Review

Leukemia (2011) 25, 909–920; doi:10.1038/leu.2011.48; published online 29 March 2011

Gene expression profiling in MDS and AML: potential and future avenues

K Theilgaard-Mönch1,2, J Boultwood3, S Ferrari4, K Giannopoulos5, J M Hernandez-Rivas6, A Kohlmann7, M Morgan8, B Porse1, E Tagliafico4, C M Zwaan9, J Wainscoat3, M M Van den Heuvel-Eibrink9, K Mills10,12 and L Bullinger11,12

  1. 1Biotech Research and Innovation Centre & Rigshospitalet, University of Copenhagen, Copenhagen, Denmark
  2. 2Department of Hematology, Lund University Hospital, Lund University, Lund, Sweden
  3. 3LRF Molecular Haematology Unit, Nuffield Department of Clinical Laboratory Sciences, John Radcliffe Hospital, University of Oxford, Oxford, UK
  4. 4Department of Biomedical Sciences University of Modena and Reggio Emilia, Modena, Italy
  5. 5Experimental Hematooncology Department, Medical University of Lublin, Lublin, Poland
  6. 6Servicio de Hematologia, Hospital Universitario de Salamanca and IBMCC, Centro de Investigacion del Cancer, Universidad de Salamanca-CSIC, Salamanca, Spain
  7. 7MLL Munich Leukemia Laboratory, Munich; Department of Microarrays and Next-Generation Sequencing, Munich, Germany
  8. 8Department of Hematology, Hemostasis, Oncology and Stem Cell Transplantation, Hannover Medical School, Hannover, Germany
  9. 9Erasmus MC—Sophia Children's Hospital, Erasmus University, Rotterdam, The Netherlands
  10. 10Center for Cancer Research & Cell Biology, Queen's University Belfast, Belfast, Northern Ireland
  11. 11Department of Internal Medicine III, University Hospital Ulm, Ulm, Germany

Correspondence: Dr L Bullinger, Department of Internal Medicine III, University Hospital of Ulm, Albert-Einstein-Allee 23, Ulm, Baden-Wurttemberg 89081, Germany. E-mail: lars.bullinger@uniklinik-ulm.de

12These authors contributed equally to this work.

Received 23 July 2010; Revised 28 January 2011; Accepted 10 February 2011; Published online 29 March 2011.

Top

Abstract

Today, the classification systems for myelodysplastic syndromes (MDS) and acute myeloid leukemia (AML) already incorporate cytogenetic and molecular genetic aberrations in an attempt to better reflect disease biology. However, in many MDS/AML patients no genetic aberrations have been identified yet, and even within some cytogenetically well-defined subclasses there is considerable clinical heterogeneity. Recent advances in genomics technologies such as gene expression profiling (GEP) provide powerful tools to further characterize myeloid malignancies at the molecular level, with the goal to refine the MDS/AML classification system, incorporating as yet unknown molecular genetic and epigenetic pathomechanisms, which are likely reflected by aberrant gene expression patterns. In this study, we provide a comprehensive review on how GEP has contributed to a refined molecular taxonomy of MDS and AML with regard to diagnosis, prediction of clinical outcome, discovery of novel subclasses and identification of novel therapeutic targets and novel drugs. As many challenges remain ahead, we discuss the pitfalls of this technology and its potential including future integrative studies with other genomics technologies, which will continue to improve our understanding of malignant transformation in myeloid malignancies and thereby contribute to individualized risk-adapted treatment strategies for MDS and AML patients.

Keywords:

gene expression profiling; acute myeloid leukemia; myelodysplastic syndrome; microarray; connectivity MAP; drug discovery

Top

Introduction

During the past decade, our understanding of the biology underlying myelodysplastic syndromes (MDS) and acute myeloid leukemia (AML) has improved tremendously. Both diseases are characterized by enormous biological and clinical heterogeneity, which is based on the accumulation of a variety of somatically acquired genetic aberrations in hematopoietic progenitor and/or stem cells (HPCs/HSCs), resulting in altered cell growth and differentiation.1, 2, 3 Today, cytogenetics still represents one of the most powerful prognostic factors in MDS and AML, with clonal chromosomal alterations detected in many MDS (~50%) and AML (~55%) patients at diagnosis. Furthermore, the incorporation of molecular genetic assays into diagnostic workflows has revealed numerous gene mutations that allowed further classification of MDS and AML into molecularly distinct subclasses. For example, in the large group of AML patients with cytogenetically normal AML (CN-AML), aberrations such as internal tandem duplications (ITD) of the FLT3 gene, as well as mutations of CEBPA and NPM1 have been shown to be of prognostic relevance, and the latter two molecular aberrations have recently been added as provisional entities into the current World Health Organization classification guidelines.4 However, the heterogeneity of MDS and AML is still not fully reflected by the current classification systems.

The advancement of genomics technologies has offered a range of experimental approaches to capture the molecular variation underlying the biological and clinical heterogeneity of MDS/AML. For example, microarray-based gene expression profiling (GEP) has been shown to be a very powerful tool that has been applied successfully in several MDS and AML studies (Table 1). In a milestone study, Golub et al.5 were able to distinguish AML and acute lymphoblastic leukemia (ALL) patients without previous knowledge of the respective leukemia classes. Significantly, this study for the first time demonstrated the power of GEP to predict known leukemia classes, and even to discover novel leukemia subclasses.


In this study, we provide a comprehensive review on the impact of GEP on MDS and AML over the last decade. The first sections focus on AML and MDS studies in adults and children, and discuss how GEP has helped (i) to identify characteristic gene expression patterns in MDS/AML subclasses, with specific cytogenetic and/or molecular genetic aberrations (class prediction), (ii) to discover novel leukemia subclasses (class discovery) and (iii) to predict response to standard therapy in both MDS and adult as well as pediatric AML (outcome prediction).6, 7 The final sections review potential applications of GEP including the prediction of drug response, drug discovery and future integrative data analysis approaches that will not only allow a refined patient management, but also provide deeper insights into the molecular mechanisms underlying MDS and AML.

GEP in adult AML

During the last years, GEP studies have shown that many cytogenetic or molecular genetic aberrations in AML are characterized by distinct gene expression profiles that, for some subclasses, even allow highly accurate prediction of the respective aberrations and can confer prognostic information independent of known genetic risk factors. Furthermore, analysis of altered gene expression can point to yet unknown molecular mechanisms underlying leukemogenesis, and thereby contribute further to the refinement of the molecular classification of AML.8, 9

Although in the early days it was sometimes doubted whether GEP could produce reproducible results, there is now strong evidence from numerous AML studies that GEP technology can easily be standardized for diagnostics, in as much distinct gene expression patterns are quite robust with respect to both high intra- and inter-platform comparability.10, 11, 12, 13, 14 Thus, GEP has been successfully applied to improve the molecular classification of AML. For example, genomics-based prediction of known leukemia classes (class prediction) has been shown to be feasible for well-defined cytogenetic AML subclasses of the World Health Organization classification guidelines such as AML with translocation t(8;21), inversion inv(16), t(15;17) and various translocations involving MLL (these are referred to as AML with t(11q23)/MLL rearrangements and include: AML with t(9;11)(p22;q23)/MLLT3-MLL; t(6;11)(q27;q23)/MLLT4-MLL; t(11;19) (q23;p13.3)/MLL-MLLT1; t(11;19)(q23;p13.1)/MLL-ELL; and t(10;11)(p12;q23)/MLLT10-MLL), thereby providing a powerful novel diagnostic tool. For example, the international multi-centre research program (MILE Study) centered around the European Leukemia Network (ELN, http://www.leukemia-net.org) performed a powerful exploratory retrospective stage I study, designed for biomarker discovery based on whole-genome expression profiles from 2143 patients with various types of chronic and acute leukemias as well as MDS. This stage I study was followed by a prospective stage II study validating the diagnostic accuracy of the GEP-based classifiers in an independent cohort of 1191 patients.15, 16 On the basis of 2096 patients, which included 542 AML cases, the stage I study achieved 92.2% classification accuracy and a median specificity of 99.7% for 18 distinct types of leukemias including AML with complex aberrant karyotype, CN-AML, t(11q23)/MLL-rearrangement, t(8;21), inv(16) and t(15;17). In accordance, in the second prospectively collected cohort of 1152 patients the classification scheme for AML and ALL cases reached a median sensitivity and specificity of 95.6 and 99.8%, respectively.16

Recently, multicenter trials have also demonstrated that distinct molecular aberrations in AML such as mutations of the NPM1 and CEBPA genes can be reliably diagnosed based on resulting characteristic gene expression signatures (Figure 1).15, 16, 17, 18 Interestingly, such gene expression signatures can also provide additional molecular insights as exemplified by the identification of the role of deregulated FLT3 expression in t(11q23)/MLL-rearranged leukemia, and might therefore better capture the underlying disease biology.19 For example, in CN-AML patients predicted to have a FLT3-ITD mutation based on the resulting gene expression patterns, the class prediction profiles outperformed the presence of the molecular marker FLT3-ITD with regard to impact on clinical outcome.20 Furthermore, in CEBPA-mutated CN-AMLs, only CEBPA double mutants, but not single mutants, seem to exhibit a characteristic gene expression profile.21

Figure 1.
Figure 1 - Unfortunately we are unable to provide accessible alternative text for this. If you require assistance to access this image, please contact help@nature.com or the author

Integrative signature for NPM1 and CEBPA mutations in CN-AML. (a) Visualization of the hierarchical clustering analysis results from a multicenter signature for CN-AML (n=233 cases, rows) and 461 differentially expressed genes (rows) that were calculated from two-group comparisons according to both NPM1 and CEBPA mutation status. Normalized gene expression values are color coded (standard deviation from mean): red indicates high and green low expression, respectively. (b) The integrated signature for NPM1 and CEBPA was further explored by principal component analysis. For all 233 CN-AML cases (colored according to NPM1 and CEBPA mutation status) each sphere represents a single gene expression profile based on the 461 differentially expressed probe sets. Ellipsoids with twofold standard deviation are drawn for wild-type NPM1, mutated NPM1 and mutated CEBPA cases (for more details see Kohlmann et al.18).

