A cross-tissue transcriptomic approach decodes glucocorticoid receptor-dependent links to human metabolic phenotypes

Abstract

Glucocorticoids, acting through the glucocorticoid receptor (GR), control metabolism, maintain homeostasis, and enable adaptive responses to environmental challenges. Their function has been comprehensively studied, leading to identification of numerous tissue-specific GR-dependent mechanisms. Abundant evidence shows that GR-triggered responses differ across tissues, however, the extent of this specificity was not comprehensively explored. It is also unknown how particular GR-induced molecular patterns are translated into profile of higher-level human traits. Here, we examine cross-tissue effects of GR activation on gene expression. We assessed changes induced by stimulation with GR agonist, dexamethasone in nine tissues (adrenal cortex, perigonadal adipose tissue, hypothalamus, liver, kidney, anterior thigh muscle, pituitary gland, spleen, and lungs) in adult male C57BL/6 mice, using whole-genome microarrays. Dexamethasone induced balanced transcriptional responses across all examined tissues with 585 identified dexamethasone-regulated transcripts, including 446 with significant treatment-tissue interaction effects. Clustering analysis revealed sixteen GR-dependent patterns, including those universal across tissues and tissue-specific. We leveraged existing gene annotations and created new annotation sets based on chromatin immunoprecipitation sequencing, recent large-scale genome-wide association studies, and human transcriptome collections. As expected, GR-dependent transcripts were associated with essential metabolic processes (glycolysis/gluconeogenesis, lipid-metabolism) and inflammation-related pathways. Beyond these, we found novel links between regulated gene patterns and human phenotypic traits, like reticulocyte count or blood triglyceride levels. Overall effects of GR stimulation are well coordinated and closely linked to biological roles of tissues and organs. Our findings provide novel insights into complex systemic and tissue-specific actions of glucocorticoids and their potential impacts on human physiology and pathology.

Keywords: Dexamethasone; Glucocorticoid receptor; Metabolic traits; Tissue-specific transcription; Transcriptional regulation.

Fig. 1

Dexamethasone treatment affects gene expression in different tissues. Barplots (mean ± SD) of log₂ ratio for the top 50 transcripts are shown for each tissue to represent overall tissue impact. ADR—adrenal cortex, FAT—perigonadal adipose tissue, HTH—hypothalamus, KID—kidneys, LIV—liver, LUN—lungs, MUS—anterior thigh muscle, PIT—pituitary gland, SPL—spleen. Significant differences in fold changes between tissues obtained by a Tukey HSD test are indicated by # (vs. Kidney; p < 0.05) and by * (vs. Hypothalamus; p < 0.05)

Fig. 2

Hierarchical clustering of dexamethasone-induced transcriptional alterations. Relative levels of all transcripts (585) with differential expression are shown as a heat map (Supplementary Table 1). Data were standardized for the expression level in each tissue. Colored rectangles represent transcript abundance 4 h after injection of DEX or CTR in the specific tissue, as indicated below the heatmap. The intensity of the color is proportional to the standardized values (between −4 and 4) from each microarray (according to the scale below the heatmap). Gene clusters are depicted as colors and letters (A-P) on the left. Clustering was performed using Pearson correlation as distance, the results are summarized on the dendrogram shown on the right. Clusters A-F are classified as down-regulated (DOWN). Clusters G-P are classified as up-regulated (UP) in response to DEX

