Secreted effectors in Toxoplasma gondii and related species: determinants of host range and pathogenesis?

Recent years have witnessed the discovery of a number of secreted proteins in Toxoplasma gondii that play important roles in host-pathogen interactions and parasite virulence, particularly in the mouse model. However, the role that these proteins play in driving the unique features of T. gondii compared to some of its nearest apicomplexan relatives (Hammondia hammondi and Neospora caninum) is unknown. These unique features include distinct dissemination characteristics in vivo and a vast host range. In this review we comprehensively survey what is known about disease outcome, the host response and host range for T. gondii, H. hammondi, and N. caninum. We then review what is presently known about recently identified secreted virulence effectors in these three genetically related, but phenotypically distinct, species. Finally we exploit the existence of genome sequences for these three organisms and discuss what is known about the presence, and functionality, of key T. gondii effectors in these three species.

Introduction

In recent years, significant progress has been made in our understanding of secreted Toxoplasma gondii effectors that have significant impacts on virulence in the mouse. To date, these effectors have all been found to be secreted from two specialized secretory organelles that are unique to the phylum Apicomplexa to which T. gondii belongs: the rhoptries and the dense granules (Figure1a). This work was greatly facilitated by pioneering early studies using proteomics to identify some of the constituents of these organelles 1,2, identifying literally hundreds of putative rhoptry and dense granule proteins. Subsequent work has shown that while some rhoptry and dense granule proteins are constituents of the organelles themselves, others are secreted from these organelles, and can be found in multiple locations post-secretion including the host nucleus (e.g. protein phosphatase 2C 3 and rhoptry protein 16; ROP16; 4) and the parasitophorous vacuole/vacuolar membrane (e.g. ROP5 and ROP18; 5–9).

Figure 1

(a) Schematic of Toxoplasma (and related apicomplexans Hammondia hammondi and Neospora caninum) highlighting the three major secretory organelles involved in secreting host-targeting and/or host-modulating effectors. (b) Phylogram based on publicly deposited internal transcribed spacer (ITS1) sequences for Toxoplasma gondii, H. hammondi and N. caninum illustrating the degree of relatedness among these species. Isospora belli is used as an out-group.

What is not known, however, is if these and other secreted effectors play any role in determining the unique phenotypic characteristics of T. gondii compared to its closest Apicomplexan relatives (summarized in Table1). Specifically, the nearest sequenced relatives of T. gondii are Hammondia hammondi 10 and Neospora caninum 11. In contrast to these other organisms T. gondii has an incredibly vast host range, being capable of infecting nearly all mammals (including humans) and birds. Moreover T. gondii is not an obligate sexual parasite. Infected intermediate hosts harbour tissue cysts that are infective to both the definitive host (members of the family felidae) and other intermediate hosts 12,13. Finally T. gondii is highly virulent in mice, while H. hammondi and N. caninum are not (see Figure2 for an illustration of this phenotype in N. caninum).

Table 1

A summary of the known host range and virulence properties of Toxoplasma gondii and the closely related species Hammondia hammondi and Neospora caninum

Figure 2

Head-to-head comparison of luciferase-tagged Neospora caninum (strain NC-1) and Toxoplasma gondii (strain S1T) in mice. Balb/c mice (three per strain) were intraperitoneally infected with 1 × 106 tachyzoites, and in vivo bioluminescence imaging was used to quantify parasite burden over the first 96 h of infection. (a) Average total flux (photons/s) indicating parasite burden over the course of infection. For the first 20 h of infection both species proliferate at similar rates, but N. caninum is then cleared by 44 h p.i. while T. gondii S1T continues to proliferate. All mice survived the infection. (b) Representative images of N. caninum and T. gondii infections represented in (a), showing the rapid clearance of N. caninum from the peritoneal cavity in comparison with T. gondii.

While it is almost certain that we have just begun to understand the molecular mechanisms of virulence in T. gondii, genome sequences of H. hammondi and N. caninum 10,11 provide a unique opportunity to conduct a preliminary analysis of whether these effectors can explain the phenotypic differences between these species. Therefore in this review we (i) compare and contrast what is known about host responses to these three organisms; (ii) review the recent literature on the molecular mechanisms, and impact on pathogenesis, of select T. gondii virulence factors; and (iii) provide new data and review recent work on the conservation of these effectors across these three species. In our conclusion we then speculate on how such a comparative analysis can inform our understanding of the selective evolution of virulence and host range in T. gondii compared to H. hammondi and N. caninum.

