Complex genomic alterations and gene expression in acute lymphoblastic leukemia with intrachromosomal amplification of chromosome 21

Abstract

We have previously identified a unique subtype of acute lymphoblastic leukemia (ALL) associated with a poor outcome and characterized by intrachromosomal amplification of chromosome 21 including the RUNX1 gene (iAMP21). In this study, array-based comparative genomic hybridization (aCGH) (n = 10) detected a common region of amplification (CRA) between 33.192 and 39.796 Mb and a common region of deletion (CRD) between 43.7 and 47 Mb in 100% and 70% of iAMP21 patients, respectively. High-resolution genotypic analysis (n = 3) identified allelic imbalances in the CRA. Supervised gene expression analysis showed a distinct signature for eight patients with iAMP21, with 10% of overexpressed genes located within the CRA. The mean expression of these genes was significantly higher in iAMP21 when compared to other ALL samples (n = 45). Although genomic copy number correlated with overall gene expression levels within areas of loss or gain, there was considerable individual variation. A unique subset of differentially expressed genes, outside the CRA and CRD, were identified when gene expression signatures of iAMP21 were compared to ALL samples with ETV6-RUNX1 fusion (n = 21) or high hyperdiploidy with additional chromosomes 21 (n = 23). From this analysis, LGMN was shown to be overexpressed in patients with iAMP21 (P = 0.0012). Genomic and expression data has further characterized this ALL subtype, demonstrating high levels of 21q instability in these patients leading to proposals for mechanisms underlying this clinical phenotype and plausible alternative treatments.

We have recently defined a recurrent chromosomal abnormality at an incidence of 1.5% in childhood B-lineage acute lymphoblastic leukemia (ALL) involving intrachromosomal duplication of chromosome 21 and amplification of the RUNX1 (AML1) gene (iAMP21) (1). These patients have a median age of 9 years, a low presenting white blood cell count, and a poor prognosis (2). Thus, on the current U.K. ALL treatment protocol, ALL 2003, these children are classified as high-risk and receive more intensive treatment. iAMP21 was identified on routine screening of childhood ALL patients for the ETV6-RUNX1 (TEL-AML1) fusion by fluorescence in situ hybridization (FISH). Although negative for the fusion, leukemic cells showed multiple RUNX1 signals, seen as clusters in interphase and in tandem duplication on the long arm of an abnormal chromosome 21 in metaphase. This abnormality cannot be defined by conventional cytogenetic analysis because the abnormal chromosome 21 adopts a range of different morphological forms. FISH with probes directed to the RUNX1 gene is currently the only detection method, which explains its prior description as “amplification of RUNX1.” However, there are several reasons why FISH detection, based solely on RUNX1 copy number, may be inappropriate. First, interpretation may be misleading, particularly in patients with a hidden high hyperdiploid clone comprising several copies of chromosome 21 (3). Second, because the observed increase in RUNX1 copy number was serendipitous, it may not be the causative mechanism. In view of the high-risk associated with iAMP21, it is important to fully characterize this abnormality to provide accurate diagnosis, particularly for ALL patients without any other high-risk clinical features.

Similar chromosome 21 amplifications have been reported in patients with acute myeloid leukemia (AML) and myelodysplastic syndrome (4–9). The most recent AML study, using BAC array-based comparative genomic hybridization (BAC aCGH), identified two common regions of amplification on 21q in 12 patients. These were at 25–30 Mb and 38.7–39.1 Mb. Oligonucleotide expression analysis revealed that most significantly overexpressed genes were located within these amplicons, implying that the changes in gene expression were entirely related to alterations in copy number (5). Similar gene expression analyses from children with high hyperdiploid ALL (10) and Down syndrome (11) have suggested that additional copies of chromosome 21 lead to overexpression of genes on chromosome 21. By using a variety of classical and innovative molecular techniques, we have been able to characterize the iAMP21 in patients with ALL and, in so doing, provide a plausible alternative therapeutic approach.

