Cancer Therapy: Clinical

Clinical Relevance of a Pharmacogenetic Approach Using Multiple Candidate Genes to Predict Response and Resistance to Imatinib Therapy in Chronic Myeloid Leukemia Dong Hwan (Dennis) Kim,1, 5 Lakshmi Sriharsha,1 Wei Xu,2 Suzanne Kamel-Reid,3 Xiangdong Liu,4 Katherine Siminovitch,4 Hans A. Messner,1and Jeffrey H. Lipton1

Abstract Purpose: Imatinib resistance is major cause of imatinib mesylate (IM) treatment failure in chronic myeloid leukemia (CML) patients. Several cellular and genetic mechanisms of imatinib resistance have been proposed, including amplification and overexpression of the BCR/ABL gene, the tyrosine kinase domain point mutations, and MDR1 gene overexpression. Experimental Design: We investigated the impact of 16 single nucleotide polymorphisms (SNP) in five genes potentially associated with pharmacogenetics of IM, namely ABCB1, multidrug resistance 1; ABCG2, breast-cancer resistance protein; CYP3A5,cytochrome P450-3A5; SLC22A1, human organic cation transporter 1; and AGP, a1-acid glycoprotein. The DNAs from peripheral blood samples in 229 patients were genotyped. Results: The GG genotype in ABCG2 (rs2231137), AA genotype in CYP3A5 (rs776746), and advanced stage were significantly associated with poor response to IM especially for major or complete cytogenetic response, whereas the GG genotype at SLC22A1 (rs683369) and advanced stage correlated with high rate of loss of response or treatment failure to IM therapy. Conclusions: We showed that the treatment outcomes of imatinib therapy could be predicted using a novel, multiple candidate gene approach based on the pharmacogenetics of IM.

Imatinib mesylate (IM) is a selective tyrosine kinase inhibitor, amplification andoverexpression of the BCR/ABL gene, point especially against BCR/ABL fusion tyrosine kinase, that has mutations in the ATP-binding site with kinase reactivation (3), achievedsuccessful treatment outcomes andimprovedthe life or overexpression of the MDR1 gene (4). quality of chronic myeloidleukemia (CML) patients (1). The It has been well establishedthat the response to imatinib activity of IM is mediated by blocking the activity of BCR/ABL therapy is strongly associatedwith the diseasefactor such as tyrosine kinase in CML cells. However, some of the patients disease stage, that is, chronic phase (CP), accelerated phase fail to achieve optimal response, anda substantial proportion (AP), or blastic crisis (BC). Besides the disease factor, several of patients develop resistance to IM (2). Several molecular determinants were known to be associated with the pharmaco- mechanisms leading to IM resistance have been proposed: kinetics of imatinib with respect to absorption, distribution, and metabolism, influencing the systemic levels or intracellular concentration of imatinib which affects the response of imatinib therapy (Supplemental Table S1; ref. 5). IM is a substrate for the Authors’ Affiliations: 1Chronic Myelogenous Leukemia Group, Department of adenosine triphosphate binding cassette (ABC) transporters, 2 Hematology/Medical Oncology, Department of Biostatistics, Princess Margaret ABCB1 andABCG2, whereas the active uptake of IM into cells is Hospital, and 3Department of Pathology, University Health Network, University of Toronto, and 4Analytical GeneticsTechnology Centre, University Health Network, mediated by the human organic cationic transporter-1 (OCT1; Toronto, Ontario, Canada; and 5Department of Hematology/Oncology, Samsung SLC22A1). Also, IM is metabolizedthrough first-pass drug Medical Center, Sungkyunkwan University School of Medicine, Seoul, Korea metabolism by the cytochrome P450 - CYP3A4 andCYP3A5. In Received1/22/09; revised 3/20/09; accepted3/23/09; publishedOnlineFirst 7/7/09. addition, it is delivered in a bound form with a plasma protein Grant support: Korea Healthcare Technology R&D Project, Ministry of Health referredto as a1-acidglycoprotein (AGP). Welfare & Family Affairs, Republic of Korea (A070001) and by a grant from the Friends to life fund, Princess Margaret Hospital foundation,Toronto, Ontario, Canada. Accordingly, the intracellular or systemic level of imatinib The costs of publication of this article were defrayed in part by the payment of page shouldbe influencedby these factors, such as the ABCB1, charges. This article must therefore be hereby marked advertisement in accordance ABCG2, SLC22A1, CYP3A4, CYP3A5,orAGP genes (6). The with 18 U.S.C. Section 1734 solely to indicate this fact. interindividual variability of five candidate genes associated Note: Supplementary data for this article are available at Clinical Cancer Research ABCB1 Online (http://clincancerres.aacrjournals.org/). with drug transport/metabolism, namely, (4, 7–9), This work was presented at the annual meeting of American Society of Hematology, ABCG2 (10–12), SLC22A1 (13–16), CYP3A4/3A5 (6, 17, 18), Atlanta, GA, USA, 2007. and AGP (19–22), couldaffect the expression of corresponding Requests for reprints: Dong Hwan (Dennis) Kim, Department of Hematology/ proteins, thus influencing the treatment outcomes of imatinib Oncology, Samsung Medical Center, Sungkyunkwan University, Seoul, Korea, therapy. We hypothesizedthat the SNPs of these five genes 135-710. Phone: 82-2-3410-3459; Fax: 82-2-3410-0041; E-mail: drkiim @ medimail.co.kr. couldpredict the outcomes of imatinib therapy in CML F 2009 American Association for Cancer Research. patients. We evaluatedits efficacy by three categories of doi:10.1158/1078-0432.CCR-09-0145 parameters: (a) response to imatinib therapy [hematologic

Clin Cancer Res 2009;15(14) July 15, 2009 4750 www.aacrjournals.org Downloaded from clincancerres.aacrjournals.org on September 29, 2021. © 2009 American Association for Cancer Research. Pharmacogenetics of Imatinib Therapy for CML

