Oncogene (2011) 30, 521–534 & 2011 Macmillan Publishers Limited All rights reserved 0950-9232/11 www.nature.com/onc ORIGINAL ARTICLE A critical role of integrin-linked kinase, ch-TOG and TACC3 in centrosome clustering in cancer cells

AB Fielding1, S Lim1, K Montgomery1, I Dobreva1 and S Dedhar1,2

1Department of Integrative Oncology, British Columbia Cancer Research Centre of the BC Cancer Agency, Vancouver, British Columbia, Canada and 2Department of Biochemistry and Molecular Biology, Life Sciences Institute, University of British Columbia, Vancouver, British Columbia, Canada

Many cancer cells contain more than two centrosomes, Introduction which imposes a potential for multipolar mitoses, leading to cell death. To circumvent this, cancer cells develop Supernumerary centrosomes, a state in which cells mechanisms to cluster supernumerary centrosomes to contain more than two centrosomes, are a common form bipolar spindles, enabling successful mitosis. Dis- characteristic of human cancer cells. Supernumerary ruption of centrosome clustering thus provides a selective centrosomes were first observed in cancer cells nearly means of killing supernumerary centrosome-harboring 100 years ago (Boveri, 1914) and have now been cancer cells. Although the mechanisms of centrosome reported in many malignancies in vivo, including clustering are poorly understood, recent genetic analyses bladder, brain, breast, bile duct, cervical, colon, head have identified requirements for both actin and tubulin and neck, liver, lung, ovarian, pancreas and prostate regulating proteins. In this study, we demonstrate that the tumors, as well as myeloma, lymphomas and leukemias integrin-linked kinase (ILK), a protein critically involved (Lingle and Salisbury, 1999; Kramer et al., 2002; Nigg, in actin and mitotic microtubule organization, is required 2002; Pihan et al., 2003; Zyss and Gergely, 2009). for centrosome clustering. Inhibition of ILK expression or Although supernumerary centrosomes can poten- activity inhibits centrosome clustering in several breast tially facilitate tumorigenesis by disrupting cellular and prostate cancer cell lines that have centrosome polarity (Nigg, 2006; Basto et al., 2008) and increasing amplification. Furthermore, cancer cells with supernumer- the likelihood of aneuploidy (Lingle et al., 2002; Ganem ary centrosomes are significantly more sensitive to ILK et al., 2009; Silkworth et al., 2009), they also impose a inhibition than cells with two centrosomes, demonstrating potential for multipolar mitoses leading to cell death that inhibiting ILK offers a selective means of targeting (Brinkley, 2001; Nigg, 2002; Ganem et al., 2009). To cancer cells. Live cell analysis shows ILK perturbation circumvent this, many cancer cells develop mechanisms leads cancer cells to undergo multipolar anaphases, to cluster supernumerary centrosomes to form bipolar mitotic arrest and cell death in mitosis. We also show spindles and hence, ensure bipolar mitosis and cell that ILK performs its centrosome clustering activity in a survival (Ring et al., 1982; Brinkley, 2001; Quintyne focal adhesion-independent, but centrosome-dependent, et al., 2005; Rebacz et al., 2007; Kwon et al., 2008). manner through the microtubule regulating proteins Disruption of centrosome clustering and induction of TACC3 and ch-TOG. In addition, we identify a specific multipolar spindle formation thus provides a selective TACC3 phosphorylation site that is required for centro- means of killing supernumerary centrosome-harboring some clustering and demonstrate that ILK regulates this cancer cells (Kwon et al., 2008; Xu and Saunders, 2008). phosphorylation in an Aurora-A-dependent manner. Although details remain poorly understood, recent Oncogene (2011) 30, 521–534; doi:10.1038/onc.2010.431; genetic analyses have identified three broad, overlapping published online 13 September 2010 mechanisms that regulate centrosome clustering (Kwon et al., 2008). First, several microtubule-interacting/ Keywords: integrin-linked kinase; centrosome cluster- regulating genes were identified, including Ncd (human ing; multipolar mitosis; TACC3; ch-TOG/XMAP215; spleen, embryo, testes protein (HSET)) and D-TACC cancer (Kwon et al., 2008). D-TACC, a protein involved in stabilizing the minus and plus ends of microtubules at centrosomes in concert with the microtubule-associated protein ch-TOG/XMAP-215/CKAP5 (Lee et al., 2001; Barros et al., 2005), was not investigated further in this report. The second broad group of genes identified to be Correspondence: Dr S Dedhar, Department of Biochemistry, Uni- required for the prevention of multipolar mitoses were versity of British Columbia, BC Cancer Research Centre, 675 West those involved either directly in actin cytoskeleton 10th Avenue, Vancouver, British Columbia, Canada V5Z1L3. E-mail: [email protected] organization, or those which contribute to interphase Received 22 January 2010; revised 4 August 2010; accepted 9 August cell shape by linking the extra-cellular-matrix to the 2010; published online 13 September 2010 actin cytoskeleton. Centrosome clustering by ILK, ch-TOG and TACC3 AB Fielding et al 522 Integrin-linked kinase (ILK) localizes to focal adhe- markers of microtubules, centrosomes and DNA to sions and is critically involved in both actin and cell enable the analysis of multipolar spindle formation. The adhesion regulation (for reviews see Legate et al., 2006; resulting phenotypes were scored as illustrated in McDonald et al., 2008). Moreover, several reports Figure 1a and the percentage of ‘de-clustered’ centro- demonstrate that ILK has a role in mitosis by regulation somes calculated for each treatment. Approximately of the microtubule cytoskeleton. Mitotic defects have 10% of control-treated BT549s contained de-clustered been detected upon ILK depletion in Drosophila S2 cells centrosomes and multipolar spindles. This is in accor- (Bettencourt-Dias et al., 2004), mouse hepatocytes dance with data from Kwon et al, who reported 14% of (Gkretsi et al., 2007) and human glioblastoma cells control-treated BT549 cells showed multipolar spindles (Koul et al., 2005; Edwards et al., 2008). It is interesting (Kwon et al., 2008). However, ILK siRNA treatment that, although the cellular phenotypes were not reported resulted in greater than 36% of the cells displaying de- in detail in these studies, all of these systems contain clustered centrosomes (Figures 1a and c). A similar supernumerary centrosomes (Weber et al., 1998; Gui- pattern was observed in MDA-MB-231 cells and also, as dotti et al., 2003; Margall-Ducos et al., 2007; Kwon an example of a different cancer type, prostate-derived et al., 2008). ILK has also been shown to localize to PC3 cells (Figures 1b and c). These results imply that centrosomes and regulate mitotic microtubule organiza- ILK participates in the clustering of supernumerary tion, likely through its influences on Aurora-A/ch-TOG/ centrosomes in both breast and prostate cancer cells. TACC3 complex formation (Fielding et al., 2008). It These experiments were performed using pericentrin to is interesting that the ILK-interacting proteins at stain for centrosomes because of the availability of a both focal adhesions (Mig-2 (Tu et al., 2003)) and the highly specific antibody against this protein. However, centrosomes (TACC3 (Fielding et al., 2008)) have been because of the potential difficulties of distinguishing implicated in centrosome clustering (Kwon et al., 2008) between true centriole-containing centrosomes and and ILK itself has been suggested as a candidate for other sites of microtubule nucleation that may recruit regulating this process (Xu and Saunders, 2008). There- components of the soluble pericentriolar-material, such fore, we sought to test whether ILK is required for as pericentrin, multipolar spindles induced by ILK centrosome clustering and if so whether inhibiting ILK perturbation were also examined with an antibody can selectively inhibit the growth of cancer cells. against centrin 2, one of the core structural components In this study, we demonstrate that ILK is required for of centrioles. This indicated that nearly all spindle poles centrosome clustering, and inhibiting ILK expression examined in ILK-depleted cells were due to the presence with small interfering RNA (siRNA) or activity with a of true centrosomes as each stained for a pair of centrin small molecule inhibitor, induces multipolar spindles in 2 foci (Supplementary Figure 1). Quantification of the several breast and prostate cancer cell lines that have percentage of de-clustered centrosomes in ILK-depleted centrosome amplification. Furthermore, cancer cells cells as judged by centrin 2 staining gave very similar with supernumerary centrosomes are more sensitive to results to the pericentrin quantifications (Supplementary ILK inhibition than cells with two centrosomes and Figure 1c). show defects in cell division resulting in decreased cell Next, we sought to determine whether inhibiting ILK proliferation. We also found that the centrosomal kinase activity using a pharmacological ILK inhibitor proteins TACC3 and ch-TOG, which co-localize, inter- could also prevent centrosome clustering and whether act with and are regulated by ILK (Dobreva et al., 2008; this effect was specific to cancer cells. Fielding et al., 2008) are required for centrosome clustering in mammalian cells, whereas key focal adhesion binding partners of ILK are not. ILK per- Pharmacological inhibition of ILK leads to multipolar forms its centrosome-clustering functions independently spindle formation in breast and prostate cancer but not of the actin cytoskeleton and by its centrosomal, normal epithelial cells microtubule-regulating partner TACC3. In addition, For analysis of the effects of the small-molecule we show that a specific TACC3 phosphorylation event is inhibitor of ILK kinase activity, QLT-0267 (character- required for centrosome clustering and that this is ized in Troussard et al., 2006), the panel of breast cells regulated by ILK, in an Aurora-A-dependent manner. lines was expanded to include a breast cancer cell line that has a low percentage of supernumerary centro- some-harboring cells (MCF7 (Kwon et al., 2008)) and two normal breast epithelial cell lines, which contain Results a very low percentage of cells with greater than two centrosomes, MCF10As (You et al., 2004) and ILK knockdown prevents centrosome clustering in breast 184-hTERTs (our own, unpublished data. See Supple- and prostate cancer cell lines mentary Table 1 for additional information). BT549 and MDA-MB-231 breast cancer cell lines have A range of QLT-0267 concentrations were added for high percentages of cells harboring greater than two 6 h and the percentage of cells displaying de-clustered centrosomes (45 and 44%, respectively (Kwon et al., centrosomes calculated. BT549 and MDA-231 cells 2008), see Supplementary Table 1 for further details). show a dose-dependent increase in mitoses containing Following control or ILK siRNA transfection these cells de-clustered centrosomes, whereas the cell lines con- were processed for immunofluorescence and stained for taining relatively low percentages of cells with

