Safeguarding the sense of smell
Given the proximity of the olfactory bulb to the upper airways, it is surprising that viruses that infect the upper airways, including influenza, rarely infect the brain. In mice, intranasal infection with vesicular stomatitis virus (VSV) does result in infection of sensory neurons in the olfactory bulb. Using VSV that expresses a fluorescent reporter, Moseman et al. examined the ability of VSV to infect distinct cell types within olfactory bulb. Although VSV infection was restricted to neurons within the olfactory bulb, they found that microglial cells that were not infected by VSV to be the key in priming T cell responses that promoted viral clearance.
The neuroepithelium is a nasal barrier surface populated by olfactory sensory neurons that detect odorants in the airway and convey this information directly to the brain via axon fibers. This barrier surface is especially vulnerable to infection, yet respiratory infections rarely cause fatal encephalitis, suggesting a highly evolved immunological defense. Here, using a mouse model, we sought to understand the mechanism by which innate and adaptive immune cells thwart neuroinvasion by vesicular stomatitis virus (VSV), a potentially lethal virus that uses olfactory sensory neurons to enter the brain after nasal infection. Fate-mapping studies demonstrated that infected central nervous system (CNS) neurons were cleared noncytolytically, yet specific deletion of major histocompatibility complex class I (MHC I) from these neurons unexpectedly had no effect on viral control. Intravital imaging studies of calcium signaling in virus-specific CD8+ T cells revealed instead that brain-resident microglia were the relevant source of viral peptide–MHC I complexes. Microglia were not infected by the virus but were found to cross-present antigen after acquisition from adjacent neurons. Microglia depletion interfered with T cell calcium signaling and antiviral control in the brain after nasal infection. Collectively, these data demonstrate that microglia provide a front-line defense against a neuroinvasive nasal infection by cross-presenting antigen to antiviral T cells that noncytolytically cleanse neurons. Disruptions in this innate defense likely render the brain susceptible to neurotropic viruses like VSV that attempt to enter the CNS via the nose.
Viral infections of the central nervous system (CNS) can be devastating when not properly contained (1, 2). Because the CNS contains irreplaceable postmitotic cells, it is protected by several physical barriers that limit pathogen access into the CNS, including the blood-brain barrier (BBB), blood cerebrospinal fluid barrier (BCSFB), and skull, among others. In addition, immune responses in this compartment are heavily regulated (3). Viruses, in turn, use several approaches to bypass these barriers such as direct infection of the BBB, invasion of peripheral nerves followed by transport into the CNS, and “trojan horse” entry via surveying immune cells (4). One especially vulnerable route that viruses use to invade the CNS is via olfactory sensory neurons (OSNs) within the nasal cavity.
OSNs lie within the mucosal upper airway surface, which is constantly exposed to environmental pathogens. However, the olfactory epithelium (OE) lining the nasal turbinates is unique in that this mucosal surface provides access for viruses to enter the CNS. Unlike the neighboring respiratory epithelium, the OE contains thousands to millions of OSNs (depending on the species) that are the predominant cell type within the olfactory neuroepithelial surface. Although the OSN cell bodies lie beneath a layer of supporting or sustentacular cells, they extend a ciliated dendrite into the mucus lined airway space. Odorant information gathered from the external environment is conveyed via OSN axons within the turbinates through the specialized cribriform plate at the front of the skull and into the olfactory bulb (OB) of the brain (5). However, this anatomical arrangement also results in OSNs serving as a direct single-cell route for neuroinvasion. Pathogens that infect OSNs can be shuttled intracellularly along OSN axons directly into the brain. The intracellular passage via OSNs into the brain allows invading pathogens to “tunnel under the castle wall” and evade classical CNS barriers that typically shield the brain. Thus, the olfactory route of infection is especially vulnerable to neurotropic viruses (6, 7).
The immune response to viruses must be appropriately balanced between pathogen clearance and limiting tissue damage. This balance is especially important in the CNS because most neurons are unable to regenerate, and damage can result in permanent damage to neural networks (1). Although virus-induced cytopathology poses a serious concern to the CNS, immune-mediated cellular damage via perforin/granzyme poses a similar threat to neuronal integrity. Therefore, noncytolytic viral clearance via cytokines, especially interferons (IFNs), are believed to play an important role in CNS viral control (8–10). Type I IFN (IFN-I) is critical to the innate defense against most viruses (11) and also plays an important role in priming adaptive immune responses (12, 13). On the other hand, IFNγ, which is primarily produced by lymphocytes [CD8+, CD4+, and natural killer (NK) cells], plays a dominant role in the control of CNS viral infections (14–17). Tumor necrosis factor–α (TNFα) has also been shown to facilitate noncytolytic clearance (18–20).
Upper airway infections are a ubiquitous part of human life, and although many viruses infect the respiratory system (e.g., influenza, rhinovirus, coronavirus, and respiratory syncytial virus), several of these viruses can also infect the OE. In this study, we used vesicular stomatitis virus (VSV) as a model for nasal infection because of its ability to target the OE preferentially over the respiratory epithelium (21)—a preference shared by herpes simplex virus (HSV) (22). Several other viruses can also infect the OE, including mouse hepatitis virus (MHV) (23), parainfluenza (24), Japanese encephalitis virus (JEV) (25), and influenza virus (6, 26). Whereas seasonal influenza is typically an infection of respiratory epithelium (27), some more pathogenic strains are clearly able to target OSNs (26). Highly pathogenic avian influenza H5N1 prefers to replicate within the OE (28).
