Article in press - uncorrected proof

J Lab Med 2009;33(5):255–266 2009 by Walter de Gruyter • Berlin • New York. DOI 10.1515/JLM.2009.045 2009/45

Molekulargenetische und zytogenetische Redaktion: H.-G. Klein Diagnostik

Array comparative genomic hybridisation in clinical diagnostics: principles and applications Array-CGH in der klinischen Diagnostik: Prinzipien und Anwendungen

Uwe Heinrich1, Imma Rost1,*, Anthony Brown2, wertvollen, genom-weiten Screeningmethode zur Auf- Tony Gordon2, Nick Haan2 and Jessica Massie2 deckung chromosomaler Vera¨ nderungen in Form von Kopienzahlvarianten (CNV) entwickelt. Die kommerziell 1 Centre for Human Genetics and Laboratory Medicine erha¨ ltlichen Plattformen beinhalten die Subtelomerregio- Dr. Klein and Dr. Rost, Martinsried, Germany nen sowie die bekannten Mikrodeletions- und Mikrodu- 2 BlueGnome Ltd., Cambridge, UK plikationssyndromregionen, das restliche Genom wird mit unterschiedlichen Auflo¨ sungen von 8 kb bis 1 Mb abge- Abstract deckt. Neben der Aufdeckung eindeutig pathogener oder harmloser CNVs kann die aCGH auch CNVs mit unklarer In the last few years, array comparative genomic hybri- klinischer Signifikanz aufdecken, welche die Interpreta- disation (aCGH) has become a valuable genome-wide tion einer aCGH-Analyse erschweren. Ihre Hauptindi- screening tool for the detection of chromosomal aber- kationsstellungen umfassen Kinder mit mentaler rations in the form of copy number variants (CNVs). Com- Retardierung, Entwicklungsverzo¨ gerung, angeborenen mercially available platforms cover the subtelomeric Fehlbildungen und neuropsychiatrischen Erkrankungen regions and all known microdeletion/microduplication wie Autismus. In dieser Patientengruppe wird die aCGH syndrome regions, as well as the rest of the genome, with zunehmend die klassische Chromosomenanalyse als resolution ranging from 8 kb to 1 Mb. Besides detecting Ersttest ersetzen und ihr Einsatz in der Pra¨ nataldiagnos- clearly pathogenic or benign CNVs, aCGH can uncover tik steht derzeit in der Diskussion. Ein weiteres viel CNVs with unknown clinical significance, thus compli- versprechendes Einsatzgebiet ist die Tumordiagnostik, cating the interpretation of aCGH results. The main wo die aCGH dem Behandler die Klassifizierung und indications include children with mental retardation, Prognosestellung verschiedener Tumorentita¨ ten erleich- developmental delay, congenital anomalies and neuro- tern wird. psychiatric disorders such as autism. In this patient group, aCGH will gradually replace conventional chro- Schlu¨ sselwo¨ rter: Array-basierte genomische kompara- mosome analysis as the frontline test, and its implemen- tive Hybridisierung (aCGH); Kopienzahlvarianten (CNV); tation in prenatal diagnostics is in discussion. Another Mentale Retardierung; Mikrodeletionen, Mikroduplikatio- promising field is cancer diagnosis, for which aCGH will nen; Pra¨ nataldiagnostik. enable clinicians to classify and prognose different tumours more accurately in the near future.

Keywords: array comparative genomic hybridisation Introduction (aCGH); copy number variant (CMV); mental retardation; microdeletions; microduplications; prenatal diagnostics. The first cytogenetic era started in 1956, when Tjio and Levan w1x determined for the first time that the exact number of human was 46 using Giemsa- Zusammenfassung stained metaphase preparations (Figure 1). In the follow- In den letzten Jahren hat sich die Array-basierte geno- ing years, several syndromes were identified as being mische komparative Hybridisierung (aCGH) zu einer caused by numerical aberrations, such as ( 21), Edward syndrome *Correspondence: Imma Rost, MD, Centre for Human (trisomy 18), (trisomy 13) and Ullrich- Genetics and Laboratory Medicine Dr. Klein and Dr. Rost, ( X). Despite the very low Lochhamer Str. 29, 82152 Martinsried, Germany resolution of approximately 20–30 Mb, some structural Tel.: q49-89-8955780 Fax: q49-89-89557878 anomalies in the form of terminal deletions, such as dele- E-Mail: [email protected] tion 5p () and 4p (Wolf- Article in press - uncorrected proof

256 Heinrich et al.: Array CGH

Figure 1 Resolution of conventional and molecular cytogenetic techniques. (A) Partial karyotype showing a trisomy 21 corresponding to a duplication of 322 genes and 46 Mb of DNA. (B) Partial karyotype of a pair and the corresponding ideograms at a 500-band level with a resolution of 4–6 Mb; the right chromosome 22 has a terminal 22q13.3 deletion (Phelan-McDermid syndrome); the deleted region has a size of ca. 6–8 Mb containing more than 100 genes. (C) FISH analysis of a metaphase showing a 22q11.2 microdeletion (DiGeorge syndrome) with the region-specific TUPLE1 probe in red and the control ARSA probe in green (the arrow denotes the missing red signal). The deleted region usually comprises 3 Mb and the TUPLE1 probe is 120 kb in length. In general, the maximum resolution of FISH depends on the probe length, ranging between 40 and 250 kb. (D) Interphase FISH representing a duplication of clone RP11-58H17 (green signals) in region Xp11.23 in a male patient. The red signal represents the X-centromere control probe DXZ1. (E) Result of an array CGH analysis corresponding to a 1.5-Mb deletion in band 15q13.3, a newly discovered recurrent . The resolution of array CGH is platform- dependent, varying between 10 kb and 1 Mb.

Hirschhorn syndrome), could be detected. The intro- imbalance could be detected in 2–7% of patients with duction of chromosomal banding techniques in the early idiopathic mental retardation (MR) and inconspicuous 1970s allowed genome analysis at a resolution of chromosome analyses w4, 5x. 5–10 Mb and led to the discovery of further deletion syn- The third era started in 1997–1998, when two groups dromes such as WAGR syndrome and Jacobsen syn- independently published the principle of array-based drome. High-resolution banding improved the resolution comparative genomic hybridisation (aCGH) w6, 7x. aCGH to 3–5 Mb and led to identification of the first micro- is based on co-hybridisation of the sample and reference deletion syndromes – Prader-Willi, DiGeorge and Smith- genomic DNA labelled with different fluorophores to Magenis syndromes. mapped DNA sequences or oligonucleotides that are The second era started in the mid-1980s, when fluoro- immobilised on a glass slide. The fluorescence of hybri- chrome-labelled DNA probes of 40–250 kb in length dised DNA is detected and expressed as a ratio that is dramatically increased the resolution w2x. With an plotted against genomic position so that an aberrant increase in the number of probes commercially available, region can be accurately mapped (Figure 2). Genomic this fluorescence in situ hybridisation (FISH) technique regions where there is more or less test sample relative rapidly became a standard method in the routine cyto- to the reference sample can be easily identified as a shift genetics laboratory, expanding the spectrum of micro- in the fluorescent ratio. The major difference with aCGH deletion syndromes identified (Williams-Beuren syndrome, compared to previous methods is replacement of the microdeletion 1p36) and revealing the first microdupli- metaphase spread with spotted DNA sequences. cation syndrome (microduplication 22q11.2). The great Although this modification precludes detection of purely disadvantage of FISH is the requirement for a tentative balanced translocations, it has the major advantage of pre-diagnosis, as FISH is a targeted diagnostic and not being a genome-wide screen at vastly improved resolu- a screening tool. A first step towards screening was intro- tion, limited only by the length of the DNA fragment spot- duced with the availability of a complete set of human ted and the genomic distance between DNA fragments. sub-telomeric probes to check for submicro- In addition, because only a DNA sample is required from scopic terminal deletions and cryptic balanced trans- a patient, much of the skill required for traditional band- locations w3x. Depending on the inclusion criteria, an ing karyotyping is removed. Using this method, copy Article in press - uncorrected proof

