reported, whether there are differences between wild-type and poc1Δ A/C linkers based on our cryoET structure was difficult to ascertain due to a limited amount of data and structural flexibility in the region (Meehl et al., 2016). However, we found that poc1Δ BBs partially lost the cartwheel pinhead and its molecular constituent, Bld10 (Fig. 4). This was unexpected given Poc1’s inner junction localization; we did not observe any WD40-like densities in the pinhead that would indicate an additional site of Poc1 localization there. This structure is reduced even in BBs with complete TMTs. One explanation for pinhead reduction is that Poc1 influences its stability from a seemingly distant location in the inner junctions. It is possible that the N- and C-termini of Poc1 extend away from the inner junctions and, like outstretched arms, interact with the pinhead and A/C linkers to stabilize them. This is ostensibly inconsistent with our prior work showing that Poc1 lacking the C-terminal tail was sufficient to rescue BB loss at high force, unless other structures like the inner scaffold, which depends more directly on the WD40 domain, were recovered and are sufficient for the rescue (Pearson et al., 2009a). Regardless, this hypothesis of outstretched termini is strengthened by the high-resolution structure at the A–B inner junction in the proximal region, which reveals extended densities that protrude from the WD40 domain, consistent with a model in which one of Poc1’s termini links with the pinhead (Fig. 2 C and Video 3). These densities reach laterally across the TMT toward the A-tubule on one side, where the pinhead resides, and toward the C-tubule and A/C linker attachments on the other side. These coil densities are absent in poc1Δ BBs, but our resolution of the poc1Δ structure is lower than that of wild-type, making it difficult to accurately compare the thin densities that only became distinguishable at higher resolution in wild-type BBs. If these densities are not extensions of Poc1, it suggests some other protein stretches across the Poc1 WD40 domain. An additional model to explain pinhead loss in the absence of Poc1 is that forces exerted onto the TMTs cannot be distributed in a manner compatible with the maintenance of the pinhead structure. Importantly, these models are not mutually exclusive. In summary, Poc1 supports the overall pinhead structure that links TMTs to the cartwheel, which appears critical for resisting forces from ciliary beating.

In the BB central core, Poc1 localizes to the A–B inner junction, creating the base of a conserved attachment site of the inner scaffold to the TMTs, referred to as the stem (Fig. 6 C). The Tetrahymena TMT structure shows that the stem is also comprised of another, unidentified protein that binds directly onto Poc1’s WD40 domain. This unidentified Poc1-interacting protein in the BB core region reveals a solenoid protein fold (Fig. 6 C and Video 4) and appears to connect the inner scaffold to the TMTs. Future experiments are needed to identify and characterize this unknown protein.

In both wild-type and poc1Δ BBs, we observed loss of C-tubules in the core region, with only 9.3% and 6.0% of the TMTs being complete, respectively (Fig. S4 A). That C-tubules were lost in wild-type BBs indicates that they are generally less stable than A- and B-tubules. Incomplete TMTs were nonetheless capable of maintaining an inner scaffold. This is perhaps due to the non-tubulin B–C inner junction structure in the core, which is still present in the absence of C-tubules, and serves as another attachment site for the inner scaffold (Fig. S4 D). Complete TMTs from poc1Δ BBs did not retain an inner scaffold density at the B–C inner junction attachment site, and the inner scaffold density was greatly reduced at the A-tubule attachment site, even though Poc1 does not localize to those attachments. Thus, we conclude that the A–B inner junction stem is likely the most critical attachment site for the organization of the inner scaffold.

