Diagnostic Genome Profiling

This commentary appears in conjunction with the Review Article by Bejjani et al.1 In these articles, The Journal of Molecular Diagnostics explores array comparative genomic hybridization (CGH) and offers two perspectives on the question of testing using whole-genome arrays or targeted arrays.

Unbiased Whole-Genome Array CGH

For several decades, cytogenetic analysis has relied on chromosome banding by karyotyping for the genome-wide detection of structural and/or numerical chromosomal anomalies in patients with various disorders and/or disabilities.2,3 High-resolution targeted fluorescence in situ hybridization (FISH) analysis was developed as an additional tool for the identification of common microdeletion syndromes, which cannot be identified by karyotyping because of its limited resolution. The FISH technology is predominantly used to confirm a clinical diagnosis because it can only screen for a limited number of genomic targets in a single hybridization experiment. Recent developments in genomic microarray technologies have resulted in a novel profiling approach called microarray-based comparative genomic hybridization or array CGH.4,5 Array CGH combines the advantages of karyotyping with that of the FISH technology, ie, a whole genomic view with high resolution. Bejjani and Shaffer,1 in this issue of The Journal of Molecular Diagnostics, review the potential of this technology in clinical diagnosis, with a focus on targeted genomic arrays. We agree with the authors in that array CGH technology is revolutionizing cytogenetics and, as such, is providing clinicians with a powerful diagnostic tool that can be used in their daily practice. We disagree, however, with another conclusion in the article that specifically designed arrays targeting genomic regions of known clinical significance offer the best option for diagnostics at present. Bejjani and Shaffer state that “genome-wide dense arrays that are currently available for research are not appropriate to use in a clinical diagnostic setting as these arrays raise a number of medical, technical, and financial concerns.”1 In contrast to Bejjani and Shaffer, based on our experience in a clinical setting,6,7 we are strongly convinced of the diagnostic power of whole-genome profiling by array CGH and believe that the whole-genome approach offers numerous advantages over targeted arrays. These advantages will be discussed below. In addition, we will briefly touch on alternative microarrays being developed and elaborate on the future prospects of genomic microarrays in clinical diagnostics. Although we use mental retardation as an example, the discussion can be seen in a wider perspective, including disorders such as autism, inborn errors of development, etc.

Genomic Microarrays: Preparation, Analysis, Validation, and Interpretation

Several microarray platforms have been developed for unbiased genome-wide DNA copy number analysis using bacterial artificial chromosome (BAC) arrays, initially with a coverage of approximately one clone per megabase6,8,9 and, currently, with a tiling resolution of approximately one clone per 100 kb.7,10 The most commonly used BAC clone sets for microarray purposes have been extensively validated using FISH and DNA fingerprinting, as well as end sequencing.9,11,12 Therefore, less than 1% of the BACs in these clone sets will map to incorrectly assigned genomic locations. Moreover, even when a clone has accidentally been mapped to a wrong position in the genome, this will not pose serious problems, let alone have diagnostic consequences. This is because the use of high-density microarrays, such as our tiling resolution genome-wide BAC arrays, allows the employment of spatial information and does not rely on single clone measurements. In our diagnostic setting, we only follow up those genomic alterations that encompass a minimum of three adjacently located BAC clones. This approach dramatically reduces the chance that a problem is caused by rare clones with an adverse position assigned to it, since it is highly unlikely that three or more of such clones are located next to each other. The intrinsic technical variability of the microarray-based method is also reduced to a minimum by using this same approach. In addition, all microarray-based findings with a potential clinical relevance (see below) are validated by an independent technique such as FISH and/or semiquantitative polymerase chain reaction (eg, multiplex ligation-dependent probe amplification13). Only after an independent confirmation are the results sent to the referring clinician, who subsequently has the parents tested for de novo occurrence. In our experience, this does not lead to significantly raised anxiety within the family as long as its members are counseled properly. The clinical geneticists and pediatricians involved are accustomed to dealing with chromosomal aberrations in their patients. Therefore, the (subtle) chromosome abnormalities identified by genomic microarray analysis, accommodated by proper explanation, will not pose serious diagnostic challenges to these clinicians.

One important issue that has been raised in recent years is the unexpectedly high frequency of inherited submicroscopic copy number variants,14,15 along with the initial difficulty in discerning between a copy number alteration causing disease versus one without clinical consequences. In the period between 2004 to 2005, we were the first to test patients with unexplained mental retardation using a tiling resolution microarray in a diagnostic setting.7 At that time, knowledge about genomic copy number variation was limited, and therefore, extensive parental analysis was necessary to distinguish between inherited variants and de novo alterations. Since then, knowledge about genomic variation has increased dramatically, and it appears that a limited number of genomic variation hotspots explain the majority of variation in the human population at the BAC clone resolution level (∼100 kb).7,16 Because these regions appear to vary both in normal controls and in patients, we have decided not to follow up genomic variations at these regions in our diagnostic setting. After having excluded these variation hotspots, it is our experience that there are either no or few regions left that display genomic copy number variation in an individual patient.7 As a consequence, parental testing is limited to those genomic alterations that have a high potential of being clinically relevant.

In all, we feel that genome-wide BAC arrays can effectively be used in a diagnostic setting, and this has now been demonstrated by several laboratories including ours.6,7,17,18,19,20 Still, one can imagine that both the workload and the costs are higher for a genome-wide microarray analysis compared with the targeted microarray analysis encompassing far less genomic clones. Therefore, the real issues include determining the advantages of the genome-wide approach versus the targeted approach and whether specific subgroups of patients can be directed toward one or the other type of analysis.

