Microarray comparative genomic hybridization in prenatal diagnosis: a review

Authors

  • S. C. Hillman,

    1. School of Clinical and Experimental Medicine, College of Medicine and Dentistry, University of Birmingham, Edgbaston, Birmingham, UK
    Search for more papers by this author
  • D. J. Mcmullan,

    1. West Midlands Regional Genetics Laboratories and Department of Clinical Genetics, Birmingham Women's Foundation Trust, Edgbaston, Birmingham, UK
    Search for more papers by this author
  • D. Williams,

    1. West Midlands Regional Genetics Laboratories and Department of Clinical Genetics, Birmingham Women's Foundation Trust, Edgbaston, Birmingham, UK
    Search for more papers by this author
  • E. R. Maher,

    1. School of Clinical and Experimental Medicine, College of Medicine and Dentistry, University of Birmingham, Edgbaston, Birmingham, UK
    2. West Midlands Regional Genetics Laboratories and Department of Clinical Genetics, Birmingham Women's Foundation Trust, Edgbaston, Birmingham, UK
    Search for more papers by this author
  • M. D. Kilby

    Corresponding author
    1. School of Clinical and Experimental Medicine, College of Medicine and Dentistry, University of Birmingham, Edgbaston, Birmingham, UK
    2. Fetal Medicine Centre, Birmingham Women's Foundation Trust, Edgbaston, Birmingham, UK
    • School of Clinical and Experimental Medicine, College of Medical and Dental Sciences, University of Birmingham, Birmingham B15 2TT, UK
    Search for more papers by this author

Abstract

G-band chromosomal karyotyping of fetal cells obtained by invasive prenatal testing has been used since the 1960s to identify structural chromosomal anomalies. Prenatal testing is usually performed in response to parental request, increased risk of fetal chromosomal abnormality associated with advanced maternal age, a high-risk screening test and/or the presence of a congenital malformation identified by ultrasonography. The results of karyotyping may inform the long-term prognosis (e.g. aneuploidy being associated with a poor outcome or microscopic chromosomal anomalies predicting global neurodevelopmental morbidity). Relatively recent advances in microarray technology are now enabling high-resolution genome-wide evaluation for DNA copy number abnormalities (e.g. deletions or duplications). While such technological advances promise increased sensitivity and specificity they can also pose difficult challenges of interpretation and clinical management. This review aims to give interested clinicians without an extensive prior knowledge of microarray technology, an overview of its use in prenatal diagnosis, the literature to date, advantages, potential pitfalls and experience from our own tertiary center. Copyright © 2012 ISUOG. Published by John Wiley & Sons, Ltd.

Prenatal Genetic Testing

Cytogenetic diagnosis using cultured cells obtained by prenatal invasive tests has been the mainstay of prenatal testing since 19661. Microscopic visualization of chromosome banding (karyotyping) in the 1970s saw this testing become routine with potential structural chromosomal abnormalities being detected down to a resolution of 5–10 Mb. In the 1990s quantitative fluorescence polymerase chain reaction (QF-PCR) was used to target specific whole-chromosome anomalies (typically involving chromosomes 13, 18, 21, X and Y), a process that may be automated, allowing a relatively cheap and rapid diagnosis in response to a ‘high-risk result’ from prenatal screening tests2.

Fluorescence in-situ hybridization (FISH) is used as an adjunct to full, conventional karyotyping and may identify submicroscopic anomalies. This technique may be used to identify interstitial microdeletions and microduplications (e.g. the ∼ 2 Mb deletion associated with DiGeorge syndrome3) or subtelomeric deletions and duplications4. However both FISH and QF-PCR are ‘targeted’ approaches capable only of assessing a limited number of loci and, in the case of FISH, only with a high degree of suspicion of a specific anomaly being present.

