Acentrosomal spindle organization renders cancer cells dependent on the kinesin HSET

All animal somatic cells contain centrosomes – organelles with a highly conserved structure and function in ciliogenesis and cell division (Bettencourt-Dias and Glover, 2007; Nigg and Raff, 2009). The centrosome consists of a microtubule (MT)-based pair of centrioles embedded in a pericentriolar protein matrix (PCM). During cell division, centrosomes organize the spindle poles by focusing spindle MT-minus ends and constitute the major driving force in assembling the bipolar spindle. Although sufficient and dominant when present, centrosomes are not absolutely essential for the assembly of a functional bipolar spindle (Basto et al., 2006; Hinchcliffe et al., 2001; Khodjakov and Rieder, 2001; Uetake et al., 2007). In cells lacking centrosomes, such as higher plants and female germ cells or somatic cells following experimental removal of centrosomes, a spindle is formed by centrosome-independent mechanisms (Gadde and Heald, 2004; Kalab and Heald, 2008; O'Connell and Khodjakov, 2007; Schuh and Ellenberg, 2007; Wadsworth and Khodjakov, 2004). In these cells, high levels of Ran-GTP accumulate in the vicinity of chromosomes and drive MT-polymerization from acentrosomal sites. Additionally, acentrosomal MT-nucleation is promoted by the conserved Augmin/HAUS-complex (Goshima et al., 2008; Lawo et al., 2009). The subsequent organization of a bipolar spindle structure with focused spindle poles is mediated by MT-associated molecular motors. HSET and its orthologs of the kinesin-14 family are key players in acentrosomal spindle formation and are evolutionarily conserved from plants to mammals (Loughlin et al., 2008; Verhey and Hammond, 2009; Wordeman, 2010). Via MT-binding sites at their N- and C-termini and a MT-minus end-directed motor activity, kinesin-14 motors crosslink and focus MT-minus ends into two spindle poles. Their crucial role in centrosome-independent pole-focusing is most obvious in acentrosomal Xenopus extracts, Drosophila S2 cells and early embryonic divisions in fly and mouse, where depletion or mutation of kinesin-14 orthologs leads to multipolar spindles and chromosome-misalignment (Endow et al., 1994; Goshima et al., 2005; Hatsumi and Endow, 1992; Mountain et al., 1999; Sköld et al., 2005; Walczak et al., 1997).

Although dispensable in non-transformed human cells with functional centrosomes, HSET is essential for bipolar cell division and survival of cancer cells with supernumerary centrosomes (Godinho et al., 2009; Kwon et al., 2008). Aberrations in centrosome number and structure are commonly observed in solid and hematologic malignancies (Giehl et al., 2005; Lingle et al., 1998; Pihan et al., 1998; Zyss and Gergely, 2009). Strikingly however, cancer cell division in the presence of supernumerary centrosomes remains predominantly bipolar (Godinho et al., 2009; Nigg, 2002) and correlates with centrosome clustering or the detachment of extra centrosomes from the spindle structure – a process often referred to as ‘centrosome inactivation’ (Basto et al., 2008; Kwon et al., 2008; Yang et al., 2008). Recent studies have demonstrated that clustering of supernumerary centrosomes occurs during a prolonged prometa- and metaphase and requires acentrosomal pole-focusing by HSET. Although HSET likely represents the active motor that drives centrosome clustering, additional spindle assembly factors such as the MT-generating Augmin/HAUS-complex were also shown to contribute to stable bipolarity in cancer cells with supernumerary centrosomes (Kwon et al., 2008; Leber et al., 2010).

Here, we show that acentrosomal MT-organization mediated by HSET is required for bipolar spindle formation in cancer cells, irrespective of normal or supernumerary centrosome number. We demonstrate that mitotic spindle assembly in the analyzed cancer cells is characterized by the formation and subsequent incorporation of pronounced acentrosomal MTOCs into the assembling spindle structure. Upon HSET-depletion, proper spindle assembly and stable pole-focusing is defective and spindle poles progressively fragment into multipolar spindles. Furthermore we show that in checkpoint-defective cancer cells acentrosomal spindle formation and HSET-dependence correlate with the activation of the DNA damage response. In summary, we propose that oncogenic transformation and/or persistent DNA damage in checkpoint-defective cancer cells can hyperactivate acentrosomal spindle assembly mechanisms and thereby render cancer cells dependent on HSET, irrespective of centrosome number.

