Microscopic chromosome analysis of cultured cells has been regarded as the standard method for prenatal cytogenetic diagnosis since its first application to prenatal testing in 1966 (Steele and Breg, 1966) and has become routine since the first use of chromosome banding (karyotyping) in the early 1970s. Karyotyping has proved to be highly reliable for the diagnosis of numerical chromosome abnormalities (aneuploidies) and large structural rearrangements [>5–10 million base (Mb) pairs] in fetal cells obtained invasively by either amniocentesis in the second trimester of pregnancy or chorionic villus sampling in the first trimester. The supremacy of karyotyping in prenatal cytogenetic diagnosis has been challenged by the introduction of molecular cytogenetic methods including interphase fluorescence in situ hybridization (FISH), quantitative fluorescent PCR and more recently, multiplex ligation-dependent probe amplification (Boormans et al., 2010) for the rapid detection of aneuploidies for chromosomes 13, 18, 21 and the sex chromosomes (Shaffer and Bui, 2007).
In addition to the common aneuploidies, many submicroscopic chromosomal rearrangements that lead to copy-number gains or losses have been shown to cause distinctive and recognizable clinical phenotypes. The sensitivity of detection of copy-number alterations has increased significantly with the advent of microarray-based comparative genomic hybridization (aCGH) with its commercially available multiple platforms. Together with improved assemblies and annotation of genome sequence data, these methods allow for the rapid identification of new syndromes that are associated with submicroscopic genomic changes in children with idiopathic intellectual disabilities, autism, developmental delay and/or multiple congenital anomalies. This has spurred interest in applying this technology to prenatal diagnosis. During the ISPD 2010 meeting in Amsterdam, The Netherlands, a very well-attended debate dealt with the current role of traditional cytogenetics (karyotyping and FISH) and whether aCGH can be considered as a replacement for this routine testing in the near future.
ARGUING IN FAVOR OF CONVENTIONAL KARYOTYPING
Karyotyping and FISH are essential adjuncts to microarray analysis. Signature Genomics has provided microarray analyses on more than 45 000 cases since 2004. Although most of these cases are newborns or older children with developmental, intellectual and/or physical disabilities, more than 2000 cases have been for prenatal evaluation. Our experience, along with that of other laboratories, has demonstrated that microarrays will detect those unbalanced aberrations that can be seen by chromosomal banding and those that require molecular methods, such as FISH, to visualize. More importantly, microarray analysis has uncovered complex karyotypes that cannot be discerned through the light microscope. This technology can be high throughput, with turnaround times as fast as 48 h from DNA extraction. Depending on the platform and genomic content, microarray analysis can be comprehensive and have high resolution with extensive genomic coverage.
The questions that need to be answered are: Is microarray analysis sufficient for the detection of cytogenetic abnormalities? Is there still a role for karyotyping and FISH in this new era of cytogenetics?
It is well recognized that microarray analysis is not useful when there is no net gain or loss of the chromosomes. Thus, balanced rearrangements, such as reciprocal and Robertsonian translocations, inversions and balanced insertions, will not be detected by this technology. On the basis of this fact, another question comes to mind. Does it matter if these balanced alterations are not detected? Indeed it may. For example, carriers of balanced Robertsonian translocations are at risk for uniparental disomy (UPD) (Shaffer, 2006). Although aCGH cannot identify UPD, single nucleotide polymorphism (SNP) arrays can detect regions or chromosomes with copy-neutral absence of heterozygosity, which may indicate an isodisomy (Faas et al., 2010). However, even SNP arrays cannot detect heterodisomy, the most common form of UPD, without testing the parents in conjunction with the fetal specimen. Therefore, although SNP arrays may detect UPD, it is only the rarest forms of isodisomy, or significant stretches of isodisomy due to recombination of a heterodisomy, that can be detected with this technology. Nonetheless, it may be important to detect these relatively rare events, and the identification of a balanced Robertsonian translocation is one clue that UPD may be present.
In addition to the detection of isodisomy, it may be important to detect balanced rearrangements, because some de novo reciprocal translocations or insertions are known to disrupt genes and can lead to abnormal phenotypes or genetic disease without detectable gains or losses at the rearrangement breakpoints (Bodrug et al., 1990). Finally, although of no consequence to the current fetus, the detection of a balanced translocation in a family will provide the couple with future risk assessments for unbalanced offspring and information that they can use for reproductive planning. In our laboratory, we have examined 56 cases of apparently balanced translocations. In 11% of cases, a gain or loss at the presumed breakpoint was detected. However, in 89% of cases, if microarray analysis were the only analysis performed, these balanced translocations would have gone undetected, and for those that were inherited, the family would not be aware of the potential for unbalanced offspring in the future.