Heinrich et al.: Array CGH 257

BAC arrays were the first to be explored and use large- insert (;200 kb) BAC clones spotted on glass slides. The major advantage of BAC arrays is that they give very robust results that are easy to interpret; this is because of the large regions of the genome included in each fea- ture. The primary disadvantage is that the resolution is limited by the size of the insert, so that even on high- density tiling arrays the maximum resolution that can be achieved is ;100 kb. This is in contrast to oligo arrays, which use short, generally 60-bp oligos so that the res- olution is only limited by the number of spots that can be physically fitted onto an array. The possible limitation with oligo arrays is that they produce noisier data because of the shorter hybridisation probe, so that smoothing (3–10 oligos) is necessary, reducing the effec- tive resolution. Nevertheless the results are good and enable detection of breakpoints down to the gene level. Recently produced oligo arrays include 1 million (Agilent Figure 2 Principle of array CGH. Technologies) and 2 million features (Roche Nimblegen Fluorescently labelled test and reference are co-hybridis- ed with Cot-1 to an array of spotted genomic DNA sequences Inc.) with an average probe spacing of 2.1 and 1.1 kb, that are immobilised on a glass slide. After washing, the hybri- respectively. However, many clinical laboratories use dised array is scanned and the image is analysed. The array more cost-effective oligo arrays with between 44K and CGH results can be expressed as a log2 ratio of test and ref- 180K features. erence fluorescent readings plotted against genomic distance Although high-resolution arrays improve the detection for the human genome. In this example the results for a single of breakpoint regions, clinically they have done little to chromosome reveal a loss of material on the p arm. improve abnormality pick-up rates over lower-resolution arrays. This is because most CNV abnormalities detect- number variations (CNVs) can be identified and classified ed, outside known disease-causing regions, are of the as either inherited (generally benign) copy number poly- order of megabases and the majority of all changes are morphisms (CNPs) or pathogenic de novo CNVs. )250 kb. For any array an assessment of the pick-up aCGH is now standard practice in many cytogenetic rate of pathologically relevant de novo copy-number laboratories, and its use and throughput are increasing. changes has to be balanced by the number of CNVs of Many smaller laboratories are seeing the benefits offered unknown clinical significance also detected. Such chang- and are venturing into setting up aCGH technologies in es that are not described in the existing CNV databases their own laboratories. The purpose of this article is to have to be confirmed as benign CNPs using another explore the integration and application of aCGH into the method. The ideal clinical platform will sensitively detect clinical diagnostic setting and examine the technical pathogenic de novo CNVs with minimum detection of aspects to ensure its successful implementation. The benign CNV regions; resolution has to be balanced with emphasis is on diagnosis of constitutional disorders sensitivity to optimise this ideal. spanning MR and dysmorphic features (DF). Now that aCGH is becoming more common, the main difficulty emerging is interpretation of the data, and information on aCGH workflow the resources currently available is discussed. This section provides a guide to aCGH in low- and high- throughput laboratories and highlights critical issues in aCGH platforms the workflow that can influence results. An outline of a clinical cytogenetics workflow is shown in Figure 3. The primary objective of aCGH in a clinical context is to increase diagnostic yield, with secondary objectives of Patient selection increasing automation and robustness, simplifying labo- ratory workflow, and cost reduction. In contrast to pure Patients are usually referred for aCGH analysis only after research, higher resolution is only useful if it delivers other cytogenetic tests have been carried out and it has more diagnoses for patients reliably and on the time been determined that they are karyotypically normal, scale required. In this context, the most common clinical despite presenting with an abnormal phenotype. This aCGH platforms are bacterial artificial chromosome preliminary process may vary widely from one laboratory (BAC) and oligonucleotide (oligo) arrays. Both are avail- to another but usually includes karyotype banding ana- able commercially. lysis, which might be at insufficient resolution to identify Article in press - uncorrected proof

258 Heinrich et al.: Array CGH

of purity, and samples that do not have the desired metrics often still yield interpretable results. The most common cause of aCGH assay noise is dif- ferential properties in DNA between reference and patient samples, typically resulting in false positive results in GC- rich regions and sub-telomeric areas. An extreme exam- ple of this effect is shown in Figure 4, where there is large variation in the ratios in the telomeric regions of the genome because of differences in the quality of the test and reference DNA. The signal variability observed is due to the sequence and structural complexity of the human Figure 3 Clinical cytogenetics workflow. genome and properties in co-hybridisation assays, com- Timings given are for a standard BAC array protocol over 3 pounded by large variation in the DNA extraction tech- working days. It is possible to reduce certain steps to complete niques. It is suspected that differing qualities of DNA can the analysis in 2 working days. result in differential labelling of test and reference sam- ples in particular regions. In some circumstances, dye- specific bias in the labelling reaction can be present in the imbalance, and might possibly include telomeric affected regions. Many laboratories find it advantageous FISH or a PCR-based technique such as multiplex liga- to create their own male and female reference DNAs by tion-dependent probe amplification (MLPA), which, unlike pooling copy-number-normal DNAs extracted using the aCGH, require a tentative pre-diagnosis. The advantage same method as for the patient DNA rather than using of conventional karyotyping and FISH techniques is that commercially available normal genomic DNA, thereby they provide spatial information about any possible minimising the technical variation and reducing noise in genomic rearrangement and can identify balanced trans- the aCGH assay. To avoid the effect of individuals with locations, which standard aCGH cannot. polymorphisms within the reference pool, it is suggested Following failure to discover a plausible cause using that at least 10 different DNA samples are used. conventional tests, patients are typically referred for It is recommended that aCGH laboratories establish aCGH on the basis of the severity of their phenotype and robust and routine workflows that ensure the properties information in the literature about the likelihood of a diag- of extracted DNA remain constant over time. If possible, nostically significant copy number imbalance. The selec- the procedures used by referring clinics should be includ- tion of patients referred can dramatically affect the ed in this workflow; for example, there is some evidence pick-up rate of any given laboratory. that variation in blood collection tubes and the duration Whether aCGH should be used earlier in the clinical that blood spends in the tubes can strongly affect aCGH cytogenetics workflow is an interesting question and is assay quality. Ideally, extracted DNAs should be analysed discussed later. as soon as possible, and otherwise stored under appro- priate conditions. Many DNA samples remain stable DNA extraction and QC enough to assay for several months (48C) or even years (–208C). The purity of the genomic DNA used at the start of the aCGH assay is the most important factor in determining aCGH experimental protocol result quality. It is also the stage over which a laboratory has most control. The main steps in the aCGH protocol are: preparation of Some spectrophotometry metrics can be used to the sample and reference DNA; labelling (usually using assess DNA purity, such as the 260 nm/280 nm absor- random priming); co-hybridisation of the labelled test and bance ratio to determine whether a sample is protein-free reference samples with Cot-1 DNA to suppress repetitive (ideally )1.8), and the 260 nm/230 nm absorbance ratio sequences; washing to remove excess hybridisation to ensure organic contaminants have been removed solution; and scanning the slide to produce an analysable (ideally )2.0). DNA should be relatively intact and of high image. These laboratory procedures are relatively simple molecular weight as evaluated by gel electrophoresis. to implement. In general the protocols from one platform Routine monitoring and recording of these metrics is rec- to another differ very little and all follow the same basic ommended to establish variation in the DNA extraction steps. process and as a first step in troubleshooting DNA quality The standard aCGH protocol for BAC or oligo arrays issues. Subsequent DNA clean-up procedures that can is spread over 3 or 4 days with an overnight step for be effective in improving DNA quality include phenol labelling and a one- (BAC arrays) or two-night step (oligo chloroform extraction to remove non-nucleic acid organic arrays) for hybridisation (Figure 3). However, the hands- material and salt removal by precipitation with on wet laboratory time is relatively short. Many labora- repeated 70% ethanol washes. However, it is not always tories find it convenient and efficient to stagger the practical to have DNA that conforms to the ideal in terms set-up of arrays, starting one round on Monday, one on Article in press - uncorrected proof

