Research ArticleCancer

Proteins Required for Centrosome Clustering in Cancer Cells

See allHide authors and affiliations

Science Translational Medicine  26 May 2010:
Vol. 2, Issue 33, pp. 33ra38
DOI: 10.1126/scitranslmed.3000915

Abstract

Current cancer chemotherapies are limited by the lack of tumor-specific targets, which would allow for selective eradication of malignant cells without affecting healthy tissues. In contrast to normal cells, most tumor cells contain multiple centrosomes, which tend to cause the formation of multipolar mitotic spindles, chromosome segregation defects, and cell death. Nevertheless, many cancer cells divide successfully because they can cluster multiple centrosomes into two spindle poles. Inhibition of this centrosomal clustering, with consequent induction of multipolar spindles and subsequent cell death, would specifically target cancer cells and overcome one limitation of current cancer treatments. We have performed a genome-wide RNA interference screen to identify proteins involved in the prevention of spindle multipolarity in human cancer cells with supernumerary centrosomes. The chromosomal passenger complex, Ndc80 microtubule-kinetochore attachment complex, sister chromatid cohesion, and microtubule formation via the augmin complex were identified as necessary for centrosomal clustering. We show that spindle tension is required to cluster multiple centrosomes into a bipolar spindle array in tumor cells with extra centrosomes. These findings may explain the specificity of drugs that interfere with spindle tension for cancer cells and provide entry points for the development of therapeutics.

Introduction

Centrosomes consist of a pair of centrioles embedded in pericentriolar material and act as microtubule-organizing centers. During mitosis, centrosomes function as spindle poles, directing the formation of bipolar spindles, a process essential for accurate chromosome segregation (1, 2). Centrosomes duplicate precisely once per cell cycle to assure spindle bipolarity, with each daughter cell receiving one centrosome on cytokinesis.

Abnormal centrosome amplification is frequent in both solid tumors and hematological malignancies and is linked to tumorigenesis and aneuploidy (37). The extent of centrosomal aberrations correlates with chromosomal instability and malignant behavior in tumor cell lines, mouse tumor models, and human tumors (3, 69). Mutation or misregulation of a variety of tumor suppressors and oncogenes directly induces chromosomal instability by disrupting the normal function and numeral integrity of centrosomes (10). Centrosome amplification can, in principle, arise from several types of cell division defects, including centrosome overduplication during a single cell cycle, de novo synthesis of centrosomes, cytokinesis failure, and cell fusion (5).

In mitosis, supernumerary centrosomes can lead to the formation of multipolar spindles, which are a hallmark of many tumor types (5). Multipolar spindles, however, are inconsistent with cell viability (11, 12). Most progeny derived from a defective mitosis will undergo apoptosis. To circumvent this problem, many cancer cells appear to have mechanisms that suppress multipolar mitoses, the best studied being clustering of supernumerary centrosomes into two spindle poles, enabling bipolar division (5, 11, 12). Bipolar spindle formation through centrosomal clustering is associated with an increased frequency of lagging chromosomes during anaphase, thereby explaining the link between supernumerary centrosomes and chromosomal instability (12).

Supernumerary centrosomes almost exclusively occur in neoplastic disorders but not in nontransformed cells. Therefore, inhibition of centrosomal clustering with consequential induction of multipolar spindles and subsequent cell death would specifically target tumor cells with no effect on normal cells, as we have proposed (11).

The mechanisms of centrosomal clustering in tumor cells are incompletely understood. It appears that cytoplasmic dynein, a minus end–directed microtubule motor protein, and NUMA, a microtubule-associated protein that has been implicated in microtubule minus end bundling, are required for centrosomal clustering (13). In addition, a recent RNA interference (RNAi) screen in Drosophila S2 cells suggested roles for the spindle assembly checkpoint (SAC), the cortical actin cytoskeleton, and the interphase cell adhesion pattern in suppressing spindle multipolarity in flies (14).

To comprehensively identify genes required to suppress multipolar mitoses in human cancers, we performed a genome-wide RNAi screen in oral squamous cell carcinoma cells (11, 13). The classes of genes discovered by this screen enabled the identification of mechanisms that suppress multipolar mitoses in human cancer cells.

Results

RNAi screen for the identification of proteins required for centrosomal clustering

We used a genome-wide RNAi screen to uncover the molecular pathways involved in clustering of supernumerary centrosomes in human cancer cells (Fig. 1). We used the oral squamous cell carcinoma cell line UPCI:SCC114 for the screen because they contain high levels of extra centrosomes, which are efficiently clustered into bipolar spindles in >95% of mitoses (11, 13).

Fig. 1

Genome-wide RNAi screen for proteins required for clustering of supernumerary centrosomes in UPCI:SCC114 cells. (A) Procedures for the primary and secondary screen. (B) Sample image from screen. Multipolar spindles after treatment with aurora-B–specific siRNA are shown. Cells were stained for microtubules (α-tubulin; green) and DNA [4′,6-diamidino-2-phenylindole (DAPI); blue]. Scale bar, 10 μm. A multipolar spindle is shown enlarged in an inset. Scale bar, 5 μm. (C) Table summarizing the screening results. (D) Gene ontology annotations for the 82 genes from the screen. (E) Multipolar mitoses resulting from siRNA-mediated knockdown of INCENP and CEP164, respectively. Scale bars, 5 μm.

We used a human small interfering RNA (siRNA) library that targets ~21,000 genes, with each gene being targeted by a siRNA pool consisting of four siRNAs directed against different regions of one gene (Fig. 1A and table S1). UPCI:SCC114 cells were reverse-transfected with siRNAs in 384-well plates. After 48 hours, images were acquired on an automated microscope (Fig. 1B).

