A gene expression signature of CD34(+) cells to predict major cytogenetic response in chronic-phase chronic myeloid leukemia patients treated with imatinib

Oregon Health & Science University Knight Cancer Institute, Portland, OR 97239, USA.
Blood (Impact Factor: 10.45). 10/2009; 115(2):315-25. DOI: 10.1182/blood-2009-03-210732
Source: PubMed

ABSTRACT In chronic-phase chronic myeloid leukemia (CML) patients, the lack of a major cytogenetic response (< 36% Ph(+) metaphases) to imatinib within 12 months indicates failure and mandates a change of therapy. To identify biomarkers predictive of imatinib failure, we performed gene expression array profiling of CD34(+) cells from 2 independent cohorts of imatinib-naive chronic-phase CML patients. The learning set consisted of retrospectively selected patients with a complete cytogenetic response or more than 65% Ph(+) metaphases within 12 months of imatinib therapy. Based on analysis of variance P less than .1 and fold difference 1.5 or more, we identified 885 probe sets with differential expression between responders and nonresponders, from which we extracted a 75-probe set minimal signature (classifier) that separated the 2 groups. On application to a prospectively accrued validation set, the classifier correctly predicted 88% of responders and 83% of nonresponders. Bioinformatics analysis and comparison with published studies revealed overlap of classifier genes with CML progression signatures and implicated beta-catenin in their regulation, suggesting that chronic-phase CML patients destined to fail imatinib have more advanced disease than evident by morphologic criteria. Our classifier may allow directing more aggressive therapy upfront to the patients most likely to benefit while sparing good-risk patients from unnecessary toxicity.

8 Reads
  • Source
    • "Microarray profiling has been used to determine whether the specific gene expression at diagnosis can predict the response to TKI treatment [61]. Analysis of CD34+ cells from newly diagnosed, treatment-naive patients with CML-CP has revealed a 75-transcript signature (50 upregulated and 25 downregulated transcripts) that predicted major cytogenetic response at 12 months with an overall accuracy of 87% — exceeding the predictive ability of the Sokal score [62]. Notably, 62% of the upregulated transcripts were potential targets of the WNT/β-catenin pathway, which is activated during BC. "
    [Show abstract] [Hide abstract]
    ABSTRACT: The introduction of BCR-ABL1 tyrosine kinase inhibitors (TKIs) for treatment of chronic myelogenous leukemia in chronic phase (CML-CP) has revolutionized therapy, altering the outcome from one of shortened life expectancy to long-term survival. With over 10 years of long-term treatment with imatinib and several years of experience with the next generation of TKIs, including nilotinib, dasatinib, bosutinib, and ponatinib, it is becoming clear that many clinical parameters have great impact on the prognosis of patients with CML. Emerging novel gene expression profiling and molecular techniques also provide new insights into CML pathogenesis and have identified potential prognostic markers and therapeutic targets. This review presents the supporting data and discusses how certain clinical characteristics at diagnosis, the depth of early response, the presence of certain kinase domain mutations, and additional molecular changes serve as prognostic factors that may guide individualized treatment decisions for patients with CML-CP.
    Journal of Hematology & Oncology 07/2013; 6(1):54. DOI:10.1186/1756-8722-6-54 · 4.81 Impact Factor
  • Source
    • "HSCT is associated with >80% survival when performed in CP, but outcomes are significantly worse for advanced disease, ∼40% in AP and <20% in BC (Hansen et al., 1998; Radich et al., 2003). There are no established molecular predictors of transplant outcome in CML, thus clinical measures such as the Sokal, Hasford, or European Group for Blood and Marrow Transplantation (EBMT) Risk Scores have been used for prognostication (Gratwohl et al., 1998; Hasford et al., 1996; McWeeney et al., 2010; Mohty et al., 2007; Sokal et al., 1984; Yong et al., 2006). The EBMT score has been validated extensively, and is used to predict outcomes prior to HSCT (Gratwohl et al., 1998; Passweg et al., 2004). "
    [Show abstract] [Hide abstract]
    ABSTRACT: MOTIVATION: Selecting a small number of signature genes for accurate classification of samples is essential for the development of diagnostic tests. However, many genes are highly correlated in gene expression data, and hence, many possible sets of genes are potential classifiers. Because treatment outcomes are poor in advanced chronic myeloid leukemia (CML), we hypothesized that expression of classifiers of advanced phase CML when detected in early CML [chronic phase (CP) CML], correlates with subsequent poorer therapeutic outcome. RESULTS: We developed a method that integrates gene expression data with expert knowledge and predicted functional relationships using iterative Bayesian model averaging. Applying our integrated method to CML, we identified small sets of signature genes that are highly predictive of disease phases and that are more robust and stable than using expression data alone. The accuracy of our algorithm was evaluated using cross-validation on the gene expression data. We then tested the hypothesis that gene sets associated with advanced phase CML would predict relapse after allogeneic transplantation in 176 independent CP CML cases. Our gene signatures of advanced phase CML are predictive of relapse even after adjustment for known risk factors associated with transplant outcomes.
    Bioinformatics 01/2012; 28(6):823-30. DOI:10.1093/bioinformatics/bts059 · 4.98 Impact Factor
  • Source
    • "This technology has made it possible to perform genome-wide searches for changes in gene expression in response to perturbations such as gene knockouts [1] and treatment with bioactive compounds [2]. It has also been useful in identifying gene expression differences associated with histologic subtypes of disease [3], clinical diagnosis [4], prognosis [5], or the efficacy of various therapeutic strategies [6]. However, a challenge for scientists performing genome-wide gene expression microarray analysis has been using these data to generate hypotheses about biological processes responsible for the patterns of differential gene expression associated with a particular trait or experimental variable. "
    [Show abstract] [Hide abstract]
    ABSTRACT: Background Identifying similarities between patterns of differential gene expression provides an opportunity to identify similarities between the experimental and biological conditions that give rise to these gene expression alterations. The growing volume of gene expression data in open data repositories such as the NCBI Gene Expression Omnibus (GEO) presents an opportunity to identify these gene expression similarities on a large scale across a diverse collection of datasets. We have developed a fast, pattern-based computational approach, named openSESAME (Search of Expression Signatures Across Many Experiments), that identifies datasets enriched in samples that display coordinate differential expression of a query signature. Importantly, openSESAME performs this search without prior knowledge of the phenotypic or experimental groups in the datasets being searched. This allows openSESAME to identify perturbations of gene expression that are due to phenotypic attributes that may not have been described in the sample annotation included in the repository. To demonstrate the utility of openSESAME, we used gene expression signatures of two biological perturbations to query a set of 75,164 human expression profiles that were generated using Affymetrix microarrays and deposited in GEO. The first query, using a signature of estradiol treatment, identified experiments in which estrogen signaling was perturbed and also identified differences in estrogen signaling between estrogen receptor-positive and -negative breast cancers. The second query, which used a signature of silencing of the transcription factor p63 (a key regulator of epidermal differentiation), identified datasets related to stratified squamous epithelia or epidermal diseases such as melanoma. Conclusions openSESAME is a tool for leveraging the growing body of publicly available microarray data to discover relationships between different biological states based on common patterns of differential gene expression. These relationships may serve to generate hypotheses about the causes and consequences of specific patterns of observed differential gene expression. To encourage others to explore the utility of this approach, we have made a website for performing openSESAME queries freely available at
    BMC Bioinformatics 09/2011; 12:381. DOI:10.1186/1471-2105-12-381 · 2.58 Impact Factor
Show more