TRIM37 prevents formation of condensate-organized ectopic spindle poles to ensure mitotic fidelity

TRIM37 is a tripartite motif (TRIM) ubiquitin ligase with an N-terminal RING, B-box, coiled-coil ubiquitin ligase domain (Brigant et al., 2018). Unique among TRIM family ligases, TRIM37 also has a TRAF (tumor necrosis factor receptor–associated factor) domain that mediates protein–protein interactions (Meitinger et al., 2020; Park, 2018). TRIM37 loss-of-function mutations cause a rare autosomal recessive disorder called mulibrey (for muscle-liver-brain-eye) nanism (Avela et al., 2000; Brigant et al., 2018). Symptoms of mulibrey nanism include pre- and postnatal growth failure, liver enlargement, deregulation of glucose and lipid metabolism (including type 2 diabetes, fatty liver, and hypertension), infertility, and, in ∼20% of cases, overgrowth of the fibrous sac that surrounds the heart (constrictive pericarditis). Mulibrey nanism patients are also plagued with numerous tumors (Karlberg et al., 2009a; Karlberg et al., 2009b; Sivunen et al., 2017). How TRIM37 loss causes the spectrum of symptoms observed in mulibrey nanism patients or renders them cancer prone is not understood.

At the cellular level, TRIM37 has been localized to peroxisomes (Kallijärvi et al., 2002; Wang et al., 2017) and centrosomes (Meitinger et al., 2020). At peroxisomes, TRIM37 was shown to monoubiquitinate the peroxisomal import receptor Pex5, which lent support to the classification of mulibrey nanism as a peroxisomal disorder (Kallijärvi et al., 2002; Wang et al., 2017). A TRIM37 knockout mouse model recapitulated a number of the phenotypes seen in mulibrey nanism patients (Kettunen et al., 2016). Trim37^−/− mice had growth defects, were infertile, and exhibited elevated fasting blood glucose and low fasting serum insulin levels. After 18 mo, they also exhibited cardiomyopathy, hepatomegaly, fatty liver, and various tumors. However, peroxisome number and morphology were normal in Trim37^−/− mice, even following a challenge that drives peroxisome biogenesis (Kettunen et al., 2016). Thus, cellular-level defects other than in peroxisomes likely underlie mulibrey nanism pathology.

TRIM37 also localizes to centrosomes (Meitinger et al., 2020), and centrosome-associated proteins feature prominently among TRIM37 interaction partners identified by proximity interaction proteomic analysis (Firat-Karalar et al., 2014; Yeow et al., 2020). Centrosomes, which are organized by small cylindrical organelles called centrioles, are the primary microtubule organizing centers (MTOCs) in animal cells (Banterle and Gönczy, 2017; Breslow and Holland, 2019; Loncarek and Bettencourt-Dias, 2018). Centrioles recruit a proteinaceous matrix called the pericentriolar material (PCM) that nucleates microtubules to catalyze assembly of the mitotic spindle (Mennella et al., 2014; Woodruff et al., 2014). To ensure that mitotic cells have precisely two centrosomes to organize the two poles of a bipolar mitotic spindle, centriole duplication is tightly regulated. In S phase, each mother centriole gives rise to a single daughter centriole in a process controlled by the kinase PLK4 (Banterle and Gönczy, 2017; Breslow and Holland, 2019; Loncarek and Bettencourt-Dias, 2018). TRIM37 was identified in an RNAi screen in HeLa cells for proteins whose knockdown led to an increase in the number of foci containing the centriolar protein centrin (Balestra et al., 2013). This phenotype has been interpreted as TRIM37 loss leading to the formation of extra centrioles by promoting centriole reduplication, in which a single mother centriole gives rise to more than one daughter centriole within a single cell cycle (Balestra et al., 2013). TRIM37Δ cells and cells expressing ligase-defective TRIM37 have also been shown to possess an ectopic PLK4-containing condensate (Meitinger et al., 2020). TRIM37 interacts with and ubiquitinates PLK4 but does not alter PLK4 levels, suggesting that rather than promoting its degradation, TRIM37-mediated ubiquitination prevents the incorporation of PLK4 into ectopic assemblies (Meitinger et al., 2020). In live-cell imaging, transiently multipolar spindles were observed in ∼20% of TRIM37Δ cells (Meitinger et al., 2016). Whether this low-penetrance spindle phenotype arises from extra centrioles or is due to microtubule organization by ectopic assemblies, such as the PLK4-containing condensate, has not been clarified.

Here, we analyze the impact of TRIM37 loss to define mitotic defects that may contribute to mulibrey nanism. We show that TRIM37 activity does not regulate centriole structure or duplication, or the ability of centrioles to template cilia formation, but instead restrains the formation of ectopic centrosomal protein assemblies that coexist with centrosomes. Cells lacking TRIM37 form a large highly ordered condensate, scaffolded by the centrosomal protein centrobin and containing PLK4, that arises by budding off of the centrosome. In ∼25% of cells, the condensate recruits other centrosomal proteins and acquires MTOC activity during mitotic entry, generating an ectopic spindle pole that elevates chromosome missegregation rates. The aberrant mitotic phenotypes of TRIM37Δ cells are suppressed by removing centrobin, which interacts with and is ubiquitinated by TRIM37. We propose that chromosome missegregation resulting from transient multipolarity and multipolar segregation caused by the presence of an ectopic noncentrosomal spindle pole is a major cellular-level pathology contributing to mulibrey nanism that explains why patients with this disorder are cancer prone.

To determine if TRIM37 loss affects centriole structure or duplication, we used expansion microscopy to analyze centrioles in WT and TRIM37Δ RPE1 cells. To assess whether selectively mutating TRIM37 ligase activity was similar to TRIM37 deletion, we also analyzed TRIM37Δ cells stably expressing transgene-encoded WT or ligase mutant TRIM37 (Lig^mut; catalytic Cys18 mutated to Arg; Fig. 1 A; Meitinger et al., 2020). After expansion, centrioles were visualized by immunostaining for acetylated tubulin, a marker for centriolar microtubules, and CEP290, a marker for the distal region of centrioles (Kong et al., 2020). Centrioles in TRIM37Δ cells appeared structurally normal (Fig. 1 B) and there was no significant difference in the distribution of mother centriole lengths between parental RPE1, TRIM37Δ, and TRIM37Δ reconstituted with WT or Lig^mut TRIM37 (Fig. 1, B and C).

To assess effects on centriole number, we analyzed large populations of cells for each condition. TRIM37Δ cells had predominantly two centrioles (single or duplicated) with a small fraction (∼5%) having more than two separated centrioles (Fig. 1, B and D). An even smaller fraction (∼1%) had partial structures labeled with acetylated tubulin that were not full centrioles (Fig. 1, B and D). The minor elevation in cells with increased centriole number was suppressed by expression of WT, but not Lig^mut, TRIM37 (Fig. 1 D). Expansion microscopy makes it straightforward to assess centriole overduplication by counting the number of procentrioles per mother centriole in S phase cells. Based on two independent experiments scoring >1,000 S phase cells with duplicating mother centrioles, no centriole overduplication was observed in TRIM37Δ or in TRIM37Δ cells expressing Lig^mut TRIM37. Thus, the ∼5% of cells with extra centrioles observed in TRIM37Δ cells do not arise from overduplication in S phase. TRIM37Δ RPE1 cells also retained the ability to assemble primary cilia following serum starvation (Fig. 1 E). Ciliation frequency and ciliary length were slightly elevated in the TRIM37Δ cell line relative to parental RPE1 cells (Fig. 1 E and Fig. S1 A); however, whether this results from TRIM37 loss or reflects clonal variation is unclear.

Collectively, these results indicate that TRIM37 is not regulating centriole duplication, structure, or the ability to assemble cilia. The minor proportion of cells with elevated centriole number observed in TRIM37Δ cells likely arise indirectly as a consequence of mitotic defects, as described below.

