J Appl Genetics (2012) 53:83–91 DOI 10.1007/s13353-011-0073-x
HUMAN GENETICS • ORIGINAL PAPER
Gene expression signatures and ex vivo drug sensitivity profiles in children with acute lymphoblastic leukemia Joanna Szczepanek & Michal Jarzab & Malgorzata Oczko-Wojciechowska & Malgorzata Kowalska & Andrzej Tretyn & Olga Haus & Monika Pogorzala & Mariusz Wysocki & Barbara Jarzab & Jan Styczynski
Received: 5 September 2011 / Revised: 1 October 2011 / Accepted: 3 October 2011 / Published online: 27 October 2011 # Institute of Plant Genetics, Polish Academy of Sciences, Poznan 2011
Abstract Introduction Causes of treatment failure in acute lymphoblastic leukemia (ALL) are still poorly understood. Microarray technology gives new possibilities for the analysis of the biology of leukemias. We hypothesize that drug sensitivity in pediatric ALL is driven by specific molecular
J. Szczepanek : M. Pogorzala : M. Wysocki : J. Styczynski Department of Pediatric Hematology and Oncology, Collegium Medicum, Nicolaus Copernicus University, Bydgoszcz, Poland J. Szczepanek : A. Tretyn Department of Plant Physiology and Biotechnology, Nicolaus Copernicus University, Torun, Poland M. Jarzab Department of Clinical and Experimental Oncology, Maria Sklodowska-Curie Memorial Cancer Centre and Institute of Oncology, Gliwice, Poland M. Oczko-Wojciechowska : M. Kowalska : B. Jarzab Department of Nuclear Medicine and Endocrine Oncology, Maria Sklodowska-Curie Memorial Cancer Centre and Institute of Oncology, Gliwice, Poland O. Haus Department of Clinical Genetics, Collegium Medicum, Nicolaus Copernicus University, Bydgoszcz, Poland J. Szczepanek (*) Department of Plant Physiology and Biotechnology, Nicolaus Copernicus University, ul. Gagarina 9, 87-100 Torun, Poland e-mail:
[email protected]
mechanisms that correlate with gene expression profiles assessed by microarray analysis. Objective The aim of the study was to determine the ex vivo resistance profiles of 20 antileukemic drugs and gene expression profiles, with relation to response to initial therapy. Patients and methods Lymphoblasts were analyzed after bone marrow biopsy was obtained from 56 patients. The profile of in vitro resistance to drugs was determined in the 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyl-tetrazoliumbromide (MTT) cytotoxicity assay. High-quality total RNA was prepared and hybridized to oligonucleotide arrays HGU133A 2.0 Chip (Affymetrix). The expression of selected genes was tested by qualitative reverse transcription polymerase chain reaction (qRT-PCR). Results and conclusions The exposure of leukemic blasts to drugs initiates a complex cellular response, which reflects global changes in gene expression. Changes in the expression of several genes are highly correlated with drug resistance. Keywords Acute leukemia . Children . Drug resistance . Gene expression profiles . Microarrays
Introduction Acute lymphoblastic leukemia (ALL) is the most commonly diagnosed cancer in children. Currently used therapeutic strategies are likely to achieve remission in about 80% of pediatric patients (Möricke et al. 2008; Jeha and Pui 2009; Pui 2009; Pui et al. 2009). The main causes of treatment failure are primary and acquired drug resistance, although the mechanisms for their formation are of complex biology. The treatment of pediatric ALL is mainly based on multidrug chemotherapy, conducted according to national and international therapeutic protocols (Pui 2009; Möricke
84
