The central nervous system (CNS) is often considered as an immune-privileged organ because it is separated from the blood by specific cellular barriers. Nonetheless, pathogens, in particular neurotropic viruses, have been selected throughout evolution for their ability to reach the brain (1). Human pathologies following CNS viral infections can be due to direct virus invasion and elicited toxicity or indirectly by mediators of neuroinflammation. For instance, some virus-associated neuronal diseases can be due to direct infection of neurons or to indirect effects triggered by CNS-supporting cells and inflammatory mediators, causing damages and dysfunctions (paralysis, cognitive deficits, ocular problems) (2, 3). Different clinical symptoms may appear depending of the site of infection and/or inflammation: meningitis for inflammation of leptomeningeal structures, myelitis for the inflammation of the spinal cord, and encephalitis for inflammation of parenchymal brain tissue (4), the latter being most common upon viral neuroinfections. Viral brain infections are generally limited by in situ innate immune response of the host and by the physical protection exerted by the brain barriers. These are complex multicellular structures forming an endothelium or an epithelium that separates the systemic circulation from the CNS. Depending on their location, the brain barriers exert different functions but they mostly control diffusion between CNS and blood and allow a precise regulation of CNS metabolism and immunity.

Importantly, immune cells have an ambivalent role during viral brain infections, as they are actors of the antiviral response but can also be viral carriers into the CNS and therefore participate in virus transmission, neuroinflammation, and associated deleterious effects. Moreover, because of the presence of viruses or inflammatory mediators in the brain, the release of cytokines and chemokines will further allow transmigration of immune cells such as monocytes, T lymphocytes, natural killer cells (NK), and dendritic cells (DCs) (5–7).

Immunosurveillance is provided by specific resident and incoming immune cell subsets. Their distribution varies depending on the nature of the epithelium and the inflammatory state of the tissue (8, 9). Among these immune cells, DCs are the very first responders following viral infection, as they are at the frontier between innate and adaptive immunity and acting as sentinels of the immune system, notably in the site of potential infection, including in the CNS. In this review, we try to shed some light on the pathogenesis of viral brain infections with a particular interest in the interactions with brain barriers and a focus on the role of DCs during viral brain invasion.

Different viral CNS access mechanisms have been characterized, involving a neuronal-mediated spread mechanism (through axonal transport) and a hematogenous-mediated viral entry route (through interaction with brain barriers) ( Table 1 ). Notably, entry mechanisms and viral tropism may determine neurological symptoms and clinical outcome (42, 43).

Immunosurveillance is provided by specific resident and incoming immune cell subsets. Their distribution varies depending on the nature and the inflammatory state of the tissue (8, 9). Among these immune cells, DCs are the very first responders following viral infection, as they are at the frontier between innate and adaptive immunity and acting as sentinels of the immune system, including in the CNS.

DCs include different subsets with specific cellular and immunological properties: the myeloid/conventional DCs (DCs), the Langerhans cells (LCs) that are the unique DC subset located in mucosal stratified (intestine and oral mucosa) and pseudostratified epithelium (example: lung) as well as skin epidermis (8, 77), and the plasmacytoid DCs (pDCs), which are unconventional DCs characterized by their ability to produce large amounts of type I interferon (IFN-I) in response to viral pathogens (78, 79) ( Table 2 ).

