α-/γ-Taxilin are required for centriolar subdistal appendage assembly and microtubule organization

The centrosome composed of a pair of centrioles (mother and daughter) and pericentriolar material, and is mainly responsible for microtubule nucleation and anchorage in animal cells. The subdistal appendage (SDA) is a centriolar structure located at the mother centriole’s subdistal region, and it functions in microtubule anchorage. However, the molecular composition and detailed structure of the SDA remain largely unknown. Here, we identified α-taxilin and γ-taxilin as new SDA components that form a complex via their coiled-coil domains and that serve as a new subgroup during SDA hierarchical assembly. The taxilins’ SDA localization is dependent on ODF2, and α-taxilin recruits CEP170 to the SDA. Functional analyses suggest that α- and γ-taxilin are responsible for SDA structural integrity and centrosomal microtubule anchorage during interphase and for proper spindle orientation during metaphase. Our results shed light on the molecular components and functional understanding of the SDA hierarchical assembly and microtubule organization.


Introduction
The centrosome, the main microtubule organizing center (MTOC) in many eukaryotic cells, participates in microtubule-related activities that include cell division (Cabral et al., 2019;Wu et al., 2012), cell polarity maintenance (Burute et al., 2017), cell signaling transduction (Barvitenko et al., 2018), and ciliogenesis (Pitaval et al., 2017;Tu et al., 2018). The centrosome is a non-membrane-bound organelle composed of a pair of orthogonally arranged centrioles and surrounded by pericentriolar material (PCM) (Delattre and Gönczy, 2004). The two centrioles, mother and daughter, are distinguished from each other by the decorations at the distal and subdistal ends of the mother centriole, called the distal/subdistal appendages (DAs/SDAs) (Tischer et al., 2021).
Accumulated data have revealed the structural characteristics and functions of DAs and SDAs. DA proteins (i.e. CEP83, CEP89, FBF1, SCLT1, and CEP164) are essential for centriole-to-membrane

