FBF1 localization differs from those of other DAP proteins

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 and Supplementary Fig. ⁴). 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 electron-dense 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, Supplementary Fig. ⁵, and Supplementary Table ²), reflecting their similar circumferential arrangements. By contrast, histograms for CEP164-FBF1 and FBF1-SCLT1 pairs revealed peaks close to 20°, 60°, etc. (Fig. 2f and Supplementary Table ²), confirming their alternating arrangements. Thus, unlike other core DAP components, which localize to the slanted DAP blades (Fig. 2a−c, g), FBF1 is instead associated with the space between DAP blades (Fig. 2d, e, g).

Open in a separate window

Fig. 2

Angular distributions of CEP83, CEP89, SCLT1, FBF1, and CEP164. a–e Two-color axial-view images of a CEP83-SCLT1, b CEP89-CEP164, c CEP164-SCLT1, d CEP164-FBF1, and e FBF1-SCLT1 pairs. CEP83 and SCLT1 a, and CEP89 and CEP164 b align well with each other. SCLT1 likely resides within each “blade-like” distribution of CEP164 signals c, whereas the angular positions of FBF1 puncta are different from those of CEP164 puncta d and SCLT1 puncta e. White bars mark SCLT1 in a or CEP164 in b–e; circles mark SCLT1 in c or FBF1 in d–e; triangles mark SCLT1 in e. f Angular analysis in the circumferential direction shows that CEP83, CEP89, SCLT1, and CEP164 exhibit similar angular distributions, while FBF1 is arranged alternately with SCLT1 or CEP164, suggesting it fills the gaps between adjacent CEP164 puncta. g Model of the localization of DAP proteins. FBF1 occupies the gaps between DAP blades, whereas C2CD3 localizes to the centriolar lumen. Bar a–e=200nm

Axial and lateral images reveal 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 ~130nm (Fig. 3b and Supplementary Fig. ⁶). The distal layer of CEP89 is close to the same longitudinal position as FBF1 (Fig. 3b and Supplementary Fig. ⁷), whereas the proximal layer is located longitudinally close to ODF2 in sDAPs (Fig. 3c and Supplementary Fig. ⁷), suggesting possible dual roles for CEP89 in DAPs and sDAPs. By overlaying the dSTORM signals over EM images of ciliated centrioles, we found that CEP164, CEP83, SCLT1, and FBF1 align well with DAPs, with FBF1 localizing closest to the ciliary membrane (Fig. 3d). CP110 and CEP97 in non-ciliated centrioles are localized at a longitudinal position, similar to SCLT1 (Fig. 3e). Collectively, the relative positions of DAP proteins in the lateral view reflect the contour of a DAP blade (Fig. 3f) where the diverse localization variations among different proteins suggest their largely different distribution ranges in space.

Open in a separate window

Fig. 3

Three-dimensional spatial map of distal appendage (DAP)-associated proteins. a Three representative two-color dSTORM images from a lateral view of each pair of DAP proteins illustrates their relative longitudinal positions. Pink and green arrowheads depict the heights of FBF1 and SCLT1, respectively. b Representative images of DAP proteins are aligned and combined according to their average longitudinal positions. The magnified image (inset) demonstrates the mapping of DAP proteins on the tilted arrangement of the DAP structure (dashed line). c Super-resolved images of the subdistal appendage protein ODF2 reveal its two-layered localization at a far lower location than that of SCLT1. d Merged super-resolution image in b overlaid with a longitudinally sliced EM micrograph. CEP83, SCLT1, and FBF1 are well aligned with the electron density of the slanted DAPs. e CP110 and CEP97 are localized longitudinally, similar to SCLT1, but with a narrower lateral width. f Combined relative localization of DAP-associated proteins in radial and longitudinal directions, revealing the envelope of the slanted arrangement of a DAP blade (dotted line). The results (presented as mean±SD) and the numbers of measured points are summarized in Supplementary Table ¹. g A 3D computational model illustrating the positioning of proteins at the DAP region with a backbone structure based on EM¹⁰. h CEP164, CEP83, SCLT1, and the distal layer of CEP89 mapped onto a blade of the 3D-reconstructed DAPs. i FBF1 localizes to gaps between two adjacent DAP blades, possibly at the proximal boundary of the ciliary pocket. Bars a, b (left), c, and e=200nm; b (inset)=50nm

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 EM¹⁰. 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 (Fig. 3i). Thus, FBF1 may serve a role distinct from other core DAP components.