By visual inspection of ~200,000 images, we scored the percentage of multipolar spindles for each RNAi pool. Two hundred mitotic cells per siRNA treatment were analyzed and scored as a hit when a highly significant difference (more than three SDs) from negative controls was observed. With these criteria and after reevaluation of each image-positive siRNA pool by direct microscopic analysis, the primary screen identified 133 candidates associated with a multipolar spindle phenotype (Fig. 1C).

To minimize the number of false positives, we performed a secondary screen, with each hit from the primary screen being targeted by each of the four siRNAs per siRNA pool in separate experiments (Fig. 1A). By using this strategy, we confirmed that 82 of the initial 133 genes induced spindle multipolarity in UPCI:SCC114 cells (Fig. 1C and table S2). Among the validated genes, 22 (26.8%) had been previously identified to play a role in mitosis, 6 (7.3%) are involved in cell adhesion, 4 (4.9%) are involved in proteolysis and ubiquitination, and 13 (15.9%) do not have a known function (Fig. 1D). Coimmunostaining with an antibody against cyclin B1 in a subgroup of positive cells confirmed that 94.2 ± 3.5% (range, 89.5 ± 0.02% to 99.0 ± 0.01%; fig. S1) of the events counted as spindle multipolarity represented cells in metaphase. To control for siRNA-mediated knockdown efficacy, we performed quantitative reverse transcription polymerase chain reaction on 13 randomly selected genes from our list of validated genes and showed a knockdown efficacy of 86.9 ± 9.8% (range, 56.4 ± 0.6% to 96.5 ± 1.0%; fig. S2A). To further demonstrate knockdown of protein, we performed Western blots in a subset of cases with comparably weak knockdown efficacies as assessed by messenger RNA (fig. S2B).

To further elucidate the mechanisms leading to spindle multipolarity, we focused on genes important for spindle assembly and mitosis progression (Table 1). Small interfering RNA–mediated knockdown of all of these genes was confirmed to induce spindle multipolarity in an independent cell line (MDA-MB-231, breast cancer) that also efficiently clusters supernumerary centrosomes into two spindle poles (12) (fig. S3).

Table 1

Validated genes required for centrosomal clustering with a function in mitosis.

View this table:

Multipolar spindle formation in UPCI:SCC114 cells is mainly caused by inhibition of centrosomal clustering

To ascertain that spindle multipolarity after siRNA-induced knockdown of the genes identified in our screen is indeed caused by inhibition of centrosomal clustering and not by the formation of acentrosomal spindle poles, we immunostained green fluorescent protein (GFP)–α-tubulin–expressing UPCI:SCC114 cells with an antibody to centrin as a centriolar marker after exposure to siRNAs for 48 hours. In >50% of cells with multipolar spindles, each spindle pole contained centrin signals for 29 of 32 siRNAs examined (90.6%), clearly indicating that spindle multipolarity in most cases is not a consequence of acentrosomal spindle pole formation (Fig. 1E, Table 1, and table S3). In contrast, only 3 of 32 siRNAs (9.4%), namely, ch-TOG, HAUS3, and CEP164, led to multipolar spindles with centrin signals at only two poles.

Multipolar mitoses by inhibition of centrosomal clustering should not arise in normal cells without supernumerary centrosomes. Consistent with this prediction, depletion of chromosomal passenger complex (CPC) (aurora-B, INCENP, survivin, and borealin) and Ndc80 complex (HEC1, SPC24, and SPC25) proteins, which were among the hits identified in our screen, did not lead to a significant induction of spindle multipolarity in normal diploid BJ fibroblasts (Table 1 and table S3). In contrast, depletion of proteins involved in cytokinesis, which induce spindle multipolarity only after cytokinesis failure via duplication of the cellular centrosome content, led to multipolarity to a certain extent in BJ fibroblasts as well. Also, knockdown of augmin complex components, disruption of which induces microtubule-dependent fragmentation of centrosomes (15), induced multipolar spindles in BJ fibroblasts. Whereas after depletion of proteins involved in cytokinesis virtually all spindle poles contained centrin signals, knockdown of augmin complex components led to a significant fraction of acentriolar spindle poles. Similar results were obtained with hTERT-RPE-1 cells, a second normal control cell line (fig. S4).

Knockdown of ch-TOG and TPX2 induces spindle multipolarity (16, 17). Whereas ch-TOG stabilizes microtubule minus ends at spindle poles, TPX2 mediates chromosome-associated microtubule formation. In our screen, depletion of both proteins led to significant spindle multipolarity in both UPCI:SCC114 cells and BJ fibroblasts (Table 1). Consistent with their mechanism of action, ch-TOG knockdown was associated with acentriolar supernumerary spindle poles, whereas depletion of TPX2 led to multipolarity with centrin signals at each pole in UPCI:SCC114 cells.

The CPC is necessary for centrosomal clustering

All four CPC components, aurora-B, INCENP, survivin, and borealin, were required for centrosome clustering in UPCI:SCC114 cells, suggesting a role for the CPC in this process (Fig. 2A). The CPC is a key regulator of mitotic events, including the control of chromosome-microtubule interactions and spindle tension at the kinetochore during metaphase as well as cytokinesis during anaphase and telophase (18). The nonenzymatic members of the complex—INCENP, survivin, and borealin—control the targeting, enzymatic activity, and stability of aurora-B. To examine whether aurora-B kinase activity is required to prevent spindle multipolarity, we treated UPCI:SCC114 cells with the aurora kinase inhibitor ZM447439 (19). Incubation with 0.5 μM ZM447439 for 24 hours resulted in 58.5 ± 1.63% multipolar spindles, whereas only 4.6 ± 0.24% of mock-treated mitoses were multipolar (Fig. 2, B and C). DLD-1 colorectal cancer cells treated with ZM447439 assemble normal bipolar spindles (19) and display microsatellite but not chromosomal instability and centrosome amplification (20). In line with these reports, we have found that only 5.3 ± 0.5% of DLD-1 cells contain supernumerary centrosomes. Accordingly, treatment with 0.5 μM ZM447439 for 24 hours led to only 12.8 ± 3.4% multipolar spindles as compared to 7.0 ± 0.82% in mock-treated DLD-1 cells. We conclude that CPC-associated aurora-B kinase activity is necessary for centrosomal clustering in cancer cells with supernumerary centrosomes.

