Single-cell transcriptomics provides insights into the stress-immune interplay and informs disease risk Authors: Sofia Benavides, Oksana Kosyk, Haiyang Wang, Alexander G Bastian, Christina Georgiou, Brian C Miller, Natalie Stanley, Anthony S Zannas Published: 2026 ABSTRACT SummaryEnvironmental stress contributes to many diseases—including cardiovascular disease, mental illness, and autoimmune disorders—through immune dysregulation. Stress operates through the hypothalamic-pituitary-adrenal and sympathoadrenal axes, leading to peripheral secretion of cortisol and norepinephrine. While these stress hormones can profoundly influence immune function, their mechanistic effects at the cell-type level remain poorly understood. To address this knowledge gap, we employed an ex vivo model investigating the transcriptomic responses of primary peripheral blood mononuclear cells and neutrophils to physiological stress levels (i.e., levels reached during in vivo stress) of cortisol and norepinephrine at single-cell resolution. We identified novel stress-hormone-dependent and cell-type-specific transcriptomic changes, including uniquely regulated genes and inflammation-related signatures within distinct innate (e.g., CD163+ monocytes/macrophages and neutrophils) and other immune cell types. The single-cell-derived signatures effectively distinguished individuals with stress-related diseases in two independent human cohorts. Our findings provide cell-type-level insights into how stress hormones modulate immune function relevant to human health.Keywords: catecholamines, CD163+ monocytes, glucocorticoids, immune system, single-cell transcriptomics, stress Introduction Environmental stress, an unavoidable characteristic of every society,1 has been linked to an increased risk for various diseases.2,3,4,5,6,7 For example, studies have established an association between excessive psychosocial stress and an increased risk of cardiovascular disease (CVD),2 mental illness,8,9 and autoimmune disorders.5 While the underlying mechanisms are poorly understood, immune dysregulation has been repeatedly highlighted as a key process shared across stress-related conditions.10,11,12 Stress operates through two distinct biological pathways: the hypothalamic-pituitary-adrenal (HPA) axis13 and the sympathoadrenal axis.14 These pathways culminate in the secretion of stress hormones—cortisol (CORT) and norepinephrine (NE)—into the periphery, where they interact with immune cells.15 Controversy exists regarding the effects of stress hormones on immune cells;16 some studies suggest an anti-inflammatory response,17,18 while others demonstrate pro-inflammatory effects.19 However, to our knowledge, all previous research on stress hormones and immune cells has focused on analyzing a mixture of cells (i.e., bulk-level analysis),20,21 thereby not accounting for cell-type-specific effects that could play a critical role in mediating stress pathology. Additionally, most studies investigating stress-induced immune dysfunction have relied on in vivo animal models. While these models are invaluable for studying complex systemic effects, they are inherently limited by species differences and confounding factors such as dynamic shifts in the relative proportion of cell types. To address these knowledge gaps, here we employ a cell model (ex vivo) approach and single-cell RNA sequencing (scRNA-seq) to directly assess the immune effects of physiological stress hormone exposure at the single-cell level. To model physiological stress in culture, primary peripheral blood mononuclear cells (PBMCs) and neutrophils isolated from human donors were exposed to the endogenous human stress hormones CORT and NE at concentrations known to be reached in human tissues during in vivo stress.22,23 This approach enabled us to model stress ex vivo and conduct a comprehensive analysis of the unique immune transcriptomic signatures of physiological stress hormone exposure at the single-cell and cell-type levels. The single-cell-derived signatures, applied to two independent human cohorts, effectively distinguished individuals with stress-related diseases. Our findings provide new insights into how stress hormones influence immune cell function and the immune cell-type-specific mechanisms underlying stress-related conditions. Results Immune single-cell transcriptional profiling after ex vivo exposure to physiological stress hormone levels To evaluate the effects of stress hormones on immune cell function, we obtained whole blood samples from four male and four female healthy donors (Table S1) and isolated PBMCs and neutrophils. Cells from each donor were treated with vehicle control (0.0001% DMSO), 100 nM CORT, or 10 nM NE (Figure 1A). These concentrations were selected based on previous literature indicating that in vivo stress exposure can result in circulating concentrations of 100 nM CORT22 and 10 nM NE.23 This robust experimental design enables within-donor comparisons, thereby accounting for potential inter-individual variability in transcriptional responses. Single-cell RNA sequencing analysis of immune cells after ex vivo exposure to stress hormones (A) Graphical representation of experimental model. PBMCs and neutrophils from healthy human donors were isolated and exposed to control (DMSO) and treatment (cortisol [CORT] and norepinephrine [NE]) groups. After 4 h of incubation for neutrophils (due to their shorter lifespan in cell culture) and overnight incubation for PBMCs, cells were prepared for single-cell RNA sequencing. (B) UMAP plot showing 62,975 cells from all conditions, colored by cell type. (C) UMAP plots are separated by treatment group, highlighting that all cell-type clusters are represented in each condition. (D) Heatmap displaying differentially expressed genes (DEGs), canonical marker genes, and genes identified in previous PBMC studies used for cell-type annotation. Heatmap is normalized by columns. See also Figures S1 and S2 and Table S1. After excluding low-quality cells and potential doublets, we retained 62,975 high-quality cells: 23,000 from the vehicle group, 20,707 from the CORT group, and 19,268 from the NE group. Quality control metrics for the retained cells are shown in Figures S1B–S1F. Cluster analysis identified 20 distinct cell clusters, visualized using a uniform manifold approximation and projection (UMAP) approach24 (Figure 1B). Although the treatment groups had similar abundances of cells across all 20 clusters, initial UMAP visualization by treatment revealed subtle treatment-induced shifts in the distribution of cell type (Figure 1C). The clusters encompassed all major cell types typically found in peripheral blood, including lymphocytes, monocytes, and neutrophils. We identified five distinct monocyte/macrophage (MonoMac) clusters: CD14− MonoMacs,25,26 which most closely resemble non-classical monocytes; CD163+ MonoMacs, resembling alternatively activated monocytes27; CD74+ MonoMacs, which align with classical monocytes25; VCAN+ MonoMacs, showing expression patterns similar to intermediate monocytes25; and IL7R+ MonoMacs, which appear to represent a specialized subset of tissue-resident monocytes.28,29 The distribution of cells by individual donor is shown in Figure S1A. Notably, cells tended to cluster by donor, particularly within the MonoMac populations. However, because samples were sequenced in a single batch, we were cautious to avoid overcorrecting for donor-level effects that may reflect genuine biological differences. MonoMac populations exhibited partial separation when clustered by sex (Figure S2A). To minimize the risk of attributing treatment effects to sex-specific differences while preserving inter-individual variability, we therefore performed a parallel analysis in which the dataset was harmonized by donor, improving integration of cells across sexes (Figure S2B). All clusters were successfully annotated using the top differentially expressed genes (DEGs), canonical markers, and reference data from previously published scRNA-seq datasets29,30 (Figure 1D). Exposure to physiological stress hormone levels alters immune cell-type proportions While stress-induced changes in immune cell type composition have been well-documented in vivo, the functional impact of CORT and NE on distinct immune cell types remains unclear. Addressing this question in culture allows for evaluating the direct effects of stress hormones on immune cell composition, free from the confounding effects of cell mobilization and redistribution observed in vivo. Using density plots, we qualitatively observed treatment-dependent changes in cell-type composition. CORT treatment resulted in a general decrease in T cell proportions and altered the distribution of MonoMac populations (Figure 2A). In contrast, NE treatment produced a distinct pattern of T cell distributions and a unique MonoMac distribution that differed from the control and CORT-treated groups (Figure 2A). Stress hormone treatment induces changes in immune cell-type composition (A) Density plots showing cell-type distributions in control and treatment groups (n = 8 donors per condition). Blue represents lower cell density, while red indicates higher density. (B) Changes in the proportions of plasmacytoid dendritic cells, naive CD8+ T cells, memory CD8+ T cells, and basophils following CORT exposure. While all cell types shown were nominally