Pinkel D, Segraves R, Sudar D, Clark S, Poole I, Kowbel D, Collins C, Kuo WL, Chen C, Zhai Y, Dairkee SH, Ljung BM, Gray JW, and Albertson DG. High resolution analysis of DNA copy number variation using comparative genomic hybridization to microarrays. Nat Genet 20: 207-211

Abstract

Gene dosage variations occur in many diseases. In cancer, deletions and copy number increases contribute to alterations in the expression of tumour-suppressor genes and oncogenes, respectively. Developmental abnormalities, such as Down, Prader Willi, Angelman and Cri du Chat syndromes, result from gain or loss of one copy of a chromosome or chromosomal region. Thus, detection and mapping of copy number abnormalities provide an approach for associating aberrations with disease phenotype and for localizing critical genes. Comparative genomic hybridization (CGH) was developed for genome-wide analysis of DNA sequence copy number in a single experiment. In CGH, differentially labelled total genomic DNA from a 'test' and a 'reference' cell population are cohybridized to normal metaphase chromosomes, using blocking DNA to suppress signals from repetitive sequences. The resulting ratio of the fluorescence intensities at a location on the 'cytogenetic map', provided by the chromosomes, is approximately proportional to the ratio of the copy numbers of the corresponding DNA sequences in the test and reference genomes. CGH has been broadly applied to human and mouse malignancies. The use of metaphase chromosomes, however, limits detection of events involving small regions (of less than 20 Mb) of the genome, resolution of closely spaced aberrations and linking ratio changes to genomic/genetic markers. Therefore, more laborious locus-by-locus techniques have been required for higher resolution studies. Hybridization to an array of mapped sequences instead of metaphase chromosomes could overcome the limitations of conventional CGH (ref. 6) if adequate performance could be achieved. Copy number would be related to the test/reference fluorescence ratio on the array targets, and genomic resolution could be determined by the map distance between the targets, or by the length of the cloned DNA segments. We describe here our implementation of array CGH. We demonstrate its ability to measure copy number with high precision in the human genome, and to analyse clinical specimens by obtaining new information on chromosome 20 aberrations in breast cancer.

Full text

letter
nature genetics volume 20 october 1998 207
High resolution analysis of DNA copy number variation
using comparative genomic hybridization to microarrays
Daniel Pinkel
1,2
, Richard Segraves
1
, Damir Sudar
2
, Steven Clark
1
, Ian Poole
3
, David Kowbel
2
, Colin Collins
2
,
Wen-Lin Kuo
1
, Chira Chen
1
, Ye Zhai
1
, Shanaz H. Dairkee
4
, Britt-marie Ljung
5
, Joe W. Gray
1,2
& Donna G. Albertson
1,2,6
1
Cancer Genetics Program, UCSF Cancer Center, University of California San Francisco, Box 0808, San Francisco, California 94143-0808, USA.
2
Life
Sciences Division, E.O. Lawrence Berkeley National laboratory, 1 Cyclotron Road, Berkeley, California 94720, USA.
3
Vysis, Inc., Downers Grove, Illinois,
USA.
4
Geraldine Brush Cancer Research Institute, California Pacific Medical Center, 2330 Clay Street, San Francisco, California 94115, USA.
5
Pathology
Department, University of California San Francisco, Box 0102, San Francisco, California 94143, USA.
6
MRC Laboratory of Molecular Biology, Hills Road,
Cambridge CB2 2QH, UK. Correspondence should be addressed to D.P. (e-mail: pinkel@cc.ucsf.edu).
Gene dosage variations occur in many diseases. In cancer, dele-
tions and copy number increases contribute to alterations in the
expression of tumour-suppressor genes and oncogenes, respec-
tively. Developmental abnormalities, such as Down, Prader Willi,
Angelman and Cri du Chat syndromes, result from gain or loss of
one copy of a chromosome or chromosomal region. Thus, detec-
tion and mapping of copy number abnormalities provide an
approach for associating aberrations with disease phenotype
and for localizing critical genes. Comparative genomic hybrid-
ization
1
(CGH) was developed for genome-wide analysis of DNA
sequence copy number in a single experiment. In CGH, differen-
tially labelled total genomic DNA from a ‘test’ and a ‘reference’
cell population are cohybridized to normal metaphase chromo-
somes, using blocking DNA to suppress signals from repetitive
sequences. The resulting ratio of the fluorescence intensities at
a location on the ‘cytogenetic map’, provided by the chromo-
somes, is approximately proportional to the ratio of the copy
numbers of the corresponding DNA sequences in the test and
reference genomes. CGH has been broadly applied to human
and mouse malignancies. The use of metaphase chromosomes,
however, limits detection of events involving small regions (of
less than 20 Mb) of the genome, resolution of closely spaced
aberrations and linking ratio changes to genomic/genetic mark-
ers. Therefore, more laborious locus-by-locus techniques have
been required for higher resolution studies
2–5
. Hybridization to
an array of mapped sequences instead of metaphase chro-
mosomes could overcome the limitations of conventional CGH
(ref. 6) if adequate performance could be achieved. Copy num-
ber would be related to the test/reference fluorescence ratio on
the array targets, and genomic resolution could be determined
by the map distance between the targets, or by the length of
the cloned DNA segments. We describe here our implementa-
tion of array CGH. We demonstrate its ability to measure copy
number with high precision in the human genome, and to
analyse clinical specimens by obtaining new information on
chromosome 20 aberrations in breast cancer.