Heinrich et al.: Array CGH 259

Figure 4 Array CGH profiles with noise. (A) A whole-genome profile with test (female) and reference (male) samples of matched quality. Here there is a tight distribution of points along the autosomal chromosomes and an expected shift in the X and Y chromosomes. (B) A sex-matched genome profile that shows peaks in copy number in sub-telomeric regions. This noise in the experiment is thought to be due to differences in test and reference DNA quality.

Tuesday and one on Wednesday. This increases the required is not linearly proportional to increased array throughput without the expense of any specialist equip- throughput. An increase in throughput can be simply ment. A stage that can cause a substantial bottleneck is achieved by using mastermixes and scaling up the num- scanning. Optimisation of settings and generation of ber of DNA extractions, labelling and clean-up reactions images for a single patient can often take 20 min, includ- carried out at the same time using plate formats and ing 5 min of hands-on time. Scanners with autoloaders corresponding clean-up reagents. The use of semi-auto- are available (e.g. Axon, Agilent, Innopsys), allowing mated liquid handling tools such as multi-channel and slides to be scanned overnight, thus providing an impor- electronic pipettes can also increase productivity. tant decrease in user intervention and hands-on time in The most difficult part of the protocol to automate is higher-throughput scenarios. application of the hybridisation solution to the arrays, and Because of the routine nature of many of the wet lab- for most platforms such automation is currently unavail- oratory steps in the aCGH workflow, the hands-on time able. Many laboratories find this part of the protocol Article in press - uncorrected proof

260 Heinrich et al.: Array CGH relatively time-consuming and it also carries the risk of Data analysis mislabelling and traceability problems. Even automated hybridisation systems such as the Tecan HSPro still Once a laboratory has established a reliable wet labora- require manual application of the samples. Hybridisation tory workflow, the critical step becomes storage (dis- stations can, however, significantly reduce the hands-on cussed above), analysis and interpretation of the large time for hybridisation and array washing. volumes of data provided by an array. A number of com- panies such as BlueGnome, Affymetrix and OGT produce software specifically for cytogenetics laboratories. Data management Data analysis comprises a number of stages, including processing of the array image (grid finding and quantifi- Record-keeping is imperative for any cytogenetics labo- cation), normalisation to remove sources of experimental ratory. For aCGH experiments, the amount of data that bias (e.g., intensity-dependent dye effects), removal of is generated is substantial, so some consideration should outliers, and automated calling of CNV regions. The be given to data management issues. sophistication of data normalisation and automated call- Not only should the final results be stored, but infor- ing are particularly important in reducing experimental mation at each stage of the experiment should also be noise and therefore false positives and negatives. Soft- recorded along with the results in a searchable database. ware should be simple to use, ideally within a single Recorded information should include: patient phenotype, package, and free the user from checking and setting of clinical details, sample type, DNA quality scores, labelling subjective tuning parameters for the aforementioned efficiency metrics, Cot-1 batch, and lot numbers of the stages of analysis. This is particularly important in a clin- kits and reagents used. If recorded consistently, this ical context to guarantee reproducibility of results. Soft- information provides the power to spot trends within the ware that both analyses array results (in single or batch data, for example, correlating phenotype with a particular modes) in a single click and then imports the data into genomic imbalance across a laboratory’s sample cohort. internal databases of consistently analysed samples is In more practical terms, it is possible to identify trends in highly desirable. the quality of results and to spot the cause of any chang- Single software packages such as the BlueGnome es, e.g. a particular batch of Cot-1 of poor quality. BlueFuse Multi package are being introduced that allow All aCGH data and the reporting of results to a clinician all these processes to be completed in a single user- require information on the human genome sequence, friendly package. These forms of software are greatly such as genomic position, gene and CNV content. Even reducing the time for sample interpretation, which until now in the post-genomic era, changes to the human recently has been one of the major hold-ups in a clinical genome sequence map are occurring with the identifi- aCGH laboratory. cation of new genes and adjustments in the position of some regions. It is therefore useful to record which build of the genome was used to make a decision in a report, Validation of aCGH results so that if the genome sequence changes in the future it will be possible to determine why a particular judgement Owing to the complex reaction kinetics of several was made. thousands of hybridisations running simultaneously, For aCGH, as for other techniques used in a cytoge- overlapping signal intensities between euploid and aneu- netics laboratory, it is imperative to have a double sign- ploid chromosomal regions can occur. Therefore, it is off procedure to ensure that no mistakes are made with very important to choose appropriate calling algorithms mixing up of samples. Samples should be tracked, ide- and threshold values to minimise false positive or nega- ally with barcodes, as they are transferred to different tive results. This is especially true for BAC arrays, vessels and on to the array slide, and these processes because most BACs, owing to their large insert size, con- should be observed by a second trained laboratory staff tain repetitive elements that can lead to cross hybridi- member. sations. In addition, BACs sometimes map to the wrong External quality control assessments to determine locus or a wrong clone can be delivered by the clone laboratory competence in performing and reporting clin- provider. ical aCGH samples are beginning. For instance, a UK To exclude methodological artefacts, aCGH results – external quality assessment (EQA) administered by especially single clone changes – not previously NEQAS is currently under development, starting with the described in the literature or databases, as well as imbal- reporting of supplied array images. It is hoped that this ances smaller than ;1 Mb, must be evaluated by an will be expanded to an exercise involving the analysis alternative method. Most laboratories use FISH because and reporting of cases from DNA samples, although this it is a standard technique in every cytogenetic laboratory. is a much more complex and expensive form of study. Other validation strategies are quantitative PCR (qPCR) Such EQA exercises are essential to obtain consistency and MLPA. qPCR is a comparably cheap but time-con- between laboratories and will eventually be widespread suming method that is not common in most cytogenetic in this field. laboratories; MLPA has the disadvantage that the probe Article in press - uncorrected proof

