Oncogene (2003) 22, 8102–8116 & 2003 Nature Publishing Group All rights reserved 0950-9232/03 $25.00 www.nature.com/onc

TACC1–chTOG–Aurora A complex in breast

Nathalie Conte1,Be´ ne´ dicte Delaval1, Christophe Ginestier1, Alexia Ferrand1, Daniel Isnardon2, Christian Larroque3, Claude Prigent4, Bertrand Se´ raphin5, Jocelyne Jacquemier1 and Daniel Birnbaum*,1

1Department of Molecular Oncology, U119 Inserm, Institut Paoli-Calmettes, IFR57, Marseille, France; 2Imaging Core Facility, Institut Paoli-Calmettes, Marseille, France; 3E229 Inserm, CRLC Val d’Aurelle/Paul Lamarque, Montpellier, France; 4Laboratoire du cycle cellulaire, UMR 6061 CNRS, IFR 97, Faculte´ de Me´decine, Rennes, France; 5Centre de Ge´ne´tique Mole´culaire, Gif-sur-Yvette, France

The three human TACC (transforming acidic coiled-coil) metabolism, including mitosis and intracellular trans- encode a family of with poorly defined port of molecules, is progressing but many components functions that are suspected to play a role in oncogenesis. remain to be discovered and characterized. We describe A Xenopus TACC homolog called Maskin is involved in here the interaction of the TACC1 protein with several translational control, while Drosophila D-TACC interacts protein partners that makes it a good candidate to with the -associated protein MSPS (Mini participate in microtubule-associated processes in nor- SPindleS) to ensure proper dynamics of spindle pole mal and tumoral cells. during cell division. We have delineated here In mammals, TACC1 (Still et al., 1999a) belongs to a the interactions of TACC1 with four proteins, namely the family of three paralogous (Popovici et al., 2001) genes microtubule-associated chTOG (colonic and hepatic that also includes TACC2/AZU1/ECTACC (Chen et al., tumor-overexpressed ) protein (ortholog of Drosophi- 2000; Pu et al., 2001; Lauffart et al., 2003) and TACC3/ la MSPS), the adaptor protein TRAP (tudor repeat ERIC1 (Still et al., 1999b; Sadek et al., 2000; McKeve- associator with PCTAIRE2), the mitotic serine/threonine ney et al., 2001; Hao et al., 2002; Sadek et al., 2003). kinase Aurora A and the mRNA regulator LSM7 Mammalian TACC have known orthologs in Drosophila (Like-Sm protein 7). To measure the relevance of the melanogaster, named tacc (Gergely et al., 2000b), and in TACC1-associated complex in human cancer we have Xenopus laevis, named Maskin (Stebbins-Boaz et al., examined the expression of the three TACC, chTOG and 1999). Drosophila tacc encodes D-TACC, a centrosomal Aurora A in breast cancer using immunohistochemistry on protein required for normal spindle function in the early tissue microarrays. We show that expressions of TACC1, embryo (Gergely et al., 2000b); D-TACC interacts with TACC2, TACC3 and Aurora A are significantly corre- the microtubule-associated MSPS (Mini SPindleS) lated and downregulated in a subset of breast tumors. protein to regulate microtubule behavior (Gergely Using siRNAs, we further show that depletion of chTOG et al., 2000b; Cullen and Ohkura, 2001; Lee et al., and, to a lesser extent of TACC1, perturbates cell 2001; Theurkauf, 2001). This complex is regulated by division. We propose that TACC proteins, which we also the Aurora A serine/threonine kinase (Giet et al., 2002). named ‘Taxins’, control mRNA translation and cell D-TACC and TACC3 associate with fly and human division in conjunction with microtubule organization , respectively (Giet et al., 2002). The and in association with chTOG and Aurora A, and that msps gene has orthologs in other species named zyg-9 in these complexes and cell processes may be affected during Caenorhabditis elegans, XMAP215 in X. laevis, and mammary gland oncogenesis. chTOG (colonic and hepatic tumor overexpressed gene) Oncogene (2003) 22, 8102–8116. doi:10.1038/sj.onc.1206972 in humans (Charrasse et al., 1995, 1998; Charrasse and Larroque, 2000; Spittle et al., 2000). The TACC1– Keywords: Aurora kinase; breast cancer; cell division; chTOG interaction has been recently described in a two- chTOG protein; microtubule; SM proteins; TACC hybrid screen in yeast (Lauffart et al., 2002). protein; tissue microarray; RNAi Maskin associates with both the 30UTR-CPE binding protein (CPEB) and the 50UTR CAP-binding protein eIF-4E to regulate mRNAs translation during oocyte Introduction maturation in Xenopus (Stebbins-Boaz et al., 1999); CPEB and Maskin are present on the mitotic apparatus The identification of protein complexes involved in of animal pole blastomeres in embryos (Groisman et al., processes associated with microtubule structure and 2000). We have recently shown that TACC1 interacts with LSM7 and SmG (Conte et al., 2002), two proteins related to the yeast Sm small nuclear ribonucleoproteins *Correspondence: D Birnbaum, U119 Inserm, 27 Bd Lei Roure, 13009 Marseille, France; E-mail: [email protected] (snRNPs). The Sm/LSM proteins associate with small The first two authors contributed equally to the work nuclear RNAs to form the core of snRNPs required Received 25 April 2003; revised 2 July 2003; accepted 10 July 2003 for a large variety of RNA maturation processes and TACC1–chTOG–Aurora protein complex N Conte et al 8103 mRNA degradation (Salgado-Garrido et al., 1999; Bouveret et al., 2000; Tharun et al., 2000). To date, few other mammalian TACC-interacting partners have been identified (Sadek et al., 2000; Lauffart et al., 2002, 2003; Steadman et al., 2002; Aitola et al., 2003). TACC proteins (designated hereafter as ‘Taxins’), under regulation by kinases, could intervene in micro- tubule-associated processes such as mitotic spindle formation (reviewed in Gergely, 2002; Lappin et al., 2002) and transport of RNAs specifically to spindles and and to subcellular areas necessary for cell polarity (Richter, 2001). Variation in the structure and/ or expression of Taxins and Taxins partners may have consequences on cellular growth (Raff, 2002). Human TACC were initially characterized as potential cancer genes (Still et al., 1999a, b; Chen et al., 2000). TACC1 is downregulated in breast cancer (Conte et al., 2002) and has been characterized as a gastric tumor antigen (Line et al., 2002a, b). TACC2/AZU1 is downregulated in transformed cell lines (Chen et al., 2000). The knockout of the mouse Tacc3 gene causes embryonic lethality associated with a high rate of apoptosis in several cell lineages (Piekorz et al., 2002). Overexpression of chTOG occurs in human hepatomas and colonic tumors (Charrasse et al., 1995). The STK6 gene, encoding the Aurora A kinase, is localized in the 20q13 chromosomal region, which is frequently amplified in , and Figure 1 Conservation of the Taxin family in metazoans. An overexpression of Aurora A leads to cell transformation unrooted phylogenetic tree (neighbor-joining method, Poisson correction, 225 sites from C-terminal region) of 10 TACC proteins (for reviews, see Bischoff and Plowman, 1999; Giet and (Taxins) is shown (Cel: C. elegans, Dme: D. melanogaster, Cin: C. Prigent, 1999; Nigg, 2001; Dutertre et al., 2002). intestinalis, Hsa: H. sapiens, Mmu: M. musculus, Xla: X. laevis, We have demonstrated and delineated here the Ocu: Oryctolagus cuniculus). Significant bootstrap values (out of interaction between TACC1 and protein partners, 500 replicates) are indicated at each node. Rabbit TACC4 (TACC4 Ocu) is actually a wrongly named TACC3 (Steadman et al., 2002). including LSM7, chTOG, Aurora A kinase and Branch lengths are proportional to time (0.1 Byr). Protostomians TRAP/PCTAIRE2BP. We have measured the relevance (i.e. Drosophila and C. elegans) and nonvertebrate chordates of the potential TACC1-associated complex in human (urochordate ascidian C. intestinalis) have only one TACC cancer by determining the simultaneous expression of member. The topology of the tree is the same using the maximum the three Taxins and of two Taxin-interactor proteins in parsimony method around 400 breast cancer samples using immunohisto- chemistry on tissue microarrays. We have further shown serine- and proline-rich motifs named SPAZ. We that inhibition of TACC1 or chTOG transcription leads screened a human mammary gland cDNA expression to abnormal cell division. Finally, in view of these library with two different TACC1 proteins (TACC1s results, we propose a preliminary model of TACC1 and TACC1spaz region) fused to the GAL4-binding function. domain. We isolated four clones using TACC1s as bait (Figure 2a). Two of them corresponded to the C- terminal region (aa 1776–1980) of the chTOG protein and the two others to LSM7, which we had already Results found in a previous screen using TACC1l as bait (Conte et al., 2002). We isolated seven clones using TACC1spaz Phylogenetic analysis of the TACC family region as bait; two of them encoded the C-terminal Through searches in databases, we identified one tacc region (aa 520–708) of TRAP(tudor repeat associator gene in the chordate nonvertebrate Ciona intestinalis with PCTAIRE 2). The TRAP gene, also called and built a phylogenetic tree of the Taxins (Figure 1). PCTAIRE2BP, encodes a protein that contains five The tree shows the relationships of the three human Tudor domains. TRAPassociates with serine/threonine Taxins and orthologs. kinase PCTAIRE2 (Hirose et al., 2000), which is expressed in terminally differentiated neurons (Hirose et al., 1997). Identification of TACC1-interactors by in vitro methods To document TACC1 interactions with these poten- We used the two-hybrid system in yeast to identify tial partners, we produced GST-fusion proteins that potential TACC1-interacting partners. TACC1 encodes were used in pull-down experiments using HA–TACC1l at least two protein isoforms, a long form, TACC1l, and (Figure 2b) or myc-TACC1s (Figure 2c). GST–chTOG, a short form, TACC1s. The latter lacks two series of GST–TRAPand GST–SmG precipitated TACC1l

