Unravelling mysteries at the perivascular space: a new rationale for cerebral malaria pathogenesis

(i). Perivascular space and CD8⁺ T cells

T cell entry into the perivascular space due to BBB breach is a key event in many infectious and autoimmune diseases [40,64]. In its simplest form, the BBB can be distinguished into two types. One type surrounds capillaries, is formed by a single layer of endothelial cells abutted by a glia limitans layer of astrocyte end feet, and is supported by a basement membrane. The second type surrounds post-capillary venules, is similar in structure, with the two layers separated by the perivascular space [40] (Figure 3). Intravital imaging studies in experimental autoimmune encephalomyelitis (EAE), an animal model for multiple sclerosis, have adeptly captured the events that unfold when T cells that promote inflammation initially gain entry [64]. T cell infiltration through the tight junctions of the venular endothelium is regulated by resident antigen-presenting cells (APCs) in the perivascular space via an antigen-dependent mechanism [65,66]. These findings are likely relevant in CM, where a similar line of enquiry is currently followed. Leucocyte infiltration into the perivascular space during CM is a distinctive feature of the venous vasculature in both mice and humans [7,21] (Figure 3). Post-mortem analyses in paediatric patients from Malawi demonstrated an infiltration of CD8⁺ T cells into the leptomeninges, pial vessels, subarachnoid space and choroid plexus [67]. Further post-mortem sample analyses showed for the first time a similar pattern of CD3⁺CD8⁺ T cell accumulation in ECM and HCM [21,68], dispelling the long-held view that this phenomenon was solely associated with ECM [69]. Granzyme B-expressing CD8⁺ T cells were found in close contact with CD31⁺ endothelium lining the luminal wall of the venous vasculature, as well as the abluminal wall in the perivascular space in paediatric HCM [21]. The perivascular space appears to increase in size during ECM and HCM. This cerebrovascular engagement of CD8⁺ T cells in children who died of CM was exacerbated by HIV coinfection. Long-lived cells such as PVMs, microglia, and astrocytes act as a reservoir for latent HIV, but how this might impact on CM pathogenesis remains unknown. Additionally, many activated Iba1⁺CD68⁺ macrophages in children who died of CM had localized both luminally and abluminally to the same sites where CD8⁺ T cells had released granzyme B, indicative of antigen recognition [21]. Expression of Iba1⁺ and CD68⁺ is a shared phenotypic feature of macrophage subsets, including PVMs, although this was not investigated. Recently, studies used a multiphoton microscopy approach to show that PVMs create a ‘vascular niche’ for CD8⁺ T cell arrest, which may kick-start the inflammatory cascade leading to BBB alteration and neuropathology in ECM [26,70] (Figure 3). Myeloid cells clustering around the vasculature, activation of CD8⁺ T cells and pathogen clearance via the recognition of peptides presented by H2K^b or H2D^b have been reported in ECM [26,71]. Repeated interactions of CD8⁺ T cells at PVM sites can lead to the release of cytotoxic mediators’ perforin and granzyme B, which would accumulate at inter-endothelial junctions over time and lead to BBB disruption in CM (Figure 3).

(ii). Perivascular myeloid cells

Perivascular myeloid cells, including PVMs, are a unique subset of immune cells that contribute to inflammation and cerebrovascular dysfunction (Figure 3). The strategic localization of PVMs beneath the endothelium in the perivascular space make them prime candidates for immunosurveillance [26]. PVMs are ideally positioned to phagocytose harmful pathogens that cross the vasculature into tissues, and are particularly abundant in the leptomeninges and subarachnoid space compared to the parenchyma. They become activated soon after pathogen entry and efficiently co-ordinate the innate and adaptive immune response [26]. Thus far, underappreciated is the role of perivascular myeloid cells in continually surveying the abluminal wall with motile process extensions during neuroinflammation. In this context, the engulfment of PbA in the perivascular space by PVMs is highly relevant [71] (Figure 3). Phagocytosis by intravascular macrophages directly from systemic circulation was first reported in 1883, with cells protruding into the vascular lumen. Similarly, PVMs have access to iRBCs in the perivascular space in the late stages of CM, when BBB disruption occurs [7]. However, it is unclear if and how PVMs sense phosphatidylserine ‘eat me’ signals and phagocytose pathogens in the early stages while localized beneath the impervious endothelial barrier [72]. Transport of antigens via an independent route – such as through the choroid plexus to the perivascular space – may be involved [73]. Of note, microglia extending processes through the intercellular junctions of the endothelium into the vascular lumen of the brain has been reported [74]. These findings provide a lead for investigating whether, and how, iRBCs may be phagocytosed by PVMs directly from the vascular lumen in the absence of a BBB breach during early-stage disease (Figure 3).