First, we assessed the relevance of HSET in two cancer cell lines, namely BT549 breast cancer and IGR39 melanoma cells, and compared it to non-transformed RPE1 cells as a control. In all cell lines the MT-associated kinesin HSET decorated MTs and poles of the mitotic spindle (Fig. 1A). In agreement with previous results (Kwon et al., 2008), efficient HSET depletion by siRNA induced multipolar mitoses in cancer cell lines, but not in RPE1 cells (Fig. 1A–C). Strikingly, however, the percentage of cancer cells with multipolar spindles did not correlate well with the frequency of supernumerary centrosomes (42% versus 21% for BT549 cells and 57% versus 18% for IGR39 cells, respectively). Furthermore, HSET-depletion did not significantly enhance the frequency of supernumerary centrosomes (Fig. 1B). Instead, our data indicated that HSET is also involved in bipolar spindle formation in cancer cells with normal centrosome number. We additionally noticed a pronounced reduction of spindle MT density in HSET-depleted cancer cells, a phenomenon which was also observed to a lesser extent in the central region of bipolar spindles in HSET-depleted RPE1 cells (Fig. 1A). By using anti-HURP antibodies which specifically visualize chromatin-associated kinetochore fibers, we were able to confirm these observations (supplementary material Fig. S1).

To discriminate between normal and supernumerary centrosome number, control- or HSET siRNA-treated cancer cells were stained with anti-γ-tubulin and anti-Cep135 antibodies to visualize bona fide centrosomes. The majority of cancer cells clearly contained two centrosomes usually associated with the poles of the bipolar spindle (Fig. 1D, illustrated by gray spindle and bars). In addition, we detected a small fraction of cancer cells (3% for both BT549 and IGR39) with bipolar spindles formed by acentrosomal poles (Fig. 1D, illustrated by blue spindle and bars, arrows indicate acentrosomal poles). In these cells either one or two free centrosomes (arrowheads) were found at some distance from the acentrosomal spindle pole(s). Finally, some cancer cells also formed multipolar spindles. Besides two centrosomal poles, these multipolar spindles contained additional pole structures devoid of bona fide centrioles (Fig. 1D, illustrated by red spindle and bars). Strikingly, this multipolar spindle phenotype in cells with two centrosomes increased dramatically upon depletion of HSET (44% in BT549 and 48% in IGR39 cells, Fig. 1D, red bars). These observations show that functional spindle poles are not necessarily organized by bona fide centrosomes in BT549 and IGR39 cancer cells, and that stable bipolarity critically depends on HSET-mediated MT-organization.

We also observed acentrosomal pole focusing and free centrosomes in cancer cells with supernumerary centrosomes (Fig. 2). As expected, mitotic spindles in control-treated cancer cells were predominantly bipolar despite the presence of multiple centrosomes. However, while clustering of extra centrosomes was only detected in a minor subpopulation of BT549 and IGR39 cells (Fig. 2, illustrated by gray spindle and bars), supernumerary centrosomes were predominantly found to be free and dislocated from the spindle (Fig. 2, illustrated by green and blue spindles and bars). Of note, cancer cells with supernumerary centrosomes contained a higher fraction of bipolar spindles with free centrosomes and acentrosomal poles (14%) than those with normal centrosome number (3%) (Fig. 1D; Fig. 2, blue bars). Upon HSET RNAi, 90% of cancer cells with supernumerary centrosomes displayed multipolar spindles. However, de-clustering of multiple centrosomes could only account for 33% and 29% of multipolar spindles observed in BT549 and IGR39 cells, respectively (Fig. 2, illustrated by brown spindle and bars). Strikingly, up to two thirds of HSET-depleted cancer cells with supernumerary centrosomes displayed multipolar spindles due to spindle fragmentation (Fig. 2, illustrated by red spindle and bars). This demonstrates that acentrosomal pole-focusing by HSET is required for bipolar spindle formation in BT549 and IGR39 cancer cells irrespective of the centrosome number. In some cancer cells acentrosomal pole formation eventually correlated with the co-existence of free centrosomes.

