Comprehensive Profiling of 8P11-12 Amplification in Breast Cancer

Comprehensive Profiling of 8p11-12 Amplification in Breast Cancer

Ve´ronique Gelsi-Boyer,1,3 Beátrice Orsetti,4 Nathalie Cervera,1 Pascal Finetti,1 Fabrice Sircoulomb,1 Carole Rouge´,4 Laurence Lasorsa,4 Anne Letessier,1 Christophe Ginestier,1 Florence Monville,1 Se´verine Esteyrie`s,1 JoseÁde´laı¨de,1 Benjamin Esterni,2 Catherine Henry,6 Stephen P. Ethier,7 Fre´de´ric Bibeau,5 Marie-Joe¨lle Mozziconacci,1,3 Emmanuelle Charafe-Jauffret,1,3 Jocelyne Jacquemier,1,3 François Bertucci,1 Daniel Birnbaum,1 Charles Theillet,4 and Max Chaffanet1

1Marseilles Cancer Institute, Department of Molecular Oncology, UMR599 Institut National de la Sante et de la Recherche Medicale and Institut Paoli-Calmettes; 2Department of Biostatistics, UMR599 and Institut Paoli-Calmettes; 3Department of BioPathology, Institut Paoli-Calmettes, Marseilles, France; 4EMI229 Institut National de la Sante et de la Recherche Medicale and 5Department of Pathology, CRLC Val d’Aurelle-Paul Lamarque, Montpellier, France; 6Laboratory of Cytogenetics, Centre Hospitalier Universitaire of Rennes, Rennes, France; and 7Karmanos Institute, Detroit, Michigan

Abstract region just telomeric to and associated with the In human carcinomas, especially breast cancer, amplification. Finally, we show that 8p11-12 chromosome arm 8p is frequently involved in amplification has a pejorative effect on survival in complex chromosomal rearrangements that combine breast cancer. (Mol Cancer Res 2005;3(12):655–67) amplification at 8p11-12, break in the 8p12-21 region, and loss of 8p21-ter. Several studies have identified Introduction putative oncogenes in the 8p11-12 amplicon. The short arm of chromosome 8 is a frequent target of genetic However, discrepancies and the lack of knowledge alterations in a wide variety of human cancers (1, 2). on the structure of this amplification lead us to think In breast cancer, the 8p12-21 region often displays complex that the actual identity of the oncogenes is not combinations of chromosomal breaks, losses, and DNA definitively established. We present here a amplification (3-11).8,9 Loss-of-heterozygosity studies have comprehensive study combining genomic, expression, identified at least two regions of loss (7, 11), whereas recurrent and chromosome break analyses of the 8p11-12 chromosome breaks are found in 6% of breast cancers at the region in breast cell lines and primary breast tumors. NRG1 locus (12, 13). Amplification of 8p11-12 occurs in 10% We show the existence of four amplicons at to 15% of breast cancers (14-17) and is often visible 8p11-12 using array comparative genomic cytogenetically as homogeneously staining region (9, 18-20). hybridization. Gene expression analysis of 123 The frequency of 8p11-12 alterations contrasts with our limited samples using DNA microarrays identified 14 genes knowledge of the genes involved, although several candidates significantly overexpressed in relation to have been suggested. No oncogene from the 8p11-12 amplifi- amplification. Using fluorescence in situ hybridization cation unit (amplicon) has been identified with certainty. The analysis on tissue microarrays, we show the role of the FGFR1 gene, which codes for a tyrosine kinase existence of a cluster of breakpoints spanning a receptor, has been evoked (17), but normal levels of FGFR1 expression in some amplified tumors have suggested that it may not be the actual or sole oncogene involved (21). Due to its importance in breast cancer, several recent studies Received 8/8/05; revised 10/26/05; accepted 11/14/05. have aimed at a better definition of the 8p11-12 amplification. Grant support: Institut National de la Sante et de la Recherche Medicale, Institut Paoli-Calmettes, Association pour la Recherche sur le Cancer (2002-2003), Ligue Ray et al. characterized the 8p11-12 amplicon in three breast Nationale Contre le Cancer (Label 2003-2006), and Ministries of Health and tumor cell lines (21). Several potential oncogenes other than Research (Cance´ropoˆles PACA and GSO); Ligue Nationale Contre le Cancer fellowship (A. Letessier) and Ministry of Research fellowship (C. Ginestier, F. FGFR1 were suggested. Reyal et al. analyzed transcriptome data Monville, F. Sircoulomb, and S. Esteyrie`s). of 130 invasive breast carcinomas using an approach based on The costs of publication of this article were defrayed in part by the payment of the calculation, for each gene, of the correlation between page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. its expression profile and that of its neighbors (22). A list of Note: V. Gelsi-Boyer and B. Orsetti have equally contributed to this work. candidate oncogenes has been established, including some Supplementary data for this article are available at Molecular Cancer Research proposed previously (21). However, this approach may have Online (http://mcr.aacrjournals.org/). Requests for reprints: Daniel Birnbaum, UMR599 Institut National de la Sante et de la Recherche Medicale, 27 Bd. Leı¨Roure, 13009 Marseilles, France. Phone: 33-4-91-75-84-07; Fax: 33-4-91-26-03-64. E-mail: [email protected] Copyright D 2005 American Association for Cancer Research. 8 http://amba.charite.de/~ksch/cghdatabase/index.htm. doi:10.1158/1541-7786.MCR-05-0128 9 http://cgap.nci.nih.gov/Chromosomes/RecurrentAberrations.

Mol Cancer Res 2005;3(12). December 2005 Downloaded from mcr.aacrjournals.org on September 27, 2021. © 2005 American Association for Cancer Research. Profiling of 8p11-12 Amplification in Breast Cancer 657

excluded isolated candidate genes and did not inform clearly on our series with 87.1% of the tumors and 100% of the cell lines the structure of the amplicon. The most extensive study of 8p showing either gains or losses. Tumors and cell lines showed amplification to date has been reported by Garcia et al. (23). similar aCGH profiles. Losses were prevalent from telomere to These authors studied 80 breast and ovarian tumors and cell lines 8p12 and gains in the 8p12-11 region. Transitions between gains by using both high-resolution array comparative genomic and losses were concentrated in a region located at the 8p12- hybridization (aCGH) and gene expression analyses. They 8p21 boundary. Individual aCGH profiles showed wavy defined a 1-Mb region of common amplification. This patterns, suggesting the existence of several peaks of amplifi- region excluded most previously proposed candidate genes, cation within this region of gain. Compilation of gains and including FGFR1, and contained four potential oncogenes, losses data from all samples further supported this idea FLJ14299, SPFH domain family, member 2, related to stomatin, (Supplementary Fig. S2A). Alignment of the regions of gains a component of lipid rafts (SPFH2/C8orf2), BRF2,and showed three recurrent interruptions and four subregions of RAB11FIP1, which showed good correlation between amplifi- amplification (Fig. 1A). The latter were designated amplicons cation and overexpression. A novel amplicon resulting from A1 to A4, from telomere to centromere. NRG1 rearrangements and associated with poor prognosis was The boundaries of the amplified subregions were positioned recently described (24). This small amplicon is included in the immediate nonamplified neighbor BAC clones. In in the 1-Mb minimal amplicon defined by Garcia et al. (23) addition, to ascertain that the subregion selected corresponded and contains two previously proposed candidate genes (22, 23). strictly to an amplification common to several tumor samples, we The four studies provided very valuable information. However, determined its core by using a stringent cutoff and considering the structure of the amplification can be refined further, only BAC clones whose log2 ratio exceeded 0.75 (Fig. 1B). especially with regard to its centromeric part and its association Amplicons and their cores were then characterized. Locations, with breaks and to the number and identity of potential BAC and gene contents, and sizes and cores of the amplicons are oncogenes it contains. detailed in Fig. 2 and Table 1. Although we defined amplicon A1 We report here a detailed genomic analysis of the 8p11-12 as a single core, aCGH profiles indicated the existence of two region in a series of breast cell lines and primary breast tumors. subpeaks, a main one centered f37.3 Mb and a second at 37.8 Using aCGH, we show that the 8p11-12 amplification can Mb, which we designated A1V(Fig. 1B). Profiles of the other be divided in four amplicons. Correlation analysis between amplicons did not show clear-cut subpeaks. According to amplification and gene expression identified several candidate National Center for Biotechnology Information Build 35 of the oncogenes. Using fluorescence in situ hybridization (FISH) Human Genome Sequence, amplicons A1 to A4 contained 17, on primary breast tumors organized in tissue microarrays, we 25, 8, and 10 genes or part of genes, respectively (Table 1; Fig. identified a cluster of breakpoints just upstream of the 2B and C). The four amplicons showed separate cores but amplification. Finally, we show the clinical significance of overlapped at their extremities: A1/A2, A2/A3, and A3/A4 8p12 amplicons in breast cancers. overlaps included 11, 2, and 3 genes, respectively (Table 1; Fig. 2B and C). The amplicons could be observed independently in Results some samples (Fig. 1A). Thus, DNA amplification at 8p11-12 aCGH Profiles Showed Multiphasic DNA Amplification presented a complex structure, suggesting multiple causal at 8p12 Associated to Loss of Distal 8p elements. We studied genomic alterations at 8p in 37 breast cell lines and 62 primary breast tumors by using a custom-made BAC Gene Expression Profiling Revealed Overexpression of array that included a near tiling-path coverage from centromere 8p11-12 Genes (43.2 Mb) to 8p21 (26.1 Mb). Examples of profiles are shown Total RNA from 34 cell lines, 89 tumors, and 4 normal in Fig. 1, and complete results for every sample analyzed are breast RNA pools were hybridized to whole-genome oligo- shown in Supplementary Fig. S1. nucleotide DNA microarrays. After filtering, we focused our Samples were not analyzed at random; several of them were analysis on 261 genes mapped to chromosome 8p, of which preselected based on either amplification or chromosomal 67 are located in 8p11-12. For each sample, gene expression rearrangement at 8p observed by either FISH or Southern levels were centered on the average levels obtained with the blotting. As a result, chromosome 8p alterations were frequent in four normal breast samples. Expression profiling data were