ARL13B lies adjacent to DAPs during ciliary initiation

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 8h serum starvation), ARL13B puncta were localized in close proximity to FBF1 puncta (Fig. 4a, b). However, after cilia had grown fully (after a 24h starvation), ARL13B signals occupied the entire primary cilium, similar to IFT88 signals (Supplementary Fig. ⁸), but were completely excluded from the ciliary base above DAPs where the proximal TZ was located (Fig. 4c and Supplementary Fig. ⁹). 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 exclusion³². EHD1, a protein involved in ciliary vesicle formation, was found to occupy a space wider than the diameter of the primary cilium (Fig. 4d, Supplementary Fig. ⁹, and ¹⁰; Supplementary Movie ¹), consistent with its presence on the ciliary pocket membrane¹⁵.

Open in a separate window

Fig. 4

Localization of other DAP-related proteins ARL13B, EHD1, and TTBK2. a During the early stages of cilia growth, ARL13B is located adjacent to FBF1 (arrows), as confirmed by correlation analysis in b. c ARL13B becomes localized to cilia during or after ciliogenesis with an exclusion zone approximately at the proximal transition zone. d The vesicle-forming protein EHD1 is localized to a peripheral region wider than that of ARL13B, presumably on the membrane of the ciliary pocket. Axial e, f and lateral g views of the distribution of TTBK2 with and without serum, showing a broader and more enriched TTBK2 distribution in both radial and longitudinal directions after serum starvation. Results are presented as mean±SD of 24, 20, and 23 centrioles for 0, 24, and 48h, respectively. Bars=200nm

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 and Supplementary Fig. ¹¹).

IFT88 and FBF1 localize to the DAP matrix

We next examined IFT88, a core IFT complex component known to be associated with DAPs²⁵. 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. ¹²). 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 (Fig. 5d, e, Supplementary Fig. ¹³, and Supplementary Movie ²), resembling the tapered shape of DAPs. Strikingly, the relative distributions of IFT88 and SCLT1 reveal abundant IFT88 proteins filling the gaps between neighboring SCLT1 molecules (Fig. 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 DAP 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 the DAM (Fig. 5g).

Open in a separate window

Fig. 5

Localization of IFT88 revealing a distal appendage matrix (DAM). a Three representative axial-view dSTORM images of IFT88 reveal a broad angular distribution without a clear nine-fold symmetric pattern. b Two representative sets of two-color images showing IFT88 spanning a longitudinal range relative to the location of SCLT1 (marked with a dashed line). The solid line in c depicts the distribution in the lower panel in b, whereas the dotted line illustrates normalized distributions averaged over nine data pairs. d, e 3D STORM imaging of IFT88 showing the tapered distribution of IFT88 that is wider at the distal end of the DAP region. f Three-dimensional two-color super-resolution image revealing IFT88 proteins in the gaps between neighboring SCLT1 regions, as shown in the lower image (derived from the upper-right image) in which nine SCLT1 puncta are numbered accordingly. g The region between adjacent DABs is designated as the DAM. h Model showing the localization of FBF1 and IFT88 at the DAM. i Two-color imaging of IFT88-FBF1 shows the angular coincidence between IFT88 and FBF1 at the DAM. j Histogram of angular spacing between IFT88 and FBF1. k, l The radial distribution of IFT88 was reduced in FBF1^−/− cells. m Model showing a functional role of FBF1 in maintaining the DAM spatial arrangement. Bars=200nm for all, except f (100nm)

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. ¹⁴) 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, l), 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 study³⁴. 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, d). Thus, without CEP164 at the DAB tips, even though FBF1 was recruited to the DAP region⁵, it was unable to reach the outermost DAM location (Fig. 6e). CEP164 depletion has been shown to affect IFT recruitment³⁵ and it is also possible that the DAM was damaged upon knockdown of CEP164, resulting in the collapse of the FBF1 ring.

