JAK-STAT signaling maintains homeostasis in T cells and macrophages

Abstract

Immune cells need to sustain a state of constant alertness over a lifetime. Yet, little is known about the regulatory processes that control the fluent and fragile balance that is called homeostasis. Here we demonstrate that JAK-STAT signaling, beyond its role in immune responses, is a major regulator of immune cell homeostasis. We investigated JAK-STAT-mediated transcription and chromatin accessibility across 12 mouse models, including knockouts of all STAT transcription factors and of the TYK2 kinase. Baseline JAK-STAT signaling was detected in CD8⁺ T cells and macrophages of unperturbed mice-but abrogated in the knockouts and in unstimulated immune cells deprived of their normal tissue context. We observed diverse gene-regulatory programs, including effects of STAT2 and IRF9 that were independent of STAT1. In summary, our large-scale dataset and integrative analysis of JAK-STAT mutant and wild-type mice uncovered a crucial role of JAK-STAT signaling in unstimulated immune cells, where it contributes to a poised epigenetic and transcriptional state and helps prepare these cells for rapid response to immune stimuli.

Fig. 1. Transcriptome effects of JAK-STAT mutants in homeostasis.

a, Outline of the experimental approach for dissecting the gene-regulatory landscape of JAK-STAT signaling under homeostatic conditions. b, Bar plots showing the number of differentially expressed genes (at a 5% FDR cutoff and FC greater than 2) between JAK-STAT mutant and wild-type mice in five immune cell types. c, Gene expression for selected IFN response genes in immune cells from JAK-STAT mutant and wild-type mice. Bar plots display the mean and standard error. d, Similarity of transcriptional effects of JAK-STAT mutant mice in T cells and macrophages, based on multi-dimensional scaling (MDS) of Spearman correlation coefficients among log₂FCs compared with wild-type mice. Results for all cell types are shown in Extended Data Fig. 2d,e. FDR, false discovery rate; log₂FC, log₂ fold change.

Fig. 2. Gene-regulatory modules underlying JAK-STAT signaling in homeostasis.

a, Outline of the analytical approach for identifying JAK-STAT gene-regulatory modules. b, Similarity of genes in terms of their differential expression patterns across JAK-STAT mutants, based on a UMAP of log₂FCs between JAK-STAT mutant and matched wild-type samples. This UMAP places genes with similar effects of JAK-STAT mutants on their transcriptome in proximity. It includes all genes with a twofold or greater change in gene expression for at least one mutant, and it places them in 16 gene clusters marked by letters A to P. c, Overlay of mutant-specific differential expression (with color-coded log₂FCs) on the gene UMAP from b. d, Dot plot showing the average log₂FC across all genes in the clusters from b, for two cell types and 12 JAK-STAT mice. e, Dot plot showing gene set enrichment for the gene clusters from b (two-sided Fisher’s exact test, corrected for multiple comparisons). The four most enriched gene sets are shown for each cluster. OR, odds ratio.

Fig. 3. Characteristic roles of JAK-STAT signaling in homeostasis.

a, Differential gene expression for IRF9 and STAT2 knockouts. Left, scatterplot of log₂FCs for the two knockouts relative to matched wild-type samples. Right, bar plots of differential expression levels relative to wild type for selected genes, displaying mean and standard error. b, Differential gene expression for STAT5 modulation and IL-2 treatment in T cells. Left, scatterplot of log₂FCs for the hyperactivated STAT5B^N642H mutant (STAT5-hyp) and STAT5 knockout relative to wild type, and for the response of wild-type T cell to in vitro IL-2 treatment at two time points. Right, gene set enrichment analysis for the differentially expressed genes (two-sided Fisher’s exact test, corrected for multiple comparisons). Upregulation, downregulation and no change are indicated by ‘+’, ‘−’ and ‘o’, respectively. c, Differential gene expression for STAT1 isoforms. Left, scatterplot of log₂FCs for the two STAT1 isoforms and for the full STAT1 knockout. Right, box plots showing STAT1 isoform effects (log₂FC) for genes with significant STAT1 effect in macrophages, grouped by the effects of full STAT1 and STAT2 knockouts. Upregulation, downregulation and no change are indicated by a ‘+’, ‘−’ and ‘o’, respectively. Box plots show the full data range, with the box indicating interquartile range and median. Bottom, bar plots showing the expression levels of selected genes affected by these mutants. d, Differential gene expression for TYK2 modulation. Left, scatterplot of log₂FCs for the TYK2 knockout and the kinase-dead TYK2^K923E mutant. Right, TYK2 mutant effects on selected IL-12 regulated genes (two-sided linear mixed models, corrected for multiple comparisons). Bottom, bar plots showing the expression levels of selected genes affected by these mutants. Mac, macrophage; r, Spearman correlation coefficient. Bar plots display the mean and standard error.

