I will review the role of persistent antigen, cross-reactive stimulation, bystander proliferation and reexposure to the maintenance of immunological memory. I will describe recent results on: (i) the role of cross-reactive stimulation on the longevity of memory; and (ii) the extension of earlier studies which focused on CD8 responses to humoral immune responses.
One of the defining features of the adaptive immune response is the ability to respond more rapidly and with greater vigour on re-exposure to infection. This feature of adaptive immunity is known as memory and is due to increased numbers of antigen specific T and B memory cells that remain after proliferation and differentiation in response to antigen. Long-term immunological memory depends on a self-renewing pool of antigen-specific T-memory (Tm) cells. It has been suggested that the Tm compartment is maintained by low-level reactivation with persistent or cross-reacting antigens but there is now persuasive evidence for the maintenance of specific memory in the absence of T-cell receptor (TCR) stimulation. In vivo labelling experiments in mice and humans have shown that ~1-5% of both CD4 and CD8 memory cells are in cycle at any one time and a similar background proliferation has recently been observed in memory B cells. This background homeostatic proliferation of CD8 Tm cells occurs in response to interleukin-15 (IL-15) and a similar (cytokine-driven) mechanism is thought to be responsible for CD4 memory T-cell proliferation, probably in response to IL7. Homeostasis of the Tm population therefore depends on balancing cell loss through death and differentiation into effector cells against input from antigen activation of naive cells and the background of homeostatic proliferation. Competition models where the Tm cells compete for space or essential growth factors have been considered but an alternative mechanism based on density dependent tm cell death through apoptosis (fratricide) has also been proposed. In this paper we compare mathematical models of these two distinct processes with the aim of distinguishing between them. We also use the fratricide model to investigate the loss in CD4 T cells that occurs in HIV infection.
Work done in collaboration with Andrew Yates and Jaroslav Stark.
V(D)J recombination involves the introduction of double-strand DNA breaks and so must be targeted specifically to appropriate sites of cleavage. This targeting is mediated by recombination signal (RS) sequences adjacent to each V, D and J gene segment. Consensus sequences have been used to describe the first seven and last nine RS nucleotide positions, but only 13% of mouse RS contain a consensus heptamer and nonamer, and the entropy averaged over nucleotide positions in an alignment of mouse RS is 0.5 and 0.67 for RS with 12- and 23-bp spacers, respectively. We hypothesized that correlation between RS nucleotides could confer sequence-specificity to the recombinase-RS interaction that would not be apparent to models of RS assuming independence of RS nucleotides. We developed an algorithm for the identification of correlations between positions in a DNA sequence alignment, including higher-order correlations and those between non-adjacent positions, and constructed a probability model of mouse RS based on the identified correlations. We compared our model (RIC) to an order 0 and an order 1 Markov model and found that while the three models predict the recombination efficiency of any RS-length sequence equally well, the RIC model is better at recognizing RS and cryptic RS in genome scans.
Work done in collaboration with Marco Davila, Garnett Kelsoe, and Thomas B. Kepler.
Our laboratory employs a diversity of biophysical, biochemical and cell biological techniques in an effort to understand mast cell signaling mediated through IgE and its high affinity cell surface receptor Fc-epsilon-RI. Binding and crosslinking of the receptors by a cognate multivalent antigen initiates a signal cascade that culminates in degranulation and the release of various effecter molecules by these cells. We are examining the dynamics of molecules involved in this signaling pathway and the role of the plasma membrane in modulating their interactions. We use structurally well-defined nanometer length scale ligands to probe the binding of IgE to its antigens and we have described this binding using realistic mathematical models. We utilize fluorescence spectroscopy and quantitative fluorescent imaging of live cells to characterize the interactions between the cell surface receptor and other membrane bound and intracellular signaling proteins. We also use model membrane systems as a basis for understanding the role of lipid rafts - ordered domains in the plasma membrane hypothesized to segregate signaling molecules by their preferential partitioning. This talk will describe some of our recent results in these areas.
Work done in collaboration with David Holowka and Barbara Baird.
We fit a mathematical model to data characterizing the primary cellular immune response to LCMV. The data enumerate the specific CD8+ T cell response to six MHC class I restricted epitopes, and the specific CD4+ T cell responses to two MHC class II restricted epitopes. The peak of the response occurs around day eight for CD8+ T cells and around day nine for CD4+ T cells. By fitting a model to the data we characterize the kinetic differences between CD4+ and CD8+ T cell responses, and among the immunodominant and subdominant responses to the various epitopes. CD8+ T cell responses have faster kinetics in almost every aspect of the response. For CD8+ and CD4+ T cells, the doubling time during the initial expansion phase is 8 h and 11 h, respectively. The half-life during the contraction phase following the peak of the response is 41 h and 3 d, respectively. CD4+ responses are even slower because their contraction phase appears to be biphasic, approaching a 35 d half-life eight days after the peak of the response. The half-life during the memory phase is 500 d for the CD4+ T cell responses, and appears life-long for the six CD8+ T cell responses. Comparing the responses between the various epitopes we find that immunodominant responses have an earlier and/or larger recruitment of precursors cells before the expansion phase, and/or have a faster proliferation rate during the expansion phase.
