An assessment of the efficacy of in situ hybridization as a quantitative method by variance components estimation

One serviceable feature of in situ hybridization is its potential for assessing relative levels of mRNA in specific regions of tissues and organs. To determine its efficacy as a quantitative technique, we applied a nested factorial design to a multifactorial experiment. Estimates of the magnitude of variance components then allowed an assessment of variation over samples of sections from the same tissue source, variation in label over 2 anatomical sites within the same section of tissue, as well as experiment‐to‐experiment variation. We found ∼ 51% of the total variance arose from experiment‐to‐experiment variation, while ∼ 21% of the total variance was due to variation in autoradiography grain density over neurons in the same brain region. Rat‐to‐rat variation accounted for approximately 11%. About 10% of the variance was due to variation between sections of tissue that were derived from the same tissue source and were hybridized in the same hybridization experiment. Variation between 2 homologous, bilaterally located brain regions located on the same tissue section (the right and left supraoptic nucleus), accounted for ∼ 5% of the total variance. The remaining unaccounted error variance was ∼ 2% of the total variance. Since an expected change in cellular content of a particular mRNA was observed as a function of experimental treatment, results suggest in situ hybridization is a useful quantitative method. Findings also indicate, however, the importance of experimental design: the use of multiple samples of tissue from the same tissue source, a sufficient number of tissue sources (animals, batches of cultured cells) to account for variations in sample sources, and the need to assess experiment‐to‐experiment and rat‐to‐rat variations. Results suggest the utility of analyzing the data of in situ hybridization experiments from the perspective of experimental design. © 1993 Wiley‐Liss, Inc.