To better understand how spindle multipolarity emerged in HSET-depleted cancer cells with two centrosomes, we used time-lapse microscopy. In control-treated BT549 cells, bipolar spindles formed within 20 minutes and most cells progressed into anaphase within 36–55 minutes (Fig. 3A, upper panel, Fig. 3B; supplementary material Movie 1). Upon HSET-depletion, cancer cells displayed severe defects in spindle assembly including a severe reduction of MT-density and progressive fragmentation of the spindle poles. Defective spindle assembly was also accompanied by a delayed progression into anaphase (Fig. 3A, lower panel; supplementary material Movie 2). While few HSET-depleted cancer cells managed to initiate anaphase within 60 minutes, meta-to-anaphase transition was usually blocked for several hours and mitosis was either aborted or followed by cell death (Fig. 3A,B). Defects in spindle assembly, mitotic progression and cell death were also observed in HSET-depleted cancer cells that maintained a bipolar spindle. Notably, neither bipolar spindle formation nor mitotic progression was impaired in HSET-depleted RPE1 cells (Fig. 3B; supplementary material Movies 3, 4). Accordingly, increased apoptosis following HSET-depletion was detected only in the cancer cell lines and was accompanied by increased levels of cleaved PARP1 in the respective cell lysates (Fig. 3C). Taken together, these results confirm that HSET is essential for spindle formation in BT549 and IGR39 cells but is dispensable in normal RPE1 cells. Moreover, spindle assembly defects in the absence of HSET correlated with progressive spindle fragmentation and a severe delay in the meta-to-anaphase transition, likely caused by activation of the spindle assembly checkpoint.

To confirm the relevance of acentrosomal spindle organization in the cancer cell lines by a second approach, we assessed spindle bipolarity in the absence of the Augmin/HAUS-complex using HAUS6 siRNA (Lawo et al., 2009). The conserved Augmin/HAUS-complex is known to be required for acentrosomal MT-generation and bipolar spindle organization in the absence of functional centrosomes (Goshima et al., 2008; Lawo et al., 2009; Wainman et al., 2009). Moreover, Augmin/HAUS facilitates bipolar cell division in cancer cells with supernumerary centrosomes (Leber et al., 2010). In agreement with previous results (Goshima et al., 2008), all HAUS6-depleted spindles displayed reduced spindle MT-density (Fig. 4A). However, we demonstrate that while HAUS6-depletion induced spindle fragmentation in mitotic cancer cells, RPE1 cells retained a bipolar phenotype (Fig. 4A–C). By using centrin-2 as additional, distal centriole marker, we could exclude the presence of extra centrosomes or centriole splitting as potential underlying causes for spindle multipolarity in HAUS6- and HSET-depleted cancer cells (supplementary material Fig. S2). Taken together, these results demonstrate that centrosomes are not sufficient for assembling functional bipolar spindles in BT549 and IGR39 cancer cells, but require acentrosomal MT-generation and -organization by the Augmin/HAUS complex and the kinesin HSET.

