Clinical Implications of Gene Dosage and Gene Expression Patterns in Diploid Breast Carcinoma

Published OnlineFirst June 15, 2010; DOI: 10.1158/1078-0432.CCR-10-0889

Clinical Human Cancer Biology Cancer Research Clinical Implications of Gene Dosage and Gene Expression Patterns in Diploid Breast Carcinoma

Toshima Z. Parris1, Anna Danielsson1, Szilárd Nemes1, Anikó Kovács2, Ulla Delle1, Ghita Fallenius1, Elin Möllerström1, Per Karlsson1, and Khalil Helou1

Abstract Purpose: Deregulation of key cellular pathways is fundamental for the survival and expansion of neoplastic cells. In cancer, regulation of gene transcription can be mediated in a variety of ways. The purpose of this study was to assess the impact of gene dosage on gene expression patterns and the effect of other mechanisms on transcriptional levels, and to associate these genomic changes with clinicopathologic parameters. Experimental Design: We screened 97 invasive diploid breast tumors for DNA copy number alterations and changes in transcriptional levels using array comparative genomic hybridization and expression microarrays, respectively. Results: The integrative analysis identified an increase in the overall number of genetic alterations dur- ing tumor progression and 15 specific genomic regions with aberrant DNA copy numbers in at least 25% of the patient population, i.e., 1q22, 1q22-q23.1, 1q25.3, 1q32.1, 1q32.1-q32.2, 8q21.2-q21.3, 8q22.3, 8q24.3, and 16p11.2 were recurrently gained, whereas 11q25, 16q21, 16q23.3, and 17p12 were fre- quently lost (P < 0.01). An examination of the expression patterns of genes mapping within the detected genetic aberrations identified 47 unique genes and 1 Unigene cluster significantly correlated between the DNA and relative mRNA levels. In addition, more malignant tumors with normal gene dosage levels displayed a recurrent overexpression of UBE2C, S100A8,andCBX2, and downregulation of LOC389033, STC2, DNALI1, SCUBE2, NME5, SUSD3, SERPINA11, AZGP1, and PIP. Conclusions: Taken together, our findings suggest that the dysregulated genes identified here are critical for breast cancer initiation and progression, and could be used as novel therapeutic targets for drug development to complement classical clinicopathologic features. Clin Cancer Res; 16(15); 3860–74. ©2010 AACR.

Copy number alterations (CNA) are a fundamental fea- However, tumorigenesis is a dynamic evolutionary pro- ture of neoplastic cells that influence crucial cancer-specific cess that promotes genetic heterogeneity and thereby pro- processes (1). It is estimated that >1% of protein-coding duces a complex combination of random and nonrandom genes in the human genome contribute to tumorigenesis, aberrations. To study complex phenotypes such as cancer, whereas only 3% of these have been implicated in sporad- a straightforward approach is to provide comprehensive ic breast carcinoma (2). In comparison with many tumor information on the disease by integrating multiple plat- types of nonepithelial origin, breast carcinomas arising forms on several biological levels (DNA-RNA-protein). In- from epithelial cells have numerous genetic alterations tegrative analyses of data produced by genome-wide (3). Many of the recurrent alterations occurring in carcino- profiling techniques such as array comparative genomic mas of the breast have been identified, including copy hybridization (array-CGH; ref. 5) and expression micro- number gains on 1q, 8p, 8q, 11q, 16p, 17q, and 20q, arrays can provide detailed information on the genomic and losses on 1p, 6q, 11q, 16q, 17p, and 22q (4). locations of recurrently altered DNA regions, the impact of these CNAs on gene deregulation, as well as enhance Authors' Affiliations: 1Department of Oncology, Institute of Clinical our understanding of these genetic events coupled with Sciences, and 2Laboratory of Clinical Pathology and Cytology, clinicopathologic parameters (6, 7). However, there are Sahlgrenska Academy at University of Gothenburg, Gothenburg, Sweden limitations associated with the array-CGH approach as it Note: Supplementary data for this article are available at Clinical Cancer will not detect inversions, balanced translocations, poly- Research Online (http://clincancerres.aacrjournals.org/). ploidy, mutations, or epigenetic modulations. It is, there- Corresponding Author: Toshima Z. Parris, Department of Oncology, Sahlgrenska Academy at University of Gothenburg, Gula stråket 2, fore, necessary to also study changes in gene expression SE-41345 Gothenburg, Sweden. Phone: 46-31-3427855; Fax: 46-31- patterns independent of CNA. 820114; E-mail: [email protected]. In the present investigation, we chose to conduct ge- doi: 10.1158/1078-0432.CCR-10-0889 nome-wide screening on a series of diploid breast carcino- ©2010 American Association for Cancer Research. mas (DBC). A previously suggested theory postulated that

3860 Clin Cancer Res; 16(15) August 1, 2010

Integrative Genomics in Diploid Breast Cancer

Materials and Methods Translational Relevance Patient and tumor specimens This study provides novel information on the spe- Primary invasive tumors from 97 DBC patients were cific DNA copy number aberrations and underlying selected from the fresh-frozen tissue tumor bank at changes in gene expression associated with clinico- the Sahlgrenska University Hospital Oncology Lab pathologic features in diploid breast carcinoma. Using (Gothenburg, Sweden). Patient stratification according an integrative genomics approach, we identified tran- to axillary lymph node status (node-negative, pN ;node- scripts whose expression patterns were directly im- 0 positive, pN1) and overall survival (>8-year survivors, pacted by gene dosage. In addition, the transcriptional henceforth termed long-term survivors; breast cancer– levels of 12 genes were consistently associated with a specific mortality within 8 years of diagnosis, henceforth more malignant phenotype in diploid breast carcino- termed short-term survivors) is shown in Supplementary ma. Integrating data on DNA and mRNA dysregulation Table S1. All samples were assessed for DNA content at in relation to breast cancer behavior is, therefore, a step the time of diagnosis from 1991 to 1999 (data not shown) toward understanding tumor initiation and progression by flow cytometry at the Laboratory for Clinical Chemistry, contributing to unfavorable prognosis. Furthermore, Sahlgrenska University Hospital. Tumor specimens with a these findings provide potential targets in the develop- DNA index of 1.0 were classified as diploid. The presence ment of individualized patient therapy for aggressive of malignant cells was assessed in all samples by evaluation breast neoplasms. of touch preparation imprints stained with May-Grünwald Giemsa (Chemicon). Highly representative specimens possessing >70% neoplastic cell content were included in the microarray and fluorescence in situ hybridization (FISH) the aneuploid state is not fundamental to malignancy as analyses. All procedures were done in accordance with the diploid and aneuploid breast tumors share many of the Declaration of Helsinki and approved by the Medical Fac- acquired genetic aberrations characteristic of the disease, ulty Research Ethics Committee (Gothenburg, Sweden). despite the increase in number of events in aneuploid dis- The clinicopathologic features of the 97 cases are shown ease (8–14). Approximately 1 in 4 diagnosed breast carci- in Table 1. noma cases have a diploid DNA content (15). Generally, diploid tumors are composed of a nearly homogenous Array-CGH mix of slow-growing cells that may permit these neo- Whole-genome tiling arrays with 38,043 reporters map- plasms to follow a more favorable clinical course than an- ping to the UCSC May 2004 hg17: NCBI Build 35 were euploid tumors (16, 17). The more aggressive phenotype manufactured as previously described (19) at the SCIBLU displayed by aneuploid tumors can possibly be explained Genomics DNA Microarray Resource Center (SCIBLU), by clonal heterogeneity represented by the presence of Department of Oncology, Lund University. Images and fewer cells expressing the steroid hormone receptors, in- raw signal intensities were acquired using an Agilent creased proliferative activity, and a higher percentage of G2505B DNA microarray scanner (Agilent Technologies) poorly differentiated tumors (18). Taken together, these and GenePix Pro 6.0.1.22 (Axon Instruments) image anal- observations suggest that despite diversity there are similar ysis software. molecular mechanisms active in the development of Data preprocessing and normalization were done using breast carcinomas, regardless of ploidy status. Thus, it is the web-based BioArray Software Environment system of particular interest to conduct integrative genomics using (BASE) provided by SCIBLU (ref. 20; http://base2.thep. diploid breast carcinomas to mitigate some of the obsta- lu.se/onk/). Further analysis to segment the data into cles associated with the heterogeneity of genetic alterations regions of gains and losses was done using the Rank in carcinomas. Segmentation algorithm with Nexus Copy Number Profes- Therefore, we investigated the copy number– and non- sional 4.1 software (BioDiscovery), as previously de- CNA-induced gene expression patterns of DBC from 97 scribed (21). Further, log2ratio thresholds for low-level patients using 38K BAC array-CGH and the Illumina gain (LLG), high-level gain/amplification (HLG), hetero- HumanHT-12 platform. With the application of this ap- zygous loss (HL), and homozygous deletion (HD) were proach on a series of DBC we were able to reduce biological set at +0.2, ≥+0.5, -0.2, and ≤-1.0, respectively, using a noise and thereby identify key putative candidate genes P value cutoff of 0.01. Genomic regions covered entirely associated with an aggressive phenotype in breast carcino- by previously reported copy number variations in the ma. Additionally, we were able to develop an approach to human genome were removed (22). Unsupervised hierar- (a) test the relationship between gene dosage and mRNA chical clustering was applied using complete linkage with levels, (b) determine the proportional or nonproportional Pearson correlation to group tumors based on detected gain/loss in CNA and gene expression patterns, and (c) de- genetic alterations (LLG and HL). The association between termine if a significant difference in expression levels exists clinicopathologic features and CNAs was analyzed using between neoplasms possessing a given genetic aberration a P value cutoff of 0.05 with the Mann-Whitney U or and those lacking these aberrations. Kolmogorov-Smirnov test, as appropriate.

