Application of array-based whole genome scanning technologies as a cytogenetic tool in haematological malignancies


Jaroslaw P. Maciejewski, MD, PhD, Taussig Cancer Institute/R40, 9500 Euclid Avenue, Cleveland, OH 44195, USA. E-mail:


Karyotypic analysis provides useful diagnostic information in many haematological malignancies. However, standard metaphase cytogenetics has technical limitations that result in the underestimation of the degree of chromosomal changes. Array-based technologies can be used for karyotyping and can supplant some of the shortcomings of metaphase cytogenetics, and include single nucleotide polymorphism arrays (SNP-A) and comparative genomic hybridization arrays (CGH-A). Array-based cytogenetic tools do not rely on cell division, have superb resolution for unbalanced lesions and allow for the detection of copy number-neutral loss of heterozygosity, a type of lesion not seen with metaphase cytogenetics. Moreover, genomic array analysis is automated and results can be objectively and systematically analysed using biostatistical algorithms. As a potential advantage over genomic approaches, metaphase cytogenetics can detect balanced chromosomal defects and resolves clonal mosaicism. Initial studies performed in various haematological malignancies indicate the potential of SNP-A-based karyotyping as a useful clinical cytogenetic detection tool. The current effort is aimed at developing rational diagnostic algorithms for the detection of somatic defects and the establishment of clinical correlations for novel SNP-A-detected chromosomal defects, including acquired somatic uniparental disomy. SNP-A can complement metaphase karyotyping and will probably play an important role in clinical cytogenetic diagnostics.

Routine and new cytogenetic technologies

Metaphase cytogenetics

Metaphase cytogenetics has become a routine test in the management of haematological malignancies in which the presence of specific chromosomal translocations is diagnostic, while, for other diseases, specific chromosomal aberrations are highly predictive of prognosis or responsiveness to targeted therapeutics. Invariant aberrations serve as clonal markers to detect and follow minimal residual disease and relapse. Moreover, metaphase cytogenetics facilitates the mapping of recurrent lesions to delineate minimally affected regions. However, metaphase cytogenetics is time consuming and technically demanding; its yield is related to the proportion of clonal cells in the tested sample (sensitivity) and size of the lesion (resolution). The sensitivity is relatively low; traditionally 2 abnormal metaphases of 20 tested are considered pathological. The resolution depends on the location of the lesion with regard to the banding pattern. The need for cellular proliferation to obtain chromosomal spreads is a limitation; metaphase cytogenetics determines the proportion of abnormal cells within the dividing progenitor pool and may not always correlate with the total percentage of malignant cells, a fact that may account for the quantitative differences between various cytogenetic methods. Due to its inherent limitations, it is likely that when metaphase cytogenetics is applied in the study of haematological malignancies many chromosomal defects remain undetected.

Complementary cytogenetic techniques

Fluorescence in situ hybridization (FISH) is widely applied in cytogenetics, in particular to precisely diagnose reciprocal translocations. Currently, polymerase chain reaction (PCR)-based methods are often utilized for detection of known fusion genes, reducing the need for cytogenetic analysis and allowing for better quantitative follow-up analyses. Due to the high background and inherent lower precision of FISH, its role in the detection of unbalanced defects is less well defined and identification of very small numbers of abnormal cells is of unclear clinical significance.

Novel array-based technologies

Various DNA array-based technologies have been introduced to facilitate the examination of the normal and malignant genome. Based on the availability of bacterial artificial chromosome (BAC) libraries, arrays with various densities of BAC probes have been generated, enabling array-based comparative genomic hybridization. The subsequent introduction of high-density oligonucleotide arrays (comparative genomic hybridization arrays; CGH-A) has enabled even more precise scanning of the genome for copy number changes. Using a similar microchip technology, single nucleotide polymorphism arrays (SNP-A), developed for whole genome association studies, have also been adopted for karyotyping. Unlike routine cytogenetics, arrays can be performed on interphase cells and consequently even archival samples can be examined (Table I). While these techniques only enable the detection of unbalanced defects and do not allow for the distinction between multiple large clones (clonal mosaicism versus compound lesions), they have superior resolution when compared to metaphase analysis. SNP-A also has the advantage of simultaneous genotyping, enabling detection of copy number-neutral loss of heterozygosity (CN-LOH), also referred as to as somatic uniparental disomy (UPD; see below). In addition to its diagnostic value in cytogenetic diagnostics, array-based karyotyping technologies constitute an excellent chromosome mapping tool, thereby allowing for delineation of boundaries of commonly deleted/duplicated regions.