We undertook this study to better understand the mechanisms by which a cytopathic virus that infects the OE is contained and ultimately prohibited from spreading throughout the CNS. Intranasal VSV infection can induce a potent encephalitis that varies in outcome depending on the age and strain of the infected animal (29). After nasal infection, VSV is transported from the airway via the OE into the CNS, where it is initially observed within the OB (30). Virus replication is usually halted at the glomerular layer of the OB where OSN projections enter (10); however, the virus can sometimes move into deeper regions of the OB and throughout the CNS. VSV that escapes this OSN-glomerular junction is believed to use retrograde axon transport and the ventricular system to move caudally in the CNS (31). It is known that IFN-I (32, 33), B cells (33, 34), and antiviral T cells (29, 35) all play a role in containing VSV. In this study, we monitored the in vivo activities of CD8+ T cells in the anatomical site where VSV is typically contained within the OB (glomeruli) and probed for the most functionally relevant source(s) of antigen presentation that result in noncytolytic VSV control in neurons. Unexpectedly, an innate immune cell that is never infected by the virus (microglia) served as a key player in the control of virus in neurons.
Virus infection of the olfactory epithelial leads to rapid neuroinvasion
To more precisely understand how neurotropic viruses invade the CNS after infection of the OE, we inoculated mice intranasally with VSV expressing DsRed (VSV-DsRed). This approach enabled us to monitor the distribution and spread of the virus in olfactory tissue. Confocal imaging of coronal whole head sections (fig. S1) revealed 24 hours after intranasal infection that there was widespread viral replication within OSNs in the OE (Fig. 1, A and C). After intranasal inoculation, VSV moves from the OE into the OB where it is typically halted at the glomerular layer (10). At day 6 after infection, we observed robust VSV-DsRed signal in outer nerve layer, which are peripheral nerves that project into the OB (Fig. 1, B and D). We also observed virus in the OB glomerular layer, indicating invasion of the brain parenchyma (Fig. 1, B and D). Quantification of viral titers in the OB revealed that VSV was very efficient at gaining access to the brain via this intranasal route (Fig. 1E). Viral titers reached ~104 plaque-forming units (PFU) at 24 hours and climbed to ~105 PFU on day 6 before being extinguished by day 8.
To protect the CNS, noncytolytic viral clearance is often favored in this compartment (36). However, because VSV is a cytopathic virus, it was unclear whether this lifecycle could be halted once initiated. To determine whether a cytopathic virus could be cleared noncytolytically from the brain, we generated a recombinant of VSV expressing iCre-recombinase (VSV-iCre). We then conducted fate-mapping studies by infecting floxed Stop TdTomato reporter (Stopfl/fl TdTomato) mice with this recombinant virus. We observed dense patches TdTomato+ cells with a definitive neuronal morphology in the OB at late time points (>40 days) (Fig. 1F and fig. S2), long after infectious virus was cleared from this compartment (Fig. 1E). These data indicate that VSV can be noncytolytically cleared from the OB, although it does not exclude the possibility that some cells are cytolytically cleared as well.
Intranasal VSV infection induces a robust immune response in the olfactory system
Except for patches of bone marrow within the turbinate bones, the naïve OE as shown in coronal head sections (fig. S1A) contains few hematopoietic cells (Fig. 2A); however, after nasal infection, widespread VSV replication within the OE provoked robust inflammatory cell infiltration. CD45+ leukocytes were observed in the OE as early as day 2, and infiltration elevated massively by day 6 after infection (Fig. 2A). At this time point, virus and leukocytes were found in OE and in the airways. This was associated with substantial OE degradation and dissociation from the lamina propria (Fig. 2, A and B).
Under steady state, the OB is defended in part by resident microglia; however, the arrival of VSV into this compartment via OSNs is associated with a large influx of peripherally derived immune cells that surrounded infected glomeruli (Fig. 2, A and C to E). This response was dominated by neutrophils and NK cells during the first 3 days, which were superseded by CD8+ T cells and Ly6Chi monocytes at day 6 after infection (Fig. 2E). CD4+ T cells also arrived at this time point. At day 8 after infection, T cell and monocyte numbers increased further within the OB (Fig. 2E). To track the recruitment of virus-specific CD8+ T cells to the OB, we seeded mice intravenously with 3000 mCerulean+ OT-I and then infected them intranasally with VSV–ovalbumbin (OVA). VSV-specific CD8+ T cells were primed and proliferated in the draining cervical lymph nodes and spleen and increased steadily over time in the OB (Fig. 2F). These data demonstrate that the OB mounts a robust immunological defense against intranasal VSV infection.
