| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Molecular Diagnostics and Genetics |
1 Department of Laboratory Medicine, Children’s Hospital Boston, Boston, MA.
2 Center for Human Genetic Research, Massachusetts General Hospital, Boston, MA.
3 Division of Genetics, Children’s Hospital Boston, Boston, MA.
4 Departments of Medicine, Neurology, Pathology, Pediatrics, Harvard Medical School, Boston, MA.
5 Department of Molecular and Human Genetics, Baylor College of Medicine, Houston, TX.
6 Howard Hughes Medical Institute, Chevy Chase, MD.
aAddress correspondence to this author at: Departments of Laboratory Medicine and Pathology, Children’s Hospital and Harvard Medical School, 300 Longwood Ave., Boston, MA 02115. Fax 617-730-0338; e-mail bai-lin.wu{at}childrens.harvard.edu.
 |
Abstract |
|---|
 Top Abstract IntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Methods: We selected >10 000 60-mer oligonucleotide features from Agilent’s eArray probe library to interrogate all subtelomeric and pericentromeric regions and 95 additional clinically relevant regions for a total of 179 loci. Sensitivity and specificity were measured for 105 patient samples, including 51 with known genomic-imbalance events, as detected by bacterial artificial chromosome–based array CGH, FISH, or multiplex ligation-dependent probe amplification.
Results: Focused array CGH detected all known regions of genomic imbalance in 51 validation samples with 100% concordance and an excellent signal-to-noise ratio. The mean SD among log2 ratios of all noncontrol features without copy number alteration was 0.062 (median, 0.055). Clinical testing of another 211 samples from individuals with developmental delay, unexplained mental retardation, dysmorphic features, or multiple congenital anomalies revealed genomic imbalance in 25 samples (11.9%).
Conclusions: This focused oligonucleotide-array CGH assay, a flexible, robust method for clinically diagnosing genetic disorders associated with genomic imbalance, offers appreciable advantages over currently available platforms.
|
Introduction |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Molecular techniques such as multiplex ligation-dependent probe amplification (MLPA) and real-time PCR are alternatives to multiple FISH assays for evaluating the genomic copy number of multiple targets(4). Microarray-based comparative genomic hybridization (CGH) offers the ability to interrogate many more genomic regions in a single assay. Early CGH arrays were composed of large-insert bacterial artificial chromosome (BAC) clones(5). BAC-based arrays have revolutionized the detection of genomic imbalance in clinical cytogenetic laboratories(6)(7) but are challenging to develop, validate, and manufacture. The fact that BAC clones in standard libraries may be inaccurately mapped could lead to diagnostic errors without careful validation(8). Additionally, BAC clone inserts average approximately 150 kb, limiting the resolution of detectable copy-number variants (CNVs) to the size of a BAC insert. Deletion breakpoints that extend beyond the BAC clone cannot be accurately determined. Once validated, BAC arrays are much more efficient than multiplex FISH analysis, but genetic information is constantly changing. Consequently, updates to a BAC-based array require successive rounds of extensive probe validation.
Oligonucleotide-based arrays offer advantages over BAC-based arrays, and many platforms are available. Oligonucleotide arrays designed for genotyping single-nucleotide polymorphisms may not provide uniform coverage at all sites of genomic imbalance(9)(10). Custom oligonucleotide arrays that are based on libraries of validated synthetic probes can interrogate clinically relevant genomic regions without the need for large-insert clone libraries. We describe an array based on Agilent’s eArray library, a large collection of 60-mer oligonucleotides specifically selected for robust copy-number analysis(11). This targeted oligonucleotide-based array provides a flexible and adaptable method for CGH to detect genomic copy-number imbalance in the clinical diagnostic laboratory.
|
Materials and Methods |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
clinical samples
After assay validation, we performed clinical array CGH testing of 211 consecutively submitted samples from presumably unrelated children. Samples were submitted after referral to specialists in the Divisions of Clinical Genetics and Developmental Medicine, and the Department of Neurology for clinical molecular-diagnostic testing. The referring diagnoses for these patients included developmental delay, mental retardation, dysmorphic features, or multiple congenital anomalies. All samples were compared with a reference sample for standard 2-color array CGH, either a 46,XY male or a 46,XX female sample. Reference DNA was purchased from Promega.
