The blood-brain barrier (BBB) is a dynamic barrier formed by endothelial cells (ECs) that controls the movement of molecules, plasma proteins, pathogens and cells between the blood and brain. The brain endothelium of the BBB is part of a cellular complex known as the neurovascular unit (NVU), which is composed of a basement membrane, pericytes, astrocytes, microglia and neurons (reviewed by Abbott and Friedman, 2012) ( Figure 1 ).

Brain ECs (BECs) vary from peripheral ECs in that they have no fenestrae, minimal pinocytic activity, a continuous basement membrane, a negatively charged luminal surface and the existence of tight junctions (reviewed by Daneman and Prat, 2015). Tight junctions and adherens junctions between BECs produce a strong BBB with a very high transendothelial electrical resistance (TEER) that is 50-100 times tighter than peripheral ECs (reviewed by Abbott et al., 2010).

BECs play an active role in the pathogenesis of a number of central nervous system (CNS) disorders such as CM. In the late stages of the intraerythrocytic cycle, Plasmodium falciparum infected red blood cells (PRBC) expressing Plasmodium falciparum erythrocyte membrane protein-1 (PfEMP1) bind to several receptors including the intercellular adhesion molecule (ICAM-1) and endothelial protein C receptor (EPCR) on brain endothelial cells in a process known as sequestration (Brown et al., 2013; Lau et al., 2015; Avril et al., 2016) ( Figure 2 ). Results from human post-mortem and in vitro studies suggest that sequestration is involved in CM pathogenesis. A high accumulation of PRBC in the cerebral vasculature was observed in the brains of patients who died from CM (MacPherson et al., 1985; Turner et al., 1994; Ponsford et al., 2012). Studies by Tripathi et al. (2006) showed a dose and time-dependent increase in ICAM-1 levels when PRBC were co-cultured with human brain endothelial cells (HBEC) (Tripathi et al., 2006).

Examination of post-mortem CM brain tissues of Vietnamese adults and Malawian children showed a reduction in junction proteins vinculin, zonula occludens-1 (ZO-1), and occludin in blood vessels where PRBC sequestration was present (Brown et al., 1999; Brown et al., 2001). In vitro coculture of PRBC from CM patients with human umbilical vein endothelial cells (HUVEC) also resulted in lower vinculin, ZO-1, and occludin levels (Susomboon et al., 2006). Similarly, a reduction in TEER was observed when PRBC were cocultured with HBEC (Tripathi et al., 2007). These findings indicated that sequestration of PRBC to BBB resulted in a reduction in junction proteins which implies loss of BBB integrity.

MicroRNA155 (miR155), a small noncoding molecule involved in neuroinflammation at the BBB, has also been shown to cause BBB disruption in vitro. Upregulation of miR155 induced by proinflammatory cytokines resulted in the reorganization of junction proteins and a 1.9 fold increase in HBEC permeability in vitro (Lopez-Ramirez et al., 2014). In ECM, genetic deletion of miR155 reduced endothelial cell activation and decreased BBB leak (Barker et al., 2017) Additionally, treatment with anti-miR155 decreased vascular leakage caused by serum from cerebral malaria patients close to basal levels in an ex vivo endothelial microvessel model (Barker et al., 2017). These results suggested that miR155 indirectly contributed to endothelial dysfunction and BBB disruption in CM ( Figure 2 ).

Transmigration of PRBC into the brain endothelium can cause BBB disruption during CM ( Figure 2 ). Studies by Adams et al. (2021) showed that incubation of PRBC expressing dual ICAM-1 and EPCR PfEMP-1 proteins, with HBEC, resulted in the internalization of PRBC by HBEC. This resulted in the swelling of the HBEC and BBB breakdown (Adams et al., 2021) ( Figure 2 ). Also, ex vivo studies showed the presence of internalized PRBC within HBEC in post-mortem tissue samples from Indian patients who died from CM. This data indicates transmigration of PRBC into the brain endothelium during CM could be a contributing factor to CM pathogenesis (Adams et al., 2021).

Altogether these studies suggest that activation of the brain endothelium during CM can cause BBB disruption. This can lead to the movement of cytokines, chemokines and parasite-derived products into the brain parenchyma where they activate cells of the NVU such as astrocytes and microglia.

