Architecture of mammalian centriole distal appendages accommodates distinct blade and matrix functional elements

Distal appendages (DAPs) are nanoscale, pinwheel-like structures protruding from the distal end of the centriole that mediate membrane docking during ciliogenesis, marking the cilia base around the ciliary gate. Here, we determined a superresolved multiplex of 16 centriole-distal-end components. Surprisingly, rather than pinwheels, intact DAPs exhibit a cone-shaped architecture with components filling the space between each pinwheel blade, a new structural element we termed the distal appendage matrix (DAM). Specifically, CEP83, CEP89, SCLT1, and CEP164 form the backbone of pinwheel blades, with CEP83 confined at the root and CEP164 extending to the tip near the membrane-docking site. By contrast, FBF1 marks the distal end of the DAM near the ciliary membrane. Strikingly, unlike CEP164 which is essential for ciliogenesis, FBF1 is required for ciliary gating of transmembrane proteins, revealing DAPs as an essential component of the ciliary gate. Our findings redefine both the structure and function of DAPs.


Introduction
All cilia grow from the distal end of centrioles/basal bodies [1][2][3][4] . To initiate primary ciliogenesis, a mature vertebrate centriole docks to the membrane, recruits membrane vesicles, regulates axoneme growth, and forms part of the ciliary base that gates the ciliary compartment, all of which depend on its distal end [5][6][7][8][9] . In particular, distal appendages (DAPs) are previously defined as nine-bladed, pinwheel-like structures protruding from the distal periphery of vertebrate centrioles 10 and are thought to be involved in many of the aforementioned steps of ciliogenesis. CEP83, CEP89, SCLT1, CEP164, and FBF1 are core DAP components recruited to centriole distal ends prior to and independently of ciliogenesis 5 . These components are assembled at the G2/M border, and assembly is dependent on C2CD3, another centriole-distal-end protein 11 . In response to cell cycle cues in G1 that stimulate ciliogenesis, other proteins including TTBK2, recycling endosome components (Rab8, Rab11, and EHD1), IFT88, and ARL13B are recruited to DAPs along with the associated membrane vesicles [12][13][14][15][16][17] . Upon membrane docking mediated by DAPs, centriole distal ends are further modified to release negative regulators of ciliogenesis, such as CP110 or CEP97 18 , which enables nucleation of the axoneme from the centriole. At the nascent axoneme bud, a specialized structure known as the transition zone (TZ) is built as part of the ciliary barrier from dozens of protein components recruited from the cytoplasm 19 , guided in part by the residential centriole-distal-end proteins CEP290 and CEP162 20,21 . The continuous growth and maintenance of the axoneme in the gated ciliary compartment are further supported by the intraflagellar transport (IFT) machinery 22 , generating a functionally intact primary cilium that serves as a sensory organelle.
By initiating ciliogenesis, DAPs are situated in a highly intricate multi-junction position that marks the border between the centriole and axoneme (or TZ), as well as the junction between the plasma membrane and the ciliary membrane [5][6][7][8][9] . This unique location raises the possibility that DAPs and their associated factors may also form part of the structural barrier gating the ciliary compartment. However, the pinwheellike morphology, as defined by electron microscopy (EM), does not intuitively suggest how DAPs can serve as a barrier, given that the large space between each pinwheel blade would allow many substances to pass through with little or no resistance. Thus, DAPs may simply function as the anchor mediating membrane docking or IFT machinery recruitment. Alternatively, however, the pinwheel shape may not morphologically reflect the full complexity of DAPs, due the limitations of EM.
Here we resolved 12 DAP-associated molecular species and four additional proteins surrounding DAPs using direct stochastic optical reconstruction microscopy (dSTORM) 23,24 , revealing a 3D nanoscopic resolution of the protein megacomplex in intact cells and reaching a level of detail not previously achieved by current structure biology techniques. We found the architecture supports a previously unrecognized structure that gates transmembrane proteins. Together, our findings reveal an unprecedented architectural and functional framework at the base of mammalian primary cilia.

