TEDC1 and TEDC2 have been reported to physically interact with delta-tubulin and epsilon-tubulin, and loss of either Tedc1 or Tedc2 in 3T3 cells results in cells with a variable number of centrioles through the cell cycle (Breslow et al., 2018). To further dissect the phenotypes of loss of TEDC1 or TEDC2 and directly compare to our original report on delta-tubulin and epsilon-tubulin, we used CRISPR/Cas9 to generate strong loss of function/null mutations in TEDC1 or TEDC2 in the same cell type and background genotype (hTERT RPE-1 TP53^-/-, which will be referred to as RPE-1 p53^-/-) as the TUBD1^-/- (delta-tubulin knockout) and TUBE1^-/- (epsilon-tubulin knockout) mutant cells (Figure 1—figure supplement 1). By immunofluorescence staining for two centriolar proteins, centrin (CETN) and CP110, we observed that TEDC1^-/- and TEDC2^-/- mutant cells had similar phenotypes to each other and to TUBD1^-/- and TUBE1^-/- mutant cells: in an asynchronously growing culture, about half of the cells had no centrioles, and half had five or more centrioles. These phenotypes were fully rescued by expression of tagged TEDC1 (TEDC1-Halotag-3xFlag) or TEDC2 (TEDC2-V5-APEX2; Figure 1, Figure 1—figure supplement 1).

Next, we checked whether the centrioles in TEDC1^-/- and TEDC2^-/- mutant cells underwent a futile cycle of centriole formation and disintegration. We synchronized cells in each stage of the cell cycle, quantified the number of cells with centrioles, and found that almost all mutant cells lacked centrioles in G0/G1 phase. Centrioles formed in S-phase and disintegrated in M (Figure 1B). The centrioles that were present in mutant cells were immature: all centrioles were positive for the procentriole marker SASS6 and negative for the mature centriole marker CEP164 (Figure 1C and D). We conclude that cells lacking TEDC1 or TEDC2 also undergo a futile cycle, similar to cells lacking delta-tubulin or epsilon-tubulin (Figure 1G).

We also examined the centriolar microtubule status of TEDC1^-/- and TEDC2^-/- mutant cells by TEM. Similar to cells lacking delta-tubulin or epsilon-tubulin, we found that centrioles in TEDC1^-/- and TEDC2^-/- mutant cells lacked compound microtubules and only had singlet microtubules. These centrioles had cartwheels and pinheads, but A-C linkers were not visible (Figure 1E and F, Figure 1—figure supplement 2). Together, these results demonstrate that loss of TEDC1 or TEDC2 phenocopies loss of delta-tubulin or epsilon-tubulin, indicating that these proteins likely act together.

Next, we investigated the localization of TEDC1 and TEDC2 to determine if they may directly act on centrosomes. TEDC1 and TEDC2 are expressed at low levels in cells (Figure 1—figure supplement 1), and we could not reproducibly localize the endogenous proteins with antibody staining. Instead, we localized the functional, tagged proteins in our rescue cell lines. We found that the tagged rescue constructs localize to centrosomes (Figure 2A and B) and the antibodies for the tags were specific (Figure 2—figure supplement 1E-J). TEDC1 and TEDC2 were enriched at centrosomes in S/G2 and colocalized with SASS6, but not centrin, indicating that TEDC1 and TEDC2 may localize to newly formed procentrioles and/or the proximal ends of parental centrioles.

To analyze TEDC1 and TEDC2 localization at higher resolution, we localized our tagged rescue constructs using three methods: a super-resolution spinning disk confocal microscope with immunofluorescence microscopy (Figure 2—figure supplement 1A, B), ultrastructure expansion microscopy (U-ExM, (Gambarotto et al., 2019), Figure 2C and D), and a second expansion microscopy method (Kong et al., 2024, Figure 2—figure supplement 1C, D). With all three methods, we observed that both proteins localize to procentrioles and the proximal ends of parental centrioles. At these regions, both proteins overlap with the centriolar microtubules. Together, these results show that TEDC1 and TEDC2 localize to centrosomes and likely directly act upon them.

To determine how TEDC1, TEDC2, TUBD1 and TUBE1 might act together, we first determined whether they are mutually required for their localization at centrosomes. We found that TEDC1 did not localize to centrioles in the absence of TEDC2, TUBD1, or TUBE1 (Figure 3A). Likewise, TEDC2 did not localize to centrioles in the absence of TEDC1, TUBD1, or TUBE1 (Figure 3B). These results indicate that these proteins are mutually required for TEDC1 or TEDC2 localization. Furthermore, overexpression of TEDC1 or TEDC2 did not rescue the centriole phenotypes in any of the other mutants, indicating that TEDC1 and TEDC2 are not downstream effectors of TUBD1 and TUBE1 (Figure 3A and B).

