Comparative analysis of copy number detection by whole-genome BAC and oligonucleotide array CGH.

Abstract

Microarray-based comparative genomic hybridization (aCGH) is a powerful diagnostic tool for the detection of DNA copy number gains and losses associated with chromosome abnormalities, many of which are below the resolution of conventional chromosome analysis. It has been presumed that whole-genome oligonucleotide (oligo) arrays identify more clinically significant copy-number abnormalities than whole-genome bacterial artificial chromosome (BAC) arrays, yet this has not been systematically studied in a clinical diagnostic setting.
To determine the difference in detection rate between similarly designed BAC and oligo arrays, we developed whole-genome BAC and oligonucleotide microarrays and validated them in a side-by-side comparison of 466 consecutive clinical specimens submitted to our laboratory for aCGH. Of the 466 cases studied, 67 (14.3%) had a copy-number imbalance of potential clinical significance detectable by the whole-genome BAC array, and 73 (15.6%) had a copy-number imbalance of potential clinical significance detectable by the whole-genome oligo array. However, because both platforms identified copy number variants of unclear clinical significance, we designed a systematic method for the interpretation of copy number alterations and tested an additional 3,443 cases by BAC array and 3,096 cases by oligo array. Of those cases tested on the BAC array, 17.6% were found to have a copy-number abnormality of potential clinical significance, whereas the detection rate increased to 22.5% for the cases tested by oligo array. In addition, we validated the oligo array for detection of mosaicism and found that it could routinely detect mosaicism at levels of 30% and greater.
Although BAC arrays have faster turnaround times, the increased detection rate of oligo arrays makes them attractive for clinical cytogenetic testing.

Full text

Page 1

RESEARCHOpen Access
Comparative analysis of copy number detection
by whole-genome BAC and oligonucleotide
array CGH
Nicholas J Neill, Beth S Torchia, Bassem A Bejjani, Lisa G Shaffer, Blake C Ballif*
Abstract
Background: Microarray-based comparative genomic hybridization (aCGH) is a powerful diagnostic tool for the
detection of DNA copy number gains and losses associated with chromosome abnormalities, many of which are
below the resolution of conventional chromosome analysis. It has been presumed that whole-genome
oligonucleotide (oligo) arrays identify more clinically significant copy-number abnormalities than whole-genome
bacterial artificial chromosome (BAC) arrays, yet this has not been systematically studied in a clinical diagnostic
setting.
Results: To determine the difference in detection rate between similarly designed BAC and oligo arrays, we
developed whole-genome BAC and oligonucleotide microarrays and validated them in a side-by-side comparison
of 466 consecutive clinical specimens submitted to our laboratory for aCGH. Of the 466 cases studied, 67 (14.3%)
had a copy-number imbalance of potential clinical significance detectable by the whole-genome BAC array, and
73 (15.6%) had a copy-number imbalance of potential clinical significance detectable by the whole-genome oligo
array. However, because both platforms identified copy number variants of unclear clinical significance, we
designed a systematic method for the interpretation of copy number alterations and tested an additional 3,443
cases by BAC array and 3,096 cases by oligo array. Of those cases tested on the BAC array, 17.6% were found to
have a copy-number abnormality of potential clinical significance, whereas the detection rate increased to 22.5%
for the cases tested by oligo array. In addition, we validated the oligo array for detection of mosaicism and found
that it could routinely detect mosaicism at levels of 30% and greater.
Conclusions: Although BAC arrays have faster turnaround times, the increased detection rate of oligo arrays makes
them attractive for clinical cytogenetic testing.
Introduction
Molecular cytogenetic techniques such as array-based
comparative genomic hybridization (aCGH) have revolu-
tionized cytogenetic diagnostics and, in turn, the clinical
management of patients with developmental delays and
multiple congenital anomalies [1,2]. These rapid, high-
resolution, and highly accurate techniques have identi-
fied numerous previously unrecognized chromosomal
syndromes [3-8], refined critical regions for established
genetic defects [9], and broadened our view of the “nor-
mal” diploid genome [10]. In addition, aCGH has given
the clinician a greater appreciation of variability in the
clinical presentation of many well-described conditions
[11,12] and allowed for the discovery of new conditions
with relatively mild phenotypes [13,14]. Furthermore,
the application of aCGH has created a paradigm shift in
genetics that has moved the description and discovery
of genetic conditions from the “phenotype-first”
approach, in which patients exhibiting similar clinical
features are identified prior to the discovery of an
underlying etiology, to a “genotype-first” approach, in
which a collection of individuals with similar copy-num-
ber imbalances can be examined for common clinical
features [15].
Originally, targeted microarrays constructed from bac-
terial artificial chromosomes (BAC) were developed for
the clinical laboratory because of their ability to clearly
* Correspondence: ballif@signaturegenomics.com
Signature Genomic Laboratories, Spokane, WA, USA
Neill et al. Molecular Cytogenetics 2010, 3:11
http://www.molecularcytogenetics.org/content/3/1/11
© 2010 Neill et al; licensee BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons
Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in
any medium, provided the original work is properly cited.

