Longitudinal transcriptomics define the stages of myeloid activation in the living human brain after intracerebral hemorrhage

Abstract

Opportunities to interrogate the immune responses in the injured tissue of living patients suffering from acute sterile injuries such as stroke and heart attack are limited. We leveraged a clinical trial of minimally invasive neurosurgery for patients with intracerebral hemorrhage (ICH), a severely disabling subtype of stroke, to investigate the dynamics of inflammation at the site of brain injury over time. Longitudinal transcriptional profiling of CD14⁺ monocytes/macrophages and neutrophils from hematomas of patients with ICH revealed that the myeloid response to ICH within the hematoma is distinct from that in the blood and occurs in stages conserved across the patient cohort. Initially, hematoma myeloid cells expressed a robust anabolic proinflammatory profile characterized by activation of hypoxia-inducible factors (HIFs) and expression of genes encoding immune factors and glycolysis. Subsequently, inflammatory gene expression decreased over time, whereas anti-inflammatory circuits were maintained and phagocytic and antioxidative pathways up-regulated. During this transition to immune resolution, glycolysis gene expression and levels of the potent proresolution lipid mediator prostaglandin E₂ remained elevated in the hematoma, and unexpectedly, these elevations correlated with positive patient outcomes. Ex vivo activation of human macrophages by ICH-associated stimuli highlighted an important role for HIFs in production of both inflammatory and anti-inflammatory factors, including PGE₂, which, in turn, augmented VEGF production. Our findings define the time course of myeloid activation in the human brain after ICH, revealing a conserved progression of immune responses from proinflammatory to proresolution states in humans after brain injury and identifying transcriptional programs associated with neurological recovery.

Fig. 1.. The myeloid response to ICH is a highly conserved two-stage process.

(A) Visual summary of sample collection from ICH patients and cDNA generation for RNA sequencing of myeloid cells from hematoma effluent and peripheral blood. Additional information can be found in Materials and Methods. (B) Principal component analysis of transcriptional profiles of CD14+ monocytes/macrophages and neutrophils from ICH patients. Each point represents a single sample; all timepoints from each donor are represented. Projections show the first two principal components, which comprise the largest proportion of the variance in overall gene expression. Additional data are provided in Fig. S4. (C) Mean expression over time of genes in one of two dynamic transcriptional modules identified in hematoma monocytes/macrophages and neutrophils, or in the remaining static genes. Each point represents the mean relative expression level of all genes either upregulated (red/orange), downregulated (blue/teal) or unchanged (gray) over time in a single sample; every monocyte/macrophage or neutrophil sample contributes one point of each color. Colored lines represent loess-smoothed regression of the data. Additional information is provided in Fig. S5. (D) Venn diagram depicting overlap in genes between the modules portrayed in (C). Gene lists can be found in Table S5.

Fig. 2.. CD14+ macrophages and neutrophils display broad transcriptional remodeling in the hematoma during the acute stage of ICH.

(A) Median level of expression of monocyte/macrophage and neutrophil genes in blood versus hematoma. Only the first (earliest) blood and hematoma samples from each patient in the dataset were included in this analysis, spanning 23 to 99 hours post-ICH; n = 21 patients for monocytes/macrophages, n = 17 patients for neutrophils. Transcriptional profiles that did not meet minimum quality standards were filtered out of the dataset (see Materials and Methods, Initial data processing), which led to fewer neutrophil profiles in the comparison than monocyte/macrophage profiles. Significantly upregulated or downregulated genes in the hematoma (BH adjusted p<0.05) are colored red and blue respectively. (B) Single sample gene set enrichment analysis (ssGSEA) of the first blood and hematoma samples from each patient depicting relative enrichment of all 53 hallmark gene sets to provide an overview of transcriptional differences between blood and hematoma cells. Key immune and metabolic pathways are highlighted in bold. Differential enrichment between blood and hematoma populations was assessed by student’s t test adjusted for multiple comparisons using the Benjamini-Hochberg (BH) method. ***: p<0.001, **: p<0.01, *: p<0.05.

Fig. 3.. CD14+ monocytes/macrophages and neutrophils are functionally and metabolically reprogrammed within the hematoma during the acute response to ICH.

(A) Canonical pathway analysis of differentially expressed genes in CD14+ monocytes/macrophages and neutrophils during the acute stage. After selecting for pathways with significant enrichment (BH adjusted p<0.05), the eight pathways with the highest enrichment Z-scores in each cell type are presented. (B-G) Differential expression of genes in glucose utilization pathways and immune factor secretion pathways during the acute stage of ICH. Color scale represents median log2 fold-change in expression in hematoma compared to blood. Blank spaces represent genes not expressed by neutrophils in either tissue. Only genes with significant differential expression between blood and hematoma in at least one of the two cell types (BH adjusted p<0.05) are included. Dashed lines between glycolysis enzymes and schematic denote genes controlling a particular enzymatic step in glycolysis. Heatmaps of expression of these genes by every sample are presented in Fig. S7 and additional differential expression data is presented in Table S6. (H) Upstream regulator analysis of predicted transcriptional mediators of the changes to gene expression during the acute stage of the ICH response. After selecting for pathways with significant enrichment (BH adjusted p<0.05), the ten pathways with the highest enrichment Z-scores in each cell type are presented. All analyses were performed using the first (earliest) blood and hematoma samples from each patient in the dataset, spanning 23 to 99 hours post-ICH (the acute stage); n = 21 patients for CD14+. monocytes/macrophages, 17 patients for neutrophils.

