Microglia in neuropathology caused by protozoan parasites

Involvement of the central nervous system (CNS) is the most severe consequence of some parasitic infections. Protozoal infections comprise a group of diseases that together affect billions of people worldwide and, according to the World Health Organization, are responsible for more than 500000 deaths annually. They include African and American trypanosomiasis, leishmaniasis, malaria, toxoplasmosis, and amoebiasis. Mechanisms underlying invasion of the brain parenchyma by protozoa are not well understood and may depend on parasite nature: a vascular invasion route is most common. Immunosuppression favors parasite invasion into the CNS and therefore the host immune response plays a pivotal role in the development of a neuropathology in these infectious diseases. In the brain, microglia are the resident immune cells active in defense against pathogens that target the CNS. Beside their direct role in innate immunity, they also play a principal role in coordinating the trafficking and recruitment of other immune cells from the periphery to the CNS. Despite their evident involvement in the neuropathology of protozoan infections, little attention has given to microglia–parasite interactions. This review describes the most prominent features of microglial cells and protozoan parasites and summarizes the most recent information regarding the reaction of microglial cells to parasitic infections. We highlight the involvement of the periphery–brain axis and emphasize possible scenarios for microglia–parasite interactions.


I. INTRODUCTION
It has long been generally accepted that the central nervous system is an immunologically privileged area. The reasons for this are: (i) no lymphatic system exists in the brain; (ii) the blood-brain barrier strictly separates blood vessel compartments from the brain parenchyma; (iii) antigen-presenting cells were not thought to be present; (iv) brain grafts were less intensively rejected compared to other body areas. Over the last three decades, however, it has become clear that diseases such as multiple sclerosis and neuromyelitis optica are autoimmune disorders in which microglia, the main immune cells of the brain, play an important role (Zrzavy et al., 2017). Microglial cells in the mammalian brain comprise 5-20% of all glial cells (Biber, Owens & Boddeke, 2014). Taking a value of 10% to approximate the proportion of microglial cells among the entire glial population, and given that the glial population exceeds the neuronal population by a factor of 10, it appears feasible that the number of microglial cells is similar to that of neurons (Graeber, 2010). While the neuroscience community has tended to focus their interest on neurons rather than neuroglial and microglial cells, recent interest in microglia has increased considerably. Indeed, while even the existence of microglial cells was questioned only 30 years ago (Dolman, 1985), today research on microglial cells is one of the most active branches of modern neuroscience (Wolf, Boddeke & Kettenmann, 2017a;Borst, Schwabenland & Prinz, 2018). Microglial cells play prominent roles in brain development, homeostasis and aging, brain immunology, neurodegeneration, neuroinflammation and cellular plasticity. For in-depth reviews of these topics, see Miyamoto et al. (2013), Erny, Hrabě de Angelis & Prinz (2017), Mosser et al. (2017), Tay et al. (2017) and Bisht, Sharma & Tremblay (2018). In this review, we focus our attention on the role of microglia in parasitosis induced by protozoan parasites.

II. MORPHOLOGICAL AND FUNCTIONAL SIGNS OF MICROGLIAL ACTIVATION (1) Morphology of microglia
Glial cells are non-neuronal cells specific to the nervous system. They comprise macroglia, including Schwann cells in the peripheral nervous system as well as oligodendrocytes with their precursors, astrocytes, ependymal glial cells in the central nervous system, and microglia. Macroglial cells develop within the neuro-ectoderm. Microglia derive from primitive erythromyeloid yolk sac precursor cells that appear in the mouse brain as early as embryonic day 8 (Ginhoux et al., 2010). For a detailed description of the history of microglia research, see Kettenmann et al. (2011) and Tremblay & Sierra (2014).
In the uninjured adult brain, microglial cells typically have a small cell body endowed with numerous radial extensions, which are highly ramified. Microglial cells occupy cellular microdomains which do not overlap, and which accordingly are only sparsely interconnected by junctions, although they are equipped with gap junctional hemichannels (Orellana et al., 2009;Umebayashi et al., 2014). Such cells were formerly termed 'resting' or 'quiescent' microglia. In fact, these cells are steadily investigating their surroundings with their processes, and are far from dormant. Therefore, these cells now are called 'surveillant' microglia. Following neuronal or tissue damage, microglial cells change their shape, reducing the number of processes and altering their immunocytochemical phenotype, thus transferring into the 'activated' state (Tremblay & Sierra, 2014;Eyo, Murugan & Wu, 2017;Wolf et al., 2017a). A similar effect can be induced by injection of killed bacteria, causing cortical inflammation and prompting microglia to remove synapses from neuronal surfaces in a process called 'synaptic stripping' (Blinzinger & Kreutzberg, 1968;Trapp et al., 2007;Kettenmann, Kirchhoff & Verkhratsky, 2013). An even higher level of microglial activation can be achieved by injection of ricin into a nerve, causing neuronal degeneration; this induces the 'phagocytic' or 'amoeboid' state of microglia (Streit, Graeber & Kreutzberg, 1988).
Both surveillant and activated microglia express ionized calcium-binding adapter molecule-1 (Iba-1), which binds to actin and plays a central role in cytoskeletal reorganization (Sasaki et al., 2001), as well as CD11b [also known as macrophage-1 antigen (Mac-1) or complement receptor 3 (CR3)] and CD45 proteins. In addition, microglia constitutively express major histocompatibility complex type II (MHC II), a typical feature of antigen-presenting cells (Walker & Lue, 2015;Hong et al., 2016) and also expressed by myeloid cells outside the brain. Recent single-cell RNA-sequencing studies have revealed highly cell-type-specific gene markers for microglia such as the adenosine diphosphate purinoceptor 12 receptor (P2Y 12 ) and the transmembrane protein 119 (TMEM119) (Bennett et al., 2016;Böttcher et al., 2019). These markers are now being used to examine morphological changes in microglia as they allow the exclusion of peripheral Iba-1-positive cells that can infiltrate the brain. In addition, specific lectins (e.g. Griffonia simplicifolia B4 isolectin, Lycopersicon esculentum tomato lectin or Ricinus communis lectin) can be used to selectively recognize microglial cells in human (Morin-Brureau et al., 2018) and rodent (Färber & Kettenmann, 2006;Schwendele et al., 2012) brains. Binding of these lectins to phagocytic microglia may be stronger than binding to surveillant microglia, perhaps suggesting a higher level of antigenicity in activated compared to surveillant microglia (Biber et al., 2014).