significant, only naive CD8+ T cells and plasmacytoid dendritic cells remained significant after multiple testing correction using the Bonferroni method. Graphs depict the deltas between CORT and DMSO. Two-sided paired t tests were used to test for significance (∗Bonferroni-adjusted p < 0.05). (C) Changes in the proportions of plasmacytoid dendritic cells, naive CD4+ T cells, memory B cells, and CD163+ MonoMacs following NE exposure. While all cell types shown were nominally significant, only memory B cells remained significant after multiple testing correction using the Bonferroni method. Graphs show the deltas between NE and DMSO. Two-sided paired t tests were used to test for significance (∗Bonferroni-adjusted p < 0.05). (D) Differences in the MLR between treatment and control groups. Two-sided paired t tests were used to test for significance (∗p < 0.05). See also Figures S3 and S4. We further quantified these changes, leveraging the strength of our experimental design that enables within-donor comparisons between control and treated cells. These comparisons revealed significant stress hormone-induced changes in cell-type proportions. Specifically, CORT treatment led to nominally significant reductions in memory and naive CD8+ T cells, basophils, and plasmacytoid dendritic cells. However, after correcting for multiple testing, only the decreases in naive CD8+ T cells and plasmacytoid dendritic cells remained statistically significant (Figure 2B). Similarly, NE treatment resulted in nominally significant decreases in plasmacytoid dendritic cells and memory B cells, as well as increases in naive CD4+ T cells and CD163+ MonoMac populations (Figure 2C). After multiple testing corrections, only the reductions in memory B cells remained statistically significant. Additional patterns of change in cell-type proportions were observed but did not reach statistical significance (Figures S3A and S3B). To validate these findings on cell-type proportions, we performed RNAScope analysis in four independent donors (two males and two females). We chose to focus on validating the CORT-driven decrease in naive CD8+ T cells because this cell type was the most abundant among the statistically significant cell populations (naive CD8+ T: ∼6.26%, memory B: ∼1.50%, and plasmacytoid dendritic cells: ∼0.29%), which allowed for greater statistical power. We observed a pattern of decreased naive CD8+ T cells across donors (Figures S4A–S4C), but the overall result did not reach statistical significance. This may reflect limitations in power and the lower sensitivity of the RNAScope assay compared to scRNA-seq. To further evaluate the potential clinical relevance of stress hormone-induced immune remodeling, we calculated the monocyte-to-lymphocyte ratio (MLR), an established prognostic marker of conditions such as cardiovascular mortality31 and cancer.32 Our analysis revealed a significant increase in MLR following CORT, but not NE, treatment (Figure 2D). These findings indicate that, even in the absence of cell mobilization and redistribution, exposure to physiological stress hormone levels can selectively remodel the immune cell composition. Exposure to physiological stress hormone levels induces cell-type-specific gene expression changes Next, we investigated the extent to which exposure to CORT and NE concentrations corresponding to physiological levels of stress alters immune cell-type-specific gene expression. We first examined the overall effects of CORT and NE, ignoring cell-type distinctions (i.e., when pooling all cell types). CORT exposure resulted in a total of 450 DEGs with an absolute log fold change (LFC) > 0.25 and a false discovery rate (FDR)-corrected p (q) < 0.05. Of these, 251 genes were downregulated, and 199 were upregulated (Table S2). Importantly, CORT increased the expression of well-characterized glucocorticoid-responsive genes such as FKBP5 and ZBTB16 (Figure 3A) across cell types, confirming the effectiveness of our ex vivo treatment. We validated several top DEGs (VSIG4, FKBP5, and ZBTB16) using qPCR (Figure 3B). We next investigated whether these CORT-induced gene expression changes differed between sex groups. To do so, we stratified the data by sex and re-ran the Wilcoxon test for the 450 significant genes identified in the combined analysis. Of these, 292 genes remained significant in males and 314 in females. The two sets overlapped by 169 genes, of which 168 were expressed in the same direction (Table S3; Figure S5A), overall indicating high concordance between sex groups. NE exposure led to a total of 472 DEGs with an absolute LFC greater than 0.25. Strikingly, 451 of these genes were downregulated, while only 21 were upregulated (Figure 3C; Table S4). We validated several top DEGs (RUN4-2, CXCL1, and IL1B) via qPCR, and although these results did not reach statistical significance, they followed a similar expression pattern as our scRNA-seq findings (Figure 3D). Of the 472 NE DEGs, 313 remained significant in males and 358 in females, with 201 genes overlapping and expressed in the same direction in both sexes (Figure S5B). Results from sex-stratified analysis are available in Table S5. Interestingly, CORT and NE treatments shared 100 common DEGs, of which 62 were differentially regulated in the same direction and 38 regulated in opposite directions (Figure 3E). CORT and NE treatments lead to cell-type-specific gene expression changes (A) Volcano plot showing bulk gene expression changes due to CORT treatment. Green dots represent significant genes with an absolute log fold change (LFC) greater than 0.25 and an FDR-corrected p value (q) of less than 0.05. The top 10 genes, based on LFC, are labeled. (B) qPCR validation of gene expression changes due to CORT treatment. Statistical significance was determined by two-sided Student’s t test. ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, and ∗∗∗∗p < 0.0001. (C) Volcano plot showing bulk gene expression changes due to NE treatment. Purple dots indicate significant genes with an absolute LFC greater than 0.25 and a q less than 0.05. The top 10 genes according to LFC are labeled. (D) qPCR validation of gene expression changes due to NE treatment. Statistical significance was determined by two-sided Student’s t test. (E) Double volcano plot showing differentially expressed genes (DEGs) in response to CORT (x axis) and NE (y axis) treatments. Genes significantly regulated by CORT are highlighted in green, those by NE in purple, and those significantly affected by both treatments are shown in orange. Significance was determined using an FDR-corrected p value threshold of 0.05 and an absolute fold change cutoff of |0.25|. (F) Heatmap of the top 10 DEGs based on LFC, categorized by cell type due to CORT exposure. (G) Heatmap of the top 10 DEGs based on LFC, categorized by cell type due to NE exposure. See also Figure S5–S7 and Tables S2, S3, S4, and S5. We next examined whether the expression patterns of the top 10 DEGs due to CORT and NE treatment vary across different immune cell types. Notably, CORT induced both shared and distinct transcriptional changes across cell types (Figure 3F). For instance, VSIG4 was markedly upregulated in progenitor cells (LFC = 24.4), while it was strongly downregulated in plasmacytoid dendritic cells (LFC = −23.4). Among all DEGs, FKBP5 was the only gene consistently upregulated across all cell types due to CORT treatment (with an absolute LFC > 0.25). Although not part of the top 10 DEGs, RNU4-2 showed consistent upregulation across most cell types except plasmacytoid dendritic cells, while CCND3 exhibited consistent upregulation across all cell types except plasmacytoid dendritic and progenitor cells. Similar to CORT, NE treatment also induced cell-type-specific gene expression patterns. For example, CXCL3 was strongly upregulated in neutrophils (LFC = 23.6) and plasmacytoid dendritic cells (LFC = 26.5), whereas it was markedly downregulated in memory CD8+ T cells (LFC = −19.9) and plasma B cells (LFC = −25.6) (Figure 3G). Among all DEGs, we identified RNU4-2 as the single gene consistently significantly upregulated in the same direction across all immune cell types due to NE treatment (with an absolute LFC > 0.25). Given the lack of a benchmark gene marker for assessing NE signaling, our analyses for the first time identify RNU4-2 as a strong candidate for further investigation in immune cells. When examining the top DEGs in a sex-stratified manner, we observed differences in expression across cell types. Notably, following CORT exposure, IL1B was broadly upregulated across most cell types in females, whereas it showed comparatively lower or reduced expression in males (Figures S6A and S6B). Despite these differences, the overall transcriptional responses remained largely cell-type specific, indicating that while the magnitude and identity of top-responsive genes vary by sex, the underlying cell-type-specific structure of the response is preserved. This study provides evidence that stress hormones have immune cell-type-specific effects on gene expression. These results may help explain the inconsistencies observed in the literature regarding the effects of stress hormones on immune function,16,18,33,34 as prior studies were conducted at the bulk level. Sex-specific pathway remodeling and divergent immune responses to CORT and NE across cell types To further