Asymmetric Centriole Numbers at Spindle Poles Cause Chromosome Missegregation in Cancer

Article

Asymmetric Centriole Numbers at Spindle Poles Cause Chromosome Missegregation in Cancer Graphical Abstract

Authors Marco R. Cosenza, Anna Cazzola, Annik Rossberg, ..., Christel Herold-Mende, Yannick Schwab, Alwin Kra¨mer

Correspondence [email protected]

In Brief Extra centrosomes are frequent in human cancers and cause chromosome missegregation via clustering into a pseudo-bipolar mitotic spindle array. Cosenza et al. now demonstrate that centriole rosettes, a transient stage of extra centrosome formation, drive chromosome missegregation in addition to centrosome clustering and are frequently found in primary tumors.

Highlights d

Asymmetric bipolar rosette mitoses induce chromosome missegregation

d

Asymmetric rosette mitoses cause unequal chromosome capture from the two spindle poles

d

Centrosome aberrations in primary malignancies often consist of centriole rosettes

Cosenza et al., 2017, Cell Reports 20, 1906–1920 August 22, 2017 ª 2017 The Author(s). http://dx.doi.org/10.1016/j.celrep.2017.08.005

Cell Reports

Article Asymmetric Centriole Numbers at Spindle Poles Cause Chromosome Missegregation in Cancer Marco R. Cosenza,1 Anna Cazzola,1 Annik Rossberg,1 Nicole L. Schieber,2 Gleb Konotop,1,3 Elena Bausch,1 Alla Slynko,4 Tim Holland-Letz,4 Marc S. Raab,1,3,5 Taronish Dubash,6 Hanno Glimm,6 Sven Poppelreuther,7 Christel Herold-Mende,8 Yannick Schwab,2 and Alwin Kra¨mer1,5,9,* 1Clinical Cooperation Unit Molecular Hematology/Oncology, German Cancer Research Center (DKFZ) and Department of Internal Medicine V, University of Heidelberg, 69120 Heidelberg, Germany 2Cell Biology and Biophysics Unit, European Molecular Biology Laboratory (EMBL), 69117 Heidelberg, Germany 3Max-Eder Group Experimental Therapies for Hematologic Malignancies, German Cancer Research Center (DKFZ) and Department of Internal Medicine V, University of Heidelberg, 69120 Heidelberg, Germany 4Division of Biostatistics, German Cancer Research Center (DKFZ), 69120 Heidelberg, Germany 5Department of Internal Medicine V, University of Heidelberg, 69120 Heidelberg, Germany 6Department of Translational Oncology, National Center for Tumor Diseases (NCT) and German Cancer Research Center (DKFZ), 69120 Heidelberg, Germany 7Carl Zeiss Application Center, 69120 Heidelberg, Germany 8Experimental Neurosurgery, Department of Neurosurgery, University of Heidelberg, 69120 Heidelberg, Germany 9Lead contact *Correspondence: [email protected] http://dx.doi.org/10.1016/j.celrep.2017.08.005

SUMMARY

Chromosomal instability is a hallmark of cancer and correlates with the presence of extra centrosomes, which originate from centriole overduplication. Overduplicated centrioles lead to the formation of centriole rosettes, which mature into supernumerary centrosomes in the subsequent cell cycle. While extra centrosomes promote chromosome missegregation by clustering into pseudo-bipolar spindles, the contribution of centriole rosettes to chromosome missegregation is unknown. We used multi-modal imaging of cells with conditional centriole overduplication to show that mitotic rosettes in bipolar spindles frequently harbor unequal centriole numbers, leading to biased chromosome capture that favors binding to the prominent pole. This results in chromosome missegregation and aneuploidy. Rosette mitoses lead to viable offspring and significantly contribute to progeny production. We further show that centrosome abnormalities in primary human malignancies frequently consist of centriole rosettes. As asymmetric centriole rosettes generate mitotic errors that can be propagated, rosette mitoses are sufficient to cause chromosome missegregation in cancer. INTRODUCTION Chromosomal instability, a condition characterized by high rates of chromosome gains and losses during cell division, is a hallmark of cancer (Hanahan and Weinberg, 2011; Lengauer et al., 1997). Anaphase-lagging chromosomes, caused by unresolved

merotelic kinetochore-microtubule attachments, are a leading source of chromosome missegregation with consequential aneuploidy and clonal heterogeneity observed in the majority of human tumors (Cimini et al., 2001; Thompson and Compton, 2008). Mechanistically, the most prominent cause of chromosome missegregation is the presence of supernumerary centrosomes (Vitre and Cleveland, 2012). Centrosomes are the major microtubule-organizing centers in mammalian cells, and they consist of a pair of centrioles embedded in pericentriolar material (PCM) (Bettencourt-Dias and Glover, 2007). Centrosomes duplicate once per cell cycle, with the newly formed daughter centrioles undergoing maturation that lasts almost two cell cycles. Therefore, each cell contains three generations of centrioles, an older and a younger mother centriole referred to as parental centrioles, and two daughter centrioles, one produced by each parent at the beginning of S phase. Only the old mother centriole is structurally fully mature and decorated with subdistal and distal appendages (Kong et al., 2014). Centrosome abnormalities are found in the vast majority of cancers where their presence correlates with karyotype abnormalities and poor prognosis (Go¨nczy, 2015; Lingle et al., 2002; Neben et al., 2003; Nigg, 2002; Pihan et al., 1998). Several studies in flies and mice provide evidence that extra centrosomes promote tumorigenesis (Basto et al., 2008; Coelho et al., 2015; Levine et al., 2017; Sabino et al., 2015; Serc¸in et al., 2016). Apart from inducing chromosome missegregation, centrosome abnormalities have been proposed to contribute to tumorigenesis in additional ways, including defects in asymmetric cell division and promotion of invasion (Basto et al., 2008; Godinho et al., 2014). In cells with extra centrosomes, lagging chromosomes and chromosome missegregation can result from transient multipolar spindle intermediates in which merotelic attachment errors accumulate before centrosomes are clustered into a pseudo-bipolar spindle array (Ganem et al., 2009; Silkworth et al., 2009). In

1906 Cell Reports 20, 1906–1920, August 22, 2017 ª 2017 The Author(s). This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).

B

A

G1

S

Centrin

Pericentrin

Merge

M

I

II Older motherr Younger mother Daughter

III

D

EM slice 2

EM slice 3

EM slice 4

6

Cell 2

6

5

4

Cell 3

Right pole

Centrin α-Tubulin

Cell 1

C

Segmentation EM slice 1

3

Cell 4

2

4

Cell 5

Left pole

2

4

5

(legend on next page)

Cell Reports 20, 1906–1920, August 22, 2017 1907

principle, extra centrosomes can arise through centriole overduplication or through the accumulation of mature centrosomes by aborted cell division, cell fusion, or centrosome clustering (Nigg, 2002). The pattern produced by these two mechanisms is different. Centriole overduplication initially leads to the formation of centriole rosettes, comprised of mother centrioles surrounded by numerous daughters, which remain engaged and function as single units within their first cell cycle and give rise to autonomous extra centrosomes only thereafter (Ganem et al., 2009; Habedanck et al., 2005; Kleylein-Sohn et al., 2007). On the other hand, aborted cell division, cell fusion, and centrosome clustering lead to the accumulation of multiple mature mother centrioles (Ganem et al., 2009; Silkworth et al., 2009). It is unknown whether, in addition to centrosome clustering, other mechanisms contribute to chromosome missegregation in cancer cells with extra centrosomes. Also, the causes of centrosome abnormalities as well as their composition in primary human tumor cells remain to be elucidated. Here we have identified a novel mechanism by which centriole rosettes cause chromosome missegregation via asymmetric chromosome capture in cell lines and primary tumors, and we provide evidence for centriole overduplication as a frequent origin of centrosome aberrations in primary cancer samples. RESULTS Bipolar Rosette Mitoses Induce Lagging Chromosomes and Chromosome Missegregation Mitoses immediately ensuing centriole overduplication are bipolar with a centriole rosette at one or both spindle poles (Ganem et al., 2009), referred to as rosette mitoses below (Figure 1A). To directly test the hypothesis that bipolar rosette mitoses contribute to chromosome missegregation, centriole overduplication was induced in U2OS cells by conditional expression of PLK4, the principal kinase that regulates centriole duplication (Habedanck et al., 2005; Kleylein-Sohn et al., 2007), or STIL, a protein involved in centriole duplication downstream of PLK4 (Arquint et al., 2012; Tang et al., 2011; Vulprecht et al., 2012). As expected, 15 hr after induction most spindles in GFP-PLK4-U2OS cells were bipolar with a centriole rosette at one or both poles (Figures S1A and S1B) (Ganem et al., 2009; Kleylein-Sohn et al., 2007). Compared to GFP-PLK4, GFP-STIL expression led to fewer cells with centriole overduplication and centriole rosettes with fewer daughters both in interphase and mitosis (Figure S2).

