A Network of Clinically and Functionally Relevant Genes Is Involved in The

A network of clinically and functionally relevant genes is involved in the reversion of the tumorigenic phenotype of MDA-MB-231 breast cancer cells after transfer of human chromosome 8

Susanne Seitz*,1, Renate Frege1, Anja Jacobsen2,Jo¨ rg Weimer2, Wolfgang Arnold3, Clarissa von Haefen4, Dieter Niederacher5, Rita Schmutzler6, Norbert Arnold2 and Siegfried Scherneck1

1Department of Tumor Genetics, Max Delbrueck Center for Molecular Medicine, Robert Roessle Str. 10, 13125 Berlin, Germany; 2Oncology Laboratory, Gynecology and Obstetrics Clinic, University Hospital Schleswig-Holstein Campus Kiel, Michalisstr. 16, 24105 Kiel, Germany; 3atugen AG, Wiltbergstr. 50, 13125 Berlin, Germany; 4Department of Hematology, Oncology and Tumor Immunology, Charite-Campus Berlin-Buch, Humboldt University, Lindenberger Weg 80, 13125 Berlin-Buch, Germany; 5Department of Gynecology and Obstetrics, University of Duesseldorf, Moorenstr. 5, 40225 Duesseldorf, Germany; 6Department of Molecular Gynecology and Oncology, Gynecology and Obstetrics Clinic, Kerpener Str. 34, 50931 Ko¨ln, Germany

Several investigations have supposed that tumor suppres- and 17p are detected frequently in more than 20–25% of sor genes might be located on human chromosome 8. We primary BC, suggesting the presence of tumor suppres- used microcell-mediated transfer of chromosome 8 into sor genes (TSG) in these regions of the genome MDA-MB-231 breast cancer cells and generated inde- (Lerebours and Lidereau, 2002). The regions of LOH pendent hybrids with strongly reduced tumorigenic in BC are usually large and complex and it is difficult to potential. Loss of the transferred chromosome results in identify candidate genes relying on allelic loss alone. reappearance of the malignant phenotype. Expression Furthermore, dissection of genetic events in LOH analysis identified a set of 109 genes (CT8-ps) differen- regions is complicated by factors such as tumor tially expressed in microcell hybrids as compared to the heterogeneity and contamination with nonmalignant tumorigenic MDA-MB-231 and rerevertant cells. Of material (Murakami, 2002). It is therefore important to these, 44.9% are differentially expressed in human breast obtain functional evidence for a TSG before laborious tumors. The expression pattern of CT8-ps was associated positional cloning attempts are made. with prognostic factors such as tumor size and grading as Microcell-mediated chromosome transfer (MMCT) is well as loss of heterozygosity at the short arm of an alternative approach to identify TSGs in sporadic chromosome 8. We identified CT8-ps networks suggesting cancers. Cancer cells showing deletions at the target that these genes act cooperatively to cause reversion of chromosomal region are preferentially used as recipient, tumorigenicity in MDA-MB-231 cells. Our findings working on the assumption that both alleles of a TSG provide a conceptual basis and experimental system to have been inactivated. The altered phenotype exhibited identify and evaluate genes and gene networks involved in by MMCT hybrids, as compared to the parental cancer the development and/or progression of breast cancer. cells, allows conclusions regarding the location and the Oncogene (2005) 24, 869–879. doi:10.1038/sj.onc.1208260 function of genes relevant for neoplastic development Published online 6 December 2004 under physiological conditions (Kugoh et al., 2002). Several lines of evidence suggest that chromosome 8 Keywords: breast tumorigenesis; microcell-mediated harbors one or more TSGs in BC. Various laboratories, chromosome transfer; expression difference analysis including ours, have reported up to 86% LOH in both familial and sporadic BC as well as in ductal carcinoma in situ (Seitz et al., 1997, 2000; Suzuki et al., 1999). LOH on chromosome 8p is also associated with higher Introduction grade and invasive behavior of tumors (Yaremko et al., 1996; Seitz et al., 2000) and has been proposed as a Breast Cancer (BC) is thought to result from the prognostic indicator to guide postoperative manage- accumulation of a number of genetic alterations, ment of patients (Tsuneizumi et al., 2002). Studying 305 including inactivation of tumor suppressors. Loss of breast tumors by CGH Rennstam et al. (2003) found heterozygosity (LOH) analysis and comparative geno- that combined loss of 8p and gain of 8q, one of the most mic hybridization (CGH) indicate that allelic losses of common coupled aberrations in BC, is correlated with human chromosomes 1p, 3p, 6q, 8p, 9p, 11p, 13q, 16q high DNA nondiploidy, high histological grade, lymph node positivity and low survival rate. *Correspondence: S Seitz; E-mail: [email protected] Functional evidence for one or more TSGs on Received 18 June 2004; revised 1 October 2004; accepted 7 October 2004; chromosome 8p stems from MMCT of chromosome 8 published online 6 December 2004 into rat prostate- and colorectal cancer cells, indicating Chromosome transfer and gene expression analysis S Seitz et al 870 the existence of a metastasis suppressor gene at 8p12– Restoration of human chromosome 8 in MDA-MB-231 p21 and a TSG at 8p22–p23, respectively (Gustafson breast cancer cells et al., 1996; Nihei et al., 2002). The critical regions defined for these putative suppressor genes encompass To identify a functional role for chromosome 8 in breast the regions most commonly deleted in BC. cancer, we introduced an intact copy of chromosome 8 Few if any studies have linked functional changes to into the MDA-MB-231 cell line. Following transfer, the presence or absence of chromosome 8 in breast eight microcell hybrid clones were isolated, expanded as tumor cells. Recently, Wilson et al. (2003) reported distinct cell lines and characterized. First, the hybrid R gene by transfer of chromosome 8 into two BC cell lines and cells were screened for the presence of the neo PCR using specific primers and the presence of the localized three chromosome 8p regions harboring transferred chromosome 8 using 19 informative MS putative in vitro growth suppressor genes. However, markers. In all hybrid cells, the presence of the these authors did not provide evidence for suppression neoR gene was detected, thus confirming that the of tumorigenicity in vivo. G418-resistant colonies arose as a result of microcell In the present study, we provide an experimental fusion (data not shown). MS analyses revealed that system to identify and evaluate genes and gene net- all hybrid clones carried the donor allele (Figure 1a works involved in BC genesis. For the first time we and c). These data are consistent with the results provide functional evidence for the presence of tumor suppressor(s) on chromosome 8(p) and show of cytogenetic analyses. As shown by FISH-MD CT60/4 hybrid cells contain an entire chromosome 8 that the suppression of the tumorigenic phenotype (donor chromosome), whereas in MDA-MB-231 cells of the BC cell line MDA-MB-231 is mediated by a only fragments of this chromosome were detected specific set of genes (CT8-ps)/networks regulated by one (Figure 1d, Weimer et al., in preparation). In addition, or more genes on chromosome 8. It is of particular CGH analyses demonstrated amplification of chromo- interest that the CT8-ps signature is also reflected in a some 8 in the hybrid clones (CT60/4, CT60/6) as panel of breast tumors and is correlated with worse compared to the parental MDA-MB-231 cells patient prognosis. (Figure 1e). In parallel, hybrids (CT60/4, CT60/6) were screened with FISH to determine whether any mouse chromosomes had inadvertently been transferred along Results with the neoR-tagged chromosome 8. Both cell lines were found to be negative for the presence of contaminating Homozygosity mapping of deletions (HOMOD) analysis mouse DNA. of breast cancer cell lines