T cells and antiviral effector molecules protect against fatal VSV encephalitis
VSV titers in the OB are controlled between days 6 and 8 after infection (Fig. 1E), which coincided with the arrival of antiviral T cells (Fig. 2, E and F). We therefore sought to better understand the role of these T cells in preventing the fatal spread of VSV from olfactory glomeruli throughout the brain. We used survival rather than viral titers as an initial screen because we wanted to identify components of the OB antiviral defense that prevent induction of fatal disease after intranasal infection. Consistent with a previous study that used Balb/c mice (nu/nu) to demonstrate a role for T cells in controlling VSV (29), we found that a fraction of C57BL/6J mice depleted of either CD8+ or CD4+ T cells developed fatal VSV encephalitis (Fig. 3A). In addition, depletion of both T cell subsets induced fatal disease in nearly all mice infected intranasally with VSV. Consistent with this observation, we found increased viral titers in the OB (Fig. 3B) and caudal brain (Fig. 3C) of T cell–depleted mice at day 8 after infection. In addition, although VSV–green fluorescent protein (GFP) was largely confined to the outer nerve layer and glomeruli of control mice (Fig. 3, D and F), T cell depletion promoted movement of VSV-GFP to other brain regions, including the rostral migratory stream and brainstem (Fig. 3, D and F). T cell depletion also allowed VSV-GFP to invade deeper layers of the OB, such as the mitral cell layer (Fig. 3E).
We next examined the effector mechanisms involved in preventing fatal VSV encephalitis. Mice deficient in TNFα, IFNγ, or perforin (PRF1) all showed modest reductions in survival comparable with that observed in mice depleted with anti-CD8 or CD4 antibodies (Fig. 3G). None of these knockout mice phenocopied the incidence of fatal disease observed in mice receiving anti-CD8/4–depleting antibodies (Fig. 3A). To determine whether expression of the IFNγR was required on infected cells, we inoculated floxed IFNγR with VSV-iCre (Fig. 3H). This, however, did not reduce survival, indicating that IFNγR is not required on infected cells to elicit antiviral control. Collectively, these data suggest that there are redundant effector mechanisms controlling VSV within the OB.
Antiviral CTLs have slowed motility within the infected OB
Although it is likely, based on our depletion studies, that CD4+ and CD8+ T cells exert complementary antiviral programs after infection (Fig. 3A), we decided to focus in more detail on the CD8+ T cell compartment because their major histocompatibility complex class I (MHC I) restriction offered the possibility of studying a broader selection of potential target cells. We initially examined VSV-specific versus bystander CD8+ T cell dynamics in the living OB at day 6 or 7 after infection. We selected these time points because viral titers in the OB declined rapidly between days 6 and 8 after infection (Fig. 1E), and we wanted to understand the mechanism by which this occurred. One day before VSV-OVA infection, B6 mice were seeded with 3000 mCerulean+ OT-I cells that served as the virus-specific CD8+ T cell population. Bystander cells were generated by seeding a separate group of B6 mice with orange fluorescent protein–positive (OFP+) P14 cells, which are specific to the glycoprotein of lymphocytic choriomeningitis virus (LCMV) (37). These mice were then infected intraperitoneally with LCMV Armstrong, and at day 7, effector OFP+ P14 cells were harvested from the spleen and transferred into VSV-GFP-OVA–infected mice 1 day before imaging. mCerulean+ OT-I (VSV-specific) and OFP+ P14 (bystander) cells were imaged simultaneously by intravital two-photon microscopy (TPM) through a thinned skull preparation (38, 39) made above the OB. This technique allowed us for the first time to analyze antiviral T cell behavior in the virally infected OB. After VSV-GFP-OVA infection, virus-specific CD8+ T cells moved rapidly along, between, and across infected axon tracts in the outer nerve layer and within the glomeruli (Fig. 4, A and B, and movies S1 and S2). Quantification of T cell track velocities in two-photon time lapses revealed that virus-specific CD8+ T cells had reduced track velocities relative to bystander T cells imaged in the same OB (Fig. 4, C and D). In addition, virus-specific CD8+ T cells also had a lower motility coefficient and higher arrest coefficient (Fig. 4D), all suggesting that these cells were encountering antigen locally (40).
CTLs engage cognate antigen on an uninfected brain-resident cell population
To more accurately assess antigen encounters by virus-specific CD8+ T cells in vivo, we devised an approach to monitor T cell calcium flux by TPM. T cell recognition of cognate peptide MHC complexes results in a rapid increase in intracellular calcium levels (41). We therefore generated OT-I and P14 cells expressing mTomato (to label the plasma membrane) and GCaMP6s—a fluorescent calcium concentration indicator protein (13). Six days after intranasal infection with VSV-OVA, calcium flux was visible in virus-specific CD8+ T cells that had infiltrated and were surveying the OB (Fig. 5, A, B, and D, and movie S3). Calcium flux was rarely observed in infiltrating bystander P14 CD8+ T cells, indicating that T cell calcium flux is a good surrogate for cognate peptide MHC recognition (Fig. 5B) (13). We also observed that calcium fluxing virus-specific OT-I CD8+ T cells had reduced instantaneous velocities relative to the cells that were not fluxing (Fig. 5C), which is consistent with studies showing that T cell antigen recognition is associated with decreased motility (40).
