Cep57 regulates human centrosomes through multivalent interactions

The Cep57-Cep63-Cep152 interaction network plays a critical role in the centrosome cycle. The observation that the Cep63/Cep152 complex forms a higher-order structure and condensates that contribute to the centrosome’s dynamic self-assembly nature (25, 26) prompted us to examine whether Cep57 also has phase-separation property. The 14 to 18 h overexpression of Cep57-mCherry in HeLa cells resulted in spherical puncta formation in the cytoplasm (Fig. 1A). Some of these puncta colocalized with microtubules. Similar phenotypes were observed when a C-terminal mEGFP or N-terminal HA was fused to Cep57 (SI Appendix, Fig. S1A). Therefore, puncta formation is less likely due to the tags. After 20 h of overexpression, these constructs decorated microtubules (SI Appendix, Fig. S1A), consistent with previously reported microtubule-binding activity (13, 27). When Cep57 expression is low, it forms cytoplasmic puncta, but when it’s high, it decorates microtubules.

To examine whether Cep57 can self-assemble, we purified superfolder GFP (sfGFP) tagged Cep57 from SF9 cells through tandem affinity purification in high salt (1 M NaCl) (Fig. 1B and SI Appendix, Fig. S1B). We observed that sfGFP-Cep57 formed spherical droplets in micron size when the salt was reduced to 200 mM without crowding reagents (Fig. 1C). The spherical morphology/boundary suggests a segregated phase of sfGFP-Cep57 condensates. Moreover, the critical concentration for phase separation is ~4 nM, close to the physiological concentration of Cep57 in cells (Fig. 1D) (32). However, sfGFP-Cep57 was prone to degradation, resulting in poor quality and thereby impeding biochemical characterization. By removing some unstructured and less conserved regions, a shorter protein construct named Cep57S was created. This construct comprises residues 56 to 264 fused to 334 to 433, exhibiting high protein quality (Fig. 1B and SI Appendix, Fig. S1C). The circular dichroism (CD) spectrum of Cep57S demonstrates that Cep57S is helical, consistent with being a coiled-coil protein (SI Appendix, Fig. S1D). The formation of Cep57S condensates occurred immediately after the reduction in salt concentration (Fig. 1E). Cy5 labeled Cep57S was highly enriched in droplets. The round shape of Cep57S condensates indicates a liquid-like state. Given that Cep57S preserved the phase separation property of the full-length protein and has a better protein quality and purity, we used Cep57S as a surrogate to characterize the biophysical properties of Cep57 phase separation.

We performed a series of image-based assays to characterize the physiochemical properties of Cep57S condensates. First, the phase transition of Cep57S was salt-sensitive. At a fixed protein concentration (8 μM), Cep57S was highly soluble in high salt (1 M to 300 mM), while droplets were visualized in low salt conditions (200, 100 mM) (Fig. 1F). Moreover, the condensate formation was reversible. Cep57S droplets formed in low salt were dissolved in high salt when protein concentration remained constant (SI Appendix, Fig. S1 E and F). Second, Cep57S displayed a concentration-dependent phase transition with a critical concentration in the submicromolar range in the presence of 200 mM NaCl (Fig. 1G). Third, Cep57S was dissolved in 1,6-hexanediol, a reagent used to test phase separation by disrupting hydrophobic interaction (SI Appendix, Fig. S1 G and H). Furthermore, Cep57S droplets coalesced and formed larger ones over time (Fig. 1H). Similar to the dynamic nature of Cep57’s interacting partner Cep63 and Cep152 in cells (26), fluorescence recovery after photobleaching (FRAP) assay showed Cep57S molecules were exchanged between the bulk environment and condensates. The younger droplets (formed for ~1 h) displayed higher dynamics when compared to the older ones (formed for ~4 h). The half-time of recovery for the younger droplets was ~230 s, while it was ~550 s for the older ones (Fig. 1 I and J). The reduction of dynamics indicates the condensate becomes hardening over time. Taken together, the formation of Cep57S assembly is salt and concentration-dependent. The morphology and dynamic properties indicate that purified Cep57S undergoes LLPS in a simplified, low-salt buffer.