The performance goals of array CGH are more stringent than
those of related array-based methods for measuring gene expres-
sion
7,8
. Optimum utility requires data over a wide copy number
range, including reliable detection of single copy changes relative
to the normal diploid state. This precision must be achieved for
hybridizations involving the entire mammalian genome, a nucleic
acid pool that has over ten times the complexity of the expressed
sequences, and which includes a significant quantity of inter-
spersed repeats. Our procedures, which differ considerably from
others
6–8
, meet these challenges when using DNA from cosmid,
P1, BAC and other large insert clones for array targets.
The sensitivity and quantitative capability of our methodology
was evaluated using test and reference genomes made by adding
various amounts of λ DNA (50 kb, 1.5×10
–5
of the human
genome) to the normal human genome. This model simulates
the behaviour of a cosmid-sized target that does not contain
repetitive sequences. The measured fluorescence ratios on the λ
targets were quantitatively proportional to the relative amounts
of λ DNA in the test and reference genomes from below single-
copy equivalent level to at least two orders of magnitude higher
(Fig. 1). Our techniques, therefore, have the sensitivity necessary
to measure copy number increases or decreases in the human
genome. The limited cross hybridization of human sequences to
a λ DNA target is shown (Fig. 3b).
Measurement precision and accuracy of array CGH was
assessed using arrays containing targets made from a set of clones
on chromosome 20 and four clones on the X chromosome. The
chromosome 20 target clones (Table 1; Fig. 2a, inset) were spaced
at an average interval of approximately 3 Mb along the entire
chromosome, and included clones specifically chosen to detect
regions of copy number change that had previously been found
Fig. 1 Comparison of the fluorescence intensity ratio on λ DNA targets to the
ratio of λ DNA in model test and reference genomes. The six model test
genomes, labelled with fluorescein, contained 400 ng of human genomic DNA
and 0, 3, 6, 60, 600 and 6000 pg of λ DNA, respectively. These corresponded to
0, 0.5, 1, 10, 100 and 1000 times the equivalent single-copy level for a 50-kb
sequence. The reference genomes for all six hybridizations, labelled with Texas
red, contained 400 ng of human genomic DNA and 6 pg of λ DNA. Fluores-
cence ratios on λ targets were normalized relative to human genomic DNA tar-
gets contained in the same arrays. (Cloned targets could not be used for
normalization because of cross hybridization between λ DNA and vector
sequences.) Ratios for the five test genomes with non-zero λ DNA are shown in
the main figure, normalized relative to the 6-pg (single copy equivalent) ratio.
Error bars indicate the standard deviations of three independent hybridiza-
tions. At some copy number levels the standard deviations are smaller than the
symbols used for the data points. The measured fluorescence ratios were quan-
titatively proportional to copy number over a dynamic range of at least 200,
but the ratio was about 30% low at the highest point. The inset shows a linear
plot of the data from the 0, 3 and 6 pg hybridizations, 0, 0.5 and 1 times the
single copy level, respectively.
Normalized flourescence ratio
λ DNA equivalent copy number
© 1998 Nature America Inc. • http://genetics.nature.com
© 1998 Nature America Inc. • http://genetics.nature.com

letter
208 nature genetics volume 20 october 1998
on chromosome 20q (ref. 3). Four clones were from a contig
9
at
20q13.2 (Table 1, region A). Six independent comparative
hybridizations of a normal human genome with itself produced
very similar fluorescence ratios on all chromosome 20 targets
(Fig. 2a), providing a precise baseline from which to measure
deviations. The measured ratios, normalized to 1.0 for each
hybridization, were 1.0±0.07 (mean±s.d. for all 6×22 targets
equals 132 measurements). The ratios were constant, although
the test and reference fluorescence intensities varied several-fold
among the targets, approximately proportional to the length of
the human DNA segment they contained.