To understand whether HSET-dependent pole-focusing in cancer cells could be a consequence of mitotic centrosome inactivation, we examined centrosome function in BT549 cancer cells by time-lapse microscopy. Although spindle MT-polymerization was observed at all mitotic centrosomes during prophase and prometaphase, individual centrosomes progressively detached from the assembling spindle during metaphase (Fig. 5A, arrowheads; supplementary material Movie 5). These free centrosomes remained dislocated from the spindle during chromosome-segregation and eventually re-established spindle attachment upon progression into anaphase. Hence, the transient detachment of centrosomes from the assembling spindle cannot reflect a centrosome-intrinsic loss of MT-nucleation activity. Acentrosomal spindle formation and centrosome inactivation have previously been reported in Drosophila embryos and Xenopus egg extracts as a result of DNA damage (Sibon et al., 2000; Smith et al., 2009; Takada et al., 2003). In these systems centrosome inactivation was consistently linked to impaired centrosomal MT-nucleation and the loss of centrosomal PCM components including γ-tubulin and Cep63. In this current study, however, we failed to detect a loss of PCM components required for mitotic centrosome function and centrosomal spindle assembly in human cancer cells (Doxsey et al., 1994; Fong et al., 2008; Gomez-Ferreria et al., 2007; Graser et al., 2007; Haren et al., 2009; Sir et al., 2011; Smith et al., 2009; Zhu et al., 2008; Zimmerman et al., 2004). Instead, we consistently detected γ-tubulin, Cep63, Pericentrin, Cep192 (the human ortholog of Drosophila SPD2) and Cep215 (the human ortholog of Drosophila centrosomin) at free mitotic centrosomes (Fig. 5B, arrowheads indicate free centrosomes). This indicates that mitotic centrosomes are functional and active in the cancer cells and that, in line with previous results (Yang et al., 2008), free centrosomes emerged by a transient detachment from the spindle structure.

To confirm that centrosomes were indeed functionally active in the cancer lines, we performed MT-regrowth assays in BT549, IGR39 and RPE1 cells. Therefore, spindle MTs cells were first depolymerized by cold treatment before MT re-polymerization was allowed at 37°C to determine the sites of MT-nucleation. Indeed, we observed rapid MT-polymerization from mitotic centrosomes – both in non-transformed RPE1 and cancer cell lines, demonstrating that centrosomes were functional in the latter (Fig. 6A). In addition, spindle MT-polymerization also occurred from acentrosomal sites in the vicinity of chromatin. However, while acentrosomal MTOCs were small and rather rare (9%) in mitotic RPE1 cells, pronounced acentrosomal MTOC structures were found in 22% and 47% of mitotic BT549 and IGR39 cancer cells, respectively (Fig. 6B, red bars). Strikingly, some of these acentrosomal MTOCs in cancer cells grew into structures equal to the size of bona fide centrosomes. When comparing the two cancer cell lines, we noticed subtle differences concerning the size and the number of acentrosomal MTOCs. While acentrosomal MTOCs were numerous and of moderate size in IGR39 cells they were less numerous but of bigger size in BT549 cells (as shown in Fig. 6A). In all cells multiple acentrosomal MTOCs were commonly detected during the first 2–5 minutes of MT-regrowth. However, after 15 minutes of MT-regrowth, multipolar spindles with acentrosomal poles were rarely observed (Fig. 6B, gray bars). This indicates that centrosomal and acentrosomal MTOCs had been incorporated into one bipolar spindle structure. In agreement with its role in acentrosomal MT-nucleation, we also localized the Augmin/HAUS-complex at acentrosomal MTOCs (supplementary material Fig. S3). In summary, this indicates that the Augmin/HAUS-complex contributes to proper spindle formation in cancer cells by facilitating acentrosomal MT-nucleation while HSET ensures spindle bipolarity by clustering and focusing acentrosomal and centrosomal MTOCs into two spindle poles.