Fig. 2

The CPC is necessary for centrosomal clustering. (A) siRNA-mediated depletion of CPC and Ndc80 complex components leads to spindle multipolarity in UPCI:SCC114 cells. Multipolar mitoses were scored 48 hours after siRNA transfection. The graph shows the average of three independent experiments; mean ± SD. Transfection with luciferase-specific siRNA (LUC) served as a control. (B) Chemical inhibition of aurora-B kinase activity by ZM447439 induces spindle multipolarity in UPCI:SCC114 cells. Cells were incubated with 0.5 μM ZM447439 for 24 hours. pH3, phospho-histone-S10 H3. Scale bars, 5 μm. (C) Mean ± SD of three independent experiments is given. ***P < 0.001, Student’s t test. (D) GFP–α-tubulin–expressing UPCI:SCC114 cells transfected with survivin- or INCENP-specific siRNA were subjected to continuous time-lapse video microscopy 24 hours after transfection. Representative examples of mononuclear cells that develop multipolar mitotic spindles are shown. Scale bars, 5 μm. Time is shown in minutes. (E) Induction of spindle multipolarity by expression of borealin and survivin deletion constructs. GFP-borealin1–141, GFP-borealinfull-length, GFP-survivin89–142, and GFP-survivinfull-length were transiently transfected into UPCI:SCC114 cells. Multipolar mitoses were scored 48 hours after transfection. Representative examples of cells with multipolar spindles and midbody localization (arrows) of GFP-borealin1–141 (top left panel) and GFP-survivin89–142 (bottom left panel) are shown. Scale bars, 5 μm. Mean percentages of spindle multipolarity ± SD are depicted on the right. ***P < 0.001, Student’s t test.

Centrosome amplification results from either several rounds of centrosome duplication within the same cell cycle or aborted cell division with subsequent duplication of both cellular DNA and centrosome content (5). Consequently, down-regulation of CPC proteins that also function in cytokinesis might lead to spindle multipolarity only indirectly after initial abortion of cell division with duplication of already amplified centrosome numbers in UPCI:SCC114 cells. To exclude this possibility, we first analyzed GFP–α-tubulin–expressing UPCI:SCC114 cells treated with siRNAs to the CPC proteins aurora-B, INCENP, survivin, and borealin by continuous time-lapse video microscopy (Fig. 2D). This analysis revealed that 13 of 52 (25%), 13 of 35 (37%), 16 of 43 (37%), and 18 of 83 (22%) mononucleated cells examined after treatment with siRNA to aurora-B, INCENP, survivin, and borealin, respectively, developed multipolar spindles before the development of cytokinesis defects. The remaining cells performed bipolar mitoses likely because of limited siRNA transfection efficacy. Second, when UPCI:SCC114 cells were transiently transfected with GFP-borealin1–141, a borealin fragment that allows for CPC localization to the spindle midzone but delocalizes the entire CPC from centromeres (21), 12.4 ± 0.9% of mitoses showed multipolar spindles 48 hours after transfection. In contrast, only 6.5 ± 0.4% of mitoses in cells transfected with GFP-borealinfull-length were multipolar (P = 2.5 × 10−5; Fig. 2E). Similar results were obtained when cells were transfected with GFP-survivin89–142, a survivin deletion construct that localizes the CPC to the spindle midzone but interferes with centromere targeting of the CPC, analogous to borealin1–141 (11.9 ± 0.4% versus 6.8 ± 0.4% multipolar spindles after transfection with GFP-survivinfull-length, P = 2.5 × 10−6) (22). Together, these findings demonstrate that inhibition of centrosomal clustering after knockdown of CPC proteins is a direct result of disturbed microtubule-kinetochore interaction rather than an indirect consequence of centrosome content duplication after cytokinesis failure.

Depletion of proteins involved in microtubule-kinetochore attachment induces spindle multipolarity

In addition to CPC proteins, three of four proteins of the Ndc80 complex (HEC1, SPC24, and SPC25) as well as CENPT were identified in our screen (Fig. 2A, Table 1, and table S2). The Ndc80 complex localizes to the outer plate of vertebrate kinetochores and serves as the contact point for kinetochore-microtubule attachment (23). CENPT is part of a DNA binding complex that associates with nucleosomal DNA in centromeric regions and connects the centromere with outer kinetochore components (24). Improperly attached microtubules are released from kinetochore binding after phosphorylation of HEC1 by aurora-B, thereby activating the SAC (25). Recent evidence suggests that phosphorylation of kinetochore proteins by aurora-B, which localizes to the inner centromere, depends on its distance from the kinetochore and thereby on centromere tension rather than microtubule-kinetochore attachment (25, 26). In contrast to the knockdown of CPC proteins, siRNA-mediated depletion of the Ndc80 complex and CENPT led to a mitotic arrest (Fig. 3A and Table 1). Previous work in PtK1 cells suggests that the SAC is not activated by multipolar spindles (27). On the other hand, the SAC is required to allow time for centrosomal clustering in cells with supernumerary centrosomes (14). Our data suggest that, as knockdown of Ndc80 complex proteins and CENPT induces spindle multipolarity and mitotic arrest, spindle tension is required for centrosomal clustering and SAC inactivation. Because depletion of CPC proteins induces multipolar spindles without mitotic arrest, aurora-B activity seems to be necessary for generation of spindle tension and SAC activation in response to a lack of tension.