ABL fusion gene transcripts was repeatedevery 3 mo regardlessof Translational Relevance cytogenetic response using quantitative BCR/ABL mRNA PCR. Sequenom MassARRAY genotyping system. In the initial step, the The current study attempted to investigate the predictive candidate genotypes were selected by the literature review. If there was role of multiple candidate gene single nucleotide poly- no available data in the literature review, we referred to the SNPs from morphisms (SNP) associated with the metabolism/trans- the Entrez SNP site (http://www.ncbi.nlm.nih.gov/sites/entrez). The port of imatinib for the treatment of chronic myeloid selection criteria for the SNP were the nonsynonymous SNPs in exon region with minor allele frequency of >0.05 to 0.1. If we couldnot find leukemia. Several biomarkers were suggested to predict any SNPs satisfying the criteria, we included synonymous SNPs in the response or resistance to imatinib therapy, such as the exon region or SNPs with minor allele frequency of <0.05. IC 50 cell assay, OCT1activity assay, or blood trough level Genotyping was undertaken using the Sequenom iPLEX platform, measurement. We hypothesized that pharmacogenetic according to the manufacturer’s instructions (www.sequenom.com; information can be utilized to predict the response to ima- Sequenom Inc.). Whole bloodsamples were obtainedaccordingto the tinib therapy. Out of 16 SNPs in 5 genes (ABCB1,multidrug declaration of Helsinki. DNA was extracted using the Puregene DNA resistance 1; ABCG2, breast-cancer resistance protein; purification Kit (Gentra Systems Inc.). The detection of SNPs was done CYP3A5, cytochrome P450-3A5; SLC22A1, human or- by the analysis of primer extension products generated from previously ganic cation transporter 1; AGP, a1-acid glycoprotein), amplifiedgenomic DNA using a Sequenom chip-basedmatrix-assisted SNPs in the ABCG2 and CYP3A5 genes correlated with laser desorption/ionization time-of-flight (MALDI-TOF) mass spec- cytogenetic response to imatinib therapy and SLC22A1 trometry platform. Multiplex SNP assays were designed using Spec- troDesigner software (Sequenom). Ninety-six-well plates containing 2.5 SNP correlated with loss of response or treatment failure ngDNAineachwellwereamplifiedbyPCRfollowingthe to imatinib therapy. Based on the results from single marker specifications of Sequenom. Unincorporatednucleotidesin the PCR analyses, we established the predictive model for the product were deactivated using shrimp alkaline phosphatase. Allele response or treatment failure to imatinib therapy.The pres- discrimination reactions were conducted by adding the extension ent results can be used to identify high-risk group of primer(s), DNA polymerase, and a cocktail mixture of deoxynucleotide patients with CML who potentially need dose escalation triphosphates and di-deoxynucleotide triphosphates to each well. of imatinib or to switch to second-generation tyrosine MassExtend clean resin (Sequenom) was added to the mixture to kinase inhibitors. remove extraneous salts that couldinterfere with MALDI-TOF analysis. The primer extension products were then cleaned and spotted onto a SpectroChip. Genotypes were determined by spotting an aliquot of each sample onto a 384 SpectroChip (Sequenom), which was sub- response, major/complete cytogenetic response (MCyR/CCyR), sequently readby the MALDI-TOF mass spectrometer. Duplicate major/complete molecular response (MMoR/CMoR)]; (b) treat- samples andnegative controls were includedto check genotyping ment failure [loss of response (LOR), primary resistance, prog- quality. Genotyping results were confirmedusing directsequencing in n ression to AP or BC]; and(c ) dose escalation for the purpose selectedsamples ( = 20). The sequences of primers are listedin Supplemental Table S2. of overcoming treatment failure including LOR. We analyzed BCR/ABL real-time quantitative PCR and tyrosine kinase domain 16 candidate gene SNPs in 5 genes associated with imatinib mutation test. Peripheral bloodsamples (5 mL) were also analyzed ABCB1, ABCG2, SLC22A1, transport andmetabolism, including using quantitative PCR to determine the levels of BCR/ABL fusion gene CYP3A5, and AGP, in 229 CML patients receiving IM. transcripts according to the manufacturer’s instructions (ABI 9700 Thermal Cycler; AppliedBiosystems) andfollowing the recommenda- tions established to standardize this procedure at the international level Patients, Materials, and Methods (23–25). GAPDH was usedas the internal control gene in the BCR-ABL quantitation. The log reduction was calculated by using our molecular Study population. A total of 229 Ph+ CML patients who receivedIM lab diagnostic baseline, which was established from f 50 untreated at the Princess Margaret Hospital (Toronto, Ontario, Canada) from CML patients in chronic phase before commencement of IM. Nested August 2000 to December 2006 were retrospectively enrolledin this PCR techniques were usedto confirm the results in selectedsamples study. The current study was approved by the Research Ethics Board of that hadundetectable BCR/ABL transcript levels. Kinase domain the University Health Network. Patient characteristics are summarized mutations were screened in any patient in advanced-phase disease. in Table 1. A total of 212 patients were startedon IM at a doseof 400 For chronic-phase patients who start treatment with IM, mutation mg/dand17 patients were startedat dosesof 600 mg/d( n = 15) screening is indicated if there is inadequate initial response or any sign or 800 mg/d(n = 2). All CP patients were startedon 400 mg/d( n = of loss of response. All bloodsamples were collectedafter informed 203), whereas patients in AP or BC startedat dosesof consent hadbeen obtainedfrom patients in accordance with the 400 mg/d(n = 9), 600 mg/d(n = 15), or 800 mg/d(n = 2). Declaration of Helsinki. Before commencing IM therapy, in addition to routine history taking Definition of response criteria and end points. Response criteria were andphysical examination, all patients hada complete bloodcell count the same as previously defined in studies using IM (2, 26, 27). Briefly, a including WBC differential as well as standard biochemistry. Baseline hematologic response was defined as normalized peripheral blood cell tests also included bone marrow evaluation for morphology, conven- counts (WBC <10 Â 109/L andplatelet count <450 Â 109/L) without tional cytogenetic analysis, and BCR/ABL mRNA reverse transcription- evidence of peripheral blasts, promyelocytes, or myelocytes, and PCR. Cytogenetic analysis was done by the G-banding technique. without evidence of extramedullary disease including disappearance Patients were regularly monitoredon an outpatient basis: biweekly of palpable splenomegaly lasting for at least 4 wk. Cytogenetic physical examinations, bloodcounts, andbiochemistry were obtained responses were categorizedas complete (CCyR; 0% Ph + cells in marrow during the first month of IM therapy and then monthly until a by conventional cytogenetics or fluorescence in situ hybridization), cytogenetic response was achieved, and then every 3 mo thereafter. partial (1-34% Ph+ cells in marrow), or minor (35-90% Ph+ cells in Until a complete cytogenetic response was confirmed, bone marrow marrow). A major cytogenetic response (MCyR) was defined as the sum evaluation andfluorescence in situ hybridization studies were of CCyR andpartial cytogenetic response (0-35% Ph + cells in marrow). performedevery 3 mo. The quantification of peripheral blood BCR/ Major molecular response (MMoR) was defined as =3 log reduction

www.aacrjournals.org 4751 Clin Cancer Res 2009;15(14) July 15, 2009 Downloaded from clincancerres.aacrjournals.org on September 29, 2021. © 2009 American Association for Cancer Research. Cancer Therapy: Clinical

Table 1. Patient and disease characteristics (N = 229)

Initial diagnosis of Ph+ CML No. of patients (%) Gender Female/male (ratio) 95/134 (42:58) Age, yMedian (range) 49.5 (14-74) Race White/non-white (ratio) 170/59 (74:26) Disease stage Chronic phase 203 (89) Accelerated phase 24 (10) Blastic crisis 2 (1) Treatment prior to IM INF 98 (43) Busulfan 17 (7) Cytarabine 19 (8) BMT 12 (5) At the time of imatinib therapy Age, yMedian (range) 52.5 (20-75) Disease duration from diagnosis, mo Median (range) 4.3 (0-231) Disease stage Chronic phase 203 (89) Accelerated phase 23 (10) Blastic crisis 3 (1) Cytogenetics t(9;22) only 200 (87) Additional abnormalities* 29 (13) Maximum dose of IM 400 mg/d 210 (92) 600 mg/d 17 (7) 800 mg/d 2 (1)