investigate hormone-driven changes in the function of specific cell types, we performed pathway analysis across major immune cell populations. Specifically, we grouped together CD8+ T cells (naive and memory), CD4+ T cells (Th17, naive, and memory), B cells (naive, memory, and plasma B cells), Monos/Macs (CD74+, CD163+, VCAN+, and IL7R+ MonoMacs), dendritic cells (plasmacytoid and myeloid), natural killer cells, and neutrophils. This grouping strategy was implemented to increase statistical power and improve the robustness of pathway enrichment analyses, particularly for rarer subpopulations with limited cell numbers when analyzed individually. CORT exposure was associated with significant upregulation of several pathways, including upregulation of inflammatory response pathways in MonoMacs, regulation of cell adhesion in CD4+ T cells, response to stimulus in CD8+ T cells, and negative regulation of cellular processes in B cells (Figure S7A; Table S6). In contrast, pathways associated with significant downregulation include response to stimulus in MonoMacs, positive regulation of cellular processes in CD4+ T cells, antigen processing and presentation of peptide antigen via MHC class Ib in both CD8+ and CD4+ T cells, lymphocyte and leukocyte activation in B cells, and positive regulation of leukocyte-mediated cytotoxicity in natural killer cells (Figure S7B; Table S6). We next stratified analyses by sex group and found that CORT induces widespread sex-specific differences in immune pathway regulation across both innate and adaptive cell types (Tables S7 and S8). While some shared responses occur, males and females frequently showed opposite or divergent pathway regulation, suggesting fundamentally different immune remodeling strategies between sex groups. Males tended to exhibit broader suppression of immune activation and signaling programs, reflecting a more restrained or stimulus-focused response. In contrast, females displayed more expansive and coordinated changes across immune, metabolic, developmental, and regulatory pathways, suggesting a more integrative and system-wide remodeling of immune function. NE exposure was associated with significant pathways in MonoMacs, including upregulation of inflammatory response and response to lipopolysaccharide (LPS) pathways, and significant downregulation of positive regulation of cellular processes. NE exposure was also associated with downregulated pathways in CD4+ T cells, including the intracellular signaling cascade and cell cycle process. No other cell types exhibited significant pathway associations after multiple-testing correction (Figures S7C and S7D). Detailed pathway analysis per major cell type is available in Table S9. NE elicited both shared and sex-divergent pathway responses, with some common activation programs but widespread suppression of core immune functions in both sexes (Tables S10 and S11). The direction and breadth of pathway regulation diverged between males and females. Again, males showed a more limited and stimulus-focused pattern characterized by selective enhancement of signaling and communication pathways. In contrast, females displayed a broad and coordinated activation profile spanning immune, developmental, and stress-response programs, indicating a more expansive remodeling of cellular function. Innate immune cell types display distinct inflammatory responses when exposed to CORT and NE Given the interest in characterizing the effects of stress on disease-related inflammation, we analyzed how CORT and NE treatments influence differential inflammatory gene expression across immune cell types. To achieve this, we utilized AUCell, a tool designed to assess whether a specific input gene set is enriched within the expression profile of each cell.35 For our analysis, we used the Hallmark Inflammatory Response gene set, a curated collection of 200 genes derived from meta-analysis of multiple studies and optimized for minimal redundancy. We selected this set due to its robust performance across diverse inflammatory contexts and its compatibility with single-cell enrichment tools.36 Our findings revealed that cells exposed to stress hormones exhibited cell-type-specific differential expression of inflammation-related genes, with MonoMac cell clusters demonstrating the most pronounced changes (Figure 4A). To quantify these differences, we compared the AUCell scores for each cell type between the control and treatment groups. Following CORT exposure, neutrophils, IL7R+ MonoMacs, and CD163+ MonoMacs showed significant enrichment of inflammation-related gene expression, with CD163+ MonoMacs exhibiting the most pronounced change (Figure 4B). Similarly, NE treatment resulted in inflammation-related gene enrichment in IL7R+ MonoMacs and CD163+ MonoMacs, with CD163+ MonoMacs again showing the greatest change (Figure 4C). Interestingly, in both treatment conditions, we observed a significant downregulation of inflammation-related genes in VCAN+ MonoMacs (Figures 4A and 4B). This highlights how the effects of stress hormones can be contrasting even between closely lineage-related cell types, such as those in the MonoMac compartment, emphasizing the importance of cell-type-specific analysis. Of note, the significant increase in inflammatory gene expression in CD163+ cells exposed to stress hormones persisted after harmonization by donor, indicating that this effect is not driven by donor-specific differences (Figures S8C and S8D, Bonferroni-corrected t test p < 0.05). CD163+ MonoMacs show a distinct inflammatory response to stress hormone treatment (A) Heatmaps of UMAPs by treatment. Heatmap shows individual cell’s AUCell score depicting inflammatory gene expression enrichment. (B) Quantification of the difference in AUCell enrichment score between CORT and control (n = 8 donors per condition) using a two-sided Student’s t test. Bonferroni-corrected p values: ∗p < 0.05, ∗∗p < 0.01, and ∗∗∗p < 0.001. (C) Quantification of the difference in AUCell enrichment score between NE and control (n = 8 donors per condition) using a two-sided Student’s t test. Bonferroni-corrected p values: ∗∗∗p < 0.001. (D) Volcano plot showing gene expression changes due to CORT treatment in CD163+ MonoMacs. Green dots represent significant genes with an absolute log fold change (LFC) greater than 0.25. The top 10 genes, based on LFC, are labeled. (E) Volcano plot showing gene expression changes due to NE treatment in CD163+ MonoMacs. Purple dots represent significant genes with an absolute LFC greater than 0.25. The top 10 genes, based on LFC, are labeled. (F) Double volcano plot showing differentially expressed genes (DEGs) in response to CORT (x axis) and NE (y axis) treatments in CD163+ MonoMacs. Genes significantly regulated by CORT are highlighted in green, those by NE in purple, and those significantly affected by both treatments are shown in orange. Significance was determined using an FDR-corrected p value threshold of 0.05 and an absolute fold change cutoff of |0.25|. See also Figures S8 and S9 and Table S6, S7, S8, S9, S10, S11, S12, and S13. Characterization and validation of CD163+ MonoMacs as a novel stress-activated innate immune cell subtype Given the observed robust inflammatory signature enrichment in CD163+ MonoMacs (F-statistic: 75.45, p = 4.02 × 10−33, Figure S9), we further characterized the stress-driven DEGs specifically in this cell type. CORT exposure resulted in a total of 477 DEGs with an absolute LFC greater than 0.25, of which 218 were downregulated and 259 upregulated (Table S12). Among the upregulated genes, we identified several well-known pro-inflammatory genes, such as IL6 (LFC = 27.1), IL1A (LFC = 4.0), and CCL4 (LFC = 3.3) (Figure 4D). NE treatment resulted in 319 DEGs, with 209 downregulated and 110 upregulated (Table S13). Similar to CORT exposure, we observed the upregulation of well-established inflammatory cytokines and chemokines, including CXCL1 (LFC = 2.5), IL1B (LFC = 3.2), and CXCL8 (LFC = 2.4) (Figure 4E). These genes—particularly IL6, IL1B, and the chemokines CXCL1 and CXCL8—are central mediators of inflammation, and their dysregulation has been implicated in the pathogenesis of various inflammatory and autoimmune diseases, highlighting the potential impact of stress-induced immune modulation. Overall, 149 genes were significantly differentially expressed in both treatment groups, of which 131 were regulated in the same direction and 18 in opposite directions (Figure 4F). Notably, the top genes upregulated in both groups include IL1B, CXCL1, CCL3, CXCL2, and CXCL3, all recognized as central inflammatory mediators. To validate the enrichment of inflammatory gene expression in CD163+ cells following CORT exposure, we performed RNAScope analysis, looking specifically for cells expressing CD163 that also express IL6 and IL1A. We observed an increase in the number of CD163+ cells expressing IL6, as well as a marked increase in IL6 expression within IL6+ CD163+ cells (Figures 5A and 5B). Additionally, while the proportion of CD163+ cells expressing IL1A did not change, there was a suggestive pattern (p = 0.08) toward increased IL1A expression within these cells (Figures 5A and 5B). To further assess whether our transcriptomic findings translate into phenotypic outcomes, we performed flow cytometry on CD163+ MonoMacs following CORT exposure. Cells from an independent donor