Astrocytes are glial cells whose end-feet surround >99% of the brain capillaries (Mathiisen et al., 2010). They can be characterised into two broad morphologies: protoplasmic and fibrous astrocytes (Andriezen and Lond, 1893). Protoplasmic astrocytes are located in the grey matter whereas fibrous astrocytes can be found in the white matter (Zhang et al., 2019). Although astrocytes are broadly classified into these two groups, astrocyte heterogeneity can be observed within and between regions. Studies by Batiuk et al. (2020) using single-cell RNA sequencing and transcriptomics observed 5 different astrocyte populations in the hippocampus and cortex of adult mice (Batiuk et al., 2020). Astrocytes are involved in many key processes in the brain. They protect and support neurons by regulating synapse formation and maintaining brain homeostasis (reviewed by Dossi et al., 2018). Astrocytes also secrete factors that are important for the formation of a functional BBB (reviewed by Dossi et al., 2018). In response to CNS injury, astrocytes undergo a gradation of cellular, molecular, and functional changes known as astrogliosis (reviewed by Dossi et al., 2018). Astrogliosis is characterised by an increase in glial fibrillary acidic protein (GFAP) and can range from mild to moderate or severe alterations in astrocytes where compact scar formations occur (referred to as glial scar) (reviewed by Dossi et al., 2018). Astrogliosis is thought to initially repair and limit the level of damage during CNS injury, however, it inhibits regeneration and later causes detrimental effects in CNS disorders.

Reactive astrocytes have recently been grouped into two different types in the adult CNS, A1 and A2. Studies by Liddelow et al. (2017) showed that A1 type astrocytes were induced by activated microglia via secretion of interleukin-1 alpha (IL-1α), tumour necrosis factor (TNF) and C1q. A1 astrocytes in vivo and in vitro released an unknown neurotoxin that caused apoptosis of neurons and oligodendrocytes (Liddelow et al., 2017). However, A2 astrocytes were induced by ischaemia and upregulated neurotrophic genes that promoted neuronal survival (Liddelow et al., 2017). These results indicate that astrocytes can play a neuroprotective or detrimental role in the brain during different neurological disorders. Astrocytes are heterogeneous and difficult to study, thus it is possible that more subtypes of reactive astrocytes exist. Reactive astrocytes may also possess unique cellular and molecular features that only occur in specific neuropathology. There is still a lot of research that needs to be done to understand the mechanisms and pathways that are involved in the induction of reactive astrocytes in CNS disorders.

Microglia are the resident immune cells of the CNS and play a vital part in the immune response (Thurgur and Pinteaux, 2019). They are restricted to the brain and self-renew throughout life without the involvement of circulating blood cells (reviewed by Arcuri et al., 2017).

Microglia are very heterogeneous, differing in population densities across brain regions, and more microglia are present in the gray matter than the white matter. White matter microglia have elongated somata aligned parallel to fibres whilst microglia in circumventricular organs have a compact morphology and microglia in the grey matter are radially ramified (reviewed by Arcuri et al., 2017). Microglia cells express the pattern recognition receptors (PRRs) that survey their microenvironment and recognise indicators for injury known as pathogen-associated molecular pattern molecules (PAMPs) and damage-associated molecular patterns (DAMPs) (reviewed by Arcuri et al., 2017). Upon acute brain injury microglia transition from a surveillance mode to an activated state where they undergo significant morphological changes, decrease the number of processes and release pro-and anti-inflammatory cytokines (reviewed by Arcuri et al., 2017).

Previously, microglia were classified into two different activation phenotypes. The M1 (classical activation) phenotype referred to a proinflammation state in which microglia release mediators TNFα, IL-1β, IL-6, reactive oxygen species (ROS) and nitric oxide (NO); whereas the M2 (alternative activation) phenotype referred to an anti-inflammatory state where microglia released trophic factors such as IL-4, IL-10 and transforming growth factor beta (TGFβ) (reviewed by Figarella et al., 2020). However, this classification did not reflect the heterogeneity of the microglia phenotype and now it is known that microglia can assume a broad spectrum of different activation profiles. This heterogeneity of the microglial phenotype depends on the anatomical region of microglia and their close interactions with cells such as microglia, astrocytes, neurons and oligodendrocytes (reviewed by Figarella et al., 2020). Initially, activated microglia were thought to be only detrimental to the CNS, however, several studies suggest that activated microglia have both detrimental and beneficial functions. Microglia can also be neuroprotective by producing factors such as brain-derived neurotrophic factor, glial cell-derived neurotrophic factor, and nerve growth factor, that can help prevent neuronal damage (reviewed by Fakhoury, 2018).

