The IFNG (IFN-γ) Genotype Predicts Cytogenetic and Molecular Response to Imatinib Therapy in Chronic Myeloid Leukemia

Imatinib is a selective tyrosine kinase inhibitor, particularly against BCR/ABL fusion tyrosine kinase, and provides therapeutic benefit to patients with chronic myeloid leukemia (CML). However, interindividual variability of response to imatinib exists (1). For investigation of interindividual variation in drug response or resistance to imatinib, an approach using single nucleotide polymorphisms (SNP) is particularly worthwhile. In the present study, we simultaneously examined multiple candidate gene SNPs in terms of their association with imatinib response and development of resistance in patients with CML.

The Abelson tyrosine kinase domain mutation is an important mechanism in imatinib resistance. However, a substantial proportion of patients with imatinib resistance do not show evidence of tyrosine kinase domain mutation. Accordingly, other mechanisms aside from Abelson tyrosine kinase domain mutation should be explored intensively. Once imatinib binds to the bcr/abl oncoprotein, it then inactivates signal transduction, leading to cell death or apoptosis. Findings from recent studies have revealed that alternative pathways, such as the src kinase–mediated pathway, are switched on when imatinib blocks the bcr/abl-mediated pathway; thus, CML cells escape cell death (2). Therefore, SNPs in the apoptosis pathway need to be evaluated in terms of the response and resistance to imatinib therapy in CML.

Besides the apoptosis pathway, other potential candidate pathways include angiogenesis pathways, cell growth pathways, WT1 (Wilms tumor gene) pathways, drug transport/metabolism pathways, or IFN signaling pathways. In the current study, we included several SNPs in these pathways with the following evidence: (a) resistance to apoptosis is regulated by the expression levels of bcr/abl fusion tyrosine kinase (3, 4); (b) increasing angiogenesis and higher vascular endothelial growth factor (VEGF) production was noted in CML patients (5–7); (c) CML that are characterized by clonal proliferation and growth factor–independent myeloid cell growth such as CSF2 (G-CSF; ref. 8), CSF3 (GM-CSF; refs. 9, 10), FLT3 (11), JAK3 (12, 13), and IL1 (14); (d) the WT1 gene may be involved in leukemogenesis of CML (15); (e) ABCB1 (multidrug resistance-1; refs. 16–18) and ABCG2 (breast cancer resistance protein) is highly expressed in CD34⁺ progenitor cells in CML patients (19), and α1 acid glycoprotein levels could reflect pharmacologic resistance to imatinib in patients with CML during blastic phase (20); and (f) IFN affects the proliferation of hematopoietic stem cells (HSC) and affects myeloid colony formation in both murine and human models (21–24).

The SNP of certain genes might affect the promoter expression of the gene, thus controlling mRNA transcription speed, or result in amino acid residue change, thus switching the functions of the corresponding protein. These SNPs could result in interindividual variability in the gene expression of certain genes. Accordingly, investigations of SNPs on the multiple candidate pathways might be helpful to discover certain SNP markers predicting treatment outcomes of imatinib therapy in CML. In the present study, 79 SNPs in multiple candidate pathways, including 6 major pathways (i.e., apoptosis, angiogenesis, myeloid cell growth, multidrug resistance, WT1 signaling pathways, and IFN signaling pathways) were evaluated in a discovery cohort (n = 229), and significant SNPs were validated in an independent validation cohort (n = 187).

The discovery cohort included 229 CML patients who began imatinib therapy between August 2000 and December 2006 at the Princess Margaret Hospital, Toronto, ON, Canada. The validation cohort included 187 CML patients who started imatinib therapy between March 2002 and December 2008 at the Samsung Medical Center, Seoul, Korea or at the Chonnam National University Hwasun Hospital, Hwasun, Korea. Patients started imatinib at doses of 400, 600, or 800 mg, depending on their disease status. Before or during imatinib therapy, blood samples were collected after informed consent was obtained from patients in accordance with the Declaration of Helsinki. The current study was approved by the Research Ethics Board of the University Health Network, University of Toronto, ON, Canada.

Candidate genotypes were selected through review of the literature and by selection of SNPs in nonsynonymous SNPs in exon regions with a minor allele frequency of >0.05 (Table 1). If the frequency was not available, it was reported from the Entrez SNP site (http://www.ncbi.nlm.nih.gov/sites/entrez).

