1
as our modeling framework. The model contained unpolarized and polarized cells, lumen, and matrix, all of which interacted on a 2- dimensional grid. To maintain the quasi-autonomous nature of our cellular agents, we added plugins to the system that could be executed from the perspective of the individual agents. Once the model was implemented, we iteratively compared the computational results with in vitro data and modified the in silico operating principles to improve the model's ability to survive falsification. Results: The model successfully reproduced a number of in vitro targeted attributes. In silico cells grew and proliferated at a similar rate to that observed in vitro, as well as polarizing and differentiating at roughly similar time intervals. The cysts produced by the model contained both single lumens and multiple lumens at rates analogous to the in vitro system. In addition, we found that lumen number and cyst size correlated with the length of time before in silico cells polarized and also with the type of cell division axis regulation. Conclusions: These results suggest that the operating principles chosen for the model are an accurate representation of the underlying in vitro operating principles. There appears to be a close link between the timing of cell polarization, differentiation, and overall cyst structure. In addition, it is possible for cyst creation to occur with little or no cell death, a result confirming recent findings within in vitro MDCK cell culture but not seen in other systems. Further in silico and in vitro research is being conducted to verify these observations. doi:10.1016/j.jcrc.2009.06.038 Immune response to influenza A Ian Price a , David Swigon a,b , Bard Ermentrout a,b , Frank Toapanta c , Ted Ross c , Gilles Clermont b,d a Department of Mathematics, University of Pittsburgh b Center for Inflammation and Regenerative Modeling, McGowan Institute for Regenerative Medicine, University of Pittsburgh c Center for Vaccine Research, University of Pittsburgh d Department of Critical Care Medicine, University of Pittsburgh Objectives: In mammals, Influenza A virus triggers innate immunity before adaptive immunity; however, an exaggerated innate response is harmful to tissue and does not further the elimination of virus. This project seeks to model the immune response and identify methods of preventing lung failure and improving recovery. The model introduces a dynamic innate immune response with relevant biological compartments, expanding upon existing models of adaptive immune response to Influenza A virus. Methods: The project uses an ODE-based model to study the dynamics of regulation between virus, immune and respiratory cells, and signaling macromolecules. The inflammatory process begins with macrophage-mediated production of cytokines and chemotaxis of immune cells into the tissue. Virus-specific immune responses such as type I interferon and NK cells are introduced subsequently. Finally, type II interferon, CTLs, and antibodies are produced to facilitate complete virus removal. The system allows for stable health and death, with initial viral load leading to each. For some choices of key parameters, unstable health and a stable chronic state with viral clearance can be attained. The model is calibrated to a data set taken from test animals that includes blood, lung and spleen cytokine levels, chemokine levels, cell populations, as well as organ function and survival. The data are divided into adult and elderly test animals, and separate parameters sets are evaluated for each. Mathematical techniques explore solutions to the inverse problem for parameter values locally and globally over parameter space. Results: Parameter sets were derived that minimize the squared error of the model to the data for the adult and elderly cases. Parameter values are compared to demonstrate how the ageing process affects immune response. Finally, we propose a treatment regimen to optimize the likelihood of survival for an unknown level of initial viral load. Conclusions: The model, as developed, gives us a metric relating initial infection, strength of various immune responses, and total lung damage. Decoding the pathways of viral immune response allows us to measure their effect on overall damage and time of recovery, and motivates experimental study of interventions successful in the model. doi:10.1016/j.jcrc.2009.06.039 A multi-reservoir model of influenza evolution David W. Dreisigmeyer a,b,c , Roni Rosenfeld d , Jay V. DePasse e , Elodie Ghedin e , Ian Price a , Gilles Clermont b a Department of Mathematics, University of Pittsburgh b Department of Critical Care Medicine, University of Pittsburgh c Department of Computational Biology, University of Pittsburgh d Department of Computer Science, Carnegie Mellon University e Department of Medicine, University of Pittsburgh Objectives: The work by Koelle et al [1] introduced the idea that the influenza virus type A (IVA) may evolve on neutral networks. A neutral network is a collection of genotypes that map (in some way) to the same phenotype even if there may be significant evolutionary change in the genotype. The virus experiences no evolutionary pressure until it enters a new neutral network. This allows the IVA genotype to diffuse over the entire neutral network. This diffusion can bring the current IVA cluster (phenotype) in contact with many other neutral networks (phenotypes). Upon entering a new network, the phenotype can change either moderately or significantly. The model can be seeded with strains from an (unmodeled) external reservoir via migration events. The phylogenetic trees emerging from the neutral network model are visually stunning. Subsequent investigations by Shih et al [2] and Suzuki [3] have cast doubt on whether IVA evolves on a neutral network. Their work suggests that IVA experiences continuous evolutionary pressure. This raises the obvious question whether neutral networkbased theory of IVA evolution in humans can be constructed in such a way that the IVA still experiences continuous positive selection. One conceivable way for this to happen is to allow IVA evolution to occur in several reservoirs. The evolution within each reservoir will be subjected to positive evolutionary pressure. On occasion, new genotypes arising in these other reservoirs could be (re)introduced to the human population. From the perspective of the human reservoir, significant evolutionary change can occur in IVA. However, this evolution is occurring in other populations. Methods: The phenotypes of the IVA strains are determined by constructing a neutral network for each of 8 RNA strands comprising the IVA genome, similar to what was done in reference [1], restricted for IVA hemagglutinin gene. Every IVA genotype is assigned to these networks. Every neutral network has associations with certain class(es) of reservoir(s) that it can infect. In our model, e33 Meeting Abstracts for the Society for Complexity in Acute Illness (SCAI)

