Whole-Genome Array CGH Evaluation for Replacing Prenatal Karyotyping in Hong Kong

Objective To evaluate the effectiveness of whole-genome array comparative genomic hybridization (aCGH) in prenatal diagnosis in Hong Kong. Methods Array CGH was performed on 220 samples recruited prospectively as the first-tier test study. In addition 150 prenatal samples with abnormal fetal ultrasound findings found to have normal karyotypes were analyzed as a ‘further-test’ study using NimbleGen CGX-135K oligonucleotide arrays. Results Array CGH findings were concordant with conventional cytogenetic results with the exception of one case of triploidy. It was found in the first-tier test study that aCGH detected 20% (44/220) clinically significant copy number variants (CNV), of which 21 were common aneuploidies and 23 had other chromosomal imbalances. There were 3.2% (7/220) samples with CNVs detected by aCGH but not by conventional cytogenetics. In the ‘further-test’ study, the additional diagnostic yield of detecting chromosome imbalance was 6% (9/150). The overall detection for CNVs of unclear clinical significance was 2.7% (10/370) with 0.9% found to be de novo. Eleven loci of common CNVs were found in the local population. Conclusion Whole-genome aCGH offered a higher resolution diagnostic capacity than conventional karyotyping for prenatal diagnosis either as a first-tier test or as a ‘further-test’ for pregnancies with fetal ultrasound anomalies. We propose replacing conventional cytogenetics with aCGH for all pregnancies undergoing invasive diagnostic procedures after excluding common aneuploidies and triploidies by quantitative fluorescent PCR. Conventional cytogenetics can be reserved for visualization of clinically significant CNVs.


Introduction
Conventional cytogenetics has been the gold standard for detecting chromosomal abnormalities in prenatal diagnosis. It enables the examination of genome-wide numerical and structural abnormalities at microscopic level, and can achieve a resolution of 5-10 Mb [1]. However, the method is labour intensive, with a turn-around time of 14 to 21 days. Various molecular cytogenetic techniques, such as Quantitative Fluorescent Polymerase Chain Reaction (QF-PCR) [2,3] and Fluorescent In Situ Hybridization (FISH) technology, could complement the detection of chromosomal abnormalities and offer faster turn-around times. However, these methods are targeted to detect specific chromosomal abnormalities and are dependent on the chromosomal probe used. In contrast, whole-genome array comparative hybridization (aCGH) not only provides high resolution detection of genomic alterations, but also allows refinement of breakpoints on chromosome rearrangements.
Chromosomal microarray to assess DNA copy number variations has been suggested as the first-tier clinical diagnostic test in the postnatal setting for individuals with developmental disabilities or congenital anomalies because of an increased diagnostic yield of 12 to 15% compared to conventional karyotyping [4,5]. The clinical utility of aCGH in the prenatal setting has been demonstrated in systematic reviews [6][7][8], a large scale prospective randomized controlled trial [9] and also in recent research [10]. The major challenge for the large-scale implementation of these techniques appears to lie in interpretation of the results [11]. Thus genetic counseling and ethical issues [12,13] are significant in offering whole-genome aCGH in prenatal diagnosis. Essentially, there is a need for consensus in determining which patient groups should be offered in routine prenatal practice, and international guidance on interpretation and reporting of copy number variations for prenatal population [11]. This evaluation study demonstrates the advantages and disadvantages of aCGH and argues that whole-genome oligonucleotide aCGH is able to replace conventional cytogenetics in prenatal diagnosis in the local population of Hong Kong.
In this study, the use of aCGH for prenatal diagnosis is evaluated in two models: (1) as a first-tier test and, (2) as a 'furthertest' analyzing prenatal samples of patients with abnormal fetal ultrasound findings and normal karyotypes. Results from aCGH are compared with those from conventional cytogenetics in order to determine the concordance of results and the additional diagnostic yield of aCGH over karyotyping.

