Real-Time PCR Gene Expression Profiling of Microarray Gene Signatures in Acute Leukemia.
Abstract Cancer subtype discovery and classification using microarray gene signatures has the potential to transform pathological diagnosis but measurement of indicator genes in routine practice remains difficult. We tested use of real-time PCR measurement of indicator genes for AML and ALL (Golub et al, Science, 1999) as a method for validation and application of microarray gene signatures. Mononuclear cells (MC) were isolated from whole bone marrow (BM) aspirates by density gradient centrifugation and sorted into unselected (total), CD34+ve and CD34-ve fractions. The mRNA in each fraction was globally amplified using a PolyA PCR method. We measured the expression profile of the 17 top ranked genes (cystatin C, leptin receptor, fumarylacetoacetate, CD33, HoxA9, adipsin, proteoglycan 1, LTC4 synthase, LYN, C-myb, MB-1, cyclin D3, SNF2, RbAp48, proteasome iota, HkrT-1 and E2A) from Golub et al (1999) by real-time PCR. All values were calibrated against control standards and normalized to the mean of three housekeeping genes (IF2-beta, GAPDH and human ribosomal protein S9). Data for all 17 genes were obtained for 4 (ALL), 26 (AML), 12 (AML remission) and 9 (morphologically normal) BM samples, each fractionated into three fractions (total MC, CD34+ve MC & CD34−ve MC). There was no significant difference in the mean of three housekeeping gene expression levels between the diagnostic groups. Comparison of the expression level of the other genes confirmed ability to separate AML and ALL, whilst the direction of expression change (increased or decreased) for each gene between AML and ALL was the same as found by Golub et al. In particular, c-myb showed largest significant increase in ALL vs AML in the total BM fraction, whilst cystain c was increased in AML in the CD34−ve fraction. hSNF2b was significantly increased in the ALL total B.M fraction and Hox-A9 was significantly increased in the AML CD34+ve B.M fraction. Furthermore expression level of LYN and CD33 was significantly increased in AML compared to remission AML, indicating ability of the method to determine activity status of disease. In addition, several of the genes provided better separation between AML and ALL when measured in the CD34+ve and −ve fractions indicating more prominent expression in cells of different maturity and that prior fractionation is diagnostically more informative. The results demonstrate ability of the method to validate gene expression signatures by an independent method, which is simple, sensitive and robust, allowing translation to routine clinical use. Whilst the present study used AML and ALL, in principle the method could be extended to any other tumor type for which gene signatures exist.