FIGURE 1. Genomic profiling of 8p11-12 amplification in breast cancers. A. Compilation of interpreted 8p11-12 aCGH profiles in breast cell lines and primary tumors exhibiting gains in this region. Regions were considered to be gained or lost when at least two consecutive BAC clones presented a log2 ratio > 0.25 (gain) and less than À0.25 (loss). Each square represents one BAC clone: green, loss; red, gain; open square normal, no square no data available. Clones are ordered by chromosomal position from telomere to centromere according to the May 2004 University of California at Santa Cruz freeze. Only samples presenting gains are shown. Samples are numbered as in Supplementary Table S1: breast cancer cell lines: 3, BT474; 4, BT483; 6, CAMA1; 8, HCC38; 9, HCC1187; 12, HCC1500; 13, HCC1569; 15, HCC1954; 16, HCC2218; 17, Hs578T; 19, MDA-MB-134; 21, MDA-MB-175; 24, MDA-MB-415; 28, S68; 29, SK-BR-3; 30, SUM44; 31, SUM52; 32, SUM149; 33, SUM185; primary tumors: 3, MA4420; 6, MA4814; 8, MA5103; 10, MA5357; 11, MA5633; 12, MA5731; 13, MA6137; 14, MA6165; 15, MA6795; 17, MA7371; 18, MA7499; 19, MA7641; 20, MA7767; 21, MA7809; 23, MA8189; 25, MA8404; 26, MA8406; 28, MA8525; 29, MA9377; 31, MA11934; 34, VA4380; 43, VA6052; 48, VA6190; 49, VA6204; 50, VA6219; 51, VA6270; 54, VA6582; 57, VA6660; 60, VA7106. Bold lines and numbers at the bottom of the Mb scale indicate regions shown as blowups in B. B. Detailed genomic profiles of four subregions of amplification at 8p11-12. The aCGH profiles of tumors or cell lines showing gains in the considered interval were plotted. Subregion 1, 35.52-38.44 Mb; subregion 2, 37.59-40.10 Mb; subregion 3, 38.65-41.35 Mb; subregion 4, 40.88-42.58 Mb. Y axis, log2 ratios; X axis, midpoint Mb position of each clone covering the subregion. The threshold used to define amplicons was log2 ratio > 0.75. Horizontal bars at the bottom of the plots, the region encompassed in the amplicon.

classified by hierarchical clustering (Fig. 3A). In data sets, BAG4, DDH domain containing 2 (DDHD2), HTPAP, Wolf- genes (rows) were ordered according to their chromosomal Hirschhorn syndrome candidate 1-like 1 (WHSC1L1), AP3M2, location, whereas samples (columns) were classified according IKBKB, POLB, VDAC3, SLC20A2, and LOC114926) displayed to similarity of their expression profiles. Expression data expression profile significantly correlated with the expression were color-coded: red for overexpression and green for of neighbor genes. Fig. 3B shows the diagram of transcriptome underexpression compared with normal breast samples. correlation scores (TCS) measured for each probe set, organized Expression profiles displayed contiguous red clusters containing according to chromosomal location. The distribution of the 67 genes of the 8p11-12 region from telomere-NRG1 to correlated genes was bimodal with a first peak at 37.7 to 38.3 centromere. Mb and a second peak at 42.1 to 42.5 Mb. Correlated gene Based on a published method (22), we defined a ‘‘tran- expression could be due to coamplification. Therefore, we scriptome correlation map’’ of the 8p11-12 region that shows compared the transcriptome correlation map to the location of the degree of coexpression of each gene with its neighbor genes. the four amplicons defined by aCGH (Fig. 3E). Genes whose With a stringent significance threshold (1/10.000), 16 of the 67 expression correlated with neighbors included seven genes genes (SPFH2, BRF2, RAB11FIP1, EIF4EBP1, ASH2L, LSM1, from A1 (SPFH2, BRF2, RAB11FIP1, EIF4EBP1, ASH2L,

FIGURE 2. Genetic map of 8p11-12 and outline of alterations found in breast cancer. A. Mb scale of the region and corresponding cytogenetic banding. B. Genes mapping in the interval of interest. C. Boundaries (lines) and cores (enlarged lines) of the four amplicons. D. Intervals with identified chromosomal breaks. Cell lines in which they have been characterized are indicated. Boxes, intervals in which breaks have been determined in primary tumors by FISHon tissue microarray. Blue boxes, number of primary tumors presenting breaks in the BPC2 region (defined by opened box). BPC1 region spanning NRG1 was reported previously (12, 13). E. Numbers in boxes, frequency of amplification observed in a given interval in tumors by FISH on tissue microarray. F. BAC clones used as FISH probes on either metaphases or tissue microarray. Pools of BAC clones used as probes in FISH on tissue microarray and regions covered are indicated as P1 to P6 (for more details, see Supplementary Table S4). Map and gene names were taken from the Build 35 from National Center for Biotechnology Information.

