MHC-II expression is a hallmark of microglial activation during acute CMV infection in brain
Perinatal MCMV infection results in brain infection and subsequent inflammatory response, characterized by infiltration of leukocytes and activation of resident microglia at the site of infection (Bantug et al., 2008; Kosmac et al., 2013; Kveštak et al., 2021). We have previously shown that activated microglia upregulate MHC-I, MHC-II, and the costimulatory molecules CD80 and CD86 (Kveštak et al., 2021). To extend these findings, we infected newborn C57BL/6 (Fig. 1A, 1C, Suppl. Figure 1A) and BALB/c (Fig. 1B, 1D, Suppl. Figure 1B) wild-type (WT) mice intraperitoneally (i.p.) with MCMV and harvested brains when MCMV replication peaks in the brain (13 days post-infection, 13 dpi) (Brizić et al., 2018b; Kveštak et al., 2021). The prototypic microglia marker ionized calcium-binding adaptor protein 1 (IBA-1), and MHC-II expression colocalized upon MCMV infection in both strains of mice, indicating that microglia express MHC-II in the brain (Fig. 1A, B). Expression of MHC-II on microglia in C57BL/6 (Fig. 1C) and BALB/c (Fig. 1D) mice was confirmed by flow cytometry. Similarly, MHC-I expression on microglia was also increased in both mouse strains following infection (Suppl. Figure 1A, B).
To compare our findings in the mouse model to congenital HCMV infection, we analyzed the expression of human MHC-II (HLA-DR) on IBA-1+ cells in cadaveric brain tissues of two cases of cCMV infection. Analogously to congenitally MCMV-infected mice, human microglia expressed MHC-II in HCMV-infected brains (Fig. 1E). We have previously shown that IFN-γ is required for early microglial activation during the acute phase of perinatal MCMV infection (Kveštak et al., 2021). To determine the requirements for HLA-DR upregulation upon HCMV infection of the fetal brain, we have used the recently developed human fetal organotypic brain slice culture (hfOBSC) platform (Rashidi et al., 2024). Whereas HCMV infection alone did not result in HLA-DR expression in hfOBSCs (Fig. 1F), IFN-γ treatment of infected hfOBSCs induced HLA-DR expression on microglia (Fig. 1F), similar to microglia in MCMV-infected newborn mice (Kveštak et al., 2021). Thus, in both humans and mice, cytomegalovirus infection in the brain results in microglia activation in an IFN-γ-dependent manner.
MCMV infection of the CNS causes persistent activation of microglia
Productive MCMV infection in the brain is resolved by weeks 3 to 4 post-infection, but MCMV DNA remains detectable in these brains for > 3 months, suggesting the establishment of MCMV latency (Brizić et al., 2018a, 2018b; Koontz et al., 2008). We detected MCMV genomes in the brains of MCMV-infected C57BL/6 and BALB/c mice at 3 months post-infection (mpi), a time corresponding to the latency (Fig. 2A). Moreover, we could not detect viral transcription from any known locus of MCMV in brain homogenates during the latent phase of infection (Suppl. Figure 2), suggesting that the virus established true latency.
Next, we performed immune response profiling by analyzing the protein expression of 48 cytokine, chemokine, and growth factor targets in mouse brains. Almost all proteins analyzed were equally or similarly expressed in control and latently MCMV-infected brains (Suppl. Figure 3). However, we detected increased levels of chemokines MCP-1, RANTES, MCP-3, IP-10, and B-cell-activating factor (BAFF) in the brains of latently MCMV-infected mice (Fig. 2B). Since these cytokines are associated with inflammatory response, we next characterized microglia following the resolution of productive infection in BALB/c mice. The number of microglia was increased in MCMV-infected mice at all time points analyzed (Fig. 2C). Furthermore, microglia in infected mice expressed relatively higher levels of MHC-I and -II at all time points analyzed, even at 6 mpi (Fig. 2D, E). MHC-II expression remained restricted to IBA-1+ microglia in the latent phase of MCMV infection, as shown by confocal microscopy (Fig. 2F). To determine if there are differences in microglia activation between different brain regions, we isolated microglia from the hippocampus, cortex, and cerebellum at 4 mpi and analyzed the expression of MHC-I and -II (Fig. 2G, H). Compared to cortex and cerebellum, hippocampal microglia expressed the highest MHC-I and -II levels. Notably, the hippocampus also had the relatively highest viral load (Brizić et al., 2024), suggesting that MCMV latency shapes the microglial activation status.