Cep57 contains an N-terminal coiled-coil domain (NTD) and a C-terminal coiled-coil domain (CTD). To map the regions responsible for LLPS, we purified Cep57N (residues 56 to 251) and Cep57C (residues 334 to 433). Both fragments were highly soluble and displayed a monodisperse profile on gel filtration in 150 mM salt (SI Appendix, Fig. S2 A and B). We, therefore, speculated that both Cep57N and Cep57C are required for phase separation. To verify this, we generated a Cep57tevS construct, into which a TEV protease cleavage site was inserted between the NTD and CTD (Fig. 2A). In the phase separation assay, Cep57tevS presented similar LLPS potency to Cep57S as both constructs have similar critical concentrations (SI Appendix, Fig. S2C). However, after 6 h of incubation with TEV protease, the formation of Cep57S condensates was abrogated (Fig. 2B). The cleavage was confirmed by SDS-PAGE analysis (SI Appendix, Fig. S2D). These results suggest that the driving forces for assembling the higher-order complex of Cep57S involve the interaction between the NTD and CTD. In agreement with our speculation, the isothermal titration calorimetry (ITC) experiment showed a micromolar-ranged interaction between Cep57N and Cep57C (Fig. 2C and SI Appendix, Table S1). To assess the importance of this N–C domain interaction in LLPS, we defined the partition coefficient of Cep57 constructs as the intensity in condensate divided by that in the bulk environment. The excess of Cep57N or Cep57C decreased the partition of Cep57S in the condensates (Fig. 2 D and E), indicating Cep57N or Cep57C alone attenuates phase separation by competing for the intermolecular interactions between Cep57S molecules. A similar inhibitory effect by Cep57C was observed on sfGFP-Cep57 (SI Appendix, Fig. S2E). Hence, Cep57 and Cep57S use similar mechanisms for phase separation.

The formation of Cep57S droplets is salt dependent (Fig. 1F), implying that electrostatic interactions may play a role in phase separation. By sequence analysis, we identified a highly conserved K/R-rich motif located C terminally to the NTD (Figs. 1B and 2F). To examine whether this highly positively charged motif contributes to phase separation, we generated a Cep57S-KR mutant by replacing six charged Lys and Arg residues in this motif with Ala (Fig. 2H). This mutation significantly lowered the partition coefficient of Cep57S-KR and critical concentration for LLPS, compared with that of Cep57S (Fig. 2 I and J and SI Appendix, Fig. S2F). This mutational analysis demonstrated that this positively charged motif contributes to the phase separation of Cep57S. To map the binding site for the K/R-rich motif, we measured the affinity of the synthesized K/R peptide to different Cep57 fragments. The ITC experiments showed the K/R peptide binds to Cep57C with a K_D of 46.7 μM, while no interaction was detected between the K/R peptide and Cep57N (Fig. 2G and SI Appendix, Fig. S2G). In addition, we noticed that Cep57C exhibited a stronger inhibitory effect than Cep57N (Fig. 2 D and E) because excess Cep57C can block the NTD and K/R motif from network formation. Our data show that the NTD, K/R motif, and CTD are required for the phase separation of Cep57. Finally, we cannot rule out that the deleted regions (aa 1 to 55, 265 to 333, and 434 to 500) may contribute to phase transition, as Cep57 has a lower critical concentration for phase separation compared to Cep57S.

Coiled-coil domains often self-assemble into high-order oligomers and have been shown to promote phase separation (33). Here, we examined the oligomeric state of the NTD and CTD. By SEC-SAXS (size exclusion chromatography coupled small angle X-ray scattering) analyses, Cep57N displayed an Rg of 36.5 Å with an estimated molecular mass (Mr) around 33.4 kDa in high salt (1 M) and an Rg of 63.3 Å with an estimated Mr around 54.4 kDa in low salt (150 mM) (Fig. 3A), while the calculated size for a monomer is 26 kDa. Thus, Cep57N is a dimer in physiological conditions. Cep57C is also a dimer in 150 mM salt based on the analysis of SEC-MALS (size exclusion chromatography coupled with multiangle light scattering). The estimated Mr (25.8 kDa) is two times larger than the monomer (12.5 kDa) (Fig. 3B).