First, genotyping was undertaken with discovery cohort samples using the Sequenom iPLEX platform (http://www.sequenom.com/; Sequenom, Inc.). DNA was extracted using the Puregene DNA Purification Kit (Gentra Systems, Inc.). Detection of SNPs was performed by analysis of primer extension products generated from previously amplified genomic DNA using a Sequenom chip-based matrix-assisted laser desorption/ionization time-of-flight (MALDI-TOF) mass spectrometry platform. The details of the primers are described in Supplementary Table S2. Ninety six-well plates containing 2.5 ng of DNA in each well were amplified by PCR following the specifications of Sequenom. Unincorporated nucleotides in the PCR product were deactivated using shrimp alkaline phosphatase. Allele discrimination reactions were conducted by the addition of extension primer(s), DNA polymerase, and a cocktail mixture of deoxynucleotide triphosphates and dideoxynucleotide triphosphates to each well. MassExtend clean resin (Sequenom) was added to the mixture for removal of extraneous salts that might interfere with MALDI-TOF analysis. 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 subsequently read by MALDI-TOF mass spectrometer. Duplicate samples and negative controls were included for evaluation of genotyping quality.

Following primary statistical analyses, significant genotypes identified as having an association with clinical outcome in the discovery cohort were processed for internal and external validation. For external validation, 26 of the significant and neighboring genotypes were genotyped in the validation cohort (n = 187) using the Sequenom iPLEX platform. Genotyping was processed at the Analytical Genetics Technology Centre, University Health Network, Toronto, Ontario, Canada (for discovery), and at Bioneer Inc., Chungwon, Korea (for external validation).

Prior to the commencement of imatinib therapy, complete blood count and standard baseline biochemistry tests were performed with bone marrow evaluation for morphology, conventional cytogenetic analysis, and BCR/ABL mRNA reverse transcription-PCR. Cytogenetic analysis was performed by the G-banding technique. Patients were monitored regularly on an outpatient basis as follows: biweekly blood counts and biochemistry were obtained during the first month of imatinib therapy, and then monthly until a cytogenetic response was achieved, and then every 3 months thereafter. Until complete cytogenetic response (CCR) was confirmed, bone marrow evaluation and/or fluorescence in situ hybridization studies were performed every 3 months. Using quantitative BCR/ABL mRNA PCR, quantification of peripheral blood BCR/ABL fusion gene transcripts was repeated every 3 to 4 months regardless of cytogenetic response.

Peripheral blood samples (5 mL) were also analyzed using quantitative PCR for determination of levels of BCR/ABL fusion gene transcripts, according to the instructions of the manufacturer (ABI 9700 Thermal Cycler; Applied Biosystems) and following recommendations established for the standardization of this procedure at the international level (25–27). BCR/ABL transcript levels were measured and presented using the international scale. Nested PCR techniques were used for confirmation of results in selected samples with undetectable BCR/ABL transcript levels. Abl tyrosine kinase domain mutations were screened in any patient in an advanced phase of disease. For patients with the chronic phase (CP) of the disease beginning treatment with imatinib, mutation screening was indicated if response was not acceptable, or if any sign of loss of response (LOR) or progression was observed.

Response criteria were the same as previously defined in studies using imatinib (1, 28, 29). Briefly, cytogenetic responses were categorized as complete (CCR; 0% Ph⁺ cells in marrow by conventional cytogenetics or fluorescence in situ hybridization), partial (1-34% Ph⁺ cells in marrow), or minor (35-65% Ph⁺ cells in marrow). A major cytogenetic response (MCR) was defined as the sum of CCR and partial cytogenetic response. Major molecular response (MMR) was defined as <0.1% of the BCR/ABL fusion gene transcript level on an international scale by quantitative PCR.

Time to treatment failure (TF) was defined as the interval between initiation of imatinib therapy and the occurrence of events, i.e., imatinib failure, including primary and secondary resistance, such as LOR. Time to LOR was defined as the interval between the date of any confirmed response and the date at which criteria for responses were no longer being met including (a) transformation from CP to accelerated phase (AP) or blastic crisis (BC); (b) loss of CCR/MCR, and (c) development of the Abelson tyrosine kinase domain mutation.

Time to transformation-free survival was defined as the interval between initiation of imatinib therapy and confirmation of progression to AP or BC, or death from any cause, whereas overall survival was calculated from the initiation of imatinib therapy until the date of death from any cause or the date of the last follow-up.