Oncogene TACC1–chTOG–Aurora protein complex N Conte et al 8104

Figure 2 Identification of TACC1 protein partners using a two-hybrid screen in yeast. (a) The TACC proteins used as baits in a previous (TACC1l – Conte et al., 2002) and present (TACC1s and TACC1spaz) yeast two-hybrid screens are depicted. Arrows point to the identified preys. (b) Cos-1 cells were transfected with HA-tagged TACC1l and GST fusion proteins were used in a pull-down experiment. Blot with anti-HA revealed precipitated TACC1l protein. (c) Cos-1 cells were transfected with myc-tagged TACC1s and GST–chTOG was used in pull-down experiment. Blot TAC anti-myc revealed precipitated TACC1s protein

(Figure 2b, lanes 3–5), whereas GST did not (Figure 2b, 1 and 2). Reciprocally, HA–TACC1l coimmunoprecipi- lane 2). GST–chTOG also precipitated TACC1s tated with myc–chTOG from lysates using anti-myc (Figure 2c, lane 2), whereas GST–TRAPprecipitated antibody (not shown). An interaction between TACC1 only TACC1l (not shown). and chTOG (Figure 3b) was also found after coimmu- noprecipitation of endogenous TACC1l from human Interactions of TACC1 with chTOG, TRAP and Aurora mammary epithelial (HME) cell lysate using different antibodies. A in vivo TRAPwas immunoprecipitated from Cos-1 cells that The in vitro results were supported by coimmunopreci- expressed both HA-tagged TACC1l and myc-tagged pitation experiments (Figure 3). Lysates of Cos-1 cells, TRAPusing anti-myc antibody. Coimmunoprecipitated cotransfected with HA-tagged TACC1l and either myc- TACC1l proteins were detected with anti-HA antibodies tagged chTOG or myc-tagged TRAP(Figure 3a), were (Figure 3c). Blotting with anti-PCTAIRE2 antibodies immunoprecipitated with anti-TACC1-C antibody and revealed that a weak amount of endogenous PCTAIRE2 detected following Western blotting with anti-tag anti- kinase was bound to the precipitated proteins (lower bodies. Blotting with anti-HA antibody revealed panel) suggesting that PCTAIRE2 participates in the TACC1l in the immunoprecipitate (Figure 3a, lanes 3 TRAP–TACC1 complex. and 4); blotting with anti-myc antibody revealed both We tested whether there was an interaction between TRAP(lane 4) and chTOG (lane 3) as TACC1- human TACC1l and Aurora A, although this interac- associated proteins. The amount of tagged-proteins tion had not been suggested from our yeast two-hybrid expressed upon transfection was controlled by Western screens. Cos-1 cells were cotransfected with HA-tagged blot on total cell lysates using the same antibodies (lanes TACC1l and either myc-tagged Aurora A or mock

Oncogene TACC1–chTOG–Aurora protein complex N Conte et al 8105 vectors. HA–TACC1l and myc-Aurora A expressions In conclusion, these experiments showed that were controlled by Western blot (Figure 3d, lane 1). TACC1l associates with chTOG, TRAPand Aurora A TACC1 was immunoprecipitated with anti-TACC1-C kinase. The interaction was demonstrated for endogen- antibody and immunoprecipitated proteins were de- ous chTOG and Aurora A. The endogenous tected by Western blotting with anti-HA (lanes 2 and 4) PCTAIRE2 kinase was found in a complex with TACC1 or anti-Aurora A (lane 4) antibody. Interaction between and TRAP, which occurs probably via its interaction TACC1 and Aurora A was also found after coimmu- with TRAP/PCTAIRE2BP. noprecipitation of endogenous TACC1 in mitotic cells (not shown). Mapping of the regions of interaction within TACC1 To delineate the region of TACC1 that interacts with each of these partners, we used TACC1 fragment- encoded proteins fused to LEX binding domain (Figure 4a). They were used as baits in two-hybrid assays against LSM7, SmG, TRAPand chTOG preys. Results allowed the delineation of the region of binding to TACC1 for LSM/SmG and TRAP(Figure 4b). N1AN1 and N3AN3 fragments interacted with LSM7/ SmG and TRAP, respectively (Figure 4b). The LSM7/ SmG binding region is thus restricted to the N-terminus of TACC1, whereas TRAPbinds to a region that includes part of the first SPAZ motif. D-TACC interacts with MSPS (Gergely et al., 2000b) via its C-terminus. We tested if this is also the case for the human TACC1 protein. We made different constructions from the C- terminal region of TACC1 (Figure 4a) and tested the interaction with chTOG using the two-hybrid proce- dure. We were not able to conclude because the C- terminal fragments activated the system in a nonspecific fashion (not shown). GST pull-down and coimmunoprecipitation experi- ments confirmed that the N-terminus region of TACC1 is indeed the site of binding of Sm proteins (not shown). GST pull-down experiments (Figure 5a) showed that GST–chTOG precipitated TACC1 C1AC1 (lane 5) and C3AC3 (lane 9) fragments. Additional experiments (not shown) showed that the two fragments were able to precipitate endogenous TACC1l (probably due to dimerization via coiled-coil domains) and that at

