Central nervous system in cerebral malaria - Nature

3 downloads 0 Views 934KB Size Report
active participant in the induction of immunopathology? ... parasitized erythrocytes in the central nervous system (CNS) and immunopathological processes con-.
Immunology and Cell Biology (2001) 79, 101–120

Review Article

Central nervous system in cerebral malaria: ‘Innocent bystander’ or active participant in the induction of immunopathology? I S A B E L L E M M E DA NA , 1 * G E E TA C H AU D H R I , 1 TA I L O I C H A N - L I N G 2 a n d NICHOLAS H HUNT1 Departments of 1Pathology and 2Anatomy/Histology, University of Sydney, New South Wales, Australia Summary Cerebral malaria (CM) is a major life-threatening complication of Plasmodium falciparum infection in humans, responsible for up to 2 million deaths annually. The mechanisms underlying the fatal cerebral complications are still not fully understood. Many theories exist on the aetiology of human CM. The sequestration hypothesis suggests that adherence of parasitized erythrocytes to the cerebral vasculature leads to obstruction of the microcirculation, anoxia or metabolic disturbances affecting brain function, resulting in coma. This mechanism alone seems insufficient to explain all the known features of CM. In this review we focus on another major school of thought, that CM is the result of an over-vigorous immune response originally evolved for the protection of the host. Evidence in support of this second hypothesis comes from studies in murine malaria models in which T cells, monocytes, adhesion molecules and cytokines, have been implicated in the development of the cerebral complications. Recent studies of human CM also indicate a role for the immune system in the neurological complications. However, it is likely that multiple mechanisms are involved in the induction of cerebral complications and both the presence of parasitized erythrocytes in the central nervous system (CNS) and immunopathological processes contribute to the pathogenesis of CM. Most studies examining immunopathological responses in CM have focused on reactions occurring primarily in the systemic circulation. However, these also do not fully account for the development of cerebral complications in CM. In this review we summarize results from human and mouse studies that demonstrate morphological and functional changes in the resident glial cells of the CNS. The degree of immune activation and degeneration of glial cells was shown to reflect the extent of neurological complications in murine cerebral malaria. From these results we highlight the need to consider the potentially important contribution within the CNS of glia and their secreted products, such as cytokines, in the development of human CM. Key words: astrocytes, cerebral malaria, cytokines, endothelium, glia, immune response, microglia, pathology.

Cerebral malaria: A common fatal complication of malaria infection Malaria still remains a serious health problem. There are an estimated 300–500 million cases and over 1 million deaths from malaria each year.1 Despite the early success of malaria eradication campaigns in the 1960s, the disease has undergone a resurgence as a result of resistance to the commonly used insecticides and drugs by mosquitoes and parasites, respectively. Furthermore, progress in human vaccine development has been slower than hoped as a result of the complexity of the parasite’s life cycle and the antigenic diversity exhibited by each stage. The resurgence of malaria infections and the lack of a suitable vaccine has brought about the realization that a greater understanding of the pathogenetic processes associated with the disease is necessary if effective modes of treatment or prevention are to be found.

Correspondence: NH Hunt, Department of Pathology, Blackburn Building (D06), University of Sydney, NSW 2006, Australia. Email: [email protected] *Present address: IM Medana, Nuffield Department of Clinical Laboratory Science, The John Radcliffe Hospital, University of Oxford, UK. Received 11 July 2000; accepted 21 November 2000.

Cerebral malaria (CM) is the most severe complication of Plasmodium falciparum infection, usually occurring in infants, pregnant women and visitors to areas endemic for malaria,2 and carries a mortality of 30–50%.3–5 Most often, patients present with fever, headache, delirium progressing to an acute febrile stupor, followed by coma.5,6 Cerebral complications may be the predominant organ pathology evident and either present abruptly (typically African children living in malaria-endemic areas), or develop as a late complication with multisystem involvement (typically non-immune adults from South-East Asia7,8). Given the different patterns of neurological involvement in African children and South-East Asian adults, there can be problems drawing parallels with neuropathological pathways. In recent years, several groups in Africa have made detailed clinical studies to identify those children at high risk of death from P. falciparum infection.9,10 African children represent more than 90% of all malaria cases and 3000 deaths, of children under 5 years of age, occur per day.1 It is now clear that neurological complications in African children with P. falciparum are unlikely to be the result of a single pathophysiological process. Rather, several different, but overlapping, pathological mechanisms occur in individuals culminating in the syndrome known as cerebral malaria.9,11–14 Factors that were found to be strongly related to mortality in

102

Table 1

IM Medana et al.

Some similarities between pathological findings in the central nervous system of mice and humans with cerebral malaria

Observations in retinal and cerebral tissue Loss of vascular cell integrity/tissue oedema Congestion of microvessels with parasitized erythrocytes Haemorrhages Mononuclear cell adherence to, or extravasation through, the vascular endothelium Astrocyte response (redistribution, astrogliosis, activation) Microglia and perivascular macrophage response (redistribution, morphological changes, activation) Cytokine expression Neurological complications, including convulsions, paralysis, coma

Murine

Human

+ –/+ + +

+ + + –/+

+ +

? +

122, 192, 193, 200 123, 126, 127, 131, 200, 201

+ +

+ +

72, 125, 142, 156, 202 5, 17, 23, 28, 35–38, 203, 204

children presenting with neurologic involvement included respiratory distress, hypoglycaemia and deep coma.9,15 This implies that there are differences in the underlying pathogenic factors in different patient subgroups. The presence of multiple pathological pathways is a feature of the disease that must be taken into account when trying to model disease progression in a single animal model. Currently, the full spectrum of pathological changes in CM mice has not been completely explored.

Neuropathological findings The major histological findings in postmortem brain tissue are cerebral microvessels congested with parasitized red blood cells (pRBC), haemorrhages and the deposition of malaria pigment, and oedema (5,16,17 reviewed in Turner18). There is doubt over whether oedema occurs pre- or postmortem.19 However, subtle mechanisms of blood–brain barrier (BBB) dysfunction have been demonstrated in a recent study of human CM, including leakage of fibrinogen into the CNS parenchyma and a reduction in the expression of the endothelial cell junction proteins ZO-1, occludin, and vinculin.20 Inflammatory cell infiltration is a variable finding.17 Occasional granulomatous lesions (Dürck’s granuloma), which occur predominantly in the white matter of the brain, also are observed.16

The retina: The tissue of choice for studying neuropathological changes during cerebral malaria? Histopathological studies of CM have commonly used brain sections. Although this is the primary organ of interest, information about the 3-D spatial relationship of cellular elements cannot be acquired because of the need to section the tissue. However, this can be overcome by using the retina as a whole-mount preparation, where the entire vascular plexus with its relationship to neighbouring structures, such as glial cells and neurones, remains intact. Unlike the CNS, the retina’s relative thinness, approximately 200 µm, allows it to be examined as a whole-mount preparation, where mild, early microvascular changes and associated cell–cell interactions

