Array Comparative Genomic Hybridization Copy Number Profiling: A New Tool for Translational Research...

7
Array Comparative Genomic Hybridization Copy Number Profiling: A New Tool for Translational Research in Solid Malignancies José Luis Costa, PhD,* ,† Gerrit Meijer, MD, PhD, Bauke Ylstra, PhD, and Carlos Caldas, MD The molecular genetic investigation of cancer is rapidly evolving because of ever-improving technology. Insights into cancer disease mechanisms are being elucidated using new chromosome-based biomarkers. Until recently, diagnostic and prognostic assessment of diseased tissues and tumors relied heavily on histologic indicators that permitted only general classifications into morphologic subtypes and did not take into account the alterations in individual gene expression or copy number. Genomic and expression profiling now allow the simultaneous interrogation of thousands of genes and offer unprecedented opportunities to obtain global molecular signatures of neoplastic cells in patient samples. One limitation of global profiling at the expression level is that acquisition and optimal transport of high-quality RNA is problematic because of its inherent instability in vitro. In contrast, tumor DNA is stable, relatively easy to transport, and can be obtained from archival paraffin tissue blocks. Thus, there is now a tremendous opportunity to globally profile copy number imbalances in tumors using array comparative genomic hybridization (CGH), which can identify at high resolution the presence of genomic copy number changes in constitutional or tumor DNA samples. Array CGH profiling has already allowed a deeper insight into the biology of a variety of tumor types and in the near future will undoubtedly prove to be a key technology leading to better cancer classification, prognosis, and outcome prediction. Semin Radiat Oncol 18:98-104 © 2008 Elsevier Inc. All rights reserved. C hromosomal copy number aberrations (CNAs) are fre- quent in solid tumors. 1 The wide range in the number and type of aberrations most likely reflects clonal selection of changes that directly affect the cardinal properties of the can- cer cell 2 and passenger events that are the result of the un- derlying genetic instability. This fits with the observation that many of these aberrations affect known oncogenes or tumor suppressor genes whose expression levels are altered by DNA copy number genomic alteration. 3-5 Classic examples in solid tumors include amplification of established oncogenes, such as ERBB2 and MYC. 6,7 Other aberrations involve loss of spe- cific regions of the genome. For example, deletions are im- portant in the inactivation of tumor suppressor genes, such as PTEN and CDKN2A, and in the elimination of the remaining wild-type allele in carriers of inherited mutations involving BRCA1, BRCA2, and TP53. 1 Array comparative genomic hybridization (CGH) allows the global profiling of such copy number imbalances in tu- mors and the precise determination of the breakpoints of regions that are gained and/or lost. The options for detecting DNA copy number changes on a genome-wide basis and at base-pair resolution are developing rapidly. Until recently, the technical prerequisites of the available methodology se- verely hampered clinical application. Karyotyping with the requirement of living cells is limited to hematologic malig- nancies and congenital disorders. As a complementary tech- nique, fluorescence in situ hybridization was developed to allow the precise analysis of a limited number of targets at a time and DNA cytometry to detect the presence of gross *Institute of Molecular Pathology and Immunology of the University of Porto, University of Porto, Porto, Portugal. †Department of Pathology, VU University Medical Center, Amsterdam, The Netherlands. ‡Breast Cancer Functional Genomics Laboratory, Cancer Research UK Cam- bridge Research Institute and Department of Oncology, University of Cambridge, Cambridge, United Kingdom. Supported by the EU-sixth framework project DISMAL (contract no. LSHC- CT-2005 to 018911). JL Costa was financially supported by a grant from “Fundação para a Ciência e Tecnologia” (SFRH/ BPD/ 20,370/ 2004). Address reprint requests to Carlos Caldas, Breast Cancer Functional Genom- ics Laboratory, Cancer Research UK Cambridge Research Institute and Department of Oncology, University of Cambridge, Li Ka Shing Centre, Robinson Way, Cambridge CB2 0RE, UK. E-mail: [email protected] 98 1053-4296/08/$-see front matter © 2008 Elsevier Inc. All rights reserved. doi:10.1016/j.semradonc.2007.10.005

Transcript of Array Comparative Genomic Hybridization Copy Number Profiling: A New Tool for Translational Research...

Page 1: Array Comparative Genomic Hybridization Copy Number Profiling: A New Tool for Translational Research in Solid Malignancies

AHTJC

Caccdms

*

S

A

9

rray Comparative Genomicybridization Copy Number Profiling: A Newool for Translational Research in Solid Malignancies

osé Luis Costa, PhD,*,† Gerrit Meijer, MD, PhD,† Bauke Ylstra, PhD,† andarlos Caldas, MD‡

The molecular genetic investigation of cancer is rapidly evolving because of ever-improvingtechnology. Insights into cancer disease mechanisms are being elucidated using newchromosome-based biomarkers. Until recently, diagnostic and prognostic assessment ofdiseased tissues and tumors relied heavily on histologic indicators that permitted only generalclassifications into morphologic subtypes and did not take into account the alterations inindividual gene expression or copy number. Genomic and expression profiling now allow thesimultaneous interrogation of thousands of genes and offer unprecedented opportunities toobtain global molecular signatures of neoplastic cells in patient samples. One limitation ofglobal profiling at the expression level is that acquisition and optimal transport of high-qualityRNA is problematic because of its inherent instability in vitro. In contrast, tumor DNA is stable,relatively easy to transport, and can be obtained from archival paraffin tissue blocks. Thus,there is now a tremendous opportunity to globally profile copy number imbalances in tumorsusing array comparative genomic hybridization (CGH), which can identify at high resolution thepresence of genomic copy number changes in constitutional or tumor DNA samples. ArrayCGH profiling has already allowed a deeper insight into the biology of a variety of tumor typesand in the near future will undoubtedly prove to be a key technology leading to better cancerclassification, prognosis, and outcome prediction.Semin Radiat Oncol 18:98-104 © 2008 Elsevier Inc. All rights reserved.