The BB and centriole A–B inner junction in the core region was proposed to be occupied by Poc16/WDR90 (Yanagisawa et al., 2014; Steib et al., 2020; Hamel et al., 2017). Poc16 is predicted to contain a DUF667 motif at its N-terminus and two to four WD40 domains in the remainder of the protein, depending on the organism (Steib et al., 2020). This is similar to the structure of the cilia axoneme A–B inner junction protein Fap20, which is a DUF667 fold-containing protein, and the two WD40 domain-containing axonemal protein, Fap52, which resides in the B-tubule lumen. Thus, Poc16 was proposed to be a fusion of a Fap20-like domain with a Fap52-like domain that uses the Fap20-like domain to bind the inner junction (Yanagisawa et al., 2014; Steib et al., 2020; Hamel et al., 2017). However, the shape of the density at the Tetrahymena A–B inner junction is that of a seven-bladed WD40 domain and not a DUF667 domain (Fig. 6 C and Video 4). In addition, an Alphafold prediction of Tetrahymena Poc16 shows a β-sandwich domain followed by four seven-bladed WD40 domains. This predicted structure does not fit into the densities at the A–B inner junction. Finally, Poc16 localizes to poc1Δ BBs in vivo while the cryoET structure shows the A–B inner junction doughnut and barrel structures to be completely absent (Fig. 6 C). This suggests that, in Tetrahymena, Poc16 resides elsewhere in the TMT core architecture and that Poc1, but not Poc16, resides at the A–B inner junction of the BB core region. Further studies are needed to define Poc16’s localization in TMTs.

Fam161A is reported to be an inner scaffold protein that binds MTs and Poc1B (Le Guennec et al., 2020; Roosing et al., 2014). Exactly how Fam161A contributes to the inner scaffold structure is unknown. Our fluorescence data show that Fam161A localizes to BBs in the absence of Poc1, albeit at a reduced level (∼50% reduction). This suggests that Fam161A has redundant mechanisms to localize to the Tetrahymena BB core region. While the vast majority of the density for the inner scaffold was lost in poc1Δ BBs, some density remained at the protofilament A03 attachment site (Fig. 6 B). We propose that in the absence of the A–B inner junction attachment established by Poc1, inner scaffold proteins attached to the A03 site are no longer organized horizontally across the TMT. This causes them to both be unstable and unlikely to reach binding partners near the TMT C-tubule. Whether Fam161A binds to the A03 attachment site or elsewhere is unknown, but in either case, the cryoET structure corroborates the fluorescence data showing a reduction of Fam161A in the core region.

At least two Fam161A populations are present in Tetrahymena BBs: one localizes at, and sometimes around, the BB core region and the second localizes to the distal end, which we assume to be the transition zone (Fig. 5 A and Fig. S5 A). This distribution of Fam161A localization may be analogous to that found at the BB and connecting cilium of the human retina (Mercey et al., 2022). Fam161A intensity decreased in poc1Δ BBs and, surprisingly, in wild-type BBs when cells were shifted to high-force conditions (Fig. 6 A and Fig. S5 B). The notion that Fam161A could be a dynamic component of the inner scaffold has not previously been reported. This observation may reflect a dynamic molecular and structural remodeling of the inner scaffold to improve force resistance, akin to force-responsive remodeling of focal adhesions (Grandy et al., 2023; Legerstee and Houtsmuller, 2021). The molecular basis for Fam161A dynamism is unclear, but precedence for such remodeling is evident with Poc5, a proposed inner scaffold protein in humans, which is only present in immature BBs in Tetrahymena (Heydeck et al., 2020). Further experimentation is needed to understand these putative remodeling events. Moreover, a greater understanding of the repertoire of inner scaffold proteins in Tetrahymena and other organisms is necessary to understand how they are used to create functional inner scaffolds.

We studied BBs that experience forces from ciliary beating, but aberrant human POC1A/B cause disease phenotypes not necessarily associated with motile cilia, namely polycystic kidneys, cone-rod dystrophy, dwarfism, and other developmental defects (Shaheen et al., 2012; Sarig et al., 2012; Shalev et al., 2012; Roosing et al., 2014; Durlu et al., 2014; Beck et al., 2014). These clinical presentations implicate POC1A/B in the function of BBs that nucleate primary cilia or centrioles that are not ciliated. Indeed, primary cilia in kidneys and photoreceptor-specialized cilia are both mechanosensitive, likely imposing force on the BB (Praetorius and Spring, 2001; Bocchero et al., 2020; Mercey et al., 2022). Moreover, centrioles aid in both spindle and cytoplasmic MT force resistance (Abal et al., 2005; Bobinnec et al., 1998). These observations suggest that Poc1 broadly functions to stabilize BBs and centrioles against forces outside of the context of ciliary beating.