The glymphatic system, an extensive meningeal lymphatic vessel network, provides a pathway of drainage from the cranium. Tracers injected into the meningeal lymphatic vessels drained into the deep cervical lymph nodes (DCLNs) [9]. As iRBCs or their products accumulate in the perivascular space, the aforementioned studies provide compelling grounds for investigating whether Plasmodium antigens in the CNS can access DCLNs through the cervical lymphatics [7,9], either contributing to the deleterious immune response in CM, or tuning it down [9].

Extravasated T cells arrest and form long-term cognate interactions with CX3CR1-bearing APCs [7,75] (Figure 3), dispelling the notion that T cell activity is confined to the vascular lumen during ECM. Despite the compelling evidence of monocyte/macrophage involvement in their studies (see earlier text), Riggle et al. subscribed to the long-held view that P. falciparum antigen was being presented to T cells by the endothelium [21]. There is evidence, based on in vitro and ex vivo studies, that the endothelium can acquire and cross-present P. falciparum and PbA antigens [76]. However, our understanding of the structure of the endothelium has undergone a significant makeover in recent years. Seminal studies have shown that process extensions of macrophages localize to inter-endothelial junctions of blood vessels and bridge neighbouring tip cells during embryogenesis [77]. Moreover, PVMs play an important role in the maintenance of inter-endothelial junctions and in limiting vessel permeability in the steady state [78]. It is conceivable that, following uptake of iRBCs, PVMs in their activated state can promote leaky tight junctions, and process and present Plasmodium antigen to CD8⁺ T cells. In this context, PVMs and vascular endothelial cells are integral parts of a single anatomical and functional unit that scavenges particulate matter from renal blood vessels and monitors trans-endothelial transport [79]. This interdependent, reciprocal vascular unit supports angiogenesis, macrophage differentiation, and the integrity of endothelial cell junctions [80]. Studies assessing the role of the PVM–endothelium unit are warranted in CM.

An important question that remains unanswered is the mechanism(s) that underlies the directed migration of antigen-specific (not antigen-ignorant) T cells towards focal areas of PVM localization along the luminal wall [26]. Of relevance, T cells transmigrating across the inner vessel wall in a highly constrained manner at APC sites in response to antigen recognition has been reported in EAE [81]. Such a mechanism could explain how focal chemokine gradients – created by PVMs at inter-endothelial junctions – may lead to antigen-specific T cell recruitment in the absence of a BBB breach [82]. Nuclear lobes of leucocytes can squeeze through junctions without rupturing endothelial stress fibres [83]. As a next step, process extensions of PVMs could provide a structural scaffold for guiding T cells along the abluminal wall. Regardless of what the mechanism might be in the early stage, disruption of endothelial tight junctions in later stages of infection will no doubt provide an access route for T cells to preferentially extravasate at circumscribed locations adjacent to PVMs.

(iii). Antigen presentation: parasite, endothelium, and PVMs

CD8⁺ T cells specific for PbA antigens migrate from the spleen to the brain, guided by chemokines such as CXCR3 [84]. CD8⁺ T cells must recognize antigen once again in the brain before killing can occur [85]; the latter happens within minutes once T cells receive a ‘license’ from an APC. Indeed, endothelial cells and parenchymal brain macrophages expressed higher levels of MHC class I during ECM [71]. The additional stimulus in the form of parasitic antigen to brain-infiltrated CD8⁺ T cells is provided by an APC, endothelium, or both, at which time cytotoxic enzymes are released. Mice deficient in granzyme B and perforin are protected against ECM [85]. Adoptively transferred wild-type CD8⁺ T cells could reverse this effect but only when a critical threshold of parasitic antigen was presented in the brain. These observations demonstrate the requirement for T cells to receive local antigen presentation signalling.