Fig. 4.. CD14+ monocytes/macrophages decrease expression of glycolytic and inflammatory genes over time.

(A) Gene ontology (GO) analysis of genes decreasing in expression over time in hematoma CD14+ monocytes/macrophages. Number of genes represented in the pathway are presented at the end of each bar. (B) Gene expression over time in blood and hematoma CD14+ monocytes/macrophages of enzymes controlling rate limiting steps in glycolysis. Gray shading represents the 95% confidence interval of the regression mean. (C) Lactate levels in hematoma effluent and peripheral blood over time after ICH. n = 8 blood samples from 8 patients and 56 hematoma samples from 18 patients. (D) Gene expression over time of secreted inflammatory cytokines by blood and hematoma CD14+ monocytes/macrophages. For all gene expression plots, n = 82 (blood) and 57 (hematoma). p values represent significance of changes to gene expression in hematoma cells or lactate levels over time in hematoma as measured by spline regression, adjusted using the BH method. Additional data presented in Table S5 and Fig. S8.

Fig. 5.. Hematoma CD14+ monocytes/macrophages acquire a transcriptional profile associated with erythrocyte phagocytosis and repair over time after ICH.

(A) Gene ontology (GO) analysis of genes decreasing in expression over time in hematoma CD14+ monocytes/macrophages. Number of genes represented in the pathway are presented at the end of each bar. (B) Gene expression over time of genes involved in efferocytosis (MERTK, HAVCR2, ITGAV) and heme degradation (HMOX1). (C-E) Gene expression over time of genes encoding anabolic metabolism (C), monocyte chemoattractants (D), the anti-inflammatory cytokine IL-10, and PTGES (E). Gray shading represents the 95% confidence interval of the regression mean. For all gene expression plots, n = 82 (blood monocytes), 57 (hematoma CD14+ monocytes/macrophages), 76 (blood neutrophils) and 49 (hematoma neutrophils) samples. (F) PGE₂ levels in hematoma effluent and peripheral blood over time after ICH. n = 8 blood samples from 8 patients and 57 hematoma samples from 18 patients. p values represent significance of changes over time as measured by spline regression, adjusted using the BH method. Additional data presented in Table S5 and Fig. S9.

Fig. 6.. PTGES and glycolytic enzyme genes are more highly expressed by CD14+ monocytes/macrophages in patients with good neurological recovery.

(A) Differential gene expression in hematoma CD14+ monocytes/macrophages during the sub-acute stage of ICH (>96 hours post-ICH). Each column represents one patient; each row represents one gene (561 total genes). The dendrogram represents agglomerative hierarchical clustering by unweighted pair group method with arithmetic mean (UPGMA). Additional data presented in Table S9. (B) Gene ontology (GO) analysis of differentially expressed genes in hematoma CD14+ monocytes/macrophages during the sub-acute stage of ICH. Number of genes represented in the pathway are presented at the end of each bar. For enrichment in poor outcome, only the 5 most significantly enriched pathways are shown, additional pathways are presented in Table S10. (C) Differentially expressed genes from metabolic and functional pathways in figures 3, 4, and 5 are displayed. (D) Expression level of PTGES by hematoma CD14+ monocytes/macrophages. (E) Expression level of HK2 by hematoma CD14+ monocytes/macrophages. (F) Lactate levels in hematoma during sub-acute stage of ICH; n = 14 total patients, lactate levels for one patient were not recorded. Lactate levels were compared by linear modeling adjusted for initial hemorrhage severity and subsequent F test for statistical significance. For all gene expression data, statistically significant differential expression was determined by linear modeling adjusted for initial hemorrhage severity (n =15 total patients); BH adjusted p<0.05 significance threshold. **: p<0.01; ***: p<0.001,

Fig. 7. HIF-mediated glycolysis by macrophages promotes reparative functions.

(A) Expression of HK2 by macrophages treated for 8 hours with ICH-associated danger molecules (ICH-DAMP: S100A8 [1 μg/mL], Thrombin [10 U/mL], and IL-1β [10 ng/mL]), HIF activator DFO (100 μg/mL), or vehicle control +/− HIF signaling antagonist echinomycin (50 nM). Points represent mean values from two replicate wells of n = 3 donors. (B) Glycolytic flux of healthy donor human macrophages stimulated with ICH-DAMP or vehicle control in+/− echinomycin (50 nM). Points represent mean values from five replicate wells of n = 4 donors. (C) PTGES expression by healthy donor macrophages under conditions described in (A). (D) IL-6 and PGE₂ production by healthy donor macrophages treated for 24 hours with ICH-DAMP or vehicle control +/− echinomycin or glycolysis inhibitor 2-DG (1 mM). All comparisons are to vehicle + ICH-DAMP. (E) PGE₂ and VEGF production by healthy donor macrophages treated for 48 hours with ICH-DAMP in the presence or absence of 100 nM PGE₂. Points represent mean values from five replicate wells of n = 6 donors. Additional cytokine data are presented in Fig. S12. All error bars represent standard deviation. ***: p<0.001, **: p<0.01