Infiltrating ductal primary breast carcinomas were obtained from patients included between in the prospective database initiated in 1981 by the Institut Curie Breast Cancer group. All patients gave original verbal consent on the use of tumor specimens for research purposes. Eleven normal breast specimens were obtained from breast mammoplasty surgery. Total RNA was extracted by the cesium chloride method and hybridized to DNA microarrays (Human Genome, Affymetrix HG-U133 set) as previously described [22]. Microarray data reported here are in accordance with MIAME guidelines and have been partially published elsewhere [22]. All samples were reviewed and classified by the same breast pathologist (AVS). Histologic nuclear grades were scored (I to III) according to Elston and Ellis [23]. Estrogen receptor (ER) and progesterone receptor (PR) expression were determined by immunohistochemistry [24] and were considered negative when less than 10% of the tumor cell nuclei showed staining. HER2 expression was evaluated by immunostaining as described previously [24] and was considered positive when IHC staining was 3+ (uniform, intense membrane staining of more than 30% of invasive tumor cells), as recommended by Wolff et al. [25]. This study was approved by the institutional review boards and ethics committee of the Institut Curie.

Human cancer cell lines MDA-MB-231, MDA-MB-468, SK-MES, HeLa and HeLa-H2B were maintained at 37°C with 5% CO2 in DMEM 4.5 g/l glucose supplemented with 10% fetal calf serum (FCS). MCF7 breast cancer cells were grown in DMEM/F12 medium with 5% FCS. All cancer cell lines are available from ATCC. HeLa-H2B cells stably expressing the mCherry-histone were kindly provided by Dr. Valérie Doye (Institut Jacques Monod, Paris). The source, tumor type and histologic characteristics of breast cancer cell lines used in this study have been described in details elsewhere [26], [27].

The coding sequence of human ATIP3a (3812 bp) was obtained by ligating partial 3′ ATIP sequence isolated from human lung cDNA library [19] with the 5′ region of IMAGE clone ID 3835953 (ATCC). Sequencing of the complete cDNA confirmed its identity with Genbank accession number NM_001001924. The full-length ATIP3 sequence was then PCR-amplified and subcloned into the BamHI and XhoI restriction sites of the pEGFP-C1 vector (Clontech) so that the initiating codon of ATIP3 is in frame with the carboxy-terminal end of the Green Fluorescent protein (GFP).

Transfection of plasmid cDNA into HeLa, HeLa-H2B, MDA-MB-231 or MCF7 cells was performed for 24 h using Lipofectamine Reagent 2000 (Invitrogen) as described by the manufacturer. Small interfering RNA (siRNA) was transfected at 50 nM for 72 h using Lipofectamine reagent 2000. Control siRNA duplex (non targeting pool #1), and specific MTUS1 siRNA#1 (on-target plus smart pool, NM020749) and siRNA#2 (on-target plus siRNA duplex, sens strand : UGG CAG AGG UUU AAG GUU A) were purchased from Dharmacon (Chicago, IL, USA). Silencing efficiency was measured by western blot and real time RT-PCR.

Total RNA was extracted from tumor samples by the cesium chloride protocol [28], and from cancer cell lines using Trizol reagent (Invitrogen) according to the manufacturer's instructions. First-strand cDNAs were synthesized using oligo(dT) primer with Superscript™ reverse transcriptase (Invitrogen) for 60 min at 37°C. cDNA (5 ng) was used for real-time PCR analysis using Fast Start SYBR Green I reagent (Roche) and LightCycler instrument. Each reaction was carried out in duplicate. EEF1G and PP1A cDNAs served as internal controls. Gene-specific primer pairs located in different exons (supplemental Table S1) were designed with Oligo6 software. Results are expressed relative to EEF1G values.

Immunohistochemistry was performed on 5 µm sections from formalin-fixed, paraffin-embedded tissue of the breast cancer samples. Heat-mediated antigen retrieval was performed in EDTA buffer pH 9 in water bath for 30 min. Immunostaining was performed on a Dako autostainer using a peroxidase-labeled polymer-based detection system (Envision plus, Dako) and diaminobenzidine as a chromogen. Monoclonal anti-MTUS1 antibodies (Abnova) were diluted 1∶40 and incubated overnight at 4°C. Slides were counterstained with hematoxylin. MCF7 cells transfected with GFP-ATIP3 were used as a positive control and non-transfected MCF7 cells were considered as negative control. MTUS1 immunoreactivity in tissue sections was scored based on the percentage of positive cells and the intensity (1 to 3) of the staining. Tumors were considered MTUS1-positive when more than 60% of the cells showed intense (2–3) staining, and were considered MTUS1-negative when less than 30% of the cells showed weak (0–1) immunostaining.

