Comparative genomic hybridization (2025)

U.S. patent number 7,238,484 [Application Number 11/017,493] was granted by the patent office on 2007-07-03 for comparative genomic hybridization. This patent grant is currently assigned to The Regents of the University of California. Invention is credited to Joe W. Gray, Anne Kallioniemi, Olli-Pekka Kallioniemi, Daniel J. Pinkel, Frederic Waldman.

Comparative genomic hybridization

Abstract

Disclosed are new methods comprising the use of in situhybridization to detect abnormal nucleic acid sequence copy numbersin one or more genomes wherein repetitive sequences that bind tomultiple loci in a reference chromosome spread are eithersubstantially removed and/or their hybridization signalssuppressed. The invention termed Comparative Genomic Hybridization(CGH) provides for methods of determining the relative number ofcopies of nucleic acid sequences in one or more subject genomes orportions thereof (for example, a tumor cell) as a function of thelocation of those sequences in a reference genome (for example, anormal human genome). The intensity(ies) of the signals from eachlabeled subject nucleic acid and/or the differences in the ratiosbetween different signals from the labeled subject nucleic acidsequences are compared to determine the relative copy numbers ofthe nucleic acid sequences in the one or more subject genomes as afunction of position along the reference chromosome spread.Amplifications, duplications and/or deletions in the subjectgenome(s) can be detected. Also provided is a method of determiningthe absolute copy numbers of substantially all RNA or DNA sequencesin subject cell(s) or cell population(s).

Prior Publication Data

Related U.S. Patent Documents

References Cited [Referenced By]

U.S. Patent Documents

Foreign Patent Documents

Other References

Primary Examiner: Fredman; Jeffrey
Attorney, Agent or Firm: Beyer Weaver LLP Haliday; EmilyM.

Government Interests

This invention was made with Government support under Grant Nos.CA-45919 and CA-44768, awarded by the National Institute of Health.The Government has certain rights in this invention.

Parent Case Text

RELATED APPLICATIONS

This application is a continuation of U.S. Ser. No. 09/644,140,filed on Aug. 22, 2000 now abandoned, which is a continuation ofU.S. Ser. No. 08/903,095, filed on Jul. 30, 1997, now U.S. Pat. No.6,159,685, which is a continuation of U.S. Ser. No. 08/468,629,filed on Jun. 6, 1995, now U.S. Pat. No. 5,721,098, which is acontinuation of U.S. Ser. No. 08/166,147, filed on Dec. 14, 1993,now abandoned, which is a continuation of U.S. Ser. No. 07/969,948,filed on Oct. 30, 1992, now abandoned, which is acontinuation-in-part of U.S. Ser. No. 07/846,659, filed Mar. 4,1992, now abandoned. Each of the applications cited above isincorporated by reference in its entirety.

Claims

The invention claimed:

1. A method of surveying all chromosomes in a human genome for thepresence of a copy number variation in one or more unique DNAsequence(s), said method comprising: (a) labeling DNA comprisingunique sequences from each chromosome from a first sample with afirst label; (b) contacting the labeled DNA with a plurality oftarget nucleic acids comprising unique sequences from eachchromosome under conditions that permit hybridization, whereineither the labeled DNA or the target nucleic acids, or both, havehad repetitive sequences, if initially present, blocked and/orremoved; and (c) comparing the intensities of the signals from thelabeled DNA hybridized to each target nucleic acid; wherein saidmethod additionally comprises performing a second survey of allchromosomes in a human genome in a second hybridization, saidsecond survey comprising: labeling DNA comprising unique sequencesfrom each chromosome from a second sample with a label, wherein thelabel is the same as or different from the first label; andcontacting the labeled DNA from the second sample with a pluralityof target nucleic acids that comprises the same unique sequences asused for the first survey under conditions that permithybridization, wherein either the labeled DNA or the target nucleicacids, or both, have had repetitive sequences, if initiallypresent, blocked and/or removed; and said comparing comprisescomparing the intensities of the signals from labeled DNAhybridized to each target nucleic acid to determine one or morecopy number variations in sequences in the first sample relative tosubstantially identical sequences in the second sample.

2. The method of claim 1, wherein the first sample comprises tumortissue and the second sample comprises a normal tissue.

3. A method of surveying all chromosomes in a human genome for thepresence of a copy number variation in one or more unique DNAsequence(s), said method comprising: (a) labeling DNA comprisingunique sequences from each chromosome from a first sample with afirst label; (b) contacting the labeled DNA with a plurality oftarget nucleic acids comprising unique sequences from eachchromosome under conditions that permit hybridization, whereineither the labeled DNA or the target nucleic acids, or both, havehad repetitive sequences, if initially present, blocked and/orremoved; and (c) comparing the intensities of the signals from thelabeled DNA hybridized to each target nucleic acid; wherein saidmethod additionally comprises: labeling DNA comprising uniquesequences from each chromosome from a second sample with a secondlabel, wherein the first and second labels are different; andwherein: said contacting comprises contacting labeled DNA from thefirst and second samples with the plurality of target nucleicacids, wherein either the labeled DNA or the target nucleic acids,or both, have had repetitive sequences, if initially present,blocked and/or removed; and said comparing comprises comparing theintensities of the signals from labeled DNA hybridized to eachtarget nucleic acid to determine one or more copy number variationsin sequences in the first sample relative to substantiallyidentical sequences in the second sample.

4. The method of claim 3, wherein said comparing step comprisesdetermining the ratio of signal intensity of the labeled DNA fromsaid first and second samples to each target nucleic acid.

5. The method of claim 4, wherein said comparing step additionallycomprises comparing the ratio for one target nucleic acid with theratio for another target nucleic acid.

6. The method of claim 3, wherein said method additionallycomprises performing a second survey of all chromosomes in a humangenome in a second hybridization, said second survey comprising:labeling DNA comprising unique sequences from each chromosome fromthird and forth samples with third and fourth labels, respectively,wherein the third and fourth labels are different from each other;and contacting the labeled DNA from the third and fourth sampleswith a plurality of target nucleic acids that comprises the sameunique sequences as used for the first survey under conditions thatpermit hybridization, wherein either the labeled DNA or the targetnucleic acids, or both, have had repetitive sequences, if initiallypresent, blocked and/or removed, and said comparing comprisescomparing the intensities of the signals from labeled DNAhybridized to each target nucleic acid to determine one or morecopy number variations in sequences in one sample relative tosubstantially identical sequences another sample.

7. The method of claim 6, wherein the second and fourth samples arethe same.

8. The method of claim 7, wherein the first and third samples eachcomprise test samples, and the second and fourth samples comprise areference sample.

9. The method of claim 8, wherein the test samples comprise tumortissues, and the reference sample comprises normal tissue.

Description

FIELD OF THE INVENTION

This invention relates generally to the field of cytogenetics, andmore particularly to the field of molecular cytogenetics. Itconcerns methods of determining the relative copy numbers ofdifferent nucleic acid sequences in a subject cell or cellpopulation and/or comparing the nucleic acid sequence copy numbersof substantially identical sequences in several cells or cellpopulations as a function of the location of those sequences in areference genome. For instance, the methods of this inventionprovide the means to determine the relative number of copies ofnucleic acid sequences in one or more subject genomes (for example,the DNA of one tumor cell or a number of cells from a subregion ofa solid tumor) or portions thereof as a function of the location ofthose sequences in a reference genome (for example, a normal humanmetaphase spread). Further, the invention provides methods ofdetermining the absolute copy number of nucleic acid sequences in asubject cell or cell population.

Although the examples herein concern human cells and the languageis primarily directed to human concerns, the concept of thisinvention is applicable to genomes from any plant or animal. Thegenomes compared need only be related closely enough to havesufficient substantially identical sequences for a meaningfulanalysis. For example, a human genome and that of another primatecould be compared according to the methods of this invention.

BACKGROUND OF THE INVENTION

Chromosome abnormalities are associated with genetic disorders,degenerative diseases, and exposure to agents known to causedegenerative diseases, particularly cancer, German, "Studying HumanChromosomes Today," American Scientist, 58: 182 201 (1970); Yunis;"The Chromosomal Basis of Human Neoplasia," Science, 221: 227 236(1983); and German, "Clinical Implication of Chromosome Breakage,"in Genetic Damage in Man Caused by Environmental Agents, Berg, Ed.,pgs. 65 86 (Academic Press, New York, 1979). Chromosomalabnormalities can be of several types, including: extra or missingindividual chromosomes, extra or missing portions of a chromosome(segmental duplications or deletions), breaks, rings andchromosomal rearrangements, among others. Chromosomal or geneticrearrangements include translocations (transfer of a piece from onechromosome onto another chromosome), dicentrics (chromosomes withtwo centromeres), inversions (reversal in polarity of a chromosomalsegment), insertions, amplifications, and deletions.

Detectable chromosomal abnormalities occur with a frequency of onein every 250 human births. Abnormalities that involve deletions oradditions of chromosomal material alter the gene balance of anorganism and generally lead to fetal death or to serious mental andphysical defects. Down syndrome can be caused by having threecopies of chromosome. 21 instead of the normal 2. This syndrome isan example of a condition caused by abnormal chromosome number, oraneuploidy. Down syndrome can also be caused by a segmentalduplication of a subregion on chromosome 21 (such as, 21q22), whichcan be present on chromosome 21 or on another chromosome. Edwardsyndrome (18+), Patau syndrome (13+), Turner syndrome (XO) andKleinfelter syndrome (XXY) are among the most common numericalaberrations. [Epstein, The Consequences of Chromosome Imbalance:Principles, Mechanisms and Models (Cambridge Univ. Press 1986);Jacobs, Am. J. Epidemiol, 105: 180 (1977); and Lubs et al.,Science, 169: 495 (1970).]

Retinoblastoma (del 13q14), Prader-Willis syndrome (del 15q11 q13),Wilm's tumor (del 11p13) and Cri-du-chat syndrome (del 5p) areexamples of important disease linked structural aberrations. [Noraand Fraser, Medical Genetics: Principles and Practice, (Lea andFebiger (1989).]

One of the critical endeavors in human medical research is thediscovery of genetic abnormalities that are central to adversehealth consequences. In many cases, clues to the location ofspecific genes and/or critical diagnostic markers come fromidentification of portions of the genome that are present atabnormal copy numbers. For example, in prenatal diagnosis, asindicated above, extra or missing copies of whole chromosomes arethe most frequently occurring genetic lesion. In cancer, deletionor multiplication of copies of whole chromosomes or chromosomalsegments, and higher level amplifications of specific regions ofthe genome, are common occurrences.

Much of such cytogenetic information has come over the last severaldecades from studies of chromosomes with light microscopy. For thepast thirty years cytogeneticists have studied chromosomes inmalignant cells to determine sites of recurrent abnormality toglean hints to the location of critical genes. Even thoughcytogenetic resolution is limited to several megabases by thecomplex packing of DNA into the chromosomes, this effort hasyielded crucial information. Among the strengths of suchtraditional cytogenetics is the ability to give an overview of anentire genome at one time, permitting recognition of structuralabnormalities such as inversions and translocations, as well asdeletions, multiplications, and amplifications of whole chromosomesor portions thereof. With the coming of cloning and detailedmolecular analysis, recurrent translocation sites have beenrecognized as involved in the formation of chimeric genes such asthe BCR-ABL fusion in chronic myelogeneous leukemia (CML);deletions have been recognized as frequently indicating thelocation of tumor suppressor genes; and amplifications have beenrecognized as indicating overexpressed genes.

Conventional procedures for genetic screening and biologicaldosimetry involve the analysis of karyotypes. A karyotype is theparticular chromosome complement of an individual or of a relatedgroup of individuals, as defined both by the number and morphologyof the chromosomes usually in mitotic metaphase. It include suchthings as total chromosome number, copy number of individualchromosome types (e.g., the number of copies of chromosome X), andchromosomal morphology, e.g., as measured by length, centromericindex, connectedness, or the like. Karyotypes are conventionallydetermined by chemically staining an organism's metaphase, prophaseor otherwise condensed (for example, by premature chromosomecondensation) chromosomes. Condensed chromosomes are used because,until recently, it has not been possible to visualize interphasechromosomes due to their dispersed condition and the lack ofvisible boundaries between them in the cell nucleus.

A number of cytological techniques based upon chemical stains havebeen developed which produce longitudinal patterns on condensedchromosomes, generally referred to as bands. The banding pattern ofeach chromosome within an organism usually permits unambiguousidentification of each chromosome type [Latt, "Optical Studies ofMetaphase Chromosome Organization," Annual Review of Biophysics andBioengineering, 5: 1 37 (1976)].

Unfortunately, such conventional banding analysis requires cellculturing and preparation of high quality metaphase spreads, whichis time consuming and labor intensive, and frequently difficult orimpossible. For example, cells from many tumor types are difficultto culture, and it is not clear that the cultured cells arerepresentative of the original tumor cell population. Fetal cellscapable of being cultured, need to be cultured for several weeks toobtain enough metaphase cells for analysis.

Over the past decade, methods of in situ hybridization have beendeveloped that permit analysis of intact cell nuclei--interphasecytogenetics. Probes for chromosome centromeres, whole chromosomes,and chromosomal segments down to the size of genes, have beendeveloped. With the use of such probes, the presence or absence ofspecific abnormalities can be very efficiently determined; however,it is tedious to test for numerous possible abnormalities or tosurvey to discover new regions of the genome that are altered in adisease.

The present invention, Comparative Genomic Hybridization (CGH)[formerly called Copy Ratio Reverse Cytogenetics (CRRC) among othernames] provides powerful methods to overcome many of thelimitations of existing cytogenetic techniques. When CGH isapplied, for example, in the fields of tumor cytogenetics andprenatal diagnosis, it provides methods to determine whether thereare abnormal copy numbers of nucleic acid sequences anywhere in thegenome of a subject tumor cell or fetal cell or the genomes fromrepresentative cells from a tumor cell population or from a numberof fetal cells, without having to prepare condensed chromosomespreads from those cells. Thus, cytogenetic abnormalities involvingabnormal copy numbers of nucleic acid sequences, specificallyamplifications and/or deletions, can be found by the methods ofthis invention in the format of an immediate overview of an entiregenome or portions thereof. More specifically, CGH provides methodsto compare and map the frequency of nucleic acid sequences from oneor more subject genomes or portions thereof in relation to areference genome. It permits the determination of the relativenumber of copies of nucleic acid sequences in one or more subjectgenomes (for example, those of tumor cells) as a function of thelocation of those sequences in a reference genome (for example,that of a normal human cell).

Gene amplification is one of several mechanisms whereby cells canchange phenotypic expression when increased amounts of specificproteins are required, for example, during development [Spradlingand Mahowald, PNAS (USA), 77: 1096 1100 (1980); Glover et al., PNAS(USA), 79: 2947 2951 (1982)], or during an environmental challengewhen increased amounts of specific proteins can impart resistanceto cytotoxic agents [Melera et al., J. Biol. Chem. 255: 7024 7028(1980); Beach and Palmiter, PNAS (USA, 78: 2110 2114 (1981)].

A major limitation of Southern analysis and related conventionaltechniques for analysis of gene amplification is that only specificsites are studied leaving the vast majority of the genomeunexamined. Conventional cytogenetic studies, on the other hand,provide a broad survey of the genome but provide little informationabout genes that may be involved in amplification events. However,the procedures of this invention overcome those limitations. Thisinvention can be used to show the normal chromosomal locations ofall regions of a genome that are amplified or deleted wherein thesize of the regions that can be detected is limited only by theresolution of the microscopy used and the organization of DNA incondensed chromosomes. Thus, this invention provides among otheruses the ability to study gene amplifications and deletions andtheir roles in tumor development, progression and response totherapy more thoroughly than was possible previously. The methodsof CGH are sufficiently rapid and simple that large numbers ofsubject nucleic acids, for example from many tumors, can beanalysed in studies for gene amplification and deletion.

The karyotypic heterogeneity in solid tumors can be extreme.Identification of commonly occurring chromosomal changes byanalysis of metaphase spreads is often difficult or impossibleusing conventional banding analysis because of the complexity ofthe rearrangements and because of the poor quality of the metaphasepreparations. CGH overcomes that limitation in that the tumornucleic acid can be studied without the requirement of preparingmetaphase spreads. Since CGH can probably be performed on singlecells by amplifying the nucleic acid therefrom, CGH can be used toinvestigate the heterogeneity of tumors by studying representativecells from different cell populations of the tumor. Alternatively,CGH of nucleic acid from a tumor extracted in a bulk extractionprocess from many cells of the tumor can reveal consistencieswithin the apparent heterogeneity. For example, the same amplifiedsequences may appear as homogeneously staining regions (HSRs)and/or double minute chromosomes (DMs) in one tumor cell but as anextension of a chromosome arm in another tumor cell. Thus, orderfrom the apparent randomness may be realized by CGHhybridization.