Results
Screening of α-taxilin and γ-taxilin as new SDA components APEX2-mediated proximal labeling is an approach that uses hydrogen peroxide (H 2 O 2 ) as an oxidant to catalyze biotin-phenol (BP), a small molecular substrate, to produce the reactive BP radical that conjugates to the endogenous proteins that are proximal to APEX2 (Hung et al., 2016b). To search for new SDA components, we labeled two previously identified SDA components CCDC68 and CCDC120 (Huang et al., 2017) by infusing them with V5-tagged APEX2. When overexpressed in HeLa cells, the V5 immunostained CCDC68 and CCDC120 fusion proteins showed a ring (top view) ( Figure 1-figure supplement 1A) or three dots (side view) (Figure 1-figure supplement 1B) at the centrosome, results that were consistent with their SDA locations when they had been stained by their specific antibodies (Huang et al., 2017). After treatment with BP and H 2 O 2 , the biotinylated endogenous proteins co-localized and encircled V5-tagged CCDC120 or V5-tagged CCDC68 (Figure 1figure supplement 1A-B), thus indicating that APEX2-mediated biotinylation had successfully marked the endogenous proteins around CCDC68 or CCDC120 at the centrosomes. As a negative control, samples without BP or H 2 O 2 showed no biotinylation signal. Furthermore, immunoblotting results also confirmed the endogenous biotinylated proteins mediated by CCDC68 and CCDC120 proximal labeling (Figure 1-figure supplement 1C-D). Finally, the biotinylated proteins were enriched by Streptavidin-coated beads and sent for mass spectrographic (MS) analysis to search for CCDC68 and CCDC120 proximal candidates.
Among those candidates for CCDC68 and CCDC120 proximal labeling, a series of previously identified centriolar proteins (marked in green in Figure 1-figure supplement 1E) were found, and thus verified the efficiency of our centrosomal proteomics approach. Since several centrosomal proteins possess the coiled-coil domain that is the basis for the protein-protein interactions essential for centrosome configuration (Andersen et al., 2003), we focused on proteins containing one or more coiled-coil domains. To determine their sub-cellular localizations, each protein candidate was tagged (e.g. V5, mNeonGreen, or pmEmerald) and expressed in U2OS cells. Other than the already known centrosome proteins, five new proteins, including α-taxilin, DACT1, DRG2, NCAPH2, and SMAP2 ( Figure 1-figure supplement 1F), were found co-localized with centrosome markers CP110 or γ-tubulin, thus proving that they resided on centrosomes. Among them, DRG2 was found in both CCDC68 and CCDC120 proximal candidates; α-taxilin, NCAPH2, and SMAP2 were found in CCDC68 proximal candidates; and DACT1 was found in CCDC120 proximal candidates (Figure 1-figure supplement 1E-F). A previous study had shown γ-taxilin co-localized with Nek2A at the centrosome during interphase (Makiyama et al., 2018). The immunofluorescence assay showed it co-localized with γ-tubulin ( Figure 1-figure supplement 1E), thus confirming its centrosomal localization.
Among the proteins that showed centrosomal localization, α-and γ-taxilin are similar in that their middle regions each has a long coiled-coil domain (Figure 1-figure supplement 2A; Makiyama et al., 2018). Both the anti-α-taxilin antibody and anti-γ-taxilin antibody, which were designed to recognize the C-terminus of each of those proteins (Figure 1-figure supplement 2A-C), were chosen to detect their specificities. Immunoblotting results showed that the two proteins' molecular weights (above 72 kDa for α-taxilin and about 70 kDa for γ-taxilin) were slightly higher than predicted (Figure 1-figure supplement 2B-C). Immunoblotting also detected significant decreases in the amount of those proteins (Figure 1-figure supplement 2B-C). The immunofluorescence assay showed that α-taxilin and γ-taxilin stained by their antibodies focused prominently at the centrosome, as indicated by their co-localizations with γ-tubulin in human RPE-1 cells. Also, the fluorescence intensities of both taxilins decreased at the centrosome after siRNA knockdown (Figure 1-figure  supplement 2D). Those results show both the antibodies' specialties and the effectiveness of siRNAs.
Detailed α-and γ-taxilin localization at the centrosome was determined using super-resolution microscopy with a 3D-structured illumination system (SIM). In the G1 phase, the top views of α-and γ-taxilin showed their ring-like patterns encompassing one centrin-3 dot that resides in the distal lumen of centrioles (Middendorp et al., 1997;Figure 1A-B). The α-and γ-taxilin side views showed three dots occupying two levels with one level (two dots) beside ODF2 and thus at the SDA ( Figure 1A-B), and the other level (one dot) assumed to be at the proximal end ( Figure 1A-E). Those patterns resembled those of the ninein group, such as ninein and CEP170 (Mazo et al., 2016). Indeed, α-taxilin clearly co-localized with the ninein and CEP170 rings in the top view, and with the three ninein and/ or CEP170 dots in the side view ( Figure 1C and E). Corresponding γ-taxilin patterns resembled those of α-taxilin by containing a smaller ring that resided inside the ninein and CEP170 signals ( Figure 1D and F).
We then measured SDA proteins' longitudinal positions by co-immunostaining with the proximal end protein C-Nap1 (Vlijm et al., 2018), which serves as a position reference ( Figure 2C-D). ODF2 covered a wide range, occupying between about 360 nm and 589 nm distances from C-Nap1, and so it supposedly resides at the base of both the DA and SDA, respectively. This is comparable to the data obtained by a dSTORM super-resolution microscopy study (Chong et al., 2020). Expect for CCDC120, the rest of the SDA proteins had two layered positions with one layer proximal to C-Nap1 and the other localizing between the two ODF2 layers. The side view of CCDC120 showed Figure 1. Structured illumination microscopy (SIM) images and characterizations of α-taxilin and γ-taxilin at the centrosomes. (A) Immunostained α-taxilin (green) and centrin-3 (magenta) in RPE-1 cells transfected with ODF2-Scarlet (red). Scale bar, 1 µm. The cartoons to the right of each set of images graphically depict the merge images. (B) Immunostained ODF2 (red) and centrin-3 (magenta) in RPE-1 cells transfected with GFP-γ-taxilin (green). Scale bar, 1 µm. (C) Immunostained α-taxilin (red) and ninein (magenta) in RPE-1 cells transfected with centrin-3-GFP (green). Scale bar, 1 µm. Figure 1 continued on next page a three-layered pattern. At the proximal end, one layer resembled ninein group proteins; while the other two layers settled about 458 nm and 636 nm from C-Nap1 and were assumed to reside at the SDA and DA structures, respectively ( Figure 2C-D).
By combining those SDA diameter and longitudinal position measurements seen in STED images, the relative localization of each SDA protein was established at the mother centriole. First, ODF2 resides closest to the centriole wall, while ninein and CEP170 reside at the tip of SDA. At the SDA structure, CCDC68 and CCDC120 localize close to ODF2, and α-taxilin and γ-taxilin are located in concentric circles between CCDC120 and ninein ( Figure 2B and E). In the longitudinal position, γ-taxilin was higher than CCDC68, CCDC120, ninein and CEP170, while α-taxilin was lower than those proteins ( Figure 2D-E).

