SMB2023 FollowMonday during the "CT01" time block. Room assignment: Cartoon Room 2 (#3147) in The Ohio Union.
Jason E. Shoemaker
University of Pittsburgh
"Sex-specific immunoregulation: Computational modeling approaches to determine why biological females may experience greater inflammation during influenza infection"
In humans, biological males and females often experience different outcomes during respiratory infections. Here, we are referring to differences in biological sex (XX and XY) and not gender, which includes behaviors and activities that are determined by society or culture in humans. During the 2009 H1N1 influenza pandemic, adult females were at greater risk than their male, age-matched counterparts for hospitalization and death. Many factors may determine sex-specific outcomes, but there is considerable evidence that sex-specific immune regulation is a key driver of enhanced disease pathology. Severe influenza infections are characterized with aggressive inflammatory responses, and studies show that male and female inflammatory responses differ when infected with a common virus. Yet, it remains unknown how the inflammatory differences emerge. In complex systems, a change in a single component’s behavior impacts the temporal response of other system components. Computational modeling is a powerful tool for determining how changes in the behavior of immune system components lead to changes in the overall system response, and computational modeling can consider multiple hypotheses on how sex-specific immune responses emerge. Our research team develops knowledge-based, mechanistic models of the lung immune system and then employs state-of-the-art optimization and Bayesian inference approaches to rigorously determine how the lung immune system is differently regulated between cohorts. These models can be constructed to consider several factors impacting sex-specific regulation simultaneously, including the effects of hormones and sex chromosome dependent gene regulation. Here, we will discuss our most recent effort where we constructed a computational model of the mouse lung innate immune system and used computational models to determine how the immune system is differently regulated in male and female mice infected with a pandemic H1N1 virus. Our results suggest that not only do the rates of key immune processes, particularly those associated with interferon induction, have to be different between males and females, but that the effectiveness of therapeutic intervention using anti-inflammatory compounds is also sex-specific. The model is now being expanded to include additional immune components and we are currently developing strategies for incorporating hormone regulation.
Additional authors: Tatum S. McGeary, University of Pittsburgh, Chemical Engineering
Elizabeth R Duke
Fred Hutchinson Cancer Center
"Intrahost mathematical modeling of CAR T cells for HIV cure"
The primary barrier to cure from human immunodeficiency virus (HIV) is a reservoir of long-lived, latently infected CD4+ T cells. This reservoir causes viral rebound when people with HIV stop taking daily antiretroviral therapy (ART). One approach to reducing viral rebound is to use T cells with HIV-specific chimeric antigen receptors (CAR T cells) to target and destroy reservoir cells after they activate. In a pilot study, four rhesus macaques (n = 4) infected with Simian-HIV (SHIV) were given a single infusion of CAR T cells during ART to induce post-rebound control after ART was interrupted. Macaques that received CAR T cells had a lower viral peak after analytical treatment interruption (ATI) than before compared to controls. To model this intervention, we first developed ordinary differential equations (ODE) to recapitulate viral loads during primary infection and post-ATI rebound in the control animals. Then, using viral parameter values from the control model, we fit candidate ODE models to plasma viral loads and the CD4+ and CD8+ CAR T cell measurements from SHIV-infected macaques that received CAR T cells. Using the best fitting version of the model, we found the parameter that modulates CAR T cell proliferation in response to SHIV correlated with significantly lower post-ATI viral peaks. We simulated the data-validated model for each macaque to find conditions in which the CAR T cell infusion achieved ART-free, SHIV remission. Although gene and cell therapy strategies for HIV cure are in the initial stages, mathematical modeling might accelerate the success of these approaches.
Additional authors: Katherine Owens, Fred Hutchinson Cancer Center; Chloe Bracis, Université Gernoble Alpes; Christopher W Peterson, Fred Hutchinson Cancer Center; Hans-Peter Kiem, Fred Hutchinson Cancer Center; Joshua T Schiffer, Fred Hutchinson Cancer Center; E Fabian Cardozo-Ojeda, Fred Hutchinson Cancer Center
Vitaly V. Ganusov
University of Tennessee
"Using mathematical modeling to determine pathways of Mycobacterium tuberculosis dissemination in mice"
Tuberculosis (TB) remains a major disease of humans claiming lives of 1.6 millions in 2021. TB is caused by bacteria Mycobacterium tuberculosis (Mtb) that are transmitted by aerosol and initiate the infection in the lung. Over time, Mtb often disseminates from the initial infection site to other parts of the lung and in some cases, to extrapulmonary sites such as lymph nodes or spleen. In mice, infection with aerosolized Mtb also initially infects the lung but over time, Mtb is typically found in many extrapulmonary tissues such as lung-draining lymph nodes (LNs), spleen, or liver. The specific pathways of Mtb dissemination from the lung to other tissues, however, remain unclear. One study (Chackerian et al. 2002) measured dissemination of Mtb Erdman in two strains of mice (B6 and C3H) and suggested that Mtb first disseminates from the lung to the LNs and then to spleen and liver. We developed several alternative mathematical models describing how Mtb could disseminate from the lung to other tissues. Interestingly, we found that these data were insufficient to establish the Mtb dissemination pathway based on model fits; for example, the models in which Mtb spreads from the lung to LNs, and then from LNs for spleen/liver or from the lung to spleen/liver and then to LNs described the data with similar quality. However, the second model predicted extremely high rate of Mtb replication in the spleen/liver and high dissemination rate to the LNs; estimating these rates in future experiments may help falsify the model. The results were similar for two strains of mice (B6 and C3H). Interestingly, while Mtb causes stronger pathology in C3H mice, we also found that the rate of Mtb replication in the lung and other tissues were smaller in C3H mice than those in B6 mice. Our best models suggest that after lung infection, most Mtb (75% or more) exiting the lung disseminate to lung-draining LNs suggesting that control of Mtb replication in the LNs with an appropriate vaccine could be a strategy to prevent systemic dissemination of Mtb.