In a localization screen of centrosomal components comparing TRIM37Δ RPE1 cells to control USP28Δ RPE1 cells, we found that centrobin (encoded by CNTROB) localized prominently to a noncentrosomal focus in TRIM37Δ cells; no such focus was observed in control USP28Δ cells (Fig. 1 F). Centrobin localizes to procentrioles and to daughter centrioles that have not yet duplicated (Zou et al., 2005) and has been implicated in ciliogenesis (Ogungbenro et al., 2018; Reina et al., 2018). In our experiments, we used USP28Δ RPE1 cells generated in parallel to the TRIM37Δ RPE1 cells as a control, because loss of USP28 inactivates a sensor pathway that triggers p53 activation in response to extended mitotic duration, which facilitates analysis of phenotypes in the presence and absence of centrosomes (Meitinger et al., 2016).

The noncentrosomal structure observed with centrobin staining also labeled for PLK4, indicating that it is the same condensate previously described based on PLK4 staining in TRIM37Δ cells (Meitinger et al., 2016; Meitinger et al., 2020). Approximately 60% of interphase TRIM37Δ cells had a condensate that was spatially distinct from the centrosome, and all condensates colabeled for centrobin and PLK4. Quantification of signal intensities indicated that condensates varied in size and that centrosomal centrobin levels were significantly reduced in TRIM37Δ cells with a condensate, whereas centrosomal PLK4 levels were not (Fig. 1 G). Similar condensates were observed in a fibroblast cell line from a mulibrey nanism patient (Fig. 1 H), consistent with biallelic loss of TRIM37 in this disorder.

Costaining with centrobin and the centriolar/centrosomal proteins SAS6, CEP152, CEP192, CDK5RAP2, and PCNT indicated that these components were not present in the condensate (Fig. S1 B), consistent with prior colabeling conducted with PLK4 (Meitinger et al., 2020). Catalytic mutants of ubiquitin ligases can act as substrate traps; consistent with this, an mNeonGreen (mNG) fusion with Lig^mut TRIM37 concentrated prominently at the condensate, in addition to localizing to centrosomes (Fig. 1 I). Thus, the condensates in TRIM37Δ cells contain centrobin and PLK4 and, when expressed, Lig^mut TRIM37, but not other centrosomal components (Fig. 1 J; Meitinger et al., 2020).

To analyze the structure of the condensates resulting from TRIM37 deletion, we employed correlative light and electron microscopy (CLEM) and super-resolution microscopy. For CLEM, we used the TRIM37Δ RPE1 cell line stably expressing Lig^mut TRIM37-mNG, which allows condensate identification by fluorescence microscopy (Fig. 2 A). CLEM of the condensates revealed a highly ordered striated morphology, with electron-dense stripes spaced ∼90 nm apart (Fig. 2 A); in the same cells, centrosomes with localized Lig^mut TRIM37-mNG and a normal pair of centrioles were present (Fig. 2 A; cell 1, centrosome, sections 1–4). Similar structures were observed in 9 out of 12 cells analyzed by CLEM. Serial sectioning of large condensates suggests formation by coalescence of smaller width structures (e.g., condensate in cell 1 in Fig. 2 A). In one cell processed for CLEM, we observed condensates that in addition to striated densities also contained regions with punctate electron densities arranged in hexagonally packed sheet-like configuration (Fig. 2 B and Fig. S1 C). In serial sections, the hexagonally packed punctate morphology transitioned to a striped morphology associated with narrowing of the width of the structure (e.g., sections 2 and 3 of condensate in Fig. 2 B).

To confirm the strikingly regular morphology of the condensates observed by CLEM, we employed three superresolution imaging approaches: expansion structured illumination microscopy (Exp-SIM), stochastic optical reconstruction microscopy (STORM), and stimulated emission depletion (STED) microscopy (Fig. 3, A–D; and Fig. S1 D). Centrobin was detected using an antibody raised against the whole protein, and PLK4 was detected using an antibody against the C-terminal 157 amino acids. Only the centrobin antibody worked for Exp-SIM, but both antibodies worked for STED and STORM. As these methods do not require a reference fluorescence signal, we analyzed TRIM37Δ cells and TRIM37Δ cells expressing Lig^mut TRIM37-mNG. All three methods revealed condensate morphologies that fell into the classes observed by CLEM: (1) “linear striated” structures, (2) hexagonally packed “punctate sheets”, and (3) “hybrid” structures with regions exhibiting both morphologies (Fig. 3, A–D). In TRIM37Δ cells expressing Lig^mut TRIM37-mNG, where a direct comparison can be made between the CLEM and superresolution analysis, predominantly linear striated structures were observed with interstripe distances nearly identical to those measured by CLEM (Fig. 3, C and D). All three condensate morphologies were observed in both cell lines, but TRIM37Δ cells had a higher percentage with punctate sheet-like morphology while linear striated condensates predominated in the TRIM37Δ cells expressing Lig^mut TRIM37-mNG (Fig. 3 C). Quantification of the interstripe and interpuncta dimensions revealed consistent values independently of the method employed and suggested that the interstripe distance in the linear structures was modestly larger than the interpuncta distance in the sheet-like structures (Fig. 3 D).

We conclude that loss of TRIM37 or its ligase activity leads to the formation of condensates that exhibit two related highly ordered morphologies. The existence of hybrid condensates suggests that there may be interconversion between these two morphologies. Linear striated structures are enriched following expression of Lig^mut TRIM37-mNG in TRIM37Δ cells, suggesting that binding of the catalytically inactive ligase may preferentially stabilize this morphology.

We next sought to understand how condensates form in the absence of TRIM37 and why there is typically one such condensate per cell. Structured illumination microscopy (SIM), in conjunction with markers for specific centriolar domains, indicated that TRIM37 is broadly localized throughout the centrosome, suggesting that it acts locally at the centrosome to prevent condensate formation (Fig. S1 E). To image condensate formation in living cells, we employed a TRIM37Δ cell line expressing Lig^mut TRIM37-mNG. As cells typically have one condensate, following mitosis, one daughter cell inherits the condensate and the other is born without a condensate (Fig. 3 E). In cells born without a condensate, Lig^mut TRIM37-mNG hyperaccumulated around the centrosome, and, after some time, a portion of the accumulated material budded off to form a condensate (Fig. 3 E and Video 1). In the daughter cell that inherited a condensate after division, no hyperaccumulation of the probe or new condensate formation occurred (Fig. 3 E and Video 1). Condensates budded off of the centrosome at varying times, suggesting that budding is not cell cycle regulated (Fig. 3 F). In addition, condensate formation and budding occurred with similar frequency following acute inhibition of PLK4 by centrinone (Fig. 3 F), indicating that PLK4 kinase activity is not required. In rare cases, we observed two condensates that fused into one (Fig. S1 F); the fused condensate was elongated and did not adopt a spherical shape characteristic of liquid-like condensates. We conclude that condensate assembly occurs at centrosomes and that this assembly is normally prevented by the ubiquitin ligase activity of TRIM37.

We next focused on the consequences of TRIM37 loss–induced condensate formation on mitosis. Expansion microscopy coupled to labeling for acetylated tubulin revealed that ∼30% of mitotic TRIM37Δ cells had chromosome configurations that suggested the presence of an extra spindle pole; of the cells with a multipolar chromosome configuration, the majority (∼75%) had two centrosomes, with the remainder having either an extra centrosome or a small focus of acetylated tubulin at the third pole (Fig. 4 A). This initial analysis suggested that ∼20–25% of TRIM37Δ prometaphase/metaphase cells have an ectopic noncentrosomal spindle pole that is likely formed by the condensate. Consistent with the idea that condensates can nucleate microtubules to form an ectopic pole, optical sectioning through immunofluorescence images of prometaphase/metaphase TRIM37Δ cells colabeled for centrobin and microtubules revealed that ∼25% of the condensates were at the center of a focused array of microtubules (Fig. 4, B–D). Acquisition of MTOC activity by condensates correlated with transition into prometaphase/metaphase (Fig. 4, B and D) and was associated with recruitment to the condensate of the microtubule nucleation-promoting centrosomal components CEP192, PCNT, CDK5RAP2, and γ-tubulin (Fig. S2 A), none of which are present on condensates in interphase (Fig. S1 B and Fig. 1 J). In microtubule regrowth assays, approximately half of the condensates in prometaphase/metaphase cells exhibited microtubule-nucleating activity (Fig. S2 B).