Work done in collaboration with Dirk Homann and Alan S. Perelson.
Recent advances indicate that the process of signaling through cell surface receptors involve highly connected networks of interacting components. Understanding the often counterintuitive behavior of these networks requires the development of mathematical models. I will discuss the application of mathematical models to understanding signaling through the immune recognition receptors. Simple models, like kinetic proofreading and serial engagement, which ignore the details of the signaling machinery, have provided considerable insight into how ligand-receptor binding properties affect signaling outcomes. More detailed models that include specific molecular components and interactions beyond the ligand and receptor are difficult to develop, but offer the hope that the vast information we have about signaling may one day be integrated into models that predict the full spectrum of signaling behavior. Both types of models will be reviewed.
In this presentation I will highlight the role of the "conceptualist" in theoretical immunology. Focusing on T-cell biology, I will describe a consistent, long-standing attempt at building a conceptual framework. (a) Balance of growth and differentiation. Self-renewal is usually regarded as an exclusive property of the earliest, most primitive cells, designated stem cells. A counter proposition is that self-renewal is a regulated activity. (b) Concomitant regulation of T cell activation and homeostasis. We proposed a dynamic, bidirectional interplay between the changing structure and size of the population of activated cells on the one hand and adaptive changes in the function of individual cells on the other. Resilience is facilitated (i) by feedback controls; (ii) by a selective, regulated incorporation of RTE into the naive population; (iii) by a selective, regulated incorporation of "homeostatically" activated naive cells into the memory pool; and (iv) by a structured replacement of memory T cells by the progeny of naive cells. (c) Proximal immune activation and HIV transmission. The dynamics of immune activation bursts and virus replication are intimately related. In particular, the concept of structured replacement of memory cells sheds light on the dual role, protective and pathogenic, of chronic immune activation in HIV infection. (d) "Activation-threshold tuning" has been proposed as a major regulatory mechanism, involved in the inhibition of autoimmune reactivity, in the control of immune responses, and in the cell-density dependent regulation of T-cell numbers. When lymphocytes are subject to recurrent stimulation they may respond by proliferation and/or differentiation or, alternatively, they may adapt and become less responsive. The onset and maintenance of the latter mode depend on certain quantitative characteristics of the stimulation, which ensure that "perturbations" of the balance between "positive" and "negative" intracellular signals do not exceed the cell's activation thresholds. (e) Cognitive capacities of lymphoid cells and lymphoid cell-organization. We proposed that lymphocytes learn from experience in real time and that the organization of groups of cells - the functional units - is guided by feedback from the microenvironment and/or the neuroendocrine system, providing "quality control".
Responses such as progression through the cell cycle, programmed cell death or survival, and cell migration are tightly controlled by intracellular reaction pathways that transduce signals perceived by cell surface receptor proteins. To selectively influence cell behaviors, then, a quantitative, mechanistic understanding of signal transduction will be needed. As a model system, we have focused experimental and theoretical approaches to the understanding of platelet-derived growth factor (PDGF) receptor signaling in fibroblasts, a central process in dermal wound repair that progresses in concert with the innate immune system (albeit on a slower time scale). A prominent feature of this system is the activation of phosphoinositide (PI) 3-kinase, which produces specific lipid second messengers (3' PIs) in the plasma membrane. We are actively studying three quantitative aspects of this pathway: 1) the networking of PI 3-kinase signaling with other pathways; 2) its spatial regulation in cells exposed to PDGF gradients, an important determinant of directed fibroblast migration; and 3) the integration of intracellular signaling, cell response, and fibroblast population dynamics during would healing. Given the similarities among signaling networks activated by diverse receptor families, insights from such studies are expected to apply to other cell/receptor systems. Along those lines, pathway crosstalk in interleukin-2/3/4 signaling in B and T cells will also be discussed.
Cells of the immune system are highly regulated by signals received from numerous cell surface receptors. We have developed a quantitative framework, the cellular calculus, for dissecting the manner in which such signals can affect T and B lymphocyte behaviour in vitro. An assumption of this platform is that lymphocytes behave as if composed of independent stochastic 'machines' governing the times to divide and times to die (Gett and Hodgkin, Nat. Immunol. 2000. 1:239). As a consequence of the stochastic variation many alternative outcomes are possible for individual cells, however the population is highly predictable. Cytokines that affect proliferation rate and survival can be shown to 'add' together in quantitative manner yielding surprisingly large effects on final cell behaviour. Similar stochastic rules can be applied to differentiation. For example generation of IgG secreting cells from naive precursors is highly predictable. The probabilities of isotype switching and development into secreting cells change with successive cell divisions and interleave independently. Cytokines alter the probability of each differentiation event while leaving intact their independent assortment. As a result cellular heterogeneity arises automatically as the cells divide (Hasbold et al. 2004. Nat. Immunol 5:55).