Full figure and legend (247K)

In addition to its role in class prediction, GEP has proved to be a very powerful means for the discovery of novel leukemia subclasses (class discovery). Initial studies screening large AML cohorts identified novel biologically and clinically meaningful subclasses, primarily through discovery of novel subclasses among the heterogeneous cohort of CN-AML cases.22, 23 Notably, a cohort of AML patients with a gene expression signature resembling CEBPA-mutated CN-AML were found to exhibit epigenetic silencing of CEBPA expression through hypermethylation of its promoter.24 Similarly, the identification of KIT wild-type core-binding factor AMLs harboring expression profiles similar to KIT-mutated core-binding factors suggests that different leukemogenic events can result in the activation and deregulation of identical sets of genes and/or oncogenic pathways.25

In accordance, gene expression signatures have been identified that reflect the heterogeneous clinical response of both CN-AML and core-binding factor AML, thereby providing further evidence that GEP-based outcome prediction in AML is feasible and can provide information in addition to cytogenetics and molecular genetics.22, 26 To discover gene expression signatures of prognostic value, a recent proof-of-principle study successfully applied a novel strategy including a combination of supervised and unsupervised data analysis and defined a 133-gene signature. Subsequent validation demonstrated the 133-gene signature to be a significant independent outcome predictor, both across all cytogenetic AML subclasses and within CN-AML cases.22 Importantly, the prognostic significance of these results was also confirmed by different study groups in independent AML cohorts despite several significant study differences including microarray platforms and patient treatment protocols.27, 28 To increase the potential clinical applicability, Radmacher et al.27 created a class-prediction algorithm allowing outcome prediction for individual patients, and recently GEP-based outcome prediction was further refined in large CN-AML cohorts.22, 26, 27, 29 Although these findings are encouraging and ongoing studies analyzing large patient cohorts are aiming to validate and improve CN-AML-specific classifiers, the rapid pace at which novel molecular markers of prognostic relevance are currently identified in AML reduces the rate of development of GEP-based outcome signatures that are truly independent of known molecular markers. In this study, an accurate outcome prediction signature that no longer needs to be independent of these markers, but has the ability to capture the prognostic impact of all possible marker combinations in a single measurement, might represent a valuable alternative to multiplex sequencing of the plethora of currently known molecular markers. This approach is particularly intriguing in as much irrelevant passenger mutations are unlikely to affect outcome prediction, whereas so-called driver mutations that are functionally and prognostically relevant will have a major impact on the gene expression profiles.

GEP in pediatric AML

Like adult AML, pediatric AML represents a heterogeneous disease where prognosis correlates with early therapy response, and cytogenetic as well as molecular genetic aberrations. Although current overall survival rates of pediatric AML patients are as high as 60%, further improvement is hampered by considerable short- and long-term toxicity of current therapeutic regimens. Thus, in pediatric AML, GEP studies have been performed with identical aims as described for in adult AML.

In a seminal study of 130 de novo pediatric AML patients, Ross et al.30 successfully identified class-discriminating gene expression signatures for ALL and AML.30 Likewise, the major prognostic AML subclasses, that is, t(15;17), t(8;21), inv(16) and t(11q23)/MLL, as well as cases classified as acute megakaryoblastic leukemia (cases exhibiting a FAB M7 morphology) were predicted with an overall classification accuracy greater than 93% using supervised learning algorithms. This high classification accuracy for the respective AML subclasses was confirmed and extended in an independent study of 237 children with pediatric AML (specificity and sensitivity for discovery of the indicated cytogenetic subclasses was 99 and 100%, respectively).31 In contrast to adult AML, no general predictive gene expression signatures were found for the molecular genetic aberrations NPM1, CEBPA, FLT3-ITD or KIT in pediatric AML. However, distinct gene expression signatures were discovered for FLT3-ITD in t(15;17) and CN-AML.31 Significantly, these studies demonstrate the potential of GEP in terms of class prediction also in childhood leukemia. Additionally, they indicate that gene expression profiles associated with specific genetic aberrations might be influenced by the cell lineage and maturation stage of the leukemia subclass, which in turn is most likely influenced by a specific combination of yet unknown genetic and epigenetic aberrations.

Nevertheless, recent studies have shown the potential to identify unique signatures for distinct genetic aberrations independently of the cellular context. For example, a combined analysis of a large series of ALL and AML patients identified a common gene expression signature for leukemias with MLL chimeric fusion genes irrespective of their lineage association, whereas AMLs with partial tandem duplications of the MLL gene (MLL-PTD) failed to cluster with those leukemias having MLL chimeric fusion genes.19, 30, 31 Another unsupervised study of 237 de novo pediatric AMLs showed profound clustering of the known relevant cytogenetic subgroups (Figure 2).31 Moreover, it was recently shown that high VEGFC expression is associated with unique gene expression profiles, and predicts adverse prognosis in pediatric as well as adult AML.32

Figure 2.
Figure 2 - Unfortunately we are unable to provide accessible alternative text for this. If you require assistance to access this image, please contact help@nature.com or the author

Unsupervised analysis of gene expression data revealed clustering of known cytogenetic subgroups in pediatric de novo AML. Pair-wise correlations between gene expression profiles of 237 de novo pediatric AML samples, which are calculated on the basis of 1608 probe sets (cutoff: 16-fold), are displayed in a correlation plot. Box colors represent the Pearson correlation coefficient with a color gradient ranging from deep blue—for a negative correlation to vivid red for a positive correlation of samples. Distinct clusters of samples are observed, which can be recognized by the red triangles indicating a high correlation of samples. The first column to the right of the correlation plot refers to the major cytogenetic subgroups to which the samples belong (dark blue: inv(16), pale red: t(8;21), bright blue t(15;17), yellow: t(7;12), light blue: 11q23-rearrangements, dark grey: normal cytogenetics, light grey: other aberrant cytogenetics and white: not available). Clear clustering of the recurrent cytogenetic subgroups is seen (based on Balgobind et al.31).

Full figure and legend (129K)

In addition to class prediction and discovery, prognostic gene expression signatures have also been reported in pediatric AML. For example, Bresolin et al.33 showed non-AML-like and AML-like gene expression signatures discriminating juvenile myelo-monocytic leukemia, with the AML-like signature being significantly associated with adverse outcome. Similarly, investigation of pediatric AML patients with FLT3-mutations allowed the identification of a RUNX3 to ATRX expression ratio that correlated with outcome, irrespective of the FLT3 mutation status,34 and a small study in pediatric acute myelomonocytic leukemia (characterized by FAB M4/M5 morphology) revealed age-associated gene expression differences that might be associated with outcome.35 Thus, like in adult AML, results of pediatric studies hold great promise that GEP-based prognostication is feasible.

In line with these findings, novel information on the molecular pathomechanisms underlying pediatric AML can be uncovered by GEP. For example, pediatric MDS and MDS-related AML showed differential expression of genes related to endocytosis and protein transport.36 Also, similarly as in adult AML, NPM1-mutated pediatric AML was associated with deregulation of homeobox genes, which significantly differed from HOX gene deregulation in MLL-rearranged pediatric AML, thereby suggesting for the first time different routes of perturbed HOX gene expression in pediatric AML subclasses.37 Insights into the function of leukaemia-associated antigens were recently gained from investigating the biology of PRAME (preferentially expressed antigen of melanoma) in pediatric AML showing PRAME-positive cases to harbor an increased expression of genes encoding ABC transporters, such as multidrug resistance proteins and decreased expression of genes encoding apoptotic proteins.38 And recently like in adult AML,24 co-clustering of CEBPA-mutated AML with CEBPA wild-type samples identified promoter hypermethylation-based CEBPA silenced cases.31

Thus, in both adult and pediatric AML, GEP studies have proven to be of value for class prediction, class discovery, outcome prediction and the identification of novel biological insights. However, especially in pediatric AML, the value of diagnostic and prognostic signatures needs confirmation. Given the rarity of pediatric AML, prospective studies of larger cohorts will require international collaborations. In this study, the COST (Cooperation in Science and Technology)-sponsored action EuGESMA (The European Genomic and Epigenetic Study on MDS and AML) as well as the European LeukemiaNet (ELN) can provide such collaborative networks of excellence (see http://www.qub.ac.uk/research-centres/EuGESMA/ and http://www.leukemia-net.org/content/home/).