Next, we analyzed the density and morphology of microglia in the hippocampal CA1 region, which is important for learning and memory-dependent processes and is known to be more susceptible to inflammation during early development (Gomez et al., 2019; Korte and Schmitz, 2016). Microglia in latently MCMV-infected mice showed increased density (Fig. 2I), and activated morphology with an increased volume of the whole cell and soma (Fig. 2J, K). Furthermore, microglial activation has been shown to represent a continuum between ramified activated and amoeboid forms (Ladeby et al., 2005). To further assess the microglial morphological changes, the total length and branch points of microglial filaments were examined in both groups. Remarkably, the microglial cells in the latently infected mice exhibited a higher total filament length and a higher number of branching points, indicating a hyper-ramified profile of activated microglial cells (Fig. 2J, K). Next, we analyzed MHC-I and -II expression on macrophages in the brain, spleen, and liver to determine if MCMV latency shapes the activation state of other tissue-resident macrophages. Notably, splenic red pulp macrophages (F4/80+CD11b−) (Fig. 2L), liver Kupffer cells (F4/80+CD11b+) (Fig. 2M), but also CNS-associated macrophages (CD45.2hiCD11b+) (Fig. 2N), did not have significantly increased MHC-I and -II expression in latently MCMV-infected mice. Altogether, these data demonstrate that microglia are selectively and persistently activated in latently MCMV-infected brains.
Latent MCMV infection causes persistent transcriptional alteration of microglia
Increased expression of MHC-I and -II on microglia in latently infected brains was still observed at 16 mpi (Fig. 3A). Since microglia were persistently activated following perinatal MCMV infection, we sorted microglia from MCMV- and mock-infected mice 16 mpi and performed RNA-seq analysis (Fig. 3B). To elucidate transcriptional differences of microglia between acute and latent infection, we also reanalyzed our previous RNA-seq data of microglia obtained 8 dpi (Kveštak et al., 2021). Individual samples from the same experimental groups clustered closely together in the PCA plot, indicating low inter-sample variability (Fig. 3C). In accordance with previous analysis, acute MCMV infection in the brain strongly reshaped the microglial transcriptome (Fig. 3C). Microglia from latently MCMV-infected brains were also transcriptionally different when compared to microglia from uninfected mice, even though this difference was smaller than that observed between microglia from acutely infected mice and uninfected mice (Fig. 3C). The gene ontology over-representation analysis (GO-ORA) showed that microglia from latently infected mice display the transcriptional signature of elevated antiviral response (Fig. 3D). Gene ontology biological process categories associated with antigen processing and presentation, negative regulation of viral process and interferon − mediated signaling pathway were enriched in microglia of latently infected mice (Fig. 3D). By comparing the list of differentially expressed (DE) genes in acute and latent infection, we identified 1613 DE genes unique to microglia from the acutely infected brain, 137 DE genes unique to microglia from latently infected brain, and 248 genes that were DE in both microglia from both acute and latently infected brains (Fig. 3E). Importantly, most of these 248 shared DE genes were associated with antiviral response, including antigen processing and presentation, interferon type I (IFN-I) and type II (IFN-II) pathways, and inflammatory response (Fig. 3F). Taken together, these data demonstrate the maintenance of a persistent antiviral state of microglia in brains of latently MCMV-infected mice.