Page 2

identify copy number changes in discrete regions of the
human genome known to play a role in genetic disease
[16]. This “less is more” idea prevailed in the early years
of clinical aCGH because the technology was new and
proof of principle was required before it could be
adopted for more widespread diagnostic use. Further-
more, the identification of copy number alterations of
unclear clinical significance was considered undesirable
to the diagnostician, the ordering physician, and the
patient’s family. Recently, the coverage of microarrays
has expanded to include more comprehensive coverage
of the human genome, leading many to suggest that
whole-genome BAC or oligo arrays are the next step in
the continued improvement in the detection rate of
cytogenetic abnormalities.
It has been presumed that whole-genome oligonucleo-
tide arrays, because they have higher resolutions, would
detect more copy number aberrations than whole-
genome BAC arrays. However, to our knowledge, there
has not been a systematic comparison of these two
whole-genome copy number screening technologies in a
clinical diagnostic environment. Therefore, to determine
which platform is most effective in identifying clinically
significant DNA copy number alterations, we designed a
whole-genome BAC array and a whole-genome oligo
array and compared the results in a blinded study of
466 clinical diagnostic specimens. In addition, we pro-
spectively evaluated 3,443 patients by the whole-genome
BAC array and 3,096 patients by the whole-genome
oligo array and compared the detection rates of clini-
cally significant abnormalities and those of unclear clini-
cal significance. Finally, we validated our oligo array
with 48 cases to determine the level of mosaicism that
can be reliably detected and compared that level to our
previously published cases analyzed using the BAC
array.
Materials and methods
Whole-genome BAC array design and aCGH
We constructed a whole-genome BAC array designed
for clinical diagnostic use using >4,600 BAC clones. All
clones were validated by FISH prior to inclusion on the
array using previously described validation procedures
[16]. Contigs of 3-6 overlapping clones were selected to
cover 1,543 genetic loci, including >150 known micro-
deletion/microduplication syndromes and increased den-
sity of coverage in the 5-10 Mb surrounding the
subtelomeric and pericentromeric regions of the gen-
ome. In addition, we placed contigs to cover >500 func-
tionally significant genes such as transcription factors
and other genes known to play important roles in devel-
opment. This coverage also includes genome-wide
representation with at least one contig in nearly every
chromosomal band at the resolution of an 850-band
karyotype. The mean gap size for the whole-genome
BAC array is ~1.6 Mb. Microarray manufacturing
and aCGH analysis using the whole-genome BAC array
were performed as previously described [13]. BAC arrays
were analyzed after a dye-swap, two-experiment analysis
[16], using sex-mismatched controls. Results were then
displayed using custom BAC aCGH analysis software
(Genoglyphix™; Signature Genomic Laboratories,
Spokane, WA).
Whole-genome Oligonucleotide Array Design and aCGH
Oligonucleotide-basedmicroarray
performed using a custom-designed, 105K-feature
whole-genome microarray manufactured by Agilent
Technologies (Santa Clara, CA) with one probe every 10
kb in regions of interest–microdeletion/microduplication
syndromes, the pericentromeric regions, subtelomeres
and genes involved in important developmental path-
ways–for an average of 50 oligos per clinical locus. In
addition, to achieve backbone coverage, we placed a
probe, on average, every 35 kb throughout the rest of
the genome between the regions of interest. Genomic
DNA labeling was performed as described for BAC
arrays, whereas array hybridization and washing were
performed as specified by the manufacturer (Agilent
Technologies). A dye swap was not performed for the
oligo arrays, and sex-matched controls were used.
Arrays were scanned and analyzed as previously
described [17]. Results of aberration calls consisting of
five or more consecutive oligos were then displayed
using custom oligonucleotide aCGH analysis software
(Genoglyphix™; Signature Genomic Laboratories). The
use of five consecutive oligos achieved a resolution of
40 kb in the regions of interest and a resolution of 140
kb in the backbone.
analysiswas
Decision Algorithm for Clinically Significant Copy Number
Reporting
We developed a decision algorithm for classifying clini-
cally significant copy number alterations, alterations of
unclear clinical significance, and alterations of no cur-
rently known clinical significance. Alterations that were
associated with established chromosomal syndromes,
were large and affected a significant amount of gene
content, or were part of a complex rearrangement such
as an unbalanced translocation, insertion, or marker
chromosome were characterized as clinically significant.
Alterations with unclear clinical significance were most
commonly those which were not currently associated
with a syndrome but which affected gene content which
may have contributed to the patient’s phenotype and
those which could not be precisely refined by the BAC
array. Alterations were considered to have no known
clinical significance if they were small, affected minimal
Neill et al. Molecular Cytogenetics 2010, 3:11
http://www.molecularcytogenetics.org/content/3/1/11
Page 2 of 14