Microarray comparative genomic hybridization (aCGH) involves hybridization of a patient's DNA onto predetermined targets representative of the whole genome (bacterial artificial chromosomes (BAC) clones of 100–200 kb in length or synthetic oligonucleotide probes of typically 25–75 bp in length) spotted onto glass slides. The patient or ‘test’ DNA (in a prenatal context obtained by amniocentesis, chorionic villus sampling (CVS) or fetal blood sampling) is extracted from the relevant sample, labeled with a fluorochrome, mixed with a reference DNA pool (labeled with a different fluorochrome) and then hybridized on the microarray slide (Figure 1)5.

Figure 1.

Diagram showing the basic principle of array comparative genomic hybridization. Differentially labeled test and reference DNA are cohybridized to the microarray and spotted with genomic clone or oligonucleotide probes in the presence of cot-1 DNA, which suppresses repetitive sequences. After hybridization and laser scanning, fluorescent ratios on each array spot are calculated and normalized so that the median log 2 ratio is 0. User-defined thresholds are set for calling copy number changes as loss or gain.

aCGH in its earlier form used BAC-based arrays. These are often ‘targeted’ to areas within the genome known to be disease specific that cover genomic regions associated with well defined microdeletion and duplication syndromes. Oligonucleotide aCGH is now more commonly used, especially in the postnatal setting. It has advantages over the BAC array technique including more flexibility in terms of probe selection, thus facilitating high probe density and customization of content. Single nucleotide polymorphism-based arrays have the advantage of being able to detect long continuous stretches of homozygosity, representing whole chromosome uniparental isodisomy (UPD), in which a fetus may inherit a duplicate of one chromosome from a parent and no chromosome from the other parent, for example as seen in rare cases of Angelman's syndrome, segmental UPD or consanguinity6. Development and application of these microarrays led to the emergence of multiple whole genome copy number variation (CNV) studies in normal populations. CNVs are segments of DNA of 1 kb or larger present at a variable copy number in comparison with a reference genome7. Redon et al.8 determined that approximately 12% of the human genome exhibits CNV. With such widespread CNV density, interpretation in a clinical setting can be challenging in terms of classifying variants as pathogenic, benign or novel variants of unknown significance (VOUS). The majority (> 99%) of benign CNVs are inherited and the vast majority of these are less than 500 kb9.

Microarray technology has several advantages over conventional, full G-band karyotyping, including improved resolution and potentially higher detection rates of chromosomal variation. In postnatal patients (children and adults) with a diagnosis of unexplained neurodevelopmental disability, the positive diagnostic yield of aCGH has been reported to be ∼ 10% higher than that of standard karyotyping6. When aCGH is applied in the setting of prenatal diagnosis, fetal cells obtained by invasive prenatal procedures do not have to be cultured and therefore results may be reported more quickly. aCGH testing is also potentially amenable to automation and high-throughput analysis.

Microarray technology does, however, have potential limitations. aCGH is only able to potentially detect ‘unbalanced chromosomal changes’. De-novo (non-inherited) balanced chromosomal rearrangements (such as reciprocal translocations or insertions) may disrupt genes and lead to phenotypic disease without detectable gains or losses at breakpoints10. However, in practice, many apparently balanced rearrangements detected by G-banding are not truly balanced at a DNA level and microarray testing can be used to detect small regions of DNA loss or gain and so clarify the exact nature of the rearrangement11.

The most significant current potential disadvantage of aCGH is the identification of novel, previously unreported, VOUS and the difficulties this may cause for clinical management, particularly in the prenatal setting. In order to facilitate the more accurate assessment and interpretation of VOUS there has been a major effort to catalog and collate genomic and clinical information. Thus the development of CNV databases, such as the International Standards for Cytogenomic Array (ISCA) Consortium from within the Database of Genotype and Phenotype (dbGaP) at the National Center for Biotechnology Information and the Database of Genomic Variants (DGV) will facilitate the resolution of VOUS into benign or pathogenic variants. Postnatally detected pathogenic cases are also recorded in the Database of Chromosomal Imbalance and Phenotype in Humans using Ensembl Resources.