TEDC1 and TEDC2 have previously been shown to physically interact with TUBD1 and TUBE1 (Breslow et al., 2018). To further probe the nature of this interaction, we first determined whether any of these proteins may form subcomplexes in cells. We expressed TEDC1-Halotag-3xFlag in each mutant cell line and determined whether immunoprecipitation of tagged TEDC1 could precipitate the other proteins. TEDC1-Halotag-3xFlag rescuing the TEDC1^-/- mutant could precipitate TEDC2, TUBD1, and TUBE1, indicating that all four proteins physically interact. TEDC1 did not interact with epsilon-tubulin in the absence of delta-tubulin, nor did it interact with delta-tubulin in the absence of TUBE1. In the absence of TEDC2, TEDC1 did not interact with TUBD1 or TUBE1. However, in the absence of TUBD1 or TUBE1, TEDC1 and TEDC2 could still interact with each other (Figure 3C).

We performed the reciprocal experiment, in which we expressed TEDC2-V5-APEX2 in each mutant cell line and determined whether immunoprecipitation of tagged TEDC2 could precipitate the other proteins. We observed similar results as our analysis with TEDC1. TEDC2-V5-APEX2 rescuing the TEDC2^-/- mutant could precipitate TEDC1, TUBD1, and TUBE1, indicating that all four proteins physically interact. TEDC2 did not interact with either tubulin in the absence of the other. In the absence of TEDC1, TEDC2 did not interact with either tubulin. However, in the absence of TUBD1 or TUBE1, TEDC2 and TEDC1 could still interact (Figure 3D).

Together, these experiments indicate that TEDC1, TEDC2, TUBD1, and TUBE1 physically interact with each other, as previously reported (Breslow et al., 2018; Huttlin et al., 2017; Huttlin et al., 2021). Furthermore, TEDC1 and TEDC2 can form a subcomplex in the absence of either tubulin.

To gain additional insight into the nature of this interaction, we used AlphaFold-Multimer (Evans et al., 2021) to predict the structure of the complex. AlphaFold-Multimer predicted that TUBD1 and TUBE1 would form a heterodimer, similar to the alpha-tubulin/beta-tubulin heterodimer, with TUBD1 at the minus-end of the heterodimer. AlphaFold also predicted that the alpha-helices of TEDC1 and TEDC2 interact with each other, and that TEDC1 and TEDC2 form an interaction surface with TUBD1. These predictions, especially at the interface between TEDC1, TEDC2, and TUBD1, yielded high confidence pLDDT and PAE scores (Figure 3E-G, Figure 3—figure supplement 1A). A similar prediction was obtained with the newly released AlphaFold 3 (Abramson et al., 2024; Figure 3—figure supplement 1B). As controls, we used AlphaFold-Multimer to predict whether TEDC1 and TEDC2 might interact with alpha-tubulin and beta-tubulin, and whether similar structures would be predicted for Xenopus TEDC1, TEDC2, TUBD1, and TUBE1. While AlphaFold-Multimer did not predict a high-confidence interaction for TEDC1, TEDC2, alpha- and beta-tubulin (Figure 3—figure supplement 1C), it did predict a high-confidence structure for Xenopus TEDC1, TEDC2, TUBD1, and TUBE1, similar to that predicted for the human proteins (Figure 3—figure supplement 1D).

Our pulldown experiments showed that TEDC1 and TEDC2 can interact in a subcomplex in the absence of TUBD1 or TUBE1, which supports the predicted structural model, in which TEDC1 and TEDC2 are predicted to directly interact with each other without being bridged by either tubulin. Further supporting this model, immunoprecipitation of TEDC2 identifies the other proteins in stoichiometric amounts (Breslow et al., 2018), and we previously showed that TUBD1 and TUBE1 physically interact (Wang et al., 2017). Given the size and shape of the tetrameric complex as predicted by AlphaFold-Multimer, it is possible that these may form a structural component of centrioles. Future work will be necessary to test these possibilities. Together, our experiments indicate that TEDC1, TEDC2, TUBD1, and TUBE1 physically interact in a complex and are recruited together to centrioles.