We analysed a panel of 11 breast cancer cell lines with 24 Consequences of chromosome 8 transfer in microsatellite (MS) markers for extended regions of MDA-MB-231 breast cancer cells hemizygosity (X5 adjacent markers). These are indicative for allelic loss in the unmatched tumor cells The eight microcell hybrids were first tested for (Goldberg et al., 2000). These markers were concen- differences in cell morphology. In general, the hybrid trated in the 8p12–p21 region previously determined cells appeared flatter than the MDA-MB-231 parental to be frequently deleted in BC (Seitz et al., 2000). cell line (data not shown). All microcell hybrids showed We identified four cell lines (MDA-MB-231, CAMA-1, decreased anchorage-dependent growth. The doubling T-47D, BT-20) showing deletions of the 8p12–p21 times for exponentially growing MDA-MB-231 cells region, which are suitable as recipients of chromosome were 28.9 h, whereas the doubling times of the eight 8 by MMCT (data not shown). microcell hybrids varied between 33.4 h (CT60/4) and 44.6 h (CT60/8) (medium 37.7 h) (Figure 2a). Next, we Selection of MDA-MB-231 as recipient cells for MMCT determined the effect of transferred chromosome 8 on breast cancer cell invasion in vitro. As shown in As demonstrated by HOMOD –analysis, the MDA- Figure 2b, hybrid CT60/4 cells showed a clear reduction MB-231 cells were found to have a region of extended of invasive cells (>97%) compared with the parental hemizygosity of about 10 Mb at chromosomal region MDA-MB-231. We characterized two microcell 8p12–p21, encompassing the critical regions defined by hybrids, CT60/4 and CT60/6, in vivo to determine their LOH analysis (Figure 1a). CGH analysis of these cells tumorigenic potential. The CT60/4 cells (Figure 2c) revealed loss of chromosomal region 8pter–8q21and did not form tumors during a 3-month observation gain of 8q21–qter (Figure 1b). In addition, we compared period and CT60/6 cells exhibited a clear reduction in the gene expression profiles of MDA-MB-231 cells with the ability to form tumors in nude mice in comparison MCF-10A breast epithelial cells and found a high to the parental cell line. Tumors formed after CT60/6 frequency (28.7%) of chromosome 8 genes differentially injection appeared following a delayed latency period expressed (data not shown). This frequency of differen- of at least 3 weeks and showed a significant reduction of tially expressed genes is significantly higher (Po0.0001) tumor volume when compared to the parental cells. as compared to all 22 283 probe sets (ps) spotted on the Next, we analysed the effect of introduction of human U133A GeneChip, where only 20.3% of the genes were chromosome 8 into MDA-MB-231 cells on cell-cycle differentially expressed in MDA-MB-231 cells. progression using FACS. The cell-cycle profile did not

Oncogene Chromosome transfer and gene expression analysis S Seitz et al 871

Figure 1 Detection of chromosome 8 in MDA-MB-231, revertant and rerevertant cells. (a) Schematic representation of microsatellite analysis of chromosome 8 in MDA-MB-231 cells and representative revertant (CT60/4) and rerevertant (CT60/4-12) cells. Gray and dotted circles indicate homozygosity and retention of both alleles in MDA-MB-231 cells, respectively. The deleted region is boxed. The presence of the donor allele (closed circle), the absence of the donor allele (open circle) and uninformative markers (striped circle) are as indicated. (b) CGH analysis of MDA-MB-231 cells with normal DNA as reference. The average ratio profiles are plotted along an ideogram with gains and losses of the respective chromosomal regions indicated by vertical lines on the right (gain) and left (loss) side of the graph. Only the chromosome 8 is shown. (c) Representative examples of microsatellite analysis at chromosome 8 locus D8S1110. The arrow indicates the donor allele. The numbers beneath each peak give the marker, the allele fragment length and peak intensity. (d) FISH-MD of chromosome 8 in MDA-MB-231 and revertant CT60/4 cells. (e) CGH analysis of revertant CT60-4 cells with MDA- MB-231 DNA as reference differ significantly between parental MDA-MB-231 and functionally revertants, anchorage-dependent growth, CT60/4 hybrid cells (data not shown). invasiveness through the matrigel-reconstituted base- ment membrane and tumorigenicity in nude mice were Re-expression of the tumorigenic phenotype in microcell tested. CT60/4-12 cells showed a reduced doubling hybrid clones time (22.9 h) in comparison to the microcell hybrids thus comparable to the parental cells (Figure 2a). A To confirm whether the introduction of the chromosome significant increase of the invasive potential in CT60/4- 8 was the cause for the observed suppression of 12 cells was demonstrated when compared with the tumorigenicity, the microcell hybrids CT60/4 and corresponding hybrid CT60/4 cells (Figure 2b). Unlike CT60/6 were cultivated for 8 weeks in G418-free CT60/4 and comparable to the parental MDA-MB-231 medium. A total of 27 subclones were isolated and cells, the clone CT60/4-12 formed tumors in nude mice analysed for the presence of the neoR gene. In all, 22 (Figure 2c). In addition, CT60/4-12 cells were morpho- subclones lost the neo marker. One of these clones logically similar to the parental MDA-MB-231 cells. (CT60/4-12) was examined in more detail. MS analyses (Figure 1a and c) and FISH-MD showed loss of the Microarray hybridization and search for candidate genes donor allele in CT60/4-12 cells. In addition, these cells on chromosome 8 show a CGH profile comparable with MDA-MB-231 cells and indicative for loss of chromosome 8 material. As a strategy to identify transcriptional targets likely to To address the question, whether the microcell contribute to the suppression of the tumorigenic hybrid cells that lost the donor chromosome became phenotype in CT60/4 cells, we used oligonucleotide

Oncogene Chromosome transfer and gene expression analysis S Seitz et al 872 a A total of 134 ps (107 genes) fulfill the criteria for 40 differential expression, defined in the Materials and methods section. These genes are listed in the supple- 30 ment. A total of 95 ps (75 genes) were upregulated and 39 ps (33 genes) were downregulated in CT60/4 cells as 20 compared to the tumorigenic MDA-MB-231 and CT60/ 4-12 cells, respectively. It is noteworthy that 43.3% of 10 the identified ps are also differentially expressed in MDA-MB-231 cells compared to the MCF-10A breast