We next classified virus-specific CD8+ T cell interactions in the virally infected OB into two groups based on velocity. “Kinetic” interactions were defined by continued T cell motility while fluxing calcium, whereas “stable” interactions were characterized by a relative loss of motility (<2 μm/min) during calcium flux (Fig. 5, E and F, and movie S3) (42). Most antigen encounters (65%) by virus-specific CD8+ T cells were defined as kinetic interactions. These cells did not arrest during peptide MHC recognition but paused briefly before continuing their migration through the OB (Fig. 5, E and F, and movie S3). Virus-specific CD8+ T cells also engaged in more stable interactions where the T cell would fully arrest and then flux calcium; however, this was only observed in 35% of the antigen encounters (Fig. 5, E and F, and movie S3). Stable interactions typically resulted in more calcium fluxes (Fig. 5G) and longer flux durations relative to the kinetic interactions (fig. S3).
Having defined the dynamics of virus-specific CD8+ T cell interactions in the infected OB, we next sought to define target cell identity. In our imaging studies, we often noted that T cells would flux calcium upon engagement of “autofluorescent” cells (Fig. 5, A and D, and movie S3). Morphologically, these autofluorescent cells did not match the size, shape, or spatial distribution of VSV-GFP infected cells that we typically observed in the OB. Because hematopoietic cells massively infiltrated the OB after VSV infection (Fig. 2), we posited that an infiltrating cell population might be responsible for presenting local antigen to virus-specific CD8+ T cells. To address this possibility, we generated bone marrow chimeras in which the CNS-resident compartment lacked MHC I, whereas the infiltrating hematopoietic compartment was MHC I sufficient. The percentage of donor-derived hematopoietic cells in the blood of these mice was 99.2 ± 0.4%, demonstrating near-complete chimerism. The overall frequency of cytotoxic lymphocyte (CTL) calcium flux dropped significantly in mice lacking MHC I on the CNS-resident compartment, indicating that antigen presentation to virus-specific CD8+ T cells relied primarily on a radiation-resistant brain cell population rather than infiltrating hematopoietic cells (Fig. 5H). VSV is known to infect neurons, but MHC I is expressed at very low levels on these cells (43, 44). We nevertheless set out to evaluate the importance CTL-neuronal interactions in controlling an intranasal VSV infection. To address this question, we generated a conditional β2-microglobulin (β2M) mouse in which the exons 2 and 3 are flanked by loxP sites (floxed β2M) (fig. S4A). β2M is an essential component of the MHC I surface complex, and thus, removal severely impairs presentation of class I peptides (45). We infected “conditional β2M” mice with VSV-iCre to remove the possibility that infected cells engage in direct antigen presentation to CD8+ T cells. Infection with VSV-iCre resulted in deletion of β2M from the genome of infected cells (fig. S4B). However, unexpectedly, VSV-iCre–infected floxed β2M mice survived intranasal infection at the rate of control animals (Fig. 5I), indicating that the antiviral pressure exerted by CD8+ T cells, which is required in part for survival, does not involve direct interactions with virus-infected cells.
Microglia within the infected OB are activated and present antigen to CTLs
Microglia are radiation-resistant CNS-resident myeloid cells, and although they are not believed to be potent antigen-presenting cells in the naïve state (46), we characterized the surface expression of antigen-presenting molecules on microglia during VSV infection (Fig. 6, A and B). Six days after VSV infection, microglia in the OB showed evidence of activation characterized by robust up-regulation of MHC I, CD80, and CD86 as well as a very modest increase in CD40 and MHC II relative to naïve microglia (Fig. 6, A and B). These changes in antigen-presenting molecules suggested that microglia were capable of presenting viral antigen to CD8+ T cells.
We next sought to directly observe in vivo whether microglia were presenting antigen to virus-specific CD8+ T cells. To accomplish this, we adoptively transferred naïve mTomato+ GCaMP6s+ OT-I cells into bone marrow chimeras in which CX3CR1gfp/+ were reconstituted with wild-type (WT) bone marrow (99.6 ± 0.31% blood chimerism). In these chimeras, radioresistant microglia are GFP positive, whereas all peripherally derived hematopoietic cells as well as perivascular and meningeal macrophages are GFP negative. Six days after intranasal infection with VSV-OVA, we used intravital microscopy to identify calcium flux induced by antiviral CD8+ T cell engagement in the OB (movie S4). We then quantified the frequency of CD8+ T cell antigen recognition (calcium flux) occurring in contact with microglia versus unlabeled space within the OB. This study revealed that ~72% of CD8+ T cell calcium flux occurred in contact with a CX3CR1gfp/+ microglia (Fig. 6, C and D, and movie S4), suggesting that microglia are indeed presenting cognate peptide MHC I complexes to virus-specific T cells. The interaction partner for the remaining CTL calcium flux events (~28%) is not known; however, we noted that the within the OB of the bone marrow chimeras, 23.4 ± 3.9% of the microglia were nonfluorescent (i.e., donor-derived). These cells might also have induced T cell calcium flux.
Microglia acquire infected OSN debris and orchestrate neuroprotective CTL activity
Our calcium imaging data suggested that microglia play an important role in presenting antigen to antiviral CD8+ T cells after VSV neuroinvasion. Neurons are the only known target of VSV after nasal infection; however, we decided to evaluate whether microglia were also infected. We sought evidence for microglia infection in two ways—by inoculating WT mice with VSV-GFP (fig. S5A) and by inoculating Stopfl/fl TdTomato mice with VSV-iCre (fig. S5B). Using either approach, flow cytometric analyses of the OB from infected mice revealed no evidence of microglia infection despite an abundance of virus in CD45-negative cells (presumably neurons) (fig. S5, A and B).