were treated with CORT for 48 h to allow for sufficient protein expression, then analyzed by gating on CD14+/CD163+ populations and quantifying IL-1A+ or IL-6+ cells. The number of IL-1A+ cells was too low to draw definitive conclusions (Figures 5C and S10); however, consistent with our scRNA-seq results, CORT treatment increased IL-6 expression in CD163+ MonoMacs (Figures 5D and S10). CD163+ inflammatory signatures are validated at the RNA and protein levels (A) Representative images of RNAScope analysis. Cells were selected for CD163+ expression (green), then IL6 (red) and IL1A (magenta) puncta were analyzed. Scale bars represent 2 μm. (B) Image quantifications of RNAScope analysis. Significance was established using a two-sided Student’s t test. ∗p < 0.05, ∗∗p < 0.01, and ∗∗∗p < 0.001. (C) Flow cytometry analysis of CD163+ cells co-expressing IL-1A. A very small proportion of cells were IL-1A+, limiting comparison between control and treatment groups. A two-sided Student’s t test found no significant difference between groups (t = 0.08, p = 0.94). (D) Flow cytometry analysis of CD163+ cells co-expressing IL-6. A two-sided Student’s t test reveals that CORT exposure increases IL-6 protein expression in CD163+ cells (t = 4.96, p = 0.003). (E) CD163+ MonoMacs significantly upregulate IL6 expression when isolated from other PBMCs (left, p = 0.049. Although the effect does not reach statistical significance after 2 days of rest, the observed trend suggests that the response is sustained (right, p = 0.057). Statistical significance was assessed using a mixed-effects model, with IL6 expression as the dependent variable, treatment as the independent variable, and donors (n = 2, three replicates per donor and condition) included as fixed effects. We next asked whether these inflammatory effects arise from direct interactions between CORT and CD163+ cells or are mediated through cell-cell communication. To address this, we magnetically isolated CD163+ cells from two additional donors and cultured them with DMSO or CORT. Similar to the flow cytometry experiment, this further validates the gene expression signature in cells isolated using a protein marker (confirming the scRNA-seq-based clustering). After overnight treatment, CD163+ cells exhibited significantly higher IL6 expression compared to controls (p = 0.049, Figure 5E). Notably, following washout of the treatment and an additional 2-day rest period, IL-6 expression remained elevated, although this did not reach statistical significance (p = 0.06, Figure 5E). These findings are consistent with a direct and sustained pro-inflammatory effect of stress hormone exposure on CD163+ cells. Taken together, these analyses identify CD163+ MonoMacs as a novel stress-activated innate immune cell subtype and further characterize distinct stress-responsive, disease-relevant gene signatures within this cell subtype. CORT and NE exposure elicited distinct yet overlapping pathway-level responses in CD163+ monocytes. Both treatments showed enrichment of upregulated pathways related to cellular activation, including responses to stimuli, signaling processes, and inflammatory pathways, consistent with an overall activated transcriptional state. CORT preferentially upregulated pathways involved in cellular and chemical stimulus responses and intracellular signaling, while downregulated pathways were enriched for immune defense processes, including responses to biotic stimuli (Figure 6A). In contrast, NE upregulated pathways associated with cytokine signaling, responses to LPS and bacterial molecules, and metabolic processes, whereas downregulated pathways were enriched for immune system processes, antigen presentation, and broader stimulus-response pathways (Figure 6B). Together, these findings suggest that while both CORT and NE promote activation-related transcriptional programs in CD163+ monocytes, they differentially shape downstream immune and inflammatory signaling. CD163+ MonoMacs show increased activation-related pathways in response to stress hormones (A) Top 15 upregulated (top) and downregulated (bottom) pathways due to CORT exposure in CD163+ MonoMacs. (B) Top 15 upregulated (top) and downregulated (bottom) pathways due to NE exposure in CD163+ MonoMacs. See also Figure S10. CD163+ MonoMac-specific stress-inflammation transcriptional risk score distinguishes disease phenotypes To explore the potential clinical and translational relevance of our findings, we sought to develop a transcriptional risk score (TRS) informed by our cell-type-level findings. By integrating the cumulative effects of multiple genomic sites, composite TRS markers are thought to capture risk for complex diseases better than single risk genes.37,38 Building on this framework, we developed a cell-type-specific (CD163+) stress-inflammation score (hereafter denoted CytSIS). This parsimonious score was generated by identifying the CORT and NE DEGs (including up- and downregulated genes) in CD163+ cells that overlapped with the hallmark inflammatory gene set (n = 10 genes), weighting each gene’s expression by its treatment effect size (summed regression coefficient across CORT and NE treatments), and summing the resultant weighted gene expression products (Figures 7A and 7B; Table S14). Notably, identified genes included TLR2, IL1B, and CCL2, all of which are well-established mediators of innate immune activation. Application of CytSIS to our scRNA-seq dataset demonstrated that stress hormone exposure robustly increased CytSIS in CD163+ cells (F-value = 248.25 p = 1.18 × 10−104) (Figure 7C), as expected. This effect did not change when repeating analyses with our harmonized dataset (F-value = 248.25, p = 1.16 × 10−104; Figure S8E). Cell-type-specific (CD163+) stress-inflammation score and its application in stress-related diseases (A) Overview of the methodology for generating the transcriptional risk score (TRS) CytSIS. The score was developed using CD163+ MonoMac inflammatory gene expression and applied across different single-cell RNA sequencing datasets, including those investigating stress-associated diseases. (B) Identification of 10 overlapping genes present in both the CORT and NE DEGs and the hallmark inflammatory gene set. These genes were used to generate CytSIS (Table S5). (C) Application of CytSIS to our dataset shows significantly higher CytSISs in stress hormone-treated groups compared to control CD163+ cells, as expected. Statistical significance was determined by a one-way ANOVA and Tukey post hoc test. ∗∗∗∗p < 0.0001. (D) Application of CytSIS to a cohort of HIV-infected individuals, categorized by cardiovascular disease (CVD) status. CytSIS is significantly higher in individuals with CVD compared to those without CVD, with scores decreasing in individuals with CVD who were treated with statins. Statistical significance was determined by a one-way ANOVA and Tukey post hoc test. ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, and ∗∗∗∗p < 0.0001. (E) 10,000 permutations were conducted, generating TRSs by randomly drawing the same number of genes (n = 3) from the hallmark inflammatory genes and weighing them by the same treatment effect sizes. For each permutation, the TRS of CD163+ cells was then compared between the CVD and non-CVD groups. CytSIS significantly outperformed the random permutations. (F) Application of CytSIS to a cohort of individuals with rheumatoid arthritis, grouped based on symptom severity (remission or symptomatic). CytSIS was significantly elevated in the symptomatic group compared to the remission group. Statistical significance was determined by a Student’s t test. ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, and ∗∗∗∗p < 0.0001. (G) 10,000 permutations were conducted, generating TRSs by randomly drawing the same number of genes (n = 10) from the hallmark inflammatory genes and weighing them by the same treatment effect sizes. For each permutation, the TRS of CD163+ cells was then compared between the remission and symptomatic arthritis groups. CytSIS significantly outperformed the random permutations. See also Figure S11 and Table S14. To provide preliminary validation in an in vivo stress context, we applied CytSIS to a publicly available RNA-seq dataset of Ly6Clow monocytes in mice undergoing chronic social defeat stress (GEO: GSE202662; n = 4 controls, n = 12 stressed). Although murine monocyte markers do not directly map onto human subset definitions, Ly6Clow monocytes are widely considered the murine counterpart of human non-classical monocytes, a population often enriched for CD163 expression.39 Consistent with this, CD163 expression was significantly higher in Ly6Clow monocytes compared to Ly6Chigh monocytes in this cohort (t = 2.21, p = 0.03). We observed a trend toward increased CytSIS in stressed mice; however, this did not reach statistical significance (t = 1.25, p = 0.27), most likely reflecting both the limited sample size and the challenges of cross-species translation. To assess the applicability of CytSIS to CVD, a known stress-related condition,2 we next applied it in an scRNA-seq PBMC dataset from a cohort of HIV-infected individuals, classified into three groups: those with no CVD, those with untreated CVD, and those with CVD treated with statins (GEO: GSE205320).40 The analysis focused on CD163+ cells (Figure S11A). CytSIS was increased in the CD163+ cells of the CVD