α-Taxilin and γ-taxilin localization at the SDA depends on ODF2
To reveal the assembly manners of α-and γ-taxilin in the SDAs, the interactions between them and other SDA proteins were characterized through immunoprecipitation. Since ODF2 is the innermost SDAs layer, it is the starting point for assembly of the other SDA components (e.g. TCHP, CCDC68, CCDC120, etc.) (Ibi et al., 2011;Huang et al., 2017). The immunoprecipitation results revealed that endogenous ODF2 interacted with both α-and γ-taxilin in RPE-1 cells ( Figure 3A-B). Correspondingly, the specific localization of ODF2 and α-or γ-taxilin via STED nanoscopy is consistent with their specific localization as indicated in Figure 2. The top view of the SDA region showed that α-and γ-taxilin encircled the ODF2 ring ( Figure 3C-D). When viewed from the side, α-and γ-taxilin localized proximal to ODF2's upper level ( Figure 3C-D).
Furthermore, immunoblot analysis showed that when ODF2 was depleted by siRNA knockdown, α-and γ-taxilin levels in those cells did not change significantly ( Figure 3E-G). However, fluorescence intensity of both α-and γ-taxilin decreased at the centrosomes after ODF2 depletion (Figure 3-figure supplement 1A-B), suggesting that ODF2 affects their centrosomal localizations without affecting their protein levels. As shown by 3D-SIM images, ODF2 depletion by siRNA treatment resulted in α-taxilin or γ-taxilin signal loss at the SDA region, while their proximal end signals were not affected ( Figure 3H-I). Moreover, those lost SDA signals could be rescued by overexpressing full-length ODF2. However, the 1-59 aa deletion mutant, which affects ODF2 localization at the SDA (Tateishi et al., 2013) and recruitment of other SDA components (Huang et al., 2017), could not rescue SDA localization of either α-taxilin or γ-taxilin after siRNA treatment ( Figure 3H-I). These data further illuminate the role of ODF2 in α-and γ-taxilin assembly at the SDAs.
The online version of this article includes the following source data and figure supplement(s) for figure 1:         Previously, TCHP was reported to reside at the SDA midzone and to interact with ODF2 and ninein (Ibi et al., 2011). Here, endogenous TCHP did not interact with either α-taxilin or γ-taxilin in HEK-293T cells (Figure 3-figure supplement 1K). Therefore, the proper localization of α-taxilin and γ-taxilin at the SDA depends on ODF2, but not on CCDC68, CCDC120, or TCHP.

α-Taxilin and γ-taxilin form a complex at the SDA via their coiled-coil domains
We then moved on to determine the relationship between α-taxilin and γ-taxilin, which belong to the taxilin family and each possesses a coiled-coil domain in their middle regions (Nogami et al., 2004). First, we detected an interaction between those two proteins after immunoprecipitation in HEK-293T cells ( Figure 4A). So, we then generated α-taxilin or γ-taxilin knockout (KO) RPE-1 cells using the CRISPR-Cas9 approach (Figure 4-figure supplement 1A-F and Figure 4B-C). Compared with that of wild-type (WT) cells, α-taxilin fluorescence intensity was less at the γ-taxilin KO cell centrosomes ( Figure 4D). Conversely, α-taxilin depletion did not affect centrosomal γ-taxilin fluorescence intensity ( Figure 4D). Mapping with ectopically expressed γ-taxilin and α-taxilin truncated mutants indicated that they interact with each other via their coiled-coil domains ( Figure 4E-H). This was further confirmed via an in vitro binding assay using a purified bacteria-produced MBP-fused γ-taxilin M region (153-464 aa) and a GST-tagged α-taxilin M region (186-491 aa) ( Figure 4I). These data suggest that γ-taxilin directly recruits α-taxilin to centrosomes via its coiled-coil domain.