www.aacrjournals.org Clin Cancer Res; 16(15) August 1, 2010 3861

Parris et al.

Table 1. Clinicopathologic characteristics of 97 patients with diploid breast carcinoma

No. of patients (%)

Characteristic pN0 long-term pN1 long-term pN0 short-term pN1 short-term survivors (n = 25) survivors (n = 25) survivors (n = 22) survivors (n = 25)

Mean age in y (range) 58 (40-78) 57 (33-77) 62 (36-79) 51 (30-78) Histologic type Ductal 18 (72) 22 (88) 18 (82) 13 (52) Lobular 4 (16) 1 (4) 1 (5) 5 (20) Other 3 (12) 2 (8) 3 (1) 7 (28) Pathologic tumor size

pT1 12 (48) 5 (20) 5 (9) 4 (16)

pT2 9 (36) 15 (60) 15 (68) 12 (48)

pT3 4 (16) 4 (16) 2 (9) 5 (20)

pT4 0 (0) 1 (4) 0 (0) 4 (16) SBR grade I 7 (28) 4 (16) 1 (5) 1 (4) II 15 (60) 12 (48) 9 (41) 12 (48) III 0 (0) 5 (20) 6 (27) 6 (24) Not available 3 (12) 4 (16) 6 (27) 6 (24) GGI Low 14 (56) 14 (56) 7 (32) 4 (16) High 9 (36) 9 (36) 9 (41) 15 (60) Not available 2 (8) 2 (8) 6 (27) 6 (24) No. of positive axillary lymph nodes 0 25 (100) 0 (0) 22 (100) 0 (0) 1-3 0 (0) 18 (72) 0 (0) 5 (20) ≥4 0 (0) 7 (28) 0 (0) 20 (80) Neoadjuvant therapy Yes 0 (0) 0 (0) 1 (5) 1 (4) No 5 (20) 6 (24) 3 (14) 5 (20) Not available 20 (80) 19 (76) 18 (82) 19 (76) Surgery Lumpectomy 12 (48) 8 (32) 11 (50) 4 (16) Mastectomy 11 (44) 16 (64) 9 (41) 19 (76) Not available 2 (8) 1 (4) 2 (9) 2 (8) Chemotherapy Yes 1 (4) 13 (52) 1 (5) 15 (60) No 20 (80) 8 (32) 18 (82) 8 (32) Not available 4 (16) 4 (16) 3 (14) 2 (8) Endocrine therapy Yes 6 (24) 17 (68) 9 (41) 12 (48) No 16 (64) 3 (12) 10 (45) 10 (40) Not available 3 (12) 5 (20) 3 (14) 3 (12) Radiotherapy Yes 10 (40) 6 (24) 7 (32) 10 (40) No 13 (52) 15 (60) 12 (55) 12 (48) Not available 2 (8) 4 (16) 3 (14) 3 (12) Estrogen receptor Negative 2 (8) 2 (8) 6 (27) 6 (24) Positive 23 (92) 23 (92) 16 (73) 19 (76) Progesterone receptor Negative 9 (36) 4 (16) 14 (64) 10 (40) Positive 16 (64) 21 (84) 8 (36) 15 (60)

(Continued on the following page)

3862 Clin Cancer Res; 16(15) August 1, 2010 Clinical Cancer Research

Integrative Genomics in Diploid Breast Cancer

Table 1. Clinicopathologic characteristics of 97 patients with diploid breast carcinoma (Cont'd)

No. of patients (%)

Characteristic pN0 long-term pN1 long-term pN0 short-term pN1 short-term survivors (n = 25) survivors (n = 25) survivors (n = 22) survivors (n = 25)

HER2 status Negative 23 (92) 22 (88) 18 (82) 23 (92) Positive 2 (8) 3 (12) 4 (18) 2 (8) Triple-negative status Negative 23 (92) 25 (100) 16 (73) 22 (88) Positive 2 (8) 0 (0) 6 (27) 3 (12)