For analysis of xenografts, 3 µm sections were cut from formalin-fixed, paraffin-embedded tissue blocks, then counterstained with hematoxylin-eosin and examined under an inverted microscope.

Cells were seeded in quadruplicate in 96-well plates at the density of 5.10³ cells per well. Cell proliferation was analyzed by incubation for 4 h at 37°C with 1 mg/ml tetrazolium salt MTT (Sigma). Cleavage products (formazan) were solubilized with DMSO and optical density was measured at 560 nm, as recommended by the manufacturer. DNA synthesis was measured using a colorimetric immunoassay (Roche) after incorporation of 5-bromodeoxyuridine (BrdU) for 4 h at 37°C. Incorporated BrdU was quantified by optical density reading at 405 nm according to manufacturer's instructions.

For clonogenicity experiments, MCF7 cells were transfected for 24 h with 0.2 µg pEGFP-C1 vector or 2 µg of GFP-ATIP3 cDNA. Similar transfection efficiency (50%) was assessed by FACS analysis and cell viability was verified using trypan blue. Cells were plated at various dilutions in 6 well plates and transfectants were selected by adding geneticin (G418, 1 mg/ml, Gibco) in fresh medium twice a week for 3 weeks. Resistant colonies were stained with 0.5% cristal violet (Sigma) and counted.

For anchorage-independent growth assays, stable transfectants were seeded onto 6-well plates on 0.4% agar gel in appropriate medium supplemented with 20% FCS, over a bottom layer of 0.6% low-melting temperature agar gel. Cells were grown for 4 weeks and the formed colonies were counted under an inverted microscope.

Five weeks old female immunodeficient CB17/scid mice were purchased from Harlan (UK). A slow-release 17-beta-estradiol pellet (60-day release, 0.72 mg per pellet, Innovative Research of America, Florida, USA) was implanted subcutaneously into the intrascapular region of each mouse three days before breast cancer cell inoculation. MCF7 cell clones (5×10⁶ cells in 100 µl PBS) were injected subcutaneously into both flanks. All procedures were performed in accordance with institutional guidelines established by the Ministère de l'Agriculture et de la Forêt, Direction des services vétérinaires (Paris, France). Tumor diameter was measured twice weekly using a calipper along two orthogonal axes: length (L) and width (W). The volume (V) of tumors was estimated by the formula V = L×(W ²)/2. Mice were sacrificed on day 38 and tumors were excised, weighed and photographed. Tumors were fixed in 4% formaldehyde solution and paraffin-embedded for histologic analysis.

Exponentially growing cells plated on coverslips were fixed in ice-cold methanol for 5 min prior to incubation for 1 hr at room temperature with primary antibodies. SK-MES cells were incubated with human anti-alpha-tubulin antibodies clone F2C diluted 1∶10 [29] and mouse anti-MTUS1 monoclonal antibodies diluted 1∶10 (Abnova, clone 1C7). HeLa, MDA-MB-231 or MCF7 cells transiently transfected (24 hrs) with pEGFP-C1 or GFP-ATIP3 were incubated with mouse anti-alpha-tubulin monoclonal antibodies (Sigma) diluted 1∶500. After extensive washing, cells were incubated for 30 min at room temperature with appropriate secondary Cy-2-conjugated anti-human and/or Cy-3-conjugated anti-mouse antibodies (Jackson Laboratories, Interchim, France) diluted 1∶500. Coverslips were mounted on glass slides using Glycergel Mounting Medium (DakoCytomation) and examined with a Leica TCS SP2 AOBS confocal microscope using appropriate filters. Image analysis was performed with NIH ImageJ software.

For live cell imaging, images were acquired on a spinning disk microscope (one image taken every 5 min for 36 hrs following transient transfection of HeLa-mCherryH2B cells) as described [30]. Multi-dimensional acquisitions were performed in time-lapse mode using Metamorph 7.1.7 software.