Open in a separate window

Fig. 6

DAB components mediating the structural arrangement of the DAM. a, b Axial view dSTORM images reveal the distribution of a CEP164 and b FBF1 at the mother centrioles upon CRISPR/Cas9 FBF1 knockout and CEP164 siRNA silencing, respectively. It is noteworthy that the FBF1 ring formed upon CEP164 knockdown becomes smaller than that of controlled cells (b, right panel). c, d Statistical analysis of the radial distribution of CEP164 and FBF1 upon perturbation. The size of CEP164 in FBF1^−/− cells is comparable to that in WT cells c, whereas the size of FBF1 is reduced upon CEP164 knockdown d. e Model showing that the arrangement of FBF1 is no longer maintained at the outer region of the DAM in the absence of CEP164. f Three-dimensional model based on all axial and lateral super-resolution images illustrating the molecular architecture on the framework of DABs, DAM, axoneme, and ciliary membrane. g Close-up view of the DAP region, suggesting that FBF1 potentially interfaces the ciliary membrane and the DAM. Bars=200nm

Subsequently, we expanded our 3D computational model to include all ciliary structures examined, including the membrane, pocket, axoneme, DABs, and the DAM (Fig. 6f, g, Supplementary Fig. ¹⁵, and Supplementary Movie ³).

DAB proteins but not FBF1 are essential for ciliogenesis

We then examined whether DABs and the DAM have distinct functions in ciliogenesis. Consistently, depletion of the DAB component CEP164, as demonstrated previously^35,36, or SCLT1, as demonstrated herein, fully abolished DABs and ciliogenesis (Fig. 7a, b and Supplementary Fig. ¹⁶). Specifically, this resulted in defective initiation (Fig. 7a), as neither CP110 removal nor TTBK2 recruitment could occur at the mother centriole in SCLT1^−/− cells (CRISPR knockout; Supplementary Fig. ¹⁴). 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 and Supplementary Fig. ¹⁶). These results suggest that unlike CEP164 or SCLT1, which are essential for cilia initiation, FBF1 acts downstream of SCLT1⁵ and is involved in cilia maturation and/or maintenance.

Open in a separate window

Fig. 7

SCLT1 in ciliogenesis vs. FBF1 in transmembrane protein gating. Deletion of SCLT1 fully inhibits ciliogenesis during the initiation stage, as neither CP110 removal nor TTBK2 recruitment occurred at the mother centriole. However, both CP110 removal and TTBK2 recruitment were observed in FBF1^−/− cells. White boxes highlight the areas at the basal bodies or centrioles shown in the insets. b EM micrographs showing the absence of DABs in SCLT1^−/− cells and visible DABs in wild-type cells (arrows). Sub-distal appendages (double arrows) remain present in SCLT1^−/− cells. c ARL13B is enriched in both WT and FBF1^−/− cilia. (n=227 and 170 cilia for WT and FBF1^−/−, respectively). d, e Overexpressed Smoothened was highly concentrated in WT cilia but was not enriched in FBF1^−/− cilia. The insets in d show enlarged images of the boxed region. e Smo-GFP intensities relative to cytoplasmic levels. (n=29 and 22 cilia for WT and FBF1^−/−, respectively). f, g SSTR3-GFP signals at cilia were lower in siFBF1 cells. (n=24 and 21 cilia for ctrl RNAi and FBF1 RNAi, respectively in g). h, i Overexpressed Smoothened is abundant in surfaced cilia in CEP128^−/−;C-Nap1^−/− cells, but less enriched in surfaced cilia in CEP128^−/−;C-Nap1^−/−;siFBF1 cells, suggesting Smoothened was not retained in FBF1-depleted cilia. (n=24 and 18 cilia for ctrl RNAi and FBF1 RNAi, respectively in i). j Enrichment of transition zone proteins TCTN2, TMEM67, RPGRIP1L, and CEP290 was unaffected in FBF1^−/− cells compared with WT cells. k, l Model showing that FBF1 plays a role in gating transmembrane proteins, presumably at the junction between the ciliary membrane and the plasma membrane. Data for e, g and i are presented as mean±s.d. ****P<0.0001, two-tailed Student’s t-test. Bars=5μm in a, d, f, h, 100nm in b, and 1μm in c, j