The absence of direct microglia infection led us to investigate alternative mechanisms of antigen acquisition and presentation by microglia. VSV can cause damage to the neurons it infects, and microglia are known to phagocytose cellular debris. We therefore evaluated whether microglia acquired antigen from VSV-infected OSNs. We labeled OSNs within the OE by injecting fluorescent cholera toxin B (CTB) intranasally into olfactory marker protein-GFP (fig. S5C) or CX3CR1gfp/+ mice (Fig. 7A) (47). Ten hours later, after CTB was transported along OSN axons into OB glomeruli (fig. S5C), we infected mice with VSV and, 6 days later, observed evidence of CX3CR1gfp/+ cells with intracellular CTB obtained from the labeled OSNs (Fig. 7A). These cells had the morphological characteristics of microglia, and this was confirmed flow cytometrically, which revealed that ~5% of microglia in the OB contained CTB (i.e., OSN debris) 6 days after infection (Fig. 7, B and C). Processing the entire OB for flow cytometry, however, did not allow us to specifically analyze the anatomically relevant microglia in the glomerular layer of the OB, which is where VSV (Fig. 2C) and CTB (fig. S5C) enter via OSN axons. We therefore quantified uptake histochemically, allowing us to focus on this anatomical site. We found 6 days after VSV infection that ~40% of Iba-1+ myeloid cells in OB glomeruli contained intracellular CTB derived from OSNs (Fig. 7D). For these studies, CTB served as a surrogate for an OSN-derived antigen. We next wanted to address whether microglia could also acquire virus particles. This was accomplished by infecting mice with VSV-PeGFP—a viral recombinant in which the phosphoprotein (P) is fused to enhanced GFP (eGFP) enabling detection of the virion itself (48). Six days after infection with VSV-PeGFP, we determined that ~10% of Iba-1+ myeloid cells in OB glomeruli contained VSV-PeGFP (Fig. 7E). Collectively, these data indicate that microglia can directly acquire antigens from virally infected OSNs.
Having demonstrated that microglia could acquire OSN-derived antigens in vivo, we next asked whether their specific removal affected cognate antigen recognition of virus-specific CD8+ T cells in the infected OB. To deplete microglia, we first generated CX3CR1-Cre-ER × Rosa Stopfl/fl diphtheria toxin receptor (DTR) mice and treated them with tamoxifen to drive DTR expression in myeloid cells. After tamoxifen cessation, we waited an additional 60 days before DT treatment to allow for peripheral monocyte turnover (49). This is a commonly used approach that allows microglial depletion while preserving CX3CR1-expressing monocytes and antiviral T cells. These mice were then seeded with naïve mTomato+ GCaMP6s+ OT-I cells and infected intranasally with VSV-OVA. Six days later, we observed that virus-specific CD8+ T cell calcium flux was significantly reduced in the microglia-depleted mice (Fig. 7F), indicating compromised viral antigen presentation when OB microglia but not monocytes were reduced (fig. S6, A and B). Because the microglia depletion efficiency varied between animals, we tested whether there was any relationship between the frequency of CD8+ T cell calcium flux and the percentage of microglia remaining in the OB. We observed a strong positive correlation (r = 0.75, P < 0.0013) between these two variables, further emphasizing the importance of microglia in fostering cognate antigen recognition and calcium flux in antiviral CD8+ T cells (Fig. 7G).
Last, we assessed the degree to which microglia contributed to the antiviral defense against an intranasal VSV infection. For this study, we used a more convenient alternative method to deplete microglia. Specifically, C57BL/6J mice were fed chow containing the CSF1R antagonist pexidartinib (PLX3397) for 1 month (50). This treatment decreased the number of microglia in the OB by 66% without reducing the recruitment of CD8+ or CD8+ OT-I T cells in day 7 VSV-OVA–infected mice (fig. S6). Despite the preservation of CD8+ T cells, VSV infection of PLX3397-treated animals resulted in frequent viral escape at day 7 after infection with elevated brain viral titers compared with control animals (Fig. 7H). Moreover, viral escape in PLX3397-treated mice was accompanied by reduced survival (Fig. 7I), which mirrored that observed in mice depleted of CD8+ T cells (Fig. 3G). Collectively, these data suggest that microglia can acquire and cross-present locally derived viral antigen to brain-infiltrating CD8+ T cells that exert antiviral pressure and contain VSV in the OB.
OSNs in the upper airway are responsible for relaying information about airborne odorants, but their contact with circulating air makes them a target for many pathogens. OSN infection provides a direct single-cell route into the CNS that bypasses the BBB and BCSFB. In this study, we sought previously unknown insights into how the propagation of a cytopathic nasal pathogen is thwarted as it enters and attempts to spread throughout the CNS. Nasal VSV infection induced massive leukocyte infiltration into the OE. The virus nevertheless managed to travel via OSN axons through the cribriform plate and into the OB. Within the OB, fate-mapping studies revealed that VSV was controlled noncytolytically after infecting neurons in glomeruli as well as deeper neuronal layers. This containment depended in part on antiviral T cells and a combination of effector molecules (perforin, IFNγ, TNFα). However, specific removal of MHC I from virus-infected cells using floxed β2M mice demonstrated that direct engagement of neurons by CD8+ T cells was not required for viral control. This unexpected observation led us to perform intravital imaging studies of cognate antigen recognition by virus-specific CD8+ T cells in infected glomeruli. These studies revealed that microglia displayed cognate peptide-MHC I, promoting CTL antigen recognition and calcium flux. Microglia were not infected by VSV but instead acquired antigen from adjacent neurons, which was then cross-presented to the infiltrating virus-specific CTLs. These interactions were functionally important, because microglia depletion markedly reduced cognate antigen recognition by CTLs, impeding viral containment within the OB and promoting the development of fatal encephalitis.