To address whether the ectopic poles observed in TRIM37Δ RPE1 cells are relevant to mulibrey nanism, we analyzed mitotic patient-derived fibroblasts. No condensates were observed in control mitotic primary fibroblasts, whereas 21 of 22 mitotic mulibrey fibroblast cells had a condensate. In 9 of these mitotic cells (∼40%), the condensate organized an ectopic spindle pole (Fig. 4 E). Thus, formation of ectopic spindle poles by condensates is observed in both RPE1 TRIM37Δ cells and in mulibrey nanism patient-derived fibroblasts.

To complement the immunofluorescence in fixed cells, we conducted live imaging of microtubules, which revealed ectopic spindle poles during ∼25% of mitoses in TRIM37Δ cells (Fig. 4, F and H; and Video 2). The formation of ectopic poles was suppressed by the expression of WT TRIM37-mNG, but not Lig^mut TRIM37-mNG (Fig. 4, F–H). Two-color live imaging of microtubules and of Lig^mut TRIM37-mNG as a condensate marker revealed three phenotypes (Fig. 4, G and H; and Video 3): (1) condensates that did not nucleate microtubules or generate an ectopic spindle pole (bipolar configuration), (2) condensates that generated an ectopic pole that ultimately clustered with one of the centrosomes to enable bipolar segregation (transient multipolarity), and (3) condensates that generated an ectopic pole that persisted into anaphase (multipolar segregation).

A small number of mitotic TRIM37Δ cells (∼5%) harbor an aberrant number of centrioles (Fig. 1, B and D), although no centriole overduplication was observed in >1,000 S phase TRIM37Δ cells. To address the origin of these extra centrioles, we analyzed a large number of mitotic TRIM37Δ cells labeled for centrobin. In TRIM37Δ mitotic cells with two centrosomes and a condensate, ∼27% exhibited a multipolar configuration with the condensate sitting at the third pole; interestingly, in ∼2% of the cells, a bipolar chromosome configuration was observed in which the condensate formed one pole and both centrosomes clustered at the other pole (Fig. 4, I and J). This configuration, which we refer to as “dominant,” would lead to generation of one daughter cell with twice the normal number of centrioles and a second daughter cell without centrioles, which would likely activate de novo centriole assembly (which lacks copy number control) in the subsequent S phase. Division of cells with this dominant configuration caused by a condensate-organized spindle pole, along with potential cytokinesis failure of multipolar division configurations, likely accounts for the generation of TRIM37Δ cells with aberrant centriole numbers. We speculate that such cells are constantly generated and being selected against, explaining the persistence of a small percentage of cells with aberrant centriole numbers in the clonally derived TRIM37Δ cell population.

We next addressed the functional consequences of ectopic condensate-based spindle poles in TRIM37Δ cells. Fixed analysis of chromosome segregation, comparing WT and TRIM37Δ RPE1 anaphase-telophase cells with two centrosomes, revealed a significantly elevated rate of multipolar segregation (∼6.5% of anaphase/telophase cells) and of lagging chromosomes (∼3% of anaphase/telophase cells; Fig. 5 A). Consistent with this observation, the frequency of cells harboring micronuclei was also significantly elevated in the TRIM37Δ cell population relative to parental RPE1 cells (Fig. 5 B).

To independently assess chromosome missegregation rates in TRIM37Δ cells, we performed FISH analysis of two chromosomes (17 and 18) in telophase cells. We found a significant number of missegregation events for these chromosomes in both bipolar and multipolar configurations (Fig. 5 C), as well as properly segregating but aneuploid cells (Fig. 5 C, bipolar aneuploid), indicative of a missegregation event in a prior division. Collectively, this analysis indicates that the ectopic spindle poles formed by condensates in TRIM37Δ cause significant chromosome missegregation, generating aneuploid progeny.

To causally link condensates to the observed mitotic defects in TRIM37Δ cells, we assessed the molecular requirements for condensate formation using inducible CRISPR/Cas9-based knockouts of CNTROB and PLK4. Inducible knockouts were generated in TRIM37Δ RPE1 cells and USP28Δ RPE1 cells, which served as controls. The strategy used to target CNTROB is shown in Fig. S3 A, and efficient generation of mutations in CNTROB after Cas9 induction in both the TRIM37Δ and USP28Δ cell lines was confirmed by sequencing (Fig. S3 A). The inducible PLK4 knockout was previously generated and shown to result in efficient generation of mutations and centrosome loss after 4-d induction (Meitinger et al., 2020).

Inducible CNTROB knockout caused loss of centrosomal centrobin signal in USP28Δ RPE1 cells, confirming efficacy of the knockout (Fig. 6 A and Fig. S3 B); no defect in centrosome number or proliferation was observed in the 4-d period after induction (not shown). In TRIM37Δ RPE1 cells, inducible CNTROB knockout led to loss of condensates labeled with centrobin and PLK4; by contrast, PLK4 signal was still detected at centrosomes (Fig. 6 A). Conducting the same analysis following inducible PLK4 knockout showed that 4-d induction of the PLK4-targeting gRNA resulted in formation of acentrosomal cells. Approximately 12% of these acentrosomal cells lacked a centrobin and PLK4-containing condensate; the remaining cells retained a condensate that labeled for both centrobin and PLK4, although condensates in TRIM37Δ iPLK4 KO cells varied in size and on average had less of both markers than condensates in TRIM37Δ cells (Fig. S4, A and B). These results suggest that PLK4 does not turn over after it is incorporated into the condensates. A caveat with interpreting this result is that centriole loss due to PLK4 protein depletion in TRIM37Δ cells delays mitosis and triggers p53-dependent arrest (Meitinger et al., 2020). This cessation of proliferation following inducible PLK4 knockout, which is not observed after removal of centrobin, confounds analysis of the role of PLK4 in condensate formation.

We conclude that centrobin is essential to build the structured condensates in TRIM37Δ cells. We thus refer to these structures as centrobin-scaffolded condensates.

TRIM37 depletion by RNAi in HeLa cells has been reported to result in supernumerary foci containing the centriolar marker centrin (Balestra et al., 2013). We confirmed that centrin, in addition to localizing to the centrosome, is present in ectopic foci in ∼60% of interphase TRIM37Δ RPE1 cells. The number of centrin foci per cell was variable, typically between one and five, and centrin foci were smaller than the condensate (Fig. 6 B). One centrin focus partially colocalized with the condensate (see magnified view of the condensates in Fig. 6 B and Fig. S4 C; 96% of the condensates had a colocalized centrin focus, n = 100). Additional centrin foci that did not colocalize with the condensate were often also present in TRIM37Δ cells (Fig. 6 B and Fig. S4 C).

Removal of centrobin using the inducible CNTROB knockout eliminated the condensate but did not eliminate centrin foci (Fig. 6 B); the frequency of cells with centrin foci and the number of centrin foci per cell were mildly elevated in TRIM37Δ cells lacking centrobin (Fig. 6 B). No PLK4 was detected in the centrin foci observed in CNTROB knockout cells (Fig. 6 B). We conclude that centrin foci in TRIM37Δ cells do not require the centrobin-scaffolded condensate to form. Thus, TRIM37 independently suppresses formation of a centrobin-scaffolded condensate and of centrin foci (Fig. 6 C).

In a significant proportion (∼20–30%) of mitotic TRIM37Δ cells, an ectopic noncentrosomal spindle pole is observed with a condensate at its center. To causally link the condensate to the formation of ectopic poles, we employed the inducible CNTROB knockout followed by fixed analysis and live imaging. Inducing CNTROB knockout in TRIM37Δ cells for 4 d suppressed the formation of ectopic spindle poles in fixed prometaphase and metaphase cells (Fig. 7, A and B). Centrin foci were still observed in these cells lacking centrobin (Fig. S4 C). Live imaging of chromosome dynamics revealed that centrobin removal suppressed both transient multipolarity, which is evident in the chromosome configuration, and the multipolar segregation observed in TRIM37Δ cells (Fig. 7, C and D; and Video 4). CNTROB knockout also suppressed the mild but statistically significant increase in mitotic duration observed in TRIM37Δ cells (Fig. 7 D). Thus, condensates built on a centrobin scaffold are responsible for the formation of ectopic spindle poles that perturb spindle assembly and chromosome segregation in the absence of TRIM37.