Microtubule cosedimentation assay was performed as described [31] with modifications. Cells were incubated for 20 min at 4°C in PEM buffer (100 mM PIPES, pH 6.9, 1 mM MgSO4, 1 mM EGTA), scraped and sonicated prior to centrifugation at 15000 rpm for 10 min, 4°C. Clarified samples were incubated with taxol (20 µM) in the presence of GTP (1 mM) and DTT (1 mM) for 45 min at 37°C and were spun at 70 000 g for 30 min at 30°C through a cushion buffer containing 40% glycerol, 20 µM taxol and 1 mM GTP. The supernatant (S) and pellet (P) fractions were collected separately and subsequently immunoblotted with anti-GFP (Sigma) or anti-MTUS1 (Abnova) antibodies. Blots were reprobed using anti-alpha-tubulin (Sigma) antibodies.

In all studies, values are expressed as mean ± standard deviation (SD). Statistical analyses were performed by unpaired Student's t test and Tukey-Kramer's test using JMP7 software. Differences were considered statistically significant at p<0.05.

The expression of MTUS1 transcripts was analyzed in a series of 151 infiltrating ductal breast carcinomas and eleven normal breast tissues. The intensities of each of three specific MTUS1 probesets (212096_s_at; 212093_s_at; 239576_at) were obtained from our previous U133A Affymetrix DNA array study [22] and compared to histologic characteristics of the tumors and clinical data for the patients (supplemental Table S2). As shown in Fig. 1A and Table 1, MTUS1 expression levels were significantly reduced in breast cancer samples as compared to normal tissue. MTUS1 levels were significantly lower in high grade tumors (grade III) as compared to grades I (p<0.0001) and II (p = 0.0032), but did not differ significantly between tumors of grades I and II (p = 0.06). Similar results were obtained with all three MTUS1 probesets (Fig. 1A and supplemental Fig.S1). As presented in Table 1, a two- to ten-fold reduction in MTUS1 expression was observed in 47.7% of the tumors, and the number of samples underexpressing MTUS1 increased with the histological grade : 38.8%, 47.5% and 84.6% in grade I, II and III, respectively. Low levels of MTUS1 also significantly correlated with tumors that developed metastasis at distant sites, but not with the occurrence of axillary lymph node metastasis (Fig. 1B). MTUS1 levels were then compared among molecular breast tumors classified clinically according to immunohistochemical detection of the surrogate markers ER, PR and HER2. As shown in Fig. 1C, MTUS1 Affymetrix probeset intensities were significantly lower in a majority (83.3%) of triple negative (ER−/PR−/HER2−) carcinomas as compared to luminal (ER+) and HER2+ tumor subgroups (Table 1), which supports the conclusion that reduced levels of MTUS1 are associated with breast cancer aggressive subtypes.

Immunohistochemical analyses were undertaken to examine the cellular distribution and expression of MTUS1 at the protein level in situ. Immunohistochemistry performed on normal breast tissue sections using monoclonal anti-MTUS1 antibodies revealed a strong cytosolic staining in luminal epithelial cells of the breast (Fig. 1D). The same results were obtained using polyclonal anti-ATIP antibodies described previously [19] (data not shown). MTUS1 immunostaining was then analyzed in 20 breast cancer samples for which paraffin-embedded tissue sections were available (Fig. 1D). Selected tumors were representative of all histological grades (10, 3 and 7 tumors of grade I, II and III, respectively), and main molecular subtypes (12 ER+, 3 HER2+ and 3 triple negative tumors). MTUS1 immunostaining in tumor sections was scored and compared to MTUS1 mRNA expression levels according to Affymetrix probeset intensities. As presented in supplemental Table S3, tumors showing low levels of MTUS1 transcripts by DNA array and/or real-time RT-PCR analysis also displayed undetectable or weak MTUS1 staining, whereas those expressing high levels of MTUS1 mRNA were attributed a high score by immunohistochemistry. Thus, results on protein expression paralleled gene array analysis.

Results obtained by DNA microarray were further confirmed and extended by real-time RT-PCR. We selected a panel of 29 representative tumor samples whose MTUS1 Affymetrix probesets intensities ranged from low (n = 11), moderate (n = 13) to high (n = 5). Expression levels of total MTUS1 transcripts, measured by quantitative PCR using oligonucleotides located in the 3′ exons shared by all ATIP members, were found to significantly correlate (r2 = 0.77) with Affymetrix probeset intensities (Fig. 1E). In order to discriminate between different MTUS1 spliced variants, real-time RT-PCR was designed to specifically amplify ATIP1 and ATIP3 mRNAs using oligonucleotides located in their specific 5′ exons, respectively. Since ATIP4 transcripts are exclusively expressed in the central nervous system [20], these were not included in the present study. Significant correlation was observed between Affymetrix probeset intensities and expression levels of ATIP3 (r2 = 0.83) (Fig. 1F), but not ATIP1 transcripts (r2 = 0.01) (Fig. 1G), indicating that ATIP3 is the major MTUS1 splice variant whose expression is regulated in breast cancer.