We have developed algorithms and computer based tools for simulating the effect of combinations of signals on lymphocyte responses to illustrate the manner of operation of costimulation and cytokine based regulation. Furthermore, our tools allow time series data of cell division and cell number to be dissected to provide kinetic parameters such as average time to first division, subsequent division time and the proportion of cells that die in each division (Deenick et al. 2003. J. Immunol. 2003:4963).
The cellular calculus modeling framework enables a quantitative dissection of proliferation, survival and differentiation data to accurately predict and simulate apparently complex cell behaviour.
Work done in collaboration with Elissa K. Deenick, Jhagvaral Hasbold, Edwin D. Hawkins, Hilary F. Todd, Lynn M. Corcoran, David M. Tarlinton, and Stuart G. Tangye.
There is significant concern about the dramatic increase in the incidence of inflammatory diseases in the past two decades in "westernized" countries. The most notable examples are allergies/asthma, diabetes, inflammatory bowel disease, autoimmunity and heart disease/atherosclerosis. A common feature among these diseases is that they are all diseases characterized by an over-exuberant inflammatory response in the diseased tissue. Unfortunately, there is a significant gap in our understanding of how the immune system normally modulates and down-regulates inflammatory responses. This is in contrast to the mechanisms of inflammation where there is a significant body of information about both the development and amplification of inflammatory responses. Why is the incidence of inflammatory diseases increasing? How are "normal" inflammatory responses, which are required to effectively handle infectious microbes, kept under control and shut down? These are the global questions our research is addressing, focusing on inflammatory responses in the airways, i. e. allergies and asthma. The significant increase in allergy and asthma in westernized countries correlates with the widespread use of antibiotics and alterations in fecal microflora. Antibiotics also lead to growth of the yeast Candida albicans. We have previously published a number of reports demonstrating that fungi can secrete potent prostaglandin-like immune response modulators (oxylipins). We have developed a novel mouse model of antibiotic-induced gastrointestinal microflora disruption that includes enhanced gastrointestinal yeast colonization and have demonstrated that antibiotic therapy can drive the development of a T cell-mediated airway allergic response to mold in genetically disparate "normal" mice. The underlying mechanism of microbial-host immunologic communication remains to be elucidated but there is data to suggest that the gastrointestinal microflora may play a role in the development of regulatory (Th3/Treg) responses, which are potent regulators of inflammatory processes.
Cell signaling pathways interact with one another to form networks. Such networks are found in all cell types and we have proposed that there may be an overall general format for signaling networks in different cell types. The central signaling network regulates multiple cellular machines that results in the expression of physiological functions. Such networking results in the appearance of regulatory motifs that facilitate signal processing and consolidation and consequently increase the ability of the network to evoke biological responses. We have analyzed a meso-scale network of a mammalian cell using graph-theory approaches. Results from these analyses will be presented. We have also analyzed the dynamics of smaller networks by biochemical computation. Such analysis highlights interesting features of signaling networks. These include the presence of gates that can allow for signal prolongation and positive feedback loops that can function of bistable switches. The physiological consequences of these types of regulation in T cell functions will be discussed.
Antigenic peptides bound to Major Histocompatibility Complex II (MHC II) molecules are the ligands that activate CD4+ helper T lymphocytes. Although production of peptide-MHC II complexes has been studied extensively in vitro, much less is known about the anatomic constraints that govern this process in vivo. We have approached this problem by developing methods that allow in vivo detection of a foreign antigen, peptide-MHC II complexes derived from this antigen, and CD4+ T cells expressing T cell antigen receptors specific for this peptide-MHC II complex.
Within several hours of subcutaneous injection, antigen was found in the draining lymph nodes within a network of thin conduits composed of collagen fibers and wrapped with reticular fibroblasts that run through the T cell-rich area. Nearby dendritic cells acquired free antigen from the conduits, displayed antigen-derived peptide-MHC II molecules, interacted with antigen-specific na´ve CD4+ T cells, and caused these T cells to produce IL-2 and proliferate. About 12 hours after antigen injection, dendritic cells displaying large numbers of antigen-derived peptide-MHC II molecules migrated from the subcutaneous injection site via a G-protein-dependent mechanism and interacted with the antigen-specific CD4+ T cells. Presentation of peptide-MHC II complexes by these migrants sustained expression of the IL-2 receptor and was necessary for the T cells to differentiate into cells capable of causing a later delayed-type hypersensitivity reaction. These results demonstrate that in the case of a soluble subcutaneous antigen, CD4+ T cells are first stimulated in the draining lymph nodes by peptide-MHC II complexes displayed by dendritic cells that acquire the antigen from the conduits when in the lymph nodes, and several hours later by different dendritic cells that migrate from the injection site.