To obtain atomic structure information on Cep57, we determined a crystal structure of the Cep57C at 2.10 Å resolution in the space group of P6₄22 (SI Appendix, Table S2). The Cep57C adopts a helix–turn–helix fold, with one protomer in the asymmetric unit, forming a symmetric dimer (Fig. 3C). The ordered residues span from 354 to 428, with Helix 1 comprising residues 356 to 387, and Helix 2 comprising residues 391 to 427. The dimer interface is composed of two universally conserved F371 and F376 in Helix 1 (Fig. 3C and SI Appendix, Fig. S3A), whose benzyl rings pack well in a near 90-degree orientation with the centroid distance of 5.6 Å, a very stable π–π stacking interaction (34). To test whether dimerization is important for LLPS, we performed structure–function analyses by replacing F371 and F376 with Ala in Cep57C and Cep57S, referred to as Cep57C-FF and Cep57S-FF. SEC-MALS data showed that the size of Cep57C-FF mutant decreased to ~17 kDa, indicating it is a mixed population of dimer and monomer (Fig. 3D). Furthermore, the FF mutation reduced the thermal stability of Cep57C by 11.5 degrees (Fig. 3E), while the secondary structure remains similar to the wild-type Cep57C as monitored by circular dichroism (SI Appendix, Fig. S3B). Therefore, FF mutation weakens the dimer formation. Next, we investigated the role of CTD dimerization in phase separation. Compared with the wild-type Cep57S, the FF mutation decreased the condensate size, partition coefficient, and critical concentration for LLPS (Fig. 3 F–H and SI Appendix, Fig. S2F). Therefore, Cep57C dimerization is essential to promote Cep57 condensation.

TPX2 and tau condensates nucleate microtubule by concentrating tubulin dimers in reconstituted systems (24, 35, 36). Cep57 CTD alone binds to and bundles microtubules (27), and the depletion of Cep57 impairs spindle microtubule assembly in cells (13). These findings suggest that Cep57 condensates may have microtubule assembly activity. To test this idea, we performed a microtubule nucleation assay in BRB80 and 5% PEG3350, which prevents protein precipitation that formed upon mixing α/β-tubulin with preexisting Cep57S condensates. As illustrated in Fig. 4A, we performed the assays using OG488-taxol to mark polymerized microtubules. Microtubules emanated from Cep57S condensates immediately after adding α/β-tubulin (Fig. 4B). Over time, microtubule asters became more evident with a morphology resembling microtubule asters found in cells (Fig. 4B). To determine whether Cep57S condensates facilitate microtubule nucleation, we determined the critical tubulin concentration for spontaneous assembly, which was 5 μM in our assay system (Fig. 4C). Cep57S condensates further lowered the tubulin threshold to ~2 μM. Next, we examined the partition of tubulin dimers using Cy5-tubulin (10 μM, 5% labeled) in the presence of nocodazole, which inhibits tubulin polymerization. Cep57S condensates concentrated α/β-tubulin dimers by about 2.6-fold (Fig. 4 D and E), indicating that the tubulin concentration in condensates is above the critical concentration for microtubule assembly when the bulk tubulin concentration is 2 μM. In contrast, non-Cep57 interacting proteins sfGFP-OLA1 (78 kDa), GAPDH (144 kDa as a tetramer), and PKM2 (232 kDa as a tetramer) demonstrate a partition coefficient of ~1.1. Therefore, Cep57 condensate selectively recruits tubulin. In addition, our data reveal that the interstitial space within Cep57 condensates can accommodate molecules of at least 11 nm in size, as estimated from the crystal structure of PKM2.

Electrostatic interactions often mediate interactions between α/β-tubulin and microtubule-associated proteins (MAPs) between positively charged motifs on MAPs and negative-charged tubulin surface (37, 38). To examine whether tubulin binds to the K/R motif of Cep57, we performed a pull-down assay. His-tagged Cep57S coeluted with α/β-tubulin, whereas the K/R mutation largely diminished tubulin binding (Fig. 4 F and G). This mutational binding analysis identified the poly-K/R motif as a tubulin binding site on Cep57, in addition to the known C-terminal domain (aa 278 to 491) (27). Moreover, the poly-K/R motif is the dominant binding site for tubulin in Cep57S since the KR mutation reduced the binding by ~65%. We further tested whether tubulin binding–deficient Cep57S-KR condensates can nucleate microtubules. Compared to the tubulin alone control, Cep57S-KR did not affect microtubule assembly (Fig. 4H), and the KR mutation reduced the partition coefficient of tubulin (Fig. 4 D and E). To delineate if LLPS of Cep57S is required for microtubule nucleation, we conducted microtubule nucleation assays on an LLPS-incompetent fragment (Cep57N_long) containing only the NTD and poly-K/R motif. This fragment demonstrated no effect on microtubule assembly (Fig. 4H).

Combined, our data demonstrate that Cep57S condensates nucleate microtubules and the K/R motif contributes to tubulin recruitment. Given the functional significance, we renamed this K/R motif the LLPS and microtubule nucleation (LMN) motif.