Results
dSTORM superresolution imaging reveals differential radial occupancies of DAP proteins 5   To systematically characterize DAPs and centriole distal ends, our dSTORM imaging   analysis focused on core DAP components (CEP83, CEP89, SCLT1, CEP164, and   FBF1), DAP-associated proteins (Cby1, TTBK2, and IFT88), other centriole-distalend proteins (C2CD3, CP110, CEP97, CEP162, and CEP290), ciliary membrane proteins (ARL13B and EHD1), and the subdistal appendage (sDAP) component ODF2. Superresolution microscopy has been used previously to localize centriolar proteins 25-29 , but systematic determination of proteins at the centriole distal end has never been performed. Here, mature centrioles oriented axially or laterally (perpendicular or parallel to the imaging plane) in human retinal pigment epithelial (RPE-1) cells were imaged. As depicted in Fig. 1a, visualization of axially and laterally oriented centrioles, respectively, allows clear determination of the radial and longitudinal distributions of centriolar components, leading to a 3D architectural map of the centriole distal end.
Analysis of axially oriented centrioles revealed radial localization of core DAP components exhibiting nine-fold symmetric ring-like patterns that differ in size/diameter, with CEP164 and FBF1 forming the largest ring, followed by SCLT1, CEP89, and CEP83 ( Fig. 1b; Supplementary Table 1). CEP164 displays a broader radial distribution with distinct outer and an inner intensity peaks ( Fig. 1b and Supplementary Fig. 1), resembling the pinwheel-like pattern characteristic of DAPs as defined by EM (Fig. 1c). By contrast, C2CD3, a protein essential for DAP assembly, occupies a surprisingly small, more compact region inside the centriole lumen than CP110, CEP97, CEP290, and CEP162 (Fig. 1b, c), all of which form a ring-like distribution of ~170200 nm in diameter (Fig. 1d, Supplementary Fig. 2; Supplementary Table 1), similar in size to the centriolar axoneme (Fig. 1c). This suggests that C2CD3 is not part of the DAP slanted structure defined by EM; rather, it is a luminal protein likely defining the distal environment of centrioles required for appendage formation.
CEP83, CEP89, SCLT1, and CEP164 form a blade-like arrangement to which FBF1 is not localized Two-color dSTORM imaging of the radial distribution further revealed distinct angular positions among the core DAP proteins (Fig. 2ae). Images of pairs of CEP83-SCLT1 (Fig. 2a), CEP89-CEP164 (Fig. 2b), and CEP164-SCLT1 (Fig. 2c) showed that they can all be allocated to tilted lines resembling the arrangement of the nine electron-dense DAP blades observed in EM images. SCLT1 was often localized within the blades of CEP164 propeller-like signals (Fig 2c; Supplementary Fig. 3). By contrast, FBF1 fills the circumferential gaps of CEP164 puncta (Fig. 2d) as well as SCLT1 puncta (Fig. 2e), likely localized to an open area between adjacent electrondense DAPs observed by EM. Histograms of angular spacing between two puncta, one from each of a pair of DAP proteins collected from multiple cilia, revealed that CEP83-SCLT1, CEP89-CEP164, and CEP164-SCLT1 pairs yielded peaks close to multiples of 40, or 360/9 ( Fig. 2f (Fig. 2ac, g), FBF1 is instead associated with the space between DAP blades (Fig. 2de, g).