Cell culture and transfection

Human RPE cells (hTERT RPE-1, ATCC-CRL-4000) were seeded on poly-l-lysine-coated coverslips and cultured at 37°C under 5% CO₂ in Dulbecco’s modified Eagle’s medium (DMEM)/F-12 mixture medium (1:1; 11330-032, Gibco, Thermo Fisher Scientific) supplemented with 10% fetal bovine serum (FBS) and 1% penicillin–streptomycin. HEK293T cells were cultured in DMEM with 10% FBS and 1% penicillin–streptomycin. Cilium formation in RPE-1 cells was induced by serum starvation for 24−48h. Lentiviral stocks were prepared according to standard procedures (PT-5144-1, Clontech, Mountain View, CA, USA). RNA interference (RNAi)-mediated knockdown of human FBF1 in RPE-1 cells was performed using 12.5μM of three different small interfering RNA (siRNA) duplex oligonucleotides (Thermo Fisher, ambion silencer select) targeting the sequences 5′-GGGAGAACCAGGCTCCAAATT-3′, 5′-GCATGGATGCTGATATCTTTT-3′, and 5′-TGAGAAGCCTGTTAAACTATT-3′. siRNA oligos were delivered using Lipofectamine RNAiMAX Reagent (13778150, Thermo Fisher). Human CEP89 (NM_032816) depletion was carried out using a combination of five different lentiviral short hairpin RNAs (RNAi core, Academia Sinica, Taiwan) with the following targeting sequences: sh1, 5′-GCACGTCAAAGATATACAGAA-3′; sh 2, 5′-CAGGAGTCATTTCAAACACAT-3′; sh 3, 5′-GAATGTAGAACTCAGTCGATA-3′; sh 4, 5′-CCAGATATAACTGGTAGAGCA-3′, and sh 5, 5′-GTAACGTTTCCGATTGTGAAA-3′. All lentiviruses were generated by transient cotransfection of HEK293T cells with packaging and envelope vectors (PT-5144-1, Clontech) using a TransIT-293 Transfection Reagent Kit (MIR 2700, Mirus, Madison, WI, USA). Stable expression of human CEP89 in RPE-1 cells was performed through lentiviral transduction and selection using 5mg/mL puromycin. To induce ciliary growth, RPE-1 was starved of serum for 48h. SSTR3-GFP (35623, Addgene, Cambridge, MA, USA⁴²) was cloned into the pLVX-Tight-Puro vector (Clontech) for tet-inducible expression, as was Smoothened-GFP (described previously³⁸). For Smo-GFP/SSRT3-GFP, lentivirus production and infection were performed as described previously⁵ using psPax2 (packaging, 12260, Addgene), pCMV-VSV-G (envelope, 8454, Addgene), and pLVXTet-On Advance (Clontech) plasmids. Cells were infected simultaneously with lentiviral vectors for rttA expression (Tet-On advance) and Smo-GFP (or SSTR3-GFP). Clonal populations displaying uniform transgene expression were isolated prior to siRNA experiments. Tet induction was performed with 1μg/mL doxycycline for 48h (during serum starvation). Cell lines used in this study were tested for absence of mycoplasma contamination using immunofluorescence with DAPI (4′,6-diamidino-2-phenylindole) staining and a PCR assay (EZ-PCR Mycoplasma Test Kit, Biological Industries).