Acentrosomal spindle formation in Drosophila embryos and Xenopus egg extracts has been linked to DNA damage and active DNA damage signaling (Sibon et al., 2000; Smith et al., 2009; Takada et al., 2003). Interestingly however, although constitutive DNA damage is commonly found in human cancer cells (Bartkova et al., 2005; Tort et al., 2006), a causative link between DNA damage and acentrosomal spindle organization in human cells has not yet been established. This prompted us to address whether Doxorubicin- (DOX) or Zeocin- (ZEO) mediated DNA breaks, and Hydroxyurea- (HU) mediated stalled replication forks, could also induce acentrosomal spindle organization in cancer cells. As expected, even low concentrations of DNA damaging agents activated the DNA damage response (DDR) in RPE1 cells, which included phosphorylation of ATM and the ATM/ATR substrates Chk1, Chk2 and p53 (Fig. 7A) as well as a robust cell-cycle arrest (supplementary material Table S1). Thus, we failed to detect mitotic RPE1 cells in the presence of Hydroxyurea, Doxorubicin or Zeocin. In contrast however, although we observed pronounced DDR stimulation upon drug treatment (Fig. 7A), BT549 and IGR39 cancer cells failed to induce a robust G2/M block (supplementary material Table S1) and a significant fraction of these cancer cells entered mitosis in the presence of pronounced DNA damage (Fig. 7B). This allowed us to study the effects of DNA damage on mitotic spindle organization. As DNA damage was shown to induce centrosome amplification and centriole splitting (Bourke et al., 2007; Dodson et al., 2004; Hut et al., 2003), we focused the analysis on cancer cells with normal centrosome number and structure – determined by two Cep135-positive centriole pairs in mitosis (Fig. 7B,C, graph). While present in only 2–3% of control-treated BT549 cells, bipolar spindles with acentrosomal poles and free centrosomes were apparent in 8–9% of cells after 16 h and in 8–28% of cells after 48 h of drug treatment (Fig. 7C, illustrated by blue spindles and bars). Notably, Doxorubicin and Zeocin were more potent than Hydroxyurea in inducing acentrosomal spindles in cancer cells with normal centrosome number. As already shown in untreated BT549 cells, we detected no changes in the centrosomal localization of Cep63, Cep192, Cep215 and Pericentrin when comparing pole-associated with free centrosomes in the presence of Doxorubicin (supplementary material Fig. S4). In addition to acentrosomal bipolar spindles, we observed a slight increase of multipolar spindles with additional acentrosomal poles upon Doxorubicin- and Zeocin-treatment (Fig. 7C, illustrated by red spindles and bars).

To address whether pronounced DNA damage could also enhance HSET-dependence, we analyzed spindle morphologies in HSET-depleted BT549 cells in the presence of Doxorubicin (Fig. 7D). Notably, exclusively cancer cells with two bona fide centrosomes were considered, as determined by centriolar Cep135-staining. As already shown in Fig. 1D, 40% of HSET-depleted cells displayed fragmented spindles in the absence of Doxorubicin. However, spindle fragmentation in HSET-depleted cells was significantly increased up to 52% upon Doxorubicin-treatment (16 h). Since this observed increase of multipolar mitoses was not due to the presence of extra centrosomes or splitting of the two centriole pairs, these results show that DNA damage promoted spindle fragmentation in HSET-depleted cells (Fig. 7D, illustrated by red spindle and bars). In a final series of experiments we attempted to address whether endogenous DNA damage signaling could be the underlying cause of HSET-dependence in BT549 cells. Therefore, we suppressed endogenous DNA damage signaling in HSET-depleted cells, and examined spindle morphologies in cancer cells with two centrosomes (Fig. 7E, spindle morphologies as illustrated in Fig. 7C). We used either Caffeine to block both ATM- and ATR-dependent DNA damage signaling, or Ku55933 to selectively block ATM-activity alone (Fig. 7E). Interestingly, we found that Caffeine but not Ku55933 treatment could partially restore spindle bipolarity in the absence of HSET. These observations suggest that constitutive ATM- and ATR-dependent DNA damage signaling could contribute to HSET-dependent spindle organization in cancer cells with normal centrosome number.

Since our results indicated that HSET may represent an attractive therapeutic cancer target, we analyzed HSET mRNA expression in primary tumor tissue in comparison to corresponding normal tissue. As illustrated in Fig. 8A, although HSET was ubiquitously expressed in all samples, its expression was significantly elevated in a broad panel of cancer tissues (Fig. 8A). In summary, our data indicate that HSET-dependent MT-organization is particularly important for mitotic cancer cells, strengthening the notion that HSET could be an attractive target for novel cancer therapies.

Although centrosome-independent MT-organizing mechanisms are active in centrosome-containing somatic cells, their relative contribution to mitotic spindle formation is ill defined. Recently, HSET has been proposed as a potential target for cancer therapy since its depletion induces centrosome de-clustering and cell death in cancer cells with supernumerary centrosomes. However, little is known about how cancer cells with supernumerary centrosomes impact tumor growth and thus whether targeting these cells would be an efficacious therapy. Here, we report a more general and critical role for the kinesin HSET in cancer cells, irrespective of centrosome number (summarized in Fig. 8B). Consequently, our results significantly extend the relevance of this kinesin in cancer and thereby suggest a novel rationale for the use of HSET-inhibition in cancer therapy.