Next, we examined the effect of the LMN motif on the microtubule assembly in cells. We performed a microtubule regrowth assay on cells whose endogenous Cep57 was knocked down by siRNA (SI Appendix, Fig. S3C). Microtubule disassembly was achieved by cold treatment in the presence of nocodazole for 1 h. After nocodazole was washed away, cells were incubated in a warm medium to allow for microtubule reassembly for 2 min. Aster formation was abolished in Cep57-depleted cells (Fig. 4 I and J). The overexpression of a siRNA-resistant Cep57 construct fully rescued this phenotype, whereas the LMN mutant failed. Through the in vitro and cellular assays, we demonstrate that Cep57 plays a role in the formation of centrosomal microtubule asters in an LMN motif-dependent manner.

The LLPS of Cep57S is driven by multiple interactions among Cep57S molecules, which may provide sites for the regulation of LLPS, activity, and expansion from the existing assembly in the centrosome cycle. Previous studies revealed that N-terminal Cep57 interacts with Cep63, and this interaction is required for the PCM localization of Cep152 (28, 39). To study the role of Cep63 in Cep57 assembly, we purified full-length Cep63 to high purity (Fig. 5A and SI Appendix, Fig. S4A). The CD analysis of Cep63 showed it is highly helical as a coiled-coil protein should be (SI Appendix, Fig. S4B). In our LLPS system, we applied the soluble D488-labeled Cep63 to the preformed condensates and observed the gradual recruitment of Cep63 to the Cep57 scaffold over time (Fig. 5 B and C). When soluble Cep57S and Cep63 in high salt were simultaneously diluted in low salt to allow cocondensation, Cep63 exhibited a high partition coefficient of 10.7 ± 1.2 (n = 166) after 6 h (Fig. 5D). Interestingly, we noticed the shape of Cep57S-Cep63 cocondensates became amorphous, implying the material property of condensates was changed to a solid-like state (Fig. 5D).

Next, we investigated how Cep63 affects the recruitment of α/β-tubulin. In the experiments, we prepared condensates 4 h before supplementing 5 μM tubulin in the presence of nocodazole. Compared to the Cep57S control, Cep63 greatly reduced tubulin concentration within the cocondensates with a partition coefficient close to 1 (Fig. 5 E and F), indicating that soluble tubulin dimers could penetrate the cocondensates, but could not be locally enriched. How did Cep63 impede tubulin concentration by Cep57S? One possibility is that Cep63 blocks the tubulin-binding LMN motif on Cep57S. However, the reported Cep63 binding motif is distant from the LMN (28), making the blocking hypothesis less likely. The other possibility is that Cep63 reduces the density of LMN motifs in the condensate so that the tubulin concentration effect is diminished. If true, Cep57S’ partition coefficient will be lowered by Cep63. To test this idea, we used the N-terminal fragment of Cep63 (Cep63N) in the phase separation assays instead of the full-length protein, whose solubility was low. Cep63N has been reported to interact with Cep57 (28). To improve the solubility of Cep63N for biophysical measurements, we created a fusion of MBP and Cep63N. Our ITC data confirmed a strong interaction between Cep57N and MBP-Cep63N (K_D = 0.6 μM, Fig. 5G). Strikingly, MBP-Cep63N dissolved Cep57S condensates as the partition of Cep57S and the droplet size reduced considerably in a concentration-dependent manner (Fig. 5 H and I and SI Appendix, Fig. S4C). In the nanometer scale, SAXS measurements reveal the internal structural change induced by MBP-Cep63N. First, the Cep57S condensate exhibits a fractal network structure (SI Appendix, Fig. S5A). The slope in the fractal region became flattened as the amount of MBP-Cep63N increased (SI Appendix, Fig. S5A), indicating the reduction of the fractal network. The time course of dissolution is about 1 h from the time-dependent scattering analysis (SI Appendix, Fig. S5B). MBP-Cep63N reduced the fractal network of Cep57S condensates in a concentration- and time-dependent manner, correlating well with the microscopic observation (Fig. 5 H and I and SI Appendix, Fig. S4C). There is a possibility that MBP may contribute to the dissolution of Cep57S condensates. However, considering all the data together, it strongly indicates that Cep63 regulates the microtubule nucleation activity of Cep57S by the impediment of LLPS via disrupting the network structure of condensates.