Abbreviations: INF, interferon; BMT, bone marrow transplantion. *Additional cytogenetic findings were detected: -Y (n = 5); double Ph+ chromosome (n = 5); t(7;8) (n = 2); t(9;22;22) (n = 2); t(2;9;22) (n = 2); t(9;22;17) (n = 1); t(1;22;18) with inv (5) (n = 1); t(7;8) with +8 and +der(22) (n = 1); t(8;17) (n = 1); t(8;16) (n = 1); inv (9q) (n = 1); -18q (n = 1); t(12;16) (n = 1); t(3;19) (n = 1); t(4;6), 47-52, +X, +6, +8, +18, +19, +der(22) (n = 1); t(17;20), +der(17), +der (20) (n = 1); +8 (n = 1); -X (n = 1). of BCR/ABL fusion gene transcripts by quantitative PCR compared estimated using the Kaplan-Meier method. Demographic and disease with the base line, andcomplete molecular response (CMoR) was characteristics were comparedaccordingto the genotypes of 16 SNPs in defined as disappearance of detectable BCR/ABL fusion gene transcripts, additive, dominant, and recessive models using m2 test, Fisher’s exact equivalent to a 5 log reduction. For major and complete cytogenetic test, or Mann-Whitney’s U-test. In univariate analyses, the treatment response, bone marrow evaluation andfluorescence in situ hybridiza- outcomes such as MCyR, CCyR, MMoR, CMoR, LOR, treatment failure, tion studies were done every 3 mo until a complete cytogenetic TFS, andOS were also comparedaccordingto the genotypes of 16 SNPs response was confirmed. For molecular response, the peripheral blood in additive, dominant, and recessive models using the log-rank test. In BCR/ABL real-time quantitative-PCR was repeatedevery 3 mo regardless multivariate analyses using Cox proportional hazard models, the disease of cytogenetic response. Time to treatment failure was defined as the stage (CP versus AP or BC), the presence of additional cytogenetical interval between the initiation of IM therapy andthe occurrence of abnormality, race (caucasian versus non-caucasian), age, andsignificant events that indicated that patients had failed IM, including primary genotypes were considered as covariates for each event: ABCG2 hematologic resistance, cytogenetic resistance, or loss of response (rs2231137) and CYP3A5 (rs776746) for MCyR andfor CCyR; ABCG2 (LOR). Time to LOR was defined as the interval between the date of any (rs2131142) for MMoR andfor CMoR; SLC22A1 (rs683369) for LOR confirmedresponse andthe dateat which the criteria for response were andtreatment failure; ABCB1 (rs1045642) for OS; and ABCG2 no longer met, such as (a) transformation from CP to AP or BC, (b) loss (rs2131142) and SLC22A1 (rs683369) for dose escalation of IM. The of CCyR/MCyR, and( c) confirmedincrease of z1 log of BCR/ABL multivariate analyses using Cox proportional hazardmodels were mRNA transcript for patients in CCyR or more. conducted using backward stepwise modeling and a P value for the Time to transformation-free survival was defined as the interval likelihoodratio test of >0.05. The hazardratios (HR) and95% between the initiation of IM therapy andconfirmation of progression to confidence intervals (95% CI) were also estimated. AP or BC or death from any cause, whereas overall survival (OS) was Basedon the results from multivariate models, we generatedtwo calculated from the initiation of IM therapy until the date of death from scoring models (a) using the disease stage and genotype data of ABCG2 any cause or the date of last follow-up. (rs2231137) and CYP3A5 (rs776746) for the prediction of MCyR and Statistical analysis. The 16 SNPs were primarily evaluatedfor CCyR; and(b ) using the disease stage and genotype data of SLC22A1 adequacy of Hardy-Weinberg Equilibrium using m2 test. The frequency (rs683369) for LOR and treatment failure. In each predictive model, a of each genotype was calculatedusing Haploview software (available at score of 1 was assignedto adversegenotype (i.e. GG genotype for http://www.broad.mit.edu/mpg/haploview). Additive, dominant, and ABCG2, rs2231137; AA genotype for CYP3A5, rs776746; or GG recessive models were tested in each SNP for the association with genotype for SLC22A1, rs683369) or advanced-disease stage, and a treatment outcomes. score 0 was assignedto referent genotype (i.e. AA or AG genotype for The medical records of all patients were reviewed until the end of July ABCG2, rs2231137; GG or GA genotype for CYP3A5, rs776746; CC or 2007. The basic demographic and disease characteristics were examined. CG genotype for SLC22A1, rs683369) or chronic-phase disease. After The cumulative incidences of hematologic response, MCyR, CCyR, summing up the scores for the predictive model of MCyR or CCyR, three MMoR, andCMoR, anddose escalation of IM were calculated risk groups were obtained: low (composite score 0; n = 27), intermediate considering the discontinuation of IM as competing risks for interest (composite score 1; n = 173), andhigh risk (composite score 2 or 3; n = events. The probability of freedom from LOR and from treatment failure 28). And for the predictive model of LOR or treatment failure two risk was estimatedandplottedusing the Kaplan-Meier method.The groups were generated: low (composite score 0; n = 198) andhigh risk probabilities of OS andtransformation-free survival (TFS) were also (composite score 1; n = 30).

Clin Cancer Res 2009;15(14) July 15, 2009 4752 www.aacrjournals.org Downloaded from clincancerres.aacrjournals.org on September 29, 2021. © 2009 American Association for Cancer Research. Pharmacogenetics of Imatinib Therapy for CML