Comparisons of cell populations containing 1−5 copies of the
X chromosome with normal female DNA showed that fluores-
cence ratios were proportional to copy number (Fig. 2b). Mea-
surement precision was high enough to resolve each of the four
copy number steps. Additional comparisons of normal male and
normal female DNA yielded normalized ratios on the X chromo-
some targets of 0.69±0.05 (mean±s.d.). Thus, single copy
decreases produce ratios that differ from diploid with high statis-
tical significance. Although the measured ratios are proportional
to copy number, the slope of the relation, 0.37, differs from the
ideal value of 0.5. The reason for this is not known, but reduced
slopes are expected if the measured fluorescence intensities have
contributions from sources other than unique sequence
hybridization
10
. Possible sources include incompletely sup-
pressed repeats (perhaps those not in the Cot-1 fraction), cross-
hybridization of closely related sequences and target
autofluorescence. Although some repeats may not be completely
blocked, suppression of common centromeric sequences is very
effective (Fig. 3b).
Application of array CGH to breast cancer provided new, high
resolution, high dynamic range information on copy number
Table 1 • Array targets and summary of regional aberrations
Source
a
Map Copy number aberration
d
Chromosome Target (library, plate, position) Locus/gene location
b
Region
c
BT474 S6 S21 S50 S59
Chromosome 20 2005(15) ref. 5 20p sub-telomere 0.034
p arm RMC20P107 DuPont B 114 H10 CDC25B 0.085
RMC20P160 DuPont A 967 A8 WI-7829 0.158
RMC20P178 DuPont A 899 F11 D20S186 0.209
RMC20P055 DuPont B 121 B6 D20S114 0.272
RMC20P099 DuPont B 39 A2 CST3 0.352
q arm RMC20P090 DuPont B 81 C11 BCLX 0.526
RMC20P117 DuPont B 72 G11 AIB4 0.548 C ●●
RMC20P037 DuPont B 76 F8 SRC 0.603 F ❍
RMC20P154 DuPont A 897 C10 D20S44 0.646 F ❍
RMC20P058 DuPont B 77 D11 TOP1 0.675 F ❍
RMC20P100 DuPont B 53 B2 SEMG1 0.694 D ●
RMC20P131 DuPont A 213 F1 D20S178 0.722 D ●●
RMC20P063 DuPont B 106 D4 PTPN1/PTP1B 0.755 B ●●●
RMC20B4135 Research Genetics 22 N5 WI-4444 0.805 A ●● ●
RMC20P4039 DuPont B 34 A6 RMC20C001 0.806 A ●● ●●
RMC20B4097 Research Genetics 133 E8 D20S211 0.807 A ●●●●●
RMC20B4130 Research Genetics 341 H15 D20S211 0.808 A ●●●●●
RMC20P070 DuPont B 58 E10 D20S120 0.815 ●● ●
RMC20P071 DuPont B 97 A2 D20S100 0.827 ●
RMC20P073 DuPont B 101 A6 PCK1 0.867 E ●● ●
RMC20P179 DuPont A 1173 H1 CHRNA4 0.948
a
The P1 or BAC clones were selected by PCR from a pooled (DuPont B) or unpooled (DuPont A) human genomic P1 library
29
or human genomic BAC library
30
(Research Genetics). The selection and mapping of many of the P1 clones have been reported
31
.
b
Location of the clones on chromosome 20 are given in fractional
length, Flpter previously. Order of the chromosome 20 clones is consistent with genetic maps.
c
Clones that define the regions of consistent copy number aberra-
tion are indicated. Those labelled A−E correspond to defined regions
3
; region F is also defined.
d
The filled circles indicate that elevated copy number was mea-
sured on that clone, the open circles indicate a loss. All other clones were at average copy number relative to the p arm of chromosome 20.
Fig. 2 Precision and accuracy of human copy number measurements. a, Fluorescence ratios on the chromosome 20 array targets for six comparisons of a normal
genome to itself. The inset schematically shows the location of the chromosome 20 array targets. The vertical set of targets on the q arm indicates the location of
four nearly contiguous clones at 20q13.2. The ratios were normalized so that the average for all targets in each hybridization was 1.0. The data points show the
mean of the six normalized ratios obtained for each target, the error bars indicate the standard deviations. They are plotted at the physical location of the clone as
determined by FISH, measured as a fraction of the chromosome length relative to the p terminus (Flpter). The solid horizontal line is drawn at ratio 1.0. b, Normal-
ized fluorescence ratios on X chromosome targets as a function of X chromosome copy number. Arrays containing four clones from the X chromosome, and four
clones on chromosome 20p (RMC20P107, RMC20P160, RMC20P178 and RMC20P099; Table 1), were hybridized with test genomic DNA from a normal male, normal
female and three cell lines containing three, four and five copies of the X chromosome. Normal female reference DNA was used for all hybridizations. The fluores-
cence ratios on each of the X chromosome targets was normalized by the mean ratio of the chromosome 20 targets in that hybridization. All measurements are
plotted, in some cases overlapping points obscure each other. The line is a linear regression through all of the data, slope 0.37, intercept 0.27.