We show that ciliary force resistance imparted by Poc1 is multifaceted. First, Poc1 binds the inner junctions of BB TMTs, linking protofilaments and promoting tubule architecture that prevents TMT disassembly. Second, Poc1 stabilizes interconnecting structural features and their attachments to the TMTs. This is exemplified by Poc1’s direct interaction with the inner scaffold via its WD40 domain in the core region, which also occurs through unclear mechanisms to stabilize the pinhead in the BB proximal region. Such functions of Poc1 lend insight to its strong conservation across phylogeny and devastating consequences when mutated in humans.

Tetrahymena thermophila strains B2086 (TSC_SD01625) and SB1969 (TSC_SD00701) were obtained from the Tetrahymena Stock Center at Cornell University. Creation of the poc1Δ and Poc1:mCherry and GFP:Poc1 strains is described in Pearson et al. (2009a). Cells were grown to mid-log phase at 30°C in 2% SPP (2% proteose peptone, 0.2% glucose, 0.1% yeast extract, and 0.003% Fe-EDTA) unless otherwise indicated. Starvation media was 10 mM Tris-HCl pH 7.4. Live and fixed cells were collected by centrifugation at 0.5 × g for 3 min unless otherwise indicated. Cell densities were determined using a Coulter Z1 cell counter with size gating of 15–45 μm.

NCBI protein Basic Local Alignment Search tool was used to identify Tetrahymena homologs of human FAM161A (accession number NP_001188472.1) and WDR90/Poc16 (accession number AAI21187.1). For FAM161A, we identified TTHERM_00052590, with an E-value of 5 × 10⁻¹¹ and 29% sequence identity. For WDR90/Poc16, we identified TTHERM_00218620, Hs > Tt 5 × 10⁻⁻⁵⁵ with 28.2% sequence identity.

The following constructs were generated to create C-terminal appended fluorescent proteins for expression in Tetrahymena cells: Poc1:HaloTag (Neo2); Fam161A:mCherry (Blasticidin); Poc16:mCherry (Blasticidin); and Poc16:GFP (NEO2). For Fam161A and Poc16 constructs, homology arms ∼400 bp each, separated by a NotI restriction site, were synthesized in pUC57 at the EcoRV site by Genscript. The upstream homology arm sat at the end of the coding portion of the gene but eliminated the termination codon. Inserts containing mCherry or GFP and blasticidin or NEO2 resistance genes were amplified by PCR from plasmid p4T2-1 (Winey et al., 2012) with the following primers: NotIp4T21US 5′-GCG​CGG​CCG​CTA​AAG​AAA​CTG​CTG​CTG​CTA​AAT​TCG-3′ or NotI p4T2 GFP US F 5′-ACA​GCG​GCC​GCT​TTA​ATG​AGT​AAA​GGA​GAA​GAA​CTT​TTC​AC-3′ and NotIp4T21DS 5′-GCG​CGG​CCG​CCT​AAC​ATG​TAT​GTG​AAG​AGG-3′. The inserts and synthesized homology arm vectors were then digested with NotI, gel-purified, and ligated using 2× Ligation Mix (Takara) and checked for insert orientation. The N-terminal GFP tag of Poc1 used in this study was created previously (Pearson et al., 2009a) and the Poc1:Halo tag construct was created by subcloning the HaloTag from p4T2-1 Sas4:HaloTag vector into the p4T2-1 Poc1:mCherry vector using EcoRI and XbaI cut sites (Ruehle et al., 2020; Pearson et al., 2009a). All plasmid sequences were verified using whole-plasmid nanopore sequencing technology (Plasmidisaurus).