Table 1. Description of Boundaries, BAC Clone Coverage, and Gene Content for A1 to A4 Amplicons, Respectively

Amplicon A1 A2 A3 A4

Chromosome location (Mb)* Start End Start End Start End Start End 35.97 38.23 37.78 39.99 39.84 41.63 41.41 42.33 36.51 37.78 38.09 38.89 39.85 41.10 41.58 42.04 Size (Mb)* Whole: 2.25 Whole: 2.21 Whole: 1.79 Whole: 0.92 Core: 1.27 Core: 0.8 Core: 1.25 Core: 0.46 Size overlap (Mb) A1/A2: 0.45 A2/A3: 0.15 A3/A4: 0.22 Distance between cores (Mb)* A1-A2: 0.31 A2-A3: 0.96 A3-A4: 0.48 BAC clones* RP11-89M20, RP11-210F15, CTB-22K20, RP11-90P5, RP11-44K6, RP11-262I23, RP11-360L9, RP11-26K8, RP11-197P20, RP11-513D5, RP11-265K5, RP11-133O7, RP11-470M17, RP11-46C23, RP11-125F4, CTD-2118N15, CTB-22K20, RP11-350N15, RP11-100B16, CTD-3082P9, RP11-104D16, RP11-45I11, RP11-108L9, RP11-90P5 RP11-359P11, RP11-495O10, RP11-51K12, RP11-48D21, RP11-282J24, RP11-313J18, RP11-723D22, RP11-648J12, RP11-284J3, RP11-465K16, RP11-231D20 RP11-44K6 RP11-681J6, RP11-360L9 Genes FKSG2, KCNU1, LOC157860, 3Vpart of GPR124, BRF2, INDO, LOC169355, GOLGA7, SLD5, FLJ14299, SPFH2, PROSC, RAB11FIP1, MGC33309, C8orf4, ZMAT4, SFRP1, DKFZp586M1819, FLJ25169, GPR124, BRF2, RAB11FIP1, ADRB3, EIF4EBP1, ASH2L, GOLGA7, SLD5, ANK1, MYST3, AP3M2, MGC33309, ADRB3, EIF4EBP1, STAR, LSM1, BAG4, DDHD2, DKFZp586M1819 PLAT, IKBK 5Vpart of POLB ASH2L, STAR, LSM1, BAG4, HTPAP, WHSC1L1, LETM2, 5Vpart of DDHD2 FGFR1, FLJ43582, TACC1, HTRA4, BLP1, ADAM9, ADAM32, ADAM18, ADAM2 INDO,5Vpart of LOC169355 Genes included in cores FKSG2, KCNU1, LOC157860, 3Vpart of ASH2L, STAR, INDO, LOC169355, C8orf4, 3Vpart of DKFZp586M1819, FLJ14299, SPFH2, PROSC, LSM1, BAG4, DDHD2, ZMAT4 FLJ25169, ANK1, MYST3 5Vpart of GPR124 HTPAP, WHSC1L1, LETM2, FGFR1, FLJ43582, TACC1

*Information written in bold concern the cores of amplicons.

LSM1, and BAG4) and nine genes from A2 (BRF2, RAB11FIP1, Chromosome Breaks Are Associated with 8p EIF4EBP1, ASH2L, LSM1, BAG4, DDHD2, phosphatidic acid Amplification phosphatase type 2 domain containing 1B (HTPAP/PPAPDC1B), Typical aCGH profiles showed sharp transitions at the and WHSC1L1). Six were common to both amplicons A1 and telomeric end of the amplified region. In most cases, these A2. Whereas A3 did not contain any correlated gene, A4 transitions lead from gain at 8p11-12 to loss of 8p21-ter and included AP3M2 and IKBKB. Other genes with significant TCS were comprised in an interval between 35 and 37 Mb were located centromeric to A4 (POLB, VDAC3, SLC20A2, and (Supplementary Figs. S1 and S2A). In some cases, illustrated LOC114926). by the SUM52 profile (Supplementary Fig. S2B), transitions were found at 32 Mb, colocalizing with a breakpoint cluster Correlation between Gene Expression and Amplification identified in the NRG1 locus (12, 13). We wondered whether Identified 14 Candidate Genes the sharp transitions at 35 to 37 Mb were associated with To identify genes whose expression levels were significantly chromosomal breaks. We used dual-color FISH analysis with modified in relation to DNA copy number changes, a total of probes covering the region between NRG1 (not included) and 42 samples (25 cell lines and 17 tumors) common to aCGH FGFR1 (Fig. 2F; Supplementary Table S4) in a series of 12 cell and RNA expression experiments were analyzed (Supplemen- lines (Set 1, Supplementary Table S1) selected based on either tary Table S1). Among them, 29 samples did not present the existence of transition in their aCGH profile or 8p11-21 amplification at any of the four amplicons and 13 exhibited A1, rearrangement detected previously by mbanding FISH (25). A2, A3, and/or A4 amplification. Samples presenting <33% Results are shown in Fig. 4 and summarized in Supplemen- of BAC clones with a log2 ratio exceeding 0.75 within the tary Table S5. Eight cell lines (MDA-MB-134, BT-20, SUM44, amplicon were not considered as amplified. HCC-1500, CAMA-1, BT-483, S68, and SUM225) showed We applied two statistical tests (discriminating score breakpoints (Fig. 4B and C) in a 4-Mb interval delimited by and Mann-Whitney) to identify genes whose pattern of BAC clones RP11-91P13 and RP11-265K5, which spanned the overexpression correlated with DNA amplification (Fig. 3C telomeric region of UNC5D to the 3Vend of FGFR1. Break- and D). Of the 67 8p11-12 genes, 24 were selected by either points involved UNC5D (35.5 Mb) in SUM44, a region one or both methods. Fourteen were selected by both tests and comprising six genes (MGC33309, ADRB3, EIF4EP, ASH2L, considered as the best candidates. FLJ14299, SPFH2, proline STAR,andLSM1; 37.7-38.1 Mb) in S68, and a region synthetase cotranscribed homologue (PROSC), BRF2, and comprising BAG4, DDHD2, HTPAP, and WHSC1L1 (38.1- RAB11FIP1 correlated with A1 amplification, whereas LSM1, 38.3 Mb) in SUM225 (Fig. 2D). DDHD2, HTPAP, WHSC1L1, FGFR1,andTM2 domain In agreement with aCGH profiles, chromosomal breaks containing 2 (BLP1/TM2D2) correlated with A2. The only defined by FISH were associated to the loss of 8p telomeric gene correlating with A3 was GOLGA7, and two genes, portions and concomitant amplification of sequences immedi- MYST3 and AP2M3, correlated with A4 (Fig. 3E). ately centromeric to these breakpoints. Gain or amplification

was achieved by either multiplication of chromosome deriva- DNA amplification were determined by the split-signal FISH tives (BT-483, CAMA-1, HCC-1500, SUM44, and SUM225) approach. Intact copies of the region were seen as a pair of or homogeneously staining region formation (MDA-MB-134 adjacent green and red signals, whereas a break resulted in split and S68; Fig. 4B and C). The data indicate the existence of two signals. A break associated with amplification was seen as an breakpoint clusters at 8p12-21 in breast cancer, one at NRG1 increased number of red signals without green signal coloca- (designated BPC1) and a second more proximal (designated lization (Fig. 4D). Of 161 interpretable results, 31 (19%) tumors BPC2). showed either a break and/or an amplification in the UNC5D- To better define BPC2, we studied primary tumors arrayed FGFR1 region. We studied TMA2 with two BAC combinations in TMA1 by FISH. We screened the region between UNC5D (P1 + P2 and P2 + P3) to refine the presence of alterations in and FGFR1 (34.84-38.65 Mb) using BAC pools as probes UNC5D and UNC5D-FKSG2 genomic region, respectively. (Fig. 2G; Supplementary Table S4). Chromosomal break and/or Breaks in the UNC5D-FGFR1 region occurred always in

FIGURE 3. Gene expression analysis of the 8p region and correlation with amplification. A total of 34 breast cancer cell lines, 89 primary tumors (123 samples), and 4 normal breast tissue RNA pools were hybridized to whole-genome Affymetrix U133 Plus 2.0 human oligoarrays. Genes located in 8p were filtered and their expression patterns were studied. Expression levels were centered on the average level obtained with normal breast RNA. Expression profiles were studied by hierarchical clustering. A. Zoom showing expression patterns of the 67 genes located between NRG1 and the centromere (8p11-12). Each row represents a gene; each column represents a sample. Genes are ordered according to their Mb position (http://genome.ucsc.edu/). Log2- transformed expression level of each gene in a single sample is relative to its abundance across normal breast sample and is depicted according to the color scale (bottom). The dendrogram of samples (above matrix) represents overall similarities in gene expression profiles. B. Correlation transcriptome map of 8p11-12 between NRG1 and the centromere in 123 breast samples. The TCS evaluates the correlation between the expression profile of each gene and those of its neighbors. The diagram shows the TCS for each gene: dot, a gene; orange line, significant threshold (equal to 0.33) for the TCS; a gene with a score exceeding the threshold is significantly correlated with its neighbors. C. Discriminating score of the 11 genes identified by supervised analysis as significantly differentially expressed between samples with amplification and samples without amplification within each of the four amplicons (A1, A2, A3, or A4). A total of 42 samples (25 cell lines and 17 tumors) common to aCGH and RNA expression experiments were analyzed: 29 samples did not present any amplification and 13 exhibited A1, A2, A3, and/or A4 amplification. Bars, genes whose expression correlates with the amplification of one amplicon. Results were color-coded according to which amplicon correlated with increased expression: green, A1; blue, A2; purple, A3; yellow, A4. The significance thresholds producing <0.1 false positive are 0.65, 0.62, and 0.77, respectively. D. Mann-Whitney test of the same data set as discriminating score. Color codes are as in discriminating score. E. Chromosomal location and limits of the four amplicons (A1-A4 from NRG1 to the centromere) as defined by aCGH. Genes significant in both discriminating score and Mann-Whitney tests are indicated in color. Loci with asterisks are not listed in Genome Browser but can be accessed in Entrez Gene.