Selecting a microarray that is able to detect the majority of pathological CNVs (sensitivity) but without detecting too many VOUS (specificity) is key. The detection of VOUS may increase the emotional burden on the parents and if prenatal diagnostic tests have been utilized in the presence of a suspected congenital malformation there is the possibility that such information, however ‘non-informative’ may increase the possibility of termination of pregnancy. In addition, the identification of CNVs of uncertain significance that then require the analysis of parental DNA for further interpretation has financial consequences and delays the availability of definitive results.

The sensitivity of aCGH is determined by the number and density of the probes and their resolution. In the clinical setting, in order that aCGH has improved detection over G-band karyotyping, it must accurately detect imbalances smaller than 5 Mb (the smallest resolution visible by light microscopy with G-band preparations). Array platforms in common clinical use have a typical resolution of 10 kb in targeted disease-specific regions of the genome and 200 kb in the genome backbone. Although most known pathogenic CNVs are 400 kb or larger, as array resolution increases smaller recurrent imbalances will be detected.

We published a contemporary critical appraisal of the literature evaluating the use of aCGH in prenatal diagnosis up to and including 200912. When aCGH was used prenatally for any indication (e.g. maternal age, parental anxiety or an identified ultrasound abnormality), pathogenic CNVs or VOUS were detected in 3.6% (95% CI, 1.5–8.5%) of cases in which conventional karyotyping was ‘normal’. When the indication for prenatal aCGH was a fetal anomaly on ultrasound, microarrays detected an additional 5.2% (95% CI, 1.9–13.9%; Figure 2) pathological CNVs or VOUS over conventional karyotyping. Heterogeneity (and hence large confidence intervals) was attributed to the varying resolution of the aCGH methodology. In addition, there was considerable variation in the literature as to whether an attempt had been made to identify and investigate the presence of benign CNVs (by reviewing parental samples).

Figure 2.

Forest plot showing meta-analysis of the detection rate of chromosomal imbalances by array comparative genomic hybridization when karyotyping is normal and chromosomal testing is performed for either any clinical indication or an abnormal ultrasound scan. Analysis is for array results showing chromosomal differences that are copy number variations which are pathogenic or of unknown significance. Reproduced from Hillman et al.12. TG, targeted genome; WG, whole genome.

Other cohort studies, published from 2009 onwards, have demonstrated an increased detection rate over G-band karyotyping ranging from 0.9 to 26.5%13–20. Some of the studies with much larger detection rates may represent selection of patients rather than being a reflection of a true prospective series. The results from the larger cohorts are therefore discussed. Fiorentino et al.19 processed 1030 prenatal samples (selected for many clinical indications including advanced maternal age, abnormal fetal ultrasound and known abnormal karyotype) both by standard G-band karyotyping and 1-Mb resolution cytochip focus constitutional BAC microarray (BlueGnome®, Fulbourn, Cambridge, UK). They detected 34 chromosomal anomalies in total (3.3% of the total cohort). Of these, nine cases would not have been detected by standard G-band karyotyping (0.9% of the cohort). Two of the nine cases had CNVs > 10 Mb and would usually be detected by good-quality cytogenetic testing. Of the remaining seven, three had CNVs resulting in findings of diseases (Charcot–Marie–Tooth disease, hereditary neuropathy with liability to pressure palsies and Duchenne muscular dystrophy) that may have been detected by chance due to the large sample size therefore representing the background prevalence within the population. Four samples had microdeletion/microduplication syndromes including two cases of DiGeorge syndrome, Sotos syndrome and 15q13.3 microdeletion syndrome. Of the nine CNVs detected, five were in fetuses with abnormal scan findings (n = 66 for abnormal ultrasound findings). The rate of potentially pathological CNVs detected when a karyotype was normal and a fetus had an abnormal scan increased from 0.9 to 7.6%.