The SNPs genotyped in the discovery cohort were primarily evaluated for adequacy of Hardy-Weinberg equilibrium using χ² test. Genotype errors and genotype frequencies were summarized using Haploview version 3.32 (Broad Institute, Cambridge, MA; available at http://www.broad.mit.edu/mpg/haploview). The SNP ABCB1 (rs2032582) had more than two alleles and was coded in two ways: (a) GG, G/−, and −/−; and (b) T/T, T/−, and −/−.

Cumulative incidences of MCR, CCR, and MMR were calculated with consideration for discontinuance of imatinib as competing risks for interest events. Probability of freedom from LOR and from TF was estimated and plotted using the Kaplan-Meier method. Probabilities of overall and transformation-free survival were also estimated using the Kaplan-Meier method. In univariate analyses, according to the 79 genotypes in genotype models, MCR, CCR, and MMR were compared using Gray test, and LOR, TF, transformation-free survival, and overall survival were compared using log rank test.

For validation of the genetic effect, we performed internal validation procedures using a bootstrap algorithm, and constructed the bootstrap confidence interval (CI). We applied bootstrap based on 500 replications. Results were obtained using the PROC SURVEYSELECT procedure in SAS version 9.1, and presented as the bootstrap hazard ratio (HR) CIs of the genetic effects adjusted by clinical factors. External validation was also performed using an independent validation cohort. Treatment outcomes were also compared according to significant genotypes using the log rank test, and adjusted by Bonferroni correction. Procedures were repeated in additive, dominant, and recessive models with and without Bonferroni's correction.

Next, multivariate analysis was performed using variables that were significantly associated with CCR and MMR, including the disease stage (CP versus AP/BC), the presence of additional cytogenetic abnormality, and IFNG genotypes (rs2068705; CC versus CT/TT genotype). Multivariate analyses using Cox proportional hazard models were conducted using backward stepwise modeling and a P value for the likelihood ratio test of >0.05. HRs and 95% CIs were also estimated.

All statistical tests were two-sided, with the significance level set as 0.05, unless otherwise stated. Statistical data were obtained using an SPSS software package (SPSS 13.0, Inc.) and SAS version 9.1 (SAS Institute). Incidence curves were obtained using R package, version 2.4.1 (available at http://cran.r-project.org/).

Out of 229 patients, 203 patients (89%) were in CP, 23 patients (10%) were in AP, and 3 patients (1%) were in BC in the discovery cohort, whereas 163 patients (90%) were in CP, 11 patients (6%) were in AP, and 7 patients (4%) were in BC in the validation cohort. Demographic and disease characteristics of the discovery and validation cohorts are described in Table 2, including age, gender, race, previous treatment prior to imatinib, cytogenetics, and dosage of imatinib. Except ethnicity and treatment prior to imatinib, two cohorts showed similar characteristics of disease.

In the discovery cohort, with a median duration of imatinib administration of 40.8 months (range, 1-86), the cumulative incidence of HR was 96% (95% CI, 92-99%) at 3 months. The cumulative incidence of MCR and CCR was 86% (81-91%) and 62% (56-69%) at 12 months after initiation of imatinib therapy, respectively. In the case of MMR, it was 33% (27-40%) and 52% (45-59%) at 1 and 2 years, respectively. Forty-six cases (27%) of TF were documented due to either resistance (n = 38) or intolerance (n = 8). Probability of freedom from LOR was 71% (63-78%) at 2 years, and 60% (51-69%) at 3 years after achievement of any response to imatinib therapy. Probability of freedom from TF was 69% (62-76%) and 58% (51-65%) at 2 and 3 years after initiation of imatinib therapy, respectively. The 5-year probability of transformation-free survival and overall survival was 90% (87-94%) and 95% (92-99%), respectively. As shown in Supplementary Table S2, treatment outcomes in the validation cohort were similar to those in the discovery cohort.

Detailed information and frequency of candidate SNPs are summarized in Table 1. Among them, five were excluded from the analysis because they were monomorphic in the current population: CASP9 (rs4645981), CYP3A5 (rs28383469), ORM (rs1126724), ORM (rs3182034), and BCL6 (rs11545363).

All results of univariate analyses are presented in Table 1. With respect to the probability of achieving CCR, the following SNPs showed significant correlation; IFNG (rs1861494 and rs2069705; Fig. 1A), FASL (rs763110), FAS (rs2234767), and VEGFR2 (rs1531289). When we divided IFNG (rs2069705) into CC versus CT/TT, the median time to CCR was 183 ± 36 days in the CC genotype group versus 273 ± 37 days in the CT/TT genotype (P = 0.003).