FIGURE 4. FISH characterization of breakpoints in 8p12 region. A. Localization of BPC2 on chromosome 8. B and C. Dual-color FISH analysis of cell lines using combinations of BAC probes. BPC2 affects chromosome 8 derivatives present in SUM44 (B) and BT-20, CAMA-1, MDA-MB-134, and S68 cell lines (C). Metaphase chromosomes of SUM44 were successively hybridized with RP11-595O22 and RP11-3P9 BAC clones (red) spanning UNC5D, combined with a centromeric probe (green). RP11-595O22 hybridizes solely to normal chromosome 8 (left), whereas red signals from RP11-3P9 are observed on both, normal, and der(8) chromosomes (right), indicating that the breakpoint is located in an interval delimited by RP11-595O22 and RP11-3P9 within the large intron 1 of UNC5D. To detect breakpoints of derivative chromosome 8 in BT-20, CAMA-1, MDA-MB-134, and S68, BAC clones located between NRG1 and FGFR1 (green) were used in combination with RP11-350N15 (red; this clone contains sequence including FGFR1 and is the most proximal BAC used). The top and bottom of FISH images correspond to results obtained with BAC clones located in the proximal and distal regions of the corresponding breakpoints, respectively. All breaks lead to the lack of 8p12-pter chromosome. D. FISH analysis of sections of tumors included in tissue microarrays. Different probe combinations were used to refine 8p12 amplification and/or breaks occurring in the region between UNC5D and FGFR1 (34.84- 38.65 Mb). Three types of patterns are illustrated: (1) tumors without break [right; FISH result obtained with P2 (green) + P5 (red) probe combination], (2) tumors with amplification [middle; FISH result obtained with P2 (green) + P5 (red) probe combination], and (3) tumors with break associated with amplification of the proximal region [left; FISH result obtained with P1 (green) + P2 (red) probe combination]. Schematic representation of signals for interpretation of results is shown below.

combination with amplification at 8p11-12 (six tumors with A1 and A3 separately because the number of samples included both break and amplification), whereas amplification was in these classes was too small. A correlation was observed with observed without break in eight tumors (Fig. 4D, middle and histologic grade and Ki-67 proliferation index. No correlation left; Supplementary Table S6). was found with the other tested factors, including age, histologic type, size of the tumor, vascular or lymph node Amplification Has Effect on Survival invasion, type of treatment, estrogen receptor, ERBB2, FGFR1, To determine the incidence of 8p11-12 amplicons in breast progesterone receptor, or p53 protein expression. Five-year cancer, we studied their occurrence by FISH on TMA1 using metastasis-free survival was significantly different between BAC pools covering A1 to A3 regions as probes (P4-P6 pools tumors without amplification and tumors with amplification at À3 in Fig. 2). Amplification at 8p was found in 13.2%, 9.4%, any of the three amplicons tested (P = 2.96 Â 10 ; Fig. 5). À5 and 11.4% of the tumors for amplicon A1 to A3, respectively Metastasis-free survival was also different (P = 1.59 Â 10 , (Table 2; Supplementary Table S6; Supplementary Fig. S3). We log-rank test) between tumors without amplification and tumors next assessed the effect of 8p amplification on disease features with amplification detected by the P5 probe. This suggests that and outcome. The presence of amplification at any of three amplification could be involved in pejorative disease evolution. amplicons (A1-A3) was correlated to histoclinical features Similar analysis was done for overall survival and disease-free (Table 2; Fig. 5). It was not possible to study amplification at survival (Supplementary Figs. S4A and S4B, respectively). Whereas for overall survival there was a trend for 8p11-12 amplification status and clinical outcome association (P = 0.236 and 0.0591), only disease-free survival was different between Table 2. Correlations between 8p11-12 Amplification tumors without amplification and tumors with amplification Determined by FISH and Histoclinical Features at any of the three amplicons tested (P = 2.29 Â 10À2)as c No amplification,* n (%) Amplification, n (%) P well as between tumors without amplification and tumors with amplification detected by the P5 probe (P = 8.15 Â 10À3, Age log-rank test). <50 39 (29) 18 (35) NS >50 96 (71) 34 (65) Histologic type Ductal 106 (79) 39 (75) NS Discussion Lobular 14 (10) 3 (6) Chromosome 8 is affected by frequent and complex Other 15 (11) 10 (19) Pathologic tumor size alterations in many cancers (1, 2). What occurs on the short pT1 60 (44) 23 (44) NS arm and specifically in region p11-21 is part of this genomic pT2 57 (42) 24 (46) ‘‘turbulence’’ (26). We have here characterized in detail pT3 18 (14) 5 (10) SBR grade 8p11-12 alterations in breast cancer using a combination of 1 41 (30) 8 (15) <0.05 aCGH, gene expression profiling, and FISH analyses. Genomic 2 58 (43) 21 (40) 3 36 (27) 23 (45) alterations consisted of amplification and chromosome breaks, Peritumoral vascular invasion which overall were found in f10% of breast tumors. We Absent 81 (60) 33 (63) NS identified four amplicons and two breakpoint clusters. Present 54 (40) 19 (37) Axillary lymph node status Negative 72 (54) 27 (52) NS Amplicons and Potential Oncogenes Positive 62 (46) 25 (48) Chemotherapy Genomic profiles obtained by aCGH showed four sub- Yes 65 (48) 26 (50) NS regions of amplification at 8p11-12. Expression analysis by No 70 (52) 26 (50) DNA microarrays was coherent with the genomic profiles. Hormone therapy Yes 28 (54) 54 (41) NS Concordance between high-level amplification and increased No 24 (46) 76 (59) gene expression has been shown by previous analyses (27, 28). ERBB2 0-1 109 (85) 40 (87) NS However, TCS analysis was not useful either to understand the 2-3 19 (15) 6 (13) structural complexity of amplicons or to determine the best FGFR1 candidate oncogenes. Negative 35 (32) 17 (39) NS Positive 76 (68) 27 (61) In subregion A1, genes with strong expression correlated Ki-67 with amplification were FLJ14233 (which codes for a protein <20 97 (80) 29 (62) <0.05 containing zing finger motifs), SPFH2, PROSC, BRF2 >20 24 (20) 18 (38) p53 (encoding a RNA polymerase III transcription initiation factor), Negative 93 (71) 39 (76) NS and RAB11FIP1 (a regulator of RAB GTPases). In contrast, Positive 38 (29) 12 (24) Estrogen receptor GPR124, which encodes a G protein-coupled receptor, could Negative 29 (22) 12 (24) NS be a tumor suppressor gene (29). A1 comprised the 1-Mb Positive 103 (78) 39 (76) minimal amplicon defined by Garcia et al. (23), with Progesterone receptor Negative 44 (34) 14 (29) NS FLJ14299, SPFH2, BRF2, and RAB11FIP1 as candidate onco- Positive 87 (76) 35 (71) genes, and the amplicon defined by Prentice et al. (24), which contains FLJ14299 and SPFH2. NOTE: P4 + P5 + P6 probes. SBR, Scarf-Bloom-Richardson, histologic grade. *P4 and P5 and P6 show no amplification. Selected genes in A2 code for WHSC1L1 and signaling cP4 and/or P5 and/or P6 show amplification. regulators DDHD2, BLP1, and PPAPDC1P. These proteins