DC subsets originate from bone marrow CD34⁺ hematopoietic stem cells giving rise to common myeloid/lymphoid progenitors (CMLPs), which then differentiate toward common myeloid progenitor (CMP) or common lymphoid progenitor (CLP). Under specific environmental cytokine conditions, concomitantly with the expression of defined transcription factors, CMPs separate from the monocyte/macrophage axis and generate common DC progenitors (CDPs), which will give rise to pDCs and conventional DC subsets (8, 80). Various DC precursors (Pre-DCs) with selective functional features are characterized by the unique expression of AXL and SIGLEC6 along with myeloid and plasmacytoid markers like CD11c and CD123 (81, 82). However, this was recently challenged, since DC (Axl⁺ Siglec 6⁺), named AS DC, was described in 2017 as a potential new functional DC subset instead of being an exclusive pre-DC progenitor (81). Nevertheless, DCs can be divided into three major DC subsets, pDC, conventional DC1 (cDC1), and conventional DC2 (cDC2), but other cell phenotypes can arise from those subsets particularly when considering tissue-resident DCs and inflammatory status (8, 80). Interestingly, LCs, which represent a DC subset exclusively present in the epidermis and upper mucosal layers, were shown to be functionally related to DCs but ontogenically closer to macrophages due to their reported embryonic origin and self-renewing capacities (83, 84). Indeed, the ontogeny of LCs has always been subject to intense debate, and new transcriptomic data suggest that LCs are rather closely related to tissue-resident macrophages. Nevertheless, LCs were recently reported to share ontogeny and transcriptional features with both macrophages and DCs (85). From a functional and phenotypic point of view, LCs have strong similarities with DCs (86). For example, LCs have the ability to capture pathogens and possess migratory abilities to initiate T-cell responses in lymph nodes. Moreover, LCs can be generated ex vivo from hematopoietic precursors in response to cytokines and cell membrane-associated ligands. Thus, LCs were found to share both macrophage and DC characteristics (85, 87), as particularly evidenced upon skin damage and inflammation or infection processes in vivo. All the current debates based on the different reported cell or animal models emphasize the importance to consider functional lineage plasticity besides ontogenetic data. Also, as LCs, other DC subsets were reported to develop under certain conditions like monocyte-derived DC (MoDC), arising in inflamed tissues and reminiscent of reported inflammatory DC subsets (88, 89). Importantly, the myeloid DC lineages (cDC1 and cDC2) are clearly distinguished from MoDCs, which are usually derived from monocytes predominantly under inflammatory conditions in vivo and in vitro ( Table 2 ) (90), although the identification of new DC subsets evolves with the recent identification of new human anogenital MoDCs present in non-inflamed tissues that take up HIV and transmit virus to CD4 T cells (91).

DCs are highly specialized professional APC populations through the expression of major histocompatibility complex (MHC) class II and costimulatory molecules that lead to naive T-cell stimulation (91–94). They have the capacity to capture antigen in the periphery and in the subepithelium to migrate to proximal lymph nodes where they prime naive T cells and engage the adaptive immune response. These cells express various pathogen sensors, including many specific PRRs and toll like receptors (TLRs), allowing major secretion of type I IFN that induces IFN-stimulated genes (ISGs) used to counteract infection (8, 95–97). However, their capacity to capture pathogens is mainly due to their expression of lectin and lectin-like receptors such the sialic acid-binding immunoglobulin (Ig)-like lectins (Siglecs) family (98) and especially C-type lectins receptors (CLRs). CLRs bind carbohydrate structures associated with viruses, fungi, or bacteria expressed by LCs, Dendritic cell-Specific Intercellular adhesion molecule-3-Grabbing Non-integrin (DC-SIGN) (CD209) and the CLR DC-associated C-type lectin-2 (Dectin-2/CLEC6A) expressed by DCs and macrophage, and galactose type C-type lectin (MGL, CD301) expressed by both DCs and macrophages (but not LCs), respectively (99, 100). Although CLRs have established antiviral functions, many viruses are capable of hijacking these receptors to their advantage. In particular, langerin escape or its hijacking by a virus seems to be much rarer in contrast to DC-SIGN that has been shown to bind the gp120 glycoprotein of HIV-1 and promotes efficient trans-infection of CD4⁺ T cells. DC-SIGN can also be used to infect certain DCs with many viruses, such as cytomegalovirus or Ebola virus but also several flaviviruses, including dengue virus and WNV (101). Thus, some DC subtypes may constitute the point of access of viruses, allowing their subsequent transmission and propagation through the body including the CNS. Although the contribution of different DCs is not clearly established, they may also represent a “Trojan horse” because they can migrate efficiently to the brain.

DCs are naturally present within the brain where they act as sentinels under steady-state conditions and during brain disease and neuroinflammation (102). They have been shown to efficiently interact with brain barriers in order to access the CNS.