As a pre-requisite to functional studies, a panel of human cancer cell lines was analyzed for MTUS1 expression by real-time RT-PCR (Supplemental Fig. S2A). Nine out of 13 breast cancer cell lines including MDA-MB-231, MCF7, Hs578T, BT549, MDA-MB-453, T-47D, ZR-75.1, MDA-MB-361 and SK-BR3 showed low to undetectable levels of MTUS1, whereas CAMA-1, ZR-75.30, BT-474 and MDA-MB-468 expressed moderate levels of MTUS1 transcripts. Highest levels of MTUS1 expression were found in the lung cancer cell line SK-MES, that was therefore selected for further study. As for primary tumors, ATIP3 expression in cancer cells paralleled that of total MTUS1 mRNA, while ATIP1 remained undetectable in every cell line examined. In agreement with our RT-PCR studies, western blot experiments using anti-MTUS1 antibodies revealed a polypeptide of 170 KDa corresponding to ATIP3 in cancer cell lines MDA-MB-468 and SK-MES, but not in MCF7 nor MDA-MB-231 (supplemental Fig.S2B).

The functional consequence of ATIP3 underexpression was investigated by transfection of specific siRNA into the ATIP3-positive MDA-MB-468 cell line. Two different sequences of siRNA successfully silenced ATIP3 expression in MDA-M468 cells both at the mRNA and protein level (Fig. 2A, left panels). As shown in Fig. 2A (right panel), ATIP3 knock-down by both siRNAs led to a significant increase (66.4±5.3% and 63±10.2% upon transfection of siRNA#1 and siRNA#2, respectively) in MDA-MB-468 breast cancer cell proliferation, pointing to an anti-proliferative effect of ATIP3.

To further confirm the inhibitory effects of ATIP3 on breast cancer cell proliferation, a GFP-ATIP3 fusion protein was generated and expressed in ATIP3-negative cells MDA-MB-231 and MCF7. Independent stable transfectants (clone 3A1 in MDA-MB-231, and clones HC1, HC6, HC7 in MCF7) expressed GFP-ATIP3 at levels similar to those of endogenous ATIP3 in MDA-MB-468 (Fig. 2B and 2D, left panel). As shown in Fig. 2B (right panels), cell proliferation measured by BrdU incorporation and MTT assay was reduced by 28±2.3% and 30±5%, respectively, in GFP-ATIP3-transfected clone 3A1 as compared to non transfected and GFP-expressing MDA-MB-231 cells. Similar results were obtained using stably transfected MCF7 clones HC1, HC6 and HC7 (data not shown).

Clonogenicity experiments were then conducted to evaluate the consequences of ATIP3 expression on cell-cell contact growth inhibition. As shown in Fig. 2C, the number of colonies resistant to neomycin were reduced by 85% upon transfection of MCF7 cells with GFP-ATIP3 as compared to GFP.

To investigate the effects of ATIP3 on anchorage-independent growth, which is a hallmark of tumorigenic properties in vitro, stably transfected breast cancer cells were allowed to grow for four weeks in soft agar. As shown in Fig. 2D, the number of colonies grown in soft agar were reduced by 75% in MCF7 cell clones transfected with GFP-ATIP3 (HC1 and HC6) as compared to GFP. Similar results were obtained using stably transfected cell clones HC7 and 3A1 (data not shown).

Stably transfected MCF7 clones (HC1, HC6 and HC7, or GFP) were injected subcutaneously into both flanks of immunodeficient mice, and tumor growth was monitored twice a week. At day 30, only 17% (6/36) of tumor xenografts developed in mice injected with GFP-ATIP3-transfected clones (HC1, HC6 or HC7), as compared with 82% (18/22) in mice injected with GFP-transfected MCF7 cells (Fig. 3A). The time-course of tumor progression (Fig. 3B) and size of the tumors (Fig. 3C, 3D) were significantly reduced in mice injected with HC1, HC6 and HC7 as compared to GFP-expressing MCF7 cells. Histological examination of the tumor nodules excised on day 38 confirmed the presence of tumor cells in each case (Fig. 3E). However a two-fold reduction in the number of cancer cells undergoing active division, assessed by microscopic examination of mitotic cells (Fig. 3F), was observed in tumors expressing GFP-ATIP3 as compared to GFP alone. Thus, restoring endogenous levels of ATIP3 expression leads to reduced breast cancer cell proliferation and delayed tumor growth in vivo.

