Host‐parasite dynamics in Chagas disease from systemic to hyper‐local scales

Trypanosoma cruzi is a remarkably versatile parasite. It can parasitize almost any nucleated cell type and naturally infects hundreds of mammal species across much of the Americas. In humans, it is the cause of Chagas disease, a set of mainly chronic conditions predominantly affecting the heart and gastrointestinal tract, which can progress to become life threatening. Yet around two thirds of infected people are long‐term asymptomatic carriers. Clinical outcomes depend on many factors, but the central determinant is the nature of the host‐parasite interactions that play out over the years of chronic infection in diverse tissue environments. In this review, we aim to integrate recent developments in the understanding of the spatial and temporal dynamics of T. cruzi infections with established and emerging concepts in host immune responses in the corresponding phases and tissues.

forms in mammals, asynchronous replication and trypomastigogenesis, asymmetric divisions, reversible transitions and formation of apparently quiescent or dormant amastigotes. [17][18][19][20][21][22] The morphological, antigenic and spatial variability, combined with active evasion strategies, presents a formidable challenge to the mammalian immune system. Nevertheless, most infections resolve to a stable chronic equilibrium of parasite replication and suppression via a combination of sustained antibody and type 1 cellular responses. The majority of people (>95%) survive acute infection and progress to a chronic, asymptomatic phase. Chagas cardiomyopathy is then estimated to develop at a rate of ~2% per year. 23 Disorders of the gastrointestinal (GI) tract develop in a smaller proportion of cases, sometimes in combination with cardiac disease. 24 Why Chagas pathology only affects a limited subset of tissues, in only a specific subset of infected people, is one of the longest-standing and most important unanswered questions in the field.
In this review, our aim is to integrate recent developments in the understanding of the spatial and temporal dynamics of T. cruzi infections with established and emerging concepts in host immune responses in the corresponding phases and tissues. The result is a view that parasite persistence occurs in a small number of privileged tissues alongside highly competent, T. cruzi-specific systemic responses, suggesting a substantial degree of compartmentalization, even within tissues. The low-level, yet perpetual chronic inflammation has the potential to become pathological, dependent on largely undefined host, parasite and environmental factors. Thus, progress in the development of anti-parasitic drugs, adjunct treatments, immunotherapies and vaccines is likely to require a much better understanding of the molecular and cellular determinants of T. cruzi persistence at the tissue-specific and even hyper-local, intra-tissue scale.

| ADVAN CE S IN S TUD IE S OF TISSUE-S PECIFIC INFEC TION DYNAMIC S
While T. cruzi must spend some time in extracellular environments of the blood and interstitial fluid to sustain infections and ensure transmission, it is predominantly an intracellular parasite of solid organs. Consequently, most of what is known of the cells and tissues targeted by T. cruzi comes from experimental animal studies. It is difficult to obtain robust data on tissue distribution in human patients, although post-mortem, transplant and biopsy results tend to be consistent with animal models. The mouse is the species of choice, but other rodents, rabbits, dogs and nonhuman primates have also demonstrated utility. 25 Tissue-specific parasite loads can be measured by a range of direct and indirect methods (reviewed in 26).
Developments in real-time bioluminescence imaging methods have underpinned much recent progress in understanding T. cruzi infection dynamics. 4,5,[27][28][29][30][31] These systems are based on transgenic parasites expressing luciferases, enabling analysis of light signals emitted by parasites in discrete anatomical locations. Major advantages include greatly reduced tissue sampling bias and the ability to monitor individual mice over time. Bioluminescence lacks the resolution necessary to visualize parasites at individual cell scale, but this can be achieved using parasites expressing fluorescent reporters, 21,32,33 an approach that becomes particularly powerful when luciferase-fluorescence fusion proteins are employed ( Figure 1). 22,34 The possibility to integrate these imaging methods with analyses of concomitant immune responses 35 holds considerable promise for advancing our understanding of T. cruzi-host interactions.