Figure 3 TACC1l interacts with chTOG, TRAPand Aurora A. (a) HA-tagged TACC1l was coexpressed with either myc-tagged chTOG or myc-tagged TRAPin Cos-1 cells. Proteinswere immunoprecipitated with anti-TACC1-C antibody and bound proteins were respectively detected with anti-HA or anti-myc antibody, revealing coimmunoprecipitation of TACC1 with chTOG (lane 3) and with TRAP(lane 4). One-tenth of the lysate used for immunoprecipitation was run for control and probed with anti-HA or anti-myc (lanes 1 and 2). (b) Proteins from HME lysates were precipitated with either TACC1-com or TACC1-C and bound proteins were detected with anti-chTOG or anti-TACC1– SPAZ1 antibody, revealing coimmunoprecipitation of endogenous chTOG with TACC1 (lanes 1 and 2). (c) Lysates of Cos-1 cells cotransfected with HA-tagged TACC1l and myc-tagged TRAP were precipitated with anti-myc antibody. Blot anti-PCTAIRE2 revealed endogenous bound kinase. (d) HA–TACC1l and either myc-Aurora A or mock vector pRK5-myc, and myc-Aurora A and pCDNA3-HA, were cotransfected in Cos-1 cells. Proteins were immunoprecipitated with anti-TACC1-C antibody. Bound proteins were detected with either anti-HA or anti-Aurora A antibody, revealing coimmunoprecipitation of TACC1 with Aurora A (lane 4) but not in controls (lanes 2 and 3). One-tenth of the lysate used for precipitation was run for control (lane 1). Throughout the figure, the þ and À symbols mean transfected and not transfected, respectively

Oncogene TACC1–chTOG–Aurora protein complex N Conte et al 8106 mitotic Caco-2 or HeLa cells, in which metaphase and telophase are better observed, respectively (Figure 6). Anti-TACC1 commercial antibody weakly stained the centrosomes and diffusely stained the cytoplasm (a–c). chTOG staining appeared in the cytoplasm, at the centrosomes and as dots on the spindle (Figure 6d–f, details in f0 and schematic representation in f00). Aurora A staining was found in the spindle microtubules, and at the centrosomes where it colocalized with TACC1 (g–i) and chTOG (j–l). TACC1 is localized at the midbody in telophase, whereas chTOG and Aurora A are not (m–r). Thus, there is a common localization of the three proteins during metaphase and anaphase at centrosomes and the spindle poles, indicating that, at least in some areas and at some periods of the , the three proteins may be in the same complex.

Expression of Taxins, chTOG and Aurora A in breast cancer A microtubule-associated function in cell division could endow a Taxin with potential oncogenic or tumor suppressor capacity. To better understand the role of Taxins in cancer, their expression, together with that of TACC1-interacting partners was studied by immuno- histochemistry (IHC) on sections of a tissue microarray Figure 4 Mapping of TACC1 interactors using the two-hybrid (TMA) containing 592 breast tumor samples. Results method in yeast. (a) TACC1 protein is schematized with its three different regions: N-terminus, central serine/proline-rich region are shown in Figure 7 and Table 1. A total of 408 cases with two SPAZ motifs and coiled-coil C-terminus. Below are the were available for the analysis of the five markers after different fragments generated. (b) Results of the mapping of LSM7, staining. For each protein, normal epithelial cells SmG and TRAPinteractions using the different regions of TACC1 present in controls and close to tumor sections were in yeast two-hybrid experiments are summarized always stained. A total of 200 (49%), 228 (56%), 232 (57%), 408 (100%) and 257 (63%) cases showed a least one interaction could be indirect. To further staining equivalent for tumor and normal cells for characterize TACC1–chTOG interaction, we carried TACC1, TACC2, TACC3, chTOG and Aurora A, out coimmunoprecipitation experiments with HA- respectively. In contrast, 208 (51%), 180 (44%), 176 tagged TACC1 fragments and myc-tagged chTOG. (43%), 0 (0%) and 151 (37%) cases showed no staining Lysates from Cos-1 cells that coexpressed HA-tagged or a downregulated expression in tumors cells, for TACC1s, HA-tagged C1AC1, C2AC2 or C3AC3 with TACC1, TACC2, TACC3, chTOG and Aurora A, myc-tagged chTOG were immunoprecipitated with anti- respectively. Aurora A staining was strong in mitotically HA antibody and bound proteins were revealed by anti- active cells. Expression of TACC3 and Aurora A was myc antibody (Figure 5b). In addition to TACC1s (lane seen mostly in mitotic cells and therefore was strongly 1), C3AC3 (lane 4) (but no longer C1AC1, lane 2) was correlated with mitotic index and histological grade able to coimmunoprecipitate the chTOG protein. We (Po0.001) (Table 1). conclude that the interaction of TACC1 with chTOG is The expression of the three Taxins varied together. mediated via the C3AC3 fragment, which happens to be Using linear regression analyses, the correlation the most conserved region among the Taxins. between the three molecules was statistically significant GST pull-down experiments also showed that GST– (Figure 8a). Similarly, the expression of TACC3, and TRAPprecipitated TACC1 N2AN3 fragment to a lesser degree those of TACC2 and TACC1, (Figure 5c, lane 3). TRAPcoimmunoprecipitated with was correlated with that of Aurora A (Figure 8b, c). SPAZ fragment (Figure 5d, lane 2), but not with For each pairwise comparison, with the same number TACC1s (Figure 5d, lane 1). These results show that of samples (N ¼ 408), we calculated a coefficient of TRAPbinds TACC1 via its N-terminus region and correlation (R), a coefficient of determination (R2) and SPAZ region extremity. a P-value: for TACC1/TACC2, R ¼ 0.70, R2 ¼ 0.49, The mapping results are summarized in Figure 5e. P-value o0.001, for TACC1/TACC3, R ¼ 0.71, R2 ¼ 0.52, P-value o0.001 (Figure 8a), for Aurora A/ TACC2, R ¼ 0.69, R2 ¼ 0.48, P-value o0.001, for Subcellular colocalization of TACC1, chTOG and 2 Aurora A Aurora A/TACC3, R ¼ 0.82, R ¼ 0.67, P-value o0.0001 (Figure 8b), for TACC1/Aurora A, R ¼ 0.71, R2 ¼ 0.50, To further document the TACC1–chTOG–Aurora A P-value o0.001 and for Aurora A/Erbin – anti-Erbin complex in living cells, we looked for the respective (Borg et al., 2000) was used as control – no correlation subcellular localization of each of its components in was seen (Figure 8c).

Oncogene TACC1–chTOG–Aurora protein complex N Conte et al 8107

Figure 5 Mapping of TACC1–partner interactions. (a) Cos-1 cells were transfected with either myc-tagged C1AC1, C2AC2 or C3AC3 fragment of TACC1, and GST-chTOG fusion proteins were used in a pull-down experiment. One-tenth of the lysate used for precipitation was run for control (lanes 1–3). Blot with anti-myc revealed precipitated TACC1 C1AC1 (lane 5) and TACC1 C3AC3 fragments (lane 9). (b) Cos-1 cells were transfected with myc-tagged chTOG and with several HA-tagged fragments of TACC1 (C1AC1, C2AC2, C3AC3) or HA-tagged TACC1s. Proteins were immunoprecipitated with anti-HA antibody and bound proteins were detected with anti-myc antibody, revealing coimmunoprecipitation of TACC1s (lane 1) and C3AC3 fragment (lane 4) with chTOG. One-tenth of the lysate used for precipitation was run for control (total lysate). (c) Cos-1 cells were transfected with myc- tagged TACC1 N2AN3 fragment and GST or GST–TRAPfusion proteins were used in a pull-down experiment. Blot with anti-myc revealed precipitated TACC1 N2AN3 fragment (lane 3). One-tenth of the lysate used for precipitation was run for control (lane 1). (d) Cos-1 cells were transfected with myc-tagged proteins as indicated. Proteins were immunoprecipitated with appropriated anti- TACC antibodies and bound proteins were detected with anti-myc antibody, revealing coimmunoprecipitation of TRAPwith the SPAZ region of TACC1 (lane 2) but not with TACC1s (lane 1). (e) Summary of the interactions. The different interactors are listed below the corresponding regions of TACC1 with which they interact: other partners or processes involved are indicated below