Next, we determined how the loss of these proteins, and the triplet microtubules themselves, affect centriole ultrastructure and protein composition. Because centrioles are constitutively formed de novo every cell cycle in our mutant cells, we incorporated two controls in our analysis: procentrioles undergoing normal parental-mediated centriole duplication in control (RPE-1 p53^-/-) cells, and centrioles formed in RPE-1 p53^-/- cells de novo in the first cell cycle after centrinone washout (Wong et al., 2015). For each of the two control and four mutant cell lines, cells were synchronized by mitotic shake off, resulting in coverslips enriched for cells in late S- and G2-phases, with a minor population in M-phase. Synchronized cells were then expanded using U-ExM and stained for centriolar markers.

We first tested whether the microtubules of mutant centrioles could be modified by acetylation of alpha-tubulin. During centriole formation, acetylation is thought to proceed from the proximal toward the distal end and from the A- to the C-tubules (Sahabandu et al., 2019). We found that antibodies against acetylated alpha-tubulin stained mutant centrioles well (Figure 4B), indicating that centrioles with only singlet A-tubules can be acetylated.

We next tested whether mutant centrioles were capable of elongating during the cell cycle. In our expansion gels of cells enriched in late S and G2 phases, we used PCNA to mark S-phase cells and co-stained with acetylated tubulin to mark centrioles. Similar to a recently published report, we also found a range of centriole lengths in S- and G2-phases (Laporte et al., 2024). In S-phase, centrioles were short in all conditions. In G2-phase, centrioles elongated in all conditions, and mutant centrioles reached approximately similar lengths as control centrioles (Figure 4A). By contrast, mutant centriole widths did not increase and centrioles remained narrow, as we previously reported (Figure 4—figure supplement 5 and Wang et al., 2017). These results indicate that centrioles with singlet microtubules can elongate to the same overall length as control centrioles in G2 phase. Consistent with this hypothesis, CEP120, a protein involved in regulating centriole length (Comartin et al., 2013; Lin et al., 2013b; Mahjoub et al., 2010), was present and properly localized within mutant centrioles (Figure 4—figure supplement 1D).

The compound microtubules of centrioles are heavily post-translationally modified, and recent studies have indicated that each tubule may acquire different modifications (Guichard et al., 2023). We checked glutamylation, a post-translational modification thought to be restricted to the outer surface of centrioles (Guichard et al., 2023). Within Chlamydomonas centrioles, glutamylation is differentially distributed between each tubule: on the C-tubule at the distal end, on all three tubules in the central core, and on the A-tubule at the proximal end (Hamel et al., 2017). In human centrioles, polyglutamylation is enriched in the proximal and central regions, and is absent in the distal region (Gambarotto et al., 2019; Mahecic et al., 2020; Sullenberger et al., 2020). We used two antibodies to detect glutamylation: the GT335 antibody, which recognizes the glutamylation branch and thus detects all polyglutamylation, and the polyE antibody, which recognizes long polyglutamate side chains with at least 2 or 3 glutamate residues (Kann et al., 2003; van Dijk et al., 2007). We found that mutant and control centrioles could be stained by GT335 (Figure 4C), indicating that mutant centrioles are at least mono-glutamylated. However, the polyE antibody did not label control procentrioles or de novo centrioles in the first cell cycle after their formation, making this antibody uninformative for our mutants (Figure 4D). These results show that centrioles with just singlet microtubules (A-tubules) can be mono-glutamylated. Moreover, similar to previous reports (Sullenberger et al., 2020), our results suggest that centriole glutamylation is a multi-step process, in which long glutamate side chains are added later during centriole maturation.

We previously demonstrated that TUBD1^-/- and TUBE1^-/- mutant centrioles fail to recruit the distal centriole protein POC5 (Wang et al., 2017). Using expansion microscopy, we found that TEDC1^-/- and TEDC2^-/- mutant centrioles also failed to recruit POC5 (Figure 4Ei). Since our original work was published, POC5 was shown to be a component of the helical inner scaffold within the central core. These results indicate that the helical inner scaffold is not properly formed in centrioles with singlet microtubules. To test the mechanisms underlying loss of POC5, we next tested whether mutant centrioles recruit WDR90, which has been proposed to localize to the inner junction between the A- and B-tubules and function in recruiting the inner scaffold (Steib et al., 2020). We found that WDR90 was not recruited to mutant centrioles, in contrast to control centrioles, in which it is recruited in G2-phase (Figure 4Eii). From these results, it is likely that mutant centrioles with singlet microtubules fail to build or stabilize the inner junction between the A- and B-tubules. In the absence of the inner junction and junctional protein WDR90, centrioles with singlet microtubules cannot form the inner scaffold. As also previously reported (Laporte et al., 2024), we failed to detect gamma-tubulin within the lumen of control or de novo-formed centrioles in S or G2-phase (Figure 4—figure supplement 1E) and thus were unable to test whether gamma-tubulin, which is recruited to the lumen of centrioles by the inner scaffold, was mislocalized in mutant centrioles.