Verification of HER2/neu gene amplification stratified into the molecular breast cancer subtypes using using FISH the five centroids [normal-like, basal-like, luminal subtype Dual-color FISH was done to confirm HER2/neu gene A, luminal subtype B, and human epidermal growth factor amplification in tumors with positive (log2ratio ≥+0.5) receptor 2/estrogen receptor negative (HER2/ER-)] and array-CGH results (11%). Touch preparation imprints genomic grade index (GGI; low, high) using estrogen were prepared from fresh cuts of frozen tumor samples receptor–positive tumors as previously described (25– on Superfrost Plus microscope slides (Erie Scientific Com- 27). All microarray data discussed in this publication are pany). Probe hybridization and posthybridization washes accessible through the National Center for Biotechnology were done according to the manufacturer's guidelines Information (NCBI) Gene Expression Omnibus (GEO ac- using the ZytoLight SPEC HER2/CEN 17 Dual Color Probe cession number GSE20486; http://www.ncbi.nlm.nih.gov/ (ZytoVision GmbH). The slides were evaluated using a geo/query/acc.cgi?acc=GSE20486). Leica DMRA2 fluorescence microscope (Leica) equipped with an ORCA Hamamatsu charged-couple devices camera Quantitative real-time PCR (Hamamatsu Corporation). Digitalized black-and-white Validation of the expression microarray data was done as images were captured at ×100 magnification using the previously described using quantitative real-time PCR Leica CW4000 software package (Leica Microsystems (qPCR) with predesigned TaqMan Gene Expression Assays Imaging Solutions Ltd). HLG was classified in cells with (Applied Biosystems; ref. 28). In brief, 82 of 97 total RNA ≥2.5-fold the centromere count. samples were used to validate the expression patterns of 16 transcripts and 3 endogenous controls, i.e., PPIA, PUM1, Gene expression analysis and HPRT1. The endogenous controls were selected based Total RNA from the same extraction was used for both on their constitutive expression using the Illumina expression profiling and subsequent validation with quan- HumanHT-12 platform. The geometric mean of the three titative real-time PCR. The RNA samples were processed at endogenous controls was used to normalize the data, SCIBLU using Illumina HumanHT-12 Whole-Genome Ex- and relative gene expression levels were calculated with pression BeadChips (Illumina), according to the manufac- the relative standard curve method. Student's t-test was turer's instructions. The expression microarrays contained used to determine the difference in expression between approximately 49,000 probes representing >25,400 RefSeq studied groups, and the Spearman correlation coefficients (Build 36.2, Release 22) and Unigene (Build 199) anno- (two-tailed) were calculated to establish the relationship tated genes. Images and raw signal intensities were ac- between microarray and qPCR expression patterns. quired using the Illumina BeadArray Reader scanner and BeadScan 3.5.31.17122 (Illumina) image analysis soft- Integrated DNA copy number and expression analysis ware, respectively. Illumina HumanHT-12 probe nucleotide sequences Data preprocessing and quantile normalization were ap- were mapped to genomic locations (NCBI Build 35) using plied to the raw signal intensities using BASE. Further data sequences downloaded from the UCSC Genome Browser processing was done in Nexus Expression 2.0 (BioDiscov- (ref. 29; http://hgdownload.cse.ucsc.edu/goldenPath/ ery) using log2-transformed, normalized expression values hg17/chromosomes/). A pairwise comparison of the Illu- and a variance filter. Normalized values from five normal mina probe and BAC clone nucleotide sequences was then breast samples profiled with Illumina HumanWG-6 Ex- conducted to generate Illumina-BAC probe pairs with pression Beadchips (GEO, accession number GSE17072) 100% sequence similarity. Illumina-BAC probes spanning were used as reference (23). Differentially expressed genes the recurrent aberrations were selected from smoothed were determined using the Benjamini-Hochberg method array-CGH data using log2ratio ± 0.2. CNA-induced genes (24) to control for the false discovery rate (FDR) with were assessed by Pearson correlation (Q< 0.05). Further, FDR-corrected P values <0.01. The diploid dataset was the change in mRNA levels as a function of DNA copy

www.aacrjournals.org Clin Cancer Res; 16(15) August 1, 2010 3863

Parris et al.

numbers was estimated by robust piecewise linear regres- pathologic features are presented in Supplementary sion. Regression slopes significantly >1, i.e., positive allo- Table S2. On average, 40.1 ± 3.9 chromosomal aberrations metric change, indicated higher gene expression levels in (range, 2-204) were detected per diploid tumor sample, in- comparison with CNA levels; the opposite was true for cluding 20.1 ± 2.0 LLG (range, 0-117), 4.9 ± 0.6 HLG slopes significantly <1, i.e., negative allometric change; a (range, 0-31), 15.1 ± 1.7 HL (range, 0-104), and 0.03 ± slope of 1 indicated a proportional gain/loss in CNA 0.02 HD (range, 0-1). Recurrent regions of gains and losses and gene expression, i.e. isometric change. Lastly, the dif- observed with a frequency of ≥25% (P < 0.01) were identi- ference between the mean relative mRNA values for tu- fied using the Rank Segmentation algorithm in Nexus Copy mors containing either gain/loss versus tumors without Number 4.1 (Fig. 1A). LLG was observed on nine chromo- these aberrations was estimated with t-test for Illumina- somal subregions on 1q22 (53%), 1q22-q23.1 (53%), BAC probes showing significant association between 1q25.3 (53%), 1q32.1 (53%), 1q32.1-q32.2 (53%), CNA and expression patterns. Statistical analyses were 8q21.2-q21.3 (33%), 8q22.3 (26%), 8q24.3 (26%), and done in R/Bioconductor. The methods are described in 16p11.2 (26%). Loss was observed on six subregions on more detail in Supplementary Materials and Methods. 11q25 (26%), 16q21 (38%), 16q23.3 (38%), and 17p12 (26%). Recurrent regions of HLG in at least 10% (P < Results 0.01) of the cases were identified on three subregions on 1q32.1 (20%), 1q32.2 (20%), and 11q13.3-q13.4 (10%). Overview of chromosomal aberrations detected by In addition, several amplicons were detected in <10% of the array-CGH in the diploid patient population patient population, i.e., 3q26.2 (5%), 8p11-p12 (6%), 8q Ninety-seven diploid breast tumors were analyzed for (5-9%), 17q (5-9%), and 20q13 (6%; Fig. 1B). These find- DNA copy number alterations using 38K BAC CGH arrays. ings are consistent with published array-CGH data on Chromosomal aberrations were detected in all tumors, in- breast carcinoma (9, 30–35) and are listed in more detail cluding all human autosomes and the X chromosome. in Table 2. These segments varied in size from small aberrations to Unsupervised clustering of the dataset according to the gain and/or loss of whole chromosomes (range, 103 kb detected LLG and HL in each tumor produced two distinct to 109 Mb; mean, 6.6 Mb). The mean number of DNA groups. Cluster 1 (n = 21) contained tumors displaying copy number alterations (± SEM) corresponding to clinico- very few aberrations in select portions of the genome

Fig. 1. Genome-wide frequency plots of gains and losses in 97 DBC samples. X-axis, genomic region from chromosome 1 to X; Y-axis, percentage of gains and losses. Green, dark green, and red bars, percentage of LLG, HLG, and HL in the given chromosomal region, respectively. A, the frequency of LLG, HLG, and HL in DBC. B, schematic overview of HLG in DBC. Statistically significant regions (P < 0.01) present in ≥10% of DBC samples are indicated with black arrows showing candidate genes located in the region. Red arrows, previously reported genomic regions in breast cancer, present in <10% of DBC samples. C, top, green and red, statistically significant regions (≥25% difference). Bottom, frequency of CNAs for the two clusters generated by unsupervised hierarchical clustering using complete linkage with Pearson correlation.

3864 Clin Cancer Res; 16(15) August 1, 2010 Clinical Cancer Research

Integrative Genomics in Diploid Breast Cancer

including exclusion of 1q gain, which is common in most than long-term survivors with pN0 tumors; LLG of the breast neoplasms, whereas cluster 2 (n = 76) contained tumors 6p25.3 and 17q12-q21.2 regions was more prevalent in with an abundance of genetic aberrations covering the entire long-term survivors with four or more PALN than in genome (Fig. 1C). In addition, the former group lacked genetic long-term survivors with pN0 tumors; lastly, LLG of alterations on chromosomes 3, 18, and 20, and loss of chro- 17q12-q21.2 was more prevalent in long-term survivors mosome 6 was more characteristic of this group. In general, with four or more PALN than in long-term survivors cluster 1 consisted predominantly of low-grade tumors (P = with one to three PALN. To validate the array-CGH results, 0.006), low GGI (P = 0.021), and long-term survivors (P = dual-color interphase FISH was done on tumors display- 0.034), whereas cluster 2 consisted predominantly of tumors ing HLG of the HER2/neu gene using the ZytoLight SPEC from short-term survivors (P = 0.034), high GGI (P = 0.035), HER2/CEN 17 Dual Color Probe. We observed a 100% short-term survivors with pN1 disease (P = 0.044), and tumors concordance between array-CGH and FISH for the assess- with gain of the 8p11-p12 locus (P = 0.044). ment of HER2/neu amplification (Fig. 2).