Sufficient amounts of rosette mitoses were available for analysis at 15 and 72 hr after the induction of GFP-PLK4 and GFP-STIL, respectively. At these time points, which were chosen for further experiments, rosette mitoses could be clearly identified by immunofluorescence staining for pericentrin and centrin (Figure 1B), as corroborated by Airyscan superresolution confocal microscopy as well as correlative light and focused ion beam-scanning electron microscopy (CLEM/FIB-SEM) with subsequent 3D image reconstruction (Figures 1C and 1D; Movie S1). Indeed, by CLEM/FIB-SEM all spindle poles identified were constituted of single rosette-type centrosomes with multiple daughter centrioles engaged to one mother, in all mitotic cells analyzed. In mitoses with extra spindle poles, centrosome clustering after transient spindle multipolarity causes chromosome missegregation by promoting merotelic kinetochore mis-attachments (Ganem et al., 2009; Silkworth et al., 2009). However, despite the presence of only two spindle poles, bipolar rosette mitoses displayed significantly increased frequencies of anaphase-lagging chromosomes in both GFP-STIL-U2OS and GFP-PLK4U2OS cells (Figures 2A and 2B). Analogous data were obtained when spontaneous rosette mitoses from unperturbed breast and prostate cancer cell lines or non-induced GFP-STIL-U2OS and GFP-PLK4-U2OS cells were examined, confirming that this effect is due to the presence of centriole rosettes and not a consequence of PLK4/STIL overexpression (Figure 2B). Livecell imaging experiments using GFP-PLK4-U2OS cells led to similar results (Figure 2C). To determine whether the observed increase in lagging chromosomes indeed reflects chromosome missegregation, we analyzed chromosome segregation among daughter cells by telophase centromere fluorescence in situ hybridization (FISH) in cytokinesis-blocked GFP-PLK4-U2OS cells (Figures 2D and 2E) (Fenech, 2007; Zijno et al., 1994). Missegregation rates of the chromosomes examined were increased and mirrored the results obtained by scoring lagging chromosomes. These findings indicate that centriole overduplication increases the rate of chromosome missegregation via anaphase-lagging chromosomes irrespective of the formation of extra spindle poles. Centriole Rosettes Induce Increased Nucleation of Centrosomal Microtubules To ascertain how rosette mitoses affect chromosome segregation fidelity, we next analyzed whether supernumerary centrioles impact on microtubule assembly. For that, microtubules were

Figure 1. Overexpression of PLK4 and STIL Leads to the Formation of Rosette Mitoses (A) Formation of daughter centriole rosettes around parental centrioles. These rosettes remain engaged and function as single units in the first mitosis after their formation (I). In the subsequent cell cycle (II), rosette centrioles disengage and themselves bud multiple daughters. Resulting mitoses are either multipolar or clustered. Only the progeny of clustered mitoses (III) is expected to harbor multiple mature mother centrioles. Proteins whose presence was used to characterize older mothers, younger mothers, and daughter centrioles are given in the table below. GT335, antibody to polyglutamylated tubulin. (B) Prophase (upper panel), metaphase (middle panel), and anaphase (lower panel) bipolar U2OS cells with one centriole rosette per spindle pole 15 hr after the induction of GFP-PLK4 expression, immunostained for centrin and pericentrin. Centriole rosettes are shown enlarged in insets. Scale bar, 5 mm. (C) Airyscan superresolution confocal microscopy images of a mitotic, bipolar U2OS cell with one centriole rosette at each spindle pole 15 hr after the induction of GFP-PLK4 expression, immunostained for centrioles and microtubules. Scale bar, 5 mm. (D) Segmentation and representative FIB-SEM (EM) slices of centriole rosettes found at the 2 spindle poles of five mitotic GFP-PLK4-U2OS cells 15 hr after transgene induction. In the segmentation images, mother centrioles are pseudo-colored in green and daughter centrioles in pink. Numbers of centrioles per rosette are denominated. All volumes were aligned in 3D and computationally sliced to display mother and daughter centrioles in the same orientation. In some instances, reslicing resulted in-out-of-volume pixels displayed in grey (see Methods section). Scale bar, 300 nm. See also Figures S1 and S2 and Movie S1.

1908 Cell Reports 20, 1906–1920, August 22, 2017

A

Centrin

30

Lagging chromosomes (%)

25

**

CREST

30

n=149 n=826

Lagging chromosomes (%)

B

Pericentrin

***

20 15 10 5 0

25

DAPI

n=830 n=148

***

n=126

**

n=117

20 15

n=145

Normal Rosette

10 5 0

STIL Tet+

PLK4 Tet+

C

STIL Tet-

D

PLK4 Tet-

DU-145

MDA-MB-231

E

40

6

**

30 25 20 n=331

10 5

Chromosome 2 Chromosome 3

Chromosome missegregation rate (%)

Lagging chromosomes (%)

n=315

35

15

Merge

PLK4 Tet-

PLK4 Tet+

n>100 per condition PLK4 TetPLK4 Tet+ PLK4 Tet+ ***

5 4

*

3 2 1 0

0

PC-3

2 2 Chr.

3 3 Chr.

Figure 2. Centriole Rosettes Induce Lagging Chromosomes and Chromosome Missegregation (A) Bipolar mitosis with centriole rosettes at spindle poles and lagging chromosome 15 hr after the induction of GFP-PLK4 expression, immunostained for centrin, pericentrin, and centromeres (CREST antibody). Spindle poles and lagging chromosome (arrow) are shown enlarged in insets. Scale bar, 5 mm. (B) Percentage of normal and rosette mitoses with lagging chromosomes during bipolar anaphase in GFP-STIL-U2OS (STIL) and GFP-PLK4-U2OS cells (PLK4) 72 and 15 hr after transgene induction, respectively (left panel). In the right panel, results for GFP-STIL-U2OS and GFP-PLK4-U2OS cells without transgene induction and unperturbed DU-145 and PC-3 prostate cancer as well as MDA-MB-231 breast cancer cell lines are shown. Data are means ± SEM from 3 independent experiments. n, total number of anaphase cells counted. (C) Live-cell imaging analysis of anaphase-lagging chromosome frequency in bipolar mitoses of GFP-PLK4-U2OS cells transiently transfected with H2BmCherry, starting 15 hr after transgene induction. Only cells without evidence of multipolar spindle intermediates were scored. Data are means ± SD from 3 experiments. n, total number of mitoses analyzed. (D) Centromere FISH with probes for chromosomes 2 and 3 of binucleated U2OS cells 15 hr after the induction of GFP-PLK4 expression and treatment with cytochalasin B to abolish cytokinesis. Scale bar, 10 mm. (E) Frequency of cytochalasin B-treated GFP-PLK4-U2OS cells with an unequal number of chromosome 2 or 3 FISH signals in the 2 sister nuclei as a direct measure of chromosome missegregation. Data are means ± SD from 3 independent experiments. Tet, tetracycline; *p < 0.05, **p < 0.01, ***p < 0.001.

Cell Reports 20, 1906–1920, August 22, 2017 1909

(legend on next page)

1910 Cell Reports 20, 1906–1920, August 22, 2017

depolymerized by cold treatment in U2OS cells conditionally expressing GFP-STIL for 72 hr, allowed to repolymerize, and subsequently counted. As expected (Godinho et al., 2014; Lingle et al., 2002), extra centrioles in interphase GFP-STIL-U2OS cells nucleated significantly more microtubules than non-induced isogenic control cells, with a linear correlation between the number of centriolar signals and the number of microtubules nucleated (Figures 3A, 3B, and S3A). Moreover, the total microtubule polymer mass as well as the number of EB1-microtubule plus end-tracking protein complexes as a readout for growing microtubule plus ends (Matov et al., 2010) was significantly increased in U2OS cells expressing GFP-STIL (Figures 3C, 3D, and S3B), indicating that interphase cells with extra centrioles contain an increased amount of microtubules. In GFP-PLK4-U2OS cells at 15 hr after transgene induction, extra centrioles almost exclusively occurred as newly generated daughters within centriole rosettes. Importantly, in such rosettes the number of nucleated microtubules linearly correlated with the number of daughter centrioles as well (Figures 3E and 3F). Similar results were obtained when synchronized GFP-PLK4U2OS cells were analyzed (Figure S3C). Also, the amount of PCM associated with interphase centriolar rosettes proportionally increased with the number of daughter centrioles present per rosette (Figures 3G and 3H). These findings suggest that daughter centrioles contribute to microtubule nucleation and supernumerary daughters impact on microtubule numbers nucleated by centriole rosettes during interphase. Asymmetric Rosette Mitoses Display Unequal Capture of Chromosomes from the Spindle Poles Immunofluorescence staining revealed that 60%–70% of rosette mitoses in U2OS cells overexpressing GFP-STIL or GFP-PLK4 bore an asymmetric number of daughters engaged to the two parental centrioles (Figure 4A), a finding confirmed by CLEM/ FIB-SEM with 4 asymmetric rosette mitoses of 5 cells analyzed (Figure 1D). Strikingly, in GFP-PLK4-U2OS cells, only rosette mitoses with an asymmetric number of daughter centrioles at the two poles displayed an increased lagging chromosome frequency (Figure 4B). Loss of daughter centrioles by depletion of centriole replication proteins can result in asymmetric mitoses, with two centrioles at one pole and only one centriole at the other. This results in an imbalance in microtubule stability between the 2 half-spindles due to asymmetric localization of centrobin, a daughter centriole-specific protein that stabilizes spindle pole microtu-