Neuroinflammation is a common characteristic of most CNS disorders and insults. It is associated with activation of astrocytes and microglia with marked production of cytokines, chemokines, proinflammatory mediators including C-X-C motif chemokine ligand 10 (CXCL10), and TGFβ, and BBB disruption in neurological diseases such as CM ( Figure 2 ).

The chemokine CXCL10 also known as interferon-γ inducible protein-10 (IP-10) is constitutively expressed in astrocytes, microglia, and neurons (reviewed by Jiang et al., 2017) and markedly increased in reactive astrocytes in CNS disorders such as Alzheimer’s disease (AD) (Xia et al., 2000). CXCL10 was significantly elevated in the serum and CSF of Ghanaian children with CM (Armah et al., 2007) suggesting that CXCL10 played a role in neuroinflammation observed during CM. Indeed, in vivo, upregulation of genes involved in chemokine production of CXCL9, CXCL10, CCL8 and CCL12 was observed in the microglia of mice infected with PbA using genome wide transcriptomic analysis (Capuccini et al., 2016). CXCR3, the receptor of CXCL10 is known to play a key role in the recruitment of T cells into the brain. Studies by Campanella et al. (2008) showed CD8⁺ T cell infiltration into the brain of CXCR3 knock-out mice infected with PbA was significantly reduced by 300% compared with wild-type mice (Campanella et al., 2008). Data from both studies suggested that upregulation of CXCL10 by activated glial cells could induce the recruitment of CD8⁺ T cells into the brain by binding to CXCR3 highly expressed on the surface of CD8⁺ T cells during ECM.

CD8⁺ T cells were previous shown to only exist in ECM, however, recent studies confirmed the presence of CD8⁺ T cells in the brains of Malawian children who died from CM (Barrera et al., 2019; Riggle et al., 2020). In these studies, CD8⁺ T cells were found in the intravascular and perivascular space of the brain but were absent from the brain parenchyma (Barrera et al., 2019; Riggle et al., 2020). CD8⁺ T cells were also observed in the leptomeninges and choroid plexus in HCM samples suggesting possible routes of entry into the superficial areas of the brain (Barrera et al., 2019). Also, granzyme B (which is a protease known to mediate cellular apoptosis) was found to be expressed by CD8⁺ T cells that were in contact with endothelial cells (Riggle et al., 2020). Results from these studies suggest that CD8⁺ T cells do not enter the brain parenchyma during HCM and thus may not have a direct influence on the neuroinflammation caused by microglia and astrocytes during HCM. However, astrocytes can extend their processes across the perivascular space in the brain, bringing them in contact with activated CD8⁺ T cells that can then target astrocyte processes by releasing granzyme B ( Figure 2 ). This could result in astrocyte activation and the release of chemokines and cytokines leading to further neuroinflammation in the brain during HCM ( Figure 2 ).

TGF-β is another cytokine that has been shown to be involved in the neuropathogenesis of CM TGF-β expression was upregulated in the brain sections of CM patients in an intravascular and perivascular distribution but not in an intraparenchymal distribution (Armah et al., 2005). In the post-mortem brain tissues of CM patients, a significant increase in TGF-β-1 expressing astrocytes was observed around the brain vessels with malaria pigment. Also, TGF-β2 expressing microglia within Dürck granulomas and ring haemorrhages, and TGF-β3 expressing endothelial cells around the brain vessels were observed in the post-mortem brain tissue of CM patients (Deininger et al., 2000). These results suggest that TGF-β expressed by glial cells contributes to neuroinflammation during CM. However, the exact role of TGF-β expressed by glial cells during CM has not been explored. In other neurological disorders, TGF-β expression by glial cells can be beneficial or detrimental depending on the disease. Overproduction of TGF-β1 in astrocytes accelerated disease progression and reduced microglia function in amyotrophic lateral sclerosis (ALS) mice (Endo et al., 2015). On the contrary, TGFβ-1 derived from astrocytes protected synapses against amyloid B oligomers (the main component of amyloid plaque in AD) (Diniz et al., 2012). Also, TGF-β treatment after intracerebral haemorrhage resulted in a reduction in microglia inflammation and an increase in functional recovery in vivo (Taylor et al., 2016). Reduced levels of TGF-β in malaria patients were associated with disease severity. TGF-β1 levels in the serum and plasma of cerebral malaria patients were significantly reduced compared to uncomplicated malaria patients (Chaiyaroj et al., 2004; Hanisch et al., 2015).