ctacpPwB

tmrDbtvrnna

hromosomal copy number aberrations (CNAs) are fre-quent in solid tumors.1 The wide range in the number

nd type of aberrations most likely reflects clonal selection ofhanges that directly affect the cardinal properties of the can-er cell2 and passenger events that are the result of the un-erlying genetic instability. This fits with the observation thatany of these aberrations affect known oncogenes or tumor

uppressor genes whose expression levels are altered by DNA

Institute of Molecular Pathology and Immunology of the University ofPorto, University of Porto, Porto, Portugal.

Department of Pathology, VU University Medical Center, Amsterdam, TheNetherlands.

Breast Cancer Functional Genomics Laboratory, Cancer Research UK Cam-bridge Research Institute and Department of Oncology, University ofCambridge, Cambridge, United Kingdom.

upported by the EU-sixth framework project DISMAL (contract no. LSHC-CT-2005 to 018911). JL Costa was financially supported by a grant from“Fundação para a Ciência e Tecnologia” (SFRH/ BPD/ 20,370/ 2004).

ddress reprint requests to Carlos Caldas, Breast Cancer Functional Genom-ics Laboratory, Cancer Research UK Cambridge Research Institute andDepartment of Oncology, University of Cambridge, Li Ka Shing Centre,

tRobinson Way, Cambridge CB2 0RE, UK. E-mail: [email protected]

8 1053-4296/08/$-see front matter © 2008 Elsevier Inc. All rights reserved.doi:10.1016/j.semradonc.2007.10.005

opy number genomic alteration.3-5 Classic examples in solidumors include amplification of established oncogenes, suchs ERBB2 and MYC.6,7 Other aberrations involve loss of spe-ific regions of the genome. For example, deletions are im-ortant in the inactivation of tumor suppressor genes, such asTEN and CDKN2A, and in the elimination of the remainingild-type allele in carriers of inherited mutations involvingRCA1, BRCA2, and TP53.1

Array comparative genomic hybridization (CGH) allowshe global profiling of such copy number imbalances in tu-ors and the precise determination of the breakpoints of

egions that are gained and/or lost. The options for detectingNA copy number changes on a genome-wide basis and atase-pair resolution are developing rapidly. Until recently,he technical prerequisites of the available methodology se-erely hampered clinical application. Karyotyping with theequirement of living cells is limited to hematologic malig-ancies and congenital disorders. As a complementary tech-ique, fluorescence in situ hybridization was developed tollow the precise analysis of a limited number of targets at a

ime and DNA cytometry to detect the presence of gross
Page 2: Array Comparative Genomic Hybridization Copy Number Profiling: A New Tool for Translational Research in Solid Malignancies

gTc

losclctacuecrfvl

rc

mirmsrrunolFrcdbl

itatc

T

ABBBBBBBBBBBBBBCCCCCEGGGIKLLMOOWW

Array CGH copy number profiling 99

enomic aberrations but without chromosome specificity.8,9

his situation dramatically changed with the introduction ofomparative genomic hybridization (classical CGH).

Comparative genomic hybridization uses differentiallyabeled tumor and reference DNAs that are cohybridizednto normal metaphase chromosome spreads on a glasslide and allows the genome-wide analysis of alterations inhromosomal copy number.10 The ratios between the 2abels along the chromosome axis allow the mapping ofhromosomal gains/amplifications and deletions acrosshe entire genome. Because no cell culturing is requirednd even formaldehyde-fixed paraffin-embedded materialan be used, this technique became one of the most pop-lar genome-scanning techniques and contributed to annormous progress in the analysis and characterization ofhromosomal changes in solid tumors.11,12 However, theesolution of classical CGH (10-20 Mb) was still a limitingactor. The sequencing of the human genome and the ad-ent of microarray technology allowed a leap in the reso-ution of cancer copy number profiling.

Almost 10 years ago, the first microarray-based CGH (ar-ay CGH) was introduced.4,13 In array CGH, the metaphasehromosomes are replaced by DNA fragments arrayed on a

able 1 Selected array CGH profile studies of solid tumors.

Tumor Type Cohort

strocytic 104 Inladder 41 Uladder 98 Ureast 25 Sreast 28 Mreast 30 Areast 31 Inreast 44 Inreast 65 Ureast 67 Ureast 89 Sreast 100 Vreast 145 Ureast 148 Vreast 171 Inervix 26 Vhondrosarcoma Molorectal 57 Inolorectal 59 Solorectal 125 Usophagous 32 Vastric 35 Vastric 46 Vlioblastoma 70 G

nsulinoma 27 Inidney 40 Uiver 34 Iniver 49 Uedulloblastomas 41 Insteosarcoma 48 Svary 26 Vilms tumour 58 B

ilms tumour 76 BBC BA

icroscope slide for which the exact chromosomal locations known from the human genome sequence. The spatialesolution is then determined by the size of the DNA frag-ents and the density of their spacing along the chromo-

omes.14 Importantly, arrays can be customized to cover anyegion of interest at any resolution. Moreover with higheresolution, amplifications and deletions that previously wentndetected are no longer missed.9,15 Many solid tumors haveow been profiled using this technique, revealing the heter-geneity and complexity of copy number aberrations under-ying those cancers (Table 1). Two such profiles are shown inigure 1. Figure 1A shows a profile from a gemcitabine-esistant mouse colon tumor with a clear amplification onhromosome 7 associated with acquired resistance to therug.15 Figure 1B shows the genomic profile of a humanreast tumor cell line (BT474) showing a variety of high-/

ow-level amplifications and deletions.Despite the great diversity of CNAs observed across tumors, it

s becoming obvious that the gains and losses have tumor- andissue-specific traits.16-18 This review will focus on the use ofrray CGH in solid malignancies, the recent advances of theechnology, and the use of CNAs patterns for subclass identifi-ation and clinical outcome prediction in cancer.