Thus, as compared to normal, the expression of shown). Cell proliferation was affected; cell counts at 24, chTOG did not vary in breast tumors, whereas the 48 and 72 h after transfection showed that mock- expression of Taxins and Aurora A was downregulated transfected HeLa cells (control) continued to proliferate in a subset of cases. at 72 h, whereas the number of TACC1 and chTOG- depleted cells was significantly diminished at the same Depletion of either chTOG or TACC1 perturbates cell period (Figure 9b). There were variations in the intensity division of the effect depending on both the target gene and the selected siRNA; for chTOG-depleted cells, the effect was To determine if downregulation of Taxins and Taxin- more pronounced, and even visible at day 2 with containing protein complexes is a mere consequence of chTOG-a siRNA, whereas cells transfected by TACC1- tumor cell proliferation or can directly affect cell c behaved as the control cells. The study of the DNA behavior, we used transfection of small interfering content of HeLa cells treated with siRNAs by fluores- RNAs (siRNAs) to deplete HeLa cells of TACC1 or cence-activated cell sorter analysis showed that the cell chTOG RNA (Figure 9). Three different siRNAs were cycle profiles were abnormal in chTOG-depleted cells used for each gene. Western blot analysis (Figure 9a) while they remained normal in TACC1-depleted cells showed that RNA depletion blocked, although not (Figure 9c). chTOG-depleted cells showed a marked completely, the production of the corresponding protein increase in the G2/M phase. (upper panel) but not of a Tubulin (lower panel) or P85 To further characterize the mitotic features of phosphatidylinositol-3-kinase, an irrelevant protein (not TACC1-depleted and chTOG-depleted cells, we did an

Oncogene TACC1–chTOG–Aurora protein complex N Conte et al 8108

Figure 6 Localization of TACC1 complexes in vivo. Anti-TACC1com, anti-chTOG and anti-Aurora A antibodies were used in immunofluorescence studies on mitotic Caco-2 at metaphase (a–l) or HeLa cells at telophase (m–r) to delineate the localization of the three proteins during mitosis. Anti-a Tubulin was used to stain mitotic microtubule structures. Panels a–c: localization of TACC1 in the cytoplasm and at the centrosomes (arrow); d–f: localization of chTOG at the (arrow) and at the mitotic spindle (f 0: details of chTOG staining at the extremities of the polar microtubules; f00: schematic representation of chTOG localization on the spindle); g–i: colocalization of TACC1 and Aurora A at the centrosomes (arrow); j–l: colocalization of chTOG and Aurora A at the centrosomes (arrows); m–o: localization of TACC1 at the midbody; TACC1 localizes to the extremities and middle of the microtubule bundles (arrows); p–r: absence of localization of chTOG at the midbody

immunofluorescence analysis of these cells using anti- Discussion TACC1-com and anti-chTOG antibodies (Figure 10). chTOG-depleted cells displayed abnormal cell divisions We have shown here that TACC1 interacts with several with multipolar spindles. Similar aberrations were also partners, namely Sm proteins, adaptor protein TRAP, observed after TACC1-RNAi but at a lower frequency microtubule-associated chTOG and Aurora A serine/ (Figure 10B g–i). Staining with anti-Aurora A showed threonine kinase. These interactions involve distinct that Aurora A was still localized at the poles of regions of the TACC1 protein. We speculate that these multipolar spindles and was not affected by chTOG interactions may be time and space-regulated variably RNAi (Figure 10A g–l) or TACC1 RNAi (Figure 10B and participate in several cellular processes. The Taxin g–l). family is conserved in bilaterians and has expanded to

Oncogene TACC1–chTOG–Aurora protein complex N Conte et al 8109

Figure 7 TACC, Aurora A and chTOG protein expression in breast carcinomas studied by IHC on tissue microarrays. (a) Hematoxylin–eosin staining of a paraffin block section (35  25 mm) from a tissue microarray containing 592 arrayed tumor and control samples. (b) TACC1 staining. (c) Aurora A staining. (Inset: blow up of Aurora A staining in mitotic cells.) (d) chTOG staining. From B–D, the first and second panels are from tumor tissue, the third panel illustrates normal breast tissue. Magnification is  50 several members in species of vertebrates. This expan- reviews, see Mack et al., 2000; Blagden and Glover, sion must have occurred after separation of chordates 2003). Our findings are in close agreement with previous from the other vertebrates, but its time is impossible to reports (Gergely et al., 2000a). It would not be precise further since only one member is identified to surprising if other components of the PCM were found date in Xenopus. Some of the interactions (i.e. with to bind to Taxins; indeed, it is already known that rabbit chTOG and Aurora A) are conserved. TACC3 binds to centrosomal A-kinase anchoring protein AKAP350 (Steadman et al., 2002). Our results on the localization of chTOG agree also closely with Potential normal function(s) of TACC1 those of Lee et al. (2001) and Cullen and Ohkura (2001) Taxins are potential centrosomal regulators of micro- on D-TACC, and those of Charrasse et al. (1998) on tubule dynamics (for a review, see Gergely, 2002). They chTOG. The current hypothesis is that D-TACC and belong to the many coiled-coil proteins contained in the MSPS cooperate to organize bipolar spindles. Upon pericentriolar material (PCM) of the centrosome (for results of RNA interference experiments Gergely et al.

Oncogene TACC1–chTOG–Aurora protein complex N Conte et al 8110 Table 1 Expression of Taxins and Aurora A as measured by IHC on TMA, and mitoses in breast tumors Protein Staining Cases Grade(%) P-value Mitotic index(%) P-value

12 3 1 2 3

TACC1 ENS 208 45 25 30 43 28 29 DE 200 44 32 24 NS 47 30 23 NS

TACC2 ENS 229 38 30 32 39 29 32 DE 179 52 27 21 NS 52 27 21 NS

TACC3 ENS 164 32 24 44% 34 23 43 DE 244 52 33 15% o0.001 53 31 16 o0.001

Aurora A ENS 153 35 24 41 34 23 43 DE 255 50 29 21 o0.001 51 32 17 o0.001

ENS: cases showing tumor cells staining equivalent to normal cells staining DE: cases showing downregulation of expression in tumor cells in comparison to staining of normal cells NS: not significant

(2003) have suggested that chTOG is also essential for be played through the interaction with these proteins spindle pole organization. Our results are in complete (Figure 11). The potential relationship of the TACC– agreement with their findings. In addition, we show that chTOG complex with RNA is reinforced by the fact that Aurora A localization at the spindle is not grossly chTOG interacts with the heterogeneous nuclear ribo- modified in chTOG-depleted cells. This suggests that nucleoprotein hnRNPA2, an RNA trafficking-associated Aurora A is localized to this site independently of protein (Barbarese et al., 2002). One of the other chTOG. Neither chTOG nor TACC1 localization interacting proteins we identified, TRAP/PCTAIRE2BP, seemed affected in the partner-depleted cells (not has Tudor domains. Tudor domains, first described in shown), although it was sometimes difficult to judge the Drosophila Tudor protein, are found in RNA-binding when the perturbation was intense. proteins and are also thought to intervene in Sm protein Independently, the Xenopus TACC protein Maskin binding (Selenko et al., 2001). Whether TRAPbinds inhibits translation of maternal mRNAs through bind- RNA or Sm proteins, or both, is currently unknown. ing to CPEB and eIF-4E (Stebbins-Boaz et al., 1999) Thus, a number of studies indicate that TACC/ and of maternal mRNAs being transported to the Maskin, chTOG/XMAP215/MSPS and Aurora kinase spindles (Groisman et al., 2000). TACC/Maskin, participate in the same functional complex(es) to chTOG and Aurora A colocalize at some specific areas promote cell division. Maskin and Aurora are also where they interact with Cyclin B1 and with the found in nondividing cells (Huang et al., 2002). A microtubules. Human chTOG associates with Cyclin unifying theory (Figure 11) would propose that one role B1 and targets the CDC2 kinase to the microtubules of Taxins may be in the control of mRNA biology in (Charrasse and Larroque, 2000). In Xenopus, Maskin conjunction with microtubule organization and in also colocalizes with Cyclin B1 at the mitotic apparatus association with chTOG and Aurora A. The TACC– (Groisman et al., 2000). During meiosis, the RNA- chTOG–Aurora or TACC–chTOG–PCTAIRE com- binding, cytoplasmic polyadenylation element binding plex(es) may serve several functions, including one CPEB (Mendez and Richter, 2001), which interacts with during cell division. Another role might be in the Maskin (Stebbins-Boaz et al., 1999), is first activated by establishment or maintenance of cell polarity (Richter, Aurora and then, to allow transcription of some 2001); a role for Taxin-associated complexes in asym- mRNAs, degraded after CDC2 phosphorylation (Men- metric cell division is also possible (Barbee et al., 2002; dez et al., 2002). During mitosis, the Maskin-induced Berdnik and Knoblich, 2002); this might be one repression of Cyclin B1 mRNA translation is overcome explanation of the hematopoietic stem cell defects in by polyadenylation after Aurora phosphorylation of Tacc3 deficient mice (Piekorz et al., 2002). TACC1 CPEB (Groisman et al., 2002). We did not find binding might also participate in the regulation of cytokinesis of TACC1 to CPEB using coimmunoprecipitation and interact at the midbody with another molecular experiments (not shown); TACC1–CPEB interaction complex, including instead of Aurora A the midbody- may not be conserved in humans, may be restricted to associated , or another mitotic kinase another Taxin, or may occur through other, RNA- such as a Polo kinase. Similarly, TACC3, whose gene is binding proteins. We showed that TACC1 interacts with highly expressed in the testis and ovary (Hao et al., 2002; Sm proteins and mapped this interaction. In humans, a Aitola et al., 2003; Sadek et al., 2003) may also interact ‘Maskin-like’ role of TACC1 in RNA metabolism might with Aurora C which plays a major role in meiosis.

