(416d) Dynamic Analysis of the DNA Sensing Pathway Predicts Host Immune Response

To effectively respond to a pathogen, the host intracellular immunity must dynamically assess and change with the environment. Dysregulation of the immune response can increase disease pathology and impair viral clearance. This can lead to chronic, systemic infections which—left untreated—can be fatal [1]. It is therefore imperative to develop a predictive model describing how pathogens invade the host and abrogate the immune response.

Significant research has gone into identifying key host proteins that perpetuate the signaling cascades responsible for innate immunity. The recently identified DNA sensing molecule cGAS recognizes viral DNA and invokes a type I interferon response. This leads to paracrine signaling of the JAK-STAT pathway and subsequent transcription of interferon stimulated genes (ISGs). ISGs effect various elements of the virus life cycle, many of which have specific mechanisms that are not yet known [2]. While this is an excellent description, it fails to accurately characterize the dynamics of the viral DNA response. Understanding these dynamics can help explain observations that are not addressed by static, qualitative descriptions. For example, it has been well established that viruses with similar genotypes exhibit distinct kinetics which alter the course of infection [3]. Modeling will elucidate the discrepancies in experimental observation like this and motivate new hypotheses.

Methods

To characterize the DNA sensing pathway’s behavior in response to viral DNA, ordinary differential equations (ODEs) are employed to simulate how host proteins coordinate and activate ISGs. The model is comprised of 16 differential equations that represent the forward and reverse rates for each species (proteins and mRNAs) in the pathway. Typical formulations for biochemical reactions are used to define these rates including mass action, Michaelis Menten, and Hill kinetics. The latter is useful for modeling highly cooperative binding events such as the interactions between cGAS, viral DNA, ATP, and GTP to form the signaling molecule cGAMP. The model contains 24 rate constants, some of which have estimates from literature while others are unknown.

Traditional parameter estimation techniques that rely on gradient descent or maximum likelihood estimation which can be insufficient for large nonlinear ODE systems [4]. To compensate, a Bayesian approach is implemented to survey the parameter landscape and return probabilistic distributions as estimates. Adaptive Markov chain Monte Carlo (MCMC) simulation provides a computationally sufficient method to estimate the posterior distributions for all the unknown parameters. Uniform priors were used to avoid biasing parameter estimation [5]. To perform simulations for MCMC, the R package Flexible Modelling Environment (FME) was utilized [6]. Data for parameter fitting comes from time-course gene expression experiments. A transcriptionally deficient variant of Herpes Simplex Virus (HSV) was used to infect human fibroblasts. This type of experiment is informative because it controls for the negative regulation from viral proteins and allows for the observation of the unabated DNA sensing pathway.

In addition to parameter fitting, global and local sensitivity analyses were performed to determine the robustness of the model. Relatively large sensitivities, while also revealing stability properties, can be effective for motivating new experimental hypotheses. A parameter sensitive in the model is a likely candidate for viral manipulation.

Results

Model simulations capture the relative timing and shape of the dynamics seen in experiment and literature [7]. Typical upregulation of interferon beta occurs within four to six hours given an input of viral DNA and follows a saturation curve. Precursor protein complexes exhibit low accumulation due to the high input and output rates. It is possible that this is an accurate description of the biology, but also could be an indication that model reduction is needed. The production of ISGs show a delayed upregulation starting at 24 hours and downregulation after 48 hours. The eventual decline is controlled by the negative regulators (SOCS1, SOC2, USP18) that overtake the interferon signal and shut down the pathway. In many instances biologically, chronic inflammation occurs from aberrant ISG production that was not controlled by the negative regulators [8]. The model predicts this behavior with continual DNA stimulation.

We show how perturbations in highly sensitive kinetic parameters abrogate immune responses making their associated proteins likely targets for viral inhibition. Feedback mechanisms are also excellent targets due to their ability to amplify signals and stabilize systems.

Future work aims to expand this model to incorporate the transcription of interferon stimulated genes (ISGs) that are positively autoregulated, as well as ISGs that are not directly activated by the interferon response. These subsets of ISGs can be determined by different knockout experiments along the DNA sensing pathway. This analysis is vital for modeling to determine the timeframes of individual ISG activation.

Additionally, the model can be implemented to simulate different types of viral infections and autoimmune diseases caused by the detection of self-DNA. Understanding the different contexts in which this signaling pathway is stimulated can provide additional information into the structure of the pathway and varying kinetics observed.

Bibliography

[1] D. Kobasa, S. M. Jones, K. Shinya, J. C. Kash, J. Copps, H. Ebihara, Y. Hatta, J. Hyun Kim, P. Halfmann, M. Hatta, F. Feldmann, J. B. Alimonti, L. Fernando, Y. Li, M. G. Katze, H. Feldmann, and Y. Kawaoka, “Aberrant innate immune response in lethal infection of macaques with the 1918 influenza virus,” Nature, vol. 445, no. 7125, pp. 319–323, Jan. 2007.

[2] J. W. Schoggins, “Interferon-stimulated genes and their antiviral effector functions,” Curr. Opin. Virol., vol. 1, no. 6, pp. 519–525, 2011.

[3] J. E. Medeiros-Filho, “Differences in viral kinetics between genotypes 1 and 3 of hepatitis C virus and between cirrhotic and non-cirrhotic patients during antiviral therapy,” World J. Gastroenterol., vol. 12, no. 45, p. 7271, 2006.

[4] M. Ashyraliyev, Y. Fomekong-Nanfack, J. A. Kaandorp, and J. G. Blom, “Systems biology: parameter estimation for biochemical models,” FEBS J., vol. 276, no. 4, pp. 886–902, Feb. 2009.

[5] K. S. Brown and J. P. Sethna, “Statistical mechanical approaches to models with many poorly known parameters,” Phys. Rev. E, vol. 68, no. 2, p. 21904, Aug. 2003.

[6] K. Soetaert and T. Petzoldt, “Inverse Modelling, Sensitivity and Monte Carlo Analysis in R Using Package FME,” J. Stat. Softw., vol. 33, no. 3, pp. 1–28, 2010.

[7] M. Fribourg, B. Hartmann, M. Schmolke, N. Marjanovic, R. A. Albrecht, A. García-Sastre, S. C. Sealfon, C. Jayaprakash, and F. Hayot, “Model of influenza A virus infection: Dynamics of viral antagonism and innate immune response,” J. Theor. Biol., vol. 351, pp. 47–57, 2014.

[8] G. Trinchieri, “Type I interferon: friend or foe?,” J. Exp. Med., vol. 207, no. 10, 2010.