Genomic DNA was extracted from whole blood for all samples with a D50K PureGene DNA-isolation reagent set (Qiagen/Gentra) according to the manufacturer’s instructions. All DNA was stored at –20 °C.
chip design
This focused oligonucleotide chip covers 179 clinically relevant regions of genomic imbalance, including all subtelomeric and pericentromeric regions, and 95 regions responsible for well-defined microdeletion/microduplication syndromes, mental retardation, and autism (for a summary of array coverage, see Supplemental Data 1 in the Data Supplement that accompanies the online version of this article athttp://www.clinchem.org/content/vol53/issue12). A total of 10 207 region-specific features and 603 quality-control and negative-control features selected from Agilent’s eArray library(11) are randomly located on the array with an average spatial resolution of <35 kb within the targeted regions. Each subtelomeric region has a minimum coverage of 5 Mb. A subset of 660 features is duplicated on each block as a quality-control measure. Arrays were manufactured with Agilent’s SurePrint Inkjet technology. In designing the targeted oligonucleotide-based array, we consulted the Database of Genomic Variants (http://projects.tcag.ca/variation/) to avoid CNVs with no apparent clinical relevance.
cgh
Oligonucleotide-array CGH was performed according to the manufacturer’s Oligonucleotide Array–Based CGH for Genomic DNA Analysis protocol (version 3; Agilent Technologies; see Supplemental Data 2 in the online Data Supplement for a summary of the protocol).
Dye-swap verification was performed on all samples with positive findings. For other confirmation assays, we carried out BAC-array CGH and FISH confirmation as described previously(12). MLPA confirmation was performed as described previously(13). The MLPA oligonucleotides for the CTNS2 (cystinosis, nephropathic) gene are as follows: exon 2, GTTTTCACACTGGGCGAAGGGAGGACT and CCTGAGCTCTGCCTCTTCCAGTAACATTG; exon 6, CCGAGGATACGCTTTCTTGTGATCC and GCAGCAGCGCCATTAGCATCATAAACC; exon 12, CAACCAAGTTTGGACTCGGGGT and CTTCTCCATCGTCTT CGACGTCGTC.
data analysis
Scanned images were quantified with Feature Extraction software (version 9.0; Agilent Technologies). We used the signal-to-noise ratio (SNR) and the normalized log2 ratio (test:reference) with 40 nonpathologic individual DNA samples to evaluate the quality and variability of each feature/target. The SNR was calculated by dividing the mean signal intensity of each feature by the mean background signal intensity. Features with a mean SNR <4 or an SD of the log2 ratio >0.1 were considered to have poor signal quality and high variability and were filtered out before further analysis. These thresholds were chosen empirically and are similar to those used in similar studies(1)(14).
We visualized the filtered data further with CGH Analytics software (version 3.4; Agilent Technologies) and evaluated the quality of each test with the quality-control metrics generated with CGH Analytics software. Copy-number aberration was indicated with the Aberration Detection Method 2 algorithm for the data that passed quality-control testing. The Aberration Detection Method 2 algorithm finds intervals of varying size with a consistent, appreciably low, or high log2 ratio. An aberration filter was set to indicate regions with at least 3 targets showing the same direction in copy-number change. The mean log2 ratio of each region of potential imbalance was calculated and compared with the SD for the whole dataset. A copy-number gain was called if the mean log2 ratio was greater than twice the SD of the whole dataset, and a loss was called if the mean was less than –2 SDs. These thresholds were chosen empirically and are similar to those used in other such studies(15)(16). Cutoff values for genomic imbalance can be adjusted and set accordingly with the threshold function of CGH Analytics software, especially when a potential mosaic scenario is encountered. Variants not known to be pathogenic were compared with the Database of Genomic Variants (http://projects.tcag.ca/variation/) to facilitate interpretation.
|
Results |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
We demonstrated a log2 ratio SD of >0.1 for 714 features (7.12%); we categorized these features as nonideal targets and excluded them from further analysis. More than 90% of the features passed the feature level filter criteria: an absolute mean log2 ratio <0.1, a log2 ratio SD <0.1, and a mean SNR >4. After excluding nonideal features, the dataset quality improved dramatically. For example, the SD of the log2 ratio dropped from >0.06 to <0.03. Features with a large log2 SD largely overlapped those with a low SNR, further validating this filtering approach. Because the nonideal targets are approximately evenly distributed across the target regions, the overall resolution of the chip is not appreciably affected.
Several key variables were used to evaluate chip quality and to describe the quality of the dataset as a whole. These variables included probe-to-probe log2 ratio noise (DLRSpread), the median signal intensity of both channels, background noise for both channels, and SNR. The following cutoffs were used to pass our quality-control testing: DLRSpread 
0.25, median signal intensity 
50, background noise 10, and SNR 15. None of the samples failed quality-control testing because of poor chip quality or problems with hybridization. Two samples failed testing because of DNA impurities; both samples passed quality-control tests after we repurified the DNA.
chip validation with blinded samples
We next blindly tested 65 samples for further chip validation. Genomic imbalance had previously been detected in 51 of the 65 samples by BAC-array CGH, FISH, karyotyping, or MLPA, or by some combination of these analyses, and these 51 samples served as positive controls for validation. The remaining 14 samples had previously been tested by targeted BAC-array CGH with nonpathologic results and thus served as negative controls for validation. All samples were traceable to the technologist who performed the hybridizations.