Oncogene Centrosome clustering by ILK, ch-TOG and TACC3 AB Fielding et al 523

Figure 1 ILK depletion by siRNA causes an increase in mitotic cells displaying de-clustered centrosomes in BT549 and MDA-MB- 231 breast cancer cells and PC3 prostate cancer cells. (a) Immunofluorescence analysis of BT549 treated with control or ILK siRNA. Cells are stained with antibodies to pericentrin (red), a-tubulin (green) and with the DNA dye Hoechst (blue). Pro-metaphase and metaphase cells (i.e. mitotic cells with condensed DNA and centrosomes separated into two or more poles but which have not begun anaphase) were found to fall into two broad categories. The first is cells with clustered centrosomes; this includes cells with the normal number of two centrosomes and cells with 42 centrosomes that are clustered into two poles allowing the cell to form a bipolar spindle. The second category show de-clustered centrosomes, where centrosomes are not arranged into two poles, microtubule staining does not show a bipolar mitotic spindle and DNA is not aligned at a metaphase plate. The percentage of each phenotype present in control or ILK siRNA-treated cells is shown below. Scale bars 5 mm. (b) Immunofluorescence examples of PC3 cells showing clustered and de- clustered centrosomes in control and ILK siRNA-treated cells, respectively. Scale bars 5 mm. (c) Quantification of the presence of de- clustered centrosomes in either control or ILK siRNA-treated BT549, MDA-MB-231 or PC3 cells. Data are mean and s.e. of at least two independent experiments. *Denotes significance at Po0.05, as determined by Student’s t-test. (d) Western blots showing ILK knockdown by siRNA. supernumerary centrosomes do not (Figure 2a). It has These results imply that ILK inhibition perturbs the previously been shown that QLT-0267 inhibits ILK ability of cells to cluster supernumerary centrosomes kinase activity with a half maximal inhibitory concen- yet, it is important that, does not itself induce the tration (IC50) of between 2–5 mM, depending on cell type formation of supernumerary centrosomes. This also (Troussard et al., 2006). This correlates well with the suggests that inhibiting ILK may have deleterious effects effect on centrosome clustering observed here, suggest- on supernumerary centrosome-containing cancer cells, ing that ILK kinase activity is likely to be the driving although leaving cells with a normal centrosome force behind the phenotypes observed. QLT-0267 was complement unperturbed. This theory was tested next. also added to non-cancerous (BPH-1) and cancerous (PC3) prostate cell lines. Reflecting the results obtained Centrosome number correlates with sensitivity to ILK with the breast cell lines, the inhibitor-induced centro- inhibition in panel of breast and prostate cell lines some de-clustering in the PC3, but not the BPH-1 cells Clonogenic assays were performed on the panel (Figure 2d). of breast cell lines to determine their sensitivity to

Oncogene Centrosome clustering by ILK, ch-TOG and TACC3 AB Fielding et al 524 the ILK inhibitor. It was found that there was a highly less than 10% (Figures 2b and c). These results significant difference between the normal cell types were reflected in the survival of the prostate cell with two centrosomes, which were unaffected by 2.5 mM lines, with the non-cancerous BPH-1 cells being only QLT-0267, and the supernumerary centrosome-contain- minimally effected whereas the proliferation of the ing BT549 and MDA-MB-231 cancer cells, where cancerous PC3 cells was significantly reduced the surviving fraction of cells at 2.5 mM QLT-0267 was (Figure 2e).