References 6, 16, 17, 20, 23, 25, 27, 35, 36, 70, 71, 191–193 18, 23, 194–198 18, 24, 26, 35, 38, 193, 199 23, 35, 36, 70, 71, 142

can be detected.21,22 Chan-Ling, Neill and Hunt23 were the first to use the retina, as a whole-mount preparation, in studies of fatal murine CM and noted that microvascular changes reflected those previously observed in the brain (see Table 1). Histopathological studies of patients that have died of human CM have not utilized postmortem retinal tissue, although ophthalmologic abnormalities have been observed in live patients with CM during retinal examinations. Many similarities were found between the retinal changes in murine CM and the human condition (see Table 2). The occurrence of retinal haemorrhages in CM has been reported to have prognostic significance in Thai patients (mainly adults) with CM,24 but not in African children.25,26 Conversely, retinal oedema is indicative of a poor prognosis in African children,25 but not in Thai patients.27

Theories on the pathogenesis of cerebral malaria Many host and parasite factors have been proposed to play a role in the induction of cerebral complications during P. falciparum infection (reviewed in Newton and Warrell28). The more prominent theories include: mechanical or sequestration, toxin, cytokine, nitric oxide (NO), reactive oxygen species (ROS), permeability, immunological hypotheses. One of the dominant hypotheses, the ‘sequestration’ hypothesis (reviewed in Berendt et al.29), suggests that sequestration causes multifocal abnormalities in cerebral blood flow (including both vascular obstruction and dilatation) leading to microheterogeneous hypoxaemia, acidosis, hypoglycaemia and other metabolic derangements affecting brain function, resulting in coma. However, this mechanism seems insufficient to explain all the known features of human CM. For example, blockage of blood flow would be expected to result in stroke-like pathology involving anoxic neuronal injury and severe residual impairment. This contrasts with the findings in the majority of human CM cases, in which prolonged coma proceeds to apparent functional recovery.30 A further observation is that some patients develop CM with severe oedema and haemorrhage, but have a virtual absence of parasitized erythrocytes in cerebral vessels.5,16,17 Conversely, sequestration has been found in patients who do not develop CM.31

Any malaria patient (Papua New Guinea)

P. falciparum (Thailand)

P. falciparum (Zambia)

P. falciparum (Thailand)

P. falciparum (Malawi)

P. falciparum (Malawi)

50 (adults)27

150 (> 6 years)24

67 (children)26

32 (14–49 years)27

56 (children)205

141 (children)25 Direct and indirect ophthalmoscopy, optic nerve colour and topography, calibre and patence of vessels out to the midperiphery, presence of retinal haemorrhages and oedema

Indirect ophthalmoscopy, optic nerve colour and topography, calibre and patence of vessels out to the midperiphery, presence of retinal haemorrhages and oedema

Retinal photography, fluorescein angiography

Indirect opthalmoscopy

Direct ophthalmoscopy following instillation of a mydriatic

Serial fundoscope examinations: presence and type of haemorrhages, exudates and arterial or venous occulsions were noted

Retinal examination

Retinal haemorrhages; cotton wool spots, intraretinal oedema, narrowed and obstructed arteries, venous distension and tortuosity, papilloedema

Retinal haemorrhages; cotton wool spots, intraretinal oedema, narrowed and obstructed arteries, venous distension and tortuosity, papilloedema

Haemorrhages, cotton wool spots, capillary non-perfusion and/or extravasation of fluorescein

Haemorrhages

Haemorrhages, retinal oedema, bilateral papilloedema

Retinal haemorrhages with and without exudates, none showed visible vessel occlusion

Retinal changes

Macular oedema, extramacular oedema, vessel abnormalities and poor outcome Retinal haemorrhage not associated with outcome

Extramacular oedema, blood glucose levels and poor outcome Papilloedema, opening pressure and poor outcome

Markers of disease severity and a generalized increase in systemic capillary permeability Superficial retinal infarcts and severely ill patients Retinal haemorrhages and poor prognosis Fluorescein leakage with coma

Haemorrhages not predictive of poor outcome

Retinal haemorrhages with high parasitaemia with schizontaemia, anaemia, elevated creatinine, reduced antithrombin III, severe manifestations of P. falciparum infection

Retinopathy and severe anaemia Retinal haemorrhages were more frequent in patients with cerebral malaria

Associations between recovery status, non-ophthalmologic predictors and retinal changes

Cotton wool spots, white or gray soft-edged opacities in the retina composed of cytoid bodies; fundus, the back portion of the interior of the eyeball, as seen by means of an ophthalmoscope; papilloedema, (choked disc), oedema of the optic disc (papilla); macula, an irregular yellowish depression on the retina, approximately 3° wide, lateral to the optic disc, the site of absorption of short wavelengths of light; poor outcome, death or recovery with neurological sequelae.

Type of malaria (country)

Summary of studies on retinal pathology and human cerebral malaria

Number of patients (age) and reference

Table 2

CNS and the pathogenesis of cerebral malaria 103

104

IM Medana et al.

Cerebral malaria as an immunopathological manifestation This hypothesis suggests that CM is the result of an overvigorous immune response (reviewed in Clark and Rockett30) originally evolved for the protection of the host, producing endothelial cell damage and alterations in blood–brain barrier (BBB) permeability5 or other manifestations that affect central nervous system (CNS) function, such as nitric oxide production.30 Clinical studies consistent with an immunopathogical aetiology of CM include the observation that CM rarely occurs in children who are protein malnourished,32 in whom there is depressed cell-mediated immunity. Histologically, in post-mortem brain sections from some patients with CM, deposition of parasite antigen, IgG, and fibrin in cerebral vessels have been demonstrated.16,33 These historical findings in human CM have been further substantiated more recently as detailed later in this review. Further evidence for the pathogenesis of CM being an immunopathological process comes from studies using murine models, as discussed below. In all likelihood, both the presence of parasitized erythrocytes in the CNS microcirculation and immunopathological reactions are needed for the full manifestations of cerebral malaria. The presence of the parasitized red blood cells might serve to ‘focus’ the reaction in the brain.34

Murine models of cerebral malaria Fatal murine cerebral malaria Clearly there are major ethical and practical barriers to the testing of hypotheses about the pathogenesis of human CM. In particular, it is impractical to evaluate the early stages of the condition, prior to its clinical manifestation. Murine models of CM have provided an alternative approach to investigation of post-mortem brain tissue of CM victims because they reflect the human condition in many respects (23,35,36; Table 1), are reproducible and allow experimental manipulation and investigation of the important events occurring during the development of CM. The most widely used system is a ‘fatal’ model of CM (FMCM).35 CBA mice infected with Plasmodium berghei ANKA strain die 7 days after parasite inoculation (p.i.), exhibiting severe cerebral symptoms, including hemiplegia, convulsions and coma. Histological features include monocytosis, oedema and petechial haemorrhages in the CNS,23,35,36 all of which have been reported in post-mortem human CM brain tissue (17,37; see Table 1).