Patients and samples
Approval was granted by the Institutional Review Board, University of Hong Kong/Hospital Authority, Hong Kong for the study to be conducted within 3 hospitals: Tsan Yuk Hospital, Queen Elizabeth Hospital, and Kwong Wah Hospital. Written informed consent was obtained from participants, who were recruited between January 2011 and November 2012. Informed consent and counseling on the benefits and limitations of the test, test methodology, reporting time, and possible test results (clinically significant, unclear clinical significance, benign) and outcomes of the investigation were explained to the participants by medical staff. Parental blood samples were obtained at the time of consent in case information on inheritance of CNV is necessary for further interpretation of prenatal results. Prenatal samples of 370 patients ( Figure 1) with indications for chromosome studies were tested using whole-genome aCGH analysis at the Prenatal Diagnostic Laboratory, Tsan Yuk Hospital. Of those 370 prenatal samples, 220 patients had been prospectively recruited for the firsttier test study prior to the invasive procedure. In addition, 150 subjects with abnormal fetal ultrasound findings were retrospectively recruited into the 'further-test' study once conventional cytogenetic results had shown a normal karyotype on invasive testing. A total of 193 parental blood samples were used to categorize unclear CNVs identified in the corresponding prenatal samples in the evaluation study. In addition to the evaluation study, abnormal findings of 12 prenatal samples tested by conventional cytogenetics and requiring characterization were assessed using aCGH.
Conventional cytogenetics was performed by Giemsa banded (G-banded) karyotyping as a clinical service on all prenatal samples at Prenatal Diagnostic Laboratory, Tsan Yuk Hospital. As per protocol, optimally, 3-5 mg of dissected chorionic villi or 30 ml amniotic fluid was obtained to set up for karyotype, QF-PCR and aCGH. Cultured cells in flasks were used when either a) there were not enough cells in the primary sample, or b) in retrospective samples for the 'further-test' study, or c) in retrospective samples required for characterization studies. Eleven placental tissue samples and 2 skin biopsy samples obtained from pregnancies which were terminated after abnormal fetal ultrasound findings were processed for aCGH analysis.
Indications for recruitment to the study Prenatal patients with clinical indications for further diagnoses were recruited to the research study. The reasons included abnormal findings on fetal ultrasound; positive Down syndrome screening; or maternal anxiety concerning advanced maternal age, family history of genetic disorder or previous child with anomalies. Some patients met more than one of the indications for study. In categorizing the indications as shown in Figure 1, ultrasound abnormality preceded positive Down syndrome screening and in turn over maternal anxiety.

Methods of DNA extraction
Cells were pelleted from 5 ml amniotic fluid by centrifugation at 4006g for 10 min. DNA was extracted by Gentra Puregene Tissue Kit (QIAGEN, USA) following manufacturer's instruction. Uncultured chorionic villi, tissue or cultured cells were pelleted by centrifugation at 30006g for 5 min, lysed in 300 ml of lysis buffer (100 mM Tris, pH 8.5, 5 mM EDTA, 0.2% SDS, 200 mM NaCl) with Proteinase K at final concentration of 2 mg/ml. The lysate was incubated at 55uC overnight, added with 7 ml of RNase A solution (Qiagen, USA) and incubated at 37uC for 60 to 120 min. DNA was precipitated by the addition of 2.5 volume of cold 100% ethanol, spooled, washed twice with 1 ml of 70% ethanol and air dried. The DNA pellet was dissolved in Tris EDTA (TE) buffer (10 mM Tris, 0.1 mM EDTA, pH 7.5). For the blood samples, 3 ml of EDTA blood were diluted to 9 ml with 1X PBS. The diluted blood was overlaid onto 6 ml of Ficoll-Paque Plus (GE Healthcare Life Sciences) and centrifuged at 6006g for 30 min. The mononuclear cells at the interphase were transferred to a fresh tube and washed twice with 15 ml of 1X PBS. Cells were pelleted by centrifugation at 4006g for 10 min. The cells were lysed in 1 ml of lysis buffer (100 mM Tris, pH 8.5, 5 mM EDTA, 0.2% SDS, 200 mM NaCl) and DNA was precipitated and  processed as above. Where DNA extraction was deemed urgent, for example in advanced gestations, a commercial kit (QIAmp DNA blood kit, Qiagen, USA) was used. The concentration of DNA samples was measured by NanoDropND-1000 spectrophotometer (NanoDrop Technologies, USA) and the quality of DNA samples was analyzed by agarose gel electrophoresis to exclude degradation or RNA contamination.