CRISPR construction of SCLT1^−/− and FBF1^−/− cells

RNA-guided targeting of genes in human cells was achieved through coexpression of the Cas9 protein with guide RNAs (gRNAs) using reagents prepared by the Church research group⁴³, which are available from Addgene (http://www.addgene.org/crispr/church/). Targeting sequences for SCLT1 and FBF1 were 5′-GGTGTCTGCCCATGGAAGAG-3′ and 5′-GGCATGGATGCTGATATCTT-3′, respectively, and these were cloned into the gRNA cloning vector (41824, Addgene) via the Gibson assembly method (New England Biolabs, Ipswich, MA, USA) as described previously⁴³. Pure SCLT1 and FBF1 knockout cell lines were then established through clonal propagation from single cells. For genotyping, the following PCR primers were used: 5′-ACGCGGATCCAACGTAAGTGTTAGAATGCT-3′ and 5′-CTAGTCTAGAAGCTAGGTGCATTTAACAAT-3′ for SCLT1 alleles, and 5′-ACGCGGATCCCCCGAGACCAGCCACCTAGG-3′ and 5′-CTAGTCTAGACCTTTCTCACTGGCTACTGA-3′ for FBF1 alleles. PCR products were cloned and sequenced.

Immunofluorescence

Cells on glass coverslips were fixed with either 4% paraformaldehyde at room temperature or methanol at −20°C after one wash with phosphate-buffered saline (PBS). Cells were then permeabilized with 0.1% PBST (PBS with 0.1% Triton X-100) for 10min and blocked with 3% bovine serum albumin (BSA) for 30min. Primary antibodies were diluted to optimized concentrations with 0.1% BSA in PBST and applied to cells. After a 1h incubation with antibody, cells were washed three times with 0.1% PBST to remove unbound antibodies, then incubated with optimally diluted secondary antibodies for 1h. Cells were finally washed three times with 0.1% PBST and stored in PBS with sodium azide at 4°C until later use.

dSTORM imaging and analysis

Super-resolution imaging was performed with a modified inverted microscope (Eclipse Ti-E, Nikon, Tokyo, Japan). Three light sources were merged onto a laser merge module (ILE, Spectral Applied Research, Richmond Hill, Ontario, Canada), which was integrated with individual controllers. Beams from a 637nm laser (OBIS 637 LX 140mW, Coherent, Santa Clara, CA, USA), a 561nm laser (Jive 561 150mW, Cobolt, Solna, Sweden), and a 405nm laser (OBIS 405 LX 100mW, Coherent) were homogenized (Borealis Conditioning Unit, Spectral Applied Research) and focused on the back aperture of a 100× 1.49 numerical aperture oil-immersion objective (CFI Apo TIRF, Nikon) for wide-field illumination of samples. A perfect focusing system (Nikon) was used to minimize z-axis drift. The 637nm and 561nm laser lines were operated at a high intensity of ~1−5kWcm⁻² to quench most of the fluorescence from Alexa 647 and Cy3B. A weak 405nm beam was used to activate a portion of the dyes that were converted from a ground state to an excited state. The collected fluorescence was cleaned by a single-band emission filter and imaged on an electron-multiplying charge-coupled device (EMCCD) camera (Evolve 512 Delta, Photometrics, Tucson, AZ, USA) with an overall pixel size of 93nm. Fiducial markers (Tetraspeck, Thermo Fisher) were used to correct the drift. Typically, 10,000−20,000 frames (for two-dimensional dSTORM) and ~50,000 frames (for 3D dSTORM) were recorded at a rate of 50−67fps. Individual single-molecule peaks were then real-time localized using a MetaMorph Super-resolution Module (Molecular Devices, Sunnyvale, CA, USA), based on a wavelet segmentation algorithm. Super-resolution images in figures were cleaned with a Gaussian filter with a radius of 0.6−1 pixels. For single-color imaging of Alexa 647, signals were filtered with a bandpass filter (700/75, Chroma, Bellows Falls, VT, USA); for two-color imaging, the Alexa 647 channel was first recorded, then the Cy3B channel was acquired with the corresponding filter (593/40, Chroma). For 3D imaging, a cylindrical lens with a focal length of 25cm was added before the camera to create astigmatism, from which the z position was determined through calibration. The point spread function (PSF)-based calibration curve was generated by scanning a fluorescent bead in the z-direction and calculating the elliptical widths (w_x and w_y).