one of which could be the selective elimination of TSG, such as SFRP1. Subregion A4 was associated to MYST3 and AP3M2. MYST3 encodes a histone acetyltransferase that becomes oncogenic in a subset of acute myeloid leukemias following chromosomal translocation (36). Golgi adaptor AP3M2 and golgin GOLGA7 may be both involved in vesicular transport. Thus, the 8p amplification contains many potentially relevant genes. For most of them, the function is poorly defined and a potential role in cancer has not been documented. Some of these genes may be neutral passengers without any role in oncogenesis (37). Alternatively, the reason to find many genes in a complex amplification unit could be that not a single one but only the combined effect of several loci provides a selective advantage to the cancer cell. Different genes of the amplification may even participate in the same pathway or cell process. Most of the proteins encoded by the 8p candidate FIGURE 5. 8p11-12 amplification and disease outcome. Kaplan-Meier curves illustrating metastasis-free survival (MFS) of 319 patients accord- oncogenes can be grouped into two categories. SPFH2, ing to 8p11-12 amplification detected at amplicon A1 to A3 by FISH RAB11FIP1, DDHD2, PPAPDC1B, TM2D2, GOLGA7, and on TMA1 with P4 to P6 pool probes, respectively (see map of Fig. 2). AP3M2 might all be involved in vesicular protein transport and P is calculated using the log-rank test. Metastasis-free survival is different between patients with and without 8p11-12 amplification (at any of membrane trafficking between the Golgi apparatus and the cell the three tested amplicons; P = 2.96 Â 10À3; orange curves) as well surface. This process might be associated with chlathrin-coated as between patients with and without A2 amplification (P = 1.59 Â 10À5; blue curves). pits and increased in the tumor cells, perhaps as a consequence of hypoxia. The RAB25 gene, which is related to the RAB11 family and regulates exocyst function, is amplified in about half breast cancers (38). The second category includes trans- may play a role in cell proliferation and survival and could cription modulators FLJ14299/ZNF703, BRF2, WHSC1L1, act as oncogenes (30, 31). However, their function is often not and MYST3. well characterized and their role will have to be further shown. Rather than a single gene, a functional module could be DDHD2 and HTPAP phospholipid enzymes may act in the selected in the 8p amplification. The structure of the same metabolic pathways. LSM1/CaSM is a potential oncogene amplification may thus be slightly different from one tumor (32). Recent data have suggested that FGFR1 is not the driving to another at the edges of each subregion (e.g., may or may not oncogene of the 8p11-12 amplicon (21, 23). Its role in the include FGFR1) depending on whether the module is complete A2 amplicon is indeed debated, because it is not always or not. A search for the absolute driver of the 8p11-12 overexpressed when amplified and not always contained in the amplification may be elusive, but targeting a single protein amplification (21). However, even if FGFR1 is not the sole or might prove efficient. Rather than aiming at determining which actual selected driver gene of an amplification unit, it might be of the amplified genes is the best candidate driver, further useful to systematically measure FGFR1 gene amplification functional studies could aim at selecting what would be the best and FGFR1 protein overexpression. Because it is a kinase therapeutic target and to which molecular subtype targeting and a transmembrane protein, FGFR1 could be used for should be applied. Our list of genes and that of other studies targeted therapy with either anti-kinases or antibodies. Anti- (21-24) are good starting points (see Supplementary Table S7). FGFR1 kinase drug PD185070 does not inhibit the proliferation of amplified SUM cell lines (21). However, in our hands, Chromosome Breaks and Potential Tumor Suppressors anti-FGFR1 kinase SU5402 inhibits the growth of other Two regions of recurrent breaks were identified on 8p: amplified cell lines, which may have come to depend on BPC1, located in the NRG1 locus (13), and BCP2, centromeric FGFR1 for proliferation.10 Breast tumor patients with FGFR1 to NRG1. In tumors, BPC2 corresponded to a 2-Mb region protein overexpression might be one day benefit from anti- from UNC5D to FGFR1. In agreement with recent data (24), FGFR1 targeted therapy. The TACC1 gene did not meet criteria we found that breaks in NRG1 can occur independently of for being a potential oncogene, which is in agreement with our amplification, but the two types of alterations are often linked. previous studies (33). BPC2 is located just telomeric to amplicon A1. Breaks at BPC2 Subregion A3 is framed by two down-regulated genes, always occurred in combination with amplification at 8p11-12. indoleamine-pyrrole 2,3-dioxygenase (INDO)andsecreted Amplifications were sometimes found in the absence of break frizzled-related protein 1 (SFRP1). SFRP1, a regulator of at BPC2. Concomitant break and amplification are likely to the WNT pathway, is a potential TSG in breast cancer (7, 17, result from the same ‘‘break-fusion-bridge’’ mechanism (39). 34, 35). The role of INDO is less clear. Amplification containing Amplification is likely to be the main oncogenic determinant, multiple amplicons may be formed by different mechanisms, but some regions may be more frequently affected by breaks than others, because they are susceptible to breakage and/or are selected because concomitant alteration provides an additional advantage. The existence of breakpoints at 8p12 associated to 10 Gelsi-Boyer et al., unpublished observations. loss-of-heterozygosity suggests that this chromosomal region is

highly prone to breakage (11). The discovery of homozygous breast tissue samples pooled in 4 RNA samples (1 sample from deletions at D8S87, within intron 1 of the UNC5D gene, in 4 women from Val d’Aurelle and 3 commercial pools of 1, 2, prostate cancer gives additional support to this idea (40, 41). and 4 normal breast RNA, respectively; Clontech, Palo Alto, Some of the BPC2 breaks target the UNC5D gene itself. CA). Samples for tissue microarrays were in two series: TMA1 This gene, which is expressed in the normal mammary gland comprised 547 nonmetastatic invasive carcinomas described (42), codes for a netrin receptor (see ref. 43 for review) and previously (Supplementary Table S2; ref. 48). TMA2 contained could be a tumor suppressor (44). The clustering of breakpoints 23 cases selected for 8p alterations, including 14 cases of suggests that the 8p12 region is sensitive to breakages and 31 altered in the UNC5D-FGFR1 region and 9 cases with a could play a general role in oncogenesis via inactivation of documented NRG1 breakpoint (13). one or several potential TSG. Both genomic profiles and FISH analysis showed that regions distal to breakpoints were Genomic Arrays and aCGH Conditions lost. Loss of 8p21-ter sequences with several other potential Custom-made genomic arrays were constructed as described TSG (e.g., DBC2; ref. 45) is also likely to be important for previously (49). For 8p, 184 genomic clones (BAC and PAC) oncogenesis. were selected, resulting in an average density of 1 clone/240 kb In conclusion, we have defined the structure of the 8p11-12 and an increased density of 1 clone/140 kb in the 8p11-12 amplification, which consists of four amplicons associated with region. Most clones were from CHORI12 (Oakland, CA), with a telomeric breaks. The amplicons include 14 candidate onco- subset of 72 genomic clones selected to enrich the 8p11.1-21.3 genes and have clinical significance in breast cancers. region, from CTB and CTD libraries obtained from Research Genetics (Huntsville, AL). Clones were ordered according to the May 2004 release of the human draft sequences.13 Materials and Methods Detailed information is compiled in Supplementary Table S3. Breast Cell Lines Probe labeling was done as described previously (49). Hybri- A total of 47 breast cell lines were studied (Set 1, Sup- dization and further processing of the arrays were done using a plementary Table S1): BrCa-MZ-01 (46), BT-20, BT-474, HS4800 hybridization station (Tecan, Lyon, France). Compre- BT-483, CAL51, CAMA-1, EFM19, HCC38, HCC202, hensive survey of genomic imbalances at 8p was done for 37 HCC1187, HCC1395, HCC1428, HCC1500, HCC1569, breast cell lines and 62 primary breast tumors. Log2 thresholds HCC1937, HCC1954, HCC2218, HME1, Hs578T, MCF-7 for gains and losses were set at 0.25 and À0.25 as described Rich,MCF-7,MCF-10A,MDA-MB-134,MDA-MB-157, previously (49). MDA-MB-175, MDA-MB-231, MDA-MB-361, MDA-MB- 415, MDA-MB-435, MDA-MB-436, MDA-MB-453, SK-BR- Purification of Nucleic Acids 3, SK-BR-7, T-47D, UACC-812, ZR-75-1, ZR-75-30 (American Genomic DNA was isolated as described previously (14). Type Culture Collection, Manassas, VA), 184B5, SUM44, Total RNA was extracted from frozen samples as described SUM52, SUM149, SUM159, SUM185, SUM190, SUM206, previously (14) and integrity was controlled by denaturing and SUM225 (47)11 and S68, a newly established breast tumor formaldehyde agarose gel electrophoresis and by microanalysis cell line (a kind gift of V. Catros, Rennes, France). All cell lines (Agilent Bioanalyzer, Palo Alto, CA). are derived from carcinomas, except MCF-10A, which is derived from atypical ductal hyperplasia, and HME1 and 184B5, which originate from normal mammary gland. The cells were grown RNA Expression Profiling of Chromosome 8 Using DNA using the recommended culture conditions. Microarrays Gene expression profiling was done with Affymetrix U133 Plus 2.0 human oligonucleotide microarrays as recommended Primary Tumors by the supplier.14 Briefly, for each sample, synthesis of the first- A total of 134 primary breast carcinomas were analyzed strand cDNA was done from 3 Ag total RNA by T7-oligo(dT) using aCGH, DNA microarrays, and/or FISH methods (Sets 2 priming followed by second-strand cDNA synthesis. After and 3, Supplementary Table S1). Of 62 primary breast purification of cDNA, an in vitro transcription combined carcinomas studied by aCGH, 31 were collected at the Institut with amplification of cRNA was used to generate the cRNA Paoli-Calmettes (Marseilles, France) and 31 at the Centre Val containing biotinylated pseudouridine, which was then purified, d’Aurelle (Montpellier, France; Sets 2 and 3, Supplementary quantified, and chemically fragmented at 95jC for 35 minutes. Table S1). Among these, 19 were selected based on FGFR1 Fragmented biotinylated cRNA was hybridized in 200 AL gene amplification status established by Southern blot (12 hybridization buffer at 45jC for 16 hours to microarrays amplified and 7 nonamplified) as positive controls of contained >47,000 transcripts and variants, including 38,500 amplification (12), 6 were selected based on NRG1 break well-characterized human genes. Automated washes of micro- (13), and 12 for alterations in the UNC5D-FGFR1 region arrays and staining with streptavidin-phycoerythrin were done observed by FISH in this study. Profiled breast samples according to the manufacturer’s instructions. Double signal included 123 samples (34 cell lines and 89 cancer tissue samples collected at Institut Paoli-Calmettes) and 11 normal