Array CGH analysis and interpretation
All samples were tested by NimbleGen CGX-135K arrays which were designed by Signature Genomics (Perkin Elmer, USA) following manufacturer's instructions. The coverage of the array has an average resolution of 140 kb across the genome and 40 kb or less in regions of clinical relevance. It evaluates over 245 known genetic syndromes and over 980 gene regions of functional significance in human development. The data were analyzed by Genoglyphix software (Signature Genomics, Spokane, USA). The gender of the prenatal samples was examined using QF-PCR [14] to determine the gender-matched reference DNA used in aCGH. The control DNA was from pooled gender-matched DNA from Promega (Madison, Wisconsin, USA).
Copy number variants (CNVs) detected by aCGH were systematically evaluated for clinical significance by comparison with information in the Signature Genomics' proprietary Genoglyphix Chromosome Aberration Database (Signature Genomics, Spokane, WA, USA), the internal laboratory database at Tsan Yuk Hospital, and the publicly available databases [Database of Genomic Variant (DGV), International Standards for Cytogenomic Arrays Consortium Database (ISCA), Children Hospital of Philadelphia database (CHOP), Database of Chromosomal Imbalance and Phenotype in Humans using Ensembl Resources (DECIPHER)]. Categorization of CNVs was based on available information concerning the fetal phenotypes and by comparison of phenotypes with known genes in the region of copy gain or loss. This was ascertained from searching Online Mendelian Inheritance in Man (OMIM), PubMed, RefSeq and the University of California Santa Cruz (UCSC) genome browser. A CNV was considered to be: (1) benign if it was reported in healthy subjects in the databases searched; if there are no genes involved; or if involved genes were unrelated to the phenotype and have no apparent clinical relevance; (2) clinically significant if it corresponded to a region known to be of clinical relevance or had a gene of clinical relevance; (3) of unclear clinical significance if there is insufficient evidence to categorize as clinically significant or benign at the time of reporting. When CNVs of unclear clinical significance were detected in a prenatal sample, parental blood samples were processed to provide additional information for interpretation.

Confirmation of CNVs
Clinically significant copy number gains and losses not detectable by karyotyping were confirmed by Fluorescent In Situ Hybridization (FISH) studies whenever possible. Microdeletion and telomeric FISH probes were obtained from Abbott Diagnostics (USA) and FISH probes from bacterial artificial chromosome clones from The Centre for Applied Genomics at the Hospital for Sick Children (Toronto, Canada). Homozygous alpha thalassemia deletion was confirmed by standard laboratory protocol [15].

Results
Clinically significant CNVs identified by aCGH in the prenatal samples from both the first-tier test study and from the 'furthertest' study are shown in Figure 1. The initial turn-around time for aCGH was 8-10 days, while conventional cytogenetics took 14-21 days (data not shown).

First-tier test study
Two hundred and twenty prenatal samples were examined using whole-genome aCGH methodology together with routine culture for G-banded karyotyping with or without QF-PCR rapid aneuploidy testing. One hundred and sixteen samples (52.7%) were positive for Down syndrome screening with no apparent fetal ultrasound anomalies at the time of testing; whilst 77 prenatal samples (35.0%) had fetal ultrasound anomalies detected prior to invasive prenatal diagnosis. Forty of these fetal malformations concerned a single organ system and 37 had abnormalities involving more than one organ system. In 27 prenatal samples (12.3%), maternal anxiety was the clinical indication for invasive testing and inclusion into the study ( Figure 1).
The clinically significant results were analyzed and categorized according to the clinical indications for the investigations. There were six cases of patients who had screened positive for Down syndrome without fetal ultrasound anomalies being evident (Table 1). Three cases had common aneuploidies (one each of trisomy 13, 18, 21) and the remaining three had other chromosomal imbalances including ring structures and trisomy 7 ( Table 3, Case no. 14, 15, and Table 4 Case no. 21). In 77 prospective samples with fetal ultrasound anomalies detected (Table 1), 38 (49.4%) showed clinically significant chromosomal imbalances. Of these, 18 were common aneuploidies (three trisomy 13, six trisomy 18, five trisomy 21 and four monosomy X). In addition, 20 had other chromosomal imbalances ( Table 2, Case no. 1-7, Table 3, Case no. 8-13, 16, Table 4, Case no. 17-20, 22 and 23). Eleven out of the 20 cases had fetal ultrasound anomalies in more than one organ system (including the neck and body fluid, central nervous system, cardiovascular, craniofacial, gastrointestinal, genitourinary, musculoskeletal, thoracic, or other anomalies including intrauterine growth restriction). Where invasive testing was indicated for maternal anxiety, no chromosomal imbalance was detected in the 27 samples studied.