Platform Reference

e tilling array 49ity of California BAC array 50ity of California BAC array 51AC array 52

ene BAC array 53cDNA array 54e BAC array 55e cDNA array 5ity of Pennsylvania BAC array 56ity of California BAC array 17d cDNA array 25AC array 27ity of California BAC array 16AC array 26e oligo array 57BAC array 23ancer BAC Array 58e BAC array 59 BAC array 60ity of California BAC array 61AC array 62BAC array 63BAC array 64nsor Array 300 65e BAC array 66ity of California BAC array 67e BAC array 68ity of California BAC array 69e BAC array 70l Genomics BAC array 71AC array 72C array 73

housniversniversMRT Bacroggilent hous housniversniverstanforysis Bniversysis B housUMC CG C housangerniversysis BUMC UMC enoSe housnivers housnivers houspectraysis BBC BA

C array 74

Page 3: Array Comparative Genomic Hybridization Copy Number Profiling: A New Tool for Translational Research in Solid Malignancies

AiAttitinTe“otdasAs

bmIgcebgotbtwcntmoC

msmiminfiotwesbdm

irttoAtTp

betCpcclasgfiotpomt

srtiiia

AMTgrfisfifcpnvo

BBD

100 J.L. Costa et al

rray CGH Applicationsn Cancer Researchrray CGH is currently the method of choice for the detec-

ion of CNAs, and it has allowed a deeper understanding ofumor biology. Until recently, it was believed that mutationsn tumor suppressor genes and oncogenes, triggered or main-ained by genomic instability, occurred predominantly in ep-thelial tumors, whereas cytogenetic aberrations predomi-ated in hematologic disorders and mesenchymal tumors.1,19

he present view, which we favor, is that CNAs are an earlyvent, and it is presumed that these are alterations thatdrive” the onset and growth of a tumor.19 Once aneuploidyccurs, additional chromosomal derangements are likely toake place. By the time a tumor is clinically identified andissected, many cell cycles have passed and what is measuredre not only causal and progressive CNAs but also the pas-enger alterations (consequence of derailed mitotic events).s a result, tumor CNA profiles harbor both tumor-tissuepecific characteristics as well as more random aberrations.

For example, there are quantitative and qualitative differencesetween epithelial tumors and hematologic and mesenchymalalignancies, such as was observed in a recent meta-analysis.20

n this study, tumors of hematopoietic and mesenchymal ori-ins clustered separately from tumors of epithelial origin, indi-ating that chromosomal aberrations of tumors with the differ-nt origins are distinct and that these differences can be detectedy array CGH.20 In other studies, CNA profiles could distin-uish known subclasses of tumors within the same tissue ofrigin. For instance, in breast cancer, it allowed the separation ofumors from the 2 familial syndromes (BRCA1 and BRCA2reast tumors) and also separated these from sporadic breastumors.21,22 This was confirmed in an independent study inhich all the BRCA tumors clustered in only 1 of the 3 main

lusters of sporadic tumors.17 Likewise in gastric adenocarci-oma, the genomic profiling of CNAs has allowed discrimina-ion of a subgroup of patients with high risk of lymph nodeetastasis and is predictive of prognosis.18 Taken together, these

bservations suggest the possibility of prospectively using arrayGH for cancer classification.DNA copy number alterations can also be useful moleculararkers for cancer prognosis or the prediction of treatment re-

ponse (Fig 2). For example, in cervical carcinogenesis, com-on CNAs were identified by using array CGH, resulting in the

dentification of potential novel biomarkers, which could ulti-ately enable a better risk stratification of high-risk human pap-

llomavirus (HPV)-positive women.23 Similarly, in head andeck squamous-cell carcinomas, array CGH allowed the identi-cation of CNAs that differ between tumors with or withoutncogene-expressing human papillomavirus.24 In breast cancer,he profiling of CNAs identified distinct loci of CNA associatedith different clinicopathologic features, including tumor grade,

strogen receptor (ER) status, TP53 mutation, gene-expressionubtype, and overall survival.16,25,26 These CNAs may provide aasis for improved patient prognosis as well as a starting point toefine important genes contributing to breast cancer develop-

ent and progression.25 o

New prognostic markers in breast cancer have also beendentified and validated using array CGH and tissue microar-ays. By studying regions commonly amplified, a subset ofhe genes (EMS1, TOP2A, CCNE1, and ERBB2) was identifiedhat could divide the series into 2 divergent outcome groupsf either long-term survivors or early disease-related deaths.27

dverse disease-related outcome was associated with amplifica-ion of TOP2A; ERBB2; and with the combined amplification ofOP2A, ERBB2, and EMS1. Amplification of CCNE1 had norognostic role in this series.27