| S TAG E 1: T. CRUZI S YS TE M I C COLONIZ ATION AND INNATE RE S P ONS E S
In vectorial transmission scenarios, infection results from contamination of the triatomine bite wound or of mucosal membranes with trypomastigotes present in the bug's faeces. Transmission may also occur orally, via contaminated food or drink, in utero, and by blood transfusion and organ transplant. Several mechanisms of host cell invasion have been described and reviewed elsewhere. 36 In humans, oedema with intense mononuclear infiltrate at the entry site in the skin (chagoma) or eye (Romaña's sign) indicates an initially very localized infection. 37 However, the true extent of trypomastigote dissemination is not clear and surprisingly little is known at the cellular level about the actual sites of primary invasion and the first cycle of intracellular parasite replication, which lasts approximately 1 week.
Experimental animal studies indicate that the route of inoculation is a key factor. Intra-peritoneal injection results in similar parasite numbers in diverse tissues after 6 days. 38 Conversely, oral transmission results in highly localized infections in the stomach or nasomaxillary tissues 31,38,39 with initial infection of the local mucosal epithelium. 40 Similarly, after conjunctival inoculation, parasites first invaded and replicated in the mucosal epithelium of the nasolacrimal ducts and nasal cavities. 41 At the end of the first intracellular cycle, trypomastigotes are released and the infection disseminates widely. T. cruzi is pan-tropic in the acute phase of infection (reviewed in 26). However, the relative intensity of infection in different cell or tissue types again varies depending on the inoculation route and inoculum size, as well as intrinsic factors such as replication rate and capacity for dissemination.
Sites reported to harbour the highest acute infection intensities include skeletal, smooth and cardiac muscle and adipose tissues. Some studies have described T. cruzi strains with an increased capacity to parasitize mononuclear phagocytes 13,42,43 or to cross the bloodbrain barrier. 12,44,45

| Sensors
The host response to T. cruzi primary infection is considered to be markedly delayed by comparison with model intra-cytosolic pathogens. 46,47 The main features of the immediate response are the induction of type I interferon signalling, and recruitment of neutrophils, macrophages and Natural Killer (NK) cells. 48 Ca 2+ mobilization, associated with invasion of myeloid cells, can activate the transcription factor NFATc1, leading to interferon gamma (IFNγ) production by NK cells and dendritic cell (DC) maturation. 49 T. cruzi also produces multiple B-cell mitogens that directly trigger a robust T-independent B-cell activation. [50][51][52] Few canonical pathogen-associated molecular patterns (PAMPs) are conserved in T. cruzi. The best characterized innate pattern recognition receptors (PRRs) for T. cruzi PAMPS are Toll-like receptor (TLR) 2 and 9. These recognize, respectively, the glycophosphatidylinositol (GPI) anchor of parasite surface proteins and parasite DNA, specifically unmethylated CpG motifs. 53,54 TLR2 + 9 double-knockout mutant mice suffer higher parasitaemias and significantly increased mortality rates (50% by day 50) compared to wild-type controls. 54 Mice lacking both MyD88 and TRIF, thus rendered incapable of any TLR-mediated responses, have uncontrolled parasitaemia and 100% mortality by 18 days of infection. 55 This may be explained by the additional involvement of TLR4 and TLR7, recognizing parasite glycoinositolphospholipids and RNA, respectively. [56][57][58] Many T. cruzi surface proteins are extensively glycosylated, 2 and several host galectins (a widely expressed family of carbohydrate-binding proteins) are able to bind them. 59 Interactions involving several different galectins may actually help T. cruzi bind to and enter host tissues, 60,61 but this does not appear to directly trigger any anti-parasitic effector activity. As an occupant of the host cell cytoplasm, it is likely that T. cruzi triggers cytosolic sensors. The best known candidate systems centre on NOD-like receptors (NLR).
Mice lacking the NOD1 receptor suffer 100% mortality to acute T. cruzi infection, although the mechanism explaining this remains obscure. 62 Studies also support a parasite-suppressive role for the