Macronuclear transformation was performed as previously described using DNA-coated particle bombardment (Bruns and Cassidy-Hanley, 2000). Transformed clones were selected using 100 μg/ml paromomycin to select for the NEO2 gene, 60 μg/ml blasticidin S for the BSR gene, or 7.5 μg/ml cycloheximide to select for the CHX gene. To increase the copy number of the tagged constructs, cells were assorted using increasing concentrations of the appropriate drug.

Cell lines created in this work expressing fluorescent fusion proteins include the following: Poc1:HaloTag in SB1969 background; Fam161A:mCherry in SB1969, SB1969 with Poc1:HaloTag, and poc1Δ; Poc16:mCherry in SB1969 background, SB1969 with Poc1:HaloTag, and poc1Δ; and Poc16:GFP in SB1969 and poc1Δ backgrounds. Note that all genes, including Poc16 and Fam161A, were tagged at their endogenous loci. The parent SB1969 cell line was obtained from the Tetrahymena Stock Center (TSC_SD00701).

Tetrahymena cells were grown in 200 ml SPP prepared with HPLC grade water to high-log cell concentrations (0.5–1.0 × 10⁶ cells/ml) at 30°C. All media and buffers were made using HPLC-grade water, as this was determined to be important for cell lysis. Cells were pelleted, washed with 1× PBS, then resuspended in cold 18.75 ml Osmo Buffer (10 mM Tris pH 7.5, 1 M sucrose, 1 mM EDTA) containing freshly prepared 1 mM phenyl methyl sulfonyl fluoride (PMSF) protease inhibitor and incubated on ice for 5 min. Samples were transferred to 50 ml conical vials and Triton X-100 was added to a final concentration of 7.5% plus 1 μl of neat β-mercaptoethanol. 7.5 ml of sterile glass beads were added to the samples and the conical vials were vortexed three times with 4-s pulses. Vials still containing the glass beads were then nutated at 4°C for 1–4 h, monitoring progress of the lysis with a brightfield microscope approximately every 30 min. Lysis was considered complete when whole cell bodies were largely disrupted, and cytoplasmic streaming was observed. This process was completed at about 4 h for wild-type cells and within about 1.5 h for the poc1Δ cells. Upon completion of cell lysis, samples were diluted threefold in 1× TE buffer and spun in a JA25.2 fixed angle rotor at 13,000 × g for 45 min to pellet the BBs. The pellet contained a looser, “fluffy” layer which, in addition to BBs, contained more cell debris and whole oral apparatuses, as well as a glassier, “hard” pellet, which contained more pure BBs, though other macromolecular complexes are found in this pellet too, including proteasomes and ribosomes, as determined by negative staining the samples. After carefully removing the supernatant and fluffy pellet, the hard pellet was washed with 500 μl of 1× TE buffer to remove any residual cell debris and then was resuspended in 500 μl of 1× TE buffer. CryoET grid preparation was performed immediately afterward to prevent sample degradation. Freeze–thaw destroys the structural integrity of the sample.

Grids for cryoET were prepared using 200 mesh Quantifoil Cu EM grids with 2 μm holes (Ted Pella). Grids were glow discharged for 1 min in negative ion mode. 4 μl of basal body sample containing BSA-coated 10 nm gold fiducials (BBI Solutions) was applied to the grid and plunge frozen in liquid ethane after a 45-s wait time using a Thermo Fisher Scientific Vitrobot. The relative humidity of the chamber was 95%, the temperature was 22°C, and the blot time was 0.5 s. Negative stains to assess sample quality were prepared immediately after freezing using 3 μl of BB sample on 400 mesh copper grids and staining with 3% uranyl acetate. Negative stained samples were imaged on a 120 kV Thermo Fisher Scientific/FEI Talos L120C microscope with a Ceta CMOS detector.

