Calibrated mitotic oscillator drives motile ciliogenesis

See allHide authors and affiliations

Science  10 Nov 2017:
Vol. 358, Issue 6364, pp. 803-806
DOI: 10.1126/science.aan8311

Taming mitosis for differentiation

The mitotic oscillator consists of molecular switches known to drive cell division forward. This conserved clocklike regulatory circuit has not previously been implicated in cellular processes other than division. Multiciliated cells generate motile cilia-powered flows that are essential for brain, respiratory, and reproductive functions. Al Jord et al. found that the mitotic oscillator was activated in a calibrated fashion in terminally differentiating progenitors of multiciliated cells (see the Perspective by Levine and Holland). The oscillator function was used to drive massive production of cilia-nucleating centrioles while avoiding mitotic commitment. Thus, mammalian postmitotic progenitors can recruit and calibrate the mitotic oscillator to impose timing and directionality of cellular differentiation instead of proliferation.

Science, this issue p. 803; see also p. 716


Cell division and differentiation depend on massive and rapid organelle remodeling. The mitotic oscillator, centered on the cyclin-dependent kinase 1–anaphase-promoting complex/cyclosome (CDK1-APC/C) axis, spatiotemporally coordinates this reorganization in dividing cells. Here we discovered that nondividing cells could also implement this mitotic clocklike regulatory circuit to orchestrate subcellular reorganization associated with differentiation. We probed centriole amplification in differentiating mouse-brain multiciliated cells. These postmitotic progenitors fine-tuned mitotic oscillator activity to drive the orderly progression of centriole production, maturation, and motile ciliation while avoiding the mitosis commitment threshold. Insufficient CDK1 activity hindered differentiation, whereas excessive activity accelerated differentiation yet drove postmitotic progenitors into mitosis. Thus, postmitotic cells can redeploy and calibrate the mitotic oscillator to uncouple cytoplasmic from nuclear dynamics for organelle remodeling associated with differentiation.

Motile cilia of multiciliated cells generate streams of vital fluids that promote respiratory, reproductive, and brain functions (1). Defects in motile ciliation or cilia number cause life-threatening diseases by perturbing cilia-based flow (14). However, the events leading to motile ciliogenesis and the mechanisms regulating cilia number are poorly explored.

Organelle biogenesis and remodeling are common to cell division and differentiation. Centrioles are microtubule-based organelles that are duplicated in dividing cells to provide daughter cells with a full centrosome (5). In differentiating multiciliated cells, centrioles are amplified to nucleate patches of 30 to 300 cilia (69). The mitotic machinery (fig. S1A) participates in centriole number control (10, 11) and regulates centriole growth (12) and dynamics (5, 1316) in dividing cells. How these steps, essential for motile ciliogenesis and cilia number control, are regulated in multiciliated progenitors is unknown. Centriole amplification follows a typical spatiotemporal pattern (9), prompting an analogy with centriole duplication. We therefore investigated if multiciliated progenitors implemented molecular mechanisms of cell division to orchestrate subcellular remodeling associated with differentiation.

Centriole amplification in mouse-brain progenitor cells is marked by three sequential phases delimited by two switchlike transitions (fig. S1B) (9). The A to G transition borders the amplification (A) phase, during which procentrioles form around centrosome-derived deuterosome platforms, and the growth (G) phase, during which all procentrioles grow synchronously from these platforms. At the A to G transition, centrosomal centrioles separate and deuterosomes disperse along the nuclear membrane to assume a perinuclear distribution (fig. S2, A to D). The G to D transition borders the G and disengagement (D) phase, during which maturing centrioles detach from their growing platforms in a wavelike manner, the primary cilium resorbs, and microtubules reorganize (fig. S2, D to H, and movie S1). Disengaged centrioles then migrate and dock apically to become basal bodies and nucleate motile cilia (movie S1). This precise spatiotemporal pattern of centriole amplification (fig. S2I) mirrors centriole dynamics of dividing cells, where duplicated centrosomes separate along the nuclear membrane at the G2 to M transition of the cell cycle before disengaging at the metaphase to anaphase transition (5).