| Signal mediators and amplifiers
A plethora of cross-talking signalling pathways are activated downstream of the PAMP/DAMP sensors described above. Signalling converges on a set of transcription factors (NF-κβ, AP-1, IRF3), which results in production of inflammatory cytokines. [73][74][75][76] Critical amongst these are the IL-12 family, IFNγ and TNF-α, the canonical drivers of type 1 immune responses required to tackle intracellular infections. IL-12 is essential for the early activation of recruited natural killer cells and their production of IFN-γ; both these cytokines are indispensable for control of parasite loads and avoidance of acute mortality. 77 TNF-α is also essential for survival of the acute stage. 78 IFNy and TNF-α activate parasite destructive effector mechanisms via autocrine and paracrine signalling. Beyond the canonical IL-12-IFNγ axis, signalling through the IL-1 receptor is essential for early (10 days p.i.) induction of the myocarditis needed to control heart parasitism. 65 The local tissue response is amplified via chemokine-driven recruitment of inflammatory monocytes, macrophages, neutrophils and, eventually, antigen (Ag)-specific CD4 + helper T and CD8 + cytotoxic T lymphocytes (Th and CTL) to the site of infection. 79 Microvascular plasma leakage into parasitized tissues is promoted further by activation of mast cells and the kallikrein-kinin system (KKS), via a mechanism involving cruzipain, a parasite-derived cysteine protease. 80 The resulting tissue oedema and upregulation of associated receptors on cardiomyocytes may increase specific susceptibility to heart invasion as the infection progresses. 81

| Innate effectors and their evasion
The infection, cell necrosis and associated inflammatory signalling result in the activation of a range of innate effector mechanisms.
There is some evidence from analysis of Beclin-1-deficient mice that host cell autophagy can provide some marginal early restraint on parasite replication. 82 Epimastigotes are complement-sensitive but trypomastigotes have effective molecular mechanisms providing resistance to complement-mediated lysis. 83 Infiltrating NK cells, in addition to being major producers of IFNγ, may have direct parasiticidal effects involving the release of cytotoxic granules. 84 An unusual population of innate-like CD8 + T cells with activation characteristics of (e.g. production of granzyme A and IFNγ) expands in the thymus of T. cruzi -infected mice. These cells appear to be driven by antigen-independent mechanisms, and adoptive transfer experiments of thymocytes from infected mice suggest they might provide protection from otherwise lethal challenge 85 ; however, the underlying mechanisms conferring this protection remain to be elucidated.
Reactive oxygen and nitrogen species (ROS, RNS) are principal effectors for T. cruzi control. These are generated by IFNγ/ TNF-α-activated macrophages and via diverse other mechanisms in non-phagocytes and extracellular compartments. 86 They are a significant cause of collateral damage in infected tissues, but high levels are necessary because T. cruzi has an extensive and highly effective antioxidant defence system. 86,87 ROS can even promote T. cruzi replication, by a mechanism proposed to depend on the increased availability of intracellular Fe 2+ ions that the parasite can utilize. 88 Nitric oxide (NO) is directly parasiticidal in vitro, 89 and inducible NO synthase (iNOS) is essential for in vivo parasite control in some models, 62,77,90 although not in others. 63,91 Despite the plethora of innate responses, the overall effectiveness of T. cruzi's evasion mechanisms renders it debatable whether they actually have any meaningful impact on most infections, apart from the induction and conditioning of the adaptive response (see below). Indeed, T. cruzi infections are 100% lethal in mice that are genetically incapable of mounting adaptive responses (SCID, RAG, nude) [92][93][94] and bioluminescence imaging studies show that parasite growth in such mice is close to exponential. 4

| S TAG E 2: ADAP TIVE RE S P ON S E S TAK E CONTROL
The infection usually peaks, in terms of total parasite numbers and the extent of tissue dissemination, at a point between 2 and 3 weeks post-infection. Over the following weeks, parasite loads are reduced by several orders of magnitude by a highly effective adaptive immune response. Although they are ultimately thought to be non-sterilizing in virtually all cases, 95 it is worth reviewing the key features at the systemic level before we consider the tissue-specific host-parasite dynamics at play in the chronic phase. We also refer readers to more in-depth reviews of adaptive immunity in Chagas disease. 46,96,97