Fig. 3

Microtubule-kinetochore attachment and sister chromatid cohesion are required to prevent multipolar spindle formation. (A) siRNA-mediated depletion of Ndc80 complex components and CENPT but not CPC proteins leads to mitotic arrest in UPCI:SCC114 cells. Mitotic indices were scored 48 hours after siRNA transfection. The graph shows the average of three independent experiments; mean ± SD. Transfection with luciferase-specific siRNA (LUC) served as a control. (B) siRNA-mediated depletion of sororin, SGOL1, or sSGOL1 leads to spindle multipolarity in UPCI:SCC114 cells. Multipolar mitoses were scored 48 hours after siRNA transfection. The graph shows the average of three independent experiments; mean ± SD. Transfection with luciferase-specific siRNA (LUC) served as a control. (C) Treatment with 0.5 μM ZM447439 (ZM) for 24 hours leads to normalization of the mitotic index in UPCI:SCC114 cells 24 hours after transfection with HEC1-, SPC24-, SGOL1-, or sororin-specific siRNA. Mitotic indices were scored 48 hours after siRNA transfection. The graph shows the average of three independent experiments; mean ± SD. Transfection with luciferase-specific siRNA (LUC) served as a control. (D) siRNA-mediated depletion of SGOL1 leads to spindle multipolarity with two centrin signals at each pole in GFP–α-tubulin–UPCI:SCC114 cells. Scale bar, 5 μm.

Sister chromatid cohesion is required to prevent multipolar spindle formation

Both shugoshin (SGOL1) and sororin are required for sister chromatid cohesion (28, 29). In addition, SGOL1 has been implicated in sensing tension at mitotic chromosomes (28, 30). sSGOL1, on the other hand, a splice variant of SGOL1, seems to protect centriole cohesion (31). We identified SGOL1 as well as sororin as additional strong hits in our screen (Fig. 3B, Table 1, and table S2). siRNA-mediated depletion of either protein led to a pronounced mitotic arrest, similar to what was seen after knockdown of Ndc80 complex proteins (Fig. 3C). To discriminate between the two shugoshin isoforms at chromatids and centrosomes as being responsible for centrosome clustering, we selectively depleted full-length SGOL1 by using an siRNA directed against the sequence encoded by exon 6, which is lacking in the sSGOL1 variant isoform. Similar to a siRNA that targets both isoforms, the full-length SGOL1-specific siRNA led to a pronounced induction of spindle multipolarity (Fig. 3B). In addition, most cells with multipolar spindles contained two centrioles at each spindle pole after SGOL1 depletion (Fig. 3D). Therefore, we conclude that inhibition of centrosomal clustering after depletion of shugoshin is a consequence of decreased spindle tension due to reduced chromatid cohesion. In line with these findings, depletion of haspin, another protein required for chromatid cohesion that localizes to centrosomes as well, leads to spindle multipolarity only after disruption of cohesion and loss of spindle tension (32).

Additional treatment of cells depleted for HEC1, SPC24, SGOL1, or sororin with the aurora kinase inhibitor ZM447439 led to a normalization of the mitotic index in the presence of spindle multipolarity (Fig. 3C). These results further support the notion that loss of centromere tension results in centrosomal declustering, with tension-induced mitotic arrest being dependent on aurora-B activity, which therefore is necessary to avoid spindle multipolarity.

Depletion of the augmin complex prevents centrosomal clustering

Augmin, a recently identified eight-subunit protein complex, is a critical factor for microtubule-based microtubule generation (15, 33, 34). The complex recruits γ-tubulin ring complexes to spindle microtubules and thereby promotes microtubule-dependent microtubule amplification. Knockdown of augmin prevents γ-tubulin localization to the spindle and triggers the SAC via loss of kinetochore microtubules and consecutive reduction of tension at kinetochores (15, 33). Our screen identified three of the eight proteins of the augmin complex—FAM29A, HEI-C, and HAUS3—as required for prevention of spindle multipolarity (Table 1 and table S2). As for the Ndc80 complex proteins, CENPT, sororin, and SGOL1, depletion of these proteins induced spindle multipolarity accompanied by mitotic arrest.

Spindle multipolarity after knockdown of augmin components is a consequence of centrosome fragmentation because microtubule depolymerization by nocodazole (to decrease pulling forces on centrosomes) reduced the number of cells with fragmented centrosomes (15). Because our findings, on the other hand, support the notion that loss of spindle tension causes spindle multipolarity by inhibition of centrosomal clustering, addition of nocodazole after knockdown of augmin complex components should increase rather than decrease the extent of spindle multipolarity in cells that efficiently cluster supernumerary centrosomes. Indeed, microtubule depolymerization by addition of nocodazole after knockdown of either HAUS3 or FAM29A led to a dose-dependent increase of spindle multipolarity in UPCI:SCC114 cells (Fig. 4A). This finding further corroborates the concept that counterbalancing forces at spindle poles are necessary for clustering of supernumerary centrosomes.

Fig. 4

Depletion of augmin complex components prevents centrosomal clustering. (A) Reduction of spindle tension by nocodazole aggravates spindle multipolarity induced by depletion of augmin complex components. Cells were incubated with increasing concentrations of nocodazole for 6 hours, starting 48 hours after transfection with FAM29A- or HAUS3-specific siRNA. Multipolar mitoses were scored 54 hours after siRNA transfection. The graph shows the average of three independent experiments; mean ± SD. Transfection with luciferase-specific siRNA (LUC) served as a control. (B) Expression of mutant HAUS3 leads to a dim spindle γ-tubulin phenotype and spindle multipolarity. GFP-HAUS31–525 was transiently transfected into UPCI:SCC114 cells. Representative examples of cells with a dim spindle γ-tubulin signal (top panel) and a multipolar spindle (bottom panel) are shown. Scale bar, 5 μm.