Oncogene Centrosome clustering by ILK, ch-TOG and TACC3 AB Fielding et al 525 These data present strong correlative evidence that confirmed that the ILK-inhibited cells did indeed spend the centrosome de-clustering activity observed upon significantly more time in mitosis. ILK inhibition leads to decreased proliferation of cancer The percentage of cells that failed in the final stages of cells. However, it is known that ILK inhibition can lead cytokinesis and underwent furrow regression to form a to cancer cell-specific decreases in cell-survival signaling, bi-nucleated progeny (category 3) was higher than such as decreased v-akt murine thymoma viral oncogene initially expected in control-treated cells (31.0%), homolog 1 (AKT473) phosphorylation (Troussard et al., although was further elevated upon QLT-0267 treat- 2006) that may partially account for the decrease in cell ment (37.7%). It is possible that this high rate of survival observed in the cancer cell lines. It could not, cytokinesis failure in the control cells may be due to the however, explain the difference seen between the cancer phototoxicity of the imaging regime. However, these cell lines with high (BT549s and MDA-MB-231) and results may at least in part be explained by observations low (MCF7) percentages of supernumerary centro- that cells with supernumerary centrosomes often pass somes. Despite this, to test whether the decreased through multipolar anaphase intermediates that can proliferation observed in the clonogenic assays was leave chromosomes stranded near the center of the cell, due, at least in part, to mitosis-related events, we referred to as lagging chromosomes (Ganem et al., 2009; performed time-lapse imaging to determine the mitotic Silkworth et al., 2009), the presence of which correlates outcomes of cells in which ILK was inhibited. with cytokinesis failure (Silkworth et al., 2009). Indeed, MDA-MB-231 cells that contain multiple centrosomes have a high rate of lagging chromosomes even in cells ILK inhibition leads to multipolar anaphases and undergoing normal bipolar anaphases (Ganem et al., cytokinesis, prolonged mitotic arrest and cell death 2009), which could account for the high rate of in mitosis cytokinesis failure in control-treated cells. MDA-MB-231 cells were treated with either dimethyl Detailed analysis of the movies revealed that 23.0% of sulfoxide (DMSO) or QLT-0267 and observed by phase QLT-0267-treated cells showed multipolar anaphase and microscopy and the fate of all cells attempting mitosis cytokinesis intermediates, as compared with just 1.2% in was categorized into six groups (Figure 3a). It is of note controls, indicative of centrosome de-clustering. However, that, only 21.3% of QLT-0267-treated cells underwent the majority of these multipolar intermediates did not go successful bipolar divisions (category 1, Figure 3c) as on to complete multipolar cytokineses, resulting in only opposed to 66.7% of DMSO-treated cells. This differ- 3.3% of QLT-0267-treated (and 0.0% of DMSO-treated) ence was mostly accounted for by three categories seen cells (category 4, Figure 3e) actually completing multipolar with increased frequency in the ILK-inhibited cells. cell divisions. This notwithstanding, cells that displayed These were cells that enter mitosis, but do not proceed multipolar anaphase/cytokinesis intermediates rarely un- to anaphase and, after prolonged arrest, exit mitosis derwent successful bipolar divisions, instead often under- (category 2 and Figure 3d), cells that had been arrested going cytokinesis failure to produce multinuclear progeny in mitosis for 4120 min at the end of the time course (Supplementary Figure 2a) or uneven bipolar divisions (category 5) and cells that became arrested in mitosis where one cell was considerably smaller than the other and subsequently underwent cell death (category 6 and (Supplementary Figure 2b). Figure 3f). As prolonged mitoses seemed to be a feature To complement this data, the mitotic index of both of the ILK-inhibited cells, the length of mitosis (from MDA-MB-231 cells and the non-cancerous MCF10A cells nuclear envelope breakdown to the onset of anaphase) was calculated subsequent to treatment with the ILK was calculated for each cell where this was possible and inhibitor (Supplementary Figure 4). MDA-MB-231 cells these data were plotted in Figure 3b. This analysis showed an increase in mitotic index upon ILK inhibition,

Figure 2 Inhibition of ILK using the small-molecule ILK-specific inhibitor QLT-0267 results in an increase in the percentage of cells displaying de-clustered centrosomes and selectively decreases the survival of cancer cell lines containing supernumerary centrosomes. (a) Quantification of the percentage of mitotic cells showing de-clustered centrosomes upon ILK inhibition in the panel of mammary epithelial cell lines. It can be seen that the normal cell types (MCF10A and 184-hTERT) and cancer cells with mostly two centrosomes (MCF7) show little or no increase in cells displaying de-clustered centrosomes. However, in cancer cells where approximately 45% of cells contain supernumerary centrosomes (MDA-MB-231 and BT549), a dose-dependent increase in the percentage of mitotic cells showing de-clustered centrosomes upon ILK inhibition is observed. Data are mean values±s.e. of at least two independent experiments. *Denotes significance at Po0.05, **Po0.01. (b) Here the ability of these cells to form colonies under conditions of ILK inhibition was tested. Clonogenic or colony formation assays test the ability of cells to undergo repeated divisions. Cells were plated at low density, DMSO or QLT-0267 added and the cells incubated for 15 days. Plates were then stained and the number of colonies counted. DMSO control counts were converted to represent 100% surviving fraction and the relative numbers of QLT-0267-treated cells plotted. It can be seen that there is a striking inverse relation between the number of centrosomes a cell contains and its survival under conditions of ILK inhibition. Bar graph shows mean and s.e. of triplicate repeats. **Denotes significance at Po0.01 and ***Po0.001 as determined by Student’s t-test. (c) Images of the stained plates shown. Overlaid numbers indicate the concentration of QLT-0267 (mM) added to well. Beneath the plates the percentages of cells that contain more than two centrosomes are shown. These figures are from the following sources; BT549, MDA231 and MCF7 (Kwon et al., 2008), MCF10A (You et al., 2004), 184-hTERT (our own, unpublished data). (d) Quantification of the percentage of mitotic cells showing de-clustered centrosomes upon ILK inhibition in normal (BPH-1) and cancerous (PC3) prostate cell lines. Data are mean values±s.e. of at least two independent experiments. *Denotes significance at Po0.05. (e) Clonogenic assays of normal and cancerous prostate cell lines treated with DMSO or the ILK inhibitor QLT-0267. Bar graph shows mean and s.e. of triplicate repeats. **Denotes significance at Po0.01.