Differential regulation of microglial and astrocyte transcriptomes at the single cell level in latently MCMV-infected CNS
Next, we used the single-cell RNA sequencing (scRNA-seq) approach to decipher the impact of latent MCMV infection on the microglial transcriptome at the individual cell level. For these experiments, newborn C57BL/6 mice were i.p. infected with MCMV and microglia were sorted for scRNA-seq (Suppl. Figure 4A). Following data pre-processing and dataset integration, a total of four distinct microglia clusters were detected using the selected clustering resolution (Fig. 4A), all of which exhibited strong expression of canonical microglia marker genes, such as C1qa, Fcrls, Hexb, P2ry12, and others (Suppl. Figure 4C) (Hammond et al., 2019). Interestingly, latent MCMV infection caused a noticeable redistribution of a subset of microglia cells from clusters 0 and 1 into cluster 2 (Figs. 4A and 4B), which suggests that latent MCMV infection transcriptionally reprograms a subset of microglia. In addition to the prominent increase in the proportion of cells in microglial cluster 2 in response to latent MCMV infection, various proportions of cells in that cluster were also characterized by a relatively high expression of genes encoding MHC molecules (e.g. B2m, H2-Aa, Cd74) and their transcriptional activators (e.g. Ciita, Nlrc5), genes involved in type I interferon (IFN-I) and type II interferon (IFN-II) signaling (i.e. Stat1, Irf1, Irf7, Irf9, Ifit2, Ifit3, Iigp1), and genes encoding pro-inflammatory chemokines (Cxcl9, Cxcl10, Ccl5) (Fig. 4C), in accordance with bulk RNA-seq analysis (Fig. 3.). Notably, a distinct microglia subset did not display any apparent changes in the expression of these infection or inflammation-associated genes, suggesting that latent MCMV infection shapes microglia towards an activated and proinflammatory phenotype; however, not in all microglial cells in brains of latently MCMV-infected mice.
Astrocytes are another type of brain-resident glial cells that can undergo activation and modulate immune responses in different brain diseases (Colombo and Farina, 2016; Giovannoni and Quintana, 2020). To investigate whether the latent MCMV infection affects astrocytes to the same or similar extent as it affects microglia, we performed scRNA-seq analysis of astrocytes, which were sorted as CD45.2−CD11b−O1−ACSA-2+ cells by flow cytometry (Suppl. Figure 4B). Four astrocyte clusters were identified in brains of mock- and MCMV-infected mice (Fig. 4D), and analyzed cells expressed canonical astrocyte markers Slc1a2, Slc1a3, Atp1b2, Sox9, Glul and Apoe (Suppl. Figure 4D) (Batiuk et al., 2020). In contrast to microglia, no pronounced changes in the number of cells within individual astrocyte clusters were detected (Fig. 4D, E). Furthermore, we did not detect a substantial increase in the expression of proinflammatory genes within any astrocyte subpopulation (Fig. 4F). These results demonstrate that while microglia are transcriptionally reprogrammed at the single-cell level to exert proinflammatory state, astrocytes exhibit homeostatic features in latently MCMV-infected brain.
Antiviral interventions limit microglial activation
Having shown that MCMV can trigger the establishment of a proinflammatory microglia population in the brain, we next investigated if available interventions can mitigate these changes. Antivirals, such as ganciclovir (GCV), are commonly used to control HCMV infection in humans (D and Rc, 1990). In mice, passive immunization protects against MCMV infection in the brain by decreasing both the viral burden and virus-induced pathology (Cekinović et al., 2008). To test if antiviral treatment can reduce microglial activation upon MCMV infection, we i.p. infected newborn BALB/c mice and subsequently treated them with either GCV or MCMV immune sera. Mice were treated with immune sera on 1 and 7 dpi, and GCV was administered daily until 14 dpi (Fig. 5A). Both approaches reduced the upregulation of microglial MHC-I and -II expression (Fig. 5B), and reduced viral load in the brain (Fig. 5C). Moreover, the antiviral treatment attenuated microglial MHC-I and -II expression long-term (Fig. 5D, E), as well as reduced latent viral load in the brain for as long as 90 dpi (Fig. 5F). Next, we analyzed how anti-viral treatment impacts neuroinflammation if we postpone it for 5 days, the time when MCMV has already reached the brain (Fig. 5G). Anti-viral treatment initiated at 5 dpi only reduced MHC-II upregulation, but had no effect on MHC-I expression nor the viral load in treated mice (Fig. 5H-I). Finally, we assessed if antiviral therapy could mitigate neuroinflammation during MCMV latency. To that aim, BALB/c mice infected with MCMV as newborns were treated with GCV starting 2 mpi (Fig. 5J). One month of antiviral treatment did not affect microglial MHC-I and -II expression (Fig. 5K) nor latent viral loads in the brain (Fig. 5L). These data indicate that timely inhibition of virus replication reduces microglial activation, while antiviral treatment during the latent phase of MCMV infection does not affect microglial activation.