Doubling time (hr) time Doubling 0 epithelial cell line (see Supplement table). MDA-MB-231 CT60/4 CT60/4-12 Of the 107 differentially expressed genes, five genes mapped to chromosome 8. Four genes were mapped b 160 outside the 8p12–p21 target region, including INDO, located at 8p11–p12, FLJ20533, CCNE2 and NOV, 120 which mapped to 8q13.3–q24.1. Only one gene, clusterin (CLU), was located within the target region. This 80 region is flanked by the markers D8S1752 and D8S259. According to the Entrez Map Viewer (http:// www.ncbi.nlm.nih.gov/mapview/map_search.cgi), a % Invasion 40 total of 131 genes has been mapped within this region, 57 of them are arrayed on the U133A GeneChip. 0 Of these 57 genes, 32 genes were scored as present, 25 MDA-MB-231 CT60/4 CT60/4-12 genes as being absent on the GeneChip. We used c reverse transcription–polymerase chain reaction (RT– 400 PCR) analysis to test eight of these 25 genes including MDA-MB-231* ) ADAM28, NEFL, GNRH1, PTK2B, EPHX2, CSR1, 3 CT60/4-12 PNOC and EXTL3 and identified two additional genes, CT60/4 the exostoses (multiple-) like 3 (EXTL3) gene and ADAM28, a disintegrin and metalloproteinase domain gene, overexpressed in CT60/4 as compared to MDA- 200 MB-231 and CT60/4-12 cells, respectively (Figure 3).

Functional proﬁling of differentially expressed genes

Tumor volume (mm Tumor volume We used Onto-Express (http://vortex.cs.wayne.edu:8080) 0 to correlate expression data with functional proﬁles 0 414855626975828996 (Khatri et al., 2002). The selected 109 genes are involved in a diverse spectrum of cellular functions with the Days post cell challenge Figure 2 Suppression of tumorigenicity after transfer of chromosome 8 into MDA-MB-231 cells. (a) Doubling times of MDA-MB- 231, representative revertant (CT60/4) and rerevertant (CT60/4-12) cells obtained in two independent experiments. (b) Cell invasion of MDA-MB-231, revertant CT60-4 and rerevertant CT60/4-12 cells MDA-MB-231 through Matrigel invasion chamber. Data are shown as the CT60/4 percentage of invasion, normalized to the number of migrated CT60/4-12 MDA-MB-231 Genes M CT60/4 CT60/4-12 MDA-MB-231 cells and obtained from the triplicate of two independent experiments; bars, s.d. (c) In vivo tumorigenicity of 339bp ADAM28 MDA-MB-231, revertant (CT60-4) and rerevertant (CT60/4-12) cells after subcutaneous injection of 1 Â 106 cells in NMRI-nu/nu mice over time. *Data are means of two independent experiments EXTL3 311bp

GAPDH 443bp

microarrays (HG-U133A) to compare the gene expres- PBGD 126bp sion proﬁles of CT60/4 microcell hybrid cells with those of the tumorigenic parental cells, MDA-MB-231, and B2M 114bp the rerevertant cells CT60/4-12. We hypothesize that 0 50 100 150 genes overexpressed in CT60/4 microcell hybrids but O.D. lower expressed or downregulated in MDA-MB-231 and CT60/4-12 cells, or vice versa, can be considered as Figure 3 RT–PCR analysis. Expression analysis of ADAM28, EXTL3 and the housekeeping genes B2M, GAPDH and PBGD in candidate genes involved in the chromosome 8-mediated parental MDA-MB-231 cells, revertant CT60/4 and rerevertant reversion of tumorigenicity in MDA-MB-231 cells. CT60/4-12 cells using RT–PCR

Oncogene Chromosome transfer and gene expression analysis S Seitz et al 873 majority of them participating in immune response, cell The numbers of literature co-occurrences between two proliferation, signal transduction, cell–cell signaling, genes varied from 1 to 358. The 40 genes cluster into five inflammatory response, neurogenesis, oncogenesis and networks (a–e). In cluster a, 31 genes including four on cell adhesion. chromosome 8 (NOV, CLU, ADAM28, INDO) are A literature search using LocusLink (http:// connected. The other four clusters consist of two or www.ncbi.nlm.nih.gov/LocusLink/list.cgi) revealed that three genes. 20 and seven of the 109 genes in our set have previously In a second step, we analysed the gene expression been associated with cancer or breast cancer, respec- data in the context of literature associations (http:// tively. www.pubgene.org/tools/ExpressionData/Analysis.cgi). The differential expression of multiple genes from The expression data are superimposed on the literature different chromosomes in CT60/4 microcell hybrids network to compute an expression score for each gene raises the question of their functional significance. To in the CT8-ps. Under the criteria of upregulation, the determine the biological relationship between the 109 top-scoring cluster included a network of interferon- differentially expressed genes and to identify coex- induced proteins with tetratricopeptide repeats (IFITs) pressed groups of genes associated in the literature, we (Figure 4b). When we analysed our gene-set under the used PubGene web tools (http://www.pubgene.org/ criteria of downregulation in CT60/4 vs MDA-MB-231 tools/Network/Subset.cgi) (Jenssen et al., 2001). We cells, the highest ranking cluster consisted of six genes, identified 40 genes which co-occurred in the literature including QPCT, CSF3, IL11, IL1A, CXCR4 and TLR4 with at least one other gene of the CT8-ps (Figure 4). (Figure 4a).

ISG20 c a CD24 ABCG2 OAS1 2 NP 2 BCHE SAA2 2 MX1 QPCT OAS2 8 CSF3 9 GAD1 INDO CXCR4 358 91 4 PTGER4 11 IL11 120 TLR4 14 IL1A HPSE LIM 2

2 46 CLU IGFBP6 MMP1 BMP4 CDKN1A 37

TFF1 NOV PTN TIMP3 ADAM28 d

2 MAGEA3

NR2F1 NDP COL1A2 6 b MAGEA6

e RIG-1 IFIT1 2 IFIT4 2 4 2 RARRES3 IFIT2

Figure 4 Literature association of genes differentially expressed after transfer of chromosome 8 into MDA-MB-231 cells. Genes differentially expressed in revertant CT60/4 vs MDA-MB-231/rerevertant CT60/4-12 cells cluster into ﬁve biological networks (a–e). Blue colored genes reﬂect genes downregulated and green colored are genes upregulated in revertant vs MDA-MB-231/rerevertant cells. The subnetwork containing IFIT1, IFIT2 and IFIT4 was ranked as the most upregulated, whereas the six-gene network containing QPCT, CSF3, IL11, IL1A, CXCR4 and TLR4 was ranked as the most downregulated one. Both subnetworks are marked. The numbers next to the links correspond to the number of literature co-occurrences between the linked genes. Only genes which co- occurred more than once in the literature are numbered