Resolving and non-cerebral malaria models Despite the similarities of fatal murine CM to human CM there are some important differences; for example, more than 95% of the mice die, whereas less than 50% of humans with CM die from the complications. Therefore the ‘resolving’ cerebral symptoms model was developed.38 DBA/2J mice inoculated with P. berghei ANKA develop mild cerebral symptoms including disturbed gait, increased excitability and varying degrees of transient limb paralysis, around the same time the mice of the ‘fatal’ model develop more severe cerebral symptoms. However, the mice recover from these cerebral complications and die between days 15–22 p.i., with

up to 80% parasitaemia and consequent severe haemolysis and anaemia.38 In this ‘resolving’ model, the mice do not normally become comatose. The fatal and resolving cerebral symptoms models can be complemented with murine models that do not result in neurological complications: CBA or DBA mice inoculated with P. berghei K173 exhibit none of the cerebral symptoms seen in the fatal or resolving models of CM, but instead die with severe haemolysis between days 15–22 after parasite inoculation.38 By comparing the fatal CM and resolving cerebral symptoms models with the non-CM models, the events that lead to neurological complications can be distinguished from those that are common to malaria infection.

The immune response to malaria While considering the evidence for immunopathogenesis in CM, it is useful to examine, in parallel, the components of the immune response that are activated in response to a malaria infection. Much of the data on the anti-malarial response comes from studies that have utilized models of murine malaria in which cerebral complications are not a feature. The two most commonly used parasite strains for these experiments have been Plasmodium vinckei and Plasmodium chabaudi. P. vinckei infections cause a fulminating, fatal infection that is uniformly lethal to non-immune mice.39 However, mice can be immunized to this parasite by repeated infections followed by drug cure,40 and it is following this procedure that some immune mechanisms have been studied. In contrast to P. vinckei infection, P. chabaudi infection is naturally resolved by the immune system.41

T-cell involvement in immunity to murine and human malaria T cells appear to be central to both malaria immunity and to the manifestations of CM. Using the P. chabaudi and P. vinckei murine models it has been shown that the primary immune response to asexual blood stage parasites is mediated by CD4+ T cells, which are required for cell mediated immunity, but not for help in antibody production.39,40,42,43 Support for this hypothesis comes from the following observations in P. vinckei-infected animals: (i) mice that do not make antibody (µ-suppressed) develop immunity with the same kinetics as normal mice;44 (ii) in vivo depletion of CD4+ T cells in B-cell deficient mice abrogates their immunity; and (iii) sera from immune mice transferred to athymic nude mice had no effect on the course of the parasitaemia (reviewed in Kumar et al.39). Similarly, CD8+ T cells do not appear essential for the primary immune response against the erythrocytic stage of the parasite.39,45,46 Studies of the role of T cells in human immunity to malaria parasites are very complex and have often resulted in conflicting evidence. It has been reported that acute infection with P. falciparum induces a temporary state of reduced immunocompetence and increased susceptibility to concomitant infections. This is accompanied by changes in many parameters of immune function, including decreased numbers of circulating T cells47,48 and depression of in vitro peripheral blood

CNS and the pathogenesis of cerebral malaria

mononuclear cell proliferation in response to malaria antigens.49,50 Explanations for these findings include: (i) migration of activated T cells from the peripheral blood to the spleen and liver; (ii) activation of CD8+ suppressor cells; (iii) the generalized physiological effects of febrile disease; (iv) a defect in the production of IL-2;8,49 (v) the effects of nitric oxide on proliferation of mononuclear cells.51 The ability of erythrocytes containing P. falciparum parasites to inhibit the activity of antigen-presenting dendritic cells52 is also a probable factor in immunosuppression. At the same time, however, there is evidence of marked T-cell activation in human CM.8 Riley and colleagues53 examined, in a malaria-endemic population with differing levels of clinical immunity to malaria, the relationship between: (i) soluble IL-2 receptors and lymphoproliferative responses to malaria antigens; and (ii) plasma IL-2 receptor levels, age, malaria parasitaemia and clinical symptoms. They found high levels of soluble IL-2 receptor in the plasma of malariainfected individuals and this was independently associated both with age and parasitaemia. As the plasma concentration of soluble IL-2 receptor is an indication of T-cell activation in vivo, this group concluded that a vigorous cellular immune response to malaria antigen occurs in vivo. To further complicate the story, a most unexpected finding is that some humans who have never been exposed to malaria have CD4+ T cells with specificity for circumsporozoite protein (CSP), or peptides from CSP, as well as many other malaria proteins.54–57 However, as discussed by Good and colleagues,56 this does not disprove the importance of CD4+ T cells in malaria immunity, in pre-exposed adult humans. It may be that other factors acting together with these T cells, but not present in non-exposed individuals, may be critical for full expression of cellular immunity to malaria.56 One such factor could be an intact spleen that has been modified architecturally by a previous malaria infection.39,40,58 It has also been hypothesized that malaria-specific T cells in non-exposed individuals could partially contribute to the varying degrees of innate resistance of adult humans to malaria parasites.58

The role of T cells in the pathology of murine and human cerebral malaria T cells have also been shown to contribute to the neurological complications associated with murine CM. P. bergheiinfected athymic nude mice do not succumb to CM, unlike heterozygous mice that are T-cell competent.59 Subsequently, Grau and colleagues60 demonstrated that it was the CD4+ T-cell subpopulation that was involved in the pathogenesis. Treating malaria-infected mice with an anti-CD4+ antibody abolished the occurrence of CM. Supplementation of thymectomised, irradiated and bone marrow-reconstituted mice with normal CD4+ T cells re-established susceptibility to the neurological symptoms. Experiments using antibody against CD8+ T lymphocytes suggested that these cells were not necessary for the pathogenesis of murine CM.60 Subsequent studies using β2microglobulin−/− mice, however, revealed that CD8+ cells do play an essential role.61 This was confirmed by further work using anti-CD8+ T-cell antibody.62 The mechanism underlying this role of CD8+ cells has not yet been determined.

105

The finding that T cells from some non-exposed humans respond in vitro to malaria parasites as well as do T cells from exposed individuals has formed the basis of a hypothesis that may explain the role of activated T cells in the disease complications of human malaria infection.63 In particular, this hypothesis could be applied to the development of cerebral complications, although some other factors would have to be operating in parallel to explain why only a small subset of people are affected by CM. Evidence suggests that malariaspecific T cells from non-exposed individuals have arisen as a result of cross-reactive stimulation by other organisms. T-cell clones expanded in vivo by cross-reactive organisms will: (i) ‘home’ to the tissues where the antigen was first encountered; (ii) be ineffective parasite killers; and (iii) contribute to pathology as a result of cytokine production in various tissue locations.8 In contrast, T-cell clones expanded in vivo by malaria parasites will: (i) not have a preferential tissue location; and (ii) more frequently recirculate through the spleen, where their anti-parasite effect is maximal.63

Monocyte and macrophage involvement in immunity to murine and human malaria Hepatomegaly and splenomegaly are characteristic features of murine and human malaria infection and have been attributed not only to erythrophagocytosis, but also to increased numbers of macrophages recruited from the circulation.64 The role of spleen macrophages and liver Kupffer cells in the host response to malaria is well established experimentally. Macrophages in the spleen and liver have been shown to eliminate parasites from mice infected with plasmodia.39,65,66 Quantitative differences in macrophages in the marginal sinuses in the spleen are thought to explain differences in the level of resistance to P. chabaudi infection. C57/B16 mice, which are resistant to infection, have more macrophages in the marginal sinuses, perhaps thereby enhancing initial confrontation with the malaria parasite in the circulation.39 It is believed that circulating pRBC become coated with antibody against parasite-derived proteins, facilitating phagocytosis. However, Ferrante and colleagues67 have shown that human macrophages effectively engulf and kill pRBC in vitro, even in the absence of serum. It was also noted that monocytes and fully differentiated macrophages were equally effective. Therefore, a more important characteristic of malaria-infected cells that may make them more ‘palatable’ to phagocytes is the marked increase in phosphatidylethanolamine and phosphatidylcholine in the outer monolayer of the RBC membrane.68,69 This disruption of the normal distribution of phospholipids across the RBC membrane also aids clearance of affected erythrocytes in other disease states, and of old RBC.68