Centrobin is essential for formation of condensates in TRIM37Δ cells, suggesting that centrobin may be a direct target of TRIM37’s ubiquitin ligase activity. Consistent with this idea, centrobin protein level was elevated ∼3.5-fold in TRIM37Δ cells (Fig. 8 A); by contrast, RNA-Seq indicated that CNTROB mRNA levels were unchanged (Meitinger et al., 2020). To determine if TRIM37 binds to and regulates centrobin stability, we used a human cell expression system to coexpress centrobin with WT or Lig^mut TRIM37. We analyzed crude extracts to assess the effect of TRIM37 ligase activity on total centrobin levels and assayed the formation of insoluble centrobin condensates by centrifugation to generate supernatant and pellet fractions; in addition, we analyzed interaction between centrobin and TRIM37 in the supernatant (Fig. 8 B). Comparison of centrobin levels in crude extracts indicated that coexpression with WT TRIM37 reduced the total amount of centrobin protein approximately threefold compared with expression of centrobin alone (Fig. 8, C and E; and Fig. S5 A). TRIM37’s ligase activity was required for the reduction in centrobin levels (Fig. 8, C and E; and Fig. S5 A). When expressed on its own, the majority of centrobin was present in the pellet (Fig. 8, D and E; and Fig. S5 B). While quantitative recovery of centrobin from the pellet was challenging even under denaturing conditions, leading to variability in measured values, coexpression of WT TRIM37 consistently increased the relative solubility of centrobin, measured as a decrease in the ratio between the amount in the pellet and supernatant (pellet/supernatant ratio; Fig. 8, D and E; and Fig. S5 B); by contrast, coexpression of Lig^mut TRIM37 did not (Fig. 8, D and E; and Fig. S5 B).

The above analysis additionally revealed that Lig^mut TRIM37, which is soluble when expressed on its own, sedimented into the insoluble pellet fraction when coexpressed with centrobin (Fig. 8, D and E). As ligase-inactive forms of ubiquitin ligases can act as substrate traps, this result is consistent with centrobin being a substrate of TRIM37. In support of this idea, Lig^mut TRIM37 coimmunoprecipitated with the centrobin that remained in the supernatant (Fig. 8 F and Fig. S5 C). WT TRIM37 is expressed at a much lower level than Lig^mut TRIM37 because TRIM37 ligase activity reduces its own stability, a common property of RING family ubiquitin ligases (de Bie and Ciechanover, 2011). Consistent with an interaction between the two proteins, WT TRIM37 could also be coimmunoprecipitated with centrobin when it was partially stabilized by treating the cells with the proteasome inhibitor MG132 (Fig. S5 D). Introduction of epitope-tagged ubiquitin into the coexpression revealed that immunopurified centrobin was ubiquitinated when coexpressed with WT, but not Lig^mut, TRIM37 (Fig. 8 G). TRIM37-dependent ubiquitination of centrobin was associated with a decrease in centrobin levels (Fig. 8 G).

The ability of centrobin to greatly decrease the solubility of Lig^mut TRIM37 suggests that they are stably associated. To assess if Lig^mut TRIM37 is also stably associated with condensates in cells, we conducted photobleaching analysis of Lig^mut TRIM37-mNG–labeled condensates in TRIM37Δ cells. Lig^mut TRIM37 exhibited no detectable turnover on condensates in ∼20 min (Fig. 8, H and I; and Video 5), in agreement with the conclusions of the coexpression analysis.

Collectively, these data indicate that TRIM37 directly binds to centrobin and ubiquitinates it to control its levels. TRIM37 may regulate centrobin levels by ubiquitination-dependent degradation, or it may modify centrobin with ubiquitin to prevent self-association, with the increase in centrobin levels in the absence of TRIM37 being an indirect consequence of the formation of stable insoluble condensates. Distinguishing these possibilities will require future biochemical and structural analysis.

Much of the prior functional analysis of TRIM37 in dividing cells has focused on cells that lack centrosomes due to inhibition of PLK4 kinase activity (Fong et al., 2016; Meitinger et al., 2016; Meitinger et al., 2020; Yeow et al., 2020). In cells lacking centrosomes, TRIM37 loss accelerates acentrosomal spindle assembly, because it leads to the formation of an ectopic array of PLK4-scaffolded foci that recruit PCM proteins and can substitute for centrosomes in promoting spindle assembly in mitosis (Meitinger et al., 2016; Meitinger et al., 2020). A prominent component of these PLK4-scaffolded foci is the centrosomal component CEP192 (Fig. 9 A). To assess if centrobin is also present in the array of foci containing PLK4, CEP192, and other centrosomal components in cells lacking centrosomes, we analyzed centrobin localization in centrinone-treated TRIM37Δ cells. We found that the majority of PLK4 foci in centrinone-treated cells contained CEP192, but not centrobin, whereas one PLK4 focus contained centrobin and not CEP192 (Fig. 9 A). Thus, centrobin staining partitioned foci labeled with the PLK4 antibody in centrinone-treated TRIM37Δ cells into two types: a centrobin-containing focus (likely the highly ordered condensate described above) and an array of PLK4-containing foci that contain other centrosomal components such as CEP192, but not centrobin. Consistent with this partitioning, knockout of CNTROB eliminated the centrobin-scaffolded condensate in centrinone-treated TRIM37Δ cells but did not affect formation of PLK4-scaffolded foci (Fig. 9, B and C). By contrast, inducible PLK4 knockout eliminated the PLK4-scaffolded foci, as shown previously (Meitinger et al., 2020), but the centrobin-containing condensate was still observed (Fig. 9, B and C). Thus, the array of PLK4-scaffolded foci in centrinone-treated TRIM37Δ cells that contain centrosomal components such as CEP192 is distinct from the structured centrobin-scaffolded condensate.

TRIM37 loss is proposed to improve mitosis in acentrosomal cells, because the PLK4-scaffolded foci act as efficient MTOCs that accelerate spindle assembly (Meitinger et al., 2016; Meitinger et al., 2020). In support of this idea, PLK4 knockout eliminated the functional improvement in spindle assembly, assessed by live imaging of chromosome dynamics, that is observed in centrinone-treated TRIM37Δ cells (Meitinger et al., 2020). We conducted the same analysis with CNTROB knockout, which does not affect the array of PLK4-scaffolded foci. The results showed that both in the presence and absence of centrobin, centrinone-treated TRIM37Δ cells exhibited significantly more efficient mitosis than centrinone-treated control USP28Δ cells (Fig. 9 D); in addition, PLK4-scaffolded foci were observed at the poles of the acentrosomal spindles with and without centrobin (Fig. 9 E). Thus, it is the PLK4-scaffolded foci and not the centrobin-scaffolded condensate that improve acentrosomal mitosis in centrinone-treated TRIM37Δ cells (Fig. 9 F).

We conclude that in cells that lack centrioles due to inhibition of PLK4 activity, TRIM37 prevents PLK4 protein from scaffolding the formation of centrobin-independent microtubule organizing foci that increase the efficiency of acentrosomal mitosis. This conclusion is consistent with the finding that TRIM37 interacts with and ubiquitinates PLK4 (Meitinger et al., 2020).

Here, we investigate TRIM37 loss of function in cells that have centrosomes with the goal of defining cellular-level defects that underlie mulibrey nanism. Our results indicate that the major function of TRIM37 in dividing cells is to prevent the formation of a centrobin-scaffolded condensate (Fig. 10). These condensates frequently acquire the ability to serve as MTOCs after mitotic entry, which elevates chromosome missegregation (Fig. 10). TRIM37 loss also triggers the formation of foci containing the centriolar protein centrin. Removal of centrobin eliminates the condensates and the ectopic spindle poles in TRIM37Δ cells while leaving centrin-containing foci intact. Thus, centrobin-scaffolded condensates, and not centrin foci, underlie the mitotic defects of TRIM37Δ cells. We propose that low frequency chromosome missegregation is a prominent cellular-level defect underlying mulibrey nanism that additionally 1explains the tumor-prone nature of mulibrey patients (Fig. 10). The overlap in phenotypic features between mulibrey nanism and mosaic variegated aneuploidy (MVA), a distinct rare human genetic disorder associated with mitotic defects, lends support to this proposal.