group as compared to the non-CVD and statin-treated CVD groups (F-value = 17.54, p = 3.26 × 10−08; Figure 6D). To confirm that this difference is driven specifically by the transcriptional effects of stress and is not due to non-specific differences in inflammation-related gene expression across groups, we conducted 10,000 permutations whereby we generated TRSs by randomly drawing genes from the hallmark inflammatory gene list expressed in this cohort (n = 4) and weighing them by the scRNA-seq-trained effect sizes. For each permutation, the TRS was then compared between the CVD and non-CVD groups. CytSIS significantly outperformed the random permutations (pperm < 0.0001) (Figure 7E). We observed similar findings when drawing from genes in the whole transcriptome (Figure S11B). We also applied CytSIS to a cohort of individuals with rheumatoid arthritis, another stress-related condition,41 who were grouped based on their symptoms into remitted or symptomatic arthritis (CellxGene Discover: https://cellxgene.cziscience.com/collections/e1a9ca56-f2ee-435d-980a-4f49ab7a952b).42 In CD163+-expressing cells (Figure S11C), CytSIS was significantly elevated in the symptomatic compared to the remitted group (T-statistic = 5.31, p = 1.36 × 10−6) (Figure 5F). We again performed 10,000 permutations of randomly generated weighted inflammatory TRSs (n genes = 10) that were compared between remitted vs. symptomatic groups. CytSIS again significantly outperformed the random permutations (pperm < 0.0001) (Figure 7G). We observed similar findings when drawing from genes in the whole transcriptome (Figure S11D). Importantly, these validation efforts rely on external disease and stress-related datasets rather than a directly measured stress-exposure human cohort and, therefore, should be interpreted as exploratory rather than definitive evidence of translational applicability. When investigating whether CytSIS translates to bulk RNA-seq cohorts from rodent psychosocial stress models, human CVD, and human rheumatoid arthritis, we found that CytSIS failed to differentiate between groups. This highlights that the score is cell-type specific and may not capture disease-relevant signals in heterogeneous bulk-tissue samples. Together, these findings suggest that cell-type-specific measures of stress-related inflammation may represent promising candidate markers of disease risk, but further validation, particularly in well-characterized stress-exposure human cohorts, is needed to establish their clinical and translational utility. Discussion Prior work underscores immune dysregulation as a potential key process linking stress with disease, but how stress hormones affect distinct immune cells and cell types has remained unknown. This study leveraged single-cell transcriptomic signatures to determine how the stress hormones CORT and NE influence immune function and the relevance of these transcriptomic signatures to stress-related conditions. Prior research shows a complex relationship between stress exposure, stress hormones, and immune dysregulation in vivo.43,44,45,46 Both human and rodent studies have shown that stress exposure induces widespread shifts in immune cell types.44,47,48 However, in vivo studies cannot disentangle whether these shifts reflect mere mobilization of particular cell types or reprogramming toward different immune cell identities. Using our ex vivo design that is devoid of the confounding effects of immune cell mobilization, we found that CORT exposure reduced plasmacytoid dendritic cells and CD8+ naive T cells (Figure 2B). The decrease in CD8+ T cells aligns with in vivo findings, where corticosterone administration in mice significantly reduced circulating CD8+ T cells, suggesting cell death or identity changes.49 While corticosterone also changes the proportions of monocytes, neutrophils, and B cells in vivo,49 the lack of significant changes in these populations in our model indicates these shifts are likely to be due to cell mobilization only. Similarly, NE exposure decreased memory B cells while increasing CD4+ T cells (nominally significant) (Figure 2C). This aligns with an in vivo study showing a trend toward increased CD4+ T cells after NE administration.49 This study also demonstrated a dramatic increase in neutrophils following NE exposure, attributable to cell mobilization.49 Mechanistically, these hormone-driven reductions in specific cell populations may reflect increased apoptosis—particularly given prior evidence that glucocorticoids can induce lymphocyte cell death50 or, alternatively, suppressed cell proliferation or survival. While our findings are consistent with these possibilities, our experimental design does not allow us to distinguish between these mechanisms, which will be important to address in future studies. Our compositional analysis also revealed that the MLR increases following CORT exposure. Elevated MLR has been reported in individuals experiencing high stress compared to those with low stress,43 suggesting that CORT may be a key driver of this pathology. Additionally, MLR has been studied as a marker of disease severity,31,33 supporting a plausible link between stress and disease. In this study, we demonstrate that stress hormones induce cell-type-specific transcriptomic changes, shedding new light that may help reconcile prior research supporting a range of pro- and anti-inflammatory effects of stress hormone exposure. For instance, some studies report that glucocorticoids reduce pro-inflammatory cytokines such as TNF and IL1B while increasing anti-inflammatory cytokines such as IL10.17 Conversely, other studies suggest that glucocorticoids can promote the secretion of pro-inflammatory cytokines such as IL1B and IL18.10 Similarly, conflicting evidence exists regarding the inflammatory effects of NE.51,52 Notably, our findings show that stress hormones can have opposite effects even within closely lineage-related cell subtypes (e.g., VCAN+ MonoMacs vs. CD163+ MonoMacs), highlighting the importance of studying distinct immune cell subpopulations in stress research. Nevertheless, we also identify a single gene differentially expressed consistently across all cell types for CORT (FKBP5) and another for NE (RNU4-2), suggesting that these genes could serve as cell-type-agnostic immune markers of stress exposure. Although the functional role of RNU4-2 in glucocorticoid or stress hormone signaling remains poorly characterized, its role in spliceosome assembly and pre-mRNA processing suggests that its consistent upregulation across immune cell types may reflect hormone-driven changes in RNA splicing dynamics. Prior work has shown that stress and glucocorticoids can influence alternative splicing and spliceosome activity across multiple systems,53 raising the possibility that alterations in small nuclear RNAs such as RNU4-2 may contribute to broader shifts in transcriptome processing. Our inflammation analysis revealed that the effects of stress hormones are cell-type-specific, with the potential to either increase or decrease inflammatory gene expression depending on the cellular context. Notably, blood-derived CD163+ MonoMacs exhibited a particularly pronounced increase in inflammation-related gene expression following exposure to both CORT and NE, underscoring their potential central role in stress-induced inflammatory responses. This finding is especially noteworthy given that CD163+ macrophages are traditionally characterized in the literature as anti-inflammatory, often described as alternatively activated macrophages.27,54 It has been well established that a group of monocytes highly express CD163, and this expression is further upregulated in response to glucocorticoids.55 However, CD163+ macrophages are also commonly observed in inflammatory and tumorigenic microenvironments, where they can contribute to disease-associated inflammation through mechanisms distinct from those of classically activated macrophages.56,57,58,59 One possible explanation for this apparent shift is that stress hormone signaling can alter transcriptional and chromatin states in myeloid cells in ways that favor the pro-inflammatory state. Stress has been shown to prime monocytes toward a hyperresponsive inflammatory phenotype through enduring transcriptomic and chromatin remodeling that heightens inflammatory signaling pathways (including mTOR/PI3K) and alters metabolic regulation, thereby sensitizing cells to inflammatory cues.43 In addition, glucocorticoids can have context-dependent effects: while generally considered anti-inflammatory, they can also increase expression of specific pro-inflammatory mediators by modulating pathways such as the NLRP3 inflammasome and allowing NF-κB activation in the presence of other stimuli, including cytokines and adrenergic signals.60 These mechanisms suggest that stress signaling can bias these cells toward a phenotype that retains CD163 expression while acquiring pro-inflammatory functionality. In particular, these cells have been implicated in pro-inflammatory and pro-atherogenic processes in atherosclerosis, where they may promote plaque progression and instability.57,58,59 Increased abundance of CD163+ macrophages has also been associated with inflammation and various disease phenotypes.61,62 Thus, our observation that CORT and NE enhance inflammation-related gene expression in circulating CD163+ MonoMacs