Montgomery et al., PNAS (USA), 80: 5724 5728 (September 1983),concerns the hybridization of labeled Cot fractionated DNAs fromtumor cell lines (a Cot fraction from which the high copy repeats,low copy repeats and single copy sequences were substantiallyremoved) to metaphase spreads from said tumor cell lines.Basically, Montgomery et al. mapped the positions of nucleic acidsequences from tumor cell lines that are very highly amplified backto tumor cell line genomes.

Total genomic DNA from one species has been used in in situhybridization to discriminate in hybrid cells between chromosomesof that species and of a different species on the basis of thesignal from the high copy repetitive sequences. [Pinkel et al.,PNAS (USA), 83: 2934 (1986); Manuelidis, Hum. Genet., 71: 288(1985); and Durnam et al., Somatic Cell Molec. Genet., 11: 571(1985).]

Landegent et al., Hum. Genet., 77: 366 370 (1987), eliminatedhighly repetitive sequences, like Alu and Kpn fragments, from wholecosmid cloned genomic sequences by blocking the highly repetitivesequences with Cot-1 DNA. The resulting probe was used for in situhybridization.

European Patent Application Publication No. 430,402 (published Jun.5, 1991) describes methods and compositions for chromosome-specificpainting, that is, methods and compositions for stainingchromosomes based upon nucleic acid sequence employing highcomplexity nucleic acid probes. In general in thechromosome-specific painting methods, repetitive sequences notspecific to the targeted nucleic acid sequences are removed fromthe hybridization mixture and/or their hybridization capacitydisabled, often by blocking with unlabeled genomic DNA or with DNAenriched for high copy repetitive sequences as is Cot-1[commercially available from Bethesda Research Laboratory,Gaithersburg, Md. (USA)]. Pinkel et al., PNAS (USA), 85: 9138 9142(1988) also describes aspects of chromosome-specific painting aswell as International Publication No. WO 90/05789 (published May31, 1990 entitled "in situ Suppression Hybridization and UsesTherefor").

Chromosome-specific repeat sequence probes and chromosome-specificpainting probes can be hybridized in situ to interphase nuclei aswell as metaphase spreads and provide information about the geneticstate of the individual targeted genomes. A limitation of suchhybridizations is that cytogenetic information is only providedfrom the regions to which the probes bind. Such hybridizations arevery useful for determining if a particular abnormality is present,for example, the deletion of a specific gene or a duplication amongother abnormalities, but it is laborious to search for currentlyunknown abnormalities on a region by region basis.

Other methods of searching for unknown genetic abnormalitiessimilarly require a lot of work. For example, looking for loss ofheterozygosity in tumor cells, requires the hybridization of manyprobes to Southern blots of tumor and normal cell DNA. The instantinvention, Comparative Genomic Hybridization (CGH), providesmethods to overcome many of the limitations of the existingcytogenetic techniques.

Saint-Ruf et al., Genes, Chromosomes & Cancer, 2: 18 26 (1990)state at page 24 that Human breast carcinomas are characterized bytwo sets of molecular anomalies. Firstly, some proto-oncogenes,such as MYC, INT2, HST, and ERBB2, are frequently found eitheramplified or overexpressed . . . . Secondly, loss of heterozygosityhas been reported, especially for 1p, 11, 13 and 17 . . . Humanbreast carcinomas are also characterized cytogenetically by variousanomalies that may be the chromosomal counterpart of the molecularanomalies: regions of amplification (HSRs) are found in more thanone-third of the tumors . . . , and various deletions, affecting,e.g., 1p, 11p, 11q, 13, and 17p, are found recurrently . . . .[Citations omitted.] Saint-Ruf et al. concluded from the reportedexperiments that although amplification of genetic material is afrequent and probably important event in breast carcinogenesis,that the relevant genes involved in such amplifications remainunknown but do not seem to correspond to the proto-oncogenescommonly considered important in breast cancer.

Since HSRs in tumors are most often not at the site of theamplified gene(s) in normal cells, standard cytogenetics does notyield any information that could assist with identification of thegene(s). CGH on the other hand permits mapping them in the normalgenome, a major step towards their identification.

Dutrillaux et al., Cancer Genet. Cytocenet., 49: 203 217 (1990)report (at page 203) that "[a]lthough human breast carcinomas areamong the most frequent malignant tumors, cytogenetic data remainscarce, probably because of their great variability and of thefrequent difficulty of their analysis." In their study of "30 caseswith relatively simple karyotypes to determine which anomaliesoccur the most frequently and, in particular, early during tumorprogression" (p. 203), they concluded that "trisomy 1q and monosomy16q are early chromosomal changes in breast cancer, whereas otherdeletions and gain of 8q are clearly secondary events." [Abstract,p. 203.] Dutrillaux et al. further state (at page 216) thatdeletions within tumor suppressor genes "characterize tumorprogression of breast cancer."

It is believed that many solid tumors, such as breast cancer,progress from initiation to metastasis through the accumulation ofseveral genetic aberrations. [Smith et al., Breast Cancer Res.Treat., 18 Suppl. 1: S 5 14 (1991); van de Vijver and Nusse,Biochim. Biophys. Acta, 1072: 33 50 (1991); Sato et al., CancerRes., 50: 7184 7189 (1990).] Such genetic aberrations, as theyaccumulate, may confer proliferative advantages genetic instabilityand the attendant ability to evolve drug resistance rapidly, andenhanced angiogenesis, proteolysis and metastasis. The geneticaberrations may affect either recessive "tumor suppressor genes" ordominantly acting oncogenes. Deletions and recombination leading toloss of heterozygosity (LOH) are believed to play a major role intumor progression by uncovering mutated tumor suppressoralleles.

Dominantly acting genes associated with human solid tumorstypically exert their effect by overexpression or alteredexpression. Gene amplification is a common mechanism leading toupregulation of gene expression. [Stark et al., Cell, 75: 901 908(1989).] Evidence from cytogenetic studies indicates thatsignificant amplification occurs in over 50% of human breastcancers. [Saint-Ruf et al., supra.] A variety of oncogenes havebeen found to be amplified in human malignancies. Examples of theamplification of cellular oncogenes in human tumors is shown inTable 1 below.

TABLE-US-00001 TABLE 1 Amplified Degree of DM or HSR Gene TumorAmplification Present c-myc Promyelocytic leukemia 20x + cell line,HL60 Small-cell lung 5 30x ? carcinoma cell lines N-myc Primaryneuroblastomas 5 1000x + (stages III and IV) and neuroblastoma celllines Retinoblastoma cell 10 200x + line and primary tumorsSmall-cell lung carcinoma 50x + cell lines and tumors L-mycSmall-cell lung carcinoma 10 20x ? cell lines and tumors c-mybAcute myeloid leukemia 5 10x ? Colon carcinoma cell lines 10x ?c-erbB Epidermoid carcinoma cell 30x ? Primary gliomas ? c-K-ras-2Primary carcinomas of lung, 4 20x ? colon, bladder, and rectumN-ras Mammary carcinoma cell 5 10x ? line SOURCE: modified fromVarmus, Ann. Rev. Genetics, 18: 553 612 (1984) [cited in Watson etal., Molecular Biology of the Gene (4th ed.; Benjamin/cummingsPublishing Co. 1987)]

Chromosomal deletions involving tumor suppressor genes may play animportant role in the development and progression of solid tumors.The retinoblastoma tumor suppressor gene (Rb-1), located inchromosome 13q14, is the most extensively characterized-tumorsuppressor gene [Friend et al., Nature, 323: 643 (1986); Lee etal., science, 235: 1394 (1987); Fung et al., Science, 236: 1657(1987)]. The Rb-1 gene product, a 105 kDa nuclear phosphoprotein,apparently plays an important role in cell cycle regulation [Lee etal., supra (1987); Howe et al., PNAS (USA), 87: 5883 (1990)].Altered or lost expression of the Rb protein is caused byinactivation of both gene alleles either through a point mutationor a chromosomal deletion. Rb-1 gene alterations have been found tobe present not only in retinoblastomas [Friend et al., supra(1986); Lee et al., supra (1987); Fung et al., supra (1987)] butalso in other malignancies such as osteosarcomas [Friend et al.,supra (1986)], small cell lung cancer [Hensel et al., Cancer Res.,50: 3067 (1990); Rygaard et al., Cancer Res., 50: 5312 (1990)] andbreast cancer [Lee et al., Science, 241: 218 (1988); T'Ang et al.,Science, 242: 263 (1988); Varley et al., Oncogene, 4: 725 (1989)].Restriction fragment length polymorphism (RFLP) studies haveindicated that such tumor types have frequently lost heterozygosityat 13q suggesting that one of the Rb-1 gene alleles has been lostdue to a gross chromosomal deletion [Bowcock et al., Am. J. Hum.Genet., 46: 12 (1990)].

The deletion of the short arm of chromosome 3 has been associatedwith several-cancers, for example, small cell lung cancer, renaland ovarian cancers; it has been postulated that one or moreputative tumor suppressor genes is or are located in the p regionof chromosome 3 (ch. 3p) [Minna et al., Symposia on QuantitativeBiology, Vol. L1: 843 853 (SCH Lab 1986); Cohen et al., N. Eng. J.Med., 301: 592 595 (1979); Bergerham et al., Cancer Res., 49: 13901396 (1989); Whang-Peng et al., Can. Genet. Cytogenet., 11: 91 106(1984; and Trent et al., Can. Genet. Cytogenet., 14: 153 161(1985)].

The above-indicated collection of amplified and deleted genes isfar from complete. As the Saint-Ruf et al. study (supra) ofoncogene amplification in cells showing cytogenetic evidence ofamplification, such as double minutes (DMs) or homogeneouslystaining regions (HSRs), indicated, the amplified genes were notknown oncogenes in most cases. As Dutrillaux et al., supraindicated, "cytogenetic data remains scarce" for "the most frequentmalignant tumors"--breast carcinomas.

Discovery of genetic changes involved in the development of solidtumors has proven difficult. Karyotyping is impeded by the lowyield of high quality metaphases and the complex nature ofchromosomal changes [Teyssier, J. R., Cancer Genet. Cytogenet., 37:103 (1989)]. Although molecular genetic studies of isolated tumorDNA have been more successful and permitted detection of commonregions of allelic loss, mutation or amplification [Fearon et al.,Cell, 61: 759 (1990); Sato et al., Cancer Res., 50: 7184 (1990);Alitalo et al., Adv. Cancer Res., 47: 235 (1986); and Schwab andAmler, Genes Chrom. Cancer., 1: 181 (1990)], such molecular methodsare highly focused, targeting one specific gene or chromosomeregion at a time, and leaving the majority of the genomeunexamined.

Thus, a research tool leading to the identification of amplifiedand deleted genes and providing more cytogenetic data regardingtumors, especially tumor progression and invasiveness is needed intumor cytogenetics. CGH provides such a molecular cytogeneticresearch tool.

CGH facilitates the genetic analysis of tumors in that it providesa copy number karyotype of the entire genome in a single step.Regions of tumor DNA gain and loss are mapped directly onto normalchromosomes. Comparisons of primary tumors with their metastases byCGH should be informative concerning cancer progression.

The ability to survey the whole genome in a single hybridization isa distinct advantage over allelic loss studies by restrictionfragment length polymorphism (RFLP) that target only one locus at atime. RFLP is also restricted by the availability andinformativeness of polymorphic probes.

The copy number karyotype determined by CGH may become as importantfor diagnostic and/or prognostic assessment of solid tumors asconventional karyotyping now is for hematologic malignancies.[Yunis, J. J., Science, 221: 227 (1983); Solomon et al., Science,254: 1153 (1991).]

SUMMARY OF THE INVENTION

Comparative Genomic Hybridization (CGH) employs the kinetics of insitu hybridization to compare the copy numbers of different DNA orRNA sequences from a sample, or the copy numbers of different DNAor RNA sequences in one sample to the copy numbers of thesubstantially identical sequences in another sample. In many usefulapplications of CGH, the DNA or RNA is isolated from a subject cellor cell population. The comparisons can be qualitative orquantitative. Procedures are described that permit determination ofthe absolute copy numbers of DNA sequences throughout the genome ofa cell or cell population if the absolute copy number is known ordetermined for one or several sequences. The different sequencesare discriminated from each other by the different locations oftheir binding sites when hybridized to a reference genome, usuallymetaphase chromosomes but in certain cases interphase nuclei. Thecopy number information originates from comparisons of theintensities of the hybridization signals among the differentlocations on the reference genome.

Two representative basic approaches are employed in CGH asillustrated herein for the analysis of subject DNAs. In an exampleof the first approach, genomic DNA from a subject cell or cellpopulation of cells is isolated, labeled and hybridized toreference chromosomes, usually in metaphase. In an example of thesecond approach, genomic DNAs from two or more subject cells orcell populations are isolated, differentially labeled, andhybridized to reference chromosomes, usually in metaphase.

The CGH methods of this invention can be qualitative and/orquantitative. A particular utility of CGH is for analysing DNAsequences from subject cell(s) or cell population(s), for examplefrom clinical specimens including tumor and fetal tissues.

An important utility of CGH is to find regions in normal genomeswhich when altered in sequence copy number contribute to disease,as for example, cancer or birth defects. For example, regions atelevated copy number may contain oncogenes, and regions present atdecreased copy number may contain tumor suppressor genes.

A representative CGH method is for comparing copy numbers ofdifferent DNA sequences in a subject cell or cell populationcomprising the steps of:

a) extracting the DNA from the subject cell or from a number ofcells of the subject cell population;

b) amplifying said extracted subject DNA, if necessary;

c) labeling the subject DNA;

d) hybridizing said labeled subject DNA in situ to referencemetaphase chromosomes after substantially removing from the labeledDNA those repetitive sequences that could bind to multiple loci inthe reference metaphase chromosomes, and/or after blocking thebinding sites for those repetitive sequences in the referencemetaphase chromosomes by prehybridization with appropriate blockingnucleic acids, and/or blocking those repetitive sequences in thelabeled DNA by prehybridization with appropriate blocking nucleicacid sequences, and/or including such blocking nucleic acidsequences for said repetitive sequences during said hybridization,wherein the DNA sequences in the labeled subject DNA that bind tosingle copy sequences in the reference metaphase chromosomes aresubstantially retained, and those single copy DNA sequences as wellas their binding sites in the reference metaphase chromosomesremain substantially unblocked both before and during thehybridization;

e) rendering the bound, labeled DNA sequences visualizable, ifnecessary;

f) observing and/or measuring the intensity of the signal from thelabeled subject DNA sequences as a function of position on thereference metaphase chromosomes; and

g) comparing the copy numbers of different DNA sequences of thesubject DNA by comparing the signal intensities at differentpositions on the reference metaphase chromosomes, wherein thegreater the signal intensity at a given position, the greater thecopy number of the sequences in the subject DNA that bind at thatposition. An analogous method can be performed wherein the subjectnucleic acid is RNA.

Further, disclosed are methods wherein two or more subject nucleicacids are analysed by CGH. Exemplary methods are those wherein thesubject nucleic acids are DNA sequences from a subject cell or cellpopulation. Analogous methods may be performed wherein the subjectnucleic acids are RNA. Such an exemplary method is that forcomparing copy numbers of different DNA sequences in one subjectcell or cell population relative to copy numbers of substantiallyidentical sequences in another cell or cell population, said methodcomprising the steps of:

a) extracting the DNA from both of the subject cells or cellpopulations;

b) amplifying said extracted subject DNAs, if necessary;

c) differentially labeling the subject DNAs;

d) hybridizing said differentially labeled subject DNAs in situ toreference metaphase chromosomes after substantially removing fromthe labeled DNAs those repetitive sequences that could bind tomultiple loci in the reference metaphase chromosomes, and/or afterblocking the binding sites for those repetitive sequences in thereference metaphase chromosomes by prehybridization withappropriate blocking nucleic acids, and/or blocking thoserepetitive sequences in the labeled DNA by prehybridization withappropriate blocking nucleic acid sequences, and/or including suchblocking nucleic acid sequences for said repetitive sequencesduring said hybridization;

e) rendering the bound, differentially labeled DNA sequencesvisualizable, if necessary;

f) observing and/or measuring the intensities of the signals fromeach subject DNA, and the relative intensities, as a function ofposition along the reference metaphase chromosomes; and

g) comparing the relative intensities among different locationsalong the reference metaphase chromosomes wherein the greater theintensity of the signal at a location due to one subject DNArelative to the intensity of the signal due to the other subjectDNA at that location, the greater the copy number of the sequencethat binds at that location in the first subject cell or cellpopulation relative to the copy number of the substantiallyidentical sequence in the second subject cell or cell populationthat binds at that location.