In February 2012 at the 32nd annual meeting of the Society for Maternal–Fetal Medicine in Dallas, Dr Ron Wapner presented data on the ‘Multicenter, prospective, masked comparison of chromosomal microarray with standard karyotyping for routine and high-risk prenatal diagnosis’ on behalf of the prenatal microarray study group, National Institute of Child Health and Human Development, Bethesda, USA21. 4401 women were enrolled with indications of advanced maternal age (46%), abnormal first- or second-trimester screening (18%), abnormal ultrasound findings (26%) and other indications (9%). A customized 44-kb oligoarray was used and all CNVs detected by microarray were reviewed by an independent genetic panel. 4340 samples were available for comparison between standard karyotyping and microarray testing. 316 (7.3%) autosomal and 57 (1.3%) sex chromosome aneuploidies were identified by both karyotyping and microarray (seven were reported as mosaic by the microarray). Overall the study found a 2.5% chance of either a known pathological CNV or VOUS with potential clinical significance when a normal karyotype was found. For those cases with abnormal ultrasound and normal karyotyping the rate was higher, at 5.8%. For those referred for maternal age or positive screening and normal karyotype the rate was reduced to 1.7%. The group concluded that the microarray technique detects additional clinically relevant information in cases with structural anomalies and in those sampled for routine indications.

Two further papers have been published in 2012. Lee et al.22 looked at the utility of both a 1-Mb BAC and 60-K oligonucleotide array in 3171 pregnancies. Although the added utility of aCGH over karyotyping was only small when there was an uneventful prenatal examination (0.52%), the proportion of cases in which additional information was provided by aCGH increased to 8.2% when a fetus had an abnormality on ultrasound scan22. Interestingly, Srebniak et al.23 used a Human CytoSNP12 array (Illumina®, Illumina Inc., San Diego, CA, USA) with a resolution of 150/200 kb to analyze DNA from 207 cases, all with fetal anomalies, and found that it detected extra information in a similar percentage to that in the Lee et al. cohort (7.7%)23.

The recently published literature seems to conclude that although aCGH technology has only a small advantage over karyotyping when employed for the prenatal population as a whole, when performed for cases of abnormal ultrasound findings the advantage is more marked, finding potentially pathological CNVs in 5 to 8.5% of cases. This is higher than our original meta-analysis estimate of 5.2%, probably because the resolution of arrays used has increased between 2009 and 2012 and larger cohorts are providing a more accurate estimate. The literature indicates that the optimal use of prenatal aCGH is determined firstly by the design and resolution of the aCGH probes and secondly by parental aCGH when a potential pathological CNV is discovered.

The variation in prenatal detection rates from international studies emphasizes the need for a large prenatal prospective cohort study in the UK. At the University of Birmingham/Birmingham Women's Foundation Trust and West Midlands Genetics Laboratory, we are recruiting to an ongoing prospective cohort of pregnant women with a suspected fetal malformation detected using ultrasound over a 3-year period. This work is funded by the children's health charity Sparks. As part of the discussion and counseling the parents are offered fetal karyotyping. Amniotic fluid, CVS or fetal blood obtained by invasive fetal sampling is utilized for DNA extraction and aCGH run in parallel to conventional full karyotyping (after QF-PCR). If a common autosomal trisomy (13, 18 or 21) or a sex chromosome aneuploidy is identified, then microarray analyses are not performed. In addition, parental samples are taken for DNA extraction in all cases. We perform a constitutional focused BAC array (BlueGnome) with a resolution of ∼ 100 kb in targeted regions and ∼ 2 Mb in the genomic backbone prenatally, and postnatally (either at delivery or after termination of pregnancy) at higher resolution using oligonucleotide arrays (ISCA design) with a resolution of ∼ 10–20 kb in targeted regions and ∼ 100–200 kb in the genomic backbone. We chose to use a lower-resolution array in the prenatal period to minimize the number of VOUS detected which require parental follow-up. Results from the targeted array are fed back to patients in a timely manner (often available up to 6 days before standard karyotype). This is an ongoing study and as of March 2012 the study has recruited over 310 pregnancies with identified structural anomalies. In addition, the Medical Research Council-funded Evaluation of Array Comparative Hybridisation study (under the Efficacy and Mechanism Evaluation scheme) aims to recruit 2000 patients with ultrasound anomalies in a multicenter cohort throughout the UK within the forthcoming few years and will use the ISCA arrays as described above prospectively in the prenatal period. As well as determining the range of CNVs (of unknown significance, pathological and benign) in the prenatal population it is hoped that these studies will add to the expanding databases linking phenotype with pathogenic CNV. In addition the type of sample tested (amniocyte, fetal blood or chorionic villus) will be studied to look at which consistently produces the most reliable results. It is hoped that these larger prospective cohorts will go some way to answering the questions regarding array detection rates and rates of VOUS. The studies will also answer questions regarding ‘rolling-out’ a prenatal array service within the National Health Service (NHS) in the UK and difficulties that may be encountered. To illustrate these potential difficulties we present two examples of difficult cases from our cohort.

