Breast Cancer Research and Treatment 78: 289–298, 2003. © 2003 Kluwer Academic Publishers. Printed in the Netherlands. Profiling breast cancer by array CGH Donna G. Albertson Comprehensive Cancer Center, University of California San Francisco, San Francisco, CA, USA Key words: array CGH, copy number aberration, genome instability Summary Breast tumors display a wide variety of genomic alterations. This review focuses on DNA copy number variations in these tumors as measured by the recently developed microarray-based form of comparative genomic hybridization. The capabilities of this new technology are reviewed. Initial applications of array CGH to the analysis of breast cancer, and the mechanisms by which the particular types of copy number changes might arise are discussed. Introduction Array CGH The development of solid tumors involves acquisition As originally described, CGH detects and maps DNA of genetic and epigenetic alterations, and the concom- sequence copy number variation throughout the entire itant changes in gene expression, that modify normal genome onto a cytogenetic map supplied by meta- growth control and survival pathways. One of the char- phase chromosomes [2]. The use of metaphase chro- acteristics of breast tumors is the great heterogeneity mosomes as the hybridization target has previously in aberrations that are found. The tumor genomes limited the resolution of CGH to 10–20 Mb, prohibited may have nearly normal or highly abnormal karyo- resolution of closely spaced aberrations, and only al- types; they may or may not contain point mutations, lowed linkage of CGH results to genomic information or epigenetic modifications, such as methylation. In and resources with cytogenetic accuracy. Array-based different tumors, expression of the same gene may be CGH, on the other hand, provides the capability to altered in multiple ways, or particular functional path- map copy number aberrations relative to the genome ways may be affected at different locations. It is now sequence, with the resolution being determined by the generally accepted that in order for a sufficient num- spacing of the clones. In array CGH, arrays of ge- ber of alterations to accumulate to cause a malignancy, nomic BAC, P1, cosmid or cDNA clones are used one or more mechanisms that work to maintain genetic as the hybridization target in place of the metaphase integrity in cells and/or to regulate cell cycle pro- chromosomes [3–6]. Relative copy number is then gression must be compromised, presumably through measured at these specific loci by hybridization of mutations that occur early in tumorigenesis [1]. Here, fluorescently labeled test and reference DNAs as in we focus on DNA copy number alterations in breast conventional CGH [4]. Since the clones used on the tumors, first reviewing the capability of the recently array contain sequence tags, their positions are ac- developed microarray-based form of comparative ge- curately known relative to the genome sequence, and nomic hybridization (array CGH) to measure and map genes mapping within regions of copy number altera- these aberrations, and then discussing the mechanisms tion can be readily identified using genome databases. by which the particular types of copy number changes Arrays comprised of large insert genomic clones might arise. such as BACs, P1’s and cosmids provide reliable 290 DG Albertson Figure 1. Comparison of three hybridizations with BT474 DNA. Test BT474 DNA and male reference DNAs were labeled either by nick translation or random priming and hybridized to an array of 1777 clones (HumArray 1.11). The log2ratios for three hybridizations are shown together, plotted according to order on the draft sequence for chromosomes 14–22 and the X. The complete data sets are available in the web supplement to Snijders et al. [31]. copy number measurements on individual clones, and sembled across regions of particular interest [9]. By thus have potential utility for both research and clin- using overlapping clones from regions of contiguous ical applications. However, preparation and spotting clone coverage copy number changes can be mapped of BAC DNA is problematic. BACs are single copy with a genomic resolution less than the clone length vectors so the yield of DNA from cultures is low (<50 kb). compared to yields from plasmid vectors, and spotting Validation of the capability of array CGH to mea- high molecular weight DNA at sufficient concentra- sure single copy gains and losses has been carried tion to obtain good signal to noise in the hybridizations out using cell strains with known monosomies or may be difficult. Therefore we have used a ligation- trisomies [6]. The measured ratios closely approx- mediated PCR procedure [7] to make a complex rep- imated the ideal ratios. For example, 13 different resentation of the BACs for spotting on the array regions of trisomy were measured in this