GEP in MDS

In contrast to the wealth of GEP data for most leukemia and lymphoma entities, the MDS have not been analyzed extensively, yet. The heterogeneity of MDS has been seen as a disadvantage for GEP-based classifications, but several studies have compared gene expression profiles between MDS and healthy individuals, between different stages of MDS or between MDS-derived AML and de novo AML (Figure 3).39, 40, 41, 42 Most of these studies have been based on CD34+ or CD133+ cell populations isolated from normal/healthy and MDS bone marrow providing valuable insights into the molecular pathomechanisms of MDS. However, other studies have proven that it is also feasible to use unfractionated mononuclear bone marrow cells to identify diagnostic or prognostic molecular signatures.16, 43

Figure 3.
Figure 3 - Unfortunately we are unable to provide accessible alternative text for this. If you require assistance to access this image, please contact help@nature.com or the author

GEP in MDS. GEP-based classification distinguishing between MDS cases with either refractory anemia or refractory anemia with ringed sideroblasts and healthy individuals (NOL). Results of a supervised class comparison analysis have been visualized by a hierarchical cluster analysis. Normalized gene expression values are color coded (standard deviation from mean): red indicates high and green low expression, respectively (unpublished data provided by Drs Hernandez-Rivas and Mills).

Full figure and legend (319K)

With regard to biological insights, an early study analyzing CD34+ MDS cells (n=55) revealed expression profiles similar to interferon-γ-induced gene expression changes in normal CD34+ cells.41 Interestingly, distinct gene expression profiles were associated with specific FAB and cytogenetic subclasses and CD34+ cells from patients with refractory anemia with ringed sideroblasts were characterized by deregulation of mitochondrial-related genes. Subsequent studies investigating ringed sideroblasts showed that expression of the iron transporter ABCB7 was low in ringed sideroblasts patients and observed a strong relationship between increasing percentage of bone marrow ring sideroblasts and decreasing ABCB7 gene expression levels,44 and confirmed the deregulation of mitochondrial function or integrity genes.45

Investigation of the MDS subgroup with loss of the long arm of chromosome 5 (5q−) also provided novel insights. Comparison of 5q− MDS stem cells versus normal HSCs (CD34+, CD38− and Thy1+) showed the upregulation of a number of significant genes in 5q− stem cells: BMI1 (which is known to have a role in HSC and neural stem cell self-renewal), ID1 (which is involved in preventing the premature differentiation of HSCs and ES cells), DNMT3A (which is involved in DNA methylation regulation) and MYC (which has a role in stem cell regulation).46 Expression of CEBPA, encoding a myeloid transcription factor, was also upregulated in the 5q− stem cell fraction, but downregulated in 5q− progenitor cells (CD34+, CD38+ and Thy1−) compared with normal HSCs and HPCs, respectively. Other studies have reported haploinsufficiency of the tumor suppressor genes SPARC and RPS14, a component of the 40S ribosomal subunit, in patients with 5q− syndrome.47, 48 Interestingly, similarities in the defective gene expression patterns observed in the CD34+ cells of patients with Diamond Blackfan anemia and patients with the 5q− syndrome were recently demonstrated, including the downregulation of multiple ribosomal genes and genes involved in translation initiation, suggesting that the 5q− syndrome represents a disorder of aberrant ribosome biogenesis.47 Downregulation of multiple ribosomal genes in MDS patients with the 5q− has been confirmed recently.49

In general, GEP analysis of a large cohort of CD34+ MDS cell samples (n=183) revealed the most deregulated gene pathways to be interferon signaling, thrombopoietin signaling and the Wnt pathways.50 Early MDS stages showed deregulation of immunodeficiency, apoptosis and chemokine signaling pathways, whereas advanced MDS was characterized by deregulation of DNA damage response and checkpoint pathways. Distinct gene expression profiles and deregulated gene pathways were identified in patients with 5q−, trisomy 8 or −7/del(7q). The deregulated pathways identified in MDS illuminate the molecular pathogenesis of this disorder, thereby providing new targets for therapeutic intervention.50

Further and larger studies, mainly on AML patients, showed that it is possible to identify existing classification groups, but also showed that some of the GEP-derived subclasses could be correlated to clinical outcome.22, 23, 51 However, these studies also highlighted the need for a large multi-center comparison of GEP-based diagnosis with standard diagnostics as a global approach to leukemia classification.15 In this study, the MILE study assessed the application of a microarray test using unfractionated bone marrow and its potential for use in the diagnosis and sub-classification of acute and chronic leukemia subclasses, including MDS.16 However, unlike for AML, in the early stages the diagnostic classifier correctly identified only ~50% of the unfractionated MDS specimens submitted to the MILE study. The prediction results of the other samples were split approximately equally between ‘none of the target classes’ and AML signatures. Interestingly, this distinction was reflected in clinical outcome in terms of a highly significant correlation with GEP-based classification (MDS, AML or ‘normal’) and time to AML transformation.43 Notably, the performance of this diagnostic classifier was not directly correlated with the amount of blast cells in the sample indicating that this morphological feature is only a proxy for the aggressiveness of the stage of disease, and further indicating that the current classification needs to be further refined. Furthermore, an improved prognostic classifier was developed that had a very high correlation with both time to AML transformation and overall survival.43 The prognostic classifier included ‘AML-associated’ genes, including HOXA9, KIT, NPM1 and WT1, whose expression correlated with the transformation process. The predictive value of gene expression signatures was also confirmed using marrow CD34+ cells from MDS patients.49 In this study, a ‘poor risk’ signature associated with AML transformation was shown to be enriched for ‘ribosomal’, MYC and WNT signaling targets.49

Overall, the analysis of MDS subclasses, diagnostics approaches and outcome prediction have been less advanced than for AML and ALL. This is due to the wide variation in numbers of blast cells and has led to two directions of analysis. Sub-population analysis of sorted immature cell populations has identified a series of genes probably related to the pathophysiology of the disease subclasses, whereas the analysis of unfractionated BM has been valuable to characterize disease classification and prognosis. Both the directions will contribute now to the molecular understanding of disrupted gene expression in MDS, which forms the basis for improved MDS therapy and patient outcome.

GEP-based prediction of drug response or resistance

Recently, GEP has been successfully applied to predict response or resistance to different therapeutic modalities in various cancer entities including leukemia. This is of great importance as in addition to the currently used cytotoxic standard chemotherapy regimens in AML, there is increasing evidence for the potential of drugs that can overcome the differentiation block of AML cells. The huge potential of differentiation therapy is particularly striking in acute promyelocytic leukemia (APL), where therapeutic regimens including all-trans-retinoic acid (ATRA) and more recently arsenic trioxide have improved the clinical outcome significantly.52, 53 This tremendous success of differentiation therapy in APL has raised the question whether AML subclasses other than APL might respond to differentiation therapy, a scenario supported by recent studies demonstrating an ATRA response in non-APL AML.54, 55, 56 However, differentiation following ATRA treatment seems to be restricted to a subgroup of non-APL AML, which might be predicted by GEP.57, 58, 59

In an elegant GEP study, Tagliafico et al.60 used AML cell lines and defined a gene expression signature capable to predict ATRA-mediated differentiation of 28 primary AMLs in-vitro. To further validate the identified ATRA response signature in a clinical context, they generated a clinical ATRA response signature using GEP data from the AMLSG Trial HD98B, which recently demonstrated that ATRA administered after intensive chemotherapy significantly improved the outcome of older non-APL AML patients.61 Strikingly, a supervised analysis showed an overlap of genes contained within both the clinical- and the in-vitro ATRA response signatures (Tagliafico E, unpublished observations). Importantly, the latter indicates a certain robustness of ATRA response signatures as in-vitro signatures based on ATRA treatment alone, and clinical signatures reflecting a synergistic effect of both ATRA and chemotherapy share a distinct set of genes (Tagliafico E, unpublished observations).

Consistent with these findings, other studies have demonstrated the potential of GEP to predict response or resistance of MDS/AML to specific drugs. For example, in an attempt to identify patients likely to respond to treatment with a farnesyltransferase inhibitor, Raponi et al.28 analyzed gene expression profiles of bone marrow samples from elderly and previously untreated AML patients enrolled in a phase II trial investigating the farnesyltransferase inhibitor tipifarnib as single agent. Comparison of gene expression profiles from responders (that is patients with complete remission and hematologic improvement) and non-responders (that is patients with progressive disease) identified a 2-gene classifier, the RASGRP1/APTX gene expression ratio, which was highly predictive of farnesyltransferase inhibitor response (sensitivity, 92.3%; specificity, 100%; negative predictive value, 92.9%; positive predictive value, 100%). Subsequent validation using quantitative PCR and an independent data set of gene expression profiles obtained from relapsed or refractory AML patients treated with single agent tipifarnib confirmed the predictive power of the identified RASGRP1/APTX classifier in terms of farnesyltransferase inhibitor response.28

Another study exemplifying the successful use of GEP to predict response to novel drugs aimed to predict a lenalidomide (revlimid) response in MDS lacking a 5q deletion, as lenalidomide treatment not only has demonstrated significant clinical response in 5q− MDS but also in a minority of MDS patients lacking 5q deletions.62, 63 In a seminal study, Ebert et al.64 compared responder and non-responder gene expression profiles from non-5q− MDS patients and identified genes associated with lenalidomide response. Notably, the identified lenalidomide response signature contained primarily genes specific to the erythroid differentiation program, which was markedly downregulated or absent in responders compared with non-responders. Consistent with this finding lenalidomde not only induced the observed response signature in CD34+ cells from healthy subjects, but also promoted their erythroid differentiation. Additional validation using an independent data set confirmed the predictive power of the lenalidomide response signature for both MDS patients with and without 5q deletions.64 The latter indicates that a distinct molecular phenotype in MDS, that is absence of erythroid differentiation, predicts clinical response to lenalidomide.