α-Taxilin recruits CEP170 to SDAs
The SDA marker proteins ninein and CEP170 occupy the peripheral region of the SDA structure (Huang et al., 2017;Chong et al., 2020), and based on the SDA protein localization pattern, α-taxilin is located upstream of those two proteins ( Figure 2A). So, we investigated whether α-taxilin is involved in recruiting ninein and CEP170 to the SDAs. An endogenous immunoprecipitation assay showed that α-taxilin interacted with CEP170 but not with ninein ( Figure 4A). Also, α-taxilin and CEP170 co-localized at both the SDAs and the proximal ends, as observed with STED nanoscopy ( Figure 5A). To identify which segment of α-taxilin was responsible for that association, we overexpressed truncated mutants of HA-tagged α-taxilin in HEK-293T cells and mapped the mutants' interactions with CEP170. CEP170 interacted with α-taxilin in cells overexpressing full-length α-taxilin, as well as the M (186-491 aa) region and C-terminus (491-546 aa) deletion mutant constructs (ΔC), but not with the N-terminus construct (1-185 aa) ( Figure 5B-C). This suggests that the α-taxilin M region, but not the N-or C-terminal regions, is required for its interaction with CEP170. An in vitro binding assay using GST-tagged α-taxilin-M and 3×FLAG-CEP170 purified from bacteria and HEK-293T cells ( Figure 5D), respectively, suggesting that α-taxilin may directly bind CEP170.
Next, we examined the effect of α-taxilin depletion on CEP170's centrosomal localization and detected a significant decrease in CEP170 fluorescence intensity in RPE-1 cells treated with α-taxilin siRNA, compared with control siRNA ( Figure 5E-F). Additionally, overexpressed siRNA-resistant full-length α-taxilin could rescue the CEP170 fluorescence intensity at the centrosomes, whereas the α-taxilin deletion mutant (ΔM2) lacking the region responsible for its SDA localization ( Figure 1G and I) failed ( Figure 5E-F). These results indicate that α-taxilin participates with CEP170 in SDA assembly.
showing ring size diameter. Data are Mean ± SD. n ≥ 7, box = 25th and 75th percentiles. (C) Representative two-color STED super-resolution images showing side view of the SDA proteins (green) and the centriole proximal end protein C-Nap1 (red). Scale bar, 500 nm. (D) A scatter plot describing the distance of SDA proteins relative to C-Nap1. Data are Mean ± SD. n ≥ 11, box = 25th and 75th percentiles. (E) Relative localization of SDA proteins in radial and lateral directions of the mother centriole. The upper dotted lines reveal the slanted arrangement of distal appendage (DA) and the lower dotted lines represent the triangular SDA structure, respectively. Data are Mean ± SD.
The online version of this article includes the following source data for figure 2: Source data 1. The diameter of subdistal appendage (SDA) proteins, including α-taxilin and γ-taxilin.
Source data 2. The longitudinal positions of subdistal appendage (SDA) proteins, including α-taxilin and γ-taxilin. . In WT RPE-1 cells, the side view showed that SDA stems were beside centriole wall and that they occupied a wide range at the subdistal position of the mother centriole ( Figure 6A). The TEM images showed that a relatively narrower range of SDAs with smaller sizes appeared in α-taxilin KO RPE-1 cells and the smallest in γ-taxilin KO RPE-1 cells ( Figure 6A; Figure 6-figure supplement 1A). These results suggest an indispensable role of α-taxilin and γ-taxilin in maintaining SDA integrity.

Figure 2 continued
Since SDAs serve mainly as microtubule anchoring sites at the centrosome, we then examined whether α-and γ-taxilin are involved in microtubule organization. We began with a microtubule regrowth assay using α-taxilin KO or γ-taxilin KO RPE-1 cells (Figure 4-figure supplement 1A-F and Figure 4B-C). After ice-induced microtubule depolymerization, an immunofluorescence assay detected microtubule dynamics after rewarming. RPE-1 cells were fixed at different periods of time (0, 5, and 10 min), during which microtubules formed arrays radiating from the centrosomes ( Figure 6B-E). The centrosomal microtubule aster in normal RPE-1 cells was visible 5 min after rewarming; and at 10 min, an extensive array of microtubules centering on the centrosome had formed (Figure 6, B and D). However, the microtubule asters were significantly impaired in the α-taxilin and γ-taxilin KO RPE-1 cells and the microtubule array densities were obviously less ( Figure 6, B and D). To confirm this phenomenon, we further conducted microtubule regrowth assays in control and in α-taxilin and γ-taxilin siRNA-depleted RPE-1 cells. Those results displayed compromised microtubule reformation in the experimental cells, compared with that in the control cells ( Figure 6-figure supplement 1B-E). The compromised microtubule regrowth, triggered by depletion of α-taxilin or γ-taxilin, could be partly rescued by overexpression of 3×FLAG-tagged full-length α-taxilin or γ-taxilin, but not by the M2 region ( Figure 1G-H) deletion mutants of either taxilin ( Figure 6B-E). This suggests that the SDA localization of both α-taxilin and γ-taxilin have indispensible roles in controlling microtubule anchoring at the interphase centrosome.
The online version of this article includes the following source data and figure supplement(s) for figure 3: Source data 1. Data of normalized α-taxilin band intensity in control-and ODF2-siRNA treated RPE-1 cells (Data provided as Mean ± SEM).
Source data 2. Data of normalized γ-taxilin band intensity in control-and ODF2-siRNA treated RPE-1 cells (Data provided as Mean ± SEM).
Source data 3. Full immunoblots labeled and unlabeled for Figure 3A, B and E.           The centrosome acts as an MTOC by controlling microtubule nucleation and anchoring. γ-Tubulin, as a member of the γ-TuRC complex, plays a major role in microtubule nucleation (Schatten and Sun, 2018). To determine the causes of the compromised microtubule reformation observed in both α-and γ-taxilin depleted cells, we examined γ-tubulin at the interphase centrosome in control-, α-taxilin, and γ-taxilin-siRNA treated RPE-1 cells. The γ-tubulin fluorescence intensity at the interphase centrosome remained unchanged following both α-taxilin and γ-taxilin siRNA-induced depletion ( Figure 6-figure  supplement 1F-G), thus suggesting that microtubule nucleation may not be affected.