Combination of axial and lateral superresolution images reveals a 3D molecular map of DAPs
The lateral view, revealed by a series of two-color dSTORM images (Fig. 3a), showed that CEP83, SCLT1, and FBF1 respectively form distinct thin layers along the proximal-to-distal axis of the centriole (Fig. 3b). Consistent with its wide radial distribution, CEP164 also occupies a broad longitudinal localization, forming a trapezoid-like shape with two tapered sides, resembling that of the DAP slanted structure observed by EM (Fig. 3b, d). This suggests that CEP164 forms the primary backbone of DAPs. Meanwhile, CEP89 occupies two layers at the centriole distal end both with a similar width but separated axially by ~130 nm ( Combining information from all relative axial and lateral positions, we built a 3D computational model to visualize the positioning of proteins at the DAP region ( Fig.   3g), with a backbone structure illustrated based on EM 10 . CEP164, CEP83, SCLT1, and the distal layer of CEP89 mapped well onto the 3D-reconstructed DAPs (Fig. 3h).
By contrast, FBF1 localizes to the distal regions of the DAP gaps between adjacent CEP164 molecules, possibly near or on the proximal boundary of the ciliary pocket 8 (Fig. 3i). Thus, FBF1 may serve a role distinct from other core DAP components.

ARL13B lies adjacent to DAPs during ciliary initiation but is excluded from the proximal transition zone upon ciliary maturation
Building upon this 3D nanoscale map of DAPs, we continued to add DAP-associated proteins to the map. We first examined the spatial relationship between DAPs and the membrane. ARL13B is a small GTPase associated with the ciliary membrane and ciliary vesicles 30, 31 . During the early stages of cilia growth (following an 8 h serum starvation), ARL13B puncta were localized in close proximity to FBF1 puncta (Fig   4a, b). However, after cilia had grown fully (after a 24 h starvation), ARL13B signals occupied the entire primary cilium, similar to IFT88 signals ( Supplementary Fig. 7), but were completely excluded from the ciliary base above DAPs where the proximal TZ was located ( Fig. 4c; Supplementary Fig. 8). These results suggest that ciliary vesicles carrying ARL13B may initially dock at the DAP region, but are later excluded from at least the proximal end of the TZ of intact cilia, consistent with the previous observation of a ciliary zone of exclusion (CIZE) 32 . EHD1, a protein involved in ciliary vesicle formation, was found to occupy a space wider than the diameter of the primary cilium ( We also examined the localization of TTBK2, a kinase interacting with CEP164 and promoting ciliogenesis 12,33 . In proliferating cells, TTBK2 was sporadically scattered with an asymmetric and incomplete coverage around the centriole (Fig. 4e). However, upon serum starvation, a broadened radial and longitudinal distribution of TTBK2 was induced, forming a thicker occupancy covering the distal regions of DAPs close to SCLT1 ( Fig. 4eg; Supplementary Fig. 10).

IFT88 and FBF1 localize to the distal appendage matrix, a newly identified region between the nine distal appendage blades
We next examined IFT88, a core IFT complex component known to be associated with DAPs 25 . We found that IFT88 occupied a broad angular distribution without clearly distinguishable nine puncta in the axial view (Fig. 5a). Lateral views revealed that in addition to being distributed along the entire ciliary axoneme, IFT88 is heavily concentrated at the ciliary base, at a longitudinal position spanning and proximal to the SCLT1 molecules ( Fig. 5b, c, and Supplementary Fig. 11). To further uncover the exact localization of IFT88 close to DAPs, we performed 3D two-color dSTORM imaging of IFT88 and SCLT1. We found that IFT88 occupied a conical distribution 5f), demonstrating that molecules can occupy or be retained in this unexplored area between DAP blades. We therefore designated the regions between the nine DAP blades as the distal appendage matrix (DAM), in which proteins such as IFT can be concentrated. In this sense, the intact DAP structure should be viewed as a conical addition at the centriole distal end consisting of the distal appendage blades (DABs) and matrix (DAM) (Fig. 5g).
The positioning of FBF1 at DAP gaps toward the ciliary membrane suggests that FBF1 localizes at the distal end of the DAM interface with the ciliary membrane ( Fig.   5h). Indeed, two-color IFT88-FBF1 dSTORM images showed that IFT88 coexists at the angular positions of FBF1 (Fig. 5i), and histograms of angular spacing revealed peaks close to multiples of 40° (Fig. 5j), confirming the positioning of FBF1 at the DAM. To check whether FBF1 and IFT88 at the DAM are functionally related, we generated the FBF1 -/-CRISPR knockout line ( Supplementary Fig. 13) and examined IFT88 localization in this region. Notably, in FBF1 -/cells, the distribution of IFT88 was reduced to a size comparable with that of CEP83 (Fig 5k, 5l), suggesting that IFT88 cannot span the DAM region without FBF1. The causal effect of FBF1 depletion on IFT88 distribution is consistent with the observation of interactions between FBF1 and IFT88 in a previous study 34 . Our results potentially imply a functional role for FBF1 in maintaining the DAM spatial arrangement (Fig. 5m).
We next tested whether DAB components perform a role in the structural arrangement of the DAM, and vice versa. Axial-view dSTORM imaging revealed no obvious change in CEP164 distribution upon FBF1 knockout (Fig. 6a, c). By contrast, knockdown of CEP164 did reduce the radial distribution of FBF1 (Fig. 6b, (Fig. 7a, b; Supplementary Fig. 15). Specifically, this resulted in defective initiation (Fig. 7a), since neither CP110 removal nor TTBK2 recruitment could occur at the mother centriole in SCLT1 -/cells (CRISPR knockout, Supplementary Fig. 13). However, intriguingly, knockout of the DAM component FBF1 did not affect ciliogenesis during the same initiation steps. Both CP110 removal and TTBK2 recruitment occurred normally in FBF1 -/cells (~70%), compared with WT cells (~80%), and yet cilia were detected in only ~14% of the population ( Fig. 7a; Supplementary Fig. 15). These results suggest that unlike CEP164 or SCLT1, which are essential for cilia initiation, FBF1 acts downstream of SCLT1 5 and is involved in cilia maturation and/or maintenance.