In mitosis, the PCM expansion is critical for bipolar spindle assembly. This process is achieved by the accumulation of scaffolding proteins. Cep63 has been reported to play a negative role in PCM maturation. Downregulation of Cep63 protein level permits PCM expansion (6). Since Cep63N led to the dissolution of Cep57S condensates, we further tested whether Cep63 represses cocondensates to recruit soluble Cep57S in the bulk environment. In effect, preexisting condensates were marked by Cy5, and the supplemented Cep57S was marked by D488 at 0.4 μM below the critical concentration to keep D488-Cep57S soluble. Images were taken 2 h after supplementing soluble D488-Cep57S into the preexisting condensates. The Cep57S assembly concentrated soluble D488-Cep57S with a partition coefficient of ~3.5 (Fig. 5 J and K). In contrast, D488-Cep57S barely accumulated at the Cep63-Cep57S cocondensates with a partition coefficient of ~1.3. Our data demonstrate that Cep63 could intervene in the Cep57S assembly, thereby suppressing its LLPS expansion. Thus, modulating LLPS by client proteins is a means to control the structural expansion and functional activity of scaffolding proteins.

Our in vitro study shows that the NTD, CTD, and LMN motif are crucial for promoting phase separation of Cep57. To investigate the significance of these multivalent interactions in the centrosomal recruitment of Cep57, we created two truncation mutations Cep57-C1 and Cep57-C2 (Fig. 6A). We used an antibody that recognizes the N-terminal region of Cep57 (residues 118 to 226) for immunofluorescence imaging assays, but not the Cep57-C1 and Cep57-C2 constructs. The expression of truncation constructs did not affect the endogenous Cep57 level as compared to the vector control (SI Appendix, Fig. S6A). However, the amount of endogenous Cep57 at centrosomes decreased, while the full-length construct increased the centrosomal Cep57 level (Fig. 6 B and C). Our data indicate that truncation mutants, which compete for the multivalent interactions of Cep57, inhibit LLPS in vitro and centrosomal recruitment in cells.

Furthermore, we investigated whether the truncation constructs have effects in centrosome duplication. HeLa cells expressing these constructs were synchronized to the M phase by thymidine and subsequent RO-3306 treatment (Fig. 6D). Cells were subjected to immunostaining 20 min after release from the G2 block. Overexpression of the empty vector control or full-length Cep57 resulted in a similar degree of centrosome amplification (Fig. 6 E and F). In contrast, overexpression of Cep57-C1, Cep57-C2, or Cep57-c.915_925dup11 [C-terminal deletion mutation found in mosaic-variegated aneuploidy (MVA) patients (29–31)] induced centrosome amplification.

Cep57 depletion causes centrosome amplification through precocious centriole disengagement in mitosis, which licenses new centrosome formation (40). This finding prompted us to investigate the role of Cep57 multivalent interactions in centriole engagement. We performed siRNA experiments against Cep57, which resulted in PCM disorganization (amorphous shape or fragmentation of PCM), centriole disengagement (lengthening the distance between the centriole pair), and centrosome amplification (more than two centrosomes in a cell) during mitosis (Fig. 6 G–J). Introducing the siRNA-resistant Cep57 construct can restore its function. In contrast, siRNA-resistant Cep57-KR and Cep57-FF showed comparable expression levels (SI Appendix, Fig. S6B) to Cep57 but failed to rescue these phenotypes to the same degree as the wild type (Fig. 6 G–J). The KR and FF mutations impair the LLPS of Cep57S in vitro and induce premature centriole disengagement during mitosis. We conclude that Cep57 maintains the PCM integrity through multivalent interactions.

Numerical and structural centrosome aberration can cause enhanced cell invasion (41, 42). To investigate the role of Cep57 LLPS in cell motility, we assessed the migration ability of U2OS cells expressing these Cep57 constructs in Boyden chambers. Although these constructs have minimal effect on cell migration, we still observed that Cep57-C1 promoted U2OS migration compared to the full length and control (SI Appendix, Fig. S6 C and D). Together, these cell assays demonstrated the biological significance of Cep57 multivalent interactions in centrosome integrity.