Monocyte involvement in the pathogenesis of murine and human cerebral malaria High levels of toxic products derived from monocyte/ macrophages have been found in the serum of both CMresistant and -susceptible mice during malaria infection (further discussed below). As significant numbers of monocytes have been found lining the walls of the blood vessels, it is conceivable that they could produce factors that result in

106

IM Medana et al.

vascular damage. Electron and light micrographs showing monocytes adjacent to endothelial cell damage23,70 support this interpretation. Previously23,38,70 it has been noted that monocytes bind and phagocytose Monastral blue particles, possibly indicating ‘activation’ of these cells. Work by Patnaik and colleagues71 has shown that in some human CM cases up to two-thirds of the cerebral vascular obstructions can be associated with mononuclear cells. These results go against the current dogma that CM is solely the consequence of plugging of cerebral vessels with pRBC. This finding of monocytes plugging brain venules in human CM has also been reported by other groups.72 It is possible that the association of macrophages with cerebral vascular endothelium may lead, through production of toxic molecules, to disruption, the consequences of which are discussed in the following sections. The potential of CD8+ T lymphocytes to damage the endothelium, given the findings in murine CM referred to earlier, should also not be forgotten.

Involvement of the central nervous system in the induction of pathology during cerebral malaria Despite our knowledge concerning the events occurring in the systemic circulation it is still hard to conceptualize how malaria infection leads to damage within the CNS. One version of the immunopathogenesis hypothesis suggests that toxic mediators, able to destroy the intraerythrocytic form of the parasite, cause non-specific tissue damage due to their untargeted mechanism of action, which is directed towards the cerebral vascular endothelium73 (Fig. 1). Ultimately, this hypothesis argues that the CNS vascular endothelium is an ‘innocent bystander’, which once damaged by non-specific mechanisms, results in cerebral oedema and haemorrhage, progressing to coma and possibly death.38 Indeed, there is evidence to suggest that cerebral microvascular injury is a central feature of fatal murine CM and human CM.20,23,36,38 A complementary hypothesis investigated in our laboratory suggests that there is active participation of CNS-specific parenchymal cells, in particular astrocytes and microglia, resulting in the development of the cerebral complications associated with malaria infection. Increased permeability of the BBB could allow malaria exoantigens, immune cells, cytokines and other proteins access to the brain parenchyma, possibly altering the immune and supportive functions of astrocytes and microglia. This might lead to local production of toxins or a decrease in their clearance, resulting in CNS dysfunction (Fig. 1).

The central nervous system: A major site of complications during malaria infection The parenchyma of the brain and retina are composed of neurons and glia. Neurons, of course, are the basic communicating units of the nervous system. Glia, such as astrocytes and microglia, have active roles anatomically, physiologically and biochemically in brain function. Furthermore, glial activation is a very common feature of many forms of brain pathology.74 Activated glia have the potential to induce toxic pathways that may cause brain pathology but, equally, it should not be forgotten that impaired glial responses can also exacerbate CNS disease.75,76

Astrocytes regulate the central nervous system milieu and have potential immune effector functions An important role of astrocytes in the normal CNS is to induce and maintain BBB properties in the vascular endothelium. The endothelial cells themselves have complete belts of tight junctions between each cell that are uninterrupted by gap junctions.77 These features of the endothelium restrain the rate of exchange of solutes between the blood and nervous tissue by reducing the effects of fluctuations in the blood plasma metabolites and other constituents, helping to maintain the unique CNS milieu conducive to neuronal functioning.77 Astrocytes also make close contact with neuronal synapses and are thought to be intimately involved in maintaining acid-base, electrolyte and neurotransmitter balance.78,79 In addition, astrocytes regulate the concentration of neurotransmitters, such as glutamate, in the extracellular fluid. Any alteration in these astrocyte functions has profound effects on normal neuronal function. Within the injured or infected brain many inflammatory mediators and their receptors are thought to be produced or expressed by astrocytes. Primary rodent astrocytes can be stimulated in vitro by a variety of agents such as viruses, IL-1, TNF-α, IFN-γ, LPS and calcium ionophore to produce IL-1, IL-380, IL-6,81,82 TNF-α,83,84 IFN-α and IFN-β.83 The few in vivo studies of cytokine production by astrocytes support a role for them in immune-mediated neurological disease. It has been demonstrated that the capacity of astrocytes to produce TNF-α85 correlates with susceptibility to experimental allergic encephalitis (EAE). Furthermore, Wahl and colleagues86 have shown that astrocytes in HIV-1 infected brains express TGF-β. This group speculated that the production of this cytokine enhances the recruitment of HIVinfected monocytic cells, thereby increasing the spread of virus into the CNS and contributing to the inflammatory changes associated with HIV-1 encephalomyelitis. Astrocytes may also contribute to an inflammatory state within the CNS through the production of chemokines such as RANTES and MCP-1,87 which are thought to be vital for leucocyte migration and activation. Normal astrocytes are also positive for TNF receptor 1 (TNFR1), TNF receptor 2 (TNFR2), IL-1 receptor 1, IFN α/β receptor, IFN-γ receptor, and macrophage colony-stimulating factor (M-CSF) receptor.88 TNF-α action through cell surface receptors induces astrocytes to produce colony stimulating factors M-CSF, granulocyte colony-stimulating factor (G-CSF) and GM-CSF, IL-6 and TNF-α itself.89 Thus, TNF-α and IL-1 are thought to orientate astrocyte function in a pro-inflammatory direction by enhancing their production of cytokines and nitric oxide, but at the same time may promote cell survival through their production of neuronal growth factors.90 However, in vivo, astrocytes are thought to act chiefly to limit the immune response within the CNS and initiate repair processes.91

Microglia are the central nervous system counterpart of the mononuclear phagocyte system Microglial cells comprise 5–20% of the total glial population in the brain92,93 and are pluripotent members of the monocyte/ macrophage lineage (see reviews by Davis et al.94 and Streit et al.95). Microglia from the normal adult CNS are unusual