Host Responses to Toxoplasma and Closely Related Species Differ Greatly

Despite a remarkable degree of genomic conservation between T. gondii and H. hammondi 10,12,14 and Neospora caninum 11 (Figure1b), there are a number of striking differences in host response and host range among these species. Here we will discuss differences in host range and pathogenicity among these species, as well as known differences or similarities in the host response in experimentally infected animals. Additional closely related species, such as Neospora hughesi, Besnotia besnotii, Hammondia triffittae and Hammondia heydorni will not be included in this review, as comparatively little is known about these parasites, and currently there are no available genome sequences for any of these species.

Variation in host range between T. gondii and closely related species

Toxoplasma gondii has a life cycle typical of tissue-dwelling coccidian parasites. Sexual reproduction occurs exclusively in the definitive host, members of the family felidae 15, while asexual reproduction occurs in a variety of intermediate hosts. Strikingly, T. gondii is capable of infecting virtually any warm-blooded animal, from birds to humans 16, and this broad intermediate host range is not only unique in comparison with closely related species such as H. hammondi and N. caninum, but also with respect to most eukaryotic parasites. An estimated one-third of the world's human population is currently infected with T. gondii, and while healthy individuals are able to control infection, those with compromised immune systems are at risk for developing life-threatening symptoms 17–20. In addition, some T. gondii strains have been found to cause severe, and even fatal, disease in immunocompetent adults 21–23. Acute infection during pregnancy often results in foetal loss, blindness, hearing loss, or severe cognitive disabilities 24,25. T. gondii is also the cause of foetal loss in a number of domestic animals including sheep, goats, and pigs 26–29. Virulence in humans has never been observed for H. hammondi, which is the most closely related extant relative to T. gondii and shares the same definitive host 12,30. While it is also assumed that H. hammondi is incapable of infecting humans, it is worth noting that given the antigenic similarity between these species 31 and that the most commonly used serological test for T. gondii infection is based on immunoreactivity to T. gondii surface antigen 1 (p30; SAG1), it is certain that if H. hammondi is capable of infecting humans such an infection would be misidentified only as a T. gondii infection. Development of a serum-based diagnostic test that could distinguish T. gondii from H. hammondi would allow for a direct test of the infectivity of H. hammondi in humans.

While the host range of T. gondii and N. caninum have been extensively studied, less is known about the host range in H. hammondi. Most H. hammondi isolates have been obtained from infected cats, but a wide variety of animals have been experimentally infected with this parasite, including cats, mice, rats, hamsters and monkeys 32,33. Importantly, however, birds appear to be refractory to H. hammondi infection 34. Another important distinction between T. gondii and H. hammondi is the inability of H. hammondi to be transmitted (at least experimentally) from one intermediate host to the other, and from one definitive host to another 12. In fact this is a key diagnostic feature to distinguish isolates of these parasites in the laboratory 12,35.

Neospora caninum does not share the cat as a definitive host, but rather utilizes canines for sexual reproduction 36. Dogs also appear to be intermediate hosts, as N. caninum infection in dogs causes a variety of neurological symptoms including encephalitis and ascending paralysis, often resulting in death 37,38. The known intermediate host range of N. caninum is more restricted than that of T. gondii and consists of dogs, cattle, water buffalo, sheep, goats and horses 38–43. With the exception of dogs and horses, all of these intermediate hosts are members of the Bovidae family. N. caninum causes abortion in cattle 44,45, much like T. gondii infection in sheep or goats. Unlike T. gondii, and similar to H. hammondi, there is no evidence that N. caninum infects humans 46; however, as with H. hammondi the antigenic similarity between T. gondii and N. caninum makes it difficult to rule out the possibility of N. caninum infections in humans 47.

These overlapping, yet distinct host ranges for T. gondii, H. hammondi, and N. caninum (see Table1), have been observed for quite some time, yet the genes (both parasite and host) responsible for these differences remain unknown. It should also be noted that while the host ranges of these parasites overlap, the pathologies and host response in overlapping intermediate hosts are not always the same. We review these data below.

Experimental models reveal both similarities and differences in host response to parasite infection in T. gondii, H. hammondi and N. caninum

As both T. gondii and N. caninum can cause spontaneous abortion in livestock, experimental infections of sheep, goats and cattle have been used to understand the pathology and modes of transmission of these parasites. Rodent models have also been developed to study infection, dissemination and transmission of T. gondii, H. hammondi, and N. caninum.