After nasal inoculation, virus is rapidly transported from the airway via the OSNs into the CNS where it is seen primarily within the OB. Viral replication is usually halted at the OSN-glomerular junction (10). Previous studies have shown that T cells play an important role in the control and clearance of VSV after CNS infection (29, 35). In addition, IFNγ, TNFα, interleukin-12, and nitric oxide synthase 1 were all shown to participate in the antiviral defense against VSV (51–55), yet it is presently unclear how T cell effector functions are regulated in the CNS. Regulation of T cell effector functions is important because adult neurogenesis is reserved for specialized compartments [such as the OE (56, 57)]; thus, when neurons are lost in the CNS during adulthood, they typically are not replaced. Because of this, it is postulated that the CNS favors the use of noncytolytic viral clearance—a well-described cytokine-mediated mechanism of eradicating an intracellular pathogen without killing the cell (18, 58). Cytokines such as IFNγ, TNFα, and IFN-I are all capable of purging viruses noncytolytically, and this mode of viral clearance has been described for many different pathogens, including hepatitis B virus (59), LCMV (60, 61), influenza virus (62, 63), MHV (64), and Sindis virus (65).
VSV is a cytopathic pathogen that is highly lytic in vitro (66); however, its interaction with the CNS appears to be more nuanced, as mice infected with VSV can sometimes harbor virus for several weeks before succumbing (Fig. 3A). Although we cannot exclude a role for CTLs using cytolytic effector molecules like perforin to control this virus (Fig. 3G), our fate mapping studies in VSV-iCre–infected Stopfl/fl TdTomato reporter mice revealed ample evidence of previously infected OB neurons that were cleared noncytolytically. These results are consistent with a previous study showing that neurotropic MHV can be noncytolytically cleansed from OB interneurons after infection (64). Antiviral cytokines like IFNγ and TNFα are known to play a direct role in noncytolytic viral clearance from the CNS (8, 9, 65) and peripheral tissues (18, 19, 60). We observed that both IFNγ- and TNFα-deficient mice have impaired survival after intranasal VSV infection, which is consistent with studies showing an important role for these cytokines in controlling other CNS infections. For example, IFNγ is known to help control measles virus (14), herpes virus (16), MHV (67), Borna virus (68), and recombinant vaccinia virus (69), whereas TNFα has a protective role against West Nile virus (WNV) (70), JEV (71), tick-borne encephalitis virus (72), and HSV (73). In our study, VSV-iCre allowed us to distinguish the direct effects of IFNγ on infected cells from other potential functions exerted by this cytokine, such as the activation and recruitment of immune cells (74–76). We observed that IFNγR expression by infected cells plays no direct antiviral role in mediating survival after intranasal VSV inoculation. Whether TNF receptor is required on infected cells in this model is presently unclear.
MHC I is important for different phases of normal brain development and for controlling ongoing plasticity (77, 78). However, MHC I expression is often undetectable on most CNS neurons (43, 44, 79), and its ability to engage in physiologically relevant levels of antigen presentation is a source of continued debate. The classical paradigm suggests that activated CTLs enter virally infected tissues and recognize cognate viral peptides presented in MHC I molecules displayed on the infected cell surface. After some viral infections, neuronal MHC I does appear to play a role in containing virus through CTL pressure, and these interactions can facilitate release of immune effector molecules and eventual neuronal damage (80). During HSV infection, CTL-derived IFNγ is believed to contribute to noncytolytic maintenance of viral latency within infected neurons (16, 81). In addition, perforin and granzyme B also participate in the noncytolytic control of this virus in neurons by degrading the HSV-1 immediate early protein (82). By contrast, CNS neurons infected by WNV are believed to be targeted directly by CTLs that use perforin-mediated cytolytic mechanisms to purge virally infected neurons (83). It is presently unclear why some CTL engagements of infected neurons result in cytopathology, whereas others do not. This could be linked to the amount of MHC I expressed by different neuronal subtypes.
Although CTLs have the capacity to engage at least some virus-infected neuronal populations, our results in VSV-iCre–infected floxed β2M mice demonstrate that MHC I expression on infected neurons does not play a role in protecting mice from fatal encephalitis (Fig. 5I). CTLs are, however, important in preventing fatal disease after intranasal VSV infection (Fig. 5I), and our intravital imaging studies of cognate antigen recognition by antiviral CTLs in the OB uncovered that microglia served as the target of at least 70% of these interactions. This was especially unexpected because we found no evidence, using two sensitive methods, that microglia were infected by VSV (fig. S5A and B). These results suggested that microglia were, in fact, acquiring antigen from adjacent neurons and cross-presenting the material via MHC I to CTLs. This conclusion is supported by our data showing that microglia within the OB engulfed the neuronal tracer CTB and fluorescent protein–tagged VSV particles—a process that supplies potentially cross-presentable MHC I antigen. Cross-presentation is thought to be important during naïve CTLs priming within secondary lymphoid organs; however, it can also occur in peripheral tissues, particularly tumors (84). In addition, both neonatal and adult microglia were previously shown to cross-present exogenous antigen (85), demonstrating that these cells do indeed have the capacity to cross-present.
