(595f) A Compartmentalized Sepsis Model Based On Neutrophil Kinetics As a Decision Supporting Tool of Therapeutic Intervention

The complex pathophysiological interaction network of sepsis and its systemic nature involve many inflammatory mediators and a number of cell types, tissues and organs. Mechanisms driving the high mortality and morbidity observed in sepsis are therefore difficult to intuit and elucidate. Consequently, diagnostic, prognostic and therapeutic options have been limited due to the high variability and non-specificity of symptoms. As a broad-spectrum immune modulating therapy, an extracorporeal blood purification by hemoadsorption (HA) device using CytoSorb^TM beads (CytoSorbents, Monmouth Junction, NJ) has shown the capacity to alter circulating cytokine levels and improve survival in experimental endotoxemia and cecal ligation and puncture (CLP) induced sepsis in animals. However, the exact mechanisms by which the HA treatment resulted in reduced organ injury and improved outcome remain unclear.

Balancing the complexity of the inflammatory response, the insufficient knowledge of mechanistic details, and the relative paucity of data, we developed a mathematical model having a conceptual abstraction of the system behaviors across multiple tissue compartments. We seek to account for available experimental evidence and provide useful perspective on the pathophysiologic mechanisms of sepsis. The model focuses on the systemic behavior of neutrophils in response to infectious challenges by phenotyping circulatory neutrophils as they respond to inflammatory signals. The network components and interactions were assembled based on qualitative domain knowledge of the acute inflammatory response, including multiple phenotypes of neutrophils and major molecular effectors in distinct compartments (blood, peritoneum, and lung). In order to capture the impaired recruitment of neutrophils, which is one of the key pathophysiologic features of severe sepsis, several lines of evidence about the mechanisms influencing neutrophil migration in sepsis were incorporated in the model. Based on prior and concurrent experimental data, and our previous simulation results, we hypothesized that the HA device adsorbs circulating neutrophils as well as pro- and anti-inflammatory mediators (e.g., IL6, IL10, TNF, IL1, IL8) from the circulation.

Serial cytokine/chemokine measurements and neutrophil counts in distinct compartments were measured in rats that underwent CLP with sham (no HA treatment) or an HA intervention of 4 hour duration 18 hours after the induction of CLP. The experimental data sets were categorized into four subgroups (HA survival, HA death, Sham survival, Sham death) and the resulting four objective functions were constructed to constrain the model behaviors separately.
We computed quantitative predictions over an ensemble of parameter sets statistically drawn from all sets that are consistent with the available experimental data. An adaptive Markov Chain Monte Carlo where the step scales are periodically optimized based on the Hessian matrix of the cost function was used to efficiently sample the plausible parameter sets. Analysis of the parameter ensembles sampled from different target distributions provided useful insights on the pathophysiologic mechanisms represented by parameter subsets strongly associated with an adverse outcome and hypothesized the effects of HA on the dynamics of the system. In addition, model simulations predict a subset of animals (in their parametric description) most likely to benefit from the HA intervention. Other factors such as treatment timing and duration are discussed based on different plausible hypotheses. The mathematical model constructed here provides a platform for generating and testing hypotheses in silico, as well as motivating further translational (experimental) studies to advance our understanding of the complex biological response to inflammation and sepsis.