Oncogene Centrosome clustering by ILK, ch-TOG and TACC3 AB Fielding et al 526 reflective of the live-cell analysis showing the induction of proliferation. Next, we investigated the mechanism by mitotic arrest, whereas the MCF10A cells showed no such which ILK influences centrosome clustering. increase. This supports the data presented in Figure 2 that the mitotic dynamics and proliferation of normal cells, the ILK’s centrosomal partners, TACC3 and ch-TOG, majority of which contain the normal number of centro- regulate centrosome clustering in human cancer cells somes, is not affected by ILK perturbation. Both actin-cytoskeleton regulating and microtubule- In summary, these data indicate that ILK inhibition organizing proteins are required for centrosome cluster- does indeed cause adverse outcomes for cancer cells ing (Kwon et al., 2008). ILK regulates the actin attempting mitosis that will result in decreased cell cytoskeleton from its location at focal adhesions (for

Oncogene Centrosome clustering by ILK, ch-TOG and TACC3 AB Fielding et al 527 reviews see Legate et al., 2006; McDonald et al., 2008) data is included in Supplementary Figure 6 and and microtubule organization through centrosomal- discussed in Supplementary material. interacting partners (Bettencourt-Dias et al., 2004; Dobreva et al., 2008; Fielding et al., 2008). Therefore, ILK regulates centrosome clustering in concert with its to begin to elucidate the mechanism by which ILK effects centrosomal partner TACC3 centrosome clustering we first tested whether any of Supplementary Figure 6 suggests that focal adhesion ILK’s known centrosomal or focal-adhesion-interacting partners of ILK are not required for centrosome partners also influenced multipolar spindle formation. clustering and that ILK is controlling centrosome Ch-TOG and TACC3 co-localize, interact with and clustering independently of the actin cytoskeleton. We, are influenced by ILK (Dobreva et al., 2008; Fielding therefore, hypothesized that ILK may be influencing et al., 2008) and are also key regulators of microtubule centrosome clustering by its microtubule-regulating dynamics (Lee et al., 2001; Gergely et al., 2003; Barros partner TACC3. To test this we simultaneously per- et al., 2005). Figure 4 shows that both TACC3 and ch- turbed both ILK and TACC3. Figure 4e shows that TOG depletion lead to significant increases in de- adding QLT-0267 in addition to TACC3 siRNA does clustered centrosomes in both BT549 and PC3 cells. not cause a further increase in the percentage of mitotic TACC3 siRNA lead to increases in de-clustered centro- cells displaying de-clustered centrosomes when com- somes on a similar scale to ILK. However, ch-TOG pared with the effect of TACC3 siRNA alone. However, siRNA lead to much greater increases in the appearance addition of QLT-0267 to Parvin/PINCH siRNA or of multipolar spindles in both cell lines (Figure 4b). This Mig-2 siRNA does result in an additive effect on agrees with the report that ch-TOG depletion can induce centrosome clustering. This suggests ILK is acting the formation of multipolar spindles by two pathways, through TACC3 and hence, the microtubule cytoskele- centrosome de-clustering and also the formation of ton. It has previously been shown that ILK co-localizes acentrosomal spindle poles (Barr and Gergely, 2008). with TACC3 at the centrosome (Fielding et al., 2008). This in turn suggests that TACC3 and ILK are only To verify that this was the case in the cell types studied involved in one of these pathways. Supplementary here, BT549 cells were stained for ILK and pericentrin Figure 1 shows that nearly all ILK and TACC3 or TACC3, as well as IgG controls (Supplementary siRNA-induced multipolar spindles stained for a pair Figure 3). This confirmed that ILK localizes to the of centrin 2 foci at each of the spindle poles present, centrosome in BT549 cells where it has overlapping whereas a significant proportion of ch-TOG siRNA- localization with TACC3. It is known that ILK treated cells contained spindle poles that did not stain perturbation leads to a loss of complex formation for centrin 2. Acentrosomal poles, however, did not between TACC3, Aurora-A and ch-TOG (Fielding account for all multipolar spindles observed with the et al., 2008). However, the precise effect that ILK has remainder of the multipolar spindles observed because on these proteins has not been elucidated. of the centrosome de-clustering (Supplementary Figure 1c). This confirms the role of ch-TOG in both the formation of acentrosomal spindle poles and centro- TACC3 phosphorylation is required for centrosome some clustering. clustering Therefore, two of ILK’s centrosomal-interacting TACC3’s microtubule-regulating functions at the cen- partners are required for centrosome clustering. We trosome are dependent on its phosphorylation on a also assessed whether ILK’s key binding partners at conserved serine residue, Ser558 in humans/Ser347 in focal adhesions are required for centrosome clustering mice (Giet et al., 2002; Barros et al., 2005; Kinoshita and whether ILK’s effects on clustering occur depen- et al., 2005; Peset et al., 2005; LeRoy et al., 2007). As dently or independently of the actin cytoskeleton. This both spindle and astral microtubules are required to

Figure 3 ILK inhibition leads to an increase in multipolar divisions, prolonged mitotic arrest and cell death in mitosis. MDA-MB-231 cells were treated with either DMSO or 5 mM QLT-0267 1 h before imaging began and time-lapse microscopy performed for 10 h with images acquired every 3 min. Movies were observed to identify cells undergoing mitosis and these analyzed in detail. (a) All identified mitotic cells were categorized into one of six groups as shown in the table. Note a large decrease in the percentage of successful cell divisions upon ILK inhibition, accompanied by increases in cells exiting mitosis without attempting anaphase or cytokinesis (category 2), multipolar cell divisions (category 4) and mitotic arrest, often resulting in cell death (categories 5 and 6). (b) Bar graph showing the length of mitosis (from nuclear-envelope breakdown (NEBD) to anaphase onset) in DMSO and QLT-0267-treated cells. QLT-0267- treated cells showed a highly significant increase in the length of mitosis. Bars show s.e., ***Po0.001 as determined by Student’s t-test. (c–f) Show frames from phase movies as examples of cell fates upon DMSO or QLT-0267 treatment. The category as defined in (a) are shown for each movie. Numbers under each frame indicate the length of time in minutes from cell rounding. Cell outlines indicate cells of interest. (c) A cell treated with DMSO starts to round up to enter mitosis in the second frame (0 min). Within 100 min of rounding it has completed anaphase (frame 5, 99 min) and in the subsequent frame undergoes a normal cell division. The final frame indicates the position of the two, attached, daughter cells at the end of the movie. (d) A QLT-0267-treated cell enters mitosis near the beginning of this movie and after prolonged mitotic arrest (429 mins) with no anaphase or cytokinesis attempts visible, exits mitosis and sticks back down as one cell (final two frames). (e) A cell treated with QLT-0267 undergoes a tripolar anaphase (clearly visible at 12 mins) and cytokinesis, resulting in three separate daughter cells. (f) A QLT-0267-treated cell rounds up and enters mitosis (second frame, 0 min) with condensed but misaligned chromosomes visible in at 36 and 342 min. After this prolonged mitotic arrest (342 min), the cell then undergoes apoptosis, visible in the final two frames.