Cattle

The prevalence of T. gondii vs. N. caninum infection in cattle is variable by region or herd. In some areas, such as southern Vietnam and western Thailand, T. gondii is more prevalent than N. caninum 48,49; however, in southern China the prevalence of N. caninum infection is slightly higher than that of T. gondii 50. Natural infection of cattle by T. gondii does occur 51,52, but is not associated with abortion 53. Experimental infection suggests that there is a low rate of abortion in cattle upon T. gondii infection 54 and that this rate increases with T. gondii strains that are typically more virulent in mice 55. Surveys of aborted calves show a strong association with N. caninum infection, but no association with T. gondii infection 56, and herds with high abortion rates tend to have a high rate of N. caninum infection 57. Experimental infection of pregnant cattle shows that N. caninum infection during early gestation is likely to cause abortion 58; however, virulence among isolates does vary, and less virulent isolates do not appear to cause abortion 59. While some earlier studies suggested that infection late during pregnancy facilitates vertical transmission, but does not cause abortion 60, more recent studies show that infection with N. caninum late in gestation can cause abortion 61. There have also been conflicting studies suggesting that horizontal transfer of N. caninum infection and abortion does not occur in subsequent pregnancies after initial infection 62, while other studies suggest that chronic infection can lead to recurrent abortions 63. Clearly more experimental work is needed to clarify these conflicting data and to take into account both the genetics of the parasite and the host. Experimental infections of cattle reveal that N. caninum disseminates to a variety of tissues including the heart, lung, kidney, skeletal muscle and perhaps most importantly the brain 47. In fact, in one study they detected N. caninum in the brain and spinal cord, but in no other surveyed location, including the gastrointestinal tract, liver, kidney, heart, lung and skeletal muscle 64.

The genetic differences between T. gondii and N. caninum responsible for the differences in virulence in a bovine host have not yet been identified. It is interesting to note that N. caninum appears more virulent in cattle than T. gondii, whereas in most other shared intermediate hosts it appears that T. gondii is more virulent than N. caninum. This could be due to the fact that cattle are the natural, and most common, intermediate host for N. caninum, and it has evolved specialized methods for evading the bovine immune response. Further studies are required to determine why N. caninum is so successful in the bovine host, whereas T. gondii is not.

Sheep

Surveys of the prevalence of N. caninum and T. gondii in sheep herds show a significantly higher proportion of sheep infected with T. gondii than N. caninum 65–68. A combination of serological studies and experimental infections demonstrate the ability of both T. gondii and N. caninum to cause abortion in sheep, particularly when infected during early pregnancy 69–73. There is also experimental evidence that both T. gondii and N. caninum cause recurrent abortions in chronically infected ewes 74,75. Histological studies of aborted foetuses, weak lambs, congenitally infected healthy lambs and experimentally infected ewes show that T. gondii and N. caninum dissemination patterns are quite similar 69–71,76,77. Aborted T. gondii-infected foetuses have lesions primarily in the brain, with some specificity for regions such as the optic tract and rostral margin of the pons 77. Experimental infection in male sheep (rams) has shown that T. gondii does infect the male reproductive organs 78, and T. gondii infection can be sexually transmitted from infected rams to uninfected ewes 79,80, but this has not been examined with N. caninum infection.

Taken together this suggests that although seroprevalence of T. gondii is higher than that of N. caninum in domestic sheep, both species are successful parasites of sheep and clearly cause similar pathology. This is in stark contrast to experimental infection in cattle, (described above), where N. caninum is clearly much more virulent than T. gondii.

Goats

The seroprevalence of T. gondii infection in goats is generally much higher than that of N. caninum 81–83. Surveys of aborted goat foetuses suggest that T. gondii infection contributes to a number of these abortions 84,85, and experimental infection confirms that T. gondii is capable of causing abortions in goats 86,87. There are very few studies examining N. caninum infection in goats, compared to the number of studies done in sheep and cattle. Experimental infection of pygmy goats during pregnancy suggests that N. caninum infection in goats does cause abortion when infection occurs early during gestation, and that abortion in these goats does not recur with subsequent pregnancies 88. As with infection in sheep, T. gondii does infect the male reproductive organs 89, and infection can be sexually transmitted 90.

As with cattle and sheep, the genetics underlying the pathology differences between N. caninum and T. gondii in goats are not known. Additional studies in each of these intermediate hosts with genetically engineered parasites may help to uncover the genes responsible for both similarities, and differences, in host range and host response in T. gondii and N. caninum.

Rodents

Several rodent models have been developed for studying T. gondii, N. caninum, and H. hammondi infections. These are particularly relevant as rodents are a natural intermediate host for both T. gondii and H. hammondi, and likely play an important role in the evolutionary history of these parasites.