Our study demonstrates that HSET and the Augmin/HAUS-complex are essential for proper spindle assembly and spindle integrity in BT549 and IGR39 cancer cells but dispensable in non-transformed RPE1 cells. However, not all cancer cells are equally HSET-dependent since depletion of this kinesin in HeLa, MCF7 and MDA231 cancer cells did not induce multipolar spindle formation in the presence of normal centrosome number (Cai et al., 2009; Kwon et al., 2008). Noteworthy, mitotic spindle formation is mediated by a balance of centrosomal and acentrosomal forces, which might vary from one cell line to another. Since HSET is usually masked by functionally dominant centrosomes in untransformed cells, HSET-dependence in the analyzed cancer cell lines implied a shift towards acentrosomal mechanisms – potentially via centrosome inactivation or hyperactivation of acentrosomal MT-polymerization. Of note, DNA damage-induced centrosome inactivation has been shown to correlate with a loss of pericentriolar proteins such as γ-tubulin or Cep63 from mitotic centrosomes (Sibon et al., 2000; Smith et al., 2009; Takada et al., 2003). In contrast, in our study, centrosomes of the analyzed cancer cells were active and all analyzed PCM components were localized correctly. Instead, spindle formation was characterized by the generation of pronounced acentrosomal MTOCs during spindle assembly. These acentrosomal MTOCs varied strongly in size and even grew to the size of bona fide centrosomes. In some cancer cells, acentrosomal MTOC organization eventually replaced centrosomal MTOC function and led to the formation of acentrosomal bipolar spindles and free centrosomes. We propose that hyperactive acentrosomal MT-polymerization could drive the formation of acentrosomal MTOCs and thereby generate a requirement for HSET-dependent pole-focusing. This would be highly reminiscent of spindle formation in meiotic cells where acentrosomal MTOCs are clustered into a bipolar spindle by the kinesin HSET/Ncd (Mountain et al., 1999; Schuh and Ellenberg, 2007). Similarly, Ncd is required for acentrosomal MTOC clustering in S2 cells lacking bone fide centrosomes (Moutinho-Pereira et al., 2009). Based on this evolutionarily conserved role of the kinesin-14 HSET in MTOC-clustering and spindle pole organization, we propose that HSET mediates bipolar spindle formation in cancer cells with two bona fide centrosomes by clustering and focusing additional acentrosomal MTOCs into the assembling spindle (Fig. 8B).

HSET has been identified as a target of Ran-GTP (Cai et al., 2009; Ems-McClung et al., 2004; Walczak et al., 1997), a key regulator of chromatin-driven MT-polymerization in mitotic cells (Clarke and Zhang, 2008; Kalab and Heald, 2008). Upon nuclear envelope breakdown, high levels of Ran-GTP mediate the release of specific MAPs, including HSET, from their binding partner importin β and thereby activate acentrosomal MT-polymerization and -organization in the proximity of chromosomes. Ran-dependent MT-polymerization is highly active until Ran-target proteins are degraded by the APC/C during anaphase (Gruss et al., 2001; Koffa et al., 2006; Nachury et al., 2001; Ribbeck et al., 2006; Silljé et al., 2006; Song and Rape, 2010; Stewart and Fang, 2005; Wiese et al., 2001). Recently, Xia et al. reported that Ran expression was elevated in human tumor tissue compared to normal tissue and proposed that Ran overexpression represents a mechanism for hyperactivation of Ran-driven spindle assembly in cancer cells (Xia et al., 2008).

We hypothesize that deregulation of Ran expression, activation or localization could also promote the formation of pronounced acentrosomal MTOCs and thereby render cancer cells dependent on HSET.