Oncogene Chromosome transfer and gene expression analysis S Seitz et al 874 Validation of the expression pattern of the MMCT- Table 1 Significant correlations between the expression of CT8 genes derived (CT8-ps) gene set in breast tumors and clinicohistopathological parameters and LOH at 8p21 To assess the relevance of the MMCT-derived CT8-ps in No of tumors BC genesis, we determined the expression profiles of 17 Parameter D NC I P-value breast carcinomas. The extent of differential expression varied between 30.3 and 59.8% in single tumors and Probe set 44.9% (23.1% down-, 21.8% upregulated) of the CT8- CT8-ps (135 ps) To2 cm 87 220 66 T ¼ 2–5 cm 175 378 200 ps were differentially expressed in all 17 tumor samples T>5 cm 90 243 77 0.001 (see Supplement table). The percentage of differentially expressed CT8-ps is significantly higher (Po0.001) as CT8-ps (135 ps) G2 127 373 154 compared to all 22 283 ps spotted on the U133A G3 225 468 189 0.019 GeneChip. Only 22.3% of them were differentially expressed in the tumors. Remarkably, the high percen- CT8-ps (135 ps) LOH 8 p 187 434 210 tage of differentially expressed CT8-ps is comparable to ROH 8 p 189 464 145 0.002 Gene that of a recently defined predictor gene set for clinical Probe set ID (location) outcome of breast cancer (van’t Veer et al., 2002). There IFI44 To2cm031 are 46 ps of this 70 gene predictor set on the U133A 214453_s_at (1p31.1) T ¼ 2–5 cm 0 1 7 GeneChip (van’t Veer-ps) and 39.5% of them are T>5 cm 0 3 1 0.042 differentially expressed in the analysed tumors. CXCL10 To2cm031 To gain insight into the prognostic relevance of the 204533_at (4q21) T ¼ 2–5 cm 0 1 7 CT8 gene set, the expression data were correlated to T>5 cm 0 0 4 0.026 established prognostic factors such as tumor size, lymph node status, histological grade and patient age LIM G2 2 5 0 at the time of diagnosis. Differential expression of all 213684_s_at (4q22) G3 9 0 0 0.005 CT8-ps was significantly correlated to higher grading (P 0.019) and higher tumor size (P 0.001) (Table 1). FLJ20637 LOH 8 p 0 3 4 ¼ ¼ 219352_at (4q22.1) ROH 8 p 2 6 0 0.031 In addition, we found a significant correlation between the CT8 expression profile and LOH at 8p (P ¼ 0.002) PSPHL LOH 8 p 3 3 1 (Table 1). Of the 17 tumors tested, nine showed LOH at 205048_s_at (7q11.2) ROH 8 p 8 0 0 0.044 the 8p target region, eight tumors did not. The highest percentage of differentially expressed genes (52.5%) was FLJ20073 To2cm020 219691_at (7q21.3) T ¼ 2–5 cm 0 0 4 found in tumors with LOH at 8p and high grading T>5 cm 0 0 3 0.011 (P ¼ 0.002) compared to tumors with LOH and lower grading or tumors without 8pLOH (data not shown). Clu To2cm310 To further evaluate the clinical relevance of the CT8- 208791_at (8p21–p12) T ¼ 2–5 cm 8 0 0 ps, we analysed the correlation between the expression T>5 cm 1 3 0 0.018 of CT8 single genes and clinico-histopathological para- ADAM28 G2 1 4 0 meters, the age at diagnosis and the 8p LOH status. As 208268_at (8p21.2) G3 7 1 0 0.032 shown in Table 1, we identified 13 genes (15 ps) whose expression pattern was significantly correlated with NOV G2 0 7 0 tumor size, age at diagnosis and 8p LOH. Three of 214321_at (8q24.1) G3 4 3 2 0.024 these genes are located on chromosome 8: CLU and NOV LOH 8 p 4 5 0 ADAM 28 on 8p, NOV on 8q. 214321_at (8q24.1) ROH 8 p 0 5 3 0.031

IFIT1 To2cm003 203153_at (10q25–q26) T ¼ 2–5 cm 0 0 8 Discussion T>5 cm 0 2 2 0.042

This study provides the ﬁrst functional evidence that one FKBP11 To2cm310 219118_at (12q13.12) T ¼ 2–5 cm 0 7 0 or more tumor suppressor’s contributing to BC tumor- T>5 cm 0 4 0 0.006 igenesis exist on chromosome 8. The microcell hybrids generated are characterized by altered morphology, MX1 To2cm031 reduced anchorage-dependent growth, invasive poten- 202086_at (21q22.3) T ¼ 2–5 cm 0 1 7 tial and substantially reduced tumorigenic capability in T>5 cm 0 3 1 0.042 nude mice. Loss of chromosome 8 in hybrid cells TIMP3 LOH 8 p 4 0 4 restored the original tumorigenic potential of MDA- 201147_s_at (22q12.3) ROH 8 p 3 5 0 0.010 MB-231 cells, implying that suppression of tumorigenicity was, in fact, caused by introduction and retention of TIMP3 p55 years 6 1 0 201148_s_at (22q12.3) >55 years 3 7 0 0.050 chromosome 8. In order to identify genes or groups of genes which I, increased; D, decreased; NC, no change in expression in breast contribute to the suppression of the tumorigenic tumors vs pooled normal breast samples