12 http://bacpac.chori.org/. 13 http://genome.ucsc.edu/. 11 http://www.cancer.med.umich.edu/breast_cell/clines/clines.html. 14 http://www.Affymetrix.com.

amplification was done by biotinylated anti-streptavidin For each gene located on the 8p11-12 region, we also antibody with goat IgG blocking antibody. Scanning was done measured the frequency of RNA overexpression (ratio superior with Affymetrix GeneArray scanner and signals were quanti- to 2) and underexpression (ratio inferior to 0.5) in the 123 fied using Affymetrix GCOS software. samples (34 cell lines and 89 tumors) compared with the All hybridization images were inspected for artifacts. average expression in the 4 normal breast RNA samples. Expression data were then analyzed by the Robust Multichip Supervised analysis was applied to identify genes that Average method in R using Bioconductor and associated discriminate between samples with and without DNA packages (50). Robust Multichip Average did the background amplification within each of our defined amplicons (A1, adjustment, the quantile normalization, and finally the A2, A3, or A4). The selection of samples was done based on summarization of 11 oligonucleotides per gene. Filtering the aCGH profiles. Samples presenting <33% of the BAC process removed from analysis the genes with low and clones with a log2 ratio exceeding 0.75 in the studied poorly measured expression as defined by expression level amplicon were not considered as amplified. For each gene of inferior to 100 units in all samples. A second filter was then the 8p11-12 region, we computed a discriminating score by applied to exclude genes showing low expression variation comparing the expression levels between the subgroup of across the samples as defined by Max-Min difference inferior samples presenting gene amplification in one of the four to 100 units for genes with Min expression value superior to amplicons and the subgroup of samples with gene amplifi- 100 and by Max-100 difference inferior to 100 units for cation within one or more of the other three amplicons genes with Min expression value inferior to 100. All data and samples without any amplification. Discriminating score were then log2 transformed for display and analysis. Before (52) was defined as discriminating score = (M1 À M2) / analysis and for each sample, gene expression levels were (S1 + S2), where M1 and S1 represent mean and SD of centered on the average levels obtained with normal breast expression levels of the gene in subgroup 1, respectively, and samples. M2 and S2 in subgroup 2. Confidence levels were estimated Oligonucleotide probe sets were mapped to the human by 2,000 iterative random permutations of samples as genome according to the Build 35 from the National Center described previously (53) with a significance threshold for Biotechnology Information Ensemble database and the producing <0.1 false positive. We applied the Mann-Whitney University of California at Santa Cruz Genome Bioinformatics test to identify significant differences in expression associated database Genome Browser.13 Of all probe sets represented on to the occurrence of amplification at either of the defined the microarrays, 1,423 were assigned to chromosome 8. After amplicons (A1-A4). The nonparametric nature of this test filtering of data, we retained for analysis 561 that focused on seemed well adapted to the nonnormal distribution of chromosome 8. Oligonucleotides assigned to 8p BAC arrays microarray data and was therefore a good complement to are listed in Supplementary Table S3 in agreement to the May the discriminating score approach. 2004 human draft sequences (40).13,14 To eliminate the redundancy of probe sets representing the same gene present Tissue Microarrays on microarrays, we selected for each gene the probe set that Tissue microarrays were prepared as described (54). presented the highest correlation with the median profile Sections (5 Am) of the resulting microarray block were made obtained with all corresponding probe sets. Therefore, we and used for FISH or immunohistochemistry analysis after retained 261 probe sets (genes) on 8p and 67 in the region transfer onto glass slides. contained between NRG1 and the centromere. Hierarchical clustering of expression data was done with the Cluster program (51) using Pearson correlation as similarity metric FISH Analysis on Breast Tumor Cell Lines and centroid linkage clustering. Results were displayed using To characterize breaks between NRG1 and FGFR1 loci, the TreeView program (51). dual-color FISH analysis was done as described previously (12), on metaphase chromosomes from 12 cell lines (Set 1, Identification of Deregulated Gene Sets Supplementary Table S1), using as probes a combination of To evaluate the correlation between the expression profile labeled BAC clones. Fixation and preparation of metaphase of each gene and that of its neighbors, we used the same spreads for FISH analysis were done on cytogenetic approach as Reyal et al. (22) focused on the 261 probe sets pellets of cultured cell lines as published (55). BAC clones (genes) located on 8p. For each probe set (gene), we selected to cover 8p12 are listed in Supplementary Table S4. computed a TCS. This score is the average of the correlation BAC clones were checked for chimerism by FISH on coefficients across all samples between RNA expression levels normal metaphases chromosomes. After counterstaining with V of this gene and the RNA levels of each of the physically Vectashield containing 4 ,6-diamidino-2-phenylindole (Vector, nearest 20 genes (10 centromeric genes and 10 telomeric Burlingame, CA), images were processed as described genes). A significance threshold for TCS was obtained by previously (25). permutation tests in using the 1,000th quantile of the random distribution (i.e., the value for which 1 of 1,000 probe sets in FISH Analysis on Tissue Microarray the random data sets was above this value). This significant Dual-color FISH analysis was directly done on tumor threshold was equal to 0.33. A probe set with a score exceeding samples arrayed in tissue microarray as published (13, 56). We the threshold was considered as significantly correlated with its used a combination of two differently labeled (biotinylated and neighbors. revealed in green, FITC; digoxigenin-labeled and revealed in

