Evidence-based genomic diagnosis characterized chromosomal and cryptic imbalances in 30 elderly patients with myelodysplastic syndrome and acute myeloid leukemia

Background To evaluate the clinical validity of genome-wide oligonucleotide array comparative genomic hybridization (aCGH) for detecting somatic abnormalities, we have applied this genomic analysis to 30 cases (13 MDS and 17 AML) with clonal chromosomal abnormalities detected in more than 50% of analyzed metaphase cells. Results The aCGH detected all numerical chromosomal gains and losses from the mainline clones and 113 copy number alterations (CNAs) ranging from 0.257 to 102.519 megabases (Mb). Clinically significant recurrent deletions of 5q (involving the RPS14 gene), 12p12.3 (ETV6 gene), 17p13 (TP53 gene), 17q11.2 (NF1 gene) and 20q, double minutes containing the MYC gene and segmental amplification involving the MLL gene were further characterized with defined breakpoints and gene contents. Genomic features of microdeletions at 17q11.2 were confirmed by FISH using targeted BAC clones. The aCGH also defined break points in a derivative chromosome 6, der(6)t(3;6)(q21.3;p22.2), and an isodicentric X chromosome. However, chromosomally observed sideline clonal abnormalities in five cases were not detected by aCGH. Conclusions Our data indicated that an integrated cytogenomic analysis will be a better diagnostic scheme to delineate genomic contents of chromosomal and cryptic abnormalities in patients with MDS and AML. An evidence-based approach to interpret somatic genomic findings was proposed.


Introduction
The identification of recurrent chromosomal abnormalities in various leukemias and the understanding of molecular defects and pathogenic mechanisms underlying these abnormalities have made cytogenetic analysis valuable in providing diagnostic and prognostic parameters for disease stratification and treatment evaluation [1]. With an average resolution of 6-10 megabases (Mb) on a 300-500 G-band level, conventional karyotyping has been the current standard for screening chromosomal abnormalities on metaphases from direct and cultured bone marrow (BM) and leukemic blood (LB) cells. This approach requires mitotic active cells and frequently encounters difficulties due to the low mitotic index and poor chromosome morphology of leukemic cells. Fluorescence in situ hybridization (FISH) tests using targeted probes to detect gene/locus-specific rearrangements have enhanced the analytical resolution to 300-800 kilobases (Kb) and extended conventional metaphase analysis into interphase cells. Current cytogenetic analysis for patients with myelodysplastic syndrome (MDS) and acute myeloid leukemia (AML) involves cell-based conventional chromosomal analysis and FISH assays using a panel of targeted probes [2,3]. We have previously validated a DNA-based genome-wide oligonucleotide array comparative genomic hybridization (aCGH) for clinical diagnosis of constitutional chromosomal abnormalities and genomic disorders in pediatric patients with mental retardation and developmental delay [4]. The clinical utility of this aCGH based on Agilent's 44K design (CGH4410B) has demonstrated an average analytical resolution of 300-500 Kb and an improved abnormal detection rate from 5-7% by conventional chromosome and FISH analyses to 12% by aCGH [5]. Evidence-based guidelines to interpret genomic findings in the pediatric patients have been proposed [6,7]. Recently, genomewide analyses using BAC-clone aCGH, oligonucleotide aCGH and SNP array have been applied in a research or an exploratory setting to profile the genomic alterations in patients with MDS and AML [8][9][10][11][12][13][14][15]. To evaluate the diagnostic value of aCGH in detecting somatic chromosomal and segmental copy number alterations (CNAs), we have performed aCGH analysis on 30 MDS and AML cases with different clonal abnormalities. The results further characterized the genomic complexity of recurrent chromosomal deletions, duplications, amplifications and cryptic aberrations. Despite its inherent limitation in detecting recurrent balanced reciprocal translocations and low level secondary clonal abnormalities, the aCGH analysis provides detailed genomic features of simple and complex chromosomal abnormalities and cryptic aberrations otherwise not detectable by conventional G-band and FISH assays. Integrated chromosome and genomic analyses and evidence-based interpretation should be a standardized cytogenomic procedure for patients with MDS and AML.

Patient Samples
The Yale cytogenetics laboratory is CLIA-approved and provides diagnostic services to patients with various hematopoietic disorders and solid tumors. Follow up aCGH analyses had been performed on 30 MDS (n = 13) and AML (n = 17) patients with clonal chromosomal abnormalities detected in > 50% of BM or LB cells. All except one (case #17) were elderly patients with ages ranging from 51 to 93 years (average 67 years, Table 1). The criteria regarding the technical feasibility and medical necessity for pursuing diagnostic aCGH was: 1) sufficient residual BM or LB sample available for DNA extraction and clonal chromosomal abnormality detected in > 50% of BM or LB cells analyzed by conventional cytogenetics, 2) presence of chromosomally unresolved complex rearrangement or marker chromosome of unknown origin, and 3) genomic aberrations suspected in addition to the age-related Y chromosome loss and other simple chromosomal abnormalities. Informed consent was obtained from patients for use of residual materials on further genomic diagnosis.