A key function of TRIM37 is to prevent the formation of ectopic centrosomal protein assemblies that acquire the ability to serve as MTOCs. In acentrosomal cells generated by centrinone treatment, TRIM37 loss enables the formation of an array of PLK4-scaffolded foci that recruit centrosomal proteins and serve as MTOCs that substitute for centrosomes in spindle assembly (Meitinger et al., 2016; Meitinger et al., 2020). In cells with centrosomes, loss of TRIM37 leads to the formation of highly ordered centrobin-scaffolded condensate that can acquire MTOC activity and reduce the fidelity of bipolar mitosis. Identifying the specific sites in centrobin and PLK4 that are ubiquitinated by TRIM37 and dissecting the structure and assembly pathways for centrobin-scaffolded condensates and PLK4-scaffolded foci are important future goals.

Centrobin-scaffolded condensates in TRIM37Δ cells are highly ordered and have either a linear striated or a hexagonally packed punctate sheet structure. Ectopic centrobin-containing structures with a striated morphology have also been reported in a recent study using RNAi-mediated TRIM37 depletion (Balestra et al., 2021). An important question raised by the highly ordered nature of these ectopic structures in TRIM37Δ cells and their formation at centrosomes is whether they reflect a normal assembly process in a specific biological context. As these assemblies are deleterious in mitotically dividing cells, we speculate that their formation may occur in a postmitotic context. Centrobin is highly expressed in testis (Lee et al., 2009; Liška et al., 2009; Zou et al., 2005), and analysis of a mutation in the CNTROB locus in rats suggests that it is involved in the process that shapes the spermatid head (Dunleavy et al., 2019; Liška et al., 2009). During spermatogenesis, centrobin localizes to a basket-like structure containing a parallel array of microtubules called the manchette that ensheaths the nucleus and to an actin- and keratin-5–based cytoskeletal plate called the acroplaxome that links the nuclear envelope to the inner membrane of the acrosome (Liška et al., 2009). Thus, one possibility is that TRIM37 controls the formation of these centrobin assemblies during spermatogenesis, explaining why the rat CNTROB locus mutant and TRIM37 loss-of-function mutants in mice and humans are sterile.

Approximately one quarter of the centrobin-scaffolded condensates formed in TRIM37Δ cells recruit other centrosomal proteins and acquire the ability to serve as MTOCs upon mitotic entry. These condensate-based ectopic poles elevate chromosome missegregation in TRIM37Δ cells (Fig. 10). Notably, centrobin removal prevents condensate formation and eliminates ectopic spindle poles in TRIM37Δ cells. However, centrobin removal does not reduce the number of centrin-containing foci present in TRIM37Δ cells. We note that a recent study, which refers to the centrin foci as “Cenpas,” suggested that they serve as MTOCs to form ectopic poles that contribute to genomic instability and that centrobin-scaffolded condensates may serve as platforms for the generation of Cenpas (Balestra et al., 2021). In contrast, the inducible CNTROB knockout analysis that we conducted in TRIM37Δ cells suggests that centrobin-scaffolded condensates rather than centrin foci/Cenpas are responsible for the observed spindle defects and argues against the idea that centrobin-scaffolded condensates are required to generate centrin foci/Cenpas. Addressing the functional significance of the centrin foci observed in TRIM37Δ cells will require a means for selectively eliminating them without affecting the centrobin-scaffolded condensate.

Why only a subset of centrobin-scaffolded condensates are activated to serve as mitotic MTOCs is currently unclear. In Drosophilamelanogaster neuroblasts, centrobin has been implicated in assembly of a robust mitotic-like MTOC that forms during interphase around one of the two centrioles. In these cells, centrobin directs the recruitment of PCM components (including asterless and the PCM matrix protein Cnn) in a PLK1-regulated manner (Januschke et al., 2013). Like the Drosophila protein, human centrobin has also been shown to be phosphorylated by PLK1 during mitosis (Lee et al., 2010), suggesting that mitotic phosphorylation of centrobin by PLK1 may enable condensates to recruit other centrosomal components and acquire microtubule-nucleating capacity. As a significant proportion of condensates do not activate to become MTOCs in mitosis, there must be an additional step that is necessary.

TRIM37 loss leads to the formation of centrobin-scaffolded condensates that promote formation of multipolar spindles with elevated rates of chromosome missegregation. In RPE1 TRIM37Δ cells, the transient multipolar state resolves to a bipolar state before anaphase in the majority of cells but ∼5–6% of mitotic cells exhibit multipolar segregation and ∼2–3% exhibit lagging chromosomes (Fig. 10). Passage through a transient multipolar spindle intermediate elevates chromosome missegregation due to formation of merotelic kinetochore microtubule attachments in which a single kinetochore becomes attached to microtubules emanating from both spindle poles (Ganem et al., 2009; Silkworth et al., 2009). Thus, the ectopic spindle pole formed by the centrobin-scaffolded condensate in TRIM37Δ cells effectively act as a low-penetrance aneuploidy generator. We suggest that characteristics of mulibrey nanism, such as reduced pre- and postnatal growth and the high frequency of tumors (Karlberg et al., 2009a; Karlberg et al., 2009b; Sivunen et al., 2017), including Wilms’ tumor, are a consequence of aneuploidy resulting from condensate-mediated multipolar spindle assembly. This notion is supported by analysis of mulibrey nanism patient-derived fibroblasts.

Slow pre- and postnatal growth and a high propensity for tumor formation, including Wilms’ tumor, are also characteristic of patients with MVA (García-Castillo et al., 2008; Hanks et al., 2004; Snape et al., 2011; Yost et al., 2017). In these patients, >25% of cells have additional or missing chromosomes. MVA syndrome can result from mutations in BUB1B or TRIP13, which lead to defects in the spindle checkpoint and cause cells to exit mitosis before all chromosomes attach to the spindle (Hanks et al., 2004; Suijkerbuijk et al., 2010; Yost et al., 2017). The third described cause of MVA is mutations in the gene encoding the centrosomal protein CEP57 (Snape et al., 2011). CEP57 mutation has been shown to disrupt the structure of the PCM matrix and cause precocious separation of mother and daughter centrioles (Aziz et al., 2018; Watanabe et al., 2019). Precociously separated daughter centrioles acquire the ability to serve as MTOCs before spindle assembly, leading to the formation of multipolar spindles (Aziz et al., 2018; Watanabe et al., 2019). Although patients with CEP57 mutation–induced MVA are limited and not yet old enough to assess cancer susceptibility, mutation, loss, or haploinsuffiency of CEP57 in mice rendered them susceptible to spontaneous and carcinogen-induced tumor formation (Aziz et al., 2018).

The similarities of the cellular and patient phenotypes resulting from mutations in TRIM37 and CEP57 lead us to propose that aneuploidy resulting from a centrobin-scaffolded condensate-dependent ectopic spindle pole explains many features of mulibrey nanism, including slow growth and susceptibility to tumor formation. We note that, as loss of TRIM37 perturbs the distribution of centrobin by rendering it largely insoluble, additional features of mulibrey nanism, such as infertility, may be related to disruption of centrobin function in gametes (Liška et al., 2009, 2013; Ogungbenro et al., 2018; Reina et al., 2018).