Recent studies have begun to shed light on the roles microglia play in host defense and immunopathology after CNS virus infection (86–93). Depletion studies have revealed that microglia participate in the antiviral defense against pseudorabies virus (88), JEV (87), MHV (89), Theiler’s murine encephalomyelitis virus (90), VSV (91), and WNV (87, 92). For example, Wheeler et al. (89) found that many microglia-depleted mice fail to control MHV, resulting in compromised survival after intracerebral infection. This study also noted that microglia-depleted animals had impaired CD4+ T cell accumulation and function within the CNS. This is noteworthy because unlike microglia from VSV-infected brains (Fig. 6A) (94), MHV infection induces a marked up-regulation of MHC II on microglia. These two models of virus infection appear to showcase an important functional divergence in the ability of microglia to support either MHC I– or MHC II–restricted T cell responses in the virally infected CNS. The importance of separating MHC I and II presentation to effector T cells in the CNS is not known; however, this is the case after VSV infection, because OB-infiltrating myeloid cells express very high levels of MHC II, whereas microglia do not (94).
Although we did not focus on virus-specific CD4+ T cell interactions in this study, our findings demonstrate a role for cross-presenting microglia in displaying cognate peptide–MHC I complexes to CTLs. We had intended to provide a more direct role for MHC I expression on microglia by conditionally deleting β2M from these cells. However, VSV infection of CX3CR1-Cre-ER × floxed β2M mice resulted in near-complete NK cell–mediated deletion of the MHC I class negative microglia similar to what was described previously in inflamed floxed β2M mice (95). Nevertheless, our intravital imaging studies revealed that microglia supported most of the CTL calcium flux activity in the OB, and cognate antigen recognition was markedly reduced in their absence. CTL calcium flux in the infected OB correlated positively with the number of microglia remaining in depleted mice. In addition, microglia depletion promoted both viral escape into the brain and increased the incidence of fatal encephalitis. Antigen presentation by uninfected microglia to infiltrating CTLs is likely a mechanism the CNS uses to elicit antiviral activity but protect neurons from direct CTL contact and potential lysis. It is also important to note that most CTL interactions with microglia were short lived (kinetic) rather than stable. It is postulated that these short-lived interactions favor CTL usage of antiviral molecules (e.g., cytokines) rather than killing behavior (42). This would provide a mechanism to control virus in adjacent neurons without engaging them.
In conclusion, this study relied on intravital microscopy to analyze cognate antigen recognition by CTLs as a nasal virus attempted to enter the brain via infection of the peripheral nervous system (i.e., OSNs). Because VSV primarily infects neurons, it was unclear how CTLs would exert their antiviral pressure on a cell population that expresses little to no MHC I. However, our studies uncovered cross-presenting microglia as the main innate immune cell type responsible for driving protective, noncytolytic CTL responses against a cytopathic CNS viral infection. These findings have important implications for our understanding of antiviral immunity, because they indicate that microglia need not be infected to present antigen on MHC I and participate in the eradication of virus from neurons. Evolutionarily, it makes sense for microglia to serve as an intermediary in the defense against a neuronal infection. Microglia are a renewable cell population, whereas most neurons are not. We therefore postulate that the CNS is designed to minimize direct engagement of infected neurons by CTLs and favor release of noncytolytic antiviral cytokines through interactions with cross-presenting intermediaries. Microglia play an important role in synaptic pruning and CNS homeostasis, and their proximity to neurons makes them an ideal candidate for this task. Given the role of microglia in cross-presenting viral antigen, factors that result in their distraction or depletion are likely to render the CNS more susceptible to infection.
MATERIALS AND METHODS
C57BL/6J (B6), B6.129(Cg)-Gt(ROSA)26Sortm4(ACTB-tdTomato,-EGFP)Luo/J (mTomato), C57BL/6-Tg(TcraTcrb)1100Mjb/J (OT-I), C57BL/6-Prf1tm1Sdz/J(PRF−/−), B6.Cg-Gt(ROSA)26Sortm14(CAG-tdTomato)Hze/J (Stopfl/fl TdTomato), B6.129P2(Cg)-Cx3cr1tm2.1(cre/ERT2)Litt/WganJ (CX3CR1-Cre-ER), C57BL/6-Gt(ROSA)26Sortm1(HBEGF)Awai/J (ROSA Stopfl/fl DTR), C57BL/6N-Ifngr1tm1.1Rds/J (IFNRfl/fl) B6.129P-Cx3cr1tm1Litt/ (CX3CR1gfp/gfp), B6.129S7-Ifngtm1Ts/J (IFN−/−), and B6;129S-Tnftm1Gkl/J (TNFa−/−) mice were purchased from the Jackson Laboratory and either used or bred in house. F1 crosses were made between CX3CR1-Cre-ER and ROSA Stopfl/fl DTR mice as well as B6 and CX3CR1gfp/gfp mice. H-2KbDb−/− (96), IFNAR−/− (32) (provided by Jonathan Sprent, formerly at The Scripps Research Institute), P14 (37), OFP+ P14 (97), mTomato+ P14, mTomato+ GCaMP6s+ P14, mCerulean+ OT-I (97), mTomato+ OT-I, CD45.1, and actin-GCaMP6s mTomato OT-I (13) and were bred and maintained under specific pathogen-free conditions at the National Institutes of Health (NIH). OMP-GFP (98) mice were provided by Leo Belluscio, NINDS, and backcrossed onto a C57BL/6J background. All mice in this study were handled in accordance with the guidelines set forth by the NIH Animal Care and Use Committee.