The factors controlling microglial morphology and functions are manifold and may include well-known modulators of the immune system like glucocorticoids. In mouse hippocampus, for example, an experimental increase of glucocorticoid level in young mice (where glucocorticoid levels are normally low) enhanced microglial ramification (Van Olst et al., 2018). Amoeboid microglia are characterized by migratory movements forming pseudopods. Two different modes of movements can be observed: retraction and protrusion of processes with or without movement of the cell body. Molecules such as ATP, NO, the chemokine stromal cell-derived factor-1 (SDF-1), bradykinin, the endoplasmic reticulum Ca 2+ sensor stromal interaction molecule 1 (STIM1), purinergic receptors, and reduced nicotinamide adenine dinucleotide (NADH) oxidase influence migratory activity (Ifuku et al., 2007;Duan, Sahley & Muller, 2009;Lipfert et al., 2013;Bayerl et al., 2016;Lim et al., 2017). ATP has a central role not only in energy metabolism but also in purinergic signaling, as it operates as transmitter or co-transmitter and can be released through gap junction hemichannels. Damaged or dying cells also release large amounts of ATP. Once released, ATP is rapidly degraded by a series of various ectonucleotidases such as nucleoside triphosphatase, nucleoside diphosphatase, 5 -nucleotidase and purine nucleoside phosphorylase. All these enzymes are expressed on the surface of microglia, together with numerous purinergic receptors such as P2X 4 , P2X 7 , P2Y 6 and P2Y 12 (Boucsein et al., 2003;Brawek & Garaschuk, 2013). P2Y 12 is expressed over the entire surface of microglial cells allowing them to sense gradients of ADP/ATP, for example from damaged cells/tissues, and eliciting directed process extension (Davalos et al., 2005;Eyo et al., 2017). In P2Y 12 -knockout mice, directional process outgrowth is delayed indicating that the P2Y 12 receptor is involved in microglial chemotaxis (Haynes et al., 2006). Acquisition of the activated morphology was initially believed to reflect the neurotoxicity of microglia. For example, inflammatory diseases of the central nervous system (CNS) involve microglia-induced neuronal impairment (Trapp et al., 2007;Suzumura, 2013). However, many studies have indicated that the role of microglia is much more complex. In fact, microglia with an activated morphology can even be beneficial to certain aspects of neurogenesis, survival, migration, and differentiation (for reviews describing the complexity of microglia-related neurotoxicity and neuroprotection see Hellwig, Heinrich & Biber, 2013;Suzumura, 2013). For example, amoeboid microglia found during embryonic development in the spinal cord, corpus callosum, retina, hippocampus, and cortex play a key role in the colonization of the brain. These activated microglia continue to proliferate during early postnatal development and start to differentiate into a ramified phenotype after migration to their final destination (Mosser et al., 2017). In a murine model of stroke (middle cerebral artery occlusion, MCAO), activated microglia accumulate at the vessel injury site together with infiltrating blood cells. In transgenic mice expressing a mutant form of herpes simplex virus thymidine kinase (HSVTK), application of the virus-static ganciclovir led to a strong reduction in numbers of microglia. Following challenge of these microglia-deficient mice with MCAO, the area of infarct was significantly larger compared to control animals, thus showing a protective role of microglia after stroke (Lalancette-Hebert et al., 2007). In the same mice, monocyte entry into the brain was reduced shortly after engraftment in the presence of microglia; in mice treated with ganciclovir and thus microglia-depleted, the loss of microglia was compensated by repopulation with Iba-1-positive cells. Haematogenous monocytes, upon entering the brain, are able to replace microglial cells by acquisition of microglial properties (Varvel et al., 2012). However, despite their phenotypic similarity to resident microglia, the engrafted parenchymal brain monocytes displayed different transcriptional patterns (Cronk et al., 2018;Shemer et al., 2018). Thus, activated microglial morphology is not necessarily always connected with deterioration, but may have a role in protection of the nervous tissue as well (Hanisch & Kettenmann, 2007;Ekdahl, Kokaia & Lindvall, 2009).
(2) Functional signs of microglial activation Besides the above-described changes in microglial morphology, activation of microglia is also accompanied by several functional changes such as the production of pro-and anti-inflammatory cytokines, and release of both NO and ROS (Färber & Kettenmann, 2006;Sierra et al., 2007;Garaschuk, Semchyshyn & Lushchak, 2018). Many of the molecular pathways involved require intracellular Ca 2+ signaling and therefore changes in intracellular free Ca 2+ concentration ([Ca 2+ ] i ) might be able to provide an early and rather sensitive measure of microglial activation. Indeed, surveillant microglia in the healthy brain are known to maintain a low basal [Ca 2+ ] i accompanied by rare 'spontaneous' Ca 2+ transient increases in a minority of cells (Eichhoff, Brawek & Garaschuk, 2011;Pozner et al., 2015;. However, in vivo damage of even a single cell was shown rapidly (on a scale of milliseconds to seconds) and reliably to produce a so-called damage-associated [Ca 2+ ] transient in nearby microglia (Eichhoff et al., 2011). This robust Ca 2+ signal occurred well before any changes in morphology and likely initiated the activation of integrins, adhesion of microglial processes to the extracellular matrix, reorganization of the actin cytoskeleton, and, finally, the extension of microglial processes to the damaged cell. This process was driven by ATP/ADP-induced P2Y 12 receptor signaling, activation of phospholipase C and Ca 2+ -dependent phosphorylation of serine/threonine kinase (Akt) as well as phosphoinositide-3-kinase (PI3K)-mediated Akt phosphorylation (Eichhoff et al., 2011;Madry & Attwell, 2015). Both acute and chronic (neuro)inflammation were shown to be accompanied by an increase in 'spontaneous' Ca 2+ signaling of microglia Pozner et al., 2015) again pointing to changes in microglial [Ca 2+ ] i as a universal hallmark of activated microglia.
What are the functional consequences of these changes in [Ca 2+ ] i ? First, they likely impact the executive functions of these immune cells, many of which are Ca 2+ dependent. For example, in cultured microglia a lipopolysaccharide-mediated steady-state increase in [Ca 2+ ] i was shown to be both necessary and sufficient for the release of NO as well as TNF-α, IL-6, IL-12 and macrophage inflammatory protein-1a (Hoffmann et al., 2003). Similar effects are likely for the release of IL-1β, the synthesis of which by the NOD-, LRR-and pyrin domain-containing protein 3 (NLRP3) inflammasome is known to be Ca 2+ dependent (reviewed in . Secondly, [Ca 2+ ] i might initiate activation-induced changes in microglial morphology via the above-described processes of cytoskeleton remodeling.
Having summarized the activation-induced changes in morphology and function of microglia, we now turn to their response to eukaryotic parasites, an issue that has received little attention to date.