Oncogene Chromosome transfer and gene expression analysis S Seitz et al 875 phenotype in hybrid CT60/4 cells, we used oligonucleo- whose expression pattern may shed light on signaling tide microarrays and RT–PCR and compared the gene pathways and critical effectors of these genes (Jenssen expression profiles of CT60/4 cells with those of the et al., 2001). Recently, it has been demonstrated that tumorigenic MDA-MB-231, and the rerevertant CT60/ gene pairs with network neighborhoods occurring more 4-12 cells. There were a total of 109 genes that met our frequently together in article abstracts can be function- criteria for differential expression. These genes represent ally related. These genes more frequently encode a variety of functional groups including cell prolifera- proteins that interact physically than do genes with tion, signal transduction, cell–cell signaling and cell dissimilar neighborhoods (Schlitt et al., 2003). We found adhesion. In addition, there are several genes with more 40 genes co-occurring at least once in article abstracts specialized roles, reflecting, as an example, stromal and with one of the 109 genes identified. They cluster into inflammatory components of the tumor tissues. five networks (a–e). Some of them contain only two One of the 109 differentially expressed genes, CLU, genes. However, network A contains a large number of and additional two genes, EXTL3 and ADAM28, both genes. In particular, gene pairs CSF3–IL1A, IL1A– identified by RT–PCR analysis, were located within the IL11, CSF3–IL11, IL1A–MMP1 and MMP1–TIMP3 in 8p12–p21 target region. All three genes have been network A co-occur 358, 120, 91, 46 and 37 times, suggested to be involved in tumorigenesis. CLU is respectively, suggesting a functional relationship be- implicated in many biological processes, such as cell tween them. This is strengthened by the data presented adhesion, lipid transport, reproduction and apoptosis here as well as by data from elsewhere in the literature. (Jones and Jomary, 2002; Trougakos and Gonos, 2002) In addition, we identified two top-scoring subnetworks and has been reported to be overexpressed in some types that were ranked according to their content of up- or of cancer (Steinberg et al., 1997; Redondo et al., 2000), downregulated genes. Under the criteria of upregula- but downregulated in others (Behrens et al., 2001; Saffer tion, the top-scoring gene cluster consists of a network et al., 2002; Scaltriti et al., 2004). This may reflect of interferon-induced proteins with tetratricopeptide differences in the relative levels of the various isoforms repeat motifs (IFITs). Two members of the IFIT gene (Pucci et al., 2004). Members of the ADAM protein family, IFIT1 and IFIT4, have been described as family have been implicated in cell–cell and cell–matrix predictors of BC outcome (Huang et al., 2003). interactions (Primakoff and Myles, 2000). EXTL3 is a Furthermore, IFIT genes have been mapped to chromo- member of the exostosis (EXT) gene family, which has some 10q23–q26, a genome region frequently deleted in been tightly associated with glycosyltransferase activities different types of cancer, including BC (Singh et al., required for the biosynthesis of heparin sulfate (Kim 1998). Interferon-induced genes are also involved in et al., 2001). For two members of the EXT-family, network c, consisting of members of oligoadenylate EXT1 and EXT2, a TSG function has been postulated. synthases (OAS1 and OAS2) and are also found in a This was proved by frequent LOH at both gene loci in gene cluster within network a (MX1, ISG20). These various human cancers and intragenic mutations in findings are intriguing in view of the role of interferons sporadic chondrosarcomas (Hogue et al., 1996). The as mediators of the antitumor response. sequence and functional similarities between the EXT- By analysing our gene-set under the criteria of and the EXTL3 genes suggest also a TSG function for downregulation, the highest ranking cluster consisted the latter. of six genes (QPCT, CSF3, IL11, IL1A, CXCR4 and Four additional genes that met the threshold criteria TLR4), all included in network a. In a recent study, for differential gene expression mapped to chromosome Kang et al. (2003) investigated the molecular basis for 8. One of these genes, INDO,ag interferon-induced osteolytic bone metastasis by selecting MDA-MB-231 gene with antiproliferative effect, was mapped near the cell subpopulations with elevated metastatic activity and 8p12–p21 target region (Logan et al., 2002). The other functionally validated genes overexpressed in these cells. three genes mapped to 8q13.3–q24.3 including NOV, They identified four genes (CXCR4, IL11, CTGF and a member of the CCN family of proteins, FLJ20533 MMP1) that are highly overexpressed in bone metas- and cyclin E2 that are involved in cell cycle regulation tases. These genes act cooperatively to cause metastasis and control of cell proliferation (Gudas et al., 1999; and are already overexpressed in the parental MDA- Strezoska et al., 2002; Borg et al., 2003). Interestingly, MB-231 cells. We found that three of the metastasis- two of the genes (NOV, CCNE2) have previously been associated genes (CXCR4, IL11 and MMP1) are down- found to be aberrantly expressed in human cancers, regulated in CT60/4 as compared to the MDA-MB-231 including BC (Gudas et al., 1999; Payton et al., 2002). and CT60/4-12 cells. These data imply a coordinate All chromosome 8 candidate genes were overex- down-regulation of these genes by genes or regulator pressed in the CT60/4 hybrids as compared to the elements, which remained as yet unidentified. It is also tumorigenic MDA-MB-231 and CT60/4-12 cells. This is of interest that Nihei et al. (2002) provided functional consistent with a tumor suppressor activity. evidence for a metastasis suppressor gene for rat For the majority of the 109 differentially expressed prostate cancer within the human chromosome region genes in CT60/4 cells, the precise function as well as the 8p21–p12. In this report, we have not assessed the underlying biological relationships between these genes consequences of chromosome 8 MMCT regarding in the course of BC development remain unknown or suppression of the metastatic ability of MDA-MB-231 unclear. We used PubGene web tools to identify groups cells in athymic mice. However, based on our findings, of genes that had co-occurred in the literature and we cannot exclude that the genes responsible for