Next, we tested whether the centriole proximal end might be properly formed in mutant centrioles. We found that the centriolar cartwheel protein, SASS6, was present within the lumen of control and mutant centrioles in S-phase. In control centrioles in G2-phase, SASS6 was restricted to just the proximal end. Surprisingly, SASS6 was elongated in all G2-phase mutant centrioles (Figure 4F, Figure 4—figure supplement 4F). We observed a similar phenotype with multiple other proximal-end proteins: CEP135, STIL, CPAP, and CEP44 (Figure 4G-I, Figure 4—figure supplement 1, Figure 4—figure supplement 3), indicating that the entire proximal end is elongated in mutant centrioles. The extended localization of proximal end proteins was not due to increased protein expression in mutant cells (Figure 4—figure supplement 2). We conclude that loss of TEDC1, TEDC2, TUBD1, or TUBE1 results in elongated proximal end domains within mutant centrioles.

Elongation of the proximal end of centrioles may also indicate an overall defect in centriole polarity. To test this hypothesis, we next determined whether these mutant centrioles might properly recruit proteins to their distal ends. We found that CETN2 and CP110, two proteins of the distal centriole, were localized to mutant centrioles and clearly marked one end of the centriole barrel in both S-phase and G2-phase (Figure 4—figure supplement 1B, C). We conclude that proximal-to-distal centriole polarity was unaffected in mutant centrioles, and proximal end elongation did not affect the recruitment of proteins to the centriole distal end. Together, these results indicate that centrioles lacking compound microtubules are unable to properly regulate the length of the proximal end.

Centrioles lacking triplet microtubules undergo a futile cycle of formation and disassembly, but the mechanisms underlying disassembly are not well-understood. We first tested whether centriole loss in mutant centrioles may be due to loss of CEP295. CEP295 promotes centriole-to-centrosome conversion, a process in which pericentriolar material is recruited to newly-formed procentrioles. Cells lacking CEP295 form centrioles that disintegrate during the cell cycle due to a failure to undergo centriole-to-centrosome conversion (Izquierdo et al., 2014). Using U-ExM, we found that CEP295 was present and normally localized within mutant centrioles in both S- and G2-phases (Figure 4—figure supplement 1F). We conclude that centriole loss in our mutants is unlikely to be due to loss of CEP295 localization, and therefore that TEDC1, TEDC2, TUBD1, and TUBE1 are likely part of a different pathway required for centriole stability through the cell cycle.

Next, we used U-ExM to visualize centriole loss during mitosis. We stained for the centriole wall (GT335), the centriole proximal end (SASS6) and the centriole distal end (CP110). In control cells, in which centrioles formed de novo after centrinone washout, multiple centrioles could be seen throughout mitosis, and SASS6 was lost from centrioles in anaphase-stage cells (Figure 5A and B). By contrast, in prometaphase stage TUBD1^-/- or TUBE1^-/- cells, we found that centrioles had a unique appearance: they were longer than normal, with an elongated proximal end marked by SASS6, and a CP110-positive cap. These two ends were connected by weak monoE staining (Figure 5C and E). This phenotype is identical to our observations of centrioles in a prometaphase TUBE1^-/- cell by TEM in our previous publication (Figure 2B in Wang et al., 2017). After metaphase, centrioles in mutant cells were either completely absent, or had a fragmented appearance (Figure 5D and F), with aggregates of staining that did not resemble true centrioles. We conclude that in our mutant cells, centrioles elongate in early mitosis to form an aberrant intermediate structure, followed by fragmentation in late mitosis.