Viruses and plaque assays
LCMV, VSV-OVA serotype Indiana (provided by Leo Lefrancois, formerly at the University of Connecticut) (99), VSV-eGFP (100), VSV-DsRed, VSV-iCre, and VSV-PeGFP (48) were propagated on BHK21 cells. Infectivity of VSV preparations was quantified by plaque assay on green monkey kidney cells (Vero). VSV titers from brains of infected mice were determined similarly. Briefly, animals were perfused with saline, brains were immersed in 1 ml of 2% FBS RPMI containing zirconia silica beads and homogenized with a FastPrep 24-5G (MP Biomedical). Homogenates were spun for 3 min at 15,000g, and supernatants were serially diluted for quantification on Vero cells. One-hundred microliters of diluted homogenate was added to Vero monolayers in flat bottom 24-well plates and incubated for ~3 hours. One-hundred microliters of 0.75% methylcellulose in DMEM (containing 10% FBS, 1% glutamine, and 1% Pen/Strep) were added to each well and incubated for 44 to 48 hours at 37°C in a 5% CO2 incubator. Cells were subsequently fixed with 4% neutral buffered formalin and stained with crystal violet, and plaques were enumerated.
The pFL-VSVBsiWI plasmid (48) was modified with insertion of Ova-IRES-iCre (101) sequence (pFL-VSV-Ova-IRES-iCre) flanked by BsiWI site at G-L junction of the plasmid. The recombinant VSV expressing OVA and iCre recombinase was recovered by co-transfection of full-length plasmids along with supporting plasmids pN, pP, and pL (carrying the coding sequences of the N, P, and L proteins of VSV) in BHK-21 cells as per the protocol published earlier (48). The recombinant virus (rVSV-OVA-iCre) was plaque purified and viral stocks were prepared using BHK-21 cells. Lastly, a Western blot was performed to confirm expression of Ova and iCre proteins in cell lysates from BHK-21 cells infected with rVSV-OVA-iCre.
Eight- to 10-week-old mice were infected intranasally by injecting 10 μL of viral diluent in each nostril. For survival, titering, immunohistochemical, flow cytometric, and microarray experiments, mice were infected with 1 × 105 PFU of VSV per nostril. For intravital imaging experiments, mice were infected with 4 × 105 PFU per nostril to ensure more uniform OB infection. For experiments involving LCMV, mice were infected i.p. with 2 × 105 PFU of LCMV Armstrong.
Transgenic mouse generation
Floxed beta-2 microglobulin (β2M) knock-in mice (floxed β2M) were generated by Cyagen Biosciences. Homology arms were derived from BAC clones RP23-285J7 and RP23-299B11 from the C57BL/6 library and LoxP sites were inserted into genomic positions flanking exons 2 and 3 of the beta microglobulin gene (fig. S4). Homologous recombination was performed in C57BL/6 ES cells before transfer into C57BL/6 embryos. Positive founder lines were backcrossed one additional generation onto the C57BL/6J background before intercrossing.
Materials and Methods
Fig. S1. Diagrams of sectioning planes through the mouse head.
Fig. S2. Parenchymal cells within the OB and brain survive VSV infection after intranasal inoculation.
Fig. S3. Virus-specific CTLs flux calcium for longer durations while undergoing stable interactions.
Fig. S4. Generation of floxed β2M transgenic mice.
Fig. S5. VSV does not infect microglia and CTB traffics from OSNs to OB glomeruli.
Fig. S6. Quantification of microglia and infiltrating cells in the OB using different microglia depletion methods.
Movie S1. Virus-specific CTLs rapidly migrate across VSV infected axon tracts within the infected outer nerve layer.
Movie S2. Virus-specific CTLs survey the infected glomerular layer with the OB.
Movie S3. Antiviral CTLs engage cognate antigen and flux calcium within the infected OB.
Movie S4. Antiviral CTLs flux calcium upon contact with microglia.
Acknowledgments: We thank A. Hoofring in the NIH Medical Arts Design Section for generating the illustration shown in fig. S1. Funding: This research was supported by the intramural program at the NIH. Author contributions: A.C.B. and E.A.M. performed all the experiments and analyzed data. D.N. generated VSV recombinants. D.B.M. provided the funding. E.A.M. and D.B.M. wrote the manuscript. Competing interests: The authors declare that they have no competing interests. Data and materials availability: All data needed to evaluate the conclusions are present in the paper or the Supplementary Materials.
- Copyright © 2020 The Authors, some rights reserved; exclusive licensee American Association for the Advancement of Science. No claim to original U.S. Government Works