A polymorphism associated with STAT3 expression and response of chronic myeloid leukemia to interferon α

Abstract

Interferon α (IFN) induces variable responses in chronic myeloid leukemia (CML), with 8–30% of early chronic phase cases achieving a complete cytogenetic response. We hypothesized that polymorphic differences in genes encoding IFN signal transduction components might account for different patient responses. We studied 174 IFN-treated patients, of whom 79 achieved less than 35% Philadelphia-chromosome (Ph) positive metaphases (responders) and 95 failed to show any cytogenetic response (more than 95% Ph-positive metaphases; non-responders). We compared 17 single nucleotide polymorphisms (SNPs) at IFNAR1, IFNAR2, JAK1, TYK2, STAT1, STAT3 and STAT5a/b between the two groups and found a significant difference for rs6503691, a SNP tightly linked to STAT5a, STAT5b and STAT3 (minor allele frequency 0.16 for non-responders; 0.06 for responders, P=0.007). Levels of STAT3 mRNA correlated with rs6503691 genotype (P<0.001) as assessed by real time quantitative PCR and therefore we conclude that rs6503691 is associated with the STAT3 expression levels and response of CML patients to IFN.

Introduction

Interferon α (IFN) induces heterogeneous responses in chronic myeloid leukemia (CML), with up to 80% of early chronic phase patients achieving hematologic remission but only 8–30% achieving complete cytogenetic remission.1–6 Although response correlates with Hasford and Sokal risk scores7,8 and may be influenced by other factors such as the presence or absence of deletions at the reciprocal ABL/BCR junction on the 9q+ chromosome,9 the molecular basis for heterogeneous responses, and indeed more broadly the mechanism of response to IFN, remains poorly understood.

The type 1 IFN receptor is heterodimeric in structure, with the two subunits encoded by the genes IFNAR1 and IFNAR2. Binding of IFN to the receptor induces activation of the JAK1 and TYK2 non-receptor tyrosine kinases which then phosphorylate STAT proteins.10 Phosphorylated STAT dimers migrate to the nucleus where they activate the transcription of target genes. Inherited single nucleotide polymorphisms (SNPs) in genes encoding components of the IFN signal transduction cascade have been associated with diseases such as systemic lupus erythematosus, athsma and Crohn’s disease.11–13 We hypothesized that polymorphic difference in this cascade might account for the different responses of CML patients to IFN.

Design and Methods

Patient samples

Initially we studied 174 pre-treatment genomic DNA (gDNA) samples from BCR-ABL-positive CML patients receiving IFN as part of the German CML studies I-III.3,14,15 Two patient groups were selected on the basis of availability of DNA and maximal response to therapy: responders (n=79) achieved a major (≤35% Philadelphia chromosome-positive metaphases) or complete cytogenetic response (0% Philadelphia chromosome-positive metaphases) whereas the non-responders (n=95) failed to show any cytogenetic response (>95% Philadelphia chromosome-positive metaphases) after a median of 38 and 22 months treatment, respectively, after initiation of treatment. Samples from an additional 245 pre-treatment CML cases for whom both DNA and cDNA were available were used to compare STAT expression levels with genotype. The study was approved by the Internal Review Boards from participating institutions and informed consent was provided according to the Declaration of Helsinki.

SNP genotyping by pyrosequencing

We studied 17 single nucleotide polymorphisms (SNPs) that were within or close to the genes encoding IFNAR1, IFNAR2, JAK1, TYK2, STAT1, STAT3 and STAT5a/b. SNPs were selected on the basis of published data indicating positive associations with one or more human diseases, or as tagged SNPs with minor allele frequencies (maf) >0.2 from the International HapMap Project (release 21; www.hapmap.org). We did not include STAT2 and STAT4 in the analysis as there have not, to our knowledge, been any reports implicating these proteins in the pathogenesis of myeloid disorders. Furthermore, because of the limited number of cases available for analysis we deliberately did not attempt to capture all genetic variation at these loci due to the loss of statistical power this would entail. Pyrosequencing was performed as described16 using primers and dispensation orders as shown in Online Supplementary Tables S1A and B. Markers were quantified using the Allele Frequency Quantification function in the SNP Software (Biotage AB, Uppsala, Sweden) and called as homozygous when one allele gave a reading of >90% and heterozygous when both alleles were called as 40–60%.

Expression analysis

Reverse transcriptase real-time PCR (RQ-PCR) was performed to quantify STAT3, STAT5a and STAT5b expression relative to GUSB expression and as an internal control for cDNA quality and quantity. Complementary DNA synthesis was performed by standard procedures and GUSB quantification was performed.17 STAT3 expression was determined by using the custom designed PerfectProbe Gene Detection Kit (PrimerDesign, Southampton, UK) (sense primer: 5′-GAAGGAGGCGTCACTTTCAC-3′; antisense primer: 5′-CTGCTGCTTTGTGTATGGTTC-3′; probe 5′FAM-CTCTTACCGCTGATGTCCTTCTCCACCCAGGTAAGAG-DABCYL3′). STAT5a and STAT5b expression was determined using the inventoried TaqMan Gene Expression Assay by Applied Biosystems (Foster City, CA, USA). PCRs were performed on the Corbett Rotor-Gene 6000 (Corbett Life Science, Cambridge, UK). After demonstrating equal amplification efficiencies for each target, samples were tested in triplicate and mean STAT levels were normalized to GUSB and compared using the 2 method.18

Statistical analysis

To investigate the distribution of baseline values between groups, univariate tests were performed by using the Mann-Whitney, Fisher’s exact or χ tests, as appropriate. The possible independent influence of rs6503691 was assessed by multiple Cox regression analysis using SAS version 9.1.3 (SAS Institute Inc., Cary, NC, USA). Real time PCR results were compared to genotype by Kruskal-Wallis analysis.

