Background
With 212 million new cases and 429,000 estimated deaths in 2015, malaria remains one of the most economically impactful infectious diseases worldwide [
1]. A small percentage of
Plasmodium falciparum infections results in severe malarial disease. However, a significant proportion of severe malaria infections includes cerebral malaria (CM), which is a leading cause of death in sub-Saharan African children and represents a major burden worldwide [
2]. CM accounts for an estimated 500,000 cases per year and correlates with high parasitemic burden, severe inflammation, and cerebral edema [
2]. Furthermore, about 20% of patients with CM die despite timely treatment [
3], and neurological sequelae in survivors is common [
4]. Several host genetic factors have been implicated in pathology. For example, mutations in the promoters of the inflammatory cytokine tumor necrosis factor (TNF), which drives the anti-malaria response of phagocytes, and the regulatory cytokine IL-10, which protects the host from excessive immunopathology, have been correlated with severe disease in both mice and humans [
5‐
10]. However, inflammatory cytokines also allow parasite sequestration and leukocyte adhesion by upregulating adhesion molecules on the vascular endothelium [
11‐
13].
The role of inflammatory cytokines increased by the absence of IL-10 has been studied extensively in the
Plasmodium chabaudi mouse model of severe malaria [
14].
P. chabaudi is a rodent parasite that leads to mild malaria in C57BL/6 (WT) mice. However, in IL-10-deficient (IL-10 KO) mice,
P. chabaudi infection leads to hyper-inflammation and death. The syndrome includes increased levels of the pro-inflammatory cytokines TNF and IFN-γ [
14] and lethal disease characterized by cerebral pathology including cerebral edema and hemorrhage [
15]. In addition, we have recently demonstrated pathological behavioral phenotypes indicative of neurological and cognitive dysfunction in this model [
16]. Strikingly, there is no significant parasite sequestration in the brains of these mice. While a few parasites have been detected in the brain vasculature via electron microscopy [
17], a more recent examination of the brain using highly sensitive luminescence technology to detect luciferin-expressing
P. chabaudi parasites did not show significant enrichment [
18]. The
P. chabaudi life cycle is synchronous. Mature schizonts disappear from the circulation almost completely and are found sequestered primarily in the liver and lungs of mice in a partially ICAM1-dependent manner [
19]. Interestingly, pathological damage within each organ in
P. chabaudi does not correspond to the degree of organ-specific sequestration of the parasite [
18]. Sequestration is a hallmark of autopsy in fatal
P. falciparum-induced CM cases [
20,
21], and specific parasite variants are associated with severe malaria [
22‐
25]; however, it is challenging to definitively prove that parasite sequestration in the brain is causal to CM.
Activated immune cells and pro-inflammatory cytokines are also strongly implicated in the mortality in human disease [
26,
27]. A low ratio of IL-10 to TNF in patients predicts more severe malaria, as do mutations in the IL-10 and TNF genes [
28,
29]. Mouse models show that this is because IL-10 is required to protect animals from lethal pathology, as it regulates the pro-inflammatory cytokines IL-12 and TNF [
30], which drive as yet poorly defined neuroimmunopathology. IL-10 KO mice lacking IFN-γ receptor signaling are also rescued from mortality, even though they exhibit higher levels of parasitemia [
31]. IL-10 is primarily made by CD4
+ IFN-γ
+ effector T cells (Teff) in
P. chabaudi infection, not Tregs, and is downstream of IL-27 [
32,
33], and we have shown that CD4 Teff are found solely within the cerebral vasculature, not in the brain parenchyma [
16].
While there are studies of host genetics and those correlating systemic inflammatory cytokines with poor outcomes in severe malaria [
26,
27], no significant inflammatory infiltrate within the brain parenchyma has been documented in human or mouse studies of the disease [
20,
21,
34‐
40]. As a result, the contribution of activated peripheral leukocytes to brain pathology has been poorly appreciated. Interestingly, despite the lack of infiltrating inflammatory cells in the brain parenchyma, we have documented increased microglial activation in this model [
16]. This is intriguing because glia are found behind the multi-layered blood-brain barrier (BBB), while activated peripheral immune cells are within the vasculature [
16]. This prompted the question of how the inflammatory cells within the vasculature could amplify cytokine production in the absence of a lymphoid structure, such as that developing in neuroimmunopathologies with parenchymal infiltrates.