Most recently, HAUS3 has been found to be mutated in breast cancer (35). When UPCI:SCC114 cells were transiently transfected with GFP-HAUS31–525, one of the mutant HAUS3 forms identified in breast cancer patients, γ-tubulin did not localize to spindle microtubules (Fig. 4B). Moreover, expression of GFP-HAUS31–525 induced spindle multipolarity in UPCI:SCC114 cells (Fig. 4B).

CEP164 is required for spindle pole integrity

CEP164 was identified as one of the strongest hits in our screen (Table 1 and table S2). CEP164 is a newly identified centrosomal protein whose depletion impairs primary cilium formation (36). It localizes to the distal appendages of mature centrioles. Centriolar appendages serve as microtubule-anchoring structures of the centrosome. Knockdown of CEP164 leads to a strong induction of spindle multipolarity in normal BJ fibroblasts and hTERT-RPE-1 cells and, in contrast to most other proteins that we identified as hits, to multipolar spindles with centrioles at only two spindle poles (Fig. 1E, Table 1, and fig. S4). Time-lapse imaging revealed that, contrary to the situation in control cells, during early mitosis in CEP164-depleted UPCI:SCC114 cells, microtubule seeds that are nucleated at extracentrosomal sites are not integrated into a bipolar spindle array, resulting in spindle multipolarity (Fig. 5A). Later, bipolar spindles disintegrate, with additional acentrosomal poles emerging by separation of microtubule bundles from spindle poles, resulting in metaphase arrest with prolongation of mitosis to 505 ± 131 min as compared to 133 ± 15 min after mock treatment (Fig. 5B).

Fig. 5

CEP164 is required for bipolar spindle formation. (A) GFP–α-tubulin–expressing UPCI:SCC114 cells transfected with luciferase (LUC)– or CEP164-specific siRNA were subjected to continuous time-lapse video microscopy 24 hours after transfection. Contrary to cells transfected with CEP164-specific siRNA (bottom panel), in control cells transfected with luciferase-specific siRNA microtubule seeds (arrows) were transported toward the poles, resulting in a bipolar spindle (top panel). (B) CEP164 is required for spindle pole integrity. GFP–α-tubulin–expressing UPCI:SCC114 cells transfected with CEP164-specific siRNA were subjected to continuous time-lapse video microscopy 24 hours after transfection. Depletion of CEP164 causes disintegration of spindle poles during metaphase. Scale bars, 5 μm. Time is shown in minutes.

Spindle tension is required for centrosomal clustering

As described above, depletion of Ndc80 and augmin complex proteins, CENPT, sororin, and SGOL1, led to a mitotic arrest of cells with multipolar spindles in metaphase, whereas knockdown of CPC components resulted in spindle multipolarity without metaphase arrest. FAM29A, SGOL1, and HEC1 have already been reported to be required for spindle tension (25, 28, 30, 33, 34). In addition, aurora-B, INCENP, and survivin have been implicated in sensing tension to create a tension checkpoint (26, 37). To directly measure tension across multipolar spindles, we determined interkinetochore distances in mitotic cells treated with siRNAs to SPC24, HAUS3, aurora-B, and SGOL1. These proteins were chosen as representative components of the Ndc80, augmin, and CPC complexes and as proteins involved in sister chromatid cohesion. Pulling forces or tension across sister kinetochores was substantially reduced in each case, as indicated by shorter interkinetochore distances in multipolar metaphase cells (Fig. 6, A and B). To a lesser but still significant extent, interkinetochore distances were also reduced in metaphase cells that remained bipolar despite treatment with siRNAs to SPC24, HAUS3, aurora-B, or SGOL1 (Fig. 6B). These findings suggest that reduced tension below a certain threshold indeed leads to the formation of multipolar spindles via reduced centrosomal clustering. Consistent with these results, the frequency of kinetochores with strong BUBR1 labeling as a marker for reduced tension (19) was significantly increased in HAUS3- and SGOL1-depleted UPCI:SCC114 cells with multipolar spindles (Fig. 6C). Because both SPC24 and aurora-B are necessary for the recruitment of BUBR1 to kinetochores (19, 38), depletion of SPC24 and aurora-B did not lead to increased kinetochore BUBR1 signals. In contrast, the localization of MAD2 to kinetochores depends on microtubule attachment, not tension (39). Accordingly, whereas microtubule depolymerization by cold treatment caused unattached kinetochores with strong MAD2 labeling, no MAD2 staining was visible at kinetochores of UPCI:SCC114 cells with multipolar spindles after treatment with siRNAs to HAUS3, SGOL, and aurora-B (Fig. 6D). Because HEC1 is required for retention of MAD2 at kinetochores (40), depletion of HEC1 served as a negative control.

