Normalisation against Circadian and Age-Related Disturbances Enables Robust Detection of Gene Expression Changes in Liver of Aged Mice

• Published in: PloS one Vol. 12; no. 1; p. e0169615
• Main Authors: Fonseca Costa, Sara S., Wegmann, Daniel, Ripperger, Jürgen A.
• Format: Journal Article
• Language: English
• Published: United States Public Library of Science 09.01.2017; Public Library of Science (PLoS)
• Subjects: Age; Aging; Aging - genetics; Animals; Bioinformatics; Biology; Biology and Life Sciences; Casein; Change detection; Circadian rhythm; Circadian Rhythm - genetics; Circadian rhythm sleep disorders; Circadian rhythms; Computational Biology - methods; Gene expression; Gene Expression Profiling; Gene Expression Regulation; Gene sequencing; Genes; Genetic aspects; Genetic Markers; Genetically modified mice; Genomes; Glucose transporter; Health aspects; Hybridization; Kinases; Life expectancy; Liver; Liver - metabolism; Mammals; Metabolism; Mice; Outliers (statistics); Phosphorylation; Physiology; Research and analysis methods; Ribonucleic acid; RNA; Robustness; Rodents; Set theory; Transcription; Transcriptome
• ISSN: 1932-6203; 1932-6203
• DOI: 10.1371/journal.pone.0169615

The expression of some genes is affected by age. To detect such age-related changes, their expression levels are related to constant marker genes. However, transcriptional noise increasing with advancing age renders difficult the identification of real age-related changes because it may affect the marker genes as well. Here, we report a selection procedure for genes appropriate to normalise the mouse liver transcriptome under various conditions including age. These genes were chosen from an initial set of 16 candidate genes defined based on a RNA-sequencing experiment and published literature. A subset of genes was selected based on rigorous statistical assessment of their variability using both RNA-sequencing and Nanostring hybridization experiments. The robustness of these marker genes was then verified by the analysis of 130 publicly available data sets using the mouse liver transcriptome. Altogether, a set of three genes, Atp5h, Gsk3β, and Sirt2 fulfilled our strict selection criteria in all assessments, while four more genes, Nono, Tprkb, Tspo, and Ttr passed all but one assessment and were included into the final set of marker genes to enhance robustness of normalisation against outliers. Using the geometric mean of expression of the genes to normalise Nanostring hybridization experiments we reliably identified age-related increases in the expression of Casein kinase 1δ and 1ϵ, and Sfpq, while the expression of the glucose transporter Glut2 decreased. The age-related changes were verified by real-time PCR and Western blot analysis. As conclusion, proper normalisation enhances the robustness of quantitative methods addressing age-related changes of a transcriptome.