Human hTERT RPE-1 TP53^−/− cells were a gift from Meng-Fu Bryan Tsou (Memorial Sloan Kettering Cancer Center) and were cultured in DMEM/F-12 (Corning) supplemented with 10% Cosmic Calf Serum (CCS; HyClone). Human HEK293T cells for lentivirus production (see below) were obtained from the ATCC and cultured in DMEM (Corning) supplemented with 10% CCS. hTERT RPE-1 and HEK293T/17 cells were authenticated using STR profiling using CODIS loci. All other cell lines used were derived from hTERT RPE-1 TP53^−/− cells. Stable TP53^−/−; TEDC1^−/− and TP53^−/−; TEDC2^−/− knockout cell lines were made in the hTERT RPE-1 TP53^−/− cells by CRISPR/Cas9 (see below). For rescue experiments, clonal knockout cell lines were rescued using lentiviral transduction (see below). All cells were cultured at 37 °C under 5% CO₂, and are mycoplasma-free (Uphoff and Drexler, 2011).

TEDC1^-/- and TEDC2^-/- cells were generated by CRISPR/Cas9 mediated gene editing using a recombinantly produced, purified Cas9 protein (Cas9-NLS, QB3 Macrolab, Berkeley) and chemically synthetized two-component gRNA (crRNA:tracrRNA, Alt-R CRISPR-Cas9 system, IDT). For increased efficiency, two gRNAs, both targeting the 5’ end of each gene, were used at the same time. Target sequences were: 5’-CGCCAAGTTCGACCGTCCGG-3’ and 5’-CGTCCAATCACCGCACGGGC-3’ for TEDC1, and 5’-CGCACAGCGACAATTGCAAT-3’ and 5’-CACCGGCGCGAGCAGCCCGC-3’ for TEDC2.

Lyophilized RNA oligos were reconstituted according to the instructions provided by the manufacturer (IDT). Briefly, oligos were reconstituted in the duplex buffer at a concentration of 200 µM. To anneal crRNA with tracrRNA, 3 µl of each (600 pmol) were mixed, heated to 95 °C, and transferred to room temperature to gradually cool. Pre-complexed crRNA and tracrRNA (550 pmol) were mixed with purified Cas9 (360 pmol), diluted with PBS to a total volume of 25 µl and incubated for 15 min at room temperature to form ribonucloprotein complexes (RNPs).

RPE1 TP53^-/- cells stably expressing GFP-centrin (Wang et al., 2017) were electroporated in a home-made electroporation buffer (Zhang et al., 2014) using Amaxa Nucleofector II (Lonza). Cells were electroporated with an equal mix of two RNPs: 50 µl of RNPs mixture was added to 2x10⁶ cells in 200 µl electroporation buffer. To facilitate the identification of electroporated cells, an mRuby2 expressing plasmid (pcDNA3-mRuby2, plasmid pTS3994) was electroporated together with RNPs.

Two days after electroporation, cells expressing mRuby2 were sorted using FACS, and single cells were plated into 96-well plates in conditioned media. Surviving clones were genotyped by PCR of genomic DNA and screened for phenotype based on centrin-GFP expression.

Primers used for genotyping were: 5’CCCTGCCGACGCAGTGATTGG3’ and 5’CAGGGAGTGGCGAGAGCACAC3’ for TEDC1 and 5’CTTGCCCGCAAGGAGGGAGAGA3’ and 5’GCAGGGCCCAGCCCAAACAGA3’ for TEDC2.

To rescue the mutations, Halotag-3xFlag-tagged TEDC1 or APEX-V5-tagged TEDC2 were introduced into the mutant cells using lentiviral transduction as described below.

Recombinant lentiviruses were made by cotransfection of HEK293T cells with the respective transfer vectors (TEDC1-Halotag-3xFlag and TEDC2-V5-APEX2), second-generation lentiviral cassettes (packaging vector psPAX2, pTS3312 and envelope vector pMD2.G, pTS3313) using calcium phosphate-mediated transfection. Briefly, transfection mixture was made with CaCl2, 2 x HBS (50 mM Hepes, 10 mM KCl, 12 mM dextrose, 280 mM NaCl, 1.5 mM Na2HPO4x7H2O, pH 7.05), and plasmids. Cells were treated with 25 µM chloroquine immediately before transfection, then the transfection mixture was added to cells. The medium was changed 5–6 hr after transfection, and viral supernatant was harvested after an additional 48 and 72 hr. Recipient cells (RPE-1 TP53^−/−; TEDC1^−/− and TP53^−/−; TEDC2^−/− and TP53^−/−; TUBD1^−/− and TP53^−/−; TUBE1^−/−) were transduced with viral supernatant and 8 µg/mL Sequabrene. Transduced cells were expanded to 10 cm dishes.