bules (Jeffery et al., 2010; Kitagawa et al., 2011; Tan et al., 2015; Zou et al., 2005). Similarly, in GFP-PLK4-U2OS mitoses, the amount of centrobin correlated with the number of daughter centrioles per rosette (Figures 4C and 4D). This finding led us to hypothesize that the relationship between daughter centriole numbers and microtubule nucleation found in interphase might have an impact in mitosis as well. Indeed, automated tracing of microtubules (Weber et al., 2012) in CLEM/FIB-SEM images of an asymmetric GFP-PLK4-U2OS rosette mitosis revealed more kinetochore microtubule bundles associated with the larger pole rosette (Figure 4E; Movie S2). However, as this technique is too complex to be applied to a significant number of cells, we probed directly the ability of asymmetric rosettes to capture chromosomes. During early mitosis, microtubule arrays emanating from spindle poles probe the space to make contact to kinetochores. Due to the stochastic nature of these search-and-capture interactions, sister kinetochores are not captured simultaneously. Instead, 1 of the 2 sister kinetochores attaches first, leading to poleward movement of the mono-oriented chromosome (Rieder and Salmon, 1998). Using Airyscan superresolution microscopy, we therefore determined the numbers of chromosomes captured by the 2 spindle poles in GFP-PLK4-U2OS early prometaphases following depolymerization of non-kinetochore microtubules. In asymmetric rosette mitoses, the spindle pole with the larger rosette captured more chromosomes, whereas no pole-to-pole bias was found in cells containing rosettes with identical daughter centriole numbers (Figures 4F and 4G), further supporting our findings for lagging chromosomes (Figure 4B). Bipolar Rosette Mitoses Do Not Interfere with Spindle Assembly Checkpoint or Kinetochore-Microtubule Attachment Stability and Do Not Alert the Error Correction Machinery Anaphase-lagging chromosomes are a consequence of merotelic kinetochore orientation, which is not detected by the spindle assembly checkpoint (SAC) (Cimini et al., 2001; Thompson and Compton, 2008). Therefore, if rosette mitoses cause lagging chromosomes by the induction of merotely, then mitosis duration should not increase in cells harboring centriole rosettes. Indeed, mitosis was not delayed in GFP-PLK4-U2OS cells 15 hr after transgene induction (Figure 5A). To further exclude that anaphase-lagging chromosomes are caused by disturbed kinetochore-microtubule attachment stability (Bakhoum et al.,

Figure 3. Extra Centrioles Lead to Increased Nucleation of Centrosomal Microtubules (A) Correlation between centriole numbers and centrosomal microtubule numbers in interphase GFP-STIL-U2OS cells 72 hr after transgene induction. (B) Centrosomal microtubule numbers in GFP-STIL-U2OS cells without and 72 hr after transgene induction. (C) Microtubule polymer mass from sum-intensity projections of z stack images from interphase GFP-STIL-U2OS cells 72 hr after transgene induction. (D) EB1 dot numbers from maximum-intensity projections of z stack images from interphase GFP-STIL-U2OS cells 72 hr after transgene induction. (E) Microtubule nucleation assay with microtubule regrowth (EB1; insets, centrioles) in GFP-PLK4-U2OS cells for 20 s after cold depolymerization. Centriole and microtubule numbers are given. Scale bars, 5 mm. (F) Correlation between centriole numbers and centrosomal microtubule numbers in interphase GFP-PLK4-U2OS cells 15 hr after transgene induction. (G) Immunofluorescence microscopy image of a U2OS cell metaphase harboring 1 centriole rosette per spindle pole 15 hr after the induction of GFP-PLK4 expression, immunostained for centrin and pericentrin. Centriole rosettes are shown enlarged in insets. Scale bar, 5 mm. (H) Correlation between centriole numbers and centrosomal pericentrin signal intensity in interphase GFP-PLK4-U2OS cells 15 hr after transgene induction. Pericentrin signal intensity was determined from sum-intensity projections of z stack images. All data are means ± SEM. Tet, tetracycline; ***p < 0.001. See also Figure S3.

Cell Reports 20, 1906–1920, August 22, 2017 1911

A

B 20

70

18 Lagging chromosomes (%)

80

STIL

50

PLK4

40 30 20 10

n=378

16 14 12

n=138

n=405

10 8 6 4 2

0

0 Asymmetric

Symmetric

Normal

Symmetric Asymmetric

D

C Centrobin

60

Merge

PLK4 Tet+

PLK4 Tet-

CP110

50 Centrobin intensity

Mitoses (%)

60

Figure 4. Asymmetric Rosette Mitoses Display Unequal Capture of Chromosomes from the Spindle Poles

***

n=170 r=0.64 p<0.00001 Mean

40 30 20 10 0

2

3

4 5 6 Centriole number

E Left half-spindle

Right half-spindle

MT=513 Centrioles=5

MT=290 Centrioles=4

F

G

MT=42 Centrioles=4

MT=78 Centrioles=6

***

Captured chromosomes (ratio)

Centrin -Tubulin CREST

Normal centrosomes

Symmetric rosettes

2009a), we assayed for the presence of cold-stable kinetochore microtubules using a previously described approach (Lampson and Kapoor, 2005). The amounts of cold-stable kinetochore microtubules were similar in induced and non-induced GFPSTIL-U2OS cells with bipolar mitotic spindle arrays (Figures 5B and S4A), indicating that microtubule-kinetochore attachment stability is not affected by centriole rosette formation. BubR1 is required for the regulation of kinetochore-microtubule attachments as well as SAC activation (Lampson and Kapoor, 2005).

1912 Cell Reports 20, 1906–1920, August 22, 2017

Asymmetric rosettes

7

(A) Percentage of bipolar rosette mitoses with identical (symmetric) or different (asymmetric) numbers of centrioles per pole in U2OS cells 72 hr after the induction of GFP-STIL (STIL) or 15 hr after the induction of GFP-PLK4 (PLK4) expression. At least 100 cells were counted per condition. (B) Percentage of U2OS cells with symmetric or asymmetric centriole rosettes at spindle poles that exhibit lagging chromosomes during bipolar anaphase 15 hr after the induction of GFP-PLK4 expression. The percentage of GFP-PLK4-U2OS cells with normal centriole content and lagging chromosomes is given for comparison. Data are means ± SEM from 3 independent experiments. n, total number of anaphase cells counted. (C) Representative images of mitotic U2OS cells without (upper panel) and 15 hr after the induction of GFP-PLK4 expression (lower panel), immunostained for CP110 and centrobin. Spindle poles are shown enlarged in insets. Scale bar, 5 mm. Tet, tetracycline. (D) Correlation between centriole numbers and centrosomal centrobin signal intensity in mitotic GFP-PLK4-U2OS cells 15 hr after transgene induction. Spindle pole centrobin signal intensity was determined from sum-intensity projections of z stack images. (E) Segmentation of centriole rosettes (mother centrioles, green; daugther centrioles, pink) and microtubule bundles (yellow) from FIB-SEM images of a GFP-PKL4-U2OS cell rosette mitosis 15 hr after transgene induction. Centriole and microtubule (MT) numbers are given. Scale bars, 500 nm. (F) Representative Airyscan superresolution confocal microscopy maximum-intensity projection image of a GFP-PLK4-U2OS rosette prometaphase cell 15 hr after transgene induction, immunostained for centrioles (centrin), microtubules (a-tubulin), and centromeres (CREST) after depolymerization of non-kinetochore microtubules. Centriole and microtubule (MT) numbers per half-spindle are given. Scale bar, 5 mm. (G) Ratios of chromosomes captured by the 2 halfspindles of GFP-PLK4-U2OS early prometaphase cells with normal centrosome content or symmetric or asymmetric centriole rosettes 15 hr after transgene induction. The box plot diagram represents ratios of chromosomes captured. ***p < 0.001. See also Movie S2.