Congestion of the brain and retinal vasculature has been documented in human cerebral malaria and is associated with poor prognoses in human cases of CM [
41,
42]. Several factors are likely to contribute to congestion in human patients: parasite sequestration, leukocyte adhesion, and coagulation defects. Parasite-infected erythrocytes can both bind to the vascular endothelium, leading to activation and vascular dysfunction, and activate the coagulation cascade [
43,
44]. Coagulation defects are also seen in both murine experimental cerebral malaria and in human cerebral malaria [
45‐
47] and can be promoted by the parasite itself [
45]. Vascular thrombi were observed in CM2 patients in Malawi, who are documented to have both sequestration and cerebral hemorrhages [
20]. This supports the finding that disseminated intravascular coagulation (DIC) was observed in 19% of CM patients and correlated with poor outcomes [
48]. However, the role of coagulation in neuropathology is obscured by contradictory outcomes in studies of the effect of the anticoagulant, heparin [
49,
50]. In clinical trials, heparin significantly reduced death in a clinical trial in children with CM in Indonesia (from 13/17 to 2/16, [
50]) and reduced patient’s coma and hospitalization time [
49]. However, it is not currently recommended for treatment due to the potential of systemic hemorrhagic side effects of this older drug, suggested by work in non-human primates [
51] and case studies of malarious soldiers in Asia with pulmonary involvement [
52], though not seen in clinical trials. The presence of monocytes and T cells in the brain vasculature [
20], but not in the brain parenchyma [
34], is also documented. This has often been interpreted as a “lack of inflammation,” despite strong evidence, both genetic and serological, that cytokines play a critical role in killing parasite and inducing pathology [
53].
In an attempt to understand the role of adherent intravascular leukocytes and coagulation in promoting neuronal malfunction, we investigated the contents of congested vessels and their effects on the brain parenchyma, as measured by gliosis. Furthermore, we tested the role of coagulation in pathology by studying the effect of anticoagulants on mortality and histological features of inflammation-driven neuropathology in P. chabaudi infection of IL-10 KO mice. We found that thrombi were prevalent throughout the brain and coincide with localization of adherent leukocytes. In addition, areas of coagulation and leukocytes co-localized with parenchymal gliosis. We also found a striking reduction of mortality and a significantly recovered parenchymal histology on elimination of coagulation suggesting a pathological role for thrombi in this model. These observations suggest an important role of coagulation in vascular congestion in CM and also implicate a novel mechanism of inflammation-induced neuropathology possibly initiated by leukocytes contained within the vasculature. These findings may be relevant because the inflammation-driven neuropathology in this model shares many features with human cerebral malaria, including intravascular leukocytes and thrombi, systemic hyper-inflammation, edema, and death.
Methods
Mice
C57BL/6J (WT) and B6.129P2-Il10tm1Cgn/J (IL-10 KO) mice (Jackson Laboratory, Bar Harbor, ME) were bred in The University of Texas Medical Branch Animal Resource Center. Experimental mice were female and between 6 and 12 weeks of age at the time of infection. All animals were kept in a specific pathogen-free housing with ad libitum access to food and water. Animals were cared for according to the Guide for the Care and Use of Laboratory Animals under Institutional Animal Care and Use Committee-approved protocols. UTMB Animal Resource Center facilities operate in compliance with the USDA Animal Welfare Act, the Guide for the Care and Use of Laboratory Animals, under OLAW accreditation, and IACUC-approved protocols.
Parasite and infection
Frozen stocks of Plasmodium chabaudi chabaudi (AS)-infected RBCs (iRBCs) (Jean Langhorne, Francis Crick Institute, London, UK) kept at − 80 °C were thawed and injected intraperitoneally (i.p.) into WT mice. Parasitized blood from these animals was diluted in Krebs-Ringer bicarbonate buffer (Sigma-Aldrich, St. Louis, MO) and normal saline to deliver 105 iRBCs i.p. in 200 μl into experimental WT or IL-10 KO mice. Thin blood smears were collected at regular intervals to monitor for peripheral parasitemia by staining with Diff-Quik (Siemens Healthcare Diagnostics, Newark, DE) or Giemsa stain (Ricca Chemical Company, Arlington, TX) and counted on a light microscope.
Animal body temperature and weight
Internal body temperatures were assessed daily during infection using rounded stainless steel rectal probes and a BIO-TK8851 digital rodent model thermometer (Bioseb, Pinellas Park, FL). Probes were sanitized with CaviCide (Metrex Research Corp., Romulus, MI) between each use. Animal weights were measured using an OHAUS Scout Pro SP601 portable balance (OHAUS, Parsippany, NJ).