Antibodies against CEP192 (1–211 aa; used at 0.5 µg/ml for immunofluorescence), SAS6 (501–657 aa; used at 0.5 µg/ml for immunofluorescence), and PLK4 (814–970 aa; used at 0.5 µg/ml for immunofluorescence) were generated as previously described (Wong et al., 2015). The following antibodies were purchased from commercial sources, with their working concentrations indicated in parentheses: anti-CENTROB (1:800 for immunofluorescence and Western blotting; 1:500 for expansion microscopy, STORM, and STED; ab70448; Abcam); anti-TRIM37 (1:2,000 for immunoblotting; A301-174A; Bethyl Laboratories); anti-CDK5RAP2 (1:1,000 for immunofluorescence; ab86340; Abcam); anti-Cep152 (1:1,000 ab183911; Abcam), anti-CEP152 (1:3,000; A302-479A; Bethyl Laboratories), anti-ARL13B (1:500; 17711–1-AP; Proteintech), anti–α-tubulin (1:500; T9026; Sigma-Aldrich), anti-PCNT (1:2,000 for immunofluorescence, 1:500 for immunoblotting; ab4448; Abcam), anti–γ-tubulin (T6557; 1:1,000; Sigma-Aldrich); anti–β-tubulin (1:500; ab15568; Abcam); anti-Centrin1 (1:1,000; 20H5; Millipore); anti-FLAG (1:1,000 for immunoblotting; F1804; Sigma-Aldrich); anti-Myc (1:5,000 for immunoblotting; monoclonal 9E10; M4439; Sigma-Aldrich); anti-HA (1:500 for immunoblotting; monoclonal antibody 16B12; BioLegend); anti-acetylated tubulin (1:4,000; T7451; Sigma-Aldrich); and anti-CEP290 (1:600; ab84870; Abcam). Anti-CENTROB and anti-Cep164 antibody (22227–1-AP; Proteintech) directly conjugated with CF647 using Mix-n-Stain CFDye antibody Labeling Kit (92238; Biotium) were used for costaining with anti α-tubulin and anti-CEP152 antibodies, respectively. Secondary antibodies for expansion microscopy anti-mouse Alexa Fluor 488 (1:800; A11029) and anti-rabbit Alexa Fluor 647 (1:800; A21245) were purchased from Thermo Fisher Scientific. In some expansion experiments, centrioles were visualized using anti-acetylated tubulin directly conjugated with CF568 using Mix-n-Stain CFDye antibody Labeling Kit (92235; Biotium). Other secondary antibodies were purchased from Jackson ImmunoResearch Laboratories and GE Healthcare.

All cell lines used in this study are listed in Table S1. RPE1 (hTERT RPE-1) cells were grown in F12/DMEM, and Lenti-X 293T cells were grown in DMEM. Both media were supplemented with 10% FBS. Fibroblast cell lines AG02602 (apparently healthy individual) and AG02506 (mulibrey nanism) were purchased from Coriell and grown in EMEM with 2 mM L-glutamine and 15% FBS. All media were supplemented with 100 µg/ml streptomycin and 100 U/ml penicillin. Cell lines were maintained in 37°C and 5% CO₂. FreeStyle 293-F cells were maintained at 37°C and 8% CO₂ on an orbital shaker platform rotating at 125 rpm. To inhibit PLK4 and deplete centrosomes, cells were treated with 150 nM centrinone (LCR-263; 150–200 nM; synthesized by Sundia MediTech) for the indicated amounts of time.

The RPE1 CEP192-NeonGreen and the RPE1 TRIM37Δ and USP28Δ knockout and inducible PLK4 knockout cell lines were described previously (Meitinger et al., 2016; Meitinger et al., 2020). The following transgenes were stably integrated into the genome using lentiviral constructs (see Table S2): mRuby2-MAP4-MBD (EF1α promoter; neomycin resistance gene), TRIM37-NeonGreen (hPGK promoter; blasticidin resistance gene), and TRIM37-C18R-NeonGreen (hPGK promoter; blasticidin resistance gene). Cell lines for inducible knockout of centrobin were generated by sequential lentiviral integration of Cas9 (Edit-R Inducible Lentiviral Cas9; Dharmacon) and a gRNA expressing plasmid, which is based on the lentiGuide-Puro plasmid (Sanjana et al., 2014). The centrobin gRNA (5′-GCT​ACA​GCA​ACA​ATT​AGC​CG-3′) targets exon 5. The lentiGuide-Puro was a gift from Feng Zhang (Massachusetts Institute of Technology, Cambridge, MA; plasmid 52963; http://n2t.net/addgene:52963; RRID:Addgene_52963; Addgene). Cas9 expression was induced with 1 µg/ml doxycycline.

Viral particles were generated by transfecting the lentiviral packaging constructs into Lenti-X-293T cells using Lenti-X Packaging Single Shots (Clontech). 48 h after transfection, virus-containing medium was harvested and added to the growth medium of cells in combination with polybrene (8 µg/ml; EMD Millipore). Cells were selected for 1 wk in 400 µg/ml G418, 5 µg/ml blasticidin, or 10 µg/ml puromycin.

Expansion microscopy analysis was performed as described previously (Kong et al., 2020; Sahabandu et al., 2019). The protocol yielded a fourfold expansion of the specimen. Briefly, cells growing on glass coverslips were fixed with 4% formaldehyde in PBS at RT for 1 h. After fixation, the coverslips were incubated at 40°C for 16 h in a solution containing 30% acrylamide (A4058; Sigma-Aldrich) and 4% formaldehyde in PBS. Cells were washed three times in PBS (10 min each wash at RT). Coverslips were placed on a parafilm-covered Petri dish floating in an ice-water bath. Precooled gelling mixture consisting of 20% acrylamide, 0.04% bis-acrylamide (A9926; Sigma-Aldrich), 7% sodium acrylate (408220; Sigma-Aldrich), 0.5% ammonium persulfate (248614; Sigma-Aldrich), and 0.5% tetramethylethylenediamine (411019; Sigma-Aldrich) was pipetted onto the coverslips, incubated on ice for 20 min, and additionally at RT for 1–2 h. Following gel polymerization, several punches were excised from each gelled sample using a 4-mm biopsy puncher (33–34-P/25; Integra Miltex). Punches were placed in a dry 50-ml tube and preheated at >90°C for 10 min. SDS solution (200 mM SDS, 200 mM NaCl, and 50 mM Tris, pH 9.0), preheated to >90°C, was added to the punches, which were boiled for 1 h at >90°C with swirling of the tube every 10 min. After boiling, SDS solution containing punches was cooled to RT, and SDS was removed by exchanging 1× PBS every 20 min for the first 2–3 h, followed by an overnight wash in 1× PBS at 4°C.

Punches were blocked in immunofluorescence (IF) buffer containing 1% BSA (A9647; Sigma-Aldrich) and 0.05% Tween-20 (P9416; Sigma-Aldrich) in PBS for 1 h at RT and incubated with primary antibody diluted in IF buffer for 48 h at 4°C. Punches were washed in PBS for 1 h and incubated with secondary antibody and 0.5 µg/ml DAPI (D1306; Thermo Fisher Scientific) in IF buffer for 24 h at 4°C. After immunostaining, samples were expanded in deionized water (dH₂O) for 2 h at RT with dH₂O exchanged every 10 min and additionally overnight at 4°C. Prior to imaging, expanded punches were mounted in Rose chambers (Kong and Loncarek, 2015; Rose, 1954).

Widefield images of expanded samples were acquired on inverted Eclipse Ti microscope (Nikon), equipped with OrcaFlash4 camera (Hamamatsu), Intensilight C-HGFIE illuminator, 60× NA 1.45 Plan Apo objective, and 1.5× magnifying tube lens. One central Z section or multiple 200- to 300-nm-thick Z sections spanning the centriole or the entire cell were acquired, as needed.

SIM of expanded samples was performed on N-SIM (Nikon), equipped with 405-, 488-, 561-, and 640-nm excitation lasers, Apo TIRF 100× NA 1.49 Plan Apo oil objective, and a back-illuminated electron-multiplying charge-coupled device (CCD) camera (DU897; Andor). 0.1-µm Z sections were acquired in 3D SIM mode and reconstructed to generate a final SIM image. Nikon NIS Elements software was used for surface rendering.

In the expanded samples, G1 cells were recognized by the presence of two unduplicated centrioles and S phase cells by two closely positioned duplicated centrioles. G2 cells were identified by two separated and duplicated centrioles. Prophase cells were identified by DNA condensation within the nucleus. Various stages of mitosis were recognized by characteristic DNA morphology.

Analysis and assembly of expansion microscopy images was performed in Photoshop (Adobe) and Fiji (National Institutes of Health). To improve the visibility of the dimmer signals and when quantitative differences between fluorescent signals were not critical, the levels of fluorescent signals were sometimes differentially adjusted between different image panels. Maximum intensity projections of multiple Z slices were presented. To illustrate centriole width, length, and number, only the central Z section through a centriole was presented.

The length of acetylated signal of mother centrioles in horizontal orientation in G1, S, and G2 phases were measured along longitudinal axes at half-width half maxima on both centriole ends (Kong et al., 2020). To determine the distance between striations of expanded centrobin-rich aggregates, the distance between the fluorescent maxima between two stripes or adjacent puncta was measured. The same strategy was used to determine the distance between adjacent stripes from STED and STORM images.