Fig. 3

The spectrum of putative transcriptional regulators among the DEX-induced gene expression patterns. The heatmap presents the statistical significance of enrichment of binding sites for transcription factors (TF) in the regulatory regions of genes from the clusters identified in this work. The information about TFs binding data were obtained from the ChEA 2022 database. The columns represent GR-responsive genes grouped in clusters labeled A to P. UP and DOWN are combined groups of all DEX-increased and DEX-decreased transcripts, respectively. For each cluster, the number of genes and the observed direction of the transcriptional regulation (‘ + ’ increase, ‘-’ decrease) are indicated. Statistically significant results are represented by colored squares: p_adj. < 0.2 (gray); p_adj. < 0.1 (light red); p_adj. < 0.05 (medium red); p_adj. < 0.01 (dark red); p_adj. < 0.001 (deep red). TFs were annotated as specified in the legend. The dots represent tissue co-localization between DEX-induced gene expression patterns and corresponding TF expression in at least one tissue associated with the gene cluster (as detailed in Supplementary Table 3). Up to 10 top statistically significant results per cluster with a minimum of two genes per term were included. Light green dots represent TF expression level in tissues based on GTEx data (mean CPM > 50) [34]. Dark green dots represent TF expression measured in our dataset (mean log₂ value for the control group > 8). Full results are listed in Supplementary Table 3

Fig. 4

Heatmap of gene overlaps between the GR-dependent transcriptional patterns and literature-based GR-dependent gene lists for various tissues and cells. Lists of genes for tissues and cell types were extracted from transcriptomic studies of GR-dependent gene expression (PMIDs are provided in the rightmost column). The chi-square test was used to assess the overlap between the clusters of GR-regulated genes (x-axis) and literature (y-axis). χ² values were transformed in the following way: log₂(χ² + 1) and are displayed in each rectangle of the heatmap (white blocks—no overlap; red blocks—higher overlap). Values in brackets within each square indicate the number of overlapping genes. Statistically significant results are indicated as colored frames: p_adj < 0.01 (green frames); p_adj < 0.0001 (purple frames). The gene clusters are labeled at the top of the columns. For each cluster, the number of genes and the observed direction of the transcriptional regulation (‘ + ’ increase, ‘ − ’ decrease) are indicated. Cluster UP and cluster DOWN represent combined groups of all DEX-induced and DEX-decreased transcripts, respectively. The row description on the right of the heatmap provides details on the gene lists obtained from the literature ('+'represents upregulation and'−'represents downregulation). The gene lists presented on the heatmap are organized according to hierarchical clustering, indicated by both row and column dendrograms. An extended table summarizing gene list overlaps is included in Supplementary Table 2

Fig. 5

GR-dependent transcriptional patterns overlap with groups of genes associated with (A) human phenotype- or (B) metabolism-related traits. A The heatmap on the left presents an overlap between the GR-regulated genes and human phenotype-associated gene lists. Lists of genes associated with phenotypes were extracted from the Pan-UK Biobank database (phenocodes are provided). B The right panel presents the overlap between the clusters of GR-dependent genes and gene lists associated with metabolite levels (based on Metabolon and Nightingale databases). The chi-square test was used to examine the overlap of the lists of the GR-regulated (columns) and phenotype- or metabolism-associated genes (rows). χ² values were transformed using the formula: log₂(χ² + 1) and are presented in each rectangle of the heatmap. The intensity of the red color is proportional to the level of overlap, as indicated in the legend (white, no overlap; red, high-level overlap). The statistical significance of the overlap is pinpointed using two thresholds: p_adj < 0.01 (green frames) and p_adj < 0.0001 (purple frames). Values in brackets in each box show the number of overlapping genes. The results with a minimum of three genes were included. The identified gene clusters are indicated on top column labels with the number of genes inside brackets. Cluster UP and cluster DOWN represent combined groups of DEX-induced and DEX-decreased transcripts, respectively. For each cluster, the number of genes and the observed direction of the transcriptional regulation (‘ + ’ increase, ‘ − ’ decrease) are indicated. The row description presented on the right of the heatmaps provides details on lists of human phenotypes and metabolomes from genome-wide association studies (GWAS). The gene lists presented on the heatmaps are organized according to hierarchical clustering, indicated by both row and column dendrograms. To reduce redundancy, for cases where the overlap contained the same genes for more than one phenotype, only the phenotype with the lowest p_adj is displayed. Full results are available in Supplementary Table 6