Another study explored the role of CNAs in breast cancery identifying associations between recurrent CNAs, genexpression, and clinical outcome in a set of aggressivelyreated early-stage tumors.16 It showed that the recurrentNAs differ between tumor subtypes defined by expressionattern and that stratification of patients according to out-ome can be improved by measuring both expression andopy number, particularly high-level amplification. Low-evel CNAs appeared to contribute to cancer progression byltering RNA and cellular metabolism.16 Array CGH has alsohown that different breast cancers progress along differentenomic pathways (HER2, cyclin D, and 8q and 20q ampli-ers)26 and allowed the identification of novel breast cancerncogenes within complex amplicons (eg, 8p12).28 Addi-ionally, there are strong indications that array CGH mayrove to be useful for the differentiation between metastasisr a second primary tumor and the identification of the pri-ary location in cases of metastases and 2 (or more) primary

umors29 (and personal observations).These are just a selected number of recent studies that

erve as examples of the potential use of array CGH in canceresearch. The variety of applications include screening ofumors for genetic aberrations, searching for genes involvedn the carcinogenesis of particular subsets of cancers, analyz-ng tumors in experimental models to provide more insightnto tumor progression, and using diagnostic classificationnd prognosis assessment.

rray CGHethods and Platforms

he majority of array CGH data available today has beenenerated by using BAC CGH arrays. However, there is aapid increase in data obtained using oligo CGH arrays. Therst array CGH platforms generally used large-insert clones,uch as the bacterial artificial chromosome (BAC), yeast arti-cial chromosome, or bacterial plasmid clones. Cosmid andosmid clones have also been used.4 Several laboratories usedomplementary DNA arrays, initially designed for expressionrofiling, as an alternative for measuring chromosomal copyumber changes.30 Even though this approach has yieldedaluable information, it cannot compete with the currentligoplatforms in terms of sensitivity and resolution.31

AC Array CGH PlatformsACs vary in length from 150 to 200 kb and because highNA concentration is required for high-quality results, most

f the platforms use polymerase chain reaction amplification
Page 4: Array Comparative Genomic Hybridization Copy Number Profiling: A New Tool for Translational Research in Solid Malignancies

Figure 1 Genome-wide oligoarray CGH profiles with ratios ordered by chromosomal position; in blue the log2 raw ratiosare depicted and in red the smoothed values. (A) A gemcitabine-resistant mouse colon tumor. The profile resulted fromthe hybridization of 300 ng of tumor DNA with 300 ng of liver DNA against an array spotted with a mouse oligolibrarycontaining 21.997 oligonucleotide probes (65 bp) representing 21.587 exon regions. A clear amplification on chro-mosome 7 is depicted and is associated with an acquired resistance to the drug.15 (B) Human breast tumor cell lineBT474 genomic DNA (500 ng) hybridized with a normal human reference (500 ng) against an array on the 4 � 44,000Agilent slide (Agilent, Santa Clara, CA). Labeling was performed by using the ENZO BioArray CGH Labeling System

(ENZO, New York, NY). A variety of high-/low-level amplifications and deletions can be observed.

Figure 2 An example of a cluster analysis of 35 gastric cancers array CGH profiles for predicting outcome. (A) Hierarchicalcluster analysis revealed 3 clusters of gastric cancers with cluster 1 containing 12 cancers, cluster 2 containing 8 cancers, andcluster 3 containing 15 cancers. Especially clusters 1 and 3 showed clearly different patterns of chromosomal aberrations,whereas cluster 2 showed an intermediate pattern of chromosomal changes. (B) A Kaplan-Meier survival plot for the 3different clusters. Deaths because of causes other than gastric cancer were treated as censored observations. Survival (dead ofdisease) was significantly better in cluster 3 than clusters 1 and 2. Kaplan-Meier survival analysis revealed a significantdifference between cluster 3 versus clusters 1 and 2 combined (P � .019; log-rank statistic, 5.54, hazard ratio, 4.1; 95%

confidence interval, 1.1-14.7). (Adapted with permission of Nature Publishing Group. 18)
Page 5: Array Comparative Genomic Hybridization Copy Number Profiling: A New Tool for Translational Research in Solid Malignancies

bdcpoarhrtB

fc(rcmscamcaaaTpfs2igbc

OOsDinrtagbiwThmAna

cC

miomGgtibepmtnucgn

fccrtwccSaldbp

FaAatfbwtwmsatou

agpi

102 J.L. Costa et al

efore spotting the arrays. Genome-wide BAC arrays vary inensity, from 2,400 clones (1Mb density arrays) to 30,000lones (tilling arrays). Difficulties when setting up BAC arraysarallel those of spotted complementary DNA arrays in termsf clone management and probe identity because of polymer-se chain reaction PCR contamination. In addition, BAC ar-ay data suffer from mapping inaccuracies of the clones to theuman genome. Also, a genome-covering array is beyond theeach of most individual laboratories, and, as a consequence,hese arrays have not been widely available. Nevertheless, theAC platform is outstandingly sensitive and precise.A few BAC platforms are currently available as a resource

or researchers. The Genosensor arrays (www.genosensorcorp.om, Genosensor, Tempe, AZ) contain 287 target DNA clonesP1, bacterial plasmid, or BAC clones) printed in triplicateepresenting loci previously shown to be important in can-er, involved in congenital syndromes, and mapped to telo-eres.26 The Spectral Chip 2600 (www.spectralgenomic-

.com, Spectral Genomics, Houston, TX) consists of 2,632lones spanning across the genome at 1-Mb intervals onverage across the human genome, with each cloneapped to a cytogenetic linear position.32 BlueGnome (www.