The probability of achieving MMR was significantly associated with IFNG (rs1861494 and rs2069705; Fig. 1B), BIRC5 (rs9904341), and FAS (rs2234978). When comparing MMR between groups with the CC genotype and the CT/TT genotype of IFNG (rs2069705), significant differences were also observed in favor of the CC genotype (median time to MMR; 358 ± 76 d in the CC genotype versus 774 ± 83 d in the CT/TT genotype; P = 0.002).

In terms of LOR, IFNG (rs2069705), IFNGR2 (rs9808753), BIRC5 (rs9904341) and ORM (rs3182041) showed significant correlation, whereas TF was strongly associated with IFNG (rs2069705), JAK3 (rs3212713), and ORM (rs3182041).

For confirmation of these results, we performed an internal validation. Results were presented with the bootstrap HR CIs of the genetic effects, as shown in Table 3. IFNG (rs1861494 and rs2069705), CASP9 (rs4645981), CASP8 (rs3834129), FASL (rs763110), and FAS (rs2234767) were useful for the prediction of achievement of CCR, and IFNG (rs2069705), FAS (rs2234767), and JAK3 (rs3212713) were helpful for MMR, whereas IFNG (rs2069705) and BRIC5 (rs9904341) were useful for LOR and TF. Accordingly, our results from univariate analyses were internally validated.

External validation was repeated for confirmation of these results using an independent external cohort derived from CML patients in Korea. Interestingly, among various SNPs with significance in univariate analyses and internal validation, correlation of the IFNG genotype with CCR and MMR was clearly shown (Table 4), despite of the different frequencies between two cohorts. As shown in Fig. 1 and Table 4, external validation demonstrated that the IFNG genotype (rs2069705) was predictive of the probability of CCR (HR, 2.17; P < 0.0001; P_corr = 0.002; Fig. 1C) or MMR (HR, 1.96; P = 0.0001; P_corr = 0.005; Fig. 1C), suggesting that the CC IFNG genotype (rs2069705) has an approximately 50% higher chance of achieving CCR or MMR compared with the CT/TT genotype.

To remove the effect of disease status on treatment outcomes, the analysis was repeated in patients confined to those in CP. We compared the rate of CCR at 12 months and of MMR at 18 months according to the IFNG genotype (rs2069705) in overall patients and in a subgroup of CML patients in CP (Fig. 2). Significantly higher response rate was noted in favor of the group with the CC genotype both in overall patients (Fig. 2A and B) and in CP patients only (Fig. 2C and D), compared with those with CT or TT genotype. In CP patients, patients with CC genotype showed 76.0% MMR rate at 18 months whereas those with the CT or TT genotype only showed 35.7% and 52.5% MMR rate at 18 months (P = 0.003; Fig. 2D).

Multivariate analysis confirmed that the CC IFNG genotype (rs2068705) and CP were associated with higher CCR and MMR. With respect to CCR, the patients with CC IFNG genotypes showed a higher probability of achieving CCR compared with the CT/TT genotype (P = 0.005; HR, 1.727; 1.183-2.519), and those in CP showed higher CCR compared with those in advanced disease stage (P = 0.023; HR, 2.024; 1.104-3.704). With regard to MMR, the group with CC IFNG genotypes showed significantly better MMR than those with the CT/TT genotype (P = 0.002; HR, 1.912; 1.258-2.915).

The current study suggested that (a) a multiple candidate pathway-based SNP approach could identify several potential predictive markers of response to imatinib therapy; (b) among them, the IFNG genotype was found to be the significant predictive marker for CCR and MMR, which was confirmed not just by internal validation, but also by external validation using independent external cohort.

Because of the successful introduction of imatinib into CML management, a dramatic improvement in response to imatinib therapy and excellent prognosis have been achieved in CML patients. However, complete eradication of the BCR/ABL-producing clone is obtained in only one quarter of the patients (30). In addition, if imatinib therapy is interrupted after achievement of complete molecular response, a BCR/ABL transcript will reappear and eventually rise up to the diagnostic level in one half of the patients (31). These interindividual variations of the response to imatinib are a practical issue in our daily practice of CML. Thus far, none of the predictive surrogate markers have been successfully applied for the prediction of imatinib response before starting imatinib therapy.