Multiple mouse strains have been utilized in developing models of T. gondii infection, including both outbred (CD-1) and inbred (Balb/c, Cba/j, C57BL6) mouse strains 91,92, and most recently, the house mouse 93. Pregnant mouse models have also been developed to better understand why T. gondii infection causes abortion 94–97. Mice infected with T. gondii have enlarged spleens and lymph nodes, caused in part by an increase in mononuclear phagocytes and CD8+ T cells, which produce interferon-γ (IFN-γ) 98. Production of innate immune effectors such as interleukin-12 (IL-12) and IFN-γ increases shortly after infection and is required for host survival and control of parasite growth 97,99,100. Immune-compromised mice (lacking both B cells and T cells) that have a larger population of natural killer cells, and are therefore able to produce higher levels of IFN-γ, have a lower parasite burden 94, once again providing evidence that IFN-γ production is essential for mouse survival following T. gondii infection. Neutralization of IFN-γ increases parasite burden in these mice, but decreases transmission of T. gondii infection to offspring 94. Symptoms of acute infection by T. gondii generally decrease after several weeks, when the adaptive immune system has had time to respond and produce antibodies and effector cells to combat T. gondii 96.

The population structure of T. gondii isolates has been studied extensively in an effort to better understand parasite virulence and host interaction. The majority of North American and European T. gondii isolates can be grouped into three main lineages that vary in virulence, as well as host responses 8,101. T. gondii strains exhibit a broad range of virulence in mice (Table1), with the most virulent type I strains being capable of killing a mouse after infection with a single parasite 8,102. Less virulent type II and type III strains of T. gondii have 50% lethal doses of >103 and 105 parasites, respectively 102. Some ‘atypical’ T. gondii strains from South America, which do not belong to any of the three major lineages, have also been shown to be highly virulent in mice 103. Comparisons of these strains and differences in host response following infection have facilitated the discovery of many parasite factors responsible for virulence and/or interaction with the host, including rhoptry proteins 5, 16 and 18 (ROPs) 4,5,7–9,104–106 and dense granule proteins 15, 24, 25 and MAF1 107–110.

In general, immunocompetent mice experimentally infected by intraperitoneal injection of N. caninum tachyzoites exhibit no signs of disease; however, immunosuppression of mice using methylprenicolone acetate (MPA) results in a range of neurological symptoms, from a slight head tilt to paralysis and death, depending on the dose of immunosuppressant 111. It also appears that subcutaneous injection of tachyzoites in inbred Balb/c mice results in a number of neurological symptoms without the use of MPA immunosuppression 112. IFN-γ-deficient mice, as well as mice lacking Toll-like receptor 4 (TLR-4) and a functional IL-12 receptor, are also susceptible to N. caninum infection by intraperitoneal injection of tachyzoites 113–116. In these mice, parasites can be found in the pancreas, liver, lung, intestine, heart and brain, while parasites are not detectable in these organs in immunocompetent mice 116. Infection of dendritic cells is likely important for the dissemination of N. caninum within the host, as adoptive transfer of N. caninum-infected dendritic cells increases parasite load as well as vertical transmission in pregnant mice 117. As head-to-head comparisons between T. gondii and N. caninum have not been conducted in mice, we tagged N. caninum strain NC-1 38 with luciferase and compared its proliferation in vivo to a highly avirulent strain of T. gondii, S1T. S1T is an F1 progeny clone derived from a cross between a T. gondii type II and type III strain and contains avirulent alleles of all five identified T. gondii virulence factors 118,119. Mice eventually control parasite proliferation and are able to survive infection with up to 1 × 106 tachyzoites of this parasite clone. As shown in Figure2, both species proliferate at a similar rate during the first 20 h post-infection, but then N. caninum is rapidly controlled while T. gondii S1T continues to proliferate. This suggests that the inability of N. caninum to be virulent in wild-type mice does not have to do with an inability to replicate within mouse cells in vivo, but rather an inability to disrupt host innate immune defences that rapidly control this parasite. It will be interesting in future studies to compare host responses to these two species during the early stages of infection.

Given this attenuated phenotype in mice, several genetically altered mouse models have been developed for N. caninum infection with tachyzoites (described above). For infections with other life stages, such as sporulated oocysts, both interferon-gamma knockout mice 120 and gerbil models of infection (Meriones unguiculatus) have been used effectively 121.