Of the 51 validation samples with a previously detected genomic imbalance (see Supplemental Data 3 in the online Data Supplement), the samples from 2 cases demonstrated aneuploidy for an entire chromosome, 3 cases involved unbalanced chromosomal rearrangements, 13 cases had subtelomeric deletions/duplications, 17 cases had interstitial deletions/duplications, and 15 cases were associated with known segmental aneuploidy regions, including Angelman/Prader–Willi syndrome (4 cases), atypical Angelman syndrome on 22q13.3 (1 case), an autism phenotype associated with duplication of 15q11-q13 (1 case), a velocardiofacial/DiGeorge syndrome region (5 cases with deletions and 2 with duplication), and Williams–Beuren syndrome (2 cases).
Across all samples, results from oligonucleotide-array CGH were consistent with the results obtained with the 4 prior methods, but oligonucleotide-based array CGH provided the most precise breakpoint boundaries. Fig. 1A
shows a genomic-imbalance event (2q37.3 deletion) identified by oligonucleotide-based array CGH, and Fig. 1B
shows the FISH confirmation of the 2q37.3 deletion.
|
Oligonucleotide-array CGH detected no appreciable imbalance events in any of the 14 negative controls, with the exception of several reported CNVs. The dye-swap scheme essentially eliminated false-positive results.
To further evaluate the confidence of each imbalance call by oligonucleotide-array CGH analysis, we calculated the mean log2 ratios for each detected imbalance region and compared them with the SD for the whole dataset. The value of the mean log2 ratio/SD indicates the separability of each imbalance event from the background noise of the whole dataset. For the majority of deletion events detected, the value was less than –3.2, whereas the value was >2.6 for the majority of gain events detected.
genomic imbalance detected in clinical samples
We used focused oligonucleotide-array CGH to test 211 clinical samples that had been ascertained to have come from individuals with developmental delay, unexplained mental retardation, dysmorphic features, or multiple congenital anomalies. In this cohort, the detection rate for genomic imbalance was approximately 11.9% (25 of 211 samples). All abnormal findings were first verified with a dye-swap array CGH and then independently confirmed by either FISH or MLPA. All the genomic-imbalance events, including CNVs with unknown significance, were divided into 3 categories, as is described below. The imbalance events associated with known disorders or likely to cause disease are listed in Table 1
.
|
genomic imbalance associated with well-defined disorders
Table 1A
lists 12 samples with 10 genomic imbalance events associated with known genetic disorders. In this group, 2 of the samples revealed a complex pattern involving both gain and loss on 2 different chromosomes (case 1) or on the same chromosome (case 2), 2 samples had a well-defined microdeletion syndrome (cases 3 and 4), 1 sample had a subtelomeric deletion (case 5), and 2 samples had whole-chromosome aneuploidy (cases 6 and 7). One dominantly inherited disorder could be diagnosed by the detection of haploinsufficiency for the relevant gene (cases 8 and 9), and 3 cases involved carriers of a recessive allele: deletion at the NPHP1 [nephronophthisis 1 (juvenile); cases 10 and 11] and CTNS loci (case 12). The last 3 cases featured deletions of <100 kb, each of which covered defined disease genes.
Oligonucleotide-array CGH was able to detect cryptic rearrangements, submicroscopic alterations, and even single-gene deletions. The smallest imbalance event detected in this study was a heterozygous genomic deletion of 3 consecutive probes covering a minimal 23-kb interval. Fig. 2A
shows the targeted array CGH data, and deletion of the 3 targets was confirmed by dye-swap hybridization (green in the forward hybridization and red in the reverse hybridization). In this case, we repeated the CGH analysis with Agilent’s 244K whole-genome oligonucleotide array and confirmed the deletion, which includes the entire CARKL (carbohydrate kinase-like) gene and part of CTNS (Fig. 2B
). The partial deletion of CTNS was independently confirmed by MLPA analysis. Fig. 2C
shows a 1-copy deletion for CTNS exons 2 and 6 and a typical dosage for CTNS exon 12. To further characterize the deletion, we amplified the deletion junction by the PCR and confirmed by sequencing (Fig. 2D
) that the deletion detected by array CGH is the common "European" deletion associated with cystinosis(17). This case further demonstrated the excellent resolution and sensitivity of the custom oligonucleotide-array CGH method.