Taken together, identification of drug response signatures by GEP holds great promise as a powerful tool for the stratification of MDS/AML patients to specific treatment modalities. Moreover, the correlation of drug response signatures from clinical trials and in-vitro studies suggests that the assessment of drug responsiveness in individual MDS/AML patients using in-vitro assays might be feasible in the near future. However, transferring the potential of GEP into therapeutic benefits for MDS/AML patients represents a highly demanding task that will require prospective GEP within large clinical trials.

GEP as a tool for identification of drugs with therapeutic potential in AML

AML emerges as a consequence of accumulating independent genetic aberrations that direct deregulation and/or dysfunction of genes resulting in abnormal activation of signaling pathways, resistance to apoptosis and uncontrolled proliferation.65, 66 So far more than 100 genetic aberrations presenting in various combinations have been identified in AML.3, 4 Consistent with this considerable heterogeneity of AML genomes, AML patients demonstrate highly variable clinical responses to conventional therapeutic regimens consisting of a cytarabine and anthracycline backbone.4, 65 Although the response to current and novel therapeutic approaches can already be predicted by GEP, the current clinical situation calls also for the development of more effective therapies that specifically target deregulated genes and/or pathways of either distinct subclasses of AML or, at best, individual AML patients.

As discussed in the previous sections, GEP studies have demonstrated that AML gene expression profiles correlate well with specific genetic aberrations and can predict clinical outcome. In other words, gene expression profiles represent genomic surrogates of distinct biological phenotypes that emerge as a consequence of specific combinations of genetic aberrations. Significantly, gene expression profiles also provide encrypted information on those genes and pathways that are deregulated and contribute to malignant transformation. Consistent with these findings, recent studies have demonstrated the feasibility of translating genomic information provided by cancer gene expression profiles into targeted therapy. A most elegant study conducted by Chang et al.67 defined gene expression signatures that correlated with activity of oncogenic pathways. Significantly, Chang et al. demonstrated that signatures for RAS, E2F and EGFR pathways allowed classification of distinct cancer entities, and were predictive for clinical outcome and response to pathway-specific drugs. Two other studies applied gene expression signatures that are associated with the inhibition of leukemic cell proliferation and viability. In one study, Corsello et al.68 defined a signature for RUNX1-RUNX1T1 (AML1-ETO) inhibition by comparing gene expression profiles of a leukemia cell line harboring the t(8;21) translocation before and after RNAi-mediated knockdown of RUNX1-RUNX1T1. Subsequently, the RUNX1-RUNX1T1 inhibition signature was matched with a library of drug expression signatures, which identified corticosteroids and dihydrofolate reductase inhibitors as novel drugs that inhibit RUNX1-RUNX1T1-associated gene expression and induce cellular differentiation and ultimately apoptosis. Another study matched the signature of the nuclear factor-κB inhibitor parthenolide with a library of drug signatures and identified two novel agents that, similarly to parthenolide, eradicated primary AML stem cells and progenitor cells in functional assays.69 Finally, a recent study reported an alternative, less biased strategy for the identification of novel drugs and drug targets in AML (Figure 4). In this study, Marstrand et al.70 identified signatures of deregulated genes in a subtype of AML, namely APL, by comparison of gene expression profiles from APL cells and their normal counterparts, that is promyelocytes, purified form healthy subjects. Subsequently, APL signatures were applied to a series of computational analyses including an in silico screen of the Connectivity MAP drug signature database.71 These analyses led to the finding that APL cells exhibit stem cell properties in terms of gene expression and transcriptional regulation, and importantly identified novel candidate drugs and targets for therapeutic interventions.70 Significantly, the identified candidate drugs including a histone deacetylase inhibitor (trichostatin A), an inhibitor of phosphoinositide 3-kinase (PI3K) (LY294002), and an inhibitor of phospholipase A2 (PLA2) (quinacrine), were all shown to inhibit proliferation and induce apoptosis of a human APL cell line (NB4).

Figure 4.
Figure 4 - Unfortunately we are unable to provide accessible alternative text for this. If you require assistance to access this image, please contact help@nature.com or the author

GEP-based identification of drugs and drug targets. The figure demonstrates a flow chart of the conceptual framework including all databases and computational tools that have been applied to identify candidate drugs and drug targets in APL. Briefly, GEO microarray gene expression profiles of acute promyelocyte leukemia cells (APL), normal promyelocytes (PM), hematopoietic stem cells (HSC), early hematopoietic progenitor cells (HPC) and APL cells differentiated by 5 days treatment with all-trans retinoic acid (APL-refractory anemia) were used to generate APL-, stemness- and APL-refractory anemia signatures. The identified signatures were subjected to a series of computational analyses using various bioinformatics tools (gene set enrichment (GSEA) analysis, promoter analysis and random Forest analysis) and publically available databases providing (i) position weight matrices for transcription factor-binding sites (JASPAR and TRANSFAC) and (ii) drug signatures (connectivity MAP (CMAP)). The identified drug candidates were validated in vitro using the human APL cell line (NB4). Significantly, all of the four identified drug candidates inhibited proliferation and induced apoptosis in the human APL cell line at micromolar concentrations.70

Full figure and legend (114K)

Collectively, the above studies hold great promise that GEP-based genomic strategies can be used to prospectively define the most effective drug combinations for individual AML patients on a rational basis.

GEP—link with other ‘omics’ screenings

During the next years, integration of GEP data with other ‘omics’ data like genomic, epigenomic and proteomic data holds great promise that ‘omics’ technologies will continue to contribute to improve the understanding of leukemia biology. For example, recent studies showed that deregulated miRNA expression has an important role in leukemogenesis,72 and in accordance with GEP large-scale genome-wide miRNA profiling revealed miRNA expression profiles characteristic for cytogenetic subclasses,73, 74 as well as molecular genetic subclasses like AML with CEBPA and NPM1 mutations or deregulated MN1 expression.73, 75, 76 However, correlative studies intersecting GEP and miRNA expression profiling have been limited as it was thought that miRNAs mainly act via translational inhibition. In this study, the recent observation that miRNAs primarily affect changes in mRNA levels as well as protein levels by destabilization of target mRNAs rather than translational inhibition, provides new avenues for combining the information from miRNA and GEP studies.77

With regard to ‘omics’ other than transcriptomics, genome-wide array CGH and/or single nucleotide polymorphism analyses have revealed a large number of leukemia-specific copy number alterations (CNAs) as well as acquired homozygosity in the form of segmental uniparental disomy (UPD).78 Parallel analysis of GEP and array CGH/single nucleotide polymorphism-defined CNAs has facilitated delineation of candidate genes in the respective genomic regions, as for example, demonstrated in AML with complex aberrations.79 In contrast to ALL, the analysis of CNAs in AML has yielded only a small number of CNAs in most genomes of adult and pediatric AML patients, and the spectrum of AML-associated CNAs seems to be mainly non-recurrent.80, 81, 82 These findings collectively illustrate the enormous genetic heterogeneity of AML and suggest that the potential of the extensive catalogs of CNAs might be better exploited by future integrative analyses.

Similarly, first comprehensive genome-wide promoter DNA methylation profiling studies in AML demonstrated that cytogenetic and molecular genetic subclasses harbor distinct methylation patterns, but also seem to share a ‘common’ DNA methylation pattern, which can be intersected with aberrant gene expression.83 Although there are numerous additional studies providing evidence for the role of aberrant DNA methylation in AML,84 first integrative studies also provided evidence that for example, an outcome predictor based on the combined DNA methylation and gene expression information can outperform predictors based solely on DNA methylation patterns or GEP.85 In this study, future efforts combining analyses of various types of epigenetic modifications and GEP in AML will provide additional insights.86

Finally, proteomic approaches have been used to address several essential questions in normal and malignant haematopoiesis including studies on drug response and survival of AML patients.87, 88, 89, 90, 91, 92, 93 In this study, integrative approaches are important, as mRNA expression does not always correlate with protein expression levels due to for example, mRNA instability, post-transcriptional regulation, translation initiation and protein stability.94, 95 Moreover, mRNA splice variants and post-translational modifications, which cannot be directly detected by microarray platforms based on 3′ IVT (in-vitro transcription)-labeling protocols, can influence several critical parameters controlling the resulting protein function, including protein localization, protein activity, interaction with other molecules (for example, other proteins, nucleic acids and lipids) and turn-over. As most cell phenotypes can be ascribed to protein expression levels and activities/interactions, combined characterization of AML phenotypes by GEP and proteomic analysis is clearly necessary for a full understanding of normal and disease-specific biological processes, as well as how the proteome is reflected by the transcriptome. Furthermore, as most targeted therapeutics applied in the clinic are designed to interact with specific proteins (for example, topoisomerase I/II inhibitors, ABL1 kinase inhibitors, FLT3 inhibitors, VEGFR inhibitors, EGFR inhibitors, protease inhibitors and so on), a critical component of small molecule drug discovery and validation (including target specificity, resistance mechanisms) is the investigation of the proteome within the context of the transcriptome. For example, a first large proteomics study in primary AML samples used reverse-phase protein arrays to detect expression and post-translational modification (phosphorylation) of 51 proteins in a cohort of 256 AML patients.92 Proteomic profiling allowed the identification of patterns associated with the cellular phenotype, cytogenetic aberrations and patient outcome. In contrast, a recent study combining GEP and proteomic analysis to investigate embryonic stem cell differentiation into hematopoietic cells showed little correlation among differentially expressed mRNA transcripts and corresponding nuclear proteins.88 This demonstrates that modulation of transcription does not directly affect levels of many nuclear proteins, and supports the need for systematic studies combining proteome and transcriptome analyses.