Oncogene Chromosome transfer and gene expression analysis S Seitz et al 876 metastatic suppression identified by Kang et al. (2003) (T-47D, BT-20, CAMA-1, MCF7) (MDC, Berlin, Germany) were regulated by genes located on chromosome 8 and and W Arnold (R193) (atugen AG, Berlin). The A9(neo8) are involved in tumor suppression. cells, a gift from N Nihei and JC Barrett (NIH, MD, USA), Remarkably, by searching recent publications and are mouse A9 cells containing a single human chromosome 8 databases, we found that expression of 44 of the CT8 derived from normal human fibroblasts and tagged with the neomycin/G418-resistance (neoR) gene. genes (40.4%) have been suggested or proved to be To confirm their identity, the cell lines were screened with correlated with cancer, BC or predictors of BC and/or the microsatellite (MS) markers D16S539, D5S818 and metastasis outcome (van’t Veer et al., 2002; Huang et al., D7S820 and the resulting fingerprints compared with the 2003; Kang et al., 2003; http://www.ncbi.nlm.nih.gov/ corresponding entries in the ATCC STR human cell line LocusLink/). database (http://www.atcc.org/Cultures/str.cfm) as described To explore the relevance of the MMCT-derived CT8- in Seitz et al. (2003). ps in breast cancer tissues, we ascertained the expression profile of 17 breast carcinomas. We found 44.9% of the DNA and RNA extraction CT8-ps as differentially expressed in the tumor samples, DNA was extracted by standard methods, involving proteinase demonstrating a significantly higher percentage as K digestion, phenol/chloroform extraction and ethanol pre- compared to all ps spotted on the U133A GeneChip. cipitation (Sambrook et al., 1989). A similar high percentage of differentially expressed ps Total RNA was isolated using TRIzol Reagent (Life as for the CT8-ps was found with 46 ps of a 70 gene Technologies, Gaithersburg, MD, USA) according to the predictor set recently reported by van’t Veer et al. manufacturer’s instructions. All tumor specimens analysed (2002). In addition, the CT8-ps expression pattern was contained more than 80% tumor cells. RNAs of the normal significantly correlated with prognostic factors such as breast samples were pooled. higher grading and larger tumor size. Taken together, these data strongly suggest the relevance of the CT8-ps HOMOD on chromosome 8 in breast cancer cell lines for breast cancer tumorigenesis. To fine-map the area of interest, we narrowed the region of The clinical relevance of the CT8-ps was further deletion by HOMOD in 11 breast cancer cell lines. A total of strengthened by analysing the correlation between the 28 MS markers covering the whole of chromosome 8 were expression of single genes within the CT8-ps and selected from the Human Genome database (http:// clinicohistopathological parameters, age of the patients www.gdb.org/). Their relative order was established using the at diagnosis and 8pLOH status of the tumors. We NCBI database (http://www.ncbi.nlm.nih.gov/) (Figure 1). detected 13 genes whose expression pattern correlated PCR primers were purchased from BioTez (Berlin, Germany). PCR reactions were carried out in a 10 ml reaction containing with tumor size (IFI44, CXCL10, FLJ20073, CLU, 40 ng of genomic DNA, Taq PCR Master Mix (QIAGEN, IFIT1 FKBP11, MX1), grading (LIM, ADAM28, NOV), Hilden, Germany) (1.5 U Taq DNA Polymerase, 1 Â PCR age at diagnosis (TIMP3) and 8pLOH (FLJ20637, buffer, 200 mM of each dNTP) and 2.5 pmol of each primer. PSPHL, NOV, TIMP3). CLU and ADAM28 are One member of each primer set was tagged with a fluorescent located on 8p, NOV on 8q. label. Initial denaturation was performed at 951C for 5 min, In conclusion, we have shown that introduction of followed by 35 cycles of 30 s at 941C, 30 s at the optimal chromosome 8 into MDA-MB-231 cells results in an annealing temperature and 1 min at 721C, with a final altered phenotype with strongly reduced tumorigenic extension at 721C for 10 min. The fluorescent PCR products potential and is accompanied by characteristic changes were mixed with an internal standard size marker and analysed in gene expression. These changes were also observed in on denaturing 6% polyacrylamide gels using an ABI 377 DNA sequencer. a panel of human breast tumors. Both, the complete set of differentially expressed genes (CT8-ps) and various genes of this gene set can be correlated with established LOH mapping in breast tumors prognostic factors and LOH at the short arm of LOH analysis of 17 tumors and corresponding nontumorous chromosome 8. The data presented here and by other tissues was performed using the MS markers NEFL and authors suggest that differentially expressed genes with- D8S1989 (http://www.gdb.org/) as described above. If the patient was not informative for the two markers, a dinucleo- in the CT8-ps are significantly associated with tumor 0 suppression. A detailed analysis of these genes and their tide repeat near NEFL was analysed using the primers 5 -6- FAM-CAGAACCATGGACAGAGT-30 and 50-TCAGGCA interactions will help to determine the extent of their ATTACAGGAGG-30. LOH was ascertained by comparing involvement in breast cancer development. nontumorous and tumor tissue allele peak sizes. Intensity or signal ratio differences of 35% or more were considered sufficient for LOH assignment. Materials and methods MMCT Cell lines MMCT was performed as described by Theile et al. (1994). One normal breast and 11 breast cancer cell lines were Briefly, donor A9neo8 cells were treated with 0.05 mg/ml obtained from the American Type Culture Collection (ATCC) colcemid in DMEM supplemented with 20% FBS for 48 h at (MCF-10A, MDA-MB-231)] or were kindly provided by J 371C. The flasks were filled with medium containing 10 mg/ml Gioanni (Centre Antoine-Lacassagne, Nice, France) (CAL51), cytocholasin B and centrifuged in a Sorvall fixed-angle rotor D Amadori (Morgagni-Pierantoni Hospital, Forli, Italy) GSA at 8000 r.p.m. for 60 min at 351C. Microcell pellets were (BRC230), I Fichtner (MT-1, MT-3, MDA-N), M Theile resuspended and filtered through 8, 5 and 3 mm polycarbonate

Oncogene Chromosome transfer and gene expression analysis S Seitz et al 877 filters. Purified microcells were attached to subconfluent 10–50 mg Cot-1 DNA. The probes were hybridized onto recipient MDA-MB-231 cells with prewarmed serum-free normal male metaphase target slides (Vysis distributor: medium containing 50 mg/ml phytohemagglutinin P (Sigma) Abbott, Wiesbaden, Germany) for 2–3 days in a moist and then fused to recipient cells with 50% polyethylene glycol chamber at 371C. Posthybridization washes were performed (MW 1500) (Roche Diagnostics GmbH, Mannheim, Ger- without formamide as follows: 5 min at 2 Â SSC at room many) for 2 min. After 2 days the cells were split, plated at a temperature, 5 min at 721C and a final 5 min wash at room density of 5 Â 105 cells per 60 mm dish and cultured for 2–3 temperature. After detection with streptavidin–fluorescein weeks to select for cells that contained a neoR-tagged human isothiocyanate and anti-digoxigenin–rhodamine antibody chromosome 8. Colonies were isolated using cloning cylinders, (both Roche Diagnostics GmbH), the slides were counter- transferred to 48-well plates and expanded for further analysis. stained with 0.1 mg/ml DAPI in antifade. CGH hybridization From the chromosome transfer experiments, eight subclones was analysed using a fluorescence microscope (Axioplan 2) were obtained. connected to a digital imaging system and evaluated with the ISIS-3 software (MetaSystems). Analysis of microcell hybrids Neo-specific PCR To confirm the presence of the introduced Anchorage-dependent growth chromosome 8 in the hybrid cells, we first performed PCR to Growth rate was determined in duplicate in parental and R detect the neo gene. For PCR, we used the same conditions as hybrid cells by plating 2.5 Â 104. cells into 60 mm culture dishes described but omitted the fluorescence labeling. The sequence containing medium supplemented with 5% FBS. Cells were R 0 of the neo primer used are: 5 -ACTGGGCACAACAGA counted every 24 h over a 5-day period and starting from 0 0 0 CAATC-3 and 5 -GCCATTTTCCACCATGATAT-3 . The day 3 after plating using the cell counter and analysis PCR products (536 bp) were separated by acrylamide gel system CASYs1/TTC (Scha¨ rfe System GmbH, Reutlingen, electrophoresis and visualized by staining with ethidium Germany). bromide.