To validate the genetic effect, we carried out bootstrap algorithm to With median follow-up of 47.3 months, 46 cases (20%) of construct the bootstrap confidence interval. The bootstrap, like cross- IM treatment failure were documented including due to validation, is a resampling technique; however, it differs in the resistance (n = 38) or intolerance (n = 8). The probability of manner of resampling. Bootstrap data sets are created by sampling freedom from LOR was 71% (95% CI, 63-78%) at 2 years, 60% with replacement, whereas cross-validation methods sample without (95% CI, 51-69%) at 3 years, and53% (95% CI, 44-62%) at 4 replacement. The bootstrap methodhas been shown to give a nonparametric maximum likelihoodestimate of the predictionerror years after any response to IM therapy hadbeen attained.The andcan correct for the bias of the estimate. We appliedbootstrap probability of freedom from treatment failure was 69% (95% basedon 500 replications. The results were generatedusing PROC CI, 62-76%), 58% (95% CI, 51-65%), and49% (95% CI, SURVERY SELECT procedure of SAS version 9.1, and presented as the 40-58%) at 2, 3, and4 years after initiation of IM therapy. The bootstrap HR confidence intervals of the genetic effects and important 5-year probability of TFS andOS was 90% (95% CI, 87-94%) clinic factors. and95% (95% CI, 92-99%), respectively. The rates of MCyR, CCyR, MMoR andCMoR were eval- Treatment outcomes ofIM therapy according to the candidate uatedat the time of 1, 1.5, 2, 3, 4, and5 y after IM therapy. genotypes. Treatment outcomes were comparedaccordingto The MCyR or CCyR was comparedin each time point accord- each candidate genotype for the nine statistical end points ing to the predictive model for MCyR or CCyR using m2 test. The using additive, dominant, and recessive models. All the cumulative incidences of MCyR and CCyR were also compared according to the predictive model for MCyR or CCyR in overall significant results in the univariate analyses are presentedin ABCG2 patients andin a group confinedto the patients in CP. Table 2. In summary, the GG genotype in SNP The serial mRNA levels of the BCR/ABL fusion transcript were also (rs2231137) andAA genotype in CYP3A5 SNP (rs776746) analyzedevery 3 mo up to 24 mo as a logarithmic-transformedformat. hadadverseimpacts on achievement of MCyR andCCyR, The differences of the BCR/ABL fusion transcript mRNA levels were whereas AC or CC genotypes in ABCG2 SNP (rs2231142) compared according to the risk score model based on the disease stage showedadverseeffect in terms of MMoR or CMoR. With respect andgenotype dataof ABCG2 (rs2231137) and CYP3A5 (rs776746) to the parameters indicating treatment failure, the GG genotype using repeatedmeasures ANOVA test. in SLC22A1 SNP (rs683369) was associatedwith higher risk of Multiple comparisons were implicated in this study. That considered, LOR or treatment failure. As regardto OS, the TT genotype in this study is exploratory and hypothesis-generating. All statistical tests ABCG2 SNP (rs1045642) correlatedwith higher risk of death. were two-sided with the significance level set at 0.05 unless otherwise stated. The statistical data were obtained using an SPSS software In the subgroup analysis confinedto chronic-phase patients, package (SPSS 13.0 Inc.) andSAS version 9.1 (SAS Institute). The similar outcomes have been noted. cumulative incidence curves were obtained using R package, version For CYP3A5 SNP (rs776746), the minor allele frequency was 2.4.1 (available at http://CRAN.R-project.org). 0.118 in the overall population. However, it differs between Caucasian (0.044) andnon-Caucasian (0.331). The Hardy- P Results Weinberg Equilibrium value in each group was appropriate both in Caucasian (P = 0.28) andnon-Caucasian ( P = 0.38). Demographic and disease characteristics prior to IM Accordingly, we did subpopulation stratification according to therapy. The demographic and disease characteristics before the CYP3A5 (rs776746). In subpopulation analysis, the MCyR the commencement of IM therapy are summarizedin Table 1. A was significant in both Caucasian (P = 0.002) andnon- total of 203 patients (89%) were in CP, 23 (10%) were in AP, Caucasian (P = 0.008). The CCyR was significant in both and3 (1%) were in BC before commencement of IM. Caucasian (P = 0.03) andnon-Caucasian ( n = 0.003). The Thegenotypefrequenciesofthecandidategenes.The stratified analysis adjusted for race also provided significant genotype frequencies for 16 candidate SNPs are summarized results (P = 0.01 for MCyR and P = 0.02 for CCyR). in the Supplemental Table S3. We failedto detectpolymor- The results of multivariate analyses are summarizedin phisms for three loci (rs28383469, rs1126724, and rs3182034). Table 3. When considering the confounding effect of the The LDs of five gene polymorphisms are shown in the disease phase to the treatment outcomes of imatinib therapy, Supplemental Fig. S1. There were strong LDs within SNPs of the GG genotype in ABCG2 (rs2231137), AA genotype in the SLC22A1 gene, especially among rs1867351, rs683369, and CYP3A5 (rs776746), andadvancedstage were significantly rs628031: D’ = 1.00 between rs1867351 andrs683369, between associatedwith lower response rate to IM therapy, especially rs683369 andrs628031, andbetween rs1867351 andrs628031. for MCyR or CCyR, whereas the GG genotype at SLC22A1 The comparison of genotype frequency according to the (rs683369) andadvancedstagecorrelatedwith higher rate of patients’ demographics did not show any differences except for LOR or treatment failure. The CC genotype in ABCG2 SLC22A1 (rs1867351, P =0.03)andAGP (rs3182041, (rs2231142) was also identified as an independent predictor P = 0.08) between the patients in CP versus those in AP/BC. of more frequent needfor IM doseescalation. Overall treatment outcomes ofimatinib therapy. The We also carriedout the haplotype analysis for the SLC22A1 treatment outcomes in overall patients are summarizedin genes. We generatedthe haplotype of SLC22A1 gene poly- Fig. 1. With a median duration of IM administration of 40.8 morphisms basedon three genotypes (rs1867351, rs683369, months, the incidence of hematologic response was 96% (95% andrs628031). The frequencies of SLC22A1 haplotypes were CI, 92-99%) at 3 months. The incidence of MCyR and CCyR 37.4% (ACG), 25.3% (GCG), 18.8% (ACA), and18.4% (AGA). was 86% (95% CI, 81-91%) and62% (95% CI, 56-69%) at 12 The SLC22A1 gene haplotype was not correlatedwith MCyR, months after initiation of IM therapy, respectively. The CCyR, LOR, or treatment failure (data not shown). incidence of MMoR was 33% (95% CI, 27-40%), 52% (95% Internal validation using bootstrap method. The results CI, 45-59%), and64% (95% CI, 57-72%) at 1, 2, and3 years, showing the bootstrap HR confidence intervals of the genetic andthe incidenceof CMoR was 29% (95% CI, 23-37%) and effects andimportant clinic factors are shown in Supplemental 32% (95% CI, 26-40%) at 2 and3 years. Table S3. The bootstrap results support our previous result.

www.aacrjournals.org 4753 Clin Cancer Res 2009;15(14) July 15, 2009 Downloaded from clincancerres.aacrjournals.org on September 29, 2021. © 2009 American Association for Cancer Research. Cancer Therapy: Clinical

Fig. 1. The probability of response to or treatment failure after imatinib mesylate theraphy in overall group. A, response to imatinib theraphy (MCyR, CCyR, MMoR, CMoR) B, probability without loss of response or treatment failure. C, probability of overall survival or free from progression to advanced phase. X axis in B, ‘‘years after any response to imatinib theraphy’’ for loss of response, and ‘‘years after starting imatinib theraphy’’ for treatment failure.