Oncogene Centrosome clustering by ILK, ch-TOG and TACC3 AB Fielding et al 528

Figure 4 ILK’s centrosomal-interacting partners TACC3 and ch-TOG are required for centrosome clustering in breast and prostate cell lines and ILK controls centrosome clustering through the same pathway as TACC3. (a, b) siRNA against TACC3 or ch-TOG was transfected into BT549 or PC3 cells. After incubation, cells were processed for immunofluorescence and stained for a-tubulin, pericentrin and Hoescht and the percentage of mitotic cells displaying de-clustered pericentrin foci calculated in both control and TACC3/ch-TOG siRNA samples. Data are mean values±s.e. of at least two independent experiments. *Denotes significance at Po0.05, **Po0.01. (c, d) Western blots show successful knockdown of TACC3 or ch-TOG protein levels by siRNA. (e) BT549 cells were treated with control, TACC3, Mig-2 or Parvin a/b/PINCH siRNA with the addition of either DMSO (light grey bars) or 5 mM QLT-0267 (dark grey bars). Data are mean values±s.e. of at least two independent experiments. *Denotes significance at Po0.05.

prevent multipolar spindle formation (Kwon et al., (Supplementary Figure 7), in agreement with previous 2008), we investigated whether TACC3 phosphorylation studies that have shown the importance of this phos- was required for centrosome clustering. A siRNA- phorylation event for spindle and centrosome localiza- resistance approach was used, whereby endogenous tion of TACC3 (Kinoshita et al., 2005; Peset et al., 2005; (human), TACC3 was depleted by siRNA and mouse LeRoy et al., 2007). The effects of expressing these TACC3 WT or TACC3 serine 347 to alanine mutant constructs against the background of endogenous (S347A), which were resistant to the siRNA, were TACC3 depletion were then assessed. Figure 5 shows expressed. Immunofluorescence analysis indicated that that, although the expression of mTACC3-WT can res- the TACC3 S347A mutant showed markedly reduced cue the effects on centrosome clustering caused by endo- localization to the mitotic spindle and spindle poles genous TACC3 depletion, a mTACC3-S347A construct

Oncogene Centrosome clustering by ILK, ch-TOG and TACC3 AB Fielding et al 529

Figure 5 Phosphorylation of TACC3 is required for centrosome clustering. (a) Western blot to show knockdown of endogenous (human) TACC3 and expression of siRNA-resistant mouse TACC3. Although there is a high degree of homology between the functional domains and phosphorylation sites of human and mouse TACC3, mouse TACC3 lacks some of the N-terminal sequence of human TACC3 and is, therefore, approximately 20 kDa smaller, which allows their differentiation by western blotting. An antibody recognizing both human and mouse TACC3 shows two bands. The upper (human) band responds to TACC3siRNA treatment where as the lower (mouse) band does not. (b) Representative immunofluorescence images of TACC3 siRNA-treated cells re-expressing either mTACC-WT (top panel) or mTACC3-SA mutant (bottom panel). Although expression of the WT mTACC3 can promote centrosome clustering under conditions of endogenous TACC3 knockdown, the mTACC3 SA mutant cannot. (c) Bar graph showing the percentage of mitotic cells with de-clustered centrosomes under different TACC3 expression conditions. In mock-transfected cells (first two bars), TACC3 knockdown causes an increase in the percentage of de-clustered centrosomes, as shown in Figure 4. Upon expression of a WT mouse TACC3 construct, which is resistant to the TACC3siRNA, the de-clustering caused by the TACC3 knockdown is largely rescued (third bar). However, re-expression of mouse TACC3 in which Ser347 (homologous to the Ser588 in human TACC3) is mutated to alanine cannot rescue the knockdown of endogenous TACC3 (fourth bar), indicating this phosphorylation site is required for centrosome clustering. *Po0.05 as determined by Student’s t-test. cannot, indicating the essential nature of this phos- ILK siRNA knockdown cells (Figure 6g). However, the phorylation site for successful centrosome clustering. significant decrease in phosphorylated TACC-3, despite an increase in total TACC-3 implies a marked effect of depleting ILK on the phosphorylation status of TACC- ILK Perturbation leads to decreased TACC3 3, agreeing with the effects of inhibiting ILK activity phosphorylation with QLT-0267. For unknown reasons the phospho- We tested whether ILK inhibition or knockdown TACC3 antibody did not detect a band in western blots affected the phosphorylation of the conserved TACC3 and, therefore, we could not test for the global effects on serine residue using a phospho-specific antibody raised pTACC levels induced by ILK depletion or inactivation. against this site. As previously reported, phospho Ser558 However, the quantitative immunofluorescence data TACC3 localized exclusively to centrosomes (Kinoshita clearly show that pTACC3 levels are reduced at the et al., 2005). This was the case in both bipolar centrosome, the functional site of pTACC3 (Kinoshita (Figure 6a, upper panel) and multipolar (Figure 6b, et al., 2005) and we are, therefore, confident that the upper panel) mitotic spindles. Upon ILK inhibition, this reduction in levels observed will have functional phosphorylation was reduced (Figures 6a and b, lower consequences for mitotic spindle organization. panels). The phosphoSer558 TACC3 signal at centro- somes was quantified from Z-stacks of mitotic cells and this indicated a significantly lower amount of phosphor- ILK regulates TACC3 phosphorylation in an ylation upon ILK inhibition (Figure 6c). Quantification aurora-A-dependent manner of cells stained with pericentrin showed that the effect Next, we investigated if the TACC3 Ser558 phosphor- was specific for pTACC3 and not due to a general defect ylation was a direct consequence of ILK kinase activity in pericentriolar material recruitment (Supplementary or not. This site has been identified as a phosphorylation Figure 5). Total cellular TACC3 (and Aurora-A and site for Aurora-A kinase (Giet et al., 2002; Tien et al., ch-TOG) levels were also unaffected (Figures 6d and e). 2004; Barros et al., 2005; Kinoshita et al., 2005; LeRoy ILK knockdown by siRNA produced comparable et al., 2007). Also, we have shown that ILK depletion or effects on TACC3 phosphorylation (Figure 6f). It is inactivation can effect Aurora-A–TACC3 interactions not clear why total TACC-3 levels are increased in the (Fielding et al., 2008) and the residues surrounding