Results and Discussion

In this study, we have validated the existence of the chromosomal abnormality iAMP21 in childhood ALL and characterized the rearrangement using whole genome analyses. Genome-wide BAC aCGH showed genomic imbalances in all 10 patients with iAMP21 analyzed. Patterns of imbalance corresponding to over- and underrepresentation of specific regions of chromosome 21 were unique to each patient (Table 1). Although all BAC clones on chromosome 21 showed gain in at least one patient, these gains most frequently involved clones between genomic positions 22.1 and 27.8 Mb (clones RP11-64I12 to RP11-90A12). The size of the most highly amplified region varied considerably between patients, from 3–8.6 to 24.0–24.1 Mb for patients 6783 and 6788, respectively. However, a common region of amplification (CRA) of ≈8.6 Mb, between clones RP11-191I6 and RP5-206A10 (genomic positions 31.5 and 40.1 Mb, respectively), was identified in all 10 patients, which was accompanied by deletions of 21q in seven patients. With the exception of one within the centromeric region (from 15.1–20.3 Mb in patient 5898), all deletions included a common region of deletion (CRD) of ≈4 Mb close to the telomere. In three patients with iAMP21, imbalances of 21q were the sole genomic changes at 1-Mb resolution. Among the other patients, no recurrent changes involving chromosomes other than 21 were identified. (Table 3, which is published as supporting information on the PNAS web site). To prove the validity of aCGH, the presence of an entire additional copy of chromosome 21 was verified in seven patients with high hyperdiploidy and additional copies of chromosome 21 (HD + 21) (example, patient 6009) (Table 1). Furthermore, no changes in copy number were observed among 50 patients with apparently normal copies of chromosome 21 (example, patient 6899) (Table 1). FISH analysis confirmed the variation in copy number along 21q in the cases analyzed by BAC aCGH (Table 1 and Fig. 1A and B). FISH identified the same CRA and CRD. The high concordance between the two procedures indicated the accuracy of BAC aCGH in the determination of copy number changes, whereas FISH analysis provided precise quantification. Between three and eight additional copies of the clones within the CRA were demonstrated by FISH, indicating a 2.5–5 fold gain. FISH data on copy number changes in an additional three patients with iAMP21 provided further confirmation of the BAC aCGH results (data not included). Using tiling-path Oligo aCGH (Fig. 1C), the extent of the CRA was refined to a region of 6.527–6.604 Mb in size (between genomic positions 33.192 and 39.796 Mb) in five patients, whereas the CRD was refined to 3.541 Mb. High-resolution genotype array analysis of single-nucleotide polymorphisms (SNP analysis) was performed on three patients for whom both diagnostic and remission samples were available, permitting comparison of germ-line and tumor genotypes (examples are shown in Fig. 3, which is published as supporting information on the PNAS web site). These analyses identified the same regions of genomic gain and loss of heterozygosity within the CRA and CRD.

Table 1.

BAC aCGH and FISH results for 10 ALL patients with iAMP21

Gains (green) and losses (red) of 21q material detected by BAC aCGH. Yellow regions correspond to those areas exhibiting fluorescent ratios within standard deviation limits (SDL). Ratio values were unavailable on several samples due to a lack of material, and on certain DNA clones due to poor ratio measurements. Where FISH was carried out, results are shown numerically as deviations from a normal copy number of 2. The asterisks indicate cases studied for gene expression by oligonucleotide array. The # indicates those cases used for further genomic profiling with Oligo aCGH array analysis. Cases 6899 and 6009 are ALL patients with an apparently normal and high hyperdiploid (tetrasomy 21) karyotype, respectively.

Fig. 1.