Accordingly, BUBR1 staining intensities at kinetochores were unchanged by the induction of GFP-STIL expression in bipolar mitotic U2OS cells as well (Figures 5C and S4B). The microtubule-depolymerizing kinesins MCAK and KIF2B increase microtubule turnover at kinetochores during mitosis to correct chromosome malorientations (Bakhoum et al., 2009b). In cancer cells where chromosome missegregation originates from hyperstable kinetochore-microtubule interactions (Bakhoum et al., 2009a) or increased microtubule assembly rates

A 400

B

n=115 n=91

-Tubulin intensity

350

Mitosis duration (min)

Figure 5. Bipolar Rosette Mitoses Do Not Interfere with Spindle Assembly Checkpoint or Kinetochore-Microtubule Attachment Stability and Do Not Alert the Error Correction Machinery

5

300 250 200 150 100

4 3 STIL Tet2

STIL Tet+ n>50 per condition

1

50

0

0

PLK4 Tet-

PLK4 Tet+

Metaphase

D 350

STIL TetSTIL Tet+ n>40 per condition

300 250 200 150 100 50 0 Prophase

Prometaphase Metaphase

1.8 1,8 1.6 1,6 1.4 1,4 1.2 1,2 1.0 1 0.8 0,8 0.6 0,6 0.4 0,4 0.2 0,2 0.0 0

Lagging chromosomes (%)

C

BUBR1 intensity

Prometaphase

n=493

Not transfected MCAK KIF2B n=153

n=543

**

**

**

Normal centrosomes

Symmetric rosettes

(Ertych et al., 2014), overexpression of MCAK or KIF2B reduced the incidence of lagging chromosomes. In contrast, chromosome missegregation was not corrected by MCAK or KIF2B overexpression in asymmetric GFP-PLK4-U2OS rosette mitoses 15 hr after transgene induction, whereas spontaneously arising lagging chromosomes in non-induced cells and symmetric GFP-PLK4-U2OS rosette mitoses became significantly reduced (Figures 5D and S4C). These findings suggest that, different from previously described causes of mis-attachments, erroneous attachments induced by centriolar rosettes cannot be corrected by increasing microtubule-kinetochore turnover. Bipolar Rosette Mitoses Contribute to the Production of Progeny from Cells with Extra Centrosomes To determine whether bipolar rosette mitoses contribute to progeny production of cells overexpressing PLK4, GFP-PLK4U2OS cells were immunostained up to 8 days after transgene induction for proteins marking consecutive centriole maturation steps. Immunostaining was performed for centrin and CP110 as early markers of centriole biogenesis (Chen et al., 2002; Kleylein-Sohn et al., 2007; Paoletti et al., 1996); CEP152 and glutamylated tubulin, which accumulate at younger mothers (Bobinnec et al., 1998; Fu et al., 2016; Wolff et al., 1992); and CEP170 and ODF2, which decorate appendages of fully mature mother centrioles only (Guarguaglini et al., 2005; Ishikawa et al., 2005; Nakagawa et al., 2001) (Figure 1A). If rosette mitoses contribute to progeny production in this system, then a significant fraction of the progeny of GFP-PLK4-U2OS cells should harbor multiple parental centrioles with only a single fully mature one (Figure 1A). Supporting this idea, we found that, among the cells with supernumerary centrioles, >80% contained only 1 CEP170-positive centriole even 8 days after continuous

Asymmetric rosettes

(A) Mitosis duration in GFP-PLK4-U2OS cells transiently transfected with H2B-mCherry without and 15 hr after transgene induction. Mitosis duration was measured from chromosome condensation to telophase. Red lines represent the median. (B) Quantification of cold-stable kinetochore fibers in GFP-STIL-U2OS cells without and 72 hr after transgene induction. Microtubule polymer mass was determined from sum-intensity projections of z stack images. (C) BUBR1 signal intensity at its best of focus of z stack images from mitotic GFP-STIL-U2OS cells 72 hr after transgene induction. (D) Percentage of bipolar mitotic U2OS cells transiently transfected with GFP-MCAK (MCAK) or GFP-KIF2B (KIF2B) with normal centrosomes or symmetric or asymmetric centriole rosettes at spindle poles that exhibit lagging chromosomes during bipolar anaphase 15 hr after the induction of GFP-PLK4 expression. Results are expressed as percentages of control. All data are means ± SEM. Tet, tetracycline; **p < 0.01. See also Figure S4.

GFP-PLK4 induction (Figures 6A, 6B, and S5A–S5C). At least 60% of these cells harbored >2 CEP152 signals. As CEP152 is loaded only late onto the centriole periphery after passage through mitosis (Fu et al., 2016), these findings are consistent with bipolar mitoses with a single centriole rosette per spindle pole directly accounting for the generation of a large portion of the progeny of GFP-PLK4-U2OS cells (Figure 1A). Consistent with these results, the relative frequency of multipolar and clustered mitoses dropped over time, whereas normal and rosette mitoses became more abundant (Figure 6C). To rule out that the majority of GFP-PLK4-U2OS cells with extra centrioles contain only 1 CEP170 signal because the presence of a mature mother suppresses the maturation of additional centrioles, centriole maturation was examined in cytokinesis-blocked GFP-PLK4-U2OS cells. As about 80% of binucleated cells contained >1 mature mother centriole, maturation of overreplicated centrioles after cell division is not inhibited in this system (Figure S5D). To visualize directly the fate of the progeny from different mitosis types, we tracked conditional HA-PLK4-HeLa cells that stably express the centriole marker Dendra-centrin2 by longterm live-cell imaging. All progeny of multipolar cells died after the second round of mitosis at the latest (Figure 6D), in accordance with earlier findings (Ganem et al., 2009). Surprisingly, a large part of the offspring of clustered mitoses died within two rounds of division as well. However, bipolar mitoses, comprising rosette mitoses in about 70% of cases, led to predominantly viable offspring, further indicating that bipolar rosette mitoses do significantly contribute to the progeny of PLK4-overexpressing HeLa cells. To determine the long-term contributions of rosette, clustered, and multipolar mitoses to the proliferation of HA-PLK4-HeLa

Cell Reports 20, 1906–1920, August 22, 2017 1913

B

A Centrin

CEP170

Merge

100 90 80

Centrin

CEP152

Cells (%)

70

Merge

60

1 x old mother

50

>2 x parents

40 30

n>100 per condition

20 10 0 1

2

3

4

6

8

Time (days)

C

D 45 35

Fate of progeny (%)

Metaphases (%)

40 Normal

30

Rosettes

25

Clustered

20

Multipolar

15

n>100 per condition

10 5 0 3 1

4 2

5 3

6 4

100 90 80 70 60 50 40 30 20 10 0

Death Successful division

P1 P2

P1 P2

P1 P2

Bipolar

Clustered

Multipolar

Time (days)

6e+7 2e+7

4e+7

bipolar clustered multipolar

0e+0

Number of cells

Proportion of cells

0.0 0.2 0.4 0.6 0.8 1.0 1.2

E

0

10

20

30

40

50

0

Generations

10

20

30

40

50

Generations

Figure 6. Bipolar Rosette Mitoses Contribute to the Production of Progeny (A) GFP-PLK4-U2OS cells 5 days after transgene induction, immunostained for centrin, CEP170, and CEP152. Scale bar, 2 mm. (B) Percentage of GFP-PLK4-U2OS cells with supernumerary centrioles that harbor 1 CEP170 (old mother centriole) and >2 CEP150 (parental centrioles) signals at the given time points after transgene induction. (C) Percentage of metaphase GFP-PLK4-U2OS cells with normal bipolar, bipolar rosette-type, clustered, or multipolar spindles at the given time points 48 hr after transgene induction. Cells were co-immunostained with antibodies to centrin and pericentrin. Data are means ± SD from 3 independent experiments. (D) Percentage of progeny from bipolar, clustered, and multipolar divisions of HA-PLK4-HeLa cells that underwent successful division. P, progeny generation. (E) Example simulation of proportions (left panel) and numbers of cells produced (right panel) by bipolar, clustered, and multipolar mitoses during the first 50 generations, as determined by branching process modeling. See also Figure S5.

cells, branching process modeling was performed, based on the relative frequencies and survival outcomes of different mitosis types from the HA-PLK4-HeLa cell experiment depicted in Figure 6D. In this model, the proportions of bipolar, clustered, and multipolar mitoses reached an equilibrium within a few

1914 Cell Reports 20, 1906–1920, August 22, 2017

cell cycles at 93.3%, 2.3%, and 4.4%, respectively (Figure 6E, left panel). In line with these proportions, HA-PLK4-HeLa cell populations with overreplicated centrioles grew due to successful bipolar rosette mitoses, whereas multipolar and clustered divisions were selected against (Figures 6E, right panel, and S5E).