Allelotyping of chromosome 8 in microcell hybrids Eight Anchorage-independent growth hybrid cells were tested for successful transfer of human To assay colony formation in soft agar, 2.5 Â 104. cells in 0.3% chromosome 8 using 19 informative MS markers as described agarose medium (3 ml) containing 10% FBS were layered onto above (Figure 1). 0.6% agar medium (4 ml) containing 10% FBS in a 60 mm culture dish. Cells were incubated at 371C for 3 weeks. In our FISH and FISH MD hands and in contrast to the results, published by Zhang et al. (1991), the MDA-MB-231 cells did not form colonies in soft The paints for a multiplex FISH were generated by micro- agar. dissection. All paints (300–500 ng each) were solved with 20 mg Cot-1 in a hybridization mix (2 Â SSC, 50% formamide, 10% dextransulphate, 1% Tween 20, pH 7.0). The following paints In vitro invasion assay were hybridized together: WCP8 labeled with Biotin-16-dUTP, Cell invasion was assayed in a 24-well BioCoat Matrigel 8p labeled with Spectrum-green-dUTP, 8q12 labeled with Invasion Chamber with a pore size of 8 mm (BD Biosciences) Texas-Red-dUTP and 8q21-qter labeled with Diethylamino- according to the manufacturer’s protocol. Briefly, 2.5 Â 104 coumarin-5-dUTP (DEAC). In all, 10 ml of each paint were cells were plated in the top chamber. The bottom chamber 1 denatured for 5 min at 75 C and prehybridized for 30 min at contained 10% FBS as a chemoattractant. After 22 h incuba- 1 37 C. The further procedure of hybridization and detection tion, cells that have migrated through the membrane were followed standard protocols. For microdissection of fluores- fixed with methanol and stained with 1% Toluidine blue. cence labeled tumor chromosomes (FISH-MD), the hybridiza- Assays were performed in duplicate. tion was performed as described by Weimer et al. (1999, 2000) and the generated paints were hybridized on unobtrusive lymphocyte metaphases as described above. An epifluorescense Tumorigenicity test microscope (Axioplan 2, Zeiss, Go¨ ttingen, Germany) and ISIS- Suspensions of 1 Â 106 cells in 100 ml serum-free culture software Version 3 (MetaSystems, Altlussheim, Germany) medium were injected subcutaneously into the left inguinal were used for Image analysis. region of 5- to 7-week-old female immune-deficient nude mice (NMRI/nu-nu). Three to five mice were injected per group. Hybridization of genomic mouse DNA Tumor growth was monitored once a week over a 12-week We used a mouse genomic paint generated by microdissection period. of 10 mouse lymphocyte nuclei. This probe has been amplified by DOP-PCR and labeled by Diethylaminocoumarin-5-dUTP Fluorescence-activated cell sorter (FACS) analysis (NEN Life Science) in a further DOP-PCR step as described by Weimer et al. (2000). Cells were harvested by trypsin digestion, collected by centrifugation at 300 g,41C for 5 min, washed with PBS at 41C, and fixed in PBS/1% formaldehyde on ice for CGH 30 min. After fixation, the cells were incubated with ethanol/ CGH was carried out according to Arnold et al. (1996) with PBS (2 : 1) for 15 min, pelleted, and re-suspended in PBS minor modifications. Briefly, 2 mg of genomic DNA was containing 40 mg/ml RNase A. After incubation for 30 min at labeled in a standard nick-labeling procedure with biotin-16- 371C, the cells were pelleted again, resuspended in PBS dUTP and digoxigenin-11-dUTP (Roche Diagnostics GmbH) containing 50 mg/ml propidium iodide and subjected to FACS. for tumor and reference DNA, respectively, in a standard nick- Data were collected and analysed using a FACScan (Becton labeling procedure. For hybridization, 250–400 ng of tumor Dickinson, Heidelberg, Germany) equipped with the CELL- and reference DNA were coprecipitated in the presence of Quest software.

Oncogene Chromosome transfer and gene expression analysis S Seitz et al 878 Preparation of cRNA and GeneChip hybridization amide gel electrophoresis and assessed by quantitative densitometry of ethidium bromide-stained bands with the use Biotin-labeled cRNA was synthesized from 5 mg of total RNA of Gel Doc 2000 system and the Quantity One 4.01 software isolated from each cell line at two distinct passage numbers. (Biorad, Munich, Germany). b-Microglobulin (B2M), Double-stranded cDNA was synthesized using the cDNA GAPDH and PBGD were used to confirm cDNA integrity Synthesis System (Roche diagnostics GmbH) and Super- and normalization of cDNA yields. All reactions were ScripttIIRNaseHÀ Reverse Transcriptase (Invitrogen, Gro- repeated at least twice. ningen, The Netherlands). In vitro transcription was carried out in the presence of biotinylated UTP and CTP (Perkin Elmer, Boston, MA, USA) by using MEGAscript T7 in vitro Statistical analyses Transcription Kit (Ambion, Cambridgeshire, UK). In the case Statistical analysis was carried out using the SPSS package of tissue samples, cRNA was amplified from 1 to 2 mg of total release 11.5. Gene expression was investigated for associations RNA using the Small Sample Labeling Protokoll vII with tumor size, node-status, grading, age at diagnosis and (Affymetrix, Santa Clara, CA, USA) with minor modifica- LOH at 8p21 using the w2 or Fisher’s exact test. Differences in tions. Biotin-labeled cRNA was purified using RNeasy spin the frequencies of differentially expressed genes among gene column (QIAGEN), its quality assessed by gel electrophoresis, sets were analysed using w2 test. For all statistical tests, a and fragmented. Hybridization and scanning of the oligonu- significance level of 0.05 (5%) was used. cleotide microarray ((Human Genome U133A Array, 22 283 probe sets); (Affymetrix)) were performed according to the manufacturer’s protocols. Chip data sets were analysed with Detection of biological relationships between differentially the Affymetrix GeneChip analysis software (Microarray Suite expressed genes Version 5.0) and exported to Microsoft EXCEL. The criteria Two PubGene Expression Analysis tools, Subset Network we called a gene ‘differentially expressed’ are as follows: a (http://www.pubgene.org/tools/Network/Subset.cgi) and the difference call of ‘increase’ or ‘decrease’, a log ratio of X1.0 or Expression Data Analysis tool (http://www.pubgene.org/tools/ pÀ1.0 (fold difference of X2.0 or pÀ2.0) and an absolute ExpressionData/Analysis.cgi) were used to search for litera- call of ‘present’ in either the baseline file or the experimental ture gene-networks in our set of differentially expressed genes. file or both. The following parameter settings were used: ‘Score-depth’: 3, ‘Size of score neighborhood’: 25, ‘Score criteria’: up- or Semiquantitative RT–PCR analysis downregulation, ‘Calculation scheme’: by gene association. In total, 800 ng RNA was subjected to random-primed reverse transcription in a 40 ml PCR reaction using Super- Acknowledgements ScripttIIRNaseHÀ Reverse Transcriptase (Invitrogen). PCR We thank M-F Santibanez-Koref for helpful comments reactions were carried out in a 22 ml reaction containing 1–3 ml on this manuscript. We are indebted to K Poppe, S Werner, of the RT–PCR reaction, Taq PCR Master Mix (QIAGEN) J Strissel, P Wamuth for technical assistance and J Fischer and 5 pmol of each primer. Primer sequences are available for performing statistical analyses. This work was partially upon request. The PCR products were separated by acryl- supported by the Deutsche Krebshilfe Grant 70-2701.

