Gene expression profiling of chemically induced rat mammary gland cancerLiang Shan, Minshu Yu and Elizabeth G.Snyderwine1Chemical Carcinogenesis Section, Laboratory of ExperimentalCarcinogenesis, Center for Cancer Research, National Cancer Institute,NIH, Bethesda, MD 20892-4262, USA1To whom correspondence should be addressed. Tel: þ1 301 496-5688;Fax: þ1 301 496-0734;Email: [email protected] to carcinogens through diet, the atmosphere andother means is generally regarded as influencing humancancer risk, but the impact of specific environmental car-cinogens on human breast cancer incidence is stillunknown. We examined whether distinct chemical carci-nogens induce a unique transcriptional profile in mam-mary gland cancer that is characteristic of the etiologicagent. Rat mammary gland cancers (n ¼ 34) were gener-ated by various carcinogens, including the food-derivedheterocyclic amines 2-amino-1-methyl-6-phenylimidazo[4,5-b]pyridine and 2-amino-3,8-dimethylimidazo[4,5-f ]quinoxaline, 7,12-dimethylbenz[a]anthracene, N-nitro-somethylurea and 4-aminobiphenyl. The histopathologyof the carcinomas was graded using a modified Scarff-Bloom-- Richardson scheme and the gene expression pro-files in the carcinomas were evaluated on a 10K cDNAmicroarray. Unsupervised hierarchical clustering analysisrevealed two major clusters of carcinomas irrespective ofthe carcinogenic agent that distinguished two groups withdifferent histopathological parameters (degree of diff eren-tiation, nuclear grade, mitotic activity, epithelial cellgrowth pattern and necrosis). Using class comparisonanalysis and hierarchical clustering of all carcinomas irres-pective of histopathology, gene expression profiles werefurther shown to be statistically differentially expressedaccording to the carcinogenic agent. These findings indi-cate that the transcriptional program in carcinomas isunique to the etiologic agent and can be observed amonga diverse set of carcinogens despite variations in carcinomahistopathology. The ability to use microarray analysis todiscern an etiology-specific profile among a pathologicallyheterogeneous group of breast carcinomas may ultimatelybe valuable in determining the role of environmentalchemical carcinogens in human breast cancer risk.IntroductionHuman breast cancer is a heterogeneous disease with respect topathology, biochemistry and etiology. While environmentalcarcinogen exposure is often regarded as contributing tohuman cancer risk, identifying specific etiological factors inbreast cancer remains a challenge in cancer research (1-- 4).The rat has served as a valuable model for understanding thedevelopment of human breast cancer because of similarities inpathology, cell of origin and hormone dependency (5,6).Furthermore, the rat model provides a means to study specificetiological factors in breast carcinogenesis in a way notfeasible in humans.Recent cDNA microarray analysis of rat mamm ary glandcancer has suggested that the profile of gene expression maybe characteristic of the etiological agent (7,8). Comparison ofthe gene expression profiles in histologically similar ratmammary gland cancers revealed carcinogen-specific geneexpression profiles in carcinomas induced by 2-amino-1-methyl-6-phenylimidazo[4,5-b]pyridine (PhIP), a heterocyclicamine in the human diet, and 7,12-dimethylbenz[a]anthracene(DMBA), an experimental carcinogen belonging to the class ofpolycyclic aromatic hydroc arbons. PhIP- and DMBA-inducedcarcinomas could be distinguished by array tools, includinghierarchical clustering and multidimensional scaling analysis.Although pioneering, these prior studies used just two carc ino-gens and only examined carcinomas of similar histopathology.These studies therefore begged the quest ion of whether micro-array analysis could distinguish breast cancers according toetiology when carcinomas were induced by multiple chemicalcarcinogens and were histologically varied. In the light of thediversity in human brea st cancers and wide variations in envir-onmental exposures among individuals, the feasibility of usingmicroarray analysis to define the etiology of human breastcancers would likely depend on the ability to categorize spe-cific carcinomas from a heterogeneous group of breast cancers.Herein we have analyzed the gene expression profiles of34 rat mammary gland carcinomas induced by carcino-gens including PhIP, DMBA, N-nitrosomethylurea (NMU),2-amino-3,8-dimethylimidazo[4,5-f ]quinoxaline (MelQx) and4-aminobiphenyl (4ABP). These carcinogens represent majorclasses of environm ental carcinogens that have been linked tovarious human cancers and include the cooked meat derivedheterocyclic amines, aromatic amines, polycyclic aromatichydrocarbons and nitrosoureas (1,9,10). The curr ent studyextends earlier work to determine whether carcinogens inducecharacteristic and etiology-specific gene expressi on profiles inmammary gland cancers that are resolvable by microarrayanalysis.Materials and methodsChemical carcinogens and rat mammary gland tumor inductionFemale Sprague-- Dawley rats (43 days old) were obtained from the NIH animalsupply (Animal Production Area, Frederick, MD). All animals were providedNIH Lab Chow and water ad libitum and housed in a NIH animal facility on a12 h light/12 h dark cycle. Mammary gland carcinomas were induced by PhIP-HCl (Toronto Research Chemicals, North York, Canada), DMBA (Sigma,St Louis, MO), NMU (Sigma), MelQx (Toronto Research Chemicals, Toronto,Canada) and 4ABP (provided by F.F.Kadlubar, NCTR, Jefferson, AR).PhIP-HCl, MelQx and 4ABP were all given at 75 mg/kg p.o. once per dayAbbreviations: 4ABP, 4-aminobiphenyl; DMBA, 7,12-dimethylbenz[a]anthracene; MDS, multidimensional scaling; MelQx, 2-amino-3,8-dimethylimidazo[4,5-f ]quinoxaline; NMU, N-nitrosomethylurea; PhIP,2-amino-1-methyl-6-phenylimidazo[4,5-b]pyridine.Carcinogenesis vol.26 no.2#Oxford University Press 2005; all rights reserved. 503Carcinogenesis vol.26 no.2 pp.503-- 509, 2005doi:10.1093/carcin/bgh330for 10 days over a 12 day period. The incidence of carcinomas for PhIPand MeIQx have been reported (11,12). Five of 11 rats treated with 4ABPdeveloped carcinomas. Other rats were treated with a single dose of DMBA(75 mg/kg p.o.) or NMU (50 mg/kg i.p.). The vehicles used to dissolve thecarcinogens were water for PhIP, 0.9% NaCl solution (pH 4.0) for NMU andcorn oil for the other carcinogens. All rats were maintained on a defined highfat diet after carcinogen exposure (9). Samples