Fig. 4. In vivo validation of baseline JAK-STAT signaling in homeostasis.

a, Spatial transcriptomics profiles of spleens from wild-type and STAT1 knockout mice, shown for samples collected after in vivo cell fixation using formaldehyde. First row: hematoxylin and eosin (H&E) stains highlighting the anatomical structures of the spleen. Second row: spatial transcriptomics profiles annotated with gene expression clusters. Third and fourth row: expression levels of the T cell marker gene Cd8a and the macrophage marker gene Cd33 in the spatial transcriptomics data (scale bars, 1 mm). b, Violin plots showing the expression of STAT1-driven genes (top-15 downregulated genes comparing STAT1 knockout and wild type based on the RNA-seq data) and housekeeping genes (Actb, Hprt and Ubc) in the spatial transcriptomics data. c, Violin plots showing the expression of the ISGs Oas3, Ifit3 and Ifit1 in Cluster 4 of the spatial transcriptomics data. d, Representative RNA-FISH images for the ISG Oas3 (yellow) and the T cell marker gene Cd3e (dark blue) in spleen samples from wild-type and STAT1 knockout mice (scale bar, 50 µm). Autofluorescence of the red pulp is visible in magenta. e, Representative RNA-FISH images for the ISG Ifit3 (yellow) and the T cell marker Cd3e (dark blue) in spleen samples from wild-type and STAT1 knockout mice (scale bar, 50 µm). Experiments comprised two mice (a–c) or three mice (d and e) as biological replicates. Box plots (b and c) show the full data range, with the box indicating the interquartile range and median.

Fig. 5. Epigenome effects of JAK-STAT mutants in homeostasis.

a, Outline of the analysis dissecting JAK-STAT modulation of the epigenome (based on ATAC-seq data) and transcriptome (based on RNA-seq data). b, Genome browser tracks showing chromatin accessibility profiles for the promoter regions of the Stat5a gene, the macrophage marker gene Cd14 and the T cell marker gene Cd28. c, Transcription factor footprinting analysis, showing differential chromatin accessibility footprints for certain JAK-STAT mutants and cell types. Shaded areas indicate the standard error. d, Bar plots showing the percentage of genes and genomic regions affected by transcriptome and epigenome changes upon JAK-STAT modulation (relative to the number of all tested genes or genomic regions), as well as the Pearson correlation between transcriptome and epigenome changes (log₂FCs of gene expression versus chromatin accessibility of the corresponding gene promoter across all tested genes). Group annotations (in blue) were manually assigned based on qualitative similarities in the transcriptome and epigenome effects. e, Scatterplot of log₂FCs for transcriptome versus epigenome changes upon STAT3 and STAT5 knockout.

Fig. 6. Abrogated baseline JAK-STAT signaling outside of the in vivo tissue context.

a, Outline of the experimental approach: ex vivo culture for 20 h with basal medium and 10% FCS without supplements or with M-CSF (to support macrophage viability) or with IFN-β stimulation either in the last 1.5 h before sample collection or for the full 20 h, followed by transcriptome profiling. b, Differential expression upon ex vivo culture compared with homeostatic conditions in wild-type cells. Bar plots display the mean and standard error of log₂FCs. c, Enrichment or depletion of JAK-STAT-related gene sets among the differentially expressed genes from b. d, Enrichment or depletion of differentially expressed genes between JAK-STAT mutants and wild type from in vivo homeostatic conditions among the differentially expressed genes from b. For example, enrichment (red) of STAT1 knockout genes for IFN-β stimulation indicates that our homeostatic STAT1 target genes are preferentially induced by IFN-β stimulation. e, Summary of inferred receptor–ligand interactions in the spleen as inferred from the Tabula Muris dataset. Interactions where CD8⁺ T cells and macrophages represent targets (that is, express the receptor) are highlighted by black arrows. f, Selected ligand–receptor interactions of CD8⁺ T cells (top) and macrophages (bottom) with other types of immune cells. NES, normalized enrichment scores. P values in c, d and f are based on two-sided random sampling, corrected for multiple comparisons.

