A putative "chemokine switch" that regulates systemic acute inflammation in humans

Abstract

Systemic inflammation is complex and likely drives clinical outcomes in critical illness such as that which ensues following severe injury. We obtained time course data on multiple inflammatory mediators in the blood of blunt trauma patients. Using dynamic network analyses, we inferred a novel control architecture for systemic inflammation: a three-way switch comprising the chemokines MCP-1/CCL2, MIG/CXCL9, and IP-10/CXCL10. To test this hypothesis, we created a logical model comprising this putative architecture. This model predicted key qualitative features of systemic inflammation in patient sub-groups, as well as the different patterns of hospital discharge of moderately vs. severely injured patients. Thus, a rational transition from data to data-driven models to mechanistic models suggests a novel, chemokine-based mechanism for control of acute inflammation in humans and points to the potential utility of this workflow in defining novel features in other complex diseases.

Figure 1

Workflow leading to a conceptual model of the “chemokine switch”. (Panel A): A schematic of the analysis workflow. Time courses of inflammatory mediators were measured in trauma patients and causal interactions inferred by DyBN analysis. The inferred network topology formed the basis a Boolean model which was simulated in silico. Results were compared to clinical trajectories to refine the model. (Panels B-D): DyBN consensus network structure for Mild, Moderate and Severe Injury. Panel E: Boolean model structure with cross-regulation among chemokines MIG, IP-10, and MCP-1.

Figure 2

Inflammatory mediator trajectories for Moderate Injury: Simulations vs. data from trauma patients. Left column: 500 simulations were run with random initial conditions. Plot shows mean plus standard error for each time step. Right column: Patient data shown as mean with standard error for each time point. Healthy volunteer (white circles) included for reference.

Figure 3

Inflammatory mediator trajectories for Severe Injury: Simulations vs. data from trauma patients. Left column: 500 simulations were run with random initial conditions. Plot shows mean plus standard error for each time step. Right column: Patient data shown as mean with standard error for each time point. Healthy volunteer (white circles) included for reference.

Figure 4

Addition of a putative node improves IP-10 simulations. A new variable, X, was added upstream of IP-10 to introduce a delay due to the observed slow rise in patient IP-10 trajectory. A spiky trajectory for X (Panel A) produces an IP-10 trajectory (Panel B) that is inconsistent with patient data. A step rise for X (Panel C) produces an IP-10 trajectory (Panel D) that matches well with patient data. Coincidentally, out of the measured inflammatory mediator profiles, the trajectory of circulating IFN- γ most closely resembles a step increase trajectory (Panel E). (Panel F): Model schematic with addition of new variable X.

Figure 5

Logical model captures differences in Moderate vs. Severe Injury patients with low MCP-1. Simulation with moderate injury and low initial MCP-1 (Panel A) shows MCP-1 remaining low, whereas simulation with severe injury and low initial MCP-1 (Panel C) shows a rise and higher sustained MCP-1 levels. Similarly, patients with low initial MCP-1 and moderate injury (Panel B) exhibit lower MCP-1 levels throughout time course as compared to patients with severe injury (Panel D). Data in (Panels B vs. D) are significantly different (P < 0.05) by Two-Way ANOVA.

Figure 6

Logical model captures differences in patient discharge in Moderate vs Severe Injury patients. Kaplan–Meier style survival curve with endpoint as steady state (simulations, Panel A) or patient discharge (patients, Panel B). Simulations of moderate injury reach steady state sooner than severe injury (top panel). Similarly, patients with moderate injury are discharged sooner than severely injured patients (bottom panel).