Coagulopathy signature precedes and predicts severity of end-organ heat stroke pathology in a mouse model

Abstract

Background: Immune challenge is known to increase heat stroke risk, although the mechanism of this increased risk is unclear.

Objectives: We sought to understand the effect of immune challenge on heat stroke pathology.

Patients/methods: Using a mouse model of classic heat stroke, we examined the impact of prior viral or bacterial infection on hematological aspects of recovery. Mice were exposed to heat either 48 or 72 hours following polyinosinic:polycytidylic acid (poly I:C) or lipopolysaccharide injection, time points when symptoms of illness (fever, lethargy, anorexia) were minimized or completely absent.

Results: Employing multivariate supervised machine learning to identify patterns of molecular and cellular markers associated with heat stroke, we found that prior viral infection simulated with poly I:C injection resulted in heat stroke presenting with high levels of factors indicating coagulopathy. Despite a decreased number of platelets in the blood, platelets are large and non-uniform in size, suggesting younger, more active platelets. Levels of D-dimer and soluble thrombomodulin were increased in more severe heat stroke, and in cases of the highest level of organ damage markers D-dimer levels dropped, indicating potential fibrinolysis-resistant thrombosis. Genes corresponding to immune response, coagulation, hypoxia, and vessel repair were up-regulated in kidneys of heat-challenged animals; these correlated with both viral treatment and distal organ damage while appearing before discernible tissue damage to the kidney itself.

Conclusions: Heat stroke-induced coagulopathy may be a driving mechanistic force in heat stroke pathology, especially when exacerbated by prior infection. Coagulation markers may serve as accessible biomarkers for heat stroke severity and therapeutic strategies.

Keywords: coagulopathy; computational analysis; heat stroke; systems biology; transcriptomics.

Figure 1

Experimental design. A, Cohort of mice was divided into three groups and received an injection of polyinosinic:polycytidylic acid, lipopolysaccharide, or sterile saline. Mice were returned to their home cages for either 48 or 72 hours, after which they were exposed to heat stress at 39.5°C, or were unheated controls at 25°C. Mice were sacrificed at T_c,max (42.4°C), 1 day following T_c,max, or 7 days following T_c,max, for a total of 36 groups (Tables S1 and S2) and N = 13 mice per group, where group is specific to immune challenge, immune incubation time, heat challenge, and time point parameters. Upon sacrifice, organs were harvested and blood was drawn from the heart to perform (B) complete blood count and (C) assay of plasma coagulation markers (see Methods)

Figure 2

Multivariate hematological signature predicts circadian differential in average core body temperature (ΔT_c), a proxy of heat stroke severity. A, Orthogonalized partial least squares regression model uses covariation in complete blood count results to separate mice on a spectrum of ΔT_c as average T_c,day – average T_c,night (color bar, °C). N = 137, four latent variables, cross‐validated using one third of input data. R ² of cross‐validation: .205, Wilcoxon cross‐validation, P‐value .008. B, Loadings on latent variable 1 (LV1) represent contributions of each factor to ΔT_c. Red bars indicate variable importance of progression ≥ 1, a greater‐than‐average contribution to correlation. C, Recoloring of existing model according to time‐point of sacrifice. LV2 represents differences between complete blood count at T_c,max, 1, 2, and 7 days. D, Loadings on LV2 represent contribution of each factor to timeline of pathology

Figure 3

Coagulation perturbations correlate with liver damage in heat‐ and viral‐challenged mice. A, Orthogonalized partial least squares regression model uses covariation in levels of circulating coagulation markers to separate mice on a spectrum of liver damage from none (blue) to high (red), as measured by levels of granzyme B (color bar, pg/mL). Gray dashed box highlights three outliers in (D). N = 30, two latent variables, cross‐validated using one third of input data. R² of cross‐validation: 0.518, Wilcoxon cross‐validation, P‐value .004. B, Recoloring of existing model according to treatment, 48 hours prior to heat challenge. Differences in treatment lie on the same axis (LV1) as the granzyme B spectrum, indicating correlation between treatment and organ damage. Gray dashed box highlights three outliers in (D). C, Loadings on latent variable 1 (LV1) represent contributions of each factor to concentration of granzyme B. Red bars indicate variable importance of progression ≥ 1, a greater‐than‐average contribution to correlation. D, Three outliers differ sufficiently to create an orthogonal axis of variation, LV2, upon which they separate from other points. Loadings on LV2 represent distinguishing factors of these outliers with extreme pathology, including extremely high values of thrombomodulin but decreased D‐dimer, potentially due to fibrinolysis resistance

Figure 4

Transcription changes in genes related to vascular injury response in heat stroke. A, Volcano plot showing differentially expressed genes (log2(fold change)) in kidneys of polyinosinic:polycytidylic acid (poly I:C)‐treated, heat‐challenged mice, and their statistical significance as adjusted P‐value. B, Volcano plot showing highly enriched gene sets (NES, normalized enrichment score) and their statistical significance as false discovery rate q‐value. C, Principal component analysis of gene transcripts separates groups by heat application (PC2, 9% of variation in gene transcripts) or poly I:C injection (PC3, 5% of variation in gene transcripts). D, Top 10 positive and top 10 negative contributing gene transcripts to heat response (PC2, top) and poly I:C response (PC3, bottom)

Figure 5

Dysregulated gene transcripts form a network of affected protein interactions. Dysregulated RNA transcripts in heat‐stroked mice as identified by combined_score from the STRING database of protein‐protein interactions, including experiments, co‐expression, and gene fusion evidence of interaction. Nodes (gene transcripts) are colored by HALLMARK gene sets, grouped as listed. Nodes were chosen as leading edge subsets of gene sets with enrichment false discovery rate q‐value < 0.25 in polyinosinic:polycytidylic acid, heated mice over saline, unheated mice. Edges were chosen as STRING combined_score > 0.99, ³³ with shorter length representing higher score (edge‐weighted, spring‐embedded layout). Nodes classified into more than one enriched gene set are colored as a pie chart ³⁴