The hallmark of the adaptive immune system, exclusive to vertebrates, is the somatic diversification of antigen receptor genes through V(D)J rearrangement, somatic hypermutation and gene conversion. In contrast, there has been no indication that the innate immune system shared by invertebrates and vertebrates alike diversify by comparable processes. Here, we report, in the gastropod snail Biomphalaria glabrata, the presence and expression of diverse immunoglobulin superfamiy (IgSF)-encoding genes termed fibrinogen-related protein genes (FREPs) generated by point mutation and a process resembling gene-conversion or rearrangement. We hypothesize a mechanism present in snails, capable of generating a diverse family of IgSF-encoding molecules and involved in inducible host defense.
The statistical methods we developed for the inference of diversification mechanisms in the absence of information on the unmodified germline or donor genes are based on the minimum description-length (MDL) model selection criterion. I will describe both the biological results with their implications for our understanding of the evolution of immunity, and the mathematical techniques used to obtain them.
Work done in collaboration with Si-Ming Zhang, Coenrad Adema, and Sam Loker.
To understand the nature of competition and other forces shaping the adaptive B cell repertoire during immune and autoimmune responses, precise estimates of the hypermutation rate and the frequency of lethal mutations are critical. Microdissection studies of mutating B cell clones provide an opportunity to estimate these values more accurately than previously possible. Each microdissection provides a number of clonally related sequences that, through the analysis of shared mutations, can be genealogically related to each other. The shapes of these clonal trees can be quite distinct for different responses (e.g., immune versus autoimmune). However, it has been difficult to relate these differences to underlying biological mechanisms such as the hypermutation rate and the frequency of lethal mutations (which have an important influence on the shape of clonal trees).
We first developed two different methods to estimate the hypermutation rate based on experimentally derived clonal trees. Both are based on a model of B cell clonal expansion (one is analytical while the other makes use of a stochastic computer simulation). These methods predict comparable mutation rates in an anti-hapten response to (4-hydroxy-3-nitrophenyl)acetyl (NP) and an autoimmune response (0.9-1.1 x 10-3 bp-1 division-1 and 0.7-0.9 x 10-3 bp-1 division-1 respectively). However, the frequency of mutations that are lethal to the cell was an assumption in these original methods. We have now extended our methodology so that the lethal frequency can be estimated directly from the experimental data along with the hypermutation rate.
By testing our extended methods on synthetic data sets, we show that precise estimates of both the hypermutation rate and the frequency of lethal mutations can be made. We also applied these improved methods to various sets of experimental data. In addition to comparing differences between these responses, we have investigated the effect of various experimental decisions such as the microdissection pick size.
Scientific advances continue to identify members of the chemokine supergene families as biologically diverse mediators of important immunologic and physiologic events. While initial investigations originally defined the biological activity of chemokines as proteins with novel chemotactic activity for specific sub-populations of leukocytes, data now supports a much broader biological role for the chemokines. The chemotactic activity of chemokines for specific leukocyte sub-populations is, in itself, an important activity, as this response provides a mechanism for the successful delivery of the appropriate leukocyte population from the lumen of the vasculature to a site of inflammation. This biological response provides the means for the accumulation of either granulocytes at foci of acute inflammation, via the activity of CXC chemokines, or the accumulation of mononuclear cells at foci of chronic inflammation, via the activity of CC chemokines. However, leukocyte chemotaxis may not be the only, or the most important, activity of the chemokine family members. A variety of reports have stressed the key role of chemokines in a variety of physiologic and pathologic situations, which may provide mechanisms for activating cytokine networks, altering the expression of adhesion molecules, increasing cell proliferation, regulating angiogenesis, promoting viral-target cell interactions, increasing hematopoiesis, stimulating mucus production, increasing the metastatic potential of tumor cells, and activating the innate immune system. The importance of chemokines as a contributing player to the immune response is further underscored by investigations that have identified viral genes that encode chemokine binding proteins. Importantly, chemokines have been shown to participate in the progression of chronic inflammation by influencing mononuclear cell chemotaxis, hematopoiesis, angiogenesis, stromal cell proliferation, matrix deposition and lymphocyte polarization. This latter activity is especially important, as specific chemokine ligand/receptor pairs have been identified in type 1 (Th1) versus type 2 (Th2) immune responses. These observations have played an important role in the design of efficacious small molecular weight antagonists to therapeutically target specific chemokine receptors, as these receptors and their ligands are likely to participate in the evolution of chronic immune responses.