Oncogene TACC1–chTOG–Aurora protein complex N Conte et al 8111 cancer progression and high-grade tumors (Ma et al., 2003). The Aurora A/BTAK/STK6/STK15 gene maps to 20q13.31 region, which is amplified to around 10–15% of breast cancers (Sen et al., 1997; Nonet et al., 2001). Overexpression of Aurora A mRNA and protein is associated with chromosome instability (for reviews, see Miyoshi et al., 2001; Dutertre et al., 2002) and frequently found in colorectal cancers (Bischoff et al., 1998). Ectopic expression of Aurora A in cultured cells induces transformation, aneuploidy and centrosome amplification (Zhou et al., 1998). Aurora A- deficient cells also show abnormal mitoses as well as chromosome segregation and cytokinesis defects (Roghi et al., 1998). These studies demonstrate that regulation of Aurora A levels is critical for centrosome function and can be altered in cancer via overexpression or loss- of-function mutation (for a review, see Bischoff and Plowman, 1999). Aurora A interacts with P53 (Chen et al., 2002), which may have something to do with the phenotype of Tacc3 mouse mutants (Piekorz et al., 2002). An Sm like protein, CaSm/LSM1, has been shown to contribute to the transformed state of pancreatic cells (Schweinfest et al., 1997). Other components of Taxin complexes may be important in oncogenesis; a recent study has shown that upregulation of SmF and hnRNPA2 is associated with tumor metastasis (Ramaswamy et al., 2003). Variations of the molecular complexes can affect centrosome function and cell division, and participate in oncogenesis (for reviews, see Kra¨ mer et al., 2002; Nigg, 2002). Our RNAi experiments showed that depletion of either TACC1 or chTOG affects cell division. chTOG depletion, in particular, leads to cells with disorganized and multipolar spindles. This type of alteration is generally antagonistic to cell viability and aberrant mitotic cells are usually eliminated by apoptosis after checkpoint processing (Brinkley, 2001). Accordingly, we observed a decrease in the number of chTOG-depleted viable cells after 2 days. To some extent, the same was Figure 8 Similar variations in expression levels of Taxins and true for TACC1-depleted cells, indicating that some Aurora A. (a) The expression levels of TACC1, TACC2, and alteration of cell division was present although the TACC3 as measured by IHC on TMA correlate, as determined by simple linear regression analysis. Similarly, (b and c) the expression phenotype was not as spectacular as for chTOG- levels of TACC2, TACC3, Aurora A and TACC1 correlated. Erbin depleted cells; it is possible that some compensation of protein (Borg et al., 2000) was used as a control of the absence of the defect is ensured by a correct level of the two other covariation Taxins. In tumors, some cells accommodate perturbations of cell division and survive (Brinkley, 2001); their resulting Potential role(s) of TACC–chTOG–Aurora A aneuploidy may endow the tumor with increased complex(es) in breast cancer aggressivity. The observed downregulation of Taxins in breast tumors might be the result of this clonal Whereas chTOG expression remains quite stable, Taxins selection towards a genetically unstable phenotype. and Aurora A are downregulated in a subset of breast Downregulation of chTOG dramatically affected cell tumors and remain strongly expressed in others. In a division; and yet, we did not observe a variation of subset of cases, Taxins and Aurora A may even be chTOG expression in breast tumors. We hypothesize overexpressed, but this was difficult to measure since that the effect of chTOG downregulation on cell staining was already very strong in normal cells. TACC3 division may be too drastic to allow survival even and Aurora A expression remains particularly strong in through modulation of checkpoints. This might be why high-grade tumors and in tumors with a high mitotic this event is not selected for, and why it was not index. Increased expression of TACC3 and Aurora A observed in our series of tumor samples. It is possible mRNA has recently been found associated with breast also that some differences exist between HeLa cells and

Oncogene TACC1–chTOG–Aurora protein complex N Conte et al 8112

Figure 9 Depletion of chTOG and TACC1 RNA in HeLa cells leads to abnomal cell division. (a) TACC1 and chTOG protein levels are reduced by siRNA treatment. Western blot analysis of the levels of TACC1 and chTOG after different periods of siRNA treatment. a Tubulin is shown as a loading control in lower lanes. Left upper panel: Western blot hybridization with anti-TACC1-com antibody shows the reduced level of TACC1 after siRNA treatment with the three different oligonucleotides (TACC1-a, TACC1-b and TACC1- c). Control lane shows the level of TACC1 in mock-depleted cells (oligofectamine only). Right upper panel: Western blot hybridization with anti chTOG antibody shows the reduced level of chTOG after siRNA treatment with the three different oligonucleotides (chTOG- a, chTOG-b, chTOG-c). (b) Counts of siRNA-treated cells. The number of living siRNA treated cells and mock-treated cells was estimated daily (24–72 h) by cell count after Trypan blue staining. Data shown are the average of three independent experiments7standart deviation. (c) Analysis of the cell cycle of siRNA treated cells by flow cytometry. M1, M2 and M3 represent the percentage of cells in the G0/G1, S and G2/M phase, respectively. Left, middle and right panels show propidium iodide incorporation of mock-depleted cells, TACC1-depleted cells after 24, 48 and 72 h, and chTOG-depleted cells after 24, 48 and 72 h, respectively

breast tumor cells with respect to sensitivity of Taxin http://www.jgi.doe.gov/programs/ciona.html), to find genomic and chTOG downregulation. sequences with similarities to the vertebrate proteins. All the Breast tumors with abnormal content of Aurora A reads were collected and assembled with the Cap3 program kinase (Anand et al., 2003) or Taxins may not react well (Huang and Madan, 1999). After the first round of assembly, to a systemic treatment by taxanes. Our preliminary we obtained a collection of contigs. Each contig was scanned with GenScan to detect exons. A contig that coded for a experiments show that Taxins transcription is indeed putative Ciona TACC protein was extended by blast against modified by treatment with taxol. The full under- Ciona reads database. We extended the cDNA until the standing of the role of Taxin-associated complexes on obtention of a 251 amino acids Ciona TACC protein, which we microtubule behavior and cell division and the measure- named TACC Cin. Other metazoan sequences were retrieved ment of the level of expression of Taxins in tumors could from all nonredundant databases such as GenBank, SwissProt, thus have a direct impact on the management of breast NCBI databases. TACC sequences accession numbers are as cancer with the possibility to refine prognosis, derive follows: O75410 TACC1 Hsa (Homo sapiens), Q9JJG0 new drugs and understand resistances to treatment. TACC2 Hsa, Q9Y6A5 TACC3 Hsa, AAK54244 TACC4 Ocu (rabbit Oryctolagus cuniculus), Q9JJ11 TACC3 Mmu (Mus musculus), NP_067289 TACC2 Mmu, AAF66735 D- TACC Dme (D. melanogaster), AAF19726 Maskin Xla (X. Materials and methods laevis), Y54E2A.3 TAC-1 Cel (C. elegans). Multiple align- ments were done with Clustalx 1.81 for Linux. Phylogeny Identification of Ciona TACC ortholog and phylogenetic analysis was done with both complete TACC protein analysis sequences and TACC C-terminal region with PhyloWin The sequence of the C-terminal region common to members (neighbor-joining-Poisson distance correction and maximum of the TACC family was blasted against the JGI Ciona parsimony methods, global gap removal and 500 bootstrap reads database (http://ghost.zool.kyoto-u.ac.jp/indexr1.html, replicates).