III. BRAIN INFECTION BY PROTOZOAN PARASITES
To compete effectively for resources or space, bacteria developed the capacity to secrete molecules (e.g. antibiotics), which are often species-specific and may not harm the host. Viruses rely on rapid replication and are most sensitive to inhibitors interfering with nucleic acid metabolism. Since cell replication in the brain is rather limited, such inhibitors can be relatively well tolerated by the host. In contrast to viral and bacterial invaders, however, protozoan parasites are more closely related to the host cells and anti-parasitic drugs often induce severe (sometimes deadly) side effects. It is thus not surprising that many of these parasites pose serious problems after invading the brain and the diseases that they cause are often harmful and difficult to treat. In many cases protozoan parasites have developed strategies to undermine the host's immune response. The rather sophisticated abilities of most of these parasites renders it difficult to present a general overview, so we instead introduce representative examples to describe their invasion strategies and the host's defense mechanisms, particularly those involving microglial activation. Note that crossing of the blood-brain (BBB) or the blood-CSF barrier (BCB) is not a simple process but involves biochemical or mechanical opening of the blood-organ border, and thus requires specific adaptations by the parasite to allow them to access this 'immune-privileged area'. Brain infection involves a dangerous balancing act on behalf of the parasite, since damage to delicate structures that are essential in body homeostasis could lead to the premature death of the host before the parasite can be transmitted to another suitable vector.

(1) Kinetoplastids
Kinetoplastids comprise a group of flagellated single-celled organisms with either one (Trypanosomatidae) or two (Bodonidae) flagella. These protozoa are characterized by a relatively large and complex organized mitochondrial DNA localized near to the flagellar pocket called the kinetoplast. We here consider only the human parasites Trypanosoma brucei, Trypanosoma cruzi and Leishmania donovani, which are the causative agents of the lethal vector-borne diseases sleeping sickness, Chagas disease and kala azar, respectively. Due to the distribution of their essential blood-sucking vectors, the Brucei group of subspecies of T. brucei (transmitted by the tsetse fly) is found exclusively in sub-Saharan Africa, T. cruzi (transmitted by reduviid bugs) is restricted to Latin America, while L. donovani (transmitted by the sand-fly) is found worldwide in the tropics and subtropics and is expanding into other areas in response to recent climatic change. T. cruzi and L. donovani are intracellular parasites, while T. brucei is free swimming in blood, lymph and cerebrospinal fluid and can move between different tissues. A cerebral localization has been described for all three parasites. For T. cruzi and L. donovani no defined cerebral symptoms are known, and a fatal outcome of these diseases is usually linked with heart failure, bowel obstruction and megacolon (Chagas disease), or opportunistic infections due to a strongly impaired immune system (kala azar). By contrast, brain infection has obvious symptoms in African trypanosomiasis, with dysregulation of the sleep-wake cycle and may progress to a deadly encephalopathy.

(a) Trypanosoma brucei
Human African trypanosomiasis is induced by two subspecies of the Brucei group, namely T. brucei rhodesiense and T. brucei gambiense. Treatment is difficult, as safe and efficient drugs are not available and severe adverse side effects are common. Drug resistance is also of increasing concern (Stein et al., 2014;Field et al., 2017).
The phenomenon of antigenic variation, i.e. the ability to express immunologically different variants of a protein on the cell membrane, is widespread among parasites and is highly advanced in African trypanosomes. Several hundred genes code for variant surface glycoproteins (VSGs) but only one VSG is expressed at a time, since this respective VSG gene is located at the single active expression site of the genome (Mugnier, Stebbins & Papavasiliou, 2016). If this gene, in Each VSG molecule comprises a protein (yellow) that is inserted via a glycosylphosphatidylinositol-inositol molecule (brown triangle and two curved lines) into the plasma membrane (black lines). Two integral membrane proteins buried within the VSG coat are shown in blue and green. A schematic representation of an antibody that only recognizes the N-terminal part of the tightly packed VSG coat, but not the internal parts of the membrane structure, is shown in red. (D) Schematic representation of part of a trypanosome. The flagellum and its origin from the flagellar pocket as well as the longitudinal orientation of the microtubules belonging to the cytoskeleton are shown. Each parasite possesses a single mitochondrion containing mitochondrial DNA. Trypanosomes do not contain peroxisomes but instead have glycosomes, i.e. organelles that contain glycolytic enzymes. The VSG surface coat is shown in yellow. Note that the VSG is continuously ingested by endocytosis (blue arrows) and mostly recycled to the plasma membrane; only a minor part is digested in lysosomes and replaced by its biosynthesis in the rough endoplasmic reticulum (rER) and Golgi apparatus (red arrows). Following an antigenic switch, however, all endocytosed 'old' VSG is processed in the lysosome and replaced by the new VSG type on the plasma membrane. Thus, following antigenic switching 'double expressors' exist for some hours until all VSG molecules in the coat are of the new antigenic type. kDNA, kinetoplast DNA; N, nucleus. a stochastic process, is exchanged for a copy of another VSG gene, mosaic genes can appear that possess part of the old and part of the new gene and hence the number of different VSG molecules is virtually limitless (Hall, Wang & Barry, 2013). The expressed VSG gene forms a surface coat consisting of 10 7 identical VSG molecules which surround the whole parasite including the flagellum and flagellar pocket (Fig. 1). All VSG molecules possess the same glycosyl phosphatidylinositol (GPI) membrane anchor (Fig. 1C), but this is buried within the surface coat and is inaccessible to antibodies (Almeida et al., 1994). The surface coat protects trypanosomes against cellular immune responses and complement reactions, while antibodies developed against the 'visible' VSG N-terminus opsonize the parasite and induce its death. In this way the host can destroy a significant part of the parasite population. Any trypanosomes expressing a different VSG type will then form a new population, resulting in an oscillating parasitaemia in the blood. In addition, trypanosomes can induce immunosuppression (Mansfield & Paulnock, 2005) and enter different organs (e.g. brain, skin, eye, testis, epididymis) (Wolburg et al., 2012;Casas-Sánchez & Acosta-Serrano, 2016).