References

Arnold N, Ha¨ gele L, Walz L, Schempp W, Pﬁsterer J, Khatri P, Draghici S, Ostermeier GC and Krawetz SA. (2002). Bauknecht T and Kiechle M. (1996). Genes Chromosomes Genomics, 79, 266–270. Cancer, 16, 46–54. Kim B-T, Kitagawa H, Tamura J-I, Saito T, Kusche-Gullberg Behrens P, Jeske W, Wernet N and Wellmann A. (2001). M, Lindahl U and Sugahara K. (2001). Proc. Natl. Acad. Pathobiology, 69, 19–23. Sci. USA, 98, 7176–7181. Borg A, Ferno M and Peterson C. (2003). Nat. Med., 9, 16–18. Kugoh H, Nakamoto H, Inoue J, Funaki K, Barrett JC and Goldberg EK, Glendening M, Karanjawala Z, Shridhar A, Oshimura M. (2002). Mol. Carcinogen., 35, 148–156. Walker GS, Hayward NK, Rice AJ, Kurera D, Tebha Y and Lerebours F and Lidereau R. (2002). Crit. Rev. Oncol. Hemat., Fountain JW. (2000). Am. J. Hum. Genet., 67, 417–431. 44, 121–141. Gudas JM, Payton M, Thukral S, Chen E, Bass M, Robinson Logan GJ, Smyth CM, Earl JW, Zaikina I, Rowe PB, MO and Coats S. (1999). Mol. Cell. Biol., 19, 12–22. Smythe JA and Alexander IE. (2002). Immunology, 28, Gustafson CE, Wilson PJ, Lukeis R, Baker E, Woollatt E, 105478–105487. Annab L, Hawke L, Barrett JC and Chenevix-Trench G. Murakami Y. (2002). Oncogene, 21, 6936–6948. (1996). Cancer Res., 56, 5238–5245. Nihei N, Kouprina N, LarionovV, Oshima J, Martin GM, Hogue DA, Clines G, Lovett M, Hansen MF and Hecht JT. Ichikawa T and Barrett JC. (2002). Cancer Res., 62, (1996). Am. J. Hum. Genet., 56, A15–A21. 367–370. Huang E, Cheng SH, Dressman H, Pittman J, Tsou MH, Payton M, Scully S, Chung G and Coats S. (2002). Oncogene, Horng CF, Bild A, Iversen ES, Liao M, Chen CM, West M, 21, 8529–8534. Nevins JR and Huang AT. (2003). Lancet, 10, 1590–1596. Primakoff D and Myles DG. (2000). Trends Genet., 16, 83–87. Jenssen TK, Laegreid A, Komorowski J and Hovig E. (2001). Pucci S, Bonanno E, Pichiorri F, Angeloni C and Spagnoli Nat. Genet., 28, 21–28. LG. (2004). Oncogene, 23, 2298–2304. Jones SE and Jomary C. (2002). Int. J. Biochem. Cell Biol., 34, Redondo M, Villar E, Torres-Munoz J, Tellez T, Morell M 427–431. and Petito CK. (2000). Am. J. Pathol., 157, 393–399. Kang Y, Siegel PM, Shu W, Drobnjak M, Kakonen SM, Rennstam K, Ahlstedt-Soini M, Baldetorp B, Bendahl PO, Cordon-Cardo C, Guise TA and Massague J. (2003). Cancer Borg A, Karhu R, Tanner M, Tirkkonen M and Isola J. Cell, 3, 537–549. (2003). Cancer Res., 63, 8861–8868.

Oncogene Chromosome transfer and gene expression analysis S Seitz et al 879 Saffer H, Wahed A, Rassidakis GZ and Medeiros LJ. (2002). Satomi S, Fukushige S, Horii A and Sato T. (1999). Int. J. Mod. Pathol., 15, 1221–1226. Oncol., 15, 443–451. Sambrook J, Fritsch EF and Maniatis T. (1989). Molecular Theile M, Hartmann S, Naundorf H, Rue D, Elbe B, Krause Cloning: A Laboratory Manual. Cold Spring Harbor H, Deppert W, Barrett JC and Scherneck S. (1994). Int. J. Laboratory Press: New York. Oncol., 4, 1067–1075. Scaltriti M, Brausi M, Amorosi A, Caporali A, D’Arca D, Trougakos IP and Gonos ES. (2002). Int. J. Biochem. Cell Astancolle S, Corti A and Bettuzzi S. (2004). Int. J. Cancer., Biol., 34, 1430–1448. 108, 23–30. Tsuneizumi M, Emi M, Hirano A, Utada Y, Tsumagari K, Schlitt T, Palin K, Runa J, Dietmann S, Lappe M, Ukkonen E Takahashi K, Kasumi F, Akiyama F, Sakomoto G, Kazui T and Brazma A. (2003). Genome Res., 13, 2568–2576. and Nakamura Y. (2002). Cancer Lett., 180, 75–82. Seitz S, Rohde K, Bender E, Nothnagel A, Pidde H, Ulrich O- van’t Veer LJ, Dai H, van de Vijver MJ, He YD, Hart AA, M, El-Zehairy A, Haensch W, Jandrig B, Koelble K, Mao M, Peterse HL, van der Kooy K, Marton MJ, Schlag PM and Scherneck S. (1997). Br. J. Cancer, 76, Witteveen AT, Schreiber GJ, Kerkhoven RM, Roberts C, 983–991. Linsley PS, Bernards R and Friend SH. (2002). Nature, 415, Seitz S, Wassmuth P, Plaschke J, Schackert HK, Karsten U, 530–536. Santibanez-Koref MF, Schlag PM and Scherneck S. (2003). Weimer J, Kiechle M and Arnold N. (2000). Cytogenet. Cell Genes Chromosomes Cancer, 37, 29–35. Genet., 88, 114–118. Seitz S, Werner S, Fischer J, Nothnagel A, Schlag PM and Weimer J, Kiechle M, Senger G, Wiedemann U, Ovens-Raeder Scherneck S. (2000). Eur. J. Cancer, 36, 1507–1513. A, Schuierer S, Kautza M, Siebert R and Arnold N. (1999). Singh B, Ittmann MM and Krolewski JJ. (1998). Genes Chromosome Res., 7, 355–362. Chromosomes Cancer, 21, 166–171. Wilson P, Cuthbert A, Marsh A, Arnold J, Flanagan J, Steinberg J, Oyasu R, Lang S, Sintich S, Rademaker A, Malford C, Trott D, Baker E, Purdie D, Newbold R and Lee C, Kozlowski JM and Lee C. (1997). Clin. Cancer Res., Chenevix-Trench G. (2003). Cancer Genet. Cytogenet., 143, 3, 1807–1811. 100–112. Strezoska Z, PestovDG and Lau LF. (2002). J. Biol. Chem., Yaremko ML, Kutza C, Lyzak J, Mick R, Recan WM and 277, 29617–29625. Westbrook CA. (1996). Genes Chromosomes Cancer, 16, Suzuki A, Shao X, Song X-Q, Hanaoka T, Lrie S, Kashinwada 189–195. M, Samara G, Close LG, Aoki T, Fujimori M, Ishikawa Y, Zhang RD, Fidler IJ and Price JE. (1991). Invas. Metast., 11, Hatori M, Hosaka M, Sakurada A, Sato M, Ohuchi N, 204–215.

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

Oncogene