Page 3

gene content, and/or were present in regions where
common copy-number variation was known to occur in
the general population. Signature’s own Genoglyphix
Chromosome Aberration Database (GCAD) was used as
a reference to assist in the interpretation of each altera-
tion. GCAD is a database of >11,000 chromosomal
abnormalities identified in >9,500 patients out of
>40,000 patients evaluated by our laboratory and con-
tains detailed statistics of each observed alteration
(breakpoint coordinates, size, gene content, etc.) as well
as clinical information pertaining to patient referral.
Fluorescence in situ Hybridization (FISH)
When possible, all copy number alterations detected by
microarray analysis were visualized by interphase and/or
metaphase FISH using a BAC probe located within the
region of gain or loss. FISH was performed as previously
described [18].
Patient Clinical Testing
To validate the custom-designed 105k oligo array com-
pared to the whole-genome BAC array, 466 cases were
run side-by-side in a platform comparison study. In
each case, the clinically validated BAC array results were
used for interpretation and reporting. Specimens with
known chromosome abnormalities, parental specimens,
and prenatal cases were excluded from the analysis.
In addition, we conducted a prospective study of 3,443
consecutive BAC microarray analyses and 3,096 conse-
cutive oligo microarray analyses in our clinical labora-
tory. The array platform used for testing in each case
was chosen by the referring physician at the time of
sample submission to our clinical diagnostic laboratory.
Cases with previously known chromosomal abnormal-
ities, parental samples, and prenatal specimens were
again excluded from the data collection.
Mosaicism Assessment
The ability of the oligo platform to detect mosaicism
was assessed on 48 patients previously known to carry
mosaic abnormalities at levels as low as 5%. The altera-
tions studied included a variety of interstitial, terminal,
and whole-chromosome copy-number abnormalities, as
well as marker chromosomes. The mosaic alteration in
each patient was initially assessed by BAC array and the
level of mosaicism determined by interphase FISH ana-
lysis when possible. In a separate experiment, mosaicism
was assessed using a dilution of cells from a male with
trisomy 21 with normal male control cells, as previously
described [19]. After FISH verification of the dilutions,
DNA was extracted from the diluted cells, labeled and
hybridized to the custom-designed oligo array as
described above.
Results
Platform comparison study
From the 466 cases analyzed by the BAC array, using
the previously described algorithm, we excluded 347
cases that only had aberrations located within regions
that contained no genes and/or aberrations that had
been established to be normal population variants by
Signature Genomic Laboratories or identified in the
Toronto Database of Genomic Variants (DGV, http://
projects.tcag.ca/variation/). After these cases were
excluded, 138 copy number alterations in 119 cases
(25.5% of the original 466 cases) remained that required
FISH analysis. These aberrations included subtelomeric
and pericentromeric gains for which FISH was required
to exclude an unbalanced translocation or a marker
chromosome. After FISH was performed, 60 aberrations
in 52 cases were classified as normal variants because
marker chromosomes and derivative chromosomes were
not identified and because these alterations were located
within regions where common copy number variation is
known to occur. Thus, alterations of potential clinical
significance according to our algorithm were identified
in 67 cases, a detection rate of 14.4%. Of these cases, 56
(12.0%) were considered to contain clinically significant
copy number alterations (Table 1), and 11 (2.4%) were
considered to contain copy number variants of unclear
clinical significance for which parental analyses were
recommended to further clarify the abnormality (Table
2). aCGH and FISH analysis performed on parental
samples revealed that six alterations of unclear signifi-
cance were inherited from a carrier parent and one was
a de novo event in the proband. The origin of the other
four unclear alterations could not be determined.
Using the oligo array, we identified 1,337 copy number
variations among the same 466 cases. Using the algo-
rithm previously described, we excluded 1,172 aberra-
tions that were located within regions that had no gene
content or those that were common copy number var-
iants. After these exclusions were made, 165 aberrations
in 138 cases (29.6%) remained that required FISH analy-
sis. After FISH analysis was performed, aberrations of
potential clinical significance were identified in 73 cases,
a detection rate of 15.7%. Of these, the same 56 (12.0%)
cases that were identified by the BAC platform were con-
sidered to contain clinically significant alterations (Table
1) and 17 (3.7%) were determined to contain copy num-
ber variants of unclear clinical significance.
Table 3 shows the six cases for which aberrations of
unclear clinical significance were identified by the oligo
array but not by the BAC array. In all six cases, the
aberrations either fell within the gaps in the BAC array
coverage or were only partially covered by one or more
BACs. The average size of the alterations that were not
Neill et al. Molecular Cytogenetics 2010, 3:11
http://www.molecularcytogenetics.org/content/3/1/11
Page 3 of 14