et al. 2010). The intensity of therapy is adjusted according to risk group, referred to as the degree of risk of relapse. Recent advances in genome-wide screening techniques and pharmacogenomic studies enable the development of future personalized therapy (Jeha and Pui 2009). One of the main causes of the inefficacy of antileukemic therapy is the occurrence of resistance to anticancer drugs. This resistance applies to both individual drugs and drug groups, differing in origin, chemical structure, and mechanism of impact on leukemic blasts. Cancer cells are characterized by a specific metabolism that allows survival in the presence of cytostatic drug. The multidrug resistance phenomenon affects the genetic properties of cancer cells, primary or acquired during the development of cancer. Sensitivity to anticancer drugs is also linked to the expression of genes involved in cell cycle regulation, DNA repair, and drug metabolism (Hoelzer et al. 2002; Holleman et al. 2004, 2006; Bhojwani et al. 2008). Among those most frequently mentioned are: expression of oncogenes (such as BCL-2, MYC, and RAS), mutations of tumor suppressor genes (e.g., the TP53 gene), the presence of specific cytogenetic changes leading to the formation of fusion and gene rearrangements (e.g., BCR-ABL, MLL, TEL-AML1, PML-RARα, AML1-ETO, CBF-MYH11β), polyclonality, and heterogeneity of tumor cells. However, the knowledge about most of these genes in the evolution of resistance to various drugs is still poorly understood. For several years, researchers have been focusing their attention on the simultaneous examination of thousands of genes of a tumor cell. Such analysis has, as its objective, the selection of genes associated with the development and course of cancer, as well as associated with the sensitivity or resistance of cells to drugs (Ferrando et al. 2002; Yeoh et al. 2002). Recent studies indicate that the associated profile of drug resistance is one of the strongest prognostic factors in acute leukemias in children (den Boer et al. 2003; Styczyński and Wysocki 2004; Styczynski et al. 2008). It is expected that the compilation of results of the analysis of gene expression, drug resistance, and response to therapy will lead to the development of individualized targeted therapy. The microarray technique is currently widely used in studies of molecular methods for the analysis of gene expression at the scale of the entire genome and transcriptome. This technique is used primarily in determining the gene expression profiles, which analyze a pool of RNA transcripts (Blohm and Guiseppi-Elie 2001). The expression profile defines a set of complex data obtained by this method. The expression pattern, developed for a limited set of genes, is called a molecular signature (Jarzab et al. 2005). The aim of this study was to correlate changes in the gene expression profile and in vitro drug resistance profiles in children with ALL.
J Appl Genetics (2012) 53:83–91
Methods Patients Clinical samples were obtained from children with ALL before treatment was initiated. Analyses were conducted on cells derived from the bone marrow of patients diagnosed with ALL de novo or relapse. Only fresh leukemic samples were taken for the study and the percentage of blasts in the sample was over 90%. Mononuclear cells were isolated from bone marrow aspirate using Ficoll-Paque (Amersham Biosciences, Piscataway, NJ). All patients were divided into two groups. For the first one including 56 patients (48 ALL de novo, 8 recurrent ALL, median age 8 years), we performed the analysis of both microarray and quantitative polymerase chain reaction (PCR) methods (Table 1). Patients in the second group were regarded as an independent validation group. For these 51 patients (44 ALL de novo, 7 recurrent ALL, median age 7.5 years), we determined the expression of selected genes in the qualitative reverse transcription polymerase chain reaction (qRT-PCR) reactions. The study was approved by the local Bioethics Committee. MTT assay The in vitro drug resistance profile of leukemic cells