Nonuniform hybridization: a potential source of error in oligonucleotide-chip experiments with low amounts of starting material.

Low amounts of starting material are a significant limitation of gene-expression profiling of microprepared pathologic specimens. Linear RNA amplification has become the method of choice to overcome this problem. Thus, transcriptomal analyses by oligonucleotide-chips or cDNA microarrays are now feasible with labeled complementary RNA generated from total RNA samples in the lower nanogram range. However, in case of oligonucleotide-chips, it has been underestimated so far that individual complementary RNA molecules are shorter in length than and display a 3' bias in comparison to the sequence stretch represented by oligonucleotides on the chip. This can lead to incorrect interpretation of raw data. We have analyzed this problem testing ex vivo-microprepared endothelial cells with Affymetrix GeneChips U133A. Only a small subset of housekeeping genes showed adequate uniform hybridization. We developed a software tool for objective evaluation of oligonucleotide-chips based on automated analysis of as well as normalization to this subset of housekeeping genes. We analyzed the gene expression profile of microprepared lymphatic vascular endothelial cells. We show that optimized normalization prevented exclusion of angiopoietin-2, a lymphatic endothelial marker, from the lymphovascular transcriptome.