Table I.   Comparison of metaphase cytogenetics with novel methods of karyotyping using array-based karyotyping. Thumbnail image of


Comparative genomic hybridization arrays rely on the difference in the copy number between differentially labelled test and reference DNA samples (Fig 1). Through competition between test (e.g. tumour) and control diploid DNA, imbalances due to copy number differences result in a shift in the fluorescence spectra. The ability to compare the hybridization signals of test and control DNA affords a high level of precision and exclusion of artefacts, a clear advantage of CGH-A over SNP-A. High-density and very precise oligo-CGH-A platforms are now available from Agilent (Barrett et al, 2004) or NimbleGen (Nuwaysir et al, 2002; Albert et al, 2003; Selzer et al, 2005). In contrast to SNP-A, CGH-A enables even or targeted distribution of probes, including areas of known copy number variants (CNVs) (Tan et al, 2007), but does not allow for detection of UPD (Table I). Of note is that using ‘standard’ control DNA, CGH-A allows for detection of germ line CNVs for which the test sample varies from the control DNA. Application of paired germ line DNA and tumour DNA from the same individual would allow for exclusion of any germ line artefacts and differences would only reflect somatic lesions.

Figure 1.

 Principles of array-based karyotyping tools. Upper panel: principle of CGH-A. The left portion of the figure depicts the work-flow. On the right, examples of data output for two exemplary chromosomes using two different density arrays are shown. Lower panel: principles of SNP-A based karyotyping. The left portion of the figure depicts the work-flow starting with fragmentation and labelling of DNA (various biochemical techniques are currently applied, see description in text). At the end of the procedure, DNA is hybridized to oligonucleotide probes corresponding to the individual alleles. Hybridization signals are read, recorded and translated into genotyping calls. Various software packages allow for generation of karyotyping maps. In the example shown, Genotyping Console (Affymetrix) was used to analyse Affymetrix 6.0 array results. Small blue dots represent individual genotyping calls either in homozygous constellation for minor or major alleles or as heterozygous calls. This information allows for assessment of LOH but hybridization intensity can also be used to determine copy number changes (blue oscillating line).


Single nucleotide polymorphism arrays rely on oligonucleotide probes corresponding to the allelic variants of selected SNPs. Hybridization of genomic DNA to both probe variants indicates heterozygosity, while a signal for only one allele is consistent with homo/hemizygosity at any given locus. In addition, the strength of fluorescence emitted from individual probes allows for the analysis of gene copy number. As SNPs are not evenly distributed across the genome, coverage of some chromosomal regions is not possible. Several generations of chips varying in density of probes have been developed, successively resulting in increased analytic precision. The most common SNP-A platforms include Illumina (Gunderson et al, 2005) and Affymetrix (Syvanen, 2005) arrays, which utilize bead or chip technology, respectively. In the Affymetrix technology, genomic DNA is digested by restriction endonucleases, amplified and labelled. In bead-based platforms, the whole genome amplification and fragmentation steps are followed by hybridization to an oligonucleotide bead array. One of two bead types correspond to each allele in the SNP locus and allelic specificity is conferred by enzymatic (allele-specific primer or single base) extension and fluorescent staining. For both array types, the read-out includes genotyping calls and hybridization signal strength, corresponding to gene copy number. Final analysis is performed using various biostatistical and genetic software packages. The CNAG (Copy Number Analyser for GeneChips®) programme (Nannya et al, 2005; Yamamoto et al, 2007) combines copy number analysis and LOH and analytic programmes are available allowing determination of the overlap between lesions with known CNVs (Fig 1, lower panel).

