Condensation of pericentrin proteins in human cells illuminates phase separation in centrosome assembly

Abstract

At the onset of mitosis, centrosomes expand the pericentriolar material (PCM) to maximize their microtubule-organizing activity. This step, termed centrosome maturation, ensures proper spindle organization and faithful chromosome segregation. However, as the centrosome expands, how PCM proteins are recruited and held together without membrane enclosure remains elusive. We found that endogenously expressed pericentrin (PCNT), a conserved PCM scaffold protein, condenses into dynamic granules during late G2/early mitosis before incorporating into mitotic centrosomes. Furthermore, the N-terminal portion of PCNT, enriched with conserved coiled-coils (CCs) and low-complexity regions (LCRs), phase separates into dynamic condensates that selectively recruit PCM proteins and nucleate microtubules in cells. We propose that CCs and LCRs, two prevalent sequence features in the centrosomal proteome, are preserved under evolutionary pressure in part to mediate liquid-liquid phase separation, a process that bestows upon the centrosome distinct properties critical for its assembly and functions.

Keywords: Cell division; Centrosome maturation; Liquid-liquid phase separation.

Fig. 1

. Endogenously GFP-tagged PCNT forms dynamic aliphatic alcohol-sensitive pericentrosomal granules during late G2/early M phases before incorporating into a largely non-dynamic mitotic PCM. (A) Time-lapse micrographs of GFP-PCNT expressed from its endogenous locus during late G2/early M phases. Arrowheads denote the fusing and splitting events of the GFP-PCNT granules. Asterisks at time 0 denote the centrosomes. Similar results were obtained from more than three biological replicates (also see Movies 1-3). (B) Quantification of PCNT granule numbers at different cell cycle stages. Data are median±first and third quartiles. n, number of cells analyzed from more than three biological replicates. Representative images are shown in Fig. S2A. (C) Time-lapse micrographs of pericentrosomal GFP-PCNT granules with or without the acute 3.5%/296 mM 1,6-hexanediol treatment. Time 0 is the time of hexanediol addition. Asterisks at time −2 min denote the centrosomes. (C′) Quantification of the results shown in C. Percentage of GFP-PCNT granules remained, with or without the acute 1,6-hexanediol treatment, as the function of time was plotted and represented as mean±95% c.i. from three biological replicates. The total number of cells analyzed for each condition is indicated. The arrowhead denotes the time of hexanediol addition (time 0). (D) Time-lapse micrographs of late G2/early M cells showing pericentrosomal GFP-PCNT granules before and after acute treatments of various aliphatic alcohols at 127 or 296 mM. Note that 296 mM (or 3.5%) of 1.6-hexanediol was also the concentration used in C and C′. Time 0 is the time of aliphatic alcohol addition. Two large red dots in each frame are the centrosomes depicted, for example, by asterisks in time −1.4 min on the top panel (127 mM 1,7-heptanediol). (E-H) Quantification of the results from the experiments shown in D. After treatments of various aliphatic alcohols at 127 mM or 296 mM, the percentage of GFP-PCNT granules remained as the function of time was plotted and represented as mean±s.e.m. from two to three biological replicates. The results after 10 min of treatment are summarized in H. The relative hydrophobicity of different aliphatic alcohols is represented as an orange gradient; the darker the color, the higher the hydrophobicity approximately is. The total number of cells analyzed for each condition is indicated. (I) FRAP analysis of endogenously GFP-tagged PCNT in RPE-1 cells during late G2/early mitosis. Only one of the centrosomes (Centrosome #2) was photobleached, and the fluorescence recovery was recorded every 5 s for ∼5 min. Dashed squares delineate centrosomes. (I′) The percentage recovery and half-life (t_1/2) after photobleaching were calculated after fitting the data with non-linear regression. Data are mean±95% c.i. n, number of centrosomes analyzed from two biological replicates. Statistical significance was determined by one-way ANOVA (B) or Student's t-test (unpaired and two-tailed) (C′,E-G). *P<0.05; **P<0.01, ****P<0.0001; ns, not significant. Scale bars: 0.5 μm (A); 5 μm (C,D); 10 μm (I); 2 μm (I, insets).