Cells were grown on poly-L-lysine-coated #1.5 glass coverslips (Electron Microscopy Sciences). Cells were fixed with −20 °C methanol for 15 min. Coverslips were then washed with PBS for 10 min and blocked with PBS-BT (3% BSA, 0.1% Triton X-100, 0.02% sodium azide in PBS) for 30 min. Coverslips were incubated with primary antibodies diluted in PBS-BT for 1 hr, washed with PBS-BT, incubated with secondary antibodies and DAPI diluted in PBS-BT for 1 hr, then washed again. Samples were mounted using Mowiol (Polysciences) in glycerol containing 1,4,-diazobicycli-[2.2.2]octane (DABCO, Sigma-Aldrich) antifade.

For cell cycle analyses in Figure 1, cells were seeded onto coverslips, then synchronized in G0/G1 by serum withdrawal for 24 hr, or in G2 with 10 µM RO-3306 (Adipogen) for 24 hr. Cells were fixed for immunofluorescence and analyzed for centrin/CP110 presence. Three biological replicates were performed.

For Figures 4 and 5, mitotic shakeoff was performed on asynchronously growing cells. One pre-shake was performed to improve synchronization. Cells were fixed for U-ExM and expanded as below.

For measuring overall centriole width or length, z-stacks of U-ExM images were measured using ImageJ (FIJI) on maximum projections. Only centrioles that were in perfect longitudinal or cross-section were measured. Three measurements were made per centriole and averaged. Measurements were adjusted for gel expansion factor. Statistical analysis was performed with Graphpad Prism.

For measuring protein position as in Figure 4—figure supplement 3 and Figure 4—figure supplement 4, maximum projections of U-ExM images of longitudinally positioned centrioles were measured using ImageJ (FIJI). The coordinates of the proximal-most and distal-most position for each protein were recorded. Three measurements were made per centriole and averaged. The recorded coordinates were used to calculate the positions of the most proximal and most distal signal for each protein, then graphed from shortest to longest centriole.

Welch’s t-test was chosen for statistical analysis in Figure 4A, which is an unpaired t test that does not assume that the two datasets have the same variance.

For ultrastructural analysis of centrosomes by TEM, RPE-1 TP53^-/-; TEDC1^-/- and RPE-1 TP53^-/-; TEDC2^-/- cells were synchronized in G2/M with 10 µM RO-3306 for 24 hrs. Cells were trypsinized, resuspended in complete media and centrifuged at 800 × g for 5 min. The pellet was collected in a 14 mL tube and fixed in 2% paraformaldehyde/2.5% glutaraldehyde (Ted Pella Inc, Redding, CA) in 100 mM cacodylate buffer, pH 7.2 for 2 hr at room temperature. Samples were washed in cacodylate buffer and postfixed in 1% osmium tetroxide (Ted Pella Inc)/1.5% potassium ferricyanide (Sigma, St. Louis, MO) for 1 hr. Samples were then rinsed extensively in dH₂O prior to en bloc staining with 1% aqueous uranyl acetate (Ted Pella Inc) for 1 hr. Following several rinses in dH₂O, samples were dehydrated in a graded series of ethanol and embedded in Eponate 12 resin (Ted Pella Inc). Ultrathin sections of 95 nm were cut with a Leica Ultracut UCT ultramicrotome (Leica Microsystems Inc, Bannockburn, IL), stained with uranyl acetate and lead citrate, and viewed on a JEOL 1200 EX transmission electron microscope (JEOL USA Inc, Peabody, MA) equipped with an AMT 8 megapixel digital camera and AMT Image Capture Engine V602 software (Advanced Microscopy Techniques, Woburn, MA).

Symmetrization of TEM images was performed with centrioleJ.

Cells stably expressing TEDC1-Halotag-3xFlag or TEDC2-V5-APEX2 were lysed in 50 mM Tris pH7.5, 150 mM NaCl, 1% Triton X-100, 1 mM DTT, Halt protease and phosphatase inhibitor cocktail (ThermoFisher Scientific) for 30 min on ice, then cleared by centrifugation at 21,000 × g for 20 min. Protein concentration was determined by Pierce BCA Protein Assay - Reducing Agent Compatible (ThermoFisher Scientific). Each cell lysate was incubated with 25 µL of equilibrated Chromotek Halo-Trap Magnetic Agarose (Proteintech) or Chromotek V5-Trap Magnetic Agarose (Proteintech) for 1 hr at 4 °C on a nutator. Beads were washed using a magnetic separator rack. Elution was performed by adding 80 µL of 2 x SDS loading buffer (100 mM Tris pH 6.8, 4% SDS, 20% glycerol, 100 mM DTT), boiling the beads for 5 min at 95 °C, then separating the eluate with a magnetic separator rack. Samples were loaded on SDS-PAGE and transferred for western blotting. Three biological replicates were performed.