Continuous IFN-γ signaling maintains microglial MHC-II expression
We have previously shown that IFN-γ is critical for early microglial MHC-II upregulation following MCMV infection in the brain (Kveštak et al., 2021), and in HCMV-infected human brain organotypic cultures (Fig. 1F). To assess if IFN-γ is required for long-term microglial MHC-II expression during MCMV latency, we i.p. infected newborn IFN-γ receptor-deficient (Ifngr1−/−) mice, and control WT mice, and subsequently analyzed microglial MHC-II expression. Increased microglial MHC-II expression was not observed in Ifngr1−/− infected mice 3 mpi (Fig. 6A). In contrast, IFN-γ signaling was only partially required for increased MHC-I expression in Ifngr1−/− mice, as the increase in MHC-I expression on microglia was lower than in WT mice (Fig. 6A).
We next investigated whether continuous IFN-γ receptor signaling is required to maintain MHC-II expression on microglia. We first generated an inducible conditional knockout mouse strain (Sall1CreERT2Ifngr1fl/fl), in which the IFN-γ receptor (IFNGR) expression in microglia can be eliminated upon tamoxifen treatment. We infected newborn Sall1CreERT2Ifngr1fl/fl mice and administered tamoxifen 3 mpi (Fig. 6B). Tamoxifen treatment efficiently decreased IFNGR expression (Fig. 6C), accompanied by loss of MHC-II (Fig. 6D) and decrease in MHC-I (Fig. 6E) expression on microglia in Sall1CreERT2Ifngr1fl/fl mice, which was not observed in control and non-treated Sall1CreERT2Ifngr1fl/fl and Ifngr1fl/fl mice (Suppl. Figure 5A, B). These results demonstrate that IFN-γ is required continuously to maintain an activated microglial state during latent MCMV infection.
Tissue-resident CD8 T cells are the major source of IFN-γ in latently MCMV-infected brains
Since our results indicated that continuous IFN-γ receptor signaling in microglia was required to maintain microglia in an activated state, we next determined the cellular source of IFN-γ. IFN-γ could be produced by lymphocytes in the brain parenchyma or the periphery, reaching the brain by blood (Ivashkiv, 2018). To differentiate between both options, we first assessed if systemic IFN-γ causes microglial activation. To test this possibility, we neutralized IFN-γ for two weeks in latently MCMV-infected mice by i.p. administration of a neutralizing IFN-γ antibody (Fig. 6F). As control of peripheral IFN-γ neutralization, we infected adult C57BL/6 mice with MCMV and neutralized IFN-γ during acute infection (Suppl. Figure 5C). In contrast to control mice, MHC-II was not upregulated on peritoneal macrophages in mice treated with IFN-γ neutralizing antibody (Suppl. Figure 5C), indicating successful neutralization of IFN-γ. However, MHC-II expression levels on microglia were unaffected by IFN-γ neutralization (Fig. 6F), indicating that local but not systemic IFN-γ production is required for the maintenance of microglial MHC-II expression. Even though we did not detect IFN-γ protein in brain homogenates of latently MCMV-infected mice (Suppl. Figure 5D), we detected increased numbers of IFN-γ transcripts (Fig. 6G).
We have previously demonstrated that NK cells are the main producers of IFN-γ during acute MCMV infection in the brain, mediating microglial MHC-II expression (Kveštak et al., 2021). To determine which cell type is a major source of IFN-γ during latency, we stimulated mononuclear cells isolated from latently MCMV-infected brains. CD8 T cells were the main producers of IFN-γ, while other cell types accounted for minimal IFN-γ production in latently infected brains (Fig. 6H, I). In accordance with the data obtained by neutralizing peripheral IFN-γ, most IFN-γ+ CD8 T cells were TRM cells, expressing CD69, or co-expressing both canonical TRM cell markers CD69 and CD103 (Fig. 6J) (Mueller and Mackay, 2016). Since we observed the highest microglial activation in the hippocampus, we assessed the numbers of CD8 T cells in different brain regions. The highest numbers of total CD8 T cells were detected in the hippocampus (Suppl. Figure 5E). Similarly, virus-specific CD8 T cells, as assessed by analyzing M38-tetramer positive cells, were the most numerous in hippocampus and displayed TRM phenotype (Fig. 6K, Suppl. Figure 5F). Overall, these data suggest that virus-specific CD8 TRM cells are major producers of IFN-γ during MCMV latency in the brain.