Oncogene TACC1–chTOG–Aurora protein complex N Conte et al 8113 Cell lines CA, USA). Cells were grown as recommended. Nine breast The HeLa, Caco-2 and Cos-1 cell lines were purchased from tumor cell lines were used as controls in the tissue microarrays the American Type Culture Collection (http://www.atcc.org/). as described previously (Ginestier et al., 2002). The HME1 (human primary mammary epithelial) cell line (Clontech C4002-1) was purchased from Clontech (Palo Alto, Cloning procedures The cloning of different fragments of TACC1 was done following PCR amplification using the Gateway system (Invitrogen). The TACC1l N-terminus region N1AN3 (aa residues 1–259) and C-terminus region C1AC3 (422–805) were both separated into five fragments (see Figure 4a) (sequences of primers are available on demand). All fragments were then subcloned into different expression vectors: pcDNA-3 HA, pRK5myc, pBTM116 and pACT2, using the Gateway system. TACC1s, which corresponds to a spliced form of TACC1l (Conte et al., 2002) and TACC1spaz (aa 203–426 of TACC1l), which corresponds to the serine and proline-rich fragment of TACC1l (Conte et al., 2002) were subcloned in pBTM116 vector for two-hybrid screen in yeast.

Two-hybrid procedures Two baits, TACC1s and TACC1spaz, fused to the LexA-BD subunit using pBTM116 vector were used to screen an oligo- dT primed human cDNA breast library cloned in pACT2 vector. Yeast two-hybrid screens were performed as previously described (Conte et al., 2002). For LexA-TACC1s and LexA-TACC1spaz respectively, 115 and 85 Trp þ Leu þ transformants were selected on plates with supplemented minimum medium that lacked tryptophan, leucine, histidine in the primary screening and contained 10 mm 3-aminotriazole (3-AT) and then tested for the b-galactosidase activity by the filter method in the secondary screening. After rescue, the DNA of the selected clones was retransformed in L40 yeast containing LexA– TACC1s, LexA–TACC1spaz or LexA fused to control proteins. Specific clones were positive for growth in histidine deficient medium and for b-galactosidase activity. Interacting partners (chTOG and TRAP, see results) were then reamplified using primers located in 50 and 30 of pACT2 vector multiple

Figure 10 Abnormal mitosis with multipolar spindles in chTOG and TACC1-depleted cells. HeLa cells were stained with anti- TACC1-com, anti-chTOG, anti-Aurora A and anti-a Tubulin antibodies were used for immunofluorescence analysis on mitotic 24, 48, 72 h siRNA-treated HeLa cells at metaphase. In all experiments, images were acquired from experimental and control cells that were treated and processed for immunofluorescence at the same time using identical settings on the confocal microscope. (a) chTOG-a siRNA-treated cells at 24 h (a–c), 48 (d–f) and 72 h (g–l). a, d, g, j: chTOG staining after 24, 48 and 72 h siRNA treatment; siRNA treatment strongly reduces chTOG levels in most cells from 48 h but small amounts of proteins are still detectable on the spindles of mitotic cells. b, e, h: a Tubulin staining after 24, 48 and 72 h siRNA treatment. At 24 h (b), most cells in mitosis exhibit normal mitotic spindles. From 48 to 72 h (e, h) most cells in mitosis have abnormal mitotic spindles with more than two spindle poles. k: Aurora A staining in chTOG-depleted cells; despite abnormal spindles, Aurora A is still localized at centrosomes. c, f, i, l: merge images. (b) TACC1-a siRNA treated cells at 24 h (a–c), 48 (d–f) and 72 h (g–l). a, d, g, j: TACC1 staining after 24, 48 and 72 h siRNA treatment. Treatment strongly reduces TACC1 levels in most cells at 48 h, but small amounts of protein are still detectable on mitotic cells. b, e, h: a Tubulin staining after 24, 48 and 72 h siRNA treatment; at 24 h (b), 48 h (e) and 72 h (h) most cells in mitosis have normal mitotic spindles. k: Aurora A staining at multipolar spindles in TACC1-depleted cells; Aurora A is still localized at centrosomes. c, f, i, l: merge images

Oncogene TACC1–chTOG–Aurora protein complex N Conte et al 8114 room temperature, rinsed several times in PBS, and then incubated with 2 mg/ml of the Alexa-conjugated either anti- rabbit or anti-mouse secondary antibody. Confocal analysis. The three-dimensional immunofluorescence images were done as previously described (Conte et al., 2002).

Breast tumor samples Tumor samples were obtained from women treated at the Institut Paoli-Calmettes. The tissue microarrays constituted of 592 samples of localized forms of breast cancer collected between 1987 and 1999 (median follow-up: 48 months) for Figure 11 Hypothetical model of one type of TACC1-associated which a sample had been frozen in liquid nitrogen. Tumors protein complex in cells. The protein complex associated with were classified according to the WHO classification; the TACC1 is shown with two potential steps while it migrates along a histoprognostic grade used was the modified Scarf Bloom microtubule track. The complex includes mRNA (e.g. Cyclin B) Richardson (SBR) grading for invasive lesions. There were 401 transported to a specialized site (e.g. spindle pole, region of ductal, 77 lobular, 40 mixed, four medullary carcinomas and polarized cell). It moves along the track bound to the microtubule- 70 other histological types. A total of 297 tumors were N þ associated molecule chTOG via TACC1 C-terminus. During and 450 were positive for estrogen receptor. transport, translation is blocked. The precise role of the LSM ring is not known. Once arrived at that site, masking is abrogated, possibly after phosphorylation by kinases (e.g. Aurora A, Histological procedures PCTAIRE2 or others) and/or degradation of TACC1 The characteristics of the antibodies used are listed in Table 2. TMA were prepared as described previously (Ginestier et al., cloning site. PCR products were subcloned in pRK5 myc- 2002). IHC was carried out on 5-mm sections of formalin- tagged and pDEST15 gateway vector for biochemical studies. embedded tissue specimens. They were deparaffinized in histolemon (Carlo Erba Reagenti, Rodano, Italia) and Biochemical procedures rehydrated in graded alcohol. Antigen enhancement was done Preparation of cell lysates, GST pull-down, coimmunopreci- by incubating the sections in target retrieval solution (Dako, pitation experiments and immunoblottings were done as Coppenhagen, Denmark) as recommended. Slides were then described (Conte et al., 2002). transferred to a Dako autostainer. Staining was done at room temperature as follows: after washing in phosphate buffer, followed by quenching of the endogenous peroxidase activity Immunofluorescence analyses by treatment with 0.1% H2O2, slides were first incubated Caco-2 and HeLa cells were grown on coverslips (2 Â 105 cells / with blocking serum (Dako) for 10 min and then with the 60 mm plates). After 24 h, cells were washed once in PBS and affinity-purified antibody for 1 h. After washes, slides were fixed for 5 min with cold methanol. After two washes, cells incubated with biotinylated antibody against rabbit Ig were permeabilized for 5 min with 5% fetal calf serum/PBS, for 20 min followed by streptavidin-conjugated peroxidase 0.1% Triton X-100 and blocked in 5% fetal calf serum/PBS (Dako LSABR2 kit). Diaminobenzidine or 3-amino-9-ethyl- for 20 min. Cells were then incubated with anti-a Tubulin, anti- carbazole was used as the chromogene, counterstained chTOG, anti-TACC1-com, anti-Aurora A antibodies for 1 h at with hematoxylin, and coverslipped using Aquatex (Merck,