Our work elucidates how human PCM protein Cep57 assembles into micron-sized biomolecular condensates through liquid–liquid phase separation by employing various tools in vitro. Both purified Cep57 and Cep57S undergo LLPS, driven by at least three interacting regions, including NTD, LMN motif, and CTD (Fig. 7). The critical concentration for LLPS of purified Cep57 is around 4 nM, similar to the cellular concentration of Cep57 (32). Cellular analyses support that the multivalent interactions of Cep57 maintain the structural organization of the PCM. Two LLPS-incompetent mutants (KR and FF) result in centriole disengagement, PCM disorganization, and centrosome amplification in Cep57-depleted cells (Figs. 6 G–J and 7). It demonstrates the importance of Cep57 assembly for the integrity of human centrosomes.

Pericentrin, a critical PCM scaffold for mitotic centrosome maturation, may utilize its conserved coiled-coil and low-complexity region for phase separation in cells (43). Caenorhabditis elegans SPD-5, Drosophila CNN, and human CDK5RAP2 are functional homologs with very low sequence similarity. Interestingly, the principles governing the self-assembly of SPD-5 and CNN scaffold are extremely similar. Both use two coiled-coil domains to promote intra- and intermolecular interactions for scaffold formation (23, 44). Similarly, Cep57 undergoes LLPS through the interaction between the dimeric NTD and CTD. These scaffolding proteins utilize minimally two dimeric/oligomeric coiled-coil domains to engage intra- and intermolecularly for PCM assembly. Upon this core principle, extra motifs such as the LMN and posttranslational modifications such as phosphorylation could further enhance this assembly process with high specificity. It has been shown that phosphorylation and dephosphorylation control the CNN and SPD-5 scaffold (24, 44, 45). As coiled-coil domains often provide binding sites for other proteins, the coiled-coil mediated assembly provides a highly regulatory nature and can also function as a structural and signal hub. Here, our data suggest that Cep63 negatively controls the Cep57 assembly in a concentration-dependent manner. Cells may regulate the dynamic structural assembly of centrosomes by fine-tuning the ratio of Cep57 and Cep63. More fascinating molecular interplays between centrosomal scaffold proteins await to be discovered.

We identified a tubulin-binding LMN motif in Cep57. This motif contributes to LLPS and facilitates microtubule assembly (Figs. 2 F–J and 4). In the presence of OG488-taxol and 5% PEG3350, the critical concentration for microtubule assembly is 4 to 5 μM. The addition of Cep57S condensates lowered this critical concentration to 2 μM, while the partition coefficient of tubulin is 2.6. These findings suggest that tubulin dimers within Cep57S droplets reach a concentration (around 5.2 μM) akin to the critical threshold for spontaneous microtubule assembly. It strongly argues that Cep57 promotes tubulin assembly by locally increasing tubulin dimer concentrations above the threshold required for spontaneous microtubule assembly. Importantly, this biochemical activity observed in the reconstitution system is mirrored in cellular contexts. The LMN motif is essential for microtubule aster formation, as observed in the microtubule regrowth assay (Fig. 4 I and J).

Cell cycle transition from the G2 to M phase is accompanied by PCM expansion, characterized by the accumulation of scaffold proteins in centrosomes to boost microtubule organizing activity for bipolar spindle formation (18). Our work finds that Cep63 limits the assembly of the Cep57S scaffold. At high molar ratios, Cep63N dissolved Cep57S condensate. The fusion of an MBP tag to Cep63N increases the solubility of this construct, raising the possibility that MBP may also play a role in dissolving the droplets. Moreover, Cep63 altered the material property of Cep57S condensates from a liquid-like to a more solid-like state, as the morphology of the Cep57S-Cep63 condensates changed to amorphous (Fig. 5D). It was reported that the percolated network of SPD-5 failed to facilitate microtubule formation (24). Here, we showed that Cep63 inhibited tubulin concentration by Cep57S condensate and suppressed the self-expansion of the condensate (Fig. 5). Biochemically, LLPS confers Cep57 microtubule nucleation activity by concentrating tubulin dimers via the LMN motif. Cep63 constrains Cep57’s activity, likely by reducing the LLPS of Cep57. Thus, compositional control of the centrosome, such as changing the ratio of Cep57 and Cep63, might be a mechanism the cell uses to remodel microtubule organization globally.