Single-axis tilt series were collected on two field emission gun 300 kV Titan Krios electron microscopes (Thermo Fisher Scientific, Inc.) at UCSF. Each scope was equipped with a Bio-Quantum GIF energy filter and a post-GIF Gatan K2 or K3 Summit Direct Electron Detectors (Gatan, Inc.). The GIF slit width was set at 20 eV. SerialEM was used for tomography tilt series data collection (Mastronarde, 2005). The data were collected in the super-resolution and dose-fractionation mode. The nominal magnification was set at 33,000. The effective physical pixel size on recorded images was either 2.70 Å (for wild-type) or 2.65 Å (for poc1 KO mutant). A dose rate of 20 electron/pixel/second was used during exposure. The accumulated dose for each tilt series was limited to 80 electron/Ų on the sample. A bidirectional scheme was used for collecting tilt series, starting from zero degree, first tilted toward −60°, followed by a second half from +2° to +60°, in 2° increments per tilt.

For tomogram reconstruction and subtomogram averaging, the dose-fractionated movie at each tilt in the tilt series was corrected of motion and summed using MotionCor2 (Zheng et al., 2017). The tilt series were aligned based on the gold beads as fiducials by using IMOD and TomoAlign (Kremer et al., 1996; Fernandez et al., 2018). The contrast transfer function for each tilt series was determined and corrected by TomoCTF (Fernández et al., 2006). The tomograms were reconstructed by TomoRec taking into account the beam-induced sample motion during data collection (Fernandez et al., 2019). A total of 63 wild-type BB tomograms from 61 tilt series, 85 poc1Δ BB tomograms from 83 tilt series were used for reconstruction and subtomogram averaging.

For subtomogram averaging, first the BBs were identified in the 6xbinned tomograms. The center of TMT and their approximate orientation relative to the tilt axis were manually annotated in a Spider metadata file (Frank et al., 1996). The initial subtomogram alignment and average were carried out in a 2× binned format (pixel size 5.4 Å for the wild-type or 5.30 Å for poc1Δ mutant). The longitudinal segment length of basal body TMTs in a subtomogram was limited to 24 nm and 50% overlapping with neighboring segments. Without using any external reference, the subtomogram alignment was carried out by a program MLTOMO implemented in the Xmipp software package (Scheres et al., 2009).

Since TMTs from BBs are a continuous filament, after obtaining the initial alignment parameters, a homemade program RANSAC was used to detect and remove any alignment outliers and to impose the continuity constraint on the neighboring segments. This corrected the misaligned subtomograms by regression. MLTOMO and Relion 3.1 or 4.0 were extensively used for the focused classification of the subtomograms (Bharat and Scheres, 2016; Zivanov et al., 2022). This was critical for determining the correct periodicity of the MIPs and for identifying structural defects or heterogeneity in the TMTs. These out-of-register subtomograms were re-centered and re-extracted. This was followed by combining all subtomograms for the next round of refinement. Based on the focusing classification, microtubule inner proteins (MIPs) at the inner junction region of the BB were found to exhibit up to 16 nm periodicity. These include the A–B and B–C inner junctions in the proximal region and the A–B inner junction in the core region. Therefore, the longitudinal length of TMT in the subtomograms was set to 18 nm, covering a complete 16 nm repeat. The final refinements, focusing on the inner junctions of the TMT, were carried out by using a workflow implemented in Relion 4.0 (Zivanov et al., 2022). The overall resolutions were reported (Fig. S1 and Table 1) based on the Fourier Shell Correlation (FSC) cutoff at 0.143 (Scheres and Chen, 2012; Rosenthal and Henderson, 2003).

The pseudo-atomic models for the inner junctions at different regions of the basal body were built in ChimeraX (Pettersen et al., 2021) by fitting previously published atomic models or models predicted by AlphaFold2 (Jumper et al., 2021) into the subtomogram averaging density maps. UCSF ChimeraX was also used for visualization and recording images.

The final average maps, including their EMDB access codes, are summarized in Table 1.