Similar to T. gondii, virulence differs among N. caninum isolates, which has been revealed by a number of comparisons 112,122,123. The NC-Liverpool strain, isolated from the brain tissue of a young dog euthanized after presenting with severe neurological symptoms, is a more pathogenic strain than the NC-SweB1 strain, isolated from a stillborn calf 122. NC-Nowra, isolated from a congenitally infected calf, is also less pathogenic than LC-Liverpool, but does cause some disease in a small portion of infected mice 124. NC-1 and NC-3 were both isolated from the tissues of congenitally infected dogs, and NC-1 is much more pathogenic than NC-3 112. No studies have been done to compare the pathogenicity of all N. caninum isolates; however, these studies suggest a wide range in ability to cause neurological disease when injected subcutaneously in Balb/c mice. Genetic crosses in the definitive canine host between these strains with distinct phenotypes could potentially lead to the identification of the virulence factors responsible. However it is not known whether they would be relevant to natural N. caninum infections as, in contrast to T. gondii, rodents do not appear to be a relevant host for N. caninum in the wild.

Relatively little work has been carried out in H. hammondi-infected mice, as there is currently no way to grow H. hammondi parasites in cell culture to perform the same types of experiments that have been done with T. gondii and N. caninum. Much of the work has been carried out in IFN-γ knockout mice. In parenteral infections in both wild-type and IFN-γ knockout mice H. hammondi is benign, resulting in chronically infected mice that show almost no symptoms of infection (based on behavioural responses to hyperinflammation or adverse neurological symptoms). However oral infections with large numbers of H. hammondi oocysts can cause severe disease 12 and even mortality 30 in Swiss-Webster mice. It is important to note that IFN-γ KO mice that are chronically infected with H. hammondi are infective to the definitive host, and rodents have been found to harbour H. hammondi in the wild 12. Given that IFN-γ is required for control of both T. gondii and N. caninum, it is intriguing that this cytokine is not required for control of H. hammondi. This could be due to as yet unidentified host innate immune responses, or it could be due to a hard-wired developmental programme. In H. hammondi that results in the spontaneous conversion from rapidly growing tachyzoites to slow-growing, encysted bradyzoites. Consistent with this latter explanation, H. hammondi-infected mice have orally infective tissue cysts in muscle and other non-CNS tissues in both wild-type and IFN-γ KO mice, and multiple groups have observed the spontaneous conversion of H. hammondi tachyzoites to infectious cysts during cultivation in vitro 12,125. Further analyses will be necessary to more fully characterize the differences in parasite development between H. hammondi and particularly T. gondii. Regardless of the root cause overall the existing work on H. hammondi indicates that it is unique compared to both T. gondii and N. caninum in terms of its behaviour in immune-deficient mice.

Secreted T. gondii Effectors Drive Strain-Specific Virulence Differences in Mice

Multiple effector proteins have been identified in T. gondii that play key roles in the interaction of this parasite with its host, particularly the mouse. Without exception, these effectors are secreted from either the rhoptries or the dense granules, and some have now been shown to interact with host cell proteins. A number of reviews have been written on the subject of these effectors and the host signalling pathways that they interface with 126–130. Here we present recent data on these effectors and their mechanism of action and discuss how they determine T. gondii strain-specific virulence phenotypes. We will then briefly speculate based on the level of conservation at the sequence and functional levels whether the absence of certain key effectors can explain some of the differences in pathogenesis between T. gondii and its near relatives that we have just outlined above.

Co-evolution of T. gondii effector proteins and host innate immune defence mechanisms

Based on gene knockout and forward genetic studies the most potent mouse virulence factors identified to date in T. gondii are rhoptry proteins 5 and 18 (ROP5/ROP18). These loci were identified using genetic crosses between canonical T. gondii strain types that differ in their virulence phenotypes in mice 5,7–9, and both belong to the rhoptry 2 kinase family protein superfamily. At least 30 family members can be found throughout the T. gondii genome, and many have undergone local tandem duplication and locus expansion events (including ROP5; 5,131). All members of the ROP2 superfamily encode putative proteins with an N-terminal domain encoding membrane-interacting amphipathic helices, and a C-terminal domain encoding either a functional kinase domain or a pseudokinase domain. The amphipathic helix domain is crucial for interaction of the rhoptry kinase with host membranes (particularly the host-derived parasitophorous vacuole (PV); 6). Importantly, knocking out the entire ROP5 locus in a highly virulent type I strain (RH) renders this parasite completely avirulent: while wild-type strains cause 100% mortality at a dose as low as 10 parasites, the ROP5 knockout parasites are completely avirulent at doses as high as 1 × 106 tachyzoites 7,5. ROP18 knockout parasites are also attenuated in mice compared to wild type, but are still capable of killing mice at doses as low as 1000 tachyzoites 132. These data further suggest that ROP5 is the more potent of the two loci in terms of impact on parasite virulence.