High levels of TGF-β are associated with anti-inflammatory effects whereas low levels of TGF-β are associated with proinflammatory effects. Thus, it is possible that during CM, the amount of TGF-β expressed by glial cells determines the severity of CM. Increased levels of TGF-β could suppress neuroinflammation in CM and low levels of TGF-β could exacerbate neuroinflammation in CM. Further studies needed to be done to determine the exact role of TGF-β in the neuropathogenesis of CM.

Coagulation factors have also been shown to be important players in inflammation in several neurological diseases including CM ( Figure 2 ). Indeed, dysregulation of coagulation has been found in both HCM and murine CM. In the post-mortem brain tissues of CM patients, sequestration was associated with microvascular thrombi and perivascular haemorrhages (Dorovini-Zis et al., 2011).

In vivo, vascular thrombi containing adherent leukocytes were observed in IL-10 knockout (IL-10 KO) mice that had been infected with the rodent malaria parasite P. chabaudi. In this study, astrocytes and microglia were also detected in the brain parenchyma clustered near vessels with thrombi (Wilson et al., 2018). Interestingly, neutralisation of TNF and coagulation caused a marked reduction in intravascular thrombi and decreased astrocyte and microglial activation in P. chabaudi IL-10 KO mice (Wilson et al., 2018). These results could indicate that leukocytes contribute to intravascular coagulation during malaria and also suggest that there is an association between inflammation, coagulation, and glial activation during murine cerebral malaria (Wilson et al., 2018). This study suggests that inflammatory leukocytes localised in thrombi could produce cytokines that can cross the BBB and activate glial cells during cerebral malaria thereby contributing to neuroinflammation in the brain parenchyma during cerebral malaria.

In addition to proinflammatory cytokines, extracellular vesicles (EVs) have also been shown to contribute to inflammation during CM ( Figure 2 ). Extracellular vesicles (EVs) are produced from different subcellular compartments and are released into the extracellular space where they can influence cells within the vasculature such as endothelial cells. Studies by Combes et al. (2004) showed significantly higher levels of endothelial EVs were present in the plasma of Malawian children suffering from CM compared to patients with severe and uncomplicated malaria (Combes et al., 2004). Elevated levels of platelet, erythrocytic, endothelial, and leukocyte-derived EVs were observed in patients with P. falciparum malaria that had neurological dysfunctions (Pankoui Mfonkeu et al., 2010).

ATP-binding cassette transporter (ABCA) 1 knockout (KO) mice, that had a reduced ability to produce EVs, were protected against ECM and did not display cerebral symptoms (Combes et al., 2005). In addition, reduced inflammation accompanied by a notable decrease in TNFα was observed in PbA infected ABCA1 KO mice compared to wild-type mice (Combes et al., 2005). Data from the studies above indicate that EVs promote inflammation in CM and ECM and could contribute to the neurological syndrome observed in severe malaria.

EVs derived from PRBC and PbA infected red blood cells (PbARBC) have also been shown to interact with cells in the NVU such as astrocytes and microglia. In vitro, coculturing of PbARBC cells with a mixed astrocyte and microglia culture resulted in an uptake of EVs by astrocytes and phagocytosis of PbARBC by microglia (Shrivastava et al., 2017). Subsequently, this caused an increase in the CXCL10/IFN inducible protein 10 (IP10) secretion (Shrivastava et al., 2017). This indicated that the internalisation of EVs from malaria-infected red blood cells extracellular vesicles (MiREVs) by glial cells could contribute to neuroinflammation observed during CM. Recent studies by Mbagwu et al. (2020) showed the uptake of MiREVs by microglia generated from human blood monocytes in the perinuclear region led to morphological changes in microglia such as retraction of processes and swelling of the cell body suggesting microglial activation (Mbagwu et al., 2020). Furthermore, treatment of microglia with MiREVs derived from supernatants from P. falciparum culture caused an upregulation in the gene expression of TNFα and the downregulation in the gene expression of the anti-inflammatory cytokine IL10 (Mbagwu et al., 2020). This suggested that MiREVs contributed to inflammation in CM by inducing upregulation of the inflammatory cytokine TNFα and decreasing the expression of the immune-suppressive cytokine IL-10 in microglia.