a
b
Normalized ratio
Normalized ratio
FLpter
Number of X chromosomes
© 1998 Nature America Inc. • http://genetics.nature.com
© 1998 Nature America Inc. • http://genetics.nature.com

letter
nature genetics volume 20 october 1998 209
aberrations on chromosome 20. Several distinct peaks and a copy
number decrease were seen after analysis of breast cancer cell line
BT474 (Fig. 3a). Labels A−E indicate locations where copy num-
ber increases in breast cancer cell lines and tumours had previ-
ously been found
3
. The deletion F, which was confirmed by
interphase FISH using clones RMC20P107 (on 20p) and
RMC20P154 (in region F), represents a new feature not previ-
ously recognized as significant in breast cancer. This region is
known to be recurrently deleted in myeloid leukaemia
11,12
. FISH
data
2
for chromosome 20 copy number in this cell line are in good
agreement with the array data (Fig. 3a), further validating our
procedures. Amplifications of cERBB2 and clone RMC20B4097
relative to clone RMC20P107 in BT474 are easily detected by
visual inspection of the hybridization images (Fig. 3b).
Analysis of a set of four breast tumours, all selected by FISH
using clone RMC20B4097 to have an amplification in region A, is
shown (Fig. 4). Two regions of elevated copy number, D and E,
had been discounted in previous work as loci that probably con-
tain important oncogenes, because high-level amplification was
not detected in tumours. In contrast, this study found both could
be highly amplified (more than threefold, relative to 20p).
Tumour S59 has an indication of the deletion F. Increases and
decreases in copy number on chromosome 20 in this set of
tumours are summarized (Table 1).
Data from nearly contiguous targets
9
in region A show the
capability of array CGH for high resolution analysis of the
boundaries and the amplitude structure of amplicons. In tumour
S21, elevated copy number was found only for RMC20B4097
and RMC20B4130, indicating that the proximal amplicon
boundary is near the neighbouring ends of RMC20B4097 and
RMC20P4039, which are separated by less than 100 kb (ref. 9).
Similarly, in S59, elevated copy number was found for the three
distal clones but not RMC20B4135. Tumour S50, on the other
hand, was elevated in copy number on all four clones in region A,
but the copy number varied twofold (Fig. 4, inset).
We have developed an array CGH procedure that provides
high resolution (approximately 40 kb) measurement of gains and
losses of DNA sequence in genomes of mammalian complexity.
Fluorescence ratios were linearly proportional to copy number
over a dynamic range of several orders of magnitude, but the
Fig. 3 Analysis of breast cancer cell line BT474. a, Copy number variation on
chromosome 20. The arrows A−E show regions of copy number increase
3
,
whereas F shows the location of copy number decrease. The data points indi-
cate the means of three hybridizations and the error bars show the standard
deviations. The points are connected by straight lines that do not provide any
information about the copy number of sequences between the measured loci.
The ratios were normalized to the average of the ratios of the six targets on
chromosome 20p. b, Montage of images of five array targets from an analysis
of breast cancer cell line BT474. The targets are λ DNA, DNA from plasmid PUC
1.77, which contains the abundant repetitive sequence just below the cen-
tromere in chromosome 1, P1 RMC20P107 on chromosome 20p, cosmid cRC-
Neu1, which contains the oncogene cERBB2 and BAC RMC20B4097, which is
located at the peak of region A (Fig. 3a). The upper band shows the DAPI
image of each target. (DAPI stains DNA in general.) The middle and lower
bands show the corresponding fluorescein and Texas red signals from the
BT474 and normal human female reference DNA, respectively. The signals on
RMC20P107 are representative signals for targets at average copy number in
the genome. In comparison, the λ target has a barely perceptible signal, less
than 10% of the single copy level, from the human probes. Similarly, the signal
on the chromosome 1 centromeric repetitive target, PUC1.77, has been sup-
pressed to below 10% of the single copy intensity level. In contrast, the fluores-
cein signal on the cERBB2 target is significantly elevated above single copy
level, whereas the Texas red reference signal is not, indicating the well-known
amplification of cERBB2 in this cell line. Finally, the fluorescein signal on
RMC20B4097 is above saturation, whereas the reference signal is comparable
to the other human targets (its somewhat increased intensity is due primarily
to its longer length).
