12th Congress of the European Hematology ... - Haematologica

12 th Congress of the European Hematology Association
tinct AML subclasses is still not fully understood. Recently, deregulatedexpression
of microRNAs (miRNAs) has been correlated with various
cancers including leukemias, and evidence was provided that miRNAs
can function both as oncogenes and tumor suppressors. Aims. In order to
determine a potential role of differential miRNA expression in AML we
profiled miRNA expression in a large series of adult AML patients to better
characterize AML on the molecular level. Methods. We analyzed 91
samples which encompass the spectrum of cytogenetic and molecular
genetic aberrations in AML using DNA microarray technology. We used
a miRNA microarray platform validated by quantitative RT-PCR and
Northern Blot analyses that contained approximately 250 human miR-
NAs. Results. By unsupervised hierarchical cluster analysis of our 91 AML
samples based on the miRNA expression of 202 filtered miRNAs we
were able to identify two large AML subgroups. While these two groups
were in part characterized by the differential expression of the miR-17-
92 cluster, which recently has been identified to be deregulated in many
cancer types, especially Myc-regulated tumors, we also identified several
miRNAs not yet known to be differentially expressed in leukemia.
Interestingly, correlation of the unsupervised cluster defined groups
revealed an association with leukemia morphology as determined by
FAB (French-American-British) classification, but there was no significant
correlation with cytogenetically or molecular-genetically defined
leukemia AML subgroups. On the other hand, by supervised analysis
using the significance analysis of microarrays (SAM) methodology we
were able to identify miRNA signatures characterizing acute promyelocytic
leukemias with a t(15;17) and normal karyotype AML carrying
mutations of NPM1. Summary/Conclusions. While these findings already
support a potential role of differential miRNA expression in AML pathogenesis,
future analyses correlating miRNA expression with global gene
expression, which are currently underway, are likely to provide additional
insights into AML biology, thereby helping to unravel the role of
miRNAs in leukemogenesis.
0890
IDENTIFICATION OF CANDIDATE GENES IN HIGH-LEVEL DNA AMPLIFICATIONS IN AML
WITH COMPLEX KARYOTYPE USING AN ARRAY-BASED APPROACH
F. Rucker, 1 L. Bullinger, 1 S. Miller, 1 H. Kestler, 1 P. Lichter, 2 K. Dohner, 1
H. Dohner1 1 University of Ulm, ULM, Germany; 2 DKFZ, HEIDELBERG, Germany
Background. Complex karyotype acute myeloid leukemia (AML), commonly
defined as the presence of three or more chromosome abnormalities
without specific fusion transcripts, is seen in approximately 10-15%
of all AML cases. In this subset of cases, genomic losses and gains are
much more frequent than balanced translocations, indicating other mechanisms
of leukemogenesis. One possible mechanism is the activation of
(proto-) oncogenes through high-level DNA amplifications. Aims. To
detect high-level DNA amplifications and to identify corresponding candidate
genes, we applied comparative genomic hybridization to microarrays
(array-CGH) in 150 cases of complex karyotype AML and correlated
the findings with gene expression profiling (GEP) data. Methods. For
array-CGH a custom-made 2.8k-microarray was used consisting of 2799
different BAC- or PAC-vectors with an average resolution of approximately
2 Mb. Hybridization experiments were performed in a dye-swap
setting; significant aberrations were defined as mean plus/minus three
times the standard deviation of a set of balanced clones for each individual
experiment. In selected cases correlation with global gene expression
studies was performed to further delineate regions harboring candidate
genes. Results. We identified 70 high-level DNA amplifications in 25 different
genomic regions. Amplifications occurring in at least two cases
mapped to (candidate genes in the amplicon) 11q23.3-q24.1 (n=16; ETS,
FLI1, APLP2); 11q23.3 (n=15; MLL, DDX6, LARG, SC5DL); 21q22 (n=9;
ERG, ETS2); 9p24 (n=4; JAK2); 13q12 (n=4; CDX2, FLT3, PAN3); 8q24
(n=4; C8FW, MYC); 12p13 (n=2; FGF6, CCND2); 11q13 (n=3; STARD10,
GARP, RAD30, DLG2), 20q11 (n=2; ID1, BCL2L1); and. 22q11 (n=2;
CHEK2, NF2 To better characterize the genomic architecture of the
amplicons, we additionally applied array-CGH using an 8.0k-microarray
with an average resolution of approximately 1 Mb. Using this approach
highly complex amplicon structures with several distinct amplicon peaks
were identified e.g. in the amplified regions 8q24, 11q23, and 13q12. Parallel
analysis of array-CGH and GEP in a subset of 43 cases displayed
overexpressed candidate genes in the critical amplified regions; for some
of these genes an oncogenic role has been implicated e.g. C8FW and
MYC in 8q24, ETS1, FLI1 and APLP2 in 11q24.1, as well as FLT3 and
CDX2 in 13q12. Summary/Conclusions. Using high-resolution genomewide
screening tools such as array-CGH, a large number of high-level
DNA amplifications was identified in AML with complex karyotype sug-
332 | haematologica/the hematology journal | 2007; 92(s1)
gesting a more general role for protooncogene activation in this AML
subset. This high-resolution technique allows the detection of complex
amplicon structures with several distinct amplicon peaks pinpointing to
selective candidate genes. In addition, correlation with GEP studies facilitates
the delineation of overexpressed candidate genes within the amplified
regions.