Oncogene Centrosome clustering by ILK, ch-TOG and TACC3 AB Fielding et al 530

Figure 6 ILK inhibition leads to a decrease in phospho TACC3 levels at the centrosome. (a, b) Immunofluorescence staining of DMSO or QLT-0267-treated cells with an antibody against pSer558 TACC3 (red), a-tubulin (green) and Hoechst (blue). pSer558 TACC3 localizes to the centrosome. It’s levels are reduced in QLT-0267-treated cells in both bipolar (a) and multipolar (b) spindles. Arrowheads indicate spindle poles/centrosomes. Contrast and brightness settings have been applied equally to all images to allow images to be displayed appropriately. However, only original, non-adjusted images were used for quantifications in the remainder of the figure. Scale bar 5 mm. (c) Quantification of pSer558 TACC3 fluorescence intensity at centrosome. Z-stack images were acquired and AxioVision software used to calculate the fluorescence intensity from each centrosome. Data are means±s.e.m. of at least 20 centrosomes and displayed as relative percent intensity compared with the DMSO control. QLT-0267-treated cells showed a significant decrease in pTACC3 staining at the centrosome, ***Po0.001. (d) Total cellular TACC3 levels were quantified in a similar fashion. These showed no significant differences upon QLT-0267 treatment. (e) Western blots of TACC/ILK-interacting partners showing that under conditions of ILK inhibition total cellular levels of Aurora-A, ch-TOG and TACC3 do not change significantly. (f, g) pSer558 TACC and total TACC were also measured in cells after ILK depletion by siRNA. ILK siRNA also caused a significant decrease in TACC3 phosphorylation. (h) Western blot to show ILK knockdown by siRNA.

Ser558 do not show a strong similarity to other known normal amounts of ILK causes a small, but insignif- ILK phosphorylation sites (Hannigan et al., 2005). icant, increase in TACC3 phosphorylation. ILK deple- Therefore, we hypothesized that ILK was likely to be tion by siRNA caused a significant reduction in TACC3 acting upstream of Aurora-A to affect TACC3 phos- phosphorylation (as shown in Figure 6). Upon over- phorylation. To test this, we overexpressed Aurora-A expression of Aurora-A in these cells, the TACC3 under conditions of ILK depletion to see whether this phosphorylation was rescued back to control levels could rescue TACC3 phosphorylation. Figure 7 shows (Figures 7a and c), thus indicating that Ser558TACC3 that the overexpression of Aurora-A in cells containing phosphorylation is regulated by ILK in an Aurora-A-

Oncogene Centrosome clustering by ILK, ch-TOG and TACC3 AB Fielding et al 531

Figure 7 Aurora-A over-expression rescues phosphorylation of TACC3 and centrosome clustering under conditions of ILK depletion. To test whether ILK’s effects on TACC3 phosphorylation were Aurora-A dependent, Aurora-A was over-expressed under ILK-knockdown conditions. (a) Bar graph of phospho-Ser558-TACC3 quantification. Cells were treated with control or ILK siRNA and either mock transfected or transfected with an Aurora-A plasmid. Cells were then stained with a mouse monoclonal antibody against Aurora-A (to identify transfected cells) and a rabbit polyclonal antibody against phosphoSer558 of TACC3. Z-stack images of the cells were then acquired and the fluorescence intensity of pTACC3 signal at the centrosomes quantified. The average intensity in mock transfected, control siRNA-treated cells was set as 100% and the other values plotted relative to this. In untransfected cells (dark grey bars), it can be seen that ILK siRNA causes a decrease in the pTACC3 signal. However, upon transfection of Aurora-A, the pTACC3 signal is rescued to control levels. This indicates ILK’s effects on TACC3 phosphorylation occur in an Aurora-A-dependent manner. ***Po0.001 as determined by Student’s t-test. (b) The percentage of mitotic cells with de-clustered centrosomes was calculated for cells treated with ILK siRNA in both cells expressing only endogenous amounts of Aurora-A and cells overexpressing Aurora-A. In ILK siRNA-treated cells, centrosome de-clustering was apparent in 41.4% of mitotic cells. In ILK siRNA-treated cells that also overexpressed Aurora-A, the number of cells displaying de-clustered centrosomes was significantly reduced, indicating that Aurora-A over-expression rescues the centrosome de-clustering defects induced by ILK depletion, as well as the decrease in phosphorylated TACC3. *Po0.05 as determined by Student’s t-test.(c) Immunofluorescence examples of Aurora-A and pTACC3- stained cells. Top panel; in cells expressing only endogenous amounts of Aurora-A, ILK siRNA leads to only a weak pTACC3 signal present at centrosomes. Upon transfection of Aurora-A (bottom panel), pTACC3 amounts at the centrosome are markedly increased. Arrowheads indicate centrosomes/spindle poles where Aurora-A and pTACC co-localize in mitotic cells. Identical microscope settings were used to acquire images shown. (d) Western blot to show successful expression of exogenous Aurora-A. As the transfected Aurora- A is tagged with mcherry, it has a significantly higher molecular weight than the endogenous Aurora-A and can, therefore, be readily distinguished on the western blot. dependent manner. Upon measuring the effect on Discussion centrosome clustering in these cells it was apparent that the overexpression of Aurora-A also rescued Centrosome clustering is an essential mechanism the centrosome de-clustering caused by ILK depletion for the survival of cancer cells (Rebacz et al., 2007; (Figure 7b). These results provide a direct mechanistic Kwon et al., 2008; Ganem et al., 2009). Mechanisms link between ILK perturbation and the observed defects involving microtubules, the actin cytoskeleton, cell– on centrosome clustering. extra-cellular-matrix interactions and the spindle assem-