Page 4

Table 1 Cases with Alterations of Clinical Significance Identified by BAC and Oligo aCGH
Pt. # ChrBand Start pos.End pos.Gain/LossSize# of BACs # of Oligos
21637
chr16p11.2 29,563,98530,066,187Loss 502,202546
21992
chr16 p11.229,563,985 30,066,187Loss502,202546
22013
chr16 p11.229,563,985 30,066,187 Loss 502,2025 46
21993
chr10 q25.2q25.3114,306,207 114,925,368 Loss 619,1613 36
21756
chr2q35219,418,281 220,060,969 Loss642,6882 43
22002
chr2q37.3 239,664,393240,400,008 Loss 735,6153 82
22334
chr10 q25.2q25.3 114,024,053115,677,301 Gain1,653,248353
21688
chr16 p13.11p12.315,056,25716,742,812Gain1,686,555361
21667
21896
chr17
chr4
p12
q35.2
14,052,297
189,407,487
15,742,271
191,133,809
Gain
Loss
1,689,974
1,726,322
3
8
67
147
22269
chr2 q12.3q13107,945,041109,784,684 Loss1,839,643 6102
21640
chr19q13.42 59,272,450 61,239,237Gain1,966,7878135
22117
chr9q33.1118,991,777121,063,590Gain 2,071,813367
22066
chr8p12p11.2138,303,146 40,515,492Gain2,212,3466104
21786
chr1q21.1144,973,942147,421,814Gain2,447,872563
22237
chr1q21.1144,973,942147,421,814Gain2,447,872563
22310
chr22 q11.2117,299,74219,770,655Loss2,470,91313205
22050
chr22q11.2117,007,81919,770,655Loss 2,762,83613 208
21936
chr1q21.1144,973,942 147,966,185Gain 2,992,243565
21719
chr2 q31.1169,823,689172,870,083Loss3,046,3948 119
22128
chr4 q34.3 179,065,989182,435,119Loss 3,369,130391
22006
chr17p11.216,723,07120,145,604Loss3,422,53318311
22174
chr1q41q42.12221,260,860 224,709,317Loss3,448,4579 178
21971
chr22q11.23q12.2 24,025,269 28,008,109Loss 3,982,8403 111
22073
chr15q11.2q13.121,208,177 26,194,049Loss 4,985,87211281
21687
chr16q12.2q21 51,912,65557,173,018 Loss5,260,36315261
21975
chr5p15.2p14.3 13,514,46418,988,928Loss5,474,4643122
21555
chr7p22.3p22.1153,6446,230,285Gain6,076,64132434
21547
chr2 q24.3q31.1168,702,606174,842,496Loss6,139,89011226
21755
chr11p12p11.237,540,68043,940,573Loss6,399,89314298
21761
chr3p14.1p12.370,738,91477,275,908Loss6,536,9946151
22151
chr15 q11.2q13.118,809,804 26,194,049Gain7,384,24514331
22322
chr8p21.3p1222,954,21230,630,828Loss7,676,6169224
21889
chr2q33.1q34202,901,021211,366,732Gain8,465,711 13281
21723
chr5q23.1q23.3 121,487,477130,306,377Loss8,818,9004 204
22337
chr1p36.22p36.139,476,88019,436,653Loss 9,959,77321436
21795
chr1 p34.2p32.341,201,83755,191,500Gain13,989,66317440
20986
chr12p13.33p12.3 84,91817,505,135Gain17,420,217 46800
21957
chr11 q23.3q25116,478,434 134,419,382Gain 17,940,948 40784
21596
chr1 q25.1q32.1173,519,967203,663,817 Gain 30,143,85025814
22055
chr3p14.1p13 71,164,16171,958,845Loss794,684346
21566*
chr17p13.2p13.1 6,081,4576,904,679Loss823,222 314
21558
chr22q13.3348,567,18549,517,230Loss950,045 654
21770
chr8p23.3202,5051,411,517 Loss 1,209,0125 96
22254
chr8p23.22,604,280 3,966,809Loss1,362,529 9 140
21937
chr7q11.23 72,404,04973,771,409Loss 1,367,36010 158
Neill et al. Molecular Cytogenetics 2010, 3:11
http://www.molecularcytogenetics.org/content/3/1/11
Page 4 of 14

