The pharmacogenetics of imatinib

Imatinib mesylate (IM), a tyrosine kinase inhibitor, is one of the first molecularly targeted therapies to have been used in the clinic. It has proven to be efficient in the treatment of chronic myeloid leukemia and also in other malignancies that involve expression of a tyrosine kinase. However, some patients can develop resistance and others suffer from toxic side effects. The pharmacokinetics of IM depends on several enzymes and transporters, and several studies have attempted to identify genetic factors associated with variable drug levels and clinical responses using a candidate gene approach. Larger and more homogenous studies are still needed to replicate the findings obtained so far, or to analyze other genetic variations to get clearer insights into how IM treatment can be tailored to each patient's genetics. Here we summarize pharmacogenetic studies of IM and highlight the genetic markers that could be used to improve the treatment and management of diseases for which IM is used.


Introduction
Imatinib mesylate (IM, also known as STI571, Glivec or Gleevec) is a competitive tyrosine kinase inhibitor commonly used for treatment of chronic myeloid leukemia (CML). It has also proven effi cient for the treatment of advanced gastrointestinal stromal tumors (GISTs), c-KIT mastocytosis and myeloproliferative disorders with rearrangement of the platelet derived growth factor receptor (PDGFR) gene. In CML, IM inhibits the tyrosine kinase activity of the fusion protein of breakpoint cluster region (BCR) gene and the ABL tyrosine kinase (BCR-ABL), which results from a t(9;22)(q34,q11) translocation known as the Philadelphia chromosome; this fusion protein has a role in leukemogenesis [1]. IM occupies the ATP-binding pocket of the ABL kinase domain; this prevents a change in conformation of the protein that would otherwise convert the molecule to its active form, and IM binding thereby leads to the apoptosis of target cells.
Th ree main criteria can be used to evaluate the response to CML treatment: complete hematological remis sion, defi ned as a normal peripheral blood count with normal spleen; complete cytogenetic response (CCyR), defi ned by absence of the Philadelphia chromosome in bone marrow metaphase analysis; and major molecular response (MMR), defi ned by the thousandfold (3 log) reduction in BCR-ABL transcript levels relative to the standardized baseline [1]. Other levels of cytogenetic or molecular response can be used [1], as mentioned in Table 1. Despite outstanding results of IM in the treatment of chronic phase CML, some patients do not achieve response criteria (for example, about 25% of patients did not achieve CCyR at 18 months [2], and some patients (about 25%) who initially responded well subsequently acquired resistance [3]. Apart from dosage errors, drug interactions and non-compliance with treatment, several mechanisms can contribute to the development of resistance: changes to the target protein (occurring through mutations or BCR-ABL amplifi cation); downstream BCR-ABL-independent pathways; and drug pharmacokinetics parameters (absorption, distribution, metabolism and excretion). In malignancies other than CML, IM inhibits the tyrosine kinase domains of KIT and PDGFRα/β. Myeloproliferative disorders with PDGFR rearrangement show great sensitivity to IM and mostly require a lower dose of IM; this is especially true of chronic eosinophilic leukemia, which involves a fusion transcript of FIP1-like1 and PDGFRA [4]. In mastocytosis, the overall response rate varies according to c-KIT mutational status and has been reported to be 18 to 36% [5]. In GISTs, IM leads to a response rate of 50 to 70%, with a 2-year overall survival rate of 70% [6].