A major advantage of SNP-A over metaphase cytogenetics and CGH-A is the ability to detect diploid stretches of homozygosity present throughout the genome (Fig 2). They can be a result of acquired somatic UPD, autozygosity or early embryonic UPD. While significant autozygosity inherited from both parents is non-clonal and unlikely to be clinically relevant for haematological malignancies, UPD can result from errors during mitosis leading to both copies of a chromosome or chromosomal region being derived from one parent. Acquired somatic UPD can be due to segmental deletions and subsequent replacement of the lost fragment by a copy of the remaining allele or mitotic recombination.

Figure 2.

 Examples of somatic UPD detected by SNP-A. (A) UPD7q analysed using 250 K Affymetrix SNP-A and CNAG v3. Red dots represent individual genotyping calls. Red and green lines correspond to the individual alleles. In UPD, one of the parental chromosomes is duplicated while the other is lost, as indicated by the deviation from the haploid signal intensity line. (B) The same lesion analysed using 6.0 Affymetrix SNP-A and Genotyping Console software. (C) UPD13, found using 6.0 Affymetrix SNP-A in bone marrow cells (upper panel) but not in sorted, non-clonal CD3+ T cells (lower panel).

Analytic principles of array-based karyotyping


The level of resolution depends on the density of the arrays, distribution of probes and biostatistical algorithms used to detect copy number changes. Theoretically, the high density arrays (e.g. 500 K probes) with an average of 10 probes per call allow for a resolution of around 50 Kb, but arrays with even higher numbers of probes can decrease the minimal detectable size of a deletion to around 25 Kb (Tan et al, 2007). Higher density arrays provide a greater resolution, an aspect in which CGH-A has been superior. Because in CGH-A probe spacing does not depend upon the location of SNPs, probes can be evenly distributed (e.g. every 6 Kb in 385 K arrays). Further targeting of individual chromosomes by custom CGH-A allows for an even higher resolution and mapping of break points within 5 Kb intervals (Carter, 2007). However, in SNP-A, resolution has been recently increased by inclusion of sequential oligonucleotide probes that target CNVs.

In SNP-A, LOH is identified through a lower than expected frequency of heterozygous calls located serially along the chromosome. For any given number of SNPs, multiple heterozygous calls would be expected. Thus, the probability of LOH detection increases with the length of homozygous stretch. Assessment of LOH by genotyping (frequency of heterozygous calls) and determination of hybridization intensity reduces the variability of results (Lamy et al, 2006; Li et al, 2007). Nevertheless, there is an expected variability in the minimal size of the area of LOH detectable, ranging from regions of >200 Kb to as low as 50 Kb (Li et al, 2007). SNP-A allows for assessment of CN-LOH (Tan et al, 2007) when stretches of homozygosity are associated with diploid hybridization intensity. In contrast, LOH due to deletion shows both a decrease in heterozygous call frequency and gene copy number.

Distinction of somatic lesions

A variety of algorithms have been developed that allow for the bioanalytic reduction in experimental variations within regions with similar copy numbers (Lamy et al, 2006; Carter, 2007; Guttman et al, 2007; Yamamoto et al, 2007). While very large invariant lesions detected by SNP-A or CGH-A are unlikely to be inherited, comparison with a DNA sample containing a germ line configuration is imperative to exclude normal CNVs and distinguish them from truly somatic defects. We and others have developed rational diagnostic algorithms aimed at minimizing false positive results (see below). For diagnostic applications, additional analysis of non-clonal cells may be expensive and is not often viable as appropriate control cells have to be selected. For example, in the case of myeloid malignancies, sorted non-clonal lymphocytes may be used as a source of germ line DNA. While buccal swabs are often used as germ line reference, leucocytes often contaminate mucosa, in particular when the white cell counts are high; contamination can be easily demonstrated by performing reverse transcription polymerase chain reaction (RT-PCR) for CD45 on buccal samples. Therefore, buccal swabs do not constitute the best source of control DNA and in an ideal case, skin fibroblasts, preferably cultured, should be used.