Animal behavior evaluation
Beginning on day 5 post-infection, daily assessments were performed on all animals using an abbreviated version of the modified SmithKline Beecham, Harwell, Imperial College, Royal London Hospital Phenotype Assessment (SHIRPA) protocol [
54]. This brief behavioral assessment was developed based on the full assessment in a previous study [
16]. Higher scores were awarded for measures showing higher functional ability. The procedures were carried out in an open testing environment away from the home cage and took approximately 5 min per animal.
The abbreviated SHIRPA used involves a selection of nine semi-quantitative tests for general health and sensory function, baseline behaviors, and neurological reflexes. We observed undisturbed behavior with the mouse placed in an inverted beaker on top of a metal grid suspended above the home cage for 3 min, during which body position and spontaneous activity were assessed. Body position scores ranged from 0 (completely flat) to 5 (repeated vertical leaping). Spontaneous activity scores ranged from 0 (none) to 4 (rapid/dart movement). At the end of the observation period, palpebral closure, which was scored from 0 (eyes closed) to 2 (eyes wide open), and qualitative grip strength, scored from 0 (none) to 4 (unusually strong), are tested by applying a gentle horizontal force on the animal’s tail as it grips the metal grid. The animal is then placed in an open arena in which several behaviors are measured. Gait is observed as the animal traverses the arena and is scored from 0 (incapacity) to 3 (normal). During movement, tail elevation is scored, ranging from 0 (dragging) to 2 (elevated). Touch escape measures the reaction to a finger stroke and is scored from 0 (no response) to 3 (escape response to approach). Palpation of the animal’s sternum determines heart rate: 0 (slow) to 2 (fast), and finally, righting reflex is scored by releasing the animal from an upside-down position near the surface and observing the responding effort to upright itself, scored from 0 (fails to right) to 3 (lands on feet). The expected score of a healthy, uninfected IL-10 KO or WT mouse is 22. A score of 15 was identified as the humane endpoint based on the finding that any female animal that drops below that score by day 9 will succumb to infection (see Additional file
1: Figure S1).
Histochemistry
Immunofluorescence of cryosections was examined after 48 h of post-fixation of mouse brains in 4% PFA and 72 h of cryoprotection in 30% sucrose. Fixed frozen sagittal sections (30 μm) were made using Tissue Plus® Optimal Cutting Temperature Compound (Fisher Healthcare, Houston, TX) and mounted on glass slides with Fluoromount mounting medium (Novus Biologicals, Littleton, CO). Sections were incubated overnight at 4 °C with primary antibodies rabbit anti-fibrinogen (catalog no. A0080, Agilent Technologies, Carpinteria, CA), rat (clone 2.2B10, catalog no 13-0300, Thermo Fisher Scientific, Waltham, MA), or rabbit (catalog no. Z0334, Agilent Technologies, Carpinteria, CA) anti-GFAP, mouse anti-CD11b biotin (clone M1/70, catalog no. 13-0112-85, eBioscience, San Diego, CA), and rat anti-CD45 biotin (clone 104, catalog no. 13-0454-85, eBioscience, Sand Diego, CA). Secondary antibodies used were goat anti-rat AlexaFluor-488 (catalog no. A11006, Thermo Fisher Scientific, Waltham, MA) and goat anti-rabbit AlexaFluor-568 (catalog no. A11011, Thermo Fisher Scientific, Waltham, MA). Streptavidin-FITC (catalog no. 11-4317-87, eBioscience, San Diego, CA) was used as a tertiary step for biotinylated antibodies. CellTrace Violet (catalog no. C34557, Thermo Fisher Scientific, Waltham, MA)-labeled CD4 T cells were adoptively transferred into IL-10 KO mice for later co-localization with brain vasculature after i.v. perfusion with DyLight488-labeled tomato lectin (catalog no. DL-1174, Vector Laboratories, Burlingame, CA). Images of immunohistochemistry (IHC) sections were taken with an Olympus IX 71 inverted brightfield microscope using a × 20 air objective, while the immunofluorescence images were taken with a confocal microscope (Olympus FV 1000) with the DAPI channel for nuclei, Alexa 488 channel for Iba1 tagged with Alexa 488, and Alexa 647 channel for CD 31 tagged with Alexa 647. IHC images of Iba1-stained sections were contrast-enhanced and segmented by threshold for microglia using ImageJ (NIH, Version 1.48u). These were used to create binary images. Individual microglia were identified using a semi-automatic algorithm employing the particle analysis function on image and average area per microglia; the microglia density and total immunoreactive area were calculated from the binary images. Area fraction of small processes is a ratio of immunoreactive area without microglia to total immunoreactive area which indicates the degree of ramification. Transformation index, and indicator of activation, was calculated as T-Index = (Perimeter2)/(4π × Area) per microglia. To quantitatively describe the degree of ramification, we calculated the area fraction of small thin processes to total immunoreactive area. Ramification could be seen in IHC images as glia with long and thin processes that appeared segmented due to branching in and out of the tissue section plane. The astrocyte-thrombus association index was defined in which the ratio of Xi (the number of astrocytes contacting a thrombus divided by the total number of thrombi) was calculated, and values were normalized based on the following equation, (Xi − Xmin)/(Xmax − Xmin), where Xmin = 1.3 (lower limit of astrocyte-thrombi interaction seen in uninfected IL-10 KO brains) and Xmax = 3.25 (~ 75% astrocyte/thrombi association) approximated the lower and upper limit of astrocytes interacting with thrombi based on our data.