The Added Value of Whole-Genome Microarrays over Targeted Microarrays

To answer this question, we first have to realize how targeted arrays are designed. In general, these arrays contain clones targeting genomic regions of known clinical significance, for example, all known microdeletion syndrome regions and all subtelomeric regions, which are known to be frequently affected in patients with mental retardation.21,22,23 As a consequence, these arrays are very well suited for screening these specific regions but only these regions. Still, this is an important clinical application with considerable diagnostic yield in patients submitted for a variety of developmental problems, as was demonstrated by the recent publication of Shaffer et al.24

The diagnostic yield of targeted as well as genome-wide microarray analysis, however, strongly depends on the clinical selection as well as the previously performed cytogenetic analyses, and this may vary considerably between countries and even between institutions within countries. In The Netherlands, like in many other countries, most patients with congenital malformations and/or mental retardation are seen by a highly trained clinical geneticist and/or pediatrician. These professionals will recognize most known microdeletion syndromes and will request specific FISH tests to confirm their diagnosis. Patients with these recognizable syndromes will, therefore, only rarely be sent in for microarray analysis. In addition, cytogenetic analysis with chromosome banding will be performed in all patients in The Netherlands with mental retardation with or without congenital malformations. For those patients in which no chromosomal aberrations are detected, a high-throughput and relatively cheap subtelomeric screen will be performed.13 Genomic microarray analysis is only indicated after the clinical evaluation (and subsequent confirmation by FISH analyses), the routine chromosome analysis, and the subtelomeric analysis fail to provide a diagnosis. As a consequence, the diagnostic yield for a targeted array, which specifically targets subtelomeric regions and known regions for microdeletion syndromes, will be low. Indeed, in our first series of 100 patients with unexplained mental retardation tested on our genome-wide tiling resolution BAC array,7 we identified only two causative genomic alterations that contained genomic regions for known microdeletion syndromes (both atypical cases) and no subtelomeric aberrations. In addition, we identified eight submicroscopic interstitial alterations localized throughout the human genome, none of which would have been identified by the targeted microarrays currently available. Similar findings have been reported by others using either 1-Mb resolution BAC arrays17,18,19,20 or, more recently, microarrays containing over 100,000 single nucleotide polymorphisms.25

The two main conclusions that can be drawn from these whole-genome diagnostic microarray approaches in mental retardation are that submicroscopic copy number alterations, both subtelomeric as well as interstitial, are responsible for a considerable proportion of mental retardation, varying between 5 and 20% of cases depending on clinical selection, and that these submicroscopic alterations occur all over the human genome. Until now, there were only a few recurrent alterations that have been identified. This poses a serious challenge for targeted microarrays and predicts that these arrays need to be updated regularly to stay abreast of all novel reported regions involved in mental retardation. Therefore, we conclude that genome-wide microarrays, especially those with a tiling resolution coverage, are of great value in the diagnosis of mental retardation and will continue to be so in the coming years. These genome-wide microarrays may actually replace routine chromosome banding as a first diagnostic screen in a significant number of clinical indications, such as mental retardation, given the increase in diagnostic yield. The pace of implementation, however, strongly depends on the finances involved and the expertise present in the various diagnostic laboratories.

Novel Genomic Microarray Approaches

Most currently used genome-wide copy number profiling microarrays are produced in academic settings, and the resolution of these microarrays varies depending on the type and number of genomic targets selected, the protocols used, and the data analysis tools used. In addition, aberrations below 100 kb cannot be detected with current resolution, which is limited by the size of the genomic fragments (ie, BACs) used as array elements. Only recently have private enterprises embarked on this novel genomic microarray application, and several companies are now offering microarrays for genome-wide copy number profiling. These microarrays encompass oligonucleotides targeting random genomic sequences26,27 or single nucleotide polymorphisms.25,28 The advantages of using such commercial platforms for diagnostic applications are numerous.

1. They provide a higher genome coverage than most microarrays generated in academia by placing up to half a million oligonucleotides on a single microarray, which will increasingly allow the detection of smaller aberrations linked to disease.

2. They can be produced in large quantities according to industrial quality standards.

3. They are available to all research and diagnostic laboratories and those without access to dedicated microarray facilities.

4. Their widespread use will rapidly generate large data sets of normal controls as well as patients with various disorders, thus greatly facilitating clinical interpretation of genome profiles.

Given these advantages, it can be expected that these commercially available microarrays will rapidly find their way to routine clinical practice.

Conclusion

Despite the natural caution associated with the implementation of new technologies in the clinical arena, genomic microarrays have been recognized as effective diagnostic tools in molecular cytogenetics for several years now. The ability to obtain quantitative copy number information on the entire human genome has already been shown to significantly improve the diagnostic yield in mental retardation. Further increases in resolution provided by commercially available ultra-high density microarrays are likely to increase this yield even more. We expect that, ultimately, genomic copy number scanning of all ∼250,000 exons in the human genome will become the standard in genomic microarray diagnostics, thus allowing the detection of causative dosage variations up to the single exon level. The current diagnostic use of targeted microarrays will likely be surpassed by the availability of affordable whole-genome arrays combined with rapid increases in our knowledge on the clinical interpretation of these microarrays.