Correlation of DNA copy number alterations to Gene expression profiling clinicopathologic features The numbers of deregulated transcripts in each group To determine if clinicopathologic features are character- comparison identified by transcriptome profiling of the ized by specific genetic alterations, recurrently altered DBC samples, as well as Gene Ontology biological pro- genomic regions were compiled and are shown in Supple- cesses associated with these genes, are shown in Table 3. mentary Table S3. A high number of genetic alterations A large number of transcripts were deregulated 1.5-fold were observed in tumors with high GGI (P = 0.0004), based on short-term overall survival, endocrine insensitiv- tumors from short-term survivors (P = 0.001), cluster 2 ity, high GGI, and triple-negative status. Similar to results (P = 0.001), and high-grade tumors (P = 0.007), in compar- on the DNA level, there were few differences between tu- ison with their corresponding counterparts. Specifically, a mors stratified according to axillary lymph node status, as higher number of LLG (P = 0.001), HLG (P = 0.005), and only the CNTNAP2 gene was differentially expressed. Sim- HL (P = 0.005) were detected in short-term survivors com- ilarly, few transcripts differed statistically between tumors pared with long-term survivors, suggesting that the number of varying Scarff-Bloom-Richardson (SBR) grade, implying of genetic alterations may have a detrimental effect on clin- that there may be some overlap between SBR grade II and ical outcome. Using the Rank Segmentation algorithm and low- and high-grade tumors. One gene (SHISA2) was up- a 25% difference cutoff, HL on 3p14.2, 17p13.1-p12, and regulated in SBR grade I compared with grade II tumors, 18q23 was more prevalent in short-term survivors than in which was also among transcripts differentially regulated long-term survivors. Further, LLG and HLG were generally between SBR grade I and grade III tumors. Six genes were more frequent in high-grade than in low-grade tumors on both lists for SBR grade I versus III and SBR grade II (P = 0.002 and P = 0.01, respectively). All forms of genetic versus III (S100A8, S100A9, CBX2, MAPT, AZGP1, PIP). aberrations, excluding HD, were more prevalent in tumors Lastly, a comparison of the gene expression levels was with high than low GGI. LLG (P = 0.04) was more prevalent done on the two clusters determined using genetic altera- in larger tumors (pT3/pT4) than in smaller tumors (pT1/ tions. In total, five transcripts were differentially regulated pT2). LLG and HLG were generally more frequent in cluster (two upregulated and three downregulated). These results 2 than in cluster 1 (P = 0.00003 and P = 0.02, respectively). show the recurrent upregulation of transcripts associated LLG of the 17q23.3-q24.1 region was detected solely in with a more malignant phenotype (short-term overall sur- high-grade and never in low-grade tumors. With respect vival, endocrine insensitivity, triple-negative status, poor to GGI status, LLG of 1q, 20q and HL of 3p14.2, 8p, 11q, tumor differentiation), i.e., UBE2C, S100A8,andCBX2, 17p12-p11.2, and 18q23 were more prevalent in high GGI whereas LOC389033, STC2, DNALI1, SCUBE2, NME5, tumors. Lastly, LLG on chromosomes 2q, 7p, 12q, 17q, and SUSD3, SERPINA11, AZGP1,andPIP are downregulated 20q was observed predominantly in larger tumors. (Supplementary Table S4). Five of these genes were also In our dataset, there was no distinct difference in the differentially regulated between tumors of high and low number of or type of observed CNAs in tumors with re- GGI, i.e., UBE2C, S100A8, STC2, SCUBE2,andSERPI- gard to their axillary lymph node status. These findings NA11. To validate these results, 16 genes showing >1.5- suggest that pN0 and pN1 tumors are biologically similar fold differential expression levels in either the receptor sta- entities. Biological differences became apparent when tus or overall survival groups were assessed with qPCR. A these tumors were further stratified by either overall sur- linear relationship was found between the Illumina and vival status and/or by the number of positive axillary qPCR results (rS = 0.97; P < 0.01). The results of the vali- lymph nodes (PALN). LLG of 6p21.32-p21.31 and HL of dation experiments are listed in Supplementary Table S5. 10q26.12 were more prevalent in short-term survivors with pN0 tumors than in short-term survivors with pN1 Stratification of the diploid dataset into molecular tumors. Furthermore, heterozygous loss of 8p23.2 was gene expression subtypes and GGI predominantly detected in tumors from patients with four The diploid dataset was stratified into the five molec- or more PALN than pN0 tumors. LLG of 8p12 was more ular breast cancer subtypes by assigning each tumor to prevalent in long-term survivors with one to three PALN one of the five centroids using the Pearson correlation

www.aacrjournals.org Clin Cancer Res; 16(15) August 1, 2010 3865

Downloaded from clincancerres.aacrjournals.org on September 30, 2021. © 2010 American Association for Cancer Research. 3866 arse al. et Parris lnCne e;1(5 uut1 2010 1, August 16(15) Res; Cancer Clin Downloaded from

Table 2. Comparison of recurrent copy number alterations in the current study with previous reports clincancerres.aacrjournals.org Current study (n = 97) Previous studies Published OnlineFirstJune15,2010;DOI:10.1158/1078-0432.CCR-10-0889 Region Size Cytoband Breast cancer-related Number Frequency % of copy number P Cytoband location Frequency Reference (Mb) location candidate genes of genes (%) variations overlap (%)

Low-level gain chr1:151803323- 0.4 1q22 DCST1, ADAM15, 26 53 73 0.009 1q11-q23 25-48 (32, 35) 152187362 EFNA1, MUC1, C1orf2, FDPS chr1:153006542- 1.1 1q22-q23.1 PMF1, CRABP2, 34 53 0 0.009 1q11-q23 48 (35) 154155681 C1orf66, NTRK1, INSRR chr1:178971831- 0.6 1q25.3 RGSL1, RNASEL, 8 53 2 0.009 1q25-q31 24-45 (32, 35) on September 30, 2021. © 2010American Association for Cancer 179547540 RGS16, DHX9 Research. chr1:197639011- 5.2 1q32.1 LAD1, ELF3, 82 53 13 0.009 1q32.1-q32.3 27-66 (32, 34) 202792328 UBE2T, JARID1B, ADIPOR1, ADORA1, MYOG, BTG2, KISS1, MDM4 chr1:202921401- 2.5 1q32.1-q32.2 SRGAP2, IKBKE, 26 53 34 0.009 1q32.1-q32.2 22-66 (32, 34) 205404357 RASSF5, MAPKAPK2, IL10, IL24, CD55, CD34 chr8:86674253- 0.5 8q21.2-q21.3 5 33 98 0.000 8q21.12-q22.3 25 (32) 87135834 chr8:102067884- 0.3 8q22.3 1 26 0 0.000 102376617 chr8:145853807- 0.2 8q24.3 RPL8 7 26 17 0.000 8q24 27-65 (30, 32, 35) lnclCne Research Cancer Clinical 146082345 chr16:31371886- 0.1 16p11.2 TGFB1I1 4 26 0 0.000 16p11.2 12 (32) 31439871