Data from the above studies suggest that MiREVs may interact with glial cells during CM and this could exacerbate neuroinflammation in CM.

Malaria derived products such as haemozoin (Hz) and heme oxygenase-1 (HO-1) also contribute to neuroinflammation in CM ( Figure 2 ). During the intraerythrocytic stages of malaria, digestion of haemoglobin by the Plasmodium parasite produces heme and this is stored in the parasite’s digestive vacuole as (Hz), a non-toxic crystalline polymer. Heme oxygenase-1 (HO-1) also degrades heme into carbon monoxide (CO), biliverdin, and iron (Schluesener et al., 2001). High levels of Hz correlate with disease severity in malaria patients (Lyke et al., 2003).

In the post-mortem brain tissues of CM patients, a significantly higher accumulation of Hz and platelets were observed, compared to the post-mortem brain tissues of severe malaria with anaemia patients and nonmalaria encephalopathy patients (Grau et al., 2003). The effect of Hz on cells of the NVU has recently been shown in vitro and in vivo.

A dose and time-dependent uptake of synthetic Hz (sHz) by astrocytes and neurons were observed in vitro, leading to apoptosis in both cells (Eugenin et al., 2019). This indicated that Hz was toxic to astrocytes and neurons and possibly contributed to the neurological sequelae observed in survivors of CM (Eugenin et al., 2019). Exposure of microglia to sHz caused a significant increase in TNFα, IL-6, IL-1β, inducible nitric oxide synthase (iNOS) mediated NO production expression and NLRP3 inflammasome activation in vitro (Velagapudi et al., 2019). However, treatment of mouse peritoneal macrophages (PM) with sHz following stimulation with lipopolysaccharide (LPS) impaired their ability to produce NO and TNFα but caused an increase in HO-1 (Taramelli et al., 1995; Taramelli et al., 2000). Interestingly, this effect was not observed in microglia treated with sHz where a reduction in NO and TNF α production was not observed. It was suggested that sHz induced oxidative stress in PM and this, in turn, inhibited the production of inflammation cytokines (Taramelli et al., 1995; Taramelli et al., 2000). HO-1 could also have suppressed iNOS production impairing NO synthesis in PM. Microglia are known to contain high amounts of antioxidant molecules and thus seem to be protected against oxidative stress (Vilhardt et al., 2017). This could be the reason sHz had no effect on microglia. Altogether results from these studies indicate that sHz could play a key role in neuroinflammation through activation of astrocytes and microglia and could contribute to the neurological damage observed in survivors of CM ( Figure 2 ).

The water protein channel Aquaporin 4 (AQP4) found in the astrocyte endfeet has been suggested to contribute to the neuropathogenesis of CM ( Figure 2 ). AQP4 plays a key part in the control of water movement into and out of the brain and has been associated with vasogenic oedema and cytotoxic oedema in CM (Mohanty et al., 2017; Huang et al., 2019).

It remains unclear whether AQP4 plays a protective or deleterious role in CM. High levels of AQP4 were found to be expressed in areas of retinal whitening in CM patients due to retinal disruption (Barrera et al., 2018). Post-mortem studies showed increased expression of AQP4 in the brain stem of CM patients (Medana et al., 2011). In ECM, AQP4 was significantly higher in PbA infected mice susceptible to neurological manifestations such as palsies, convulsions, or ataxia (Ampawong et al., 2011). Thus, it was proposed that AQP4 played a detrimental role in ECM, and above a certain AQP4 expression threshold, neurovascular pathology occurred in mice (Ampawong et al., 2011). On the other hand, a reduced expression of AQP4 and a higher degree of oedema was observed in AQP4 knockout mice with ECM suggesting that AQP4 protected mice from oedema in ECM (Promeneur et al., 2013).