In the conventional dose range, a fourfold inter-patient variability has been reported both in the systemic exposure for a given dose and in the dose required to achieve a specifi c target level [7]. Variation in drug concentration may result in excessive toxicity or suboptimal anticancer eff ect. Reduced IM effi cacy has been Abstract Imatinib mesylate (IM), a tyrosine kinase inhibitor, is one of the fi rst molecularly targeted therapies to have been used in the clinic. It has proven to be effi cient in the treatment of chronic myeloid leukemia and also in other malignancies that involve expression of a tyrosine kinase. However, some patients can develop resistance and others suff er from toxic side eff ects. The pharmacokinetics of IM depends on several enzymes and transporters, and several studies have attempted to identify genetic factors associated with variable drug levels and clinical responses using a candidate gene approach. Larger and more homogenous studies are still needed to replicate the fi ndings obtained so far, or to analyze other genetic variations to get clearer insights into how IM treatment can be tailored to each patient's genetics. Here we summarize pharmacogenetic studies of IM and highlight the genetic markers that could be used to improve the treatment and management of diseases for which IM is used. linked to low drug exposure; the main reported toxicities of IM are neutropenia, superfi cial edema, nausea, muscle cramps and rashes [1-3]. Trough IM plasma levels (C min ) have been reported to be associated with MMR in CML patients following standard drug dose [8,9] A plasma threshold of 1,002 ng/ml had the best sensitivity and specifi city to predict MMR [8]. Patients with C min below this threshold have less chance to achieve MMR compared with patients with C min above it. Likewise, higher circulating levels of IM correlated with a better response rate and with a longer time to progression in patients with advanced GISTs with C min of more than 1,100 ng/ml [10]. Observed inter-patient pharmacokinetic variability may be due to patients' genetics. Polymorphisms in genes related to IM absorption, distribution, metabolism and excretion may aff ect the drug's bioavailability and consequently the response to the drug. Th e oral bioavailability of IM depends on its gastrointestinal absorption and on how much of it survives the extensive fi rst pass metabolism that it encounters. Approximately 95% of IM in the human body is bound to plasma proteins, *All patients were in chronic phase except for the study of Kim et al. [25], in which 203 patients were in chronic phase and 26 patients in accelerated phase or in blast crisis. † The base substitution and rs number taken from the National Centre for Biotechnology information SNP database [53] are indicated for polymorphisms positively associated with IM response. For the remaining SNPs for which no association was found, only the rs number is indicated. ‡ Functional impact in vivo: association had been found between polymorphisms and transcript levels of candidate genes or with trough IM concentration. § CyR, cytogenetic response; MR, molecular response; the cumulative incidence of major cytogenetic response (MCyR), CCyR, MMR, complete molecular response (CMR), loss of response, treatment failure and overall survival was measured. ¶ MMR in patients with CCyR. ¥ Polymorphisms with demonstrated functional eff ect in vitro.
mainly serum albumin and α1 acid glycoprotein (AGP). Removal of IM is mediated by the P glycoprotein (P-gp, also called ABCB1 or MDR1) and breast cancer resistance protein (BCRP), and uptake is mediated by human organic cation transporter 1 (hOCT1). IM is mostly metabolized by the cytochrome P450 (CYP) proteins CYP3A4 and CYP3A5 [11]. Various studies have analyzed the polymorphisms in genes encoding these proteins in relation to IM pharmacokinetics and response, as detailed below and in Table 1.

Pharmacokinetic determinants
Cancer cells have the ability to become resistant to multiple diff erent drugs known as the multi drug resistance (MDR) phenomenon. Th is is due, among other mechanisms, to increased transport of drugs out of the cell. Mahon et al. [12] showed that IM is a substrate of P-gp and could thus be aff ected by MDR. Th is was confi rmed by the facts that CML patients who did not achieve a CCyR had higher levels of expression of ABCB1 and that there was a correlation between BCR-ABL and ABCB1 transcript levels in these patients [13]. However, in GIST patients, the expression level of ABCB1 did not seem to impair the initial response to IM [14]. ABCB1 contains 28 exons and is located on chromo some 7q21.1. Th e silent C3435T polymorphism (rs1045642) in exon 26 of ABCB1 was the fi rst poly morphism to be described; it was reported to modify P-gp expression [15] ( Table 1). Decreased substrate specifi city, lower levels of intestinal P-gp and reduced levels of ABCB1 mRNA in 3435TT individuals were reported [15][16][17]. Consistent with this, increased bioavailability of some P-gp substrates has been observed to be associated with the 3435TT genotype in several studies, although confl icting reports have subsequently appeared showing higher or no correlation with ABCB1 expression or P-gp activity [18]. Th e C3435T polymorphism is in linkage disequilibrium with a silent C1236T polymorphism (rs1128503) in exon 12 and a non-synonymous substitution G2677T/A (rs2032582) in exon 21, which leads to an amino acid change of Ala893 to serine or threonine; this may also contribute to the observed functional eff ect of variation at the C3435T polymorphism or at associated haplotypes.