The DAM protein FBF1 is associated with ciliary gating of transmembrane proteins
To explore how FBF1 contributes to cilia maintenance, we examined whether it is required for ciliary gating. No significant difference in ciliary enrichment of ARL13B was detected between WT and FBF1 -/cilia (Fig. 7c). However, in striking contrast, although overexpressed Smoothened was highly concentrated in WT cilia, as reported previously 37 no such enrichment was observed in FBF1 -/cilia (Fig. 7d, e).
Consistently, the lack of ciliary enrichment was also seen in siFBF1 cells overexpressing SSTR3, another transmembrane protein known to localize to cilia ( Fig. 7f, g). These results suggest that either the ciliary entry or retention of transmembrane proteins (Smoothened or SSTR3) is defective in FBF1 -/cilia. We repeated the assay in CEP128 -/-;C-Nap1 -/cells known to grow fully surfaced cilia 38 , allowing clear distinction between the ciliary membrane and the plasma membrane. In the presence of FBF1, exogenously expressed Smoothened was highly enriched in surfaced cilia, but when FBF1 was depleted, as was seen in CEP128 -/-;C-Nap1 -/-;siFBF1 cells, the level of Smoothened at surfaced cilia was much lower, although still detectable (Fig. 7h, i)

Discussion
In summary, using superresolution microscopy, we determined the detailed molecular architecture of centriolar DAPs and the surrounding structure at the ciliary base, thereby bridging the gap between structural biology and cell biology, which is essential for precise functional studies. Our work provides an unprecedented roadmap to functional investigation based on more than 12 molecular elements at the interfaces of the centriole and cilium, and the plasma and ciliary membranes. Most importantly, instead of a pinwheel-like structure based on previous EM studies, our findings reveal that DAPs have a conical-shaped architecture composed of nine DABs within an embedded matrix that we termed the DAM. Both DABs and the DAM have distinct molecular elements that serve distinct functions. DABs contain CEP164, SCLT1, CEP89, and CEP83, all of which play an essential role in cilia initiation, whereas the previously unexplored DAM is filled with IFT molecules and FBF1, which is intriguingly found to facilitate ciliary gating of transmembrane proteins. As FBF1 is known to be associated with the apical junction complex that forms the barrier of epithelia 40 , it may also play a similar role at the junction between the ciliary membrane and the plasma membrane at the base of the ciliary pockets to gate the ciliary compartment.
Accumulation of IFT88 molecules in the gaps between DABs indicates a unique molecular arrangement in this region. In order to have protein molecules concentrated/localized at a specific region of the cell, where they can dynamically enter/attach and exit/detach, the region must either be gated with a structural barrier, or alternatively include a scaffold that can bind to these protein molecules with the required affinity. In either case, dedicated structures must be present in this. In our current study, we found that FBF1 and IFT88 were concentrated/localized at the space between DABs, outside of the ciliary compartment, in a region previously thought to be empty based on EM studies. We thus conclude that, rather than being empty, the region is equipped with structures that are able to trap protein molecules to form the DAM. We use the term 'matrix' to indicate a surrounding structure or material in which something develops, rather than pointing to any defined architecture.