Table 2 List of proteins, cell organelles and protein tags studied by various procedures, and characteristics of the corresponding antibodies Protein/tag Antibody Origin Designation Experiment/dilution

Aurora A kinase Mmab C. Prigent, Rennes 35C1 (Cremet et al., 2003) IHC 1/25 IF 1/500 WB 1/1000 chTOG Rpab C. Larroque, Montpellier IHC 1/300 IF 1/500 WB 1/1000 Erbin Rpab JP. Borg, U119 2 (Borg et al., 2000) IHC 1/25 Estrogen receptor Mmab Novocastra Laboratories 6F11 IHC 1/60 Ki67 Mmab Dako Corporation KI-67 IHC 1/100 Microtubules, spindle (a Tubulin) Mmab Sigma DM 1A IF 1/500 WB 1/500 PCTAIRE2 kinase (C-16) Rpab Santa Cruz Biotechnology sc-175 WB 1/200 TACC1 Rpab Upstate Biotechnology Euromedex TACC1-com, 07-229 IHC/1/500 IF 1/250 TACC1 Rpab N. Conte, U.119 TACC1-C (Conte et al., 2002) WB 1/300 TACC1 Rpab N. Conte, U.119 TACC1-SPAZ1 (Conte et al., 2002) WB 1/300 TACC2 Rpab Upstate Biotechnology Euromdex 07-228 IHC 1/40 TACC3 Rpab Upstate Biotechnology Euromedex 07-233 IHC 1/100 Myc Mmab Santa Cruz Biotechnology 9E10 WB 1/200 HA rPab Roche Diagnostics 3F10 WB 1/1000

Rpab: rabbit polyclonal antibody; Mmab: mouse monoclonal antibody; rPab: rat polyclonal antibody IHC: immunohistochemistry; IF: immunofluorescence; WB: Western blotting

Oncogene TACC1–chTOG–Aurora protein complex N Conte et al 8115 Darmstadt, Germany) mounting solution. The slides were temperature. During incubation, 175 ml OPTIMEM 1 medium evaluated under a light microscope by a pathologist (JJ). was mixed with 10 ml siRNA at 20 mm. The two mixtures were Immunoreactivities were classified by estimating the quick combined and incubated for 20 min at room temperature for score (Q) as previously described (Ginestier et al., 2002). complex formation. Cells were assayed 24–48–72 h after transfection. Oligofectamine alone was used for mock trans- Statistical analyses fection. Specific silencing was confirmed by Western blot analysis. Transfected cells grown in six-well plates were To assess the relationship between two variables assumed to be trypsinized and viable cells were counted daily with Trypan related (i.e. coregulated molecules), simple linear regression blue. Cells were then lysed in conditions described in analyses were carried out using Excel Software (Microsoft). biochemical procedures. Following cell lysis, proteins were For these tests, (0,0) points were removed; the relationship quantified by the Bradford reagent (Bio-Rad). Equal amounts tested was thus for cases with at least one positive value. Each of protein were separated on 7.5 % polyacrylamide gels and result is given with: ‘N’ the sample size, ‘a’ the slope of the transferred to nitrocellulose. Immunostainings were done with 2 regression line, the P-value and ‘R ’, the coefficient of anti-TACC1-com and anti-chTOG at concentrations described determination. Thus, for each positive comparison a linear in Table 2. Cells treated for RNAi were analysed by flow relationship can be determined (e.g. y ¼ 0.8x þ 20 means cytometry. Briefly, cells were harvested and washed in cold TACC2 ¼ 0.8Aurora A þ 20). PBS. Cells were then fixed in cold 70% ethanol. After two washes in PBS, cells were treated with RNAse A (250 mg/ml) RNA interference procedures and propidium iodide was added. After incubation for 30 min 1 To obtain target-specific siRNA duplexes, we used the design at 30 C, cell cycle was analysed by flow cytometry. Immuno- program of Eurogentec (Seraing, Belgium) (http://www.eur- fluorescence analyses of transfected cells were done with anti- ogentec.com/code/en/sirna.htm). Selected siRNA sequences TACC1-com, anti-a Tubulin, anti-Aurora A and anti-chTOG were submitted to a BLAST search against the antibodies as described above. sequence to ensure that only one gene of the human genome was targeted. All RNA oligos were made using the standard Acknowledgements phosphoramidite solid-phase synthesis technology. Sequences We thank F Birg and D Maraninchi for discussions and of oligonucleotides are available on demand. siRNA duplex support. We are especially grateful to J-PBorg, F Coulier, formation was performed as described by the provider. Human S Descamps, S Fekairi, A Gonc¸alve` s and M Leveugle for HeLa cells were grown at 371C in Dulbecco’s modified Eagle discussions and help. We thank D Weil for the CPEB medium supplemented with 10% FCS. The day before construct. This work was supported by INSERM, Institut transfection, cells were trypsinized, diluted with fresh medium Paoli-Calmettes, and grants from the Ligue Nationale contre and transferred to six-well plates (150 000 cells per well). le Cancer (LNCC). The laboratories of BS, CPand DB Transient transfection of siRNAs was carried out using are supported by a ‘Label’ from the LNCC. BD and CG are Oligofectamine (Life Technologies). A volume of 12 ml currently recipients of a fellowship from the ministry of OPTIMEM 1 medium (Life Technologies) and 3 ml Oligofec- Research, and NC is a recipient of a fellowship from tamine per well were preincubated for 5–10 min at room Association pour la Recherche contre le Cancer.

References

Aitola M, Sadek CM, Gustafsson JA and Pelto-Huikko M. Charrasse S, Schoeder M, Gauthier-Rouvie` re C, Ango F, (2003). J. Histochem. Cytochem., 51, 455–469. Cassimeris L, Gard DL and Larroque C. (1998). J. Cell Anand S, Penrhyn-Lowe S and Venkitaraman AR. (2003). Science, 111, 1371–1383. Cancer Cell, 3, 51–62. Chen H-M, Schmeichel K, Mian LS, Lelie` vre S, Petersen OW Barbarese E, Kosturko L, D’Sa C and Carson JH. (2002). J. and Bissell MJ. (2000). Mol. Biol. Cell, 11, 1357–1367. Neurochem., 81 (Suppl. 1), 93–94. Chen SS, Chang PC, Cheng YW, Tang FL and Lin YS. (2002). Barbee SA, Lublin AL and Evans TC. (2002). Curr. Biol., 12, EMBO J., 21, 4491–4499. 1502–1506. Conte N, Charafe-Jauffret E, Delaval B, Ade´ laı¨ de J, Ginestier Berdnik D and Knoblich JA. (2002). Curr. Biol., 12, 640–647. C, Geneix J, Isnardon D, Jacquemier J and Birnbaum D. Bischoff JR, Anderson L, Zhu Y, Mossie K, Ng L, Souza B, (2002). Oncogene, 21, 5619–5630. Schyver B, Flanagan P, Clairvoyant F, Ginther C, Chan CS, Cremet JY, Descamps S, Ve´ rite´ F, Martin A and Prigent C. Novotny M, Slamon DJ and Plowman GD. (1998). EMBO (2003). Mol. Cell. Biochem., 243, 123–131. J., 17, 3052–3065. Cullen CF and Ohkura H. (2001). Nat. Cell Biol., 31, 637–642. Bischoff JR and Plowman GD. (1999). Trends Cell Biol., 9, Dutertre S, Descamps S and Prigent C. (2002). Oncogene, 21, 454–459. 6175–6183. Blagden SPand Glover DM. (2003). Nat. Cell Biol., 5, 505–511. Gergely F. (2002). BioEssays, 24, 915–925. Borg JP, Marchetto S, Le Bivic A, Ollendorff V, Jaulin- Gergely F, Draviam VM and Raff JW. (2003). Genes Dev., 17, Bastard F, Saito H, Fournier E, Ade´ laı¨ de J, Margolis B and 336–341. Birnbaum D. (2000). Nat. Cell. Biol., 2, 407–414. Gergely F, Karlsson C, Still I, Cowell J, Kilmartin J and Bouveret E, Rigaut G, Shevchenko A, Wilm M and Se´ raphin Raff JW. (2000a). Proc. Natl. Acad. Sci. USA, 97, B. (2000). EMBO J., 19, 1661–1671. 14357–15352. Brinkley BR. (2001). Trends Cell Biol., 11, 18–21. Gergely F, Kidd D, Jeffers K, Wakefield JG and Raff JW. Charrasse S and Larroque C. (2000). Exp. Cell Res., 254, (2000b). EMBO J., 19, 241–252. 249–256. Giet R, McLean D, Descamps S, Lee MJ, Raff JW, Prigent C Charrasse S, Mazel M, Taviaux S, Berta P, Chow T and and Glover DM. (2002). J. Cell Biol., 156, 437–451. Larroque C. (1995). Eur. J. Biochem., 234, 406–413. Giet R and Prigent C. (1999). J. Cell. Science, 112, 3591–3601.