Fig. 6

Spindle tension is required for centrosomal clustering. (A) UPCI:SCC114 cells were transfected with siRNAs to HAUS3, SGOL1, SPC24, and aurora-B for 48 hours. Luciferase-specific siRNA (LUC) served as a control. Cells were stained for CREST (green), HEC1 (red), and DNA (DAPI; blue). Enlarged images of representative sister kinetochore pairs are shown as insets. Scale bar, 5 μm. (B) Interkinetochore distances from multipolar (left table) and bipolar (right table) spindles are given as mean and SD. P values were calculated by two-tailed Student’s t test relative to bipolar metaphase cells after transfection with luciferase-specific siRNA. (C) UPCI:SCC114 cells were transfected with siRNAs to HAUS3, SGOL1, SPC24, and aurora-B for 48 hours. Luciferase-specific siRNA (LUC) served as a control. Scale bar, 5 μm. (D) UPCI:SCC114 cells were transfected with siRNAs to HAUS3, SGOL1, HEC1, and aurora-B for 48 hours. Chilling on ice for 10 min (Cold) served as a positive control. Scale bar, 5 μm. (E) Model of centrosomal clustering (left) and mechanisms involved in its inhibition via reduction of spindle tension (right). Centrosome clustering is brought about by microtubule tension-dependent uniform positioning of individual centrosomes resulting in the formation of two spindle poles (left). Spindle tension can be disrupted by reduction of chromatid cohesion via depletion of SGOL1 or sororin (1); disturbed microtubule-kinetochore attachment after knockdown of CENPT, CPC, or Ndc80 complex components (2); reduced microtubule-based microtubule generation after depletion of augmin complex constituents (3); or disturbed microtubule-centrosome attachment after CEP164 knockdown (4) as shown in our screen.

The extracellular matrix is connected to the intracellular actin cytoskeleton of mitotic cells via retraction fibers, which put astral microtubules and thereby the mitotic spindle under tension (41). If spindle tension is in fact required for bipolar spindle formation via centrosomal clustering in cells with supernumerary centrosomes, then interference with cell adhesion should reduce traction forces onto spindle microtubules, leading to spindle multipolarity in the presence of extra centrosomes. To address this issue, we selectively analyzed spindle multipolarity in adherent versus nonadherent SW480 colon cancer cells, which contain amplified centrosomes (20) and are adhesion-defective due to an adenomatous polyposis coli (APC) mutation (42). The proportion of cells with multipolar spindles was significantly higher in the nonadherent cell population, confirming the role of tension (Fig. 5A). A summary diagram of the mechanisms identified in our screen that led to inhibition of centrosomal clustering via reduction of spindle tension is depicted in Fig. 6E.

Depletion of proteins involved in cytokinesis induces spindle multipolarity only after cytokinesis failure via duplication of the cellular centrosome content

Our screen also identified a requirement for genes implicated in cytokinesis for centrosomal clustering: PRC1, MKLP1, ECT2, and anillin (Table 1 and table S2) (43). Disruption of PRC1, MKLP1, ECT2, and anillin is known to result in failure of cytokinesis with subsequent binucleation and duplication of the centrosome content.

Proteins required for cytokinesis have been previously suspected to lead to spindle multipolarity only secondary after cytokinesis failure (14). Analysis of UPCI:SCC114 cells by continuous time-lapse video microscopy clearly revealed that 93, 89.5, and 93% of mononucleated cells performed bipolar divisions, whereas 92.9, 93.1, and 100% of binucleated cells developed multipolar spindles after treatment with siRNA to ECT2, anillin, and MKLP1, respectively (fig. S6). In contrast, transfection with a PRC1-specific siRNA led to spindle multipolarity of mononucleated cells in 34% of cases. It has been shown recently that, in addition to its role in central spindle formation and cytokinesis, PRC1 is required for the establishment of tension across kinetochores, at least in fission yeast (44).

Multipolar spindle formation leads to apoptotic cell death irrespective of the underlying mechanism

Most progeny of spontaneously occurring multipolar cell divisions die (12). Also, induction of spindle multipolarity by griseofulvin and a HEI-C–specific siRNA leads to cell death via apoptosis (11, 45). Analysis of UPCI:SCC114 cells treated with siRNAs to HEC1, aurora-B, survivin, sororin, SGOL1, HAUS3, FAM29A, PRC1, MKLP1, and anillin—representing the major complexes and mechanisms identified to be involved in the induction of spindle multipolarity in our screen—revealed an invariable induction of apoptosis by both Annexin V staining and poly(ADP-ribose) polymerase (PARP) cleavage (Fig. 7, A to C). In contrast, knockdown of these proteins had only a minor effect on the viability of nontransformed BJ fibroblasts without supernumerary centrosomes (Fig. 7D). Also, nonadherent SW480 colon cancer cells showed significantly more apoptosis than their adherent counterparts (fig. S5, B and C). Thus, multipolar spindles lead to apoptosis independent of the mechanism of multipolarity induction.

Fig. 7

Multipolar spindle formation leads to apoptotic cell death irrespective of the underlying mechanism. UPCI:SCC114 cells were transfected with siRNAs to HEC1, survivin, sororin, and SGOL1 (for 72 hours) or FAM29A, HAUS3, aurora-B, anillin, PRC1, and MKLP1 (for 96 hours). Luciferase-specific siRNA (LUC) served as a control. Subsequently, cells were stained with Annexin V–FITC (fluorescein isothiocyanate) and propidium iodide and analyzed by flow cytometry. (A) Representative example 72 hours after transfection with a HEC1-specific siRNA. (B) Table showing the percentage of Annexin V–positive UPCI:SCC114 cells 72 hours (top panel) and 96 hours (bottom panel) after transfection with the respective siRNA. Mean and SD of two independent experiments are given. (C) UPCI:SCC114 cells were transfected with siRNAs to HEC1, survivin, sororin, and SGOL1 (for 72 hours, left panel) or FAM29A, HAUS3, aurora-B, PRC1, anillin, and MKLP1 (for 96 hours, right panel). Luciferase-specific siRNA (LUC) served as a control. Lysates were immunoblotted with an antibody to PARP. Actin served as a loading control. CF, cleaved PARP fragment. (D) Table showing the percentage of Annexin V–positive BJ fibroblasts 72 hours (left panel) and 96 hours (right panel) after transfection with the respective siRNA. Mean and SD of three independent experiments are given.