Heinrich et al.: Array CGH 261 sets that are commercially available only cover the sub- SNP chip analyses revealed a CNV median of 81 kb telomeric and the known microdeletion/microduplication (mean 206 kb), resulting in only 5 Mb. Another feature of regions. FISH has the advantage of being a robust tech- these studies was the finding of a reciprocal ratio nique, allowing clear confirmation of deletions and the between CNV size and frequency, i.e., the smaller the localisation of a duplication, especially for parental inser- CNV, the higher is its frequency. tional translocation. FISH is also a fast method for ana- Most CNVs are benign variants inherited from one of lysing the parents of a patient with a proven imbalance. the parents. It is a great challenge for cytogeneticists to Disadvantages of FISH are the need for metaphase chro- discriminate between benign and pathogenic CNVs, and mosomes, possibly low signal intensities or cross hybri- in some cases a clear correlation cannot be drawn disation, and difficulty in clear detection of a duplication. between a detected CNV and the patient’s symptoms, An alternative validation strategy is to use a different making genetic counselling difficult w15x. Lee et al. w16x array platform such as an oligo array to validate the listed criteria for the interpretation of newly detected results of a BAC-based array. The practicality of this CNVs. The most important factor in favour of a patho- approach was tested in a pilot study w8x at The Centre genic CNV is its de novo origin. Another factor is the size for Human Genetics and Laboratory Medicine in Martins- of the imbalance, because approximately 80% of patho- ried, Germany. As a result of this pilot study, a routine logical aberrations are greater than 1 Mb w17x and – as was established in terms of validating a single deletion mentioned above – the smaller the CNV, the more likely by FISH and duplications or several imbalances by oligo- it is to be benign. However, it should be remembered that based aCGH w9x. This approach is a fast and reliable val- there are several new recurrent microdeletion/micro- idation method. A further advantage is the more detailed duplication syndromes smaller than 500 kb, such as the description of a proven imbalance because of the higher 200-kb 2p15p16.1 microdeletion syndrome w18x and the resolution available for an oligo array. 100-kb 14q11.2 microdeletion syndrome w19x. Moreover, even very large imbalances of several Mb in size can be benign, as exemplified by a familial 14.5-Mb deletion in Interpretation a three-generation family w20x. The gene content, espe- cially the number of genes cited in the OMIM database Since the first days of conventional cytogenetics, a chro- (http://www.ncbi.nlm.nih.gov/omim/), is a further criteri- mosomal imbalance was automatically associated with a on, but the most convincing proof of causality between patient’s phenotype. However, besides apparently patho- imbalance and phenotype is the description of compar- genic imbalances, it soon became clear that there is able patients with comparable aberrations or with a point substantial variation in human chromosomes, even in in a haploinsufficient gene in the affected euchromatic regions w10x. In patients with MR, patho- region. genic aberrations can be found in approximately 17% Factors in favour of a benign CNV are parental origin, and benign variants in ;2%. Using subtelomeric screen- low gene content, small size and the finding in several ing, this ratio reversed, because pathogenic aberrations patients with different symptoms. Although inherited can be detected in 2% of the patients and variants in CNVs are benign in most cases, a pathogenic CNV can 10.4% w11x. be inherited from a healthy parent; a classical example is This reversal in the ratio of pathogenic to benign imbal- X chromosomal imbalances inherited from a healthy ances has been dramatically aggravated by the discovery mother. Other possible mechanisms involving autosomal of CNVs via aCGH w12, 13x. CNVs are defined as DNA imbalances include the following w21x. segments of more than 1 kb occurring at a variable copy Variable expressivity and reduced penetrance: number in comparison to a reference genome; CNVs with These phenomena have been known for years in families a frequency of at least 1% in the general population are with microdeletion 22q11.2 syndrome, in which carriers also referred to as CNPs. The first analyses of CNVs in of the same microdeletion often show very variable the normal population were carried out with low-resolu- symptoms; sometimes a supposedly healthy parent tion BAC arrays. For example, Iafrate et al. w13x used a displays microsymptoms that were previously over- 1-Mb BAC array to screen 39 normal and non-consan- looked. Since the introduction of aCGH, the list of such guineous individuals and identified 255 copy-variable syndromes has increased and includes the microdupli- loci. With the use of higher-resolution arrays, this number cation 7q11.23 syndrome w22x, 1q21.1 microdeletion/ increased dramatically. Redon et al. w14x analysed DNA microduplication syndrome w23x, and 3q29 micro- from the 270 individuals from the HapMap project both duplication syndrome w24x. with a whole-genome tiling-path BAC array and a 500K Unmasking of a recessive allele: This mechanism single nucleotide polymorphism (SNP) chip, finding 1447 was initially proposed by Hatchwell w25x to explain dif- CNVs representing 12% of the genome. Owing to the ferent phenotypes in monozygotic twins with a 22q11.2 large size of BAC clones, the size of CNVs is usually microdeletion syndrome. In this scenario, a deletion overestimated. Using the BAC array, Redon et al. w14x inherited from one parent combines with a recessive found a CNV median of 228 kb (mean 341 kb) resulting point mutation – inherited from the other parent or de in 24 Mb of variable genome per individual, whereas the novo – in a hemizygous gene. Despite being a very rare Article in press - uncorrected proof