(Continued on the following page) www.aacrjournals.org Downloaded from

Table 2. Comparison of recurrent copy number alterations in the current study with previous reports (Cont'd) clincancerres.aacrjournals.org Published OnlineFirstJune15,2010;DOI:10.1158/1078-0432.CCR-10-0889 Current study (n = 97) Previous studies Region Size Cytoband Breast cancer-related Number Frequency % of copy number P Cytoband location Frequency Reference (Mb) location candidate genes of genes (%) variations overlap (%)

Heterozygous loss chr11:131205533- 0.2 11q25 1 26 24 0.000 11q22.3-q25 14-40 (32, 33) 131405980 chr16:59474046- 2.2 16q21 CDH8 1 38 9 0.004 16q21-q25 20-40 (9, 30, 32, 33) 61647365 chr16:61922081- 1.7 16q21 CDH11 1 38 10 0.004 16q21-q25 20-40 (30, 32–34)

Research. chr16:80651087- 1.8 16q23.3 HSD17B2, CDH13, 4 38 15 0.004 16q21-q25 18-40 (30, 32, 33) 82423783 HSBP1 chr17:11053362- 1.0 17p12 MAP2K4 4 26 7 0.006 17p13-p11 16-40 (9, 32–35) 12022428 chr17:12335195- 3.1 17p12 ELAC2, PMP22, 14 26 56 0.006 17p13-p11 16-40 (9, 32, 33, 35) 15480166 TRIM16 High-level gain/amplification chr1:197688254- 3.6 1q32.1 LAD1, ELF3, UBE2T, 60 20 13 0.001 201299020 JARID1B, ADIPOR1,

ADORA1, MYOG, Cancer Breast Diploid in Genomics Integrative BTG2, KISS1,

lnCne e;1(5 uut1 2010 1, August 16(15) Res; Cancer Clin MDM4 chr1:204387556- 0.6 1q32.2 CD34 2 20 36 0.001 205011692 chr11:69286642- 1.2 11q13.3-q13.4 FGF4, FGF3, 7 10 11 0.000 11q13.3-q13.4 22-77 (9, 31, 32, 35) 70504982 TMEM16A, FADD, PPFIA1, CTTN 3867 Published OnlineFirst June 15, 2010; DOI: 10.1158/1078-0432.CCR-10-0889

Parris et al.

Fig. 2. Schematic overview of HER2/neu gene amplification. A, genomic profile covering chromosome 17 for tumors 8295 (black dots) and 8491 (gray dots).

Green, BAC probes containing the HER2/neu gene; X-axis, array-CGH log2ratio; Y-axis; genomic position along chromosome 17. B, a zoom-in of chromosome 17 spanning from the centromere to the HER2/neu locus. C and D, dual-color FISH hybridization of the ZytoLight SPEC HER2/CEN 17 Dual Color Probe with HER2 (green-labeled fluorochrome) and CEN 17 (orange-labeled fluorochrome) for tumors 8295 and 8491, respectively. Tumor 8295 is shown with amplification of both the centromere and HER2 locus, whereas 8491 is amplified at the HER2 locus only.

3868 Clin Cancer Res; 16(15) August 1, 2010 Clinical Cancer Research

Integrative Genomics in Diploid Breast Cancer

coefficient (25). Subsequently, no tumors (0%) were clas- tumorigenesis via gene deregulation. These results suggest sified as normal-like, whereas 7 (7%) were classified as bas- that despite diversity DBCs contain common genetic al-like, 1 (1%) as luminal subtype A, 78 (80%) as luminal anomalies that may mediate malignant tumor behavior. subtype B, and 11 (11%) as HER2/ER-. As expected, all tu- Furthermore, these observations support the theory that mors classified as both HER2-positive and estrogen recep- aberrant gene expression levels displayed in breast carcino- tor–negative (n = 4) were correctly classified in the HER2/ ma are, in part, induced by CNAs and other mechanisms ER- subtype. The remaining HER2-positive tumors were of transcriptional control (3). either classified in the HER2/ER- (n =1)ortheluminal CNAs are a common feature of genetic instability in subtype B groups (n = 6). In addition, basal-like (86%) breast carcinomas (3). The finding that a progressive in- and HER2/ER- (64%) tumors were predominantly asso- crease in genetic aberrations is seemingly predictive of un- ciated with short-term overall survival; basal-like (67%) favorable clinical outcome has been shown in several and HER2/ER- (73%) tumors were more prevalent in studies (12, 14). In the present study, this pattern was as- cluster 2 than in cluster 1. Eighty-one percent of tumors sociated with tumors exhibiting aggressive behavior, i.e., grouped in cluster 1 were of luminal B subtype. tumors originating from short-term survivors, larger tu- Stratification of the 81 estrogen receptor–positive tumors in mors, poorly differentiated tumors, tumors with high our dataset into the GGI groups produced a slight overlap GGI, and tumors grouped in cluster 2. The presence of re- between the low and high GGI groups, whereas two SBR current genetic alterations at specific loci was also reflective grade I and four grade III tumors were classified with high of an adverse effect on patient clinical outcome. Among the and low GGI, respectively. As expected, SBR grade II tumors recurrently altered regions identified here, transcripts on were a mix of both GGI groups. In general, low GGI was pre- chromosome 1 are of particular interest, as we have shown dominantly associated with long-term overall survival (P = that patients with tumors lacking genetic alterations on 0.008), long-term survivors with one to three PALN (P = chromosomes 1, 3, 18, and 20 have better prognoses. Al- 0.009), SBR grade I (P = 0.009), pT1 (P = 0.012), cluster 1 though CNAs are a fundamental feature of tumorigenesis, (P = 0.02), and one to three PALN (P = 0.025), whereas high we have shown that changes in gene dosage do not neces- GGI was associated with four or more PALN (P = 0.001), sarily relate to the amount of mRNA gene product. short-term survivors with four or more PALN (P = 0.005), To assess the clinical significance of disease-associated short-term survivors with pN1 disease (P = 0.006), short- gene deregulation, we integrated copy number data with term overall survival (P =0.008),pT2 (P = 0.018), and cluster expression patterns. We were, therefore, able to show that 2(P =0.02). only a fraction of the genes altered at the DNA level were actually transcribed at abnormal levels. Relative mRNA Integrated DNA copy number and gene levels of 48 of 360 (13%) transcripts located within the expression analysis 18 common regions of genetic alteration were directly im- To determine the effect of CNAs on gene deregulation in pacted by CNAs, thus further emphasizing the notion that DBC, a correlation analysis was done in three steps using many genetic alterations are bystander mutations that may identified CNAs. These analyses showed that the regulation not reflect biological effect. These findings are consistent of several breast cancer–related candidate genes and genes with previous reports, which showed that the percentage not previously associated with breast cancer were located of overexpressed genes can range from 6% to 44% de- within recurrent CNAs listed in Table 2. In total, 1,161 pending on the level of gene dosage (6, 7, 32). However, Illumina-BAC probe pairs contained nucleotide sequences the coregulation and expression of adjacent genes may within the 18 observed CNAs, of which 149 probe pairs play an important role in tumorigenesis (7). To further (52 unique genes and 3 Unigene clusters) were identified characterize these transcripts, we showed a variation in as having a significant correlation between DNA and rela- the overall rate of gene transcription given a specific copy tive mRNA levels (Supplementary Table S6). For 38 genes number. This may be a reflection of the robust induction the expression levels were as expected given the CNA log2- of gene transcription, as a single gene must bypass several ratio; 3 genes showed higher expression levels and 14 genes checkpoints before it is expressed. We hypothesize that showed lower expression levels than expected considering transcripts with expression levels deviating from expected the CNA log2ratio. Comparison of relative mRNA levels values (higher and lower) given gene dosage may be con- between samples showing no CNAs and gain/loss of trolled by multiple molecular mechanisms including specific genomic regions identified 128 of 149 probe pairs smaller CNAs not detected using BAC arrays. (47 genes and 1 Unigene cluster) with significantly differ- Recurrent chromosomal regions of gain and amplifica- ent expression levels between the two groups. Figure 3 tion are common on 1q in several types of human neo- illustrates an example of the integration analysis for the plasms, ranging from carcinomas of the breast to HER2/neu gene in amplified and nonamplified tumors. sarcomas (36). Here, we show that 40 of 48 (83%) of the transcripts correlating with gene dosage were located Discussion on chromosome 1q consisting of five clusters of adjacent genes. We find ARHGEF11, CLK2, DHX9, EFNA1, RAB7L1, Our integrative genomics approach showed that diploid RABIF, RAG1AP1, RIPK5, RNASEL, RNPEP, and SNRPE of breast neoplasms exploit different strategies to promote particular interest as the elevated mRNA levels of these