The deregulated progression into mitosis of BT549 and IGR39 cancer cells with exogenously induced DNA damage, allowed us to detect that acentrosomal spindle assembly is a consequence of DNA damage in human cells. We excluded DNA damage-mediated effects on mitotic centrosomes, such as centrosome amplification and centriole splitting, by strictly focusing on cells with two intact centrosomes. Moreover, we excluded centrosome inactivation as underlying cause for acentrosomal spindle organization. Interestingly, both cancer lines in this study expressed mutant p53 (R249S for BT549 and C229fs for IGR39), a key mediator of the DNA damage response and cell-cycle regulation. Since p53 is also critical for a functional G2/M checkpoint (Bunz et al., 1998), we hypothesize that the loss of p53 function in BT549 and IGR39 cancer cells contributed to deregulated cell-cycle progression and entry into mitosis following DNA damage. Interestingly, we observed that HU-mediated stalled replication forks were less potent in inducing acentrosomal spindle organization than Doxorubicin- and Zeocin-mediated DNA breaks. This could indicate that the cell-cycle stage during which the DNA damage occurs as well as the specific repair mechanism that is induced, determines whether or not the balance between acentrosomal and centrosomal mechanisms is shifted in the subsequent cell division. Since the DNA damage response was constitutively active in the studied cancer cell lines and suppression of ATM/ATR could partially prevent spindle fragmentation following HSET depletion, we suggest that active DNA damage signaling contributed to the regulation of mitotic spindle assembly and the observed HSET-dependence in cancer cells.

In summary, we propose that oncogenic transformation and/or constitutive DNA damage signaling can induce the hyperactivation of acentrosomal MT-nucleation and spindle organization in cancer cells with defective cell-cycle checkpoints, possibly by regulating Ran expression, activation or localization. Consequently, HSET, as a key player of acentrosomal spindle formation, is upregulated in tumor tissue, rendering cancer cells more sensitive to HSET-inhibition than non-transformed cells.

RPE1 (hTERT-immortalized human Retinal Pigment Epithelial cells), BT549 (breast cancer) and IGR39 (melanoma) cells were grown in DMEM/F12 (RPE1), DMEM (IGR39), or RPMI (BT549) supplemented with 10% FCS at 37°C, 5% CO₂. To generate α-tubulin-GFP expressing cell lines, RPE1 and BT549 cells were infected with lentiviral constructs expressing α-tubulin-GFP-pLKO-TREX [GenBank accession number BC010494, (Moffat et al., 2006)] and selected in the presence of 0.5 µg/ml Geneticin (Invitrogen, Carlsbad, CA). Protein expression was induced by the addition of 200 ng/ml Doxycycline for 24 h (Sigma, St Louis, MO). Cells with high expression levels were enriched by FACS-sorting. 2 mM Thymidine (Sigma) was used to synchronize cells in G1/S. Cells were treated with 0.1 µM Doxorubicin (Adriamycin, Pfizer, New York, NY), 50 µg/ml Zeocin (Invitrogen) or 0.1 mM Hydroxyurea (ABCR GmbH & Co., KG, Karlsruhe, Germany) to induce DNA damage and genotoxic stress. 5 mM Caffeine and 10 µM Ku55933 (Calbiochem) were used to suppress ATM- and ATR-activity.

RPE1, BT549 and IGR39 cells were grown on coverslips before microtubules were depolymerized by putting the cells in ice for 60 minutes. MT-regrowth was induced by rapidly supplementing the cells with pre-warmed medium and shifting them to 37°C. Cells were fixed after 2 and 15 minutes of MT-regrowth and processed for immunofluorescence microscopy.