|
genomic imbalance probably causing a disorder
Table 1B
lists 5 samples with interstitial deletions/duplications that may be clinically relevant. The list includes 3 interesting cases: a 3.3 Mb duplication at 17p11.2, which is similar to that of a recently reported 17p11.2-duplication syndrome(18), and 2 cases of a 546 kb de novo deletion at 16p11.2, which is within a region of frequently observed cytogenetic polymorphism but is not observed in the CNV database. Neither individual with the de novo 546-kb 16p11.2 deletion had specific dysmorphic features, but the absence of this deletion in the parents suggests that the deletion is responsible for the phenotype of developmental delay.
genomic CNVS with unknown significance
Nine patients had unreported CNVs with relatively small genomic deletions of between 50 and 200 kb (data not shown). The clinical significance of these imbalance events is unclear. Although this custom oligonucleotide array was designed to avoid CNVs, many new CNV loci have been reported since the design of the array (http://projects.tcag.ca/variation/).
|
Discussion |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Uniformly manufactured arrays facilitate reproducible results in the clinical laboratory. On BAC-based arrays, each BAC clone may include both unique and repetitive sequences. Each BAC must be propagated in culture, introducing possible variability in BAC inserts or DNA contamination between batches and potentially affecting the reproducibility and consistency of the manufactured arrays. In contrast, 60-mer oligonucleotide probes are synthesized robotically in situ and have a fixed GC content and melting temperature that facilitate uniform hybridization. Although all targets in Agilent’s eArray library were designed in silico with narrow melting temperature ranges, the actual performance of each target needed to be evaluated in practice. Some features exhibited large variability, possibly due to properties of these unique sequences. Genomic CNVs unquestionably contribute to the variability in the performance of some probes.
The sensitivity and specificity of the oligonucleotide-array platform were excellent because of the high SNRs and low SDs. We observed the SDs of all targets to be consistently <0.1—with the majority being <0.08—with this oligonucleotide-based array. Given the nice separability of imbalance events above the baseline, we can identify genomic imbalance events with a high level of confidence. In addition to the 100% concordance between the oligonucleotide-array CGH results and the results generated by other methods for both positive and negative samples, we were able to identify smaller imbalance events (unreported CNVs; data not shown) that were not detected with other methods.
Oligonucleotide platforms can quickly and easily accommodate changes in genomic coverage. Manufacturing costs are not prohibitive and the list of available probe sequences extends the length of the genome. Thus, updates to oligonucleotide-based arrays can be accomplished more quickly and with less postproduction validation than BAC-based arrays, which require new BAC clones to be individually validated and DNA to be prepared from each clone before chips can be manufactured.
This focused oligonucleotide-based array CGH platform detected all genomic imbalance events in the 65 validation samples, with 100% concordance with BAC-based array CGH, FISH/karyotyping, or MLPA. Coverage of clinically relevant loci (see Supplemental Data 1 in the online Data Supplement) is equivalent to other BAC-based targeted array CGH platforms(19)(20). The enhanced sensitivity and specificity of oligonucleotide-array CGH compared with other methods are attributable to better resolution and the custom design, respectively.
Despite the superior technical performance of oligonucleotide-based array CGH, clinical interpretation can be challenging. We encountered several scenarios: (a) known genomic-imbalance events with well-documented clinical relevance (Table 1A
) that were verified with dye-swapping and either FISH or MLPA (on smaller regions); (b) known genomic-imbalance events with possible clinical relevance (Table 1B
) that were verified by dye-swapping and the 244K whole-genome array for further characterization of the genomic-imbalance events; and (c) novel genomic imbalance with uncertain clinical relevance, cases of which were verified by dye-swapping, 244K whole-genome array analysis, and examination of parental samples to determine whether the imbalance was a de novo event.
Adding to the complexity of interpretation is that an apparent familial variant can actually be pathogenic. For example, a deletion variant that appears in a healthy parent and an affected child could be acting as a recessive allele in the parent. Likewise, a de novo region of genomic imbalance could be a previously undescribed nonpathogenic CNV or a case of nonpaternity. Dozens of CNVs are present in every human genome(21)(22).
Although this custom oligonucleotide array has an average resolution of approximately 35 kb within the targeted regions, the fact that array formats continue to evolve provides ample opportunities to increase resolution through the selection of additional probes from available oligonucleotide databases or the design of custom oligonucleotide arrays. The flexibility of the oligonucleotide-array design allows individual laboratories to interrogate different genomic regions at the most biologically relevant level of resolution. For example, arrays can be designed to detect single-exon deletions and duplications in a gene of interest.