While paired germ line/tumour analysis constitutes a desired gold standard, it is possible to minimize its need in the context of particular circumstances. First, evidence for the clonal nature of the defects can be derived from a simple analysis of the data output (Fig 3, upper panel). In men, a physiological lack of paired X chromosomes may serve as a guide for the assessment of hybridization intensity should a haploid chromosome set be present in all cells. Comparison to this standard may enable recognition of the loss of chromosomal material (greater than haploid intensity) and lead to the conclusion that there is contamination with non-clonal cells, confirming the somatic nature of the observed defect. A similar principle may be applied while evaluating genotyping calls in conventional deletions and in CN-LOH: germ line lesions show an expected total lack of heterozygous genotyping calls. For somatic defects, while LOH may be easily recognizable, the presence of non-clonal diploid cells introduces contaminating heterozygous calls, highly suggestive that the chromosomal lesion in question is not of germ line origin. Such important details aid the diagnosis of acquired chromosomal defects.

Figure 3.

 Bioanalytic algorithms for exclusion of germ line-encoded stretches of homozygosity and CNVs. Upper panel: distinction of somatic versus germ line LOH through bioanalytic evaluation of data output. The size of the clone carrying the chromosomal defect can be interpreted from various data output patterns, enabling confirmation of a germ line or somatic lesion. Generally, lesions in large clones cannot be distinguished from germ line changes unless they affect specific chromosome types, are of a certain size, located in certain regions of the chromosome or most stringently, confirmed by testing for non-clonal DNA. We have proposed an algorithm that can help alleviate the need for paired testing in every case. The smaller size (smaller proportion of cells affected in the sample) of the pathological clone carrying a chromosomal lesion can be identified by the presence of residual heterozygous SNP calls and smaller than expected decrease in the hybridization density stemming from the contamination with diploid non-clonal cells. In such a situation, it is highly unlikely that the lesion seen is clonal as it would generate a perfect result including lack of heterozygous alleles and an adequate decrease in the copy number as seen when X chromosome is analysed in males. Lower panel: proposed diagnostic algorithm applied to clinical specimens.

The distinction between acquired somatic UPD and an inherited form of homozygosity can be best accomplished through comparison of DNA from the malignant tissue with remission or constitutional DNA although, to a degree, it can perhaps be inferred from the size and location of the affected region. Due to technical reasons, biological variability and imperfection of current biostatistical algorithms, stretches of possible UPD are frequently encountered in control samples. However, a high number of consecutive SNPs in a homozygous configuration suggest a true germ line UPD. Such UPD events do not represent clonal lesions, can be spurious or have an early embryonic origin. In some individuals, large regions can be affected by UPD; 15% of Hapmap controls showed areas of homozygosity, all smaller than 2 Mb. UPD regions >2 Mb were found only in 12% (Mohamedali et al, 2007; Gondek et al, 2008). Our analysis of large cohorts of healthy individuals show that current algorithms detect stretches of LOH distributed across the genome in around 8% of controls. The size and location of these areas constitute the best criteria for their distinction from acquired UPD; on average their size is 8·7 Mb and most are interstitial. Thus, any area of CN-LOH < 25 Mb [95% confidence interval (CI)] that does not involve the telomere can be excluded from the analysis of clonal lesions, as it is most likely present in the germ line. Any lesions that fall outside these parameters should be confirmed as clonal by studying non-clonal DNA. In contrast, large and/or telomeric defects do not require confirmation as they do not occur in non-clonal control DNA (e.g. UPD9p or UPD7q).


The sensitivity of SNP-A relates to the ability to detect chromosomal lesions when analysed cell populations represent a mixture of clonal and nonclonal cells. Depending on the biostatistical detection algorithm, in dilution studies the presence of c. 25% of abnormal cells with uniform unbalanced lesions may be detectable by 250 K SNP-A (Gondek et al, 2008) and a similar sensitivity is probably achievable with comparably dense CGH-A. Clearly, sorting for the relevant cell population may help as, for example, removal of nonclonal lymphocytes in myelodysplastic syndromes (MDS) may enrich the abnormal myeloid cells harbouring the lesion (Gondek et al, 2008). Metaphase cytogenetics and FISH enable estimation of the contribution of non-clonal cells in the sample and thereby assessment of the clonal size. In SNP-A such an analysis is not possible directly. Detection of copy number changes and CN-LOH allows for extrapolation that the clone is significant in size; at best 20% clonal cell admixture allows for detection of the corresponding clonal lesion. For LOH, a combined analysis of genotyping calls is helpful: 100% of clonal cells would result in an absolute absence of heterozygous calls and a maximal gradient between diploid and haploid signal. However, dilution of clonal with normal cells will lead to various degrees of blurring of the LOH up to its absolute loss, should the clone be small. The sensitivity is highly dependent upon the bioanalytic algorithm used, thus various software packages may provide discrepant results at the detection threshold. In addition, sensitivity for deletions may be higher than that of UPD.