Persistently activated microglia enhance control of latent virus
Having established that long-term microglia activation is a hallmark of latent CMV infection, we hypothesized that such microglia have enhanced functionality, corresponding to the emerging concepts of innate immune cells adaptation to different pathological conditions (Divangahi et al., 2021). To determine the role of activated microglia during MCMV latency in the brain, we used PLX5622, a colony-stimulating factor 1 receptor (CSF1R) inhibitor, to deplete microglia (Xu et al., 2020). C57BL/6 mice were infected as newborns, and after the establishment of latency, mice were fed with a PLX5622-formulated diet (further referred to as PLX-diet) or a control diet (Fig. 7A). Microglial numbers were strongly reduced in mice fed with the PLX-diet for 2 weeks (Fig. 7B). Notably, PLX-mediated depletion of microglia resulted in MCMV reactivation in the brain (Fig. 7C), demonstrating the importance of microglia in preventing MCMV reactivation. To further evaluate the role of persistently activated microglia in virus control, we have performed an intracranial challenge with MCMV in latently infected and control mock-infected mice (Fig. 7D). In control mice, microglia depletion did not significantly affect virus titers upon intracranial challenge (Fig. 7E). In sharp contrast, microglial depletion had a major role in controlling productive MCMV infection upon intracranial challenge of latently infected mice (Fig. 7E), resulting in ~ 100-fold increase in virus titer when compared to the non-depleted group of mice. Altogether, these data indicate that activated microglia have an important role in controlling both productive as well as preventing reactivation of latent MCMV in the brain.
Microglia compromise synaptic connectivity of neurons in the hippocampus during latent infection
As the resident macrophages in the brain, microglia can remove pathogens, cell debris, but also synaptic connections between neurons under pathological conditions (Boche et al., 2013; Demuth et al., 2023). The next step was, therefore, to analyze whether microglia engulf and digest synaptic terminals, followed by degradation in lysosomes (Demuth et al., 2023). Since the total volume of lysosomes in microglia is proportional to their phagocytic activity (Demuth et al., 2023), we first investigated the volumetric changes of microglial lysosomes in infected mice compared to control mice. We detected increased lysosome volume labeled with lysosome-associated membrane protein-1 (LAMP-1) in microglia in the CA1 subregion of the hippocampus (Fig. 7F, G). Next, we analyzed Homer-1, a synaptic scaffolding protein in postsynaptic terminals that regulates glutamatergic synapses and spine morphogenesis (Tao-Cheng et al., 2014), in microglia in all experimental groups. We detected an increased number of Homer-1 puncta in lysosomes in microglia from latently infected mice, demonstrating that activated microglia excessively phagocytose excitatory postsynaptic terminals of hippocampal neurons during MCMV latency (Fig. 7G).
To investigate whether the increased phagocytosis of postsynaptic terminals of excitatory neurons is reflected in the reduced synaptic connections, we evaluated the dendritic spine density as a morphological indicator of the hippocampal neurons. Spines are dendritic protrusions that carry the majority of excitatory synapses in the hippocampus, and changes in spine density can provide information about changes in the connectivity of hippocampal neurons (Demuth et al., 2023; Gabele et al., 2024; Hosseini et al., 2018). Spines were counted on the apical dendrites of CA1 pyramidal neurons. Notably, the density of spines was reduced in the apical dendrites of CA1 neurons of latently infected mice (Fig. 7H). Intriguingly, the depletion of microglia during latency restored the numbers of dendritic spines to the levels observed in control uninfected mice (Fig. 7H), demonstrating that activated microglia associated with latent CMV infection are the cause of dendritic spine loss on the neurons. These data indicate that persistent microglial activation compromises synaptic connectivity in the hippocampus.