may offer a mechanistic link between environmental stress and common conditions such as CVD. This hypothesis is further supported by the application of our TRS, CytSIS, in cohorts with stress-associated diseases, highlighting the potential role of stress-induced inflammation in disease pathogenesis. While we observe a robust inflammatory response in CD163+ MonoMacs following stress hormone exposure, the mechanisms by which these peripheral signals influence brain function and contribute to mental health disorders remain incompletely defined. Emerging evidence suggests that activated macrophages can interact with the blood-brain barrier and remodel it through the release of proteolytic enzymes, including matrix metalloproteinases, which degrade tight junction and extracellular matrix components and increase vascular permeability.63 In this context, stress-induced activation of CD163+ MonoMacs may facilitate immune-to-brain communication through both cytokine signaling and vascular remodeling, providing a potential mechanism linking peripheral inflammation to central nervous system dysfunction. Notably, the macrophage populations implicated in these processes share phenotypic features with CD163+ monocytes, supporting the relevance of this mechanism to our observed cell state.63 Our dataset provides a valuable resource for developing TRSs, which are increasingly being used as biomarkers for disease risk and progression.37,38 By leveraging cell-type-specific transcriptional responses to stress, we can create more precise tools for identifying individuals at risk of stress-associated diseases. For example, we demonstrated that CD163+ cell-specific transcriptional signatures effectively distinguish between healthy and diseased individuals. This approach provides a potential framework for leveraging single-cell datasets to promote early risk stratification and personalized intervention strategies. These findings may not only advance the development of disease biomarkers but also highlight potential therapeutic targets for mitigating the impact of stress and other environmental exposures on health. While our ex vivo exposures provide a controlled view of how stress hormones influence immune cells, the timing of these assays (4 h for neutrophils and overnight for PBMCs) represents only acute stress conditions and may not fully capture the dynamics of in vivo stress responses. Acute stress triggers rapid neuroendocrine activation that mobilizes leukocytes, redistributes immune subsets, and can transiently enhance innate inflammatory responses, as seen with short-term catecholamine and glucocorticoid signaling in circulating immune cells.16 In contrast, chronic stress engages sustained HPA and sympathetic activity that reshapes immune function over longer timescales, often leading to more profound transcriptional and chromatin remodeling, altered metabolic states, and skewed cytokine profiles that contribute to persistent inflammatory or immunosuppressive phenotypes.43 Because our design examines acute hormone exposures, the gene expression changes observed likely reflect the initial phase of stress hormone-driven immune activation rather than the full spectrum of adaptations that occur with repeated or chronic exposure in vivo. Nevertheless, identifying these early and sustained responses, particularly in CD163+ MonoMacs, reveals the intrinsic specific responsiveness of these cells to stress signals and highlights molecular pathways that chronic stress may repeatedly engage in, providing a basis for future studies to assess how transient responses evolve into persistent immune alterations in stress-related diseases. Limitations of the study A limitation of this study is that the experiments cannot definitively establish the directionality of the relationship between stress hormones and inflammation in human disease models. While our data demonstrate that CORT and NE can modulate inflammatory responses in specific immune cell types, it remains possible that inflammation itself contributes to altered stress hormone levels in vivo. Importantly, responses such as those in Figure 7D, including their modulation by statins, suggest that stress hormone-driven immune signatures may be pharmacologically tunable, providing a framework for future studies to dissect causal pathways and test interventions in vivo. Additionally, our ex vivo design does not capture the full dynamics of in vivo hormone fluctuations. This reductionist approach, however, is necessary to remove confounders present in human cohorts by exposing immune cells to physiologically relevant stress hormone concentrations in a controlled setting. Despite these simplifications, the transcriptional signatures we observed recapitulate patterns seen in disease cohorts, supporting translational relevance. Our study observed decreases in certain immune cell-type proportions following CORT and NE exposure. While prior studies have shown that glucocorticoids can promote apoptosis in specific immune cell populations (e.g., lymphocytes), which could contribute to these shifts, our experiments cannot definitively distinguish whether the changes reflect increased cell death, reduced proliferation, or other mechanisms. Dissecting the precise cellular processes underlying these alterations will be an important focus of future work. Resource availability Lead contact Requests for further information and resources should be directed to the lead contact, Anthony S. Zannas (anthony_zannas@med.unc.edu). Materials availability This study did not generate new unique reagents. Data and code availability Raw data have been deposited in GEO as GEO: GSE334840. Processed scRNA-seq data generated in this study have been deposited in Zenodo at https://doi.org/10.5281/zenodo.19787279. The analysis code is available at DOI: https://doi.org/10.5281/zenodo.21359854. Acknowledgments We thank Laura Huff, PhD, for editorial assistance; Rachel DuMez-Kornegay for designing Figures 1A, 5A, and 6; and J. Ashley Ezzell (UNC Histology Research Core) for histology support. We also thank the UNC Blood Donation Center for their assistance with donor recruitment. Microscopy was performed at the UNC Neuroscience Microscopy Core (RRID:SCR_019060), supported by NIH P50 HD103573. The Zeiss LSM 980 was funded by NIH S10 OD032388. This research was supported by the National Heart, Lung, and Blood Institute of the NIH under award number R01HL163031 (to A.S.Z.) and T32NS007431 (to S.B.). B.C.M. is supported by the Burroughs Wellcome Fund Career Award for Medical Scientists and NIH K08CA248960. Author contributions Conceptualization, S.B., O.K., and A.S.Z.; methodology, S.B., O.K., N.S., B.C.M., H.W., and A.G.B.; investigation, S.B. and O.K.; funding acquisition, A.S.Z.; project administration, S.B. and A.S.Z.; supervision, A.S.Z.; writing, S.B., O.K., and A.S.Z. Declaration of interests B.C.M. has consulted for Cellarity, LifeOmic, and Telix Pharmaceuticals. A.S.Z. has consulted for Techspert. STAR★Methods Key resources table REAGENT or RESOURCESOURCEIDENTIFIERAntibodiesPerCP/Cyanine5.5 anti-human CD14BioLegendRRID: AB_893252Alexa Fluor(R) 700 anti-human CD16BioLegendRRID: AB_493748PE/Dazzle(TM) 594 anti-human CD64BioLegendRRID: AB_2564185PE anti-human CD206 (MMR)BioLegendRRID: AB_571910Brilliant Violet 605(TM) anti-human CD163BioLegendRRID: AB_2562712Brilliant Violet 785(TM) anti-human CD68BioLegendRRID: AB_2800879PE/Cyanine7 anti-human IL-6BioLegendRRID: AB_2572041LIVE/DEAD™ Fixable Near IR (780)ThermoFisherCat#: L34994IL-1α Antibody, anti-human, APCMiltenyi BiotecCat#: 130-137-758CD163 Monoclonal Antibody (GHI/61), APCThermoFisherRRID: AB_2534503Biological samplesHuman peripheral blood (healthy donors, age 19–28)UNC Blood Donation CenterNot publicly availableChemicals, peptides, and recombinant proteinsTSA Plus Cyanine 5Akoya BiosciencesCat# NEL74500KTTSA Plus FluoresceinAkoya BiosciencesCat# NEL74100KTTSA Plus Cyanine 3Akoya BiosciencesCat# NEL74400KTFluorogel II with DAPIElectron Microscopy SciencesCat# 17985-50HistoGelThermoFisherCat# HG-4000-012eBioscience Intracellular Fixation & Permeabilization Buffer SeteBioscienceCat# 88-8824-00MonensineBioscienceCat# 00-4505-51Brefeldin A SolutioneBioscienceCat# 00-4506-51LymphoprepSTEMCELL TechnologiesCat# 18061RPMI 1640 Medium, no phenol redGibcoCat# 11835030DPBS, no calcium, no magnesiumGibcoCat# 14190144Fetal Bovine SerumN/ACortisol (Hydrocortisone)Sigma AldrichCat# H4001-5GNorepinephrineVWR LIFE SCIENCE SERADIGMCat# 89510-194DMSOSigma AldrichCat# 472301EDTAInvitrogenCat# AM9260GACK Lysing BufferGibcoCat# A1049201Serum, Type AB, Human, From Male DonorMP BiomedicalsCat# ICN2930949Critical commercial assaysEasySep™ Human APC Positive Selection Kit IISTEMCELL TechnologiesCat# 17621RNeasy Mini KitQiagenCat# 74106SuperScript IV VILO cDNA Synthesis KitInvitrogenCat# 11756500RNAscope Multiplex Fluorescent Detection KitACD BioCat# 323110Parse WT Kit (Single-cell RNA-seq)Parse BiosciencesN/AParse Fixation KitParse BiosciencesN/ADeposited dataRaw scRNA-seq dataset (this study)GEOGSE334840Processed scRNA-seq dataset (this study)Zenodohttps://doi.org/10.5281/zenodo.19787279CytSIS Code (this study)Zenodohttps://doi.org/10.5281/zenodo.21359854CVD cohort datasetGEOGSE205320Rheumatoid arthritis datasetCellxGene Discoverhttps://cellxgene.cziscience.com/collections/e1a9ca56-f2ee-435d-980a-4f49ab7a952bIn-vivo rodent