Although the exact mechanism of imatinib resistance is not fully understood, it has been explained with several proposed mechanisms, such as the acquisition of point mutations in the Abl kinase domain, BCR-ABL amplification at the genomic or transcript level (32), overexpression of other tyrosine kinases (i.e., SRC kinase family; ref. 33), variability in the amount and function of the drug influx protein OCT-1, or resistance at the level of “leukemic stem cells” (LSC; ref. 34). Because they are in a quiescent state, LSCs are known to be resistant and insensitive to tyrosine kinase inhibitors, and thus, do not require BCR-ABL signaling for survival (35). Thus, imatinib resistance mechanisms cannot be explained in a simple and singular way, but will be more complex than we speculate. Based on the complexity of the imatinib resistance mechanism, multiple candidate gene SNPs should be examined for their association with imatinib response and resistance.

Of the 79 candidate genotypes in the current study, the IFNG genotype was found to be an excellent predictive marker of response to imatinib. The IFNG genotype (rs2069705) was consistently confirmed as predictive of achievement of CCR and MMR following imatinib therapy by internal and external validation processes both in Canadian and Korean patients with CML. As shown in Fig. 2, not just in the overall population but also in CP patients, the IFNG genotype (rs2069705) was a very predictive marker for CCR at 12 months and MMR at 18 months following imatinib therapy for CML.

Therefore, what is the role of IFN signaling in imatinib therapy for CML? Prior to the introduction of imatinib, IFN-α was widely used as the standard drug for CML patients who were not available for allogeneic transplantation. The exact mechanism of action of IFN-α in the treatment of CML patients has not been fully revealed. A recent study suggested that IFN signaling activates dormant HSCs in vivo. Priming of HSCs with IFN-α or chronic IFN-α stimulation followed by 5-fluorouracil treatment was shown for the induction of HSC loss (36). Similar to the action of IFN-α, findings from several investigations have suggested that IFN-γ could promote hematopoiesis and activate the expansion and proliferation of HSCs. IFN-γ promotes ex vivo expansion of the earliest CD34+ hematopoietic precursors (37), and thus, stimulates the early stage of myelopoiesis in the presence of certain growth factors (38). In addition, it induces in vivo and in vitro expansion of Lin^negSca1⁺c-kit⁺ progenitor cells (39). Kurz and colleagues reported that IFN-γ stimulates CD34+ cells, which could induce tryptophan degradation and neopterin formation (40). This might explain the induced proliferation of HSCs and LSCs by IFN-γ, which results in sensitization of CML LSCs to imatinib, as myeloid cytokines (i.e., granulocyte colony-stimulating factor or granulocyte-macrophage colony-stimulating factor) decrease quiescent CML LSCs and promote their elimination by imatinib (41, 42).

The IFNG gene is located on chromosome 12. The SNP of rs2069705 is a regulatory sequence affecting IFN-γ expression as a promoter. In the present study, the level of mRNA or protein of IFN-γ, according to the IFNG genotype (rs2069705), was not assessed. Thus, further studies will be needed to evaluate the implications of the CC/CT/TT genotype on the expression level of IFN-γ, although other studies suggested a strong association between the IFNG genotype (rs2069705) and a susceptibility to systemic lupus erythematosus (43), which implies a strong association between the IFNG genotype and the IFN-γ expression levels.

Not just in imatinib therapy, but also in nilotinib therapy, IFN signaling can be involved in the mechanism of action of TKI therapy. A previous study revealed that nilotinib inhibits the proliferation and function of CD8+ T-lymphocytes through T-cell receptor signaling, thus resulting in a decreased release of IFN-γ (44). In addition, the inhibitory effect of nilotinib was reported to be two times stronger compared with imatinib. Accordingly, in future studies, it might be interesting to look at the effect of IFNG SNPs in nilotinib therapy for CML.

In conclusion, findings from the current study suggest that (a) a multiple candidate pathway-based SNP approach can identify several potential predictive markers of response to imatinib therapy; (b) among them, the IFNG genotype was found to be the most significant SNP marker for the prediction of CCR and MMR, which was confirmed not just by internal validation, but also by external validation in the independent external cohort. This finding proposes the potential involvement of the IFN signaling pathway in the mechanism of action of imatinib, particularly in CML LSCs. Further detailed genetic and functional studies of IFNG genes will be helpful to reach a clear conclusion on the role of the IFNG gene in the mechanism of action of imatinib therapy in CML.

No potential conflicts of interest were disclosed.

Grant Support: Korea Healthcare Technology R&D Project, Ministry of Health Welfare & Family Affairs, Republic of Korea (A070001), the Korea Health 21 R&D Project, Ministry of Health & Welfare, Korea (01-PJ10-PG6-01GN16-0005), and by the Friends to Life Fund, Princess Margaret Hospital Foundation, Toronto, ON, Canada.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.