Further-test study after normal karyotyping
In 150 patients with normal karyotyping and abnormal fetal ultrasound findings, nine clinically significant CNVs were identified using aCGH. All of the patients had fetal ultrasound detected anomalies in more than one organ system (Table 5). These included one 1p32 microdeletion resulting in NFIA haploinsufficiency [16]; one 22q11.2 microdeletion, one deletion in chromosome 14 resulting in paternal uniparental disomy 14-like phenotype; one 8p23.1 microdeletion; one unbalanced translocation detected after an apparent normal karyotype and four samples with microdeletion in 16p13.3 resulting in Hemoglobin Bart's disease. With the exception of the unbalanced translocation which was undetected on chorionic villus karyotyping, but identified by aCGH of the amniotic fluid and cytogenetic study, the remaining CNVs detected were ,5 Mb in size and submicroscopic, beyond the detection resolution of conventional karyotyping. The additional diagnostic yield of aCGH over conventional cytogenetics was found to be 6%.

CNV of unclear clinical significance
The detection rate for CNVs of unclear significance identified during the evaluation study using the 135 K whole-genome array was 2.7% (10/370), with 1.4% (3/220) detected for the first-tier test and 4.7% (7/150) for 'further-test' ( Table 6). Six of these 10 samples had CNVs associated with microdeletions or microduplications with incomplete penetrance and variable expressivity ( Table 6, Case no. 1, 2, 4-7). Eight samples involved genes or loci which may be associated with neurodevelopmental problems ( Table 6, Case no. 1-8). Two samples involved genes associated with structural abnormalities of organ systems that may or may not have relevance to the phenotype ( Table 6, Case no. 7, 8). One de novo microdeletion of 1q21.1 with susceptibility for thrombocytopenia-absent radius (TAR) was classified as a CNV of unclear significance having determined that the fetal radii were present on the follow-up ultrasound scan (Table 6, Case no. 7). Where CNVs of unclear clinical significance were identified, the incidence of de novo CNVs was found to be 0.9% of the total in the study (3/370).

Characterization by aCGH on samples with abnormal karyotypes
During the study period, in addition to the 370 samples for the evaluation study, twelve prenatal samples were identified with abnormal karyotypes during conventional cytogenetics. These were further examined by aCGH. Breakpoints were established for two samples with unbalanced translocations and three samples with interstitial deletions. One sample used aCGH to confirm suspected deletion of chromosome 16 pter which had been undetermined by G-banding. This sample was found to have a balanced translocation between chromosome 2 and chromosome 10 which was inherited from the father. It was reassuring that no submicroscopic changes were evident. Three samples showed complex rearrangements, one involved a marker chromosome derived from chromosome X, one involved a marker chromosome derived from chromosome 15 which had the same array findings as Case no. 8 in Table 3. One complex chromosomal rearrangement involved translocation of a 1.76 Mb segment from chromosome 19 to a duplicated segment of chromosome 2. These three complex rearrangements would not have been accurately detected by conventional karyotyping.

Common benign CNVs
A total of 563 samples (370 prenatal and 193 parental) were performed using aCGH for the first-tier test, 'further-test' studies and for abnormal chromosomal characterization. These studies established common benign CNVs at eleven loci in patients of Hong Kong ( Table 7). The most common locus with chromosomal gain or loss found in Hong Kong was at 8p11.23, and was detected at a frequency of up to 78% (Table 7, no. 5) reflecting the homogeneity of the local population. The proportion of parental samples required to be performed was 144 out of 193 (74.6%) in the first half period of evaluation and 49 out of 193 (25.4%) in the second half. The knowledge of common benign CNVs in the local population reduced the need for parental sample testing and could minimize cost if aCGH use in prenatal diagnosis is launched as clinical service.