ambridgebluegnome.com, BlueGnome, Cambridge, UK)lso provides a 1-Mb resolution genome-wide array.33 Ancademic version with �5,000 human BAC clones with anverage resolution of 1 Mb is produced and maintained athe University of Pennsylvania (www.genomics.upenn.edu/eople/faculty/weberb/CGH/html).21 The University of Cali-

ornia (cancer.ucsf.edu/array/, San Francisco, CA) makes amaller human BAC array (HumArray3.2) that contains,464 BAC clones spotted in triplicate, with an average spac-

ng between clones of 1.4 Mb.34 Another platform, the Wholeenome SubMegabase Resolution Tiling Array-SMRT (www.ccrc.ca/arraycgh/, Vancouver, British Columbia, Canada),onsists of tiling path resolution microarrays.35

ligoarray CGH Platformsligoarray CGH platforms are characterized by single-

tranded 25 to 85 mer oligonucleotide elements on the array.ifferent oligoarrays combine different labeling and hybrid-

zation techniques, and all have yielded high-resolution copyumber measurements. Affymetrix is a commercial oligoar-ay CGH platform that contains short 25 mer oligonucleo-ides photolithographically synthesized on the arrays (www.ffymetrix.com, Affymetrix, Santa Clara, CA). These are sin-le-channel arrays, which means that only test DNA needs toe labeled and hybridized. The labeling of the test sample

nvolves a restriction enzyme-based complexity reduction,hich precludes the use of suboptimal DNA quality samples.he technical noise per element on the array is relativelyigh, which gets compensated by the large amount of ele-ents on the array, currently 2,000,000. An advantage of theffymetrix system compared with other systems is that singleucleotide polymorphism (SNPs) are detected in parallel,llowing allelotyping.

An alternative approach to microarrays on slides is the single-hannel Illumina’s BeadArray Technology (Illumina, San Diego,

A), which is based on 3-�m silica beads that self-assemble in s

icrowells. The beads are randomly assembled, and each beads covered with hundreds of thousands of copies of a specificligonucleotide that act as the capture sequences in one of Illu-ina’s assays. Illumina’s Infinium Assays and Whole-Genomeenotyping BeadChips offer genome-wide coverage. Whole-enome genotyping can also yield allelic information on dele-ions, duplications, and amplifications, which have implicationsn cancer therapeutics. The Illumina BeadArray platform haseen shown to be suitable to genotype and detect loss of het-rozygosity (LOH) from low-quality DNA from formalin-fixedaraffin-embedded (FFPE) tissues.36 The combined measure-ent of allelic ratios and normalized intensities provides detec-

ion of aberrations while allowing the identification of copy-eutral genetic anomalies. Such copy-neutral anomalies (eg,niparental disomy and mitotic recombination) have not re-eived much attention and involve the loss of 1 allele and theain of the remaining allele, giving rise to LOH but no copyumber alteration.37

The first dual-channel commercial oligoarray CGH plat-orm was introduced by Agilent Technologies (www.agilent.om, Agilent, Santa Clara, CA).38 Agilent array platforms areonstituted of 60 mer oligonucleotides that are synthesized di-ectly on the slides. The price tag has been recently lowered withhe introduction of the 4 � 44,000 arrays (soon 4 � 200,000)here on a single slide 4 different genome-wide hybridizations

an be performed, reducing the overall costs. A second dual-hannel oligonucleotide platform is available from Nimblegenystems Inc. (www.nimblegen.com, Madison, WI). It providesrrays containing 2,100,000 oligonucleotides (50-75 mer inength) photolithographically synthesized on the array. The pro-uction of the Nimblegen and Agilent arrays is extremely flexi-le, and the companies are currently offering near-single base-air resolution arrays for individual chromosomes.

uture Developmentsnd Improvementsll platforms described previously have proven highly valu-ble and capable of producing high-quality and high-resolu-ion genome-wide data. However, the algorithms requiredor the analysis of array CGH data for tumor profiling lagehind. This is in contrast with the expression array field inhich agreement on how to perform the analysis, on which

ools, and for what purposes is being reached.39 To beginith preprocessing (consisting of 3 steps: normalization, seg-entation, and calling), there is no consensus on how these

teps should be performed ahead of further downstreamnalysis. For dual-channel array CGH analysis, normaliza-ion is either performed by taking the median of all the datar by taking the mode or the average or it can be done bysing spike-in controls.Many different algorithms are available for segmentation,40

nd still new ones are emerging.41 For the actual calling ofains, losses, and amplification after the segmentation step,ragmatic solutions are chosen currently,16,42 and better call-

ng algorithms are a current focus of development.43-45 A

erious objection against the use of called data is the fact that
Page 6: Array Comparative Genomic Hybridization Copy Number Profiling: A New Tool for Translational Research in Solid Malignancies

habestomdtbacsplffrgpvon

gtSpmtuppHbwc

AWa

R

1

1

1

1

1

1

1

1

1

1

2

2

2

2

2

2

2

2

2

2

3

Array CGH copy number profiling 103

ardly any bioinformatics tools are available for downstreamnalysis of the called datasets. The analytic tools currentlyeing used for downstream processing of copy number ab-rrations were originally developed for the analysis of expres-ion data that have a continuous rather than ordinal charac-er. However, copy number aberrations are ordinal (0, 1, 2, 4,r many chromosomes), and, hence, the expression bioinfor-atics tools are not directly applicable to called array CGHata. We thus argue that array CGH-tailored bioinformaticsools need to be developed that take statistical advantage ofoth the ordinal nature and the positional dependency of therrayed elements.46 Our opinion is not shared by the entireommunity. Because there is no agreement here, the down-tream analysis might produce results that are algorithm de-endent and sometimes not reproducible. To illustrate this