As the discovery of ROP5 and 18 as virulence effectors, multiple groups have demonstrated that they target the same host defence mechanism, namely immunity-related GTPases (IRGs; 104). Multiple lines of evidence indicate that ROP5 and ROP18 act in close collaboration, particularly with respect to host IRG proteins. Specifically, members of the ROP5 family (which have kinase-like folds but lack catalytic activity) bind to IRG proteins which, at least in one study, blocked IRG oligomerization and activation. Subsequently, ROP18-driven phosphorylation of ROP5-tethered IRG proteins renders them functionally inactive 104,133. Additionally it has been shown that the kinase activity of ROP18 depends on the presence of ROP5, further emphasizing the important interactions between these distinct gene products 134. Interestingly this all happens on the PV membrane, where in the absence of either ROP5 or ROP18 these proteins are loaded on the PV where they lead to its disruption and eventual parasite destruction. Recent work has also indicated that ROP5 and ROP18 may be a part of a much larger complex, which includes the secreted rhoptry kinase ROP17 135. ROP17 knockout parasites have reduced virulence in mice, and ROP17 and ROP18 have preferences for distinct phosphorylation sites on distinct IRG proteins 135. Additionally, ROP17 has a particularly strong preference for oligomerized IRG proteins, which again distinguishes it from ROP18 which has a preference for monomeric IRG proteins 134,135. These data provide strong support for the idea that the ROP2 superfamily of secreted kinases and pseudokinases have expanded due to strong selective pressure to interact with difference components of the mouse IRG repertoire. Genes encoding IRG proteins can be found in tandem arrays on multiple chromosomes of the mouse, and the overall IRG gene content varies significantly across even very closely related mouse species 93. IRGs are not conserved in humans, suggesting either that (i) the ROP5/ROP18 complex (and any other T. gondii effectors that target the IRG host resistance pathway) do not play a role in human infection; or (ii) ROP5 and ROP18 have additional targets that are conserved across multiple species (including mice and humans). Support for the former hypothesis can be found in the fact that ROP5 and ROP18-dependent resistance to IFN-γ stimulation of host cells only occurs in mouse cells, and not human cells 136. However other work suggests that the ROP2-related effectors may have other targets in the host cell, including ATF6β 132 and guanylate binding proteins 137,138.

In addition to ROP5 and ROP18, effectors are consistently being identified that directly modulate host cell signalling. These include ROP16, a non-ROP2-family rhoptry protein which is secreted into the host cell and traffics to the host cell nucleus due to the presence of a canonical mammalian nuclear localization signal. In the host cell cytoplasm (prior to nuclear translocation) it directly phosphorylates both STAT3 and STAT6 139–141, leading to dramatic changes in innate immune signalling at the transcriptional and translational level 4,139. The result of this activation of innate immune signalling is a slight decrease in parasite virulence 8, possibly through an attenuation of IL12 signalling that reduces the hyperinflammation that is known to be at least partially responsible for mouse mortality in T. gondii infections 142. Another is GRA15 108, which, depending on the allele, is capable of activating nuclear factor κ B (NFκB). Importantly ectopic expression in host cells of the type II allele of GRA15 can activate NFκB, although the mechanism for this activation is not yet clear. Other dense granule proteins modulate host cell signalling, including GRA25 110, MAF1 107, GRA6 143 and GRA24 109 (see Table2).

Table 2

A sampling of known secreted virulence factors in each of the three major European and North American Toxoplasma gondii clonotypes, their mechanism(s) of action and their degree of conservation in Hammondia hammondi and Neospora caninum

		Strain-specificity
Gene	Mechanism	I	II	III	H. hammondi ortholog?	N. caninum ortholog?
ROP18	IRG phosphorylationa	Active	Active	Inactive (low expression)	Yes	No (pseudogene)
ROP5	IRG bindingb	Active	Less active	Active	Yes	Yes
ROP16	STAT3/6 phosphorylation (I,III)c	Active	Inactive	Active	Yes	Yes
GRA15	NFκB activation (II)d	Inactive	Active	Inactive	Yes	No (pseudogene)
GRA25	CCL2/CXCL1 inductione	ND	Active	Less active (low protein expression)	Yes	Yes
MAF1	Host mitochondrial associationf	Active	Inactive (low expression)	Active	Yes	Yes
GRA24	P38α MAP kinase activationg	None			Yes	No (undetectable by BLAST)

156;

104;

140,141;

108;

110;

107;

109.