dSTORM imaging was performed in an imaging buffer containing Tris-HCl, NaCl (TN) buffer at pH 8.0, and an oxygen-scavenging system consisting of 60−100mM mercaptoethylamine (MEA) at pH 8.0, 0.5mgmL⁻¹ glucose oxidase, 40μg mL⁻¹ catalase, and 10% glucose (Sigma-Aldrich, St. Louis, MO, USA). Lateral position drift was measured during the acquisition and corrected by frame-by-frame correlation of the fiducial markers with ImageJ. Chromatic aberration between long and short wavelength channels was compensated with a customized algorithm relocating each pixel of a 561nm image to its targeted position in the 647nm channel with a predefined correction function obtained by parabolic mapping of multiple calibration beads. The system resolutions for Alexa 647 and Cy3B were characterized with fixed samples incubated in an imaging buffer of 80mM MEA and evaluated via the analysis of localization precision as a function of the photon count, the noise level, the pixel size of a camera, and the width of a PSF of single-molecule emission (Supplementary Fig. ¹)⁴⁴. For axial 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 super-resolved 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. ²). 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. ⁵. 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.

Transmission electron microscopy

RPE-1 cells grown on coverslips made of Aclar film (Electron Microscopy Sciences, Hatfield, PA, USA) were fixed in 4% paraformaldehyde and 2.5% glutaraldehyde with 0.1% tannic acid in 0.1M sodium cacodylate buffer at room temperature for 30min, postfixed in 1% OsO₄ in sodium cacodylate buffer for 30min on ice, dehydrated in a graded series of ethanol, infiltrated with EPON812 resin (Electron Microscopy Sciences), and embedded in the resin. Serial sections (~90nm thickness) were cut on a microtome (Ultracut UC6; Leica, Wetzlar, Germany) and stained with 1% uranyl acetate and 1% lead citrate. Samples were examined on a transmission electron microscope (Tecnai G2Spirit Twin, FEI, Hillsboro, OR, USA).

Three-dimensional printing model

To reconstruct a 3D model of DAPs, we first drew the DAP structure using 3D CAD design software (SOLIDWORKS, Waltham, MA, USA) then fabricated a solid model with a 3D printer (Form 1+, Formlabs, Somerville, MA, USA). The dimensions of the model were based on the localization positions obtained from the super-resolution results in Fig. 1b, d, Fig. 2a-f, and Fig. 3a, c, e, f, and on information from previous studies^7,10. For the 3D printed model, DAP proteins in each “blade” were printed as spherical shapes with their centers at the mean positions, then painted in the appropriate colors. The printed model, including microtubule triplets and doublets, DABs, TZ, and ciliary pocket, has an overall size of 85mm in width (diameter) by 155mm in height at a 150,000:1 scale. To properly place FBF1 in the gaps between two adjacent DAPs, an auxiliary annular scaffold, which does not exist in the actual complex, was added.

We thank Erich Nigg, Christopher Westlake, Steve Caplan, and Tang Tang for sharing reagents. This work was supported by the Ministry of Science and Technology, Taiwan (Grant Number 103-2112-M-001-039-MY3), Academia Sinica Career Development Award, Academia Sinica Nano Program, the University of Alabama at Birmingham (UAB) Hepato/Renal Fibrocystic Diseases Core Center (HRFDCC) Pilot Award (NIH 5P30DK074038-09) to J.-C.L.; NIH grant GM088253, American Cancer Society grant RSG-14-153-01-CCG, and the Geoffrey Beene Cancer Research Center grant to M.-F.B.T.