Discussion

An association between supernumerary centrosomes and cancer has been known for ~100 years (46). Nevertheless, how multiple centrosomes affect division, chromosomal stability, and survival of tumor cells is only beginning to be understood. Here, we report a genome-wide RNAi screen performed in human cancer cells designed to uncover the mechanisms by which spindle multipolarity is suppressed in tumor cells. Our experiments lead to the following conclusions. First, inhibition of centrosomal clustering constitutes the predominant but not exclusive mechanism of multipolar spindle formation in human cancer cells with supernumerary centrosomes. Second, spindle tension seems to be the principal means by which multiple centrosomes are clustered into a bipolar spindle and multipolarity is prevented (Fig. 6E). Third, regardless of the mechanism of induction, spindle multipolarity uniformly leads to cell death via apoptosis.

Multipolar spindle formation can, in principle, occur via inhibition of centrosomal clustering as well as by generation of acentrosomal spindle poles. With the notable exception of CEP164, ch-TOG, and HAUS3 depletion, all other hits that scored positive in our screen were the consequence of centrosomal declustering, as centrioles were detectable at every spindle pole. Depletion of ch-TOG and augmin complex components leads to extra acentrosomal spindle poles by centrosome-microtubule detachment on generation of spindle tension (15, 16, 33, 45), similar to what has been described for knockdown of NUMA, a protein that has been implicated in multipolar spindle formation earlier (15, 47). In light of our results showing that, depending on the depleted subunit, between 20% (HAUS3) and 70% (FAM29A) of multipolar spindles contain centrin signals at each pole, we propose that the augmin complex prevents spindle multipolarity by both assurance of spindle pole integrity and prevention of centrosomal declustering in cells with multiple centrosomes.

In Drosophila oocytes, which lack centrioles, the minus end–directed, dynein-like motor protein NCD is necessary for microtubule bundling at spindle poles, whereas in Drosophila S2 cells with supernumerary centrosomes NCD is required for centrosomal clustering (14), also suggesting that centrosomal clustering in cells with multiple centrosomes and spindle pole focusing in cells with a normal centrosome content (or even without centrosomes) occur by similar mechanisms. Our screen identified dynein, another minus end–directed motor, thereby confirming earlier results by the Saunders laboratory (13). PRC1 is a static microtubule-bundling protein that has now been found to be required for the prevention of spindle multipolarity. PRC1 is a microtubule-associated protein with potent microtubule-bundling activity, most prominently during central spindle formation. In addition, PRC1 is important for the establishment of kinetochore tension (44). Our data show that a related mechanism seems to be responsible for multipolar spindle formation after CEP164 depletion, as spindle poles disintegrate when tension is applied, ultimately resulting in excess acentrosomal spindle poles in cells with and without supernumerary centrosomes. This concept is further supported by the finding that CEP164 localizes to distal centriolar appendages, which are speculated to serve as microtubule-anchoring sites at spindle poles (36).

A recent RNAi screen in Drosophila S2 cells identified microtubule- and actin-associated proteins, microtubule motors, and adhesion and SAC proteins to be involved in centrosomal clustering (14). In many Drosophila cell types, including S2 cells, functional centrosomes are only present during cell division (48). On mitotic exit, centrosomes disassemble, producing interphase cells containing centrioles that lack microtubule-nucleating activity. These findings suggest that it is difficult to compare Drosophila with mammalian cancer cells with regard to the mechanisms that prevent spindle multipolarity and centrosomal clustering. Nevertheless and despite the fact that we did not find any overlapping hits between our screen in human cancer cells and the screen performed in Drosophila S2 cells, the mechanisms of centrosomal clustering seem to be partially conserved.

Most important, our results suggest that in human cancer cells spindle tension seems to be the principal means by which multiple centrosomes are clustered into a bipolar spindle and multipolarity is prevented. This model fits well with the observation by Kwon and co-workers (14) that in Drosophila S2 cells the interphase adhesion pattern contributes to spindle pole positioning via cortical pulling forces exerted on spindle poles by astral microtubules. Further, low concentrations of microtubule-stabilizing agents reduce spindle tension and lead to multipolarity in cancer cells (49). Also, CPC activity at the kinetochore is necessary for the correction of merotelic kinetochore-microtubule attachments, which, as long as uncorrected, cause reduced spindle tension (18).

We and others have shown that inhibition of centrosomal clustering can induce cell death selectively in cancer cells with supernumerary centrosomes. More specifically, apoptosis is induced through centrosome declustering by the microtubule-interacting drug griseofulvin (11) and knockdown of the minus end–directed motor protein HSET as shown by Kwon and co-workers (14). We now find that induction of spindle multipolarity by depletion of HEC1, survivin, FAM29A, SGOL1, anillin, PRC1, ECT2, and CEP164 leads to cell death in cancer cells with supernumerary centrosomes as well. In line with these results, small-molecule inhibition of HEC1 suppresses tumor cell growth in culture and in animals (50). Moreover, knockdown of the augmin complex component HEI-C also induces apoptosis in HeLa cells (45). From these findings, it may be concluded that spindle multipolarity uniformly leads to cell death irrespective of the underlying mechanism. Corroborating these results, single-cell fate analysis has shown that almost all of the progeny of spontaneously multipolar cells die, regardless of tissue of origin or whether the cells are mononucleated or polynucleated (12).