Oncogene Centrosome clustering by ILK, ch-TOG and TACC3 AB Fielding et al 532 bly checkpoint are required for successful centrosome prostate cancer cells lines and that ILK performs this clustering (Quintyne et al., 2005; Kwon et al., 2008). As function by its microtubule-related functions at the ILK regulates microtubules, the actin cytoskeleton and centrosome, specifically by regulating TACC3 phos- cell–extra-cellular-matrix interactions it has been pro- phorylation. Our data provide further evidence to the posed as a good candidate to also regulate centrosome theory that disrupting centrosome clustering can clustering (Xu and Saunders, 2008). selectively inhibit the proliferation of supernumerary This study shows that ILK is required for centrosome centrosome-harboring cancer cells (Rebacz et al., 2007; clustering in both breast and prostate cancer cell lines, as Kwon et al., 2008; Xu and Saunders, 2008). In this genetic or pharmacological inhibition of ILK leads to an regard, our results provide an exciting proof of concept increase in the presence of mitotic cells displaying de- for targeting centrosome clustering in cancer cells and clustered centrosomes. As supernumerary centrosomes pave the way towards a new class of targeted cancer are generally only a feature of cancer cells (Nigg, 2002; therapies. Kwon et al., 2008; Zyss and Gergely, 2009), we next tested for the potential cancer-cell-specific effects of ILK perturbation. ILK inhibition caused almost complete Materials and methods loss of cell proliferation in the cancer-cell lines contain- ing supernumerary centrosomes, whereas non-malig- Cell culture, siRNA transfections and small molecules nant breast epithelial lines were unaffected by an equal MDA-MB-231, BT549, PC3, DU145 and BPH-1 cells were concentration of the ILK inhibitor. This shows that the cultured in DMEM þ 10% FCS þ P/S, MCF7 in DMEM þ inhibition of ILK causes cancer-specific defects in cell 10% FCS þ P/S þ insulin (10 mg/ml), 184-HTERT in mam- survival that reflect its centrosome de-clustering activity. mary epithelial basal medium, MCF10A in DMEM/F12 þ 10% Live cell analysis revealed that ILK inhibition causes FCS (Invitrogen, Carlsbad, CA, USA). Silentfect (Biorad, a large decrease in successful cell divisions, largely Cambridge, MA, USA) was used according to manufacturer’s instructions to transfect siRNA. QLT-0267 (Quadra Logic attributable to cells becoming arrested in mitosis for Technology Inc., Vancouver, BC, Canada) was diluted in prolonged periods followed by either mitotic exit or cell DMSO and added to cells for 6 h unless otherwise stated. death. Failed cytokinesis and multipolar cell divisions Latrunculin A (Sigma, St Louis, MO, USA) was re-suspended also occur more frequently in cells treated with the ILK in DMSO and added for the final 2 h before to cell fixation, as inhibitor. All of these outcomes can be attributed to the described in Kwon et al. (2008). centrosome de-clustering activity of the ILK inhibitor as cells that fail to cluster supernumerary centrosomes are Antibodies and siRNAs known to have extended mitoses (Rebacz et al., 2007; Antibodies were the same as in previous studies (Dobreva Kwon et al., 2008), undergo apoptosis during mitotic et al., 2008; Fielding et al., 2008) except for rabbit anti-centrin arrest (Rebacz et al., 2007), undergo failed cytokinesis 2 (Santa Cruz Biotechnology, Inc. Santa Cruz, CA, USA), (Silkworth et al., 2009) and execute multipolar cell rabbit anti-chTOG (LifeSpan Biosciences, Inc. Seattle, WA, divisions (Kwon et al., 2008). Each of these phenomena USA), rabbit anti-b Parvin (Proteintech Group, Inc. Chicago, will also result in decreased cell proliferation and IL, USA) and rabbit-anti-Mig-2 (Abcam Inc. Cambridge, MA, USA). Anti-pSer558 TACC3 was a kind gift from survival and, therefore, the centrosome de-clustering Dr Kazuhisa Kinoshita (Kinoshita et al., 2005). siRNA details activity observed upon ILK inhibition can be directly are included in Supplementary methods. related to the decrease in cell survival observed in the clonogenic assays. Microscopy This study illustrates that two centrosomal proteins, Details are provided in Supplementary methods. TACC3 and ch-TOG, which in concert with Aurora-A, are critical for mitotic spindle formation and organiza- Clonogenic assays tion (Gergely et al., 2003; Cassimeris and Morabito, Clonogenic assays were performed as described by Franken 2004; Barros et al., 2005; Kinoshita et al., 2005; Barr et al. (Franken et al., 2006) and colonies of 200 or more cells and Gergely, 2008) are required for centrosome cluster- were counted. ing. Further investigation demonstrated that a con- served serine residue on TACC3 was required for Plasmids and transfections centrosome clustering and that ILK was regulating Described in Supplementary methods. Ser558 TACC3 phosphorylation, in an Aurora-A- dependent manner. Disruption of this phosphorylation leads to destabilized spindle and astral microtubules Conflict of interest and errors in microtubule attachment to centrosomes (Giet et al., 2002; Barros et al., 2005; Kinoshita et al., SD is a scientific consultant for Quadra Logic Technology Inc. 2005; Peset et al., 2005; LeRoy et al., 2007; Albee and (QLT Inc), Vancouver. Wiese, 2008). These phenotypes will lead to multipolar spindle formation as both spindle and astral micro- tubules are required for centrosome clustering (Kwon Acknowledgements et al., 2008). In summary, this report shows that ILK is required This work was supported by grants to SD from the Canadian for centrosome clustering in both human breast and Institute of Health Research (CIHR) and the Canadian Cancer

Oncogene Centrosome clustering by ILK, ch-TOG and TACC3 AB Fielding et al 533 Society Research Institute, with funds raised by the Canadian Dr Kazuhisa Kinoshita for the kind gift of anti-phospho- Cancer Society. We thank Quadra Logic Technology Inc Ser558 TACC3 antibody and Dr Peter Lansdorp for the (QLT Inc), Vancouver, BC for supplying QLT-0267. We thank Aurora-A plasmid.