CNS and the pathogenesis of cerebral malaria

107

Figure 1 Micrographs showing astrocyte and microglial changes in fatal murine cerebral malaria (FMCM) in control (D0), day 3 (D3), D5 and D7 retinae and the presence of MHC class II-positive monocytes, perivascular cells and parenchymal microglia at the terminal stage (D7) of FMCM. (a, d, g, j) Astrocytes labelled with an antibody against the glial fibrillary acidic protein (GFAP). Scale bar = 50 µm. (a) Astrocytes in the uninfected mouse retina were evenly distributed, with a predominantly stellate morphology and sent numerous processes to contact both arteries and veins. (d) Loss of even astrocyte distribution was evident at day 3 p.i. (g) Dense patches of astrocytes in which individual processes were indistinguishable, suggestive of gliosis, at day 5 p.i. (j) Decreased astrocyte ensheathment of vessels was co-localized with vessels containing adherent Monastral blue-containing monocytes, visible as dark particules within the lumen of the vessel. (b, e, h, k) Microglia labelled with the Griffonia simplicifolia (GS) lectin. Scale bar = 10 µm. (b) In the uninfected mouse retina microglia displayed long, fine processes typical of resting microglia. (e) At day 3 p.i., increased numbers of microglia were labelled with the GS lectin and distensions on their processes became more evident, predominantly when in close association with the vessels (arrowheads). (h) Day 5 p.i., showing microglia with shorter and thicker processes. (k) High magnification image of a ramified microglia in close association with a retinal vessel displaying an enlarged soma and prominent distensions along the processes at the terminal stage of the disease. Microglia displayed a range of morphologies at the onset of cerebral symptoms including amoeboid microglia with vacuoles (see Fig. 2c). (c, f, i, l) MHC class II expressing cells at the terminal stage of FMCM. Scale bar = 50 µm. (c) Intravascular and marginating monocytes containing phagocytosed Monastral blue (arrow) and expressing MHC class II on the plasma membrane. The curved arrow points to an extravasated Monastral blue positive monocyte that is expressing MHC class II. (f) MHC class II positive perivascular cells. (i) Shows a small cluster of MHC class II resident microglia adjacent to a retinal vessel. (l) Focal regions of MHC class II expression on resident microglia found at the edge of the retina.

108

IM Medana et al.

tissue macrophages and are aptly named ‘resting’ microglia because they appear to have downregulated macrophage properties.93,96 Ramified microglia have further properties distinct from other macrophages in that they have high proliferative potential and neurones can regulate their activation.97,98 The function of the ramified microglia in the normal CNS is still unknown, but roles in CNS homeostasis, including extracellular fluid cleansing, neurotransmitter inactivation and neurotrophic mechanisms, and in molecular signalling to macroglial cells have all been suggested. Just like other tissue macrophages, microglia, when stimulated by substances such as LPS and IFN-γ,99,100 can secrete IL-1,100,101 IL-6 and TNF-α.81,84,102 IL-1 released by microglia has been shown to be an astrocyte mitogen.100,103 TNF-α has been shown in some studies to have a cytotoxic effect on oligodendrocytes and to destroy myelin structure,84,99 and is induced on microglia in demyelinating lesions. In addition, both IL-1 and TNF-α have been found on activated microglia surrounding senile plaques in Alzheimer’s disease.104,105 In vivo and in vitro evidence also implicates microglia as the major antigen-presenting cell of the CNS. Microglia have been shown to express major histocompatibility class I and II molecules as well as costimulatory molecules such as B7, ICAM-I and the αXβ2 integrin, which may activate T cells in the presence of specific antigen.74 Activated microglia may enhance monocyte migration to the CNS through the production of chemokines, a mechanism thought to be operating in HIV-1 encephalitis.87 Major histocompatibility complex class II molecules can be rapidly induced by cytokines such as IFN-γ.106,107 Furthermore, activated microglia can act as efficient antigen-presenting cells to CD4+ T cells in vitro.99,108,109 Immunohistochemical identification of MHC class II expression in experimental models and human postmortem brain is a common finding. During T cell-mediated auto-immune diseases of the nervous system, microglia rapidly express MHC class II antigens.110 Major histocompatibility complex class II reactive microglia have also been found in affected areas in post-mortem brains of patients with senile dementia of the Alzheimer’s type,111–114 Parkinson’s disease, Huntington’s chorea and multiple sclerosis.115,116 However, it should not be automatically assumed that microglial expression of MHC class II antigens equates with antigen presentation and T-cell activation. Matsumoto and colleagues117 found in vitro that microglia at low cell concentrations effectively present myelin basic protein (MBP) to MBP-specific T cells. Conversely, at very high cell concentrations T-cell suppression occurred, perhaps in part attributable to the release of inhibitory cytokines by microglia. It also has been highlighted by some authors118 that the CNS is not a particularly conducive environment for T-cell activation. In vivo observations of constitutive expression of MHC class II on microglia in animals resistant to EAE suggest that MHC expresssion on an appropriate cell within the CNS may not be necessarily stimulatory for T cells. A mechanism by which MHC class II expression could downregulate disease has been proposed by McCombe and colleagues.119 Nonspecialized CNS antigen-presenting cells that do not produce costimulatory signals (i.e. B7, ICAM-1, LFA-1, LFA-3 molecules) that are necessary for T-cell activation may lead to activation-induced apoptosis of encephalitogenic T cells. Indeed, most T cells lose their proliferative capacity soon

after infiltration into the CNS during EAE.120 As a consequence approximately half the T cells in acute EAE lesions show signs of apoptosis at the recovery stage of the disease.120,121 From our studies in FMCM mice with fatal cerebral symptoms, intravascular monocytes (Fig. 1c, arrow), perivascular cells (Fig. 1f) and parenchymal microglia (Fig. 1i) were found to express MHC class II. In addition, small numbers of MHC class II positive microglia were found at the edge of the retina (Fig. 1l). Major histocompatibility complex class II expression was focal in its distribution, similar to our earlier observations on BBB breakdown23 and loss of astrocyte ensheathment of vessels.122 Major histocompatibility complex class II expression was examined using the monoclonal M5 antibody (American Type culture collection, Manassas, VA, USA). Retinae were immersion fixed for 30 min in 70%(v/v) ethanol (in 0.1 mol/L phosphate buffer, pH 7.4), then washed three times in PBS containing 0.1% (v/v) Triton X-100 for 30 min. The retinae were then incubated overnight with anti-MHC class II, washed, then incubated for 2 h with biotinylated anti-rat Ig (1:200; Amersham International, Little Chalfont, UK). The tissue was washed for 30 min, then incubated in StreptavidinFITC (1:100, diluted in PBS, Sigma E-2886) for 1.5 h at room temperature. Major histocompatibility complex class II expressing cells concentrated at the BBB could modulate the immune response by presenting antigen to CD4+ T cells. Antigen presentation occurring locally, resulting in T-cell activation, could result in a sustained or amplified immune response that could contribute to cerebral complications. In human CM there has been a report of cuffs of lymphocytes and microglia around veins.123 Activated microglia, however, are not always perceived to be destructive agents in the induction of brain pathology. Work by Bruce and colleagues,75 in particular, has shown that reduction of microglial activation in mice lacking TNF receptors coincides with greater neuronal damage following ischaemic injury. Furthermore, the implantation of activated microglia into injured spinal cord enhances neurite outgrowth, demonstrating their important role in trophic support.