At low drug concentrations, microtubule poisons induce greater cell death in cancer cells than in nontransformed cells (51). The same low concentrations of these drugs lead to spindle multipolarity in cancer cells (49). Also, antimitotic drugs with different mechanisms of action trigger very similar cell death responses (52). These findings together with our data suggest that the specificity of microtubule-interacting drugs for cancer cells might be a consequence of their interference with spindle tension and subsequent inhibition of centrosomal clustering in tumors cells with supernumerary centrosomes. Therefore, drug discovery based on the results presented here might be a promising avenue in developing more specific anticancer drugs.

Materials and Methods

RNAi screen

For the primary screen, siGENOME SMART pools (Dharmacon, Thermo Fisher) were transferred into 384-well plates at a final concentration of 50 nM (table S1). Current annotation of siRNA reagents can be found at http://www.genomernai.org. After incubation with the transfection reagent (Dharmafect-1), UPCI:SCC114 cells were added to the mixture of siRNA and transfection reagent using an automated Multidrop Combi dispensing system (Thermo Scientific). After 48 hours, the cells were stained with a mouse monoclonal antibody to α-tubulin (DM1A, Sigma-Aldrich) followed by an antibody to mouse Alexa 488 (Invitrogen) and Hoechst 33342 (Invitrogen). All steps were done with a Biomek FX Laboratory Automation Workstation. Images for the evaluation of the primary screen were taken with an automated BD Pathway 855 imaging system (Becton-Dickinson) equipped with a 20×/0.75 numerical aperture objective (Olympus) and a monochrome digital black and white camera (Orca-ER, Hamamatsu). For each siRNA treatment, nine areas in each well were imaged to obtain representative data for each perturbation.

Cell culture

UPCI:SCC114 cells (oral squamous cell carcinoma; S. M. Gollin) were cultured in Dulbecco’s minimum essential medium (DMEM) (Invitrogen) supplemented with 10% fetal calf serum (FCS) (Biochrom AG). UPCI:SCC114 cells stably expressing GFP–α-tubulin were maintained under selective pressure by the addition of geniticin (PAA Laboratories). MDA-MB-231 [breast adenocarcinoma; American Type Culture Collection (ATCC)], DLD-1, and SW480 cells (colorectal adenocarcinoma; ATCC) were cultured in RPMI 1640 (Invitrogen) supplemented with 10% FCS; BJ fibroblasts (diploid foreskin fibroblasts; ATCC) were cultured in MEM (Invitrogen) supplemented with 10% FCS, 1% sodium pyruvate, and 1% nonessential amino acids (PAA Laboratories); and hTERT-RPE-1 cells (retinal pigmented epithelium; ATCC) were cultured in DMEM-F12 (Invitrogen) supplemented with 10% FCS and 7.5% sodium bicarbonate (PAA Laboratories). When indicated, the aurora inhibitor ZM447439 (Tocris) or nocodazole (Sigma-Aldrich) was added to the culture medium. ZM447439 and nocodazole were dissolved in DMSO.

Supplementary Material

www.sciencetranslationalmedicine.org/cgi/content/full/2/33/33ra38/DC1

Materials and Methods

Fig. S1. Cyclin B1 staining of cells with multipolar spindles for the identification of metaphase arrest.

Fig. S2. siRNA-mediated knockdown efficacy of randomly chosen screening hits in UPCI:SCC114 cells.

Fig. S3. Spindle multipolarity in MDA-MB-231 cells.

Fig. S4. Spindle multipolarity in hTERT-RPE-1 cells.

Fig. S5. Spindle multipolarity and apoptosis in SW480 colon cancer cells.

Fig. S6. Induction of spindle multipolarity by depletion of proteins involved in cytokinesis.

Table S1. siRNA sequences.

Table S2. Validated genes required for centrosomal clustering.

Table S3. Spindle multipolarity in diploid fibroblasts, mitotic index, and centrin staining after siRNA-mediated knockdown of validated genes required for centrosomal clustering.

References

Footnotes

  • * These authors contributed equally to this work.

  • Citation: B. Leber, B. Maier, F. Fuchs, J. Chi, P. Riffel, S. Anderhub, L. Wagner, A. D. Ho, J. L. Salisbury, M. Boutros, A. Krämer, Proteins required for centrosome clustering in cancer cells. Sci. Transl. Med. 2, 33ra38 (2010).

References and Notes

  1. Acknowledgments: We thank S. M. Gollin for providing UPCI:SCC114 cells; S. Hennemann, B. Schreiter, and J. Czaja for help with image analysis; and M. Czeranski and H. Utecht for technical assistance. We are indebted to S. S. Taylor, E. A. Nigg, and S. Lens for providing antibodies to BUBR1, CEP164, and cDNAs of survivin and borealin, respectively, and S. Poppelreuther (Carl Zeiss), U. Engel (Nikon Imaging Center), and F. Bestvater as well as M. Brom (Light Microscopy Facility, German Cancer Research Center) for technical support. Funding: This study was supported by Deutsche Forschungsgemeinschaft grant KR 1981/3-1 (A.K.) and a Marie-Curie Excellence Grant (EU) and the Helmholtz Alliance for Systems Biology (M.B.). Author contributions: B.L., F.F., M.B., and A.K. conceived and designed the RNAi screening strategy. B.L. and F.F. performed the genome-wide RNAi screen. M.B. and A.K. contributed to the analysis of the primary screen. B.L., J.C., and P.R. did the visual analysis of the screening images. B.L., B.M., F.F., P.R., S.A., and L.W. performed all functional experiments and analyzed the data. J.L.S. provided the centrin antibody and wrote the manuscript. A.D.H. wrote the manuscript. A.K. conceived and designed all experiments and wrote the manuscript. All authors read and contributed to editing the final manuscript. Competing interests: A.K., B.L., B.M., F.F., and M.B. have filed a patent related to the results reported in this paper.
View Abstract

Navigate This Article