Further disclosed are methods of quantitatively comparing copynumbers of different DNA sequences in one subject cell or cellpopulation relative to copy numbers of substantially identicalsequences in another subject cell or cell population. Arepresentative method is that comprising steps (a) through (e) ofthe method immediately detailed above and the following stepsof:

f. measuring the intensities of the signals from each of the boundsubject DNAs and calculating the ratio of the intensities as afunction of position along the reference metaphase chromosomes toform a ratio profile; and

g. quantitatively comparing the ratio profile among differentlocations along the reference metaphase chromosomes, said ratioprofile at each location being proportional to the ratio of thecopy number of the DNA sequence that bind to that location in thefirst subject cell or cell population to the copy number ofsubstantially identical sequences in the second cell or cellpopulation.

Said representative methods can further comprise comparing copynumbers of different DNA sequences in more than two subject DNAswherein the comparing is done pairwise between the signals fromeach subject DNA.

This invention further discloses methods to determine the ratio ofcopy numbers of different DNA sequences in one subject cell or cellpopulation to copy numbers of substantially identical sequences inanother cell or cell population wherein the steps of (a) through(f) as described above are performed as well as the followingsteps:

g. determining the average copy number of a calibration sequence inboth subject cells or cell populations, said calibration sequencebeing substantially identical to a single copy sequence in thereference metaphase cells; and

h. normalizing the ratio profile calculated in (f) so that at thecalibration position, the ratio profile is equal to the ratio ofthe average copy numbers determined in (g), the normalized ratioprofile at any other location along the reference metaphasechromosomes thereby giving the ratio of the copy numbers of the DNAsequences in the two subject DNAs that bind at that location. Thatmethod can be extended to further subject nucleic acids as forexample determining the ratio of copy numbers of DNA sequences inmore than two subject DNAs wherein the comparing is done pairwisebetween signals from each subject DNA.

Further disclosed are methods for comparing copy numbers ofdifferent DNA sequences in a test cell or cell population, saidmethod comprising applying steps (a) through (e) of theabove-described methods and

f. observing and/or measuring the intensities of the signal fromeach subject DNA, and the relative intensities, as a function ofposition along the reference metaphase chromosomes wherein one ofthe subject cells or cell populations is the test cell or cellpopulation and the other is a normal cell or cell population;and

(g) comparing the relative intensities among different locationsalong the reference metaphase chromosomes, wherein the greater therelative intensity at a location, the greater the copy number ofthe sequence in the test cell or cell population that binds to thatlocation, except for sex chromosomes where the comparison needs totake into account the differences in copy numbers of sequences inthe sex chromosomes in relation to those on the autosomes in thenormal subject cell or cell population.

A related representative method is that for comparing the copynumber of different DNA sequences in a test cell or cell populationcomprising applying steps (a) through (e) of the above describedmethods wherein one of the subject cells or cell populations is thetest cell or cell population, and the other is a standard cell orcell population wherein the copy numbers of the DNA sequences thatbind to different positions on the reference metaphase chromosomesis known and steps:

f. measuring the intensities of the signals from each of the boundsubject DNAs and calculating the ratio of intensities as a functionof position along the reference metaphase chromosomes to form aratio profile;

g. adjusting the ratio profile at each location along the referencemetaphase chromosomes by multiplying the ratio profile by the knowncopy number of DNA sequences in the standard cell or cellpopulation that bind there; and

h. comparing the adjusted ratio profiles at different locationsalong the reference metaphase chromosomes wherein the greater theadjusted ratio profile at a location, the greater the copy numberof the DNA sequence in the test cell or cell population that bindsthere.

Another related representative method is that for determining theratios of the copy numbers of different DNA sequences in a testcell or cell population, said method comprising applying steps (a)through (f) of the immediately above-described method and the stepsof adjusting the ratio profile at each location along the referencemetaphase chromosomes by multiplying the ratio profile by the knowncopy number of sequences that bind there; and calculating the ratioof the copy number of a DNA sequence in the test cell or cellpopulation that binds to one location on the reference metaphasechromosomes to the copy number of a sequence that binds to anotherlocation by dividing the adjusted ratio profile at the location ofthe first sequence by that at the location of the second. Saidrepresentative method can be extended to determine the copy numberof different DNA sequences in a test cell or cell populationwherein steps (a) through (f) as described above are followed andthen the following steps of adjusting the ratio profile at eachlocation along the reference metaphase chromosomes by multiplyingthe ratio profile by the known copy number of DNA sequences in thestandard cell or cell population that bind there;

determining the copy number of a calibration sequence in the testcell or cell population that is substantially identical to a singlecopy sequence in the reference cells; and

normalizing the adjusted ratio profile so that at the location ofthe calibration sequence on the reference metaphase chromosomes,the normalized, adjusted ratio profile is equal to the copy numberof the calibration sequence determined in the above step, the valueof the normalized, adjusted ratio profile at another location thenbeing equal to the copy number of the DNA sequence in the test cellor cell population that binds at that location. That method can beanalogously performed wherein two or more calibration sequences areused, and the adjusted ratio profile is normalized to get the bestfit to the copy numbers of the ensemble of calibration sequences.Preferably, the copy number of the calibration sequence isdetermined by in situ hybridization. Those methods can comprise insitu hybridizing probes for more than one calibration position andnormalizing to obtain the best fit of the ratio profile to thecalibration positions. The standard cell or cell populationpreferably have normal genomes. In many applications of CGH, thereference metaphase chromosomes are normal.

Further, this invention concerns the use of antenna cell lines. Anexemplary method is for detecting amplification of a certainsequence or group of sequences in a subject cell or cellpopulation, comprising essentially steps (a) through (e) of theabove-described methods wherein the in situ hybridization istargeted to antenna cells in which the DNA sequence(s) to be testedfor is or are amplified, and examining the reference cell forregions that are hybridized significantly more intensely thanothers, the presence of such regions indicating amplifications ofthe sequence(s) which are being tested. The chromosomes of saidantenna cell lines may be in interphase or in metaphase.

When a single labeled subject nucleic acid is being hybridized, orif multiple labeled subject nucleic acids are hybridizedsequentially, it is important that the binding sites on thereference genome not be saturated prior to observing and/ormeasuring the signal intensity(ies). In the case of a singlelabeled subject nucleic acid, nonsaturation can be effected in anumber of ways, for example, by stopping the hybridization, byproviding insufficient subject nucleic acid, and/or by providing asufficient amount of unlabeled nucleic acid which is sufficientlycomplementary to the reference chromosomes to competitively preventsaturation of sites therein by the labeled subject nucleicacid.

When there are two or more labeled subject nucleic acids, thosesubject nucleic acids can be hybridized in situ to the referencegenome sequentially or simultaneously. Simultaneous in situhybridization is preferred in that saturation of the targetedbinding sites in the reference genome will not interfere with theprocedure. When sequential in situ hybridization is used, it mustbe performed under conditions wherein the individual hybridizationsare stopped well before the binding sites on the referencechromosomes are saturated.

Objects of this invention are to detect sequence copy numberimbalances throughout an entire genome in one hybridization, to mapgains and/or losses of sequences in a genome, and/or to provide acopy number karyotype of a subject genome.

Further, an object of this invention is to enable the detection ofrelative copy number differences that are common to a number ofdifferent cells and/or cell populations. For example, CGH methodscan be used wherein DNAs extracted from cells of many differenttumors are combined and labeled; the hybridization of thosecombined labeled DNAs to normal condensed chromosomes, provides forthe rapid identification of only those copy number changes thatoccurred in most of the tumors. Less frequently occurringvariations would be averaged out. Thus, this invention furtherprovides for a CGH method wherein two or more of the subjectnucleic acids that were extracted from different cells and/or fromnumbers of cells from different cell populations, are labeled thesame, and hybridized to a reference spread under conditions whereinrepetitive sequences are removed and/or suppressed and whereinsequence copy number differences that are common in said combinedlabeled nucleic acid sequences are determined.

Another object of this invention is to provide the means ofcytogenetically analysing archived chromosomal material, that is,fixed material from, for example, biopsied tissue specimens,preferably cataloged and keyed to medical records of patients fromwhom the specimens were taken, and archaeological chromosomalmaterial. Such chromosomal material cannot, of course, bekaryotyped according to traditional means in that no live cells arepresent to culture and from which to prepare chromosomal spreads.However, the nucleic acid can be extracted therefrom and amplifiedby a polymerase chain reaction (PCR) procedure or by a non-PCRprocedure and tested by the methods of this invention.

This invention further provides for a method to detectsimultaneously an ensemble of amplifications and/or deletions in atumor wherein the results can be used to determine the subsequentbehavior of that tumor. Said determination is made by associatingthe patterns of amplifications and/or deletions in tumor cells withthe behavior of that tumor. Such associations can be made bytesting, for example, as indicated immediately above, DNA fromarchived tumor tissue keyed to medical records, or when fresh tumorspecimens are tested by CGH and the patients are followed. Further,such associations can be made with CGH methods wherein there aremore than one subject cell and/or cell population, for example, oneor more tumors.

Another object of this invention is to provide a method ofanalyzing cells from a suspected lesion at an early stage ofdevelopment. An advantage of the methods of this invention is thatonly a few cells are necessary for the analysis. The earlydetection of amplifications and/or deletions in cells from a lesionallow for early therapeutic intervention that can be tailored tothe extent of, for example, invasiveness known to be associatedwith such genetic rearrangements. Further, such early detectionprovides a means to associate the progression of the cells with thegenetic rearrangements therein detected by the methods of thisinvention.

Tumors can be karyotypically heterogeneous containing thereinvarious populations of cells each having different types of geneticrearrangements. As indicated above tumor cells are difficult toculture, and it is not clear that cultured cells are representativeof the original tumor cell population. This invention provides themeans to by-pass the culturing obstacle and allows geneticcharacterization of tumor cells and thus, of the heterogeneity oftumors by testing cells from different subregions thereof accordingto the methods of this invention. Bulk extraction of the nucleicacid from many cells of a tumor can also be used to test forconsistent amplifications and/or deletions within a tumor.

It is another object of this invention to provide methods ofdetecting amplifications and/or deletions of nucleic acid sequenceswherein certain cell lines termed herein "antenna cell lines", areused to enhance the sensitivity of the detection.

It is still further an object of this invention to provide methodsof prenatal or perinatal analysis wherein the nucleic acid of thechild's cells is extracted and tested according to the methods ofthis invention. In one embodiment of CGH, such material is humanand hybridized to a normal human metaphase spread to detect whetherany deletions and/or amplifications are therein present, forexample, an extra copy of chromosome 21, diagnostic for Downsyndrome. Test kits for performing CGH methods are alsoprovided.

BRIEF DESCRIPTION OF THE FIGURES

The file of this patent contains at least one drawing executed incolor. Copies of this patent with color drawings will be providedby the Patent and Trademark Office upon request and payment of thenecessary fee.

FIG. 1 illustrates the results of a CGH hybridization of DNA fromthe BT-474 human breast cancer cell line to a metaphase spread ofnormal peripheral blood lymphocyte human chromosomes. The BT-474cell line is known to have a 13-fold c-erbB-2 amplification. TheDNA from that cell line was labeled with digoxigenin-11-dUTP andstained with fluorescein isothiocyanate (FITC); signals from thehybridization of the cell line DNA are green in thephotomicrograph. A chromosome 17 peri-centromeric repeat probe(cosmid cK17.10) was labeled with biotin-14-dATP and stained withTexas Red; signals from that probe's hybridization are red. Thechromosomal DNA was counterstained with4,6-diamidino-2-phenylindole (DAPI) resulting in a bluecounterstaining. The photomicrograph was taken using a multicolorimage analysis system after contrast stretching and pseudocolordisplay.

The green signals indicating amplified sequences in the BT-474 cellline are seen in FIG. 1 at the following loci: 17q12 (the erbB-2locus), 17q22 q23 and 20q13-ter. The latter two sites werepreviously unrecognized sites of amplification in that cell line.One centromeric repeat is non-specifically stained green.

FIG. 2 schematically illustrates the general approach used inperforming the methods of this invention--Comparative GenomicHybridization (CGH). The reference chromosome spread is hybridizedwith various nucleic acid mixtures, either simultaneously or atdifferent times, to obtain the desired information. Representativemixtures could include unlabeled sequences designed to blocksequences in the various other nucleic acid pools, for example, thehigh-copy repetitive sequences in human genomic DNA; unlabeledcompetitor nucleic acid to prevent saturation of the target sitesfor the labeled mixtures, for example, human genomic DNA within afactor of 10 of the concentration used for the labeled subjectnucleic acids (see FIG. 5); and one or more pools of sequences ofdifferent origin that are differently labeled so that their bindingcan be independently assessed, for example, tumor and normalgenomic DNA (see FIGS. 6 and 7). The information on the sequencefrequency of the labeled pools is obtained by analysis of theintensity of the individual signals and/or the differences inratios of intensities among the signals as a function of positionalong the reference chromosomes.

FIG. 3 outlines general aspects of the CGH procedure used inExample 1, infra. The reference chromosome spread, in this examplenormal human chromosomes, is first hybridized for about one hourwith a high concentration of unlabeled human genomic DNA (FIG. 3a).That prehybridization blocks many of the high copy repetitivesequences in the chromosomes so that the high copy repetitivesequences in the labeled subject nucleic acid, in this case labeledtumor DNA, will not substantially contribute to the signal duringthe subsequent hybridization. The labeled tumor DNA, and perhapssome competitor DNA or other comparison nucleic acid are thenhybridized to the target reference spread (FIG. 3b). Cot-1 DNA canbe included in the hybridization as in Example 1, below to blockmore effectively the centromeric repetitive sequences in thelabeled subject nucleic acids.

FIG. 3 is representative of one way of reducing signals fromrepetitive sequences. Other methods are detailed herein infra. Ineach of the CGH methods including the procedures outlined in therest of the figures, some means of reducing the signal from therepetitive sequences is used, but not specifically indicated in thefigures. It is important for CGH that the signal from each subjectnucleic acid be dominated by sequences that bind to well-definedloci. Total suppression of the signal from the genomic repeats isnot necessary, but the poorer the suppression, the less able theprocedure is to detect small differences in sequence frequency.

FIG. 4 illustrates the procedure used in Example 1, for whichrepresentative results are shown in the photomicrographs of FIGS. 1and 8. As shown in FIG. 4a, labeled human tumor DNA is hybridizedto a normal human chromosome spread. [Please note as indicated inthe description for FIG. 3, provisions were made to suppress thesignal from the repetitive sequences although those provisions arenot specifically indicated in the figure. Example 1 details onepreferred method to suppress the hybridization signals fromrepetitive sequences.] In this representative example, the tumorDNA is assumed to contain a region wherein some sequences arehighly amplified, for example, an amplicon containing an oncogene.The amplified sequences in the tumor DNA may be clustered andintegrated in some tumor chromosomes; they may be integrated intomultiple places in the tumor genome; or, they may exist asextra-chromosomal elements. The sequences of the amplicon will mapto some chromosomal location in the reference genome, which in thiscase is a normal human genome.

FIG. 4b illustrates the kinetics of the build-up of the signal on atarget reference chromosome. The signal builds more rapidly in theamplified region since more copies of those sequences are availablefor hybridization. If the reaction is stopped before the targetchromosome is saturated, or if insufficient labeled DNA is added toachieve saturation, then the genomic region that was amplified inthe tumor will appear higher in intensity on the normal chromosomeas illustrated by the dark band on the left reference chromosome.The more intensely labeled region (dark band) indicates thelocation and extent of the amplicon as reflected in the referencegenome. Thus, the amplification is detected without prior knowledgeof its existence, and the origin of the amplified sequences ismapped in the normal human genome.

If the reaction illustrated in FIG. 4b is allowed to proceed tosaturation of the target sites, contrast is lost, as shown by therepresentative reference chromosome on the right. Thus, in thisembodiment of CGH, it is important to stop the hybridization beforesaturation of the target or provide insufficient probe forsaturation. The graphs schematically show the build-up of thehybridization signal in the region that was amplified (graph onright) and in the remainder that was unamplified (graph on theleft). The arrows connect the chromosomal regions with the times ofobservation on the kinetic curve.

FIG. 5 illustrates an embodiment of CGH that avoids the potentialsaturation of the target as shown in the right portion of FIG. 4b.In this representative example, the reference nucleic acid is ahuman chromosome spread; the subject nucleic acid is labeled tumorDNA. If unlabeled human genomic DNA is included with the labeledtumor DNA in excess, in this case at a five-fold higherconcentration than that of the labeled tumor DNA, then anysaturation of the target will be due to a combination of labeledand unlabeled copies of the nucleic acid sequences, rather thanjust labeled copies as shown in the right of FIG. 4b. [Once again,as indicated in FIGS. 3 and 4 the means of reducing the signal fromrepetitive sequences is not indicated in this figure, but it isassumed that some protocol is performed to remove substantially therepetitive sequences that would bind to multiple loci in thereference genome and/or to block such sequences from binding to thetarget.]

