Validating reference genes using minimally transformed qpcr data: findings in human cortex and outcomes in schizophrenia

Background: It is common practice, when using quantitative real time polymerase chain reaction (qPCR), to normalise levels of mRNA to reference gene mRNA which, by definition, should not vary between tissue, with any disease aetiology or after drug treatments. The complexity of human CNS means it unlikely that any gene could fulfil these criteria. Methods: To address this issue we measured levels of mRNA for six potential reference genes (GAPDH, PPIA, SNCA, NOL9, TFB1M and SKP1) in three cortical regions (Brodmann's areas (BA) 8, 9 and 44) from 30 subjects with schizophrenia and 30 age and sex matched controls. We used a structured statistical approach to examine the characteristics of these data to determine their suitability as reference genes. We also analysed our data using reference genes selected by rank as defined using the average of the standard deviation of pair-gene Delta Ct and the BestKeeper, NormFinder and geNorm algorithms to determine if they suggested the same reference genes. Results: Our minimally derived data showed that levels of mRNA for all of the six genes varied between cortical regions and therefore no gene fulfilled the absolute requirements for use as reference genes. As levels of some mRNA for some genes did not vary with diagnoses within a cortical region from subjects with schizophrenia compared to controls, we normalised levels of mRNA for all the other genes to mRNA for one, two or three reference genes in each cortical region. This showed that using the geometric mean of at least two reference genes gave more reproducible results. Finally, using the reference gene ranking protocols the average of the standard deviation of pair-gene Delta Ct, BestKeeper, NormFinder and geNorm we showed that these approaches ranked potential reference genes differently. We then showed that outcomes of comparing data from subjects with schizophrenia and controls varied depending on the reference genes chosen. Conclusions: Our data shows that the selection of reference genes is a significant component of qPCR study design and therefore the process by which reference genes are selected must be clearly listed as a potential confound in studying gene expression in human CNS. This should include showing that, using minimally derived qPCR data, levels of mRNA for proposed reference genes does not vary with variables such as diagnoses and CNS region.