The Virtual Cell is a computational modeling framework that has been designed for cell biologists. It facilitates the organization of experimental data into quantitative hypotheses and the generation of predictions from them. A key feature of the Virtual Cell is that it permits the incorporation of experimental microscope images within full 3D spatial models of signal transduction networks. It also serves as a computational tool for the analysis of experiments such as local photorelease of caged second messengers or the translocation of GFP-linked signaling molecules. Recently added features include facilities for representing electrophysiological models, capturing reaction data from public databases, and sharing models both via access control lists and export/import of XML documents. The use of the Virtual Cell will be illustrated with several example models.
The immune system depends on coordinated intercellular interactions of different cell types. The "immunological synapse" (IS) is the contact area of cell-cell conjugates where information is transferred between immune system cells by clustering of signaling molecules. The cellular and molecular events taking place during contact occur in sequential stages that involve dramatic changes in cell polarity and dynamic redistribution of cell membrane receptors. Recent studies have emphasized the importance of cell asymmetry, cytoskeletal dynamics, membrane organization and molecular patterning in setting thresholds for the T cell activation process. However, the mechanisms driving the formation of T cell-antigen presenting cell synapses are currently not understood. Laboratory experiments tend to isolate and study one or two molecular or cellular interactions at a time. Mathematical and computational modeling is increasingly becoming an essential tool to integrate the information from many experiments, thus complementing experimental and conventional techniques. We created a simulation model that enables us to explain the dynamics of the different cellular events taking place sequentially in immune synapses. Using this simulation, we offer several following new insights into IS behavior, which have not been answered so far by mathematical models.
Work done in collaboration with Shulamit Kotzer, Mali Salmon-Divon, Catarina Rodrigues De Almeida, Petter H÷glund, and Daniel Davis.
With the advent of micro-array and other technologies, immunologists have to deal with far more detailed information. Immunology has entered a new phase where modeling can become mainstream. Modelers on their side have to develop receiving structures for all that information to help analyzing the experiments.
In this talk a framework will be described to analyze the information flow accompanying the cognate interaction, based on analysis of signaling cascades and their interaction with the regulation of the genetic activity.
Dendritic cells (DC) are the most efficient antigen presenting cells (APC) and have been successfully used to induce immune responses to tumors, viruses and alloantigens. DC are known to influence the differentiation of na´ve T helper cells into Th1 or Th2 effectors, as well as regulatory T cell populations. Some of the features that determine the ability of DC to differentially activate Th1 or Th2 differentiation are known, but new molecules of importance are likely to be defined. The ability to predict whether a given DC population will reliably induce either Th1 or Th2 responses would be extremely valuable in the context of tumor immunology, autoimmunity, transplantation and allergic diseases in which DC are either being used or considered as therapeutic options. We have been studying the therapeutic potential of DC populations to prevent autoimmune diabetes in the non obese diabetic (NOD) mouse and have identified a DC subset that can protect young pre-diabetic NOD mice from the development of diabetes. The therapeutic DC populations altered the Th1/Th2 balance in vivo and induced a population of Th2 cells that may be responsible for the observed protection. We are using microarray analysis to identify genes that determine the therapeutic potential of DC subsets. Based on these data we have used a novel computer simulation system to analyze this experimental system with the aim of identifying novel mechanisms that define the therapeutic potential of DC in immunotherapy.
Despite the vast amount of information about the immune system has accumulated, we still understand very little about how the immune system functions to provide protection or to promote pathologies. While we are adept at teasing out and characterizing the various pieces of complex immune responses, we are relatively inept at reassembling these pieces into a coherent perception of immune system function. At its most basic level, the immune system is a large collection of many simple, autonomous members, each of which reacts individually to the state of the local environment according to a set of internal rules. As such, it qualifies as a complex, adaptive system, and it displays the inherent characteristics of such systems, i.e., the ability to adjust to stimuli, to maintain coherence in the face of change, and to learn. These are important competitive advantages that complexity bestows on the immune system, making it remarkably effective under a variety of conditions.
Clearly, the immune system is a rich network of interactive agents. While this is commonly acknowledged, it is routinely ignored. Immune responses are usually treated as small, isolated, linear arrays of cause-and-effect activities. The problem is that most biologists do not know how to deal with complex, networked processes. They find them too complicated and too unpredictable. The fact that causality and outcome are often obscured by the many different functional options within a network poses a major problem for investigators. Nevertheless, a true understanding of immune function requires an appreciation for how complex, adaptive networks operate.