In addition to the dominant γ-TuRC templated mechanism, recent evidence showed that liquid–liquid phase separation grants tubulin-associated proteins activity for microtubule nucleation. Such cases include TPX-2, BugZ, Tau, and SPD-5/TPXL-1/ZYG-9 condensates (24, 35, 36, 46). Assisted by TPXL-1 and ZYG-9, SPD-5 cocondensate can concentrate tubulin to promote microtubule aster formation (24). Analogous to TPX-2, BugZ, and Tau, our data demonstrate that the Cep57 scaffold can locally enrich α/β-tubulin dimers and catalyze microtubule nucleation. These findings suggest that multivalent interactions-mediated assembly of PCM scaffold proteins can directly support the function of the mitotic centrosome, in addition to organizing the centrosome architecture and recruiting microtubule-organizing factors. It seems that not only the assembly mechanism but the LLPS-mediated functionality of PCM scaffold proteins are evolutionarily similar from worms to humans despite low sequence similarity. Ultimately, the synergy between these microtubule nucleating PCM scaffolds and γ-TuRC makes a robust mitotic centrosome for faithful cell division.

Cep57 mutations are genetically linked to mosaic-variegated aneuploidy (MVA), a rare disease characterized by an abnormal number of chromosomes. The MVA syndrome manifests in various disorders, including skeletal anomalies, microcephaly, and childhood cancers (29, 31, 47, 48). The disease mutations on Cep57 all lead to protein truncations lacking the C-terminal coiled-coil domain (c241c>t; c.520_521delGA; c.915_925dup11) (30, 31). Cellular studies from patient-derived cells revealed that Cep57 disease mutation leads to centrosome amplification (5, 29).

Of the highest occurrence, the Cep57 disease mutation c.915_925dup11 promoted centrosome abnormalities in MVA patient-derived cells, and homozygous Cep57 ^{(c.915_925dup11/c.915_925dup11)} and heterozygous Cep57 ^{(+/c.915_925dup11)} mouse MEF cells (29). Moreover, the heterozygous mutation enhanced cancer incidence in the mouse model (29). In our cell assays, overexpression of this disease construct resulted in centrosome amplification. Based on our biochemical analyses, this construct alone cannot phase separate but can inhibit the LLPS of Cep57. Our work provides a molecular mechanism for Cep57 mutation-mediated MVA syndrome. Truncation mutations in the Cep57 scaffold disrupt the integrity of the centrosome by interfering with its multivalent interaction network. Our findings suggest LLPS is essential in human centrosomes, and crippling LLPS by mutations leads to centrosome aberration, manifested as a human genetic disease. Primary microcephaly is a genetic disorder caused by mutations in a collection of proteins involved in DNA repair and centrosome function. However, molecular mechanisms for microcephaly mutations in centrosomal proteins, such as Cep135, Cep152, and WDR62, remain clouded (49). Interestingly, the disease mutations identified in these genes all resulted in protein truncation (50–52). It is worth investigating whether these proteins assemble by multivalent interactions and, if so, what is the effect of the truncation. Our study on human PCM scaffold proteins provides insight for a better understanding of centrosome-associated diseases at the molecular level and paves the road toward disease treatment.

All the constructs were cloned from the human cDNA library into a pCool2-His, pmCherry-N1, pmEGFP, or HA vector plasmid. Cep57 and Cep63 constructs were overexpressed by Escherichia coli Rosetta 2, which were lysed by Avestin Emulsflex C3 homogenizer at 4 °C. Clarified lysate was purified by affinity chromatography (Ni-NTA agarose, Qiagen), followed by gel filtration chromatography (Superdex 200, Cytiva) in 1 M NaCl. SF9 cells overexpressing sfGFP-Cep57 were disrupted by Dounce homogenization. Clarified lysate was purified by tandem affinity purification for 6× His tag and twin-strep-tag (Strep-Tactin Sepharose resin, IBA Lifesciences). Purified proteins were concentrated, flash-frozen by liquid nitrogen, and stored at −80 °C. Protein concentrations were determined by UV absorbance at 280 nm.

The polybasic LMN peptide was synthesized by Kelowna International Scientific Inc., and dissolved in ITC buffer [150 mM potassium phosphate pH 7.5, 1 mM 2-mercaptoethanol, and 10% glycerol (w/v)]. All samples were dialyzed against the ITC buffer. The measurements were conducted at 25 °C or 15 °C as listed in SI Appendix, Table S1 (MicroCal iTC200, GE Healthcare). Curve fitting was accomplished in Origin® according to a single binding site model.

Cep57S were incubated in a Ni-NTA spin column (Qiagen) for 30 min, followed by three cycles of rebinding at 4 °C, and washed three times by binding buffer [1X PBS pH 8.0, 1 mM 2-mercaptoethanol, 20 mM imidazole, 0.1% Triton X100, and 10% glycerol (w/v)]. 0.1 mg of α/β-tubulin (Cytoskeleton, Inc.) was then incubated in the column for 30 min, followed by three cycles of rebinding at 4 °C. After five washes, proteins were eluted using the binding buffer with 400 mM imidazole and subjected to SDS-PAGE analysis.