ack of consensus, when reviewing the published literatureor the way clustering of array CGH data are performed, weound different methods, ranging from using raw normalizedatios,18 individual BAC clones, segmentation values,16,17,47

ene regions defined by chromosomal bands25 or 100 cloneser chromosome,48 and most using algorithms initially de-eloped for expression arrays. Hence, a consensus is requiredn subjects such as normalization and calling. These are sig-ificant challenges for biostatisticians and bioinformaticians.In conclusion, the uses of array CGH profiling in cancer

ene discovery have been well demonstrated for a number ofumor types, (lymphoma, leukemia, sarcoma, lung, and breast).uch systematic analyses, when integrated with existing androposed cancer genome projects, will provide invaluable infor-ation on the chromosomal mechanisms associated with al-

ered gene expression and will provide a more comprehensivenderstanding of disease pathology and new therapeutic ap-roaches. Array CGH is also giving strong steps toward its ap-licability for outcome prediction and treatment response.owever, the field is still evolving and integration of copy num-er, methylation, chromatin modification, and expression dataill undoubtedly lead to better cancer classification, prognosti-

ation, and treatment response prediction.

cknowledgmentse would like to thank Thijs Krugers for his help with the

rray hybridizations.

eferences1. Albertson DG, Collins C, McCormick F: Chromosome aberrations in

solid tumors. Nat Genet 34:369-376, 20032. Hanahan D, Weinberg RA: The hallmarks of cancer. Cell 100:57-70,

20003. Albertson DG, Ylstra B, Segraves R: Quantitative mapping of amplicon

structure by array CGH identifies CYP24 as a candidate oncogene. NatGenet 25:144-146, 2000

4. Pinkel D, Albertson DG: Array comparative genomic hybridization andits applications in cancer. Nat Genet 37:S11-S17, 2005

5. Pollack JR, Sorlie T, Perou CM: Microarray analysis reveals a majordirect role of DNA copy number alteration in the transcriptional pro-gram of human breast tumors. Proc Natl Acad Sci U S A 99:12963-12968, 2002

6. Alitalo K, Schwab M, Lin CC: Homogeneously staining chromosomal

regions contain amplified copies of an abundantly expressed cellular 3

oncogene (c-myc) in malignant neuroendocrine cells from a humancolon carcinoma. Proc Natl Acad Sci U S A 80:1707-1711, 1983

7. Slamon DJ, Godolphin W, Jones LA: Studies of the HER-2/neu proto-oncogene in human breast and ovarian cancer. Science 244:707-712, 1989

8. Hermsen MA, Meijer GA, Baak JP: Comparative genomic hybridization:A new tool in cancer pathology. Hum Pathol 27:342-349, 1996

9. Oostlander AE, Meijer GA, Ylstra B: Microarray-based comparativegenomic hybridization and its applications in human genetics. ClinGenet 66:488-495, 2004

0. Kallioniemi A, Kallioniemi OP, Sudar D: Comparative genomic hybrid-ization for molecular cytogenetic analysis of solid tumors. Science 258:818-821, 1992

1. Forozan F, Karhu R, Kononen J: Genome screening by comparativegenomic hybridization. Trends Genet 13:405-409, 1997

2. Weiss MM, Hermsen MA, Meijer GA: Comparative genomic hybridisa-tion. Mol Pathol 52:243-251, 1999

3. van den Ijssel P, Tijssen M, Chin SF: Human and mouse oligonucleoti-de-based array CGH. Nucleic Acids Res 33:e192, 2005

4. Coe BP, Ylstra B, Carvalho B: Resolving the resolution of array CGH.Genomics 89:647-653, 2007

5. van de Wiel MA, Costa JL, Smid K: Expression microarray analysis andoligo array comparative genomic hybridization of acquired gemcitab-ine resistance in mouse colon reveals selection for chromosomal aber-rations. Cancer Res 65:10208-10213, 2005

6. Chin K, DeVries S, Fridlyand J: Genomic and transcriptional aberra-tions linked to breast cancer pathophysiologies. Cancer Cell 10:529-541, 2006

7. Fridlyand J, Snijders AM, Ylstra B: Breast tumor copy number aberra-tion phenotypes and genomic instability. BMC Cancer 6:96, 2006

8. Weiss MM, Kuipers EJ, Postma C: Genomic profiling of gastric cancerpredicts lymph node status and survival. Oncogene 22:1872-1879, 2003

9. Mitelman F, Johansson B, Mertens F: Fusion genes and rearrangedgenes as a linear function of chromosome aberrations in cancer. NatGenet 36:331-334, 2004

0. Jong K, Marchiori E, van der Vaart A: Cross-platform array comparativegenomic hybridization meta-analysis separates hematopoietic and mes-enchymal from epithelial tumors. Oncogene 26:1499-1506, 2007

1. Jonsson G, Naylor TL, Vallon-Christersson J: Distinct genomic profilesin hereditary breast tumors identified by array-based comparativegenomic hybridization. Cancer Res 65:7612-7621, 2005

2. van Beers EH, van Welsem T, Wessels LFA: Comparative genomic hybrid-ization profiles in human BRCA1 and BRCA2 breast tumors highlightdifferential sets of genomic aberrations. Cancer Res 65:822-827, 2005