www.aacrjournals.org Clin Cancer Res; 16(15) August 1, 2010 3869

Parris et al.

Table 3. Total deregulated transcripts and their involvement in Gene Ontology biological processes

Group comparisons Gene expression Gene Ontology biological process associated with changes differentially regulated genes (1.5-fold change) (number of transcripts) 2-fold 1.5-fold change change Up Down Up Down

Short-term survivors vs. 8 9 47 51 Mitosis, cell division, cell cycle, positive regulation of long-term survivors transcription; DNA-dependent, DNA repair, negative regulation of signal transduction Estrogen receptor positive vs. 122 77 353 287 Response to estradiol stimulus, antiapoptosis, regulation receptor negative of inflammatory response, regulation of cell proliferation, epidermal growth factor receptor (EGFR) signaling pathway, negative regulation of mitotic cell cycle, negative regulation of cell proliferation Progesterone receptor positive 35 6 133 37 Potassium ion and chloride transport, cell aging, response to vs. receptor negative estrogen stimulus, regulation of cell proliferation, antiapoptosis Receptor positive vs. negative 131 77 364 289 Response to estradiol stimulus, regulation of inflammatory response, chloride transport, negative regulation of cell proliferation, negative regulation of mitotic cell cycle

pN1, short-term survivors vs. 4 6 4 8 Cellular metabolic processes, cyclin catabolic process, long-term survivors positive regulation of exit from mitosis, tyrosine catabolic process, endothelial cell migration

pN0, short-term survivors vs. 6 21 17 36 Response to estradiol stimulus, cholesterol efflux and long-term survivors homeostasis, negative regulation of TGFB3 production, regulation of cell-cell adhesion, regulation of blood vessel size by renin-angiotensin, regulation of gene-specific transcription, immune response-regulating cell surface receptor signaling pathway, apoptotic chromosome condensation, regulation of cytokine-mediated signaling pathway, negative regulation of blood vessel endothelial cell migration, regulation of epithelial cell proliferation, regulation of MAPKKK cascade 1-3 positive axillary lymph nodes, 12 1 12 1 Nucleosome positioning and assembly, response to estradiol short-term survivors vs. stimulus, blood vessel morphogenesis, endothelial cell long-term survivors migration, calcium-dependent cell-cell adhesion, cell migration ≥4 positive axillary lymph nodes, 4 11 4 13 Vitamin D metabolic process, tyrosine catabolic process, short-term survivors vs. cellular metabolic process long-term survivors

Short-term survivors, pN0 vs. 1 15 1 15 Nucleosome assembly, response to estradiol stimulus, estrogen 1-3 positive axillary receptor signaling pathway, regulation of transcription; lymph nodes DNA-dependent, Notch signaling pathway Short-term survivors, ≥4 vs. 4 14 4 14 Regulation of blood vessel size by renin-angiotensin, 1-3 positive axillary C21-steroid hormone metabolic process, regulation of lymph nodes vasodilation, positive regulation of inflammatory response, response to toxin Short-term survivors, ≥4 vs. 2 1 2 1 Transmission of nerve impulse, neuron recognition, cellular

pN0 positive axillary lymph alcohol metabolic process, collagen catabolic process nodes HER2 negative vs. positive 3 13 3 15 Induction of apoptosis by oxidative stress, negative regulation of immature T-cell proliferation in the thymus, fibroblast growth factor receptor (FGFR) signaling pathway, positive regulation of mitogen-activated protein kinase activity, positive regulation of cell adhesion, EGFR signaling pathway, positive

(Continued on the following page)

3870 Clin Cancer Res; 16(15) August 1, 2010 Clinical Cancer Research

Integrative Genomics in Diploid Breast Cancer

Table 3. Total deregulated transcripts and their involvement in Gene Ontology biological processes (Cont'd)

regulation of epithelial cell proliferation, cell-cell signaling, mammary gland development pN0 vs. pN1 0 1 0 1 Transmission of nerve impulse, neuron recognition, cell adhesion pN0 vs. ≥4 positive axillary 0 1 0 1 Transmission of nerve impulse, neuron recognition, lymph nodes cell adhesion 1-3 vs. ≥4 positive axillary 0 3 0 3 Induction of apoptosis lymph nodes Long-term survivors, 0 9 0 9 Tyrosine catabolic process, cellular metabolic process, pN0 vs. ≥4 positive axillary biosynthetic process, transport lymph nodes Long-term survivors, 1-3 vs. 0 10 0 10 Response to estradiol stimulus, TGFBR signaling ≥4 positive axillary pathway, transport lymph nodes Cluster 2 vs. cluster 1 1 0 2 3 Regulation of hormone secretion, peptide hormone processing, antigen processing and presentation, neuropeptide signaling pathway, intracellular protein transport SBR grade I vs. II 1 0 1 0 Multicellular organismal development SBR grade I vs. III 26 7 28 10 Cyclin catabolic process, regulation of integrin biosynthetic process, positive regulation of exit from mitosis, regulation of mitosis, cell-cell signaling, tyrosine catabolic process SBR grade II vs. III 6 5 6 7 Regulation of integrin biosynthetic process, positive regulation of microtubule polymerization, response to estradiol stimulus, inflammatory response High vs. low GGI 45 14 181 105 Cell cycle, mitosis, cell division, DNA replication, phosphoinositide-mediated signaling, DNA repair, cell proliferation, regulation of cyclin-dependent protein kinase activity, protein amino acid phosphorylation, response to DNA damage stimulus, cell-cell signaling, cell differentiation, FGFR signaling pathway Triple negative vs. 106 149 294 347 Response to estradiol stimulus, regulation of cytokine non–triple negative biosynthetic process, regulation of inflammatory response, DNA replication checkpoint, positive regulation of interleukin-4 and 8 production, positive regulation of exit from mitosis, anti-apoptosis, cell-cell signaling, mammary gland development, regulation of cell proliferation, negative regulation of tumor necrosis factor production

genes combined with increased copy number may pro- deregulation of 12 genes, i.e., UBE2C, S100A8, CBX2, mote oncogenic activity. Several of these genes regulate LOC389033, STC2, DNALI1, SCUBE2, NME5, SUSD3, similar cellular processes predominantly involved in cell SERPINA11, AZGP1,andPIP, in breast carcinomas with motility, protein amino acid phosphorylation, RNA splic- a malignant phenotype. Functional studies have implicat- ing, and cell-cell signaling. ed several of these genes in a role in tumor cell growth, Alternative cellular processes by which breast neoplastic motility, and progression in multiple cancer types (37– cells mediate abnormal relative mRNA levels in this study 44). Here, we could associate all 12 genes with progester- are unclear. Interestingly, our data suggest the consistent one expression and 11 with estrogen expression (UBE2C

www.aacrjournals.org Clin Cancer Res; 16(15) August 1, 2010 3871

Parris et al.