Increased nuchal translucency (NT)
Among all 382 prenatal samples, 27 of them had NT of 3.5 mm or above. In seven of these 27 prenatal samples Down syndrome screening results were not available. The remainders were screened positive for Down syndrome. There were three findings of trisomy 21, in addition to three samples with clinically significant CNVs detectable by karyotyping. There were three samples with CNVs of unclear clinical significance ( Table 6, Case no. 3, 4, 6), one of which had structural abnormality on fetal ultrasound scan. Therefore aCGH did not yield increased detection of clinically significant CNVs compared to karyotyping in this specific number of samples with increased nuchal translucency.

Discussion
This evaluation study demonstrated the effectiveness of the whole-genome oligonucleotide aCGH in prenatal diagnosis for the  Table 3. Clinically significant CNVs other than common aneuploidies detected in the first-tier test study with additional information provided by aCGH over karyotyping.

Case
No.
Gest (w)  [20] aCGH Evaluation for Replacing Prenatal Karyotype PLOS ONE | www.plosone.org analysis of chromosome imbalance at high resolution. During the first-tier test, a detection rate of 20% was determined amongst patients with clinical indications for testing. Clinically significant imbalances were found to be common aneuploidies (9.5%), whilst 10.5% involved other chromosomal abnormalities (Table 1). Rapid aneuploidy testing such as QF-PCR is able to exclude common aneuploidies prior to testing using aCGH. This would make aCGH more cost effective. Scott et al. also proposed the use of combined QF-PCR and aCGH as first-line prenatal diagnostic testing [17]. Our detection rate of 10.5% clinically significant CNVs other than common aneuploidy was double that of Scott et al. (2013). It is considered that this may be related to different patient characteristics, different indications for prenatal testing, in addition to using whole-genome rather than targeted array. CNVs only detectable by aCGH were also higher in our evaluation study (3.2% in this study versus 1.2% pathogenic CNVs) [17]. Array CGH helped to precisely delineate breakpoints, characterize marker chromosomes and detect mosaicism within a shorter time frame compared to G-banded cytogenetics. There were five samples with chromosome mosaicism in the study. Mosaic trisomy 19 and mosaic trisomy 7 ( Table 2, Case no. 2 and Table 4 Case no. 21) were results confined to placental mosaicism where studies on skin fibroblast and in the amniotic fluid sample respectively indicated normal karyotypes. Of interest was a mosaic ring chromosome 18 finding with aCGH of one copy number gain for the chromosome 18 segment (Table 3, Case no. 13), not matching the mosaic ratio, but the FISH study found a small number of double rings in the cell population which explained the array result.
Four complex chromosomal rearrangements were determined in the first-tier test study, and three complex rearrangements were identified from prenatal samples with abnormal karyotypes requiring characterization. This highlighted the advantage of higher resolution aCGH in chromosome analysis. There were two samples with identical complex rearrangement comprising 2-copy gain at the Prader Willi/Angelman syndrome region (15q11.2-q13.2) and one copy gain at the 15q13.3 Microdeletion region (15q13.2-q13.3, Figure 2). Karyotyping of the first sample (Figure 2 Sample A; Table 3, Case no. 8) showed 15q+ while the chromosomal gain in the second sample (Figure 2 Sample B) performed using aCGH for characterization was in a marker chromosome. Although aCGH showed the same chromosomal imbalances, the phenotypes of the 2 samples were very different. In the latter sample choroid plexus cyst was the only anomaly. Since other CNVs detected in Case no. 8 (Table 3) were benign and apparently common in the local population, the severe phenotype may have been caused by small mutation in another chromosomal region or an interruption of the region 15q11.2-q13.2 by a complex rearrangement. The 15q proximal region is known for its instability due to the presence of repeating DNA elements [18,19] which may give rise to the triplication from centromere to breakpoint (BP) 4 plus duplication from BP4 to BP5 in these samples. Further investigation into the association of phenotypes and the manifestation of the copy number gain and imprinting [20] can be made on similar samples with rearrangements in the region. The finding also demonstrated the advances of complementary use of both aCGH and karyotyping in analyzing samples with multiple significant CNVs.
There were 4 cases of Hemoglobin Bart's disease with clinically significant aCGH findings in the 'further-test' study. While routine prenatal screening for thalassaemia by mean corpuscular volume (MCV) is offered in our locality, one case presented late in the third trimester (Table 5, Case no. 9) and was a result of non-paternity. This Table 3. Cont.