set of [6]. The procedure yields sufficient spotting solution cells, and the mean of the log2ratios of these tri- (0.8 µg/µl DNA in 20% DMSO) from 1 ng of BAC somic chromosomal regions was 0.49 ± 0.05. This DNA to make tens of thousands of arrays, and the value is close to, but slightly less than, the ex- ratios measured on arrays comprised of BAC repres- pected log2ratio = 0.58 for the 3/2 copy num- entations are essentially identical to ratios previously ber ratio that results from a trisomy. In female/ reported on DNA from the same whole BAC [6]. De- male comparisons, the mean log2ratios on the X chro- tailed protocols for array production and hybridization mosome were 0.72 ± 0.08, compared to the expecta- are available elsewhere [6, 8]. tion of 1.0. These underestimations of the magnitude We have assembled arrays of ∼2500 BACs and of copy number deviations are most likely due to in- P1’s for scanning the entire genome for copy num- complete suppression of repetitive sequences or errors ber changes. These arrays provide resolution of in background subtraction [4]. ∼1.4 Mb. Each clone is printed in triplicate in a Three replicate hybridizations with BT474 cell line 12 mm × 12 mm square area. A typical hybridization DNA were carried out to assess the reproducibility requires 200–300 ng of tumor DNA, although dilu- of array CGH measurements [6]. In two of the hy- tion tests have produced successful hybridizations bridizations, test and reference DNAs were labeled with as little as 3 ng of DNA. Best results have been by random priming, while the third was labeled by obtained with DNA extracted from frozen specimens nick translation. Figure 1 shows the results for chro- by conventional methods. It is also possible to use mosomes 14–22 and the X. Nearly identical ratios DNA extracted from paraffin embedded specimens, were obtained on each clone (average S.D. of the as well as DNA obtained by back extraction after log2ratio = 0.08), including those clones reporting Trizol purification of RNA, but results are more vari- high level amplifications (chromosomes 17 and 20) able. Higher resolution arrays have also been as- and losses (chromosome 20). Thus, it is generally Profiling breast cancer by array CGH 291 not necessary to perform the array measurement more losses of a small portion of distal 2q, a portion than once on a given specimen. of 7q and all of 16q. In the other breast tumor The ratios measured in tumor cells are likely to (Figure 3(d)), a greater variety of aberrations was reflect the presence of admixed normal cell DNA, seen. Alterations involving chromosome arms in- tetraploid DNA content, and/or tumor heterogeneity. cluded gain of 3q, gain of 5q and a higher level gain Therefore to make copy number measurements on tu- of 5p. In addition, amplification on chromosome 20q mor specimens, it is necessary to detect alterations that was observed, as were losses of whole chromosomes are lesser in magnitude than one expects for single (chromosomes 13, 14 and 22). This wide variety copy changes in diploid genomes. In order to as- of tumor genomic alterations is likely to reflect the sess the capability to detect small ratio changes, a many different solutions adopted by individual tumors series of measurements was carried out on dilutions to escape normal protective mechanisms for growth of a trisomic cell strain with normal DNA. Copy control and senescence and will be discussed further number aberrations were reliably detected in mix- below. tures comprised of >50% normal DNA, indicating Array CGH provides a higher resolution map- the capability to detect ratios much less than 1.5 ping of amplicons and indicates that amplicons may expected for a trisomy (log2ratio = 0.58). Further- be simple or highly complex. In some cases, the more, as expected from these experiments, log2ratios identification of oncogenes has been facilitated, be- of 0.47 ± 0.08 and 0.32 ± 0.07 are clearly visible in cause they have been found to be up-regulated when the tumor samples shown in Figure 3(c) and (d), present at elevated copy number in very focal amp- respectively. licons (e.g., CMYC in COLO320, Figure 2(b)). The amplicon encompassing ERBB2 frequently ap- pears as a simple peak in the copy number pro- Copy number alterations in breast file, while others such as those on 11q including tumor genomes CCND1 are often highly complex (Figure 4). Amp- lification of CCND1 is generally accompanied by Studies of breast tumors by conventional karyotyping, amplification of several distinct adjacent copy num- fluorescent in situ hybridization (FISH), spectral kar- ber peaks, as well as loss of copy number on distal yotyping or MFISH, chromosomal CGH and array 11q. In breast cancer, overexpression of ERBB2 is CGH have found a wide range in the number and almost always associated with amplification of the variety of types of chromosome level alterations in gene [11].