Unfortunately, the human brain has difficulty dealing with networked activities. However, it has conceived a tool for which networked activities provide no problems at all: the computer. Unlike the human brain, a computer can patiently and efficiently track an unlimited number of interacting agents, and follow them to an outcome. This leads to the question: Can we use a computer to study the function of complex adaptive systems, in general, and the function of the immune system, in particular? This is a central question of our current studies in theoretical immunology.
To employ the computer for this purpose, we used the Repast software library to develop a prototypic computer simulation of the immune response to a generalized viral infection. We needed to simulate a phenomenon that involved large numbers of different elements operating in a defined space, and these elements needed to interact with each other in defined ways via direct contact or via secreted signals that diffused through the environment. The Repast software was exceptionally useful for the creation of a computer simulation of this response. It is important to note that we did not attempt to model all of the elements of an immune response. Rather, we chose to simulate its design principles. Regardless of the accuracy of he simulator with regard to the operative features of the immune response, the simulation represents a complex adaptive system that functions in silico, and can be studied as such. Our current studies involve the refinement of this simulator to more accurately reflect the design principles of the immune system. We plan to use the simulator to explore key, formative patterns of agent behavior that develop within complex adaptive systems, to evaluate how information flows through complex adaptive systems and how it is used for decision making as immune responses evolve, and to evaluate the strengths and weaknesses of clinical and experimental tools (such as biopsy and gene chip analysis) that are currently in use.
Work done in collaboration with Virginia Folcik.
I will summarize work using various labeling approaches, eg Brdu, d-glucose, CSFE, to dissect out the kinetics of T cells both during health and during viral infection.
The CFSE dye dilution assay is widely used to determine the number of divisions that labeled cells have undergone both in vitro and in vivo experiments. The literature contains several methods for estimating parameters of cell division and death from CFSE data. I will describe these methods, discuss their performance with different data sets, and analyze their advantages and limitations. I will also discuss the limits on the amount of information provided by the CFSE data.
Work done in collaboration with V. V. Ganusov, S. S. Pilyugin, R. de Boer, K. Murali-Krishna, R. Ahmed, and R. Antia.
The contribution of the thymus for T-cell homeostasis is still poorly understood, despite some recent experimental and modeling advances. I describe a mathematical model of the peripheral effects of experimental removal of the thymus (thymectomy) in macaques. By monitoring the changes in phenotypic T cell markers as well as in the numbers of T cell receptor excisional circles (TREC), a marker for recent thymic emigrants (RTE), we have evidence that surgical thymectomy in juvenile macaques has little quantitative impact. However, the thymic output was measurable at 0.32% and 0.21% per day for CD4+ and CD8+ cells, respectively. I will compare these values with other parameters in T-cell dynamics and discuss their relevance for T-cell homeostasis.
The specificity and sensitivity of T cell recognition is vital to the immune response. Ligand engagement with the T cell receptor results in the activation of a complex sequence of signalling events, both on the cell membrane and intracellularly. Feedback is an integral part of these signalling pathways, yet is often ignored in standard accounts of T cell signalling such as McKeithan's kinetic proofreading model. In this talk, we shall present a mathematical model which shows that these feedback loops can explain the ability of the T cell receptor to discriminate between ligands with high specificity and sensitivity, as well as provide a mechanism for sustained signalling. The model also explains the recent counter-intuitive observation that endogenous 'null' ligands can significantly enhance T cell signalling.
Work done in collaboration with Jaroslav Stark, Cliburn Chan, and Andrew J. T. George.
Mathematical modeling of signal transduction pathways has opened the possibility of predicting a wide range of cellular behaviors. Historically most of these models were cast as ordinary differential equations (ODEs) that described how the concentrations of molecular species change with time. A relatively new alternative to ODE modeling is a Bayesian network representation of signal transduction. Bayesian networks are probabilistic graph models that describe causal relationships between variables. In the context of signal transduction, Bayesian networks have the advantage that they can describe systems that are noisy, nonlinear, and underspecified. In addition, this class of models can be applied to systems where no experimental kinetic data is available. In this talk I will describe how Bayesian networks can be applied to a dataset describing stem cell differentiation. The Bayesian network derived from this dataset accurately predicts a number of known signal transduction pathways and suggests novel interconnections between these established pathways. Next, I will show how a Bayesian network can be used to efficiently and rationally choose future experiments based on the existing data and the current model. In this way, Bayesian networks provide a natural alternative or intermediate point between raw experimental data and fully mechanistic, kinetic models.