B

A n=7

n=9

100

100

90

90

80

80

70

70

60

60

50

50

40 30 20

30 20

n>50 cells per sample

10 0

40

Median 1 x old mother 2 x parents

Acute Acute AML ALL myeloid lymphoid leukemia leukemia

Centriole rosettes

Multiple MM myeloma

10 GlioGBM

blastoma

Colon CC cancer

Additional types of extra centrioles

Old mother Young mother Daughter

1 x old mother 2 x parents

>1 x old mother and/or >2 x parents

Centrin

CEP170

Merge Merge

Centrin

CP110

Merge

Centrin

CEP170

Merge

CP110

GT335

Merge

Centrin

CEP152

Merge

Centrin

CP110

Merge

Centrin

CEP170

Merge

Centrin

CEP152

Merge

Normal

n=10

0

Acute myeloid leukemia

n=9

Extra centrioles with 2 parents (%)

Extra centrioles with 1 old mother (%)

n=35

C Pericentrin

Merge

Clustered

Glioblastoma

Rosette

Normal

Centrin

D

E

Mitoses (%)

100 90 80 70 60 50 40 30 20 10 0

n=104

Rosette

Clustered

Multipolar

Mother centriole Daughter centriole

(legend on next page)

Cell Reports 20, 1906–1920, August 22, 2017 1915

Fitting the prediction of this model, recent data suggest that each tumor has an intrinsic centrosome number distribution that reflects a dynamic equilibrium between ongoing overproduction and selection against cells with extra centrosomes (Wong et al., 2015). Aberrant Centrosomes Frequently Consist of Centriole Rosettes in Primary Human Malignancies To examine whether centriole rosettes occur in primary tumor cells, samples from patients with solid tumors and hematologic malignancies were immunostained for centriole maturation markers. In all tumor types examined, many tumor cells with multiple extra centrioles contained only 2 centrioles with parental markers (CEP152) and 1 mature mother centriole decorated with an appendage marker (CEP170). This pattern is compatible with the presence of two bona fide centriole rosettes arranged around 1 old and 1 young mother (Figures 7A, 7B, and S6). Multiple myeloma, glioblastoma, and colon cancer samples additionally contained tumor cells with centrosome aberrations characterized by multiple parental centrioles with or without appendages, compatible with the presence of multiple mature centrosomes and later maturation stages in fractions of these tumor types. These findings suggest that the majority of centrosome aberrations in the primary tumor types examined constitute centriole rosettes representing early stages of centriole overduplication (Figures 1A and 7A, bottom panel). However, especially in solid tumors (multiple myeloma, glioblastoma, and colon cancer), other types of centrosome aberrations are also found, which likely originate from the accumulation of mature centrioles by centrosome clustering or aborted cell division. Most mitoses in early passage (p2–4) primary glioblastoma cells with supernumerary centrioles were bipolar with a single centriole rosette at each pole (Figures 7C and 7D). We conclude that extra centriole signals in primary human malignancies often represent daughter centriole rosettes around a pair of parental centrioles, resulting in the frequent formation of rosette mitoses. DISCUSSION We have shown here that, in addition to centrosomal clustering (Ganem et al., 2009; Silkworth et al., 2009), bipolar rosette mitoses with unequal numbers of centrioles per spindle pole contribute to chromosome missegregation due to asymmetric

microtubule nucleation and unequal capture of chromosomes from spindle poles (Figure 7E). Also, centriole rosettes as well as rosette mitoses are frequently found in primary tumor cells. Generation of a single daughter per parental centriole is thought to be ensured by localized stabilization of PLK4 after interaction with STIL (Ohta et al., 2014). Overexpression of either of the 2 proteins deregulates this mechanism and results in centriole rosette formation (Arquint et al., 2012; Habedanck et al., 2005; Kleylein-Sohn et al., 2007; Tang et al., 2011; Vulprecht et al., 2012). Interestingly, activation of PLK4 is delayed at younger mother centrioles compared to the older ones (Sillibourne et al., 2010): PLK4 activation occurs in G1/S and G2 phases at older and younger mother centrioles, respectively, in accordance with the observation that daughter centriole assembly is initiated at the older mother centriole first (White et al., 2000). These findings might explain our observation of the frequent presence of asymmetric numbers of daughter centrioles at the 2 rosettes of bipolar mitotic spindles. Centrosomes are not required for cell division, but alterations of centrosome numbers, including acentriolar spindles and multipolar mitoses, cause increased chromosome missegregation (Ganem et al., 2009; Silkworth et al., 2009; Sir et al., 2013). Even the loss of a single daughter centriole in normal bipolar spindles leads to imbalanced kinetochore-microtubule attachment stability and aberrations in cytokinesis (Tan et al., 2015). Now, our results show that extra daughter centrioles at spindle poles, particularly if asymmetric, also result in anaphase-lagging chromosomes and chromosome missegregation. Our immunofluorescence quantification experiments reveal that daughter centrioles contribute to the recruitment of PCM proteins involved in microtubule nucleation both in interphase and mitosis, in line with what has been described for de-novoformed centrioles within the first cell cycle of their generation (La Terra et al., 2005). Each extra daughter centriole seems to linearly add to the total microtubule mass nucleated by centriole rosettes. Consequently, during rosette mitosis, the 2 spindle poles display asymmetric microtubule nucleation if they bear unequal numbers of daughter centrioles. Accordingly, asymmetric centriole rosettes generate a bias in kinetochore capture that favors binding to the spindle pole with more daughter centrioles. This imbalance is likely at the basis of the observed increase in anaphase-lagging chromosomes and chromosome missegregation induced by asymmetric rosette mitoses.

Figure 7. Aberrant Centrosomes in Primary Tumor Cells (A) Percentage of primary tumor cells with multiple centrioles that harbor only 2 centrioles with parental markers (CEP170/ODF2) and 1 mature mother centriole (CEP150/GT335). The numbers of tumor samples examined per tumor type are given (n). The multiple myeloma samples analyzed were Salmon and Durie stage IIA (n = 1), stage IIIA (n = 7), and unknown (n = 2); the colon cancer samples were UICC stage II (n = 1), stage III (n = 1), stage IV (n = 6), and unknown (n = 1), respectively. The bottom panel exemplifies the composition of centriole rosettes versus other, more mature types of centrosome aberrations. See also Figure 1A. (B) Bone marrow mononuclear cells from a healthy individual (upper row), primary acute myeloid leukemia (middle panel), and early passage primary glioblastoma cells (lower panel) immunostained for centrin, CP110, CEP170, glutamylated tubulin (GT335), and CEP152. Scale bars, 2 mm. (C) Representative images of mitotic early passage glioblastoma cells with a normal (upper panel), rosette-type (middle panel), and clustered (lower panel) centrosome configurations, immunostained for centrin and pericentrin. Insets show spindle poles. Scale bar, 5 mm. (D) Percentage of metaphase early passage glioblastoma cells with supernumerary centrioles that contain bipolar rosette-type, bipolar clustered, or multipolar spindle arrays. Cells were co-immunostained with antibodies to centrin and pericentrin. Data are means ± SD from 2 independent experiments. (E) Centriole rosettes cause chromosome missegregation via the generation of 2 asymmetric spindle poles, which produce unbalanced numbers of microtubules on mitotic half-spindles. This spindle asymmetry causes kinetochore-microtubule attachment errors by favored binding of microtubules from the more prominent spindle pole. See also Figure S6.

1916 Cell Reports 20, 1906–1920, August 22, 2017

Anaphase-lagging chromosomes are caused by unresolved merotelic attachments, in which a single kinetochore simultaneously contacts microtubules from both spindle poles. Merotelic attachments are a serious danger for chromosome segregation fidelity, as they are not detected by the SAC (Cimini et al., 2001). Most merotelic attachments are corrected before anaphase onset by KIF2B and MCAK, which stimulate detachment of maloriented microtubule fibers from kinetochores, permitting a new round of search and capture (Bakhoum et al., 2009b). However, in our experiments, overexpression of KIF2B or MCAK was not effective in lowering the lagging chromosome frequency in rosette mitoses. This can be explained once the asymmetric rosette binding bias is taken into account, as the corrective action of releasing maloriented kinetochore fibers has no effect on asymmetric microtubule dynamics and will likely result in another biased capture from the more prominent spindle pole. Therefore, asymmetric rosette mitoses can have a strong impact on chromosome segregation fidelity, because the mitotic machinery is unable to efficiently correct the orientation errors induced by unequal microtubule numbers from spindle poles. Recent data have demonstrated that high levels of chromosomal instability compromise viability and multipolar mitoses lead to cell death or cell-cycle arrest (Ganem et al., 2009, 2014; Weaver et al., 2007). Here we show that rosette mitoses, on the other hand, produce viable offspring, likely due to the induction of only moderate chromosome missegregation. In accordance with the literature, multipolar mitoses were negatively selected. However, live-cell fate tracking of mitotic progeny showed that centrosome clustering was also characterized by decreased offspring viability, although to a lesser extent than multipolar mitoses. Possibly, survival after centrosomal clustering requires adaptation to the clustering process to grant survival of the offspring, as accumulated extra centrosomes can lead to cell-cycle arrest (Ganem et al., 2014). In contrast, the continued proliferation of rosette mitosis offspring allows chromosomal aberrations to be propagated within the population. Accordingly, mathematical modeling based on measured cell division type outcomes demonstrates that the burden imposed by centriole overduplication does not ultimately hinder population growth, even in cell lines like HeLa, which do not tolerate centrosome amplification. Therefore, centriole rosettes could represent an important contributor to chromosome missegregation in early cancer stages. It will be interesting to analyze the respective contributions of rosettes and clustered mitoses to tumorigenesis in vivo. Clearly, multipolar cell divisions are long known to occur in many types of cancers. However, a detailed structural or functional characterization of centrosomal amplification pattern is missing, because almost all studies that have examined centrosome aberrations in primary tumor tissues rely on single-antigen immunostainings, mostly against the PCM proteins pericentrin or g-tubulin. Many studies have noted a positive correlation between centrosome size and karyotype abnormalities in several tumor types, yet the mechanistic basis for this correlation has remained unclear (Lingle et al., 2002; Neben et al., 2003; Nigg, 2002; Pihan et al., 1998). As we showed that centriole rosettes lead to expanded PCM signals, these findings might be explained by the model presented here. One study has shown