Resolution beyond that provided by this custom oligonucleotide array may not be necessary or desirable in all cases, however. A majority of the CNVs detected on high-resolution oligonucleotide platforms are <150 kb(23)(24), and the detection of these CNVs complicates the clinical interpretation. CNV databases are highly valuable research tools, but they have limitations in the clinical setting. CNVs in databases are often determined by a variety of methods that may not provide accurate breakpoints. These databases are often not annotated with clinical information and thus do not indicate genotype-phenotype associations.
As the knowledge of genotype-phenotype associations in disorders of genomic imbalance accumulates, however, precise and accurate determination of the genomic sequences involved is critically important in the clinical diagnostic setting. FISH and BAC clones can detect imbalance equal to or larger than the size of clone insert, typically approximately 150 kb; thus, smaller deletions or duplications can be missed on a BAC-based array. Even genome-wide tiling BAC arrays do not have a resolution of less than approximately 46 kb(25)(26).
Oligonucleotide-based arrays provide an unprecedented degree of resolution for determining both the size and breakpoints of regions of genomic imbalance. Our current design was able to precisely define a heterozygous deletion of 23 kb, a level of resolution not possible on a BAC-based array. This improvement in resolution could be important in the clinical interpretation of cases in which the inclusion or exclusion of certain genes may predict clinical features of the condition.
In conclusion, the custom oligonucleotide array that we have described offers a sensitive and specific clinical assay for detecting genomic imbalance. The platform can easily accommodate new information to keep pace with research advances. This methodology offers a proof of principle that well-designed oligonucleotide-based arrays offer substantial technical advantages over currently available platforms and at a comparable cost. We anticipate widespread adoption of this methodology in the clinical diagnostic setting.
|
Acknowledgments |
|---|
Financial disclosures: None declared.
Acknowledgments: We thank the referring physicians and genetic counselors at the Children’s Hospital Boston for their cooperation, Dr. Arthur L. Beaudet of Baylor College of Medicine for thoughtful discussion of the study, and Dr. Arthur R. Brothman and Emily Aston of the University of Utah for FISH confirmation of clinical cases.
|
Footnotes |
|---|
1 Nonstandard abbreviations: FISH, fluorescence in situ hybridization; MLPA, multiplex ligation-dependent probe amplification; CGH, comparative genomic hybridization; BAC, bacterial artificial chromosome; CNV, copy-number variant; SNR, signal-to-noise ratio.
2 Human genes: CTNS, cystinosis, nephropathic; NPHP1, nephronophthisis 1 (juvenile); CARKL, carbohydrate kinase-like.
|
References |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
The following articles in journals at HighWire Press have cited this article:
|
|
 |
|
  O. Saydam, Y. Shen, T. Wurdinger, O. Senol, E. Boke, M. F. James, B. A. Tannous, A. O. Stemmer-Rachamimov, M. Yi, R. M. Stephens, et al. Downregulated MicroRNA-200a in Meningiomas Promotes Tumor Growth by Reducing E-Cadherin and Activating the Wnt/{beta}-Catenin Signaling Pathway Mol. Cell. Biol., November 1, 2009; 29(21): 5923 - 5940. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 Y. Shen and B.-L. Wu Microarray-Based Genomic DNA Profiling Technologies in Clinical Molecular Diagnostics Clin. Chem., April 1, 2009; 55(4): 659 - 669. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 D T Miller, Y Shen, L A Weiss, J Korn, I Anselm, C Bridgemohan, G F Cox, H Dickinson, J Gentile, D J Harris, et al. Microdeletion/duplication at 15q13.2q13.3 among individuals with features of autism and other neuropsychiatric disorders J. Med. Genet., April 1, 2009; 46(4): 242 - 248. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
L.-J. C. Wong, D. Dimmock, M. T. Geraghty, R. Quan, U. Lichter-Konecki, J. Wang, E. K. Brundage, F. Scaglia, and A. C. Chinault Utility of Oligonucleotide Array-Based Comparative Genomic Hybridization for Detection of Target Gene Deletions Clin. Chem., July 1, 2008; 54(7): 1141 - 1148. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
E Aston, H Whitby, T Maxwell, N Glaus, B Cowley, D Lowry, X L Zhu, B Issa, S T South, and A R Brothman Comparison of targeted and whole genome analysis of postnatal specimens using a commercially available array based comparative genomic hybridisation (aCGH) microarray platform J. Med. Genet., May 1, 2008; 45(5): 268 - 274. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