To further characterize ATIP3 at the molecular level, its subcellular localization was analyzed in SK-MES cancer cells that express high levels of ATIP3 and no detectable ATIP1. Confocal microscopy analyses (Fig. 4A) allowed the detection of endogenous ATIP3 proteins as filamentous structures in the cytoplasm, with a staining pattern similar to that of microtubules. Incubation with anti-alpha-tubulin antibodies revealed colocalization of ATIP3 with the microtubule network during interphase. During mitosis, endogenous ATIP3 staining decorates the mitotic spindle in metaphase and anaphase, and remains strongly associated with the intercellular bridge during cytokinesis (Fig. 4A). To confirm that the ATIP3 isoform indeed colocalizes with microtubules, human cell lines were transiently transfected with the GFP-ATIP3 fusion protein and analyzed by fluorescence microscopy. As shown in Fig. 4B, GFP-ATIP3 colocalizes with alpha-tubulin at the microtubule cytoskeleton and microtubule organizing center (MTOC) in cells expressing low levels of the fusion protein. Higher levels of expression of the GFP-ATIP3 construct led to the formation of bundles which is a characteristic feature of stabilized microtubules, suggesting that ATIP3 may regulate the dynamics of microtubule polymerization. The same intracellular localization was observed upon transfection of GFP-ATIP3 into HeLa, RPE1, MDA-MB-231 and MCF7 cells (not shown). As observed for endogenous ATIP3, the GFP-ATIP3 fusion protein also localizes at the mitotic spindle during mitosis (not shown), and stains the intercellular bridge during cytokinesis (Fig. 4B).

Microtubule co-sedimentation assays were then performed on cells transfected with either GFP-ATIP3 or GFP. In this assay, microtubules are stabilized in the presence of taxol and GTP, and proteins that associate with microtubules are captured in the pellet fraction. As shown in Fig. 4C (upper panel), the whole pool of GFP-ATIP3 co-sediments with stabilized microtubules in the pellet whereas the soluble GFP protein is only detected in the supernatant fraction. Co-sedimentation of GFP-ATIP3 with microtubules was observed in both MCF7 and MDA-MB-231 cells either stably or transiently transfected with GFP-ATIP3 (data not shown).

Whether endogenous ATIP3 also co-sediments with microtubules was further evaluated in SK-MES, MDA-MB-468 and CAMA-1 cancer cells, that express high to moderate levels of ATIP3 (supplemental Fig.S2). As shown in Fig. 4C (lower panel), the whole pool of endogenous ATIP3 was detected in the microtubule pellet and no residual soluble protein was found in the supernatant. Thus, ATIP3 associates with stabilized microtubules.

The localization of ATIP3 at the mitotic spindle and intercellular bridge, together with its anti-proliferative effects in vitro and in vivo, prompted us to investigate the effects of this protein in essential steps of cell division. To this end, GFP-ATIP3 was transiently transfected into HeLa-H2B cells, that stably express a red fluorescent protein-tagged Histone H2B and constitute a useful tool to monitor mitosis by live cell imaging. Time-lapse fluorescence videomicroscopy analyses revealed the association of GFP-ATIP3 with duplicated centrosomes during prophase, and further confirmed its localization at the mitotic spindle and intercellular bridge at different steps of mitosis in living cells (Fig. 5A, supplemental video S1 and supplemental video S2). In addition, measuring the time-lapse between chromosome condensation and the end of cytokinesis revealed that cells expressing GFP-ATIP3 need a longer time to achieve mitosis (187±138 min, n = 35) as compared to non-transfected cells (81±12 min, n = 9) and cells expressing GFP (59±7 min, n = 16) (Fig. 5B). More specifically, the time spent in metaphase (from chromosome condensation in prometaphase to the beginning of anaphase) was significantly increased in GFP-ATIP3 (135.9±129.7 min, n = 33) as compared to GFP-transfected cells (19±5.6 min, n = 16) (Fig. 5C). Thus, ATIP3 overexpression delays the progression of mitosis by prolonging the step of metaphase.