Fig. 3. Robust piecewise linear regression analysis showing the relationship between gene dosage and transcriptional levels for the HER2/neu gene in

97 DBC samples. The Illumina log2 (ILMN_2352131) and BAC log2ratio (RP11-909L6) were compared using Pearson correlation. Shaded areas, 95% confidence intervals of the regression line.

excluded). Similarly, Gruvberger et al. (45) identified In summary, we have shown the feasibility of using mi- S100A8 and STC2 as estrogen receptor antagonist and re- croarray technology as a tool to identify common biological sponsive genes. Furthermore, Berlingieri et al. (46) revealed features among breast carcinomas that can be used as poten- that suppression of ERBB2 inhibits UBE2C activity and tial therapeutic targets. We identified 18 recurrent regions of thereby cell growth. genetic alteration in our dataset, of which 48 transcripts were Another intriguing finding was the identification of a abnormally expressed. In addition, the expression levels of single transcript (CNTNAP2) differently deregulated be- 12 genes displaying normal copy numbers were associated tween the axillary lymph node groups. This transcript with malignant phenotypes in breast carcinoma. This ap- spans 2.3 Mb of chromosome 7, which makes it the largest proach identified good candidates for further investigations known gene in the human genome, and is also located at a using an independent series of multiple cancer types to as- region containing a common fragile site (FRA7I; refs. 47, sess the biological relevance of elevated DNA and/or mRNA 48). We did not, however, observe loss of the 7q35-q36.1 levels on protein gene product. Taken together, this infor- chromosomal region spanning the CNTNAP2 gene that mation can potentially be used to establish cost-effective could explain downregulation in pN0 tumors. Expression targeted and individualized treatment regimens that will of the CNTNAP2 protein has been primarily implicated in directly benefit a specific patient based on the genetic/ disorders of the vertebrate nervous system for its functions transcriptomic profile of the associated tumor to target as cell adhesion molecules and receptors (49). In cancer, specific cellular pathways perturbed in the tumor without the CNTNAP2 promoter has been shown to be hyper- affecting the activity of nonneoplastic cells. methylated in 82% of pancreatic cancers in comparison with 3% in normal pancreas (50). Hypermethylation of Disclosure of Potential Conflicts of Interest the CNTNAP2 promoter may be one mechanism by which the transcriptional levels of this gene may be silenced. To No potential conflicts of interest were disclosed. determine the methylation status of this gene and to establish its role in breast tumor development and progression, Acknowledgments further methylation studies using breast neoplasms are warranted. We thank Marcela Davila for invaluable bioinformatics support.

3872 Clin Cancer Res; 16(15) August 1, 2010 Clinical Cancer Research

Integrative Genomics in Diploid Breast Cancer

Grant Support The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. Grants from the King Gustav V Jubilee Clinic Cancer Research Founda- tion (K. Helou) and the Wilhelm and Martina Lundgren Research Founda- Received 04/07/2010; revised 05/19/2010; accepted 06/03/2010; tion (T. Parris). published OnlineFirst 06/15/2010.

References 1. Vogelstein B, Kinzler KW. The multistep nature of cancer. Trends 20. Saal LH, Troein C, Vallon-Christersson J, Gruvberger S, Borg A, Genet 1993;9:138–41. Peterson C. BioArray Software Environment (BASE): a platform 2. Futreal PA, Coin L, Marshall M, et al. A census of human cancer for comprehensive management and analysis of microarray data. genes. Nat Rev Cancer 2004;4:177–83. Genome Biol 2002;3, SOFTWARE0003. 3. Albertson DG, Collins C, McCormick F, Gray JW. Chromosome 21. Moore SR, Persons DL, Sosman JA, et al. Detection of copy number aberrations in solid tumors. Nat Genet 2003;34:369–76. alterations in metastatic melanoma by a DNA fluorescence in situ 4. Tsuda H. Gene and chromosomal alterations in sporadic breast hybridization probe panel and array comparative genomic hybridi- cancer: correlation with histopathological features and implications zation: a southwest oncology group study (S9431). Clin Cancer for genesis and progression. Breast Cancer 2009;16:186–201. Res 2008;14:2927–35. 5. Pinkel D, Segraves R, Sudar D, et al. High resolution analysis of DNA 22. Iafrate AJ, Feuk L, Rivera MN, et al. Detection of large-scale variation copy number variation using comparative genomic hybridization to in the human genome. Nat Genet 2004;36:949–51. microarrays. Nat Genet 1998;20:207–11. 23. Lim E, Vaillant F, Wu D, et al. Aberrant luminal progenitors as the 6. Hyman E, Kauraniemi P, Hautaniemi S, et al. Impact of DNA candidate target population for basal tumor development in BRCA1 amplification on gene expression patterns in breast cancer. Cancer mutation carriers. Nat Med 2009;15:907–13. Res 2002;62:6240–5. 24. Benjamini Y, Hochberg Y. Controlling the false discovery rate: a 7. Reyal F, Stransky N, Bernard-Pierrot I, et al. Visualizing chromo- practical and powerful approach to multiple testing. J R Stat Soc somes as transcriptome correlation maps: evidence of chromosomal 1995;57:289–300. domains containing co-expressed genes-a study of 130 invasive 25. Hu Z, Fan C, Oh DS, et al. The molecular portraits of breast ductal breast carcinomas. Cancer Res 2005;65:1376–83. tumors are conserved across microarray platforms. BMC Genomics 8. Farabegoli F, Santini D, Ceccarelli C, Taffurelli M, Marrano D, 2006;7:96. Baldini N. Clone heterogeneity in diploid and aneuploid breast 26. Loi S, Haibe-Kains B, Desmedt C, et al. Definition of clinically distinct carcinomas as detected by FISH. Cytometry 2001;46:50–6. molecular subtypes in estrogen receptor-positive breast carcinomas 9. Hicks J, Krasnitz A, Lakshmi B, et al. Novel patterns of genome through genomic grade. J Clin Oncol 2007;25:1239–46. rearrangement and their association with survival in breast cancer. 27. Sotiriou C, Wirapati P, Loi S, et al. Gene expression profiling in breast Genome Res 2006;16:1465–79. cancer: understanding the molecular basis of histologic grade to 10. Hicks J, Muthuswamy L, Krasnitz A, et al. High-resolution ROMA improve prognosis. J Natl Cancer Inst 2006;98:262–72. CGH and FISH analysis of aneuploid and diploid breast tumors. Cold 28. Karlsson E, Delle U, Danielsson A, et al. Gene expression variation Spring Harb Symp Quant Biol 2005;70:51–63. to predict 10-year survival in lymph-node-negative breast cancer. 11. Rennstam K, Baldetorp B, Kytola S, Tanner M, Isola J. Chromosomal BMC Cancer 2008;8:254. rearrangements and oncogene amplification precede aneuploidi- 29. Kent WJ, Sugnet CW, Furey TS, et al. The human genome browser zation in the genetic evolution of breast cancer. Cancer Res 2001; at UCSC. Genome Res 2002;12:996–1006. 61:1214–9. 30. Andre F, Job B, Dessen P, et al. Molecular characterization of breast 12. Ried T, Heselmeyer-Haddad K, Blegen H, Schrock E, Auer G. cancer with high-resolution oligonucleotide comparative genomic Genomic changes defining the genesis, progression, and malig- hybridization array. Clin Cancer Res 2009;15:441–51. nancy potential in solid human tumors: a phenotype/genotype 31. Fridlyand J, Snijders AM, Ylstra B, et al. Breast tumor copy number correlation. Genes Chromosomes Cancer 1999;25:195–204. aberration phenotypes and genomic instability. BMC Cancer 2006; 13. Tanner MM, Karhu RA, Nupponen NN, et al. Genetic aberrations 6:96. in hypodiploid breast cancer: frequent loss of chromosome 4 and 32. Haverty PM, Fridlyand J, Li L, et al. High-resolution genomic and amplification of cyclin D1 oncogene. Am J Pathol 1998;153:191–9. expression analyses of copy number alterations in breast tumors. 14. Tirkkonen M, Tanner M, Karhu R, Kallioniemi A, Isola J, Genes Chromosomes Cancer 2008;47:530–42. Kallioniemi OP. Molecular cytogenetics of primary breast cancer 33. Loo LW, Grove DI, Williams EM, et al. Array comparative genomic by CGH. Genes Chromosomes Cancer 1998;21:177–84. hybridization analysis of genomic alterations in breast cancer 15. Truong K, Vielh P, Guilly MN, et al. Quantitative FISH analysis on subtypes. Cancer Res 2004;64:8541–9. interphase nuclei may improve diagnosis of DNA diploid breast 34. Naylor TL, Greshock J, Wang Y, et al. High resolution genomic cancers. Diagn Cytopathol 2002;26:213–6. analysis of sporadic breast cancer using array-based comparative 16. Habermann JK, Doering J, Hautaniemi S, et al. The gene expression genomic hybridization. Breast Cancer Res 2005;7:R1186–98. signature of genomic instability in breast cancer is an independent 35. Nessling M, Richter K, Schwaenen C, et al. Candidate genes predictor of clinical outcome. Int J Cancer 2009;124:1552–64. in breast cancer revealed by microarray-based comparative 17. Ried T, Just KE, Holtgreve-Grez H, et al. Comparative genomic genomic hybridization of archived tissue. Cancer Res 2005;65: hybridization of formalin-fixed, paraffin-embedded breast tumors 439–47. reveals different patterns of chromosomal gains and losses in 36. Knuutila S, Bjorkqvist AM, Autio K, et al. DNA copy number fibroadenomas and diploid and aneuploid carcinomas. Cancer Res amplifications in human neoplasms: review of comparative genomic 1995;55:5415–23. hybridization studies. Am J Pathol 1998;152:1107–23. 18. Leers MP, Nap M. Steroid receptor heterogeneity in relation to 37. Arai K, Takano S, Teratani T, Ito Y, Yamada T, Nozawa R. S100A8 DNA index in breast cancer: a multiparameter flow cytometric and S100A9 overexpression is associated with poor pathological approach on paraffin-embedded tumor samples. Breast J 2001;7: parameters in invasive ductal carcinoma of the breast. Curr Cancer 249–59. Drug Targets 2008;8:243–52. 19. Jonsson G, Staaf J, Olsson E, et al. High-resolution genomic profiles 38. Chapman EJ, Kelly G, Knowles MA. Genes involved in differentiation, of breast cancer cell lines assessed by tiling BAC array comparative stem cell renewal, and tumorigenesis are modulated in telomerase- genomic hybridization. Genes Chromosomes Cancer 2007;46: immortalized human urothelial cells. Mol Cancer Res 2008;6: 543–58. 1154–68.