262 Heinrich et al.: Array CGH event, there are some descriptions in the literature. base (Database of Chromosomal Imbalance and Examples are a patient with Wolf-Hirschhorn syndrome Phenotype in Humans using Ensembl Resources) and the due to a 4p16.3 deletion and Wolfram syndrome due to ECARUCA database (European Cytogeneticists Associ- a mutation in the WFS1 gene w26x and a patient with ation Register of Unbalanced Chromosome Aberrations). Smith-Magenis syndrome due to a 17p11.2 deletion and To aid in interpreting CNVs, some companies provide sensorineural deafness due to a mutation in the MYO15A additional databases of samples. For example, Blue- gene w27x. Gnome provides an additional database of 1400 samples Position effects: Genomic position can affect the of developmentally delay (DD)/MR that offer another ref- clinical manifestation of an aberration by locating genes erence for the occurrence of CNVs. The advantage of this close to the telomere or to heterochromatin, making them data set is that all samples have been prepared using sensitive to gene silencing, or by deleting a regulatory same method, have been analysed on the same aCGH sequence, thus affecting gene expression. The former platform, run against the same reference and consistently case was identified in a patient with Wolf-Hirschhorn syn- analysed. Such consistent data are very useful for clas- drome w28x and the latter in a patient with Townes-Brocks sifying a CNV of unknown significance until larger exter- syndrome carrying a balanced translocation with one nal databases of normal populations are available. breakpoint lying near the SALL1 gene w29x. : Genomic imprinting refers to the inheritance of two homologous chromosomes or chro- Applications mosomal regions from the same parent. Classical exam- ples are paternally inherited 15q11.2 microdeletions Mental retardation/developmental delay leading to Prader-Willi syndrome, whereas maternal inheritance leads to . MR is defined as an IQ of -70 and affects 2–3% of the Internet databases are very helpful tools for the inter- population, whereas the term DD refers to younger chil- pretation of CNVs (Figure 5). The Database of Genomic dren (usually -5 years), when IQ testing is not reliably Variation (DGV) is a useful resource and collects CNVs practicable. In more than 50% of cases, the causes of found in the normal population that have been described MR or DD are not known w31x. Conventional chromo- in the literature. Because many of the CNVs listed in DGV some analysis detects aberrations in approximately 3.7% have not been validated or characterised, it should be of patients w32x and the introduction of subtelomeric used with caution. For instance, microduplication in Xq28 screening added another 2.5–7% w33, 34x. including the MeCP2 gene is listed as a variant, although First studies using low-resolution aCGH detected it is no doubt a pathogenic CNV. DGV is also not ideal imbalances in 7–15% of patients with MR/DD and nor- because it comprises numerous studies of the HapMap mal karyotype w35, 36x. Using high-resolution aCGH, this cohort of samples and hence can overestimate the CNP diagnostic yield could be increased to 20% w37, 38x, thus frequency. Ideally, consistently analysed data from the doubling the number of aberrations detected compared same platform should be used to form a reference CNV to conventional karyotyping; Table 1 lists some examples database. To improve our understanding of genetic var- of new syndromes detected by aCGH. Therefore, it is iation in human health, a multicentre study known as the conceivable that chromosome analysis will be replaced Human Variome Project has been initiated to collect and by aCGH in the near future as the frontline analysis in publish all the relevant data w30x. As other large-scale this patient group. In a large meta analysis of 36,325 DD/ CNV studies of normal populations are completed, such MR patients investigated before the advent of aCGH, as the Wellcome Trust Case Control Consortium Hochstenbach et al. w52x estimated that only ;0.8% of (WTCCC: http://www.wtccc.org.uk/) and the 1000 all aberrations would have escaped detection by aCGH Genomes re-sequencing project (http://www. because of their balanced nature (translocations, inver- 1000genomes.org/page.php), a more complete picture of sions) or triploidy. Their conclusion was that aCGH using human CNV will emerge. Databases collecting pathogen- arrays with a resolution of 30–70 kb can detect ;99% ic CNVs and clinical data include the DECIPHER data- of all pathogenic anomalies and that subsequent karyo- typing increases this yield by only 0.23%.

Multiple congenital anomalies and dysmorphic features

Multiple congenital anomalies (MCA) and/or DF in new- borns are strongly indicative of a chromosomal aberra- tion w53x. Lee et al. w54x demonstrated that the detection rate in patients with MCA and/or DF was even higher than in patients with DD/MR (8.4% vs. 3.9%). The com- bination of DD/DF and congenital heart defect (CHD) is Figure 5 Infobox: Internet resources. another important indication for aCGH use. Thienpont Article in press - uncorrected proof

Heinrich et al.: Array CGH 263

Table 1 Examples of recurrent microdeletion and microduplication syndromes discovered by array CGH.

Locus Size Phenotype References

Microdeletion 1q21.1 1.35 Mb Moderate MR, microcephaly, autism, reduced penetrance w39, 40x 3q29 1.6 Mb Moderate MR, microcephaly, dysmorphic features, autism w41x 15q13.3 1.5 Mb Moderate MR, mild dysmorphic features, seizures w42x 15q24 1.7–3.9 Mb Moderate MR, low birth weight, microcephaly, hypospadias, w43x loose connective tissue 17q12 1.2–1.4 Mb Multiple renal cysts, diabetes mellitus w44x 17q21.31 500–650 kb Moderate MR, dysmorphic features, pronounced hypotonia w45–47x Microduplication 1q21.1 1.35 Mb Moderate MR, dysmorphic features, macrocephaly, autism, w39, 40x reduced penetrance 7q11.23 1.5 Mb Severe speech delay, mild dysmorphic features, mild MR, w48x autistic features 16p11.2 600 kb Association with autism, reduced penetrance w49x 17p11.2 3.7 Mb Mild MR, infantile hypotonia, failure to thrive, autistic features, w50x congenital anomalies Xq28 0.2–2.2 Mb MR, recurrent infections, hypotonia, speech delay w51x et al. w55x found a genomic imbalance in 30% of their and complex rearrangements with more than three patients. Even in patients with an isolated CHD, imbal- breakpoints in another 18%. The authors explained the ances as a direct cause or a genetic risk factor play an different detection rates for imbalances in terms of acci- important role w56x. dental findings of a translocation being indeed balanced and the high rate of imbalances in patients attributed to Characterisation of cytogenetic findings selection bias.

Breakpoint determination of an imbalance using chro- Autism spectrum disorders mosome analysis is only a rough approximation because the resolution ranges between 5 and 10 Mb. To warrant Autism spectrum disorders (ASDs) are a group of neu- genotype-phenotype correlations, precise delineation of rodevelopmental disorders which all have a dysfunctional the chromosome region concerned is desirable. Our own reciprocal social interaction in common with autism lying study w9x included seven cases with known chromosomal at the severe end of the spectrum. The prevalence of imbalances for which one of the breakpoints had to be ASD is approximately 37 in 10,000 individuals, with a corrected in each case after aCGH analysis, and in three four-fold over-representation of males w58x. It has long cases both breakpoint positions had to be revised. In been recognised that ASDs have a strong genetic com- some cases, the nature of an aberration had to be ponent w59x. Chromosome analysis reveals an abnor- changed. For example, one case had a presumed 1p36.3 mality in 6–7% of ASD cases, the most prevalent being terminal deletion, but subtelomeric screening revealed duplications in region 15q11–q13 w60x. Single gene that this was an interstitial deletion. In a second example, – for example, in the genes SHANK3 w61, 62x, a simple 7p15 deletion emerged as a complex aberration CNTNAP2 w63x and CACNA1C w64x – make up another with an additional deletion in 3q25. 3–4%. However, in ;90% of ASD patients the aetiology Another important application of aCGH is characteris- remains unknown. This group can be further subdivided ation of an additional chromosome of unknown origin into complex autism with additional symptoms such as (marker chromosome). In many cases, even time-con- DF and essential autism in the absence of any further suming efforts to characterise the marker chromosome symptoms w65x. with different FISH probes and staining techniques do In recent years it has been demonstrated that de novo not lead to an identification. In prenatal cases, identifi- deletions and duplications play an important role in the cation of a marker chromosome is particularly important cause of autism in general and specifically in complex because precise characterisation can be the decisive autism. In one of the first aCGH studies in this patient factor in providing a more accurate prognosis. The same group, Jaquemont et al. w66x found pathogenic imbal- is true for de novo translocations that appear to be bal- ances in 8 out of 27 individuals with complex autism. anced. De Gregori et al. w57x analysed 14 foetuses with However, even in essential autism, aCGH can detect rel- a de novo translocation using oligo arrays and demon- evant imbalances in up to 10% of cases w67x. It can be strated that all translocations were balanced, even at assumed that the increase in resolution and the broader high resolution. In contrast, for clinically conspicuous application of aCGH will further increase the detection of patients with a karyotypically balanced reciprocal trans- imbalances, leading to the identification of new disease- location, imbalances were detected in 40% of patients relevant genes. Article in press - uncorrected proof