Since the MT is a dominant feature in the averaged TMT structure, the overlapping MT signal could interfere with the analysis of the inner scaffold. To eliminate this potential problem, a soft-edged binary 3D mask was initially generated and imposed onto the TMT average. This removed MT backbone density from the average and kept only the inner scaffold structure. Based on the TMT subtomogram location and their 3D refinement parameters defined during the refinement, the masked average volume containing only the inner scaffold structure was then backfit into the tomograms. The backfitting took into account the translations needed to put the TMT into the register of its 16 nm periodicity in the inner junction region. The resulting model was a barrel-shaped structure containing only the inner scaffold.

To analyze the arrangement of the inner scaffold, the inner scaffold volume calculated above was projected on a plane orthogonal to the longitudinal axis of the BB. This was followed by calculating the Fourier transform of the projection. The resulting Fourier showed characteristic layer lines, indicating the nature of its helical assembly. The Fourier had a strong signal at 8 nm layer line, indicating the 8 nm periodicity of the inner scaffold (Fig. S4 F). A line scan along the 8 nm layer line showed the maximal intensity near the center, crossing the meridian, strongly indicating its Bessel order close to zero (DeRosier and Moore, 1970; Klug et al., 1958). In real space, this translates into the inner scaffold repeat unit forming a “zero start” helix, a closed circular ring. The inner scaffold is arranged as a stack of multiple rings that are 8 nm apart perpendicular to the BB longitudinal axis. An example process is illustrated in Fig. S4 E. The analysis was repeated on 63 wild-type BB datasets at different longitudinal locations in the core region and had a consistent result.

To confirm the above conclusion, auto-correlations are calculated while rotating the BB longitudinal axis in a 40° step on the regenerated inner scaffold volume that has nearly intact ninefold symmetry. The nine correlation maxima have minimal translation along the BB longitudinal axis, indicating the repeat units have a rise close to zero. They are on the same plane perpendicular to the longitudinal axis.

Finally, manual tracing and inspecting connection is conducted by following the inner scaffold ring around the inner circumference of BB. It shows consistent results that the inner scaffold is arranged as a longitudinal stack of closed rings in Tetrahymena (Fig. 5 C).

Cells grown to mid-log density in 2% SPP were pelleted and resuspended in 10 mM Tris-HCl pH 7.4 starvation medium and distributed equally between two flasks for each cell line. Both flasks for each cell line were incubated at 30°C for 12 h, then one flask of each cell line was temperature-shifted to 38°C. After 12 h, the cells both at 30^oC and 38°C were harvested and stained by immunofluorescence as described below. For experiments with NiCl₂ treatment, starved cells were treated with 300 μM freshly prepared NiCl₂ for 1 h at 30°C to verify inhibition of ciliary beating before shifting to 37°C for 16 h. Image analysis was performed as follows: maximum projected images through half a cell volume were generated for each image. 15 BBs from 10 cells (150 BBs total) were analyzed for each condition in each replicate. To determine the signal from each BB, a 0.65 × 0.65 μm square region of interest was centered over the BB signal of interest, followed by two squares of the same size placed immediately next to the BB to determine the local background signal. Integrated densities of all boxes were measured and the average of the two background boxes was subtracted from its corresponding BB box to determine the fluorescence value of that BB. The average value of BB fluorescence at 30°C was normalized to 1 for each cell line. Three biological replicates were performed. To determine the amount of protein at BBs in wild-type versus poc1Δ cells, sum-projected images of the entire cell volume were created and a custom region of interest was drawn tightly around the cell edge. Five 25 × 25 μm squares were placed around each cell to measure the local background signal. Integrated densities and areas of each region of interest were measured and integrated density per area values were calculated. The average integrated density per area of the background boxes was subtracted from the cell’s integrated density per area and then multiplied by the cell’s area to calculate the total cell fluorescence signal. These values were then used to normalize the BB fluorescence intensities determined above for 15 BBs in five cells per cell line at 30°C. This step was necessary to account for slight differences in assortment level between cell lines.