| T-and B-cell activation
T. cruzi cycles between the cytosolic and extracellular compartments and, accordingly, its control is critically dependent on the generation and deployment of Ag-specific CTL to infected tissues and antibody production by B cells. This is evidenced by relevant gene disruption and antibody-mediated depletion experiments in mice. [98][99][100][101][102] Mature DCs in the spleen and lymph nodes draining infected tissues, conditioned by the inflammatory environment, activate parasite Ag-specific CD8 + and CD4 + T cells from the naïve pools. 47,103 Activated T cells then migrate to sites of infection to exert effector mechanisms or, in some cases, begin differentiation to memory subsets. 102,104 A number of factors may impinge on the quality and magnitude of the T-cell response, including parasite-driven immature thymocyte apoptosis 105 and direct and indirect modulation of DC-T cell interactions. [106][107][108][109] In terms of antigen specificity, the murine T-cell repertoire is focussed mainly on a small number of immunodominant epitopes from highly expressed surface proteins, [110][111][112] but in humans there is evidence of a broader hierarchy 97,113 and immunodominance appears not to directly contribute to chronicity.
The role of CD4 + T cells is not well characterized, but the association between HIV infection and life-threatening acute T. cruzi relapse in humans 114 indicates they are critical for parasite control.
Accordingly, mice that are specifically incapable of mounting CD4 + T-cell responses experience 100% acute lethality of T. cruzi infection. 115 This has been linked to loss of support for parasite-specific CD8 + T-cell cytotoxicity against intravenous delivered splenocytes loaded with parasite antigens from the ASP-2 gene, 99  Although activation of Tfh responses to T. cruzi infection has not been explored in detail, it is reasonable to hypothesize that they are required for the production of parasite-specific antibodies and ultimate control of the infection. In line with this, IL-6, which supports Tfh differentiation, 125,126 is required for the control of parasitaemia and splenocyte recall response to parasite antigens, 127 but not for T-cell independent polyclonal activation of B-cell responses. 51 The initial B-cell response in the spleen is estimated to be at least 10-fold higher as compared to LNs draining infected tissues, 52 and a robust T. cruzi-specific antibody response is still generated there alongside the aforementioned polyclonal B-cell activation and non-specific hyper-gammaglobulinaemia. The parasite-specific antibody response is presumably driven by B-cell activation involving T-cell collaboration because it is accompanied by a robust germinal centre B-cell response and production of parasite-specific classswitched antibodies. 52 The specific and non-specific splenic B-cell responses appear to be either differentially regulated or carried out by different B-cell compartments because only the latter depend on the cytokine B-cell activating factor (BAFF). 128 Activation of auto-reactive T-and B-cell clones, the latter leading to the production of autoantibodies, is a well-described phenomenon during T. cruzi infection. 129 Polyclonal B-cell activation, host molecular mimicry by parasite proteins and bystander activation caused by tissue damage have been postulated as underlying mechanisms. 129 There is broad evidence and consensus that parasite persistence is required to sustain these autoimmune responses. [130][131][132] Nevertheless, the significance of autoantibodies and auto-reactive T cells for Chagas disease pathogenesis and the mechanisms involved in their production during T. cruzi infection remain major unresolved questions.
It has been suggested that the non-specific polyclonal B-cell activation contributes to delay the generation of T. cruzi -specific B-cell responses, thus contributing to parasite escape and establishment of chronic infections. 133,134 Polyclonal B-cell activation is associated with rapid, innate-like production of IL-17 and IL-10 135,136 but the wider relevance is unclear as both protective 136 The kinetics of the adaptive response depend to some extent on the early parasite load, 99 but in most cases, it coincides with the second or third intracellular cycle of parasite replication and is considered relatively delayed. 47 Nevertheless, substantial immune memory can be generated quite rapidly: mice whose infections were cured by benznidazole anti-parasitic chemotherapy starting 4 or 14 days after infection were then able to restrict acute parasite loads in challenge infections by 85% and >99%, respectively. 35 Notably though, very few of these animals achieved sterile cure and they progressed to chronic phase infections that were comparable to primary infections in naïve mice. This raises important questions about what is mediating memory responses to secondary infections, for example whether they are T cell-dependent or independent.