Histopathological studies describing glial changes in human cerebral malaria Most of the histopathological studies that describe glial changes in human CM post-mortem tissue were performed at a time when glial cells were still only considered as the ‘glue’ that held neurons together. Therefore the possible role of glial cells in the induction and/or perpetuation of CM is unlikely even to have been considered. Today, despite the plethora of studies investigating the immunocompetence of glial cells, there is a lack of studies that have used markers specific for these cells in human CM. Most of the descriptions that comment on glial cell changes have only been from changes in glial cell nuclei on haematoxylin and eosin sections. A number of these histological descriptions of human CM have shown progressive proliferation of glial cells and mesenchymal adventitious tissue in the absence of major parenchymal changes.124 Another study reported that, near necrotic foci, glial cells proliferate and develop around

CNS and the pathogenesis of cerebral malaria

thrombosed capillaries and may form malaria granulomas, especially in subcortical regions, corpus callosum, cerebellum, brain stem, and in the spinal cord.2 Janota and Doshi123 described microglial changes in human CM using the neutral lipid stain Sudan III and IV and silver impregnation. They found evidence for the involvement of microglia at the terminal stage of the disease. Microglia were found to be concentrated around capillaries in the grey matter, microglia containing neutral lipid and iron pigment were found in large perivascular lesions in the white matter, and microglia were associated with perivenular haemorrhages. More recently, perivascular macrophage activation (assessed by the expression of the macrophage scavenger receptor and sialoadhesin) was investigated in brain tissue from adult Vietnamese P. falciparum malaria cases. These molecules were expressed at very low levels in control brain tissue, but were greatly upregulated in non-cerebral as well as cerebral malaria cases.125 In another recent study, parenchymal microglial activation, assessed by expression of two peptides of the S-100 superfamily produced by activated monocytes (MRP8 and MRP14), was shown to be widespread throughout the brain in seven CM patients who had contracted malaria while travelling in Africa.126 In both these studies, activated microglia and perivascular macrophages were found in white and grey matter.

Histopathological studies describing glial changes in murine cerebral malaria Morphological and distributional changes in astrocytes and microglia were investigated during fatal murine CM using the retinal whole-mount technique.122,127 Astrocytes were visualized using antibodies against the glial fibrillary acidic protein, the major intermediate filament specific for astrocytes. Microglia and the vasculature were visualized using the Griffonia simplicifolia lectin, which labels α-D-galactose residues, or reacted for the nucleoside diphosphatase (NDPase) enzyme.

Astrocytes In the fatal murine CM model, astrocytes lost their even distribution from day 3 p.i. (compare Fig. 1a with 1d), progressing to gliosis (day 5 p.i.; Fig. 1g). At the terminal stage of the disease (day 7 p.i.) there was a loss of astrocyte processes contacting retinal vessels (Fig. 1j; Fig. 2a,b), often along segments containing adherent monocytes (see arrowheads). The loss of astrocytes could have major consequences in the CNS by contributing to the massive focal BBB breakdown and disruption of normal neuronal function at the terminal stage of the disease. To confirm that astrogliosis and degeneration were associated with the development of cerebral complications, astrocytes were visualized during the progression of disease in the ‘resolving’ cerebral symptoms and the non-CM murine models of malaria. In the resolving model, astrocytes showed small regions of unevenly distributed astrocytes from day 3 p.i. At day 7 p.i., there was a decrease in astrocyte ensheathment of vessels, coincident with margination of monocytes containing Monastral blue (Fig. 2d,e). However, these features were mild and infrequent compared to those observed in the fatal CM model. The

109

astrocyte response peaked at day 7 p.i., the time at which the mice showed mild cerebral symptoms. Although loss of astrocyte distribution and some loss of astrocyte degeneration were still evident until death (days 15–22 p.i.), these features were not as striking as those found at day 7 p.i. No major morphological changes in astrocytes were seen in noncerebral malaria mice (compare Fig. 2a,b with 2g,h) until the mice were about to die (days 15–22 p.i.). At the time of death a reduced wrapping of vascular segments and some loss of regularity in distribution were evident in 36% of mice. These changes were minor compared to those found in the fatal and resolving models of cerebral malaria. From these results it was evident that loss of astrocyte ensheathment of vessels was associated with the big increase in BBB permeability that occurs at the time of neurological complications in the murine models of malaria. This loss of astrocyte processes was associated with the adherence of monocytes to the vascular endothelium, suggesting that toxic products produced by monocytes might play a role in astrocyte degeneration.

Microglia Changes in microglial morphology and redistribution towards the venous side of the circulation were seen within 2–3 days p.i. (Fig. 1e), that is, at least 3 days before the onset of cerebral symptoms and 4 days before death in the fatal murine model of cerebral malaria. Morphological changes included retraction of ramified processes, soma enlargement, an increasingly amoeboid appearance and vacuolation (Fig. 1e,h,k and 2c). These features were maximal when the mice displayed fatal cerebral complications on day 7 p.i. (Figs 1k and 2c). Another interesting finding was the increased α-D-galactose residues on microglia in focal regions during the progression of FMCM. This observation may have clinical significance because antibodies against αD-galactose residues have been found in the CSF of patients suffering from human cerebral malaria.128 Ligation of these residues on activated microglia might alter their biological function, leading to neurocomplications within the CNS. The acquisition of new membrane oligosaccharide epitopes has previously been shown to permit cells to respond to new ligands with selective and coordinated changes in their secretory profile. Indeed, ligation of α-D-galactose residues on macrophages in vitro leads to TNF-α production.129,130 Furthermore, TNF-α production by microglia is a feature of FMCM (see later sections). We confirmed that the microglial responses were specific to the FMCM by comparing them with those in the resolving and non-CM models. In resolving CM, progressive morphological changes in microglia were observed during the course of the disease, but the distribution and frequency of these changes were not as striking as those found in the fatal model. From day 5 p.i. microglia showed an increase in staining intensity and displayed thicker and shorter processes. These morphological changes were predominantly found in close association with the retinal vessels. The morphological changes peaked at day 7 p.i., the time at which these mice displayed transient cerebral complications (Fig. 2f). These changes included the development of amoeboid microglia with vacuoles within their processes. Like those changes

110

IM Medana et al.

Figure 2 Micrographs displaying the degree of astrocyte degeneration and microglial activation at day 7 p.i. in the fatal cerebral malaria model (a, b, c), the resolving cerebral symptoms model (d, e, f), and the non-cerebral malaria model (g, h, i). Day 7 p.i. is a time at which fatal CM mice fit, become hemiplegic, comatosed and eventually die; resolving cerebral symptom mice show varying degrees of behavioural changes, but do not become comatosed and recover; and non-CM mice show no cerebral complications. (a, d, g) Astrocytes labelled with an antibody against GFAP. (b, e, h) Same field of view as (a, d, g), but using transmitted light illumination to visualize monocytes containing phagocytosed Monastral blue. Monastral blue had been injected i.v. 2 h prior to sacrifice of the mice. (c, f, i) Microglia labelled with the Griffonia simplicifolia (GS) lectin. (a) In the fatal cerebral malaria model there was a striking loss of astrocyte processes on the retinal vessels and in adjacent regions of the retinal parenchyma. (b) This loss of astrocytes was associated with monocyte margination of retinal veins (arrowheads). (d) Although there was a reduced ensheathment of vessels with astrocyte processes at day 7 p.i. in the resolving cerebral symptoms model, these changes were mild (compare a, d). (e) Similarly there was a marked reduction in the number of monocytes marginating within retinal vessels (arrowheads). (g) There was no apparent loss of astrocyte processes contacting retinal vessels at day 7 p.i. in the non-cerebral malaria model. (h) No monocytes containing Monastral blue are apparent in this micrograph although, occasionally, they could be found in the retinal vessels. (c) Several different morphological types of GS positive cells were found in the retina at the terminal stage of FMCM, because increased expression of GS lectin binding sites is indicative of activation of resident microglia and is also associated with monocytes. These included an intensely staining, ramified microglia with prominent distensions (see Fig. 1k); a population with stout processes, large soma and extensive vacuolization (arrow) and a small population of round cells (arrowhead) that are likely to be extravasated monocytes. (f) In the resolving model, microglia showed an increase in staining intensity and thicker and shorter processes. These morphological changes were predominantly found on microglia in close association with the retinal vessels. (i) There were no major morphological changes in microglia in the non-cerebral malaria model at day 7 p.i. An increase in tortuosity of microglial processes associated with retinal vessels was variable between individual mice and was mild and infrequent compared with the other malaria models with cerebral complications.