At the early stages of the reaction, the amplified region willbuild up faster than elsewhere in the chromosome (for example ifthe sequence is amplified five-fold, it would build up 5 times asfast) and will be detectable as in the left portion of FIG. 4b.However as the reaction proceeds to saturation, the unamplifiedregions of the chromosome reach only one-fifth (1/5) of theintensity shown in FIG. 4b, because most of the sites are filled byunlabeled copies of the sequences. On the other hand, a sequencethat was amplified five-fold in the tumor would reach one-half(1/2) of the saturation intensity since an equal number of labeledand unlabeled copies of those sequences are present. Thus, contrastis maintained according to this embodiment at all stages of thereaction, although it changes as the reaction proceeds.

FIG. 6 illustrates an embodiment of CGH designed to enhance itssensitivity in detecting small changes in copy number of varioussequences. When a CGH procedure as indicated in FIG. 5 is followed,intrinsic variation in the saturation levels, or rate of signalbuild-up at different positions in the reference genome may not beindicative of abnormal gain or loss of sequences. Such intrinsicvariations would interfere with interpretation of intensitydifferences as indicating differences in copy number of thesequences. This CGH embodiment overcomes that potential problem byproviding a mixture of labeled subject nucleic acid, in this casetumor DNA labeled with a green fluorochrome, and a differentlylabeled competitor nucleic acid in this case normal human genomicDNA labeled with a red fluorochrome. The two differently labeledDNAs are simultaneously hybridized to the chromosome spread. [Onceagain, removal of the repetitive sequences and/or blocking of thesignal therefrom is performed but not illustrated.] Changes in theratio of green to red along each of the chromosomes in thereference spread then indicate regions of increased or decreasedsequence copy number in the tumor. Those ratio changes may resultin color variations from red to yellow to green on the referencespread.

FIG. 7 graphically and schematically explains the kineticsunderlying the CGH embodiment illustrated in FIG. 6. In the centeris one of the chromosomes of the reference chromosome spread, anormal human chromosome in this case. The darkness of the shadingon the reference chromosome shows the ratio of green to redintensity along the chromosome.

In the amplified region, the green/red ratio is much higher than inthe normal region, whereas in the deleted region the green/redratio is less than in the normal region. The arrows from examplesof each of the different green/red intensity regions point tokinetic curves that indicate the build-up of green (solid line forthe tumor DNA) and red (dashed line for the normal DNA) signalsduring the hybridization. In the normal region, upper left graph,the red and green signals build together. (They have beennormalized to be equal for the purposes of this explanation.) Inthe amplified region, upper right, the green (tumor) signal buildsup much more rapidly than the red (normal) signal, the green/redratio being approximately the level of amplification (given thenormalization to the normal part of the chromosome).

In the lower left of FIG. 7, the signal build-up for the duplicatedregion is shown; the green (tumor) signal is 50% brighter than thered (normal) signal. In the lower right, the build-up for a deletedregion is schematically described; the green (tumor) signal is 50%dimmer than the red (normal) signal. The ratio approach of this CGHembodiment further normalizes for the frequent finding thathybridization to some chromosomes in a spread is intrinsicallybrighter than that for others because of differences in the localhybridization environment.

FIG. 8 illustrates an example of how a deletion can be detectedusing CGH. A deletion is simulated by employing DNA from a humanprimary breast carcinoma (XX) as a subject genome and a normal malechromosome spread (XY) as the reference genome. The absence of theY-chromosome in the tumor DNA was detected, as would acytogenetically significant deletion, by the hybridization. DNAfrom the primary breast carcinoma was labeled withdigoxigenin-11-dUTP and stained with fluorescein isothiocyanate(FITC) (green signals). The normal male peripheral blood lymphocytemetaphase was counterstained with 4,6-diamidino-2-phenylindole(DAPI) (blue). The picture was taken from a multicolor imageanalysis system (QUIPS) after image thresholding and contraststretching. The green chromosomal fluorescence level on allchromosomes was increased to make the absence of this fluorescenceon the Y-chromosome (arrow) more readily visible. The Y-chromosomeis only stained with the DAPI counterstain.

FIG. 9 presents an idiogram of chromosome 1 from the breast cancercell line 600 MPE, the karyotype for which was published by Smithet al., JNCI, 78: 611 615 (1987).

FIG. 10A is a photomicrograph showing the comparative genomichybridization (CGH) of DNA from a 45, X0 cell line (green) and anormal human female DNA (red) to a normal human male referencespread. The reddish color of the X chromosome, pointed out by thelarge arrow, as compared with the autosomes reflects the lowerrelative copy number of the X chromosome sequences in the 45, X0cell line. Faint staining of a small part of the Y chromosome,pointed out by the small arrow, is a result of the binding ofhomologous sequences in the pseudo-autosomal region.

FIG. 10B graphically illustrates the correlation of the number of Xchromosomes in five fibroblast cell lines and the averagegreen-to-red ratio of the X chromosome(s) relative to the sameratio for the autosomes.

FIGS. 11A E illustrates green-to-red fluorescence ratio profiles ofchromosomes 1, 9, 11, 16 and 17 after comparative genomichybridization with breast cancer cell line 600PE (green) and with anormal DNA (red). The profiles reflect the relative copy number ofthe chromosomal regions. Fluorescence in situ hybridization (FISH)with 16p and 16q cosmid probes to interphase and metaphase 600PEcells indicated that there were two signals with 16p cosmid probesand one signal from the 16q cosmid probes. That information on theabsolute copy number of those loci provided by FISH permitsinterpretation of the ratio 1.0 as indicating that there are twocopies of the sequence throughout the genome.

The dip in the profile at 1p34 through 1p36 may represent apreviously unsuspected small interstitial deletion; however, thatobservation has not yet been independently verified with specificprobes for that region.

Centromeric and heterochromatic regions of the genome are notincluded in the analysis because the Cot-1 DNA partially blockssignals in those regions, and the large copy number polymorphismsbetween individual sequences at those loci effect unreliable ratiodata.

FIGS. 12(A) and 12(B) respectively provide green-to-redfluorescence ratio profiles of chromosome 8 (A) and chromosome 2(B) after comparative genomic hybridization respectively with COLO320 HSR (human colon adenocarcinoma cell line) and NCI H69 (smallcell lung carcinoma cell line) cell line DNAs (green) and withnormal human DNA (red). The inserts illustrate the overlaid greenand red fluorescence images of the chromosomes, and the chromosomalmedial axis drawn by the image analysis program used.

In FIG. 12(A), the myc locus at 8q24 shows a highly elevatedgreen-to-red ratio, which is consistent with the known high levelamplification of myc in the COLO 320HSR cell line.

In FIG. 12(B), three regions of amplification are seen onchromosome 2. The signal at 2p24 corresponds to the location ofN-myc known to be amplified in the NCI-H69 cell line. The two otherregions with a highly increased green-to-red fluorescence ratio, at2p21 and 2q21, were not previously known to be amplified in theNCI-H69 cell line.

FIG. 13 is a photomicrograph of a comparative genomic hybridization(CGH) with BT-20 (breast cancer cell line) cell line DNA (green)and normal DNA (red) to a normal human metaphase spread. Loss ofDNA sequences in the tumor cell line DNA relative to normal DNA areshown by red whereas gain of DNA sequences in the tumor cell lineare shown in green.

DETAILED DESCRIPTION

Comparative Genomic Hybridization (CGH) has also been termed CopyRatio Reverse Cytogenetics (CRRC), competition hybridization andquantitative in situ ratio karyotyping (QUIRK). Further, in theembodiment wherein fluorochromes are used as labels, it has beentermed competition FISH (fluorescence in situ hybridization). CGHspecifically provides methods whereby amplifications, duplicationsand/or deletions can be identified in an immediate overview of agenome.

CGH provides methods for determining variations in the copy numberof different elements in a mixture of nucleic acid sequences (forexample, genomic DNA isolated from a tumor) as a function of thelocation of those sequences in the genome of a reference organism(for example, the genome of a normal cell from the same species).The methods comprise the use of in situ hybridization of thenucleic acid sequence mixture to a chromosome spread of thereference organism, and measuring the intensity of thehybridization at different locations along the target chromosomes.Exemplary methods are schematically outlined in FIGS. 2 7. Thoseillustrative examples are not exhaustive but suggest the wide rangeof variations and other uses of the basic approach.

As the figure descriptions indicate, it is critical that signalsfrom repetitive sequences do not dominate the signal from thesubject nucleic acid pool, and that they be removed from the poolor that their signals be suppressed as necessary. It is preferredto exclude sequences from the hybridization or block sequences inthe hybridization mixture that could bind to multiple clearlyseparated positions on the chromosomes, for example, sites that areon different chromosomes, or that are on the same chromosome butare well-separated. In many applications of CGH, it is the highcopy repetitive sequences, such as Alu, Kpn, Lines, andalpha-satellites among others, that are removed from the labeledsubject nucleic acid and/or which are blocked and/or the bindingsites therefor are blocked. Described herein are methods to removeand/or block those repetitive signals. It should be noted thatnucleic acid sequences in the labeled nucleic acid that bind tosingle copy loci are substantially retained in the hybridizationmixture of labeled subject nucleic acids, and such single copysequences as well as their binding sites in the referencechromosome spread remain substantially unblocked relative to therepetitive sequences that bind to multiple loci (that is, loci thatare visually distinguishable) both before and during thehybridization.

The methods of this invention provide the means to identifypreviously unknown regions of amplification and deletion. Forexample, one embodiment of CGH as detailed in Example 1 hereinprovides an efficient method that gives an immediate overview of agenome identifying all regions that are amplified greater thanabout five-fold to ten-fold as well as at least large deletions.More sensitive embodiments that can identify smaller amplificationsand deletions are also disclosed.

Nanogram quantities of the subject nucleic acids are required forthe CGH methods of this invention. Paraffin embedded tumor sectionscan be used as well as fresh or frozen material. Snap frozenmaterial from normal and malignant tissue are preferred for mRNAisolation.

Standard procedures can be used to isolate the required nucleicacid from the subject cells. However, if the nucleic acid, forexample, DNA or mRNA, is to be extracted from a low number of cells(as from a particular tumor subregion) or from a single cell, it isnecesary to amplify that nucleic acid, by a polymerase chainreaction (PCR) procedure or by a non-polymerase chain reaction(non-PCR) procedure. PCR and preferred PCR procedures are describedinfra. Exemplary non-PCR procedures include the ligase chainreaction (LCR) and linear amplification by use of appropriateprimers and their extension (random priming).

Some of the various embodiments of CGH are illustrated,particularly in FIGS. 2 7. In the embodiment illustrated in FIGS. 6and 7, wherein a subject nucleic acid, in this case, human genomicDNA, that is labeled differently from another subject nucleic acid,amplifications and/or deletions are indicated by a change in ratiobetween the different signals, rather than just a change in signalintensity.

The representative examples concerning CGH of Examples 1, 2 and 3below involve the hybridizations of tumor cell line DNA to normalhuman metaphase spreads. However, there are many permutations andcombinations of pairwise and multiple hybridizations of differentnucleic acids from different genomes all of which are considered tobe within the scope of this invention.

For example, CGH could be used to hybridize labeled DNA from atumor cell line to metaphase spreads of that same cell line toestimate the level and pattern of amplification in each cell line,comparing those results to hybridizations of said tumor cell lineDNA to a normal human metaphase spread. Alternatively, labeledtumor cell line DNA and differently labeled human genomic DNA couldbe simultaneously hybridized to a metaphase spread of a tumor cellline metaphase spread. Further, DNA from a primary tumor and thatfrom its metastasis could be differently labeled and hybridized ina CGH method to a normal human metaphase or to a related tumor cellline metaphase. Those are just some of the many examples ofCGH.

Although the examples herein concern the hybridizations of the DNAfrom breast cancer cell lines and primary tumors to normal humanmetaphase spreads, it will be clear to anyone skilled in the artthat CGH is not limited to studying genomes of cancer cells or tothe results of hybridizing abnormal genomes to normal genomes. CGHpermits the comparison of nucleic acid sequence copy frequencies ofany two or more genomes, even genomes of different species if theirnucleic acid sequences are sufficiently complementary to allow formeaningful interpretation. It should be noted regardinginterspecies comparisons that the information obtained by CGHincludes not only an assessment of relative copy number but alsothat of sequence divergence.

It will also be clear to those skilled in the art thathybridization with nucleic acid other than chromosomal DNA, such asmessenger RNA (mRNA) or complementary DNA (c-DNA) of subject cellscan be used to determine the location and level of expression ofgenes in those cells. Conventional methodology is used to extractmRNA from a cell or cell population, and to synthesize in vitroc-DNA by reverse transcription.

CGH does not require the preparation of condensed chromosomes, forexample, metaphase, prophase or other condensed chromosomal states,of the subject genomes. Thus, genomes from which metaphase,prophase or otherwise condensed chromosomal spreads are difficult,time-consuming or not possible to prepare at least in good quality,for example, genomes of tumor cells or fetal cells can be studiedby CGH.

In CGH, labeled subject nucleic acids, for example, labeled tumorDNA, is hybridized to a reference genome, for example, a normalhuman metaphase spread, under conditions in which the signal fromamplified, duplicated and/or deleted nucleic acid sequences fromthe labeled nucleic acid can be visualized with good contrast. Suchvisualization is accomplished by suppressing the hybridization ofrepetitive sequences that bind to multiple loci including the highcopy interspersed and clustered repetitive sequences, such as, Alu,Kpn, Lines, alpha-satellites among others, using unlabeled totalhuman genomic nucleic acid, preferably DNA, and/or therepeat-enriched (Cot-1) fraction of genomic DNA, and/or by removingsuch repetitive sequences from the hybridization mixture. Inproviding the detection sensitivity required, the extent ofsuppression of the hybridization of repetitive sequences and/orremoval thereof can be adjusted to the extent necessary to provideadequate contrast to detect the differences in copy number beingsought; for example, subtler copy number changes may require thesuppression or removal of lower level repetitive sequences.

When combining more than one labeled nucleic acid in ahybridization mixture, the relative concentrations and/or labelingdensities may be adjusted for various purposes. For example, whenusing visual observation or photography of the results, theindividual color intensities need to be adjusted for optimumobservability of changes in their relative intensities. Adjustmentscan also be made by selecting appropriate detection reagents(avidin, antibodies and the like), or by the design of themicroscope filters among other parameters. When using quantitativeimage analysis, mathematical normalization can be used tocompensate for general differences in the staining intensities ofdifferent colors.

The kinetics of the CGH hybridizations are complicated. Since thesubject nucleic acids are frequently double stranded, complementarysequences will reassociate in the hybridiztion mix as well ashybridizing to the target. Such reassociation may result in a morerapid decrease in concentration of the high copy sequences than thelow copy ones, thereby making the signal intensity variations onthe reference chromosomes less pronounced than the copy differencesin the original subject DNAs. In addition, non-specific binding ofthe labeled subject DNAs to the slide, coverslip, etc. maygenerally reduce the concentration of that labeled subject nucleicacid during the hybridization. Those skilled in the art willrecognize numerous methods of optimizing the quantitative aspectsof CGH, such as, mathematical correction of digital images,supplying freshly denatured subject DNA during the hybridization,and adding unlabeled genomic DNA in excess to dominate thereassociation rates.

The resolution of CGH is presently at a level that can be seenthrough a light microscope, as is traditional cytogenetic staining.Thus, if a small sequence in a subject nucleic acid is amplified,to be seen as a signal in a subject genome, it must be amplifiedenough times for its signal to be able to be visualized under alight microscope. For example, the locus for erbB-2 which isrelatively small (very approximately, a few hundred kb), needs tobe amplified at least greater than five times to be visuallydistinguishable under a light microscope when the CGH embodimentused in Example 1 is employed. On the other hand, if a largesection of a chromosome is present at increased frequency in asubject nucleic acid, the signal from that region would show up inthe reference genome at a much lower level of amplification.

The term "labeled" is herein used to indicate that there is somemethod to visualize nucleic acid fragments that are bound to thetarget, whether or not the fragments directly carry some modifiedconstituent. A section infra entitled "Labeling the Nucleic AcidFragments of the Subject Nucleic Acids" describes various means ofdirectly labeling the probe and other labeling means by which thebound probe can be detected.