| Adaptive effector mechanisms
Lymphocytes contribute to control of T. cruzi by production of type 1 cytokines that amplify the prior, innate ROS and RNS production in infected tissues. Their signature, direct effector mechanisms are also crucial. These include the principal CTL effector pathways, namely perforin-mediated delivery of granzymes and FasL-induced apoptosis. In particular, granzymes cause fatal oxidative damage to T. cruzi, which can be mitigated by ROS scavenging drugs or overexpression of parasite antioxidant genes. 140 This may potentially be accelerated in humans by granulysin-mediated delivery of granzymes directly into intracellular parasites themselves. 140 Mice do not have a granulysin gene but in most cases still achieve good control of parasite levels, so immune pressure may be more focussed on extracellular amastigotes after host cell apoptosis and on clearance of trypomastigotes. These canonical pathways are essential in some experimental models [140][141][142] but dispensible in others. 98,143 The difference is likely explained by other pathways providing sufficient compensatory effector capacity in lower parasite load or virulence scenarios.
T. cruzi-specific lytic and neutralizing antibodies are normally detected in humans and animal models. [144][145][146][147] These are mostly produced in the spleen; antibody secretion by bone marrow cells obtained from acutely infected mice is below detection level. 52

| Deactivating/Regulatory mechanisms
The strong and sustained systemic inflammation, host cell lysis and tissue parasite killing in this control phase cause potentially dangerous levels of collateral tissue damage. Infections may become overtly symptomatic and in some cases fatal, particularly if the CNS is involved. 153 Tissue-protective immune regulatory pathways are therefore initiated to dampen the inflammatory response, to the benefit of the remaining parasites, which form the founding populations of the chronic infection reservoirs (Figure 3).
The factor with the strongest evidence for an important regulatory role is probably the cytokine IL-10. Early studies of IL-10 deficiency using high virulence Tulahuen strain parasites reported better control of acute T. cruzi parasitaemia at the expense of rapidly fatal (~2 weeks p.i.) pathogenic inflammation, for example TNF-α-mediated toxic shock. 154,155 More recent studies point to greater complexity. Rôffe et al (2012) reported IL-10 was essential to protect against later mortality (3-6 weeks p.i.) associated with poor control of Colombiana strain tissue parasite loads and increased myocarditis intensity. In still lower virulence scenarios, the absence of IL-10 has been associated with reduced CTL effector function but without any increased mortality. 156 Both CD8 + and CD4 + T cells are IL-10 sources, and a high proportion simultaneously produce IFNγ, 157 likely supported by IL-27 production 158 and potentially in direct response to parasite shed trans-sialidase. 123 B cells also produce IL-10, 135 and overall IL-10 production is lower in B1 B-cell-deficient mice early during infection. 159 CD11b + B1 B cells from asymptomatic, infected individuals show increased capacity to produce IL-10 compared to those with cardiac disease symptoms. 138 In addition, recent data show that when compared to non-infected donors, chronically