The thymus is the primary lymphoid organ supplying new lymphocytes to the periphery. Recent advances in characterizing thymic function confirm the importance of thymus to T-cell diversity in the periphery of both children and adults during both health and disease. Clinical and morphologic studies of HIV-infected patients indicate that the thymus is affected by HIV. Thymic dysfunction and thymic involution occur during HIV disease and have been associated with rapid progression in infants infected perinatally with HIV. In vitro information of thymic organ culture, thymic epithelial cell culture, the SCID-hu mouse system and SHIV infection of primates have supported HIV-induced thymic damage. The mechanisms underlying this could be many, including direct thymocyte killing by the virus, apoptosis, or disruption of thymic stromal architecture. T cell receptor excision circles (TREC) have been developed as a marker of new thymic emigrants. Decreases in TREC concentrations have been found in both HIV-infected pediatric and adult patients. Mathematical models have suggested that thymic infection in children is more severe than in adults, particularly during infection with strains that use CXCR4 as coreceptor. Recent evidence suggests that thymic recovery may be achieved in some patients as a result of potent antiretroviral therapy. Extensive thymic damage may, however, hamper immune reconstitution, particularly in pediatric patients. I will summarize evidence for thymic involvement during HIV infection in children and in adults, the role of thymic infection in disease progression, and the contribution of the thymus to immune restoration following potent antiviral therapy. Immunologic interventions aiming at restoring thymic function in AIDS patients will also be reviewed.
PathSim (PathogenSimulation) is a systems biology modeling tool designed for the exploration of human host responses to viral pathogens. It is based on two key components, a multi-scale anatomical viewer and an agent-based simulation engine. Although our engine will eventually be generic, our first implementation involves Epstein-Barr virus (EBV) infection of the Waldeyer's tonsilar ring. In this model, the anatomy is represented as a collection of tissues, each with its own properties that determine both 3D visual rendering and agent behavior during simulation. Agents represent the active and potentially mobile elements within the simulation such as virions (EBV) and immune cells (B and T lymphocytes). Agents are localized at mesh points representing a small region within a tissue. Each mesh point is assigned to a class that further refines its properties. Vertices that form a 3-D mesh controlling the local motion of agents connect the mesh points. In order to account for agent transport via blood and lymph fluids a simple model of the circulatory and lymphatic systems is also represented in the simulation.
The simulation engine is constructed as a discrete-time, cellular automaton. At each time step agent motion, activation, aging, and interaction are controlled by a set of stochastic state transition rules. These rules are based on our current understanding of host responses to EBV. The simulation results to date demonstrate qualitative agreement with data reported in the literature for the acute phase of EBV infection, acute infectious mononucleosis. A comparison to the standard model is included.
Work done in collaboration with J. McGee1, K. Lee1, R. Laubenbacher1, and K.A. Duca1.
In the life sciences, the development of rigorous models and databases of biological phenomena provides major benefits for biological research, drug design, and education. A grand goal in biology is to produce integrated information-rich biological databases that capture the complexity of reality. A common class of such databases can be characterized as integrating diverse information including: spatial representations of physical systems and phenomena, abstract data such as gene expression data and annotations, temporal dimension for time series, multiple levels of scale (from anatomical to cellular to molecular), and multiple runs of simulation output, and experimental results.
However, the current major shortcoming is the lack of effective user interfaces and visualizations for information-rich databases that enable biologists to gain insight. The true utility of the databases will come to fruition when biologists are able to explore and navigate them and relate effects between space, abstract data, and time across levels of scale. Current virtual environments and information visualizations lack the usability and support for such complex information-rich databases.
PathSim is an example of an information-rich model with associated databases. The main goal of PathSim is to model a variety of viral agents in human and animal hosts, from initial infection to viral clearance. PathSim allows an end-user to explore the physiology and dynamics of infections and immune system response. As an interface to this system, we are constructing and evaluating information-rich virtual environments (IRVEs) for the PathSim project. This interface framework can also be applied to other similar information-rich databases in the life sciences that share these characteristics.
An IRVE combines the capabilities of virtual environments and information visualization to support integrated exploration of both spatial and embedded abstract data. Biologists can view the simulated physical structures of the model in a 3D virtual environment, interact with visually embedded abstract data, navigate across levels of scale, choose data for display, and control simulation run management all within an integrated environment. For example, a user might decide to examine the effect of titer on the course of infection. Within the IRVE, the user deposits virions in the locations to be infected. After the simulation commences, the user revisits the IRVE to view signaling events initiated by virus deposition at the molecular level. Later, the user examines how fast the virus is spreading, killing cells, or recruiting immune cells to the vicinity. All activities are viewable in the virtual environment, with interactive links and data export to a suite of analytic tools also possible.
This work is discovering critical new methods for display and interaction in multi-scale IRVEs that are usable and useful for biologists. The system user interface operates on a wide range of hardware, from standard desktop displays to high-performance immersive CAVEs. The system will eventually be public and open for use in other applications.
Work done in collaboration with N.F. Polys, D. Bowman, K.A. Duca, R. Laubenbacher, and C. North.