3. Wilting S, Snijders P, Meijer G: Increased gene copy numbers at chro-mosome 20q are frequent in both squamous cell carcinomas and ade-nocarcinomas of the cervix. J Pathol 209:220-230, 2006

4. Smeets SJ, Braakhuis BJM, Abbas S: Genome-wide DNA copy numberalterations in head and neck squamous cell carcinomas with or withoutoncogene-expressing human papillomavirus. Oncogene 25:2558-2564, 2006

5. Bergamaschi A, Kim YH, Wang P: Distinct patterns of DNA copy num-ber alteration are associated with different clinicopathological featuresand gene-expression subtypes of breast cancer. Genes ChromosomesCancer 45:1033-1040, 2006

6. Chin SF, Wang Y, Thorne NP: Using array-comparative genomic hy-bridization to define molecular portraits of primary breast cancers.Oncogene 26:1959-1970, 2007

7. Callagy G, Pharoah P, Chin SF: Identification and validation of prog-nostic markers in breast cancer with the complementary use of array-CGH and tissue microarrays. J Pathol 205:388-396, 2005

8. Garcia MJ, Pole JCM, Chin SF: A 1Mb minimal amplicon at 8p11-12 inbreast cancer identifies new candidate oncogenes. Oncogene 24:5235-5245, 2005

9. Ruiz MIG, van Cruijsen H, Smit EF. Genetic heterogeneity in patientswith multiple neoplastic lung lesions: A report of three cases. J ThoracOncol 2:12-21, 2007

0. Pollack JR, Perou CM, Alizadeh AA: Genome-wide analysis of DNA copy-number changes using cDNA microarrays. Nat Genet 23:41-46, 1999

1. Ylstra B, van den Ijssel P, Carvalho B: BAC to the future! or oligonucle-

Page 7: Array Comparative Genomic Hybridization Copy Number Profiling: A New Tool for Translational Research in Solid Malignancies

3

3

3

3

3

3

3

3

4

4

4

4

4

4

4

4

4

4

5

5

5

5

5

5

5

5

5

5

6

6

6

6

6

6

6

6

6

6

7

7

7

7

7

104 J.L. Costa et al

otides: a perspective for micro array comparative genomichybridization (array CGH). Nucleic Acids Res 34:445-450, 2006

2. Kitsiou-Tzeli S, Sismani C, Ioannides M: Array-CGH analysis and clin-ical description of 2q37.3 de novo subtelomeric deletion. Eur J MedGenet 50:73-78, 2007

3. Algar EM, St Heaps L, Darmanian A: Paternally inherited submicroscopicduplication at 11p15.5 implicates insulin-like growth factor II in over-growth and Wilms’ tumorigenesis. Cancer Res 67:2360-2365, 2007

4. Snijders AM, Nowak N, Segraves R: Assembly of microarrays for ge-nome-wide measurement of DNA copy number. Nat Genet 29:263-264, 2001

5. Ishkanian AS, Malloff CA, Watson SK: A tiling resolution DNA microar-ray with complete coverage of the human genome. Nat Genet 36:299-303, 2004

6. Lips EH, Dierssen JW, van Eijk R: Reliable high-throughput genotypingand loss-of-heterozygosity detection in formalin-fixed, paraffin-embeddedtumors using single nucleotide polymorphism arrays. Critical review ofpublished microarray studies for cancer outcome and guidelines on statis-tical analysis and reporting. Cancer Res 65:10188-10191, 2005

7. Andersen CL, Wiuf C, Kruhoffer M: Frequent occurrence of uniparen-tal disomy in colorectal cancer. Carcinogenesis 28:38-48, 2007

8. Barrett MT, Scheffer A, Ben-Dor A: Comparative genomic hybridizationusing oligonucleotide microarrays and total genomic DNA. Proc NatlAcad Sci U S A 101:17765-17770, 2004

9. Dupuy A, Simon RM: Critical review of published microarray studiesfor cancer outcome and guidelines on statistical analysis and reporting.J Natl Cancer Inst 99:147-157, 2007

0. Willenbrock H, Fridlyand J: A comparison study: Applying segmenta-tion to array CGH data for downstream analyses. Bioinformatics 21:4084-4091, 2005

1. Venkatraman ES, Olshen AB: A faster circular binary segmentationalgorithm for the analysis of array CGH data. Bioinformatics 23:657-663, 2007

2. Chen W, Houldsworth J, Olshen AB: Array comparative genomic hy-bridization reveals genomic copy number changes associated with out-come in diffuse large B-cell lymphomas. Blood 107:2477-2485, 2006

3. Engler DA, Mohapatra G, Louis DN: A pseudolikelihood approach forsimultaneous analysis of array comparative genomic hybridizations.Biostatistics 7:399-421, 2006

4. van de Wiel MA, Kim KI, Vosse SJ: CGHcall: Calling aberrations forarray CGH tumor profiles. Bioinformatics 23:892-894, 2007

5. Wang P, Kim Y, Pollack J: A method for calling gains and losses in arrayCGH data. Biostatistics 6:45-58, 2005

6. van de Wiel MA, van Wieringen WN: CGHregions: Dimension reduc-tion for array CGH data with minimal information loss. Cancer Inform2:55-63, 2007

7. Neve RM, Chin K, Fridlyand J: A collection of breast cancer cell lines forthe study of functionally distinct cancer subtypes. Cancer Cell 10:515-527, 2006

8. Jong K, Marchiori E, Meijer G: Breakpoint identification and smoothingof array comparative genomic hybridization data. Bioinformatics 20:3636-3637, 2004

