Social status alters chromatin accessibility and the gene regulatory response to glucocorticoid stimulation in rhesus macaques

Abstract

Low social status is an important predictor of disease susceptibility and mortality risk in humans and other social mammals. These effects are thought to stem in part from dysregulation of the glucocorticoid (GC)-mediated stress response. However, the molecular mechanisms that connect low social status and GC dysregulation to downstream health outcomes remain elusive. Here, we used an in vitro GC challenge to investigate the consequences of experimentally manipulated social status (i.e., dominance rank) for immune cell gene regulation in female rhesus macaques, using paired control and GC-treated peripheral blood mononuclear cell samples. We show that social status not only influences immune cell gene expression but also chromatin accessibility at hundreds of regions in the genome. Social status effects on gene expression were less pronounced following GC treatment than under control conditions. In contrast, social status effects on chromatin accessibility were stable across conditions, resulting in an attenuated relationship between social status, chromatin accessibility, and gene expression after GC exposure. Regions that were more accessible in high-status animals and regions that become more accessible following GC treatment were enriched for a highly concordant set of transcription factor binding motifs, including motifs for the GC receptor cofactor AP-1. Together, our findings support the hypothesis that social status alters the dynamics of GC-mediated gene regulation and identify chromatin accessibility as a mechanism involved in social stress-driven GC resistance. More broadly, they emphasize the context-dependent nature of social status effects on gene regulation and implicate epigenetic remodeling of chromatin accessibility as a contributing factor.

Keywords: chromatin accessibility; dominance rank; epigenomics; gene regulation; social status.

Fig. 1.

Dominance rank and Dex treatment both induce widespread changes in gene expression and chromatin accessibility. (A) Social group formation for one study group. Each line represents a different female and starts at the date of her introduction. Females introduced earlier were higher-ranking at the time of sampling (Pearson’s r = −0.67, P = 8.8 × 10⁻⁷; n = 9 groups and 43 females). (B) Schematic of the in vitro GC challenge. Paired PBMC samples from each female were treated with either 1 μM Dex (Dex+), a synthetic GC, or 0.02% ethanol (the vehicle for Dex, Dex−) and subjected to a 90-min incubation period before RNA and DNA extraction. (C) Example response to Dex challenge for FKBP5, a known GC-induced gene, for chromatin accessibility in intron 4 of isoform FKBP5-201 (P = 7.36 × 10⁻¹⁸; this region also falls in intron 5 of isoform FKBP5-202) and for gene expression (P < ×10⁻¹⁰⁰). (D) Dex treatment is strongly associated with the first PC of gene expression data (paired T₄₂ = 80.61, P = 1.12 × 10⁻⁴⁷) and (E) with the first three PCs of chromatin accessibility (PC1 T₄₂ = −3.85, P = 3.95 × 10⁻⁴; PC2 T₄₂ = 4.49, P = 5.44 × 10⁻⁴; PC3 T₄₂ = −4.93, P = 1.35 × 10⁻⁵). Social status is significantly associated with PC3 for both (F) gene expression (Pearson’s r = 0.77, P = 3.33 × 10⁻¹⁸) and (G) chromatin accessibility (Pearson’s r = 0.48, P = 3.41 × 10⁻⁶).

Fig. 2.

Differences in the stability of social status effects on gene expression and chromatin accessibility after Dex treatment. (A) The effects of social status on gene expression are stronger before Dex treatment (paired T₂₂₇₈ = 17.33, P = 2.63 × 10⁻⁶³), but status effects on chromatin accessibility are similar before and after Dex treatment (paired T₄₀₈₂ = 2.69, P = 0.007). “Scaled difference” is the difference between the absolute value of the effect of social status in the Dex− and Dex+ conditions scaled by the sum of the effect sizes: $(\left|{\beta }_{status\hspace{0.17em}in\hspace{0.17em}Dex-}\right|-|{\beta }_{status\hspace{0.17em}in\hspace{0.17em}Dex+}\left|\right)/(\left|{\beta }_{status\hspace{0.17em}in\hspace{0.17em}Dex-}\right|+|{\beta }_{status\hspace{0.17em}in\hspace{0.17em}Dex+}\left|\right)$. Solid vertical lines represent the median difference in scaled effect sizes. (B) We observed an excess of genes with large social status effects on gene expression in the Dex− versus the Dex+ condition. In contrast, we observed no consistent bias in the number of social status-associated genes in the Dex− versus Dex+ condition as effect size increased. The x axis depicts mutually exclusive bins of effect sizes for the standardized effects of status on gene expression (blue) or chromatin accessibility (pink). The y axis represents the log₂-transformed ratio (±SE) of the number of genes falling in each bin for the Dex− compared with the Dex+ condition.

Fig. 3.

Dex treatment affects the correlation between rank-associated chromatin accessibility and rank-associated gene expression levels. Standardized rank effects on chromatin accessibility and standardized rank effects on the expression of the closest gene are correlated in the Dex− condition (Pearson’s r = 0.23, P = 1.51 × 10⁻¹⁹⁹; n = 16,677 region–gene pairs) (A) but negligibly so after Dex treatment (r = 0.10, P = 2.76 × 10⁻⁹; n = 16,677) (B). (C and D) Pearson’s correlation (±SE) between the standardized effect of social status on chromatin accessibility and the standardized effect of social status on gene expression for the closest gene (y axis) is stronger in the Dex− condition compared with the Dex+ condition, whether conditioning on the gene expression association with dominance rank (C, x axis) or the chromatin accessibility association with dominance rank (D, x axis). CA, chromatin accessibility; GE, gene expression.

Fig. 4.

The relationship between rank effects and Dex effects on gene expression, chromatin accessibility, and TF accessibility. (A) The effects of rank and Dex treatment on gene expression were uncorrelated (Pearson’s r = 0.007, P = 0.47), unlike (B) the effects of rank and Dex treatment on chromatin accessibility (Pearson’s r = 0.18, P = 5.78 × 10⁻¹⁴⁷). (C) TF binding sites enriched in Dex-induced regions (n = 2,874 regions, FDR < 20%) also were enriched in regions that were more accessible in high-ranking females (n = 544 regions, FDR < 20%; Pearson’s r = 0.64, P = 1.95 × 10⁻⁴¹). Conversely, TF binding sites enriched in Dex-repressed regions (negative values on the x axis; n = 2,905 regions, FDR < 20%) were also enriched in regions that were more accessible in low-ranking females (negative values on the y axis; n = 540 regions, FDR < 20%). Each point (n = 369 TFBS) represents the log₂-fold enrichment of one TF motif in Dex-induced compared with Dex-repressed regions (x axis) and rank-induced versus rank-repressed regions (y axis). TF motifs closely associated with GC regulation (GRE, AP-1, and NF-κB) are shown in colored circles.