Fig. 2

. N-terminal segments of PCNT phase separate in a concentration-dependent manner in cells. (A) Alignments of 169 pericentrin orthologous proteins from vertebrates (167), fruit fly (1) and budding yeast (1), colored by the ClustalX coloring scheme in Jalview (Table S1). Conservation scores, locations of the predicted CCs, PACT motif, LCRs, putative dynein and γ-tubulin binding domains (BDs) of human PCNT, and the epitopes of the anti-PCNT antibody (Abcam, ab4448) are noted below the alignments. (B) Conservation scores within or outside of CC or LCRs of human PCNT. Data are median with the third quartile. (C) Representative time-lapse micrographs of GFP-PCNT (2-1960) condensates and GFP-PCNT (1954-3336) scaffolds 24 h post Dox induction in RPE-1 cells. Brackets denote an area with dynamic rearrangement of GFP-PCNT (2-1960) condensates; PCNT (1954-3336) scaffolds are non-dynamic (also see Movies 4, 5). (D) FRAP analyses of GFP-PCNT (2-1960) condensates and GFP-PCNT (1954-3336) scaffolds in RPE-1 cells. Dashed circles denote the bleached sites. Data are mean±95% c.i. n, number of condensates/scaffolds analyzed from more than three biological replicates. The percentage of recovery and half-life (t_1/2) after photobleaching were calculated after fitting the data with non-linear regression. (E) Quantification of relative protein concentrations in live cells expressing various GFP-tagged PCNT segments after Dox induction (see Fig. S4 for details and representative images). PCNT (2-1960) and PCNT (854-1960) phase separated after reaching their respective critical concentrations (Csats), the concentrations of the light phase at which LLPS just occurred. Csats are mean±95% c.i. n, number of cells analyzed from three biological replicates. Statistical significance was determined by the Student's t-test (unpaired and two-tailed). ****P<0.0001. a.u., arbitrary unit. Scale bars: 5 μm.

Fig. 3

. GFP-PCNT (854-1960) undergoes LLPS, coalesces and moves toward the centrosome. (A) Time-lapse micrographs of GFP-PCNT (854-1960) condensates in RPE-1 cells. Arrowheads denote the fast fusing and splitting events of the spherical condensates. Similar results were obtained from three biological replicates. (B) Time-lapse micrographs of GFP-PCNT (854-1960) expressed in RPE-1 cells stably expressing miRFP670-CETN2 (insets, magenta arrowheads denote the centrosomes). Time-lapse imaging started 3.5 h post Dox induction. The time when the first condensates formed is marked as time 0. White arrowheads denote the examples of two condensates moving around the nucleus toward the centrosome. Similar results were obtained from more than three biological replicates (also see Movies 6, 7). (C) FRAP analyses of different ages of GFP-PCNT (854-1960) condensates in RPE-1 cells. Dashed circles delineate the bleached sites. Data are mean±95% c.i. n, number of condensates analyzed from three biological replicates. The percentage of recovery and half-life (t_1/2) after photobleaching were calculated after fitting the data with non-linear regression. The highly mobile nature of young condensates prevented us from tracking the same condensates consistently beyond 5 min in the recovery phase of the FRAP assay. Scale bars: 2 μm (A); 10 μm (B); 1 μm (C, young condensates); 2 μm (C, old condensates).

Fig. 4

. GFP-PCNT (854-1960) condensates coalesce and move toward the centrosome in a dynein- and MT-dependent manner. (A-C) Quantification of the size (A), number (B), distance to the centrosome (C) and distance to the centrosome at time 0 (the start of phase separation) (D) of GFP-PCNT (854-1960) condensates over time in the cells treated with DMSO vehicle, ciliobrevin D and dynarrestin mix (50 μM each), or nocodazole (8.3 μM). GFP-PCNT (854-1960) was expressed from a Dox-inducible promoter in RPE-1 cells. Data were aligned at the onset of phase separation (time 0) of individual condensates. The fusing of PCNT (854-1960) condensates and their movement toward the centrosome were attenuated upon dynein inhibition or MT depolymerization. Initiation of phase separation also occurred closer to the centrosome with intact dynein activity and MTs (D). Data are mean±95% c.i. (A-C) or median±the first and third quartiles (D). n, number of cells (A-C) or condensates (D) analyzed from three (DMSO and dynein inhibition) or two (nocodazole) biological replicates. Statistical significance was determined by the F-test that compares the slopes of fitted lines between data sets via linear regression (A-C) or by one-way ANOVA for the distance comparison at time 0 in D. ***P=0.0001, ****P<0.0001.

