Identification of Genes That Potentially Allow Prediction of Interferon Response in Chronic Phase CML Diagnostic Samples through Gene Expression Analysis.

Interferon (INF) was the first drug to significantly prolong survival in chronic myeloid (CML) and has been extensively used to treat CML patients over the past years. Today, in the Imatinib era INF still has a role as additional agent in patients with a suboptimal response to Imatinib. However, not all patients will respond to INF and the reasons for this heterogeneity are largely unknown. The aim of the study was to identify biomarkers that allow prediction of future response to INF in CML diagnostic samples. Microarray-based gene expression analysis enables the study of thousands of genes in one single experiment and has been shown to be a useful tool to identify biomarkers that can assign samples to specific categories. Gene expression profiles in diagnostic samples from 15 CML chronic phase patients were analyzed using cDNA microarrays with 7458 genes. Blood samples were collected at diagnosis and all patients were treated with alpha-interferon. Seven patients were responders, defined as patients achieving a complete or major cytogenetic response at 12 months or earlier after initiation of interferon. Eight patients that had no cytogenetic response at 6–18 months were regarded as non responders. A list of 61 genes differentially expressed in both sample groups was generated by sorting those genes that showed the largest difference in median expression between the groups. The top 20 genes were subsequently used in an “all pair” selection procedure to identify those genes that are highly predictive of response to INF. The genes selected by this procedure were LTF, PRG2, JARID1A, NRGN, RNASE2 and DEFA4. The accuracy of these genes in predicting response to INF was determined in a “leave one out cross validation” procedure and the estimated error rate was found to be 0,13. The expression level of these six genes was verified with real-time PCR. The data generated by real-time PCR were used in a principal component analysis that separated the samples in two clusters.We conclude that it might be possible to use microarray-based gene expression analysis to predict future treatment outcome in CML diagnostic samples.