Fig. 7. Partial restoration of wild-type signaling upon stimulation of JAK-STAT mutant T cells.

a, Outline of IFN-β stimulation experiments and analyses in JAK-STAT knockout cells cultured ex vivo. This figure focuses on T cells, while corresponding results for macrophages are shown in Extended Data Fig. 10. b, Grouping of genes based on the observed IFN-β stimulation effects in wild-type and mutant cells. Lines correspond to the mean transcriptional change across all genes in each group. c, Prevalence of the five gene groups from b in each JAK-STAT mutant. Genes with significant but minor differences of stimulation effects between wild-type and mutant cells were not assigned to any group (marked in black). d, Differential gene expression heatmap for IRF9 knockout and wild-type T cells upon IFN-β stimulation, annotated with the grouping of differentially expressed genes (rows). e, Share of genes for which the JAK-STAT mutant effect reverts the IFN-β stimulation effect. This is calculated as the percentage of all genes with an IFN-β stimulation effect in wild-type cells, the total number of which is shown in brackets. f, Mean differential gene expression (log₂FC) upon IFN-β stimulation across 80 core ISGs. Box plots show the full data range, with the box indicating interquartile range and median. g, Share of genes for which the IFN-β stimulation reverts the JAK-STAT mutant effect, relative to all genes with a JAK-STAT mutant effect in unstimulated cells (shown in brackets). MUT, mutant; WT, wild type.

Extended Data Fig. 1. Study overview.

Illustration of the study design and key results as a graphical abstract (a) and a summary table of transcriptome and epigenome alterations in JAK-STAT mutant cells (b). We hypothesized that baseline JAK-STAT signaling underlies immune cell homeostasis, beyond its well-established role during immune stimulation. We analyzed homeostatic JAK-STAT signaling through transcriptome and epigenome profiling in a large collection of JAK-STAT mutants and validated our results by depriving cells from their cellular context in ex vivo culture. Our molecular maps of homeostatic JAK-STAT signaling reveal substantial roles of JAK-STAT signaling in specific homeostatic JAK-STAT complexes and regulated gene modules.

Extended Data Fig. 2. Transcriptome effects of JAK-STAT mutants in five immune cell types.

(a) Expression levels of two classical IFN target genes across cell types, mutants and sample processing laboratories. Downregulation of genes upon JAK-STAT mutants (differences to wildtype) strongly exceeded experimental variability across laboratories (variability within wildtype samples, indicated by the shape and size of points). (b) Effects of mutants on gene expression levels of JAK-STAT genes (two-sided linear mixed models, corrected for multiple comparisons). (c) Analysis of a pan-STAT signature (two-sided linear mixed models, corrected for multiple comparisons), showing only genes affected by multiple STAT mutants are shown. No gene was affected by all STAT mutants. (d-e) Transcriptional effects of JAK-STAT-mutant mice in all cell types, visualized by multi-dimensional scaling (MDS) (d) of Spearman correlation coefficients among log₂FCs compared to wildtype mice (e). FC: fold change; padj: adjusted p-value.

Extended Data Fig. 3. Differential expression of core ISGs between JAK-STAT mutant and wildtype mice.

Dot plot showing log₂FCs for the 50 genes with the largest log₂FC (two-sided linear mixed models, corrected for multiple comparisons). Genes are grouped based on the clusters derived from the UMAP analysis. padj: adjusted p-value.

Extended Data Fig. 4. Differential expression of genes regulated by STAT1 isoform-only mutants in macrophages.

Dot plot showing log₂FCs for STAT1 isoform-only regulated genes grouped based on STAT1 and STAT2 mutant effects in macrophages (two-sided linear mixed models, corrected for multiple comparisons). padj: adjusted p-value.

Extended Data Fig. 5. Spatial transcriptomics profiles of spleens from wildtype and STAT1 knockout mice.