The phrase "antenna cell line" is herein used to indicate areference genome that has one or more known significant geneticaberrations, for example, a cell line known to have an oncogenethat is highly amplified, for example, in large homogeneouslystaining regions (HSRs). The amplified regions of that cell linewould thus provide a much bigger target site than a normalchromosome spread. Thus, observation of the signal from such alarge target site would be easier in that on average the signalwould be brighter from amplified target sequences in the referencegenome as provided by such an antenna cell line. A subject nucleicacid extracted from, for example, a number of tumor cells, could betested by a CGH hybridization to such an antenna cell line to seeif it also contained amplification(s) of the oncogene known to beamplified in the cell line.

When an antenna cell line is used as the reference genome, thereare instances wherein it can be used in interphase rather than as achromosome spread. For example, if one is checking to see if acertain oncogene is amplified or not in the subject nucleic acid,interphase CGH is sufficient. However, the maximum amount ofinformation is provided when condensed chromosome spreads areused.

A base sequence at any point in the genome can be classified aseither "single-copy" or "repetitive". For practical purposes thesequence needs to be long enough so that a complementary probesequence can form a stable hybrid with the target sequence underthe hybridization conditions being used. Such a length is typicallyin the range of several tens to hundreds of nucleotides.

A "single-copy sequence" is that wherein only one copy of thetarget nucleic acid sequence is present in the haploid genome."Single-copy sequences" are also known in the art as "uniquesequences". A probe complementary to a single-copy sequence has onebinding site in haploid genome. A "repetitive sequence" is thatwherein there is more than one copy of the same target nucleic acidsequence in the genome. Each copy of a repetitive sequence need notbe identical to all the others. The important feature is that thesequence be sufficiently similar to the other members of the familyof repetitive sequences such that under the hybridizationconditions being used, the same fragment of probe nucleic acid iscapable of forming stable hybrids with each copy.

Herein, the terms repetitive sequences, repeated sequences andrepeats are used interchangeably.

The phrase "metaphase chromosomes" in herein defined to encompassthe concept of "condensed chromosomes" and is defined to mean notonly chromosomes condensed in the prophase or metaphase stage ofmitosis but any condensed chromosomes, for example, those condensedby premature chromosome condensation or at any stage in the cellcycle wherein the chromosome can be visualized as an individualentity. It is preferred that the chromosomes in the referencegenome be as long as possible but condensed sufficiently to bevisualized individually.

A subject nucleic acid is herein considered to be the same asanother nucleic acid if it is from a member of the same sex of thesame species and has no significant cytogenetic differences fromthe other nucleic acid. For example, the DNA extracted from normallymphocytes of a human female is considered for the purposes ofthis invention to be the same nucleic acid as that of DNA fromnormal cells of a human female placenta.

The following abbreviations are used herein:

Abbreviations

TABLE-US-00002 AAF N-acetoxy-N-2-acetyl-aminofluorene ATCC AmericanType Culture Collection BN bicarbonate buffer with NP-40 BRLBethesda Research Laboratories bp base pair CCD charge coupleddevice CGH Comparative Genomic Hybridization Chr. chromosomal CMLchronic myelogenous leukemia CRRC Copy Ratio Reverse CytogeneticsDAPI 4,6-diamidino-2-phenylindole dATP deoxyadenosine triphosphateDCS as in fluorescein-avidin DCS (a commercially available cellsorter grade of fluorescein Avidin D) dCTP deoxycytosinetriphosphate dGTP deoxyguanosine triphosphate DI DNA index DMdouble minute chromosome dNTP deoxynucleotide triphosphate dTTPdeoxythymidine triphosphate dUTP deoxyuridine triphosphate EDTAethylenediaminetetraacetate E/P estrogen/progesterone FISHfluorescence in situ hybridization FACS fluorescence-activated cellsorting FITC fluorescein isothiocyanate HPLC high performanceliquid chromatography HSR homogeneously staining region ISCNInternational System for Cytogenetic Nomenclature IB isolationbuffer kb kilobase kDa kilodalton LOH loss of heterozygosity Mbmegabase met. metastasis min minute ml milliliter mM milliMole mmmillimeter ng nanogram NIGMS National Institute of General MedicalSciences NP-40 non-ionic detergent commercially available fromSigma as Nonidet P-40 (St. Louis, MO) PBS phosphate-buffered salinePCR polymerase chain reaction PHA phytohemagglutinin PI propidiumiodide Pl. pleural PMSF phenylmethylsulfonyl fluoride PN mixture of0.1 M NaH.sub.2PO.sub.4 and 0.1 M buffer Na.sub.2HPO.sub.4, pH 8;0.1% NP-40 PNM Pn buffer plus 5% nonfat dry milk buffer(centrifuged); 0.02% Na azide QUIRK quantitative in situ ratiokaryotyping Rb-1 retinoblastoma tumor suppressor gene RFLPrestriction fragment length polymorphism RPM revolutions per minuteSD Standard Deviation SDS sodium dodecyl sulfate SSC 0.15 MNaCl/0.015 M Na citrate, pH 7 Td doubling time ug microgram ulmicroliter um micrometer uM micromole VNTR variable number tandemrepeat

Resolution of differences in copy number can be improved by the useof image analysis and by averaging the results from hybridizationsof a subject nucleic acid to multiple condensed chromosome spreads.Using such methods, the background signal (noise) can bedifferentiated from actual nucleic acid sequence copy numberdifferences.

Image Analysis:

An image analysis system, preferably computer-assisted, can be usedto enhance and/or accurately quantitate the intensity differencesbetween and/or among the signals from a hybridization and thebackground staining differences for more accurate and easierinterpretation of results. Image analysis and methods to measureintensity are described, for example, in Hiraoka et al., Science,238: 36 41 (1987) and Aikens et al., Meth. Cell Biol., 29: 291 313(1989). In such an image analysis system, it is preferred to use ahigh quality CCD camera whose intensity response is known to belinear over a wide range of intensities.

The components of a particular quantitative image processing system(QUIPS) are described in Example 1 under the subheadingFluorescence Microscopy and Interpretation of Results. Asexemplified in Example 1, a computer-assisted image analysis systemwith a filterwheel is used so that the images from the signals andcounterstaining of the DNA are superimposed on one image.Pseudocolors, that is, colors that are not exactly spectrallyconverted, can be displayed. Contrast stretching, wherein thedifferences between the intensity levels of the signals andbackground staining differences are enhanced by adjusting controlsof the image analysis system. Thresholding can also be used whereinthe background staining can be assigned a value close to zero so itwould barely appear in the processed image from such a system.Similarly, computer analysis permits substraction of background,smoothing of fluctuations in the signals, accurate intensity andratio calculations and the ability to average signals onchromosomes in multiple spreads.

Absolute Copy Numbers:

Hybridization of the subject DNAs to the reference chromosomesgives information on relative copy numbers of sequences. Someadditional normalization is required to obtain absolute copy numberinformation. One convenient method to do this is to hybridize aprobe, for example a cosmid specific to some single locus in thenormal haploid genome, to the interphase nuclei of the subject cellor cell population(s) (or those of an equivalent cell orrepresentative cells therefrom, respectively). Counting thehybridization signals in a representative population of such nucleigives the absolute sequence copy number at that location. Giventhat information at one locus, the intensity (ratio) informationfrom the hybridization of the subject DNA(S) to the referencecondensed chromosomes gives the absolute copy number over the restof the genome. In practice, use of more than one reference locusmay be desirable. In this case, the best fit of the intensity(ratio) data through the reference loci would give a more accuratedetermination of absolute sequence copy number over the rest of thegenome.

Thus, the CGH methods of this invention combined with otherwell-known methods in the art can provide information on theabsolute copy numbers of substantially all RNA or DNA sequences insubject cell(s) or cell population(s) as a function of the locationof those sequences in a reference genome. For example, one or morechromosome-specific repeat sequence or high complexity paintingprobes can be hybridized independently to the interphase nuclei ofcells representative of the genomic constitution of the subjectcell(s) or cell population(s). Whole chromosome painting probes arenow available for all the human chromosomes [Collins et al.,Genomics, 11: 997 1006 (1991)]. Specific repeat-sequence probes arealso available [Trask et al., Hum. Genet., 78: 251 (1988) andreferences cited therein; and commercially available from Oncor(Gaithersburg, Md., USA)]. Hybridization with one or more of suchprobes indicates the absolute copy numbers of the sequences towhich the probes bind.

For such interphase analysis, painting probes with a complexity offrom about 35 kb to about 200 kb, are preferred; probes from about35 kb to about 100 kb are further preferred; and still morepreferred are probes having a complexity of from about 35 kb to 40kb, for example, a cosmid probe. Exemplary of such locus-specificpainting probes are any cosmid, yeast artificial chromosomes(YACS), bacterial artificial chromosomes (BACs), and/or p1 phageprobes as appropriate, preferably to the arms of a selectedchromosome. Such cosmid probes, for example, are commerciallyavailable from Clontech [South San Francisco, Calif. (USA)] whichsupplies cosmid libraries for all the human chromosomes. Anotherexample of a cosmid probe that could be used in such methods ofthis invention would be a 3p cosmid probe called cC13-787 obtainedfrom Yusuke Nakamura, M.D., Ph.D. [Division of Biochemistry, CancerInstitute, Toshima, Tokyo, 170, Japan]. Its isolation and mappingto 3p21.2-p21.1 is described in Yamakawa et al., Genomics, 9(3):536 543 (1991). Another example would be a 3q cosmid probe namedJ14R1A12 obtained from Wen-Lin Kuo [Biomedical Department, P.O. Box5507 (L-452), Lawrence Livermore National Laboratory Livermore,Calif. 94550 (USA)]. For interphase analysis, preferred repeatsequence probes are centromeric-specific and/orperi-centromeric-specific repeat sequence probes. Such acentromeric-probe is, for example, the chromosome 17peri-centromeric repeat probe (cosmid ck17.10) and the alphasatellite repeat probe for the centromeric region of chromosome 8,both of which are described in Example 1 infra. A variety of repeatsequence probes are commercially available from Oncor[Gaithersburg, Md. (USA)]. However, the locus-specific paintingprobes are preferred over the repeat sequence probes for themethods of this invention to determine absolute copy numbers ofnucleic acid sequences.

Further, when the subject nucleic acid sequences are DNA, thereference copy numbers can be determined by Southern analysis. Whenthe subject nucleic acid sequences are RNA, the reference copynumbers can be determined by Northern analysis.

Those reference copy numbers or reference frequencies provide astandard by which substantially all the RNA or DNA sequences in thesubject cell(s) or cell population(s) can be determined. CGHmethods are used to determine the relative copy numbers of the restof the sequences. However, absolute copy numbers require a standardagainst which the results of CGH can be determined. Otherwise theCGH procedures would have to be highly standardized and quantitatedto see differences in the absolute copy numbers of sequences in agenome, for example, haploidy, triploidy, octaploidy, wherein thereare 1, 3 and 8 copies of each of the chromosomes, respectively.

PCR and Microdissection:

The mechanics of PCR are explained in Saiki et al., Science, 230:1350 (1985) and U.S. Pat. Nos. 4,683,195, 4,683,202 (both issuedJul. 18, 1987) and U.S. Pat. No. 4,800,159 (issued Jan. 24, 1989).]PCR offers a rapid, sensitive and versatile cell-free molecularcloning system in which only minute amounts of starting materialare required.

A preferred PCR method to amplify the subject nucleic acids fortesting by CGH is a PCR adapter-linker amplification [Saunders etal., Nuc. Acids Res., 17 9027 (1990); Johnson, Genomics, 6: 243(1990) and PCT 90/00434 (published Aug. 9, 1990).] The labeledsubject nucleic acid could be produced by such a adapter-linker PCRmethod from a few hundred cells; for example, wherein the subjectnucleic acid is tumor DNA, the source DNA could be a few hundredtumor cells. Such a method could provide a means to analyse by CGHclonal sub-populations in a tumor.

Another preferred PCR method is a method employing a mixture ofprimers described in Meltzer et al., "Rapid Generation of RegionSpecific Probes by Chromosome Microdissection and theirApplication: A Novel Approach to Identify Cryptic ChromosomalRearrangements," Nature--Genetics, 1(1): 24 28 (April 1992).Microdissection of sites in the reference metaphase spread thatproduce signals of interest in CGH, would permit PCR amplificationof nucleic acid sequences bound at such sites. The amplifiednucleic acid could then be easily recovered and used to probeavailable libraries, as for example, cosmid libraries, so that theamplified sequences could be more rapidly identified.

High copy repetitive sequences can be suppressed in amplifying thesubject nucleic acid by PCR. The PCR primers used for such aprocedure are complementary to the ends of the repetitivesequences. Thus, upon proper orientation, amplification of thesequences flanked by the repeats occurs. One can further suppressproduction of repetitive sequences in such a PCR procedure by firsthybridizing complementary sequences to said repetitive sequenceswherein said complementary sequences have extendednon-complementary flanking ends or are terminated in nucleotideswhich do not permit extension by the polymerase. Thenon-complementary ends of the blocking sequences prevent theblocking sequences from acting as a PCR primer during the PCRprocess. Primers directed against the Alu and L1 repetitive DNAfamilies have allowed the selective amplification of humansequences by interspersed repetitive sequence PCR (IRS-PCR) [Nelsonet al., PNAS, 86: 6686 (1989); Ledbetter et al., Genomics, 6: 475(1990)].

Archived Material

An important aspect of this invention is that nucleic acids fromarchived tissue specimens, for example, paraffin-embedded orformalin-fixed pathology specimens, can be tested by the methods ofCGH. Said nucleic acid cannot, of course, be prepared intochromosome spreads for traditional cytogenetic chemical staining.Also, it is difficult for large enough restriction fragments to beextracted from such material for other conventional research tools,such as Southern analysis. However, the nucleic acid from suchspecimens can be extracted by known techniques such as thosedescribed in Greer et al., Anatomic Pathology, 95(2): 117 124(1991) and Dubeau et al., Cancer Res., 46: 2964 2969 (1986), and ifnecessary, amplified for testing by various CGH methods. Suchnucleic acid can be amplified by using a polymerase chain reaction(PCR) procedure (described above), for example, by the methoddescribed in Greer et al., supra wherein DNA from paraffin-embeddedtissues is amplified by PCR.

A particular value of testing such archived nucleic acid is thatsuch specimens are usually keyed to the medical records of thepatients from whom the specimens were taken. Therefore, valuablediagnostic/prognostic associations can be made between the revealedcytogenetic state of patients' nucleic acid material and themedical histories of treatment and outcome for those patients. Forexample, information gathered by CGH can be used to predict theinvasiveness of a tumor based upon its amplification and/ordeletion pattern matched to associations made with similar patternsof patients whose outcomes are known.

Analogously, other nucleic acid that is fixed by some method, as,for example, archaelogical material preserved through naturalfixation processes, can also be studied by CGH procedures. Asindicated above, copy number differences between species provideinformation on the degree of similarity and divergence of thespecies studied. Evolutionarily important linkages and disjunctionsbetween and among species, extant or extinct, can be made by usingthe methods of CGH.

Tumor Cytogenetics

CGH provides the means to assess the association between geneamplification and/or deletion and the extent of tumor evolution.Correlation between amplification and/or deletion and stage orgrade of a cancer may be prognostically important because suchinformation may contribute to the definition of a genetically basedtumor grade that would better predict the future course of diseasewith more advanced tumors having the worst prognosis. In addition,information about early amplification and/or deletion events may beuseful in associating those events as predictors of subsequentdisease progression. Gene amplification and deletions as defined byCGH to, for example, normal metaphase spreads (genomic site,intensity of the signal and/or differences in signal ratios, andnumber of different genomic sites at which the copy numberdifferences occur) can be associated with other known parameterssuch as tumor grade, histology, Brd/Urd labeling index, hormonalstatus, nodal involvement, tumor size, survival duration and othertumor properties available from epidemiological and biostatisticalstudies. For example, tumor DNA to be tested by CGH could includeatypical hyperplasia, ductal carcinoma in situ, stage I III cancerand metastatic lymph nodes in order to permit the identification ofassociations between amplifications and deletions and stage.

The associations made may make possible effective therapeuticintervention. For example, consistently amplified regions maycontain an overexpressed gene, the product of which may be able tobe attacked therapeutically (for example, the growth factorreceptor tyrosine kinase, p185.sup.HER2).

CGH hybridizations of nucleic acids from cells of primary cancersthat have metastasized to other sites can be used to identifyamplification and/or deletion events that are associated with drugresistance. For example, the subject nucleic acids to be analysedcould be selected so that approximately half are from patientswhose metastatic disease responded to chemotherapy and half frompatients whose tumors did not respond. If gene amplification and/ordeletion is a manifestation of karyotypic instability that allowsrapid development of drug resistance, more amplification and/ordeletion in primary tumors from chemoresistant patients than intumors in chemosensitive patients would be expected. For example,if amplification of specific genes is responsible for thedevelopment of drug resistance, regions surrounding those geneswould be expected to be amplified consistently in tumor cells frompleural effusions of chemoresistant patients but not in the primarytumors. Discovery of associations between gene amplification and/ordeletion and the development of drug resistance may allow theidentification of patients that will or will not benefit fromadjuvant therapy.