Target cell selection

An important technical question is whether whole bone marrow, blood or purified cell populations should be used for array-based karyotyping. Intuitively, arrays performed on purified CD34+ cells should be more accurate and sensitive but our experience, also reported in other studies (Gondek et al, 2008; Starczynowski et al, 2008), shows that CD34 selection does not lead to significant diagnostic gain. Our studies have shown that the application of whole blood or lymphocyte-depleted blood provides comparable results to those of total bone marrow samples. Separation of mononuclear cells may not be useful as it may enrich for non-clonal lymphocytes and depletes myeloid cells that may be derived from the malignant clone. In contrast, if mature neutrophils are produced by normal cells while the clonal cells are mostly immature and contained in mononuclear cell population, using whole bone marrow or blood may decrease the sensitivity of detection and lead to discrepancies between array-based karyotyping and metaphase cytogenetics.

Diagnostic applications

Diagnostic algorithms

In most clinical applications, either total or mononuclear cells are used as a source of DNA. In general, SNP-A and CGH-A show a good concordance with metaphase cytogenetics for detection of previously known unbalanced chromosomal defects and may also allow for identification of lost chromosomal material, for example, metaphase-detected monosomy 17 may in fact represent a pure deletion of 17p alone as the 17q material may be translocated to other chromosomes and escape detection by traditional means (Jasek et al, 2008).

For clinical karyotyping platforms, increasing probe density to 250–500 K does not appear to result in a higher detection rate for microdeletions or UPD (Mohamedali et al, 2007); when Affymetrix 6.0 (Affymetrix, Santa Clara, CA, USA; over 900 K SNP probes) and 250 K arrays were compared, there was remarkable concordance in the diagnostic yield and additional defects were found using higher density arrays (Huh et al, 2008). Thus, very high-density arrays add significant cost but may provide only a marginal (if any) diagnostic gain. However, due to a more even/dense distribution of probes, they enable a more precise mapping of chromosomal defects and CNVs.

The diagnostic rules for clinical cytogenetic diagnosis and investigative applications of arrays may vary, but in both applications, the distinction of artefacts and germ line changes from somatic clonal lesions is of great importance. In constitutional CGH-A or SNP-A, analysis to validate aberrations by alternative approaches, such as FISH, is usually required. Confirmation of novel somatic chromosomal defects detected by array technologies may be important, but this requirement may depend on the size of the defect location. We proposed that if metaphase cytogenetics and SNP-A show a concordant result with regard to a defect, no further analysis of germ line lesions are needed for that particular site (Fig 3, lower panel). In addition, in the case of a non-informative metaphase cytogenetics examination, very large recurrent deletions or gains are unlikely to be constitutional. Should array karyotyping reveal microdeletions and gains not detectable by metaphase cytogenetics, the diagnostic algorithm needs to include comparison with databases of known CNVs and internal control samples; such changes can be excluded without a need for testing germ line DNA. Unfortunately, the available reference databases contain only limited sets of controls. It is likely that, with time, such resources will become available but in the meanwhile a laboratory-generated database of controls may be helpful. Some of the currently available software packages provide information about the degree of overlap with known CNVs but the frequency of each CNV in the general population is difficult to assess. Depending on the size of the control cohort, CNVs with a finite frequency can be identified and excluded. The remaining defects, if diagnostically significant (e.g. due to their location in important chromosomal areas), should be confirmed as somatic.