For the future, the challenge remains to discover the pertinent proteomic alterations, which determine the phenotypic changes identified by GEP studies. Although changes in expression of a particular mRNA transcript may not correlate with similar modulation of the corresponding protein expression or activity, it may be possible to integrate GEP and protein profiling data to define a set of transcripts (that is a gene expression signature), whose alteration corresponds to meaningful changes at the protein level and activity. In this sense, hierarchical clustering of GEP data might be used to guide development of more focussed proteomic profiling strategies, which will greatly reduce analysis time and cost. Alternatively, proteomics may eventually be used to guide development of more specific GEP approaches. It is evident that strategies, utilizing both GEP and other ‘omics’ profiling, will further elucidate clinically relevant modulation of specific cellular pathways/functions allowing us to understand and exploit differences between normal and disease-specific processes and phenotypes. Combination of these powerful methodological approaches might also be fundamental for the development of more efficacious therapeutic regimens.

Top

Conclusions/perspectives

The vast genetic heterogeneity and variable clinical responses among MDS/AML patients is only partially explained by current cytogenetic and molecular genetic diagnostics, as these technologies do not provide diagnostic information at the genomic level. Hence, implementation of genomics technologies is needed to further improve the current World Health Organization classification system of MDS/AML, and most importantly to develop more effective therapies to specifically treat distinct subclasses of MDS/AML or, optimally, individual patients.

Microarray-based GEP as well as new technologies such as exon microarrays and the large-scale implementation of next generation sequencing will provide additional insights into the altered transcriptome in MDS/AML with regard to for example, alternative splice variants as well as 3′ mRNA modifications leading to altered mRNA stability and miRNA processing.96 Nevertheless, despite the robustness and high reproducibility of GEP-derived findings it remains questionable whether microarray-based GEP will have an important role in the diagnosis of MDS, and AML and whether it might be implemented to the clinic in the near future. With the notable exception of the MILE study,16 so far there has been a relative paucity of other prospective and confirmatory studies, although recent studies performed within large AML study groups also suggest the potential clinical usefulness of GEP.17, 18 Although critics might state a lack of clear-cut advantages in comparison with standard cytogenetic and molecular genetic diagnostics, GEP-based diagnostic platforms might be useful for the identification of patients with masked fusion proteins (like t(15;17) and t(8;21)) that are not detected by conventional cytogenetics or fluorescence in situ hybridization.97, 98 In addition, GEP-based diagnostic platforms can provide information comparable to that of multiple different cytogenetic and molecular genetic analysis within a single experiment. Thus, a fully integrated and robust microarray platform such as the AMLprofiler, which is currently evaluated within a prospective clinical trial, might deliver faster results and improve patient classification by allowing determination of the presence of t(8;21), t(15;17), inv(16), CEBPA double mutants, NPM1 mutations, as well as high expression levels of genes that recently have been shown to predict poor outcome (that is MECOM (EVI1), MN1, BAALC and ERG) within a single test (Skyline Diagnostics, Rotterdam, Netherlands http://www.amlprofiler.com/site/).75, 99, 100 Future developments like the AMLprofiler might help overcome the current lack of therapeutic consequences even in MDS/AML cases, where microarray-based classifiers could improve classification or prognostication provided future research efforts can define and validate robust surrogates for the respective leukemia classes.

However, the question remains why some genomic aberrations can be easily detected by class prediction, whereas others are more difficult to predict. Supposedly, some genetic aberrations such as t(15;17), t(8;21) or inv(16) are ‘drivers’ (that is primary events) of the leukemic phenotype and therefore, most likely have a strong effect on gene regulation and can dramatically change a cell's transcriptome. In accordance, NPM1 and CEBPA mutations display distinct signatures, whereas FLT3 mutations likely represent ‘passengers’ (that is secondary events) that occur in combination with many other genetic aberrations and thus, are more difficult to detect based on expression patterns due to more prominent overlying ‘driver’ signatures. Furthermore, yet unknown hits and FLT3 aberrations might both result in the deregulation of an identical tyrosine kinase-signaling cascade thereby hampering the possibility for an accurate discrimination. And finally, there is evidence that gene expression profiles also depend on patient age, lineage-specificity and maturation stage of the leukemic cells, and the leukemic cell of origin. For example, leukemias arising from committed hematopoietic progenitor cells (HPCs) versus HSCs following transformation with the oncogenic fusion gene MLL-AF9, display significantly different gene expression signatures pointing to the different origin.101 However, part of these current challenges might be solved by (i) future studies applying unbiased whole-genome transcriptome sequencing and large combined data sets that will allow the molecular heterogeneity of AML to be taken into account, and (ii) by studies comparing pediatric and adult AMLs and investigating for example, age and cell of origin-dependent differences.

With regard to the impact of GEP on further deciphering leukemia biology and discovering both novel drugs and druggable targets, GEP data will need to be integrated with genomic, epigenomic and proteomic data, and correlated with clinical information. In this study, new systems biology approaches combining various ‘omics’—technologies continue to contribute to the exploration of MDS and AML. Furthermore, data from functional screens will need to be taken into account, as for example, RNA interference screens can reveal novel disease mechanisms, such as non-oncogene addiction.102 In addition, most of the current studies in MDS and AML have been performed with ‘bulk’ tumor material. In order to even better understand the pathomechanisms of MDS/AML, sophisticated analyses of prospectively purified tumor subpopulations such as leukemic stem cells might shed additional light on the molecular basis of MDS/AML.103 Finally, future analyses will also need to focus on the influence of the tumor microenvironment on gene expression profiles in MDS/AML, as there is accumulating evidence of the utmost importance of the stem cell niche, despite it being yet poorly understood.104

Ultimately, GEP will continue to contribute to an improved understanding of the complex nature underlying MDS and AML. However, the vast complexity of GEP-based data requires large studies/cohorts to obtain valid data of clinical significance, and these will only be feasible by combining future efforts within well-structured leukemia research networks such as the above mentioned COST initiative and the ELN.

Top

Conflict of interest

The authors declare no conflict of interest.