The first case represents a VOUS. The patient underwent prenatal testing by amniocentesis for a fetus with a complex cardiac anomaly (truncus arteriosus). Standard G-band karyotyping was normal. Focused constitutional array (BlueGnome) noted a duplication on the X chromosome of a male fetus (4–5 BAC clone, 0.5–1.0 Mb duplication at Xp22.32 to Xp22.31; Figure 3) (RP11-60N3-> RP11-769N24)x2, (RP11-44F2)x2 mat). This was confirmed using a higher-resolution 2.7-M array (Affymetrix®; Affymetrix Microarray Solutions, Santa Clara, CA, USA). The duplication was maternally inherited (the mother is of normal intelligence) and no male family members had the same duplication. Owing to the possibility of X-inactivation the duplication may have had a phenotypic effect in a male baby where it had none in the mother. X-inactivation studies were inconclusive. The duplication involved the gene NLGN4 (a member of the neuroligin family). Because of the involvement of neuroligins in synapse function they have been suggested as candidate genes for autism and mental retardation24, 25. The parents were counseled and the uncertainty regarding the involved gene and its potential effect on the baby was conveyed to the parents. They opted to continue the pregnancy. The baby was born at 40 weeks' gestation by emergency Cesarean section and within the following week underwent cardiac surgery.

Figure 3.

X chromosome profile showing a 0.5–1.0 Mb duplication between Xp22.32 and Xp22.31, represented by elevation of 4–5 bacterial artificial chromosomes (BAC) using a BlueGnome constitutional focused BAC array (RP11-60N3-> RP11-769N24)x2, (RP11-44F2)x2 mat).

The second case is one of an uncommon benign CNV. This patient underwent invasive testing by CVS, as the fetus had a nuchal translucency of 3.4 mm and megacystis. Conventional G-banding was normal but the focused constitutional array revealed a deletion 6q27 of ∼ 300 kb. This deletion may have potentially disrupted part of the TBP gene, potentially predisposing the baby to neurodevelopmental morbidity26. Analysis of the maternal blood showed that the deletion was inherited (the mother is of normal intelligence). The patient and her partner were seen and counseled at this point so that they could be made aware that further testing was taking place. Further studies using FISH, higher-resolution 60-kb oligonucleotide array and subtelomeric multiplex ligation-dependent probe amplification were performed. The outcome was that the TBP gene was intact and the deletion was likely to represent a rare benign CNV.

In both these cases there was a delay in providing the couple with a definitive result and follow-up studies were financially costly. Both cases may have had the potential to increase parental anxiety, underlining the need for clear and timely communication between the laboratory, geneticists and fetal medicine specialists.