Genomic analysis of DNA and cell suspension from patient 5989. (A) BAC aCGH results: chromosome 21 is positioned horizontally, with the centromeric to telomeric positions running from left to right, respectively. Dye swap experiments 1 and 2 are shown by the blue and red lines, respectively. Double deviation of both these experiments from a normal value of 1.00 demonstrates loss or gain of DNA material. Deviation of the red and blue line >1.00 shows loss or gain of copy number, respectively. (B) Examples of the FISH confirmation of aCGH data: each numbered FISH probe corresponds to the same highlighted clone in A. (C) Oligo aCGH data for this patient. Chromosome 21 is positioned as in A. The scatter plot demonstrates mean log intensity ratios at 5,000-bp intervals along chromosome 21. Segmentation analysis is shown as red horizontal lines.

Combining the results from these genomic analyses has highlighted regions of variable gain along 21q in patients with iAMP21 and identified a CRA covering a large genomic region of 6 Mb, containing the RUNX1 gene. This CRA was found to be telomeric of the first of the two amplified regions described for AML (25–30 Mb) but overlapping with the second (38.7–39.1 Mb) (5). The majority of patients showed a 3.5-Mb CRD, telomeric of the CRA. The SNP data suggested that the amplification was derived from a single chromosome. Overall, the results indicated the highly variable nature of this abnormality, reflecting considerable instability of chromosome 21, thus making it difficult to determine the causative event.

Global gene expression profiling, using the Affymetrix U133A oligonucleotide array containing 22,283 probes sets, was performed on eight patients with aCGH results. The CRA was represented on this Affymetrix GeneChip by 96 probe sets in total, including 40 well characterized genes and six ORFs. The genes located within the CRA and CRD are indicated in Tables 4 and 5, which are published as supporting information on the PNAS web site. From a total of 768 probe sets within the CRA, 321 (42%) were present (or marginal) and up-regulated (Fig. 4, which is published as supporting information on the PNAS web site). Of the 46 sequences from the genes and ORF, 13 were up-regulated in at least 75% of patients. The CRD was represented on the GeneChip by 83 probe sets, containing 33 genes, three ORFs, and three ESTs. From a total of 664 probe sets, 462 (70%) were absent. An absent flag was carried in 22 of these 39 gene sequences in at least 75% of patients. When compared to all children with ALL (n = 89) from our previously reported analysis (12), 14 (10%) of the top 150 genes significantly overexpressed in patients with iAMP21 were located within the CRA, for which there was a strong correlation with the Taqman data (Table 6, which is published as supporting information on the PNAS web site). As shown in Table 4, 51 (53%) of the 96 probe sets within the CRA had a 1.5-fold increase in expression. This observation suggested that overexpression of these genes corresponded to the gain of genomic material. However, it was noted that 47% of the probe sets from the CRA were not overexpressed. To examine the effect of the gain of chromosomal material more closely, we calculated the mean and median expression of the genes within the CRA in the eight iAMP21 and six patients with other subtypes of ALL (Fig. 2A). The mean expression levels of the genes contained within the CRA was higher in patients with iAMP21 (t test, P = 0.00903) and those with HD + 21 (t test, P = 2.02e-7) compared with the other subtypes. These observations support previous reports demonstrating that large-scale genomic alteration does result in changes in expression of genes within these regions (13, 14), but we could not correlate all gene expression changes with alteration at the genomic copy level. There was no linear correlation between the degree of amplification and expression; this may have arisen from heterogeneity of amplification within the region or other regulatory mechanisms influencing gene expression, such as epigenetics and biofeedback regulation. We have recently reported partial acquired isodisomy in patients with AML (15), whereas others have reported disomy of chromosome 21 in cases of Down syndrome and ALL (16). Thus, it is plausible that this type of mechanism may contribute to variations in expression.

Fig. 2.

Box plot diagrams illustrating LGMN expression (A) and expression of those genes within the CRA (B), compared to other ALL subtypes. On the x axis are shown seven ALL subtypes, BCR-ABL, E2A-PBX1, T-ALL, HD + 21, ETV6-RUNX1, iAMP21, and others. The y axis represents the relative gene expression level of either LGMN or all those genes within the CRA. Each box plot shows the distribution of expression levels from 25th to 75th percentile. The median is shown as a line across the box, whereas the + is the calculated mean expression level for the particular subtype. The dotted line indicates the inner fence, and a value outside the outer fence is shown as an asterisk.