Cell and in vivo labeling
Some infected IL-10 KO and WT animals were injected with 2 × 10
6 CTV
+ CD4 T cells 3.5 h before sacrifice (i.p.) and 40 μg of DyLight488 labeled
Lycopersicon esculentum (tomato) Lectin (catalog no. DL-1174, Vector Laboratories, Burlingame, CA) 20 min before sacrifice (i.v.). CellTrace Violet (catalog no. C34557, Thermo Fisher Scientific, Waltham, MA) labeling was performed as previously described [
55].
Anti-TNF antibody treatment
Mice receiving anti-TNF antibody (clone XT3.11, Bio X Cell, West Lebanon, NH) were treated with 0.2 μg/day for 5 days starting on day 5 post-infection (days 5–9). Untreated mice received isotype rat IgG1 as a control.
CLARITY and optical clearing
Fixed brain sections (IL-10 KO and WT) were subjected to the passive CLARITY optical clearing method [
56] for large-scale labeling and imaging. In brief, mice were anesthetized and perfused transcardially with a mixture of 4% (wt/vol) PFA, 4% (wt/vol) acrylamide, 0.05% (wt/vol) bis-acrylamide, and 0.25% (wt/vol) VA044 (hydrogel solution) in PBS. Brains were extracted and incubated in hydrogel solution at 4 °C for 3 days. Solution temperature was then increased for 3 h to 37 °C to initiate polymerization. Hydrogel-embedded brains were sectioned into 2-mm-thick sagittal sections and placed in clearing solution (sodium borate buffer, 200 mM, pH 8.5) containing 4% (wt/vol) SDS) for 3 weeks at 40 °C under gentle agitation. Samples were immunostained for GFAP to assess astrogliosis. After immunostaining, samples were optically cleared using increasing serial concentrations (10–100%) of 2,2′-thiodiethanol (TDE) in Milli-Q water (EMD Millipore, Darmstadt, Germany) to achieve optimal refractive index matching with tissue.
Microscopy
Fixed cryosections (30 μm thickness, fluorescent or confocal microscopy) were imaged with a Nikon Eclipse 80i epifluorescence microscope and a Fluoview 1000MPE system configured with an upright BX61 microscope (Olympus, Center Valley, PA). Fixed, CLARITY-processed sections (2 mm thickness, two-photon confocal microscopy) were imaged using a Prairie Ultima IV (Prairie Technologies/Bruker, Middleton, WI) upright multiphoton microscope. For two-photon fluorescence microscopy, a × 10 0.3 N.A. objective (UPLFL10X, Olympus) and a × 25 1.05 N.A. super-objective (XLSLPLN25XGMP, Olympus) were used for image collection. Illumination for excitation of fluorescence was provided by a femtosecond laser (Mai Tai, SpectraPhysics, Santa Clara, CA) tuned to 800 nm. Fluorescence was collected using a two-photon standard M filter set including filters with bandwidth 604 ± 45 nm, a filter with bandwidth 525 ± 70 nm, and dichroic mirror cutoff at 575 nm. Samples were mounted on a 30-mm cage plate (CP06, ThorLabs, Newton, NJ) between two #1.5 cover glass. To visualize large regions of optically cleared brain tissue using two-photon microscopy, image stack mosaic and stitching were applied. Image stack stitching was done with a 10% overlap on a field of view of 2327.3 × 237.3 μm providing 232.73 μm of co-registration in X and Y coordinates. Images were analyzed using ImageJ (FIJI), Olympus Fluoview FV1000-ASW 2.0 Viewer (confocal), Imaris Image Analysis Software (confocal and two-photon microscopy; Bitplane USA, Concord, MA), and NIS Elements (confocal; Nikon Instruments, Melville, NY). Positive fibrinogen and elevated GFAP staining in each field was quantified by applying a signal intensity threshold and the percent area covered was calculated via the outlined areas of positive staining that met the signal intensity threshold per field of view. The percentage of total area included was calculated using ImageJ software (FIJI, NIH).