264 Heinrich et al.: Array CGH

Complex diseases complex disorders. This high detection rate and falling array costs will gradually lead to replacement of conven- Owing to the high frequency of many CNVs and their tional cytogenetics as frontline analysis in these patients. gene content (CNVs are especially enriched in genes With a growing understanding of the nature of CNVs, it involved in immune response, xenobiotica metabolism, is likely that aCGH will be used in prenatal diagnostics in cell adhesion, and sensory perception w68x) it was soon the near future. Another important field of application is hypothesised that CNVs, similar to SNPs, are involved in tumour diagnostics, for which aCGH will lead to better the aetiology of complex diseases. Accordingly, there is classification and more accurate prognosis of different a growing list of common diseases associated with CNV tumours w77x. w69x. For example, CNV of the CCL3L1 gene is associ- ated with variability in susceptibility to HIV-1 infection w70x or rheumatoid arthritis w71x. Similarly, FCGR3B CNVs References predispose to glomerulonephritis w72x and some auto- immune disorders such as systemic lupus erythe- matosus, microscopic polyangiitis and Wegener 1. Tjio JH, Levan A. The chromosome number of man. Here- ditas 1956;42:1–6. granulomatosis w73x. The use of new high-resolution 2. Pinkel D, Straume T, Gray JW. Cytogenetic analysis using chips covering all known CNV regions will lead to a deep- quantitative, high-sensitivity, fluorescence hybridization. er understanding of the association between CNV and Proc Natl Acad Sci USA 1986;83:2934–8. common diseases. 3. Flint J, Wilkie AO, Buckle VJ, Winter RM, Holland AJ, McDermid HE. The detection of submicroscopic chromo- Prenatal diagnosis somal rearrangements in idiopathic mental retardation. Nat Genet 1995;9:132–40. Despite many limitations, conventional cytogenetics is 4. Flint J, Knight SJ. The use of telomere probes to investigate submicroscopic rearrangements associated with mental still the gold standard in the detection of foetal anoma- retardation. Curr Opin Genet Dev 2003;13:310–6. lies. These techniques are limited by failure to cultivate 5. Ravnan JB, Tepperberg JH, Papenhausen P, Lamb AN, cells, long reporting times, low-resolution chromosome Hedrick J, Eash D, et al. Subtelomere FISH analysis of banding and culture artefacts. In contrast, aCGH does 11688 cases: an evaluation of the frequency and pattern of not require cell cultivation, has short reporting times and subtelomere rearrangements in individuals with develop- allows higher-resolution analysis. Further advantages of mental disabilities. J Med Genet 2006;43:478–89. aCGH are better detection of mosaicism (lower limit 6. Solinas-Toldo S, Lampel S, Stilgenbauer S, Nickolenko J, ; Brennert A, Do¨ hner H, et al. Matrix-based comparative 15%), identification of all known microdeletion/micro- genomic hybridization: biochips to screen for genomic w x duplication syndromes and automation 74 . For exam- imbalances. Genes Chromosomes Cancer 1997;20:399– ple, Le Caignec et al. w75x found eight rearrangements 407. (16%) in 49 foetuses with multiple malformations, with at 7. Pinkel D, Segraves R, Sudar D, Clark S, Poole I, Kowbel D, least five of these considered as causative. In another et al. High resolution analysis of DNA copy number variation study of 300 unselected prenatal diagnoses, 15/58 using comparative genomic hybridization to microarrays. Nat Genet 1998;20:207–11. imbalances detected (5%) were pathologic, whereas w x 8. Heinrich U, Locher M, Hinrichsen T, Sendelbach K, Polten three CNVs (1%) were of uncertain significance 76 . A, Rost I. Validation of BAC-array CGH using a different In spite of the advantages, the possibility of a report array platform. Eur J Hum Genet 2008;16(Suppl 2):125. with uncertain clinical significance is a great problem for 9. Heinrich U, von Voss H, Rost I. Array CGH in der Routine- two reasons. First, phenotypic information on the foetus diagnostik bei geistiger Entwicklungssto¨ rung. J Lab Med is only accessible in a very restricted way by ultrasound, 2008;32:298–307. in contrast to postnatal cases. Second, providing a prog- 10. Barber JC. Directly transmitted unbalanced chromosome abnormalities and euchromatic variants. J Med Genet nosis in prenatal diagnostics is very important in enabling 2005;42:609–29. the parents to decide on whether to terminate the preg- 11. Rauch A, Hoyer J, Guth S, Zweier C, Kraus C, Becker C, et nancy. In his conclusions, Friedman w74x lists only two al. Diagnostic yield of various genetic approaches in currently valid indications for aCGH in prenatal diagnosis: patients with unexplained developmental delay or mental pregnancies at high risk for a pathogenic CNV, such as retardation. Am J Med Genet A 2006;140:2063–74. a de novo translocation/inversion, or multiple malforma- 12. Sebat J, Lakshmi B, Troge J, Alexander J, Young J, Lundin tions detected sonographically. An additional scenario P, et al. Large-scale copy number polymorphism in the human genome. Science 2004;305:525–8. includes aCGH being offered in a clinical trial. 13. Iafrate AJ, Feuk L, Rivera MN, Listewnik ML, Donahoe PK, Qi Y, et al. Detection of large-scale variation in the human genome. Nat Genet 2004;36:949–51. Conclusions 14. Redon R, Ishikawa S, Fitch KR, Feuk L, Perry GH, Andrews TD, et al. Global variation in copy number in the human genome. Nature 2006;444:444–54. Introduction of aCGH into routine clinical diagnostics has 15. Van Ravenswaaij-Arts CMA, Kleefstra T. Emerging micro- substantially improved the detection rate of chromoso- deletion and microduplication syndromes; a counseling par- mal imbalances in patients with MR/DD, MCA, ASD and adigm. Eur J Med Genet 2009;52:75–6. Article in press - uncorrected proof