Discussion
SDAs are conserved structures located at the subdistal end of the mother centriole, and their formation, together with that of the DAs, marks centriole maturity (Uzbekov and Alieva, 2018). They play important roles in microtubule organization and spindle arrangement and participate in various biological processes such as cell division and cell differentiation (Hall and Hehnly, 2021). Here, we applied APEX2-mediated proximity-labeling to centrosome proteomics by using CCDC68 and CCDC120 as baits (Hung et al., 2016b;Huang et al., 2017). The results show a variety of possible proteins, some of which are known centrosomal proteins like γ-tubulin and γ-taxilin, and several new centrosomelocalized proteins are also identified (Figure 1-figure supplement 1A-B), such as α-taxilin, DACT1, NCAPH2, DRG2, and SMAP2. Among these proteins, α-taxilin and γ-taxilin are both located at the SDA and the proximal end of mother centriole (Figures 1-3). However, other already known SDA proteins, such as ODF2, ninein and CEP170, are not included in the datasets, probably because of their relatively low levels in the cells (for ODF2, Figure 3A-B) or their relatively longer distances from CCDC68 and CCDC120 (for ninein and CEP170, Figure 2). Figure 4A, B, C, F and H.    . The GST-α-taxilin-M was stained with Coomassie brilliant blue (CBB) and the 3×FLAG-CEP170 was pulled down and IB using FLAG antibody. (E) Confocal images of immunostained CEP170 (red) and γ-tubulin (magenta) in control-or α-taxilin-siRNA treated RPE-1 cells, and those cells rescued with siRNA-resistant GFP-tagged α-taxilin-FL or the α-taxilin M2 deletion mutant (△261-300 aa) (green). Scale bar, 5 µm. (F) Comparisons of CEP170 fluorescence intensities at the centrosomes in (E). Statistical significance was determined with one-way ANOVA with three replicates. Data are Mean ± SEM. n > 100; *, p < 0.05; ***, p < 0.001; n.s., not significant.

Source data 3. Full immunoblots labeled and unlabeled for
The online version of this article includes the following source data for figure 5: Source data 1. Data of centrosomal CEP170 fluorescence intensity in control and α-taxilin siRNA treated RPE-1 cells, and rescued by overexpressed fulllength α-taxilin or the α-taxilin M2 deletion mutant (Data provided as Mean ± SEM). Figure 5C and D. The SDA is a comprehensive structure that consists of a spreading radial distribution of concentric proteins, and SDA assembly follows a hierarchical relationship based on protein distributions (Chong et al., 2020). Comparing SDA protein diameters, those of α-and γ-taxilin are larger than those of CCDC68 and CCDC120, but smaller than those of ninein and CEP170. In the longitudinal position, γ-taxilin is relatively higher while α-taxilin is relatively lower than the other SDA proteins. Associated with their relative localization with other SDA proteins, they are supposed to reside in the middle SDA zone ( Figure 8A-B).