Because the bootstrap algorithm provides a nonmodel based Then, we comparedthe cumulative incidenceof MCyR and confidence interval of the hazard ratio, it is more robust than CCyR according to the predictive model using two genotypes the model-based estimates. and disease stages. The median time to MCyR was significantly Predictive models for the response or treatment failure to different (P = 0.003): 106 days (95% CI, 66-146 days) for the imatinib therapy. As the above results showing that the ABCG2 low-risk group, 146 days (95% CI, 110-181 days) for the genotype (rs2231137) and CYP3A5 genotype (rs776746) were intermediate-risk group, and 279 days (95% CI, 224-334 days) significantly associatedwith response to imatinib therapy for the high-risk group. The median time to CCyR was also together with disease stage, we generated the predictive model different in favor of the low-risk group (P < 0.001; Fig. 3A): incorporating information of two genotypes anddiseasestages. 168 days (95% CI, 148-188 days) for the low-risk group, Three risk groups were obtained: low (composite score 0; 239 days (95% CI, 186-292 days) for the intermediate-risk n = 27), intermediate (composite score 1; n = 173), andhigh group, and 962 days (95% CI, 354-1,569 days) for the high-risk risk (composite score 2 or 3; n =28).TheMCyRwas group. Also, a similar result was notedwhen analysis was significantly different according to the three risk groups at confinedto the patients in chronic phase with respect to MCyR 1 year (P < 0.001), 1.5 years (P < 0.001), 2 years (P < 0.001), or CCyR (Fig. 3B). The cumulative incidence curves of MMoR 3 years (P = 0.001), and4 years (P = 0.04; Fig. 2A), whereas the according to this predictive model are available in Fig. 3C. CCyR was also different according to the three risk groups at Basedon the multivariate result showing the GG genotype 1 year (P < 0.001), 1.5 years (P < 0.001), 2 years (P < 0.001), for SLC22A1 (rs683369) and advanced stage as adverse risk and3 years ( P = 0.001; Fig. 2B). factors for LOR or treatment failure, we generatedthe predictive In order to confirm the above result in the group restricting to model for LOR and treatment failure incorporating the chronic-phase disease, we generated another predictive model SLC22A1 genotype anddiseasestage. As mentionedin the incorporating information of the two genotypes of ABCG2 statistical section, two risk groups were generated: low (rs2231137) and CYP3A5 (rs776746), andassignedscore 1 to (composite score 0; n = 198) andhigh risk (composite score adverse genotype and score 0 to referent genotype. Three risk 1; n = 30). The 3-year probabilities of freedom from LOR were groups were generatedafter summing up the scores: low 72 F 4% and54 F 11% for the low- andhigh-risk groups, (composite score 0; n = 30), intermediate (composite score 1; respectively (P < 0.001; Fig. 3D). A similar result was noted n = 191), andhigh risk (composite score 2; n = 7). Similar results when analysis was confinedto patients in chronic phase for were notedfor MCyR andCCyR as shown in Fig. 2C andD. LOR (P < 0.001; Fig. 3E). The 3-year probabilities of freedom

Table 2. Results of the univariate analyses for the treatment outcomes according to the candidate genotypes

Outcomes Gene, rs number Referent genotype Adverse genotype P HR (95% CI) MCyR ABCG2, rs2231137 AA or AG GG 0.05 0.68 (0.46-1.00) MCyR CYP3A5, rs776746 GG or GA AA 0.01 0.34 (0.14-0.83) CCyR ABCG2, rs2231137 AA or AG GG 0.02 0.63 (0.42-0.94) CCyR CYP3A5, rs776746 GG or GA AA 0.03 0.39 (0.16-0.96) MMoR ABCG2, rs2231142 AA AC or CC 0.004 0.40 (0.20-0.81) CMoR ABCG2, rs2231142 AA AC or CC 0.006 0.42 (0.21-0.85) LOR SLC22A1, rs683369 CC or CG GG 0.0008 4.86 (1.74-13.52) Treatment failure SLC22A1, rs683369 CC or CG GG 0.02 3.24 (1.18-8.89) OS ABCB1, rs1045642 CC or CT TT 0.04 3.70 (0.96-14.7)

Clin Cancer Res 2009;15(14) July 15, 2009 4754 www.aacrjournals.org Downloaded from clincancerres.aacrjournals.org on September 29, 2021. © 2009 American Association for Cancer Research. Pharmacogenetics of Imatinib Therapy for CML

Table 3. Multivariate regression model based on Cox’s proportional hazard model for each end point

P, HR (95% CI) Candidate genes, SNP ID Prognostic factors ABCG2; ABCG2; CYP3A5; SLC22A1; Disease risk rs2231137 rs2231142 rs776746 rs683369 Adverse group GG genotype CC genotype AA genotype GG genotype AP or BC

Response to imatinib mesylate therapy, P, HR (95% CI)

MCyR 0.04, 0.67 — 0.02, 0.34 —— (0.45-0.98) (0.14-0.83) CCyR 0.03, 0.64 — 0.04, 0.39 — 0.01, 0.47 (0.43-0.95) (0.16-0.95) (0.26-0.84) MMoR — — 0.08, 0.36 — 0.04, 0.47 (0.11-1.13) (0.23-0.95) CMoR — 0.005, 0.48 ——— (0.28-0.80)

Resistance to imatinib mesylate therapy, P, HR (95% CI)

LOR — — — 0.001, 5.47 0.011, 2.57 (1.95-15.36) (1.24-5.32) Treatment failure — — 0.06, 2.40 0.009, 3.86 0.002, 2.67 (.96-6.01) (1.39-10.69) (1.45-4.91) Transformation-free survival — — — — 0.003, 9.76 (2.18-43.65) Death — — — — 0.003, 9.76 (2.18-43.65)

Abbreviations: MCyR, major cytogenetic response; CCyR, complete sytogenetic response; MMoR, major molecular response; CMoR, complete molecular response; AP, accelerated phase; BC, blastic crisis.

from treatment failure were 68 F 4% and42 F 10% for the levels. The current study included 16 SNPs in 5 genes low- andhigh-risk groups, respectively ( P < 0.001). The result associatedwith imatinib metabolism andtransport, andthis of analysis confinedto chronic phase was also similar for multiple candidate genes approach may gain a benefit of treatment failure (P = 0.006). excluding potential effects of gene-to-gene interaction between Serial changes ofthe BCR/ABL fusiontranscript levels genes. according to the predictive model. We also appliedrepeated A previous study suggested that not only ABCB1 but also measures ANOVA model on the BCR/ABL fusion gene transcript ABCG2 is overexpressedin hematopoietic stem cells andthat mRNA levels according to the ABCG2 genotype (rs2231137), imatinib can be a substrate for ABCG2 (29). The ABCG2 gene, CYP3A5 genotype (rs776746), anddiseasestage. The P value locatedon chromosome 4q22, encodesfor a 655 amino-acid for the overall comparison of repeatedmeasures was <0.0001 polypeptide [breast cancer resistance protein (BCRP)] that is (Fig. 3F). capable of pumping imatinib out of cells (30). ABCG2 is expressednot only in the hematopoietic stem cells but also in the hepatic cells or intestinal epithelial cells, affecting the Discussion bioavailability of imatinib (30, 31). Whether ABCG2 is involvedin the resistance mechanism of imatinib is a matter In the current study, we have shown that (a) the treatment of debate (32–36). Gardner et al. reported that the ABCG2 SNP outcomes of imatinib therapy can be predicted using the SNP Q141K (rs2231142 in the current article) is associatedwith the approach in CML patients; (b) the response to imatinib was intracellular level of imatinib in vitro model (6). Several SNPs strongly associatedwith SNPs for ABCG2 (rs2231137) and were known to modify the transporter activity of ABCG2 CYP3A5 (rs776746), anddiseasestage; and( c) the treatment (11, 12). In the present study, the ABCG2 genotype V12M failure including LOR was associated with SNP for SLC22A1 (rs2231137) was foundto be significantly associatedwith (rs683369) anddiseasestage. Thus, the present studysuggested cytogenetic response to imatinib therapy. Similar to a previous that the pharmacogenetic variability of individuals might result using an in vitro model (6), the ABCG2 genotype Q141K predict the treatment outcomes of imatinib therapy in CML (rs2231142) was associatedwith molecular response to patients. imatinib therapy in univariate analyses although it was not Picardet al. reportedthat trough plasma levels of imatinib are statistically significant in multivariate analyses. associatedwith cytogenetic andmolecular response to IM Of interest, two SNPs in the ABCG2 gene showedgood therapy in CML patients, suggesting an importance of individual correlation with different parameters of treatment response. pharmacokinetic differences on the response to IM therapy (28). The GG genotype for ABCG2, rs2231137 was associatedwith a Hence, interindividual variations in uptake and clearance by lower CCyR, whereas the non-AA genotype for ABCG2, transporter protein, metabolic inactivation, and/or plasma rs2231142, was also associatedwith a lower MMoR. The bindinginhibition couldleadto changes in systemic imatinib linkage disequilibrium between these two SNPs was significant