We recently examined 55 marker chromosomes for which samples were submitted to our laboratory for further molecular characterization. Of these cases, only 14 of 26 nonmosaic markers were detected by microarray analysis, leaving 46% of array results as apparently normal. This may reflect that the markers are mainly heterochromatin, but the lack of detection does not completely exclude a possible phenotypic effect. In addition, only 12 of 29 mosaic markers were detected by array analysis in our laboratory. It is well documented that low-level mosaicism may not be detected by this technology (Ballif et al., 2006; Neill et al., 2010). Thus, in the majority of these marker cases, chromosome analysis was crucial for their detection.
At Signature Genomics, we perform FISH for every abnormal microarray result. We do not consider this a confirmation of the array finding but rather a visualization of the abnormal chromosome. Chromosome visualization is essential for identifying the type of rearrangement, assessing additional family members with the same FISH probes as used on the proband and providing accurate genetic counseling for pregnancy management and future reproductive planning. Signature Genomics has identified a number of cases in which FISH evaluation revealed more complex rearrangements than suspected based on the array results.
DOUBLE SEGMENTAL IMBALANCES
A typical microarray pattern for an unbalanced translocation, or double segmental imbalance, is a gain on one chromosome and a loss on a different chromosome (Figure 1A). However, in many cases, the chromosome gains and losses are independent events (Figure 1B), which can only be determined by visualizing the rearranged chromosomes by FISH (Figure 1C). Our laboratory has identified cases in which a gain and loss are apparently on the same chromosome. In some cases, both imbalances are located on the same chromosome as revealed by FISH (Figure 2A and B), and in other cases, one abnormality is on one homolog, and the other imbalance is on the other homolog (Figure 2C and D).
SINGLE SEGMENTAL IMBALANCES
In some cases, single segmental imbalances can also be unbalanced translocations. These unbalanced translocations can have a microarray pattern of a single gain of a chromosome segment (Figure 3A and B) or a single segmental loss (Figure 3C and D). Many of these single segmental imbalances are revealed to be insertions after FISH analysis (Neill et al., 2011). Finally, the same microarray pattern of a single segmental gain (Figure 4) may be revealed by FISH to represent a marker chromosome or duplication. Thus, a gain seen by microarray analysis may represent a duplication, an insertion, a marker chromosome or an unbalanced translocation, and FISH is critical for distinguishing between these possibilities.
On the basis of these findings, there is a strong role for FISH or karyotyping after an abnormal finding by microarray analysis to clarify the rearranged chromosomes further. Single segmental imbalances may be revealed to be duplications or deletions, insertions, marker chromosomes or unbalanced translocations. Thus, chromosome visualization after microarray analysis is essential for delineating the rearrangement and assessing for further potential imbalance. Chromosome analysis in conjunction with a microarray will aid in identifying balanced rearrangements and will help find marker chromosomes that may not be detectable by microarray analysis because of low-level mosaicism or inadequate coverage near the pericentromeric regions on the microarray. Together, these technologies allow for assessing additional family members for rearrangements identified and providing accurate genetic counseling to the family.
In this discussion, I have restricted my comments to the scientific aspects of the limitations of microarray testing. Issues relating to the cost of performing multiple assays, identification of unclear or unwanted information by microarrays and the ethical or moral obligations of providing as much information as possible to the family after undergoing an invasive testing procedure are all aspects that must also be considered in prenatal testing.