red, rhodamine) DNA from BAC pools P1 (34.8-35.4 Mb), P2 hybridization analysis of 38 breast cancer cell lines: a basis for interpreting complementary DNA microarray data. Cancer Res 2000;60:4519 – 25. (35.8-36.6 Mb), P3 (36.6-37.0 Mb), P4 (37.6-38.0 Mb), P5 6. Rummukainen J, Kytola S, Karhu R, Farnebo F, Larsson C, Isola JJ. (38.1-38,6 Mb), and P6 (40.5-41 Mb; Supplementary Table S4). Aberrations of chromosome 8 in 16 breast cancer cell lines by comparative BAC positions were ascertained by metaphase FISH. genomic hybridization, fluorescence in situ hybridization, and spectral karyo- Break characterization was based on the split-signal FISH typing. Cancer Genet Cytogenet 2001;126:1 – 7. approach (25, 57). Tumor sections present on tissue microarray 7. Charafe-Jauffret E, Moulin JF, Ginestier C, et al. Loss of heterozygosity at microsatellite markers from region p11-21 of chromosome 8 in microdissected were first hybridized with biotinylated and digoxigenin-labeled breast tumor but not in peritumoral cells. Int J Oncol 2002;21:989 – 96. BAC combinations P1 + P5 to determine the presence of 8. Struski S, Doco-Fenzy M, Cornillet-Lefebvre P. Compilation of published alterations (breaks and/or amplification) in the [UNC5D- comparative genomic hybridization studies. Cancer Genet Cytogenet 2002;135: FGFR1] genomic region. P1 + P2, P2 + P5, and P2 + P3 63 – 90. were then used to refine the alterations in [UNC5D], [UNC5D- 9. Teixeira MR, Pandis N, Heim S. Cytogenetic clues to breast carcinogenesis. Genes Chromosomes Cancer 2002;33:1 – 16. FGFR1], and [UNC5D-FKSG2](UNC5D excluded from these 10. Ferti AD, Stamouli MI, Panani AD, Ferti AD, Raptis SA, Young BD. intervals) genomic regions, respectively. Amplification gener- Molecular cytogenetic analysis of breast cancer: a combined multicolour ated in subregions A1 to A3 was studied by dual-color FISH fluorescence in situ hybridisation and G-banding study of uncultured tumour using P4 to P6 probes, respectively. Counterstaining treatment cells. Cancer Genet Cytogenet 2004;149:28 – 37. 11. Adams J, Williams SV, Aveyard JS, Knowles MA. Loss of heterozygosity and images were processed as described in the previous section. analysis and DNA copy number measurement on 8p in bladder cancer reveals two Areas enriched for tumor cells were identified by reference to mechanisms of allelic loss. Cancer Res 2005;65:66 – 75. near-adjacent sections stained with H&E, and fluorescence 12. Ade´laı¨de J, Huang HE, Murati A, et al. A recurrent chromosome was scored on a minimum of 50 nuclei per tumor. Two translocation breakpoint in breast and pancreatic cancer cell lines targets the neuregulin/NRG1 gene. Genes Chromosomes Cancer 2003;37:333 – 45. observers (C. Ginestier and V. Gelsi-Boyer) read the tissue 13. Huang HE, Chin SF, Ginestier C, et al. A recurrent chromosome breakpoint microarray independently. in breast cancer at the NRG1/neuregulin 1/heregulin gene. Cancer Res 2004;64: 6840 – 4. Immunohistochemical Analysis 14. Theillet C, Ade´laı¨de J, Louason G, et al. FGFR1 and PLAT genes and DNA amplification at 8p12 in breast and ovarian cancers. Genes Chromosomes Immunohistochemistry was done and analyzed as described Cancer 1993;7:219 – 26. previously (48, 54). Results were evaluated under a light 15. Courjal F, Theillet C. Comparative genomic hybridization analysis of breast microscope by two pathologists (E. Charafe-Jauffret and tumours with predetermined profiles of DNA amplification. Cancer Res 1997;57: J. Jacquemier). 4368 – 77. 16. Ade´laı¨de J, Chaffanet M, Imbert A, et al. Chromosome region 8p11-p21: refined mapping and molecular alterations in breast cancer. Genes Chromosomes Statistical Methods Cancer 1998;22:186 – 99. Distributions of molecular markers and other categorical 17. Ugolini F, Ade´laı¨de J, Charafe-Jauffret E, et al. Differential expression assay variables were compared using standard m2 tests or Fisher exact of chromosome arm 8p genes identifies Frizzled-related (FRP1/FRZB) and fibroblast growth factor receptor 1 (FGFR1) as candidate breast cancer genes. test. Overall survival, disease-free survival, and metastasis-free Oncogene 1999;18:1903 – 10. survival intervals were calculated from the date of diagnosis. 18. Saint-Ruf C, Gerbault-Seureau M, Viegas-Pequignot E, Zafrani B, Malfoy B, For overall survival, death was scored as an event; for disease- Dutrillaux B. Recurrent homogeneously staining regions in 8p1 in breast cancer and lack of amplification of POLB, LHRH, and PLAT genes. Cancer Genet free survival, first locoregional recurrence or first metastasis Cytogenet 1991;52:27 – 35. was scored as an event; for metastasis-free survival, first distant 19. Pandis N, Jin Y, Gorunova L, et al. Chromosome analysis of 97 primary metastasis was scored as an event. Patients without event were breast carcinomas: identification of eight karyotypic subgroups. Genes Chromo- censored at the time of last follow-up for each survival somes Cancer 1995;12:173 – 85. measurement methods or at death for disease-free survival and 20. Bernardino J, Apiou F, Gerbault-Seureau M, Malfoy B, Dutrillaux B. Characterization of recurrent homogeneously staining regions in 72 breast metastasis-free survival. For graphical presentation, follow-up carcinomas. Genes Chromosomes Cancer 1998;23:100 – 8. was truncated at 120 months. Survival curves were derived 21. Ray ME, Yang ZQ, Albertson D, et al. Genomic and expression analysis of the from Kaplan-Meier estimates and the curves were compared by 8p11-12 amplicon in human breast cancer cell lines. Cancer Res 2004;64:40 – 7. log-rank tests. 22. Reyal F, Stransky N, Bernard-Pierrot I, et al. Visualizing chromosomes as transcriptome correlation maps: evidence of chromosomal domains containing co-expressed genes—a study of 130 invasive ductal breast carcinomas. Cancer Acknowledgments Res 2005;65:1376 – 83. We thank F. Birg, D. Maraninchi, and C. Mawas for encouragement and L. Ursule 23. Garcia MJ, Pole JC, Chin SF, et al. A 1 Mb minimal amplicon at 8p11-12 in for help. breast cancer identifies new candidate oncogenes. Oncogene 2005;24:5235 – 45. 24. Prentice LM, Shadeo A, Lestou VS, et al. NRG1 gene rearrangements in References clinical breast cancer: identification of an adjacent novel amplicon associated with 1. Emi M, Fujiwara Y, Nakajima T, et al. Frequent loss of heterozygosity for loci poor prognosis. Oncogene 2005;24:7281 – 9. on chromosome 8p in hepatocellular carcinoma, colorectal cancer, and lung 25. Letessier A, Mozziconacci MJ, Murati A, et al. Multicolour-banding cancer. Cancer Res 1992;52:5368 – 72. fluorescence in situ hybridisation (mbanding-FISH) to identify recurrent 2. Knowles MA, Shaw ME, Proctor AJ. Deletion mapping of chromosome 8 in chromosomal alterations in breast tumour cell lines. Br J Cancer 2005;92:382 – 8. cancers of the urinary bladder using restriction fragment length polymorphisms 26. Lucito R, Healy J, Alexander J, et al. Representational oligonucleotide and microsatellite polymorphisms. Oncogene 1993;8:1357 – 64. microarray analysis: a high-resolution method to detect genome copy number 3. Adeyinka A, Kytola S, Mertens F, Pandis N, Larsson C. Spectral karyotyping variation. Genome Res 2003;13:2291 – 305. and chromosome banding studies of primary breast carcinomas and their lymph 27. Hyman E, Kauraniemi P, Hautaniemi S, et al. Impact of DNA amplification node metastases. Int J Mol Med 2000;5:235 – 40. on gene expression patterns in breast cancer. Cancer Res 2002;62:6240 – 5. 4. Davidson JM, Gorringe KL, Chin SF, et al. Molecular cytogenetic analysis of 28. Pollack JR, Sorlie T, Perou CM, et al. Microarray analysis reveals a major breast cancer cell lines. Br J Cancer 2000;83:1309 – 17. direct role of DNA copy number alteration in the transcriptional program of 5. Forozan F, Mahlamaki EH, Monni O, et al. Comparative genomic human breast tumors. Proc Natl Acad Sci U S A 2002;99:12963 – 8.