F I G U R E 2
Overview of host-parasite interaction dynamics in Trypanosoma cruzi infections. Chart illustrates some alternative course of infection scenarios for acute and chronic phase total parasite burdens, feeding into potential clinical outcomes, which range from longterm non-progression to severe Chagas disease affecting the heart and/or gastrointestinal tract. The chronic equilibrium scenarios are the product of many temporally overlapping host-parasite interactions within and between the various organs targeted for infection by T. cruzi. Three possible, non-mutually exclusive modes of persistence at the tissue or tissue sub-domain level are illustrated above the chart, continual persistence, dormancy/reactivation and episodic re-invasion.
increased proportion of immature transitional CD24 high CD38 high and naïve B cells able to produce IL-10 upon in vitro re-stimulation. 160 This suggests B cell-intrinsic IL-10 signalling might be important to regulate the intense adaptive immune response, as is the case for other parasitic infections, 161 but direct mechanistic evidence is required to support this hypothesis.
Transforming growth factor beta (TGF-β), another potent regulatory and tissue-protective cytokine, can be activated from its latent form by a T. cruzi protease (cruzipain) in vitro. 162 TGF-β signalling to T cells reduces the risk of late acute mortality, and this appears to involve inhibition of cell proliferation rather than suppression of inflammatory cytokine production. 106,163 Other factors potentially contributing to early inhibition of adaptive immune effector responses include suppressor of cytokine signalling (SOCS), 164 regulatory CD4 + T cells (Tregs) 106,165 and induction of various regulatory/suppressive myeloid cell phenotypes, such as expression of iNOS-limiting arginase. 109,166,167 The overall result of these deactivating pathways is the avoidance of potentially life-threatening levels of inflammation and tissue damage at the expense of incomplete clearance of the infection (Figures 2 and 3). The situation at the tissue-specific level, however, is likely to be more complex because only a subset of tissues serve as privileged sites for T. cruzi persistence in the chronic phase. 26

| S TAG E 3: THE 1% AND THE CHRONI C HOS T-PAR A S ITE EQU ILIB RI UM
In the chronic phase, blood parasitaemia is typically sub-patent and tissue parasite loads are between 0.1% and 1% of their levels in the acute phase. 4,29 Animal imaging studies 4  background. It appears that the GI tract, mainly the large intestine and stomach, is a universal site of continual parasite persistence in mice. 29,30,170 The well-studied parasite strain CL Brener is also commonly detected in the skin of BALB/c mice 38  parasite loads <1% of those in primary infection controls, and fully sterile protection is seen in around half of the animals. 35 The determinants of both categories of protection remain to be defined, but it is likely critically dependent on the T CM population. It is likewise an open question whether T CM -derived effectors contribute to suppression of tissue parasite numbers during chronic infections, particularly in organs subject to cycles of episodic re-invasion and clearance.
The mechanisms of immune evasion sustaining the host-parasite equilibrium during perpetual chronic infection are not necessarily the same as those that prevent sterile clearance in the acute to chronic transition, which resemble a conserved host tissue-protective, anti-inflammatory programme. They are also harder to study, both from the parasite perspective, owing to the scarcity of T. cruzi foci in tissues, and from the host perspective because of the need for conditional intervention techniques that can be applied after acute infections have been brought under control. The essentiality of CD8 + T cells for continued suppression of parasite numbers in the chronic phase is reasonably clear. Parasite loads rebound rapidly upon treatment with anti-CD8 antibodies, almost to the level seen with pan-adaptive immunosuppression using cyclophosphamide, 174,178 although depletion of CD8 + NK cells and DCs may contribute to the relapse, in addition to CTLs. Unlike in the acute phase, experimental anti-CD4 treatment has no effect on chronic parasite loads. 178 Nevertheless, severe reactivation of Chagas disease in HIV co-infected patients indicates CD4 + T cells are vital for control and the frequent presentation of meningoencephalitis points to a specific role in protection against invasion of the CNS. 114 There are various non-exclusive hypotheses for how a small subpopulation of parasites reliably evades sterilization in the face of the sustained adaptive immune pressure. However, in our view none currently has compelling evidence supporting a mechanistic explanation so this will remain an active area of investigation.