Blood vessels within the choroid plexus are fenestrated to allow the formation of cerebrospinal fluid (CSF) from serum as it passes though their epithelial cells into the ventricles; these fenestrations allow easier penetration by trypanosomes (Schmidt, 1983;Wolburg et al., 2012), which also takes place in a cyclical process related to the parasite's population in the blood (Mogk et al., 2014b). In infected individuals, the choroid plexus, the CSF, the circumventricular organs (CVOs) and the pia mater are permanently infiltrated with trypanosomes, which may control their cell density by inducing differentiation from a long slender to a short stumpy morphology and the removal of the latter form by apoptosis (Figarella et al., 2005), i.e. without inducing inflammation. This adaptation by the parasite considerably reduces interaction with the immune system while retaining access to blood vessels to facilitate reinfection. Infection of adjacent brain areas occurs readily after a tsetse fly bite, as observed in a rodent model (Mogk et al., 2014a). It was also observed in rodents that trypanosomes penetrate the BBB early after infection (Frevert et al., 2012). This would seem more difficult to execute, because of (i) the bloodstream flow counteracting firm attachment of trypanosomes to the endothelial cells prior to BBB penetration, as observed for immune cells (Lutz et al., 2017), and (ii) a requirement for biochemical signals that physically open tight junctions. However, trypanosomes can induce immune cells to release such factors [e.g. C-X-C chemokine ligand 10 (CXCL-10); Kristensson et al., 2010] and bioluminescence studies have revealed trypanosomes within the cerebellum and the olfactory bulb (Myburgh et al., 2013). Interestingly, experimental placement of trypanosomes into the brain did not result in brain infection but in a rapid death of the parasites (Bafort, Schmidt & Molyneux, 1987;Wolburg et al., 2012). These data are consistent with results from co-cultivation of freshly isolated microglia cells and trypanosomes, where microglia were activated (e.g. NO release) and phagocytically active (Figarella et al., 2018). One may speculate that the observed latency of brain infection in humans (late disease stage) could reflect the ability of microglia to destroy parasites that occasionally cross the BBB by an unknown mechanism. At the same time, a meningeal infection will already be manifest, and trypanosomes localized within the CVO could cause the observed sleeping symptoms.
The final disease stage may occur with a delay of months (T. b. rhodesiense) or years (T. b. gambiense) by an as yet unknown mechanism. It is tempting to speculate that it may involve a weakening of the immune system in response to the continued infection of an increasingly malnourished patient, and be initiated by a massive brain infiltration of parasites from the meninges via the Virchow-Robin spaces, where the BBB is significantly less well organized. It should be kept in mind that while many patients suffering from African trypanosomiasis do die due to concomitant infections, so-called 'healthy' carriers may live symptomless for years or even decades (Mogk et al., 2017) before symptoms appear again.
(b) Trypanosoma cruzi American trypanosomes are primarily intracellular parasites that hide from the immune system by infiltrating host cells, especially macrophages and muscle cells ( Fig. 2A). T. cruzi is taken up by reduviids (kissing bugs) during a blood meal and develops in the insect's midgut. Infectious parasites are released in the faeces and infect humans either by penetrating intact mucosal membranes (e.g. the conjunctiva) or via skin irritations, often produced by the victim's scratching a bite site. Once in the body, the parasite enters a cell (usually a skin macrophage) and differentiates from the flagellated trypomastigote to the amastigote form with a reduced flagellum. Amastigotes multiply by binary fission and, following formation of trypomastigotes, kill the host cell and enter the bloodstream. Here infection of further macrophages occurs and the parasites are passively translocated to different organs, where released trypomastigotes will penetrate other host cells, especially heart muscle cells. The acute phase is often symptomless, apart from a generalized fatigue and unspecific symptoms such as fever, headache, or muscle pain. In about 90% of patients these symptoms disappear within the next 2 months. However, some patients die from myocarditis or meningo-encephalitis due to a massive inflammation mainly associated with infections via oral contamination, e.g. consumption of food contaminated with the faeces of the bug. Following the acute phase, the disease becomes chronic with continuous infection of new host cells by trypomastigotes released from dying infected cells. The trypomastigote form cannot divide, but can infect both host cells and the reduviid insect. During the chronic phase, patients may develop life-threatening cardiac insufficiencies and/or disorders of the digestive tract. Cell death induced by dividing amastigotes can lead to inflammation, lesions within the affected organs and fibrosis.
Involvement of the CNS can take place during the acute phase of the disease, in which a diffuse meningoencephalitis may occur, visible as lesions in the choroid plexus and meninges, the appearance of cerebral oedema, scattered inflammatory foci and perivascular infiltrates of lymphocytes. Amastigotes can be detected within astroglia and in the centre of microglial nodules (Chimelli, 2011). Parasites are usually not observed during the chronic phase, while microglial nodules and small lymphocyte aggregates have been documented (Da et al., 2000). Impairment of neuronal function is in general attributed to chronic hypoxia induced by the Chagas-typical cardiomyopathy, but does not support the existence of a defined brain stage in Chagas disease. Hence, although the parasite gains access to the brain parenchyma, its distribution and damaging abilities seem well restricted by the immune system and especially the microglial activities in this area. This view is supported by the massive increase in amastigote-infected astroglia and macrophages in brain tissue and the occurrence of an acute haemorrhagic encephalitis observed in immune-suppressed or HIV-positive patients (Fig. 2B) (Mortara et al., 1999;De Almeida et al., 2011).
(c) Leishmania spp. Leishmaniasis is induced by many different species of the genus Leishmania (Fig. 2C). This parasite is transmitted by sand-flies during a blood meal and similar to T. cruzi invades host cells, primarily macrophages. Depending on the species, the disease may appear in a cutaneous, mucocutaneous or visceral form. The former is characterized by skin ulcers that may or may not undergo self-healing, while the latter leads to severe infections of inner organs (e.g. bone marrow, liver, and spleen), organ enlargement and dysfunction. Brain infection is not usually clinically conspicuous, but neurological involvement has been reported repeatedly in cases of visceral leishmaniasis, as well as parasites occurring in the choroid plexus, CSF and the meninges (Melo et al., 2017). In mice experimentally infected with Leishmania donovani, parasites have also been detected within the brain, where infected leukocytes were used as a 'Trojan horse' to enter the CNS (Melo et al., 2017). Thus, as in T. cruzi infections, cerebral leishmaniasis can occur, but is largely controlled by the immune response in the brain. It is interesting to note that cell-culture experiments with Leishmania clearly showed a 'higher phagocytic ability and cytotoxic potential' of microglia as compared with macrophages (Ramos et al., 2014(Ramos et al., , p. 1052.

(2) Apicomplexa
The phylum Apicomplexa comprises a large group of protozoa which contain a specific organelle called an apicoplast. The apicoplast is a complex structure surrounded by four membranes and necessary for invasion of the parasite into a host cell. It also contains its own DNA and hosts four metabolic pathways: synthesis of fatty acids, isoprenoids, haem and iron-sulfur clusters. Apicomplexa possess a complex life cycle that includes asexual (sporozoites, merozoites) and sexual (gametes) stages, and several are important parasites of humans inducing diseases such as malaria (Plasmodium spp.), toxoplasmosis (Toxoplasma gondii), babesiosis (Babesia spp.), coccidiosis (Coccidia spp.), theileriosis (Theileria spp.), cryptosporidiosis (Cyptosporidia spp.), among others. We here consider only malaria and toxoplasmosis.