The G2 to M and metaphase to anaphase transitions are spatiotemporally coordinated by the mitotic oscillator, centered on the cyclin-dependent kinase 1–anaphase-promoting complex/cyclosome (CDK1cyclin B1-APC/CCDC20) axis (cyclin B1 and CDC20 are mitotic coregulators) (1719). By immunostaining differentiating progenitors in vitro and in vivo, we observed the expression of mitosis-specific regulators (17) and the presence of mitotic phosphorylations (20) (Fig. 1 and figs. S3 to S6), suggesting a transient exit from quiescence and mitosis machinery reactivation. The sequential profile of mitotic phosphorylations (Fig. 1C) suggested the involvement of G2 to M and metaphase to anaphase regulators in A to G and G to D transitions, respectively. To confirm the sequential activity of mitosis regulators in differentiating progenitors, we performed short incubations with pharmacological modulators of the CDK1-APC/C axis (fig. S1C). We used phosphorylated histone H3 [pH3 (pSer10)] and phosphorylated vimentin [pVim (pSer55)] as mitotic machinery–activity readouts. Inhibition of the G2 to M CDK1 inhibitors Wee1-like protein kinase (WEE1) and membrane associated tyrosine/threonine kinase 1 (MYT1) (17) with PD 166285 increased pVim phosphorylation in G- and D-phase cells (Fig. 2A and fig. S7, A to C). Massive pH3 phosphorylations also appeared in G- and D-phase cells, some of which showed prophase-like chromosome condensations (Fig. 2, A and B, and fig. S7C, red dots). Coinhibiting WEE1-MYT1 and APC/C with PD 166285 and proTAME (21) further augmented pVim and pH3 phosphorylations and increased prophase-like figure incidence in D-phase cells (Fig. 2, A and B, and fig. S7, A to C). Metaphase- and anaphase-like figures appeared in this condition (Fig. 2, B and C). Inhibiting APC/C alone with proTAME increased pH3 and pVim phosphorylations and triggered prophase- to anaphase-like figures only in D-phase cells (Fig. 2, A to C, and fig. S7, A to C). Differentiating progenitors with mitosis-like events presented hyperphosphorylations comparable to those observed during mitosis in cycling progenitors (Fig. 2D and fig. S7D). Cotreating cells with the CDK1 inhibitor RO-3306 abolished hyperphosphorylations and mitotic events (fig. S7, C and E). Thus, CDK1 is controlled and calibrated by WEE1-MYT1 at the A to G transition and by APC/C at the G to D transition, confirming the proposed parallel between mitotic G2 to M–metaphase to anaphase transitions and multiciliated A to G–G to D transitions.

Fig. 1 Mitosis machinery is transiently activated in differentiating multiciliated progenitors.

(A and B) Immunostaining profiles of CDK1, p27KIP1 (a CDK inhibitor), pH3, and pVim in cultured CEN2-GFP+ cells at different stages of cilia-nucleating centriole formation during postmitotic multiciliated-cell differentiation. (C) Horizontal bars illustrate immunoreactivity of mitosis markers relative to differentiation stages. C, centrosome stage (black); A phase, centriole amplification phase (gray); G phase, centriole growth phase (orange); D phase, centriole disengagement phase (green); and MBB, multiple basal body stage (black). The nucleus is outlined with dashed lines, and “x” marks GFP aggregates. Scale bars, 2 μm.

Fig. 2 CDK1-APC/C axis is conserved in differentiating multiciliated progenitors and can trigger mitosis.

(A) Representative immunoreactivities of pH3 and pVim in A-, G-, and D-phase CEN2-GFP+ progenitors incubated for 3 hours with dimethyl sulfoxide (DMSO) (control); PD 166285 with or without proTAME; or proTAME alone. Prophase-like nuclei are outlined in yellow. DAPI, 4′,6-diamidino-2-phenylindole. (B) Quantifications of prophase- and metaphase-like chromosome condensations in the same conditions. In cells with metaphase-like figures, centriole phase is nonidentifiable. (C) Representative images of metaphase- and anaphase-like chromosome condensations in differentiating progenitors with pharmacologically induced mitosis. (D) pH3 quantifications in differentiating progenitors with pharmacologically induced mitosis compared to control cycling or differentiating progenitors. Error bars represent mean ± SD. (E) CEN2-GFP+ and histone H2B-RFP+ dynamics during proTAME-induced mitosis starting from G phase (RFP, red fluorescent protein). Mitosis entry (prophase-like nucleus) occurs during D phase (13:20) followed by metaphase-like (13:30 to 14:00) and anaphase-like (14:20) states. Arrowheads indicate micronuclei after mitotic exit (16:20). Time in hours:min; scale bars, 2 μm. *P = 0.0402; **P > 0.0018; ***P = 0.0009; ****P < 0.0001; ns, not significant.