Immune response to influenza A

Embed Size (px)

Citation preview

e33Meeting Abstracts for the Society for Complexity in Acute Illness (SCAI)

as our modeling framework. The model contained unpolarized andpolarized cells, lumen, and matrix, all of which interacted on a 2-dimensional grid. To maintain the quasi-autonomous nature of ourcellular agents, we added plugins to the system that could beexecuted from the perspective of the individual agents. Once themodel was implemented, we iteratively compared the computationalresults with in vitro data and modified the in silico operatingprinciples to improve the model's ability to survive falsification.Results: The model successfully reproduced a number of in vitrotargeted attributes. In silico cells grew and proliferated at a similarrate to that observed in vitro, as well as polarizing and differentiatingat roughly similar time intervals. The cysts produced by the modelcontained both single lumens and multiple lumens at rates analogousto the in vitro system. In addition, we found that lumen number andcyst size correlated with the length of time before in silico cellspolarized and also with the type of cell division axis regulation.Conclusions: These results suggest that the operating principleschosen for the model are an accurate representation of theunderlying in vitro operating principles. There appears to be aclose link between the timing of cell polarization, differentiation,and overall cyst structure. In addition, it is possible for cyst creationto occur with little or no cell death, a result confirming recentfindings within in vitro MDCK cell culture but not seen in othersystems. Further in silico and in vitro research is being conducted toverify these observations.

doi:10.1016/j.jcrc.2009.06.038

Immune response to influenza AIan Price a, David Swigon a,b, Bard Ermentrout a,b, Frank Toapanta c,

Ted Ross c, Gilles Clermont b,d

aDepartment of Mathematics, University of PittsburghbCenter for Inflammation and Regenerative Modeling, McGowan Institute

for Regenerative Medicine, University of PittsburghcCenter for Vaccine Research, University of PittsburghdDepartment of Critical Care Medicine, University of Pittsburgh