found at day 5 p.i., these changes were only evident on a small subpopulation of microglia, predominantly in close association with the retinal vessels. Intensely stained microglia with short processes in close association with the retinal vessels and ramified parenchymal microglia with prominant soma were still evident until death (day 15–22 p.i.). However, these features were not as striking as those found at day 7 p.i. There were no major changes in microglia observed in the non-CM model until day 10 p.i., but once the changes

appeared they persisted until death (days 15–22 p.i.). These changes were mild and infrequent compared with those found in the FMCM model (Fig. 2i). Although there were no morphological changes in microglia until day 10 p.i., it was evident that the tortuosity of microglial processes associated with the retinal vessels had increased from day 5 p.i. At day 10 p.i., ramified microglia displayed shorter processes in the non-CM models compared with ramified microglia from uninfected mice. These shorter microglial processes rarely

CNS and the pathogenesis of cerebral malaria

became thickened, never contained vacuoles and persisted until the terminal stage of the disease. Thus, it can be concluded that the changes in microglial morphology accurately reflected the extent of neurological complications in murine malaria infection and suggest a role for these cells in the course of the neurological disease.

Experimental manipulations Further studies using intracarotid injection of arabinose to open the BBB127,131 showed that the early glial changes could be replicated by an increase in BBB permeability alone. However, this was not the case for glial changes occurring later in the course of the disease. Furthermore, glial changes could be ameliorated following treatment with the immunosuppressive agent, dexamethasone, if given early in the course of the disease.127,131 On the basis of these observations it was concluded that: (i) astrocytes and microglia are involved in the pathogenesis of FMCM because their morphological and distributional changes occur substantially before the onset of cerebral symptoms and the degree of morphological changes reflect the extent of neurological complications; (ii) the initial changes in astrocyte and microglial distribution may be a consequence of the increase in blood–retina barrier permeability; and (iii) the immune response triggered by specific malaria parasites (PbA) may be responsible for the loss of astrocyte ensheathment of vessel segments. Furthermore, it was hypothesized122,127 that functional changes in glia, such as the release of cytokines, probably induced by an increase in BBB permeability or circulating products of the malaria parasite, may interfere with CNS functions, leading to cerebral symptoms and death. To examine this further, functional involvement of astrocytes and microglia during FMCM was investigated by assessing: the production, the temporal appearance, and the cellular source of cytokine mRNA and protein in the brain (the results are presented below in the context of previous findings).

Cytokines in the host anti-malarial response and the pathogenesis of cerebral malaria Tumour necrosis factor-α Clark and colleagues predicted that pathology observed in malaria could be caused by excess production of cytokines such as TNF-α (reviewed in Clark et al.132). Parasite exoantigen(s) derived from malaria parasites that infect rodents and humans can stimulate the release of TNF-α from monocytes in vitro.133,134 Similarly, injection of exoantigens into mice induces TNF-α secretion in vivo.134 This TNF-α secretion is thought to control parasitaemia in vivo. TNF-α has been shown to inhibit parasite survival in some rodent malaria models66,135 and to enhance human neutrophil killing of P. falciparum.136,137 Taverne and colleagues66 demonstrated the anti-parasitic effect of TNF-α using transgenic mice carrying a human TNF/β globulin gene construct. Plasmodium yoelii-infected mice displayed a 10-fold reduction in parasitaemia compared with controls and inhibition was even more striking in mice infected with P. chabaudi. This result was thought to be due to augmentation of monocyte/

111

macrophage phagocytosis or killing by toxic products released by activated macrophages, such as reactive oxygen species138 and reactive nitrogen intermediates.139 Perhaps surprisingly, treatment of P. berghei ANKA-infected mice with an anti-TNF-α antibody does not affect parasitaemia.140 Experiments by Grau and colleagues140,141 led to the hypothesis that excessive TNF-α production plays a critical role in the pathogenesis of murine CM. This is supported by the following findings: (i) elevated serum TNF-α is only found at the time of the neurological syndrome; (ii) a single injection of anti-TNF-α antibody ameliorates the cerebral complications; and (iii) administration of recombinant TNF-α to a CM-resistant strain of mouse induces neurological features similar to CM. Furthermore, de Kossodo and Grau142 also found an upregulation of TNF-α mRNA in the brains of CM-susceptible mice and decreased levels of IL-4 (which antagonizes the effects of TNF-α). However, Eling and co-workers143 did not find that a neutralizing antibody against TNF-α was protective in their murine CM model. Several anti-cytokine antibodies can prevent CM in the murine models. For example, administration of anti-IFN-γ or a combination of anti-IL-3 and anti-GM-CSF ameliorate FMCM. Taken together, these observations led Grau and colleagues144 to hypothesize that, in FMCM, IFN-γ activates macrophages and IL-3 and GM-CSF increase the production of monocytes from haemopoietic precursors, culminating in this excessive TNF-α production. This conclusion may be relevant to the human condition because TNF-α has been implicated as a cause of at least some of the pathology associated with human CM,145 with plasma levels rising in severe cases.141,146–148 Kwiatkowski and colleagues148 undertook a study involving 178 Gambian children. They found that plasma TNF-α levels in children that had survived CM were twice as high as normal levels. In the fatal cases they were 10-fold higher than normal, leading to the conclusion that excessive TNF-α production may predispose humans to CM and its fatal outcome. Paradoxically, however, a trial of an anti-TNF antibody, comprising more than 600 Gambian children with cerebral malaria, was found to be associated with a significant increase in neurological sequelae. One explanation put forth for this finding was that the antibody may act to retain TNF within the circulation and thereby prolong its effect.149 A clinical trial with an anti-TNF polyclonal Fab fragment has been piloted in Thailand.150 Another unexplained finding is that circulating TNF-α levels can be higher in humans infected with Plasmodium vivax, which does not cause cerebral malaria, than in patients with P. falciparum infection with cerebral complications.151 This can also be observed in murine models, where higher plasma levels of TNF-α are found in models of malaria that do not develop cerebral symptoms, and this finding has led to the suggestion that TNF-α produced locally within the CNS may contribute more to the sequelae than that found in the plasma.152,153 Experiments by Clark and colleagues152 involving administration of recombinant human TNF-α into a CMsusceptible mouse strain resulted in a syndrome like that in P. vinckei, another rodent model of malaria that does not produce cerebral manifestations. However, increasing TNF-α production through endotoxin treatment hastens the onset of cerebral symptoms in mice infected with P. berghei ANKA.154

112

IM Medana et al.