www.aacrjournals.org 4755 Clin Cancer Res 2009;15(14) July 15, 2009 Downloaded from clincancerres.aacrjournals.org on September 29, 2021. © 2009 American Association for Cancer Research. Cancer Therapy: Clinical

(D¶ = 0.67). This potential linkage disequilibrium between of White et al. (42) The group with the GG genotype of rs2231137 andrs2231142 may explain why both these SNPs SLC22A1 (rs683369) showeda 5.5-foldand3.9-foldhigher associate with response to IM, either CCyR or MMoR. The risk of treatment failure or loss of response to IM therapy, reason one was associatedwith CCyR andanother with MMoR respectively (Table 3). The interindividual variability of the may be due to different frequencies of minor alleles in each SNP. SLC22A1 genotype or gene expression couldaffect the In the present study, the SLC22A1 genotype was well variability of OCT1 protein activity (38). Controversy still correlatedwith treatment outcomes, especially for LOR or remains on whether the SLC22A1 genotype coulddetermine treatment failure. The human organic cation transporter, OCT1, the levels of SLC22A1 mRNA transcription or OCT1 activity was known to mediate active influx of imatinib into cells, thus a (13–15), thus facilitating further study to investigate the critical determinant of intracellular imatinib concentration correlation of the SLC22A1 genotype andOCT1 activity. (37). White et al. suggestedthat the interpatient variability of Although it was not statistically significant in the multivariate OCT1 activity was an important determinant of molecular analysis, the SLC22A1 genotype (rs683369) was associated response to IM (38, 39). They reportedthat patients with low with different MMoR, suggesting that the SLC22A1 genotype OCT1 activity showedhigher failure rates to achieve MMoR by couldaffect the response to imatinib therapy (Supplemental 18 months (82% versus 17% of failure rate) andto achieve a 2- Fig. S2). Maybe because of the relatively low frequency of the log reduction of BCR/ABL mRNA transcript by 12 months SLC22A1 GG genotype (18%), it couldnot be identifiedina (45% versus 8% of failure rate; ref. 38). Other studies also multivariate analysis. suggesteda significant association of SLC22A1 mRNA tran- The cytochrome P450 (CYP) 3A (CYP3A) family is the scription levels with imatinib response (40, 41). Wang et al. predominant drug metabolizing enzyme and accounts for reportedthat the group with high baseline OCT1 expression approximately 30% of hepatic CYP and>70% of intestinal showedbetter progression-free ( P = 0.01) or overall survival CYP expression. Imatinib is predominantly metabolized by (P = 0.004) comparedwith those with low baseline, andthat CYP3A4 andCYP3A5 in the liver (17, 18, 43, 44). This study the pre-imatinib OCT1 expression level was much higher in showedthat the AA genotype in the CYP3A5 genotype those achieving CCyR at 6 months than noncytogenetic (rs776746) hada 66% and61% lower rate of MCyR andCCyR responders (27). Our result is quite consistent with the result comparedwith the referent genotype group, suggesting an

Fig. 2. The differences of response rates to imatinib theraphy according to the risk score in overall CML patients. Notably, differences in response rates were noted according to the risk score based on the disease stage and ABCG2/CYP3A5 genotypes in terms of major (A) or complete cytogenetic responses (B). Similar patterns were also noted when we only included the ABCG2/CYP3A5 genotypes in the risk model without the information of disease stage (C and D).

Clin Cancer Res 2009;15(14) July 15, 2009 4756 www.aacrjournals.org Downloaded from clincancerres.aacrjournals.org on September 29, 2021. © 2009 American Association for Cancer Research. Pharmacogenetics of Imatinib Therapy for CML

Fig. 3. The risk score model based on the ABCG2, CYP3A5,orSLC22A1 genotypes, together with disease stage, may predict the response or resistance to imatinib therapy for CML. The predictive model for the complete cytogenetic response based on the information of the ABCG2/CYP3A5 genotype and disease phase (A, B)waswell correlated with clinical outcomes in the overall group (N = 229) and in chronic-phase patients (n = 203).The model also showed higher incidence of major molecular response in the low-risk group compared with the high-risk group (C). In addition, the predictive model based on the information of the SLC22A1 genotype and disease phase was well correlated with the loss of response (D, E ) in the overall group and in-chronic phase patients. Furthermore, the predictive model based on the ABCG2/CYP3A5 genotypes and disease stage showed good correlation with different level of the serial change of BCR/ABL fusion transcript levels following imatinib therapy (F). important role of the CYP3A5 genotype (rs776746) in predict- failure) couldbe reasonably acceptedas statistical endpoints ing response to IM therapy. for pharmacogenetic markers to predict the outcomes to One of the interesting findings was that the ABCB1 genotype imatinib therapy in CML patients. These endpoints couldbe did not associate with any response or outcome parameters. adopted in a future studies evaluating the power of other Several previous investigations suggestedthat ABCB1 gene candidate SNP markers to predict the treatment outcomes of (MDR1) overexpression may confer the resistance of imatinib imatinib therapy. in cell line model (4, 37). The ABCB1 gene, which belongs to Because of the limitation of the retrospective study, the the ABC superfamily, encodes for a transmembrane glycopro- information of the Sokal score or percentage of Ph+ metaphase tein (P-glycoprotein) capable of pumping imatinib out of the was not available, which were suggestedas potential predictive tumor cell (45). Accordingly, P-glycoprotein was suggested as a factors for the response to imatinib therapy in CML. Further potential determinant of intracellular concentration of imatinib studies should consider these variables as important covariates. through ATP-dependent efflux pathway (46). C3435T The other issue is that the disease stage is linked to the dose (rs1045642), C1236T (rs1128503), andG2677T or A administered. This may confound any analysis where nonlin- (rs2032582) have been the most frequently investigatedpoly- earity of pharmacokinetics may be important. However, morphisms for the ABCB1 gene (46–48). However, unlike subgroup analysis confinedto the chronic-phase patients other study results, we could not identify any association of reproduced a similar result to those from overall patients ABCB1 gene polymorphisms with clinical outcomes of (Fig. 2C andD andFig. 3B andE). imatinib therapy for CML. The approach targeting other A potential debate is the use of GAPDH as a control gene for genotypes of the ABCB1 gene besides C3435T, C1236T, and determining BCR/ABL transcript level. GAPDH was usedas G2677T or A genotypes will be necessary. internal control gene to monitor BCR-ABL transcript levels for Several end points were included in the present study to the past 7 years in our institute. Andnow our institute has evaluate which endpoint coulddiscriminatethe patients by their switchedto BCR as internal control gene. Our validation study risk based on individual differences of pharmacogenetic showed that the log reduction difference by using GAPDH and markers. At least three events (i.e. CCyR, LOR, andtreatment BCR from 95% of patients is within 0.5 log (data not shown).