0891
AN INTERNATIONAL MULTI-CENTER RESEARCH STUDY TO DEFINE THE APPLICATION
OF MICROARRAYS IN THE DIAGNOSIS AND SUBCLASSIFICATION OF LEUKEMIA (MILE
STUDY): A REPORT ON 2926 CASES
T. Haferlach, 1 A. Kohlmann, 2 K. Mills, 3 W.K. Hofmann, 4 T. te Kronnie, 5
G. Basso, 5 J. Hernandez Rivas, 6 J. Downing, 7 A. Yeoh, 8 J. De Vos, 9
M.C. Béné, 9 C. Preudhomme, 9 E. Macintyre, 9 T. Kipps, 10
L. Wieczorek, 2 ,R. Foa11 1 MLL, Munich Leukemia Laboratory, MUNICH, Germany; 2 ROCHE Molecular
Systems, PLEASANTON, USA; 3 Department of Haematology, CARDIFF,
United Kingdom; 4 Charité, BERLIN, Germany; 5 Oncohematology, PADOVA,
Italy; 6 Hematología, SALAMANCA, Spain; 7 St.Jude Children’s Research Hospital,
MEMPHIS, USA; 8 National University of Singapore;, SINGAPORE,
Singapore; 9 Institut de Recherche en Biothérapie, MONTPELLIER, France; 10 University
of California, SAN DIEGO, USA; 11 Università La Sapienza, ROME,
Italy
Background. Microarray analysis can identify differentially expressed
genes associated with distinct clinical and therapeutically relevant classes
of both pediatric and adult leukemias. Aims. Recently, 11 MILE
(Microarray Innovations in Leukemia) study centers from the European
Leukemia Net (seven laboratories), USA (three laboratories), and Singapore
(one laboratory) completed their analysis phase of archived patient
samples using whole genome microarrays. Methods. Using a standardized
laboratory protocol for Affymetrix HG-U133 Plus 2.0 microarray analysis,
the classification accuracy of gene expression profiles for 16 acute and
chronic leukemia subclasses (mature B-ALL with t(8;14), Pro-B-ALL with
t(11q23)/MLL, c-ALL/Pre-B-ALL with t(9;22), T-ALL, ALL with t(12;21),
ALL with t(1;19), ALL with hyperdiploid karyotype, c-ALL/Pre-B-ALL
without t(9;22), AML with t(8;21), AML with t(15;17), AML with
inv(16)/t(16;16), AML with t(11q23)/MLL, AML with normal karyotype
or other abnormalities, AML complex aberrant karyotype, CML, CLL),
MDS, as well as non-leukemia/healthy volunteers as a control group is
compared to the current routine diagnostic workup of each center. Results.
All participating laboratories demonstrated high proficiency of microarray
analysis in a so-called pre-phase where cell lines had been tested during
laboratory standardization and proficiency analysis. The subsequent
MILE study Stage I included in total 2156 samples. Strict quality acceptance
criteria of the gene expression results were met in >98% of the samples.
Gene expression profiles from this study were further combined
with previous microarray data obtained by the research groups from
Munich (Haferlach) and Memphis (Downing). The emerging data set of
2926 patient samples and subsets thereof then was used to train linear
discriminant classification algorithms to assess the prediction accuracy for
18 distinct sample classes. The average accuracy of three 30-fold crossvalidations
resulted in 94.44% concordant calls comparing goldstandard
diagnosis and microarray result. Miscalls between the classes were predominantly
observed in the distinction between MDS samples, characterized
by an underlying AML-like gene expression signature, and AML
with normal karyotype. For an independent test data set with 92 specimens
covering 12 of the 18 classes, the accuracy was 87/92 (94.57%).
Moreover, based on distinct gene expression profiles, CLL can be further
classified into IgVH mutated or unmutated subgroups (98.41% accuracy
of cross-validation with fixed probe sets, n=289 samples), or ZAP70 positive
or negative cases (95.51%, n=256 samples), respectively. As a next
step, a customized microarray for leukemia classification was designed
using 1,449 probe sets. Conclusions. This international multi-center study
demonstrated a very high accuracy of leukemia classification using
whole-genome gene expression profiling and indicated the feasibility of
a routine diagnostic application of microarrays. As started in December
2006 additional 2,000 samples will be prospectively analyzed in the MILE
study Stage II with a custom research AmpliChip Leukemia microarray
to test these accuracy estimates in a blinded study.