Results of clinical application

Recently, the application of array-based karyotyping technologies has been described for various haematological malignancies (Table II). These studies utilized arrays with increasing densities as they became available, including 10, 50, 250, 500 K and Affymetrix 6.0 arrays containing almost two million SNP and CNV probes as well as various densities of BAC or oligonucleotide CGH arrays. Target diseases included multiple myeloma (MM) (Gutierrez et al, 2004; Walker et al, 2006), chronic lymphocytic leukaemia (CLL) (Tyybakinoja et al, 2007a), acute lymphoblastic leukaemia (ALL) (Mullighan et al, 2007), acute myeloid leukaemia (AML) (Fitzgibbon et al, 2005; Raghavan et al, 2008; Tiu et al, 2008), MDS (Gondek et al, 2007a,b; Mohamedali et al, 2007) and myeloproliferative syndrome (MPD) (Gondek et al, 2007c) (Table II). Overall, it can be concluded that the addition of array-based karyotyping increases diagnostic yield when combined with routine metaphase cytogenetics (Huh et al, 2008). However, due to the established value of traditional karyotyping, its ability to resolve balanced translocations and a proportion of cases in which array-based approaches fail to identify the lesions detected by metaphase cytogenetics, both traditional and array-based karyotyping should be combined for a comprehensive analysis of the malignant genome rather than replace metaphase analysis with SNP-A or CGH-A-based cytogenetic technologies (Huh et al, 2008; Tiu et al, 2008).

Table II.   Summary of the most important studies of haematological malignancies utilizing CGH-A and SNP-A technologies.
Target disease NMethodGerm line confirmationReference
  1. AML, acute myeloid leukaemia; CLL, chronic lymphocytic leukaemia; MM, multiple myeloma; MDS, myelodysplastic syndrome; MPD, myeloproliferative disease; U, unclassifiable.

AML13SNP-A (10 K)NFitzgibbon et al (2005)
AML27SNP-A (10 K)NRaghavan et al (2008)
AML140SNP-A (250 K and 6.0)YTiu et al (2008)
AML26Array based CGHNTyybakinoja et al (2007a)
ALL242SNP-AYMullighan et al (2007)
CLL70SNP-AYPfeifer et al (2007)
CLL20Array based CGHNTyybakinoja et al (2007b)
MM30SNP-AYWalker et al (2006)
MM74Array based CGHNGutierrez et al (2004)
MDS38Array based CGHNO’Keefe et al (2007)
MDS44Array based CGHYStarczynowski et al (2008)
MDS/MDS-MPD-U174SNP-AYGondek et al (2008)
MDS119SNP-AYMohamedali et al (2007)

In general, the results obtained in clinical applications so far demonstrate a larger complexity of chromosomal defects than judged by metaphase cytogenetics due the presence of previously cryptic lesions. SNP-A also resolved cases in which metaphase cytogenetics was unsuccessful and the overall detection rate of karyotypic aberrations was higher by array analysis. The most common newly identified defects included small deletions and gains of chromosomal material but in a minority of studies, a stringent distinction of somatic defects from CNVs has been provided. Of particular importance are recurrent lesions such as microdeletions involving specific genes present in patients sharing similar phenotypic features. Examples of such defects include microdeletions in 11q spanning CBL (Dunbar et al, 2008), 21q spanning RUNX1 (AML1) or microdeletions in chromosome 4q involving TET2 (Jankowska et al, 2009), among many others under intense investigations.

Acquired somatic UPD is a new type of chromosomal lesion frequently identified by SNP-A (Fig 2). Principally, the pathophysiology of UPD may include various mechanisms that are frequently associated and shared with deletions. Acquired somatic UPD may lead to the duplication of an activating somatic mutation or homozygosity for a disease-prone minor allele present in the germ line DNA. UPD can also result in increased or decreased gene expression due to duplication of a methylation pattern. A similar mechanism may operate in deletion with loss of the inactivated or unmethylated allele. In addition, deletion may result in haploinsufficiency or loss of the intact allele with the remaining allele deficient due to somatic polymorphism or an inactivating somatic mutation.