At the time of writing aCGH is a more expensive technique than conventional G-band karyotyping. However, if the published detection rates are accurate then aCGH may prove to be the more cost-effective option, especially as the cost of aCGH is decreasing rapidly. A comprehensive economic evaluation is required to investigate the extra cost per case identified using aCGH compared with G-band karyotyping and whether this is a cost that would be acceptable to national healthcare services. Different models will be required for different health economies. Within this analysis the cost of follow-up testing would need to be included when a CNV that is pathological, a VOUS or an uncommon benign CNV (as in the case above) is detected. It may be that currently standard G-banding with ‘intelligent’ adjuvant FISH testing (for example 22q deletion for cardiac anomaly) may be more cost-effective than performing aCGH with the risk of VOUS. This economic evaluation would need to take into account firstly the diminishing cost of aCGH and secondly that the number of VOUS will decrease as databases such as DGV increase and large centers establish robust internal mapping of benign CNVs.

At present different countries are exploring if and how prenatal aCGH should play a role and are coming to markedly different conclusions. Lee et al.22 in Taiwan concluded that they would recommend aCGH as a routine prenatal screening tool for everyone undergoing invasive testing. Conversely the American College of Obstetricians and Gynecologists concluded that it is useful when a fetal anomaly is found on ultrasound but is currently not a substitute for G-band karyotyping27. Similarly the Italian Society of Human Genetics recently reported that aCGH should not be used as a substitute for conventional karyotyping and only as an adjuvant in specific cases (sonographic fetal anomalies, apparently balanced chromosomal rearrangements or supernumerary marker chromosomes)28.

The Future

The future of microarray technology in the prenatal setting may be to target specific regions in which clinical interpretation is non-equivocal (i.e. trisomies, common well documented microdeletion/duplication syndromes and recently discovered microdeletion/duplication syndromes uncovered by advancing technology). The number of targeted regions would then increase as information from the genetic community as a whole increases. This would eliminate the concern regarding VOUS in the prenatal setting. In this way aCGH could be individualized to healthcare systems and regulatory frameworks in different countries.

In 1997, cell-free fetal DNA (cffDNA) was discovered in maternal plasma29. This opened up the possibility of non-invasive testing of fetal DNA, eliminating the risk of miscarriage (∼ 1%) associated with invasive testing. Since then clinical applications have included fetal sex determination30, fetal rhesus status in mothers with a history of hemolytic disease of the newborn31 and a very few single-gene disorders32. Fetal DNA accounts for just 3–10% of the total free DNA in the maternal serum, the vast majority being maternal. The proportion of cffDNA increases with gestation and disappears after delivery. The ultimate goal will be the ability to perform high-resolution fetal chromosomal analysis on DNA obtained from the maternal circulation.

Next-generation sequencing (NGS) has the ability to review the entire genome. It can detect single-gene mutations and CNVs and it provides tens of thousands of sequences per chromosome. The genome is fragmented, sequenced in short reads and reassembled to be compared with a genomic database33. At present its major disadvantages are its high cost per sample and that analysis takes several days per sample34, but it has been suggested that costs will fall dramatically as NGS is used more frequently; an estimated cost by the end of 2012 is $ 1000 per case35. However, problems with interpretation of the data generated and determining the data's clinical relevance suggest that large-scale studies will have to be undertaken before it can take its place in the clinical setting, especially in prenatal cases.

Conclusion

Microarray technology is at present a useful adjunct to other methods of chromosomal testing in the prenatal setting, especially when there is risk of a complicated chromosomal abnormality (as is the case when a structural abnormality is detected on ultrasound). The concern regarding VOUS is currently a very real one but will diminish with time. With the cost of microarray technology declining it will come into line with G-band karyotyping. The increased detection rate of microarray technology means that it is likely to become the most cost-effective option in the future.

Acknowledgements

S. H. is funded as a Clinical Research Fellow by the children's medical research charity Sparks.

Ancillary