(a) Plasmodium falciparum
Malaria is induced by several species of Plasmodium, leading to different forms of the disease. Among these, P. falciparum is the most important, as it is responsible for the most dangerous form of malaria (Fig. 2D). It is transmitted by mosquitos (Anopheles spp.), which inject infective sporozoites into the bloodstream while taking a blood meal. Sporozoites first infect liver cells, thereby escaping the immune system. Within hepatocytes, the parasite multiplies by a process called schizogony and eventually differentiates to form merozoites. By lysing the cell, a merosome containing up to 90000 merozoites is released into the blood. A merozoite attaches to the surface molecules of an erythrocyte, orientates itself to the apical position, and with the help of the apicomplexan machinery enters the red blood cell by invagination of the membrane, thus forming a parasitophorous vacuole. Within this vacuole the parasite develops via trophozoites (the ring stage) to form schizonts that rupture the cell and infect new erythrocytes.
Within an erythrocyte, the parasite feeds on hemoglobin but also takes other nutrients from blood, such as glucose, fatty acids, water, minerals, etc. Since it is surrounded by the parasitophorous vacuole membrane, it must produce and insert the respective transporters and membrane proteins not only here but also into the erythrocyte's membrane. Some of the latter proteins (e.g. the well-known knob protein) attach to endothelial proteins lining the blood vessels thus leading to sequestration of infected erythrocytes to certain organs. Although normally advantageous to the parasite, this may lead to major complications, such as blocking capillaries in the brain, leading to stroke and sudden death of patients. Cerebral malaria is thus a major complication of the disease and is characterized by impaired consciousness and coma that is often fatal. Clinically, brain swelling, intracranial hypertension and eye changes involving pupil size and reactivity or ocular movement are commonly observed (Idro et al., 2010). Interestingly, neuronal necrosis is limited in cerebral malaria and the main complications of P. falciparum infection arise from sequestered parasitized red blood cells (pRBCs). These cells express P. falciparum erythrocyte membrane protein-1 (PfEMP-1) on their plasma membranes, the major mediator of parasite adhesion. This protein is encoded by a multicopy gene family, var. Individual domains of PfEMP-1 bind to different receptors on endothelial cells, of which intercellular adhesion protein-1 (ICAM-1), endothelial protein C receptor (EPCR), and CD36 are the most prominent (Turner et al., 2013;Hsieh et al., 2016). Recently, it was shown that pRBCs from patients with cerebral malaria bind at higher rates to brain endothelial cells than those from patients with uncomplicated malaria (Storm et al., 2019). Interestingly, using an engineered three-dimensional human endothelial microvessel model, it was also demonstrated that pRBC binding is heterogeneous and depends on the expressed type of PfEMP-1 as well as on several physiological variables, such as blood flow speed and endothelial preactivation by TNF-α (Bernabeu et al., 2019). Besides sequestration of pRBCs, rosetting of RBCs and other pRBCs contribute to microvasculature obstruction and endothelial dysfunction thus impairing perfusion and inducing hypoxia. The pathogenesis of cerebral malaria is multifactorial and also involves the host's immune response (Mandala et al., 2017;Wolf, Sherratt & Riley, 2017b).
In an animal model of cerebral malaria, microglial cells proliferated during the onset of the infection and showed upregulation of 'genes involved in immune responses and chemokine production' following the onset of cerebral malaria symptoms (Capuccini et al., 2016, p. 1). In vivo in mice it was shown that parasite-specific CD8 + T cells interact with endothelial cells in an interferon-gamma (IFN-γ ) dependent manner, and that this was responsible for activating endothelial cells in the brain microvasculature (Swanson et al., 2016). Moreover, CD8 + T cells expressing perforin were responsible for vascular permeability breakdown inducing brain swelling and oedema (Huggins et al., 2017). Therefore, although cerebral malaria is primarily a problem of a disrupted blood supply to the brain, the CNS can be affected in a broader sense and is likely to be protected by its defense machinery, including the microglia. This suggestion is supported by the fact that while patients that survive cerebral malaria may recover fully (Muntendam et al., 1996), studies in children have demonstrated neurological deficits or epilepsy as long-term sequelae (Brewster, Kwiatkowski & White, 1990).

(b) Toxoplasma gondii
Toxoplasma gondii is not transmitted by blood-sucking insects, rather infection usually occurs by eating insufficiently cooked infected meat (e.g. pork, lamb or venison) or by smear contamination with cat faeces. Felids are the primary host for Toxoplasma gondii (Fig. 2E). Following uptake in infected prey, the parasite survives stomach digestion and enters the epithelial cells of the small intestine. Here it undergoes sexual reproduction leading to cysts containing infectious sporozoites. Cysts are released in the faeces and survive for several months until ingested by homeothermic animals. Digestive enzymes open the cyst thereby releasing sporozoites to infect surrounding epithelial cells. Like plasmodium, the parasite replicates in a parasitophorous vacuole and is, following rupture of the host cell, released as highly infective merozoites, here usually called tachyzoites. Tachyzoites spread within the bloodstream by entering macrophages or dendritic cells, using them as 'Trojan horses' to enter various tissues, especially muscle, brain and eye. Dendritic cells in particular become more motile after infection, thus accelerating dissemination of the parasite (Lambert et al., 2006). After entering cells within these organs, the membrane of the parasitophorous vacuole is rearranged to form a cyst and the parasite differentiates to form slowly dividing bradyzoites that persist in the body. There is evidence that free tachyzoites present in the vascular compartment are able to invade endothelial cells in different tissues, with higher infection rates in microvasculature with a diameter <10 μm, from where they cause tissue invasion (Konradt et al., 2016). As in T. cruzi, there is an acute phase that, although often undetected, may manifest as unspecific symptoms like headache, swollen lymph nodes, fatigue, etc. Symptoms may be increased or severe in people with a weak or compromised immune system (HIV, chemotherapy) and may include confusion, coordination problems, encephalitis as well as lung and vision problems, the latter may even lead to retina necrosis. Since the parasite can cross the blood-placenta barrier, it can induce miscarriage or stillbirth, or the infant may suffer from severe symptoms or may experience them later in life. It is estimated that up to 50% of the population in European countries (www.euro.who.int) and one-third of the World's population (Pappas, Roussos & Falagas, 2009) are infected with Toxoplasma, 95% of which are 'healthy' carriers as the immune system controls the disease.