Top

References

  1. Corey SJ, Minden MD, Barber DL, Kantarjian H, Wang JC, Schimmer AD. Myelodysplastic syndromes: the complexity of stem-cell diseases. Nat Rev Cancer 2007; 7: 118–129. | Article | PubMed | ISI | ChemPort |
  2. Nimer SD. Myelodysplastic syndromes. Blood 2008; 111: 4841–4851. | Article | PubMed | ISI | ChemPort |
  3. Dohner K, Dohner H. Molecular characterization of acute myeloid leukemia. Haematologica 2008; 93: 976–982. | Article | PubMed | ISI | ChemPort |
  4. Dohner H, Estey EH, Amadori S, Appelbaum FR, Buchner T, Burnett AK et al. Diagnosis and management of acute myeloid leukemia in adults: recommendations from an international expert panel, on behalf of the European LeukemiaNet. Blood 2010; 115: 453–474. | Article | PubMed | ISI |
  5. Golub TR, Slonim DK, Tamayo P, Huard C, Gaasenbeek M, Mesirov JP et al. Molecular classification of cancer: class discovery and class prediction by gene expression monitoring. Science 1999; 286: 531–537. | Article | PubMed | ISI | ChemPort |
  6. Baldus CD, Bullinger L. Gene expression with prognostic implications in cytogenetically normal acute myeloid leukemia. Semin Oncol 2008; 35: 356–364. | Article | PubMed | ISI |
  7. Wouters BJ, Lowenberg B, Delwel R. A decade of genome-wide gene expression profiling in acute myeloid leukemia: flashback and prospects. Blood 2009; 113: 291–298. | Article | PubMed | ISI | ChemPort |
  8. Bullinger L, Valk PJ. Gene expression profiling in acute myeloid leukemia. J Clin Oncol 2005; 23: 6296–6305. | Article | PubMed | ISI | ChemPort |
  9. Bacher U, Kohlmann A, Haferlach T. Current status of gene expression profiling in the diagnosis and management of acute leukaemia. Br J Haematol 2009; 145: 555–568. | Article | PubMed | ISI |
  10. Kohlmann A, Schoch C, Dugas M, Rauhut S, Weninger F, Schnittger S et al. Pattern robustness of diagnostic gene expression signatures in leukemia. Genes Chromosomes Cancer 2005; 42: 299–307. | Article | PubMed | ISI | ChemPort |
  11. Kohlmann A, Kipps TJ, Rassenti LZ, Downing JR, Shurtleff SA, Mills KI et al. An international standardization programme towards the application of gene expression profiling in routine leukaemia diagnostics: the Microarray Innovations in LEukemia study prephase. Br J Haematol 2008; 142: 802–807. | Article | PubMed | ISI | ChemPort |
  12. Kohlmann A, Haschke-Becher E, Wimmer B, Huber-Wechselberger A, Meyer-Monard S, Huxol H et al. Intraplatform reproducibility and technical precision of gene expression profiling in 4 laboratories investigating 160 leukemia samples: the DACH study. Clin Chem 2008; 54: 1705–1715. | Article | PubMed | ISI | ChemPort |
  13. Nilsson B, Andersson A, Johansson M, Fioretos T. Cross-platform classification in microarray-based leukemia diagnostics. Haematologica 2006; 91: 821–824. | PubMed | ISI | ChemPort |
  14. Shi L, Reid LH, Jones WD, Shippy R, Warrington JA, Baker SC et al. The MicroArray Quality Control (MAQC) project shows inter- and intraplatform reproducibility of gene expression measurements. Nat Biotechnol 2006; 24: 1151–1161. | Article | PubMed | ISI | ChemPort |
  15. Haferlach T, Kohlmann A, Schnittger S, Dugas M, Hiddemann W, Kern W et al. Global approach to the diagnosis of leukemia using gene expression profiling. Blood 2005; 106: 1189–1198. | Article | PubMed | ISI | ChemPort |
  16. Haferlach T, Kohlmann A, Wieczorek L, Basso G, Te Kronnie G, Bene MC et al. Clinical utility of microarray-based gene expression profiling in the diagnosis and subclassification of leukemia: report from the International Microarray Innovations in Leukemia Study Group. J Clin Oncol 2010; 28: 2529–2537. | Article | PubMed | ISI | ChemPort |
  17. Verhaak RG, Wouters BJ, Erpelinck CA, Abbas S, Beverloo HB, Lugthart S et al. Prediction of molecular subtypes in acute myeloid leukemia based on gene expression profiling. Haematologica 2009; 94: 131–134. | Article | PubMed | ISI |
  18. Kohlmann A, Bullinger L, Thiede C, Schaich M, Schnittger S, Dohner K et al. Gene expression profiling in AML with normal karyotype can predict mutations for molecular markers and allows novel insights into perturbed biological pathways. Leukemia 2010; 24: 1216–1220. | Article | PubMed | ISI |
  19. Armstrong SA, Staunton JE, Silverman LB, Pieters R, den Boer ML, Minden MD et al. MLL translocations specify a distinct gene expression profile that distinguishes a unique leukemia. Nat Genet 2002; 30: 41–47. | Article | PubMed | ISI | ChemPort |
  20. Bullinger L, Dohner K, Kranz R, Stirner C, Frohling S, Scholl C et al. An FLT3 gene-expression signature predicts clinical outcome in normal karyotype AML. Blood 2008; 111: 4490–4495. | Article | PubMed | ISI | ChemPort |
  21. Wouters BJ, Lowenberg B, Erpelinck-Verschueren CA, van Putten WL, Valk PJ, Delwel R. Double CEBPA mutations, but not single CEBPA mutations, define a subgroup of acute myeloid leukemia with a distinctive gene expression profile that is uniquely associated with a favorable outcome. Blood 2009; 113: 3088–3091. | Article | PubMed | ISI | ChemPort |
  22. Bullinger L, Dohner K, Bair E, Frohling S, Schlenk RF, Tibshirani R et al. Use of gene-expression profiling to identify prognostic subclasses in adult acute myeloid leukemia. N Engl J Med 2004; 350: 1605–1616. | Article | PubMed | ISI | ChemPort |
  23. Valk PJ, Verhaak RG, Beijen MA, Erpelinck CA, Barjesteh van Waalwijk van Doorn-Khosrovani S, Boer JM et al. Prognostically useful gene-expression profiles in acute myeloid leukemia. N Engl J Med 2004; 350: 1617–1628. | Article | PubMed | ISI | ChemPort |
  24. Wouters BJ, Jorda MA, Keeshan K, Louwers I, Erpelinck-Verschueren CA, Tielemans D et al. Distinct gene expression profiles of acute myeloid/T-lymphoid leukemia with silenced CEBPA and mutations in NOTCH1. Blood 2007; 110: 3706–3714. | Article | PubMed | ISI | ChemPort |
  25. Luck SC, Russ AC, Du J, Gaidzik V, Schlenk RF, Pollack JR et al. KIT mutations confer a distinct gene expression signature in core binding factor leukaemia. Br J Haematol 2010; 148: 925–937. | Article | PubMed | ISI |
  26. Bullinger L, Rucker FG, Kurz S, Du J, Scholl C, Sander S et al. Gene-expression profiling identifies distinct subclasses of core binding factor acute myeloid leukemia. Blood 2007; 110: 1291–1300. | Article | PubMed | ISI | ChemPort |
  27. Radmacher MD, Marcucci G, Ruppert AS, Mrozek K, Whitman SP, Vardiman JW et al. Independent confirmation of a prognostic gene-expression signature in adult acute myeloid leukemia with a normal karyotype: a Cancer and Leukemia Group B study. Blood 2006; 108: 1677–1683. | Article | PubMed | ISI | ChemPort |
  28. Raponi M, Lancet JE, Fan H, Dossey L, Lee G, Gojo I et al. A 2-gene classifier for predicting response to the farnesyltransferase inhibitor tipifarnib in acute myeloid leukemia. Blood 2008; 111: 2589–2596. | Article | PubMed | ISI | ChemPort |
  29. Metzeler KH, Hummel M, Bloomfield CD, Spiekermann K, Braess J, Sauerland MC et al. An 86-probe-set gene-expression signature predicts survival in cytogenetically normal acute myeloid leukemia. Blood 2008; 112: 4193–4201. | Article | PubMed | ISI | ChemPort |
  30. Ross ME, Mahfouz R, Onciu M, Liu HC, Zhou X, Song G et al. Gene expression profiling of pediatric acute myelogenous leukemia. Blood 2004; 104: 3679–3687. | Article | PubMed | ISI | ChemPort |
  31. Balgobind BV, Van den Heuvel-Eibrink MM, De Menezes RX, Reinhardt D, Hollink IH, Arentsen-Peters ST et al. Evaluation of gene expression signatures predictive for cytogenetic and molecular subtypes of pediatric acute myeloid leukemia. Haematologica 2011; 96: 221–230. | Article | PubMed | ISI |
  32. de Jonge HJ, Valk PJ, Veeger NJ, Ter Elst A, den Boer ML, Cloos J et al. High VEGFC expression is associated with unique gene expression profiles and predicts adverse prognosis in pediatric and adult acute myeloid leukemia. Blood 2010; 116: 1747–1754. | Article | PubMed | ISI |
  33. Bresolin S, Zecca M, Flotho C, Trentin L, Zangrando A, Sainati L et al. Gene expression-based classification as an independent predictor of clinical outcome in juvenile myelomonocytic leukemia. J Clin Oncol 2010; 28: 1919–1927. | Article | PubMed | ISI |
  34. Lacayo NJ, Meshinchi S, Kinnunen P, Yu R, Wang Y, Stuber CM et al. Gene expression profiles at diagnosis in de novo childhood AML patients identify FLT3 mutations with good clinical outcomes. Blood 2004; 104: 2646–2654. | Article | PubMed | ISI | ChemPort |