Each of the three major T. gondii clonal lineages has a distinct repertoire of virulence factor alleles

It is important to note that most of the effectors described above were identified based on strain-specific differences in gene sequence and/or phenotypic effect. Toxoplasma is a sexual species, and what this work clearly demonstrates is the significant impact that sexual recombination can have on parasite pathogenesis. Specifically, the major European and North American clonotypes used in these studies (clonotypes I, II and III) appear to be siblings 144 based on polymorphism analyses. All three clonotypes have regions of their genomes (and sometimes entire chromosomes) that are shared between two of the clonotypes. For example, type II and type III strains share a nearly identical copy of chromosome IX, while types I and II strains share a nearly identical copy of chromosome IV. All three clonotypes share a nearly identical copy of chromosome Ia 144–146. Consistent with this, each of these strain types harbour a different complement of the virulence effectors described above, and in most cases 2 of the 3 strains will have nearly identical alleles compared to a divergent 3rd clonotype (e.g. the type II alleles for ROP16 and GRA15 are divergent compared to types I and III, while type III has the most divergent allele for ROP18; Figure3). As shown in Table2, both type I and type II parasites harbour a ‘virulent’ allele of ROP18 8,9, while both types I and III strains harbour the ‘virulent’ allele of ROP5 7,5. This fact might be surprising given that type III strains are typically less virulent in mice than types I and II 147, but we now know that the sexual recombination event(s) that generated the progenitors of these lineages led to random (or possibly nonrandom via selection of the progeny for a particular virulence phenotype) re-assortment of the virulent and avirulent (or less virulent) versions of these alleles, and therefore virulence in each strain type is driven by what complement of alleles they harbour (see Table2 for a list of the virulence genes harboured by types I, II and III). As an example, type III strains have an allele of ROP18 that is essentially null due to a 107 bp deletion in the ROP18 promoter 10, but when either the type I or type II allele of ROP18 (which both have a functional promoter) is introduced into this strain it increases virulence in mice by up to 4 logs 8,9. These data demonstrate the enormous impact that even single genetic crosses can have on the fitness and virulence phenotype of haploid parasites like T. gondii.

Figure 3

Neighbour-joining trees of known Toxoplasma gondii secreted effectors in the three canonical T. gondii lineages (types, I, II and III), Hammondia hammondi and Neospora caninum. All trees are scaled identically, and the bar indicates sequence distance (0·1 substitutions/site). In N. caninum, orthologs of ROP18 11 and GRA15 are pseudogenes and therefore were omitted from the alignments.

Thus far the ability to use strain type (and therefore the presence or absence of certain key virulence effectors) to predict disease outcome in humans, or a significant link between strain type and host range, has been elusive. However it is clear from the work described above that lacking any number of virulence effectors (or having the ‘avirulent’ allele) does not necessarily render a parasite unsuccessful. For example, type III strains are highly prevalent in the USA and North America, and it has been postulated that they are the most prominent T. gondii isolate in the world 148. Perhaps this is due to reduced virulence and therefore increased transmission from host to host, but this has not been directly tested. Also, type I and II strains are also highly dominant in Europe and North America, suggesting that while they each harbour a different complement of alleles and key virulence loci, this has not had a significant impact on their ability to become so dominant in these regions. In addition, this suggests that there exist other virulence effectors in the T. gondii genome that are shared across most T. gondii lineages that could be termed ‘core’ effectors that enable T. gondii to evade host defences. We anticipate the recent advances in genetic manipulation of T. gondii and related species 149,150 may allow for these core effectors to be identified in a more unbiased way through the generation of strain collections where every putative secretory protein has been deleted.

Conservation of Known T. gondii Secreted Effectors in H. hammondi and N. caninum

The identification of key host-interacting effectors (like ROP5/18 and others) begs the question as to whether or not they play a role in determining the dramatic life cycle, host range and virulence differences between T. gondii and its near relatives. With the recent sequencing of both the H. hammondi 10 and N. caninum 11 genomes, one can begin to make comparisons across these three closely related species to begin to address this question. Our recent sequencing of the H. hammondi genome showed that T. gondii and H. hammondi are over 99% syntenic 10, while estimates of synteny between T. gondii and N. caninum are closer to 85% 10,11. When looking at the known T. gondii virulence effectors in H. hammondi and N. caninum, a different story emerges depending on the comparison. Starting with H. hammondi, this species harbours clear orthologs of multiple T. gondii effectors (ROP18, ROP5, ROP16, GRA15, GRA25, MAF1 and GRA24) while N. caninum appears to lack orthologs of ROP18 and GRA15 due to pseudogenization and appears to be missing GRA24 entirely (11 and Boyle, unpublished; Table2 and Figure3). Interestingly, it was recently reported that complementation of N. caninum strain NC-1 with the ROP18 allele from T. gondii type I dramatically increased the virulence of N. caninum in mice 151. This is an interesting, and entirely surprising result, given what is now known about the coordination between ROP18 and ROP5 in T. gondii virulence. It will be interesting to dissect this complementation phenotype further.