Like the AML study (5), our work showed differential expression of genes located outside the CRA, leading to expression variation of many genes unassociated with, but flanking genes important in cancer pathogenesis. However, these genes were not consistent between patients. When the expression profiles from patients with iAMP21 were compared to the original cohort of 89 pediatric ALL patients (12) in an unsupervised analysis, the patients with iAMP21 did not cluster together (data not shown). Due to correlation between gene expression and loss or gain of chromosomal material, a supervised cluster analysis was carried out to take into consideration that expression profiles were influenced by ALL samples with rearrangements and gains of chromosome 21. The global gene expression profiling of the eight iAMP21 patients was compared: to the full cohort (n = 89); to a subgroup of patients with the ETV6-RUNX1 fusion (n = 21); and to a subgroup of patients with HD + 21 (n = 23). When the gene list was compiled in this manner, patients with iAMP21 exhibited a distinctive expression pattern (Fig. 4). Using SAM, with a cutoff level for false discovery rate (FDR) of 10%, the three comparisons yielded 4,174, 4,768, and 5,147 probe sets, respectively. The top 150 probe sets (FDR = 5.3%) were used for comparison against other ALL samples, and the top 100 for comparisons against patients with the ETV6-RUNX1 fusion (FDR = 0.54%) and HD + 21 (FDR = 0.81%). The gene lists with full annotation are presented in the supplemental data (Tables 7–9, which are published as supporting information on the PNAS web site). Comparison of all three lists identified 11 genes that were uniquely overexpressed in iAMP21 patients, of which only two, C21orf66 and ATP50, were within the CRA. One of the other overexpressed genes was legumain (LGMN). As shown in Fig. 2B, LGMN expression was significantly elevated in those with iAMP21 when compared to other subtypes of ALL (t statistics = 4.38; df = 7; P = 0.0012). Using a similar approach, 12 genes outside the CRD were shown to be expressed at significantly lower levels (Table 2).

Table 2.

Significant differentially expressed genes in patients with iAMP21 (n = 8)

Gene	Probe identifier	Gene name	Chromosomal location	UniGene cluster
Up-regulated
LGMN	201212_at	Legumain	14q32.1	Hs.18069
C1orf54 (FLJ23221)	219506_at	Chromosome 1 open reading frame 54	1q21.2	Hs.91283
STK17B	205214_at	Serine/threonine kinase 17b (apoptosis-inducing)	2q32.3	Hs.88297
BHLHB2	201170_s_at	Basic helix-loop-helix domain containing, Class B, 2	3p26	Hs.171825
ARPC5L	220966_x_at	Actin related protein 2/3 complex, subunit 5-like	9q33.3	Hs.132499
TBCD	211052_s_at	Tubulin-specific chaperone d	17q25.3	Hs.464391
LSM7	204559_s_at	LSM7 homolog, U6 small nuclear RNA associated (S. cerevisiae)	19p13.3	Hs.512610
GADD45B	209304_x_at	Growth arrest and DNA-damage-inducible, beta	19p13.3	Hs.110571
C20orf111	209020_at	Chromosome 20 open reading frame 111	20q13.11	Hs.75798
C21orf66	221158_at	Chromosome 21 open reading frame 66	21q21.3	Hs.473635
ATP5O	200818_at	ATP synthase, H+ transporting, mitochondrial F1 complex, O subunit (oligomycin sensitivity conferring protein)	21q22.1-q22	Hs.409140
Down-regulated
NIPBL	207108_s_at	Nipped-B homolog (Drosophila)	5p13.2	Hs.481927
BAT2	208132_x_at	HLA-B associated transcript 2	6p21.3	Hs.436093
CDYL	203100_s_at	Chromodomain protein, Y-like	6p25.1	Hs.269092
GFOD1	219821_s_at	Glucose-fructose-oxidoreductase domain containing 1	6pter-p22.1	Hs.484686
KIAA0265	209256_s_at	KIAA0265 protein	7q32.2	Hs.520710
CAMSAP1	212710_at	Calmodulin regulated spectrin-associated protein 1	9q34.3	Hs.522493
PELI2	219132_at	Pellino homolog 2 (Drosophila)	14q21	Hs.105103
KIAA0100	201729_s_at	KIAA0100 gene product	17q11.2	Hs.151761
SS18	216684_s_at	Synovial sarcoma translocation, Chromosome 18	18q11.2	Hs.404263
FEM1B	212367_at	Fem-1 homolog b (C. elegans)	15q22	Hs.362733
RNF146	221430_s_at	Ring finger protein 146	6q22.1-q22.3	Hs.267120
MBD1	208595_s_at	Methyl-CpG binding domain protein 1	18q21	Hs.405610