Visualization of transcriptomics profiles for spleen samples from wildtype and STAT1 knockout mice (two biological replicates) that were formaldehyde-fixed in vivo, sliced and stained with H&E (first column; partially overlapping with plots presented in Fig. 4). K-means clustering identified six gene expression clusters (second column; partially overlapping with plots presented in Fig. 4). Spatial distribution of gene expression is further shown for the housekeeping gene Actb (third column) and the ISGs Oas3, Ifit3, and Ifit1 (fourth to sixth column).

Extended Data Fig. 6. JAK-STAT modulated chromatin accessibility at promoters of transcriptional regulator genes.

(a, b) Chromatin accessibility (ATAC-seq signal) transformed to z-scores for JAK-STAT genes in CD8⁺ T cells (a) and macrophages (b). (c–e) Differential accessibility results for JAK-STAT genes (c), JAK-STAT target genes (d), and other transcriptional regulators (e). padj: adjusted p-value. P-values in panels c-e are based on two-sided linear mixed models corrected for multiple comparisons.

Extended Data Fig. 7. Comparison of transcriptome versus epigenome changes for each of the JAK-STAT mutants.

Scatterplots of log₂FCs for differential gene expression and differential promoter chromatin accessibility across the twelve JAK-STAT mutants. The Pearson correlation (r) is indicated in each plot.

Extended Data Fig. 8. Integrative analysis of JAK-STAT mutant effects on the transcriptome and epigenome.

(a) Enrichment of regulator binding regions (from public ChIP-seq datasets) in differential chromatin regions (two-sided Fisher’s exact test, corrected for multiple comparisons). Grey dots represent examples where both enrichment and depletion were identified for different ChIP-seq datasets. (b) Enrichment of regulator binding profiles (from transcription factor binding motif analyses) in differential chromatin regions (two-sided hypergeometric test, corrected for multiple comparisons). (c) Enrichment of gene sets in differential chromatin regions (two-sided logistic regression, corrected for multiple comparisons). (d) Enrichment of gene sets in differentially expressed genes (two-sided chi squared test, corrected for multiple comparisons). OR: odds ratio; padj: adjusted p-value.

Extended Data Fig. 9. Loss of baseline JAK-STAT signaling outside of the in vivo tissue context.

(a) Average sample-to-sample Spearman correlation among homeostatic, context-deprived and stimulated immune cells, visualizing the effects of ex vivo culture on wildtype immune cells. The average correlation among samples under homeostasis is shown as baseline indicated by a black line. (b) SingleR similarity scores comparing cultured and homeostatic cells to external reference profiles from the ImmGen consortium. (c, d) Inferred receptor-ligand interactions of T cells and macrophages in single-cell transcriptome data from Tabula Sapiens. Examples of ligand-receptor interactions (c) between CD8⁺ T cells (left), macrophages (right) and various splenic cells (rows) (two-sided random sampling, corrected for multiple comparisons), and the total number of receptor-ligand interactions inferred from Tabula Sapiens (d). padj: adjusted p-value.

Extended Data Fig. 10. Partial restoration of baseline JAK-STAT signaling upon stimulation of JAK-STAT mutant macrophages.

(a) Prevalence of the five gene groups from Fig. 7b in each JAK-STAT mutant. This figure focuses on macrophages, while similar results for T cells are shown in Fig. 7b. (b) Share of genes for which the JAK-STAT mutant effect reverts the IFN-β stimulation effect. This is calculated as the percentage of all genes with a IFN-β stimulation effect in wildtype cells, the total number of which is shown in brackets. (c) Enrichment of core ISGs and IFN-β target genes among genes for which JAK-STAT mutants revert stimulation effects (two-sided Fisher’s exact test, corrected for multiple comparisons). This panel is the only panel in this figure that shows results for both macrophages and T cells. (d) Mean differential gene expression (log₂FC) upon IFN-β stimulation across 68 core ISGs. Box plots show the full data range, with the box indicating interquartile range and median. (e) Share of genes for which the IFN-β stimulation reverts the JAK-STAT mutant effect, relative to all genes with a JAK-STAT mutant effect in unstimulated cells (shown in brackets). OR: odds ratio; padj: adjusted p-value.