Fig. 4 Copy number variation on chromosome
20 in four breast tumours. Regions A−F are
labelled as in Fig. 2. All specimens were selected
by FISH for amplification in Region A. Array
measurements, normalized to the average of
the six targets on the 20p, reveal other well-
defined regions of copy number change. The
data points show the means and the error bars
show the standard deviations of three indepen-
dent measurements of each tumour. The inset
for S50 shows an expanded view of the copy
number changes for the four targets in region
A. The horizontal bars indicate the approxi-
mate length and relative locations of the tar-
gets, the heights show the measured ratios.
a
b
FLpter
Normalized ratio
DAPI
Fluorescein
(test)
Texas red
(reference)
lambda
1 cen
P107
cERBB2
B4097
Normalized ratio
Normalized ratio
Normalized ratio
Normalized ratio
FLpter FLpter
© 1998 Nature America Inc. • http://genetics.nature.com
© 1998 Nature America Inc. • http://genetics.nature.com

letter
210 nature genetics volume 20 october 1998
slope of the relationship was somewhat reduced from ideal in the
human genome. Single-copy decreases and increases from
diploid were reliably detected. Within current measurement
capability, all of the approximately 100 human genomic cosmid,
P1 and BAC clones that we have used for targets have provided
useful information. More work is needed, however, to determine
if subtle clone-specific hybridization variation, due perhaps to a
clone’s exact repertoire of repetitive sequences, average base com-
position and so on, will impact the ability to consistently achieve
the high measurement precision we have demonstrated.
Array CGH measurements of one cell line and a small set of
breast tumours has provided new information on copy number
aberrations involving chromosome 20. Most specifically, regions
D, E and F were highlighted. More generally, the array CGH pro-
files of chromosome 20 (Fig. 4, Table 1) show the complexity of
the copy number structure that can occur in this small region of
the genome. These observations, together with the previous data
obtained by FISH (refs 2,3), suggest that if copy number changes
occur on chromosome 20 in breast cancer, multiple regions may
be involved. Moreover, several frequently amplified candidate
oncogenes not contained in the regions identified in our study
also map to chromosome 20q. These include AIB1 (ref. 13) at
20q12, and CAS (ref. 14), TFAP2C (ref. 15) and BTAK (ref. 16), all
at 20q13. Thus, a higher resolution array will probably reveal even
more complex events than those shown (Fig. 3 and Fig. 4). Copy
number increases on 20q have been associated with high S phase
fraction, high grade and early recurrence in node negative breast
cancer patients
17
, with aggressive tumour behaviour
18
, and with
immortalization or extension of life span of primary epithelial
cells in culture
19–21
. Thus, improved understanding of these aber-
rations might provide significant new information on genes that
may contribute to one or more aspects of these phenotypes. Aber-
rations in other regions of the genome have also revealed signifi-
cant complexity on close examination
4,22–24
. Therefore, the
number of genes worthy of identification, and the complexity of
the genetic aberrations that need to be considered in genotype/
phenotype correlational studies, may be significantly greater than
previously recognized. Data from region A in tumours S21, S50
and S59 show that array CGH can provide high-resolution, quan-
titative information to assist in understanding the biological and
clinical consequences of these abnormalities and to refine the
locations of the critical genes they contain.
Although the emphasis of this publication has been on the
analysis of cancer, the capabilities of array CGH make it ideal for
clinical genetic applications such as identification of disease genes,
association of genetic aberrations with phenotype and analysis of
cytogenetically cryptic copy number changes, for example, those
that may occur near chromosome telomeres or translocation
breakpoints. Thus, important applications in multiple areas sug-
gest the value of implementing genome-wide array CGH.
Methods
Target clones. cERBB2, cosmid cRCNeu1 was obtained from R. White.
X chromosome, one clone for the BTK locus at Xq21−23 (RMCXP001) was
selected from the DuPont B P1 library (coordinates 58H8); three non-over-
lapping BAC clones, 590B9, 602N22 and 689C14 from library BC, for the
Oculocerebrorenal syndrome locus at Xq26.1. Chromosome 20, see Table 1.
Specimens. λ DNA 15612-013 was purchased from GIBCO-BRL. Breast
cancer cell line BT474 was grown using standard techniques. Normal
human male and female DNA was isolated from lymphocytes obtained
from a blood bank. Cell lines GMO4626, GMO1415E and GMO5009C,
which have 3, 4 and 5 copies, respectively, of the X chromosome in a
diploid background, were obtained from NIGMS Human Genetic Mutant
Cell Repository, Coriell Institute for Medical Research. Genomic DNA was
isolated from cells and cell lines using standard techniques. Snap-frozen
blocks of primary breast tumour tissue from tumours with a copy number
increase in region A (determined using FISH with clone RMC20B4097)
were cryosectioned (4 µ) and stained with haematoxylin and eosin to
determine the distribution of malignant to non-malignant cells in the sec-
tion. Blocks were trimmed to remove areas containing normal ducts and
lobules, and 10−20 sections of the trimmed block were collected in a
microfuge tube for DNA extraction. Genomic DNA was isolated using
QUIamp tissue kits (29304, Qiagen), including incubation with RNAase A,
following the instructions of the manufacturer, except that lysis times were
sometimes extended depending on the condition of the tissue. DNA was
eluted from the column in water after 5 min incubation at 70 °C. The elu-
tion step was repeated to increase yield.