  35. Jo A, Tsukimoto I, Ishii E, Asou N, Mitani S, Shimada A et al. Age-associated difference in gene expression of paediatric acute myelomonocytic lineage leukaemia (FAB M4 and M5 subtypes) and its correlation with prognosis. Br J Haematol 2009; 144: 917–929. | Article | PubMed | ISI |
  36. Roela RA, Carraro DM, Brentani HP, Kaiano JH, Simao DF, Guarnieiro R et al. Gene stage-specific expression in the microenvironment of pediatric myelodysplastic syndromes. Leuk Res 2007; 31: 579–589. | Article | PubMed | ISI |
  37. Mullighan CG, Kennedy A, Zhou X, Radtke I, Phillips LA, Shurtleff SA et al. Pediatric acute myeloid leukemia with NPM1 mutations is characterized by a gene expression profile with dysregulated HOX gene expression distinct from MLL-rearranged leukemias. Leukemia 2007; 21: 2000–2009. | Article | PubMed | ISI | ChemPort |
  38. Goellner S, Steinbach D, Schenk T, Gruhn B, Zintl F, Ramsay E et al. Childhood acute myelogenous leukaemia: association between PRAME, apoptosis- and MDR-related gene expression. Eur J Cancer 2006; 42: 2807–2814. | Article | PubMed | ISI |
  39. Mano H. DNA micro-array analysis of myelodysplastic syndrome. Leuk Lymphoma 2006; 47: 9–14. | Article | PubMed | ISI |
  40. Pellagatti A, Fidler C, Wainscoat JS, Boultwood J. Gene expression profiling in the myelodysplastic syndromes. Hematology 2005; 10: 281–287. | Article | PubMed | ISI |
  41. Pellagatti A, Cazzola M, Giagounidis AA, Malcovati L, Porta MG, Killick S et al. Gene expression profiles of CD34+ cells in myelodysplastic syndromes: involvement of interferon-stimulated genes and correlation to FAB subtype and karyotype. Blood 2006; 108: 337–345. | Article | PubMed | ISI | ChemPort |
  42. Qian J, Chen Z, Wang W, Cen J, Xue Y. Gene expression profiling of the bone marrow mononuclear cells from patients with myelodysplastic syndrome. Oncol Rep 2005; 14: 1189–1197. | PubMed | ISI |
  43. Mills KI, Kohlmann A, Williams PM, Wieczorek L, Liu WM, Li R et al. Microarray-based classifiers and prognosis models identify subgroups with distinct clinical outcomes and high risk of AML transformation of myelodysplastic syndrome. Blood 2009; 114: 1063–1072. | Article | PubMed | ISI | ChemPort |
  44. Boultwood J, Pellagatti A, Nikpour M, Pushkaran B, Fidler C, Cattan H et al. The role of the iron transporter ABCB7 in refractory anemia with ring sideroblasts. PLoS ONE 2008; 3: e1970. | Article | PubMed |
  45. Nikpour M, Pellagatti A, Liu A, Karimi M, Malcovati L, Gogvadze V et al. Gene expression profiling of erythroblasts from refractory anaemia with ring sideroblasts (RARS) and effects of G-CSF. Br J Haematol 2010; 149: 844–854. | Article | PubMed | ISI |
  46. Nilsson L, Eden P, Olsson E, Mansson R, Astrand-Grundstrom I, Strombeck B et al. The molecular signature of MDS stem cells supports a stem-cell origin of 5q myelodysplastic syndromes. Blood 2007; 110: 3005–3014. | Article | PubMed | ISI | ChemPort |
  47. Pellagatti A, Hellstrom-Lindberg E, Giagounidis A, Perry J, Malcovati L, Della Porta MG et al. Haploinsufficiency of RPS14 in 5q- syndrome is associated with deregulation of ribosomal- and translation-related genes. Br J Haematol 2008; 142: 57–64. | Article | PubMed | ISI | ChemPort |
  48. Boultwood J, Pellagatti A, Cattan H, Lawrie CH, Giagounidis A, Malcovati L et al. Gene expression profiling of CD34+ cells in patients with the 5q- syndrome. Br J Haematol 2007; 139: 578–589. | Article | PubMed | ISI | ChemPort |
  49. Sridhar K, Ross DT, Tibshirani R, Butte AJ, Greenberg PL. Relationship of differential gene expression profiles in CD34+ myelodysplastic syndrome marrow cells to disease subtype and progression. Blood 2009; 114: 4847–4858. | Article | PubMed | ISI |
  50. Pellagatti A, Cazzola M, Giagounidis A, Perry J, Malcovati L, Della Porta MG et al. Deregulated gene expression pathways in myelodysplastic syndrome hematopoietic stem cells. Leukemia 2010; 24: 756–764. | Article | PubMed | ISI |
  51. Gutierrez NC, Lopez-Perez R, Hernandez JM, Isidro I, Gonzalez B, Delgado M et al. Gene expression profile reveals deregulation of genes with relevant functions in the different subclasses of acute myeloid leukemia. Leukemia 2005; 19: 402–409. | Article | PubMed | ChemPort |
  52. Wang ZY, Chen Z. Acute promyelocytic leukemia: from highly fatal to highly curable. Blood 2008; 111: 2505–2515. | Article | PubMed | ISI | ChemPort |
  53. Grimwade D, Mistry AR, Solomon E, Guidez F. Acute promyelocytic leukemia: a paradigm for differentiation therapy. Cancer Treat Res 2010; 145: 219–235. | PubMed |
  54. Schlenk RF, Frohling S, Hartmann F, Fischer JT, Glasmacher A, del Valle F et al. Phase III study of all-trans retinoic acid in previously untreated patients 61 years or older with acute myeloid leukemia. Leukemia 2004; 18: 1798–1803. | Article | PubMed | ISI | ChemPort |
  55. Bug G, Ritter M, Wassmann B, Schoch C, Heinzel T, Schwarz K et al. Clinical trial of valproic acid and all-trans retinoic acid in patients with poor-risk acute myeloid leukemia. Cancer 2005; 104: 2717–2725. | Article | PubMed | ISI | ChemPort |
  56. Estey EH, Thall PF, Pierce S, Cortes J, Beran M, Kantarjian H et al. Randomized phase II study of fludarabine + cytosine arabinoside + idarubicin +/− all-trans retinoic acid +/− granulocyte colony-stimulating factor in poor prognosis newly diagnosed acute myeloid leukemia and myelodysplastic syndrome. Blood 1999; 93: 2478–2484. | PubMed | ISI | ChemPort |
  57. Meester-Smoor MA, Janssen MJ, Grosveld GC, de Klein A, van IWF, Douben H et al. MN1 affects expression of genes involved in hematopoiesis and can enhance as well as inhibit RAR/RXR-induced gene expression. Carcinogenesis 2008; 29: 2025–2034. | Article | PubMed | ISI | ChemPort |
  58. Glasow A, Barrett A, Petrie K, Gupta R, Boix-Chornet M, Zhou DC et al. DNA methylation-independent loss of RARA gene expression in acute myeloid leukemia. Blood 2008; 111: 2374–2377. | Article | PubMed | ISI |
  59. Walsby EJ, Gilkes AF, Tonks A, Darley RL, Mills KI. FUS expression alters the differentiation response to all-trans retinoic acid in NB4 and NB4R2 cells. Br J Haematol 2007; 139: 94–97. | Article | PubMed | ISI |
  60. Tagliafico E, Tenedini E, Manfredini R, Grande A, Ferrari F, Roncaglia E et al. Identification of a molecular signature predictive of sensitivity to differentiation induction in acute myeloid leukemia. Leukemia 2006; 20: 1751–1758. | Article | PubMed | ISI | ChemPort |
  61. Schlenk RF, Dohner K, Kneba M, Gotze K, Hartmann F, Del Valle F et al. Gene mutations and response to treatment with all-trans retinoic acid in elderly patients with acute myeloid leukemia. Results from the AMLSG Trial AML HD98B. Haematologica 2009; 94: 54–60. | Article | PubMed | ISI | ChemPort |
  62. List A, Dewald G, Bennett J, Giagounidis A, Raza A, Feldman E et al. Lenalidomide in the myelodysplastic syndrome with chromosome 5q deletion. N Engl J Med 2006; 355: 1456–1465. | Article | PubMed | ISI | ChemPort |
  63. Raza A, Reeves JA, Feldman EJ, Dewald GW, Bennett JM, Deeg HJ et al. Phase 2 study of lenalidomide in transfusion-dependent, low-risk, and intermediate-1 risk myelodysplastic syndromes with karyotypes other than deletion 5q. Blood 2008; 111: 86–93. | Article | PubMed | ISI | ChemPort |
  64. Ebert BL, Galili N, Tamayo P, Bosco J, Mak R, Pretz J et al. An erythroid differentiation signature predicts response to lenalidomide in myelodysplastic syndrome. PLoS Med 2008; 5: e35. | Article | PubMed | ChemPort |
  65. Estey E, Dohner H. Acute myeloid leukaemia. Lancet 2006; 368: 1894–1907. | Article | PubMed | ISI |
  66. Gilliland DG, Jordan CT, Felix CA. The molecular basis of leukemia. Hematol Am Soc Hematol Educ Program 2004; 80–97.
  67. Chang JT, Carvalho C, Mori S, Bild AH, Gatza ML, Wang Q et al. A genomic strategy to elucidate modules of oncogenic pathway signaling networks. Mol Cell 2009; 34: 104–114. | Article | PubMed | ISI | ChemPort |
  68. Corsello SM, Roti G, Ross KN, Chow KT, Galinsky I, DeAngelo DJ et al. Identification of AML1-ETO modulators by chemical genomics. Blood 2009; 113: 6193–6205. | Article | PubMed | ISI |
  69. Hassane DC, Guzman ML, Corbett C, Li X, Abboud R, Young F et al. Discovery of agents that eradicate leukemia stem cells using an in silico screen of public gene expression data. Blood 2008; 111: 5654–5662. | Article | PubMed | ISI | ChemPort |