Oncogene TACC1–chTOG–Aurora protein complex N Conte et al 8116 Ginestier C, Charafe-Jauffret E, Bertucci F, Eisinger F, Geneix Nigg EA. (2001). Nat. Rev. Mol. Cell Biol., 2, 21–32. J, Bechlian D, Conte N, Ade´ laı¨ de J, Toiron Y, Nguyen C, Nigg EA. (2002). Nat. Rev. Cancer, 2, 1–11. Viens P, Mozziconacci MJ, Houlgatte R, Birnbaum D and Nonet GH, Stampfer MR, Chin K, Gray JW, Collins CC and Jacquemier J. (2002). Am. J. Pathol., 161, 1223–1233. Yaswen P. (2001). Cancer Res., 61, 1250–1254. Groisman I, Huang Y-S, Mendez R, Cao Q, Theurkauf W and Piekorz RP, Hoffmeyer A, Duntsch CD, McKay C, Nakajima Richter JD. (2000). Cell, 103, 435–447. H, Sexl V, Snyder L, Rehg J and Ihle JN. (2002). EMBO J., Groisman I, Jung MY, Sarkissian M, Cao Q and Richter JD. 21, 653–664. (2002). Cell, 109, 473–483. Popovici C, Leveugle M, Birnbaum D and Coulier F. (2001). Hao Z, Stoler MH, Sen B, Shore A, Westbrook A, Flickinger Biochem. Biophys. Res. Commun., 288, 362–370. CJ, Herr JC and Coonrod SA. (2002). Mol. Reprod. Pu JJ, Li C, Rodriguez M and Banerjee D. (2001). Cytokine, Develop., 63, 291–299. 13, 129–137. Hirose T, Kawabuchi M, Tamaru T, Okumura N, Nagai K Raff JW. (2002). Trends Cell Biol., 12, 222–225. and Okada M. (2000). Eur. J. Biochem., 267, 2113–2121. Ramaswamy S, Ross KN, Lander ES and Golub TR. (2003). Hirose T, Tamaru T, Okumura N, Nagai K and Okada M. Nat. Genet., 33, 1–6. (1997). Eur. J. Biochem., 249, 481–488. Richter JD. (2001). Proc. Natl. Acad. Sci. USA, 98, Huang X and Madan A. (1999). Genome Res., 9, 868–977. 7069–7071. Huang YS, Jung MY, Sarkissian M and Richter JD. (2002). Roghi C, Giet R, Uzbekov R, Morin N, Chartrain I, Le EMBO J., 21, 2139–2148. Guellec R, Couturier A, Dore´ e M, Philippe M and Prigent Kra¨ mer A, Neben K and Ho AD. (2002). Leukemia, 16, 767– C. (1998). J. Cell. Sci., 111, 557–572. 775. Sadek CM, Jalaguier S, Feeney E, Aitola M, Damdimopoulos Lappin TR, Mullan RN, Stewart JP, Morgan NA, Thompson AE, Pelto-Huikko M and Gustafsson J-A. (2000). Mech. A and Maxwell AP. (2002). Leuk. Lymphoma, 43, 1455–1459. Develop., 97, 13–26. Lauffart B, Gangisetty O and Still IH. (2003). Genomics, 81, Sadek CM, Pelto-Huikko M, Tujague M, Steffensen KR, 192–201. Wennerholm M and Gustafsson JA. (2003). Gene Exp. Lauffart B, Howell SJ, Tasch JE, Cowell JK and Still IH. Patterns, 3, 203–211. (2002). Biochem. J., 363, 195–200. Salgado-Garrido J, Bragado-Nilsson E, Kandels-Lewis S and Lee MJ, Gergely F, Jeffers K, Peak-Chew SY and Raff JW. Se´ raphin B. (1999). EMBO J., 18, 3451–3462. (2001). Nat. Cell. Biol., 3, 643–649. Schweinfest CW, Graber MW, Chapman JM, Papas TS, Line A, Slucka Z, Stengrevics A, Li G and Rees RC. (2002a). Baron PL and Watson DK. (1997). Cancer Res., 57, Cancer Genet. Cytogenet., 139, 78–83. 2961–2965. Line A, Stengrevics A, Slucka Z, Li G, Jankevics E and Rees Selenko P, Sprangers R, Stier G, Buhler D, Fischer U and RC. (2002b). Br. J. Cancer, 86, 1824–1830. Sattler M. (2001). Nat. Struct. Biol., 8, 27–31. Ma XJ, Salunga R, Tuggle JT, Gaudet J, Enright E, McQuary Sen S, Zhou H and White RA. (1997). Oncogene, 14, 2195–2200. P, Payette T, Pistone M, Stecker K, Zhang BM, Zhou YX, Spittle C, Charrasse S, Larroque C and Cassimeris L. (2000). Varnholt H, Smith B, Gadd M, Chatfield E, Kessler J, Baer J. Biol. Chem., 275, 20748–20753. TM, Erlander MG and Sgroi DC. (2003). Proc. Natl. Acad. Steadman BT, Schmidt PH, Shanks RA, Lapierre LA and Sci. USA, 100, 5974–5979. Goldenring JR. (2002). J. Biol. Chem., 363, 195–200. Mack GJ, Ou Y and Rattner JB. (2000). Microscop. Res. Stebbins-Boaz B, Cao Q, de Moor CH, Mendez R and Richter Tech., 49, 409–419. JD. (1999). Mol. Cell, 4, 1017–1027. McKeveney PJ, Hodges VM, Mullan RN, Maxwell P, Still IH, Hamilton M, Vince P, Wolfman A and Cowell JK. Simpson D, Thompson A, Winter PC, Lappin TR and (1999a). Oncogene, 18, 4032–4038. Maxwell AP. (2001). Br. J. Haematol., 112, 1016–1024. Still IH, Vince Pand Cowell JK. (1999b). Genomics, 58, Mendez R, Barnard D and Richter JD. (2002). EMBO J., 21, 165–170. 1833–1844. Tharun S, He W, Mayes AE, Lennertz P, Beggs JD and Parker Mendez R and Richter JD. (2001). Nat. Rev. Mol. Cell. Biol., R. (2000). Nature, 404, 515–518. 2, 521–529. Theurkauf WE. (2001). Nat. Cell Biol., 3, E159–E161. Miyoshi Y, Iwao K, Egawa C and Noguchi S. (2001). Int. J. Zhou H, Kuang J, Zhong L, Kuo WL, Gray JW, Sahin A, Cancer, 92, 370–373. Brinkley BR and Sen S. (1998). Nat. Genet., 20, 89–93.

Oncogene