Array fabrication. Cloned genomic DNA for targets was isolated from bac-
terial cultures (500 ml) using Qiagen maxi kits (12162) following the
instructions of the manufacturer, except that the volume of lysis buffer was
increased 1.5−2-fold. The DNA was resuspended in TE buffer (400 µl,
10 mM Tris, 1 mM EDTA, pH 8), extracted with phenol, chloroform and
isoamyl alcohol (25:24:1) and precipitated with ethanol. After resuspen-
sion in TE buffer, the DNA concentration was determined using a Hoefer
TKO 100 fluorometer. This procedure typically yielded 40−80 µg of P1 or
BAC DNA of sufficient purity to produce targets that bound effectively to
the slides, and which had acceptably low autofluorescence. Target solutions
were made by precipitating DNA (10 µg) in ethanol and dissolving the pel-
let in water (1 µl), followed by addition of DMSO (4 µl) containing nitro-
cellulose (approximately 0.4 µg/µl). The nitrocellulose solution was pre-
pared by dissolving a nitrocellulose membrane (Gibco BRL 41051-012) in
DMSO. Target solutions were stable and were used over periods over 1 yr
without degradation in performance. Glass, quartz and fused silica slides
were cleaned by incubation in 50% concentrated sulfuric acid/50% hydro-
gen peroxide overnight, washed 10 times in distilled water, air dried and
heated to 90 °C for 15 min. The surfaces of the cleaned slides were coated
by immersion in 95% acetone with 0.1% aminopropyltrimethoxy silane
for 2 min at RT, after which they were washed five times in acetone and air
dried. Quadruplicate spots (200−400 µm diameter) of each target solution
were made by depositing the target solutions onto slides using capillary
tubes. After air drying, the target spots were primarily DNA, containing
less than 15% nitrocellulose by mass. They were invisible on the slide sur-
face. As compared with other methods
6,7
, the nitrocellulose substantially
improved the ability to retain hybridizable DNA fragments in the targets.
Signal intensities covering a dynamic range of at least 10
4
could be achieved
by varying probe concentration. However, fragments smaller than approx-
imately 2 kb were not bound effectively, limiting attachment of small PCR
products, oligonucleotides and so on. No additional denaturation of target
DNA was required before hybridization.
Comparative hybridization. Test and reference genomic DNA were
labelled by nick translation with fluorescein dCTP (DuPont NEN NEL424)
and Texas red dCTP (DuPont NEN NEL426), respectively. Unincorporated
nucleotides were removed using a Sephadex G-50 spin column. Labelled
DNA (200−400 ng) was mixed with Cot-1 DNA (35−70 µg; Gibco BRL)
and precipitated with ethanol. (The amount of Cot-1 DNA was based on
fluorometric determinations in our laboratory. We recommend checking
each lot of Cot-1, because use of a sufficient amount is critical to obtain
adequate suppression of repetitive sequences.) The precipitated DNA was
dissolved in hybridization mix (10 µl) to achieve a final composition of
50% formamide, 10% dextran sulfate, 2×SSC, 2% SDS and 100 µg tRNA.
The hybridization solution was heated to 70 °C for 5 min to denature the
DNA, then incubated at 37 °C for approximately 1−5 h to allow blocking of
the repetitive sequences. A wall enclosing the array (approximately
0.5 cm
2
) was made with rubber cement, and the resulting well filled with
hybridization mix. No coverslip was used. The slide was placed in a small
sealed slide box containing 50% formamide (200 µl) and 2×SSC to prevent
evaporation. Hybridization proceeded at 37 °C for 16−72 h on a slowly
rocking table (2−4 cycles/min) to actively transport the hybridization mix
over the array
25
. This resulted in a several-fold increase in signal intensity.
After hybridization, the slide was washed once in 50% formamide, 2×SSC,
pH 7, at 45 °C for 15−30 min, and once in 0.1 M sodium phosphate buffer
with 0.1% NP40, pH 8 at room temperature for 5−10 min. Excess liquid
was drained from the slides and the array mounted in an antifade solu-
tion
26
containing DAPI (1 µg/ml) to counterstain the DNA targets. A glass
coverslip was sealed in place with nail polish. If necessary, the surface of the
© 1998 Nature America Inc. • http://genetics.nature.com
© 1998 Nature America Inc. • http://genetics.nature.com

letter
nature genetics volume 20 october 1998 211
cover slip and back surface of the slide were cleaned with lens cleaner
(D53,881 Edmund Scientific) to remove fluorescent debris and films.