It is possible that the time point of injury and the type of oedema present in CM could determine whether AQP4 is beneficial or detrimental in cerebral malaria. Indeed, in vivo studies have shown AQP4 improves vasogenic oedema in the initial stages of spinal cord contusion through the reabsorption of water (Huang et al., 2019). At the middle stages of SCI, AQP4 facilitates the development of cytotoxic oedema, and inhibition of AQP4 formation reduces cytotoxic oedema and alleviates motor functions in vivo (Huang et al., 2019). Further studies need to be done to determine the exact role of AQP4 at the initial stage of CM and as the disease progresses.

S-100B, a marker of astrocyte brain injury is associated with an increased risk of seizures during CM ( Figure 2 ). In severe falciparum malaria, high levels of the astrocyte marker S-100B in the cerebrospinal fluid (CSF) of Kenyan children and Vietnamese adults were linked with an increased risk of recurrent seizures (Medana et al., 2007). Similarly, serum S-100B levels were markedly elevated in children with temporal lobe epilepsy, a condition characterised by recurrent seizures, compared to healthy controls (Calik et al., 2013). Results from both studies indicated that astrocytes contributed to seizures that occurred in CM and temporal lobe epilepsy. S-100B has also been suggested to contribute to neuroinflammation that occurs in CM. Treatment of co-cultures of astrocytes and neurons with S-100B resulted in apoptosis of neurons and this was dependent on NO produced from astrocytes (Hu et al., 1997). Also, S-100B induced IL-1β expression in microglia in vitro (Liu et al., 2005; Bianchi et al., 2011). Since S-100B is derived from astrocytes, these results indicate that astrocytes can coordinate certain elements of neuroinflammation caused by neurons and microglia. Thus, it is possible that during CM, S100B contributes to neuroinflammation in the brain by targeting microglia and neurons. However, the use of S-100B in the detection of brain damage is controversial and several factors affect the accuracy of S-100B as a biomarker for brain injury. S-100B has a short serum half-life, differs with age in children and extracerebral sources such as epithelial cells and adipose may contribute to S-100B serum levels (Ingebrigtsen et al., 1995; Woertgen et al., 2002; Gazzolo et al., 2003; Thelin et al., 2016; Bai et al., 2018). Hence, this raises questions concerning the reliability of S100B as a marker for brain injury. Thus, in future studies, the influence of age, sampling time and S-100B kinetics on S-100B levels should be considered when using S-100B as a biomarker for CNS damage in CM.

Neuroglobin (NGB) is a globin protein expressed in neurons and astrocytes. In neurons, neuroglobin has been shown to play a protective role in different pathologies such as AD and TBI by eliminating ROS. NGB expressed by astrocytes has also been detected under pathophysiological conditions in astrocytes.

Studies by DellaValle et al. (2010) showed that NGB was expressed by reactive astrocytes in murine models of cerebral malaria, TBI and autoimmune encephalitis (model of multiple sclerosis) but absent in the brain of healthy control mice (DellaValle et al., 2010). NGB was only observed in regions of the brain with severe pathology and astroglial scar, and BBB leakage was observed in all models with NGB astrocytes (DellaValle et al., 2010). Thus, it is possible that during ECM, following BBB damage, reactive astrocytes express and release NGB that might contribute to neuronal protection ( Figure 2 ). However, the exact role of NGB expressed by astrocytes in CM has not been explored and further studies need to be done to understand the function of NGB secreted by astrocytes during CM. Altogether the studies in this review indicate that astrocytes and microglia contribute to the neuropathogenesis of cerebral malaria. Cross talk occurs between microglia and astrocytes through the secretion of molecules such as cytokines, chemokines, growth factors, NO and ROS (reviewed by Matejuk and Ransohoff, 2020).

Normal astrocyte-microglia cross-talk in disease is essential to support neuronal survival and function after acute injury whereas abnormal astrocyte-microglia cross-talk may promote neuroinflammation and neurological damage (reviewed by Matejuk and Ransohoff, 2020). Glial cells are heterogeneous, thus the impact of astrogliosis and microglia activation in the brain during CM will not be an all-or-nothing phenomenon but will occur in a context-dependent manner and this is regulated by specific signaling cascades associated with specific insults from the environment. The glial phenotype during CM could depend on the severity of disease, source of injury and region of the brain CM is occurring in.