ARGUING AGAINST CONVENTIONAL KARYOTYPING
In the last 2–3 years, the robustness of molecular karyotype technologies, such as aCGH, has become obvious to the medical and laboratory community involved in prenatal diagnostic testing. There are at least two situations for which the augmentation of conventional karyotype with genome-wide arrays has become the gold standard for accurate prenatal diagnosis and proper genetic counseling. The first one is in the setting of a fetal sonographic abnormality with a normal karyotype. Several articles reporting either single cases (Maitz et al., 2008; Vetro et al., 2008; Choy et al., 2010) or cohorts of fetuses (Tyreman et al., 2009; van den Veyver et al., 2009; Faas et al., 2010) highlight the increased diagnostic yield of the molecular karyotype. General practice suggests that sonographic anomalies, such as increased nuchal translucency, association of two or more echographic markers of aneuploidy, amniotic fluid volume alteration and/or intrauterine growth retardation associated with major structural abnormalities may benefit from this technology. The second circumstance for which the genome-wide array is now considered essential is the finding of chromosome rearrangements for which the clinical significance remains ill-defined, such as in cases of apparently balanced de novo chromosome rearrangements, supernumerary marker chromosomes of unknown origin and structural abnormalities not obviously classifiable as clinically irrelevant polymorphisms. An example is provided by the case of an amniocentesis performed because of advanced maternal age in which conventional cytogenetics detected a 15q+. Although the 15q+ was inherited from a normal mother, FISH using a painting probe demonstrated that it was completely chromosome 15. FISH analysis of the Prader-Willi syndrome/Angelman syndrome region showed only one signal on the abnormal chromosome. Doubts about the pathogenic versus benign nature of this finding were definitely solved when the aCGH revealed the exact genetic content of the additional material, showing that a relatively small (1.4 Mb) region containing only blocks of pseudogenes was amplified several times becoming microscopically visible. In fact, amplification of that region has been reported in normal individuals (Fantes et al., 2002). Evidence regarding the increased diagnostic yield of this technique with respect to conventional karyotype makes its use tempting in a routine cytogenetics practice, although the debate on possible pitfalls of this approach is still ongoing essentially concerning the possible detection of copy-number variations (CNVs) of uncertain or unknown clinical significance.
Considering the frequency of some genomic disorders with strong impact on the postnatal phenotype, Ogilvie et al. (2009) estimated that there is a 1-in-300 to 1-in-600 chance of detecting a known disability-causing CNV that would be otherwise undetected by karyotype. Beyond the pessimistic point of view of Shuster (2007) envisaging prenatal diagnosis by microarray as a means to search and destroy fetuses that are less than perfect, most geneticists and genetic counselors converge on the serious problem arising from CNVs of uncertain significance. It is clear that targeted platforms containing probes covering both regions whose deletion/duplication are clearly associated with fully penetrant diseases, and regions traditionally considered at risk (such as the subtelomeric and pericentromeric ones) may significantly reduce the risk of uncertain findings. However, such platforms not only require continuous updating for pathogenic deletions/duplications but also essentially may not significantly reduce the risk of having an abnormal child. In fact, CNVs reported to be responsible for an abnormal phenotype, or at least for intellectual disability, are scattered all over the genome and are not only located in regions mediated by a specific DNA architecture favoring occurrence of recurrent deletions/duplications. Thus, a backbone between largely recognized ‘at risk’ regions is required, and this backbone should have probe spacing dense enough to detect rare causative cryptic imbalances, including those associated with reciprocal translocations (De Gregori et al., 2007) and neocentromeric supernumerary marker chromosomes (Marshall et al., 2008), for which there are no expected regions of imbalance. Moreover, a dense genomic backbone should allow detection of mosaic imbalances present in at least 20% of cells, as the detection of mosaic rearrangements strictly depends on the number of consecutive probes having an abnormal log2 ratio. From all these considerations, it is clear that a high-resolution array platform covering the whole genome would provide much more informative results than one containing only low coverage limited to few disease-associated regions. To try to estimate this risk, we compared two recently published articles in which the authors applied whole-genome platforms to two cohorts of fetuses/abortions with major sonographic abnormalities and a normal karyotype (Tyreman et al., 2009; Faas et al., 2010). Both studies used high-resolution SNP-array platforms (Affy 6.0 and Affy 250k, respectively). Tyreman et al. (2009) reported about 106 fetuses in which they found 35 rare CNVs. As this was a retrospective study, the parents could not be examined. The authors considered CNVs clinically relevant if they were not reported in control individuals (both public available and internal databases were used), contained obvious dosage-sensitive disease genes, or overlapped with previously reported pathogenic CNVs. They found 10 CNVs that were causative for the sonographic anomaly, five novel and five syndromic. CNVs of uncertain significance were found in 13 of the cases (12%). In a similar cohort of 30 fetuses/abortions, Faas et al. (2010) detected five abnormalities consisting of two UPDs and three CNVs, all novel. After parental screening, they found that among the remaining six CNVs, four were inherited from a normal parent and thus considered likely benign, while the two de novo, corresponding to 6%, were of unclear significance. In our laboratory, we examined 63 fetuses and abortuses all with a normal karyotype and with different ultrasound abnormalities, by using genome-wide oligo aCGH platforms with a resolution ranging from 100 to 40 kb on average (Agilent both 60K and 180K) (unpublished results). Clinically relevant CNVs, two syndromic and three novel, were detected in five cases (10%). After parental analysis, one further CNV was determined to be inherited and was considered benign, while the last one was de novo and, in the absence of similarly reported cases, was considered to be of unclear significance. Putting together our study and that of Faas et al., we can conclude that whole-genome high-resolution platforms leave a margin of uncertainty in about 3.2% of cases, with about 10% of clinically relevant CNVs identified, 72% of which were novel.