In vivo experiments in rodents showed that, after entering the brain, tachyzoites preferentially interact with and infect neurons (Cabral et al., 2016;Mendez & Koshy, 2017). However, parasites have also been detected within astrocytes and microglia (Carruthers & Suzuki, 2007;Alvarado-Esquivel et al., 2015). In in vitro assays using dendritic cells, infected microglia increased their motility and became sensitive to T-cell-mediated killing, thus paradoxically spreading the parasite further in the CNS (Dellacasa-Lindberg et al., 2011). In addition, microarray analysis revealed upregulation of proteins involved in the immune response and in inhibition of apoptosis in brain cells upon the onset of infection (Blader, Manger & Boothroyd, 2001). Thus, there appear to be myriad interactions between the host and parasite and it is astounding that the parasite burden in this case is so well controlled and that health problems are, in most cases, negligible.
Interestingly, numerous studies have revealed that toxoplasma-infected mice lose their inborn aversion to cat odors (reviewed in Ingram et al., 2013), suggesting parasite-induced reorganization of neuronal synapses. In support of this, a recent study found clear indications for changes in fear memory by analyzing neurotransmitter levels in different brain areas as well as cortex and amygdala interactions (Ihara et al., 2016). It thus seems that T. gondii have evolved adaptations to influence their transmission success by changing the brain functions of their hosts. Other studies have linked toxoplasmosis to mental disorders such as depression, mood or bipolar disorders and schizophrenia (Fekadu, Shibre & Cleare, 2010), although such correlations could also arise from the inflammatory response to infection (Mendez & Koshy, 2017;Li et al., 2019b).

(3) Amoebidae
Prominent members of the Amoebidae causing human diseases include Naegleria fowleri (primary amoebic meningoencephalitis), Acanthamoeba spp. (granulomatous amoebic encephalitis; keratitis) and Entamoeba histolytica (dysentery). Amoebae change their morphology by forming pseudopods with the help of microtubules. In this way, extrusions are formed at one position of the cell and cytoplasm moves into the newly formed space. Pseudopods are simultaneously retracted on another side of the cell, allowing an overall crawling-like movement often driven by chemotaxis.
N. fowleri is a thermophilic, bacteriophagic amoeba (Fig. 2F) often found in soil or warm fresh water like rivers, ponds, lakes or poorly chlorinated swimming pools. People can become infected following water-based activities such as swimming, snorkeling, diving or surfing, where water accidently enters the nasal cavity. The protozoon first attaches to the nasal mucosa, before traveling along the olfactory nerve to the cribriform plate. This plate is part of the ethmoid bone and separates the nasal cavity from the brain. It has a spongy structure that allows penetration of the olfactory nerve and thereby of the moving parasite. The porosity of this structure is higher in children and young adults (Krmpotić-Nemanić et al., 1998), perhaps accounting for the higher number of infections in these groups. Alternatively, children and young adults may be more exposed to contaminated water because they are more likely to take part in outdoor activities than other groups. Once in the olfactory bulb, N. fowleri induces reactions of the innate immune system, but may be resistant to these and can destroy neurons and the brain structure, hence the common name 'brain-eating amoeba'.
Considering the number of people generally in contact with the same contaminated water, only a small proportion seem to develop a fatal amoebiasis. The reason for this is not well understood, but it seems likely that the immune system has a major role. From in vitro assays and studies on experimental animals and on human infections, it was clearly shown that neutrophils and macrophages (and less often microglia) infiltrate the brain areas infected with N. fowleri and that complement also is activated (Oh et al., 2005;Marciano-Cabral & Cabral, 2007). However, highly pathogenic N. fowleri strains can compromise complement-induced cell lysis by the surface expression of complement regulatory proteins and by shedding of the C5b-C9 membrane attack complex of complement in the form of vesicles (Toney & Marciano-Cabral, 1992. The cytotoxic activity of macrophages also may fail because secreted lytic factors such as TNF-α with or without the assistance of interleukins fail to harm the parasite (Fischer-Stenger, . Amoebae possess 'food cups' on their pseudopodia able to take up bacteria, yeast or tissue of infected animals. In addition, the parasite secretes proteases, phospholipases and pore-forming proteins (naegleriapores), which together enable invasion and cell/tissue destruction. N. fowleri infection can proceed rapidly, leading to death within 4-7 days. N. fowleri thus provides an example where adaptation and long-term infection control are not possible: either the parasite is destroyed immediately by the immune response or it kills the host within days.

IV. MICROGLIAL REACTIONS AGAINST PARASITIC INFECTIONS
Studies performed in recent decades have shown that the brain is not the immune-privileged organ it was originally thought to be. Indeed, it has been demonstrated that systemic inflammation leads to microglial activation and inflammation within the brain (Hoogland et al., 2015). The brain communicates with the periphery in several ways that involve at least two pathways. One pathway is through afferent nerves, while the other utilizes Toll-like receptors in cells of the CVOs and choroid plexus (Dantzer et al., 2008). In both cases, PAMPs induce the production of pro-inflammatory cytokines, which can enter the brain either via transporters or by diffusion (Vitkovic et al., 2000). In simple terms, microglia can detect a peripheral infection through humoral information carried by soluble molecules secreted by invaders or immune cells. Detection of a parasite by the host immune system outside the brain thus likely triggers a microglial reaction, as has been shown for peripheral injection of the bacterial cell wall component lipopolysaccharide (LPS) (Cunningham, 2013;Hoogland et al., 2015). If triggered, a sequence of events occurs: (i) up-regulation of IL-1β, TNF-α, and TGF-β1 (after 2-6 h); (ii) increase of Iba-1 expression (after 12 h); and (iii) microglial de-ramification (after 24-48 h) (Biesmans et al., 2013;Norden et al., 2016). Note, however, that LPS can induce variable responses of the immune system ranging from mild flu-like symptoms to sepsis (Hoogland et al., 2015).
In contrast to bacteria, many parasites have evolved strategies to co-exist with their host. Consequently, many parasitic infections (with notable exceptions according to the virulence factors of determinate strains) are long-lasting diseases that take months or even years to progress. One such strategy is evasion of the host immune system, which can be achieved by several mechanisms including: (i) invasion of 'immune-privileged' tissue such as the CNS or adipose tissue; (ii) development of intracellular replication stages that make the parasites 'invisible'; (iii) changes to surface antigens during the course of infection through antigenic variation; or (iv) development of temporarily inactive forms, which possess a low metabolic rate and do not replicate (quiescence). Under such scenarios, peripheral inflammation during the early stages of parasitic infections may be weaker as compared with that induced by LPS and would be detected by microglia only in areas to which cytokines diffuse first. Therefore, it is likely that while such parasites are located in the periphery, any microglial reaction would not be widespread but instead be heterogeneous, showing an activated phenotype only in regions to which humoral information is accessible. Upon parasitic invasion of the brain, however, microglial activation would become much more evident.