Source data 2. Full immunoblots labeled and unlabeled for
Among the currently known SDA components, ODF2 is the most closely located to the centriolar microtubules ( Figure 8C) and it likely recruits TCHP to assist ninein assembly at the SDA (Ibi et al., 2011). The ODF2 1-60 aa N-terminus sequence is responsible for its association with CCDC120 (Huang et al., 2017). However, this association may not be direct, as a yeast two-hybrid assay indicated. Besides, CCDC68 lies between ODF2 and CEP170, and how CCDC68 is recruited to the SDA is yet unknown (Huang et al., 2017). Our data suggest that ODF2 recruits α-and γ-taxilin, similarly to how it recruits CCDC120 (Huang et al., 2017). Since the ODF2 ring is much smaller than those of both α-and γ-taxilin, most likely the interactions between ODF2 and the taxilins are indirect ( Figure 8A-B). Although TCHP, CCDC68, and CCDC120 lie between ODF2 and the taxilins ( Figure 8A-B), no interactions among taxilins with those SDA proteins, except for ODF2, are detected by immunoprecipitation assays. Therefore, ODF2 recruits α-taxilin and γ-taxilin through a new pathway ( Figure 8C).
The functions of SDA components are critical for understanding this structure. Ninein is responsible not only for microtubule nucleation by docking the γ-TuRC at the centrosomes (Delgehyr et al., 2005), but also for forming a microtubule-anchoring complex with CEP170 at the SDA periphery (Pizon et al., 2020). CCDC68 and CCDC120 also participate in microtubule anchoring (Huang et al., 2017). Similarly, compromised microtubule reformation following ice-depolymerization was observed in both α-taxilin and γ-taxilin depleted cells, likely because of depressed microtubule anchoring ability, as γ-tubulin intensity at the centrosome is not influenced by either α-taxilin or γ-taxilin siRNA knockdown. Therefore, our data suggest that α-taxilin and γ-taxilin serve as microtubule anchoring regulators at the centrosomes, functioning via a new pathway that is independent of CCDC68, CCDC120, and ninein.
In addition to acting as microtubule anchoring centers during interphase, SDAs also function as spindle regulators during mitosis. CEP170 interacts with centrosome-associated kinesins such as KIF2A, KIF2C, KIFC3, and spindle microtubule-associated KIF2B (Welburn and Cheeseman, 2012;Maliga et al., 2013), which regulates spindle assembly and cell morphology. In our study, both α-taxilin and γ-taxilin depletion result in decreased astral microtubule length and increased spindle misorientation.

Figure 7 continued on next page
As the hierarchical assembly of γ-taxilin, α-taxilin, and CEP170 is established, we speculate that α-taxilin and γ-taxilin regulate spindle orientation mainly by controlling CEP170's centrosomal localization. In addition to CEP170, the most upstream SDA protein, ODF2, is shown to regulate spindle orientation via microtubule organization and stability (Hung et al., 2016a). Spindle disorganization has been extensively correlated with cell differentiation, cancer and neurological diseases such as microcephaly and lissencephaly (Noatynska et al., 2012), as well as with tubular organ diseases (Zhong and Zhou, 2017). Whether α-and γ-taxilin are involved in development and diseases needs further investigation.
The online version of this article includes the following source data and figure supplement(s) for figure 7: Source data 1. Spindle angles of wild-type (WT), α-taxilin knockout (KO) HeLa cells, and cells overexpressed with indicated α-taxilin full-length and deletion mutants (Data provided as Mean ± SEM).
In conclusion, our results both show α-taxilin and γ-taxilin to be new SDA proteins and shed new light on how SDAs are assembled. However, other yet unknown components exist at the SDA upper zone and participate in SDA assembly, and they and their functions need to be identified. Additionally, how α-taxilin and γ-taxilin function in development and diseases warrants further investigation. Software, algorithm Image J NIH (Schneider et al., 2012) Software The identity of the cell lines has been authenticated by ATCC with a 100% match. Cells were validated to be negative for mycoplasma contamination. All cells were incubated at 37 °C with a 5% CO 2 atmosphere. The HeLa, U2OS, and HEK-293T cells were transfected using polyethylenimine (#23966-1, Polysciences, Inc, Warrington, PA, USA), while the RPE-1 cells were transfected using Lipofectamine 3000 (#L3000-015, Invitrogen), both according to the manufacturer's instructions.

APEX2-mediated proximal labeling of CCDC68 and CCDC120
Once CCDC68-V5-APEX2 and CCDC120-V5-APEX2 were constructed, a large-scale proteomic experiment was conducted in cultured HEK-293T cells in which both constructs were over-expressed for 24 hr. Following treatment with BP (final concentration of 50 µM) for 30 min and H 2 O 2 (final concentration of 1 mM) for 1 min, the biotinylated proteins were enriched by Streptavidin-coated beads (Streptavidin Sepharose High Performance, #17-5113-01, Cytiva, Marlborough, MA, USA) for mass spectrographic analysis.