Thus, the point is whether we are able to deal with 3.2% of the results having uncertain clinical significance or if, by adopting specific targeted platforms, we are able to reach an acceptable compromise between an increased diagnostic yield (thanks to the identification of cryptic causative CNVs) and a low detection rate of CNVs with uncertain significance. In fact, van den Veyver et al. (2009) had only 1% uncertain results on a cohort of 300 fetuses with different ascertainments by using targeted platforms at different resolution, and this group could confirm these percentages on a larger cohort (Breman et al., 2010). Targeted platforms in which dosage-sensitive genes associated with late-onset diseases have been eliminated might also be desirable to reduce ethical concerns. The finding of the more common CNVs associated with incomplete penetrance (Ben-Shachar et al., 2009; Vissers et al., 2010) could be communicated to the family, even if the present scarcity of data for some of them may hamper the definition of the correct risk of abnormal phenotype. In any case, genetic counselors are already familiar with the problem of incomplete penetrance related to the DiGeorge deletion or the William–Beuren syndrome region duplication (Berg et al., 2007; Calderon et al., 2009).
Common protocols for the application and interpretation of genomic arrays in prenatal diagnosis would be desirable and hopefully be capable of decreasing the risk of unexpected findings. It seems likely that in the very near future, the availability of shared databases specifically dedicated to prenatal diagnosis coupled with the growing amount of data regarding CNVs and dosage-sensitive genes could make it easier to interpret genomic arrays. A possible workflow for prenatal molecular karyotype, completely eliminating conventional cytogenetics is as follows: Fetal sampling and parental blood collection → DNA extraction and quality controls → SRY PCR for fetal sex determination → Microsatellite analysis to exclude maternal contamination → Genome-wide array analysis and quality controls on results → Evaluation of detected CNVs → Possible evaluation of parental DNAs → Reporting (available in 7 days if parental DNAs have to be analyzed).
It must be noted that, depending on the type of platform used, when a sufficient number of consecutive probes is considered for a positive call, quality assessment of the experiment must be carefully evaluated. Only experiments that pass quality control steps should be taken into consideration. Molecular karyotyping does not require any further confirmation of causative CNVs. Moreover, in a diagnostic setting, it is not important to know if the causative deletion detected in the fetal genome is associated with a translocation or the amplification to a supernumerary marker chromosome. So once again, cytogenetics is not necessary to better understand what risk does the fetus have of a serious malformation or a limited quality of life? This is what parents need to know to make an informed decision to continue or to interrupt the pregnancy. On the contrary, conventional cytogenetics, implemented by FISH, must be performed on the parents to understand whether the fetal deletion/duplication is due to an unbalanced segregation of a balanced rearrangement present in one parent (translocation, insertional translocation, inversion or complex rearrangement) thus increasing the recurrence risk. In conclusion, genome-wide array analysis may act as a substitute for conventional cytogenetics in prenatal diagnosis, although the latter remains irreplaceable to complete diagnosis and proper genetic counseling to the parents.
This work represents the efforts of many individuals. LGS thanks the following persons at Signature Genomic Laboratories for their helpful discussions and for providing figures to illustrate the complex cases that can be identified by microarray analysis: Britt Ravnan, Roger Schultz, Beth Torchia, Trilochan Sahoo, Allen Lamb, Blake Ballif, Bassem Bejjani, Justine Coppinger, Sarah Alliman, Mindy Preston, Annie Morton, Anne Bandholz, Nicholas Neill and Jill Rosenfeld Mokry. LGS also thanks Aaron Theisen and Erin Dodge for preparing the manuscript for submission. AV and OZ thank Dr Roberto Ciccone for extensive discussions on this topic.
Conflict of Interest
Lisa G. Shaffer is an employee of Signature Genomic Laboratories, a subsidiary of PerkinElmer.