Cells were fixed in 1.5% formaldehyde for 4 min at RT and postfixed in methanol for 4 min at −20°C. Cells were washed in 1×PBS for 30 min and incubated with IF buffer (1% BSA and 0.05% Tween-20) for 15 min. Samples were incubated in primary anti-CENTROB or anti-PLK4 antibodies overnight at 4°C. Following the wash in 1×PBS, samples were incubated with anti-mouse or anti-rabbit antibodies conjugated with CF647 (20042 and 20045; Biotium) diluted to 1:800 in IF buffer for STORM. Anti-mouse or anti-rabbit secondary antibodies conjugated with Abberior STAR RED (STRED-1001 and STRED-1002; Abberioir) diluted to 1:200 in IF buffer were used for STED.

Before STORM imaging, coverslips were mounted in Attofluor Cell chambers (A7816; Thermo Fisher Scientific) and layered with 100-nm tetra-spectral fluorescent microspheres (T7279; Thermo Fisher Scientific), which served as fiducial markers. Samples were imaged in STORM buffer consisting of 25 mM β-mercaptoethylamine (30070; Sigma-Aldrich), 0.5 mg/ml glucose oxidase (G2133; Sigma-Aldrich), 67 µg/ml catalase (C40; Sigma-Aldrich), and 10% dextrose (D9434; Sigma-Aldrich) in 100 mM Tris, pH 8.0. 3D STORM imaging was performed on Nikon N-STORM4.0 system using an Eclipse Ti2 inverted microscope; a super resolution high power Apo TIRF 100× NA 1.49 oil objective; 405-, 561-, and 647-nm excitation laser launch; and a back-illuminated electron-multiplying CCD camera (DU897; Andor). The 647-nm laser line (∼150 mW at the fiber and ∼90 mW before the objective lens) was used to promote fluorophore blinking. The 405-nm laser was used to activate fluorophores. The signals of fiducial markers were recorded using a 561-nm laser. 20,000 to 30,000 time points were acquired every 20 ms at a 50-Hz frame rate. NIS Elements (Nikon) was used to analyze the data.

Prior to STORM imaging, the position of the CF647-labeled target protein (PLK4 and centrobin) was recorded in a widefield mode. A rainbow Z–color-coding scheme was used for signal presentation. The signals closer to the coverslip were presented in red and those further from the coverslip in blue, typically spanning 650 nm of a working Z imaging range. The original Z–color-coding scheme was preserved on image panels. 3D STORM data are presented as a projection of the entire 3D volume.

Coverslips with immunolabeled cells were mounted in a homemade mounting medium (90% glycerol, 100 mM Tris, pH 8.0, and 0.1 mg/ml p-phenylenediamine; 695106; Sigma-Aldrich). 2D STED imaging was performed using STEDYCON (Abberior Instruments) assembled on Eclipse Ti2 inverted microscope (Nikon), 100× NA 1.45 Plan Apo objective. Avalanche photo detectors (650–700 nm, 575–625 nm, and 505–545 nm; DAPI detection) were used to detect the signals. Browser based control software (Abberior Instruments) was used to generate STED images.

RPE1 TRIM37Δ TRIM37-C18R-mNG cells grown on glass coverslips were mounted in Attofluor Cell chambers (A7816; Thermo Fisher Scientific) in CO₂ Independent Medium (Gibco). FRAP experiments were conducted on a custom-assembled workstation centered around an inverted Eclipse Ti microscope (Nikon), equipped with a back-illuminated electron multiplying CCD camera (DU888; Andor); 100× NA 1.45 Plan Apo objective, Yokogawa spinning disk (Yokogawa Electric Corporation); 405-, 488-, 561-, and 640-nm laser launch (MLC400; Agilent); and a 2× relay lens. A 488-nm photobleaching laser (OBIS 488-nm LX 30 mW; Coherent) was run at 20-Hz repetition rate. Collimated laser beam was attenuated, delivered through back epi-port of the microscope and expanded to fill the back aperture of the objective. One or two ∼200-ms laser pulses were used to bleach the fluorescent signals. Before and immediately after bleaching, fluorescence images were recorded in a confocal mode collecting 0.2-µm Z sections. Recovery of the fluorescent signal was recorded by acquiring images at 1-min time intervals.

CLEM analysis was performed as previously described (Kong and Loncarek, 2015). RPE1 TRIM37Δ TRIM37-C18R-mNG cells grown on glass coverslip were first imaged using a confocal microscope, and 200-nm-thick Z sections through the entire volume of a target cell were recorded to register the position of TRIM37-C18R-mNG containing condensates in the cell. A fixative consisting of 2.5% glutaraldehyde (G5882; Sigma-Aldrich) and 0.25% formaldehyde (15686; Electron Microscopy Sciences) in PBS (pH 7.4) was added to the cells, and the position of the condensate was recorded again after fixation. Differential interference contrast (DIC) images were acquired to determine the position of the condensate in the cell. Samples were fixed for additional 1 h at RT, washed in PBS for 30 min (10 min each wash), prestained with 1% osmium tetroxide (19100; Electron Microscopy Sciences) and 1% uranyl acetate (22400; Electron Microscopy Sciences), dehydrated in graded ethanol series, and then embedded in EMbed-812 resin (13940; Electron Microscopy Sciences). 80-nm serial sections of the target cells were sectioned, placed on the formvar-coated copper grids (2330P-XA; SPI Supplies), and further contrasted with uranyl acetate and lead citrate. Imaging was performed using a FEI Spirit transmission electron microscope, operating at 80 kV. The analysis and alignment of the serial sections and image analysis and was performed in Adobe Photoshop and Fiji.

To measure the distance between electron-dense stripes and puncta in striated and punctuated aggregates, the distance between centers of adjacent electron-dense regions were measured from electron micrographs.

For immunofluorescence, 10,000 cells per well were seeded into 96-well plates 1 d before fixation. Cells were fixed in 100 µl ice-cold methanol for 7 min at −20°C. Cells were washed twice with washing buffer (PBS containing 0.1% Triton X-100) and blocked with blocking buffer (PBS containing 2% BSA, 0.1% Triton X-100, and 0.1% NaN₃) overnight. After blocking, cells were incubated for 2 h with primary antibody in fresh blocking buffer (concentrations as indicated above). Cells were washed three times with washing buffer before 1-h incubation with the secondary antibody and DNA staining Hoechst 33342 dye. Finally, cells were washed three times with washing buffer before inspection. Images were acquired on a CV7000 spinning-disk confocal system (Yokogawa Electric Corporation) equipped with a 40× (0.95 NA) or a 60× (water, 1.2 NA) U-Plan Apo objective and a 2,560 × 2,160 pixel scientific complementary metal oxide seminconductor (sCMOS) camera (Andor). Image acquisition was performed using CV7000 software.

Live-cell imaging was performed on the CQ1 spinning-disk confocal system (Yokogawa Electric Corporation) equipped with a 40× 0.95 NA U-Plan Apo objective and a 2,560 × 2,160 pixel sCMOS camera (Andor) at 37°C and 5% CO₂. Image acquisition and data analysis were performed using CQ1 and Fiji software, respectively.

Cells were seeded into 96-well polystyrene plates at 10,000 cells/well 24 h before imaging unless indicated otherwise. Imaging conditions varied according to the experimental setup. For imaging SiR-DNA, 5 × 2–µm Z sections in the RFP or FarRed channel (25% power, 150 ms) were acquired in each field at 5-min intervals for 6–24 h. The DNA marker SiR-DNA was added 2 h before imaging at a working concentration of 0.5 µM. For live imaging of TRIM37-mNG and mRuby-MAP4-MBD, 8 × 1.2–µm Z sections in the GFP and/or RFP channel (50% power, 150 ms) were acquired in each field at 4- to 15-min intervals for 6–12 h. DMSO or centrinone treatment was conducted for three cell cycles before start of imaging, unless noted otherwise.