Fig. 5

. GFP-PCNT (854-1960) condensates selectively recruit endogenous PCM proteins and clients. (A) Schematic of the recruitment assay to show the timeline of Dox induction and immunostaining. (B) Immunofluorescence of PCNT N-terminus (PCNT N-term), γ-tubulin, CEP215, CEP192, cytoplasmic dynein ICs (Dynein ICs), PLK1 or ribosomal protein S6 (RPS6) in RPE-1 cells after Dox induction to form GFP-PCNT (854-1960) condensates. Fold enrichment of fluorescence signals in the condensate (Con) relative to those in the cytoplasm (Cyt) was quantified. Data are mean±95% c.i. The mean of fold enrichment is noted. n, number of cells analyzed from at least two biological replicates per protein. Statistical significance was determined by Student's t-test (unpaired and two-tailed). Scale bars: 5 μm.

Fig. 6

. GFP-PCNT (854-1960) condensates nucleate MTs in cells. (A) Schematic of the MT renucleation assay to determine whether GFP-PCNT (854-1960) condensates nucleate MTs. (B) Anti-α-tubulin immunofluorescence of the cells containing GFP-PCNT (854-1960) condensates after MT renucleation in maximum intensity-projected and single optical section views. MTs were renucleated within and on the surface of the condensate (arrows). The asterisk in the merged channels denotes the centrosome in which MT renucleation was robust. Similar MT renucleation in GFP-PCNT (854-1960) condensates was also observed in live cells (see Fig. 7; Movie 8). (C) MT renucleation also occurred in small PCNT condensates (asterisks), which also recruited endogenous PCNT. (D) Quantification of α-tubulin density (intensity per area) in GFP-PCNT (854-1960) condensates (Con) and in the surrounding cytoplasm (Cyt) after MT renucleation. Data are mean±95% c.i. The mean of fold enrichment is noted. n, number of condensates analyzed from three biological replicates. Statistical significance was determined by Student's t-test (unpaired and two-tailed). a.u., arbitrary unit. Scale bars: 10 μm and 2 μm (insets).

Fig. 7

. GFP-PCNT (854-1960) condensates nucleate MTs in live cells. (A) Schematic of the MT renucleation assay in live cells. GFP-PCNT (854-1960) condensates in EB3-tdTomato-expressing RPE-1 cells were formed after Dox induction. MTs were then depolymerized by nocodazole (Noc). A pre-wash image was taken [Noc (+)], followed by time-lapse imaging at 1-min intervals after nocodazole was washed out on the microscope stage to follow MT renucleation [Noc (−)]. (B) Single optical sections of time-lapse micrographs of EB3-tdTomato-labeled MT plus ends on the surface of a GFP-PCNT (854-1960) condensate during MT renucleation. Time 0 was the time immediately after nocodazole was washed out (also see Movie 8). Insets show the EB3-tdTomato channel only. Similar results were obtained from three biological replicates. Scale bars: 10 μm.

Fig. 8

. Model for PCNT phase separation in centrosome assembly. PCNT is a linear multivalent protein that phase separates through its CCs and LCRs during its co-translational targeting to the centrosome when the nascent PCNT polypeptides are in close proximity in the polysome. The resulting PCNT granules/condensates promote PCM assembly by (1) selectively concentrating PCM proteins and clients; this will facilitate PCM assembly and limit the biochemical reactions at the centrosome (e.g. MT nucleation and kinase activities) from occurring elsewhere in the cytoplasm; and (2) enabling a liquid-to-gel/solid-like transitioning process during centrosome assembly; this process provides the PCM components with a thermodynamically favored pathway to assemble into a micron-sized membraneless, and yet spatially organized, PCM.