www.aacrjournals.org 4757 Clin Cancer Res 2009;15(14) July 15, 2009 Downloaded from clincancerres.aacrjournals.org on September 29, 2021. © 2009 American Association for Cancer Research. Cancer Therapy: Clinical

Accordingly, the internal and external reproducibility of (rs683369). Further studies are warranted to validate the role measurement of BCR/ABL transcript level was maintainedin of these SNPs in the early identification of individuals with CML the current study. who may not respondoptimally to standardIM therapy. In conclusion, with a novel, multiple candidate gene approach basedon the pharmacogenetics of IM, we identifiedseveralSNP candidates in patients with CML that are potential predictors of Disclosure of Potential Conflicts of Interest clinical response to IM, the ABCG2 (rs2231137) or CYP3A5 genotype (rs776746); andof resistance to IM, SLC22A1 No potential conflicts of interest were disclosed.

References 1. Kantarjian HM, Cortes JE, O’Brien S, et al. Long-term 17. Bolton AE, Peng B, Hubert M, et al. Effect of rifam- interaction with ABCG2 on primitive hematopoietic survival benefit and improved complete cytogenetic picin on the pharmacokinetics of imatinib mesylate stem cells. Leukemia 2007;21:1267 ^ 75. and molecular response rates with imatinib mesylate (Gleevec, STI571) in healthy subjects. Cancer Che- 33. Houghton PJ, Germain GS, Harwood FC, et al. in Philadelphia chromosome-positive chronic-phase mother Pharmacol 2004;53:102 ^ 6. Imatinib mesylate is a potent inhibitor of the ABCG2 chronic myeloid leukemia after failure of interferon-a¤ . 18. Dutreix C, Peng B, Mehring G, et al. Pharmacokinet- (BCRP) transporter and reverses resistance to topote- Blood 2004;104:1979 ^88. ic interaction between ketoconazole and imatinib can and SN-38 in vitro. Cancer Res 2004;64:2333^ 7. 2. Kantarjian HM, Talpaz M, Giles F, O’Brien S, mesylate (Glivec) in healthy subjects. Cancer Chemo- 34. Jordanides NE, Jorgensen HG, Holyoake TL, Cortes J. New insights into the pathophysiology of ther Pharmacol 2004;54:290 ^ 4. Mountford JC. Functional ABCG2 is overexpressed chronic myeloid leukemia and imatinib resistance. 19. FitosI,VisyJ,ZsilaF,MadyG,SimonyiM.Selective on primary CML CD34+ cells and is inhibited by ima- Ann Intern Med 2006;145:913^23. binding of imatinib to the genetic variants of human tinib mesylate. Blood 2006;108:1370 ^ 3. 3. Gorre ME, Mohammed M, Ellwood K, et al. Clinical alpha1-acid glycoprotein. Biochim Biophys Acta 35. NakanishiT,Shiozawa K, Hassel BA, Ross DD.Com- resistance to STI-571 cancer therapy caused by 2006;1760:1704^12. plex interaction of BCRP/ABCG2 and imatinib in BCR- BCR-ABL gene mutation or amplification. Science 20. Gambacorti-Passerini C, Zucchetti M, Russo D, et ABL-expressing cells: BCRP-mediated resistance to 2001;293:876 ^80. al. Alpha1 acid glycoprotein binds to imatinib imatinib is attenuated by imatinib-induced reduction of 4. Mahon FX, Belloc F, Lagarde V, et al. MDR1 gene (STI571) and substantially alters its pharmacokinetics BCRPexpression. Blood 2006;108:678 ^84. overexpression confers resistance to imatinib mesy- in chronic myeloid leukemia patients. Clin Cancer Res 36. Ozvegy-Laczka C, HegedusT,Varady G, et al. High- late in leukemia cell line models. Blood 2003;101: 2003;9:625 ^ 32. affinity interaction of tyrosine kinase inhibitors with 2368 ^73. 21. Gambacorti-Passerini C, le Coutre P, Zucchetti M, the ABCG2 multidrug transporter. Mol Pharmacol 5. Haiat S, Decleves X, Mittaine B, et al. Determination D’Incalci M. Binding of imatinib by alpha(1)-acid gly- 2004;65:1485^95. of imatinib plasma levels by high performance liquid coprotein. Blood 2002;100:367 ^ 8; author reply 8 ^ 9. chromatography (HPLC): evaluation of pharmacoki- 37.ThomasJ,Wang L, Clark RE, Pirmohamed M. Active 22. Gambacorti-Passerini C, Barni R, le Coutre P, et al. transport of imatinib into and out of cells: implications netic variability and haematological consequences. Role ofalpha1acid glycoproteininthe in vivo resistance Blood 2006;108:1375a. for drug resistance. Blood 2004;104:3739 ^ 45. of human BCR-ABL(+) leukemic cells to the abl in- 38. WhiteD,SaundersV,DangP,etal.MostCML 6. Gardner ER, Burger H, van Schaik RH, et al. Associa- hibitor STI571.JNatl Cancer Inst 2000;92:1641 ^ 50. tion of enzyme and transporter genotypes with the patients who have a suboptimal response to imatinib 23. Branford S, Fletcher L, Cross NC, et al. Desirable have low OCT-1 activity: higher doses of imatinib pharmacokinetics of imatinib. Clin Pharmacol Ther 2006;80:192 ^201. performance characteristics for BCR-ABL measure- may overcome the negative impact of low OCT-1 ac- ment on an international reporting scale to allow tivity. Blood 2007;110:4064 ^ 72. 7. IllmerT, Schuler US,Thiede C, et al. MDR1gene poly- consistent interpretation of individual patient response morphisms affect therapy outcome in acute myeloid 39.White D, Saunders V, Dang P, et al. OCT-1-mediated and comparison of response rates between clinical leukemia patients. Cancer Res 2002;62:4955 ^ 62. influx is a key determinant of the intracellular uptake of trials. Blood 2008;112:3330 ^ 8. imatinib but not nilotinib (AMN107): reduced OCT-1 8. Kim DH, Park JY, Sohn SK, et al. Multidrug resis- 24. Muller MC, Saglio G, Lin F, et al. An international activity is the cause of low in vitro sensitivity to imati- tance-1gene polymorphisms associated with treat- study to standardize the detection and quantitation of nib. Blood 2006;108:697 ^704. ment outcomes in de novo acute myeloid leukemia. BCR-ABL transcripts from stabilized peripheral blood Int J Cancer 2006;118:2195 ^ 201. 40. Crossman LC, Druker BJ, Deininger MW, preparations by quantitative RT-PCR. Haematologica PirmohamedM,WangL,ClarkRE.hOCT1and 9. Gurney H,Wong M, Balleine RL, et al. Imatinib dispo- 2007;92:970 ^ 3. sition and ABCB1 (MDR1, P-glycoprotein) genotype. resistance to imatinib. Blood 2005;106:1133^ 4; au- thor reply 4. Clin Pharmacol Ther 2007;82:33 ^40. 25. Branford S, Cross NC, Hochhaus A, et al. Rationale for the recommendations for harmonizing current 10. Yanase K, Tsukahara S, Mitsuhashi J, Sugimoto Y. 41. Wang L, Giannoudis A, Lane S, Williamson P, methodology for detecting BCR-ABL transcripts in Functional SNPs of the breast cancer resistance pro- Pirmohamed M, Clark R. Expression of the uptake patients with chronic myeloid leukaemia. Leukemia tein-therapeutic effects and inhibitor development. drug transporter hOCT1is an important clinical deter- 2006;20:1925 ^30. Cancer Lett 2006;234:73^ 80. minant of the response to imatinib in chronic myeloid leukemia. Clin Pharmacol Ther 2008;83:258 ^ 64. 11. Morisaki K, Robey RW, Ozvegy-Laczka C, et al. Sin- 26. Druker BJ, Talpaz M, Resta DJ, et al. Efficacy and gle nucleotide polymorphisms modify the transporter safety of a specific inhibitor of the BCR-ABL tyrosine 42. White DL, Saunders VA, Dang P, et al. Most CML activity of ABCG2. Cancer Chemother Pharmacol kinase in chronic myeloid leukemia. N Engl J Med patients who have a suboptimal response to imatinib 2005;56:161 ^ 72. 2001;344:1031 ^ 7. have low OCT-1 activity; higher doses of imatinib may overcome the negative impact of low OCT-1 ac- 12. Honjo Y, Morisaki K, Huff LM, et al. Single- 27. Baccarani M, Saglio G, Goldman J, et al. Evolv- nucleotide polymorphism (SNP) analysis in the ABC ing concepts in the management of chronic myeloid tivity. Blood 2007;110:4064 ^ 72. half-transporter ABCG2 (MXR/BCRP/ABCP1). Cancer leukemia: recommendations from an expert panel on 43. Peng B, Lloyd P, Schran H. Clinical pharmacokinet- Biol Ther 2002;1:696 ^ 702. behalf of the European LeukemiaNet. Blood 2006; ics of imatinib. Clin Pharmacokinet 2005;44:879 ^ 94. 10 8:1809 ^ 20. 13. Sakata T, Anzai N, Shin HJ, et al. Novel single 44. Marin D, Marktel S, Bua M, et al.The use of imatinib nucleotide polymorphisms of organic cation trans- 28. PicardS,TitierK,EtienneG,etal.Troughimatinibplas- (STI571) in chronic myelod leukemia: some practical porter 1 (SLC22A1) affecting transport functions. malevelsareassociatedwithbothcytogeneticandmo- considerations. Haematologica 2002;87:979 ^88. Biochem Biophys Res Commun 2004;313:789 ^ 93. lecular responses to standard-dose imatinib in chronic 45. Juranka PF, Zastawny RL, LingV. P-glycoprotein: 14. Schinkel AH, Jonker JW. Polymorphisms affecting myeloid leukemia. Blood 2007;109:3496 ^ 9.29. Bur- multidrug-resistance and a superfamily of membrane- function of the human organic cation transporter ger H, vanTol H, Boersma AW, et al. Imatinib mesylate associated transport proteins. FASEB J 1989;3: hOCT1 (SLC22A1): what are the consequences? (STI571) is a substrate for the breast cancer resis- 2583^ 92. Pharmacogenetics 2002;12:589 ^ 90. tance protein (BCRP)/ABCG2 drug pump. Blood 46. IllmerT, Schaich M, Platzbecker U, et al. P-glycopro- 2004;104:2940 ^ 2. 15. Kerb R, Brinkmann U, Chatskaia N, et al. Identifica- tein-mediated drug efflux is a resistance mechanism tion of genetic variations of the human organic cation 30. Robey RW, Polgar O, Deeken J, To KW, Bates SE. of chronic myelogenous leukemia cells to treatment transporter hOCT1and their functional consequences. ABCG2: determining its relevance in clinical drug re- with imatinib mesylate. Leukemia 2004;18:401 ^8. Pharmacogenetics 2002;12:591 ^ 5. sistance. Cancer Metastasis Rev 2007;26:39 ^ 57. 47. Fromm MF. The influence of MDR1 polymorphisms 16. Itoda M, Saito Y, Maekawa K, et al. Seven novel 31. Doyle LA, Ross DD. Multidrug resistance mediat- on P-glycoprotein expression and function in humans. single nucleotide polymorphisms in the human ed by the breast cancer resistance protein BCRP Adv Drug Deliv Rev 2002;54:1295 ^310. SLC22A1 gene encoding organic cation transporter (ABCG2). Oncogene 2003;22:7340 ^ 58. 48. Kimchi-Sarfaty C, Oh JM, Kim IW, et al. A ‘‘silent’’ 1 (OCT1). Drug Metab Pharmacokinet 2004;19: 32. Brendel C, Scharenberg C, Dohse M, et al. Imatinib polymorphism in the MDR1gene changes substrate 308^12. mesylate and nilotinib (AMN107) exhibit high-affinity specificity. Science 2007;315:525^ 8.