For Figure 4—figure supplement 2, samples were lysed in 50 mM Tris pH7.5, 150 mM NaCl, 1% Triton X-100, 1 mM DTT, Halt protease and phosphatase inhibitor cocktail (Thermo Fisher Scientific) for 30 min on ice, then cleared by centrifugation at 21,000 × g for 20 min. Protein concentration was determined by Pierce BCA Protein Assay - Reducing Agent Compatible (Thermo Fisher Scientific). Equal amounts of protein (20–40 µg) were loaded per lane. For Figure 3, samples were loaded after pulldowns.

Proteins were separated by SDS-PAGE and transferred to nitrocellulose (LiCOR Biosciences) in transfer buffer (192 mM Glycine, 25 mM Tris, 20% ethanol). Membranes were blocked with 5% milk in TBST (137 mM NaCl, 25 mM Tris, 2.7 mM KCl, 0.1% Tween-20) at room temp for 1 h, then washed three times with TBST for 5 min each wash. Membranes were incubated with primary antibodies overnight at 4°C on a nutator. The next day, membranes were washed three times with TBST for 5 min each wash and incubated with secondary antibodies at room temperature for 2.5 hr. Membranes were washed again with TBST for 5 min each wash and then imaged with the LiCOR Odyssey XF imager and analyzed using Image Studio (LiCOR Biosciences). Three biological replicates were performed.

Primary antibodies used for immunofluorescence and U-ExM and dilutions in PBS-BT: mouse IgG2b anti-acetylated-tubulin, clone 6-11B-1 (1:1000,Sigma-Aldrich Cat# T6793, RRID:AB_477585), rabbit anti-acetyl-α-tubulin (Lys40) (1:100, Cell Signaling Technology Cat# 5335, RRID:AB_10544694), mouse IgG2b anti-centrin3, clone 3e6 (1:1000, Novus Biological, RRID:AB_537701), mouse IgG2a anti-centrin, clone 20H5 (IF 1:200, UExM 1:500, EMD Millipore, RRID:AB_10563501), rat anti-Cep120 (1:1000, gift from Moe Mahjoub Betleja et al., 2018), rabbit anti-Cep135 (1:500, Proteintech Cat# 24428–1-AP, RRID:AB_2879543), rabbit anti-Cep295 (1:1000, Sigma-Aldrich Cat# HPA038596, RRID:AB_10672720), rabbit anti-Cep44 (1:100, Proteintech Cat# 24457–1-AP, RRID:AB_2879557), rabbit anti-CENPJ (1:500, Proteintech Cat# 11517–1-AP, RRID:AB_2244605), rabbit anti-CP110 (IF 1:200, UExM 1:2000, Proteintech Cat# 12780–1-AP, RRID:AB_10638480), mouse IgG1 anti-Flag, clone M2 (1:500, Sigma-Aldrich Cat# F1804, RRID:AB_262044), mouse IgG1 anti-gamma-tubulin, clone GTU-88 (IF 1:1000, UExM 1:500, Sigma-Aldrich, RRID:AB_477584), mouse IgG2a anti-PCNA (1:500, BioLegend, RRID:AB_314692), rabbit anti-POC5 (for IF: 1:500, Bethyl Laboratories, RRID:AB_10949152), rabbit anti-POC5 (for U-ExM: 1:500, Thermo Fisher Scientific Cat# A303-341A (also A303-341A-T), RRID:AB_10971172), mouse IgG1 anti-polyglutamylation, clone GT335 (1:500, AdipoGen Cat# AG-20B-0020, RRID:AB_2490210), rabbit anti-polyglutamate-chain, polyE (1:500, AdipoGen Cat# AG-25B-0030, RRID:AB_2490540), mouse IgG2b anti-SASS6 (1:200, Santa Cruz Cat# sc-81431, RRID:AB_1128357), rabbit anti-STIL (1:500, Abcam Cat# ab89314, RRID:AB_2197878), mouse IgG2a anti-V5 (1:00, Thermo Fisher Scientific Cat# R960-25, RRID:AB_2556564), rabbit anti-WDR90 (1:100, Thermo Fisher Scientific Cat# PA5-61943, RRID:AB_2649628), chicken anti-GFP antibody (Aves Cat# GFP-1020, RRID:AB_10000240).