References

Albee AJ, Wiese C. (2008). Xenopus TACC3/maskin is not required maskin is required for centrosome-dependent microtubule assembly for microtubule stability but is required for anchoring microtubules in mitosis. J Cell Biol 170: 1047–1055. at the centrosome. Mol Biol Cell 19: 3347–3356. Koul D, Shen R, Bergh S, Lu Y, de Groot JF, Liu TJ et al. (2005). Barr AR, Gergely F. (2008). MCAK-independent functions of ch-Tog/ Targeting integrin-linked kinase inhibits Akt signaling path- XMAP215 in microtubule plus-end dynamics. Mol Cell Biol 28: ways and decreases tumor progression of human glioblastoma. 7199–7211. Mol Cancer Ther 4: 1681–1688. Barros TP, Kinoshita K, Hyman AA, Raff JW. (2005). Kramer A, Neben K, Ho AD. (2002). Centrosome replication, Aurora A activates D-TACC-Msps complexes exclusively at genomic instability and cancer. Leukemia 16: 767–775. centrosomes to stabilize centrosomal microtubules. J Cell Biol Kwon M, Godinho SA, Chandhok NS, Ganem NJ, Azioune A, 170: 1039–1046. Thery M et al. (2008). Mechanisms to suppress multipolar Basto R, Brunk K, Vinadogrova T, Peel N, Franz A, Khodjakov A divisions in cancer cells with extra centrosomes. Genes Dev 22: et al. (2008). Centrosome amplification can initiate tumorigenesis in 2189–2203. flies. Cell 133: 1032–1042. Lee MJ, Gergely F, Jeffers K, Peak-Chew SY, Raff JW. Bettencourt-Dias M, Giet R, Sinka R, Mazumdar A, Lock WG, (2001). Msps/XMAP215 interacts with the centrosomal Balloux F et al. (2004). Genome-wide survey of protein kinases protein D-TACC to regulate microtubule behaviour. Nat Cell Biol required for cell cycle progression. Nature 432: 980–987. 3: 643–649. Boveri T. (1914). Zur frage der entstehung maligner tumoren (The Legate KR, Montanez E, Kudlacek O, Fassler R. (2006). ILK, PINCH Origin of Malignant Tumors). Gustav Fischer, Jena. and parvin: the tIPP of integrin signalling. Nat Rev Mol Cell Biol 7: Brinkley BR. (2001). Managing the centrosome numbers game: from 20–31. chaos to stability in cancer cell division. Trends in Cell Biology 11: LeRoy PJ, Hunter JJ, Hoar KM, Burke KE, Shinde V, Ruan J et al. 18–21. (2007). Localization of human TACC3 to mitotic spindles is Cassimeris L, Morabito J. (2004). TOGp, the human homolog mediated by phosphorylation on Ser558 by aurora A: A novel of XMAP215/Dis1, is required for centrosome integrity, spindle pharmacodynamic method for measuring aurora a activity. Cancer pole organization, and bipolar spindle assembly. Mol Biol Cell 15: Res 67: 5362–5370. 1580–1590. Lingle WL, Barrett SL, Negron VC, D0Assoro AB, Boeneman K, Liu Dobreva I, Fielding A, Foster LJ, Dedhar S. (2008). Mapping the W et al. (2002). Centrosome amplification drives chromosomal integrin-linked kinase interactome using SILAC. J Proteome Res 7: instability in breast tumor development. Proceedings of the National 1740–1749. Academy of Sciences of the USA 99: 1978–1983. Edwards LA, Woo J, Huxham LA, Verreault M, Dragowska WH, Lingle WL, Salisbury JL. (1999). Altered centrosome structure is Chiu G et al. (2008). Suppression of VEGF secretion and associated with abnormal mitoses in human breast tumors. Am J changes in glioblastoma multiforme microenvironment by Pathol 155: 1941–1951. inhibition of Integrin-linked kinase (ILK). Mol Cancer Ther 7: Margall-Ducos G, Celton-Morizur S, Couton D, Bregerie O, 59–70. Desdouets C. (2007). Liver tetraploidization is controlled by a Fielding AB, Dobreva I, McDonald PC, Foster LJ, Dedhar S. (2008). new process of incomplete cytokinesis. J Cell Sci 120: 3633–3639. Integrin-linked kinase localizes to the centrosome and regulates McDonald PC, Fielding AB, Dedhar S. (2008). Integrin-linked mitotic spindle organization. J Cell Biol 180: 681–689. kinase—essential roles in physiology and cancer biology. J Cell Franken NAP, Rodermond HM, Stap J, Haveman J, van Bree C. Sci 121: 3121–3132. (2006). Clonogenic assay of cells in vitro. Nature Protocols 1: Nigg EA. (2006). Origins and consequences of centrosome aberrations 2315–2319. in human cancers. Int J Cancer 119: 2717–2723. Ganem NJ, Godinho SA, Pellman D. (2009). A mechanism linking Nigg EA. (2002). Centrosome aberrations: cause or consequence of extra centrosomes to chromosomal instability. Nature 460: cancer progression? Nat Rev Cancer 2: 815–825. 278–U146. Peset I, Seiler J, Sardon T, Bejarano LA, Rybina S, Vernos I. Gergely F, Draviam VM, Raff JW. (2003). The ch-TOG/XMAP215 (2005). Function and regulation of Maskin, a TACC family protein is essential for spindle pole organization in human somatic protein, in microtubule growth during mitosis. J Cell Biol 170: cells. Genes Dev 17: 336–341. 1057–1066. Giet R, McLean D, Descamps S, Lee MJ, Raff JW, Prigent C et al. Pihan GA, Wallace J, Zhou Y, Doxsey SJ. (2003). Centrosome (2002). Drosophila Aurora A kinase is required to localize D-TACC abnormalities and chromosome instability occur together in pre- to centrosomes and to regulate astral microtubules. J Cell Biol 156: invasive carcinomas. Cancer Res 63: 1398–1404. 437–451. Quintyne NJ, Reing JE, Hoffelder DR, Gollin SM, Saunders WS. Gkretsi V, Mars WM, Bowen WC, Barua L, Yang Y, Guo L et al. (2005). Spindle multipolarity is prevented by centrosomal clustering. (2007). Loss of integrin linked kinase from mouse hepatocytes Science 307: 127–129. in vitro and in vivo results in apoptosis and hepatitis. Hepatology 45: Rebacz B, Larsen TO, Clausen MH, Ronnest MH, Loffler H, Ho AD 1025–1034. et al. (2007). Identification of griseofulvin as an inhibitor of Guidotti JE, Bregerie O, Robert A, Debey P, Brechot C, Desdouets C. centrosomal clustering in a phenotype-based screen. Cancer (2003). Liver cell polyploidization: A pivotal role for Research 67: 6342–6350. binuclear hepatocytes. J Biol Chem 278: 19095–19101. Ring D, Hubble R, Kirschner M. (1982). Mitosis in a cell with multiple Hannigan G, Troussard AA, Dedhar S. (2005). Integrin-linked kinase: centrioles. J Cell Biol 94: 549–556. A cancer therapeutic target unique among its ILK. Nat Rev Cancer Silkworth WT, Nardi IK, Scholl LM, Cimini D. (2009). Multipolar 5: 51–63. spindle pole coalescence is a major source of kinetochore mis- Kinoshita K, Noetzel TL, Pelletier L, Mechtler K, Drechsel DN, attachment and chromosome mis-segregation in cancer cells. PLoS Schwager A et al. (2005). Aurora A phosphorylation of TACC3/ ONE 4: e6564.

Oncogene Centrosome clustering by ILK, ch-TOG and TACC3 AB Fielding et al 534 Tien A-C, Lin M-H, Su L-J, Hong Y-R, Cheng T-S, Lee Y-CG et al. Weber RG, Bridger JM, Benner A, Weisenberger D, Ehemann V, (2004). Identification of the substrates and interaction proteins of Reifenberger G et al. (1998). Centrosome amplification as a possible aurora kinases from a protein-protein interaction model. Mol Cell mechanism for numerical chromosome aberrations in cerebral Proteomics 3: 93–104. primitive neuroectodermal tumors with TP53 mutations. Cytogenet Troussard AA, McDonald PC, Wederell ED, Mawji NM, Cell Genet 83: 266–269. Filipenko NR, Gelmon KA et al. (2006). Preferential dependence Xu FL, Saunders WS. (2008). Actin and microtubules: Working of breast cancer cells versus normal cells on integrin-linked kinase together to control spindle polarity. Cancer Cell 14: 197–199. for protein kinase B/Akt activation and cell survival. Cancer Res 66: You F, Chiba N, Ishioka C, Parvin JD. (2004). Expression of an 393–403. amino-terminal BRCA1 deletion mutant causes a dominant growth Tu Y, Wu S, Shi X, Chen K, Wu C. (2003). Migfilin and Mig-2 link inhibition in MCF10A cells. Oncogene 23: 5792–5798. focal adhesions to filamin and the actin cytoskeleton and function in Zyss D, Gergely F. (2009). Centrosome function in cancer: guilty or cell shape modulation. Cell 113: 37–47. innocent? Trends Cell Biol 19: 334–346.

Supplementary Information accompanies the paper on the Oncogene website (http://www.nature.com/onc)

Oncogene