Gurney et al. [19] reported a higher IM clearance in CML and GIST patients with the TT genotype at each of three positions in the ABCB1 gene (positions 1236, 2677 and 3435). Reduction in dose as a result of toxicity also tended to be less commonly needed in these individuals [19]. We and our colleagues [20] found a higher frequency of MMR in CML patients carrying the 1236T allele and in patients with non-G genotypes at position 2677. However, we could not confi rm these results in a larger patient cohort [21]. Only in the subgroup of patients treated with both IM and cytarabine (also called AraC) was an association with the ABCB1 2677 polymorphism found, but, contrary to the initial fi nding [20], the presence of allele 2677G (GA or GG or GT) was associated with a higher rate of MMR [21]. Th is fi nding is surprising given that AraC is not a substrate of P-gp. Th e result is nevertheless similar to that reported by Deenik et al. [22], who observed a lower probability of cumulative MMR for CML patients who were homozygous for the ABCB1 alleles 3435T and 2677T, as well as for those having at least one 1236T allele. Finally, Ni et al. [23] reported a higher resistance rate in homozygous ABCB1 1236T patients and in those with at least one 3435T allele, whereas better CCyR was observed for patients with the AG/AT/AA genotype at position 2677. Th ree other studies conducted with CML patients did not fi nd an association between ABCB1 polymorphisms and MMR [24][25][26].
BCRP, which, like P-gp, causes removal of IM, is encoded by the ABCG2 gene located on chromosome 4q22. BCRP is expressed not only in hematopoietic stem cells but also in hepatic and intestinal epithelial cells, and this aff ects the bioavailability of IM [27]. Th e SNPs G34A (rs2231137), which substitutes a valine for a methionine, and C421A (rs2231142), which substitutes a glutamine for a lysine, are the most frequent non-synonymous polymorphisms in ABCG2 across various populations (Table 1) [28]. Experiments in vitro using various drug substrates showed that the Gln141Lys substitution aff ects the transport effi ciency of BCRP and may result in altered pharmacokinetics and drug-resistance profi les [28]. Human embryonic kidney cells transfected with the 141Lys variant of BCRP showed greater IM accumulation in vitro [29]. Consistent with this, a lower IM clearance and a higher dose-adjusted IM trough concentration were found in CML patients with the 141Lys variant [26,30], whereas the 141Gln variant reduced the response to IM [25]. In addition, the 34GG genotype was associated with a lower frequency of cytogenetic response [25], although the functional eff ect of this variant is less clear. In contrast to the reports on CML, pharmacokinetic IM parameters were not signifi cantly diff erent across genotypes of Gln141Lys polymorphism in patients with GISTs and KIT-positive soft tissue sarcoma [29].
