Circadian clocks are resounding in peripheral tissues

Abstract

Circadian rhythms are prevalent in most organisms. Even the smallest disturbances in the orchestration of circadian gene expression patterns among different tissues can result in functional asynchrony, at the organism level, and may to contribute to a wide range of physiologic disorders. It has been reported that as many as 5%-10% of transcribed genes in peripheral tissues follow a circadian expression pattern. We have conducted a comprehensive study of circadian gene expression on a large dataset representing three different peripheral tissues. The data have been produced in a large-scale microarray experiment covering replicate daily cycles in murine white and brown adipose tissues as well as in liver. We have applied three alternative algorithmic approaches to identify circadian oscillation in time series expression profiles. Analyses of our own data indicate that the expression of at least 7% to 21% of active genes in mouse liver, and in white and brown adipose tissues follow a daily oscillatory pattern. Indeed, analysis of data from other laboratories suggests that the percentage of genes with an oscillatory pattern may approach 50% in the liver. For the rest of the genes, oscillation appears to be obscured by stochastic noise. Our phase classification and computer simulation studies based on multiple datasets indicate no detectable boundary between oscillating and non-oscillating fractions of genes. We conclude that greater attention should be given to the potential influence of circadian mechanisms on any biological pathway related to metabolism and obesity.

Figure 1. Summary of the Microarray Analysis of Circadian Periodicity and Phase in Murine BAT, iWAT, and Liver

The red line marks the 0.05 cutoff for the p-value in Fisher's g-test. The Roman numerals represent the grouping of all expressed genes based on the calculated circadian phase, displayed in zeitgeber time: I, ZT0; II, ZT4; III, ZT8; IV, ZT16. To produce the heat map, all expression profiles have been z-scored. There seems to be no dependence between periodicity (indicated by p-value) and overall level of gene expression (Figure S4).

Figure 2. Results of the Simulation Experiment

Plot (A) shows distribution of p-values and plot (B) shows respective autocorrelation coefficients for raw data and datasets with different proportion of genes left intact, while for the rest of the genes periodicity is eliminated by random permutation of time points. The dotted line indicates the p = 0.05 significance cutoff. The raw data is derived from the liver circadian expression microarray analysis.

Figure 3. Expression Profiles and Periodicity Analysis of Basic Circadian Rhythm Genes in iWAT

The first column plots are raw expression values as reported by the Affymetrix MAS5 algorithm. The second column plots show the same data after preprocessing (central value adjustment, polynomial smoothing, and trimmed mean subtraction). The third column presents the periodograms resulting from DFT analysis. Highlighted is a single peak corresponding to the circadian rhythm (two complete cycles in 48 h). For each probeset, representing a particular gene, the results of Fisher's g-test (p_f), permutation test (p_p), and autocorrelation (r) with circadian lag are listed to the right.

Figure 4. Experimental qRT-PCR Verification of Selected Microarray Expression Profiles in Liver.

The qRT-PCR expression profiles for selected transcripts represented on microarray (probe names given in brackets).

Figure 5. Experimental Verification of Selected Microarray Expression Profiles in Liver (continued from Figure 4).

The alternative qRT-PCR expression profiles presented on Figure 4 are consistent with microarray data in phase and visual presence of diurnal oscillation pattern.