Conventional Karyotyping and FISH Testing
Conventional chromosome analysis was performed on submitted BM and LB specimens using our laboratory's standardized protocols. Routine FISH tests were performed using a MDS/AML panel of commercial probes for the 5q (EGR1 gene at 5q31), 7q (D7S486 at 7q31), 8q (MYC at 8q24) and 20q (D20S108 at 20q12) loci and for the RUNX1T1 (ETO, 8q21. 3 [4,16].

The aCGH Analysis and Data Analysis
Genomic DNAs were extracted from the residual BM or LB specimens using Puregene Kit by following manufacturer's instruction (Qiagen Inc., Valencia, CA). DNA concentration was measured using a NanoDrop spectrophotometer (ND-1000, Thermo Fisher Scientific Inc., Waltham, MA) and high molecular weight DNA quality was verified by agarose gel electrophoresis. For each aCGH analysis, 2.5 ug of test genomic DNA from the patient and 2.5 ug of control DNA from a sex-matched or -mismatched healthy individual were used following the manufacturer's protocol for the Agilent Human Genome aCGH microarray 44K kit (Agilent Technologies Inc., Santa Clara, CA). This laboratory has validated the aCGH procedure to offer 99% sensitivity and 99% specificity with an average analytical resolution of 300-500 Kb using the log 2 ratio from five to seven contiguous probes, and also demonstrated its capacity in detecting 25%, 33% and 50% level of mosaicism [4]. The differential labeling of test and control DNAs, comparative hybridization onto 4x44K Agilent slides, posthybridization wash, slide scanning, image feature extraction were processed as previously described [4,16]. Data was analyzed using Agilent's DNA Analytical (version 4.0) with the built-in ADM-2 algorithm set at threshold value of 6, a cut off value of 0.25, and a filter of six probes. All CNAs except the recognized copy number variants from the Database of Genomic Variants http:// projects.tcag.ca/variation/ were recorded. The base pair designations from the Agilent 44K array are according to the March 2006 Assembly (NCBI36/hg18) on the UCSC Human Genome browser http://genome.ucsc. edu/. The aCGH finding from each case was compared with the chromosomally detected clonal abnormality to further define the breakpoint and the gene content involved. Raw data from the 30 cases were loaded onto the Nexus5 Software (BioDiscovery, Los Angles) to evaluate the genome-wide distribution and relative frequency of chromosomal and genomic alterations.

Results
The detected clonal chromosomal abnormalities of the 30 patients were listed in Table 1. All recurrent chromosomal deletions and translocations were confirmed by FISH tests using targeted probes. The aCGH analysis detected all main clone numerical chromosomal abnormalities and chromosomally-observed segmental deletions and duplications. The observed numerical abnormalities included gains of chromosomes 1, 4, 5, 6, 7, 8, 9, 11, 13, 14, 18, 19, 20 and 22, and losses of chromosomes Y, 5, 7, 13 and 18. No genomic aberration was noted in two patients with a balanced translocation (cases #10 and #21, Table 1). In five cases (cases #21, #26-27, #29-30, Table 1), sideline clones with unbalanced chromosomal abnormalities were not detected by aCGH. A total of 113 CNAs with size ranging from 0.257 to 102.519 Mb were found in 23 patients (an average of 4.8 CNAs per case and of 20.241 Mb per CNA, see Additional file 1, Table S1). The genome-wide distribution of chromosomal and genomic copy number alterations was plotted using the Nexus5 software ( Figure 1). Cases sharing similar cytogenetic abnormalities, such as the 5q deletion or monosomy 5 (n = 9), 7q deletion or monosomy 7 (n = 3), trisomy 8 (n = 3), 20q deletion (n = 2), loss of Y (n = 3) and other abnormalities (n = 10), were grouped together to estimate the frequency of other associated aberrations. It was noted that 5q deletions showed many associated recurrent aberrations, especially the 12p deletion (5/9 cases), 17p deletion (3/9) and 17q deletion (2/9). The size distribution and percentage of detected CNAs listed in Table 2 Figure 2A, Additional file 1, Table S1). The case-by-case comparison of chromosomal and genomic findings not only defined the 5q- (9) 7q-(3) 7q (3) +8 (3) 20q- (2) -Y(3) Others (   chromosomal abnormalities, but also demonstrated the genomic heterogeneity of recurrent chromosomal abnormalities. Chromosomally observed recurrent deletions of 5q, 7q, 9q, 12p, and 20q were all characterized by aCGH and confirmed in ten, three, two, six, and three cases, respectively. Of the 10 cases (cases #1~9 and #13 in Table 1) with a 5q deletion or monosomy 5, a simple deletion of various sizes was seen in seven cases and compound deletions of two or three segments were found in three cases (Additional file 1, Table S1). Microdeletions of 1.510 to 3.615 Mb encompassing the NF1 gene at 17q11.2 were found in three cases (cases #4, #5 and #22). FISH analysis using BAC clone probes for the NF1 and RHOT1 genes confirmed these deletions ( Figure 2B). It was interesting to observe a transition pattern in case #4 with deletions of NF1 alone, NF1 and RHOT1, and RHOT1 alone in 11%, 12.5% and 34% of cells, respectively. In case #13, a deletion involving the RHOT1 gene but not the NF1 gene was noted and confirmed by FISH. These results suggested that the 17q11.2 region may contain hot spots for initiating deletions from either one or both proximal and distal orientations.
The dmin chromosomes observed in case #19 were further defined as a 4.277 Mb amplification of 8q24.13-q24.21 containing genes TRIB1, MYC and CCDC26. From the log2 ratio of 2.53 for this amplified region, it was estimated that there were average four dmin chromosomes in each leukemic cell. Compound deletions, triplications or quadruplications for 11q were noted in cases #1 and #13, which resulted in complex intrachromosomal rearrangements carrying amplified segments of the MLL gene ( Figure 2B). The derivative chromosome 6, der(6)t(3;6)(q21.3;p22.2), in case #12 was further defined with breakpoints involving the RAB43 gene at 3q21.3 and the KIAA0319 gene at 6p22.2; and the isodicentric chromosome of Xp in case #27 was resulted from break and fusion distal to the PHKA1 gene at Xq13.1 (Additional file 1, Table S1).

