ORIGINAL ARTICLE doi:10.1111/evo.12171 IMMUNE EVASION AND THE EVOLUTION OF MOLECULAR MIMICRY IN PARASITES Amy Hurford1,2,3,4 and Troy Day1,5 1Department of Mathematics and Statistics, Queen’s University, Kingston, Ontario, Canada 2Department of Biology, Memorial University of Newfoundland, St. John’s, Newfoundland and Labrador, Canada 3Department of Mathematics and Statistics, Memorial University of Newfoundland, St. John’s, Newfoundland and Labrador, Canada 4E-mail: [email protected] 5Department of Biology, Queen’s University, Kingston, Ontario, Canada Received December 3, 2012 Accepted May 8, 2013 Data Archived: Dryad doi: 10.5061/dryad.3vf7p Parasites that are molecular mimics express proteins which resemble host proteins. This resemblance facilitates immune evasion because the immune molecules with the specificity to react with the parasite also cross-react with the host’s own proteins, and these lymphocytes are rare. Given this advantage, why are not most parasites molecular mimics? Here we explore potential factors that can select against molecular mimicry in parasites and thereby limit its occurrence. We consider two hypotheses: (1) molecular mimics are more likely to induce autoimmunity in their hosts, and hosts with autoimmunity generate fewer new infections (the “costly autoimmunity hypothesis”); and (2) molecular mimicry compromises protein functioning, lowering the within-host replication rate and leading to fewer new infections (the “mimicry trade-off hypothesis”). Our analysis shows that although both hypotheses may select against molecular mimicry in parasites, unique hallmarks of protein expression identify whether selection is due to the costly autoimmunity hypothesis or the mimicry trade-off hypothesis. We show that understanding the relevant selective forces is necessary to predict how different medical interventions will affect the proportion of hosts that experience the different infection types, and that if parasite evolution is ignored, interventions aimed at reducing infection-induced autoimmunity may ultimately fail. KEY WORDS: Antigenicity, autoimmunity, clonal deletion, immunopathology, pathogen, within-host. Parasites that are molecular mimics express proteins which re- immunological processes that protect the host against autoimmu- semble host proteins (also termed “self proteins”). There are nu- nity leave mimics protected from an immune response as well. merous examples of parasites that mimic human cells (Oldstone The process that prevents autoimmunity is clonal deletion. Dur- 1998) including Campylobacter jejuni, which causes food poi- ing clonal deletion lymphocytes are tested against self-proteins, soning and mimics a human ganglioside (Ang et al. 2004), human and highly self-reactive lymphocytes are destroyed as a means cytomegalovirus, which infects 10% of infants under 6 months of preventing an immune response to self (Goodnow et al. 2005; age (Landolfo et al. 2004) and mimics endothelial receptors Hogquist et al. 2005). Therefore, parasites that are molecular (Michelson 2004) and the protozoan Trypanosoma cruzi,which mimics presumably obtain a fitness advantage over their non- mimics myocardial cells (Petkova et al. 2000). Parasites that are mimicking counterparts through the evasion of an immune re- mimics are likely to evade the immune system because the host sponse (Damain 1964; Young et al. 2002; Elde and Malik 2009; is averse to harming itself. Autoimmunity refers to host tissue Wolfl¨ and Rutebemberwa 2008). damage caused by the host immune system. To the extent that Although these benefits of mimicry seem clear (Begon et al. mimicking parasites are structurally similar to self-proteins, the 1990, p. 963; Elde and Malik 2009), the costs of such mimicry are C 2013 The Author(s). Evolution C 2013 The Society for the Study of Evolution. 2889 Evolution 67-10: 2889–2904 A. HURFORD AND T. DAY less obvious. Such costs presumably exist, however, because there Model are a great number of parasite species that have not evolved to be To derive our mathematical model, we begin by describing the molecular mimics. Although hypotheses for why mimicry does between-host epidemiological spread of a parasite. Different par- not evolve have been presented, understanding why such mimicry asite genotypes are assumed to correspond to different parasite is not a ubiquitous feature of most organisms is a long-standing, phenotypes and are represented in the epidemiological model as and unanswered, problem in evolutionary ecology (Ruxton different parameter values. We derive an expression for parasite et al. 2004). Our goal is to use a mathematical model to explore fitness as it depends on the between-host spread (section “Epi- two hypotheses for how molecular mimicry is selected against in demiological