Th e hOCT1 protein, encoded by the solute carrier family 22 gene (SLC22A1) on chromosome 6q26 [31], may also aff ect the response to IM by regulating its uptake. CML patients with low hOCT1 activity showed lower probability of achieving MMR [32], whereas those with high baseline level showed better survival and higher probability of achieving a cytogenetic or molecular response [24,33,34]. Given the important predictive role of SLC22A1 expression, some investigators have focused on gene variations that may aff ect hOCT1 function Dulucq and Krajinovic Genome Medicine 2010, 2:85 http://genomemedicine.com/content/2/11/85 [35,36]. For example, the Arg61Cys, Cys88Arg and Gly401Ser amino acid replacements resulting from the corresponding genetic polymorphisms were shown to aff ect the transport of hOCT1 substrates [30]. Bazeos et al. [37] identifi ed an association between Gly401Ser substi tution (G1201A, rs34130495) and the transcription level of the SLC22A1 gene (Table 1). In addition, CML patients with the 1201GA genotype had higher probability of achieving MMR than patients with the 1201GG genotype. Th ese results, however, need to be confi rmed in a larger cohort given the low frequency of the A allele. Indeed, White et al. [38] were not able to replicate this fi nding. Kim et al. [25] showed instead that the SLC22A1 C480G polymorphism (rs683369), which leads to Leu160Phe replacement, is associated with IM response in CML. Patients with the 480GG genotype showed a higher risk of treatment failure or loss of response than other genotypes. Finally, Takahashi et al. [26] showed that the SLC22A1 A1222G polymorphism (rs628031), which leads to Met408Val replacement, is associated with molecular response in the group of patients treated with IM doses lower than 400 mg. Patients with the 1222GG genotype showed a higher MMR rate.
Uptake of imatinib has also been described, in vitro, to be mediated to a modest extent by SLCO1B3, which encodes the organic anion-transporting polypeptide 1B3 [39], but Takahashi et al. [26] did not fi nd any association with IM response (Table 1).

Metabolizing enzymes
Among CYP3A5 polymorphisms, the CYP3A5*3 allele is particularly interesting, as it appears with a suffi ciently high frequency and has a clear functional role. Th e CYP3A5*3 allele is defi ned by the A6986G substitution (rs776746), which generates a cryptic splice site and the introduction of a premature stop codon [40]. Individuals who are homozygous for this allele have reduced CYP3A5 levels and reduced metabolic capacity. In the study of Kim et al. [25], the 6986AA genotype had an adverse impact on achievement of CCyR, whereas Takahashi et al. [26] did not fi nd an association between this allele and dose-adjusted IM trough concentration or clinical response (Table 1). In contrast, Sailaja et al. [41] found a higher frequency of the 6986GG genotype in CML patients with a minor or poor hematological response.

Binding proteins
Th e role of AGP in the mechanisms of IM resistance is not yet clear [42]. In GIST patients, an association was found between high plasma AGP levels and a lower clearance of IM and its metabolite [43]. Kim et al. [25] did not observe any association between AGP polymorphisms and IM response in CML patients.

Pharmacodynamic determinants
IM tyrosine kinase targets represent potential pharma codynamic determinants. Any modifi cation in these targets could modulate the effi cacy of IM and aff ect its mechanism of action.
Acquired point mutations in the tyrosine kinase domain of BCR-ABL are the most frequent mechanism of acquired resistance to IM in CML [44]. Th ese mutations should be distinguished from the polymorphisms in the ABL gene that could be responsible for primary resistance. However, their role is not yet clear. For example, the Lys247Arg amino acid replacement resulting from an adenine-to-guanine substitution does not seem to be functional [45]. Ernst et al. [46] found six diff erent polymorphisms in CML patients with failure of treatment or suboptimal IM response, but the clinical impact of these variations still needs to be investigated.
Approximately 95% of GIST patients express the receptor tyrosine kinase KIT and 86% of GISTs contain c-KIT activating mutations that lead to a ligandindependent activation of the tyrosine kinase. Th ese somatic mutations mostly occur in the juxtamembrane domain (encoded by exon 11) and extracellular domains (exon 9). Th e target kinase mutations in exon 11 are associated with a better overall partial response rate using standard Southwest oncology group response criteria [47].
Some GIST patients without c-KIT mutations show alterations in the juxtamembrane domain (encoded by exon 12) or activation loop domain (exon 18) of PDGFRA, with those with exon 18 mutations having a poor response to IM [47]. Resistance due to the somatic mutations in the tyrosine kinase domain of PDGFRA has also been described in a few cases of chronic eosinophilic leukemia [48]. IM-associated edema is believed to involve a disruption of PDGF signaling. Th e role of PDGFR polymorphisms in the risk of developing severe edema during IM treatment from CML was analyzed by Bruck et al. [49], but no signifi cant association was found.