Clearly then, the presence of a given virulence effector in the H. hammondi genome is not sufficient to explain its highly avirulent phenotype in mice 12. We have also shown that H. hammondi ROP18, ROP5, ROP16 and GRA15 are functional effectors when expressed in relevant T. gondii genetic backgrounds 10,152. Specifically, H. hammondi has a functional ROP18 ortholog that is highly effective at increasing the virulence of essentially ROP18-null T. gondii strains (such as the type III strain CTG; 10). Through our analysis of the H. hammondi genome we found that there was a 107 bp deletion in the type III ROP18 promoter that is responsible for this difference. The ROP18 genes in type I and type II T. gondii strains, as well as the H. hammondi isolate HhGerCat041, harbour this 107 bp sequence, and we found it to be sufficient to ‘resurrect’ the type III T. gondii ROP18 promoter when inserted into the proper location just upstream of the transcriptional start site 10. As for ROP5, this locus is duplicated and expanded in the HhCatGer041 H. hammondi isolate, containing approximately 10 copies 10. Importantly, complementing a T. gondii ROP5 knockout with H. hammondi ROP5 paralogs (1-1 and 2-1; Figure3) dramatically increased virulence of this attenuated strain, clearly demonstrating that the H. hammondi ROP5 orthologs are functional as virulence genes. In contrast, the ROP5 locus has not significantly expanded in N. caninum, having only two copies. Moreover, the one ROP5 ortholog that has a complete sequence uninterrupted by a sequence gap is highly divergent from both H. hammondi and T. gondii ROP5 isoforms (Figure3). It is not known, however, if N. caninum ROP5 could complement virulence defects in T. gondii ROP5 knockout strains as for those from H. hammondi.

Hammondia hammondi ROP16 and GRA15 are also functional effectors. Specifically, expressing H. hammondi ROP16 in type II T. gondii significantly increased STAT6 phosphorylation and translocation to the nucleus 152, which is similar to the type I and III T. gondii ROP16 alleles 4,139–141,153. Importantly, the STAT6 induction was significantly higher than type II T. gondii expressing an additional copy of a type II ROP16 allele, providing strong evidence that the ‘active’ allele of ROP16 with respect to STAT6 activation is ancestral to the T. gondii/H. hammondi split. Interestingly, however, the H. hammondi ROP16 gene promoter had a 16 bp deletion compared to T. gondii ROP16, and this 16 bp was essential for ROP16 promoter function 152,154. It is therefore possible that in H. hammondi ROP16 is fully functional with respect to STAT6 activation, but is poorly expressed. Similar to ROP16, expressing H. hammondi GRA15 in a type I strain significantly increased NFκB translocation to the nucleus 152,154, and this is similar to what has been shown previously for T. gondii GRA15 alleles from type II strains. This also suggests that GRA15-driven NFκB activation is also an ancestral phenotype. In addition based on our analyses the GRA15 promoter appears to be fully functional 152,154.

One caveat of these H. hammondi heterologous expression studies is that we do not yet know if any of the effectors in question are actually expressed in this species. Determining this is hampered by the fact that long-term cultures of H. hammondi (in contrast to T. gondii and N. caninum) cannot be generated. However next generation ultra-deep sequencing of the H. hammondi transcriptome from short-term cultures (which can only be generated from cat-derived oocysts; 12,155) could provide new insights into the importance of gene content vs. gene deployment in the relatively avirulent phenotype of H. hammondi compared to T. gondii.

Conclusions and Future Prospects

Comparative analyses between T. gondii and its near relatives represent an exciting approach to identify ‘core’ virulence factors in T. gondii. The advent of rapid genome sequencing technologies and high-throughput genetic manipulation techniques in this organism will not only facilitate genome-by-genome comparisons but will also open the door to cross-species complementation experiments to determine the role of individual loci in T. gondii virulence and host range. While the present review focused exclusively on studies in whole animals, in vitro studies, such as those recently conducted by Beiting et al. 156, will also be crucial in identifying the host responses that may be differentially suppressed/targeted by distinct species like T. gondii and N. caninum RNAseq will provide a robust tool to neutralize the problems with growing large numbers of H. hammondi tachyzoites, as small-scale infections can be used to make RNAseq libraries (including host and parasite RNA) suitable for ultra-deep sequencing. As a clearer picture of the differences in both gene content (e.g. gene gain/loss) and gene deployment (e.g. expression) across the local phylogeny of T. gondii emerges, new hypotheses regarding the evolutionary events that led to the emergence of a species of parasite with the capacity for near global dominance will be generated and tested.