Areas of the brain with microglial accumulation have been reported for many different parasitic diseases. A post-mortem examination from an 81-year-old male patient infected with the amoeba Balamuthia mandrilaris showed microglial cells surrounding necrotic lesions in the thalamus, midbrain, pons, medulla, and cerebellum, where trophozoites and cysts were also found (Itoh et al., 2015). In experimental leishmaniasis in rodents or in naturally infected dogs, high levels of Iba-1 or a high density of microglia were observed in the prefrontal cortex or in the ependymal/subependymal areas, respectively (Melo & Machado, 2011;Portes et al., 2016). Microgliosis is also a hallmark of experimental cerebral malaria. In this case, activated microglia were identified in the olfactory bulb and along the rostral migratory stream in the parenchyma, clustered near vessels with thrombi caused by infected erythrocytes (Capuccini et al., 2016;Hoffmann et al., 2016;Wilson et al., 2018). In fatal cases of cerebral malaria caused by P. falciparum in English travelers, accumulations of microglia were observed around small veins in the brain parenchyma (Janota & Doshi, 1979). In a rodent model of a chronic T. gondii infection, brain inflammation characterized by microglia and astrocyte activation has been documented (Hermes et al., 2008), and in a post-mortem analysis of patients with acquired immunodeficiency syndrome, toxoplasma cysts were found to be surrounded by activated microglia (Nebuloni et al., 2000). Similarly, in both rodents and humans, the presence of microglial hyperplasia and microglial nodules was recorded in the late stages of trypanosomiasis (Chianella et al., 1999;Rock et al., 2004;Pittella, 2013). Accumulations of T. cruzi amastigotes in the centre of microglial nodules have been reported in humans (Pittella, 2013). Interestingly, microglia containing amastigotes were also detected in rodents infected with T. cruzi (Bombeiro et al., 2012;Morocoima et al., 2012).
Following activation, microglia stimulate an inflammatory response by synthesis of pro-inflammatory cytokines, chemokines, free radicals and complement, which subsequently induce activation of other glial cells and, eventually, recruitment of mononuclear leukocytes into the CNS (Goldmann & Prinz, 2013). The goal is to fight and control injuries locally and to avoid expansion of the infection. Interestingly, in the case of T. gondii, a parasite that can persist within the host during their entire life, the experimental data indicate that microglia can act in defense in two ways: (i) microglial activation effectively inhibits parasite growth through the production of NO and IFN-γ (Lüder et al., 1999;Kawanokuchi et al., 2006;Wang & Suzuki, 2007;Sa et al., 2015); (ii) microglial activation in the cortex and hippocampus of infected mice led to neuronal apoptosis (Zhang et al., 2014). In the latter case, treatment with minocycline reduced microglial activation and consequently neuronal cell death (Zhang et al., 2014). Clearly, the microglial response to parasites involves fine-tuned coordination to allow elimination of the invading parasites while at the same time avoiding cerebral damage. Any disruption of this balance could lead to progression of the neuropathology.
So, how is the crosstalk between host cells and parasites mediated, i.e. which cytokines and chemokines are involved in protozoan infections? Table 1 provides a summary of the cytokines, chemokines and some other factors that are known to be secreted by host immune cells under in vitro conditions upon contact with selected protozoans. Notwithstanding the fact that each infection probably activates parasite-specific signaling pathways, the available in vitro data suggest that infections by almost all of these parasites induce the secretion of the pro-inflammatory mediators TNF-α and/or IL-6 in host cells. Both molecules are involved in the acute immune response, fever induction, and contribute strongly to the development of sickness behavior (Dantzer et al., 2008). Microglia possess receptors for both molecules and can also synthesize and secrete them, meaning that TNF-α and IL-6 represent both inducer and effector molecules for microglia (reviewed in Rock & Peterson, 2006). Another cytokine worth highlighting is IFN-γ . Production of this cytokine by resident microglia is essential in the suppression of cerebral growth of T. gondii (Sa et al., 2015). Conversely, IFN-γ secreted by CD8 + T cells in response to trypanin (a factor secreted by T. brucei) seems to favor the peripheral proliferation of T. brucei but also induces up-regulation of inducible NO synthase (iNOS) in microglia (Olsson et al., 1993;Girard et al., 2000). Up-regulation of IFNγ production leads to NO-dependent killing of Leishmania parasites (Liew et al., 1990). In vitro observations showed that microglial cells represent a hostile environment for this parasite (Baetas- Da-Cruz et al., 2004;Ramos et al., 2014). Moreover, cytokines TNF-α and IL-12 were produced by both macrophages and microglia after co-incubation with three different Leishmania species. However, only macrophages produced TGF-β (a potent negative regulator of immune responses) (Ramos et al., 2014). Experiments on IL-4 receptor-deficient BALB/c mice (CD11c cre IL-4Ra −/lox ) demonstrated that this interleukin is involved in parasite clearance; CD11c cre IL-4Ra −/lox mice showed a higher susceptibility to L. major infection with a prominent increase in parasite burdens in the brain and other organs, reduced IL-12 and NO synthesis, but increased IL-10 production (Hurdayal et al., 2013).
Taken together, it seems plausible that, in the early stages of infection, microglia receive humoral signals that stimulate their transition into an 'alerted' state. This prepares microglial cells to attack any parasites that enter the brain. Based on in vitro observations, it can be hypothesized that direct interaction between parasites and microglia induces activation of the latter leading to elimination of the parasite either via phagocytosis, by release of NO, or by a combination of these. A conclusive analysis of microglia-parasite interactions will depend on direct in vivo observations, which remains a challenging task. Nonetheless, recent advances in high-resolution imaging techniques have made such in vivo experiments possible. For example, intra-vital imaging using minimally invasive two-photon microscopy allows longitudinal single-cell-level monitoring of microglial structure and function in the living brain. Structural changes in microglia have been assessed previously using mice expressing green fluorescent protein (GFP) in the presence of C-X3-C motif chemokine receptor 1 (CX 3 CR1) or the Iba-1 promoter, while cells expressing genetically encoded Ca 2+ -sensitive dyes were recently used for in vivo monitoring of microglial function (Pozner et al., 2015;. In vivo analysis of experimentally infected animals accompanied by post-hoc immunohistochemistry using recently discovered microglia-specific markers could enable a better understanding of the role of microglia in protozoan infections. In addition, generation of new mouse lines utilizing promotors of these recently identified microglia-specific markers could allow more precise in vivo characterization of microglial behavior and motility, opening new avenues for further research.

