The quality of microarray datasets
The datasets GSE92913 contained 7 SD samples and 7 control samples (Fig. 1A). The median and standard deviation distributions of each sample were consistent (Fig. 1B). The principal component analysis (PCA) showed statistically significant difference in gene expression between SD group and control group (Fig. 1C). The density maps described the expression of genes in each sample, which showed no abnormal samples (Fig. 1D). These results suggest that the quality of datasets GSE92913 meets the criteria for further analysis.
Functional analysis of GSEA on SD
We performed GO and KEGG analysis based on the GSEA method. Several functions were statistically significant enriched after SD, indicating these functions were affected by SD. The biological process showed that protein folding, response to topologically incorrect protein, and ncRNA metabolic process were up-regulated, while sleep, vitamin metabolic process, hormone metabolic process, cellular aldehyde metabolic process and other metabolic processes were significantly suppressed (Fig. 2A, B). The KEGG showed that PPAR signaling pathway and RNA transport were activated, while drug metabolism, pentose and glucuronate interconversions, bile secretion, glutathione metabolism and purine metabolism were significantly suppressed after SD (Fig. 2C, D). These results indicate that the changes of liver after SD are mainly related to metabolic process.
GO and KEGG analysis on SD
Based on the screening criteria, we obtained 374 DEGs in the liver after SD, including 179 up-regulated genes and 195 down-regulated genes (Fig. 3A). DEGs were mainly involved in several kind of metabolic processes, such as steroid, small molecule, monocarboxylic acid, hormone, glycogen and triglyceride metabolic process. Interestingly, pathways including cholesterol metabolism, transport of small molecules and circadian rhythm were significantly enriched (Fig. 3B, C). Therefore, we showed all significant entries related to circadian rhythms (Fig. 3D). There results suggest that metabolism and circadian rhythm are affected after SD, which may be important factors for liver damage.
Analysis of circadian genes in liver after SD
As the circadian rhythm pathway is altered after SD, we focused on exploring the effects of circadian genes on liver function. We obtained 60 differently expressed circadian genes in liver after SD (Fig. 4A). Functional analysis showed that circadian genes were engaged in regulation of lipid metabolic process, response to extracellular stimulus, triglyceride metabolic process, glucose homeostasis, lipid localization, small molecule, androgen metabolic process and response to glucagon. These results support the important role of circadian genes on metabolism during SD. Besides, circadian genes were also involved in the regulation of circadian sleep/wake cycle, aging, and regeneration (Fig. 4B, C). In addition, there was a strong association between these circadian genes according to the results of PPI network, suggesting their synergistic functions (Fig. 4D). These results suggest that genetic disruption of circadian rhythms after SD may be an important cause of metabolic disorders.
Relationship between circadian genes
To explore the correlation between circadian genes, we obtained 16 most important genes in the PPI network based on MCODE analysis, suggesting their key role in liver injury after SD (Fig. 5A). Pearson correlation analysis in R software showed most of circadian genes share high correlation (Fig. 5B), demonstrating their consistent function after SD.
Verifying the expression of circadian genes
We showed the expression of 16 most important circadian genes in GSE92913 (Fig. 6A), which were verified in mice liver after SD in datasets GSE9441 and GSE114845. Overall, the expression of circadian genes in the validation datasets were relatively consistent with the predicted one. In particular, the expression of Per3, Nedd4l, Cry1, Fkbp5, Tsc22d3 and Cdkn1a showed the same trend and their differences were statistically significant in two liver datasets (Fig. 6B, C). Under physiological conditions, the proteins corresponding to these genes were lowly expressed except for PER3 in human liver, which corresponds exactly to their altered transcript levels after SD (Fig. 6D). Meanwhile, we found that all of detectable circadian genes in human blood of GSE98582 were significantly changed (Fig. 6E). We also found that many circadian genes were differentially expressed in the hippocampus (GSE33302) and brain (GSE9442) after SD, indicating their potential value in memory (Fig. 6F, G). However, in heart (GSE42323) and lungs (GSE42324), most circadian genes expressed with no significance (Fig. 6H, I), probably because they were not applicable to these two organs.
SD altered the liver immune infiltration
As we known, immune function is suppressed after SD. Here, we revealed the specific immune changes in the liver after SD. The immune cells presented different amounts in the liver (Fig. 7A), with down-regulation of CD4 naïve T cells and M2 macrophages, and up-regulation of M0 macrophages, CD4 memory resting T cells, activated NK cells, resting dendritic cells and resting mast cells after SD (Fig. 7B). Six circadian genes exhibited a relationship with immune cells (Fig. 7C). Specifically, we independently verified that in sleep deprived mice, high and low Cry1 associated DEGs in livers showed an association with metabolic function and circadian rhythms, which supported the influence of Cry1 on metabolism and circadian rhythms (Fig. 7D).