Page 5

detected by the BAC array was 1.12 Mb (range: 289 kb -
1.42 Mb).
In two cases, the oligo microarray identified additional
complexity that was not recognized by the BAC array.
In patient 21566, the BAC array identified one intersti-
tial deletion of 17p13.2p13.1, whereas oligo array analy-
sis identified that deletion and an additional interstitial
deletion in the same band (data not shown). In patient
21897, the BAC array identified a 6.8 Mb terminal dele-
tion of 5p, whereas oligo array analysis identified that
deletion and a 1.4 Mb duplication proximal to the
deleted region (Figure 1).
Prospective Diagnostic Comparison
Of the 3,443 diagnostic specimens analyzed using our
whole-genome BAC array, 605 (17.6%) had copy num-
ber alterations. Using the previously described algo-
rithm, 365 (10.6%) had abnormalities that were classified
as clinically significant, whereas 240 (6.9%) had copy
number variants of unclear clinical significance.
Of the 3,096 diagnostic specimens analyzed using our
whole-genome oligo array during the same time period,
698 (22.5%) had copy number alterations. Using the
previously described algorithm, 477 (15.9%) of these
cases were determined to contain alterations considered
to be clinically significant and 221 (7.0%) were deter-
mined to contain copy number variants of unclear clini-
cal significance (Table 4).
The increased number of cases with clinically signifi-
cant alterations detected by the oligo array was found to
be statistically significant using a Fisher’s Exact Test
(OR = 1.5359, p < .0001). The increased number of
cases with alterations of unclear significance detected by
the oligo array was not statistically significant (OR =
1.0259, p = 0.8090).
Mosaicism Assessment
All but three of the 48 previously known mosaic altera-
tions were detected by the oligo array. FISH analysis
estimated that the proportion of uncultured cells carry-
ing the alteration was 24% in the first case, while the
proportion in cultured cells was 6%. In the second case,
5% of cells were found to carry the alteration by FISH
(data not shown). The proportion of cells carrying the
alteration in the third case could not be determined
because FISH confirmation was not possible on the
Table 2 Cases with Alterations of Unclear Significance Identified by BAC and Oligo aCGH
Pt. #Chr BandStart pos. End pos. Gain/LossSize# of BACs # of OligosInheritance
22365
chr9p23 9,881,3859,984,838 Loss103,453215Paternal
21702
chr16 p12.2 21,486,89721,641,890Loss 154,9932 16Paternal
21883
chr16p12.1 21,974,396 22,338,234 Gain363,8383 49Maternal
21860
chr16p12.121,907,270 22,338,234Loss430,9643 50 Unknown
22009
chr22q11.2119,069,125 19,770,655Loss 701,5304 70 Paternal
22102
chr16 p13.11 14,981,04416,166,985Gain 1,185,941360Maternal
22246
chr10p11.22 31,591,31032,792,762 Loss1,201,4522 40Unknown
21893
chrX p11.32 45,930,65246,382,140Gain451,4883 34Maternal
22273
chrXq28153,355,101 154,317,591Gain962,4905 46Unknown
10245
chr3p14.167,727,84169,101,769 Loss1,373,928 225 Unknown
22348
chr13q22.274,989,69975,378,640Loss388,941351De novo
Table 1: Cases with Alterations of Clinical Significance Identified by BAC and Oligo aCGH (Continued)
21710
chr15q13.2q13.328,741,81830,186,356Loss1,444,538364
21722
chr15q13.2q13.3 28,741,818 30,226,376Loss1,484,558 3 65
21739
chr15q13.2q13.328,741,81830,226,376Loss 1,484,558365
21787
chr15q13.2q13.328,741,818 30,226,376Loss1,484,558365
21897*
chr5p15.28,511,5929,888,817Gain1,377,225 3429
21884
chrXq28152,676,750153,059,428Gain382,678344
21592
chrXp22.33q28701 154,888,083 Gain154,887,382325 6888
22087
chrYp11.32262,57857,715,879 Gain 57,453,30149 49
22285
chrXp21.131,759,55131,830,811 Loss71,260211
22244
chr22 q13.3349,342,96149,514,486Loss 171,525256
*additional alterations identified by oligonucleotide aCGH.
Neill et al. Molecular Cytogenetics 2010, 3:11
http://www.molecularcytogenetics.org/content/3/1/11
Page 5 of 14