Class Prediction Models of Thrombocytosis Using Gene Profiling.

Human blood platelets contain unique and reproducible molecular signatures consisting of relatively small number of individual mRNA transcripts. Towards the goal of developing a molecular classification of platelet disorders, we have now developed and applied a unique microarray platform for the study of thrombocytosis. A 432-member gene list encompassing a subset of platelet- and leukocyte- restricted transcripts (with Arabidopsis probe elements as normalization controls) was used for the manufacturing of a customized platelet oligonucleotide (70-mer) gene chip (POC). In silico gene rank-intensity plots from sample RNAs [normal platelets (N=5); essential thrombocythemic platelets (ET; N=6); leukocytes (N=3)] hybridized to the Affymetrix HU133a gene chip validated the discriminatory capacity of the POC, which was then applied to a broader patient cohort. Gel-filtered platelets from normal controls (N=28), patients with ET (N=18), or patients with reactive thrombocytosis (RT; N=17) were used for RNA isolation, amplification, and Cy3:Cy5 co-hybridization using human universal RNA as reference. Genotyping for the presence of the JAK2 V617F mutation was established by pyrosequencing analysis using both genomic (leukocyte) DNA and platelet-derived mRNA. Initial two-way non-parametric ANOVA identified a 131-member gene list of differentially-expressed transcripts among the three subgroups (p<0.05), with no gender differences in healthy controls. Stepwise discriminant analysis delineated a 14-gene subset that appropriately classified 94% of original cases. Leave-one-out cross-validation using 3 of the 14 markers correctly classified 94% ET and 100% RT patients, independent of the presence of the JAK2 V617F mutation. Leave-one-out cross-validation using 6 of the 14 markers also correctly subclassified 88% ET patients with JAK2 V617F mutation (N = 8) and 90% of ET patients without such mutation (N = 10). These data (1) provide the first evidence for the existence of discriminatory molecular signatures among etiologically-diverse causes of thrombocytosis, (2) demonstrate the presence of distinct ET transcriptomic profiles substratified by JAK2 V617F genotype, and (3) provide the foundation for development of a clinically-useful classification scheme of genetic platelet disorders.