Unlike the AML report, in which a control cohort of normal karyotype AML patients was used, our report benefited from comparisons with patient groups who had either gained an entire chromosome 21 as part of a high hyperdiploid karyotype or in association with an ETV6-RUNX1 fusion. By using these subgroups for comparison, it was possible to identify a unique subset of over- and underexpressed genes in patients with iAMP21 relative to those with/without other chromosome 21 aberrations. This finding demonstrated that comparative information on the loss or gain of chromosomal material is essential when interpreting expression data. Curiously, RUNX1 expression in ETV6-RUNX1 positive and iAMP21 patients was comparable, which may be due to the inability of the global gene expression profiling platform used in this study to distinguish between wild-type RUNX1 and ETV6-RUNX1 fusion transcripts. Additional copies of the ETV6-RUNX1 fusion and RUNX1 are common findings in patients with the translocation, t(12;21)(p13;q22) (3), which may contribute to the elevated RUNX1 expression levels in ETV6-RUNX1-positive patients. These observations suggest that there are common processes leading to duplication and translocation, further strengthening the hypothesis that genomic instability of a region on 21q creates a cascade of events leading to or sustaining leukemogenesis.

Although the processes that lead to ALL appear to affect a common genomic region of chromosome 21, there is disparity in the outcome to treatment. HD + 21 and ETV6-RUNX1 have an excellent survival rate on current chemotherapy protocols, whereas iAMP21 patients have a poor outcome. Recently, two papers have correlated gene expression patterns with in vitro chemosensitivity of blast cells and both demonstrated that these patterns were predictive of outcome in childhood ALL (17, 18). Although the expression patterns in our patients did not accurately reflect those associated with poor clinical outcome and chemo-resistance (including Asparaginase), there were a number of similarities including over- (IGHM, CD44, IGFBP7, RPS9, and MAFF) and underexpression (TCF4, F8A, and TAF5) of a number of genes. Of these, MAFF overexpression is known to correlate with steroid resistance and F8A down-regulation with insensitivity to asparaginase. We have shown that the gene LGMN is overexpressed in ALL samples with iAMP21. LGMN is a lysosomal cysteine protease that specifically cleaves after the asparagine residue and participates in antigen processing (19). Cancer cells expressing LGMN have been shown to invade extracellular spaces. Overexpression in a number of aggressive cancers correlates with invasiveness, dissemination, and poor outcome (20, 21). We hypothesize that lymphoblasts expressing LGMN may enter extravascular spaces. These cells survive because of suboptimal cytotoxic levels, which may lead to subsequent relapse.

We conclude that the CRA on chromosome 21 represents the only detectable recurrent finding in patients with iAMP21. Expression profiling did not show significant overexpression of RUNX1 in these patients, suggesting that it is unlikely to be the target gene. Overall, the increase of gene expression within the CRA was a result of the genomic copy number gain within this region, suggesting that these genes may be important in leukemogenesis. However, no single causative gene was identified. Outside the CRA, overexpression of LGMN was demonstrated. We hypothesize that this gene may contribute to the poor clinical outcome and treatment response observed in iAMP21 patients. In addition to the 6.5–6.6 Mb CRA, there was associated genomic imbalance in patients with iAMP21, in particular deletions affecting the subtelomeric region (CRD) of chromosome 21. These data have provided information that will be used to develop an improved diagnostic test. The expansion of this innovative study may uncover other molecular and cellular mechanisms underlying this clinical phenotype, demonstrating a pivotal role of chromosome 21 instability in the initiation of acute leukemia.