Once a new region of amplification or deletion has been discoveredby CGH, it can be studied in more detail using chromosome-specificpainting [Pinkel-et al., PNAS (USA), 85: 9138 9142 (1988); EPPublication No. 430,402 (Jun. 5, 1991)] with a collection of probesthat span the amplified or deleted region. Probes to amplifiedregions will show more signals than centromeric signals from thesame chromosome, whereas probes to nonamplified regions will showapproximately the same number of test and centromeric signals. Forexample, the amplified regions on 17q22 23 and 20qter (discussed asnewly discovered regions of amplification in Example 1) showvariability in size from tumor to tumor using CGH (the 17q22 23region more markedly); it can be expected that the regioncontaining the important gene(s) can be narrowed by mapping theregions of amplification in multiple tumors in more detail to findthe portion that is amplified in all cases. Probes for thosestudies can be selected, for example from specific cosmid librariesproduced by the National Laboratory Gene Library Project and/orfrom the National Institute of Health (NIH) genomic researchprojects.

The c-erbB-2 oncogene, also referred to as HER-2 or neu, encodesfor a 185 kilodalton (Kd) protein. Studies have reported c-erbB-2gene amplification in human mammary tumor cell lines. [Kraus etal., EMBO J. 6: 605 610 (1987); van de Vijver et al., Mol. CellBiol., 7: 2019 2023 (1987).] Also, c-erbB-2 gene amplification inhuman breast cancer has been shown to be associated with diseasebehavior, and may be a predictor of clinical outcome. [Slamon etal., Science, 235: 177 182 (1987); Berger et al., Cancer Res., 48:1238 1243 (1988); Zhou et al., Cancer Res., 47:6123 6125 (1987);and Venter et al., Lancet, 11: 69 71 (1987)]. C-erbB-2 has alsobeen shown to be amplified in ovarian cancers. [Alitalo and Schwab,Advances in Cancer Res., 47: 235 281 (1986).]

C-myc is a proto-oncogene which is the cellular homolog of thetransforming gene of the chicken retrovirus MC29. In humans, c-myclies on the long arm of chromosome 8, at band 124, and spans about5 kilobase pairs. The myc protein is a phosphoprotein present inthe nucleus. The normal function of c-myc is unknown; however, italso certainly plays a role in cell division, and is expressed innormally growing cells as well as in tumor cells. It is now widelybelieved that translocations involving c-myc lead to alteredtranscription of the gene, contributing to malignanttransformation.

Sequences from N-myc member of the myc gene family have been shownto be amplified as much as a thousandfold in some neuroblastomas.N-myc amplifications are usually seen in the later stage III and IVtumors. Some small-cell lung carcinomas also have amplified mycgenes in double minute chromosomes (DMs) and homogeneously stainingregions (HSRs). Myc has also been shown to be amplified in coloncancer. [Alitalo and Schwab, supra.] Again such amplifications arefound in late stages of tumor development, in the so-called variantcells that exhibit a more malignant behavior. Amplifications caninvolve either c-myc, N-myc or another member of the myc genefamily, L-myc. [Watson et al., supra at pp. 1084 1086].

In addition, overexpression has been observed for thep-glycoprotein gene family associated with multi-drug resistanceand for drug metabolizing enzymes such as P450 containing enzymesand glutathione S-transferase. [Fairchild and Cowan, J. RadiationOncol. Biol. Phys., 20: 361 367 (1990).]

Identification of amplified and/or deleted genes is important tothe management of cancer, for example, breast cancer, for severalreasons:

1) to improve prognostication;

2) to detect amplification and/or deletion events that areassociated with the development of drug resistance; and

3) to improve therapy.

For example, in regard to improving prognostication, in breastcancer the amplification of oncogenes, such as int-2, erbB-2 andmyc occur frequently and have been associated with aggressivegrowth and poor prognosis in some studies. [Schwab and Amier,Genes, Chromosomes & Cancer, 1: 181 193 (1990).] In regard toreason (2), gene amplification has clearly been shown to lead todrug resistance in vitro (for example, amplification of thedihydrofolate reductase gene confers resistance to methotrexate),and is likely to occur in patients undergoing therapy as well (forexample, as a result of over expression of glutathioneS-transferase and p-glycoprotein). [Fairchild and Cowan, supra].Thus, the identification of resistance-linked genes would have amajor impact on therapy by allowing therapy modification asresistance-related gene amplification occurs. Therapy could beimproved by targeting for specific therapy, tumors that overexpressspecific amplified genes. Prenatal Diagnosis

Prenatal screening for disease-linked chromosome aberrations (e.g.,trisomy 21) is enhanced by the rapid detection of such abberrationsby the methods and compositions of this invention. CGH analysis isparticularly significant for prenatal diagnosis in that it yieldsmore rapid-results than are available by cell culture methods.

Removal of Repetitive Sequences and/or Disabling the HybridizationCapacity of Repetitive Sequences

The following methods can be used to remove repetitive sequencesand/or disable the hybridization capacity of such repetitivesequences. Such methods are representative and are expressedschematically in terms of procedures well known to those ofordinary skill the art, and which can be modified and extendedaccording to parameters and procedures well known to those in theart.

Bulk Procedures. In many genomes, such as the human genome, a majorportion of distributed (or shared) repetitive DNA is contained in afew families of highly repeated sequences such as Alu. Thesemethods primarily exploit the fact that the hybridization rate ofcomplementary nucleic acid strands increases as their concentrationincreases. Thus, if a mixture of nucleic acid fragments isdenatured and incubated under conditions that permit hybridization,the sequences present at high concentration will becomedouble-stranded more rapidly than the others. The double-strandednucleic acid can then be removed and the remainder used in thehybridizations. Alternatively, the partially hybridized mixture canbe used as the subject nucleic acid, the double-stranded sequencesbeing unable to bind to the target. The following are methodsrepresentative of bulk procedures that are useful for disabling thehybridization capacity of repetitive sequences or removing thosesequences from a mixture.

Self-reassociation. Double-stranded nucleic acid in thehybridization mixture is denatured and then incubated underhybridization conditions for a time sufficient for the high-copysequences in the mixture to become substantially double-stranded.The hybridization mixture is then applied to the referencechromosome spread. The remaining labeled single-stranded copies ofthe highly repeated sequences may bind throughout the referencechromosome spread producing a weak, widely distributed signal.

Use of blocking nucleic acid. Unlabeled nucleic acid sequenceswhich are complementary to those sequences in the hybridizationmixture whose hybridization capacity it is desired to inhibit areadded to the hybridization mixture. The subject nucleic acids andblocking nucleic acid are denatured, if necessary, and incubatedunder appropriate hybridization conditions. The sequences to beblocked become double-stranded more rapidly than the others, andtherefore are unable to bind to the reference spread when thehybridization mixture is applied to the spread. In some cases, theblocking reaction occurs so quickly that the incubation period canbe very short, and adequate results can be obtained if thehybridization mix is applied to the spread immediately afterdenaturation. Further, the probe and the target can besimultaneously denatured in some cases. A blocking method isgenerally described in the context of Southern analysis by Sealy etal., "Removal of Repeat Sequences form Hybridization Probes",Nucleic Acid Research, 13:1905 (1985). Examples of blocking nucleicacids include genomic DNA, a high-copy fraction of genomic DNA andparticular sequences as outlined below.

i. Genomic DNA. Genomic DNA contains all of the nucleic acidsequences of the organism in proportion to their copy-number in thegenome. Thus, adding genomic DNA to the hybridization mixtureincreases the concentration of the high-copy repeat sequences morethan low-copy sequences, and therefore is more effective atblocking the former.

ii. High-copy fraction of genomic DNA. Fractionating the genomicDNA to obtain only the high-copy sequences and using them forblocking can be done, for example, with hydroxyapatite as describedbelow.

Removal of Sequences.

Hydroxyapatite. Single- and double-stranded nucleic acids havedifferent binding characteristics to hydroxyapatite. Suchcharacteristics provide a basis commonly used for fractionatingnucleic acids. Hydroxyapatite is commerically available [e.g.,Bio-Rad Laboratories, Richmond, Calif. (USA)]. The fraction ofgenomic DNA containing sequences with a particular degree ofrepetition, from the highest copy-number to single-copy, can beobtained by denaturing genomic DNA, allowing it to reassociateunder appropriate conditions to a particular value of C.sub.ot,followed by separation using hydroxyapatite. The single- anddouble-stranded nucleic-acid can also be discriminated by use of S1nuclease. Such techniques and the concept of C.sub.ot are explainedin Britten et al., "Analysis of Repeating DNA Sequences byReassociation", in Methods in Enzymology, 29: 363 418 (1974).

Reaction with immobilized nucleic acid. Removal of particularsequences can also be accomplished by attaching single-stranded"absorbing" nucleic acid sequences to a solid support.Single-stranded source nucleic acid is hybridized to theimmobilized nucleic acid. After the hybridization, the unboundsequences are collected and used in CGH. For example, human genomicDNA can be used to absorb repetitive sequences from the subjectnucleic acids. One such method is described by Brison et al.,"General Method for Cloning Amplified DNA by Differential Screeningwith Genomic Probes," Molecular and Cellular Biology, 2: 578 587(1982). Briefly, minimally sheared human genomic DNA is bound todiazonium cellulose or a like support. The source DNA,appropriately cut into fragments, is hybridized against theimmobilized DNA to C.sub.ot values in the range of about 1 to 100.The preferred stringency of the hybridization conditions may varydepending on the base composition of the DNA.

Prehybridization. Blocking of repeat sequence binding sites in thereference genome by hybridization with unlabeled complementarysequences will prevent binding of labeled sequences in the subjectnucleic acids that have the potential to bind to those sites. Forexample, hybridization with unlabeled genomic DNA will render thehigh-copy repetitive sequences in the reference genomedouble-stranded. Labeled copies of such sequences in the subjectnucleic acids will not be able to bind when they are subsequentlyapplied.

In practice, several mechanisms can be combined to produce thedesired contrast and sensitivity.

Labeling the Nucleic Acid Fragments of the Subject NucleicAcids

There are many techniques available for labeling single- anddouble-stranded nucleic acid fragments of the subject nucleicacids. They include incorporation of radioactive labels, e.g.Harper et al. Chromosoma, 83: 431 439 (1984); direct attachment offluorochromes or enzymes, e.g. Smith et al., Nuc. Acids Res., 13:2399 2412 (1985), and Connolly et al., Nuc. Acids Res., 13: 44854502 (1985); and various chemical modifications of the nucleic acidfragments that render them detectable immunochemically or by otheraffinity reactions, e.g. Tchen et al., "Chemically Modified NucleicAcids as Immunodetectable Probes in Hybridization Experiments,"PNAS, 81: 3466 3470 (1984); Richardson et al., "Biotin andFluorescent Labeling of RNA Using T4 RNA Ligase," Nuc. Acids Res.,11: 6167 6184 (1983); Langer et al., "Enzymatic Synthesis ofBiotin-Labeled Polynucleotides: Novel Nucleic Acid AffinityProbes," PNAS, 78: 6633 6637 (1981); Brigati et al., "Detection ofViral Genomes in Cultured Cells and Paraffin-Embedded TissueSections Using Biotin-Labeled Hybridization Probes," Virol., 126:32 50 (1983); Broker et al., "Electron Microscopic Visualization oftRNA Genes with Ferritin-Avidin: Biotin Labels," Nuc. Acids Res.,5: 363 384 (1978); Bayer et al., "The Use of the Avidin BiotinComplex as a Tool in Molecular Biology," Methods of Biochem.Analysis, 26: 1 45 (1980); Kuhlmann, Immunoenzyme Techniques inCytochemistry (Weinheim, Basel, 1984). Langer-Safer et al., PNAS(USA), 79: 4381 (1982): Landegent et al., Exp. Cell Res., 153: 61(1984); and Hopman et al., Exp. Cell Res., 169: 357 (1987). Thus,as indicated, a wide variety of direct and/or indirect means areavailable to enable visualization of the subject nucleic sequencesthat have hybridized to the reference genome. Suitable visualizingmeans include various ligands, radionuclides, fluorochromes andother fluorescers, chemiluminescers, enzyme substates orco-factors, particles, dyes and the like. Some preferred exemplarylabeling means include those wherein the probe fragments arebiotinylated, modified with N-acetoxy-N-2-acetylaminofluorene,modified with fluorescein isothiocyanate or other fluorochromes,modified with mercury/TNP ligand, sulfonated, digoxigeninated orcontain T-T dimers.

A preferred method of labeling is tailing by terminal transferaselabeling. Another preferred method is random priming with mixedsequence primers followed by polymerase extension. This has theadditional feature of amplifying the amount of subject DNA, ifseveral cycles are used, which is useful when only a small amountof DNA was originally obtained from the subject cell or cellpopulation.

The key feature of labeling is that the subject nucleic acidfragments bound to the reference spread be detectable. In somecases, an intrinsic feature of the subject nucleic acid, ratherthan an added feature, can be exploited for this purpose. Forexample, antibodies that specifically recognize RNA/DNA duplexeshave been demonstrated to have the ability to recognize probes madefrom RNA that are bound to DNA targets [Rudkin and Stollar, Nature,265:472 473 (1977)]. The RNA used is unmodified. Nucleic acidfragments can be extended by adding "tails" of modified nucleotidesor particular normal nucleotides. When a normal nucleotide tail isused, a second hybridization with nucleic acid complementary to thetail and containing fluorochromes, enzymes, radioactivity, modifiedbases, among other labeling means, allows detection of the boundnucleic acid fragments. Such a system is commercially availablefrom Enzo Biochem [Biobridge Labeling System; Enzo Biochem Inc.,New York, N.Y.(USA)].

Another example of a means to visualize the bound nucleic acidfragments wherein the nucleic acid sequences do not directly carrysome modified constituent is the use of antibodies to thymidinedimers. Nakane et al., ACTA Histochem. Cytochem., 20 (2):229(1987), illustrate such a method wherein thymine-thymine dimerizedDNA (T-T DNA) was used as a marker for in situ hybridization. Thehybridized T-T DNA was detected immunohistochemically using rabbitanti-T-T DNA antibody.

All of the labeling techniques disclosed in the above referencesmay be preferred under particular circumstances. Further, anylabeling techniques known to those in the art would be useful tolabel the subject nucleic acids in of this invention. Severalfactors govern the choice of labeling means, including the effectof the label on the rate of hybridization and binding of thenucleic acid fragments to the chromosomal DNA, the accessibility ofthe bound nucleic acid fragments to labeling moieties applied afterinitial hybridization, the mutual compatibility of the labelingmoieties, the nature and intensity of the signal generated by thelabel, the expense and ease in which the label is applied, and thelike.

Several different subject nucleic acids, each labeled by adifferent method, can be used simultaneously. The binding ofdifferent nucleic acids can thereby be distinguished, for example,by different colors.

In Situ Hybridization.

Application of the subject nucleic acids to the referencechromosome spreads is accomplished by standard in situhybridization techniques. Several excellent guides to the techniqueare available, e.g., Gall and Pardue, "Nucleic Acid Hybridizationin Cytological Preparations," Methods in Enzymology, 21: 470 480(1981); Henderson, "Cytological Hybridization to MammalianChromosomes," International Review of Cytology, 76: 1 46 (1982);and Angerer et al., "in situ Hybridization to Cellular RNAs," inGenetic Engineering: Principles and Methods, Setlow and Hollaender,Eds., Vol. 7, pgs. 43 65 (Plenum Press, New York, 1985).

Generally in situ hybridization comprises the following majorsteps: (1) fixation of tissue or biological structure to beexamined, (2) prehybridization treatment of the biologicalstructure to increase accessibility of target DNA, and to reducenonspecific binding, (3) hybridization of the mixture of nucleicacids to the nucleic acid in the biological structure or tissue;(4) posthybridization washes to remove nucleic acid fragments notbound in the hybridization and (5) detection of the hybridizednucleic acid fragments. The reagents used in each of these stepsand their conditions of use vary depending on the particularsituation.

Under the conditions of hybridization wherein human genomic DNA isused as an agent to block the hybridization capacity of therepetitive sequences, the preferred size range of the nucleic acidfragments is from about 200 bases to about 1000 bases, morepreferably about 400 to 800 bases for double-stranded,nick-translated nucleic acids and about 200 to 600 bases forsingle-stranded or PCR adapter-linker amplified nucleic acids.