29. Yamamoto Y, Irie K, Asada M, Mino A, Mandai K, Takai Y. Direct binding putative tumor suppressors controlling cell death commitment. Proc Natl Acad of the human homologue of the Drosophila disc large tumor suppressor gene to Sci U S A 2003;100:4173 – 8. seven-pass transmembrane proteins, tumor endothelial marker 5 (TEM5), and a 44. Engelkamp D. Cloning of three mouse unc5 genes and their expression novel TEM5-like protein. Oncogene 2004;23:3889 – 97. patterns at mid-gestation. Mech Dev 2002;118:191 – 7. 30. Schweinfest CW, Graber MW, Chapman JM, Papas TS, Baron PL, Watson 45. Hamaguchi M, Meth JL, von Klitzing C, et al. DBC2, a candidate for a tumor DK. CaSm: an Sm-like protein that contributes to the transformed state in cancer suppressor gene involved in breast cancer. Proc Natl Acad Sci U S A 2002;99: cells. Cancer Res 1997;57:2961 – 5. 13647 – 52. 31. Rew DA. Small RNAs: a new class of genome regulators and their 46. Mo¨bus VJ, Moll R, Gerharz CD, et al. Differential characteristics of two new significance. Eur J Surg Oncol 2003;29:764 – 5. tumorigenic cell lines of human breast carcinoma origin. Int J Cancer 1998;77: 32. Fraser MM, Watson PM, Fraig MM, et al. CaSm-mediated cellular 415 – 23. transformation is associated with altered gene expression and messenger RNA 47. Forozan F, Veldman R, Ammerman CA, et al. Molecular cytogenetic analysis stability. Cancer Res 2005;65:6228 – 36. of 11 new breast cancer cell lines. Br J Cancer 1999;81:1328 – 34. 33. Conte N, Charafe-Jauffret E, Delaval B, et al. Carcinogenesis and 48. Jacquemier J, Ginestier C, Rougemont J, et al. Protein expression profiling translational controls: TACC1 is down-regulated in human cancers and associates identifies subclasses of breast cancer and predicts prognosis. Cancer Res 2005;65: with mRNA regulators. Oncogene 2002;21:5619 – 30. 767 – 79. 34. Ugolini F, Charafe-Jauffret E, Bardou VJ, et al. WNT pathway and mammary 49. Orsetti B, Nugoli M, Cervera N, et al. Genomic and expression profiling of carcinogenesis: loss of expression of candidate tumor suppressor gene SFRP1 in chromosome 17 in breast cancer reveals complex patterns of alterations and novel most invasive carcinomas except of the medullary type. Oncogene 2001;20: candidate genes. Cancer Res 2004;64:6453 – 60. 5810 – 7. 50. Irizarry RA, Hobbs B, Collin F, et al. Exploration, normalization, 35. Armes JE, Hammet F, de Silva M, et al. Candidate tumor-suppressor and summaries of high density oligonucleotide array probe level data. genes on chromosome arm 8p in early-onset and high-grade breast cancers. Biostatistics 2003;4:249 – 64. Oncogene 2004;23:5697 – 702. 51. Eisen MB, Spellman PT, Brown PO, Botstein D. Cluster analysis and 36. Borrow J, Stanton VP, Jr., Andresen JM, et al. The translocation display of genome-wide expression patterns. Proc Natl Acad Sci U S A 1998;95: t(8;16)(p11;p13) of acute myeloid leukaemia fuses a putative acetyltransferase 14863 – 8. to the CREB-binding protein. Nat Genet 1996;14:33 – 41. 52. Golub TR, Slonim DK, Tamayo P, et al. Molecular classification of 37. Masayesva BG, Ha P, Garrett-Mayer E, et al. Gene expression alterations cancer: class discovery and class prediction by gene expression monitoring. over large chromosomal regions in cancers include multiple genes unrelated to Science 1999;286:531 – 7. malignant progression. Proc Natl Acad Sci U S A 2004;101:8715 – 20. 53. Bertucci F, Finetti P, Rougemont J, et al. Gene expression profiling for 38. Cheng KW, Lahad JP, Kuo WL, et al. The RAB25 small GTPase determines molecular characterization of inflammatory breast cancer and prediction of aggressiveness of ovarian and breast cancers. Nat Med 2004;10:1251 – 6. response to chemotherapy. Cancer Res 2004;64:8558 – 65. 39. Coquelle A, Pipiras E, Toledo F, Buttin G, Debatisse M. Expression of fragile 54. Ginestier C, Charafe-Jauffret E, Bertucci F, et al. Distinct and complementary sites triggers intrachromosomal mammalian gene amplification and sets information provided by use of tissue and DNA microarrays in the study of breast boundaries to early amplicons. Cell 1997;89:215 – 25. tumor markers. Am J Pathol 2002;161:1223 – 33. 40. Kent WJ, Sugnet CW, Furey TS, et al. The human genome browser at UCSC. 55. Courtay-Cahen C, Morris JS, Edwards PAW. Chromosome translocations in Genome Res 2002;12:996 – 1006. breast cancer with breakpoints at 8p12. Genomics 2000;66:15 – 25. 41. Prasad MA, Trybus TM, Wojno KJ, Macoska JA. Homozygous and frequent 56. Chin SF, Daigo Y, Huang HE, et al. A simple and reliable pretreatment deletion of proximal 8p sequences in human prostate cancers: identification protocol facilitates fluorescent in situ hybridisation on tissue microarrays of of a potential tumor suppressor gene site. Genes Chromosomes Cancer 1998;23: paraffin wax embedded tumour samples. Mol Pathol 2003;56:275 – 9. 255 – 62. 57. van der Burg M, Poulsen TS, Hunger SP, et al. Split-signal FISH 42. Livesey FJ. Netrins and netrin receptors. Cell Mol Life Sci 1999;56:62 – 8. for detection of chromosome aberrations in acute lymphoblastic leukemia. 43. Thiebault K, Mazelin L, Pays L, et al. The netrin-1 receptors UNC5H are Leukemia 2004;18:895 – 908.

Mol Cancer Res 2005;3(12). December 2005 Downloaded from mcr.aacrjournals.org on September 27, 2021. © 2005 American Association for Cancer Research. Comprehensive Profiling of 8p11-12 Amplification in Breast Cancer

Véronique Gelsi-Boyer, Béatrice Orsetti, Nathalie Cervera, et al.

Mol Cancer Res 2005;3:655-667.

Updated version Access the most recent version of this article at: http://mcr.aacrjournals.org/content/3/12/655

Supplementary Access the most recent supplemental material at: Material http://mcr.aacrjournals.org/content/suppl/2006/01/04/3.12.655.DC1

Cited articles This article cites 57 articles, 22 of which you can access for free at: http://mcr.aacrjournals.org/content/3/12/655.full#ref-list-1

Citing articles This article has been cited by 33 HighWire-hosted articles. Access the articles at: http://mcr.aacrjournals.org/content/3/12/655.full#related-urls

E-mail alerts Sign up to receive free email-alerts related to this article or journal.

Reprints and To order reprints of this article or to subscribe to the journal, contact the AACR Publications Subscriptions Department at [email protected].

Permissions To request permission to re-use all or part of this article, use this link http://mcr.aacrjournals.org/content/3/12/655. Click on "Request Permissions" which will take you to the Copyright Clearance Center's (CCC) Rightslink site.