Case
No. highlighted one potential use of the test to diagnose Hemoglobin Bart's disease in circumstances of non-paternity [21]. Uncertain CNVs tend to be a concern for clinicians in counseling. Adequate pre and post-test information and counseling by trained counselors, and a team approach involving obstetricians, clinical geneticists and laboratory scientists in CNV interpretation is beneficial and is adopted in our setting. In this evaluation study, both CNVs of clinical significance and CNVs of unclear significance were reported to referring doctors. In the 10 cases with CNVs of unclear clinical significance, 5 of the pregnancies, all with major ultrasound abnormalities, were terminated (Table 6). Maternal cell contamination was not determined during the evaluation study. However the impact of contamination on the interpretation of prenatal microarray has been reported [22]. Taking this into account, our laboratory has adopted the policy to routinely exclude maternal cell contamination by examining STR markers of 13, 18 and 21 in fetal and maternal samples. Various authorities already approved the offering of aCGH as an adjunct diagnostic tool in prenatal cases with fetal ultrasound abnormalities [23][24]. The additional diagnostic yield of 6% in our further-test study is consistent with 5.2 to 10% increased detection by aCGH in fetuses with ultrasound abnormalities and normal karyotype reported in the literature [6,7]. Data from a review study showed the overall detection rate of about 1% for significant submicroscopic aberrations in low risk pregnancies [10]. This consequently caused microarrays to be identified as a first-line diagnostic test in women choosing the test irrespective of clinical indication [9,10,25]. This is in line with the latest recommendations from the American College of Obstetricians and Gynecologists and the Society of Maternal-Fetal Medicine [26]. Based on these findings and reports in the literature [6][7][8][9][10]25] on the potential of aCGH to detect chromosomal abnormalities beyond G-banded karyotyping, it is proposed that whole-genome aCGH would be suitable to replace conventional cytogenetics in prenatal diagnosis in Hong Kong (Figure 3). In a prenatal diagnostic setting, women screened positive for fetal Down syndrome may be offered an option of having noninvasive prenatal testing for fetal trisomy assessment by maternal plasma DNA if there is no ultrasound abnormality. In cases where ultrasound examination shows fetal abnormalities which require additional information from aCGH, an invasive diagnostic test could be performed. Rapid aneuploidy testing of the sample with QF-PCR would exclude common aneuploidies, triploidy and maternal contamination before aCGH analysis. This rapid test would identify triploidy which cannot be identified by aCGH. Cultures would be set up for cytogenetic study and karyotyping for all clinically significant CNVs (20.0%, Table 1) detected by aCGH if the CNV is large (.10 Mb). In cases where the CNV is small (,10 Mb), metaphase and interphase FISH could be performed. Conventional cytogenetics performed for abnormal QF-PCR findings (9.5% of samples, Table 1) would be needed to assess whether there is a parental balanced translocation carrier state in order to determine recurrence risk. This approach would reduce conventional cytogenetic testing by around 80% of the cases in Hong Kong. The disadvantages would include the potential for non-detection of balanced translocation or an inversion carrier status, low level mosaicism or small heterochromatic marker chromosomes of the fetus. Whilst non-detection of balanced translocation or inversion carrier status would be unlikely to affect the health of the fetus, low level mosaicism detection is a limitation of all prenatal investigation techniques. Further cost-benefit analysis and review of the staffing requirements of a cytogenetic laboratory may help to define the value of using microarrays for our prenatal diagnostic service provision.
NimbleGen has, however, phased out production of oligonucleotide 135 K arrays in favor of transitioning to Agilent oligonucleotide 8660 K array, which has a lower backbone resolution. It is therefore anticipated that fewer CNVs of unclear significance will be detected for prenatal samples, with a shorter hybridization and faster turn-around time expected. Further studies will be required to confirm these effects.

Conclusions
This evaluation study showed that the whole-genome 135 K aCGH platform increased the diagnostic yield of 3.2% using aCGH over conventional cytogenetics in the first-tier test study, and by 6.0% in the 'further-test' study for the Hong Kong population. It also offered a higher resolution karyotyping for prenatal diagnosis in both study models and results are comparable to recent published studies. It is proposed that aCGH should replace karyotyping for use in prenatal testing where invasive procedures are required, after excluding common aneuploidies and triploidies by quantitative fluorescent PCR. Conventional cytogenetics can be reserved for visualization of clinically significant CNVs.