Heinrich et al.: Array CGH 265

16. Lee C, Iafrate AJ, Brothman AJ. Copy number variations 34. Ravnan JB, Tepperberg JH, Papenhausen P, Lamp AN, and clinical cytogenetic diagnosis of constitutional disor- Hedrick J, Eash D, et al. Subtelomere FISH analysis in ders. Nat Genet 2007;39:S48–54. 11,688 cases: an evaluation of the frequency and pattern of 17. Hojer J, Dreweke A, Becker C, Go¨ hring I, Thiel CT, Peippo subtelomere rearrangements in individuals with develop- M, et al. Molecular karyotyping in patients with mental retar- mental disabilities. J Med Genet 2006;43:478–89. dation using 100K single-nucleotide polymorphism arrays. 35. Menten B, Maas, N, Thienpont B, Buysse K, Vandesompele J Med Genet 2007;44:629–33. J, Melotte C, et al. Emerging patterns of cryptic chromo- 18. Slavotinek AM. Novel microdeletion syndromes detected by somal imbalance in patients with idiopathic mental retar- chromosome microarrays. Hum Genet 2008;124:1–17. dation and multiple congenital anomalies: a new series of 19. Zahir F, Firth HV, Baross A, Delaney AD, Eydoux P, Gibson 140 patients and review of published reports. J Med Genet WT, et al. Novel deletions of 14q11.2 associated with devel- 2006;43:625–33. opmental delay, cognitive impairment and similar anomalies 36. Lu XL, Shaw CA, Patel A, Li J, Cooper ML, Wells WR, et al. in three children. J Med Genet 2007;44:556–61. Clinical implementation of chromosomal microarray analy- 20. Filges I, Ro¨ thlisberger B, Noppen C, Boesch N, Wenzel F, sis: summary of 2513 postnatal cases. PLoS ONE 2007; Necker J, et al. Familial 14.5 Mb interstitial deletion 2:e327. 13q21.1–13q21.33: clinical and array CGH study of a 37. Fan YS, Jayakar P, Zhu H, Barbouth D, Sacharow S, Mora- benign phenotype in a three-generation family. Am J Med les A, et al. Detection of pathogenic gene copy number var- Genet A 2009;149A:237–41. iation in patients with mental retardation by genomewide 21. Feuk L, Carson AR, Scherer SW. in the array comparative genomic hybridization. Hum Mutat human genome. Nat Genet 2006;7:85–97. 2007;28:1124–32. 22. Somerville MJ, Mervis CB, Young EJ, Seo EJ, del Campo 38. de Vries BB, Pfundt R, Leisink M, Koolen DA, Vissers LE, M, Bamforth S, et al. Severe expressive-language delay Janssen IM, et al. Diagnostic genome profiling in mental related to duplication of the Williams-Beuren locus. N Engl retardation. Am J Hum Genet 2005;77:606–16. J Med 2005;353:1694–701. 39. Mefford H, Sharp A, Baker C, Itsara A, Jiang Z, Buysse K, 23. Mefford H, Sharp A, Baker C, Itsara A, Jiang Z, Buysse K, et al. Recurrent rearrangements of chromosome 1q21.1 and et al. Recurrent rearrangements of chromosome 1q21.1 and variable pediatric phenotypes. N Engl J Med 2008;359: variable pediatric phenotypes. N Engl J Med 2008;359: 1685–99. 1685–99. 40. Brunetti-Pieri N, Berg JS, Scaglia F, Belmont J, Bacino CS, 24. Lisi EC, Hamosh A, Doheny KF, Squibb E, Jackson B, Galc- Sahoo T, et al. Recurrent reciprocal 1q21.1 deletions and zynski R, et al. 3q29 interstitial microduplication: a new duplications associated with microcephaly and develop- syndrome in a three-generation family. Am J Med Genet mental and behavioural abnormalities. Nat Genet 2008;40: 2008;146:601–9. 1466–71. 25. Hatchwell E. Monozygotic twins with chromosome 22q11.2 41. Willat L, Cox J, Barber J, Cabanas ED, Collins A, Donnai D, deletion and discordant phenotype. J Med Genet 1996; et al. 3q29 microdeletion syndrome: clinical and molecular 33:261. characterization of a new syndrome. Am J Hum Genet 26. Flipsen-ten Berg K, van Hasselt PM, Eleveld MJ, van der 2005;77:154–60. Wijst SE, Hol FA, de Vroede MA, et al. Unmasking of a hemi- 42. Sharp A, Mefford HC, Li K, Baker C, Skinner C, Stevenson zygous WFS1 gene mutation by chromosome 4p deletion RE, et al. A recurrent 15q13.3 microdeletion syndrome of 8.3 Mb in a patient with Wolf-Hirschhorn syndrome. Eur associated with mental retardation and seizures. Nat Genet J Hum Genet 2007;15:1132–8. 2008;40:322–8. 27. Liburd N, Ghosh M, Riazuddin S, Naz S, Khan S, Ahmed Z, 43. Sharp AJ, Selzer RR, Veltman JA, Gimelli S, Gimelli G, Stria- et al. Novel mutations of MYO15A associated with profound no P, et al. Characterization of a recurrent 15q24 microde- deafness in consanguineous families and moderately severe letion syndrome. Hum Mol Genet 2007;16:567–72. hearing loss in a patient with Smith-Magenis syndrome. 44. Bellanne´ -Chantelot C, Clauin S, Chauveau D, Collin P, Dau- Hum Genet 2001;109:535–41. mont M, Douillard C, et al. Large genomic rearrangements 28. Bergemann AD, Cole F, Hirschhorn K. The etiology of Wolf- in the hepatocyte nuclear factor-1 beta (TCF2) gene are the Hirschhorn syndrome. Trends Genet 2005;21:188–95. most frequent cause of maturity-onset diabetes of the 29. Marlin S, Blanchard S, Slim R, Lacombe D, Denoyelle F, young type 5. Diabetes 2005;54:3126–32. Alessandri JL, et al. Townes-Brocks syndrome: detection of 45. Sharp AJ, Hansen S, Selzer RR, Cheng Z, Regan R, Hurst a SALL1 mutation hot spot and evidence for a position effect in one patient. Hum Mutat 1999;14:377–86. JA, et al. Discovery of previously unidentified genomic dis- 30. Cotton RGH, Auerbach AD, Axton M, Barash CI, Berkovic orders from duplication architecture of the human genome. SF, Brooks AJ, et al. The human variome project. Science Nat Genet 2006;38:1038–42. 2008;322:861–2. 46. Shaw-Smith C, Pittman AM, Willatt L, Martin H, Rickman L, 31. Curry CJ, Stevenson RE, Aughton D, Byrne J, Carey JC, Gribble S, et al. Microdeletion encompassing MAPT at chro- Cassidy S, et al. Evaluation of mental retardation: recom- mosome 17q21.3 is associated with developmental delay mendations of a consensus conference: American College and learning disability. Nat Genet 2006;38:1032–7. of Medical Genetics. Am J Med Genet 1997;72:46–77. 47. Koolen DA, Vissers LE, Pfundt R, de Leeuw N, Knight SJ, 32. Shevell M, Ashwal S, Donley D, Flint J, Gingold M, Hirtz D, Regan R, et al. A new chromosome 17q21.31 microdeletion et al. Practice parameter: evaluation of the child with global syndrome associated with a common inversion polymor- developmental delay: report of the Quality Standards Sub- phism. Nat Genet 2006;38:999–1001. committee of the American Academy of Neurology and the 48. Somerville MJ, Mervis CB, Young EJ, Seo EJ, del Campo Practice Committee of the Child Neurology Society. Neu- M, Bamforth S, et al. Severe expressive-language delay rology 2003;61:367–80. related to duplication of the William-Beuren locus. N Engl J 33. Flint J, Knight S. The use of telomere probes to investigate Med 2005;353:1694–701. submicroscopic rearrangements associated with mental 49. Weiss LA, Shen Y, Korn JM, Arking DE, Miller DT, Fossdal retardation. Curr Opin Genet Dev 2003;13:310–6. R, et al. Association between microdeletion and microdu- Article in press - uncorrected proof