This current work provides a functional extension of our previous finding of a hierarchical relationship among SCLT1, CEP164, and FBF1 5 . While SCLT1 mediates the recruitment of both FBF1 and CEP164 to centrioles, recruitment of each is independent of the other. Thus, loss of FBF1 alone is completely different from loss of SCLT1, which leads to removal of both FBF1 and CEP164 from centrioles. This previously determined 'nonlinear' assembly hierarchy is consistent with the observations in our current study; Both SCLT1 and CEP164 are components of DABs, and knockout of either prevents cilia initiation. By contrast, FBF1 is a matrixassociated protein occupying the space between the blades, and loss of FBF1 affects cilia maturation/maintenance (the integrity of the ciliary membrane), rather than cilia initiation.
Our previous genetic studies showed that SCLT1 acts upstream of CEP164, whereas our superresolution studies showed that SCLT1 is "sandwiched" by the signal from the 'blade-like' CEP164, not immediately obvious how this localization relationship occurs. The initial recruitment of CEP164, which depends on CEP83 and SCLT1, may be completely separate from the propagation of CEP164 along the structure that forms the final localization structure. In particular, CEP164 harbors the coiled-coil domains known to mediate self-oligomerization. That is, in the absence of CEP83/SCLT1, CEP164 molecules were not detected at centrioles. However, once initial CEP164 molecules are recruited, they can recruit other CEP164 molecules through protein-protein interaction or self-oligomerization processes. A similar concept can be seen in the PCM organization surrounding the centriole; only the inner layer of molecules is directly recruited by the centriole, whereas molecules in the outer layer are recruited through self-assembly or self-oligomerization 41 . Our previous determination of the assembly hierarchy was based on genetics, rather than the actual, stepwise, biochemical assembly pathway. The findings of the present study suggest that the basic framework of the DAP structure can in principle be assembled in the cytoplasm or in the DAP region initiated by recruited seed molecules to form the final molecular architecture.         imaging of DAPs, a ring pattern of labeled SCLT1 or FBF1 was usually sought to identify those with appropriate orientations. For lateral imaging, a rod-like pattern of SCLT1 or FBF1 was used to assure a lateral view of the centriole-cilium.
To determine the diameter of DAPs and their associated proteins shown in Fig. 1b, the position of individual superresolved puncta forming a ring-shaped pattern was first measured and then fitted with a circle to estimate its diameter. For CEP164, Cby1, IFT88, and TTBK2, where their distributions were relatively dispersed, their radial positions were described with a radius defined as the distance between the puncta and the center, yielding a radius histogram that was fitted with a Gaussian or mixed Gaussian curve (Supplementary Fig. 1). To determine how DAP proteins aligned with each other (Fig. 2ae), angular analysis was performed. Briefly, the angular position of each DAP punctum was recorded, and angle differences between each pair of puncta for two protein species indicated how they were arranged relative to each other. The concept is described in detailed in Supplementary Fig. 4. To align dSTORM and EM images in Fig. 3d, FBF1 signals of dSTORM images were placed at the tip of the high electron density signals of slanted DAPs in EM images.

3D printing model
To reconstruct a 3D model of DAPs, we first drew the DAP structure using 3D CAD