Mass spectrometry
To identify proteins, the Coomassie-stained proteins bands of each sample were cut out of the gels and destained with a solution of 25 mM ammonium bicarbonate in 50% acetonitrile. After dithiothreitol reduction and iodoacetamide alkylation, the proteins were digested with porcine trypsin (Sequencing grade modified; Promega, Madison, WI, USA) overnight at 37 °C. The resulting tryptic peptides were extracted from the gel pieces by using 200 µl of acetonitrile (with 0.1% formic acid). The samples were dried in a vacuum centrifuge concentrator at 30 °C and re-suspended in 10 µl of 0.1% formic acid. The peptides, resolved using 0.1% formic acid, were loaded into a trap column (Acclaim PepMap 100 75 µm × 2 cm nanoViper, C 18 , 3 µm, 100 Å, Thermo Fisher Scientific) and connected to an analytical column (Acclaim PepMap RSLC 75 µm × 15 cm nanoViper, C 18 , 2 µm, 100 Å, Thermo Fisher Scientific) on a nanoflow HPLC Easy-nLC 1200 system (Thermo Fisher Scientific) using a 75 min LC gradient at 280 nl/min. Buffer A consisted of 0.1% (v/v) formic acid in H 2 O and Buffer B consisted of 0.1% (v/v) formic acid in 80% acetonitrile. The gradient was set as follows: 4-8% B in 5 min, 8-20% B in 45 min, 20-30% B in 10 min, 30-90% B in 13 min, 90% B in 2 min.
Proteomic analyses were performed on a Thermo Orbitrap Fusion Lumos mass spectrometer (Thermo Fisher Scientific) using a nano-electrospray ion source with electrospray voltages of 2.2 kV. Xcalibur software in profile spectrum data type was used for data-dependent acquisition. The MS 1 full scan was set at a resolution of 60,000 at m/z 200, AGC target 5e4 and maximum IT 50 ms by an Orbitrap mass analyzer (300-1,500 m/z), and that was followed by MS 2 scans generated by HCD fragmentation at a resolution of 15,000 at m/z 200, AGC target 5e4 and maximum IT 45 ms. The fixed first mass of the MS 2 spectrum, the isolation window, and the normalized collision energy (NCE) were set at 110.0 m/z, 1.6 m/z, and NCE 30%, respectively.

Mass spectrometry data analysis
Proteome Discoverer 2.1 software (Thermo Fisher Scientific) was used to align the mass spectrometry data with Uniprot Homo sapiens (Human). Enzyme specificity was set to trypsin, with a maximum of 2 missed trypsin cleavage sites. Precursor mass tolerance was set to 10 ppm and the fragment ion mass tolerance to 0.02 Da. Also, carbamidomethylation of cysteine was the fixed modification, while oxidation of methionine and protein N-terminal acetylation were variable modifications. Percolator algorithm was used to calculate a 1% false discovery rate at the peptide and protein levels.

Electron microscopy
RPE-1 cells grown on ACLAR 33 C film were fixed by adding double-strength fixative (i.e. 1× fixative: 2.5% glutaraldehyde in 0.1 M PB buffer [pH 7.4]) with an equal volume of culture medium for 3 min at room temperature (RT). After removing the mixed fixative, cells were fixed with 1× fixative for 1 hr at RT, then held overnight at 4 °C. Then, the cells were washed four times (8 min each) with 0.1 M PB buffer, then post-fixed in 1% OsO 4 and 0.8% K₄Fe(CN)₆ for 1 hr at RT in the dark. After rinsing four times (8 min each) with distilled water, the cells were stained in 1% aqueous uranyl acetate overnight at 4 °C. Following several washes with distilled water, the cells were dehydrated by first using a graded alcohol series (30%, 50%, 70%, 85%, 95%, and 100%, 5 min for each) and then 100% acetone (twice at 5 min each). Subsequently, the cells were infiltrated, embedded in EMbed 812 resin (Electron Microscopy Sciences, Hatfield, PA, USA), and polymerized at 65 °C for 24 hr. After removing the ACLAR 33C film, the resin blocks were trimmed and sectioned using an EM UC7 ultramicrotome (Leica Microsystem, Wetzlar, Germany) with an ultra 35° diamond knife (Diatome Ltd, Nidau, Switzerland). Single-slot copper grids were used to collect 70-nm-thick serial sections that were double stained with uranyl acetate and lead citrate. The grids were then inspected using a Tecnai G 2 Spirit BioTWIN transmission electron microscope (Thermo Fisher Scientific) at 120 kV, and images were captured with an attached Orius 832 CCD camera (Gatan, Inc, Pleasanton, CA, USA).