The following antibodies were used for immunofluorescence microscopy and western blotting. Anti-HSET (Bethyl Laboratories, Montgomery, TX), anti-HSET (Abcam, Cambridge, UK), anti-γ-tubulin (GTU-88, Sigma), anti-α-tubulin-FITC (DM1A, Sigma), anti-PARP1 (NO.9542), anti-phospho-ATM-S1981 (NO.4526), anti-phospho-Chk1-S317 (NO.2344), anti-phospho-Chk2 -T68 (NO.2661), anti-phospho-p53-S15 (NO.9284), anti-γH2AX-S139 (NO.2577) (all Cell Signaling Technology, Danvers, USA), anti-Cep63 (Acris), anti-Pericentrin (Abcam), anti-centrin-2 (Santa Cruz), anti-Cep135 (Kleylein-Sohn et al., 2007), anti-CP110 (Kleylein-Sohn et al., 2007) anti-HAUS6 (Lawo et al., 2009) and anti-Cep215 (Graser et al., 2007), anti-Cep192 (Schmidt et al., 2009), anti-HURP (Silljé et al., 2006). Secondary Alexa-Fluor 568-conjugated anti-rabbit and Alexa-Fluor 680-conjugated anti-mouse antibodies were purchased from Invitrogen. Hoechst 33342 (Sigma) was used to stain DNA. Secondary anti-mouse and anti-rabbit HRP-coupled antibodies were purchased from GE Healthcare (Chalfont St Giles, UK). For western blotting, equal amounts of cell lysate were separated by SDS-page using 4–12% gradient gels (Invitrogen) and processed for western blotting using Supersignal West Dura chemiluminescent reagents (Thermo Scientific, Rockford, IL).

A pool of four oligos was used for HSET siRNA (Dharmacon, ON-TARGETplus, set of four, L-004958-00-0010). HAUS6 was depleted using the target sequence 5′-CCATTGTCAGATGTTGCAAAGAAT-3′ (Invitrogen). An siRNA duplex targeting luciferase was used as a GL2 control (Qiagen AG). 50 nM siRNA oligos were transfected into cells using Lipofectamine RNAiMAX (Invitrogen) according to the manufacturer's instructions.

For immunofluorescence microscopy, cells were grown on glass coverslips and fixed with ice-cold methanol for 10 minutes at −20°C. Antibody incubations and washings were performed as described previously (Meraldi et al., 1999). For time-lapse microscopy, cells were grown on Ibidi slides (Ibidi, Martinsried, Germany) and co-transfected with GL2 control siRNA or HSET siRNA and a H2B-mCherry plasmid for 48 h, synchronized by a single Thymidine block (20 h), and released into fresh medium supplemented with 20 mM Hepes (Invitrogen). Expression of α-tubulin-GFP was induced for 24 h by the addition of 200 ng/ml Doxocycline. Cells were analyzed using a Zeiss Axioplan microscope or a spinning disc microscope (Olympus IX81) equipped with a Yokogawa CSU-X1 scan head, an Olympus 60×/1.45 or a 100×/1.4 n.a. oil immersion objective, an ASI MS-2000 with Z-piezo stage and a Cascade II EM-CCD camera (Photometrics, Tucson, AZ). For live cell imaging, the setup was enclosed in a heating box and temperature was set at 37°C and controlled with a ‘Box’ element (Life Cell Imaging, Basel, Switzerland). MetaMorph software 7.6 (Molecular Devices) was used to collect data. Confocal sections were collected at 0.2 µm- (fixed cells) or 0.5 µm- (living cells) steps through the whole cell volume and projected into one plane using ImageJ (National Institutes of Health, Bethesda, MD). All images were processed equivalently for control and experimental samples and assembled using ImageJ and Adobe Photoshop (Adobe, San Jose, CA). Movies were assembled in ImageJ and are displayed at 15 frames per second.

HSET mRNA expression was retrieved from an internal collection of publically available, oncology-relevant Affymetrix Human U133 Plus 2.0 arrays. HSET (Gene ID 3833, Probeset 209680_s_at) MAS5 summarized expression is plotted as log2 values. Statistical analysis was done based on log2 values using an unpaired Student's t-test.

Unless defined differently results are expressed as the means ± s.d. from an appropriate number of experiments as indicated in the figure legends. The statistical analysis was done using an unpaired Student's t-test. P>0.05 was considered not significant (n.s.) (Graphpad, La Jolla, CA).

We would like to thank Prof. S. Gasser for generously sharing reagents and providing conceptual advice. We thank Dr L. Pelletier, Dr A. Krämer, Dr F. Stegmeier, Y. W. Chan, D. A. Guthy, M. Hattenberger and E. Billy for generously sharing reagents. Special thanks to Dr A. Littlewood-Evans for critical comments on the manuscript. We also thank Dr L. Gelman and H. Kohler for technical help with microscopy and FACS-sorting.