Mounted slides were stable for weeks.
Imaging and analysis. Twelve bit fluorescence intensity data were obtained
using a CCD camera (Sensys, Photometrics, equipped with a Kodak KAF
1400 chip) coupled to a 0.5× or 1× magnification optical system
27
(0.5×
was obtained using an Ercona 25 mm focal length, f/1 C mount video lens
combined with a Cannon 50 mm focal length, f/1.2 photographic lens; 1×
was obtained using a pair of 25-mm focal length video lenses). Center to
center pixel spacing referred to the array was approximately 7 µm in at 1×
magnification, approximately 14 µm at 0.5×. Data from an area up to
14 mm×14 mm could be imaged in a single exposure at 0.5× magnifica-
tion. A Chroma Technology P8100 multiband DAPI/Fluorescein/Texas red
combined with a 625-nm short pass filter was mounted between the lenses
to block excitation light. The use of the 625-nm filter substantially reduced
interference from red autofluorescence originating in the glass slides and
coverslips, permitting their use in place of quartz or fused silica. Excitation
light was supplied by a HBO 100 mercury arc lamp. Kohler illumination
optics provided uniform excitation intensity over the entire imaged area.
The back side of the slide was coupled to the hypotenuse of a right angle
fused silica prism with index matching oil (Cargille 06350 fused silica
matching liquid). Excitation light entered one surface of the prism, passed
through the slide and the array, underwent total internal reflection at the
outer surface of the cover slip and passed back through the array. Images of
the three fluorochromes were collected sequentially using a computer con-
trolled filter wheel to select the appropriate Chroma Technology 8100
series excitation filter for DAPI, fluorescein or Texas red. Signals from test
and reference DNA (fluorescein and Texas red, respectively) at the single
copy level were detectable with exposures as short as 0.01 s. However,
quantitative data were acquired with exposure times of 0.3−2.0 s. Images
were analysed with custom software that segmented the array targets based
on the DAPI staining, estimated and subtracted the background in the flu-
orescein and Texas red images, and calculated the total intensity and the
intensity ratio of fluorescein and Texas red for each target. In addition, the
Pearson’s ‘r’ correlation
28
of a scatter plot of the fluorescein versus Texas red
signal intensities for the pixels in each target was calculated. Data from tar-
gets with ‘r’ values below 0.8 were discarded, because this value indicated
an unreliable measurement, usually due to fluorescent debris. For each
clone, ratios on replicate targets (typically four) with r>0.8 were averaged.
Greater that 95% of target spots on an array met this criterion.
Acknowledgements
Work supported by NIH grants HD 17665, CA 45919 and P50 CA 58207; by
U.S. DOE DE-AC03-76SF00098; California BCRP grants 1IB-003 and
2RB-0225; NIST ATP 94-05-0021; and Vysis. S.C. was supported by a
postdoctoral fellowship from the U.S Army DAMD 17-96-1-6165 and Y.Z.
received postdoctoral support from NIH training grant CA 09215. We thank
U.J. Kim for the Xq26.1 clones.
Received 21 July; accepted 17 August, 1998.
1. Kallioniemi, A. et al. Comparative genomic hybridization for cytogenetic
analysis of solid tumors. Science 258, 818–821 (1992).
2. Tanner, M.M. et al. Increased copy number at 20q13 in breast cancer: defining
the critical region and exclusion of candidate genes. Cancer Res. 54, 4257–4260
(1994).
3. Tanner, M.M. et al. Independent amplification and frequent co-amplification of
three nonsyntenic regions on the long arm of chromosome 20 in human breast
cancer. Cancer Res. 56, 3441–3445 (1996).
4. Bärlund, M. et al. Increased copy number at 17q22−q24 by CGH in breast cancer
is due to high-level amplification of two separate regions. Genes Chromosomes
Cancer 20, 372–376 (1997).
5. Ning, Y. et al. A complete set of human telomeric probes and their clinical
application. Nature Genet. 14, 86–89 (1996).
6. Solinas-Toldo, S. et al. Matrix-based comparative genomic hybridization:
biochips to screen for genomic imbalances. Genes Chromosomes Cancer 20,
399–407 (1997).
7. Schena, M., Shalon, D., Davis, R.W. & Brown, P.O. Quantitative monitoring of
gene expression patterns with a complementary DNA microarray. Science 270,
467–470 (1995).
8. Lockhart, D.J. et al. Expression monitoring by hybridization to high density
oligonucleotide arrays. Nature Biotechnol. 14, 1675–1680 (1996).
9. Collins, C. et al. Positional cloning of ZNF217 and NABC1: Genes amplified at
20q13.2 and overexpressed in breast carcinoma. Proc. Natl Acad. Sci. USA 95,
8703–8708 (1998).