Example 1 provides details of a preferred hybridization protocol.Basically the same hybridization protocols as used forchromosome-specific painting as described in Pinkel et al., PNAS(USA), 85: 9138 9142 (1988) and in EP Pub. No. 430,402 (publishedJun. 5, 1991) are adapted for use in CGH.

The following representative examples of performing CGH methods ofthis invention are for purposes of illustration only and are notmeant to limit the invention in any way.

EXAMPLE 1

DNA from Breast Cancer Lines Hybridized to Normal MetaphaseSpreads

In this Example, methods of this invention to analyse genomes byComparative Genomic Hybridization (CGH) are exemplified byhybridizations of breast cancer cell lines to normal metaphasespreads. The target metaphase spreads were pre-hybridized withunlabeled human placental DNA to block the high copy repeatsequences. In this representative example, the hybridizationmixture containing the extracted labeled DNA from the cell linescontained unlabeled, repeat-enriched Cot-1 blocking DNA [obtainedfrom Bethesda Research Laboratories (BRL), Gaithersburg, Md.(USA].

The experiments outlined below include in the hybridization mixturefor the subject genomes, that is, the breast cancer cell line DNAs,chromosome-specific repeat sequence probes and chromosome-specificpainting probes. Those probes labeled with biotin were included asan adjunct for identifying chromosomes in the metaphasepreparations. The experiments were first performed without thosechromosome-specific probes. Then each chromosome of interest wasmeasured to determine its length which was considered along withother factors to determine its probable identity. Thechromosome-specific probes were then used in the hybridizationmixture to confirm the identity of the chromosome of interest.However, such probes are not necessary as the chromosomes couldhave been identified by the DAPI banding of the counterstain or byother chemical staining, such as staining with quinacrine, by askilled cytogeneticist.

Cell Lines and Isolation of DNA:

Six established breast cancer cell lines: BT-474, SK-BR-3, MCF-7,MDA-MB-361, MDA-MB-468 and T-47D were obtained from the AmericanType Culture Collection [Rockville, Md. (USA)]. The breast cancercell line 600MPE cell line was kindly provided by Dr. Helene S.Smith [Geraldine Brush Cancer Research Center, San Francisco,Calif. (USA)]. Cell lines were grown until they became confluent.Cells were then trypsinized, pelleted by centrifugation at 1500 RPMfor 5 minutes and washed twice in phosphate buffered saline. TheDNA was then isolated as described by Sambrook et al., MolecularCloning: A Laboratory Manual, Vol. 2: 9.16 9.19 [Cold Spring HarborLaboratory Press, Cold Spring Harbor, N.Y. (USA) 1989].

Details concerning the established human breast cancer cell linesused herein are as follows:

TABLE-US-00003 BT-474 Originated from a human primary cancer;obtained from the ATCC, catalog # HTB 20; SK-BR-3 Originated from ahuman metastatic breast adenocarcinoma derived from a pleuraleffusion; obtained from the ATCC catalog # HTB 30; MDA-MB-361Originated as a metastatic tumor to the brain; obtained from theATCC, catalog # HTB 27; MCF-7 Originated from a human metastaticpleural effusion; obtained from the ATCC, catalog # HTB 22; T-47DOriginated as a human metastatic pleural effusion; obtained fromthe ATCC catalog # HTB 133; 600 MPE Originated as a humanmetastatic pleural effusion; kindly provided by Dr. Helene S. Smith[Geraldine Brush Cancer Research Center, San Francisco, CA (USA)];and MDA-MB-468 Originated as a metastatic pleural effusion;obtained from the ATCC, catalog # HTB 132.

Preparation of Normal Lymphocyte Metaphases:

Normal peripheral blood lymphocytes were stimulated by PHA,synchronized by methotrexate treatment and blocked in metaphaseusing 0.05 ug/ml colcemid. Cells were then centrifuged, washed andincubated in 75 mM KCl at 37.degree. C. for 15 minutes. Cells werethen fixed in methanol:acetic acid (3:1) and dropped onto slides.The slides were stored under nitrogen at -20.degree. C.

DNA Labeling:

Cell line DNAs were labeled with digoxigenin-11-dUTP using nicktranslation [Rigby et al., J. Mol. Biol., 113: 237 (1977); Sambrooket al., supra]. The optimal size of the probe fragments after nicktranslation and before denaturing was 400 800 bps. As indicatedabove, chromosome-specific probes were used in dual-colorhybridizations to verify the identification of chromosomes ofinterest in the metaphase spreads. Representative examples of suchchromosome-specific reference probes labeled with biotin-14-dATPinclude the following:

1) a chromosome-specific painting probe for chromosome 20 preparedby the PCR adapter-linker method as described in PCT/US90/00434published Aug. 9, 1990;

2) a chromosome 17 peri-centromeric repeat probe (cosmid ck17.10)isolated by Anne Kallioniemi from a chromosome 17 cosmid libraryfrom Los Alamos National Laboratory [Albuquerque, N.M. (USA)]; anequivalent chromosome-specific repeat sequence probe for chromosome17 is commercially available from Oncor [Gaithersburg, Md. (USA)];and

3) an alpha satellite repeat probe specific for the centromericregion of chromosome 8 [kindly provided by Dr. Heinz-Ulrich G.Weier; University of California Medical Center, Lab for CellAnalysis, San Francisco, Calif. (USA)]; that probe was generated byDr. Weier using PCR with primers WA1 and WA2 as described in Weieret al., Hum. Genet., 87: 489 494 (1991).

Ones skilled in the art recognize that there are many otherequivalent probes available that could be used for the confirmationpurposes described. For example, whole chromosome painting probesare now available for all the human chromosomes [Collins et al.,Genomics, 11: 997 1006 (1991)]. Also available are repeat sequenceprobes that hybridize intensely and specifically to selectedchromosomes [Trask et al., Hum. Genet., 78: 251 (1988) andreferences cited therein].

Pretreatment and Prehybridization of Slides:

Lymphocyte metaphase preparations were first denatured in 70%formamide/2.times.SSC (1.times.SSC is 0.15 M NaCl, 0.015 MNaCitrate), pH 7, at 70.degree. C. for 2 minutes and dehydrated ina sequence of 70%, 85% and 100% ethanol. The slides were then airdried and treated with 10 ug/50 ml Proteinase K [BoehringerMannheim GmbH, Indianapolis, Ind. (USA)] for 7.5 minutes at37.degree. C. in a buffer containing 20 mM Tris and 2 mM CaCl.sub.2(pH 7.5). Ethanol dehydration was then done as described above, andthe slides were prehybridized with ten ul of a hybridizationmixture, consisting of 20 ug unlabeled human placental DNA[obtained from Sigma, St. Louis, Mo. (USA); size of the fragmentsis 200 700 bps] in 56% formamide, 10% dextran sulphate and2.times.SSC (pH 7) for 60 minutes at 37.degree. C. Before theprehybridization mixture was applied to the slides, it wasdenatured in a 70.degree. C. water bath for 5 minutes. Afterprehybridization, the slides were washed once in 2.times.SSC anddehydrated with ethanol as described above.

Hybridization:

Five ug of unlabeled, repeat-enriched Cot-1 blocking DNA [BRL,Gaithersburg, Md. (USA)] and 60 ng of digoxigenin labeled cell lineDNA and 20 60 ng of biotin-labeled reference probes (forverification of chromosome identification) were mixed together and1/10 vol of 3M Na-acetate was added. DNA was precipitated by adding2 volumes of 100% ethanol followed by centrifugation in amicrocentrifuge for 30 minutes at 15,000 RPM. Ethanol was removedand the tubes were allowed to dry until all visible ethanol hadevaporated. Ten ul of hybridization buffer consisting of 50%formamide, 10% dextran sulphate and 2.times.SSC (pH 7) was thenadded, followed by careful mixing. DNAs in the hybridization bufferwere then denatured for 5 minutes at 70.degree. C. followed by a 60minute renaturation at 37.degree. C. The hybridization mixture wasthen added to the prehybridized lymphocyte metaphase slides.Hybridization was carried out under a coverslip in a moist chamberfor 3 4 days at 37.degree. C.

Immunofluorescent Probe Detection:

The slides were washed three times in 50% formamide/2.times.SSC, pH7, twice in 2.times.SSC and once in 0.1.times.SSC for 10 minuteseach at 45.degree. C. After washing, the slides wereimmunocytochemically stained at room temperature in three steps (3045 minutes each). Before the first immunocytochemical staining, theslides were preblocked in 1% BSA/4.times.SSC for 5 minutes. Thefirst staining step consisted of 2 ug/ml Texas Red-Avidin [VectorLaboratories, Inc., Burlingame, Calif. (USA)] in 1%BSA/4.times.SSC. The slides were then washed in 4.times.SSC,4.times.SSC/0.1% Triton X-100, 4.times.SSC, and PN (a mixture of0.1 M NaH.sub.2PO.sub.4 and 0.1 M Na.sub.2HPO.sub.4, pH 8, and 0.1%Nonidet P-40) for 10 minutes each and preblocked with PNM (5%Carnation dry milk, 0.02% Na-azide in PN buffer) for 5 minutes. Thesecond antibody incubation consisted of 2 ug/ml FITC-conjugatedsheep anti-digoxigenin [Boehringer Mannheim GmBH, Indianapolis,Ind. (USA)] and 5 ug/ml anti-avidin [Vector Laboratories,Burlingame, Calif. (USA)] in PNM followed by three PN washes, 10minutes each. After the PNM block, the third immunochemicalstaining was done using rabbit anti-sheep FITC antibody (1:50dilution) (Vector Laboratories) and 2 ug/ml Texas Red-Avidin inPNM. After three PN washes, nuclei were counterstained with 0.8 uM4,5-diamino-2-phenylindole (DAPI) in an antifade solution.

Fluorescence Microscopy and Interpretation of Results:

A Nikon fluorescence microscope [Nikon Inc., Garden City, N.Y.(USA)] equipped with a double band pass filter [Chroma Technology,Brattleboro, Vt. (USA)] and a 100.times. objective was used forsimultaneous visualization of the FITC and Texas Red signals.Hybridization of the breast cancer cell line DNAs was seen as amore or less uniform faint green background staining of allmetaphase chromosomes with the exception of the Y-chromosome. Asthe breast cancer cell lines are of course of female origin, theydid not contain Y chromosomal DNA. The absence of said greenstaining of the Y chromosome of the metaphase spread (seen in FIG.8) is exemplary of the manner in which a cytogeneticallysignificant deletion would be visualized. Using a fluorescencemicroscope, amplified sequences can be seen as bright green dots orbands along the chromosome arms.

To facilitate the display of the results and to improve thesensitivity of detecting small differences in fluorescenceintensity, a digital image analysis system (QUIPS) was used. QUIPS(an acronym for quantitative image processing system) is anautomated image analysis system based on a standard Nikon MicrophotSA [Nikon Inc., Garden City, N.Y. (USA)] fluorescence microscopeequipped with an automated stage, focus control and filterwheel[Ludl Electronic Products Ltd., Hawthorne, N.Y. (USA)]. Thefilterwheel is mounted in the fluorescence excitation path of themicroscope for selection of the excitation wavelength. Specialfilters [Chroma Technology, Brattleboro, Vt. (USA)] in the dichroicblock allow excitation of multiple dyes without image registrationshift. The microscope has two camera ports, one of which has anintensified CCD camera [Quantex Corp., Sunnyvale, Calif. (USA)] forsensitive high-speed video image display which is used for findinginteresting areas on a slide as well as for focusing. The othercamera port has a cooled CCD camera [model 200 by PhotometricsLtd., Tucson, Ariz. (USA)] which is used for the actual imageacquisition at high resolution and sensitivity.

The cooled CCD camera is interfaced to a SUN 4/330 workstation [SUNMicrosystems Inc., Mountain View, Calif. (USA)] through a VME bus.The entire acquisition of multicolor images is controlled using animage processing software package SCIL-Image [Delft Centre forImage Processing, Delft, Netherlands]. Other options forcontrolling the cameras, stage, focus and filterwheel as well asspecial programs for the acquisition and display of multicolorimages were developed at the Division of Molecular Cytometry[University of California, Medical Center; San Francisco, Calif.(USA)] based on the SCIL-Image package.

To display the results of the comparative hybridization, two orthree consecutive images were acquired (DAPI, FITC and Texas Red)and superimposed. The FITC image was displayed after using thethresholding and contrast enhancement options of the SCIL-Imagesoftware. Exercising such options reduces the overall chromosomalfluorescence to make amplified sequences more readily visible. Forexample, using thresholding and contrast stretching, it waspossible to enhance the contrast and quantification between thefaint green background staining and staining originating from theamplified sequences in the cell lines. Alternatively, to facilitatethe detection of deletions, it is possible to increase the overallchromosomal fluorescence and make areas of reduced fluorescenceappear darker. The red color was used for reference probes to helpin the identification of chromosomes.

After identification of the chromosomes based on the use ofreference probes in a dual-color hybridization, a site ofamplification was localized by fractional length measurements alongthe chromosome arm (fractional length=distance of the hybridizationsignal from the p-telomere divided by the total length of thechromosome). The band location of the signal was then approximatedfrom the fractional length estimate based on the ISCN 1985idiograms [Harnden and Klinger, An International System forCytogenetic Nomenclature, Karger Ag, Basel, Switzerland(1985)].

Results:

The results from the hybridizations are compiled in Table 2 alongwith other information known about the cell lines. Amplification at17q12 (erbB-2 locus) and approximately 8q24 (MYC locus) was seen inlines showing amplification of erbB-2 and MYC whenever the level ofamplification was greater than about five- to ten-fold using thisCRCC method. In addition, amplification of several megabase wideregions was seen in three cell lines at 17q22 23 and in three linesat 20qter; those amplifications were previously unknown sites ofamplification and were not expected from other studies. All linesshowing amplification showed amplification at more than one site.Evidence for co-amplification may be clinically important sinceco-amplification has been observed previously [van de Vijver etal., Mol. Cell Biol. 7: 2019 2023 (1987); Saint-Ruf et al.,Oncogene, 6: 403 406 (1991)], and is sometimes associated with poorprognosis [Borg et al., Br. J. Cancer, 63: 136 142 (1991)].Amplification at 17q22 23 has also been seen using probe DNA fromprimary tumors.

TABLE-US-00004 TABLE 2 Results of Testing Breast Cancer Cell Linesfor Amplification Known Ampli- Growth Hormone ampli- fication Cellrate; receptor fication detected Line Origin -Td E/P (level) by CGHBT-474 Primary 48 96 hr +/- erbB-2 17q12 Cancer (13X) (erbB-2),17q22 23, 20qter SK-BR-3 Pl. Effusion ? ? erbB-2 17q12 (9X)(erbB-2), 8q21, MYC 8q23 24.1 (10X) (MYC), 20qter MDA-MB- Brainmet. <96 hr -/+ erbB-2 17q22 23 361 (4X) MCF-7 Pl. Effusion<48 hr +/+ erbB-2 17q22 23, (none) 20qter T-47D Pl. Effusion ?+/+ erbB-2 None (none) 600MPE Pl. Effusion ? ? erbB-2 None (none)MDA-MB- Pl. Effusion ? ? erbB-2 None 468 (none)

EXAMPLE 2

Hybridizations with two different labeled subject DNAs asschematically outlined in FIGS. 6 and 7 were performed. One of thelabeled subject DNAs hybridized was a cell line DNA as described inExample 1 and similarly labeled. The other labeled subject DNA washuman genomic DNA labeled with biotin-14-dATP.

The protocols were essentially the same as in Example 1 except thatno chromosome-specific reference probes were used, and the sameamount of the labeled human DNA as the labeled cell line DNA, thatis, 60 ng, was hybridized. Of course, reference probes could beadded to the hybridization mixture, but they need to be differentlylabeled to be distinguishable.

The results showed the normal DNA with a red signal and the cellline DNA with a green signal. The green to red ratios weredetermined along each chromosome. Amplification was indicated by anarea where the signal was predominantly green whereas deletionswere indicated by more red signals than in other areas of thechromosomes.

Exemplary, CGH results using breast cancer cell line 600MPE DNA andnormal human DNA were as follows. As indicated above, thehybridization was performed using 5 ug Cot-1 DNA, 60 ng ofdigoxigenin labeled 600MPE cell line DNA, and 60 ng of biotinylatednormal human genomic DNA. The 600MPE DNA was detected with FITC(green) and the genomic DNA with Texas Red-Avidin (red).

The 600MPE breast cancer cell line, the karyotype for which waspublished by Smith et al., JNCI, 78: 611 615 (1987), contains onenormal chromosome 1 and three marker chromosomes with chromosome 1material in them: t(1q:13q), 1p(p22) and inv(1)(p36q21). Thus, thecell line is disomic for the p-telomere-p22, trisomic forp22-centromere and tetrasomic for the q-arm of chromosome 1. Anidiogram of chromosome 1 showing those different areas isillustrated in FIG. 9.