To validate the dormant mitotic capacity of differentiating progenitors, we live imaged centriole and chromosome dynamics with APC/C inhibition. Differentiating cells entered mitosis after the G to D transition, as suggested by immunostainings (Fig. 2E; fig. S8, A and B; and movie S2). Mitosis was characterized by cyclin B1 accumulation and degradation; nuclear envelope breakdown and reformation; and unduplicated chromosome condensation and segregation (figs. S8, B to F; S9; and S10, A and B). Mitosis duration depended on proTAME concentration (fig. S10C), as in cycling cells (21). Exit from the metaphase-like state and karyokinesis completion suggested that APC/C inhibition was transient, as in cycling cells (21, 22), and that CDK1 hyperactivation eventually reactivated APC/C. Co-incubating cells with proTAME and the synergistic APC/CCDC20 inhibitor, apcin (22), increased mitotic entry and blocked mitotic exit (fig. S10, C to F). Thus, CDK1-APC/C–dependent molecular switches, which drive each stage of mitosis forward, are conserved in postmitotic progenitors of multiciliated cells. However, physiological CDK1 activity is actively restrained, consequently preventing mitosis.

To test if CDK1 controls centriole amplification dynamics, we live imaged centrin 2–green fluorescent protein (CEN2-GFP+) progenitors and pharmacologically targeted mitosis regulators. Inhibiting CDK1 or PLK1 (polo-like kinase 1) (17) hindered A to G and G to D transitions (Fig. 3A and fig. S11, A and B) and impeded multiciliated-cell formation (fig. S12A). The A to G transition was delayed and the A phase prolonged (Fig. 3A and movie S3). Because procentrioles are generated exclusively during the A phase (9), we investigated if displacing the A to G transition affects centriole number. Inhibiting CDK1 or PLK1 increased deuterosome number per cell and subsequent centriole number (Fig. 3B and fig. S12, B to D). It also prevented centrosomal centriole distancing and deuterosome dispersion around the nucleus (fig. S13). The G to D transition was also delayed and failure to initiate D phase led to deuterosome regrouping, suggesting an arrest of the centriole dynamic (Fig. 3A and fig. S14, A and B). In cells passing the G to D transition, D phase decelerated and centriole disengagement and migration were incomplete (Fig. 3A, fig. S14C, and movie S4), leading to partial motile ciliation (fig. S14, C to E). Conversely, inhibiting WEE1-MYT1 with PD 166285 accelerated A to G and G to D transitions and multiciliated-cell formation (Fig. 3A, figs. S11B and S12A, and movie S5). Because PD 166285 can affect other kinases, we inhibited WEE1 with MK 1775 and observed comparable tendencies (fig. S12A). Acceleration of the A to G transition decreased deuterosome, centriole, and motile cilia number (Fig. 3B and fig. S12, D to H). Centrosomal centriole distancing, deuterosome dispersion, and motile ciliation were comparable to controls (figs. S13 and S14, C to E), suggesting that accelerated transitions led to efficient centriole growth, disengagement, and docking. CDK1 activity is therefore fine-tuned to control centriole number, growth, and disengagement for accurate motile ciliation while avoiding mitotic commitment.

Fig. 3 CDK1-APC/C axis tunes centriole number and controls centriole amplification in postmitotic differentiating multiciliated progenitors.

(A and C) Box (25 to 75%) and whisker (10 to 90%) plots of A (gray), G (orange), and D (green) phase durations in differentiating CEN2-GFP+ progenitors incubated 22 to 24 hours with DMSO (controls); RO-3306, BI 2536 (PLK1 inhibitor), and PD 166285; and proTAME with or without apcin. Lines indicate medians, and crosses indicate means. (B) Quantification of final centriole number in differentiating CEN2-GFP+ progenitors at the D phase after 72 hours of incubation with DMSO (control), CDK1 inhibitors (RO-3306 and CGP74514A), or BI 2536; or after 24 hours of incubation with DMSO (control) or PD 166285. Error bars represent mean ± SD. (D and E) Mitosis machinery is active, yet calibrated, in terminally differentiating progenitors of multiciliated cells to orchestrate the massive production of cilia-nucleating centrioles (D) while avoiding the mitotic commitment threshold (E). Deregulating the calibration alters centriole number, maturation, and motile ciliation and can drive the differentiating progenitor into abnormal mitosis. *P = 0.018; **P < 0.006; ***P ≤ 0.0009; ****P < 0.0001; ns, not significant.