Dressman et al. [50] analyzed the eff ect of 68 polymorphisms in 26 genes on the cytogenetic response to IM. Th ey found a signifi cant association between the rs2290573 polymorphism located in an intron of a putative tyrosine kinase gene, DKFZP434C131, and the major cytogenetic response (MCyR) in a subset of IMtreated patients. Patients homozygous for the C allele had a signifi cantly lower MCyR rate and a higher risk of disease progression than patients with other genotypes. It is unknown whether this polymorphism has a functional eff ect or whether it is a genetic marker in linkage disequilibrium with another polymorphism that is functional.
A substantial proportion of patients with IM resistance do not have BCR-ABL tyrosine kinase domain mutations, suggesting the involvement of additional mechanisms in IM resistance. Activation of other signaling pathways when IM blocks the BCR-ABL-mediated pathway might facilitate cell death avoidance in CML [51]. Recently, Kim et al. [51] analyzed a variety of polymorphisms in the genes of the apoptosis, angiogenesis, cell growth, Wilms tumor gene or interferon (IFN) signaling pathways in CML patients. An association obtained both in test and validation cohorts is particularly interesting. Th e CC genotype of the rs2069705 polymorphism in the IFNG (IFN-γ) gene was associated with a higher rate of molecular and cytogenetic response, suggesting a potential involvement of the IFN-γ signaling pathway in the mechanism of IM action in CML.

Clinical relevance of the pharmacogenetics of IM
Given that there is a large variability in response rate and IM systemic exposure following a standard drug dose, pharmacogenetic studies may provide insights into the role of genetic components in this variability. Focusing on a variety of genes whose products are essential for IM levels and action, these studies may identify potential pharmacokinetic and/or pharmacodynamic markers of IM response. Th ese markers, complementing existing ones such as drug plasma concentrations, could allow the prediction, for each individual, of a lack of effi cacy or excess toxicity, leading fi rst to pharmacogenetically guided prospective clinical trials and ultimately to personalized treatment. Pharmacogenetic IM studies have been conducted so far mainly in patients diagnosed with CML or GIST, probably because of the higher incidence of these two diseases. Th e effi cacy and toxicity of IM seem to depend on both IM pharmacokinetics, infl uenced by several enzymes and transporters, and IM pharmacodynamics, infl uenced by mutational status of the target. Several polymorphisms aff ecting the pharmaco kinetic determinants of IM have been identifi ed. Nevertheless, the data are not yet suffi ciently conclusive to translate into individual drug dose adjustment (several reasons for this are outlined below). Meanwhile, trough IM plasma levels could help physicians to defi ne the best IM dose [8]. In addition, the determination of hOCT1 activity before initiation of IM therapy may also be helpful [38].

Concluding remarks
Despite several groups attempting to demonstrate the impact of candidate gene polymorphisms, confl icting results remain. Th ese discrepancies could be explained, at least in some cases, by diff erent response criteria, study sample size, IM dosage and treatment protocols. Most studies have focused on ABCB1 polymorphisms, and constitutive or compensatory changes in expression of other ABC transporters, or IM-induced changes in ABCB1 expression, may confound the observed results in these studies. Th e results are not always supported by the expected functional eff ect of a given polymorphism, and they still require replication. It will be necessary to target other genotypes beyond those already analyzed to more comprehensively estimate the eff ect of the genes from the analysis of both individual polymorphisms and haplotypes.
It seems clear that the eff ect of IM depends on several genes. An approach involving multiple candidate genes may thus give the benefi t of including the potential eff ects of gene-gene interactions, but this has not been much explored and usually requires larger studies. Further studies are clearly needed to elucidate the real impact of candidate gene polymorphisms on the IM response and to what extent the use of second generation tyrosine kinase inhibitors (nilotinib and dasatinib) may eventually overcome the resistance imposed by certain genetic variations.