More recent work by Grau and colleagues (reviewed in Lucas et al.155) has investigated the contribution of the two TNF receptors (TNFR1 and TNFR2) to the development of the fatal complications associated with FMCM. This group found that TNFR2 was critically involved in the cerebral complications because (i) there is upregulation of TNFR2, but not TNFR1, on the cerebral vasculature of CMsusceptible mice, and (ii) mice genetically deficient for TNFR2 were protected from CM, while TNFR1-deficient mice were susceptible. Also, there was an association between protection against cerebral complications and the absence of TNFR2 and an upregulation of ICAM-1 (see below). As membrane-bound TNF can effectively signal through both TNFR1 and TNFR2, whereas soluble TNF can only effectively signal through TNFR1, it was further hypothesized that the membrane form of TNF, rather than the soluble form, may have a critical role in CM. Using in situ hybridization our group showed an increase in expression of TNF-α mRNA as early as day 3 p.i., which is several days before the onset of cerebral symptoms, although TNF-α protein was not detectable until at least day 5 p.i.156 At day 5 p.i., small numbers of cells with the morphological characteristics of microglia and astrocytes were positive for TNF-α protein. In addition, endothelial cells and monocytes adherent to the cerebral vasculature were also positive. At a time when the FMCM mice were showing severe cerebral symptoms (day 7 p.i.) there was a massive increase in the number of the above-mentioned cells positive for TNF-α. The early upregulation of TNF-α mRNA indicates that initiation of the host response within the CNS occurs early in the disease process. An event that occurs at approximately the same time as the appearance of the TNF-α mRNA, and just prior to the appearance of TNF-α protein, is an increase in permeability to plasma proteins of the BBB.23,70 An increase in CNS barrier permeability would allow the passage of parasite antigens into the CNS. Soluble parasite antigens are discharged from infected erythrocytes at the time of merozoite release and some, at least from P. falciparum, stimulate the release of cytokines from macrophages.134,157 Thus, it may be the movement of plasma proteins and parasite antigens into the CNS that stimulates the early upregulation of TNF-α in the brain. TNF-α mRNA and protein levels were analysed in the brains of six pediatric patients from Malawi that had died from cerebral malaria. In this study, TNF-α was detected in >50% of samples from the CM cases. TNF-α protein could be co-localized with neurones or the cerebral vasculature, but not with glial cells. However, TNF-α mRNA induction was not specific for CM and did not correlate with parasite sequestration.125 TNF-α produced by monocytes or glia could influence neurocomplications through its capacity to induce the release of other cytokines, ROS or nitric oxide (which all have been implicated in the pathogenesis of FMCM). TNF-α primes murine microglia for enhanced superoxide production and potentiates glutamate receptor-mediated neurotoxicity.158 In conjunction with IFN-γ, TNF-α potentiates murine microglial cell release of nitric oxide.159–161 However, nitric oxide162 and phagocyte-derived ROS163 do not appear to be critical in the pathogenesis of murine CM.

Interferon-γ IFN-γ plays an important role in anti-parasite immunity. This cytokine has been shown to reduce hepatocyte invasion by malaria sporozoites.164 There is also evidence for IFN-γmediated protection against both hepatic and asexual blood stages of malaria.138,165,166 Riley and colleagues157 suggested that localized induction of IFN-γ by insoluble cell-surface antigen in the liver or at sites of parasite sequestration efficiently activates parasite killing. In vitro, IFN-γ promotes killing of parasitized erythrocytes by reactive oxygen species138 and nitrogen intermediates.139 A protective effect of in vivo IFN-γ treatment, inhibiting the level of parasitaemia, has been observed in the P. chabaudi and P. vinckei murine models.135,167 P. chabaudi primary blood stage infection is resolved by activation of Thl-type CD4+ T cells, which produce IL-2 and IFN-γ. In this setting IFN-γ is thought to activate macrophages to produce large amounts of nitric oxide that can kill malaria parasites directly. Like many studies of human malaria, reports on the importance of plasma levels of IFN-γ in immunity and pathology of CM are often conflicting. In patients with P. falciparum, malaria serum IFN-γ levels do not appear to be associated with parasitaemia, disease severity or complications.168 For example, an increase of IFN-γ concentrations in the serum of patients infected with P. falciparum was reported in some studies,148,169 but not in others.146,170 However, it has been demonstrated that there are differences in the levels of IFN-γ in the serum of acutely infected individuals, depending on their immune status. Non-immune individuals have a high peak of IFN-γ in serum during the malaria infection compared to semi-immune subjects who do not have significant changes in their IFN-γ serum levels.169,171 It also has been shown that mononuclear leucocytes from non-immune subjects produced more IFN-γ when stimulated with malarial antigen compared with those from semiimmune individuals. These differences were not found when the cells were cultured with mitogen.168,172 Although IFN-γ appears to play a role in parasite immunity it could also be deleterious to the host by virtue of its pro-inflammatory actions.49 Treatment with anti-IFN-γ monoclonal antibody of mice susceptible to CM prevents cerebral complications.144 Furthermore, C57B1/6 mice infected with P. berghei ANKA die 7 days postinoculation with the parasite, showing cerebral symptoms like those described for CBA mice. In contrast, C57B1/6 mice that cannot express IFN-γ (IFN-γ ‘gene knockout’ mice) do not develop petechial haemorrhages, monocyte adherence to the vascular endothelium, or display cerebral complications.61,173,174 In addition, there is significant accumulation of IFN-γ mRNA in the brains of CM-susceptible mice compared with CM-resistant and uninfected mice.142 Overproduction of IFN-γ in the brain of P. berghei ANKA-infected mice could lead to the activation of endothelial cells, macrophages and possibly glial cells. Adhesion molecules such as ICAM-1, which have known involvement in CM (see Table 3), would also be upregulated on endothelial cells. IFN-γ may upregulate TNF receptors on cerebral vascular endothelial cells, rendering them more susceptible to TNF-α, and could also prime macrophages to release more TNF-α.175

CNS and the pathogenesis of cerebral malaria

Table 3

113

Central nervous system (CNS) changes during fatal murine cerebral malaria

Stage of disease

Features of disease stage

CNS observations

Early (days 2–3 p.i.)

No detectable parasitaemia in blood smear No monocyte adherence to the vascular endothelium 4–5 days before the onset of cerebral symptoms Cerebral complications can be ameliorated by dexamethasone treatment

↑ extra vascular albumin ↓ cerebral vascular endothelial cell viability ↑ VCAM-1 on CNS microvessels ↑ TNF-α mRNA on small numbers of cerebral vascular endothelial cells and cells within the cerebral parenchyma Morphological and distributional changes in astrocytes and microglia ↑ α-D-galactose residues on microglia

Prior to the onset of cerebral symptoms (days 4–5 p.i.)

1–2% parasitaemia in blood smear Cerebral symptoms cannot be prevented with dexamethasone treatment Treatment with endotoxin hastens the onset of cerebral symptoms and death

Terminal stage (days 6–7, p.i.)