Objectives: In mammals, Influenza Avirus triggers innate immunitybefore adaptive immunity; however, an exaggerated innate responseis harmful to tissue and does not further the elimination of virus. Thisproject seeks to model the immune response and identify methods ofpreventing lung failure and improving recovery. The modelintroduces a dynamic innate immune response with relevantbiological compartments, expanding upon existing models ofadaptive immune response to Influenza A virus.Methods: The project uses an ODE-based model to study thedynamics of regulation between virus, immune and respiratorycells, and signaling macromolecules. The inflammatory processbegins with macrophage-mediated production of cytokines andchemotaxis of immune cells into the tissue. Virus-specific immuneresponses such as type I interferon and NK cells are introducedsubsequently. Finally, type II interferon, CTLs, and antibodies areproduced to facilitate complete virus removal. The system allowsfor stable health and death, with initial viral load leading to each.For some choices of key parameters, unstable health and a stablechronic state with viral clearance can be attained.

The model is calibrated to a data set taken from test animals thatincludes blood, lung and spleen cytokine levels, chemokine levels,cell populations, as well as organ function and survival. The data are

divided into adult and elderly test animals, and separate parameterssets are evaluated for each. Mathematical techniques exploresolutions to the inverse problem for parameter values locally andglobally over parameter space.Results: Parameter sets were derived that minimize the squared errorof the model to the data for the adult and elderly cases. Parametervalues are compared to demonstrate how the ageing process affectsimmune response. Finally,wepropose a treatment regimen tooptimizethe likelihood of survival for an unknown level of initial viral load.Conclusions: The model, as developed, gives us a metric relatinginitial infection, strength of various immune responses, and totallung damage. Decoding the pathways of viral immune responseallows us to measure their effect on overall damage and time ofrecovery, and motivates experimental study of interventionssuccessful in the model.

doi:10.1016/j.jcrc.2009.06.039

A multi-reservoir model of influenza evolutionDavid W. Dreisigmeyer a,b,c, Roni Rosenfeld d, Jay V. DePasse e,

Elodie Ghedin e, Ian Price a, Gilles Clermont b

aDepartment of Mathematics, University of PittsburghbDepartment of Critical Care Medicine, University of PittsburghcDepartment of Computational Biology, University of PittsburghdDepartment of Computer Science, Carnegie Mellon UniversityeDepartment of Medicine, University of Pittsburgh

Objectives: The work by Koelle et al [1] introduced the idea that theinfluenza virus type A (IVA) may evolve on neutral networks. Aneutral network is a collection of genotypes that map (in some way)to the same phenotype even if there may be significant evolutionarychange in the genotype. The virus experiences no evolutionarypressure until it enters a new neutral network. This allows the IVAgenotype to diffuse over the entire neutral network. This diffusioncan bring the current IVA cluster (phenotype) in contact with manyother neutral networks (phenotypes). Upon entering a new network,the phenotype can change either moderately or significantly. Themodel can be seeded with strains from an (unmodeled) externalreservoir via migration events. The phylogenetic trees emergingfrom the neutral network model are visually stunning.

Subsequent investigations by Shih et al [2] and Suzuki [3] havecast doubt on whether IVA evolves on a neutral network. Their worksuggests that IVA experiences continuous evolutionary pressure.

This raises the obvious question whether neutral network–basedtheory of IVA evolution in humans can be constructed in such a waythat the IVA still experiences continuous positive selection. Oneconceivable way for this to happen is to allow IVA evolution tooccur in several reservoirs. The evolution within each reservoir willbe subjected to positive evolutionary pressure. On occasion, newgenotypes arising in these other reservoirs could be (re)introducedto the human population. From the perspective of the humanreservoir, significant evolutionary change can occur in IVA.However, this evolution is occurring in other populations.Methods: The phenotypes of the IVA strains are determined byconstructing a neutral network for each of 8 RNA strandscomprising the IVA genome, similar to what was done in reference[1], restricted for IVA hemagglutinin gene. Every IVA genotype isassigned to these networks. Every neutral network has associationswith certain class(es) of reservoir(s) that it can infect. In our model,