"Agent Based Model Investigating Latent and Naïve In Vitro M. Tuberculosis Infection Dynamics"
Prior to COVID-19, tuberculosis (TB) was the leading cause of death due to a single infectious agent – Mycobacterium tuberculosis (Mtb). The hallmark of Mtb infection is the formation of granulomas – unique microenvironments orchestrated by the immune response to contain Mtb and localize host-pathogen interactions. The host immune status is an important determinant in the formation of granulomas during Mtb infection. Approximately 90% of individuals infected with Mtb harbor granulomas that control bacterial spread, resulting in asymptomatic disease known as latent TB infection (LTBI). In vitro granuloma models have helped to understand granuloma development as they allow for highly controllable and high time-resolution investigations into granuloma formation. Specifically, an in vitro model that generates 3D granuloma-like structures through infection of human donor PBMCs with Mtb has shown that cells from LTBI donors better control Mtb growth compared to cells from naïve donors (those never exposed to Mtb before). But identifying mechanisms behind these differences is challenging, using experimental data alone. Here, we present a complementary approach using our agent-based model of these in vitro granulomas to help elucidate differences between LTBI and naïve host cell responses. Our computational model mimics Mtb infection through interactions between virtual macrophages, CD4+ T cells and Mtb. Mechanisms include Mtb growth, macrophage phagocytosis resulting in Mtb death or macrophage infection, macrophage and T cell activation, T cell proliferation, and cytokine/chemokine diffusion and degradation. The model is implemented using Repast Simphony. The model has been calibrated to published data from LTBI and naïve donor cells. Model outputs are calibrated to fall within A) 0.5-1.5x the experimental intracellular bacterial fold change for 3, 4, 5, 7, and 8 days post infection, B) 0.6-1.8x the experimental total cell fold change at day 7 post infection, and C) day of first granuloma formation (day 3 or 4 post infection for LTBI and day 5 or 6 post infection for naïve). We used Latin Hypercube Sampling (LHS) along with the Alternating Density Subtraction (ADS) method to perform iterative calibrations to identify a robust parameter region in which at least 75% of our parameter sets passed our criteria. We calibrate parameters for both LTBI and naïve datasets in parallel after the first LHS-ADS calibration iteration. Calibration was complete after 5 iterations with 89% and 84% parameter sets passing our criteria for LTBI and naïve groups respectively. Results show that starting at Day 2 post infection, LTBI-like simulations had a significantly higher number of activated TB-specific CD4+ T cells than the naïve-like simulations. This early activation of CD4+ T cells corresponded with an early increase in the number of total activated macrophages and activated infected macrophages in the LTBI-like simulations. Macrophage activation in the naïve group seemed to lag by approximately 3 days behind the LTBI group. Interestingly, the total number of infected macrophages was lower in the LTBI simulations, but despite less total infected macrophages throughout the infection, LTBI-like simulations controlled bacteria better than the naïve-like simulations having both less intracellular and extracellular bacteria by Day 8. Parameters that may have contributed to the quick activation of TB-specific CD4+ T cells and infected macrophages in LTBI-like simulations include lower CD4+ T cell deactivation probability and lower cytokine thresholds for macrophage activation. Our computational model, calibrated to LTBI and naïve experimental data, shows that the quick activation of TB-specific CD4+ T cells in LTBI-like simulations results in early and sustained activation of infected macrophages that leads to more bacterial control in LTBI-like simulations compared to naïve-like simulations. Despite having less overall infected macrophages, having a greater percentage of activated infected macrophages means that LTBI-like simulations can control Mtb infection better than naïve-like simulations.
Additional authors: Alexa Petrucciani1; Israel Guerrero2; Charles Renshaw2; Maria Montoya2; Eusondia Arnett2; Larry S. Schlesinger2; Elsje Pienaar1,3 1 Weldon School of Biomedical Engineering, Purdue University, West Lafayette, IN; 2 Texas Biomedical Research Institute, San Antonio, TX; 3 Regenstrief Center for Healthcare Engineering, Purdue University