to 20 drugs was determined with the use of the four-day in vitro 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyl-tetrazoliumbromide (MTT) drug resistance assay. The concentrations of the tested drugs are given in Table 2. The assay conditions were similar as that described previously (Styczynski et al. 2000, 2002). Cytotoxicity was determined by measuring the ability of the cells to cleave the soluble compound MTT (Serva, Heidelberg, Germany) into an insoluble salt. This reaction is characteristic only for living cells. The LC50 values, the concentration of drugs that was lethal to 50% of the cells, was used as a measure for the in vitro drug cytotoxicity in each sample. The LC50 values used to define cells as being sensitive or resistant to each agent were those previously associated with a good or bad treatment outcome in patients with ALL as described by den Boer et al. (2003). Based on the MTT cytotoxicity assay, patients were classified as resistant, sensitive, or intermediate resistant. Samples with apoptosis during the MTT assay were excluded from the analysis. Preparation of RNA samples Mononuclear (leukemic) cells were isolated from each bone marrow aspirate by centrifugation over Ficoll. Samples (6–10 × 106 isolated blasts) were homogenized in RLT buffer (Qiagen, Hilden, Germany). Cell lysates
J Appl Genetics (2012) 53:83–91
85
Table 1 Characteristics of the analyzed groups of patients
Number of patients
Age (median, years)
Sex
Type of leukemia
Microarray group
Validation group
N ALL ALL relapse N ALL ALL relapse N ALL ALL relapse
56 48 8 8 7.5 11.5 29 F:27 M 22 F:21 M 2 F:6 M
51 44 7 7 7 9 22 F:29 M 19 F:25 M 3 F:4 M
ALL – acute lymphoblastic leukemia, M–male, F–female
were stored at −80°C. Total RNA was extracted using TRIZOL® Reagent (Invitrogen Life Technologies, Paisley, UK) and then digested with DNase I, RNase-free (Fermentas, Hanover, MD). The preparations were purified further by extraction with phenol and chloroform. RNA purity was assessed by spectrophotometry (NanoDrop® ND-1000, NanoDrop Technologies, Wilmington, DE). Total and amplified RNA integrity was assessed by capillary electrophoresis using the Agilent 2100 Bioanalyzer (Agilent, Palo Alto, CA).
Microarray analysis For each test, 5 μg of total RNA was used as a template. Total RNA through a series of reactions (including purification, reverse transcription, labeling, in vitro transcription, fragmentation) was prepared adequately for hybridization. The procedure for the preparation of nucleic acid was carried out in accordance with the recommendation of the manufacturer (Affymetrix Incorporated, Santa Clara, CA), based on a set of reagents for hybridization reactions for the selected type of microarray. Fragmented fluorescently labeled cRNA probes were synthesized and hybridized to Affymetrix Human Genome U133A 2.0 oligonucleotide arrays. Arrays were grouped according to the specific profile of the MTT assay of resistance to the drug. For each gene, a median of expression was identified. Groups were analyzed according to their drug resistance (sensitive, moderately sensitive, moderately resistant, resistant). The ratio between resistant and sensitive patients was taken as a measure of changes in the level of gene expression. Preliminary analysis of microarrays included: (1) evaluating the quality of the hybridization to the array; (2) unsupervised analysis and presentation of the main sources of variations; and (3) supervised analysis, having as its objective the selection of genes associated with specific features in the test set.
Table 2 Drugs and concentration cut-off values for drug resistance profile Drug
Producer
Sensitive profile
Intermediate profile
Resistant profile