| Antigenic diversity
African trypanosomes famously evade host immunity using a sys- Very little is known about how variant copy expression may be controlled at the individual cell level, that is amongst amastigotes and amongst trypomastigotes. Investigating this is difficult, because gene control is mainly post-transcriptional and suitable variant-specific monoclonal antibodies are lacking. Available evidence suggests that within a class of surface proteins, expression in individual parasites is not strictly mono-allelic. For example, trypomastigotes can co-express at least two members of the mucin 183 and GP85 families. 182 The finding that a specific mucin-associated surface protein (MASP) peptide was only expressed in ~5% of parasites indicates that neither is expression totally promiscuous at the protein level. 184 Mechanisms controlling the expression of parasite surface proteins may therefore vary between gene families or sub-families. T. cruzi may also regulate its antigenic repertoire expression between infection phases and between different host cell types. This requires much deeper analysis because currently there are insufficient data to rule out clonal antigenic variation as a mechanism contributing to perpetual immune evasion.

| Parasite dormancy
Many pathogens use dormancy or metabolic quiescence as an immune evasion strategy. 185,186 At the population level, T. cruzi amastigotes can rapidly decrease their replication rate in response to changes in in vitro culture conditions, but this is a function of a longer G 1 phase rather than exit from the cell cycle. 187 Individual non-replicating amastigotes also occur spontaneously in vitro and are less susceptible to the anti-parasitic drug benznidazole. 21 The frequency of 4-day replication arrested in vitro amastigotes has been estimated to be approximately 0.1%-6%, depending on the parasite strain. 188 The in vivo relevance of these phenomena remains almost completely unknown and will be hard to establish definitively, not least because neither amastigote DNA/kDNA replication, nor differentiation to constitutively non-replicating trypomastigotes is synchronized. 22 Nevertheless, it is reasonable to suspect that slowly replicating or transiently arrested intracellular parasites could have a selective advantage under immunological pressure and play a role in sustaining chronic infections.

| Cytokine-mediated suppression of type 1 responses
T. cruzi may continue to benefit from the above-mentioned conserved negative-feedback mechanisms that damp down the acute inflammatory response. However, administration of blocking antibodies targeting IL-10 signalling had no discernible effect on chronic T. cruzi infections. 174 This is in stark contrast to the well-established role of IL-10 in promoting chronicity of infections with the related parasite Leishmania spp., 189 which predominantly infects professional antigenpresenting cells. Chemical inhibition of the TGF-beta type I receptor significantly alleviated cardiac pathology and function in chronically infected mice, but this was not associated with any change in heart parasite loads. 190 It should be noted that parasite loads in the chronic phase are often close to the limit of detection, which means that in these types of intervention experiment it is relatively clear when immunity is compromised, but difficult to conclusively demonstrate a significant enhancement of infection control.
Cytokine gene expression in heart tissue from human patients with severe chronic Chagas cardiomyopathy remains strongly polarized to a type 1 profile. 191 Type 2 cytokine (IL-4, IL-5, IL-13) expression is reported as undetectable, 191 and while it is a feature of some animal models, this is apparently not at the expense of IFNγ production. 170 Interestingly, helminth co-infection is associated with reduced control of T. cruzi in a subset of patients, potentially as a result of a modulation of the cytokine balance. 192 Overall, cytokine-mediated suppression of anti-parasitic type 1 inflammation likely influences the host-parasite equilibrium and long-term disease progression, but there is little evidence that it explains T. cruzi chronicity.

| Immunological exhaustion
T-cell exhaustion is a feature of many infectious and non-infectious diseases that involve chronic antigen stimulation and this has been a recent focus of research in the Chagas disease field. Analysis of PBMCs from Chagas patients revealed increased frequencies of CD4 + and CD8 + T cells expressing exhaustion markers, for example PD-1 + , CTLA4 + , or TIM-3 + . [193][194][195] Experimental studies suggest the development of exhaustion characteristics may be promoted by suboptimal B cell,  or IL-10 156 immune responses.
Nevertheless, in chronically infected children and mice, both effector and memory CD8 + T cells retain cytotoxic capacity and there is little to no evidence of functional exhaustion. 119,152,174,177,197 Infection chronicity may also promote dysregulation of the Tfh and B-cell compartments, for example, distinct phenotypes and frequencies of these have been noted between symptomatic and asymptomatic T. cruzi-infected individuals. 198,199 Whether these alterations reflect a process of B-cell exhaustion which negatively impacts parasite control remains to be further elucidated. In summary, while deterioration of lymphocyte functional capacity may potentially be associated with progression from asymptomatic to symptomatic disease states, via progressively loosened control of parasite loads, exhaustion does not seem to be a core reason for parasite persistence per se.