Binding of N-formyl peptide ligands to the N-formyl peptide receptor (a G-protein coupled receptor) on human neutrophils initiates a signal transduction cascade that begins with G-protein activation and ultimately leads to actin polymerization (necessary for chemotaxis) and oxidant production (necessary for killing bacteria). Using this system, we have characterized a variety of agonist and antagonist ligands, gathering kinetic data not only for ligand/receptor binding and processing but also for downstream signaling events (G-protein activation, actin polymerization, and oxidant production). Ligand-receptor binding, receptor upregulation, internalization, and desensitization occur on similar time scales (within seconds of ligand presentation). Maximal actin polymerization occurs within ten seconds and maximal oxidant production within 300 seconds, well before equilibrium binding. With these data in hand we hypothesize that these receptor level events greatly determine the character of cellular responses. We use these data in models to answer the following questions: What binding model is consistent with the data? We find that a two-site binding model is sufficient to describe the binding data for all ligands, but is not able to predict differences in ligand potency. What ligand specific parameters are important in signal generation, and what are possible explanations for differences in ligand potency? For example, ligand-receptor association and dissociation rate constants are highly correlated with response generation, but we also find that ligand-dependent receptor conformations must be included in the model to account for differences in potency. Given that we have a model that accurately fits the data, can we predict cellular responses? And finally, how can these models be used to suggest new experiments to further elucidate the mechanisms of signal transduction?
Mature Dendritic Cells (DC) are some of the most prolific producers of IL-12, a major cytokine regulator of T helper and NK cell responses. IL-12 is a 70 kDa heterodimer (p70) comprised of independently-regulated disulfide-linked 40 kDa (p40) and 35 kDa (p35) subunits. Since the characterization of IL-12 in 1989, the majority of IL-12p70/p40 related PubMed citations refer to only one subunit. We asked whether the bioactivity of IL-12 could be determined from the concentration of the p70 subunit alone or whether the relative concentrations of both p40 and p70 and their competitive binding with the IL-12 receptor are essential for determining IL-12 bioactivity. Using a mathematical model of IL-12 subunit production by DC, we explored the effects of varying levels of prototypical Th1 (IFN-), Th2 (IL-4), and inflammatory mediators (PGE2) on the potential DC-derived IL-12 signal. Simulations using this model illustrate that the concentration of p70 alone is not indicative of IL-12 bioactivity. Rather, the bioactivity of IL-12 produced by mature DC is the cumulative effect of individual IL-12 subunit production and the competitive interaction of these subunits with the IL-12 receptor. Furthermore, the data suggests that species derived from IL-12 p40 may function as physiological antagonists of IL-12p70, particularly in the presence of PGE2.
When the body is infected with a pathogen such as bacteria, it mounts an acute inflammatory response to rid the body of the invading pathogen and restore health. Uncontrolled inflammation results in tissue damage, organ dysfunction and death whereas an inadequate response could result in persistent or recurrent infection. Though much has been learned about the physiological pathways of acute inflammation, due to the complex nature of the process, this knowledge has not led to effective therapies against improper systemic inflammation. We consider a simple 3 dimensional ODE model (consisting of an invading pathogen and early and late pro-inflammatory mediators) that simulates the various clinical outcomes. We analyze the bifurcation plots to determine the effect of the key parameters in taking the system to different outcomes and discuss various therapeutic strategies suggested by the model.
There is considerable empirical evidence to show reduced immune competence in older individuals. To date, the mechanisms which underly these experimental observations are unclear and mathematical modelling may provide a powerful tool in testing ageing hypotheses.
A common mathematical approach is to use a system of differential equations to describe the interactions between components of the immune system. Such models typically assume the components are well-mixed and that reactions between components occur at a prescribed constant rate. These assumptions may not be valid for immune reactions within lyphoid tissue where the distribution of reactants is both sparse and spatially inhomogeneous.
Here we use an alternative approach: we model the early stages of the immune response on the basis of evidence from recent studies which suggest immune surveillance by T cells is a stochastic process. We also describe how this approach can be used as a new framework for modelling ageing of the immune system.
Work done in collaboration with SP. Preston, SL Waters, OE Jensen, DI Pritchard, and PR Heaton.
We have developed a mathematical model to predict the outcome of macrophage infection with Mycobacterium tuberculosis relative to the biochemical state of the macrophage. The model consists of two physical scales. One captures biochemical macrophage activation of iNOS-derived nitric oxide coupled to regulation of intracellular iron. The second physical level represents the intracellular population of mycobacteria responsive to, and influencing, the macrophage state. Using this dual-level model we examine context-dependent responses to different macrophage activation states. We applied statistical sensitivity analyses to elucidate important model features in each context. Controlled comparisons between wild-type and various knockout cases allowed us to test the influence of particular interactions on bacterial load. We draw conclusions about the nature of the interaction between M. tuberculosis and its host macrophage with these results and suggest future experiments to test our predictions.