For immunoblotting experiments shown in Fig. 8 A, a similar number of asynchronously growing cells were harvested from 6-well dishes with 2× Laemmli SDS sample buffer at 80–90% confluence and lysed by sonication. For every sample, ∼10–30 µg protein/lane was run on Mini-PROTEAN gels (Bio-Rad) and transferred to polyvinylidene difluoride (PVDF) membranes using a TransBlot Turbo system (Bio-Rad). Blocking and antibody incubations were performed in TBS-T + 5% nonfat dry milk. Detection was performed using HRP-conjugated secondary antibodies (GE Healthcare) with WesternBright Sirius (Advansta) or SuperSignal West Femto (Thermo Fisher Scientific) substrates. Membranes were imaged on a ChemiDoc MP system (Bio-Rad).

10,000 cells per well were seeded into 96-well plates on the day before the experiment. The plate was incubated for 40 min on ice to depolymerize the microtubules. The plate was then transferred to RT, and the medium was replaced with 200 µl prewarmed (37°C) medium. After 1-min incubation, cells were fixed with 100 µl ice-cold methanol and incubated for 5 min at −20°C. Immunostaining was performed as described above.

For FISH, 250,000 cells per well were seeded into a 6-well plate with 18-mm coverslips 1 d before analysis. Cells were washed with 2 ml PBS and then fixed with Carnoy’s fixative (methanol/acetic acid = 3:1) for 15 min at RT. Plates were then stored overnight at −20°C. For each coverslip, 5 µl probe (Metasystem XCE 17 green and XCE 18 green) was applied and covered with a slide. The samples and probes were then denatured by heating the slides on a 75°C hot plate for 2 min, sealed with rubber cement, and incubated at 37°C in a humidified chamber overnight. The next day, coverslips were washed with 0.4× SCC at 72°C for 2 min, RT 2× SCC and 0.05% Tween-20 for 30 s PBS. Coverslips were then incubated with DAPI (1:1,000 in PBS) for 10 min and washed with PBS for 1 min. Each coverslip was sealed on a slide with 20 µl anti-fade solution (ProLong Gold; Thermo Fisher Scientific) and stored at 4°C in the dark. 12 × 200 nm Z sections in the DAPI (80% power, 500 ms) and GFP channel (80% power, 500 ms) were acquired with a CQ1 spinning-disk confocal system (Yokogawa Electric Corporation) equipped with a 40× 0.95 NA U-Plan Apo objective and a 2,560 × 2,160 pixel sCMOS camera (Andor). Image acquisition and data analysis were performed using CQ1 and Fiji software, respectively.

For immunoprecipitation assays, FLAG-tagged TRIM37 and Myc-tagged centrobin (see Table S2) were expressed in the FreeStyle 293-F cells (Thermo Fisher Scientific). The empty 5Myc plasmid (CS2P #17095; Addgene) or 3FLAG plasmid (p3XFLAG-CMV-7.1, E7533; Sigma-Aldrich) were used as negative controls. Cell transfection was performed using FreeStyle MAX Reagent and OptiPRO SFM according to the manufacturer’s guidelines (Thermo Fisher Scientific). 20 ml of cells at 10⁶ cells/ml were transfected with a total of 25 µg DNA constructs. 43–48 h after transfection, cells were harvested and washed with PBS. For MG132 treatment, 10 µM MG132 (M8699; Sigma-Aldrich) was added 6 h before harvesting the cells. The cells were resuspended in lysis buffer (20 mM Tris/HCl, pH 7.5, 150 mM NaCl, 1% Triton X-100, 5 mM EGTA, 1 mM DTT, 2 mM MgCl₂ and EDTA-free protease inhibitor cocktail; Roche) and lysed in an ice-cold sonicating water bath for 5 min. Crude extract was taken right after the sonication and suspended in SDS sample buffer. After 15-min centrifugation at 15,000 ×g. at 4°C, the whole-cell lysates were separated to supernatant and pellet and taken for the cosedimentation assay. The rest of supernatant was incubated with Pierce Anti-c-Myc magnetic beads (Thermo Fisher Scientific) for 2 h at 4°C. The beads were washed five times with lysis buffer and resuspended in SDS sample buffer. For immunoblotting, equal volumes of samples were run on Mini-PROTEAN gels (Bio-Rad) and transferred to polyvinylidene fluoride membranes using a TransBlot Turbo system (Bio-Rad). Blocking and antibody incubations were performed in TBS-T plus 5% nonfat dry milk or in TBS-T plus 5% BSA. Immunoblotting was performed as described above.

To detect ubiquitination of centrobin by TRIM37, FreeStyle 293-F cells were transfected as described above with DNA constructs encoding Myc-tagged centrobin, HA-tagged ubiquitin, and FLAG-tagged TRIM37, and cells were harvested after 48 h. The plasmid encoding HA-ubiquitin was a gift from Edward Yeh (University of Missouri, Columbia, MO; plasmid 18712; http://n2t.net/addgene:18712; RRID:Addgene_18712; Addgene). Cells were lysed in 20 mM Tris/HCl, pH 7.5, 150 mM NaCl, 1% Triton X-100, 5 mM EGTA, 1 mM DTT, 2 mM MgCl₂, EDTA-free protease inhibitor cocktail (Roche), and 5 mM N-ethylmaleimide, and immunoprecipitation and immunoblotting were performed as described above.

Statistical analysis was conducted using Prism v8 (GraphPad). P values were determined by t tests. Unpaired t tests assuming equal standard deviation were performed (Fig. 1, C and G; Fig. 7 D; Fig. 9 D; Fig. S1 A; Fig. S3 B; and Fig. S4 B; ns, P > 0.05; *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001).

Fig. S1 shows the centrosomal localization of TRIM37 and further characterization of the condensates in TRIM37Δ cells. Fig. S2 shows centrosomal component localization to centrobin condensates at ectopic spindle poles and nucleation activity of mitotic condensates in a microtubule repolymerization experiment. Fig. S3 shows the strategy used to engineer inducible CNTROB knockout and knockout validation by genotyping and centrosome immunofluorescence. Fig. S4 shows the immunofluorescence analysis of the inducible PLK4 knockout and of centrin in the inducible CNTROB knockout. Fig. S5 shows some additional coexpression analysis of the regulation of centrobin by TRIM37. Video 1 shows the birth of a condensate in TRIM37Δ RPE1 cells expressing the condensate marker Lig^mut-TRIM37-mNG and the microtubule-binding domain of MAP4 fused to mRuby2. Video 2 shows spindle assembly in control (USP28Δ) and TRIM37Δ RPE1 cells expressing the microtubule-binding domain of MAP4 fused to mRuby2. Video 3 shows TRIM37Δ RPE1 cells expressing an mNG fusion of WT or ligase mutant (C18R) TRIM37 and a fusion of mRuby2 with the microtubule-binding domain of MAP4. Video 4 shows chromosomes labeled with SiR-DNA in control (USP28Δ) and TRIM37Δ RPE1 cells with and without induced centrobin knockout. Video 5 shows the recovery after photobleaching of condensates in TRIM37Δ RPE1 cells expressing Lig^mut-TRIM37-mNG. Table S1 lists the human cell lines used in this study. Table S2 lists the plasmids used in this study.

We thank A. Shiau and D. Jenkins for imaging support, J. Anzola for help with coexpression analysis in FreeStyle 293F cells, and O. Shoshani for help with FISH.

This work was supported by grants from the National Institutes of Health to K. Oegema (GM074207) and A. Desai (GM074215). F. Meitinger received support from the German Science Foundation (ME 4713/1-1). M. Ohta was supported by the Japan Society for the Promotion of Science. J. Loncarek was supported by the Intramural Research Program of the National Institutes of Health, National Cancer Institute. A. Desai and K. Oegema received salary support from the Ludwig Institute for Cancer Research.

The authors declare no competing financial interests.

Author contributions: The study was conceptualized by K. Oegema, A. Desai, and J. Loncarek. F. Meitinger engineered cell lines and conducted immunofluorescence and live imaging experiments. D. Kong and J. Loncarek conducted the CLEM, expansion, STORM, and STED analyses. M. Ohta conducted coexpression-based biochemical experiments. J. Loncarek and M. Ohta conducted the photobleaching analysis. The manuscript draft was prepared by A. Desai, K. Oegema, and J. Loncarek and finalized with input from F. Meitinger, D. Kong, and M. Ohta.