  70. Marstrand TT, Borup R, Willer A, Borregaard N, Sandelin A, Porse BT et al. A conceptual framework for the identification of candidate drugs and drug targets in acute promyelocytic leukemia. Leukemia 2010; 24: 1265–1275. | Article | PubMed | ISI |
  71. Lamb J, Crawford ED, Peck D, Modell JW, Blat IC, Wrobel MJ et al. The Connectivity Map: using gene-expression signatures to connect small molecules, genes, and disease. Science 2006; 313: 1929–1935. | Article | PubMed | ISI | ChemPort |
  72. Chen J, Odenike O, Rowley JD. Leukaemogenesis: more than mutant genes. Nat Rev Cancer 2010; 10: 23–36. | Article | PubMed | ISI | ChemPort |
  73. Jongen-Lavrencic M, Sun SM, Dijkstra MK, Valk PJ, Lowenberg B. MicroRNA expression profiling in relation to the genetic heterogeneity of acute myeloid leukemia. Blood 2008; 111: 5078–5085. | Article | PubMed | ISI | ChemPort |
  74. Li Z, Lu J, Sun M, Mi S, Zhang H, Luo RT et al. Distinct microRNA expression profiles in acute myeloid leukemia with common translocations. Proc Natl Acad Sci USA 2008; 105: 15535–15540. | Article | PubMed | ChemPort |
  75. Langer C, Marcucci G, Holland KB, Radmacher MD, Maharry K, Paschka P et al. Prognostic importance of MN1 transcript levels, and biologic insights from MN1-associated gene and microRNA expression signatures in cytogenetically normal acute myeloid leukemia: a cancer and leukemia group B study. J Clin Oncol 2009; 27: 3198–3204. | Article | PubMed | ISI | ChemPort |
  76. Marcucci G, Radmacher MD, Maharry K, Mrozek K, Ruppert AS, Paschka P et al. MicroRNA expression in cytogenetically normal acute myeloid leukemia. N Engl J Med 2008; 358: 1919–1928. | Article | PubMed | ISI | ChemPort |
  77. Guo H, Ingolia NT, Weissman JS, Bartel DP. Mammalian microRNAs predominantly act to decrease target mRNA levels. Nature 2010; 466: 835–840. | Article | PubMed | ISI | ChemPort |
  78. Maciejewski JP, Mufti GJ. Whole genome scanning as a cytogenetic tool in hematologic malignancies. Blood 2008; 112: 965–974. | Article | PubMed | ISI | ChemPort |
  79. Rucker FG, Bullinger L, Schwaenen C, Lipka DB, Wessendorf S, Frohling S et al. Disclosure of candidate genes in acute myeloid leukemia with complex karyotypes using microarray-based molecular characterization. J Clin Oncol 2006; 24: 3887–3894. | Article | PubMed | ISI | ChemPort |
  80. Bullinger L, Kronke J, Schon C, Radtke I, Urlbauer K, Botzenhardt U et al. Identification of acquired copy number alterations and uniparental disomies in cytogenetically normal acute myeloid leukemia using high-resolution single-nucleotide polymorphism analysis. Leukemia 2010; 24: 438–449. | Article | PubMed | ISI | ChemPort |
  81. Radtke I, Mullighan CG, Ishii M, Su X, Cheng J, Ma J et al. Genomic analysis reveals few genetic alterations in pediatric acute myeloid leukemia. Proc Natl Acad Sci USA 2009; 106: 12944–12949. | Article | PubMed |
  82. Walter MJ, Payton JE, Ries RE, Shannon WD, Deshmukh H, Zhao Y et al. Acquired copy number alterations in adult acute myeloid leukemia genomes. Proc Natl Acad Sci USA 2009; 106: 12950–12955. | Article | PubMed |
  83. Figueroa ME, Lugthart S, Li Y, Erpelinck-Verschueren C, Deng X, Christos PJ et al. DNA methylation signatures identify biologically distinct subtypes in acute myeloid leukemia. Cancer Cell 2010; 17: 13–27. | Article | PubMed | ISI | ChemPort |
  84. Plass C, Oakes C, Blum W, Marcucci G. Epigenetics in acute myeloid leukemia. Semin Oncol 2008; 35: 378–387. | Article | PubMed | ISI |
  85. Bullinger L, Ehrich M, Dohner K, Schlenk RF, Dohner H, Nelson MR et al. Quantitative DNA methylation predicts survival in adult acute myeloid leukemia. Blood 2010; 115: 636–642. | Article | PubMed | ISI | ChemPort |
  86. Neff T, Armstrong SA. Chromatin maps, histone modifications and leukemia. Leukemia 2009; 23: 1243–1251. | Article | PubMed | ISI | ChemPort |
  87. Rees-Unwin KS, Morgan GJ, Davies FE. Proteomics and the haematologist. Clin Lab Haematol 2004; 26: 77–86. | Article | PubMed | ISI |
  88. Williamson AJ, Smith DL, Blinco D, Unwin RD, Pearson S, Wilson C et al. Quantitative proteomics analysis demonstrates post-transcriptional regulation of embryonic stem cell differentiation to hematopoiesis. Mol Cell Proteomics 2008; 7: 459–472. | PubMed | ISI | ChemPort |
  89. Harris MN, Ozpolat B, Abdi F, Gu S, Legler A, Mawuenyega KG et al. Comparative proteomic analysis of all-trans-retinoic acid treatment reveals systematic posttranscriptional control mechanisms in acute promyelocytic leukemia. Blood 2004; 104: 1314–1323. | Article | PubMed | ISI | ChemPort |
  90. Jin L, Xiao CL, Lu CH, Xia M, Xing GW, Xiong S et al. Transcriptomic and proteomic approach to studying SNX-2112-induced K562 cells apoptosis and anti-leukemia activity in K562-NOD/SCID mice. FEBS Lett 2009; 583: 1859–1866. | Article | PubMed | ISI |
  91. Onono FO, Morgan MA, Spielmann HP, Andres DA, Subramanian T, Ganser A et al. A tagging-via-substrate approach to detect the farnesylated proteome using two-dimensional electrophoresis coupled with Western blotting. Mol Cell Proteomics 2010; 9: 742–751. | Article | PubMed | ISI |
  92. Kornblau SM, Tibes R, Qiu YH, Chen W, Kantarjian HM, Andreeff M et al. Functional proteomic profiling of AML predicts response and survival. Blood 2009; 113: 154–164. | Article | PubMed | ISI | ChemPort |
  93. Weissinger EM, Schiffer E, Hertenstein B, Ferrara JL, Holler E, Stadler M et al. Proteomic patterns predict acute graft-versus-host disease after allogeneic hematopoietic stem cell transplantation. Blood 2007; 109: 5511–5519. | Article | PubMed | ISI | ChemPort |
  94. Varambally S, Yu J, Laxman B, Rhodes DR, Mehra R, Tomlins SA et al. Integrative genomic and proteomic analysis of prostate cancer reveals signatures of metastatic progression. Cancer Cell 2005; 8: 393–406. | Article | PubMed | ISI | ChemPort |
  95. Tian Q, Stepaniants SB, Mao M, Weng L, Feetham MC, Doyle MJ et al. Integrated genomic and proteomic analyses of gene expression in mammalian cells. Mol Cell Proteomics 2004; 3: 960–969. | Article | PubMed | ISI | ChemPort |
  96. Mayr C, Bartel DP. Widespread shortening of 3′UTRs by alternative cleavage and polyadenylation activates oncogenes in cancer cells. Cell 2009; 138: 673–684. | Article | PubMed | ISI | ChemPort |
  97. Choughule A, Polampalli S, Amre P, Shinde S, Banavali S, Prabhash K et al. Identification of PML/RARalpha fusion gene transcripts that showed no t(15;17) with conventional karyotyping and fluorescent in situ hybridization. Genet Mol Res 2009; 8: 1–7. | Article | PubMed | ISI |
  98. Rucker FG, Bullinger L, Gribov A, Sill M, Schlenk RF, Lichter P et al. Molecular characterization of AML with ins(21;8)(q22;q22q22) reveals similarity to t(8;21) AML. Genes Chromosomes Cancer 2011; 50: 51–58. | Article | PubMed | ISI |
  99. Barjesteh van Waalwijk van Doorn-Khosrovani S, Erpelinck C, van Putten WL, Valk PJ, van der Poel-van de Luytgaarde S, Hack R et al. High EVI1 expression predicts poor survival in acute myeloid leukemia: a study of 319 de novo AML patients. Blood 2003; 101: 837–845. | Article | PubMed | ChemPort |
  100. Schwind S, Marcucci G, Maharry K, Radmacher MD, Mrozek K, Holland KB et al. BAALC and ERG expression levels are associated with outcome and distinct gene and microRNA expression profiles in older patients with de novo cytogenetically normal acute myeloid leukemia: a Cancer and Leukemia Group B study. Blood 2010; 116: 5660–5669. | Article | PubMed | ISI |
  101. Krivtsov AV, Sinha AU, Stubbs MC, Kung A, Armstrong S. Cell of origin influences leukemia stem cell phenotype. Blood (ASH Annu Meet Abstr) 2009; 114: 3459.
  102. Scholl C, Frohling S, Dunn IF, Schinzel AC, Barbie DA, Kim SY et al. Synthetic lethal interaction between oncogenic KRAS dependency and STK33 suppression in human cancer cells. Cell 2009; 137: 821–834. | Article | PubMed | ISI | ChemPort |
  103. Majeti R, Becker MW, Tian Q, Lee TL, Yan X, Liu R et al. Dysregulated gene expression networks in human acute myelogenous leukemia stem cells. Proc Natl Acad Sci USA 2009; 106: 3396–3401. | Article | PubMed |
  104. Lane SW, Scadden DT, Gilliland DG. The leukemic stem cell niche: current concepts and therapeutic opportunities. Blood 2009; 114: 1150–1157. | Article | PubMed | ISI | ChemPort |
Top

Acknowledgements

This work is based on the joint research activities under the framework of the European Program for Cooperation in Science and Technology (COST, Action BM0801, WG1) and the European LeukemiaNet funded by the 6th Framework Program of the European Community (WP13 Gene Profiling).