9. Ichimura K, Mungall AJ, Fiegler H: Small regions of overlapping dele-tions on 6q26 in human astrocytic tumours identified using chromo-some 6 tile path array-CGH. Oncogene 25:1261-1271, 2006

0. Veltman JA, Fridlyand J, Pejavar S: Array-based comparative genomichybridization for genome-wide screening of DNA copy number inbladder tumors. Cancer Res 63:2872-2880, 2003

1. Blaveri E, Brewer JL, Roydasgupta R: Bladder cancer stage and outcomeby array-based comparative genomic hybridization. Clin Cancer Res11:7012-7022, 2005

2. Mastracci TL, Shadeo A, Colby SM: Genomic alterations in lobularneoplasia: a microarray comparative genomic hybridization signaturefor early neoplastic proliferationin the breast. Genes ChromosomesCancer 45:1007-1017, 2006

3. Han W, Han MR, Kang JJ: Genomic alterations identified by arraycomparative genomic hybridization as prognostic markers in tamox-ifen-treated estrogen receptor-positive breast cancer. BMC Cancer

6:92, 2006

4. Yao J, Weremowicz S, Feng B: Combined cDNA array comparativegenomic hybridization and serial analysis of gene expression analysis ofbreast tumor progression. Cancer Res 66:4065-4078, 2006

5. Nessling M, Richter K, Schwaenen C: Candidate genes in breast cancerrevealed by microarray-based comparative genomic hybridization ofarchived tissue. Cancer Res 65:439-447, 2005

6. Naylor TL, Greshock J, Wang Y: High resolution genomic analysis ofsporadic breast cancer using array-based comparative genomic hybrid-ization. Breast Cancer Res 7:R1186-1198, 2005

7. Chin SF, Teschendorff AE, Marioni JC: High-resolution array-CGH andexpression profiling identifies a novel genomic subtype of ER negativebreast cancer. Genome Biol 8:R215, 2007

8. Morrison C, Radmacher M, Mohammed N: MYC amplification andpolysomy 8 in chondrosarcoma: array comparative genomic hybridiza-tion, fluorescent in situ hybridization, and association with outcome.J Clin Oncol 23:9369-9376, 2005

9. Swede H, Bartos JD, Chen N: Genomic profiles of colorectal cancersdiffer based on patient smoking status. Cancer Genet Cytogenet 168:98-104, 2006

0. Kim MY, Yim SH, Kwon MS: Recurrent genomic alterations with im-pact on survival in colorectal cancer identified by genome-wide arraycomparative genomic hybridization. Gastroenterology 131:1913-1924, 2006

1. Nakao K, Mehta KR, Fridlyand J: High-resolution analysis of DNA copynumber alterations in colorectal cancer by array-based comparativegenomic hybridization. Carcinogenesis 25:1345-1357, 2004

2. Ishizuka T, Tanabe C, Sakamoto H: Gene amplification profiling ofesophageal squamous cell carcinomas by DNA array CGH. BiochemBiophys Res Commun 296:152-155, 2002

3. Weiss MM, Kuipers EJ, Postma C: Genomic alterations in primarygastric adenocarcinomas correlate with clinicopathological characteris-tics and survival. Cell Oncol 26:307-317, 2004

4. Buffart TE, Carvalho B, Hopmans E: Gastric cancers in young and elderlypatients show different genomic profiles. J Pathol 211:45-51, 2007

5. Korshunov A, Sycheva R, Golanov A: Genetically distinct and clinicallyrelevant subtypes of glioblastoma defined by array-based comparativegenomic hybridization (array-CGH). Acta Neuropathol (Berl) 111:465-474, 2006

6. Jonkers YM, Claessen SM, Feuth T: Novel candidate tumour suppressorgene loci on chromosomes 11q23-24 and 22q13 involved in humaninsulinoma tumourigenesis. J Pathol 210:450-458, 2006

7. Wilhelm M, Veltman JA, Olshen AB: Array-based comparative genomichybridization for the differential diagnosis of renal cell cancer. CancerRes 62:957-960, 2002

8. Steinemann D, Skawran B, Becker T: Assessment of differentiation andprogression of hepatic tumors using array-based comparative genomichybridization. Clin Gastroenterol Hepatol 4:1283-1291, 2006

9. Patil MA, Gutgemann I, Zhang J: Array-based comparative genomichybridization reveals recurrent chromosomal aberrations and Jab1 as apotential target for 8q gain in hepatocellular carcinoma. Carcinogenesis26:2050-2057, 2005

0. McCabe MG, Ichimura K, Liu L: High-resolution array-based compar-ative genomic hybridization of medulloblastomas and supratentorialprimitive neuroectodermal tumors. J Neuropathol Exp Neurol 65:549-561, 2006

1. Man TK, Lu XY, Jaeweon K: Genome-wide array comparative genomichybridization analysis reveals distinct amplifications in osteosarcoma.BMC Cancer 4:45, 2004

2. Mayr D, Kanitz V, Anderegg B: Analysis of gene amplification andprognostic markers in ovarian cancer using comparative genomic hy-bridization for microarrays and immunohistochemical analysis for tis-sue microarrays. Am J Clin Pathol 126:101-109, 2006

3. Natrajan R, Little SE, Sodha N: Analysis by array CGH of genomicchanges associated with the progression or relapse of Wilms’ tumour.J Pathol 211:52-59, 2007

4. Natrajan R, Williams RD, Hing SN: Array CGH profiling of favourablehistology Wilms tumours reveals novel gains and losses associated with

relapse. J Pathol 210:49-58, 2006