Clin Cancer Res 2009;15(14) July 15, 2009 4758 www.aacrjournals.org Downloaded from clincancerres.aacrjournals.org on September 29, 2021. © 2009 American Association for Cancer Research. Clinical Relevance of a Pharmacogenetic Approach Using Multiple Candidate Genes to Predict Response and Resistance to Imatinib Therapy in Chronic Myeloid Leukemia

Dong Hwan (Dennis) Kim, Lakshmi Sriharsha, Wei Xu, et al.

Clin Cancer Res 2009;15:4750-4758.

Updated version Access the most recent version of this article at: http://clincancerres.aacrjournals.org/content/15/14/4750

Supplementary Access the most recent supplemental material at: Material http://clincancerres.aacrjournals.org/content/suppl/2009/07/10/1078-0432.CCR-09-0145.DC1

Cited articles This article cites 47 articles, 22 of which you can access for free at: http://clincancerres.aacrjournals.org/content/15/14/4750.full#ref-list-1

Citing articles This article has been cited by 9 HighWire-hosted articles. Access the articles at: http://clincancerres.aacrjournals.org/content/15/14/4750.full#related-urls

E-mail alerts Sign up to receive free email-alerts related to this article or journal.

Reprints and To order reprints of this article or to subscribe to the journal, contact the AACR Publications Subscriptions Department at [email protected].

Permissions To request permission to re-use all or part of this article, use this link http://clincancerres.aacrjournals.org/content/15/14/4750. Click on "Request Permissions" which will take you to the Copyright Clearance Center's (CCC) Rightslink site.

Downloaded from clincancerres.aacrjournals.org on September 29, 2021. © 2009 American Association for Cancer Research.