Table 1A.Summary of genotyping analysis.

Results and Discussion

Initially we genotyped 12 SNPs and compared the allele frequencies between responders and non-responders. As shown in Table 1A, only one SNP (rs6503691) in exon 1 of STAT5b showed a significant difference with a maf of 0.16 for non-responders versus 0.06 for responders (P=0.0066, odds ratio 0.36, 95% confidence intervals 0.17–0.76). Typing of an additional 5 SNPs in the same genomic region (rs6503695, rs16967611, rs9900213, rs17500235, rs17591972) failed to reveal any other significant associations (Table 1A). It is notable that this SNP has been recently reported to be associated with the risk of developing breast cancer.19 We evaluated the impact of rs6503691 in more detail by taking other prognostic factors into account. On univariate analysis, the leukocyte count, percentage blasts, spleen size, Sokal score and rs6503691 genotype were all significantly associated with response (Table 1B). On multivariate analysis, however, rs6503691 genotype fell marginally below the level of significance (P=0.056; Table 1C).

Inspection of the HapMap data shows that rs6503691 falls in a region of strong linkage disequilibrium at 17q21 that includes the entire STAT5A gene as well as the 5′ end of STAT5B and the 3′ end of STAT3 (Figure 1A). Potentially then, this SNP could be linked to other variants that might influence the expression of any of these three genes. We therefore compared rs6503691 genotype with STAT5A, STAT5B and STAT3 mRNA levels in 245 pre-treatment CML cases. As shown in Figure 1B, STAT3 expression was strongly related to rs6503691 genotype (P<0.0001) but no differences were seen for STAT5A or STAT5B. Strikingly, a nearby polymorphism has recently been reported to be linked to both STAT3 mRNA levels and the response of metastatic renal cell carcinoma to IFN.20 BCR-ABL is known to activate STAT321 and elevated expression of SOCS3, a known STAT3 target, confers IFN resistance to CML cells.22 Taken together, these results indicate that polymorphic differences in STAT3 expression levels may be a determinant of response to IFN in CML, and that the marginal lack of significance on multivariate analysis may have been due, at least in part, to limited sample numbers.

Table 1B.Clinical characteristics and univariate analysis of responders and non-responders.

Table 1C.Multiple logistic regression analysis.

In the past decade, treatment of patients with CML has been transformed by the introduction of imatinib and other second generation tyrosine kinase inhibitors (TKIs). Nevertheless, IFN still remains relevant and its use as part of combination therapy with TKIs has attracted considerable interest, supported by favorable early clinical results and the recent demonstration that IFN activates dormant hemopoietic stem cells in vivo.23 Furthermore, discontinuation of imatinib in cases that have achieved complete molecular remission (CMR) does not always lead to relapse, and it has been suggested that sustained CMR may be influenced by prior treatment with IFN.24 We suggest that the impact of rs6503691 should be evaluated in these novel settings.

Figure 1.(A) HapMap data showing the position of rs6503961 in relation to STAT5A, STAT5B and STAT3, plus linkage disequilibrium in this region. (B) STAT3 but not STAT5A or STAT5B expression normalized to GUSB is significantly related to rs6503691 genotype in 245 patients with CML (P=<0.0001; Kruskall-Wallis test). n.s = not significant.

Acknowledgments

we are grateful to all those who contributed to the sample and data collection at the CML trial office in Mannheim, Germany.

Footnotes

• Funding: the study was supported by Deutsche Krebshilfe, Leukaemia Research (UK), the Wessex Cancer Trust, the Lady Tata Memorial Trust, the Competence Network ‘Acute and chronic leukemias’, sponsored by the German Bundesministerium für Bildung und Forschung (Projektträger Gesundheitsforschung; DLR e.V.- 01 GI9980/6), the German José-Carreras-Leukämiestiftung (H03/01) and the European LeukemiaNet within the 6th European Community Framework Programme for Research and Technological Development.
• The online version of this article has a supplementary appendix.
• Authorship and Disclosures SK, AH and NC designed the study. SK, KW, TE, AC, HW designed and performed the laboratory analysis. RH, AR and AH provided samples and clinical data. SK and NC analyzed the data. SK and NC wrote the paper, and all authors contributed to the final version.
• The authors reported no potential conflicts of interest.

• Received May 16, 2009.
• Revision received June 23, 2009.
• Accepted June 26, 2009.