Materials and Methods

Patients.

In this study, 14 patients with iAMP21, defined in accordance with published cytogenetic and FISH criteria (1), with DNA and/or RNA available were identified among those registered to the U.K. ALL treatment trials: ALL97/99, MRD PILOT, or ALL2003 for children aged 1–18 years, or UKALLXII for adults aged 15–55 years. Genome and expression studies were applied to these patient samples as indicated in Table 3. Each center obtained informed consent from patients or their parents.

Cytogenetic Analysis.

Diagnostic bone marrow and/or peripheral blood samples from all patients in this study were analyzed by standard cytogenetic methods in the U.K. regional cytogenetics laboratories. RUNX1 copy number was determined by using the LSI TEL/AML1 ES Dual Color Translocation FISH probe (Abbott Diagnostics, Maidenhead, U.K.). This information is provided in Table 3.

BAC aCGH and FISH Confirmation.

For 10 patients, genomic copy number variation was assessed by using a commercially available BAC aCGH system (Spectral Genomics, Genosystems). The arrays comprised 2,621 genomic clones positioned at ≈1-Mb intervals throughout the genome. Of these, 26 were located along 21q from position 15.1 Mb (centromeric) to 46.9 Mb (telomeric). The positions of genes and BAC clones were determined by using the National Center for Biotechnology Information (NCBI) MapViewer for Homo Sapiens, Build 35, version 1 (www.ncbi.nlm.nih.gov/mapview).

Pooled DNA extracted from peripheral blood of 10 healthy donors, sex matched to the test sample, was used as the reference (Promega) and processed according to the manufacturer's instructions. On the basis of control experiments, a normal range of 0.8–1.2 was used for the analysis of patients with iAMP21, a range broader than one calculated on the basis of 2× SD for each clone calculated, in the normal-versus-normal hybridizations. In an attempt to improve sensitivity, fluorescence ratio outside the limit of 2× SD (standard deviation limits, SDL), but within standard cutoff values of between 0.8 and 1.2, were also recorded for comparisons with FISH confirmatory data.

For nine patients, DNA copy number changes detected by aCGH were validated by using FISH probes from the same BAC clones as spotted on the array (Genosystems) (Table 1). Where possible, 200 interphase nuclei per probe were analyzed by two independent analysts, and images were recorded by using macprobe software (Applied Imaging, Newcastle, U.K.) (further details of aCGH and FISH analysis are given in Supporting Text, which is published as supporting information on the PNAS web site).

Genomic Oligonucleotide Arrays.

Five patients (all analyzed with BAC aCGH of the same sample) were analyzed with high-density oligonucleotide-based CGH (Oligo aCGH) arrays (NimbleGen Systems, Madison, WI), designed with probes tiled through chromosome 21. Sequences (NCBI build 35.1) were repeat-masked, and oligonucleotides were selected at a minimal spacing distance of 60 bp from both the forward and reverse strands, resulting in ≈190,000 features along the length of the chromosome. The arrays were synthesized as described (22), and standard labeling, hybridization, and image capture was performed in the NimbleGen Systems Service Laboratory, in a similar manner to that described by Selzer et al. (23). Data were extracted from scanned images by using nimblescan extraction software (NimbleGen Systems), which allows automated grid alignment, extraction, and generation of data files. Segmentation analysis of data sets indicated deletion and amplification breakpoints. Corrections for optical noise, background adjustments, and normalization were performed by using bioconductor as described (24). After a loss correction for probe GC content, the log₂ ratios were averaged in windows ranging from 500 to 5,000 bp to produce the final segmentations (25). Further details are provided Supporting Text.