Condensate assembly was conducted by diluting concentrated samples in the assay buffer [20 mM Tris-HCl pH 8.0, 1 mM 2-mercaptoethanol, 1 mM PMSF, and 10% glycerol (w/v)]. For microtubule assembly assays, Cep57S condensates were prepared in BRB80 buffer supplemented with 150 mM NaCl, 10% glycerol, and 5% PEG3350 at 22 °C. After 4 h of incubation, Cep57S condensates were supplemented with 1 mM Mg/GTP, 0.5 μM OG-488 taxol (Thermo Fisher Scientific), and α/β-tubulin (Cytoskeleton, Inc.). For the tubulin concentration assay, 66 μM nocodazole (Selleckchem) was added to inhibit microtubule nucleation before supplementing 5 μM Cy5-tubulins (5% labeled). Each assay was repeated three times. The size and intensity of the condensates were quantified by applying “image threshold” and “particle analysis” functions on FIJI (53). All assembly assays were repeated three times.

The FRAP measurements were conducted by “photobleaching” and “time series” functions of ZEN blue 3.1 on ZEISS LSM 800, using a 63× oil objective. The degree of photobleaching was set to 50%. After photobleaching, the images were taken every 15 s for young droplets and 30 s for old droplets. Data fitting was performed by Prism8 using a one-phase association model [% recovery = Plateau * (1 – e^−Time/Tau)].

SEC-MALS experiments were performed by using Superdex 200 Increase 5/150 GL (Cytiva) equipped with three detectors: an Agilent 1260 DAD (UV), a Wyatt DAWN-HELEOS-II detector (MALS), and a Wyatt Optilab T-Rex differential refractive index detector (RI) with a buffer containing 20 mM HEPES pH7.5, 150 mM NaCl, and 1 mM DTT. Data were analyzed by Wyatt Astra software (version 7.3.1).

SEC-SWAXS experiments were performed at the TPS 13A BioSWAXS beamline of the National Synchrotron Radiation Research Center, Taiwan (54, 55) in a buffer containing 20 mM Tris-HCl pH 8.0, 150 mM/1 M NaCl, and 1 mM 2-mercaptoethanol. The frame data of well-overlapped SAXS profiles were averaged and subtracted with buffer scattering using the TPS 13A SWAXS Data Reduction Kit (Ver. 3.6), and analyzed using ATSAS 3.1.3 (56).

The His-Cep57C crystal was grown in a screen containing 0.1 M Sodium chloride, 0.1 M HEPES pH 7.5, and 1.6 M Ammonium sulfate at 22 °C by the hanging drop vapor diffusion method. Diffraction data were collected at 100 K at beamline TLS 13B, NSRRC, Taiwan. Data were processed by HKL2000 (57). The crystal structure of Cep57C was solved by the molecular replacement program Phaser-MR in Phenix (58) using a previously determined structure (PDB ID: 4L0R) as the search model. Structural refinement and model building were carried out using Phenix (58) and Coot (59) iteratively. The structure display was created by PyMOL (Schrodinger).

Transient transfections of siRNA (MDBio, Inc.) and plasmid DNA were performed using TransIT-X2 (Mirus Bio) or FuGENE®HD (Promega). For the microtubule regrowth assay, cells were cold-treated with nocodazole, followed by microtubule reassembly. Primary antibodies are listed in SI Appendix, Table S3.

Statistical analyses for this study were conducted using Prism (GraphPad). Two-tailed student’s t test was used to compare two groups, while one-way ANOVA along with Dunnett’s multiple comparison test was used for comparing more than two groups. The data were presented as mean ± SD, and the P-values were indicated in the figure legends.

K.-T.L. and H.-C.C. designed research; H.-W.Y., P.-P.C., T.-C.Y., S.-L.L., Y.-T.C., W.-P.L., T.C., and J.M.P. performed research; L.H.-C.W., Y.-C.L., and K.-C.H. contributed new reagents/analytic tools; H.-W.Y., P.-P.C., T.-C.Y., T.C., J.M.P., K.-T.L., O.S., U.-S.J., and H.-C.C. analyzed data; and H.-W.Y., O.S., U.-S.J., and H.-C.C. wrote the paper.

The authors declare no competing interest.