For immunofluorescence and U-ExM, AlexaFluor conjugated secondary antibodies (Thermo-Fisher) were diluted 1:1000 in PBS-BT. Goat anti-Mouse IgG1, 488 (1:1000, Thermo Fisher Scientific Cat# A-21121, RRID:AB_2535764), Goat anti-Mouse IgG2a, 488 (1:1000, Thermo Fisher Scientific Cat# A-21131, RRID:AB_2535771), Goat anti-Mouse IgG2b, 488 (1:1000, Thermo Fisher Scientific Cat# A-21141, RRID:AB_2535778), Goat anti-rabbit IgG (H+L), 488 (1:1000, Thermo Fisher Scientific Cat# A-11034 (also A11034), RRID:AB_2576217), Goat anti-Mouse IgG1, 568 (1:500, Thermo Fisher Scientific Cat# A-21124, RRID:AB_2535766), Goat anti-Mouse IgG2a, 568 (1:500, Thermo Fisher Scientific Cat# A-21134, RRID:AB_2535773), Goat anti-Mouse IgG2b, 568 (1:500, Thermo Fisher Scientific Cat# A-21144, RRID:AB_2535780), Goat anti-rabbit IgG (H+L), 568 (1:500, Thermo Fisher Scientific Cat# A-11036 (also A11036), RRID:AB_10563566), Goat anti-Mouse IgG3, 594 (1:500, Thermo Fisher Scientific Cat# A-21155, RRID:AB_2535785), Goat anti-rat IgG (H+L), 594 (1:500,Thermo Fisher Scientific Cat# A-11007 (also A11007), RRID:AB_10561522), Goat anti-Mouse IgG1, 647 (1:500, Thermo Fisher Scientific Cat# A-21240, RRID:AB_2535809), Goat anti-Mouse IgG2a, 647 (1:500, Thermo Fisher Scientific Cat# A-21241, RRID:AB_2535810), Goat anti-Mouse IgG2b, 647 (1:500, Thermo Fisher Scientific Cat# A-21242, RRID:AB_2535811), Goat anti-rabbit IgG (H+L), 647 (1:500, Thermo Fisher Scientific Cat# A32733, RRID:AB_2633282), Goat anti-Mouse, Star Red (1:200, Abberior Cat# STRED-1001, RRID:AB_3068620), Goat anti-rabbit, Star Orange (1:200, Abberior Cat# STORANGE-1002, RRID:AB_3068622), Goat anti-chicken, Alexa 488 (Thermo Fisher Scientific Cat# A-11039, RRID:AB_2534096).

Primary antibodies used for Western blotting and dilutions in TBST: rabbit anti TUBD1 (1:1000, Sigma-Aldrich Cat# HPA027090, RRID:AB_1858457), rabbit anti TUBE1 (1:1000, Sigma-Aldrich Cat# HPA032074, RRID:AB_10601216), rabbit anti C14orf80 (1:1000, Sigma-Aldrich Cat# HPA039049, RRID:AB_2676320), rabbit anti C16orf59 (1:1000, Sigma-Aldrich Cat# HPA055389, RRID:AB_2732595), mouse IgG2b anti SASS6 (1:200, Santa Cruz Biotech Cat# sc-81431, RRID:AB_1128357), rabbit anti STIL (1:2000, Abcam Cat# ab89314, RRID:AB_2197878), rabbit anti CENPJ/CPAP (1:1000, Proteintech Cat# 11517–1-AP, RRID:AB_2244605), rabbit anti POC5 (1:1000, Thermo Fisher Scientific Cat# A303-341A (also A303-341A-T), RRID:AB_10971172), mouse IgG2a anti V5 (1:1000, Thermo Fisher Scientific Cat# R960-25, RRID:AB_2556564), mouse IgG1 anti Flag, clone M2 (1:2000, Sigma-Aldrich Cat# F1804, RRID:AB_262044). Secondary antibodies used for Western blotting: 680 Donkey anti rabbit (H+L) (1:20,000, Thermo Fisher Scientific Cat# A10043, RRID:AB_2534018), 800 Donkey anti rabbit (H+L) (1:20,000, Li-COR Cat# 926–32213, RRID:AB_621848), 680 Donkey anti mouse (H+L) (1:20,000, Thermo Fisher Scientific Cat# A10038, RRID:AB_11180593), 800 Donkey anti mouse (H+L) (1:20,000, Li-COR Cat# 926–32212, RRID:AB_621847).

The cell lines generated in this work are available through contacting the corresponding author (Jennifer T. Wang, Department of Biology, Washington University in St. Louis).