GeneChip Human Mapping 10K Array.

The GeneChip mapping assay protocol (Affymetrix) was used to produce the 10,000 SNP array results for three iAMP21 patients as described (26, 27). The protocol was adapted such that the purification of PCR product was performed by using the Ultrafree-MC filtration column (Millipore, Billerica, MA). Signal intensity data were analyzed by the GeneChip DNA analysis software (gdas), which uses a model algorithm to generate SNP calls. Signal values are normalized across each array to the median value, and copy number ratios and changes in SNP calls between leukemia and germ-line remission bone marrow were annotated by using a program written in visual basic. Noise was reduced by zeroing negative signal values, and using mean signal values in a running window of five SNPs.

Global Expression Profiling.

RNA Extraction and probe preparation.

Global expression profiling was carried out on bone marrow aspirates from eight patients (seven with aCGH results). RNA was extracted with TRIzol (Invitrogen) followed by a second ethanol precipitation, before quality assessment using the Agilent 2100 Bioanalyser (Agilent Technologies, Waldbronn, Germany). Fluorescently labeled cRNA probes were synthesized and hybridized to Affymetrix HG-U133A oligonucleotide arrays according to the manufacturer's instructions. The arrays were scanned on a GeneArray scanner (Agilent Technologies), and the intensities of the fluorescence signals were captured and analyzed with Affymetrix mas 5.0 software. No scaling was applied. Further detailed descriptions of the procedure and the raw Affymetrix files are given in Supporting Text.

Gene expression analysis.

genespring 6.0 (Silicon Genetics, Redwood City, CA) was used for raw data normalization. First, the data were normalized to the median per sample, using all genes not marked absent. Each gene was then divided by the median of its measurements in all samples (i.e., across all arrays). If the median of the raw values was <10, then each measurement for that gene was divided by 10. Signal intensities were log transformed for statistical analysis. Genes called absent in all samples were removed to exclude those with minimal variation across the experiments. Probe sets passing the filter were used to find statistically significant differentially expressed genes between the subgroups studied. Significance Analysis of Microarrays (SAM) was applied to the normalized and log-transformed data. We used default settings and selected the significant genes based on the d-score with a maximum FDR of 5.3%. These were compared with a data set of 89 children with ALL, of whom 21 had an ETV6-RUNX1 fusion, 23 had high hyperdiploidy comprising at least one additional copy of chromosome 21 (HD + 21) and 45 had no abnormality of chromosome 21. For 80 patients, including one patient with a iAMP21, the gene expression pattern has been reported (12).

Both unsupervised and supervised analyses were used and the results visualized in a two-way hierarchical cluster. Normalized gene expression values were used to obtain the mean and median expression values of genes within the defined amplicon. Significance in the differences of expression between the different subgroups was tested by using a t test.

Quantitative RT-PCR.

Real-Time quantitative RT-PCR (qRT-PCR) was carried out to assess the expression of genes situated within the amplicon (SOD1, OLIG2, IFNAR2, IL10RB, ITSN1, CRYZL1, RUNX1, TTC3, ERG and ETS2) for six patients, using the Taqman Gene Expression Assays (Applied Biosystems) according to the manufacturer's instructions. Appropriate positive and negative control RNA samples were tested in parallel. The comparative C_t method was used for quantitation of relative gene expression. The average C_t value of the endogenous control gene, GAPDH, was subtracted from the average experimental gene C_t value to give the ΔC_t value. Differences between control and test were carried out by using ΔΔC_t. Concordance between the qRT-PCR and global expression profile was demonstrated after calculation of the correlation coefficients between the level of expression as quantified by both qRT-PCR and Affymetrix expression arrays.

Abbreviations

ALL

acute lymphoblastic leukemia

AML

acute myeloid leukemia

aCGH

array-based comparative genomic hybridization

CRA

common region of amplification

CRD

common region of deletion

FDR

false discovery rate.