Finally, we monitored centriole dynamics with disrupted APC/C activity (fig. S11, C and D). ProTAME with or without apcin treatments did not affect the A to G transition (A-phase length, deuterosome number, and centriole dynamics; Fig. 3C; figs. S12, B and C, and S13; and movie S6). The G to D transition was significantly affected. It was slightly delayed, and cells that failed to initiate D phase regrouped their deuterosomes (Fig. 3C and fig. S14B). As expected, some cells passing the G to D transition underwent mitosis (fig. S11, C and D, and movie S6). In cells with a D phase spared from mitotic entry, D-phase duration increased in a dose-dependent and synergistic manner (Fig. 3C and movie S6) and led to partial motile ciliation (fig. S14, D and E), as with CDK1 and PLK1 inhibition. This suggests that CDK1 activates APC/CCDC20. Together with PLK1, APC/C triggers the G to D transition and controls synchronous centriole disengagement required for functional migration, docking, and ciliation. In return, APC/C dampens CDK1 activity, thus preventing mitosis in differentiating progenitors.

This study reveals that the CDK1-APC/C mitotic oscillator is summoned after progenitor cell division to drive terminal differentiation instead of proliferation (Fig. 3D). Postmitotic progenitors redeploy the robust mitotic clocklike regulatory circuit to drive the orderly progression of centriole production—number control, growth, and disengagement—and provide multiciliated cells with a sized patch of centrioles competent for motile ciliation. This finding aids understanding of the development of multiciliated cells and motile cilia-powered flows crucial for organism homeostasis (1).

Although centrosome duplication in cycling cells is coupled to cell division (1016), multiciliated progenitors dampen the mitosis machinery to drive centriole dynamics but avoid nuclear division (Fig. 3E). This is consistent with studies in mammalian cycling cells showing that CDK1 couples nuclear and cytoplasmic events at mitotic entry (23, 24) and with studies in Drosophila showing the uncoupling of nuclear-cytoplasmic events by experimentally dampening CDK1 (2527). Thus, calibration of the mitosis machinery to uncouple cytoplasmic from nuclear processes exists physiologically in mammalian cells. This mechanism can be used by postmitotic progenitors to impose timing and directionality in the control of cytoplasmic events such as organelle remodeling associated with differentiation. By contrast, this kind of calibration could allow cycling cells to undergo pathological centriole amplification linked to cancer and microcephaly (5, 2830).

Supplementary Materials

Supplementary Text

Materials and Methods

Figs. S1 to S14

References (3144)

Movies S1 to S6

References and Notes

  1. Acknowledgments: We thank all members of the Spassky laboratory for comments and discussions. We thank A. Aguilar, P. Bastin, M. Bornens, B. Durand, O. Gavet, and M. Piel for their comments on the project. We thank X. Morin for the pCAAGS-H2B-RFP, CMV-CEN2-TagRFP, and pCAAGS-CEN2-GFP-mCherry plasmids and T. Caspary for the Arl13b-GFP plasmid. We thank N. Menezes for her contribution on the image analysis bioinformatic scripts; F. Delestro Matos from the IBENS Bioinformatics Platform for collaborating on the conception and design of illustrations; A.-K. Konate and R. Nagalingum for administrative support; the IBENS Animal Facility for animal care; and Le Service des Animaux Transgéniques (SEAT) (UMS3655, PFEP, Institut Gustave Roussy, Villejuif, France) for generating the Arl13b-GFP mouse strain. We thank the IBENS Imaging Facility, with grants from Région Ile-de-France (NERF 2011-45), Fondation pour la Recherche Médicale (FRM) (DGE 20111123023), and Fédération pour la Recherche sur le Cerveau–Rotary International France (2011). The IBENS Imaging Facility and the team received support from Agence Nationale de la Recherche (ANR) Investissements d’Avenir (ANR-10-LABX-54 MEMO LIFE, ANR-11-IDEX-0001-02 PSL* Research University). A.K. and J.S.-T. are supported by UPMC, INSERM, and GEFLUC. The Spassky laboratory is supported by INSERM, CNRS, l’École Normale Supérieure (ENS), ANR (ANR-12-BSV4-0006), European Research Council (ERC Consolidator grant 647466), FRM (FRM20140329547), Cancéropôle Ile-de-France (2014-1-PL BIO-11-INSERM 12–1), and Fondation Pierre-Gilles de Gennes (FPGG03). A.M. is funded by Association pour la Recherche sur le Cancer (ARC PJA-20131200184) and ANR (ANRJC JC-15-CE13-0005-01). A.A.J. received fellowships from the French Ministry of Higher Education and Research, FRM (FDT20150531994), and Labex MEMOLIFE. All data are in the manuscript and the supplementary materials. The authors declare no competing financial interests.
View Abstract

Navigate This Article