During the leukocyte recruitment at the BBB, immune cells including DCs interact with various actors. First, cytokines and chemokines, released during CNS inflammation such as CCL2, can chemoattract circulating DCs at brain barriers. Then, contact between the upregulated cellular adhesion molecules (CAMs) P-selectin, E-selectin, vascular cell adhesion molecule-1 (VCAM-1), and intercellular adhesion molecule-2 (ICAM-2) on endothelial cells with P-selectin glycoprotein ligand-1 (PSGL-1) and DC-SIGN on DCs allows rolling and adhesion on the BBB apical surface (103). These steps are followed by potential transmigration, facilitated by interaction of DC-SIGN with CAMs or tight junction proteins. Interestingly, DCs can also express tight junction proteins to facilitate this step without a deleterious effect on the BBB integrity (103) ( Figure 3A ). Importantly, neuroinfiltration may depend on the DC state. Indeed, during activation and arrest steps, migration of mature DCs across the BBB can be enhanced by interaction of CCR7 with CCL19 and CCL21 ( Figure 3B ), whereas immature DCs can also be attracted by CCR2, CCR3, and CCR5, which interact with CCL2, CCL3, and CCL5, respectively ( Figure 3C ) (6, 104). Interestingly, immature DCs have better capacity to cross the BBB than mature DCs in part because of the different expressions of specific traffic signals at their surface (105, 106). In vitro, the transmigration of granulocyte-macrophage colony-stimulating factor (GM-CSF)-induced DCs across the BBB was shown to be increased in the presence of CCL3. In addition, GM-CSF-matured DCs secrete metalloproteinases, allowing transmigration across the BBB with redistribution of tight junction molecules (107). Regarding the BCSFB, molecular mechanisms of DC interaction are less well-defined (108, 109). As in the blood compartment, DCs are mainly recruited in the CSF by chemoattraction through the release of CCL2, CCL3, CCL4, and CXCL12, the latter of which being critical during this recruitment step (110). Following increased concentration of DCs in the CSF, they can interact by the expression of CAMs on their surface and on the surface of choroid plexus epithelial cells (5). All these interactions support BCSFB immunosurveillance functions with DC’s high APC role and recruitment in brain parenchyma.

Brain DCs represent 1% of immune cells found in the brain and are sparsely found in the CSF (less than 1% of CSF cells) (5, 111, 112). Macrophages and DCs are found naturally in non-pathological conditions in the choroid plexus stroma (113), in sites of plasticity and neurogenesis, and where the BBB is absent to provide immunosurveillance (114). Through their innate properties, brain DCs can activate encephalitogenic T cells (self-antigen-reactive T cells) or release pro-inflammatory cytokines to the detriment of the CNS homeostasis; their loss leads to dysregulation of immune tolerance and excessive inflammatory response (115). Consequently, in a context of adaptive immune response, DCs are important potential inducers of primary T cells (113). For example, they are among the APCs, present in dural sinuses, one compound of the meninges, and they allow the presentation of antigen from CSF to T cells in cervical lymph nodes (115, 116).

During neuroinflammation, the number of DCs increases in the CNS and particularly in the CSF where they can more easily reach lymph nodes to activate the immune response. They participate indirectly to exacerbate neuroinflammation (74, 110). For example, they are involved in autoimmune diseases as in the murine model of multiple sclerosis (MS) named experimental autoimmune encephalomyelitis (EAE) (117). MS is a chronic CNS inflammatory disease, often studied with the EAE mouse model, based on the reactivity of T cells to endogenous myelin epitopes (encephalitogenic T cells). At the EAE onset, DCs from the CNS support inflammatory T cells at the peak of the disease, then become poor APC and prime the development of regulatory T cells (118–120). Notably, it was shown that DCs promote myelin-reactive T cells during the reactivation phase of EAE (121) leading to the proposition to target DCs for therapeutic approaches in MS. The accumulation of DCs during the neuroinflammation translating the EAE is becoming better understood. These immune cells can be mainly from peripheral bone marrow-derived precursors or from activation of resident microglia that will show DC characteristics, reflecting the wide range of DC precursors (120, 122). Another illustration is the role of DC during Guillain–Barré syndrome. They are the main APCs in the CSF, whether they are myeloid or plasmacytoid; they can present auto-antigen from spinal nerve, activate T and B cells, and finally lead to the autoimmune disease (123, 124).

Neuroinflammation can also be due to CNS injuries such as strokes or seizures. In these cases, different DC subpopulations can be recruited due to the release of various cytokines and chemokines by damaged cells. cDCs and pDCs have different sensitivities to chemokines according to their receptor expression (125). Studies have found peripheral derived DCs in the necrotic zone after stroke injury, but also resident DCs associated with T cells in the penumbral zone where they have a more immunomodulatory function after activation probably due to release of IFNγ (126). Others have shown the presence of DCs from bone marrow-derived population that exert pro-inflammatory actions with the production of cytokines after ischemic lesions (120, 127). Epilepsy is a consequence of infection or autoimmune disease, and during extended seizure, there are glial cell activation and upregulation of CAM expression on endothelial cells (128, 129). Also, in epileptic experimentation on rats, DCs can be found in blood vessels until 24 h after the seizure (130). In humans, the study of chronic epileptic patients has found the presence of DCs and T cells in blood vessels (131). In this epileptic context, DCs seems to maintain the chronic inflammation, source of seizures (132).