Percentile Prednisolone – PRE Vincristine – VCR L-asparaginase – ASP Idarubicin – IDA Daunorubicin – DNR Doxorubicin – DOX Mitoxantrone – MIT
Jelfa, Jelenia Góra Eli-Lilly, Indianapolis Medac, Hamburg Farmitalia, Milan Rhône-Poulenc Rhorer, Paris Farmitalia, Milan Jelfa, Jelenia Góra
0–33 ≤1.5 μg/ml ≤0.2 μg/ml ≤0.05 IU/ml ≤0.04 μg/ml ≤0.08 μg/ml ≤0.43 μg/ml ≤0.02 μg/ml
33–67 1.5–45 μg/ml 0.2–0.95 μg/ml 0.05–0.3 IU/ml 0.04–0.09 μg/ml 0.08–0.23 μg/ml 0.43–0.95 μg/ml 0.02–0.05 μg/ml
67–100 ≥45 μg/ml ≥0.95 μg/ml ≥0.3 IU/ml ≥0.09 μg/ml ≥0.23 μg/ml ≥0.95 μg/ml ≥0.05 μg/ml
Etoposide – VP16 Melphalan – MELF Cytarabine – ARAC Fludarabine – FLU Cladribine – CDA Thioguanine – TG Thiotepa – THIO Treosulfan – TRE Cyclophosphamide - CYKL Bortezomib - BOR Topotecan – TOPO Clofarabine – CLO Busulfan – BUS
Bristol-Myers Squibb, Princeton Alkeran, Glaxo Wellcome, Parma Cytosar, Pharmacia & Upjohn Fludara, Schering AG, Berlin Biodribin, Bioton, Warsaw Sigma, no A4882 Lederle, Riemser, Greifswald Ovastat, Medac, Hamburg 4-HOO-Cyclophosphamide, Asta Medica Velcade, Janssen Pharmaceutica N.V. Glaxo Smith Kline-Beecham Evoltra, Bioenvision, Edinburgh Busilvex, Pierre Fabre Médicament
≤0.04 μg/ml ≤0.5 μg/ml ≤0.3 μg/ml ≤0.17 μg/μl ≤0.16 μg/μl ≤2.0 μg/μl ≤0.45 μg/μl ≤0.04 μg/μl ≤0.6 μg/μl ≤290 nM ≤0.1 μg/ml ≤0.02 nM ≤12 mg/ml
0.04–1.7 μg/ml 0.5–1.3 μg/ml 0.3–1.2 μg/ml 0.17–0.8 μg/μl 0.16–1.5 μg/μl 2.0–5.0 μg/μl 0.45–1.50 μg/μl 0.04–0.5 μg/μl 0.6–1.4 μg/μl 290–800 nM 0.1–0.5 μg/ml 0.02–0.13 nM 12–35 mg/ml
≥1.7 μg/ml ≥1.3 μg/ml ≥1.2 μg/ml ≥0.8 μg/μl ≥1.5 μg/μl ≥5.0 μg/μl ≥1.50 μg/μl ≥0.5 μg/μl ≥1.4 μg/μl ≥800 nM ≥0.5 μg/ml ≥0.13 nM ≥35 mg/ml
86
In the initial phase, all analyses were performed on the entire group. They were performed after filtration of the output of more than 50,000 probe sets, probes for which less than 10% of the samples showed a change in the ratio of median >1.5×. The use of filters reduced the analysis to a set of 13,835 probes. The normalization of data was conducted based on the GCRMA method. qRT-PCR To validate the array results, we performed quantitative RTPCR on selected genes using samples from 107 patients. We used random hexamer priming and High Fidelity Reverse Transcriptase (Roche Diagnostics, Basel, Switzerland) to generate cDNA. Real-time PCRs were performed using FastStart SYBR Green Master (Roche Diagnostics, Basel, Switzerland) and the Mastercycler ep realplex (Eppendorf, UK), as described in the manufacturer’s instructions. The expression of the housekeeping gene, glyceraldehyde-3phosphate dehydrogenase (GAPDH), was used as an endogenous control for normalization. Melting curve analyses were performed to verify the amplification specificity. Quantitative PCR reactions were carried out on a dilution series of a cDNA to obtain standard curves for the control and target genes. Based on these curves, the reaction efficiencies were calculated. The relative level of expression was determined on the basis of the method of Pfaffl (2001).
J Appl Genetics (2012) 53:83–91
Results In vitro drug sensitivity correlation matrix Based on the analysis of LC50 values, correlation sensitivity matrices were prepared. Correlation between the basic profiles of sensitivity to prednisolone, vincristine, and Lasparaginase, and profiles for daunorubicin and etoposide were observed. Similar profiles were also found for idarubicin, mitoxantrone, doxorubicin, melphalan, and thioguanine, and cytarabine, fludarabine, cladribine, and clofarabine. Treosulfan, topotecan, and busulfan showed a weak correlation (Fig. 1). Microarray analysis For the 56 samples, a correlation analysis of drug sensitivity with the profile of gene expression was performed. For each drug, genes that correlated with LC50 values at p < 0.001 were selected for further analysis. The number of these genes with p