Cells were pelleted and fixed in 3.2% paraformaldehyde 0.5% TritonX-100 in 1× PHEM buffer for 10 min at room temperature. Fixed cells were then pelleted, washed twice with 0.1% BSA-PBS, and incubated in primary antibody solution in 1% BSA-PBS either for 2 h at room temperature or overnight at 4°C. Cells were then washed three times with 0.1% BSA-PBS and incubated in a secondary antibody solution in 1% BSA-PBS either for 1 h at room temperature or overnight at 4°C. Cells were washed twice in 0.1% BSA-PBS and then once in 1xPBS before mounting on coverslips with Citifluor AF1 mounting medium (Citifluor). Coverslips were sealed to the slides using clear nail polish. Primary antibodies used in this work are documented in Table 2. Secondary antibodies were derived from goat and conjugated to Alexa Fluor 488, 549, or 647 (Invitrogen) and used at 1:1,000. HaloTags were visualized with JaneliaFluor549 or 646-conjugated Halo Ligands (Promega). Labeling of HaloTags was performed on live cells for 0.5–2 h at 30°C at a final concentration of 100 μM.

Confocal fluorescence images were acquired using a Nikon Ti Eclipse inverted microscope with a Nikon 100× Plan-Apo objective, NA 1.45, at 23°C and Andor iXon X3 camera, and CSU-X1 (Yokogawa) spinning disk. Images were acquired using Slidebook6 or Slidebook23 imaging software and analyzed using ImageJ image analysis software (https://imagej.net). All images were acquired with exposure times between 50 and 500 ms, depending on the experiment and the channel of acquisition.

1 ml of cells was pelleted at 0.5 × g for 3 min at room temperature and then resuspended in 0.5 ml of formaldehyde/acrylamide solution (1.4%/2%) for 2–5 h at 37°C with nutation. Cells were pelleted again using the same settings and 5 μl of the pellet was placed on a piece of parafilm laid flat within a humid chamber lying on ice. 35 μl of monomer solution (19% Sodium Acrylate, 10% Acrylamide, 0.1% N′,N′-methylenebisacrylamide, 1× PBS)/0.5% APS/0.5% TEMED mixture was added on top of the 5-μl cell drop and an 18 × 18 mm #1.5 coverglass was laid gently on top to form a thin layer of gel between the glass and parafilm. Gels were formed on ice for 5 min and then transferred to 37°C for 1 h. After gelation, the coverslips were flipped over, with the gel remaining adhered to the glass. Several 4 or 6 mm–diameter punches were cut into the gel using a homemade, 3D printed punch, which created multiple gel “discs” from a single sample. The gels were then transferred to 1.7-ml microcentrifuge tubes and 1.5 ml of denaturation buffer (200 mM SDS, 200 mM NaCl, 50 mM Tris in water, pH 9) was added. The tubes were incubated at 95°C for 1.5 h. After denaturation, the gels were expanded twice in 50 ml conical vials with double distilled water, then overnight in 50 ml of ddH₂O.

Gels were stained as follows. A single gel disc was transferred to a 12-well plastic dish and washed with 1× PBS two times for 15 min. PBS was removed carefully to avoid damage to the gel and 200–500 μl of primary antibody solution in 2% BSA-PBS was added to the wells and incubated for 2 h at 37°C, or overnight at 4°C, with gentle rocking. Gels were washed three times with 1 ml of PBS-Tween20 0.1%, for 10 min each. The washing solution was gently removed and 200–500 μl of secondary antibody solution in 2% BSA-PBS was added to the wells and incubated for 2 h at 37°C with gentle rocking. Gels were then washed twice with 1 ml of PBS-Tween20 0.1%, for 10 min each and re-expanded in ddH20 as described above.

For imaging, gels were sandwiched between two poly-L-lysine coated 22 × 60 mm #1.5 coverslips and then clamped to the stage mount on either confocal or SIM microscopes. Gels remained in distilled water; no mounting medium was used. Expansion factors obtained in these experiments ranged from 2.5 to 3.75 fold, using the diameter of the BB as a molecular ruler. All scale bars in U-ExM images represent the scale of the acquired image and are not corrected for expansion factor.