10. Mohapatra, G. et al. Analysis of brain tumor cell lines confers a simple model of
relationships among fluorescence in situ hybridization, DNA index, and
comparative genomic hybridization. Genes Chromosomes Cancer 20, 311–319
(1997).
11. Bench, A.J. et al. A detailed physical and transcriptional map of the region of
chromosome 20 that is deleted in myeloproliferative disorders and refinement
of the common deleted region. Genomics 49, 351–362 (1998).
12. Wang, P.W. et al. Refinement of the commonly deleted segment in myeloid
leukemia with a del(20q). Genes Chromosomes Cancer 21, 75–81 (1998).
13. Anzick, S.L. et al. AIB1, a steroid receptor coactivator amplified in breast and
ovarian cancer. Science 277, 965–968 (1997).
14. Brinkmann, U., Gallo, M., Polymeropoulos, M.H. & Pastan, I. The human CAS
(cellular apoptosis susceptibility) gene mapping on chromosome 20q13 is
amplified in BT474 breast cancer cells and part of aberrant chromosomes in
breast and colon cancer cell lines. Genome Res. 6, 187–194 (1996).
15. Williamson, J.A. et al. Chromosomal mapping of the human and mouse
homologues of two new members of the AP-2 family of transcription factors.
Genomics 35, 262–264 (1996).
16. Sen, S., Zhou, H. & White, R.A. A putative serine/threonine kinase encoding
gene BTAK on chromosome 20q13 is amplified and overexpressed in human
breast cancer cell lines. Oncogene 14, 2195–2200 (1997).
17. Tanner, M. et al. Amplification of chromosomal region 20q13 in invasive breast
cancer: prognostic implications. Clin. Cancer Res. 1, 1455–1461 (1995).
18. Courjal, F. et al. DNA amplifications at 20q13 and MDM2 define distinct subsets
of evolved breast and ovarian tumours. Br. J. Cancer 74,1984–1989 (1996).
19. Reznikoff, C.A. et al. Long-term genome stability and minimal genotypic and
phenotypic alterations in HPV16 E7-, but not E6-, immortalized human
uroepithelial cells. Genes Dev. 8, 2227–2240 (1994).
20. Savelieva, E. et al. 20q gain associates with immortalization: 20q13.2 amp-
lification correlates with genome instability in human papillomavirus 16 E7
transformed human uroepithelial cells. Oncogene 14, 551–560 (1997).
21. Salinas-Toldo, S., Dürst, M. & Lichter, P. Specific chromosomal imbalances in
human papillomavirus-transfected cells during progression toward immortality.
Proc. Natl Acad. Sci. USA 94, 3854–3859 (1997).
22. Bièche, I. et al. Two distinct regions at 17q11−q12 involved in human primary
breast cancer. Cancer Res. 56, 3886–3890 (1996).
23. Karlseder, J. et al. Patterns of DNA amplification at band q13 of chromosome 11
in human breast cancer. Genes Chromosomes Cancer 9, 42–48 (1994).
24. Reifenberger, G. et al. Refined mapping of 12q13-15 amplicons in human
malignant gliomas suggests CDK/SAS and MDM2 as independent amplification
targets. Cancer Res. 56, 5141–5145 (1996).
25. Albertson, D.G., Fishpool, R.M. & Birchall, P.S. Fluorescent in situ hybridization
for the detection of DNA and RNA. in C. elegans: Modern Biological Analysis of
an Organism (eds Epstein, H.F. & Shakes D.C.) 339−364 (Academic Press, New
York,1995).
26. Johnson, G. et al. A simple method of reducing the fading of immuno-
fluorescence during microscopy. J. Immunol. Methods 43, 349–350 (1981).
27. Wittrup, K.D., Westerman, R.J. & Desai, R. Fluorescence array detector for large-
field quantitative fluorescence cytometry. Cytometry 16, 206–213 (1994).
28. Press, W.H., Teukolsky, S.A., Vetterling, W.T. & Flannery, B.P. Numerical Recipes
in C: the Art of Scientific Computing 2nd edition636–639 (Cambridge University
Press, Cambridge, 1994).
29. Shepherd, N.S. et al. Preparation and screening of an arrayed human genomic
library generated with the P1 cloning system. Proc. Natl Acad. Sci. USA 91,
2629–2633 (1994).
30. Kim, U.J. et al. A bacterial artificial chromosome-based framework contig map
of human chromosome 22q. Proc. Natl Acad. Sci. USA 93, 6297–6301 (1996).
31. Stokke, T. et al. A physical map of chromosome 20 established using
fluorescence in situ hybridization and digital image analysis. Genomics 26,
134–137 (1995).
© 1998 Nature America Inc. • http://genetics.nature.com
© 1998 Nature America Inc. • http://genetics.nature.com