V. POSSIBLE SCENARIOS FOR MICROGLIA-PARASITE INTERACTIONS
The 'brain stage' of protozoan parasite diseases is largely under-investigated and the underlying molecular mechanisms remain unclear. Consistently, medical treatment remains a risky venture since the available drugs may  (Mammari et al., 2014) CCL, chemokine (C-C motif) ligand; CXCL, chemokine (C-X-C motif) ligand; GRO, growth-regulated oncogene; IL, interleukin; MCP, monocyte chemoattractant protein; NO, nitric oxide; SERPIN, endothelial plasminogen activator inhibitor; TgPrx3, Toxoplasma gondii peroxiredoxin 3; TNF-α, tumor necrosis factor α. a Microglial cells did not produce the anti-inflammatory cytokine TGF-β. b Plasmodium berghei causes cerebral malaria in rodents. c NO is a gaseous signaling molecule. d TgPrx3 is an antioxidant enzyme that protects cells from oxidative stress caused by hydroperoxides. e MCP-1 is also known as CCL2. f GROα is also known as CXCL1. g Serpin E1 is also known as endothelial plasminogen activator inhibitor or plasminogen activator inhibitor-1 (PAI-1).
themselves lead to CNS damage. Can we draw any conclusions that apply to all of these parasites, i.e. are infection strategies and host responses similar, or do we have to treat each parasite as having developed a unique infection mechanism?
In terms of infecting the brain, there appears to be a clear difference between parasites that live freely in the host and intracellular parasites, which can use a 'Trojan horse' strategy by residing within host cells. In addition, it is not in the parasite's interest for the transmission cycle to be interrupted by rapid death of the host following parasite entry to the CNS. N. fowleri provides an example of a parasite that lives freely in the host bloodstream and tissues and does not require a second host to complete its life cycle. Host defenses involve the migration of immune cells (including microglia) to the invasion area and their concerted action releasing complement proteins and cytotoxic compounds to kill the invaders (Fig. 3). N. fowleri has evolved counter-defence mechanisms to render complement reactions ineffective while producing cytotoxins to destroy microglia and other host cells.
The situation is different for intracellular parasites that enter the brain inside infected host cells (e.g. T. cruzi, L. donovani, T. gondii). In these infections, microglia can (1) by humoral signals including interleukins (ILs) and tumor necrosis factors (TNFs), or (2) by direct contact with parasites (e.g. Naegleria fowleri). Activation by humoral signals leads either to morphological changes in the microglia or to up-regulation of activation markers. Direct contact with a parasite induces the microglia to become phagocytically active and to release cytotoxic compounds, especially NO. Activated microglia are able to interact with parasites or infected cells and release cytotoxic metabolites. In both cases, the final step is the phagocytic uptake of the remnants of lysed parasites by microglia.
be activated by humoral signals. Chemotaxis allows the microglia to identify the infected cells, which are then surrounded and either killed by cytotoxic compounds or through the production of pro-inflammatory molecules such as NO, TNF-α, etc. (Fig. 3), or are embedded within microglial nodules. It is interesting here to note that Leishmania that infect macrophages can survive, while its uptake by microglia kills the parasite. A balance may be achieved whereby infected cells remain within the CNS but are continuously controlled by microglia to prevent unlimited parasite growth or excessive inflammation. This may reflect an evolutionary adaptation by the parasite: the host is symptomless while the parasite survives in a 'safe' compartment, retaining the potential to re-infect the blood or to modify the host's behavior, as discussed above for toxoplasmosis and African trypanosomiasis.
Malaria is a particularly interesting case. Since the causative parasite, Plasmodium spp. infects erythrocytes, which do not enter the CNS, the primary issue in cerebral malaria is the risk of stroke. However, the accumulation of microglia at sites of injury is a prominent feature in cerebral malaria. Clearly, perivascular involvement recruits microglia and it might be admissible to speculate that cytokines and NO release are primarily used to prevent vessel clogging.
T.b. gambiense and T.b. rhodesiense, form a separate group, since they are free-moving in the host, but rapid death of the host would interrupt their transmission cycle. It is thus not surprising that involvement of the brain takes several months (T.b. rhodesiense) or several years (T.b. gambiense). It is interesting to speculate how this takes place, because it was not thought that the parasite could cross the BBB during the first stages of infection. Experimental data, however, suggest that trypanosomes may indeed cross the BBB earlier in infection but are neutralized by the action of microglia (Fig. 3) (Figarella et al., 2018). Most important is the parasite's ability to cross the BCB at an early time point, whereby it can cause meningoencephalitis. This term describes an inflammation of both the meninges and the brain, i.e. two distinct areas separated by the BBB. The BCB is formed by tight junctions between epithelial cells in the choroid plexus and pituicytes in the CVOs. Endothelial cells in the choroid plexus are also linked by a special sort of tight junctions (Pfeiffer, Mack & Wolburg, 2017), but they are fenestrated and this appears to allow parasites to enter the stroma of the choroid plexus. Electron micrographs of infected rodents show that parasites accumulate in this space outside the bloodstream, where they have access to the epithelial cells and their tight junctions. Since parasites appear in the CSF, they must either use a para-or transcellular route to move into the ventricle. CSF from all four ventricles combines and flows into the subarachnoid space (SAS) allowing access to the meninges, especially the pia mater (Schmidt, 1983;Mogk et al., 2017). However, note that the SAS and the Virchow-Robin space are continuous with the stroma of the choroid plexus (Brightman & Reese, 1969;Pfeiffer et al., 2017). As a consequence, monocytes or parasites could access the SAS from the stroma (Wolburg & Mack, 2014;Pfeiffer et al., 2017). The existence of inflammatory cells in the CSF was traditionally thought to be due to their transmigration through the choroid plexus epithelium, and thus the potential alternative of a direct route from the stroma to the SAS has not been investigated experimentally. We consider it likely that African trypanosomes reside permanently between cells of the pia mater and within the CVOs. Following weakening of the immune system, a massive infiltration of parasites into the brain could take place that may overwhelm microglia function and eventually lead to a deadly encephalopathy.

VI. CONCLUSIONS
(1) Most protozoan parasites in the brain seem to be controlled by the innate immune system including microglia, although conclusive experimental evidence is still limited. As long as the elements of the immune system work in a coordinated manner, the host can remain symptomless.
(2) The situation escalates, however, as soon as the immune system is weakened. This may happen by (i) parasite-driven processes, like antigenic variation or active immune suppression; (ii) co-infections with viruses (e.g. HIV), bacteria, protozoa or worms; or (iii) immune suppression due to medical treatment, e.g. after organ transplantation.
(3) Microglia are likely to have an essential role in the control of chronic parasitic infection and the suppression of excessive inflammation. However, this has not been tested directly; in vivo investigations are now needed to understand the role of microglia in protozoan infections.