Top-down BB views with normal TMTs (nine distinctly visible TMTs or continuous staining when the resolution was low) and abnormal TMTs (less than nine distinctly visible TMTs or that have areas of missing staining when the resolution of TMTs was low), were selected across four expanded cells for each cell line and temperature. TMTs were visualized with anti-acetylated tubulin antibody (Table 2). In total, 80 BBs were counted each for wild-type and poc1Δ at 30°C, and 58 BBs and 52 BBs were counted for wild-type and poc1Δ, respectively, at 38°C.

SIM imaging was performed on a Nikon N-SIM system using a Ti2 inverted microscope with a 100× CF160 Apo-chromat superresolution/TIRF NA 1.49 objective with correction collar (Nikon Instruments) and a sCMOS camera (ORCA-Flash4.0, Hamamatsu). Images were collected at 25°C and reconstructed using the slice reconstruction algorithm (NIS Elements).

Experiments were performed with three biological replicates. The total number of cells and/or BBs analyzed is described in the figure legends or in the Materials and methods section. Prism6.2e (GraphPad Software) was used for graphing and statistical analysis. All data sets were tested for normality using the D’Agostino-Pearson omnibus normality test. Normally distributed data were analyzed using the two-tailed, unpaired Student’s t test, whereas non-normal data were analyzed using the Mann–Whitney test. Exact P values are reported on each graph. Error bars represent standard deviation.

Fig. S1 shows the resolution achieved for each structure solved in this work by Fourier shell correlation. Fig. S2 shows a comparison of structural features of the Tetrahymena BB compared with other species. Fig. S3 depicts how Poc1 is required for ciliary force resistance and that its WD40 domain does not reside in the A/C linker. Fig. S4 shows the classification of subtomograms for wild-type and poc1Δ BBs in the central core, revealing differential stability of C-tubules, the lack of Poc1 in the B–C inner junction, that the inner scaffold is a set of stacked rings, and that poc1Δ complete TMTs cluster in the proximal end of the BB. Fig. S5 shows that Fam161A is lost in the core region in wild-type and poc1Δ BBs, and that BB defects in poc1Δ at high force are associated with loss of Poc16. Video 1 shows wild-type BB-aligned cryo-ET tilt series. Video 2 shows A–B inner junction from wild-type BB subtomogram average in the proximal region. Video 3 shows poc1Δ BB aligned cryo-ET tilt series. Video 4 shows A–B inner junction from wild-type BB subtomogram average in the core region.

All data reported in this manuscript are available in the article and supplementary material. Further information is available from the corresponding author.

The authors would like to thank Mark Winey for the helpful conversations and partial funding of Sam Li during the course of the work. In addition, the University of Colorado School of Medicine Cryo Electron Microscopy Shared Resource Center, especially Brian Wimberly, Dave Farrell, Peter Van Blerkom, and Eduardo Romero Camacho for help and expertise in preparing and screening EM grids. We thank Jose-Jesus Fernandez (CINN-CSIC, Spain) for advice on cryoET image processing, David Bulkley and Eric Tse (UCSF) for assistance with cryoET data collection, Tom Goddard (UCSF) for advice on AlphaFold and ChimeraX, and UCSF Wynton HPC for computation support during this study.

This research was funded by National Institutes of Health (NIH) R35GM140813 and the W.M. Keck Foundation (C.G. Pearson), NIH R35GM118099 (D.A. Agard), NIH 2R01GM127571 (M. Winey), and NIH-NIGMS 5F32GM122239 (M.D. Ruehle).

Author contributions: M.D. Ruehle: Conceptualization, Data curation, Formal analysis, Funding acquisition, Investigation, Methodology, Validation, Visualization, Writing—original draft, Writing—review & editing; S. Li: Data curation, Formal analysis, Investigation, Software, Validation, Visualization, Writing—review & editing; D.A. Agard: Methodology, Supervision, Writing—review & editing; C.G. Pearson: Conceptualization, Funding acquisition, Project administration, Resources, Supervision, Validation, Writing—review & editing.