With respect to parasite antigens, iRBCs release an array of parasite factors, including waste products and mature merozoites upon completion of the cell cycle when the RBC bursts and merozoites egress (Box 1). In addition, iRBCs also secrete EVs that contain diverse cargo, including proteins, nucleic acids, regulatory microRNAs and lipids [86]. Because of iRBC sequestration via PfEMP1, these parasite products concentrate at the endothelium where they can directly act on endothelial cells through either activation of Toll-like receptors or inflammasomes. Thus, parasite products not previously considered to serve as parasite antigens may in fact contribute to HCM via an array of mechanisms (Box 1) [16,87]. Additionally, other proinflammatory cytokines, liberated in response to parasite factors, can induce endothelial activation, including the induction of iRBC receptor ICAM-1. This process, which is regulated by rapid nuclear translocation of p65 NFκB subunit, in turn would augment even further sequestration of iRBCs either directly or via platelet bridging [88], leading to microvascular obstruction (Figure 1), impairment of the endothelial barrier (Figure 2), and ultimately resulting in vascular leakage and microthrombi [89]. While the source of antigen cross-presented by brain endothelial cells has been demonstrated to be primarily of merozoite origin in ECM [76], a deeper understanding of how iRBC antigens, in addition to PfEMP1, are presented to the PVM-endothelium unit, and lead to inflammation, is critical for understanding how HCM arises and may inform strategies for adjunct therapy.

Due to their physical location at the interface between blood and surrounding brain tissue, endothelial cells are exposed early to blood-borne microbial components released in the circulation (Figure 4). They express a wide range of pattern-recognition receptors [90], allowing brain endothelial cells to act as early innate immune responders in P. falciparum infection and potentially contributing to CM pathogenesis [91]. Some of the aforementioned mechanisms (see Disruption of the BBB) are further amplified by the capacity of endothelial cells to act as semi-professional APCs [92], or to have macrophage-like functions [93], or uptake and present parasitic antigens during ECM [71,94,95] (Figure 4). Using luciferase-expressing Plasmodium strains PbAluc and PyYMluc, the ability of endothelial cells to cross-present a specific epitope (SQLLNAKYL) was shown to be reflective of the aptitude of parasites to sequester in the brain and cause ECM [95]. Although PbA sequestration does not lead to vessel occlusion, even a modest accumulation of iRBCs within the vasculature and the perivascular space may provide a localized source of antigen for cross-presentation by endothelial cells, leading to CD8⁺ T cell recruitment, proliferation, and activation [96]. CD8⁺ T cells can home onto their cognate antigen on or behind the BBB, a phenomenon dependent on luminal expression of MHC class I by cerebral endothelium [97]. iRBCs and/or parasite antigens can be accessed from the cerebral microvasculature through endothelial erythrophagocytosis [63], or trogocytosis, where antigens are extracted from the surface of iRBCs [98] (Figure 4). Both of these processes can lead to antigen presentation. An additional conceptual link between APC function, antigen presentation, and CM pathogenesis is provided by CXCL10 [99], which stabilizes T cell–brain endothelial cell adhesion during ECM and is a serum biomarker that predicted CM mortality in Ghanaian children [100,101].

Imaging applications in other models have begun to provide insights into when and where antigen presentation takes place in vivo, and for how long [66]. CD4⁺ T cells entering the brain have been shown to arrest along the vascular endothelium, where polarization of T cell receptor (TCR) and adhesion molecules occur [81], suggesting that these fixed points are sites of immune synapses. Following contact, antigen-specific T cells penetrate the CNS during EAE [64]. In peripheral lymphoid organs, the type of contact that occurs between APCs and T cells determines the functional outcome, with short contacts inducing tolerance and long contact activating T cells [102]. In this context, studies showed that PbA-primed CD8⁺ T cells first enter the brain in the early stage and arrest to inner vessel ‘hotspots’ either directly adjacent to, or in the near vicinity of, PVMs [26], the identity of which was established using reporter mice [103]. PbA-specific CD8⁺ T cells arresting to the endothelial wall and becoming activated in a cognate peptide-MHC class I-dependent manner during ECM have been reported [71]; however this study utilized non-reporter Kb^−/−Db^−/− mice and so, whether PVMs, additionally to endothelium, had a role in antigen presentation could not be ascertained [26]. Nevertheless, all the aforementioned studies collectively suggest that only CD8⁺ T cells that recognize cognate antigen are retained in the brain. T cells became aggressive in the late stages of ECM, their interaction rates increased at hotspots, potentially signifying their search for an exit into the perivascular space [26].