that centriole overduplication with rosette formation occurs in primary human papillomavirus-associated anal neoplasms (Duensing et al., 2008). Interestingly, although later stages of centrosome aberrations with more than two older mother centrioles resulting from centrosome accumulation were more frequent in this solid tumor type, only centriole rosettes resulting from early-stage centriole overduplication correlated with cell division errors. In line with these results, the HPV16-E7 oncoprotein was found to rapidly trigger the formation of centriole rosettes by centriole overduplication, with subsequent cell division errors (Duensing et al., 2007). These and our present findings argue for a contribution of rosette-type divisions to the development of chromosome missegregation. Whereas many acute leukemias harbor normal karyotypes, the vast majority of multiple myelomas, glioblastomas, and colon cancers show high levels of chromosomal aberrations (Mitelman et al., 2017). In acute leukemia samples, cells containing centrosome aberrations other than rosettes were rare, suggesting that most of the progeny from cells with centrosome aberrations is not viable in these disorders, thereby providing a possible explanation for their comparatively stable karyotypes. On the other hand, multiple myeloma, glioblastoma, and colon cancer samples contained larger fractions of cells with accumulated parental centrioles, suggesting the occurrence of productive divisions of cells with centriole rosettes and in accordance with the well-established higher frequency of karyotype abnormalities in these tumors. Taken together, these data highlight an additional aspect of centrosome composition in mitosis: both mother and daughter centrioles fine-tune spindle microtubule dynamics and thereby ensure mitotic fidelity. Therefore, tight regulation of centriole duplication is important for symmetric spindle formation to allow for efficient kinetochore-microtubule attachment correction and, thus, protect from chromosome missegregation. EXPERIMENTAL PROCEDURES Primary Tumor Samples Informed consent was obtained according to the Declaration of Helsinki. The study was approved by the institutional review board. Bone marrow mononuclear cells were isolated from aspirates by Ficoll density gradient centrifugation. For multiple myeloma samples, CD138+ plasma cells were further purified using magnetic-activated cell sorting (MACS) beads according to the manufacturer’s instructions. Both bone marrow mononuclear cells and CD138+ plasma cells were cytospun on slides and fixed in ice-cold methanol/acetone. Early passage primary glioblastoma cells were acquired from the Department of Neurosurgery of the University of Heidelberg (Karcher et al., 2006). Cells between passages 2 and 4 were grown on coverslips and fixed in ice-cold methanol. Colon cancer spheres were generated as previously described (Dieter et al., 2011). Superresolution Microscopy Superresolution microscopy images were acquired using a 633 1.4 numerical aperture (NA) Plan Apochromat objective (Zeiss) on an LSM 880 confocal microscope (Zeiss) equipped with an AiryScan detector unit (Zeiss) at the Zeiss Laboratories in Munich, Germany. Images presented in figures are maximumintensity projections of 0.18-mm z stacks encompassing whole cells. FIB-SEM Cells were grown on dishes with a coverslip marked by a coordinate system (Mattek) to facilitate correlative light and electron microscopy. GFP-PLK4U2OS cells were transiently transfected with an H2B-mCherry-IRES-neo3

Cell Reports 20, 1906–1920, August 22, 2017 1917

plasmid (gift of L. No¨tzold); 33 hr after transfection, transgene expression was induced. At 15 hr post-induction, cells were lightly fixed with 0.5% glutaraldehyde and 4% paraformaldehyde in 0.1 M PHEM buffer (60mM PIPES; 25mM HEPES; 10mM EGTA; 2mM Magnesium chloride; pH 6.9), and quenched with 150 mM glycine in 0.1 M PHEM buffer. Cells of interest were then selected by their fluorescent signal at the Zeiss Axio Observer.Z1, and their position within the coordinate system was noted in order to find back the same cells at the FIB-SEM. For a detailed description of the preparation for FIB-SEM, see Chatel-Chaix et al. (2016). In brief, cells were processed in a PELCO BioWave Pro microwave processor where they were fixed with 2.5% glutaraldehyde with 0.1% malachite green oxalate salt in 0.1 M PHEM buffer, postfixed with 1% osmium tetroxide and 0.8% potassium hexacyanoferrate (III), en bloc stained with 0.5% uranyl acetate in water, dehydrated in an ethanol series, and infiltrated with Durcupan resin. Samples were polymerized at 60
C for 48 hr and then mounted for FIB-SEM. An Auriga 60 FIB-SEM (Zeiss) microscope with Atlas3D software (FIBICS) was used to collect the 3D data. All acquired images were aligned using the TrakEM2 plugin of FIJI. Centrioles were segmented by thresholding with Amira (FEI), and microtubules were extracted with the Amira XTracing Extension 6.0. For comparison and visualization, centrioles were realigned in 3D using Amira software. Within this software, the Slice function can be freely rotated in order to place all mother centrioles orthogonal to the image plan, and all daughter centrioles parallel to the image plan. When the centrioles were close to the edges of the volume, the 2D image showed empty pixels (out of volume areas) that were displayed as a constant grey value (see Figure 1D cell 1 and cell 3). Live-Cell Imaging GFP-PLK4-U2OS cells were plated on glass-bottom culture dishes (Ibidi) and transiently transfected with an H2B-mCherry-IRES-Neo3 plasmid using TurboFect (Life Technologies). At 24 hr after transfection, cells were induced for 15 hr and then transferred to a Zeiss Cell Observer and imaged under controlled environmental conditions at 37
C and 5% CO2 atmosphere. The z stack series was captured at 5-min intervals overnight with a 203 0.8 NA Plan Apochromat objective (Zeiss). HA-PLK4-Dendra-centrin2-HeLa cells were induced for 48 hr. Subsequently, doxycycline was washed out and cells were imaged for 72 hr. The z stack series was captured at 20-min intervals with a 403 1.3 NA Plan Apochromat objective (Zeiss). Videos were analyzed using Zen Blue 2011 software (Zeiss). Microtubule Regrowth Assay Microtubule regrowth assays were performed as previously described (Godinho et al., 2014). Briefly, cells grown on coverslips were incubated with icecold medium for 1 hr, then allowed to repolymerize microtubules for 30 s at 37
C in fresh medium, and immediately fixed in methanol. Cells were stained as described above, and microtubule numbers sprouting from the centrosomes were assessed manually. Branching Process Modeling To model the proliferation of bipolar, clustered, and multipolar mitoses of HAPLK4-Dendra-centrin2-HeLa cells, a multitype branching process (Mode, 1971) was simulated. To estimate the offspring distribution of this branching process, it was assumed that only bipolar and clustered mitoses can produce viable offspring, whereas multipolar mitoses lead to dead progeny only. The exact offspring distribution was estimated from real data. To start the analysis, we first simulated the number of cells produced by bipolar, clustered, and multipolar mitoses for the first 50 generations, under the initial condition that there were 35 bipolar, 40 clustered, and 25 multipolar mitoses in generation 0. The results of this simulation are presented in Figure 6E (right panel). The number of bipolar mitoses in the nth generation was defined as the number of offspring of bipolar type that was produced by both bipolar and clustered mitoses from generation ðn  1Þ. Similarly, the number of clustered or multipolar divisions in the nth generation was defined as the number of clustered or multipolar offspring produced by mitoses of all types in generation ðn  1Þ. Next, we applied the Kesten-Stigum theorem (Kesten and Stigum, 1966) to the branching process. After verifying all assumptions of this theorem, the following results were obtained: (1) the entire population either goes extinct or, asymptotically, the number of cells of each type in the nth generation grows