Ammonia assay
Tissue and serum ammonia was quantified using a commercial colorimetric ammonia assay kit (ab83360, Abcam, Cambridge, MA). Briefly, brain and liver samples were collected from infected IL-10 KO and WT mice at the peak of behavioral symptoms, washed in cold PBS, resuspended in 100 μl assay buffer, and homogenized using a Dounce homogenizer to produce single-cell suspensions. After 2–5 min of centrifugation at 4 °C, cells were counted via hemocytometer and seeded into a 96-well plate to provide 1–5 × 104 cells/well. Serum samples were counted and seeded directly into plates without processing (5–10 μl/well). The colorimetric assay was conducted using OxiRed probe. Color change was recorded at OD 570 nm using a spectrophotometer microplate reader and compared to an ammonium chloride standard curve (detects 0–10 nmol/well) after 60 min of incubation at 37 °C.
Statistics
Where indicated, groups were compared by t test (2 groups) or one-way ANOVA (3 or more groups), followed by post hoc Bonferroni method or Tukey’s test to identify significance between individual groups. Each point represents the average value per animal after analysis of 10 fields, unless otherwise specified. Statistical analysis was performed in Prism (GraphPad, La Jolla, CA), *p ≤ 0.05, **p ≤ 0.01, and ***p ≤ 0.001. Error bars represent ± SEM.
Discussion
The presence of peripheral immune cells adherent within the vasculature in mouse models of CM and in brain vessels on autopsy of cerebral malaria patients [
66] suggest that such cells play an important role in mediating neuropathology [
67]. Current paradigms to explain CM pathogenesis support an important role for inflammation in the generation and amplification of neuropathology but do not explain the derivation of these cytokines in the brain. The derivation and contribution of cerebral thrombi to CM pathology is also poorly understood. The vascular findings in this study suggestive of pervasive (Fig.
1) and complete (Fig.
2) blockade of the vasculature by inflammation-induced thrombi are striking. These abnormalities have not been described in
P. chabaudi infection before. Coagulation is clearly of major relevance for our understanding of pathological mechanisms in cerebral malaria [
21,
58,
68]. Potentially pathogenic serum levels of both pro- and anticoagulation proteins have been documented in human CM [
69,
70]. Systemic inflammation has also recently been shown to contribute to intravascular clotting via mechanisms involving neutrophils and monocyte interaction with platelets in CM [
71,
72], linking inflammation and clotting, which in turn promote sequestration. Recent studies also show that the anticoagulation endothelial protein C receptor (EPCR) may bind the parasite and be downregulated, thus promoting clotting and suggesting a mechanism for the induction of coagulation by
P. falciparum sequestration [
45,
73]. Interestingly, studies point to the bi-directional amplification of the clotting cascade and inflammation suggesting an important intersection that is likely to be crucial to pathology in CM [
58].
The data presented here confirm that inflammatory cells within the vasculature can drive both clot formation and activation of cells in the brain parenchyma in the absence of local parasite adhesion. Studies of
Plasmodium berghei (ANKA) (PbA) infection have established the importance of the inflammatory response in the development of neurocognitive dysfunction [
74‐
76]. PbA infection shows pathogenic immune cell accumulation in cerebral blood vessels as a result of inflammatory TNF and IP-10 secretion [
77,
78] and intercellular adhesion molecule-1 (ICAM-1) on the vascular endothelium [
79]. PbA infection has also been shown to induce astrocyte activation and degeneration near sites of monocyte vascular adhesion [
62,
80]. However, the signals leading to the breakdown of local astrocyte barrier function in malaria have not yet been defined. The activation of astrocytes is a feature of many neurological diseases, including cerebral malaria [
81,
82]. Our results demonstrate a causal link between hyper-inflammation, hyper-coagulation, glial cell activation, and mortality (Figs.