stressGeoGSE202662OligonucleotidesqPCR primersSee Table S15N/AHs-CD3E probeAdvanced Cell Diagnostics (ACD Bio)Cat# 553971Hs-CD163 probeAdvanced Cell Diagnostics (ACD Bio)Cat# 417061Hs-NELL2 probeAdvanced Cell Diagnostics (ACD Bio)Cat# 1098201-C2Hs-IL6 probeAdvanced Cell Diagnostics (ACD Bio)Cat# 310371-C2Hs-CD8A probeAdvanced Cell Diagnostics (ACD Bio)Cat# 560391-C3Hs-IL1A probeAdvanced Cell Diagnostics (ACD Bio)Cat# 556791-C3Software and algorithmsPythonPython Software Foundationv3.9.25ScanpyOpen-sourcev1.9.3HarmonyOpen-sourceN/AFlowJoBD BiosciencesN/AImageJ (Cell Counter plugin)NIHN/AAUCellBioconductorv1.32.0 Experimental model and study participant details Human peripheral blood samples were obtained from healthy donors, aged 19–28, at the UNC Blood Donation Center with informed consent and IRB approval. Eight healthy volunteers participated in the scRNA-sequencing study and q-PCR experiments, and four independent donors in the RNAscope experiments. Method details Isolation of PBMCs and neutrophils PBMCs and polymorphonuclear neutrophils (PMNs) were isolated from human peripheral blood samples as previously described.64 12 mL of fresh peripheral blood was collected using vacutainer EDTA tubes, diluted with 12 mL of DPBS supplemented with EDTA (2mM) and FBS (1%), and then gently layered over Lymphoprep (STEMCELL Technologies, USA) in a 50 mL conical tube. Samples were centrifuged at 800g for 15 min at room temperature with breaks off. After centrifugation, PBMCs in the white foamy layer were carefully transferred to a new tube. After being washed with DPBS twice, samples were then centrifuged at 250g for 10 min, and the pellets were resuspended in cell culture media (RPMI supplemented with 5% human serum). Cell counts were performed on the automated cell counter, and 250.000 cells/well were seeded into 24-well plates, incubated at 37°C in a 5% CO2 incubator for 1 h, and then treated with 100 nM Cortisol, 10 nM Norepinephrine, or 0.0001% DMSO (vehicle control) overnight. Given their shorter lifespan in cell culture,65 PMNs were recovered from the red blood cells layer following 1 × ACK lysis buffer digestion procedure and washed twice with 1× PBS and resuspended in cell culture media. Cells were then seeded into 24-well plates at a seeding density of 250,000 cells/well, incubated at 37C in a 5% CO2 incubator for 1 h, and then treated with 100 nM cortisol, 10 nM norepinephrine, or 0.0001% DMSO for 4 h. CD163+ cell isolation and culture Peripheral blood mononuclear cells (PBMCs) were isolated from whole blood as described above. CD163+ cells were subsequently enriched using a magnetic separation approach according to the manufacturer’s instructions (STEMCELL Technologies, Cat# 17621), in combination with a CD163 antibody (Molecular Probes, Ref# A15723). Following isolation, CD163+ cells were plated and allowed to equilibrate prior to treatment. Cells were then exposed to vehicle control or CORT. For acute exposure experiments, cells were treated overnight and collected the following day for downstream analyses. To assess the persistence of treatment effects, a subset of cells underwent treatment for 24 h, after which the media was replaced to remove CORT, and cells were cultured for an additional 2 days prior to collection. Flow cytometry PBMCs were cultured for 48 h in supplemented RPMI with their respective treatments. For the last 16 h of incubation, cells were cultured with Golgi block consisting of 10.6uM Brefeldin A and 2uM Monensin (ThermoFisher). Cells were then transferred to tubes and incubated in Fc Block containing 1× Human TruStain FcX (BioLegend). Cells were then incubated with the extracellular antibodies (Table S15, key resources table). Cells were fixed with IC Fixation buffer (eBiosciences) and permeabilized with 1× Permeabilization Buffer (eBiosciences). Cells were then incubated in Fc Block once again before addition of intracellular antibodies (Table S15, key resources table) Samples were then run on a Symphony A3 flow cytometer (BD), and data was analyzed with FlowJo Software. RNA extraction and qPCR Total RNA was extracted with an RNeasy kit (Qiagen, cat. No 74106) from cells exposed to DMSO, cortisol, and norepinephrine. The Super Script IV VILO cDNA kit (Invitrogen, cat.no 11756500) was used to synthesize cDNA. mRNA levels of the following geneswere assessed by q-PCR: VSIG4, FKBP5, ZBTB16, RNU4-2, CXCL1, IL1B, IL6. Detailed information on Q-PRC primer can be found in Table S15. Reactions were performed with two replicates in a 96-well plate using the Quant Studio 6 (Applied Biosystems). The housekeeping gene GAPDH was used as a reference control. RNAScope Cell collection PBMCs were isolated from four healthy donors (two male, two female) and treated as described above. Following overnight incubation, cells were collected into centrifugation tubes and centrifuged at 300 × g for 5 min. The resulting cell pellets were resuspended in 10% neutral buffered formalin (NBF) and fixed for 15 min at room temperature. After fixation, cells were centrifuged again at 300 × g for 5 min. The supernatant was carefully removed, and the pellets were resuspended in 20–30 μL of pre-warmed HistoGel (Thermo Scientific, HG-4000-012). While still in liquid form, the cell-HistoGel mixture was pipetted onto a parafilm to form a small gel droplet encapsulating the PBMCs. The droplet was allowed to solidify and then immersed in 10% NBF for an additional hour to ensure thorough fixation. Subsequently, the gel droplet was transferred to 70% ethanol and stored until further processing. Histology Histogel cell pellets were processed using a Sakura Tissue-Tek VIP 5 Tissue Processor, embedding in paraffin, and sliced at 5um. In situ hybridization was performed using ACD Bio’s RNAScope technology per the kit instructions (Multiplex Fluorescent Detection Kit, ACD Bio, #323110). Briefly tissue sections were deparaffinized and hydrated, followed by antigen retrieval for 15 min at 100°C. After protease incubation for 30 min at 40°C, probes were hybridized for two hours a 40°C. Probe signal was amplified, then finally detected using TSA Plus fluorophores at 1:1500. Slides were coverslipped using Fluorogel II with DAPI (Electron Micrscopy Sciences, #17985-50). Probes can be found in key resources table. Imaging Imaging was performed using a Zeiss LSM 980 confocal microscope equipped with a 64× oil immersion objective. Quantification of RNA signals was conducted using the ImageJ Cell Counter plugin, enabling the enumeration of individual transcript puncta per cell. Single-cell RNA sequencing library preparation and sequencing Cells were removed from the incubator, and the plates were centrifuged at 500g for 5 min for pelleting. Cells were washed with 1× PBS and carefully detached from each well by gentle pipetting, then resuspended in 1× PBS to create a single cell suspension. Cells were fixed immediately using a Fixation Kit (Parse Biosciences), counted, and stored at −80°C as per the manufacturer’s protocol. The single-cell libraries were constructed according to the Parse WT kit manufacturer’s protocol. 4166 cells per sample (for a total of 100,000 cells from 24 samples) were barcoded using the reverse transcription barcoding approach. Barcoded cDNA was amplified and verified via TapeStation for size distribution following fragmentation, end repair, A-tailing, and adaptor ligation to complete library construction. All libraries were sequenced as paired-end 150-bp reads using the Illumina NovaSeq X Plus 300 platform with a reading depth of 83-104M reads per sample. Quantification and statistical analysis Preprocessing and quality control Raw sequencing data were processed using the Parse Biosciences split-pool pipeline according to the manufacturer’s instructions (Parse Biosciences). Reads were aligned to the human reference genome GRCh38, and UMI-collapsed gene–cell count matrices were generated using the Parse workflow. Downstream analyses were conducted in Python (v3.9.25) using Scanpy (v1.9.3). Cells expressing fewer than 300 genes and genes detected in fewer than 5 cells were excluded. Mitochondrial genes were identified based on the “MT-” prefix, and per-cell quality control metrics were computed. Cells with more than 8,000 detected genes, more than 20,000 total counts, or greater than 15% mitochondrial gene expression were removed to exclude low-quality cells and potential doublets (Figure S1). The remaining high-quality cells were retained for downstream analyses. Normalization and feature selection Gene expression counts were normalized per cell to a total of 10,000 counts using Scanpy’s normalize_total function and subsequently log-transformed using log1p. Highly variable genes were identified using Scanpy’s dispersion-based method with parameters min_mean = 0.0125, max_mean = 3, and min_disp = 0.25. Raw log-normalized expression values were stored prior to subsetting to highly variable genes. Downstream analyses were restricted to the highly variable gene set. Expression values were scaled to unit variance and zero mean with a maximum value of 10. Dimensionality reduction and clustering Principal component analysis (PCA) was performed on the scaled highly variable genes. The top 36 principal components were used to compute the neighborhood graph with 20 nearest neighbors. Clustering was conducted using the Leiden algorithm with a resolution parameter of 