Immunofluorescence
Cells grown on 18 × 18 mm coverslips were fixed and permeabilized in pre-chilled methanol for 10 min at -20 °C, and then washed three times (10 min each time) in PBS. The cells were then blocked with 4% BSA for 30 min at RT, and then incubated with the first antibody (diluted in 4% BSA) at 4 °C overnight. The slides were then washed three times in PBS buffer for 10 min each time, blocked with 4% BSA for 30 min at RT, and then incubated with the second antibody (Alexa Fluor [Invitrogen, Carlsbad, CA, USA]) for 1 hr at RT. Finally, the cells were stained with DAPI.
Confocal and STED images were acquired using either a TCS SP8 STED 3 X microscopy system with a 100 × 1.4 NA APO oil objective lens and LAS X v.2.0 software (Leica Microsystem), or a STEDYCON microscope platform (Abberior Instruments, Göttingen, Germany). Huygens v.14.10 software (Scientific Volume Imaging, Hilversum, Netherlands) was used for STED nanoscopy image deconvolution. An N-SIM Microscope system equipped with a 100 × 1.49 NA APO oil objective lens (Nikon, Tokyo, Japan) was used to acquire 3D-SIM images. NIS-Elements AR v.4.51 software (Nikon) was used both to acquire the images and for three-dimensional reconstruction. Images were then processed in Photoshop (Adobe, San Jose, CA, USA).

Immunoprecipitation and immunoblots
For immunoprecipitation, HEK-293T or RPE-1 cells were lysed in immunoprecipitation lysis buffer (50 mM HEPES, 250 mM NaCl, 0.1% Nonidet P-40, 1 mM DTT, 1 mM EDTA, and 10% glycerol, pH = 7.4) on ice for 30 min and then the lysates were centrifuged at 12,000 g for 15 min at 4 °C. Then, either Protein A-Sepharose beads or Protein G Sepharose beads (ab193259 and ab193256, respectively, Abcam, Cambridge, United Kingdom Amersham Biosciences) were added to the supernatants, which were then individually incubated with the appropriate antibodies overnight at 4 °C. The beads were then washed with immunoprecipitation lysis buffer and collected in SDS loading buffer (50 mM Tris-HCl, pH = 7.4, 2% SDS, 100 mM DTT, 0.025% bromo blue, 10% glycerol), which was then boiled at 100 °C for 10 min to obtain the immunoprecipitation samples.
For immunoblots, SDS-PAGE was used to separate protein samples, which were then transferred onto polyvinylidene difluoride membranes (Sigma, Burlington, MA, USA). The membranes were blocked with 5% non-fat milk for 30 min, and then incubated with primary antibodies either overnight at 4 °C or for 2 hr at room temperature. After incubation, the membranes were then incubated with peroxidase-AffiniPure goat anti-rabbit or goat anti-mouse IgG (H + L) secondary antibodies (1:5000) (#111-035-003 and #115-035-003, respectively, Jackson ImmunoResearch, WestGrove, PA, USA) for 1 hr at RT.

Microtubule regrowth assay
RPE-1 cells were grown on 18 × 18 mm coverslips and then the coverslips were embedded on ice for 30 min to depolymerize the cytoplasmic microtubules. The cells were then brought back to 37 °C to allow the microtubules to reform. Cells at 0, 5, and 10 min after rewarming were fixed in 4% paraformaldehyde (pre-warmed to 37 °C) for 10 min at 37 °C. The cells were immunostained with α-and γ-tubulin antibodies.

Spindle orientation analysis
Spindle orientation was measured from the centrosome pairs' x, y, and z coordinates relative to the slide. Spindle angles measurements were derived from the fixed and immunostained HeLa cells, which were immunostained with an anti-γ-tubulin antibody for spindle poles (red) and an anti-α-tubulin antibody for mitotic spindles (green). The z-stacks and the spindle pole coordinates were measured usinge LAS X v.2.0 software (Leica Microsystem). The spindle angles were then measured using 3D coordinate geometry.

Measurements and statistical analysis
Immunofluorescence intensities and the diameters of the ring-like SDA protein structures were measured using Image J software v.1.48 (NIH) (Schneider et al., 2012). The statistical significances among different groups were determined by two-tailed Student's t-tests or one-way ANOVAs. The data were graphed in Prism 5 (GraphPad, San Diego, CA, USA).