dynamics”). Next, we formalize each of our two parasites. hypothesis in terms of between-host spread and parameters corre- There exists substantial evidence that parasites which are sponding to the within-host reproduction of the parasite (section molecular mimics are more likely to induce an autoimmune re- “Formalizing the costly autoimmunity and the mimicry trade-off sponse in their host (Kirchoff 1993; Appelmelk et al. 1996; Kohm hypotheses”). We then link the within-host and between-host pa- et al. 2003; Ang et al. 2004; Rahbar et al. 2006; Dobbs et al. rameters that have been defined (section “Relating the parasite 2008). Our first hypothesis is therefore that hosts which develop phenotype to the types of infections”), and finally, we derive an an autoimmune response are less productive for the parasite be- expression that quantifies mimicry (section “Defining mimicry”). cause they produce fewer new infections. Thus, the risk of in- Definitions of the notation used are provided in Table 1. ducing autoimmunity selects against mimicry in parasites (the “costly autoimmunity hypothesis”; Damain 1964). The autoim- mune response occurs because, although clonal deletion in hosts EPIDEMIOLOGICAL DYNAMICS removes a large number of self-reactive lymphocytes, some self- The epidemiological model describes the rate of change in the reactive lymphocytes nevertheless remain in circulation (Wing number of hosts that are susceptible to an infection, S,orthatare 2006). These lymphocytes might then be activated by parasites experiencing one of three different kinds of infectious states (de- that are molecular mimics and cross-react with self tissues, caus- noted with I , A,andU), or either of two different postinfection ing autoimmunity (von Herrath and Oldstone 1995, 1996; Ohashi states (denoted Z and R; Fig. 1). For the three different types of 1996; Hausmann and Wucherpfennig 1997). infectious states, the type of disease that is induced depends on The second hypothesis is that the shape of the parasite’s the nature of the interaction between the parasite and the host’s proteins is constrained, such that becoming a mimic necessarily immune system. The first type of infection that may occur is an entails a compromise in other aspects of parasite function (the acute infection. An acute infection occurs if lymphocytes effec- “mimicry trade-off hypothesis”). There is evidence for such a tively react with the parasite, but not with normal host cells (i.e., trade-off in influenza A, for example, because the viral nucleo- self). We use I to denote the number of these infections. The protein that is targeted by cytotoxic T cells also plays a critical role second type of infection is one that leads to autoimmune disease. in viral replication (Docherty et al. 2006; Ye et al. 2006). Conse- This can occur if lymphocytes react with both the parasite and with quently, there might be little scope for modifying this nucleopro- self (i.e., cross-react). We use A to denote the number of hosts tein to evade the immune response without also compromising its that experience an infection that is cross-reactive and may lead function in viral replication and transmission. to autoimmunity. The third type of infection is an uncontrolled To determine the conditions under which each of the above infection. An uncontrolled infection can occur if the immune re- hypotheses might provide a plausible explanation for the lack sponse is ineffective against the parasite, and we use U to denote of mimicry in parasites, we first construct a general mathemat- the number of such infections in the population. ical model describing the epidemiological dynamics of parasite Hosts that experience an acute infection or an infection that spread (section “Model”). We then determine the evolutionarily may lead to autoimmune disease may clear the parasite, thus be- stable level of parasite mimicry under each hypothesis, and these coming noninfectious. We use R to denote the number of hosts results are used to determine if there are characteristics of protein that are postacute infection and no longer infectious or susceptible, expression or predictions specific to each hypothesis that allow us and we use Z to denote the number of hosts that have autoim- to distinguish between them (sections “General properties of the munity, but are no longer infectious. It is assumed that hosts that parasite ESS” and “Selection for mimicry”). Finally, we consider develop autoimmunity cannot be reinfected, because the contin- the potential consequences of certain public health interventions ued immune response to self prevents the establishment of the on the evolution of parasite mimicry and on the occurrence of parasite. We use a general model formulation where, following the different