3 and
4). Gliosis across multiple areas of the brain was observed in infected IL-10 KO mice, with astrocytes and microglia associating highly with the vasculature compared to the WT group—yet both microglial and astrocyte gliosis were significantly reduced upon LMWH treatment, indicating this direct link.
This is important because resolution of CM in African children and Asian adults can be resistant to anti-malarial drug treatment, suggesting that parasite alone does not cause the full cerebral malaria syndrome. Furthermore, it is not yet clear how parasite adhesion alone drives the neuropathology evident from patient symptoms [
83]. However, because of the overlap of inflammation with parasite-dependent factors, determining the independent contributions of each presents an ongoing challenge to investigators. The impact of parasite adhesion to the vascular endothelium on coagulation, vascular integrity, and congestion has been shown in in vitro endothelial cultures and animal models of cerebral malaria [
19,
43,
67,
84,
85]. Sequestration is seen in most fatal pediatric and adult CM cases [
20,
21] and is used as a critical hallmark of disease. We chose to study the role of inflammatory cytokines in isolation from the potential contribution of sequestration using an inflammation-induced cerebral malaria model. The results confirm that inflammation can cause many of the pathological changes seen in CM, though not all.
In this study, we show that both the congestion phenotype associated with intravascular clotting and astrocyte activation can be reversed via neutralization of TNF (Fig.
5), or anticoagulant therapy (Fig.
6). Serum TNF concentration correlates with severity of human malaria [
86]. However, TNF blockade has thus far proven ineffective in preventing death in childhood cerebral malaria [
87,
88]. As different reagents displayed differential effects, the timing, dose, or precise antigenic specificity of treatments may yet be improved for adjuvant therapy. Strikingly, these data also show that fatal neurological disease in IL-10 KO mice is dependent on intravascular coagulation, as it can be prevented by LMWH treatment (Fig.
6). This demonstrates a central role for thrombi in driving the disease mortality and promoting neuropathology in
P. chabaudi infection of IL-10 KO mice. As anti-TNF and anticoagulants have similar effects in this model, it is likely that cytokines and the coagulation cascade promote each other, as in other systems. Despite the WHO recommendation against the use of heparin since 1984, citing excessive bleeding [
89], there are several clinical trials showing significant beneficial effects of anticoagulant usage on mortality and length of coma in human CM [
49,
50,
90,
91]. Selection of treatments with relatively moderate anticoagulation activity is likely essential to achieving therapeutic goals while avoiding hemorrhagic complications. LMWH, as the name implies, involves only the activity of the smaller heparin proteins, which act with higher specificity on factor Xa, exhibit less thrombin inhibition, and produce a more reliable therapeutic profile. Our studies show that LMWH treatment is protective within the context of hyper-inflammatory cerebral malaria and prevents intravascular thrombi formation in the brains of mice exhibiting behavioral dysfunction (Fig.
6). This is particularly important in that both astrocyte and microglial activation were dependent on this coagulation event to some degree (Figs.
6 and
7). Activation of microglia has been shown to be an important component of neuroinflammation and behavioral dysfunction associated with PbA infection [
92‐
94]. Widespread microglial activation, not always restricted to areas of parasite sequestration, has also been identified in cases of human CM [
95,
96]. However, these findings are novel in the context of
P. chabaudi infection. Furthermore, the spatial relationship of intravascular coagulation with glial cell activation is also previously unknown in any malaria infection and should be examined in human CM autopsy samples.
Efforts to manipulate the inflammatory response and clotting cascade have provided mixed results in clinical trials to date [
97‐
99], highlighting the importance of understanding the interactions between various arms of the host response within the pathogenesis of cerebral malaria. In summary, our experiments support the importance of intravascular coagulation and leukocytes producing inflammatory cytokines in malaria-induced cerebral pathology. The activation of surveilling microglia and vascular/neuronal-supportive astrocytes downstream of systemic inflammation could promote the generation of neuropathology secondary to malaria infection. Identification of both T cells and monocytes within fibrin clots suggests a new working model where inflammatory cells promote cerebral damage even from their localization within the cerebral vasculature. It is possible that leukocytes within the structure of intravascular thrombi serve to amplify pathological inflammatory cytokines leading to immunopathology in the brain. These data demonstrate the interaction of the anti-parasitic and hemostatic elements of host defense, promoting a new appreciation of the interplay between mechanisms important for development of fatal cerebral malaria.