0.5. Two-dimensional visualization of the data was generated using Uniform Manifold Approximation and Projection (UMAP) with default parameters. No minimum cluster size threshold was applied. Batch correction and integration Primary analyses were conducted without batch correction. To assess potential donor-specific effects, a parallel integration analysis was performed using Harmony after PCA, harmonizing by donor identity. Clustering results following Harmony integration showed modest differences in cluster boundaries compared to the non-integrated analysis. Harmony-integrated results are presented in the supplemental information, whereas non-integrated data were used for the primary analyses. Differential gene expression analysis Differential gene expression analysis was performed at single-cell resolution using Scanpy’s rank_genes_groups function with the Wilcoxon rank-sum test. Cluster-specific marker genes were identified by comparing each Leiden cluster to all other cells. For treatment comparisons, cells were grouped by treatment condition and all cells within each condition were pooled across donors. Differential expression between treatment groups (CORT or NE) and control (DMSO) was performed using the Wilcoxon rank-sum test with use_raw = True. Analyses were conducted both across the full dataset and within individual cell types by subsetting cells by cluster prior to testing. p-values were adjusted for multiple testing using the Benjamini–Hochberg false discovery rate (FDR) correction. Genes were ranked based on adjusted p-values and log fold change, and a two-sided adjusted p-value < 0.05 and absolute log fold change > 0.25 was considered statistically significant. Compositional analysis For each sample, cell type frequencies were calculated as the proportion of cells assigned to a given cluster relative to the total number of cells in that sample. Differences in cell type composition between treatment and matched control samples were assessed using two-sided paired t-tests, pairing each donor’s treatment sample with its corresponding control sample. Bonferroni correction was applied to account for multiple comparisons, and an adjusted two-sided p-value <0.05 was considered statistically significant. The monocyte-to-lymphocyte ratio (MLR) was calculated by summing all clusters classified as monocytes (CD163+ MonoMac, CD74+ MonoMac, IL7R+ MonoMac, VCAN+ MonoMac, CD14− MonoMac) and dividing by the sum of clusters classified as lymphocytes (Naive CD4+ T, Memory CD4+ T, Naive CD8+ T, Memory CD8+ T, Th17, Naive B, Memory B, Plasma B). Statistical differences between treatment and control groups were assessed using paired two-sided t-tests with Bonferroni correction. CytSIS Development of the cell-type-specific Stress-Inflammation Score (CytSIS) Informed by our findings, we developed a transcriptional risk score (TRS) termed CytSIS. By integrating the cumulative effects of multiple genomic sites, this composite marker approach is thought to better capture risk for complex diseases as compared to single risk genes.37,38 CytSIS was generated by identifying the differentially expressed genes (DEGs) resulting from CORT and NE treatments in CD163+ MonoMacs that overlapped with the Hallmark Inflammatory Response gene set (n = 10 genes) (Table S14). For each DEG, the treatment effect size was calculated by summing the regression coefficients from CORT and NE treatments compared to the control, as determined through the Wilcoxon rank-sum test. Each gene was thus weighted by its treatment effect size, generating a score per cell. Application of CytSIS To apply the CytSIS score, we first created a subset with cells in the cohorts that expressed the CD163 gene (non-zero values). We then calculated the product of each individual gene’s weight (trained using our scRNA-seq treatment dataset) and its gene expression value for each cell (derived from the human cohort dataset). This calculation was performed only for the genes that were available in each dataset (4 genes for the CVD cohort and 10 genes for the arthritis cohort). The final CytSIS score for each cell was obtained by summing these products across all genes in the set, yielding a single score that reflects the relative stress-related inflammatory activity of the cell. External cohorts Human cohorts used for CytSIS testing were obtained from publicly available data repositories(GEO: GSE205320, CellxGene Discover: https://cellxgene.cziscience.com/collections/e1a9ca56-f2ee-435d-980a-4f49ab7a952b). Cardiovascular disease (CVD) All participants in this study were part of the Women’s Interagency HIV Study (WIHS).40 From the initial 1,865 participants in the WIHS vascular sub-study, 32 participants were selected for scRNA-seq analysis of PBMCs. The presence of carotid artery focal plaque was used to classify participants into four groups of eight participants: HIV−, HIV+CVD−, HIV+CVD+, and HIV+CVD+ on cholesterol-reducing treatment (CRT). For our analysis, we focused on the role of stress in CVD and, to rule out confounding by other disease factors, we excluded the HIV− group. Rheumatoid arthritis (RA) Participants from this study consist of patients meeting the American College of Rheumatology (ACR) classification criteria and were recruited from UCSF rheumatology between2016 and 2020.42 PBMCs were collected for scRNA-seq. The disease Activity in 28 joints using CRP (DAS28-CRP) was used to stratify the patients in remitted (n = 2) or symptomatic (n = 14) groups with low or moderate disease activity according to the 2019 updated ACR recommendation on disease activity measures.66 Published: August 13, 2026 Footnotes Supplemental information can be found online at https://doi.org/10.1016/j.xgen.2026.101330. Supplemental information Document S1. Figures S1–S11 and Tables S1, S14, and S15 Table S2. Results of the Wilcoxon rank-sum test comparing cortisol-treated and control cells, related to Figure 3 Table S3. Results of the Wilcoxon rank-sum test comparing cortisol-treated and control cells in a sex-stratified manner (tab S3A for males, tab S3B for females), related to Figure 3 Table S4. Results of the Wilcoxon rank-sum test comparing norepinephrine-treated and control cells, related to Figure 3 Table S5. Results of the Wilcoxon rank-sum test comparing norepinephrine-treated and control cells in a sex-stratified manner (tab S5A for males, tab S5B for females), related to Figure 3 Table S6. Pathway analysis of major immune cell types following cortisol exposure, related to Figure 4 Table S7. Pathway analysis of major immune cell types following cortisol exposure in males only, related to Figure 4 Table S8. Pathway analysis of major immune cell types following cortisol exposure in females only, related to Figure 4 Table S9. Pathway analysis of major immune cell types following norepinephrine exposure, related to Figure 4 Table S10. Pathway analysis of major immune cell types following norepinephrine exposure in males only, related to Figure 4 Table S11. Pathway analysis of major immune cell types following norepinephrine exposure in females only, related to Figure 4 Table S12. Results of the Wilcoxon rank-sum test comparing cortisol-treated and control CD163+ cells, related to Figure 4 Table S13. Results of the Wilcoxon rank-sum test comparing norepinephrine-treated and control CD163+ cells, related to Figure 4 Document S2. Article plus supplemental information References Associated Data Supplementary Materials Document S1. Figures S1–S11 and Tables S1, S14, and S15 Table S2. Results of the Wilcoxon rank-sum test comparing cortisol-treated and control cells, related to Figure 3 Table S3. Results of the Wilcoxon rank-sum test comparing cortisol-treated and control cells in a sex-stratified manner (tab S3A for males, tab S3B for females), related to Figure 3 Table S4. Results of the Wilcoxon rank-sum test comparing norepinephrine-treated and control cells, related to Figure 3 Table S5. Results of the Wilcoxon rank-sum test comparing norepinephrine-treated and control cells in a sex-stratified manner (tab S5A for males, tab S5B for females), related to Figure 3 Table S6. Pathway analysis of major immune cell types following cortisol exposure, related to Figure 4 Table S7. Pathway analysis of major immune cell types following cortisol exposure in males only, related to Figure 4 Table S8. Pathway analysis of major immune cell types following cortisol exposure in females only, related to Figure 4 Table S9. Pathway analysis of major immune cell types following norepinephrine exposure, related to Figure 4 Table S10. Pathway analysis of major immune cell types following norepinephrine exposure in males only, related to Figure 4 Table S11. Pathway analysis of major immune cell types following norepinephrine exposure in females only, related to Figure 4 Table S12. Results of the Wilcoxon rank-sum test comparing cortisol-treated and control CD163+ cells, related to Figure 4 Table S13. Results of the Wilcoxon rank-sum test comparing norepinephrine-treated and control CD163+ cells, related to Figure 4 Document S2. Article plus supplemental information Data Availability Statement Raw data have been deposited in GEO as GEO: GSE334840. Processed scRNA-seq data generated in this study have been deposited in Zenodo at https://doi.org/10.5281/zenodo.19787279. The analysis code is available at DOI: https://doi.org/10.5281/zenodo.21359854.