266 Heinrich et al.: Array CGH

plication at 16p11.2 and autism. N Engl J Med 2008;358: 63. Strauss KA, Puffenberger EG, Huentelman MJ, Gottlieb S, 667–75. Dobrin SE, Parod JM, et al. Recessive symptomatic focal 50. Potocki L, Chen KS, Park SS, Osterholm DE, Withers MA, epilepsy and mutant contactin-associated protein-like 2. N Kimonis V, et al. Molecular mechanism for duplication Engl J Med 2006;354:1370–7. 17p11.2 – the homologous recombination reciprocal of the 64. Splawski I, Timothy KW, Sharpe LM, Decker N, Kumer P, Smith-Magenis microdeletion. Nat Genet 2000;24:84–7. Bloise R, et al. Ca(V)1.2 calcium channel dysfunction causes 51. Van Esch H, Bauters M, Ignatius J, Jansen M, Raynaud M, a multisystem disorder including arrhythmia and autism. Cell Hollanders K, et al. Duplication of the MECP2 region is a 2004;119:19–31. frequent cause of severe mental retardation and progressive 65. Miles JH, Takahashi TN, Bagby S, Sahota P, Vaslow DF, neurological symptoms in males. Am J Hum Genet 2005; Wang CH, et al. Essential versus complex autism: definition 77:442–53. of fundamental prognostic subtypes. Am J Med Genet A 52. Hochstenbach R, van Binsbergen E, Engelen J, Nieuwint A, 2005;135:171–80. Polstra A, Poddighe P, et al. Array analysis and karyotyping: 66. Jaquemont ML, Sanlaville D, Redon R, Raoul O, Cormier- workflow consequences based on a retrospective study of Daire V, Lyonnet S, et al. Array-based comparative genomic 36,325 patients with idiopathic developmental delay in the hybridisation identifies high frequency of cryptic chromo- Netherlands. Eur J Med Genet 2009;doi:10.1016/ somal rearrangements in patients with syndromic autism j.ejmg.2009.03.015. spectrum disorders. J Med Genet 2006;43:843–9. 53. Coco R, Penchaszadeh VB. Cytogenetic findings in 200 67. Sebat J, Lakshmi B, Malhotra D, Troge J, Lese-Martin C, children with mental retardation and multiple congenital Walsh T, et al. Strong association of de novo copy number anomalies of unknown cause. Am J Med Genet 1982;12: mutations with autism. Science 2007;316:445–9. 155–73. 68. Cooper GM, Nickerson DA, Eichler EE. Mutational and 54. Lee C, Iafrate AJ, Brothman AR. Copy number variations selective effects on copy-number variants in the human and clinical cytogenetic diagnosis of constitutional disor- genome. Nat Genet 2007;39:S22–9. ders. Nat Genet 2007;39:S48–54. 69. Estivill E, Armengol L. Copy number variants and common 55. Thienpont B, Mertens L, de Ravel T, Eyskens B, Boshoff D, disorders: filling the gaps and exploring complexity in Maas N, et al. Submicroscopic chromosomal imbalances genome-wide association studies. PLoS Genet 2007;3: detected by array CGH are a frequent cause of congenital e190. heart defects in selected patients. Eur Heart J 2007;28: 70. Gonzales E, Kulkarni H, Bolivar H, Mangano A, Sanchez R, 2778–84. Catano G, et al. The influence of CCL3L1 gene-containing 56. Erdogan F, Larsen LA, Zhang L, Tu¨ mer Z, Tommerup N, segmental duplication on HIV-1/AIDS susceptibility. Science Chen W, et al. High frequency of submicroscopic genomic 2005;307:1434–40. aberrations detected by tiling path array comparative 71. McKinney C, Merriman ME, Chapman PT, Gow PJ, Harrison genome hybridisation in patients with isolated congenital AA, Highton J, et al. Evidence for an influence of chemo- kine-ligand 3-like 1 (CCL3L1) gene copy number on sus- heart disease. J Med Genet 2008;45:704–9. ceptibility to rheumatoid arthritis. Ann Rheum Dis 2008; 57. De Gregori, Ciccone R, Magini P, Pramparo T, Gimelli S, 67:409–13. Messa J, et al. Cryptic deletions are a common finding in 72. Aitman TJ, Dong R, Vyse TJ, Norsworthy PJ, Johnson MD, ‘‘balanced’’ reciprocal and complex chromosome rearrange- Smith J, et al. Copy number polymorphism in Fcgr3 predis- ments: a study of 59 patients. J Med Genet 2007;44:750– poses to glomerulonephritis in rats and humans. Nature 62. 2006;439:851–5. 58. Fombonne E. Epidemiology of autistic disorder and other 73. Fanciulli M, Norsworthy PJ, Petretto E, Dong R, Harper L, pervasive developmental disorders. J Clin Psychiatry 2005; Kamesh L, et al. FCGR3B copy number variation is asso- 66(Suppl 10):3–8. ciated with susceptibility to systemic, but not organ-specif- 59. Abrams BS, Geschwind DH. Advances in autism genetics: ic, autoimmunity. Nat Genet 2007;39:721–3. on the threshold of a new neurobiology. Nat Genet Rev 74. Friedman JM. High-resolution array genomic hybridization 2008;9:341–55. in prenatal diagnosis. Prenat Diagn 2009;29:20–8. 60. Marshall CR, Noor A, Vincent JB, Lionel AC, Feuk L, Skaug 75. Le Caignec C, Boceno M, Saugier-Veber P, Jaquemont S, J, et al. Structural variation of chromosomes in autism spec- Joubert M, David A, et al. Detection of genomic imbalances trum disorder. Am J Hum Genet 2008;82:477–88. by array based comparative genomic hybridisation in fetus- 61. Durand CM, Betancur C, Boeckers TM, Bockmann J, es with multiple malformations. J Med Genet 2005;42: Chaste P, Fauchereau F, et al. Mutations in the gene encod- 121–8. ing the synaptic scaffolding protein SHANK3 are associated 76. Van den Veyver IB, Patel A, Shaw CA, Pursley AN, Kang with autism spectrum disorders. Nat Genet 2007;39:25–7. SH, Simovich MJ, et al. Clinical use of array comparative 62. Moessner R, Marshall CR, Sutcliffe JF, Skaug J, Pinto D, genomic hybridization (aCGH) for prenatal diagnosis in 300 Vincent J, et al. Contribution of SHANK3 mutations to cases. Prenat Diagn 2009;29:29–39. autism spectrum disorder. Am J Hum Genet 2007;81:1289– 77. Kallioniemi A. CGH microarrays and cancer. Curr Opin Bio- 97. technol 2008;19:36–40.