1918 Cell Reports 20, 1906–1920, August 22, 2017

geometrically at the rate rn , with r = 1:334; and (2) for populations with geometric growth, the proportions of bipolar, clustered, and multipolar mitoses stabilize and approach the following limits: 93.3% for bipolar, 2.3% for clustered, and 4.4% for multipolar mitoses (Figure 6E, left panel). These limits depend only on the offspring distribution of bipolar and clustered mitoses and not on the initial number of cells in generation 0. Note that the proportion of bipolar mitoses in the nth generation was defined as the number of bipolar mitoses in this generation divided by the total number of mitoses. The proportions of clustered and multipolar mitoses were defined in the same way. Statistical Analysis Comparisons between two groups were calculated using two-tailed Student’s t tests. Comparison among three or more groups was estimated by one-way repeated-measure (RM) ANOVA, and single groups were compared by Mann-Whitney U test. Correlation between two variables was calculated using Pearson Product Moment Correlation. Data are reported as mean ± SEM or SD. Statistical values can be found in the figures and figure legends. Statistical calculation was performed with SigmaPlot version 12.5 software (*p < 0.05, **p < 0.01, and ***p < 0.001). SUPPLEMENTAL INFORMATION Supplemental Information includes Supplemental Experimental Procedures, six figures, and two movies and can be found with this article online at http://dx.doi.org/10.1016/j.celrep.2017.08.005. AUTHOR CONTRIBUTIONS A.K. and M.R.C. initiated the project. M.R.C. designed and performed experiments and data analysis. A.C., A.R., G.K., and E.B. performed experiments. M.S.R., T.D., H.G., and C.H.-M. provided tumor samples. S.P. contributed to superresolution microscopy experiments. N.L.S. and Y.S. contributed to correlative light and focused ion beam-scanning electron microscopy. A.S. and T.H.-L. performed mathematical modelling. A.K. conceived and supervised the study and wrote the paper. ACKNOWLEDGMENTS We thank B. Kraft and S. Duensing for comments on the manuscript; A. Baumann and M. Kirsch for technical assistance; L. No¨tzold, A. Jauch, and D. Compton for reagents; M. Weiss and the DKFZ light microscopy core facility for assistance with light microscopy analysis; and P. Machado and the EMBL Electron Microscopy Core Facility for assistance with electron microscopy analysis. M.R.C. was supported by the Helmholtz International Graduate School for Cancer Research of the German Cancer Research Center; A.C. was supported by the ERASMUS programme of the European Union; M.S.R., G.K., and E.B. were supported by the Max Eder programme of the Deutsche Krebshilfe; and A.K. is supported by DFG grant KR 1981/4-1 and Sander Foundation grant 2014.027.1. Received: October 13, 2016 Revised: April 12, 2017 Accepted: July 28, 2017 Published: August 22, 2017 REFERENCES Arquint, C., Sonnen, K.F., Stierhof, Y.D., and Nigg, E.A. (2012). Cell-cycleregulated expression of STIL controls centriole number in human cells. J. Cell Sci. 125, 1342–1352. Bakhoum, S.F., Genovese, G., and Compton, D.A. (2009a). Deviant kinetochore microtubule dynamics underlie chromosomal instability. Curr. Biol. 19, 1937–1942. Bakhoum, S.F., Thompson, S.L., Manning, A.L., and Compton, D.A. (2009b). Genome stability is ensured by temporal control of kinetochore-microtubule dynamics. Nat. Cell Biol. 11, 27–35.

Basto, R., Brunk, K., Vinadogrova, T., Peel, N., Franz, A., Khodjakov, A., and Raff, J.W. (2008). Centrosome amplification can initiate tumorigenesis in flies. Cell 133, 1032–1042.

Ishikawa, H., Kubo, A., Tsukita, S., and Tsukita, S. (2005). Odf2-deficient mother centrioles lack distal/subdistal appendages and the ability to generate primary cilia. Nat. Cell Biol. 7, 517–524.

Bettencourt-Dias, M., and Glover, D.M. (2007). Centrosome biogenesis and function: centrosomics brings new understanding. Nat. Rev. Mol. Cell Biol. 8, 451–463.

Jeffery, J.M., Urquhart, A.J., Subramaniam, V.N., Parton, R.G., and Khanna, K.K. (2010). Centrobin regulates the assembly of functional mitotic spindles. Oncogene 29, 2649–2658.

Bobinnec, Y., Moudjou, M., Fouquet, J.P., Desbruye`res, E., Edde´, B., and Bornens, M. (1998). Glutamylation of centriole and cytoplasmic tubulin in proliferating non-neuronal cells. Cell Motil. Cytoskeleton 39, 223–232.

Karcher, S., Steiner, H.H., Ahmadi, R., Zoubaa, S., Vasvari, G., Bauer, H., Unterberg, A., and Herold-Mende, C. (2006). Different angiogenic phenotypes in primary and secondary glioblastomas. Int. J. Cancer 118, 2182–2189.

Chatel-Chaix, L., Cortese, M., Romero-Brey, I., Bender, S., Neufeldt, C.J., Fischl, W., Scaturro, P., Schieber, N., Schwab, Y., Fischer, B., et al. (2016). Dengue Virus Perturbs Mitochondrial Morphodynamics to Dampen Innate Immune Responses. Cell Host Microbe 20, 342–356.

Kesten, H., and Stigum, B.P. (1966). A limit theorem for multidimensional Galton-Watson processes. Ann. Math. Stat. 37, 1211–1223.

Chen, Z., Indjeian, V.B., McManus, M., Wang, L., and Dynlacht, B.D. (2002). CP110, a cell cycle-dependent CDK substrate, regulates centrosome duplication in human cells. Dev. Cell 3, 339–350. Cimini, D., Howell, B., Maddox, P., Khodjakov, A., Degrassi, F., and Salmon, E.D. (2001). Merotelic kinetochore orientation is a major mechanism of aneuploidy in mitotic mammalian tissue cells. J. Cell Biol. 153, 517–527. Coelho, P.A., Bury, L., Shahbazi, M.N., Liakath-Ali, K., Tate, P.H., Wormald, S., Hindley, C.J., Huch, M., Archer, J., Skarnes, W.C., et al. (2015). Over-expression of Plk4 induces centrosome amplification, loss of primary cilia and associated tissue hyperplasia in the mouse. Open Biol. 5, 150209. Dieter, S.M., Ball, C.R., Hoffmann, C.M., Nowrouzi, A., Herbst, F., Zavidij, O., Abel, U., Arens, A., Weichert, W., Brand, K., et al. (2011). Distinct types of tumor-initiating cells form human colon cancer tumors and metastases. Cell Stem Cell 9, 357–365. Duensing, A., Liu, Y., Perdreau, S.A., Kleylein-Sohn, J., Nigg, E.A., and Duensing, S. (2007). Centriole overduplication through the concurrent formation of multiple daughter centrioles at single maternal templates. Oncogene 26, 6280–6288. Duensing, A., Chin, A., Wang, L., Kuan, S.F., and Duensing, S. (2008). Analysis of centrosome overduplication in correlation to cell division errors in highrisk human papillomavirus (HPV)-associated anal neoplasms. Virology 372, 157–164. Ertych, N., Stolz, A., Stenzinger, A., Weichert, W., Kaulfuß, S., Burfeind, P., Aigner, A., Wordeman, L., and Bastians, H. (2014). Increased microtubule assembly rates influence chromosomal instability in colorectal cancer cells. Nat. Cell Biol. 16, 779–791. Fenech, M. (2007). Cytokinesis-block micronucleus cytome assay. Nat. Protoc. 2, 1084–1104. Fu, J., Lipinszki, Z., Rangone, H., Min, M., Mykura, C., Chao-Chu, J., Schneider, S., Dzhindzhev, N.S., Gottardo, M., Riparbelli, M.G., et al. (2016). Conserved molecular interactions in centriole-to-centrosome conversion. Nat. Cell Biol. 18, 87–99. Ganem, N.J., Godinho, S.A., and Pellman, D. (2009). A mechanism linking extra centrosomes to chromosomal instability. Nature 460, 278–282.

€ckiger, I., Kitagawa, D., Kohlmaier, G., Keller, D., Strnad, P., Balestra, F.R., Flu and Go¨nczy, P. (2011). Spindle positioning in human cells relies on proper centriole formation and on the microcephaly proteins CPAP and STIL. J. Cell Sci. 124, 3884–3893. Kleylein-Sohn, J., Westendorf, J., Le Clech, M., Habedanck, R., Stierhof, Y.D., and Nigg, E.A. (2007). Plk4-induced centriole biogenesis in human cells. Dev. Cell 13, 190–202. Kong, D., Farmer, V., Shukla, A., James, J., Gruskin, R., Kiriyama, S., and Loncarek, J. (2014). Centriole maturation requires regulated Plk1 activity during two consecutive cell cycles. J. Cell Biol. 206, 855–865. La Terra, S., English, C.N., Hergert, P., McEwen, B.F., Sluder, G., and Khodjakov, A. (2005). The de novo centriole assembly pathway in HeLa cells: cell cycle progression and centriole assembly/maturation. J. Cell Biol. 168, 713–722. Lampson, M.A., and Kapoor, T.M. (2005). The human mitotic checkpoint protein BubR1 regulates chromosome-spindle attachments. Nat. Cell Biol. 7, 93–98. Lengauer, C., Kinzler, K.W., and Vogelstein, B. (1997). Genetic instability in colorectal cancers. Nature 386, 623–627. Levine, M.S., Bakker, B., Boeckx, B., Moyett, J., Lu, J., Vitre, B., Spierings, D.C., Lansdorp, P.M., Cleveland, D.W., Lambrechts, D., et al. (2017). Centrosome Amplification Is Sufficient to Promote Spontaneous Tumorigenesis in Mammals. Dev. Cell 40, 313–322.e5. Lingle, W.L., Barrett, S.L., Negron, V.C., D’Assoro, A.B., Boeneman, K., Liu, W., Whitehead, C.M., Reynolds, C., and Salisbury, J.L. (2002). Centrosome amplification drives chromosomal instability in breast tumor development. Proc. Natl. Acad. Sci. USA 99, 1978–1983. Matov, A., Applegate, K., Kumar, P., Thoma, C., Krek, W., Danuser, G., and Wittmann, T. (2010). Analysis of microtubule dynamic instability using a plusend growth marker. Nat. Methods 7, 761–768. Mitelman, F., Johansson, B., and Mertens, F. (2017). Mitelman database of chromosome aberrations and gene fusions in cancer. https://cgap.nci.nih. gov/Chromosomes/Mitelman. Mode, C.J. (1971). Multitype branching processes: Theory and applications, Volume 34 (American Elsevier Pub. Co.).