The comparative genomic hybridizations of this example apparentlyidentified three different regions on chromosome 1 that could beseparated according to the intensities of green and red colors. Theq-arm of chromosome 1 had the highest intensity of green color(tumor DNA). The region from band p22 to the centromere was thesecond brightest in green, and the area from the p-telomere to bandp22 had the highest intensity of red color (normal DNA). Thosehybridization results were consistent with the traditionalcytogenetic analyses of that cell line stated immediatelyabove.

However, further studies with CGH, as presented in Example 3,indicated that CGH analysis of Example 2, as well as the publishedkaryotype, were partially in error. The CGH analysis of Example 3motivated additional confirmatory experiments, as describedtherein, leading to correction of the original CGH results and thepublished karyotype.

EXAMPLE 3

Copy Number Karyotypes of Tumor DNA

In the representative experiments of CGH in this example,biotinylated total tumor DNA (cell line and primary tumor DNA) anddigoxigenin-labeled normal human genomic DNA are simultaneouslyhybridized to normal human metaphase spreads in the presence ofunlabeled blocking DNA containing high-copy repetitive sequences,specifically unlabeled Cot-1 blocking DNA [BRL, Gaithersburg, Md.(USA)]. The following paragraphs-detail the procedures used for therepresentative CGH experiments of this example.

DNA Labeling:

DNAs used in this example were labeled essentially as shown abovein Example 1. DNAs were labeled with biotin-14-dATP ordigoxigenin-11-dUTP by nick translation [Rigby et al., supra;Sambrook et al., supra]. The optimal size for double stranded probefragments after labeling was 600 1000 bp.

Pretreatment of Metaphase Spreads:

Lymphocyte metaphase preparations were denatured, dehydrated andair dried, treated with Proteinase K and dehydrated again asdescribed in Example 1.

Comparative Genomic Hybridization:

Sixty ng of biotinylated test DNA, 60 ng of digoxigenin-labelednormal DNA and 5 .mu.g of unlabeled Cot-1 DNA (BRL) were ethanolprecipitated and dissolved in 10 .mu.l of 50% formamide, 10%dextran sulfate, 2.times.SSC, pH 7. The probe mixture was denaturedat 70.degree. C. for 5 minutes, allowed to reanneal at 37.degree.C. for 60 minutes and hybridized to normal male metaphasechromosomes for 3 4 days at 37.degree. C.

Immunofluorescent Probe Detection:

The slides were washed as described above in Example 1, andimmunocytochemically stained at room temperature in threethirty-minute steps: (I) 5 .mu.g/ml FITC-Avidin [VectorLaboratories, Inc., Burlingame, Calif. (USA)] and 2 .mu.g/mlanti-digoxigenin-Rhodamine (Boehringer Mannheim GMbH); (II) 5.mu.g/ml anti-avidin (Vector Laboratories); and (III) 5 .mu.g/mlFITC-avidin. Nuclei were counterstained with 0.8 .mu.M4,5-diamino-2-phenylindole (DAPI) in antifade solution. A Zeissfluorescence microscope equipped with a double band pass filter[Chroma Technology, Brattleboro, Vt. (USA)] was used forsimultaneous visualization of FITC and rhodamine signals.

Digital Image Analysis System and Fluorescence Ratio Profiles

The QUIPS system essentially a described above in Example 1 wasused to analyse quantitatively the fluorescence signals.Fluorescence ratio profiles along the chromosomes were extractedusing WOOLZ software package [developed at MRC, Edinburgh,Scotland] as follows: the DAPI image is used to set themorphological boundary of each chromosome by thresholding. Thechromosome outline is smoothed by a n number of opening and closingoperations, a modified Hilditch skeleton is calculated and taken torepresent the medial axis of the chromosome. The DAPI image isexpanded outwards in all directions until the intensity fieldlevels off (when background is reached) or begins to rise (due toan adjacent chromosome). The intensity profile of each image alongthe medial axis and within the expanded DAPI image is thencalculated by summing the green and red fluorescence pixel valuesalong the sequence of lines perpendicular to and spaced at unitdistance along the medial axis. Modal green and red intensityvalues corresponding to the expanded DAPI image are taken torepresent the background fluorescence and used as the intensityorigin.

Cell Lines:

TABLE-US-00005 5637 Originated from a human primary bladdercarcinoma; obtained from ATCC, catalog # HTB 9 SK-BR-3 Originatedfrom a human metastatic breast adenocarcinoma, derived from apleural effusion; obtained from the ATCC, catalog # HTB 30 Colo 205Originated from a human colon adenocarcinoma; obtained from theATCC, catalog # CCL 222 NCI-H508 Originated from a human cecumadenocarcinoma; obtained from the ATCC, catalog # CCL 253 SW480Originated from a human colon adenocarcinoma; obtained from theATCC, catalog # CCL 228 SW620 Originated from a human lymph nodemetatasis of a colon adenocarcinoma; obtained from the ATCC,catalog # CCL 227 WiDr Originated from a human colonadenocarcinoma; obtained from the ATCC, catalog # CCL 218 SK-N-MCOriginated from a human neuroblastoma (metastasis to supra-orbitalarea); obtained from the ATCC, catalog # HTB 10 CaLu3 Originatedfrom a human lung adenocarcinoma, derived from a pleural effusion;obtained from the ATCC, catalog # HTB 55 CaLu6 Originated from ahuman anaplastic carcinoma, probably lung; obtained from the ATCC,catalog # HTB 56 NCI-H69 Originated from a human small cell lungcarcinoma; obtained from the ATCC, catalog # HTB 119 COLO 320HSROriginated from a human colon adenocarcinoma; obtained from theATCC, catalog # 220.1 600 PE Originated from a human breastcarcinoma; obtained from Dr. Helene Smith and Dr. Ling Chen[Geraldine Brush Cancer Research Center, San Francisco, CA (USA)].This is the same as the 600 MPE cell line described in Examples 1and 2. BT-20 Originated from a human breast carcinoma; obtainedfrom ATCC, catalog # HTB 19

The following are five fibroblast cell lines with total chromosomalnumber and X chromosomal number in parentheses, which were obtainedfrom the NIGMS repository [Camden, N.J. (USA)]:

TABLE-US-00006 GMO1723 (45, XO) GMO8399 (46, XX) GMO4626 (47, XXX)GMO1415E (48, XXXX) GMO5009B (49, XXXXX).

Results and Discussion:

Demonstrated herein is CGH's capability of detecting and mappingrelative. DNA sequence copy number between genomes. A comparison ofDNAs from malignant and normal cells permits the generation of a"copy number karyotype" for a tumor, thereby identifying regions ofgain or loss of DNA.

Demonstrated is the use of dual color fluorescence in situhybridization of differently labeled DNAs from a subject tumorgenome and a normal human genome to a normal human metaphase spreadto map DNA sequence copy number throughout the tumor genome beingtested. Regions of gain or loss of DNA sequences, such asdeletions, duplications or amplifications, are seen as changes inthe ratio of the intensities of the two fluorochromes (used in thisrepresentative example) along the target chromosomes. Analysis oftumor cell lines and primary bladder tumors identified 16 differentregions of amplification, many in loci not previously known to beamplified. Those results are shown in Table 3 below.

The tumor DNA is detected with the green fluorescing FITC-avidin,and the normal DNA with the red fluorescing rhodamineanti-digoxigenin. The relative amounts of tumor and normal DNAbound at a given chromosomal locus are dependent on the relativeabundance of those sequences in the two DNA samples, and can bequantitated by measurement of the ratio of green to redfluorescence. The normal DNA in this example serves as a controlfor local variations in the ability to hybridize to targetchromosomes. Thus, gene amplification or chromosomal duplication inthe tumor DNA produces an elevated green-to-red ratio, anddeletions or chromosomal loss cause a reduced ratio. The Cot-1 DNAincluded in the hybridization inhibits binding of the labeled DNAsto the centromeric and heterochromatic regions so those regions areexcluded from the analysis.

The fluorescence signals were quantitatively analyzed by means of adigital image analysis system as described above. A softwareprogram integrated the green and red fluorescence intensities instrips orthogonal to the chromosomal axis, subtracted localbackground, and calculated intensity profiles for both colors andthe green-to-red ratio along the chromosomes.

The ability of CGH to quantitate changes in sequence copy numberthat affect an entire chromosome was tested with the above-listedfive fibroblast cell lines having 1 to 5 copies of the X chromosomeand two copies of each autosome. Hybridization of DNA from the45,X0 cell line (in green) together with normal female DNA (in red)resulted in a uniform green-red staining of the autosomes whereasthe X chromosome appeared more red (FIG. 10A). Hybridizations withDNA from cell lines carrying 2, 3, 4 or 5 copies of the Xchromosome resulted in an increasingly strong green fluorescencefrom the X chromosome in relation to the autosomes. The averagegreen-to-red fluorescence ratio of the X chromosome (FIG. 10B),when normalized to the average ratio for the autosomes within thesame metaphase spread, increased linearly with the increasingnumber of X chromosomes [correlation coefficient (r)=0.978]. Thus,CGH can quantitatively distinguish a change of plus or minus onecopy of a chromosome at least up to 4 copies.

Experiments showed that CGH could generate a complete copy numberkaryotype for a near-diploid breast cancer cell line, 600PE.According to the published karyotype for 600PE [Smith et al., JNCI,78: 611 (1987)], 600PE is near-diploid with five marker chromosomeshaving four copies of the q-arm of chromosome 1, monosomy 16, anddeletions of 9p, 11q and 17p. CGH using biotinylated 600PE DNA (ingreen) and normal digoxigenin-labeled DNA (in red) revealed thefollowing relative copy number changes: gain of 1q and loss of 9p,16q, 17p and distal 11q. The green-to-red ratio profiles for thoseaberrant chromosomes are shown in FIG. 11. Only the q-arm ofchromosome 16 showed decreased relative copy number suggesting that16p was not deleted. That observation was subsequently confirmed byfluorescence in situ hybridization (FISH) to 600PE interphase cellsusing cosmid probes for the p- and q-arms of chromosome 16 [16p and16q cosmid probes provided by Los Alamos National Laboratory, LosAlamos, N.Mex. (USA)]; two signals per nucleus for the 16p cosmidprobe and one for the 16q cosmid probe permitted calibration of agreen-to-red ratio of 1.0 as indicating two copies of asequence.

Thus, if the absolute copy number of any point in the tumor genomeis known, relative copy numbers can be converted to actual copynumbers at all loci. The CGH results differed from the originallypublished karyotype in the region of 16p and proximal 1p. Thatdiscrepancy was resolved by locus-specific chromosome-specificpainting (FISH) that indicated that the components of one of themarker chromosomes had been misinterpreted by conventionalcytogenetic analysis.

CGH with DNAs from two fibroblast cell lines [GMO5877 and GMO1142Afrom the NIGMS repository] detected small interstitial deletionsaround the RB1 locus in 13q--del(13)(pter>q14.1::q21.2.>qter) and del(13)(pter>q14.1::q22.1>qter). On the basis of the CGH analysisand measurement of the deletion size as a fraction of the length ofchromosome 13 (total length 111 Mb), those deletions were estimatedto span about 10 and 20 megabases (Mb), respectively. Thus it ispossible that CGH can be used to screen DNA samples from solidtumors in order to identify large physical deletions that mayuncover recessive mutant tumor suppressor genes.

CGH was evaluated for its ability to detect increased gene copynumber with cell lines that contained previously reportedamplification of oncogenes. FIG. 12A shows CGH with DNA from acolon cancer cell line COLO 320HSR, known to contain more than a50-fold amplification of a 300 kb region around the myc oncogene[Kinzku et al., PNAS (USA), 83: 1031 (1986)]. The expected highgreen-to-red ratio at 8q24 corresponding to the location of myc isclear. The height of the peak does not quantitatively reflect thelevel of amplification because the fluorescent signal spread over aregion of the chromosome that is larger than the length of theamplicon. That is apparently a result of the complex organizationof the target DNA in the denatured chromosomes.

The eight-fold amplification of the erbB2 oncogene in the SK-BR-3breast cancer cell line also was detectable with CGH as ahybridization signal at 17q12 (Table 3). High level amplificationssuch as those also could be detected in single color-hybridizationswith the use of only labeled tumor DNA.

Cytogenetic and molecular studies of primary tumors and cell linesoften reveal homogeneously staining regions and double minutechromosomes that do not involve known oncogenes [Saint-Ruf et al.,Genes Chrom. Cancer., 2: 18 (1990); Bruderlein et al., Genes Chrom.Cancer, 2: 63 (1990)]. CGH allows straightforward detection andmapping of such sequences. Table 3 contains a summary of theanalysis with CGH of 11 cancer cell lines. Data in Table 3 is basedon the visual inspection of a large number of metaphase spreads andon detailed digital image analysis of four to six metaphases foreach sample.

TABLE-US-00007 TABLE 3 Mapping of amplified sequences inestablished cancer cell lines and primary tumors by CGH Cytogeneticevidence of Specimen Origin Amplif. by CGH* gene amplif..sup.+ Celllines: 5637 Bladder 3p25, 6p22 DM SK-BR-3 Breast 8q24 (myc), 8q21,17q12 (erbB2), 20q13 Colo 205 Colorectal 6p21, 6q24 NCI-H508Colorectal 14q12 13 DM SW480 Colorectal 8q24 (myc) DM SW620Colorectal 16q21 23 HSR WiDr Colorectal 8q23 24 (myc) SK-N-MCNeuroblastoma 8q24 (myc) DM CaLu3 Small cell lung 8p12 21, 8qtel,HSR 17q12 (erbB2) CaLu6 Small cell lung 13q32 34 NCI-H69 Small celllung 2p24 (N-myc), 2p21, 2q21 Primary tumors: UR140 Bladdercarcinoma 16q21 22 UR145 Bladder carcinoma 6p22 *The oncogene mostlikely involved in this amplification is shown in parentheses..sup.+Cytogenetic information based on the ATCC Catalogue of CellLines & Hybridomas (1992). DM = double minute chromosomes, HSR= homogeneously staining regions.

Sixteen amplified loci were mapped, many at regions of the genomewhere amplification had not previously been suspected. Thus, alarge variety of genes may be amplified during cancer initiationand progression. In five of the 11 cell lines, more than one locuswas amplified. Two or three separate loci on the same chromosomewere amplified in four cell lines, which suggests a spatialclustering of chromosomal locations that undergo DNA amplification(Table 3 and FIG. 12A).

CGH was also applied to identify and map amplified DNA sequences inuncultured primary bladder tumors. Of the seven tumors tested, twoshowed evidence of DNA amplification but the loci were not the same(Table 3). Thus, a number of previously unsuspected genomic regionsthat might contain genes important for cancer progression have beenidentified by CGH. Further studies will elucidate which of thoseloci contain novel oncogenes and which represent coincidental,random DNA amplification characteristic of genomic instability.

The detection and mapping of unknown amplified sequences thattypically span several hundred kilobases (kb) to a few Mbdemonstrated the usefulness of CGH for rapid identification ofregions of the genome that may contain oncogenes. Analogously,detection of deletions may facilitate identification of regionsthat contain tumor suppressor genes.

Further studies are necessary to establish to what extent alleliclosses in tumors are caused by physical deletions. In clinicalspecimens, the detection of small copy number differences is moredifficult than with cell lines because of the admixture of DNA fromcontaminating normal cells and because of intratumor heterogeneity.As indicated above, using PCR to prepare tumor DNA from a smallnumber of tumor cells (as a tumor clonal sub-population) may assistin resolving that problem. Like RFLP, CGH emphasizes the detectionof aberrations that are homogeneous in a cell population andaverages those that are heterogeneous.

At the current stage of development of CGH, sensitivity isprimarily limited by the granularity of the hybridization signalsin the metaphase chromosomes. Further improvements in sensitivitywill be achieved by optimization of the probe concentration andlabeling, and by the averaging of the green-to-red fluorescenceratios from several metaphase spreads.

The descriptions of the foregoing embodiments of the invention havebeen presented for purposes of illustration and description. Theyare not intended to be exhaustive or to limit the invention to theprecise form disclosed, and obviously many modifications andvariations are possible in light of the above teachings. Theembodiments were chosen and described in order to best explain theprinciples of the invention and its practical application to enablethereby others skilled in the art to best utilize the invention invarious embodiments and with various modifications as are suited tothe particular use contemplated. It is intended that the scope ofthe invention be defined by the claims appended hereto. Allreferences cited herein are incorporated by reference.

* * * * *

References

• ndsu.nodak.edu/instruct/mcclean/plsc431/eukarychrom/eukaryo3.htm