Discussion
Cytogenetics has played a major role in both diagnosis and prognosis in patients with hematological malignancies. The newly revised WHO classification of acute myeloid leukemias categorizes AML into subgroups with recurrent genetic abnormalities, with MDS features, treatment related and "not otherwise specified"; the last one is by far the largest category [1]. Cytogenetics is key components of both the international prognostic scoring (IPSS) and the WHO classification-based prognostic scoring systems (WPSS) for the myelodysplastic syndromes [17,18]. However, within all categories prognostic variation is observed, likely among others due to low sensitivity of conventional cytogenetics [12]. Our data demonstrated that aCGH can detect chromosomal, segmental and cryptic aberrations in LB and BM cells from MDS and AML patients. As summarized in Table 2, the aCGH detected an average of 4.8 CNAs per case ranging from 0 to 22, of which an approximately 44% were less than 10 Mb and not evident by conventional cytogenetic analysis. It is noteworthy while on one hand some relatively large deletions and duplications (> 10 Mb) were missed or unresolved by conventional chromosomal analysis, on the other hand chromosomally detected sideline clones were missed by aCGH. Therefore, an integrated cytogenomic approach using chromosome analysis, FISH assay and aCGH would definitely improve the analytical resolution and abnormality detection rate for MDS and AML patients.
The interpretation of the clinical significance of somatic CNAs has been challenge especially for cases with numerous genomic aberrations. An evidence-based approach for interpreting constitutional genomic aberrations could be adopted for somatic CNAs with the classification of research evidence into four levels [6,7]. Level I evidence is derived from recognized diseasecausing clonal chromosomal abnormalities in the WHO classification [1] or from well-designed systematic studies; level II evidence comes from series of diagnostic studies; level III evidence is derived from cohort design observational studies or basic research of gene functions and disease mechanisms on in-vitro systems and/or model animals, and level IV is based on case reports or expert's opinions. Genomic findings without evidence were documented as unknown significance and uninterpretable in the laboratory's database but not presented in the diagnostic report. We propose a workflow chart to illustrate the integration of chromosome, FISH and aCGH procedures and the algorithm for interpreting results (Figure 3).
Several systematic genomic studies of MDS and AML patients have revealed recurrent and hidden CNAs and provided level I or II evidence for diagnostic interpretation of genomic findings [8][9][10][11][12][13][14][15]. Rucker FG et al. [8] performed 2.8 K (2,799 BAC/PAC clones) aCGH analysis of 60 AML cases with complex karyotypes and found that the most frequent losses were of 5q (77%), 17p (55%) and 7q (45%) as well as the most frequent gains of 11q (40%) and 8q (38%). Suela J et al. [9] used the 44K oligonucleotide aCGH to study 100 consecutive de novo AML cases and noted most recurrent changes of losses of chromosomes 5, 7, 17 and 20 and gain of chromosome 8. Tyybakinoja et al. [10] used the same 44K oligonucleotide aCGH to analyze 26 AML patients with a normal karyotype and detected deletions of 8q24.3, 12p12.3, 1q41, 18q21.32 12p13.2 and duplications of 3p21.3 and 8q24 in four cases. Akagi T et al. [11] used SNP chip on 38 AML/MDS patients with a normal karyotype and detected CNAs of 3p, 5p, 5q, 7q, 11q, 17p and 17q as well as trisomies of chromosomes 8, 21 and 22. Starczynowski DT et al. [12] used 32K (32,433 BAC clones) tiling aCGH on CD34+ cells from 44 MDS patients and observed that frequent cryptic alterations included gains of 11q and 17q and losses at 2q and 5q. Walter MJ et al. [13] reported that, of 86 adult patients with de novo AML analyzed by SNP array, 40% of patients with an abnormal karyotype also had additional CNAs and 24% of patients with normal cytogenetics had CNAs. Interestingly, similar analysis performed on 111 children with de novo AML noted a low burden of genomic alterations [14]. A recent exploratory study on 30 cases of