Ganem, N.J., Cornils, H., Chiu, S.Y., O’Rourke, K.P., Arnaud, J., Yimlamai, D., The´ry, M., Camargo, F.D., and Pellman, D. (2014). Cytokinesis failure triggers hippo tumor suppressor pathway activation. Cell 158, 833–848.

Nakagawa, Y., Yamane, Y., Okanoue, T., Tsukita, S., and Tsukita, S. (2001). Outer dense fiber 2 is a widespread centrosome scaffold component preferentially associated with mother centrioles: its identification from isolated centrosomes. Mol. Biol. Cell 12, 1687–1697.

Godinho, S.A., Picone, R., Burute, M., Dagher, R., Su, Y., Leung, C.T., Polyak, K., Brugge, J.S., The´ry, M., and Pellman, D. (2014). Oncogene-like induction of cellular invasion from centrosome amplification. Nature 510, 167–171.

Neben, K., Giesecke, C., Schweizer, S., Ho, A.D., and Kra¨mer, A. (2003). Centrosome aberrations in acute myeloid leukemia are correlated with cytogenetic risk profile. Blood 101, 289–291.

Go¨nczy, P. (2015). Centrosomes and cancer: revisiting a long-standing relationship. Nat. Rev. Cancer 15, 639–652.

Nigg, E.A. (2002). Centrosome aberrations: cause or consequence of cancer progression? Nat. Rev. Cancer 2, 815–825.

Guarguaglini, G., Duncan, P.I., Stierhof, Y.D., Holmstro¨m, T., Duensing, S., and Nigg, E.A. (2005). The forkhead-associated domain protein Cep170 interacts with Polo-like kinase 1 and serves as a marker for mature centrioles. Mol. Biol. Cell 16, 1095–1107.

Ohta, M., Ashikawa, T., Nozaki, Y., Kozuka-Hata, H., Goto, H., Inagaki, M., Oyama, M., and Kitagawa, D. (2014). Direct interaction of Plk4 with STIL ensures formation of a single procentriole per parental centriole. Nat. Commun. 5, 5267.

Habedanck, R., Stierhof, Y.D., Wilkinson, C.J., and Nigg, E.A. (2005). The Polo kinase Plk4 functions in centriole duplication. Nat. Cell Biol. 7, 1140–1146.

Paoletti, A., Moudjou, M., Paintrand, M., Salisbury, J.L., and Bornens, M. (1996). Most of centrin in animal cells is not centrosome-associated and centrosomal centrin is confined to the distal lumen of centrioles. J. Cell Sci. 109, 3089–3102.

Hanahan, D., and Weinberg, R.A. (2011). Hallmarks of cancer: the next generation. Cell 144, 646–674.

Cell Reports 20, 1906–1920, August 22, 2017 1919

Pihan, G.A., Purohit, A., Wallace, J., Knecht, H., Woda, B., Quesenberry, P., and Doxsey, S.J. (1998). Centrosome defects and genetic instability in malignant tumors. Cancer Res. 58, 3974–3985. Rieder, C.L., and Salmon, E.D. (1998). The vertebrate cell kinetochore and its roles during mitosis. Trends Cell Biol. 8, 310–318. Sabino, D., Gogendeau, D., Gambarotto, D., Nano, M., Pennetier, C., Dingli, F., Arras, G., Loew, D., and Basto, R. (2015). Moesin is a major regulator of centrosome behavior in epithelial cells with extra centrosomes. Curr. Biol. 25, 879–889. Serc¸in, O¨., Larsimont, J.C., Karambelas, A.E., Marthiens, V., Moers, V., Boeckx, B., Le Mercier, M., Lambrechts, D., Basto, R., and Blanpain, C. (2016). Transient PLK4 overexpression accelerates tumorigenesis in p53-deficient epidermis. Nat. Cell Biol. 18, 100–110. Silkworth, W.T., Nardi, I.K., Scholl, L.M., and Cimini, D. (2009). Multipolar spindle pole coalescence is a major source of kinetochore mis-attachment and chromosome mis-segregation in cancer cells. PLoS ONE 4, e6564. Sillibourne, J.E., Tack, F., Vloemans, N., Boeckx, A., Thambirajah, S., Bonnet, P., Ramaekers, F.C., Bornens, M., and Grand-Perret, T. (2010). Autophosphorylation of polo-like kinase 4 and its role in centriole duplication. Mol. Biol. Cell 21, 547–561. €tz, M., Daly, O., Morrison, C.G., Dunning, M., Kilmartin, J.V., and Sir, J.H., Pu Gergely, F. (2013). Loss of centrioles causes chromosomal instability in vertebrate somatic cells. J. Cell Biol. 203, 747–756. Tan, C.H., Gasic, I., Huber-Reggi, S.P., Dudka, D., Barisic, M., Maiato, H., and Meraldi, P. (2015). The equatorial position of the metaphase plate ensures symmetric cell divisions. eLife 4, e05124. Tang, C.J., Lin, S.Y., Hsu, W.B., Lin, Y.N., Wu, C.T., Lin, Y.C., Chang, C.W., Wu, K.S., and Tang, T.K. (2011). The human microcephaly protein STIL interacts with CPAP and is required for procentriole formation. EMBO J. 30, 4790– 4804. Thompson, S.L., and Compton, D.A. (2008). Examining the link between chromosomal instability and aneuploidy in human cells. J. Cell Biol. 180, 665–672.

1920 Cell Reports 20, 1906–1920, August 22, 2017

Vitre, B.D., and Cleveland, D.W. (2012). Centrosomes, chromosome instability (CIN) and aneuploidy. Curr. Opin. Cell Biol. 24, 809–815. Vulprecht, J., David, A., Tibelius, A., Castiel, A., Konotop, G., Liu, F., Bestvater, F., Raab, M.S., Zentgraf, H., Izraeli, S., and Kra¨mer, A. (2012). STIL is required for centriole duplication in human cells. J. Cell Sci. 125, 1353–1362. Weaver, B.A., Silk, A.D., Montagna, C., Verdier-Pinard, P., and Cleveland, D.W. (2007). Aneuploidy acts both oncogenically and as a tumor suppressor. Cancer Cell 11, 25–36. €ller-Reichert, Weber, B., Greenan, G., Prohaska, S., Baum, D., Hege, H.-C., Mu T., Hyman, A.A., and Verbavatz, J.-M. (2012). Automated tracing of microtubules in electron tomograms of plastic embedded samples of Caenorhabditis elegans embryos. J. Struct. Biol. 178, 129–138. White, R.A., Pan, Z., and Salisbury, J.L. (2000). GFP-centrin as a marker for centriole dynamics in living cells. Microsc. Res. Tech. 49, 451–457. Wolff, A., de Ne´chaud, B., Chillet, D., Mazarguil, H., Desbruye`res, E., Audebert, S., Edde´, B., Gros, F., and Denoulet, P. (1992). Distribution of glutamylated alpha and beta-tubulin in mouse tissues using a specific monoclonal antibody, GT335. Eur. J. Cell Biol. 59, 425–432. Wong, Y.L., Anzola, J.V., Davis, R.L., Yoon, M., Motamedi, A., Kroll, A., Seo, C.P., Hsia, J.E., Kim, S.K., Mitchell, J.W., et al. (2015). Cell biology. Reversible centriole depletion with an inhibitor of Polo-like kinase 4. Science 348, 1155– 1160. Zijno, A., Marcon, F., Leopardi, P., and Crebelli, R. (1994). Simultaneous detection of X-chromosome loss and non-disjunction in cytokinesis-blocked human lymphocytes by in situ hybridization with a centromeric DNA probe; implications for the human lymphocyte in vitro micronucleus assay using cytochalasin B. Mutagenesis 9, 225–232. Zou, C., Li, J., Bai, Y., Gunning, W.T., Wazer, D.E., Band, V., and Gao, Q. (2005). Centrobin: a novel daughter centriole-associated protein that is required for centriole duplication. J. Cell Biol. 171, 437–445.