The first detection of recurrent acquired UPD in haematological disorders was in polycythaemia vera (PV) (Kralovics et al, 2002), which enabled the subsequent discovery of the JAK2 V617F mutation (Kralovics et al, 2005). However, systematic studies revealing a high frequency of acquired somatic UPD involving various chromosomes and precise mapping of regions affected were not performed until SNP-A became available. The utility of SNP-A arrays to detect UPD in AML was initially reported in a study of 60 AML patients using 10 K arrays (Raghavan et al, 2005). SNP-A also enabled the very efficient identification UPD9p (Gondek et al, 2007c; Yamamoto et al, 2007). By analogy to this lesion, UPD in other areas of the genome can indicate the presence of activating mutations in important genes, e.g. homozygous MPL mutations and UPD1p (Szpurka et al, 2009); homozygous mutations in CEBPA (UPD19q) (Fitzgibbon et al, 2005) or UPD13q and FLT3 mutations (Raghavan et al, 2005) (Table III). Recently, UPD11q was shown to harbour biallelic missense mutations in CBL (Dunbar et al, 2008). Similarly, inactivating mutations in a homozygous constellation were seen involving NF1 in patients with juvenile myelomon cytic leukaemia (JMML) with UPD17p (Flotho et al, 2008) and TP53 in AML patients with UPD17q cooperating with deletions of chromosomes 5 and 7 (Jasek et al, 2008) (Table II). The presence of homozygous mutations, such as in cases of AML with FLT3-ITD, may be associated with worse prognostic features as compared to heterozygous genotypes (Whitman et al, 2001).

Table III.   Biallelic mutations coinciding with seqmental UPD.
Area of UPDGeneDiseaseReference
  1. MPD, myeloproliferative disease; (s)AML, (secondary) acute myeloid leukaemia; JMML, juvenile myelomonocytic leukaemia; MDS, myelodysplastic syndrome; CMML, chronic myelomonocytic leukaemia.

UPD9qJAK2MPDKralovics et al (2002, 2005)
UPD13qFLT3-ITDAMLRaghavan et al (2005)
UPD17qNF1JMMLFlotho et al (2008)
UPD17pTP53sAMLJasek et al (2008)
UPD21qRUNX1AMLRaghavan et al (2005)
UPD19qCEBPAAMLFitzgibbon et al (2005); Raghavan et al (2005)
UPD4qKITAMLUnpublished observations
UPD1pMPLMDS/MPD, MPDSzpurka et al (2009)
UPD11pWT1AMLRaghavan et al (2005)
UPD11qCBLCMML, sAMLDunbar et al (2008)

Various studies have attempted to investigate the impact of CGH-A and SNP-A-identified lesions on various clinical outcomes (Mohamedali et al, 2007; Gondek et al, 2008; Starczynowski et al, 2008). Due to the complexity of karyograms and large number of areas involved, such analyses will require very large cohorts of patients. We have described how UPD7 shows a similarly poor prognosis as deletions of the corresponding regions. Similarly, UPD17q (Jasek et al, 2008) and UPD11q (Makishima et al, 2008) have poor prognosis and, if occurring in combination with otherwise known low risk defects, determine the overall prognosis. In a recent series of patients with AML who were serially tested, those who showed residual chromosomal abnormalities following remission induction had shortened relapse-free survival (unpublished observations). Similarly, serial testing also demonstrated that progression to more malignant phenotype is associated with appearance of new additional defects (unpublished observations).


Single nucleotide polymorphism arrays can be adopted as a cytogenetic diagnostic tool allowing for an increase in the diagnostic yield of routine metaphase cytogenetics. The probe density required may vary but it appears that for the diagnosis of recurrent unbalanced lesions and UPD, arrays with around 250–500 K probes may be sufficient. SNP-A and traditional cytogenetic analysis should not be contrasted, as they both fulfil distinct important roles and can complement each other. A combined evaluation of metaphase karyotyping and SNP-A results is needed. With larger cohorts of patients tested, recurrent microdeletions and gains as well as area of acquired somatic UPD will be identified and their impact on clinical outcomes will be delineated and can ultimately be incorporated in current prognostic schemes. While current array-based technologies may be still too expensive for routine applications, several reference laboratories in the United States have started to offer arrays as clinical test and it is likely that in certain diseases this test will become standard, particularly when certain recurrent defects (such as, for example, somatic UPD) will be shown to convey prognostic information. The cost of these technologies is probably going to further decrease due to wider use and automation while metaphase cytogenetics will remain expensive and labour-intense.