MDS, myeloproliferative neoplasia (MPN) or AML detected genomic aberrations in 24 cases with concordant aCGH/chromosome results in 83% of samples and new CNAs in 47% of cases; and normal aCGH results were noted in four cases with low percentage clonal abnormality (0.7~5%) detected by chromosome analysis [15]. The recurrent genomic alterations and some known genes within the alterations from these studies were summarized in Table 3. These studies and our data indicated that genomic analysis can detect CNAs in 15-40% of AML cases with a normal karyotype, in 40-90% of MDS/AML cases with abnormal clones and in almost 100% of cases with complex chromosomal findings.
Basic researches toward the understanding of gene functions and disease-causing mechanisms or case reports with unique clinical and genomic findings could provide level III and IV evidence for reporting genomic alterations. Detailed genomic analyses of the malignant cells in MDS and AML patients is likely to yield specific types or regions of recurrent chromosomal and genomic abnormalities providing further evidence for disease association and allowing identification of to date unknown candidate genes. Eventually, further classification of disease and targeted treatment can potentially result from such knowledge. The 5q deletion is possibly the best understood recurrent chromosomal deletion in myeloid malignancies. Two common deletion regions (CDR) have been identified conferring either a good (CDR1) or a poor (CDR2) prognosis, but usually changes are more complex [19]. All 5q deletions in our 10 cases and in another study of 12 MDS cases with an isolated 5q deletion by Evers C et al. [20] involved the loss of the RPS14 gene at 5q33.1, which support the causal role of RPS14 haploinsufficiency and were specified in the diagnostic reports [21]. Other haploinsufficiency genes or miR-145 and miR146a at 5q could also have causal or modifying effects [22][23][24], and should also be referred in the report. A der(6)t(3;6)(q21;p22) in a patient with AML at relapse and a t(1;6)(q21;p22) in another patient were reported [25]. We reported here a der (6)  recent study suggested that the SAMD9, SAMD9L and Miki (LOC253012) as the candidate genes for 7q21.3 [26]. Two out of the three cases with 7q deletions in our patients had the 7q21.3 deletion. As reported in the literature [16,27,28] and also shown from our cases #1, #13, # 19 and #23, genomic analysis has been used effectively to define dmin chromosomes of 8q24 and complex 11q rearrangements with or without the MLL gene amplification in simple or complex karyotype. The presence of cytogenetic unresolved marker chromosome in our case #23 and the reported hidden abnormalities associated with t(15;17) justified further genomic analysis for this obvious balanced rearrangements [29,30]. Microdeletions at 17q11.2 involving the NF1 gene are considered to be recurrent cryptic alterations in three of our cases (#4, #5 and #22) and also documented in several reports [9,[11][12][13]31]. The presence of clustered flanking repetitive sequences of the NF1 locus is a likely explanation for recurrence of both constitutional and the somatic deletions [32]; further characterization of the noted transition pattern for the somatic 17q11.2 deletion could lead to better understanding of its mutagenesis mechanism.
In conclusion, our current diagnostic application of aCGH and accumulated evidence from previous studies support an integrated cytogenomic approach with evidence-based interpretation in MDS and AML patients with 1) > 20% blast cells with or without clonal chromosomal abnormality, 2) with complex chromosomal abnormalities, and 3) with simple and balanced rearrangements or a normal karyotyope but suspected cryptic abnormalities. This cytogenomic approach may not only provide a better diagnostic scheme to delineate breakpoints and gene contents of chromosomal and cryptic abnormalities in patients with MDS and AML, but hopefully allow identification of disease-causing or modifying candidate genes, and eventually lead to improved prognostification and treatment of patients with MDS and AML.

Additional material
Additional file 1: Table S1. Segemental copy number alterations detected in the 30 MDS/AML patients.