DCs are also important players during brain infections. During protozoal infection of the brain, caused by Toxoplasma gondii, DCs are major factors of neuroinflammation. First, the intracellular parasite accesses the brain through “Trojan horse” mechanisms using migratory leukocytes, including DCs (120, 133, 134). Next, during T. gondii-induced encephalitis, DCs found in the brain are mature APCs, produce pro-inflammatory interleukin (IL)-12 cytokines, and can provide Th1 response from CD4⁺ T cells (135, 136). Here, DCs participate in the initiation of immune response against the parasite, but they also maintain neuroinflammation and can contribute in T. gondii-induced encephalitis chronicity. For bacterial meningitis, the more severe CNS infection etiology, DCs are in high concentration in patient CSF. They lead to neuroinflammation by Th1 immune response induction, but they also exert a key role in the regulation of the host immune evasion of some bacterial strains such as Escherichia coli K1 (120, 137).

In conclusion, it appears that 1) viral brain infection can be mediated by several mechanisms and lead to neuroinflammation with various consequences including leukocyte recruitment; 2) as a result, immune cells and particularly DCs can interact with brain barriers and invade the CNS to act against the viral infection; 3) they can also support neuroinflammation; 4) and they can be a “Trojan horse” during virus CNS entry. Indeed, a viral infection can lead to general inflammation as seen with a cytokine storm that can have a deleterious effect on the CNS with chronic neuroinflammation leading to various sequelae (171, 172). This inflammation, as well as by DC innate properties, can lead to recruitment of DCs in the CNS (125), sometimes also carrying the virus itself in the brain (40, 72). Once in the CNS, DCs can directly act against viruses, for example, with IFN production (173), they also contribute to neuroinflammation by T-cell activation for instance (170). For each of these conditions, different DC populations can be recruited (120). However, to tackle these pathological manifestations, we need a better characterization of CNS inflammation following infections. Thus, there is a major need to better dissect the role of each cell type in the establishment of neuroinflammation processes by considering both neuronal and immune cells including DCs. Novel technologies allowing quantitative single-cell mRNA sequencing and proteomic profiling of inflammatory markers may reveal specific DC characteristics linking these cells with the establishment of inflammation and CNS disorders. These data will critically contribute to deepen our current knowledge on brain viral infection. Moreover, DCs are known for their role in priming T-cell responses. It will be, therefore, of interest to further characterize how DC-dependent antiviral T responses can lead to neuroinflammatory pathologies and what are the antigenic drivers of such conditions. Clarifying the role of DCs as major players in neuroinflammation as well as deeper investigations on how they are carrying viruses to distal organs will undoubtedly render the possibility to consider them as a therapeutic target in order to rapidly control the possible pathological outcomes of infections. Gathering more data on metabolic and immunobiological features of DC subsets at sites of primary viral infection will also allow to transpose and adapt novel antiviral therapies, for example, by a topical drug-based application targeting early events of skin DC infection.

Moreover, a flavonoid, apigenin, was proposed to regulate inflammation by a reduction of α4-chain expression by DC and limiting the BBB cross (174). In the context of MS, there are some studies as the use of prestimulation of DC TLR9 to increase neuroinflammation control and immune regulation of the course of the disease in EAE (175). The use of DC-vaccine was also proposed to support the treatment of neurodegenerative disease as Alzheimer’s disease (176). Nonetheless, in the context of viral brain infection, there is a lack of evidence for the use of DC in care. Also, the regulation or the blocking of DC entry into the CNS should be explored, for example, to limit viral carrying. Increasing work on different subsets of DCs depending on the viral infection etiology could be a key issue to understand and treat neuroinfections that are considerable public health concerns.

SS conceived the study. OC, GM, FB, YS, and SS wrote the review. PVD corrected and amended the article. All authors contributed to the article and approved the submitted version.

This work was publicly funded through ANR (the French National Research Agency) under the “Investissements d’avenir” program with the reference ANR-16-IDEX-0006 » and by la Région Occitanie through the PhD funding program with the reference R19068FF.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.