Modeling the dynamics of neuroendocrine-immune interactions in collagen-induced arthritis

Rheumatoid arthritis (RA) is a chronic inflammatory disease that affects about 1% of the population in the developed world, and is primarily characterized by synovial inflammation and cartilage deterioration, typically in the joints of the hands and feet (1). The hypothalamic-pituitary-adrenal (HPA) axis, a critical component of the neuroendocrine system, is thought to be an important regulator of disease activity. It has been hypothesized that adrenal insufficiency; the inability of the HPA axis to mount an adequate anti-inflammatory in response to ongoing chronic inflammation in RA might contribute to disease pathophysiology (2). In addition, circadian variation in clinical symptoms has been well documented in RA patients. Moreover, a disruption in the natural circadian rhythms of major mediators of the HPA axis, such as cortisol has been observed (3). A mechanistic understanding of the interplay between the circadian rhythms of the HPA axis and the immune system might provide useful insights into RA pathologies and optimization of treatment regimens. Animal models of RA have been extensively used to study disease pathogenesis and in the discovery of anti-arthritic drugs (4). A rodent model of collagen induced arthritis (CIA), which involves immunization of the animal with an emulsion of the complete Freund's adjuvant and type II collagen, is considered to be the gold standard in vivo model for RA (5). In this work we discuss the development of a mathematical model that aims to evaluate the dynamic interactions between the HPA axis and pro-inflammatory cytokines, and their downstream effects on paw edema, a key disease end-point in CIA. In particular, we investigate the changes in circadian secretion of corticosterone (CST) and pro-inflammatory cytokines after induction of CIA, that ultimately result in circadian variability in paw edema. Moreover, we recapitulate the phenomenon of adrenal insufficiency that has been observed in both CIA as well as RA patients. These simulations will aid in the characterization of neuroendocrine-immune response in animal models of RA. Our model builds on earlier work (6-9) and is comprised of three components; a central oscillatory compartment that simulates the circadian secretion of CST in the HPA axis, a peripheral compartment that captures the downstream effects of secreted CST on cytokine expression, and finally, a compartment that accounts for the disease endpoints, specifically, paw edema. The oscillations in the central compartment are generated due to the negative feedback between CST and corticotropin-releasing hormone (CRH)/adrenocorticotropic hormone (ACTH) using a modified Goodwin oscillator model that incorporates Michaelis-Menten kinetics to express the degradation terms in the hypothalamus, pituitary, and adrenals. The HPA axis is positively modulated by the expression of cytokines in the peripheral compartment, which in turn is balanced by the negative feedback of CST. This results in an interlinked positive and negative feedback circuit that enables the investigation of the dynamics of HPA axis-cytokine interactions upon the onset of inflammation. In this model, we assume that the severity of paw edema is in direct correlation to the expression of pro-inflammatory cytokines. Furthermore, we explicitly model the emergence of adrenal insufficiency in response to chronically elevated cytokines by postulating an indirect response mechanism that effectively results in stimulating the clearance of ACTH in the HPA axis. Disease progression is initiated by a delay in the onset of the increased pro-inflammatory cytokine expression by using a simple transduction model (6). Model predictions demonstrate that elevated cytokine expression results in circadian variability in paw edema progression, reminiscent of the circadian rhythms of disease symptoms exhibited by RA patients. Notably, our model predicts that an increase in mean secretion of CST after the induction of the disease is accompanied by a decrease in the amplitude of the CST oscillation. This phenomenon persists even after the development of adrenal insufficiency and the consequent drop in mean CST secretion toward levels present prior to the onset of the disease. This suppression of circadian rhythm has been experimentally observed in a number of important neuroendocrine and HPA axis mediators during adjuvant induced arthritis, including; ACTH, CST and adrenocortical ornithine decarboxylase (10). Further analysis of possible alterations in the phase of oscillation of CST and the cytokines during disease progression is currently ongoing. In conclusion, our model captures key features of chronic inflammation in animal models of RA. In doing so, it highlights the importance of accounting for circadian rhythms while studying the neuroendocrine-immune system interactions in RA.