| Local and hyper-local immune privilege
The realization that long-term T. cruzi infections exhibit an unexpectedly high degree of spatio-temporal dynamism 4 indicates that host responses and evasion mechanisms, including those set out above, need to be studied more intensively at the tissue-specific level.
Motile trypomastigotes probably traffic between tissues in both blood and lymph, but there is also evidence that a significant amount of parasite trafficking between tissues may occur inside SLAMF1 + myeloid cells, akin to a Trojan horse strategy. 200 Consequently, parasites from privileged reservoir sites, such as the digestive tract, may seed other, less permissive sites such as the heart, resulting in episodic cycles of re-invasion and locally sterilizing host responses ( Figure 2). 26 Tissue-specific variability in permissiveness is consistent with divergent responses observed in different secondary lymphoid organs. 201 Moreover, when chronically infected mice are immunosuppressed, the infection relapses first in the GI tract and then disseminates to other organs. 29 Host microbiota may also play a role: its composition can be modulated by T. cruzi infection, 202 but it is not yet known whether this in turn influences anti-parasite immunity in barrier tissues.
To keep up with the parasite, effector cells must be continually deployed to infection foci in many organs. There appears to be no problem with T-cell homing and entry into infected tissues, 203 which is dependent on expression of integrins including VLA-4 and LFA-1, 174,204 and CXCR3 chemokine receptor signalling. 205 After extravasation though, the distinct microenvironment of each organ potentially drives phenotypic changes to infiltrating cells, and in some cases, the effect may be tolerogenic and incompatible with local sterilization. For example, skeletal muscle bulk CTLs recovered from early chronic phase mice produced less IFNγ and had greatly diminished cytotoxic activity compared to splenic CTLs. 206 Similar results have been reported for cardiac muscle compared to blood. 203 Intriguingly, splenic CTLs adoptively transferred from one chronically infected mouse to another retained a high IFNγ response phenotype if they migrated to spleen or lung tissue, but lost it if they migrated to skeletal muscle or liver. 206 More recently, however, direct ex vivo analysis of parasite-specific CTLs without antigen re-stimulation showed cells from chronically infected skeletal muscle tissue had equal or even greater effector capacity (production of IFNγ, TNF-α, granzyme B) than spleen-derived cells. 174 To our knowledge, detailed analysis of CTLs in smooth muscle has yet to be conducted.
There is thus likely to be further compartmentalization of response and evasion dynamics at the intra-organ level, perhaps even down to the hyper-local scale of individual infected cells' microenvironments. For example, the muscular, neuronal and mucosal layers of the GI tract, a key site of T. cruzi persistence, have distinct immunological microenvironments that respond differently to Salmonella infection. 207 Recent work has highlighted differences in the cellular composition of perivascular and parenchymal inflammatory infiltrates in T. cruzi-infected skeletal muscle. 171,178 Large, apparently immunologically invisible parasite nests even occur immediately adjacent to severely inflamed blood vessels, which led these authors to suggest leucocytes might fail to migrate through the parenchyma to infected cells because chemoattractant signalling is too weak in low parasite load settings. 171 Immune evasion may also operate at the level of physical interaction between T cells and parasite antigen-presenting cells, for example via manipulation of MHC class I or II expression 107,[208][209][210] or by parasitism of muscle cells, which are poor activators of NF-κB upon T. cruzi infection 76 and, in the case of skeletal muscle, do not normally express MHC class I. 211

PEER R E V I E W
The peer review history for this article is available at https://publo ns.com/publo n/10.1111/pim.12786.

DATA AVA I L A B I L I T Y S TAT E M E N T
Data sharing not applicable to this article as no data sets were generated or analysed during the current study.