www.aacrjournals.org Clin Cancer Res; 16(15) August 1, 2010 3873

Parris et al.

39. Cheng CJ, Lin YC, Tsai MT, et al. SCUBE2 suppresses breast tumor Nakagawara A. UbcH10 is the cancer-related E2 ubiquitin-conjugating cell proliferation and confers a favorable prognosis in invasive breast enzyme. Cancer Res 2003;63:4167–73. cancer. Cancer Res 2009;69:3634–41. 45. Gruvberger S, Ringner M, Chen Y, et al. Estrogen receptor status in 40. Cunha IW, Carvalho KC, Martins WK, et al. Identification of genes breast cancer is associated with remarkably distinct gene expression associated with local aggressiveness and metastatic behavior in soft patterns. Cancer Res 2001;61:5979–84. tissue tumors. Transl Oncol 2010;3:23–32. 46. Berlingieri MT, Pallante P, Sboner A, et al. UbcH10 is overexpressed 41. Fujita T, Ikeda H, Taira N, Hatoh S, Naito M, Doihara H. Overex- in malignant breast carcinomas. Eur J Cancer 2007;43:2729–35. pression of UbcH10 alternates the cell cycle profile and accelerate 47. McAvoy S, Ganapathiraju SC, Ducharme-Smith AL, et al. Non- the tumor proliferation in colon cancer. BMC Cancer 2009;9:87. random inactivation of large common fragile site genes in different 42. Hale LP, Price DT, Sanchez LM, Demark-Wahnefried W, Madden JF. cancers. Cytogenet Genome Res 2007;118:260–9. Zinc α-2-glycoprotein is expressed by malignant prostatic epithelium 48. Smith DI, Zhu Y, McAvoy S, Kuhn R. Common fragile sites, extremely large and may serve as a potential serum marker for prostate cancer. genes, neural development and cancer. Cancer Lett 2006;232:48–57. Clin Cancer Res 2001;7:846–53. 49. O'Dushlaine C, Kenny E, Heron E, et al. Molecular pathways involved in 43. Hassan MI, Waheed A, Yadav S, Singh TP, Ahmad F. Prolactin neuronal cell adhesion and membrane scaffolding contribute to schizo- inducible protein in cancer, fertility and immunoregulation: structure, phrenia and bipolar disorder susceptibility. Mol Psychiatry 2010. function and its clinical implications. Cell Mol Life Sci 2009;66: 50. Omura N, Li CP, Li A, et al. Genome-wide profiling of methylated 447–59. promoters in pancreatic adenocarcinoma. Cancer Biol Ther 2008;7: 44. Okamoto Y, Ozaki T, Miyazaki K, Aoyama M, Miyazaki M, 1146–56.

3874 Clin Cancer Res; 16(15) August 1, 2010 Clinical Cancer Research

Clinical Implications of Gene Dosage and Gene Expression Patterns in Diploid Breast Carcinoma

Toshima Z. Parris, Anna Danielsson, Szilárd Nemes, et al.

Clin Cancer Res 2010;16:3860-3874. Published OnlineFirst June 15, 2010.

Updated version Access the most recent version of this article at: doi:10.1158/1078-0432.CCR-10-0889

Supplementary Access the most recent supplemental material at: Material http://clincancerres.aacrjournals.org/content/suppl/2010/06/16/1078-0432.CCR-10-0889.DC1

Cited articles This article cites 48 articles, 17 of which you can access for free at: http://clincancerres.aacrjournals.org/content/16/15/3860.full#ref-list-1

Citing articles This article has been cited by 10 HighWire-hosted articles. Access the articles at: http://clincancerres.aacrjournals.org/content/16/15/3860.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://clincancerres.aacrjournals.org/content/16/15/3860. Click on "Request Permissions" which will take you to the Copyright Clearance Center's (CCC) Rightslink site.